The Economics of Airline Institutions, Operations and Marketing, Volume 2 (Advances in Airline Economics)

ADVANCES IN AIRLINE ECONOMICS The Economics of Airline Institutions, Operations and Marketing

This page intentionally left blank

ADVANCES IN AIRLINE ECONOMICS

The Economics of Airline Institutions, Operations and Marketing VOLUME 2 Edited by

Darin Lee LECG, LLC Cambridge, USA

AMSTERDAM • BOSTON • HEIDELBERG • LONDON • NEW YORK • OXFORD PARIS • SAN DIEGO • SAN FRANCISCO • SINGAPORE • SYDNEY • TOKYO

Elsevier Radarweg 29, PO Box 211, 1000 AE Amsterdam, The Netherlands Linacre House, Jordan Hill, Oxford OX2 8DP, UK First edition 2007 Copyright © 2007 Elsevier B.V. All rights reserved No part of this publication may be reproduced, stored in a retrieval system or transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without the prior written permission of the publisher Permissions may be sought directly from Elsevier’s Science & Technology Rights Department in Oxford, UK: phone (+44) (0) 1865 843830; fax (+44) (0) 1865 853333; email: [email protected]. Alternatively you can submit your request online by visiting the Elsevier web site at http://elsevier.com/locate/permissions, and selecting Obtaining permission to use Elsevier material Notice No responsibility is assumed by the publisher for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions or ideas contained in the material herein. Because of rapid advances in the medical sciences, in particular, independent verification of diagnoses and drug dosages should be made Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the Library of Congress British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: 978-0-444-53027-1 For information on all Elsevier publications visit our website at books.elsevier.com Printed and bound in The Netherlands 07 08 09 10 11 10 9 8 7 6 5 4 3 2 1

Working together to grow libraries in developing countries www.elsevier.com | www.bookaid.org | www.sabre.org

Contents

Advances in Airline Economics, Volume II

vii

Preface

ix

List of Contributors

xi

1 Institutions, Regulation, and the Evolution of European Air Transport

1

Jan K. Brueckner and Eric Pels

2 Wage Determination in the US Airline Industry: Union Power Under Product Market Constraints

27

Barry T. Hirsch

3 Toward Rational Pricing of the US Airport and Airways System

61

Daniel P. Kaplan

4 An Interpretative Survey of Analytical Models of Airport Pricing

89

Leonardo J. Basso and Anming Zhang

5 What if the European Airline Industry had Deregulated in 1979?: A Counterfactual Dynamic Simulation

125

Purvez F. Captain, David H. Good, Robin C. Sickles and Ashok Ayyar

6 State Aid to European Airlines: A critical Analysis of the Framework and its Application

147

Pietro Crocioni and Chris Newton

7 The Implications of the Commercialization of Air Transport Infrastructure

171

Kenneth Button

8 The Role of Regional Airlines in the US Airline Industry

193

Silke Januszewski Forbes and Mara Lederman

9 Airport Substitution by Travelers: Why Do We Have to Drive to Fly? Gary M. Fournier, Monica E. Hartmann and Thomas W. Zuehlke

209

CONTENTS

vi

10 Assessing the Role of Airlines and Airports in Multi-airport Markets

235

Jun Ishii, Sunyoung Jun and Kurt Van Dender

11 Airline Ticket Taxes and Fees in the United States and European Union

255

Joakim Karlsson, Amedeo Odoni, Célia Geslin and Shiro Yamanaka

12 Are Passengers Willing to Pay More for Additional Legroom?

275

Darin Lee and María Jose´ Luengo-Prado

13 Assessing the Potential Success of the Low-Cost Business Models

in Asian Aviation Markets

287

David Gillen and Natthida Taweelertkunthon

14 Pricing Strategies by European Traditional and Low Cost Airlines: Or,

when is it the Best Time to Book on Line?

319

Claudio A. Piga and Enrico Bachis

15 The Long-Run Distributional Effects of Industry and Carrier Changes

in the US Air Transport Market

345

Aisling Reynolds-Feighan

16 Air Travel Demand Elasticities: Concepts, Issues and Measurement

365

David Gillen, William G. Morrison and Christopher Stewart

Index

411

Advances in Airline Economics,

Volume II

Darin Lee, Editor

The second in a new series of books on the economics of the airline industry. The series is comprised of a collection of original, cutting-edge research papers from an international panel of distinguished contributors. Volume 2 focuses on topics related to airline institutions, operations and marketing such as wage determination, airport and airway pricing, taxation and low-cost carrier business strategies.

This page intentionally left blank

Preface

Following the deregulation of the US airline industry in 1978, papers analyzing the industry were a regular feature in prominent economics journals such as the American Economic Review, the Quarterly Journal of Economics and the Journal of Political Economy. While important research on the airline industry continued throughout the mid to late 1990s, the number of academic economists actively pursuing research on the airline industry had clearly wanted. After a brief hiatus however, there has been a resurgence in academic research on the airline industry, both in North America and throughout the rest of the world. Indeed, one could argue that the changes over the past several years – including the growing prominence of ’low-cost carriers’ and their interplay with the ’legacy carriers’, the advent of the Internet and its affect on industry pricing and the proliferation of international alliances – have resulted in changes that rival those following the passage of the Airline Deregulation Act. The purpose of the Advances in Airline Economics series is to provide a comprehensive overview of the current state of economic research on the airline industry. Each volume will consist of several previously unpublished research papers written by an international panel of distinguished academic and industry economists, as well as a select number of reprints of influential papers by prominent researchers in the field. Volume 2: The Economics of Airline Institutions, Operations and Marketing features 16 essays covering topics such as airline industry wage determination, airport and airway pricing, taxation and low-cost carrier business strategies. I hope readers find the essays contained in Volume 2 enlightening and that they help to stimulate further debate and research in this continually evolving and fascinating industry. Darin Lee, Editor

This page intentionally left blank

List of Contributors

A. Ayyar

Chicago Partners, LLC, New York.

E. Bachis L. J. Basso

Nottingham University Business School. ∗

Sauder School or Business, The University of British Columbia. Department of Civil Engineering, Universidad de Chile.

J.K. Brueckner

Department of Economics, University of California, Irvine, 3151 Social Science Plaza, Irvine, CA 92697.

K. Button∗

University Professor and Director of the Aerospace Policy Research Center, School of Public Policy, George Mason University, Fairfax, Virginia.

P.F. Captain P. Crocioni

∗

Ernst & Young, LLP, Houston, TX. Senior Economist, Chief Economist Team, Office of Communications (Ofcom).

S.J. Forbes

Department of Economics, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0508, USA.

G.M. Fournier∗

Department of Economics, Florida University, Tallahassee, FL 32306-2180.

C. Geslin

Department of Civil and Environmental Engineering, Massachusetts Institute of Technology.

D. Gillen∗

Sauder School of Business and Director, Centre for Transportation Studies, University of British Columbia.

D.H. Good

Indiana University, Bloomington, IN.

M.E. Hartmann

Economics Department, St. Thomas University, 2115 Summit Avenue, St. Paul, Minnesota, 55105.

B. Hirsch∗

E.M. Stevens Distinguished Professor of Economics, Trinity University, San Antonio, TX 78212.

J. Ishii

Department of Economics, University of California, Irvine, CA 92697-5100.

S. Jun

Department of Economics, University of California, Irvine, CA 92697-5100.

D. Kaplan∗

Director, LECG, LLC.

J. Karlsson∗

Division of Aviation, Daniel Webster College, 20 University Drive, Nashua, New Hampshire 03063-1300.

M. Lederman∗

Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, Ontario, Canada, M5S 3E6.

LIST OF CONTRIBUTORS

xii

D. Lee∗

LECG, LLC, 350 Massachusetts Ave. Suite 300.

M.J. Luengo-Prado

Department of Economics, Northeastern University, 301 Lake Hall, Boston, MA 02115-5000, USA.

W.G. Morrison

School of Business and Economics, Wilfrid Laurier University, 75 University Avenue West, Waterloo, ON Canada N2L 3C5.

C. Newton

Director, Frontier Economics, UK.

A. Odoni

Massachusetts Institute of Technology, Room 33-219, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139.

E. Pels

Department of Spatial Economics, Free University of Amsterdam, De Boelelaan 1105, 1081 HV Amsterdam.

C.A. Piga∗

Economics Department, Loughborough University, Leicestershire, LE11 3TU, UK.

A. Reynolds-Feighan∗

School of Economics and Geary Institute, University College Dublin, Belfield, Dublin 4.

R.C. Sickles∗

Corresponding author. Rice University, Houston, TX.

C. Stewart

School of Business and Economics, Wilfrid Laurier University, 75 University Avenue West, Waterloo, ON Canada N2L 3C5.

N. Taweelertkunthon

Centre of Transportation Studies, University of British Columbia.

K. Van Dender∗

Department of Economics, University of California, Irvine, CA 92697-5100.

S. Yamanaka

Department of Civil and Environmental Engineering, Massachusetts Institute of Technology.

A. Zhang

Sauder School of Business, The University of British Columbia.

T.W. Zuehlke

Department of Economics, Tallahassee, FL 32306-2180.

Florida

University,

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

1 Institutions, Regulation, and the Evolution of European Air Transport Jan K. Brueckner∗ and Eric Pels†‡

ABSTRACT This paper provides an overview of the institutional and regulatory developments underlying European airline deregulation. The paper has argues that the old flag-carrier regime led to a proliferation of airlines and airlines routes, leading low traffic densities in European networks and thus high operating costs. While international alliances and open skies agree­ ments helped to boost traffic densities, the low-cost carriers unleashed by deregulation, though generating substantial passenger benefits through lower fares, threaten to drain traf­ fic out of the major carriers’ networks. A defensive response is needed, and part of this response must involve concentration of the major carriers’ traffic on fewer routes through network reorganization and cross-border mergers. The paper also highlights the need for additional policy steps, especially formation of a Common Atlantic Aviation Area and new rules for airport operations.

1 INTRODUCTION An economy’s institutional and regulatory structure can have profound impacts on economic activity. This impact has been significant in the aviation sector, and the impact is starkly revealed when institutions and regulations change abruptly as a result of deliberate public policy decisions. Such a change occurred in the United States in the late 1970s when the previous regulatory structure governing airline operations was ∗ Department of Economics, University of California, Irvine, 3151 Social Science Plaza, Irvine, CA 92697, Phone: (949) 824-0083, Fax: (949) 824-2182, e-mail: [email protected]. † Department of Spatial Economics, Free University of Amsterdam, De Boelelaan 1105, 1081 HV Amsterdam, e-mail: [email protected]. ‡ We thank Darin Lee for his helpful comments. The usual disclaimer applies, however.

2

JAN K. BRUECKNER AND ERIC PELS

abruptly eliminated. Over the next decade, airline route structures were reorganized, flight frequencies increased, many new airlines began operations (with many ultimately failing), and real airfares began a long secular decline that has continued to the present. Europe is now reaping some of the benefits of its own process of airline deregulation, a process that has been more gradual than in the United States. The most noteworthy change is the explosive growth of low-cost carriers, whose share of European traffic, though still relatively small, has shown a remarkable upward trend. These carriers are not burdened by the high labor costs of the major European carriers, and they are exploiting the new opportunities for route entry in the most aggressive fashion, serving many routes that lie entirely outside their home countries. While the major airlines have been slow to exploit the freedoms granted by dereg­ ulation, strong forces are at work behind the scenes that will ultimately reshape these carriers’ operations. The new possibility of cross-border mergers within the EU will lead to consolidation of the industry, with some former flag carriers disappearing and others growing while reorganizing their route systems to achieve greater efficiency. Many observers argue that such consolidation is sorely needed to reduce the number of European airlines, which is viewed as needlessly inflated under the flag-carrier regime.1 The first major step toward industry consolidation has been achieved with the recent completion of the Air France–KLM merger, and this event is bound to be followed by other combinations of existing carriers. Deregulation faced a more difficult challenge in Europe than in the United States. because the process had to dismantle an international institutional structure, as opposed to the purely domestic one in the United States. In particular, while freeing the domestic operations of its carriers, Europe had to sweep away the web of bilateral agreements between its countries, which governed international traffic. Such an achievement was only possible, of course, because of existence of a supra-national authority like the EU. The legacy of this old institutional structure is still very much in evidence, with European airline service still reflecting the patterns established under the old flag-carrier regime. Restructuring of Europe’s aviation sector will undoubtedly take time, although this process will be accelerated by the formidable competitive pressure emanating from the low-cost carriers. The purpose of the present paper is to provide an overview of the recent evolution of air transport in Europe, with special attention to the impact of public policies. The goal is to show how institutions and regulatory history affected the initial conditions for the process of European deregulation, while exploring how changes in these institutions and regulations have begun to transform the aviation sector. While it attempts to predict the course of the aviation sector’s future evolution, the paper also discusses further regulatory changes that are needed to fully realize the benefits of European deregulation.

1

As of 1998, the EU had 13 major carriers serving a population of 374 million, while the United States had seven major carriers serving a population of 269 million. Thus, the number of carriers per million people was almost 50 per cent higher in the EU relative to the United States (0.035 vs. 0.026). In this count, the EU carriers are Austrian, Sabena, SAS, Finnair, Air France, Lufthansa, Olympic, Aer Lingus, Alitalia, KLM, TAP, Iberia, British Airways. Major US carriers are American, United, Delta, Northwest, Continental, US Airways, and TWA.

EUROPEAN AIR TRANSPORT

3

The discussion starts in section 2 by showing how the old flag-carrier regime affected the structure of European airline networks. It is argued that this regime precluded the emergence of efficient hub-and-spoke networks, which concentrate traffic on relatively few routes. Instead, the old regime led to a profusion of point-to-point airline routes, with too many carriers providing service. By leading to relatively low traffic densities, this point-to-point system prevented European carriers from fully exploiting economies of traffic density, partly contributing to their high operating cost per passenger. Sections 3 and 4 argue that the traffic deficiency of European airlines was partly remedied in the 1990s by regulatory changes that occurred in parallel with the main course of EU deregulation. These changes allowed the emergence of immunized airline alliances along with the linked phenomenon of US open skies agreements, which were signed by a number of EU countries. Alliances and open skies agreements provided a notable stimulus to international traffic between the United States and EU countries, raising passenger flows within the networks of European carriers in a beneficial fashion. Section 5 explores the initial effects of EU deregulation itself, discussing the growth of low-cost carriers and the competitive threat that they constitute. The discussion argues that, to compete against these new entrants, major carriers need to restructure their route networks to achieve greater efficiency, while renegotiating costly labor contracts in the current US fashion. It is argued that the new freedom to pursue cross-border mergers provides one path to more rational route systems, with the new potentially larger carriers able to concentrate traffic in pan-European hub-and-spoke networks. Section 6 points out that negotiation of a Common Atlantic Aviation area may be prerequisite to such mergers. Such an agreement would eliminate the threat of losing US traffic rights following a merger, which may impede some otherwise attractive combinations of carriers. Section 7 argues that a reform of airport institutions may be needed to realize the full benefits of deregulation. The current rigid system for allocating airport slots must be replaced by a system capable of delivering slots to the carriers best able to use them, with a slot auction system being an attractive possibility. Airport congestion must also be attacked, either by appropriate use of the slot system or by congestion pricing. Finally, airport privatization may be required to ensure that airports operate efficiently, although the exercise of market power by privatized airports may be a concern.

2 EUROPEAN NETWORK STRUCTURE BEFORE DEREGULATION The structure of European airline networks in the period prior to deregulation was governed largely by geography and by the institutional and regulatory features of the old regime. Since the current network structure under deregulation partly reflects the heritage of the past, it is important to understand the sources of that heritage.

2.1 Hub-and-Spoke Versus Point-to-Point Networks To begin, it is helpful to discuss the nature of a common airline network structure that has emerged over the last several decades, especially in the United States. This

JAN K. BRUECKNER AND ERIC PELS

4

A

H B

C

Figure 1 HS and PP Networks.

structure is known as a hub-and-spoke (HS) network, and it is illustrated in Figure 1. For simplicity, suppose that an airline serves four cities, denoted H, A, B, and C, as in the Figure. To serve these cities, the airline could operate a point-to-point (PP) network, under which each pair of cities is connected by an airline route, allowing nonstop service in each city-pair market. Under a PP network, airline routes in Figure 1 would consist of both the solid lines and the dotted lines, with a total of six routes being operated by the carrier. Under an HS network, by contrast, the airline uses city H, which is centrally located, as a hub, and it operates just three routes, indicated by the solid lines in the Figure. While passengers in city-pair markets AH, BH, and CH still benefit from nonstop service, passengers in city-pair markets AB, BC, and AC must now make a connecting trip, changing planes at the hub H on the way to their eventual destinations. In the United States, prior to deregulation, the structure of airline networks was largely determined by regulators, who controlled entry and exit on individual routes. In the interest of providing convenient service to the public, regulators encouraged extensive provision of nonstop service, leading to a structure that resembled the PP network. With deregulation, however, airlines were free to choose the routes they served, and as a result, the pursuit of profit maximization dictated that routes be reorganized an economically efficient manner. Economic efficiency, along with desire to serve the full range of city sizes, dictated the formation of HS networks. While such a network obviously allows a carrier to operate fewer routes, as seen in Figure 1, the true source of the efficiency gain is a phenomenon known as “economies of traffic density.” With economies of density, cost per passenger falls on an airline route as the traffic volume on the route rises. This effect arises in part because high traffic volumes allow the use of larger aircraft, which have a lower cost per seat mile.2 In addition, the fixed cost of airline operations at the endpoints of the route (operation of ticket counters and other ground facilities) can be spread over more passengers as traffic density rises.

2 Without adjustment of aircraft sizes or flight frequencies, larger traffic volumes translate into higher load factors, which also reduce cost per passenger. Average load factors indeed rose following deregulation.

EUROPEAN AIR TRANSPORT

5

By concentrating traffic on the spoke routes in and out of the hub, the HS network reduces cost per passenger on these routes.3 Because of this cost reduction, the cost of transporting passengers in city-pair markets AH, BH, and CH, who make nonstop trips, clearly falls relative to the PP case. However, passengers in the remaining city-pair markets, who must connect at the hub, have longer flight distances than under the PP network. But, because the cost of carrying these passengers along the spoke routes is relatively low due to high traffic densities, the overall cost of transporting them is likely to be lower than under the PP network. The upshot is that the total cost of carrying passengers among the six cities in Figure 1 will be lower under the HS network than in the PP case. An additional benefit of the HS structure is that the high traffic volumes on the spoke routes allow an increase in flight frequency relative to the PP case. Offsetting this gain, however, is the reduced convenience of travel for connecting passengers, who could make a nonstop trip under the PP network but undergo a time-consuming transfer at the hub under the HS network.4

2.2 European Network Structure Under the Regulated Regime With this background, consider the structure of European airline networks under the regulated regime. Initially, it is useful to focus just on intra-European traffic, considering intercontinental traffic later. First, observe that, as European carriers were public enterprises, their incentives for profit maximization were relatively weak. Losses could be covered by government subsidies, and any profits accrued to the government and not to private owners. As a result, European carriers had little incentive to hold down labor costs, allowing their workers to enjoy the benefits of an uncompetitive environment and unlimited government support. In addition, the carriers had little incentive to achieve operational efficiencies. Against this backdrop, airline operations within Europe were governed by an extensive regulatory structure. Airline service between any two European countries was regulated by a bilateral agreement between the two countries. These agreements typically specified the routes that could be flown and the allowable capacities on these routes. The identities of the carriers providing service were also specified, with the chosen carriers usually being the two flag carriers of the countries involved. With this structure ruling out competition between the carriers on routes between European countries, and with little concern for the magnitude of profit, air fares were set in a mechanistic fashion. Generally, fares corresponded to those set under fare “conferences” organized by the International Air Transport Association (IATA). At these conferences, carriers determined mutually agreeable fares for tens of thousands of

3 A downside to hub operations that has gained prominence recently is the potential for reduced aircraft

utilization, a result of the need to expand aircraft ground time at the hub in order to facilitate passenger

connections. However, this effect does not offset the many advantages of HS networks.

4 For a discussion of the effects of US deregulation on airline networks, see Morrison and Winston (1985,

1995). For a discussion of economies of traffic density as well as empirical evidence, see Brueckner and

Spiller (1994) and Caves et al. (1984).

JAN K. BRUECKNER AND ERIC PELS

6

G

E

D A

C

B

F

Figure 2 European Airline Networks.

international city-pair markets. The pricing of intra-European trips under the old regime relied mechanically on these IATA fares.5 As a result of the web of bilateral agreements between European countries, each flag carrier operated a radial route network connecting its home city to the major cities of the other countries, as seen in Figure 2. In the Figure, each square represents a different European country, with the major cities indicated by A, B, C, D, and E. For convenience, let the countries be identified by a lower case letter matching the given city, so that city B is contained in country b, and so on. As can be seen, country a’s flag carrier operates routes from its home city A to the major cities B, C, D, and E of the other countries, with country b’s flag carrier serving A, C, D, and E from its home city B (these latter routes are indicated by the dotted lines in the Figure). As a result, city-pair market AB is served by the flag carriers of countries a and b, and similarly for other markets. In this market, each carrier’s flight capacity is governed by the bilateral agreement between countries a and b, and fares are set at the IATA level. Country a’s flag carrier also serves domestic endpoints within that country, as seen in Figure 2 (these cities are unlabeled and shaded in grey). Routes to Africa (country f)

5

For a discussion of IATA fares, see O’Connor (1989).

EUROPEAN AIR TRANSPORT

7

and the United States (country g) are also shown in the Figure, but these are considered after the discussion of intra-European traffic patterns.

2.3 HS Networks in Europe? The route network operated by country a’s flag carrier is clearly radial in nature, with routes emanating from the home city A to many destinations, both outside the country and within it. While the network thus seems to resemble the HS network of Figure 1, a question is whether the network indeed functions in the HS manner, with the carrier transporting significant volumes of connecting passengers who change planes at city A. The answer to this question is negative: despite their radial structure, European route networks prior to deregulation did not function as true HS networks, carrying large volumes of connecting traffic. Instead, these networks functioned mostly as point-to­ point networks, with connecting traffic apparently modest in volume. Several observations help to explain this pattern. First, the pattern of flag carrier service between countries meant that a given country’s carrier could not attract connecting passengers flying between a second and a third country. To understand this point, observe that while country a’s airline could provide connecting service between B and D via city A, passengers in the BD city-pair market already enjoyed nonstop service between these cities, which was provided by the flag carriers of countries b and d. As a result, a connecting trip on country a’s airline would hold little attraction. While this conclusion might been have been overturned in a competitive environment, where country a’s carrier could have attempted to attract BD connecting passengers by substantially undercutting the nonstop BD fare, the weak profit motive felt by flag carriers would have made such an action unlikely. Another potential group of connecting passengers, those traveling between domestic cities within country a, would also find such a trip unattractive. Two key features of the European setting account for this conclusion. First, compared to the United States, the spatial size of European countries is relatively small. As a result, domestic travel between different cities within country a may involve a relatively short distance, making airline travel unappealing, a conclusion that applies even more strongly to a circuitous, inconvenient connecting trip through city A. The effect of relatively short domestic distances is compounded by the availability of widespread and convenient rail service within Europe. Rather than flying between two domestic endpoints, a preferred choice would be to simply take the train, using a route indicated by the curved line in Figure 2. These obstacles to connecting airline travel by domestic passengers were compounded by the nature of the pricing environment. Because of the weak profit motive felt by flag carriers, they had little incentive to make connecting trips more attractive by offering relatively cheap fares. By contrast, a third group of passengers represented more plausible candidates for connecting intra-European air travel. This group consists of passengers traveling between a small city in one country and a city in a second country, either large or small. For example, a passenger traveling between one of the small domestic endpoints in country a, shown in Figure 2, and city C in country c would find a connecting trip via city A on country a’s flag carrier to be an appropriate choice. Similarly, a passenger traveling

8

JAN K. BRUECKNER AND ERIC PELS

between two small endpoints, one in country a and one in country c, would need to make a connecting trip using both flag carriers. The passenger originating in country a would change planes at city A and would change both planes and airlines (switching to country c’s carrier) at city C. Because both types of connecting passengers make international trips within Europe that involve at least one small endpoint, their total number was likely to be relatively small compared to the total volume of intra-European traffic. As a result, connecting traffic within Europe under the old regime was undoubtedly of limited importance. With connecting traffic limited, European airline networks thus functioned mainly as pointto-point networks, with HS operations of little importance despite the radial form of the networks. The US air travel market, by contrast, offers much greater scope for HS networks, mainly as a result of a different geography. First, while the relatively compact size of Europe means that many major cities are so close together that a circuitous con­ necting trip would be unacceptable, the spatial expanse of the United States leads to a greater average distance between cities. Greater distances tend to reduce the circu­ ity of connecting trips, with layover time also being less significant compared to total travel time. Second, the US population, which is comparable in size to that of Europe, lies within a single national boundary. There is thus no analog to flag carrier system, which automatically generates nonstop service between most pairs of major European cities. As a result, even in the US city-pair markets involving relatively large end­ points, nonstop service may not be available, with passengers forced to rely instead on connecting travel.6 Third, the large physical size of the United States, as well as the limited nature of rail service, means that air travel is usually necessary for trips between one small domestic endpoint and another, unlike in the European case. Such travel by necessity requires a connecting trip. Moreover, since these small-endpoint trips occur between cities in the same country, they presumably involve larger pas­ senger volumes than for analogous trips in Europe, which are often international in nature. Thus, connecting passengers in the United States come from two groups of travelers who, in Europe, would enjoy nonstop service or shun air travel altogether: passengers traveling in some city-pair markets involving medium and large size cities, and travelers making trips between small endpoints. The presence of these groups of passengers allows HS networks to play a more important current role in the United States than they did under the old regime in Europe. The point-to-point nature of European airline networks under the old regime was undesirable from an efficiency perspective. In effect, these networks involved the oper­ ation of too many airline routes. With traffic dispersed over this large number of routes instead of concentrated on fewer segments, European carriers were unable to fully exploit economies of traffic density. One result was a higher cost per passenger than could have been achieved under a more efficient HS-style route structure. This cost escalation compounded the underlying problem of high labor costs, which resulted from union

6

An example might be the Boston-Portland, OR city-pair market.

EUROPEAN AIR TRANSPORT

9

power coupled with public ownership of the carriers. The upshot was notoriously high airline operating costs throughout Europe.7 A second deleterious effect of the inadequate traffic densities caused by the pointto-point network structure lay in the area of service quality. As mentioned above, one byproduct of large traffic densities is high flight frequency, which raises the convenience of air travel. By depressing densities, reliance on a PP route system imposed a cost in this dimension of passenger convenience. It is important to note that the source of these inefficiencies lies both in geography and in the fundamental institutional aspects of the old regime, neither of which was easily changed. The fact that Europe is divided into separate nations, with each naturally operat­ ing its own flag carrier under the old regime, helped to predetermine the nature of airline networks, leading to an excessive number of airlines and airline routes. This outcome, com­ bined with the relative unattractiveness of domestic air travel within individual countries (a consequence of a compact geography and good rail service), helped to depress traffic densities, leading to high cost per passenger and relatively low flight frequencies.8 Mirroring the flag-carrier system of the old regime, the European air traffic control (ATC) system was similarly balkanized. Each country operated its own ATC authority, and control over each international flight within Europe was handed from one ATC authority to another as the flight progressed through European air space. Relative to a system like that in the United States, which is uniform across a broad geographical area, the presence of many separate national ATC systems introduced various inefficiencies. Coordination problems between the different systems contributed to the problem of flight delays within Europe. Moreover, the sovereignty of each country over its own air space and existence of many restricted military areas undoubtedly tended to generate inefficient flight paths, with greater circuity than necessary. Both effects contributed to high airline operating costs as well as reducing the convenience of air travel within Europe.

2.4 Intercontinental Aspects of Network Structure for European Carriers Under the old regime, Europe’s flag carriers operated many international routes to other continents, with the intercontinental routes to North America being the most important. Service on these routes was governed by bilateral agreements similar to those existing between European countries, with the routes and carrier identities specified along with flight capacities. While the observations on network structure based on Figure 2 sometimes remain relevant in the intercontinental case, important exceptions arise. The most important 7 Oum and Yu (1998) offer evidence on cost differences between EU and US carriers. They compute a composite output measure, which represents passenger, freight, and mail volumes carried by the airlines, and divide total input cost by this measure. For 1995, the resulting average unit cost was 0.95 for US carriers and 1.28 for EU carriers, for an EU cost premium of more than 30 per cent (these numbers are scaled so that the value for American Airlines is 1.00). Higher EU costs are due to the combined effects of higher input costs and lower productivity. Marin (1998) computes productivity measures for EU and US carriers, and under one measure, technical efficiency for the period 1985–1989 averaged 0.83 for US carriers and 0.69 for EU carriers. 8 For further discussion of European industry under the old regime, see Doganis (1985, 2001), Good et al. (1993), McGowan and Seabright (1989), and Neven and Roller (1996).

10

JAN K. BRUECKNER AND ERIC PELS

observation is that, because of the distances involved and weak condition of many non-European flag carriers, these carriers provided spotty or nonexistent service to many important destinations outside their home countries. This fact provided service opportunities for European carriers that did not exist in the case of intra-European traffic. To understand this point, return to Figure 2, and consider the intercontinental routes from country a to the African country f and the United States (country g). While a bilateral agreement may have existed between the African country and the United States, no carrier from either country may have provided the service that the agreement allowed. However, country a’s flag carrier may have served the African country under its own bilateral, while also serving the United States, as shown in the Figure. In this situation, that carrier could provide connecting service from the African city F to the US city G via its home airport in A.9 Such connecting service by European carriers to countries with weak flag carriers appears to have been commonplace, and it may have involved Western endpoints in Europe rather than in the United States. In this case, the endpoint G would instead be a city in another European country, whose relatively small flag carrier did not serve the African country. Country a’s large flag carrier could then have provided connecting service linking the African country to the European neighbor. Such connecting service is beneficial from a network perspective, helping to raise traffic densities for a’s flag carrier on important routes like that to city G. However, the volume of traffic involved is likely to be low given that the other endpoint is in Africa, or some similar location, that does not generate or attract much traffic compared to the United States. As a result, the salutary network effects of this connecting traffic were likely to have been small. Thus, the previous conclusion that European airline networks functioned largely as point-to-point networks is largely unaltered when intercontinental routes are considered.

3 THE IMPACT OF AIRLINE ALLIANCES 3.1 The Emergence of Alliances In the 1990s, the old regime of European air transport was altered by several new develop­ ments. The first of these changes, which coincided with the initial major steps in European airline deregulation, was the emergence of international airline alliances. The ultimate effect of these alliances was to raise the number of intercontinental passengers carried by Euro­ pean airlines, with beneficial effects on their traffic densities and hence costs per passenger (and ultimately profit). The growth of alliances, however, has generated regulatory con­ cerns both in Europe and North America. As discussed further below, these concerns have been resolved mostly in favor of the alliances, allowing their growth to proceed. The fundamental force driving the emergence of international alliances is globalization of the world economy, which has spurred intercontinental business travel while also stimulating leisure trips. In competing for this new breed of international passengers,

9 An example of this phenomenon is air service to and from India, which appears to be disproportionately provided by non-Indian carriers, despite the enormous size of the country.

EUROPEAN AIR TRANSPORT

11

airlines have sought to enhance the convenience and attractiveness of intercontinental trips. An obstacle to achieving this goal, however, is the fact that many international trips cannot be carried out using just one airline. Travelers are thus forced to make an “interline” trip, typically flying on two airlines (and occasionally more than two). This need for interline travel arises because no existing airline is large enough to serve most of the world’s endpoints. While the desire to better serve international passengers creates an incentive to build such an airline through cross-border mergers, airline regulation has historically ruled out such combinations, even though the last round of European deregulation makes intra-EU mergers feasible, as discussed further below. Short of a merger, airlines can try to serve more international destinations by extending their own route networks, but such efforts are hampered by unwillingness to acquire the necessary equipment and labor force and by existing bilateral agreements, which limit the number of carriers that can provide service on any given international route. With these avenues to improving international service blocked, airlines instead attempted to improve the quality of interline trips by forming alliances. Under a typi­ cal arrangement, the alliance partners attempt to coordinate their schedules in order to ease interline connections at gateway airports. While this coordination reduces passen­ ger layover times, the airlines have also strived to rearrange gate facilities to shorten walking distances. Alliance partners have also worked to improve baggage transfers for their passengers, reducing the problem of mishandled luggage that plagues traditional interline travel. Finally, the frequent flier programs of the partner airlines are typically merged, allowing passengers to earn more miles than under a usual interline trip and giving elite members reciprocal access to the alliance partners’ airport lounges. All of these changes serve to make interline travel more like a trip on a single airline, and the resulting improvement in travel convenience has allowed alliances to capture a growing share of international traffic. The major alliances are built around pairings of large US and European carriers. The key partners are United and Lufthansa for the Star Alliance, American and British Airways for the Oneworld alliance, Delta and Air France for the Skyteam alliance, and Northwest and KLM for the “Wings” alliance.10

3.2 The Effect of Alliances on Fares Alliances also generate economic benefits for interline passengers by lowering the fares they pay. The fact that airline cooperation reduces, rather than increases, interline fares may appear counterintuitive. However, the reason for this outcome in the case of interline trips is that such travel is a “joint product” resulting from the combined efforts of two carriers. Economic theory shows that cooperation between the providers of a joint product leads to a price lower than the one emerging under noncooperative behavior. To understand this point more fully, note that the airlines relied on IATA fares in pricing traditional interline trips. Such fares can be viewed as the result of noncooperative

10 It is expected that the latter alliance will be blended into the SkyTeam alliance as a result of the Air France–KLM merger.

12

JAN K. BRUECKNER AND ERIC PELS

behavior, where each airline specifies (in the context of an IATA fare conference) the amount it requires to carry a passenger over its portion of an interline journey, with the total interline fare equal to the sum of these amounts for both airlines. The problem with this fare-setting process is that, in determining its own required revenue from an interline passenger, an airline does not consider that a high revenue requirement hurts the other airline by raising the overall fare for the trip, which in turn depresses traffic and reduces the other airline’s profit. If the airlines were instead able to cooperate in setting the interline fare, with a goal of maximizing their joint profit, each would recognize the harm done to the other when it attempts to extract extra revenue from the interline passenger. Each airline would then restrain its own pursuit of higher revenue, and the overall interline fare would fall. Moreover, the combined profits of the carriers would rise relative to that earned under the IATA fare.11 In order to engage in this kind of cooperative pricing of interline trips, the carriers must enjoy “antitrust immunity,” which legalizes interfirm cooperation that would otherwise be disallowed. Such immunity is granted formally by the US regulatory authorities and through a less-formal process by the European Commission. Antitrust immunity is granted to carrier pairs and not to alliances more generally, and most of the pairings in the key alliances are immunized. With antitrust immunity leading, via airline cooperation, to lower interline fares, the benefits of alliance travel are enhanced. With lower fares and greater convenience reducing the full economic “cost” of interline travel for the passenger, the volume of such trips has grown in step with the expansion of alliances.12 As a result, traffic flows within the networks of the European alliance partners have expanded, and the resulting gains in traffic density have reduced cost per passenger and enhanced airline profits. To better understand the pattern of alliance traffic, consider Figure 3, which also highlights the regulatory concern that alliances have generated. Suppose that a US passenger wants to travel from city I, a small or medium-size endpoint, to city J overseas, which is not served by a US carrier. To do so, the passenger would fly on a US alliance member from I to city G, the airline’s hub, connecting to one of the carrier’s transatlantic flights to city A, the home airport of the carrier’s European alliance partner (the US airline’s routes are shown as dotted lines). At city A, the passenger would then connect to one of the partner’s flights to J. Note that while Figure 3 shows city J as being located in a third country, it could alternatively be located in country a itself. Note also that the trip pattern would be similar if the origin for the US passenger were the hub city G rather than the smaller endpoint I (in both cases, travel on the two airlines would be required). Finally, observe that a key feature of alliances is implicit in Figure 3. In particular, alliances effectively link the networks of two different carriers, making a trip within the combined network equivalent to a trip on a single airline.

11 For empirical evidence on this fare effect as well as a general discussion of the economics of alliances, see Brueckner and Whalen (2000). 12 For evidence on interline traffic growth, see US Department of Transportation (1999, 2000) (both studies can be found on the DOT website). As an example of the kind of data presented, the 1999 study shows that traffic between Northwest’s US gateways and Amsterdam, the hub of its partner KLM, increased nine-fold between the pre-alliance year 1992 and 1998. However, origin-destination traffic on these hub-to-hub routes showed only a modest increase, testifying to the huge surge in interline traffic resulting from the alliance.

EUROPEAN AIR TRANSPORT

13

G I

A

J

Figure 3 Travel on an Airline Alliance.

3.3 Regulatory Concerns Engendered by Alliances While clarifying the nature of a typical interline alliance trip, Figure 3 also shows a feature of alliances not considered up to this point. In particular, the Figure shows that, because the European alliance partner also serves the route between G and A, the two airlines provide overlapping service on this route. While this fact means that the US interline passenger could just as well have used the European alliance partner for both the transatlantic portion of his journey and the onward flight to J, this overlap has broader implications. In particular, the overlap may have consequences for a different group of passengers, namely, those making nonstop trips between the major cities G and A. These passengers obviously can make their journey on one airline or the other, having no need for interline travel. Normally, this choice would enhance a passenger’s prospects, with competition between the two carriers guaranteeing an affordable fare. However, antitrust immunity gives the carriers full scope for cooperation in the fare-setting process, and on a route where overlapping service is provided, this cooperation may be anticompetitive. In other words, the carriers’ license to cooperate may be used in a collusive manner in the AG city-pair market, with the carriers raising the fare in an anticompetitive fashion, knowing that passengers may have no alternative choice of service. This concern has motivated regulatory action on alliances. The European Commission recently gave its approval to the Star and Wings alliances after a multi-year inquiry, rec­ ognizing that the AG-type markets were relatively small in each case, and that mild mea­ sures could address anticompetitive concerns. By contrast, antitrust immunity for Ameri­ can and British Airways has been denied by US regulators. The regulators argued that the large size of the AG-type overlap markets, which consist of the heavily traveled routes between US gateways and London’s Heathrow airport, meant that losses from potential

JAN K. BRUECKNER AND ERIC PELS

14

anticompetitive behavior by the alliance partners could be substantial. Rather than being mild, the proposed remedy was so draconian (involving a substantial slot divestiture at Heathrow) that the carriers rejected it, settling instead for an unimmunized alliance. International alliances will prosper as long as the regulatory environment prevents cross-border mergers between the US and European carriers. In the absence of such mergers, alliances provide the only means by which airlines can compete for a larger share of international traffic. While alliances have not generated a fundamental change in the nature of European airline networks, they have led to a beneficial growth in traffic densities within the existing networks of the alliance partners. This growth has reduced, but not eliminated, the problem of inadequate traffic flows, which is caused, as explained above, by route proliferation under the flag-carrier regime as well as the compact geography and good rail service of European countries.

4 THE EFFECT OF OPEN SKIES AGREEMENTS The 1990s witnessed the signing of a host of “open skies” agreements between the US and European countries. In a typical case, an open skies agreement completely eliminates the capacity and route restrictions of the prior bilateral agreement. The US carrier is then allowed to provide unlimited service to any endpoint in the other country, and that country’s carrier(s) are allowed to fly anywhere in the United States, with capacities and frequencies of their choosing. In addition, the most-liberal open skies agreements provide unlimited “beyond” rights (or fifth freedom rights), allowing one country’s carrier(s) to provide continuing service beyond the other country to additional destinations, service that may be used both by the US passengers and local passengers originating in the other country. The proliferation of open skies agreements is intimately tied to the growth of airline alliances. In particular, as a condition for signing such an agreement, the European country typically demands that the US regulators grant antitrust immunity to the country’s flag carrier and its US alliance partner. This requirement grows out of a fear that the much larger size of the US carriers will confer an unfair advantage under open skies unless a mechanism exists to provide the smaller European carrier, which may lack the resources to massively expand service, with equivalent effective access to US endpoints. Antitrust immunity, which effectively allows the two carriers to act as a single airline in providing interline service, achieves this goal. Traffic between the United States and the open skies signatories grew more rapidly in the 1990s than on other international routes, partly reflecting the elimination of service restrictions.13 But this traffic growth partly reflects the favorable effects of immunized alliances themselves, effects that arise only because open skies and antitrust immunity are linked. As mentioned above, the traffic growth associated with open skies and alliances has been beneficial for European carriers, helping to raise traffic densities and generally strengthen their operations. Moreover, even though the beyond rights associated with

13

See US Department of Transportation (1999, 2000).

EUROPEAN AIR TRANSPORT

15

open skies agreements raised potentially negative consequences for European carriers, who stood to lose traffic that they previously carried, this outcome has not materialized to any significant degree. Rather than exercising their beyond rights, US carriers typically relied instead on their alliance partners to provide such service, deploying their resources elsewhere.

5 THE IMPACT OF EUROPEAN DEREGULATION Following the lead of the United States, Europe in the late 1980s launched its own process of airline deregulation. The process proceeded in stages, with a sequence of three deregulation “packages” introduced by the EU over the succeeding decade. Deregulation culminated with the “third package,” introduced in 1993, which by 1997 removed the last restrictions limiting the activities of European carriers. Currently, European airlines enjoy complete pricing freedom and the freedom to enter and exit routes anywhere in the EU, including domestic routes in another country. In addition, previous prohibitions on cross-border mergers within the EU were removed, so that the old flag-carrier regime, where airlines are associated with particular countries, can in principle be replaced by a system of broader ownership. In effect, European carriers now enjoy exactly the same freedoms within the boundaries of the EU as do carriers within the United States, despite the presence of the European national borders. The response to deregulation proceeded slowly. The initial liberalization in the first half of the 1990s apparently produced little effect, with route structures and fares showing little change relative to the old regime.14 By the end of the 1990s, however, dramatic impacts of the new regime were becoming evident. The most striking change was the launching and subsequent explosive growth of low-cost carriers, especially EasyJet and Ryanair, both based in the British Isles. The growth of these carriers was partly fueled by acquisition of other, less-successful low-cost operators, although a number of these lesser carriers still compete for business. The low-cost carriers have followed the model of Southwest Airlines in the United States by relying on flexible work rules to generate high labor productivity, by flying just one or two aircraft types to economize on maintenance and crew training, by emphasizing fast aircraft turnarounds to maximize daily usage hours, and by serving large city-pair markets but doing so from secondary airports. This airport strategy avoids the congestion that plagues major European airports, facilitating the carriers’ quick-turnaround standard, and it also economizes on airport charges, which are lower at secondary airports. In the United States, low-cost carriers mostly operate point-to-point networks. For example, although some Southwest passengers make connecting trips, the airline appears not to explicitly schedule its operations to facilitate connections.15 Because of this point-to-point strategy, Southwest is unable to serve small endpoints, which would not generate enough traffic to justify point-to-point operations. Such service is instead left to

14

See Commission of the European Communities (1996) for details.

Although JetBlue’s network also offers mainly point-to-point service, AirTran and ATA appear to rely

more on connecting passengers. For a discussion of Southwest’s service patterns and market-entry decisions,

see Boguslaski, Ito and Lee (2004).

15

16

JAN K. BRUECKNER AND ERIC PELS

the network carriers, which link small endpoints to their hub airports. Because all traffic to and from the small endpoint, regardless of its origin or destination, travels along the spoke route to the hub, the volume is large enough to justify service by the network carrier. By shunning such endpoints, low-cost carriers in the United States thus follow a “cherry-picking” strategy, serving only the most attractive markets. Their European counterparts, which also favor a point-to-point style of operation, have in effect adopted the same strategy. The likely impact of this type of competition on the major EU carriers provides a key to predicting the subsequent course of European deregulation. Some clues as to the effect of low-cost competition come from observing the US case. Evidence for the United States shows that, in attempting to preserve market shares, network carriers dramatically reduce their fares in city-pair markets also served by lowcost carriers, despite their cost disadvantage. Because low-cost competition has spread to an ever greater number of the network carriers’ markets, the result has been severe downward pressure on their profits. This pressure, combined with the effects of the recent overall slump in air travel, has helped push several major US airlines into bankruptcy, while threatening other network carriers with the same fate.16 With the US example providing guidance, it is possible to speculate about the likely effect of the low-cost revolution in Europe. Recall from above that, under the old regime, the flag-carrier system, geography, and intermodal competition led to the operation of too many airlines routes, most with inadequate traffic densities. Low densities, compounded by high labor costs and various operating inefficiencies, in turn led to exorbitant operating costs for European airlines. With such costs, EU carriers had to rely on high IATA fares to avoid substantial losses. This negative picture was improved somewhat in the 1990s by the traffic stimulus provided by international airline alliances. Moreover, the trend toward full or partial privatization of EU carriers, which has proceeded apace with deregulation in the 1990s, has strengthened the profit motive, and helped to hold down labor costs at a number of airlines. But the growth of low-cost competition is likely to produce the same dramatic impact on the fortunes of EU carriers as has occurred in the United States.17 First, by draining traffic in the large city-pair markets out of the major carriers’ net­ works, low-cost competition will exacerbate the problem of inadequate traffic densities, putting upward pressure on cost per passenger. Second, as the major carriers attempt to cut their fares to stem the traffic loss, the resulting downward pressure on revenue will interact with higher costs to cut profits. The picture is thus similar to the US case, but the EU carriers’ plight is compounded by their lower operating efficiency relative to US airlines. One effect of these developments is likely to parallel the US experience. In particular, EU carriers are likely to attack the problem of high labor costs by asking for wage concessions from their workers. Such concessions have been extracted mostly through

16

For evidence on the competitive effects of low-cost carriers in the United States, see Morrison (2001). The more favorable profit positions of EU relative to US carriers in recent years poses a puzzle given their higher operating costs. Possible explanations include the lower EU penetration of low-cost carriers relative to the US case, which results in less competitive pressure, and the greater reliance of EU carriers on transcontinental traffic, which tends to generate a higher profit than domestic traffic. 17

EUROPEAN AIR TRANSPORT

17

the bankruptcy process in the United States, although American Airlines gained a broad reduction in the wages of its workforce through a credible threat of bankruptcy. However, greater labor militancy in Europe relative to the United States may make this process more difficult, and its result less effective, than in the US case. A second likely response to low-cost competition is a push for greater operating efficiency through cross-border mergers between EU carriers. By allowing replacement of the flag-carrier system, such mergers would allow a rationalization of European route networks. The current proliferation of airlines and routes would be reduced, with the merger partners reorganizing their point-to-point operations in favor of US-style HS networks. Traffic densities would rise, reducing cost per passenger and improving profits. Greater densities would in turn lead to higher flight frequencies on key routes, although some passengers would be forced to make more circuitous connecting trips. The Air France–KLM merger is likely to generate some of these beneficial effects, at least in the long run. But these gains may come at the expense of a reduction in competition on routes that were jointly served by the two carriers prior to the merger. However, the widely perceived need for consolidation of the industry led EU regulators to discount these possible negative effects in approving the merger.18 Through efficiency improvements, industry consolidation may lead to some reduction in the current disadvantage EU carriers face relative to the low-cost competition, putting the airlines more or less in the situation of the US network carriers prior to the latest upheaval. But ultimate survival in the midst of the European low-cost revolution requires more draconian cost reductions of the kind currently being secured by US airlines. Whether EU carriers will be able to gain such reductions is an open question. It should be noted that European deregulation is likely to have an impact on a segment of the aviation sector that lies mostly outside the purview of government regulators: charter operations. Partly in response to high European airfares under the old regime, a substantial share of leisure passengers used charter flights rather than scheduled service to reach their vacation destinations. With deregulation putting downward pressure on fares, it is likely that leisure travelers will increasingly opt for scheduled air service rather than using charter flights. However, this change will unfold gradually as the effects of deregulation take hold. Finally, EU deregulation has been accompanied by planned changes in Europe’s air traffic control system. The changes are designed to foster greater coordination between ATC personnel in different countries under a proposal known as Single European Sky, thus reducing delays and eliminating excessive flight distances. It is expected that the proposal will be implemented in 2004, although its effectiveness in reducing ATC fragmentation in Europe remains to be seen.

6 THE ROLE OF A COMMON ATLANTIC AVIATION AREA While internal deregulation of transport in the EU is now complete, intercontinental service outside of the EU is still governed by the various bilateral agreements, some of 18

See Brueckner and Pels (2005) for an analysis of these anticompetitive effects, which include those resulting from consolidation of the Northwest–KLM and SkyTeam alliances.

18

JAN K. BRUECKNER AND ERIC PELS

which have been liberalized through open skies agreements. This intercontinental traffic is exceedingly important for EU carriers, accounting for a much greater share of their total traffic than in the case of US carriers.19 Thus, internal airline deregulation within the EU affects a smaller share of the airline sector than did US deregulation. Accordingly, most observers argue that the last step in the deregulation process must be elimination of the remaining restrictions on the important North American routes through creation of what is known as a Common Atlantic Aviation Area. Under this proposal, individual bilateral agreements would be replaced by a single agreement governing traffic between the United States and the EU as a whole. Anticipating such an agreement, the European Court of Justice ruled in a widely noted 2002 decision that existing bilateral agreements are illegal under EU law because they award US service rights only to the given country’s flag carrier, effectively discriminating against other EU airlines (in other words, the agreements contain a “nationality clause”). Under a common aviation area, this restriction would disappear, with any EU carrier able to provide US service from any European endpoint. At first glance, such new freedom would appear to hold little value for European carriers. For example, a carrier like Lufthansa would appear to have little incentive to provide US service from Paris, a city where Air France operates most of the flights. With limited Lufthansa operations in Paris, few connecting opportunities would be available for the airline’s passengers, making US service unattractive. Since other EU carriers similarly lack the incentive to initiate service from the home airports of other airlines, the gain from eliminating the nationality clause in existing bilaterals would not appear to be substantial. This argument, however, overlooks the effect of the nationality clause on the incen­ tives for cross-border mergers within the EU. The problem is that, because bilateral agreements give traffic rights to a country’s national airline, another carrier acquiring control of that airline through a merger may lose these traffic rights, thus being unable to replicate existing service to the United States. Although some remedy might ultimately be available in such a case, uncertainty about the disposition of international traffic rights greatly reduces the incentive for airline mergers within the EU.20 However, if existing bilaterals were replaced by a common aviation area, with any EU carrier able to operate any route to the United States, then this merger disincentive would be eliminated. An acquiring carrier would be free to operate all of its merger partner’s previous US routes, removing the potential merger penalty inherent in the current system.21 Cross-border mergers hold the key to survival of many major European carriers in the face of the ongoing low-cost revolution, and a key ingredient to facilitating such mergers is the kind of route-authority liberalization inherent in a common aviation area.

19

See Good et al. (1993).

This uncertainty appears to partly explain the structure of the Air France–KLM merger, where the two

carriers will initially operate as separate entities but under common ownership. This arrangement does not

jeopardize KLM’s traffic rights to the United States.

21 For a discussion of the effect of a common aviation area on potential European mergers, see Brattle Group

(2002).

20

EUROPEAN AIR TRANSPORT

19

Until such an agreement is in place, the mergers that are needed to achieve consolidation of the European industry may be delayed.22

7 THE INTERACTION BETWEEN AIRPORT OPERATIONS AND AIRLINE DEREGULATION The deregulation of European air transport has the potential for leading to dramatic improvements in the functioning of the aviation sector within the EU. However, full exploitation of the benefits of deregulation may be blocked if the operating procedures and pricing policies of EU airports are not reformed. The airport slot allocation system, airport congestion, and the determination of airport charges are key issues that may help determine the course of deregulation in the EU.

7.1 The Slot Allocation Mechanism As argued above, Europe currently has too many carriers and too many routes. In a fully deregulated market, these problems would vanish over time as redundant routes are dropped and inefficient carriers disappear, either through bankruptcy or mergers. These developments would mean that some airports now serving as the hubs of smaller flag carriers would lose traffic, while the secondary airports served by low-cost carriers would gain passengers. In addition, as these carriers gain an ever-larger share of European traffic, pressure will build to extend their services to the major airports. Pressure to increase traffic at the major endpoints will also come from the flag carriers (or their descendants created via mergers) as these carriers attempt to concentrate traffic in more efficient, HS-style route structures. While the demand for capacity thus can be expected to grow at the larger airports, this outcome may be blocked by the current slot allocation system. This system controls landing rights at the great majority of European airports, with a carrier needing a landing slot for a particular time of day in order to operate a flight at that time. The problem is that slots are allocated using “grandfather rights.” In other words, carriers that used their slots last year have the right to continue using the slots this year. As a result, current slot allocations reflect a heritage from the past, with slot holdings largely reflecting past allocations to the pre-deregulation flag carriers. This slot allocation system implies that inefficient, high-cost airlines can have access to an airport even though a new low-cost carrier or an efficient, former flag carrier could use the slot much more productively. To fully realize the benefits of deregulation, the slot allocation system must avoid this outcome by allocating slots to the carriers best able to use them. Under a market system, such a carrier would be one willing to pay the highest amount to acquire the slot. Given this fact, an efficient allocation system could rely on the price mechanism, auctioning scarce airport slots to the highest bidder. Since airlines are currently granted the rights to use specific slots but do not actually own them, such an auction system is

22

As of mid-2006, an agreement had not been reached, reached, despite intensive negotiations by EU and US officials.

JAN K. BRUECKNER AND ERIC PELS

20

institutionally feasible. By contrast, if the airlines themselves had actual ownership of the slots, such a system would not be workable. A slot auction system will generate substantial revenues, and a key question is who will receive these revenues or, equivalently, who will organize the auction. Individual airport authorities could acquire control over the slots and thus the right to organize an auction, but as discussed below, this arrangement may give the airports considerable market power. Moreover, at a number of airports in Europe, the national government currently limits the number of available slots, in which case it would be natural for the government to organize the auction. The revenues from slot auctions could be used to finance capacity, or to invest in other airport facilities that improve passenger benefits (for instance, airport accessibility). Alternatively, it could be argued that since airlines need matched pairs of slots, one at the origin airport and one at the destination airport for a given flight, the auctions should be implemented at a European level. While such “network auctions” are theoretically very complex, they are in essence no different from the spectrum-rights auctions held in the United States, which were generally considered to be a success.23 The alternative of uncoordinated slot auctions by individual national governments would be an improvement over the current slot allocation system. But since such auctions would not take into account the carriers’ need for matched pairs of slots at airports in different countries, the resulting slot allocation may not be fully efficient, preventing the full benefits of deregulation from being realized.

7.2 Airport Congestion Airport congestion may also reduce the benefits of airline deregulation. When a lack of airport capacity causes delays, airlines and passengers incur congestion costs in the form of higher operating expenses and wasted personal time. At slot-constrained airports, congestion is determined partly by the slot allocation system, which assigns slots by time of day and thus determines the daily time pattern of airport usage. Given this fact, it could be argued that excessive congestion at European airports is a result of a failure of the slot allocation system, with too many slots allocated at peak periods at a number of airports. However, other factors may contribute to existing congestion levels, absolving the slot system from some of the blame. For example, congestion at a smaller airport that is not slot-constrained may cause a flight from that airport to arrive late at a large airport, disrupting the pattern of arrivals and causing excess congestion. Delays due to in-flight congestion of the airspace, whose management is the responsibility of the air-traffic control system, may similarly cause late arrivals, disrupting traffic and generating airport congestion. Regardless of the apportionment of blame for congestion at EU airports, it must be recognized that, because of the high demand for air travel at the most convenient times during the day, some level of congestion during these peak periods should be tolerated. In other words, a conservative allocation of slots that totally eliminates airport congestion throughout the day is not in society’s interest. It is difficult, however, for authorities

23

Rassenti et al. (1982) developed a numerical model for airport slot auctions in a network setting.

EUROPEAN AIR TRANSPORT

21

running a traditional slot allocation system, or managing a slot auction, to tell exactly how much peak-hour congestion should be tolerated. In other words, it is hard to know how many peak-hour slots to allocate relative to the airport’s design capacity, or how many peak slots to sell under an auction system. This indeterminacy could be solved by the alternate system of airport congestion pricing. Under such a system, the first step is to calculate the external congestion costs that are generated when an airline operates another flight at the airport. These external costs equal the increased operating cost for other airlines plus the value of the extra time lost by their passengers when the given airline schedules another flight, adding to congestion at the airport. Since each airline fails to take these external costs into account, it over-schedules peak-hour flights. A congestion-pricing system corrects this problem by charging the airline a fee per flight equal to the external congestion costs it generates. Faced with this fee, the airline reduces peak flights, partly alleviating airport congestion. The congestion fee could also include other external costs beyond those directly related to congestion, such as the costs of environmental damage from airline flights (noise and pollution).24 Under a congestion-pricing system, slots are no longer used. As long as a carrier can pay the appropriate congestion fee at a given time of day, it gains airport access at that time. It is important to recognize that, because the congestion fee captures all the external costs generated by a flight, the number of peak flights, and the corresponding level of congestion, end up being the correct ones from society’s point of view.25 Even though slots are absent, there is an important equivalence between the congestion-pricing and slot allocation systems. In particular, a slot allocation system replicates the outcome under congestion pricing if the total slots allocated over the day match the flight totals chosen by the airlines when faced with congestion fees. The problem, however, is that there is no guarantee that this correspondence will actually be realized, given that choosing the number of slots to allocate is mostly a matter of guess­ work. For example, a well-meaning slot allocation manager may mistakenly allocate or sell too few peak-hour slots on the belief that peak congestion needs to be dramatically restricted. By contrast, the congestion-pricing system automatically generates the correct flight totals over the course of the day. It does so by basing congestion fees on hard evidence regarding congestion costs, which is derived from engineering data on the air­ port’s congestion properties along with data on airline operating costs and information on the value of passenger time. Note that the potential for misallocation inherent in a slot allocation system exists even when the manager relies on the price system, running a slot auction, to distribute slots among the airlines. While an auction guarantees that the slots the manager chooses to sell are allocated efficiently, going to the carriers who value them most, the problem

24

See Daniel (1995) and Brueckner (2002) for analyses of airport congestion pricing. Congestion fees need not impose a larger financial burden on the airlines than existing landing fees. The latter fees, which are constant over the day at a typical airport, can be reduced at off-peak hours while being increased during peak periods. Despite this fact, airlines usually oppose any kind of new fee system. In addition, the general aviation lobby in the United States strongly opposes congestion pricing, which would effectively exclude many small aircraft from busy airports at peak hours by imposing prohibitive costs. 25

JAN K. BRUECKNER AND ERIC PELS

22

of selecting the number of slots to sell still involves guesswork. Use of the auction mechanism provides no guidance in making this quantity choice. It has been argued that appropriate congestion fees cannot be computed reliably, creating an equally serious drawback for a system of congestion pricing. Some experts would dispute this point, however, arguing that reliable operating-cost and value-of-time information can be gathered to compute appropriate fees. Such concerns, along with a common preference for quantity restrictions over the price mechanism on the part government regulators, mean that use of a slot allocation system is likely to continue. However, reliance on such a system should include a recognition of its potential pitfalls.

7.3 Airport Prices While slot auctions or congestion fees could provide substantial new revenue sources for airports, the large institutional changes needed to implement such systems may not occur soon. Therefore, it is useful to consider the current system of airport pricing, recognizing that piecemeal, temporary changes may be beneficial on the path to broader reform. Given that the provision of airport capacity exhibits constant returns to scale for large airports and increasing returns for smaller facilities, economic theory says that airport charges should roughly cover the cost of operations for major airports.26 However, existing charges, which include landing fees based on aircraft weight, occasional noise surcharges, and facility rents paid by airlines and airport retailers, often bear little relation to airport costs. As a result, airports in some cases incur losses that must be subsidized by general tax revenues, while profits are earned in other cases, indicating an excessive level of charges. In a deregulated environment, airport charges that are too high put inappropriate upward pressure on the fares charged by the carriers, leading to an unwarranted economic transfer from passengers to the airport authorities. Charges that are too low, on the other hand, force the general public to subsidize users of the air transport system, while also prolonging the lifespan of inefficient carriers, whose operations may be fostered by cheap airport fees. Both problems are exacerbated when airports are operated inefficiently, with labor and capital costs higher than the levels that could be incurred under best-practice methods. A potential solution to the joint problems of inappropriate airport charges and oper­ ational inefficiencies is airport privatization. Private airports have an incentive to keep operational costs as low as possible and to set their prices to at least recover costs. But while airport privatization eliminates inefficiency and the need for taxpayer subsidies, it may confer market power on the airport operator, raising concerns about excessive airport charges. These concerns may be especially strong for airports that enjoy high passenger demand, because they are important destinations (or origins) for business and leisure traffic. Abuse of the resulting market power will be reflected in the level of airport charges, which the airport authority may set too high, or in airport capacity, which the authority may set too low (by limiting expansion, for example).

26

See Doganis (1992) for evidence.

EUROPEAN AIR TRANSPORT

23

A natural remedy for potential airport market power is government regulation of airport charges. Various characteristics of the aviation sector, however, make the result of such regulation uncertain and its use potentially counterproductive. First, price regulation may lead to under-investment in airport capacity, potentially exacerbating the problem of airport congestion. Second, it is not completely clear that airports will actually abuse their market power, in which case regulation of charges would be inappropriate. Airports may restrain their charges because the profitability of complementary activities (shopping, catering etc.) is negatively affected when they are set too high, a consequence of the resulting loss in passenger volumes. More generally, an airport may recognize that if charges are set too high, it may lose the totality of an airline’s operations, either because the carrier relocates to another more affordable facility or because it is forced into bankruptcy. This threat of a dramatic revenue loss may help to restrain the level of charges levied by the airport.27 The potential exercise of airport market power remains a problem under both an auction-based slot allocation system and a congestion-pricing system. If a private airport authority controls the slot auction, it has an incentive to limit the number of slots sold in an attempt to extract more auction revenue. Similarly, the authority could charge congestion-sensitive landing fees but set these fees at an excessive level in an attempt to extract additional revenue. These problems could be overcome if the government ran the slot auction or the congestion-pricing system, with the privatized airport operator reaping the resulting revenue. In pursuit of profit, the operator would then minimize airport operating costs as well as making appropriate capacity investments.

8 CONCLUSION This paper has provided an overview of the institutional and regulatory developments underlying European airline deregulation. It is hoped that by clarifying the nature of the air transport system as it existed at the outset of deregulation, particularly the structure of airline networks, the paper allows a better understanding of the evolutionary process initiated by this important policy action. The paper has argued that the old flag-carrier regime led to a proliferation of airlines and airlines routes, with one effect being ineffi­ ciently low traffic densities in European networks. By raising cost per passenger, these low densities amplified the problem of high labor expenses, contributing to the high operating costs of European carriers. While international alliances and open skies agree­ ments helped to boost traffic densities, the low-cost carriers that have been unleashed by deregulation, though generating substantial passenger benefits through lower fares, threaten to drain traffic out of the major carriers’ networks. A defensive response is needed, and part of this response must involve concentration of the major carriers’ traffic on fewer routes through network reorganization and cross-border mergers. While the paper has also highlighted the need for additional policy steps, especially formation of a Common Atlantic Aviation Area and new rules for airport operations, one further

27

See also Starkie (2001) for a discussion of the consequences of airport price regulation.

24

JAN K. BRUECKNER AND ERIC PELS

recommendation is in order. This recommendation relates to the task of measuring the effects of deregulation. The problem is that, currently, the EU lacks a systematic means for tracking changes in airfares paid by European passengers. Since the ultimate goal of deregulation is to reduce the cost of air travel for passengers by generating a more efficient transportation sector, this measurement deficiency is a critical problem. To better grasp this point, consider the case of the United States, where the Department of Transportation collects extensive data on airfares that allow researchers to investigate a host of questions regarding the performance of the air transport sector. This data source, known as the Passenger Origin and Destination Survey, is generated from a 10 per cent quarterly sample of all airline tickets. The survey indicates the origin and destination cities for a passenger, the route traveled and the carriers used, and the overall fare paid for the trip. Given the nature of the data, average fares in individual city-pair markets can be measured and tracked over time, and the effect of competition in the market and other factors can be evaluated. An alternative to using such data is to rely on the published airfares available in various sources. These data, however, do not reflect the actual fares paid by traveling passengers. For example, some published fares may hardly ever be used, making them irrelevant in any attempt to measure the performance of air transport sector. Alternatively, some researchers have collected private survey data on fares, but the volume of such data is necessarily limited. With air transport deregulation now achieved in the Europe, a high priority is for the EU to institute a system that allows its effects on fares to be measured. The relevant EU authorities should create a data collection system like the ticket-sampling system used in the United States. Such a system imposes a slight cost on the airlines, who must carry out the actual ticket sampling and report the detailed results, while also generating some cost for the government authority. However, without the resulting ability to track fares, the EU can never fully evaluate the success of its historic deregulation effort.28

REFERENCES Brattle Group, 2002. The Economic Impact of an EU-US Open Aviation Area, Brattle Group, Washington, D.C. Boguslaski, C., Ito, H. and Lee, D., 2004. Entry patterns in the Southwest Airlines route system, Review of Industrial Organization, 25(3), 317–350. Brueckner, J.K. and Spiller, P.T., 1994. Economies of traffic density in the deregulated airline industry, Journal of Law and Economics 37, 379–415. Brueckner, J.K. and Whalen, W.T., 2000. The price effects of international airline alliances, Journal of Law and Economics 43, 503–545. Brueckner, J.K., 2002. Airport congestion when carriers have market power, American Economic Review 92, 1357–1375.

28 The EU should also collect data analogous to the “service segment” data compiled by the US DOT (known as database T100). These data provide detailed information about airline operations on individual non-stop route segments (flight frequency, total seat capacity, etc.). Such data are useful in tracking route entry and exit by the airlines.

EUROPEAN AIR TRANSPORT

25

Brueckner, J.K. and Pels, E., 2005. European airline mergers, alliance consolidation, and consumer welfare, Journal of Air Transport Management, 11, 27–41. Caves, D.W., Christensen, L.R., and Tretheway, M.W., 1984. Economies of density versus economies of scale: Why trunk and local service costs differ, Rand Journal of Economics 15, 471–489. Commission of the European Communities, 1996. Impact of the Third Package of Air Transport Liberalization Measures, Communication of the Third Package of Air Transport Liberalization Measures, 22.10.1996 COM (96) 514, final, Brussels. Daniel, J.L., 1995. Congestion pricing and capacity at large airports: A bottleneck model with stochastic queues, Econometrica 63, 327–370. Doganis, R., 1985. Flying Off Course: The Economics of International Airlines, George Allen & Unwin, London. Doganis, R., 1992. The Airport Business, Routledge, London. Doganis, R., 2001. The Airline Business in the 21st Century, Routledge, London. Good, D.H., Roller, L.-H., and Sickles, R., 1993. U.S. airline deregulation: Implications for European transport, Economic Journal 103, 1028–1041. Marin, P.L., 1998. Productivity differences in the airline industry: Partial deregulation versus short run protection, International Journal of Industrial Organization 16, 395–414. McGowan, F. and Seabright, P., 1989. Deregulating European airlines, Economic Policy 9, 282–344. Morrison, S.A. and Winston, C., 1985. The Economic Effects of Airline Deregulation, Brookings Institution, Washington, D.C. Morrison, S.A. and Winston, C., 1995. The Evolution of the Airline Industry, Brookings Institution, Washington, D.C. Morrison, S.A., 2001. Actual, adjacent, and potential competition: Estimating the full effect of Southwest Airlines, Journal of Transport Economics and Policy 35, 239–256. Neven, D.J. and Roller, L.-H., 1996. Rent sharing in the European airline industry, European Economic Review 40, 933–940. O’Connor, W.E., 1989. Introduction to Airline Economics, 4th edition, Praeger, New York. Oum, T.H. and Yu, C., 1998. Winning Airlines: Productivity and Cost Competitiveness of the World’s Major Airlines, New York, Kluwer Academic Press. Rassenti, S.J., Smith, V.L, and Buffin, R.L., 1982. A combinatorial auction mechanism for airport time slot allocation, Bell Journal of Economics 13, 402–417. Starkie, D., 2001. Reforming U.K. airport regulation, Journal of Transport Economics and Policy 35, 119–135. U.S. Department of Transportation, 1999. International Aviation Developments (First Report): Global Deregulation Takes Off, Office of the Secretary, U.S. Department of Transportation, Washington, D.C. U.S. Department of Transportation, 2000. International Aviation Developments (Second Report): Transatlantic Deregulation—The Alliance Network Effect, Office of the Secretary, U.S. Depart­ ment of Transportation, Washington, D.C.

This page intentionally left blank

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

2 Wage Determination in the US Airline Industry: Union Power Under Product Market Constraints∗ Barry T. Hirsch†

ABSTRACT The chapter analyzes wages in the US airline industry, focusing on the role of collective bargaining in a changing product market environment. Airline unions have considerable strike threat power, but the exercise of bargaining power is constrained by the financial health of carriers. Since airline deregulation, compensation has waxed and waned in response to the industry’s economic environment. Airline workers capture sizable rents following good times and provide concessions following lean times. Compensation at legacy carriers has been restructured, some from within and some from outside of bankruptcy, but it remains to be seen whether compensation will continue its long-run movement toward opportunity costs. Evidence from the CPS for 1995–2006 shows that wage premiums for airline industry workers, particularly for pilots, remain with existing premiums almost entirely a union phenomenon. Much of the gap in wage scales between major and mid-size carriers was erased in the mid-2000s concessionary cycle, but these rates remain much higher than rates

∗ Helpful comments were received from Gary Fournier, Jim Gillula, Darin Lee, Nick Rupp, and seminar participants at the University of Kentucky and the Southern Economic Association meetings. I thank Dan Kasper and Darin Lee for providing Form 41 labor cost data. Numbers reported in Table 3 reflect work done with Michael Wachter and James Gillula, presented in United Airlines’ 1113(c) bankruptcy hearing. The views expressed in this chapter are mine and need not reflect the opinion of others. Work for this chapter was conducted while the author was at Trinity University, San Antonio, Texas. † W.J. Usery Chair of the American Workplace, Department of Economics, Georgia State University, Atlanta, GA 30302; e-mail [email protected].

BARRY T. HIRSCH

28

at regional carriers. Compensation levels at regional carriers may approximate opportunity cost – the compensation that would be necessary to attract and retain qualified employees throughout much of the industry. Because unions retain bargaining power at the major carriers, wages are likely to head upward as carriers’ financial health returns. Such wage levels may or may not be sustainable in the inevitable next downturn.

1 INTRODUCTION The air transportation industry has realized rapid growth throughout its history. Despite this growth, carrier profitability since deregulation has proven volatile and corporate viability far from certain. With one exception, every major carrier at the time of deregu­ lation in 1978 has either failed, had its operations merged into another airline, or been in bankruptcy protection. The exception, American Airlines, narrowly avoided bankruptcy in 2003 following wage concessions from its unions. As is the case for most companies, labor compensation among airlines accounts for a substantial share of total costs.1 In much of the industry, compensation is determined through collective bargaining; thus, workers’ pay may deviate substantially from oppor­ tunity costs. While union density economy-wide has sharply declined, the airline industry has remained highly unionized. The percentage of workers who are union members in the air transportation industry was 49.2% over the 1973–1978 regulatory period and 49.4% in 2005. Union coverage rates for flight personnel and ground workers are higher. In contrast, private sector union density economy-wide fell from 24.2% in 1973 to 7.8% in 2005.2 No private sector industry has union density as high as does air transportation.3 More than any other private industry, airlines face unions who possess substantial bargaining power, that power emanating from the ability of a strike to shut down and bankrupt a carrier. Of course, it is not in the interests of workers and their unions to destroy their employers, so union demands are constrained by the financial health of carriers. Hence, the airline industry has developed a compensation pattern in which its union workers “tax” potential profits following the onset of good times, but agree to moderate contractual pay increases or provide wage and benefit concessions following the onset of bad times. For many if not most airlines and their unions, this product market

1 During much of the 1980s and 1990s, labor costs accounted for about a third of total expenses. This share peaked at 38% in 2002, a level not seen since 1979 and in the earlier regulatory period. The labor cost percentage fell substantially after 2002, to 36% in 2003, 30% in 2004, and 26% in 2005 (24% in 2005:4). Fuel costs, which accounted for 12% of total expenses in 2002, rose to 23% of expenses in 2005 (Air Transport Association, 2006, Labor and Fuel tabs). 2 Union density figures are compiled from the Current Population Survey (CPS). The 1973–1978 air trans­ portation figure is in Hirsch and Macpherson (2000, p. 136), while the 2005 figures for air transportation and the private sector are from Hirsch and Macpherson (2003, updated annually at www.unionstats.com). Prior to 2003, the air transportation industry included air courier services (e.g., largely nonunion FedEx), which were small during 1973–1978. Were these included in the 2005 figure, union density would be about 10 percentage points lower. 3 Two industries have higher union density, the predominantly public railroad transportation industry and the entirely public US Postal Service (www.unionstats.com).

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

29

union wage cycle has been accompanied by a contentious labor relations environment with no small amount of distrust on all sides.45 Following the strong financial health of the industry in the late 1990s, generous labor contracts and high labor costs took force in the early 2000s. The increased compensa­ tion was accompanied by a “perfect storm” of negative events – a recession in 2001, sharp declines in traffic following the 11 September, 2001 attacks, Internet pricing, increasing market shares among “low-cost carriers” and concomitant declines among hub-based legacy carriers, and, more recently, high fuel costs over a sustained period. The convergence of high operating costs and intense price competition resulted in bankruptcies among four legacy carriers (US Airways, United, Delta, and Northwest) and several mid-size and regional carriers. During 2004–2006, wages and benefits among the legacy carriers have been falling, either under the threat of or following bankruptcy.6 The recent restructuring of labor costs in an increasingly competitive airline industry has been substantial. Lower labor costs, a decrease in debt burden among carriers emerging from bankruptcy, relatively strong demand, and reduced capacity among the legacy carriers have improved major carriers’ financial prospects. At the same time, high fuel costs and, more fundamentally, the emergence of more competitive product markets and a high level of price competition, have served to keep profits at low levels, at least through mid-year 2006. But the future will not be an extension of the present. During the two decades following airline deregulation, periods of union wage concessions have been followed by rebounding wages as airlines’ profits recovered. Continuation of this pattern requires not only the presence of strong unions that can appropriate company profits, but also product market innovations and a degree of pricing power that generate profits to be taxed.7 The purpose of this chapter is to examine the role of unions and describe recent wage determination in the airline industry. A key question posed in this study is whether or 4 There is a large industrial organization literature on the airline industry, but little on the airline labor market. References to past studies are provided in Hirsch and Macpherson (2000); see, for example, Card (1998), Crémieux (1996), and Johnson (1995). Nay (1991) provides an early statement on union wage cycles in the airline industry. 5 Airlines differ from other US private sector industries in that collective bargaining is governed not by the National Labor Relations Act (NLRA), but by the Railway Labor Act (RLA) of 1926, amended in 1936 to apply to the airline as well as railroad industry. As compared to the NLRA, the RLA provides more specificity as to the negotiation and mediation procedures that parties must adopt in a labor dispute prior to a strike. The bargaining structure that evolved under the RLA was decentralized, with separate unions by craft and carrier-specific contracts. 6 Companies in bankruptcy cannot unilaterally void their union contracts and implement lower pay but, under U.S.C. § 1113(c), can request that a bankruptcy judge do so. The company must show that wage and benefit cuts are necessary for the company to successfully emerge from bankruptcy and that the cuts are equitable. The equity provision can be examined through a comparison of contract rates with estimates of market compensation and by showing how pay cuts are distributed across employee groups. In most cases, the employer and union agree on new wage and benefit terms, often with prodding from the judge, prior to a decision being made on voiding a contract. 7 Although not addressed in this chapter, union wage demands may be constrained by a company’s level of debt, given that increasing leverage reduces liquidity. Knowing this, union companies’ optimal debt levels will be higher than for nonunion companies. For theory, supporting evidence, and references to prior literature, see Matsa (2006).

30

BARRY T. HIRSCH

not the current decrease in wages and benefits in the industry represents a permanent shift in the level of compensation, or whether there will be a resumption of the historical cycle of rising union wage premiums following the onset of good times and subsequent wage concessions following lean times. In order to address this question, it is critical to estimate the level of opportunity cost wages in the airline industry. Subject to a number of caveats, well know in the labor economics literature, compensation in competitive markets will tend toward opportunity wages – what similar workers in similar jobs might have obtained in alternative employment.8 Compensation cannot be expected to fall below a competitive level, at least not for any sustained period. Thus, a reasoned judgment as to the pattern of future wages in the airline industry requires that we know how airline wages diverge from opportunity wages. To estimate this divergence, it is necessary not only to compare the wages of unionized airline employees at legacy carriers to wages elsewhere in the airline industry, but also to the opportunity wages outside the industry. The plan of the chapter is as follows. Section 2 provides a brief discussion of how the level and dispersion in airline labor costs have changed over time. Section 3 provides analysis on overall airline industry wage differentials using the Current Population Survey (CPS) for 1995 through 2006, focusing on the unionized sector of the air transportation industry. Section 4 follows with a more detailed focus on union and nonunion CPS wage differentials by airline industry “craft” (pilots, flight attendants, mechanics, fleet service, agents, and other). In Section 5, carrier contract data are presented that permit a comparison of pay by craft at major carriers with pay for those same occupations at regional airlines. A final section of the chapter addresses, but does not fully answer, the two questions stated above. First, given the evidence, what is the level of opportunity cost wages? Second, will the future be one in which earnings move toward opportunity costs or will we continue to observe cycles in which union wages rise well above and subsequently fall toward opportunity costs, depending largely on airlines’ ability to pay?

2 AIRLINE LABOR COSTS OVER TIME Labor cost is only one of many determinants of an airline’s financial health, but it is an important one. In each of the three years from 2001–2003, the four airlines with the highest compensation per employee (Form 41 salaries and benefits, as presented

8 A wage premium is defined here as payments to labor beyond long-run opportunity costs; that is, what workers could have earned in an alternative job path entailing similar investments in training and similar working conditions. Employees’ current pay is often greater than the pay they could get at an alternative job. These short-run premiums (quasi-rents) derive from costs associated with job mobility, firm- and industryspecific skills, and implicit contracts in which earnings deviate from spot marginal products. Efficiency wage theory proposes that in some settings, wages in excess of opportunity cost may lower per unit costs and are thus consistent with profit maximization. The reasoning is that in workplaces with high monitoring costs, voluntary effort may increase in response to high wages, either to reduce the risk of firing or as a result of positive reciprocity (these explanations seem to apply less readily to union than nonunion workplaces). For discussion, see Cahuc and Zylberberg (2004, pp. 353–360).

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

31

below) were US Airways, United, Northwest, and Delta. Dispersion in compensation across airlines was relatively high. Not coincidentally, these four airlines ended up in bankruptcy protection.9 American, a close fifth in labor cost per employee during 2001–2003, went to the brink of a bankruptcy filing in 2003, backing off from filing following concessions from its unions. Wage and benefit concessions at high-cost carriers have led to declines since 2003 in average industry compensation per employee and in pay dispersion across airlines. Much of the analysis in the chapter utilizes the CPS, the monthly household survey of individuals conducted jointly by the US Bureau of the Census and Bureau of Labor Statistics. The CPS, however, is not ideally suited to track year-to-year changes in airline labor costs among large national carriers. First, sample sizes of air transport workers each year are not large. Second, a portion of air transport workers do not work for passenger airlines and, among those working for airlines, one cannot differentiate employees of major versus regional or other airline services. And third, the CPS allows one to measure earnings, but not the dollar cost of benefits. Form 41 data reported by certificated carriers to the Department of Transportation (DOT) is better suited than CPS data to track airline labor costs over time. Figure 1 presents average total compensation (real salaries and benefits, in 2005 dollars) per

$95,000

18.0

Compensation (2005$)

16.0 $85,000 15.0 $80,000 14.0 $75,000 13.0 $70,000 12.0 $65,000

Coefficient of Variation (CV)

17.0

$90,000

11.0

$60,000

10.0 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

Year Compensation

Coefficient of Variation (CV)

Compiled from figures reported by US Department of Transportation, Bureau of Transportation Statistics, Form 41. See Appendix Table A1 and text for a listing of the numbers and further details.

Figure 1 Airline Wages and Benefits, Level and Dispersion, 1990–2005.

9 US Airways entered bankruptcy protection in 2002 and again in 2004, United in late 2002, and Delta and Northwest on the same day in September 2005.

32

BARRY T. HIRSCH

airline employee during 1990–2005, and the dispersion across airlines in average com­ pensation (Appendix Table A1).10 The data include all major and most national carriers.11 The number of carriers differs by year, as smaller airlines move in and out of the industry or mergers occur; there was a maximum of 19 in 2000 (with 12 in 1990 and 13 in 2005).12 Calculations are based on weighted averages, with airline employment by year as weights. Compensation and, subsequently, costs per available seat mile (ASM) are expressed in 2005 dollars using the CPI-U (current series). As is evident in Figure 1, average real compensation among carriers (shown by the “diamonds”) increased briskly through 1994, stayed relatively flat (or fell slightly) throughout the rest of the 1990s, and then increased after 1999. One sees large increases in average real compensation in 2000, 2001, and 2002, followed by small, moderate, and large decreases in 2003, 2004, and 2005, respectively. The real level of compensation in 2005 is similar to that seen in the mid-1990s. Also shown in Figure 1 (see the “squares”) is the dispersion in compensation across carriers, measured by the employment weighted coefficient of variation. High pay disper­ sion generally produces cost differences that cannot be sustained. Dispersion decreases when low-pay airlines play catch-up and/or when high-pay airlines fall back toward the pack. Pay dispersion had declined during the late 1990s, but increased following pay hikes taking effect in 2001–2003. Wage and benefit concessions since 2004 among the high-cost carriers have reduced pay dispersion. Figure 2 provides an alternative measure of labor costs constructed from Form 41 data, measuring cents per ASM, in 2005 dollars, for 1990–2005. Obviously, labor costs per seat mile are affected not only by costs per worker, but also by employment, productivity, airline capacity, and the like. The pattern evident in Figure 2 (see the “diamonds”) is one of a gradual but steady decline in real labor costs per ASM from 1990 through 1997, modest increases in costs until 2000, an upward break with sharply increased costs in 2001 and 2002, followed by substantial decreases after 2002, from 4.7 cents per ASM in 2002 to 3.3 cents in 2005. The recent decline in costs per seat mile came about not only through decreases in compensation per worker, but by steep declines in employment at the legacy carriers. Figure 2 also shows the dispersion across carriers in labor costs per ASM. The coefficient of variation (the “squares”) stayed constant at about 20 through the 1990s, but declined sharply in 2003–2005 to about 15. Undue weight should not be placed on this single statistic, but by this measure the cost structure across airlines was more similar

10

The values shown in Figures 1 and 2 are provided in Appendix, Table 1. Daniel Kasper and Darin Lee of LECG kindly made available Form 41 information on compensation and labor costs per ASM by airline. Employment data for certificated carriers, used to construct weights, were obtained from the Bureau of Transportation Statistics at http://www.bts.gov/programs/airline_information/number_of_employ­ ees/certificated_carriers/index.html. 11 The DOT defines a national airline as having at least $100 million in annual revenue and a major airline $1 billion. 12 In 2000, included airlines accounted for 91% of total employment among all DOT 41 “major” and “national” certificated carriers (excluding Airborne Express, FedEx, and UPS). In 1990 and 2005, the corresponding numbers were 87% and 81%, respectively. The “low” figure in 2005 reflects sharp declines in employment at the included large legacy carriers included in Figure 1, coupled with growth in employment among small airlines not included.

33

5.00

26

4.75

24

4.50

22

4.25

20

4.00

18

3.75

16

3.50

14

3.25

12

3.00

Coefficient of Variation (CV)

Labor Cost per ASM (2005$)

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

10 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005

Year Labor Costs per ASM

Coefficient of Variation (CV)

Compiled from figures reported by US Department of Transportation, Bureau of Transportation Statistics, Form 41. See Appendix Table A1 and text for a listing of the numbers and further details.

Figure 2 Labor Costs and CV Per Available Seat Mile, 1990–2005.

in 2005 than at any time since at least 1990. All else the same, similar cost structures across airlines should be associated with more stable prices and financial outcomes.

3 WAGE DIFFERENTIALS IN THE AIR TRANSPORTATION INDUSTRY: MEASUREMENT AND DATA Are unionized airline workers paid wages above long-run opportunity cost? If so, how large are these premiums? Do wage premiums vary across airline crafts? And do nonunion as well as union airline workers receive premiums? These seemingly straight­ forward questions are not easy to answer, at least not in a precise manner. The difficulty arises from a combination of methodological issues (e.g., what are the appropriate com­ parison groups for airline workers) and data limitations (e.g., company level data do not provide measures of worker attributes, while public data on individuals and their attributes do not permit one to easily examine differentials within the industry – say across major versus regional carriers or among those at passenger airlines versus air freight companies). Of course, limitations arise to some greater or lesser degree in all research endeavors. For the research questions posed in this chapter, a variety of evi­ dence allows one to paint a reasonably clear picture of wage determination in the airline industry. This section follows and extends the approach utilized by Hirsch and Macpherson (2000) in their study of wage determination in the airline industry from 1973 through

34

BARRY T. HIRSCH

1997. The CPS analysis uses individual wage and salary worker data for September 1995 through May 2006.13 Following Hirsch and Macpherson, six air transportation groups of workers are identified using the CPS – five airline crafts plus a residual category. The five craft groups are pilots, flight attendants, mechanics, fleet service (ramp and utility) workers, reservation agents and clerks, and an “other” category. For the five airline craft groups, comparison groups of individuals comprising nonairline workers employed in specific sets of occupations are identified. Workers in those occupations serve as comparison (control) groups in order to measure relative wages. For the “Other” category of airline workers, a comparison group of nonairline workers across the economy is used. Relative wage differentials between the air transport and comparison group workers are estimated within a regression framework, controlling for measurable worker, location, and job characteristics. Section 4 and Appendix Table A2 describe the construction of the CPS comparison groups for each of the six air transport groups. The empirical approach is as follows. Separate wage equations by craft are estimated, with each regression sample from the 1995–2006 CPS including both an airline “treat­ ment” group (pilots, etc.) and a large comparison group of workers. From each wage equation, whose coefficients are determined largely by the nonairline comparison group, I calculate log wage differentials for union and nonunion air industry workers, rela­ tive to measurably similar comparison group workers outside the air transport industry. Industry-wide wage differentials based on the full sample of air transport workers are calculated based on the weighted average across the six groups, using fixed air transport employment weights over the time period. In addition to controlling for a typical set of worker human capital, demographic, and location characteristics reported in the CPS (e.g., schooling, age, region), an occupational skill level and working condition variable, constructed by the BLS, is matched to the CPS. The principal purpose of this additional control variable is to account for occupational skill differences not captured by worker schooling and experience measures. Specifically, let ln Yigt = kgt Xikgt + g U Air igt + g N Air igt + c Occcit +  ln Skilli + it 

(1)

where ln Y is the natural logarithm of the wage, i designates individual, g indexes six airline craft groups and their corresponding control groups, and t is year. Included in X are k worker and labor market control variables (listed below) with k the corresponding coefficients. A “level of work” variable, Skill, defined at the detailed occupation level is shown separately since results are presented with and without its inclusion. Air is a dummy variable set to 1 for each of the g air industry craft groups. Air is interacted with index variables designating whether a worker is covered (U ) or not covered (N ) by a collective bargaining agreement. Occ is a set of c broad occupation dummies used only in the regression for “Other” air industry workers – the nonspecified craft group

13 September 1995 is the first month following 1994 CPS revisions in which imputed earners, excluded from the analysis (see below) can be identified. May 2006 was the latest CPS file released when this paper’s empirical analysis was executed. This 10-plus year period includes both low and high points in airline wage-profit cycles.

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

35

and an economy-wide control group. Year dummies for 1997 through 2006 are included in X, with 1995–96 the omitted base period. Coefficients g and g provide estimates of log wage differentials by airline craft group g for union and nonunion workers, respectively, in both cases as compared to the appropriate comparison group made up as a mix of union and nonunion workers.14 Weighted averages of these coefficients thus provide estimates of the airline wage differentials d of interest. That is du = wug g

(2)

dn = wng g

(3)

Estimates of the air transport log wage differentials are shown for union air transport workers (du ), nonunion air transport workers (dn ), and union and nonunion combined. The differentials are estimated with and without control for ln Skill. The weights wg for union and nonunion workers represent the CPS employment shares of the six air industry worker groups fixed over the 1995–2006 period.15 CPS wage differentials were also estimated by year, but are not presented. Little systematic pattern is found, presumably due to large year-to-year variation in the air transport industry samples reporting earnings. The air transport and comparison group samples include full-time nonstudent wage and salary workers ages 18 and over. To enhance the relevance of the comparison group, excluded are workers with less than a high school degree (with the GED categorized as high school) and education beyond a masters degree. No education restrictions are placed on the air transport sample. In the regression analysis, the relatively few airline workers with less than a high school degree are assigned to the high school category and those few with a degree beyond the masters level are assigned to the masters category. With the exception of flight personnel (pilots and flight attendants), full-time status is defined as reporting 35 or more usual hours per week on the principal job. The reporting of hours worked by flight personnel presents a problem, however, since some report only paid flight hours, while others report all hours away from home. Flight personnel who report 15 or more hours worked per week are retained as full-time workers. Usual weekly earnings reported by pilots and flight attendants is only weakly related to their reported weekly hours worked, ruling out the calculation of an hourly wage based on weekly earnings divided by weekly hours (construction of the wage is described below). Approximately 25–30% of workers in the CPS are either unwilling or unable to report their earnings. These individuals have weekly earnings “allocated” by the Census based on an imputation procedure in which nonrespondents are assigned the earnings of a “donor” with an identical set of match characteristics (Hirsch and Schumacher, 2004).

14 That is, union status is not included as a control in X. The assumption here is that the opportunity cost wage for each group is best approximated by an implicitly weighted average of union and nonunion wages. 15 The weights are calculated prior to omission of imputed earners and using CPS employment weights. Use of fixed weights over time means that changes in the wage gap estimates result from wage changes and not from worker mix changes. Since imputation rates can differ across airline craft groups, weights are determined prior to the exclusion of earnings nonrespondents.

36

BARRY T. HIRSCH

All those with imputed earnings are excluded from the analysis. It is important that they be excluded in order to avoid severe attenuation toward zero in wage gap estimates with respect to the airline industry and union status. Neither industry nor union status is a match criterion used to assign a donor’s earnings to a nonrespondent. Hence, air transport industry nonresponents will typically be assigned the earnings of nonairline donors. Union nonrespondents will typically be assigned the earnings of nonunion donors. Broad rather than detailed occupation is an imputation match criterion, thus nonresponding pilots (aircraft mechanics, etc.) will typically not be assigned the donor’s earnings of other pilots (aircraft mechanics, etc.). Unless imputed earners are excluded (or an explicit bias correction method used), wage differentials with respect to industry, union status, and other nonmatch criteria will be seriously attenuated. Hirsch and Schumacher (2004) show that “match bias” (i.e., the attenuation in coefficient estimates) roughly equals the proportion of nonrespondents.16 Imputed earners cannot be identified in the CPS between January 1994 and August 1995. Hence the analysis in this chapter begins with September 1995. Public use files of the CPS include an edited usual weekly earnings measure that is topcoded (i.e., capped). For years prior to 1998 (and after 1989), weekly earnings are capped at $1,923 ($100,000 annually) and for years since 1998 at $2,885 ($150,000 annually). Apart from pilots and a few managerial or professional workers, few air transport industry workers have top-coded earnings. For nonpilot air transportation workers and all comparison group workers with weekly earnings at the cap, they are assigned the estimated mean earnings above the cap based on year and gender-specific estimates that assume a Pareto distribution for earnings beyond the median (see Hirsch and Macpherson, 2006, p. 6; posted at www.unionstats.com). Values are moderately higher than 1.5 times the cap, with somewhat smaller female than male means and growth over time. Top-coded earnings among pilots is widespread, more so than for any other occupation in the CPS, but the right tail of their earnings distribution is probably less skewed than implied by the Pareto distribution (i.e., fewer extremely high earnings). During September 1995–August 1996, the first 12 months used in our sample, 20.0% of pilots who reported earnings were above the $1,923 weekly earnings cap, as compared to 9.6% in 2000 and 14.3% in 2005 with the higher $2,885 cap that began in 1998 (Hirsch and Macpherson, 2006).17 Because many senior pilot contracts are for amounts not far above the cap, I assign pilot means above the cap that are much lower than the Pareto means. For the years prior to 1998, pilots with weekly earnings greater than $1,923 have their earnings set at 1.25 times the cap, or $2,404. For years beginning in 1998, pilots with weekly earnings greater than $2,885 have their earnings set at 1.25 times the cap, or $3,606. These estimates seem likely to be conservative (i.e., produce too low a pilot/nonpilot wage differential). Moreover, use of the same multiple for all years

16

Match categories include education, age, gender, race, hours worked, broad occupation, and receipt of tips,

commission, or overtime. Bias due to imperfect matching (e.g., a PhD matched to an earnings donor from the

BA or above category) is analyzed in Bollinger and Hirsch (2006).

17 The pilot sample in 2005 has an unrepresentative number of high earners as compared to earlier years and

2006.

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

37

fails to capture some of the highest pilot contract increases and subsequent concessions realized over this period.18 For air transport workers apart from pilots and flight attendants, plus all comparison group workers, the wage is defined as follows. Hourly earnings are calculated as equal to usual weekly earnings (which includes typical overtime, tips, and commissions) divided by usual hours worked per week.19 For pilots and flight attendants, reported hours worked per week are ignored, since variability across workers contains little information.20 For pilots, the wage is calculated as weekly earnings divided by 40, while for flight attendants, weekly earnings are divided by 36. This approach is explained below. Finally, the earnings measure for all workers is converted to 2005 dollars using the CPI-U (Current Series). One cannot avoid making some rather arbitrary assumption as to how job-related hours among flight personnel compare to work hours among other workers. The earnings measure that is used implicitly assumes that job-related time spent by pilots (flight hours, wait time, and travel time) entails similar disutility on average as does 40 hours of paid work plus nonpaid travel time for nonflight air transport and comparison group workers. For flight attendants, the assumption is that a typical week is equivalent to 36 hours of work in comparison group jobs. Pilots’ mean reported hours worked in the CPS is 40.9 overall (and 39.7 for union pilots). Flight attendants’ mean reported hours worked in the CPS is 32.5 overall (and 32.0 for those unionized). If the hours assumptions of 40 and 36 for pilots and flight attendants, respectively, overstate (understate) the disutility associated with hours worked by flight personnel, then the wage differential estimates for these groups are too low (high). Included in X – the control variables – are education dummies (5) reflecting levels from a minimum of a high school degree (including a GED) through a masters degree, potential experience in quartic form separately and interacted with gender (experience being proxied by the minimum of age minus years schooling minus 6 or years since age 16), gender, race/ethnicity (4), foreign born, region (8), metropolitan area size (6, with nonmetro the base), year dummies (10), and broad occupation dummies (11, included only for the “Other” group regression). In addition to the control variables in X, earnings differentials are estimated with an included occupation or job duties variable, ln Skill, compiled by the BLS for the approx­ imately 500 Census occupation groups. As described in Pierce (1999), the unit of anal­ ysis for the National Compensation Survey is the detailed occupation cross-classified by work level. The NCS uses the Census occupation codes (COC) included in the CPS. For each Census occupation, 10 job attribute factors are defined, each with various levels.

18 Regression estimates of pilot earnings premiums are about .10 log points higher when Pareto means rather than the more conservative 1.25 multiples are used. 19 A small number of individuals do not report usual hours worked per week and instead have their hours worked value imputed (i.e., assigned) by the Census. These individuals are excluded from the estimation sample. For nonflight personnel and all comparison group workers, individuals who report “variable” weekly hours have the wage determined by usual weekly earnings divided by hours worked last week, which adds noise to the independent variable but is unlikely to bias coefficients. 20 Commercial airlines hire few part-time pilots or flight attendants. Recall that those reporting less than 15 hours worked per week are excluded from the sample.

38

BARRY T. HIRSCH

These are as follows: knowledge (9 levels), supervisory controls (5), guidelines (5), com­ plexity (6), scope and effect (6), personal contacts (4), purpose of contacts (4), physical demands (3), work environment (3), and supervisory duties (5). Each of these job attribute factors and levels were awarded “quality points” by BLS analysts in order to develop a single occupational job attribute index. This occupational job factor index is highly correlated with earnings (Pierce, 1999). Allegretto et al. (2004) have previously merged this BLS index with the CPS and used it in their study of public school teacher salaries. The job factor index, referred to here as “Skill” (but which measures a broad range of attributes, as indicated above), was obtained from the BLS for 1990 COC, used in the CPS through 2002. Beginning in 2003, the CPS adopted 2000 COC, many of which cannot be mapped one-to-one with the 1990 COC. Codes used for the five airline crafts could be mapped cleanly. For the remaining air transport industry workers and all comparison group workers beginning in 2003, each worker was assigned a 1990 COC based either on a direct match to their current COC or from a probabilistic mapping between 1990 and 2000 COC provided by the Census. “Skill” was then matched to each worker’s assigned 1990 COC. Included in the earnings equations is ln(Skill). Its coefficient  represents an elasticity; e.g.,  = 025 implies that earnings increase 2.5% for each 10% increase in the skill index. The BLS skill index does not cover Census occupations that are exclusively federal (e.g., Postal Service workers), private household, and agriculture, forestry, and fishing occupations. The exclusion of these occupations from the analysis reduced sample sizes very little. Earnings equation results (not including “Skill”) with and without these occupations are nearly identical.

4 EARNINGS IN AIR TRANSPORTATION AND AMONG COMPARISON GROUPS: DESCRIPTIVE EVIDENCE This brief section provides descriptive evidence from the CPS on earnings differentials between union and nonunion air transportation workers and “comparable” workers out­ side the airline industry. A subsequent section examines the earnings premium estimates derived from the regression analysis. Table 1 provides the CPS sample sizes, employment weights, mean earnings (in 2005$), and “BLS Skill values” for the air transportation industry, for each airline craft group (separately by union status), and for the corresponding comparison groups. As noted previously, the earnings sample for September 1995 through May 2006 includes only those who respond to the earnings question and not those whose earnings have been imputed (assigned) by the Census. The CPS sample size of air transportation industry workers over the period 1995–2006 is 6,835, with roughly equal numbers of pilots, flight attendants, and mechanics (about 900 each). The sample size of agents (reservation agents, gate agents, and stores employees) is moderately larger (about 1,200), while the sample size of the fleet service workers (i.e., baggage handlers, cleaners, and other ground workers) is substantially smaller. The residual group of “Other” air transport workers is quite large, about 2,600. Estimates of industry wage differentials are based on the weighted average of estimates across the six employee groups (jointly and separately by union status). The group weights (shown in rows labeled “Weights”) are calculated

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

39

Table 1 CPS Mean Wages (2005$) and Skill Index, by Airline Craft and Comparison Groups Union and Nonunion

Union

Nonunion

Comparison Groups

All air transport Wage Skill index N

$22.88 1,274 6,835

$27.03 1,241 2,971

$19.68 1,299 3,864

$19.29 1,282 877,302

Pilots Wage Skill index N Weight

$43.09 2,225 879 0.125

$49.38 2,225 580 0.197

$30.86 2,225 299 0.072

$27.30 1,864 76,702

Flight attendants Wage Skill index N Weight

$21.24 933 893 0.134

$22.09 933 648 0.219

$18.99 933 245 0.070

$15.56 896 278,593

Mechanics Wage Skill index N Weight

$23.73 1,577 924 0.131

$26.58 1,578 465 0.156

$20.84 1,577 459 0.112

$18.75 1,204 36,419

Fleet service Wage Skill index N Weight

$15.38 508 343 0.042

$17.67 446 155 0.046

$13.49 560 188 0.038

$12.85 437 16,050

Agents Wage Skill index N Weight

$16.41 733 1,203 0.175

$17.84 727 523 0.157

$15.32 737 680 0.189

$13.69 661 14,682

Other Wage Skill index N Weight

$20.23 1,312 2,643 0.393

$21.55 1,019 609 0.225

$19.83 1,400 2,034 0.520

$19.29 1,282 877,302

Means are compiled from the CPS monthly earnings files, September 1995–May 2006. Wages, shown in 2005$, measure the hourly earnings for non-flight personnel, calculated over the sample (of size N) excluding imputed earners. For flight personnel, wages are calculated based on weekly earnings and an assumed 40 hours week for pilots and 36 hours week for flight attendants. The airline and comparison worker groups are described in the text. The BLS Skill index points are described in text and in Pierce (1999). Group weights, calculated from CPS employment weights for the CPS sample including imputed earners, are used to compile the overall industry and combined craft wage differentials shown in Table 2.

40

BARRY T. HIRSCH

from the CPS sample, including those who do not report earnings, using the employment weights that the Census attaches to each surveyed worker. For most craft groups, the raw mean union wages are considerably higher than for the comparison group, while the nonunion air transport means are modestly higher. Interesting are means of the BLS occupational skill index. In some cases, including the overall industry comparison to the economy-wide comparison group, the skill index mean for the air transport and comparison groups are highly similar. Where the skill values are similar, the suggestion is that the comparison group is closely matched to the air transport treatment group. Where there is a difference, it illustrates the potential importance of the skill index control for more precise estimates of wage differentials. For example, aircraft mechanics (engine and non-engine) have higher occupational skill ratings than do the other mechanics with whom they are compared. Apart from pilots and flight attendants, who have unique CPS occupation codes, mean values of Skill are not identical for union and nonunion workers within a craft, since craft groups can include more than one CPS occupation (e.g., mechanics include aircraft engine mechanics, aircraft mechanics excluding engine, and mechanic supervisors) and union and nonunion workers need not be equally distributed across these detailed occupations.

5 EARNINGS DIFFERENTIAL ESTIMATES IN THE AIR TRANSPORT INDUSTRY AND BY AIRLINE CRAFT In this section, earnings differential estimates between air transport workers and “compa­ rable” workers and levels of work economy-wide are examined. The results are presented in Table 2, first the differentials for the entire air transportation industry and then for each airline worker group. Estimates of d are presented separately for union and nonunion workers and from earnings equations with and without inclusion of the BLS skill index. Estimates are for the entire 1995–2006 period, since sample sizes by year and by craft are too small to reliably identify year-to-year movements. Appendix Table A3 provides information identical to that shown in Table 2, except that results are estimated for the years 2003–2006 rather than 1995–2006. The reason for showing estimates beginning in 2003 is a change that year in CPS occupation and industry definitions (the switch from 1990 to 2000 Census codes). Because of similarity in results, discussion is restricted to Table 2, apart from noting that small sample sizes for the 2003–2006 estimates reduce their reliability, particularly for the individual crafts. It is worth emphasizing that our CPS analysis includes only wages and salaries and not benefits. Economy-wide, unionized workers realize a “benefits premium” that is larger than the wage premium (Freeman, 1981). In the airline industry, union contracts among the major airlines provide benefit levels well beyond those seen economy-wide for full-time workers in the private sector (Wachter, 2004). As evident in Table 1, the overall skill index rating for workers within the airline industry is nearly identical to that seen for the economy-wide comparison group (1,274 versus 1,282). This makes the comparison of benefits among major carriers with the average economy-wide particularly relevant, even though such analysis lacks worker and job controls. Were it possible

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

41

Table 2 CPS Log Wage Differentials by Airline Worker Group and Union Status, 1995–2006 Group

Skill Index Included

All

Union

Nonunion

No Yes

0.130 0.108

0.249 0.226

0.040 0.019

No Yes

0.189 0.154

0.283 0.249

0.072 0.037

No Yes

0.290 0.245

0.407 0.365

0.066 0.013

No Yes

0.209 0.182

0.246 0.224

0.110 0.070

No Yes

0.189 0.115

0.297 0.223

0.080 0.005

No Yes

0.114 0.112

0.220 0.219

0.008 0.006

No Yes

0.118 0.107

0.182 0.171

0.069 0.058

No Yes

0.038 0.036

0.132 0.147

0.010 0.003

All air transport industry

Industry crafts

Pilots

Flight attendants

Mechanics

Fleet service

Agents

Other

See note to Table 1. Estimates are based on the CPS monthly earnings files, September 1995–May 2006. The “Industry Crafts” group includes the five airline crafts but excludes “Other”, while “All Industry” includes “Other” as well. The “All Industry” and “Industry Craft” differentials are compiled based on the weighted average of their component parts, using the employment weights shown in Table 1 (with separate weights by union status). Estimation of differentials explained in text. Differentials are shown with and without control for BLS occupational skill index, ln Skill. Other control variables are education dummies (5) reflecting levels from a minimum of a high school degree (including a GED) through a masters degree, potential experience in quartic form separately and interacted with gender, gender, race/ethnicity (4), foreign born, region (8), metropolitan area size (6, with non-metro the base), year dummies (10), and broad occupation dummies (11, included only for the “Other” group regression).

to estimate a total compensation differential within a regression framework (i.e., with controls), the strong suggestion is that compensation premiums would exceed the wage premiums presented in Table 2.21 21

A minor caveat is that our CPS economy-wide sample of full-time workers excludes those with schooling less than a high school and greater than a masters degree, while the BLS benefits sample makes no such restriction.

42

BARRY T. HIRSCH

5.1 Industry Differentials The top line of Table 2 contains the earnings differential results for the air transport industry over the combined 1995–2006 period, separately by union status and both with and without inclusion of ln Skill. The industry differential d is the weighted average across wage differentials estimated for the five airline craft groups and a remaining “Other” air transport workers group. The “standard” log wage differential d for 1995–2006, compiled from estimated wage gaps absent control for Skill, is 0.13 log points.22 The “expanded” earnings differentials, compiled from regressions that control for Skill, reduce d by 0.02 log points, from 0.13 to 0.11. The smaller expanded gap estimate reflects not only that airline occupations tend to have somewhat higher skill (and other job attribute) ratings than do comparison group workers, but also that these higher skills are not fully accounted for by CPS measures such as schooling and potential experience. The average differential across all air transport workers masks what are large earn­ ings premiums for union workers and little apparent earnings advantage for nonunion workers. Our preferred measure of d is the expanded measure, which controls for Skill. Nonunion air transport workers have an estimated d of only 0.02, indicating that nonunion earnings in the industry are roughly comparable to earnings realized by similar workers (union and nonunion) performing comparable levels of work outside the indus­ try. By contrast, union air transport workers realize a substantial earnings premium of 0.23 log points, well above the level dictated by comparability and a competitive labor market.23 A sizable share of the air transport sample (in particular the nonunion sample) is in the “Other” category, which includes workers in a wide range of occupations and some working for air transport companies other than airlines. In general, these workers tend to have smaller wage advantages than do the traditional airline crafts. If the weighted average is constructed from just the five “craft” groups (the row labeled “Industry Crafts” in Table 2), higher estimates of d are obtained – a combined union and nonunion earnings advantage (controlling for Skill) of 0.15, a union premium of 0.25, and a nonunion differential of 0.04. In short, the air transportation industry is a high wage industry, with earnings premiums concentrated among union workers, particularly workers in the standard airline craft groups.

5.2 Pilots Earnings differential estimates by airline worker group are included in Table 2. Pilots are first examined. The CPS pilot category includes “aircraft pilots and flight engineers”

22

All differentials are presented as log point wage gaps. Percentage gap estimates can be obtained by

[ed – 1]100, where d is the log point gap. For reasons of space and because sample sizes are large, standard

errors are not presented. Standard errors for the industry gaps are approximately 0.007. Standard errors vary

across craft group, but are approximately 0.015 (but somewhat larger for fleet service and smaller for “other”

workers).

23 The airline union wage advantage is higher, but the same order of magnitude, than are economy-wide union–

nonunion wage gaps during this period (Hirsch and Macpherson, 2006, Table 2a). Note that the union airline

differential compares unionized air transport workers to a mix of union and nonunion nonairline workers.

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

43

(pre-2003 the occupation is labeled “airplane pilots and navigators”). The comparison group for pilots includes full-time workers outside the air transportation industry in occupations within the following broad categories beginning in 2003: business and financial operations, computer and mathematical, architecture and engineering, and life, physical science, and social science occupations. Pre-2003, the categories are labeled professional specialty occupations (except health) and technologists and technicians (except health). As for all the comparison group samples, workers with less than a high school education or a degree beyond a masters are excluded. The largest earnings premiums for any airline craft group are found for union pilots. Absent control for the BLS skill index, the earnings differential for all pilots is 0.290; with ln Skill included as a control the differential is 0.245. The earnings premium (with Skill included) is driven principally by unionized pilots, estimates of d for union pilots being a sizable 0.365 and for nonunion pilots being 0.013, effectively zero (corresponding estimates without the Skill control are 0.407 and 0.066). There are no doubt some unmeasured differences in skill and experience between union and nonunion pilots, the latter more likely to have less flying experience and to pilot smaller planes. Having said that plane size differs, however, it is not clear how large rate differentials with respect to aircraft size would be in a competitive labor market. Whatever those differences, they could not account for such large earnings premiums.24 Finally, it is important to note that estimates of pilot earnings differentials are sensitive to the assumed level of mean earnings above top-coded weekly earnings. As stated previously, pilots’ mean earnings above the cap have been “conservatively” assigned as being equal to 1.25 times the top-code amount. This compares to the approximate 1.7 times the cap for men (and 1.6 for women) based on the Pareto distribution, the multiples used for the comparison group sample (shown at www.unionstats.com). Had the Pareto distribution estimates been used for pilots, estimated earnings premiums for union pilots would be about 0.10 log points higher than those shown. Because annual pilot sample sizes in the CPS are not large and the number of pilots at the top-code varies quite a bit from year to year, it is difficult to reliably estimate the time pattern of changes in pilot earnings using the CPS.

5.3 Flight Attendants Flight attendants earnings are compared to those of a comparison group of workers scat­ tered across occupations within the broad categories of sales, service, and administrative support. I focus on the earnings equation results that include the BLS occupational

24

Reinforcing the finding of a large earnings premium for unionized pilots is the evidence that union pilots at the major carriers have quit rates that are close to zero (Wachter, 2004). No doubt a part of this low quit rate reflects the fact that wage scales for pilots display substantial growth with respect to seniority, but seniority cannot be transferred across airlines with union contracts. As pilot layoffs have become common at the legacy carriers, some have been willing to “start over” at FedEx or at other carriers where greater job security is expected (Dade, 2006). Although the wage is not the only determinant of the quit rate, a quit rate close to zero is hard to imagine absent a sizable premium. Economy-wide quit rates in the private sector (including part-time workers) are about 25% annually, while for private transportation and public utilities the rate is over 15% (US BLS, 2006).

44

BARRY T. HIRSCH

skill index, whose inclusion lowers estimates of flight attendant earnings differen­ tials by about 0.02–0.03 log points. The estimate for combined union and nonunion flight attendants is an earnings premium of 0.18 log points. As is the case for pilots, the premium varies by union contract coverage. Unionized flight attendants realize a 0.22 premium, whereas nonunion flight attendants have a small earnings advantage of 0.07. Several of the caveats that arise with respect to pilot earnings premium estimates do not arise for flight attendants. First, all but a few transportation attendants employed in the air transport industry are likely to work for passenger airlines, whereas pilots and some other air transport crafts are employed in air freight or some other air transport support industries. Second, neither skill requirements nor adverse working conditions systematically increase with plane size (this issue is discussed in Section 5). And third, the entire 0.22 log point difference between union and nonunion flight attendants is likely to represent a premium. Although unionized flight attendants are concentrated at major carriers and nonunion flight attendants at nonunion midsize and regional carriers (Delta’s nonunion flight attendants are an exception), large airlines should be able to attract and retain productive flight attendants at wages similar to those received by nonunion workers at small airlines.

5.4 Mechanics Aircraft mechanics include workers in the air transportation industry whose detailed occupations are aircraft engine mechanics, aircraft mechanics (except engine), and mechanic supervisors. The comparison group includes workers in all mechanic occu­ pations (including supervisors) employed outside the air transportation industry. The aircraft mechanic occupations are awarded higher skill index ratings than are all other mechanic occupations; hence wage differential estimates with a control for ln Skill are substantially lower, by about 0.07 log points, than those excluding ln Skill. Over the 1995–2006 period, the estimate of d for mechanics with the skill index (union and nonunion combined) is 0.11 log points, compared to a 0.19 estimate without the skill index. As with pilots and flight attendants, the wage premium story is really a union story. Over the entire period, the log wage premium for union aircraft mechanics (with skill included) is 0.22, as compared to effectively zero (0.01) for nonunion aircraft mechanics. There exist skill differences (e.g., licenses to work on different planes) among aircraft mechanics within the air transport industry that are not observed. If unmeasured skills are positively correlated with union status, estimates of within-industry union–nonunion wage differences are overstated, although wage gap estimates for all mechanics need not be biased. As with the other employee groups, a sizable wage premium for union, but not nonunion, aircraft mechanics is observed. Airlines have limited opportunity to substitute nonunion for union pilots or flight attendants when facing large within-industry wage differences. In contrast, airlines have some ability to substitute away from their union­ ized mechanics by outsourcing scheduled maintenance and other work to specialized companies employing licensed aircraft mechanics (Goodwyn [2006] reports a recent example of union “in-sourcing”). Although an airline’s mechanics union will bargain to

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

45

limit such substitution, the possibility of substitution should constrain the magnitude of union premiums among mechanics.25

5.5 Fleet Service (Ramp) Workers There are a variety of ground workers, apart from mechanics, who service airplanes. These include baggage handlers, airplane cleaners (utility workers), workers who guide planes into and out of their gates, and workers who refuel airplanes. Fleet service or ramp workers category includes those who work in the air transportation industry and whose occupations are freight, stock, and material handlers; and vehicle washer and equipment cleaners. The comparison group of workers includes those employed outside the air transportation industry in the following occupational categories: nonconstruction laborers and freight, stock, and material handlers; garage workers; washer-cleaners; and packers. Inclusion of the skill index has little effect on estimates of d for ramp workers, the log wage gap being 0.11 with or without control for ln Skill.26 As with other crafts, nonunion fleet service workers appear to be paid roughly their opportunity costs, with a d estimate of 0.01. Unionized fleet service workers realize an estimated wage premium of 0.22 log points. As with mechanics, carriers facing high contract rates among fleet service workers have incentive to outsource some of this work, substituting lower-cost contract workers for their own union employees.

5.6 Reservation Agents and Stores Employees Airlines have a large number of customer service employees – ticket reservation agents outside of airports, ticket and gate agents within airports, and “stores” employees who oversee the recording and distribution of supplies and parts. “Agents” are defined as those employed in the air transportation industry within the following occupations: reservation and transportation ticket agent; shipping, receiving, traffic clerks; stock clerk and order filler; and customer service representative. The comparison group includes workers outside the air transportation industry employed in the same occupations as above, plus those listed as other information and record clerks and as order clerks.

25

When faced with a strike by mechanics in August 2005, Northwest eliminated a large share of their mechanics jobs through outsourcing and hired (in advance of the strike) replacement workers for the remaining mechanics jobs. One should be reluctant to generalize from the Northwest example. Northwest mechanics were represented by a “rebel” union (AMFA) which had unseated the IAM, in the process alienating IAM’s remaining Northwest workers and Northwest’s other unions. Northwest’s unionized workers crossed the AMFA picket lines and allowed Northwest to continue operations despite the strike (Carey, 2005). An agreement, ratified in November 2006, allow striking AMFA workers to receive limited amounts of layoff or separation pay. Those accepting layoff status can bid on open technician positions. 26 The small effect of the Skill index is not surprising, since those working in and outside the air transportation industry (i.e., the treatment and comparison groups) are drawn largely from the same Census occupations and hence are assigned the same Skill values. It seems unlikely to me that the combination of required skills and adverse working conditions in these occupations is so different that it should lead to highly disparate wages in and outside the air transportation industry.

46

BARRY T. HIRSCH

The combined union–nonunion wage differential estimate of d is 0.11 with control for ln Skill. Union–nonunion differences here are less than seen with other crafts, with a union wage advantage relative to the nonairline comparison group of 0.17 and a nonunion wage advantage of 0.06.

5.7 Other Air Transport Industry Workers The “other” or miscellaneous category of air transport workers, all those not included in the previously discussed five crafts, are distributed over a broad range of occupations. The comparison group includes the entire sample of full-time workers outside of the air transportation industry (recall that the sample excludes those with very low and high education levels and for whom the BLS skill index is not defined). Most of these air transport workers are nonunion. Inclusion of the occupational skill index has little effect on estimates of the differential (broad occupation dummies are included in the “other” earnings equation). The overall earnings differential for the “other” group of workers is 0.04. This reflects an earning premium of 0.15 for the small union portion of the group and a zero estimate for the large nonunion group. More so than for the five airline craft groups, a nontrivial number of the air transportation industry workers in the “other” occupation category are likely to work at companies other than a passenger airline. Taking the occupational groups as a whole, a clear pattern emerges from the earnings analysis in this section. First, there exists a sizable earnings premium among union workers in air transportation, relative to a mix of union and nonunion comparison group workers. Second, nonunion air transport workers appear to realize little premium com­ pared to similar workers doing similar levels of work outside the industry. Although there was evidence of rent sharing among nonunion as well as union airline workers during the pre-1978 airline regulation period and in the immediate years after deregu­ lation (Hirsch and Macpherson, 2000), any remaining rents are now small. What might be labeled an airline earnings premium is for the most part a union premium.

6 WAGE DIFFERENCES ACROSS AIRLINES: DO REGIONAL AIRLINE WAGES APPROXIMATE OPPORTUNITY COSTS? In prior sections, average compensation among national carriers reported in DOT Form 41 data have been used to measure changes over time in industry labor costs, while worker data from the CPS have been examined to estimate wage differentials for union and nonunion workers throughout the air transport industry, relative to similar workers and jobs outside the industry. In this section, evidence on contractual “top rates” for airline workers across national carriers and regional airlines is reviewed. On the basis of this and prior evidence, a question that is explored is whether opportunity cost wage rates for major carriers might be approximated by the rates currently seen at regional airlines. Airline contract data are not publicly available (i.e., not reported to the government), but are assembled by trade groups. The Airline Industrial Relations Conference (Air Con­ ference) is made up of a consortium of scheduled national airlines. Each provides their

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

47

labor contracts by craft to the Air Conference, which assembles contract information and terms of employment in their database. Member airlines are provided access to contract information from all participating airlines. J. Glass & Associates (a consultancy division overseen by Ford & Harrison, LLP) manages a Regional Airline Association database with union and nonunion contract rates and terms of employment for regional airlines. Wage schedules for national and regional carriers, drawn from the Air Conference and Glass & Associates databases, are provided in a December 2004 analysis by Michael Wachter (2004), who provided expert testimony for United Airlines in their 1113(c) bankruptcy hearing.27 It’s useful to recall the economic setting. In late 2004, highcost airlines were in a serious financial situation following September 11, a recession, Internet pricing, increased price competition and rising market shares of low-cost carriers, and rising fuel prices. United had received wage concessions in 2003 following their 2002 bankruptcy filing, but were asking for further concessions to help achieve a viable business plan that would allow them to emerge from bankruptcy protection. US Airways, which had preceded United into bankruptcy, had recently obtained reduced wage scales for their employees in their bid to emerge from bankruptcy. American had earlier received concessions from their workers that allowed them to avoid bankruptcy. Delta and Northwest had relatively high contract rates at this time despite prior wage concessions; these subsequently would be reduced further leading up to and following their 2005 entry into bankruptcy. A relatively “healthy” Continental was less threatened by bankruptcy, but did receive salary relief from its unions. Compared with the legacy airlines, Southwest was the anomaly in 2004 and remains so today. Although having high contractual top rates, Southwest has a smaller share of employees at its top rates, lower benefit costs, higher labor productivity, lower operating costs, lower debt, and defined contribution rather than under-funded defined benefit pension plans. Evidence presented in Wachter (2004) allows one to compare wage rates at the major carriers to those among similar crafts at regional airlines. The analysis included two sets of United Airlines contract rates, those in force in December 2004 and those then proposed by the airline (or, in the case of pilots, rates from a tentative agreement). The existing December 2004 rates reflected pay concessions United’s unions had agreed to previously in 2003. The proposed United rates were nearly identical to those implemented previously at US Airways. In Table 3, United’s post-2004 rates are compared to rates at other major airlines, at mid-size national airlines, and those at regional airlines. The significance of United’s post-2004 rates is that they, along with those at US Airways, can be thought of as setting, at least for a couple of years, a new industry standard.28 It is this “standard” that is compared to wage schedules at mid-size and regional carriers. Salary schedules are provided for eight separate worker crafts, summarized in Table 3. These are pilots, flight attendants, mechanics, utility workers, ramp workers, stores

27 The Wachter analysis was conducted with assistance from James Gillula of Global Insight and from me.

Views expressed in this chapter are my own and need not reflect those of United, Michael Wachter, or Global

Insight.

28 Delta and Northwest, which entered bankruptcy in 2005 with pay rates well above United, eventually

implemented wage schedules similar to United’s post-2004 rates. American and Continental, which received

concessions from their unions outside of bankruptcy, have rates somewhat above those of United.

Table 3 Contractual Top-Rates and Log Wage Differentials between United Airlines and Major, Mid-Size, and Regional Carriers, December 2004 Employee Group

United Airlines Dec 2004

United Proposal Dec 2004

Average at Majors

Average at Mid-size

Average at Regionals

United Post-2004 vs. Majors

United Post-2004 vs. Mid-size

United Post-2004 vs. Regionals

Pilots (monthly) FAs (monthly) Mechanics Utility Ramp Stores CSR RSR

$12,374 $3,073 $31.09 $17.50 $21.06 $21.06 $21.75 $21.01

$10,554 $2,895 $29.82 $16.65 $19.61 $19.61 $20.25 $19.56

$12,627 $3,447 $31.99 $16.83 $20.09 $19.86 $20.31 $19.58

$11,311 $2,863 $28.39 $13.49 $17.05 $16.82 $16.86 $15.95

$5,861 $2,198 $22.73 $12.39 $13.69 $13.93 $13.82 N/A

−0.179 −0.175 −0.070 −0.011 −0.024 −0.013 −0.003 −0.001

−0.069 0.011 0.049 0.210 0.140 0.153 0.183 0.204

0.588 0.275 0.271 0.296 0.359 0.342 0.382 N/A

The dollar rates are reported in Wachter (2004). The United proposed wages are treated as a rough approximation of prevailing wages at the major carriers following wage concessions at major airlines during 2005 and 2006. The last three columns show the difference in log of the average wages between the United proposal and the respective averages at the majors, mid-size, and regional airlines. All wages shown reflect top-rates (maximum seniority). For pilots and flight attendants, monthly rates are based on a 75-hours yield. Other rates shown are hourly rates. RSR (reservation service representatives) are not employed by most regional carriers. Averages at six major carriers (excluding United), mid-size, and regional airlines are unweighted. A list of carriers included in each category is provided in the text and in footnote 29.

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

49

employees, customer service representatives (CSR), and reservation service representa­ tives (RSR). Recall that the CPS analysis had five airline craft groups. The first three groups in Table 3 – pilots, flight attendants, and mechanics – align exactly with CPS categories. The next two – utility (cleaners) and ramp workers – were included in the CPS fleet service worker category. The next three categories – stores, CSR, and RSR employees – were included in the CPS agent category. The first two columns of Table 3 provide the top-rate salary rates by craft for United before and after December 2004. It is the post-2004 rate that will be treated as an approximation of the industry standard for major and national airlines. The next three columns present the unweighted average of rates at the six major airlines other than United (American, Continental, Delta, Northwest, Southwest, US Airways), midsize national airlines, and at regional airlines.29 The use of top rates (maximum seniority) means that what is being compared is a wage contract structure across airlines rather than payroll cost differences. At the legacy airlines, a large proportion of workers are at the top rates. Generally, the average wage within an airline will be below the top rate, although this need not be true since average wages include overtime pay and possible pay supplements (e.g., international rates for flight attendants) not included in the top rate. New or expanding airlines will have fewer workers with high seniority, hence differences in top rates across carriers may not fully reflect current payroll differences between legacy and low cost carriers. Over time, there should be a narrowing payroll cost gap as the age structure of the legacy and low cost carriers becomes more similar. For all crafts other than pilots and flight attendants, hourly wage rates are provided. The rate for mechanics is an “all-in top rate” that includes the top-step base pay and maximum license, skill, line, and longevity pay. For pilots at the major and mid-size national carriers, the pay shown is a monthly pay based on a 75-hour yield for a 12th year captain flying a weighted average of planes within each airline’s fleet. Pilot rates at regional carriers are for a 12th year captain flying a weighted average of 50-seat and larger jets. Pay shown for flight attendants is likewise based on a 75-hour yield, with top-step base and incentive pay (but excluding international pay). The last three columns of Table 3 provide the log wage differential between United’s post-2004 rates of pay and the end-of-year 2004 pay in the three sectors of the airline industry – other major airlines, mid-size national airlines, and regional carriers. As evident from these columns, United’s proposed (and eventual) rates for pilots, flight attendants, and, to a lesser extent, mechanics, would be below the average of other major airlines. In 2005 and 2006, rates at Delta and Northwest would decrease to a level similar to those at United. The United top rates for pilots, flight attendants, and mechanics were similar to those seen at mid-size national airlines. Rates for the remaining crafts would be similar to December 2004 rates at other major airlines and above rates at mid-size carriers. The importance of Table 3 stems from the results reported in the final column, the difference in log wages between the United proposed wages, a stand-in for what would emerge as the national carrier “industry standard,” and the unweighted averages of up

29

The mid-size national airlines are AirTran, Alaska, America West, ATA, Frontier, JetBlue, and Midwest. The regionals are Air Wisconsin, Allegheny, American Eagle, Atlantic Southeast, Chautuaqua, Comair, ExpressJet, Horizon, Independence, Mesa, Mesaba, MidAtlantic, Midwest Connect, Piedmont, Pinnacle, PSA, SkyWest, and Trans States. Rates were not provided for all airlines in every craft.

50

BARRY T. HIRSCH

to 18 regional airlines (but fewer in some crafts). These craft-specific wage gaps are large – 0.59 log points (80%) for pilots, 0.28 (32%) for flight attendants, 0.27 (31%) for mechanics, 0.30 (34%) for utility, 0.36 (43%) for ramp workers, 0.34 (41%) for stores workers, and 0.38 (46%) for gate agents (there are few reservation agents among the regional airlines).30 Do wages at regional airlines approximate opportunity costs? By opportunity cost wages, what is meant is a compensation structure that in a competitive labor market could in the long run attract and retain a labor force with “appropriate” skills (appropriate mean­ ing profit-maximizing, with there being a trade-off between compensation and produc­ tivity). Wachter (2004) suggests that the wage scales seen at regional airlines provide an approximation of opportunity cost wages, based on the similarity of jobs at the major and regional carriers and small differences between regional wage scales and average published wage rates within the same broad occupation categories economy-wide. Although this claim is difficult to establish in a rigorous manner, it is plausible. The earlier CPS analysis implements comparability through the estimation of earnings models intended to compare air transport workers with measurably similar workers in broadly similar jobs outside the air transport industry. Jobs are made statistically equivalent, at least in principle, through measures of worker and job attributes. The analysis found rough equivalency between nonunion wages among craft workers in the air transportation industry and “comparable” workers outside the industry. Given high rates of union coverage among major carriers (Delta being the exception), then many of these nonunion workers are employed by mid-size and regional carriers, as well as air transportation companies other than certificated carriers. The CPS analysis cannot compare identical jobs, but it does approximate what airline workers might have earned in an alternative career path (i.e., long-run opportunity costs). The regional airline comparison seen in Table 3 provides an alternative way to assess comparability. The analysis compares workers in the same (i.e., comparable) occupation within the same industry. Such a wage comparison provides a control for skills, worker preferences, and working conditions. This is an important advantage, but has the disadvantage that wages for all jobs within an industry may be impacted by noncompetitive wage determination (e.g, union bargaining power), resulting in wage levels above opportunity cost.31 Although comparing wages within the industry provides a seemingly precise jobs match, it cannot be asserted that jobs and workers within the industry are literally identical. Some (unknown) portion of each of the gaps between wages at the major and at regional carriers reflects differences in worker skills or in the nature of the job. Even were there competitive wage determination throughout the airline industry, wage differences between the major and regional carriers would remain. Today’s wages at regional

30 The stated percentage differences are calculated using the “low” regional airline average wage as the base. The log gaps (times 100) provide a value in between percentages calculated using the low and high wages as base. The log of the average wage across airlines (used in the calculation) is a little larger than the average of log wages. 31 In particular, the concern is that union power in the industry (both union coverage and the threat of organizing uncovered workers) raises the level of nonunion wages and regional carrier wages to exceed opportunity costs.

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

51

carriers are presumably higher than they would be were airline labor markets perfectly competitive (e.g., absent union coverage or the threat of organizing). A competitive or opportunity cost wage structure throughout the industry might produce salaries at the major carriers that are similar to or slightly above what regional carriers pay today. How similar are workers and jobs at the major versus the regional carriers? The answer to this question will differ across craft. Clearly, pilots at major carriers are flying larger jets and typically have greater flying experience. A competitive wage structure would produce higher salaries for pilots flying larger planes and with more experience. But it does not follow that the wage gradient with respect to plane size or experience would be as large as is evident today, or that current salaries at regional carriers would not be sufficient to attract to major carriers a sufficient number of pilots skilled at flying large jets. Pilots are highly skilled and there is considerable trust placed on pilots not to make errors. But competitive salaries are determined at the margin and there are large numbers of individuals who want to be pilots – and who are able and willing to acquire the skills necessary to be a pilot.32 Regional airlines are able to attract qualified applicants, even where the possibility of moving to a major carrier at far higher pay is low. A high proportion of these pilots could and would acquire the licensing to fly larger planes were such jobs available. Nothing in the analysis permits us to say that a competitive salary for pilots of large jets would be precisely at the pay level seen today for regional jets. But we know that current levels of pay at major carriers (as seen in the 0.59 log wage premium for major carriers over regional carriers) far exceed the compensation necessary to attract and retain qualified pilots. And we know from the CPS analysis that unionized pilots (averaged across the entire air transport industry, including regionals and freight carriers) realize substantial earnings premiums, 0.37 log points, relative to what they might have earned in alternative career paths. Thus, salaries seen today for regional pilots may well be a reasonable approximation of what average salaries would be for airline pilots across a competitive airline industry labor market. It is worth noting that a fully competitive labor market would allow movement of pilots and other workers across carriers, without placing workers at a starting wage scale (zero seniority).33 If an industry were such that most worker skills were firm specific (i.e., not valued at other firms), then we would expect there to be promotion from within and little hiring at other than a junior level. Yet far more than in most industries, worker skills in the airline industry are highly transferable across firms. The skills of a pilot (or flight attendant or mechanic) that are valued at one airline are also valued at other airlines. What makes airline skills difficult to transfer and minimize cross-carrier mobility of airline workers are not differences in skill requirements, but provisions in union contracts that strictly tie pay to seniority with one’s current employer. Absent worker mobility across carriers, we cannot know what would be a competitive wage structure. Mobility would act to limit demands for “high” pay since carriers can hire

32 Ideally, we would like to measure applicant queues and turnover at regional carriers to help determine if

wages are at market clearing levels. Currently, there is an excess supply of pilots throughout the industry.

During 2005 FedEx had 14,000 applications for the 420 pilot jobs it filled. UPS hired 233 pilots from among

10,000 applicants, 8,000 of whom were passenger pilots (Dade, 2006).

33 Defined benefit pension plans also reduce mobility across firms, but these are becoming increasingly rare

for new private sector workers.

52

BARRY T. HIRSCH

experienced workers from other carriers. Likewise, mobility would limit the ability of financially strapped carriers to pay below opportunity cost wages, since workers could move to other carriers. Flight attendant pay is approximately 0.28 log points higher at major carriers than at the regionals. The principal difference between the two jobs is that flight attendants at the majors work on larger planes, work in teams rather than solo, and service longer but fewer flights. It is not at all clear that the differences in required skills and the desirability of working conditions between working on a large jet versus a small plane should lead to higher pay at the major carriers. Many (but not all) flight attendants will prefer to work as part of a team and to service longer but fewer flights. Some flight attendants will prefer the more limited range of travel typical of regional airlines; while others will prefer the wider range of travel to larger cities, both in the US and abroad. If regional airlines can readily attract and retain flight attendants at their current levels of pay, absent the expectation that they will move to higher paying job at a major carrier, then it is reasonable to argue that competitive pay for flight attendants would not be higher (or substantially higher) than the pay seen today at the regional carriers. The 0.22 log point CPS wage premium estimated for union flight attendants (and 0.18 for all flight attendants) reinforces the argument that wage scales at regional carriers roughly approximate opportunity costs. As with pilots, one would expect to see aircraft mechanics who service larger planes and work at major carriers to be paid more than mechanics at regional carriers. But it does not follow that competitive differentials between rates at major and regional carriers need be 31% for mechanics (as in Table 3). The estimated CPS premium for union aircraft mechanics relative to other types of mechanics economy-wide (conditional on schooling, age, and the occupational skill index, in which aircraft mechanics are rated highly) is 0.22 log points. If the CPS estimate is correct, then it suggests that competitive wage rates for aircraft mechanics would be lower throughout the industry, with a smaller wage gap between the regional and major carriers. The remaining airline crafts – ground workers (ramp and utility employees), agents, and stores employees – receive wage rates from 0.30 to 0.38 log points higher at the major airlines than at regional airlines. The CPS premium for union workers in these occupations was about 0.20 log points. Although there are some differences in the nature of these jobs between large and small carriers, the requisite skills and working conditions should not require such substantial wage differentials as seen in Table 3. Obviously, competitive wage levels across the entire industry would not precisely equal what is seen today at regional airlines, but these levels might not be too bad an approximation. In short, both the CPS analysis, comparing unionized workers throughout the air transport industry with similar workers and levels of job skill outside the airline industry, and the within industry comparison of contract scales at the majors and the regional carriers, indicate a wage structure in the airline industry that is well above opportunity cost. Were the labor market in the airline industry a competitive one, absent unions that possessed and exercised strong bargaining power, we would see a substantially lower level of compensation in the industry, coupled with easy worker mobility across carriers and few contractual restrictions on outsourcing. Whether airline compensation is headed toward opportunity costs is addressed in the final section of the chapter.

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

53

7 THE FUTURE – OPPORTUNITY COST WAGES OR UNION WAGE CYCLING? Workers in the air transportation industry are relatively highly paid. Some of this high pay reflects the training and skills required for jobs in the airline industry. Analysis in this chapter, however, provides evidence of substantial wage premiums in the air transportation industry. The premiums are realized primarily by union employees at major and mid-size airlines. Wages at regional carriers are substantially lower than at large national carriers and may roughly approximate what would be opportunity cost earnings for the rest of the airline industry. There is little evidence for earnings premiums among nonunion workers in the industry, many of whom work in occupations outside the traditional airline craft groups of pilots, flight attendants, mechanics, fleet service, and agents. Pay premiums are particularly large for unionized pilots, but are also substantial for other union workers. Compensation premiums are tied closely to union bargaining power (i.e., the ability to inflict costs through a strike), a strike threat power that appears unparalleled in private industry in the US. Union ability to acquire wage gains, however, depends crucially on the financial health of carriers. During good times, unions capture or “tax” some sizable share of potential profits. During bad times, unions moderate these demands or agree to wage concessions when firm survival and jobs are threatened. Because wages and benefits are negotiated and specified in long-term contracts, but the future financial fortunes of the firm are sometimes poorly predicted, changes in compensation (either through new contracts or renegotiation of contracts) lag changes in the product market. What might stop or substantially weaken this lagged wage-profit cycling; that is, a variable union tax on carrier profits? In most US industries, such cycles are not readily evident because union density is low and competition forces product prices to approximately reflect opportunity costs (including a normal return on capital). Were there a sufficient number of carriers that had competitive or opportunity cost compensation and were air fares not determined in partially segmented markets, pricing would consistently reflect the costs at these low-cost carriers. There would be little ability for unions to acquire and sustain noncompetitive wages and benefits. Certainly the airline industry is not immune to such forces. Increased penetration of low-cost carriers has limited the pricing power of major carriers, particularly on heavily traveled point-to-point routes.34 There also exists competition across cities (carriers) when customers prove willing to drive to alternative airports (Fournier et al., 2007). And during time periods and in markets with excess capacity, price competition can be keen.35 Such price competition clearly constrains union bargaining power, but does not eliminate it.

34 Low-cost carriers, however, have low penetration in low-density origin-to-destination routes, which continue to depend heavily on the hub-based legacy carriers. Analysis by Darin Lee shows that lower cost carriers’ share of domestic passengers increased from about 10% in 1990 to about 30% in 2004. He also estimates that by 2004, 75% of domestic passenger “trips” were exposed to LCC competition. See http://www.darinlee.net/stats.html. 35 Busse (2002) finds that carriers in the worse financial condition, particularly those highly leveraged, are most likely to start price wars. Carriers which most directly compete with price war leaders are more likely to join the price war.

54

BARRY T. HIRSCH

The bankruptcies of major carriers and concomitant downward adjustment in labor compensation throughout much of the industry can certainly be seen as a move toward a more competitive wage structure. The increased penetration of low cost carriers and the enhanced competition within major city hub airports suggest that price competition will constrain the future growth in labor costs. I do not predict, however, an imminent or smooth transition to an era of opportunity cost compensation. The airline industry remains highly unionized. As in the past, airlines are likely to reestablish market positions where they can be profitable, possibly through innovations or in other ways that cannot be anticipated. As this is written, airlines are close to the point of solvency due to healthy demand and reduced capacity. If fuel prices moderate and seating capacity remains tight, airlines will resume profitability. Then the test begins. Unions retain substantial bargaining power and will attempt to make up for wage and benefit concessions. How this bargaining will play out is hard to predict, but it seems unlikely that airlines can hold compensation down to the levels acquired in 2005–2006. Of course, the major carriers and their unions will not return quickly to the types of contracts seen in the early 2000s. It is clear to companies and their unions that such rates cannot be sustained. Union demands will be less affected by the opportunity cost of labor than by carriers’ ability to pay. It is unclear whether or not demands well in excess of opportunity costs will be or can be rejected by management.36 For the foreseeable future, airline unions will continue to tax carrier profits and we shall see a continuation of union wage cycling, hopefully at levels more sustainable than in the past. The labor relations environment in the airline industry has long been contentious and characterized by mutual distrust.37 The bankruptcies at US Airways, United, Delta, and Northwest, along with labor concessions at American, Continental, and several smaller airlines, have further strained management–labor relations.38 There are mutual benefits to a more cooperative labor relations environment, however, if it can pro­ duce a high level of productivity and sustainable labor costs. Although there is little reason to expect such an outcome industry-wide, there are forces that may well push some carriers in that direction. The financial troubles at major and mid-size carriers, coupled with increasing product market competition, make unions and their members more aware than ever that their long-run well-being requires a financially healthy

36 An interesting example of such a dynamic can be seen at largely nonunion FedEx, which is currently highly profitable. At the same time that pilot salaries have fallen in the passenger airline industry, union pilots at FedEx emphasized their company’s profitability and pushed for large pay increases. A lengthy impasse began in 2004, and remained unresolved until a four-year agreement with industry-leading pay was reached in August and approved in October 2006 (their 1999 contract remained in force during this period). Similarly, a new UPS pilots’ contract was ratified in August 2006 (the prior contract became amendable at the end of 2003). The new contract runs through 2011 and includes substantial pay increases. Pay scales for senior pilots at the cargo carriers now exceed those for the major carriers. Some employed passenger pilots, concerned about job security at their current employers, have tried to “start over” at UPS or FedEx, despite the loss of seniority and a large initial sacrifice in pay (Dade, 2006). 37 Exceptions include Southwest Airlines and recent labor relations at Continental. For a discussion, see Gittell et al. (2004). 38 Survey evidence in 2005 indicates high levels of anger and militancy among pilots and flight attendants (Comstock, 2006).

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

55

employer. At the same time, Southwest and, to a lesser extent, Continental, provide exam­ ples where relatively cooperative labor relations environments are possible and jointly beneficial.39 Although it is possible that labor relations at the legacy carriers will improve, this is just a possibility.40 Nothing guarantees that such environments will emerge, or even that cooperative labor relations can be maintained at Southwest or elsewhere in the industry. What can be predicted is that in absence of the emergence of more cooperative labor relations environments, we are likely to observe a return to wage and profit cycles, increased penetration of carriers that can achieve and maintain low costs (a function of productivity as well as labor compensation), and an uncertain future for the remaining legacy airlines. Whatever the labor relations path followed in the airline industry, all parties must continue to respond to the competitive forces that have been and will continue to ultimately shape the direction of the industry.

APPENDIX

Table A1 Average Compensation and Labor Costs per Available Seat Mile, Major and National Airlines, 1990–2005 Year

Salaries and Benefits (2005$)

CV

1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

$73,244 $74,164 $76,447 $77,586 $81,012 $81,147 $81,893 $80,764 $80,156 $79,318 $81,656 $86,339 $89,984

13.5 11.1 13.1 17.3 17.2 15.3 17.4 15.8 14.1 12.8 12.1 16.6 15.7

Labor Costs, cents per ASM (2005$) 4.70 4.66 4.54 4.43 4.41 4.33 4.24 4.23 4.29 4.30 4.36 4.60 4.66

CV 21.3 18.2 18.2 20.5 20.3 20.4 23.4 21.3 19.7 20.1 18.4 19.1 19.1 (Continued)

39 Little is known about the labor relations environments or human resource strategies employed by low-cost carriers, apart from Southwest. This topic is the subject of research by the Labor and Employment Relations Association (LERA) Airline Industry Council (Gittell and Kochan, 2006). The limited evidence to date suggests that there exists substantial heterogeneity across carriers. 40 An example of mutually beneficial labor-management cooperation can be seen at American Airlines, a legacy carrier with a rocky labor relations history. A redesign of maintenance operations, based in large part on worker input, has led to significant “in-sourcing” of maintenance from other airlines (Goodwyn, 2006).

BARRY T. HIRSCH

56

Table A1 Average Compensation and Labor Costs per Available Seat Mile, Major and National Airlines, 1990–2005—Cont’d Year

Salaries and Benefits (2005$)

CV

2003 2004 2005

$89,213 $87,279 $82,098

15.6 14.8 14.1

Labor Costs, cents per ASM (2005$) 4.28 3.80 3.27

CV 17.7 16.1 14.8

These numbers are shown in Figures 1 and 2. Based on compensation, employment, and ASM information reported by the US Department of Transportation, Bureau of Transportation Statistics, Form 41. Daniel Kasper and Darin Lee kindly provided current dollar figures by carrier and year on compensation and cost per ASM. Current dollar figures were converted to 2005$ and weighted averages were formed based on carrier employment counts provided in Form 41 data. CV measures the weighted coefficient of variation (standard deviation divided by mean). See text for further details.

Table A2 Construction of Airline Craft and Comparison Groups in the CPS Air transportation groups

Workers employed in air transport industry (code 421 for 1995–2002; code 6070 for 2003–2006)

CPS Occ code Occupation name Pilots 1995–2002 226 Airplane pilots and navigators 2003–2006 9030 Aircraft pilots and flight engineers Flight attendants 1995–2002 463 Public transportation attendants 2003–2006 4550 Transportation attendants Mechanics 1995–2002 508 Aircraft engine mechanics

515 Aircraft mechanics, except engine

503 Supervisors, mechanics and repairers

2003–2006 7140 Aircraft mechanics and service technicians 7000 First-line supervisors/managers of mechanics, installers, and repairers Fleet service (ramp) workers 1995–2002 883 Freight, stock, and material handlers, n.e.c. 887 Vehicle washers and equipment cleaners 2003–2006 9620 Laborers and freight, stock, and material movers, hand 9610 Cleaners of vehicles and equipment Agents and stores 1995–2002 318 Transportation ticket and reservation agents

364 Traffic, shipping, and receiving clerks

365 Stock and inventory clerks

2003–2006 5410 Reservation and transportation ticket agents and travel clerks 5610 Shipping, receiving, and traffic clerks 5620 Stock clerks and order fillers 5240 Customer service representatives Other air transport workers: All full-time air transport workers not in one of the above crafts

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

57

Table A2 Construction of Airline Craft and Comparison Groups in the CPS—Cont’d Comparison groups Comparison group workers are not employed in the air transportation industry Occupation codes (occupation codes and names posted at www.unionstats.com) Pilot comparison group 1995–2002 43–79; 173; 213–233; 235 2003–2006 500–1960 Flight attendant comparison group 1995–2002 243–285; 303–389; 433–447; 456–459 2003–2006 3600–3650; 4000–4160; 4300–4650; 4700–4960; 5000–5930 Mechanic comparison group 1995–2002 505–549 2003–2006 7000–7620 Fleet service comparison group 1995–2002 883; 889; 885–888 2003–2006 9620; 9360; 9610; 9640 Agents and stores comparison group 1995–2002 318; 323; 364; 365; 327 2003–2006 5410; 5420; 5610; 5620; 5350 Other worker comparison group: All full-time workers not in air transport (plus other sample criteria) See text for additional details and discussion on the CPS sample.

Table A3 CPS Log Wage Differentials by Union and Nonunion Airline Worker Groups, 2003–2006 Group

Skill Index Included

All

Union

Nonunion

No Yes

0.132 0.117

0.240 0.224

0.042 0.027

No Yes

0.220 0.194

0.284 0.261

0.131 0.102

No Yes

0.318 0.285

0.422 0.391

0.131 0.096

No Yes

0.156 0.148

0.209 0.207

0.009 −0.013

No Yes

0.229 0.180

0.321 0.271

0.092 0.042

No Yes

0.200 0.197

0.126 0.126

0.258 0.254

All air transport industry

Industry crafts

Pilots

Flight attendants

Mechanics

Fleet service

(Continued)

BARRY T. HIRSCH

58

Table A3 CPS Log Wage Differentials by Union and Nonunion Airline Worker Groups, 2003–2006—Cont’d Group

Skill Index Included

All

Union

Nonunion

No Yes

0197 0173

0.229 0.206

0167 0142

No Yes

−0007 −0006

0.085 0.093

−0043 −0045

Agents

Other

Identical to Table 2, except that the sample period is restricted to the years 2003–2006 following a new set of occupation and industry codes. Sample sizes for specific airline crafts are very small. See note to Table 2 and discussion in the text.

REFERENCES Air Transport Association. 2006. U.S. Airline Cost Index: Major & National Passenger Carriers, Fourth Quarter 2005, Washington, DC, July 14, http://www.airlines.org/economics/finance/ CostIndex.htm. Allegretto, S.A., Corcoran, S.P., Mishel, L. 2004. How does teacher pay compare? Methodological challenges and answers. Washington, DC: Economic Policy Institute. Bollinger, C.R., Hirsch, B.T. 2006. Match bias from earnings imputation in the CPS: The case of imperfect matching. Journal of Labor Economics 24 (3), 483–519. Busse, M. 2002. Firm financial condition and airline price wars. RAND Journal of Economics 33 (2), 298–318. Cahuc. P., Zylberberg, A. 2004. Labor Economics. Cambridge, Mass: The MIT Press. Card, D. 1998. Deregulation and labor earnings in the airline industry. In Regulatory Reform and Labor Markets, edited by J. Peoples. Norwell, Mass: Kluwer Academic Publishing, 183–230. Carey, S. 2005. Northwest, striking mechanics are to meet on talks. Wall Street Journal, October 12, p. B5. Comstock, P. 2006. Work-related views of pilots and flight attendants: Turbulence ahead? Per­ spectives on Work (The magazine of the LERA) 9 (2), 57–59. Crémieux, P. 1996. The effect of deregulation on employee earnings: Pilots, flight attendants, and mechanics, 1959–1992. Industrial and Labor Relations Review 49 (2), 223–42. Dade, C. 2006. Why some passenger pilots take huge pay cuts to fly cargo. Wall Street Journal, June 2, p. A13. Freeman, R.B. 1981. The effect of unionism on fringe benefits. Industrial and Labor Relations Review 34 (4), 489–509. Fournier, G.M., Hartmann, M.E., Zuehlke T. 2007. Airport substitution by travelers: Why do we have to drive to fly? In Advances in Airline Economics, Vol. 2: The Economics of Airline Institutions, Operations and Marketing, edited by D. Lee, Amsterdam: Elsevier. Gittell, J.H., Kochan, T.A. 2006. Low cost competition in the global airline industry. Perspectives on Work (The magazine of the LERA) 9 (2), 55–56. Gittell, J.H., von Nordenflycht, A., Kochan, T.A., 2004. Mutual gains or zero sum? Labor relations and firm performance in the airline industry. Industrial and Labor Relations Review 57 (2), 163–180.

WAGE DETERMINATION IN THE US AIRLINE INDUSTRY

59

Goodwyn, W. 2006. American Airlines “insources” maintenance work. All Things Considered. National Public Radio, Dec. 7, at http://www.npr.org/templates/story/story.php?storyId= 6594273&sc=emaf. Hirsch, B.T., Schumacher, E.J., 2004. Match bias in wage gap estimates due to earnings imputation. Journal of Labor Economics 22 (3), 689–722. Hirsch, B.T., Macpherson, D.M. 2000. Earnings, rents, and competition in the airline labor market. Journal of Labor Economics 18 (1), 125–55. Hirsch, B.T., Macpherson, D.M. 2003. Union membership and coverage database from the Current Population Survey: Note. Industrial and Labor Relations Review 56 (2), 349–54 (accompanying database at http://www.unionstats.com, updated annually). Hirsch, B.T., Macpherson, D.M. 2006. Union Membership and Earning Data Book: Compilations from the Current Population Survey (2006 Edition). Washington, DC: The Bureau of National Affairs. Johnson, Nancy Brown. 1995. Pay levels in the airlines since deregulation. In Airline Labor Relations in the Global Era: The New Frontier, edited by P. Cappelli. Ithaca, NY: ILR Press, 101–15. Matsa, D.A. 2006. Capital structure as a strategic variable: Evidence from collective bargaining. Working paper, September 28. Available at SSRN: http://ssrn.com/abstract=933698. Nay, L.A. 1991. The determinants of concession bargaining in the airline industry. Industrial and Labor Relations Review 44 (2), 307–23. Pierce, B. 1999. Using the National Compensation Survey to predict wage rates. Compensation and Working Conditions (Winter), 8–16. US Bureau of Labor Statistics. 2006. Job openings and labor turnover: April 2006. News Release, Table 8, at http://www.bls.gov/news.release/jolts.toc.htm. Wachter, M.L. 2004. Declaration and Expert Report of Michael L. Wachter. Exhibit 103, in the United States Bankruptcy Court for the Northern District of Illinois, Eastern Division, In re Chapter 11, UAL Corporation, et al., Case No. 02-B-48191, December.

This page intentionally left blank

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

3 Toward Rational Pricing of the US Airport and Airways System Daniel P. Kaplan∗�1

ABSTRACT The taxes and fees flights presently pay for use of the airports and airways in the United States presently bear little relationship to the cost of providing those services. As a result these prices generate significant distortions. Most notably, the present system of prices actually subsidizes flights operating at congested airport, and thereby foster delay. If the prices of the airport and airways system better reflected costs, there would not only be fewer delays, but more prudent investment decisions as well. This chapter considers how such a pricing system might be implemented and estimates the subsidies and overcharges to commercial aircraft operating at New York’s LaGuardia airport in 2004.

1 INTRODUCTION AND OVERVIEW For decades, US airlines have complained about the delays and the costs at the nation’s airports, as they fretted about the pace of improvements to the Air Traffic Control (ATC) system. These concerns have spawned a variety of policy proposals ranging from the imposition of congestion fees to the privatization of airports and the ATC system. Regardless of their merits, these proposals all face substantial political opposition. For example, air carriers believe they already pay too much for airport services, and therefore, wholeheartedly resist paying yet another charge. Organized labor, meanwhile, wants to keep ATC jobs in the public sector.

∗ Director, LECG, LLC 1725 Eye Street NW, Suite 800, Washington, DC 20006, Direct: 202-973-9877, Main: 202-466-4422, Fax: 202-466-4487, [email protected] 1 The author appreciates the comments of Darin Lee and David Gillen and the assistance of Zachary Kaplan. The chapter has benefited from conversations with Sandy Rederer, Mark Kahan, and Dorothy Robyn.

62

DANIEL P. KAPLAN

Despite the various complaints about the performance of the airport and airway system, there has been surprisingly little analysis of the prices users are currently paying.2 Price plays a central role in the efficient operation of a market by determining how much is produced and how that output is allocated. Price serves neither function in the case of the airport and airways system. Because prices for airport and airways services bear little relationship to cost, they actually foster delay by subsidizing low-valued services at congested facilities. Moreover, the system charges different users vastly different prices for the same service, thereby suppressing an important indicator of the value of additional capacity and distorting investment decisions. More rational pricing, therefore, has the potential not only to produce short-run gains at congested facility, but lead to more sensible long-run investments. Various government agencies in the United States (federal, state, and municipal) own and operate the airports and the ATC system, and there are significant monopoly elements in the provision of these services. Consequently, as either an owner or regulator, the continued role of the federal government in the pricing of the airport and airways system is virtually assured. A range of alternative pricing systems would likely lead to significant efficiency gains. In addition to discussing the deficiencies of the existing system, this chapter advocates a particular approach to pricing, which should reduce many of the existing distortions and could be implemented at relatively low cost. The potential impact of the new system is illustrated by considering the effect its adoption would have on prices currently paid by flights at New York’s LaGuardia Airport. Key features of the proposed system include: • Airport and airways services would be segregated into three separate activities and users would face charges for each. In addition to a fee for the use of an airport’s airfield, a flight would pay distinct fees for terminal services at each airport as well as for the en route control services it uses in traveling between them. • Flight operators would pay a price at least equal to the average variable cost (an approximation of marginal cost) of the relevant services, although those flying during periods when a facility is congested would pay more than users during uncongested periods. • Total revenue from the fees for any service would not exceed the operating costs of providing that service. Though aircraft operators would pay higher fees at certain periods, users during less congested periods would pay less. Accordingly, the proposed system does not involve congestion fees as the term has typically been used. • All system users would face the same set of fees, ending the existing distinction both between passenger and freight carriers and between commercial and private aviation. • Because airports and the ATC system offer complementary services, prices for both would be jointly established. This chapter consists of seven additional sections. Section 2 provides an overview of the taxes and fees aircraft operators currently pay for the use of the airport and airways system. Using FAA data, Section 3 provides rough estimates of the cost of

2 Michael Levine, however, did consider a number of these issues in an article in the 1960s. See “Landing Fees and the Airport Congestion Problem,” Journal of Law and Economics XII (April 1969), pp. 79–108.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

63

providing these services and highlights the wide discrepancy between costs and price. This misalignment is largely caused by prices being more closely tied to the value of the service to the user than the cost of producing the service. Section 4 discusses some basic issues the federal government should consider in designing a new price framework including the complementarities between ATC and airport services. Section 5 describes the proposed pricing system. New York City’s LaGuardia Airport is one of the nation’s most congested, and Section 6 estimates the gap between the prices paid by different flights and the costs of the services they receive under the existing system of charges. Though passenger carriers overall pay substantially more than the cost of the service they receive, a move to a more cost-based system would reduce the profitability of a substantial number of LaGuardia flights. Thus a move to a more cost-based system could reduce congestion at the same time that it reduced the overall amount carriers paid for the use of the airport and airways system. A concluding section offers some remarks about implementing the new system along with a discussion of the special treatment that might be afforded to flights serving small communities as well as for some private aviation.

2 EXISTING TAX AND FEE SYSTEM Two sets of agencies provide airport and airways services. The major (but certainly not the exclusive) role of the Federal Aviation Administration (FAA) is providing ATC services. The FAA is financed by general fund revenues as well as a variety of aviation taxes flowing through a trust fund. Receipts from these aviation taxes, however, reasonably approximate the cost of ATC services. These ATC services are provided in conjunction with the services of airports, which are, for the most part, owned and operated by various municipal governments. Airports provide a variety of services, but the analysis here focuses exclusively on the airfield. Airports have traditionally recovered these airfield costs through landing fees.

2.1 Federal Aviation Administration A combination of general fund revenues and aviation related taxes fund the ATC system. Technically, these taxes are paid into the Airport and Airway Trust Fund (Trust Fund), and the Trust Fund is the FAA’s major funding source. The Congress established the Trust Fund in 1970 to procure navigational aids and to develop airports. Since it established the Trust Fund, Congress has changed some rates and created some new taxes, but the present panoply of aviation taxes is largely consistent with those originally adopted.3 Though the taxes have remained more or less the same, the Trust Fund’s role has changed significantly. Now most Trust Fund revenues are used to underwrite the FAA’s operations, with only a portion devoted to the Trust Fund’s original purpose

3

There were aviation taxes prior to 1970, but these taxes went directly to the general fund. By creating the Trust Fund, the airlines hoped they would benefit directly from the fees they paid.

DANIEL P. KAPLAN

64

Table 1 Aviation Excise Tax Revenues, FY 2004 (In Millions) Excise Tax

Domestic Passenger

Domestic Freighter

General Aviation $52 $20

Foreign Carriers

TOTAL

$64 $21 $746

$4,929 $1,747 $1,539

Domestic passenger ticket tax Domestic flight segment tax International arrival and departure tax Domestic cargo and mail Fuel tax Alaska-Hawaii

$4�813 $1�706 $793 $28 $423 $59

$452 $60

$317

$4 $9 $12

$484 $810 $70

TOTAL

$7�893

$512

$317

$856

$9,579

Percent of total Percent of domestic

82.4% 90.5%

5.3% 5.9%

3.3% 3.6%

8.9%

Note:

1) Fractional ownership payments of $51.9 million in Ticket Tax; $20.2 million in Segment Tax; and

$8 million in fuel taxes are included under General Aviation. 2) The tex rate for commerical fuel is 4.3 cents per gallon, and the tax rate for general aviation av gas is 19.3 cents per gallon and 21.8 for jet fuel. Source: FAA.

of underwriting infrastructure investments.4 Operating the ATC system is the most significant of the FAA’s activities, but it also oversees airport improvements and the safety of aircraft and airlines, in addition to performing a variety of other aviation services.5 Receipts from nearly a dozen aviation related taxes flow into the Trust Fund, and they totaled over $9.58 billion dollars in FY 2004, of which $8.04 billion were raised by taxes on domestic service (Table 1).6 The tax rates applicable to any given flight, however, depend very much on its purpose. Flights by a commercial passenger carrier, a freight operator, and a corporate jet, each making precisely the same demand on the ATC system pay, very different amounts.7 4

See Federal Aviation Administration, “Budget in Brief, Fiscal Year 2006.” Trust Fund revenues pay for the FAA’s capital expenditures (specifically outlays for the Airport Improvement Program and for Facilities and Equipment) with the balance of Trust Fund revenues to be used to support FAA operations. In recent years, however, the budget has established a Trust Fund contribution to the FAA, which has exceeded Trust Fund revenues. As a result, the uncommitted balance of the Trust Fund has been reduced. (The General Fund contributed 26 per cent of the FAA revenues in 2004.) Also see General Accounting Office, Airport and Airway Trust Fund, GAO-03-979, September 2003, p. 12. 5 The ATO, a performance-based organization, was established in 2004 to incorporate functions relating to operating the ATC system. It does not have a separate funding source. 6 This is approximately equal to the FAA expenditures on operating the ATC system. Expenditures on ATC are not the same as operating costs, because the federal government does not distinguish between current and capital expenditures. International flights use some of the same services consumed by domestic flights. 7 Users of the transportation service nominally pay some of these taxes. Nevertheless, the identity of the person paying a tax has relatively little to do with the incidence of the tax. The following discussion assumes all taxes are paid by the entity operating the flight, which is unlikely to be the case.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

65

The ticket tax, which is a 7.5 per cent tax on domestic airfares, accounts for over half of the revenue generated by the various Trust Fund taxes.8 Commercial passenger carriers also pay a $3.10 per passenger segment tax. Carriers transporting cargo domes­ tically, whether by freighter or in the belly of a passenger aircraft, pay a 6.25 per cent fee on the value of the air transportation provided. In addition, commercial carriers also pay a 4.3 cent per gallon fuel tax. General aviation flights, on the other hand, pay only a fuel tax: 21.8 cents per gallon for jets and 19.3 cents per gallon for non-jets. The taxes on passenger and cargo revenues are valued based: taxes are tied to the amount the passenger or shipper pays for air service, which in turn is a reasonable approximation of the value consumers attach to the service. While not as directly tied to value, the segment tax is also value based because the amount paid increases with the number of passengers, and the more passengers onboard the greater the value of the flight. Even the amount a flight pays in fuel taxes is highly correlated to a flight’s value, because larger aircraft not only carry greater numbers of passengers but they also burn more fuel. Likewise both fares and fuel use tend to be correlated with distance. While within any category of flights, the taxes paid by a flight tend to be highly correlated with its value, there may be little if any correlation between taxes and values across categories of flights. In 2006, a general aviation jet flying from New York to Los Angeles paid $545 in fuel taxes, while a commercial B-767 flying the same route and using the same ATC services paid $2,500 dollars more.9 More generally, commercial passenger airlines, however, pay 90 per cent of the taxes associated with domestic aviation taxes, while accounting for only 60.7 per cent of the flights and 70.3 per cent of the flying hours of the principal users of the air traffic control system (Table 2).10 This disparity between their share of taxes and their share of activity has prompted claims of unfairness by, among others, the major carriers’ trade association.11

8 For all the Trust Fund taxes as of March 2006, see “Current Aviation Excise Tax Structure” on the Federal Aviation Administration Website (http://www.faa.gov/about/office_org/headquarters_offices/aep/aatf/media/ Simplified_Tax_Table.xls). 9 James C. May, “Smart – and Fair – Skies: A Blueprint for the Future,” Speech to the International Aviation Club, Washington, DC, April 18, 2006. (http://www.airlines.org/news/speeches/speech_4-18-06.htm). The Air Transportation Association, the trade group of the commercial carriers, calculates that a Buffalo to Philadelphia roundtrip flight on a commercial carrier would pay $900, while a business aircraft could pay as little as $22. Wall Street Journal, June 1, 2006. p. 1. 10 The output estimates include only those flights captured in the Enhanced Traffic Management System (ETMS), which records ATC use of flights operating under instrument flight rules (IFR). IFR flights tend to be the most intensive users of the en route control system. Though the vast majority of commercial passenger flights are recorded by ETMS, it includes about 35 per cent of general aviation flights. See Federal Aviation Administration, “Air Traffic Organization: Airports Data for Stakeholders, November 15, 2005.” 11 For Fiscal 2004, the Air Traffic Organization, which operates ATC, accounted for approximately 75 per cent of ATC outlays. See Federal Aviation Administration, Performance and Accountability Report, FY 2004, p. 90. (http://www.faa.gov/about/office_org/headquarters_offices/aba/offices/financial_management/ performance_accountability/media/2004_PAR.pdf) The aviation taxes fund FAA activities other than ATC operations.

DANIEL P. KAPLAN

66

Table 2 Use of Air Traffic Control for Domestic Service, FY 2004 User Classification

Passenger carriers Freighters Fractionals/non-sched part 135 General aviation-turbine Other TOTAL

Flights

10�746 901 1�809 2�884 1�361 17�701

En Route Activity Miles Flown

Hours

6�098�171 458� 328 644�476 659�734 343�393 8�204�101

15�933 1�257 2�211 2�063 1�204 22�668

Source: FAA.

When it originally established the aviation taxes, Congress probably gave little con­ sideration to the relative benefits and burdens they imposed on various segments of the industry. In any case, the regulatory structure established by the Civil Aeronau­ tics Board almost certainly limited the impact of the taxes on either fares or service.12 Moreover, private business jets were not very plentiful, and the express cargo services (i.e., services provided by the likes of Federal Express and UPS), which account for the bulk of the freighter service in the United States, had yet to be introduced. In short, in developing the current array of taxes, the government could not have possibly conceived the uses of the airspace or the demands on the system in the twenty-first century. The aviation taxes were also established with little regard for the cost of operating the ATC system. Nevertheless, in fiscal 2004 the revenues generated by these taxes were roughly in line with FAA expenditures on ATC. As noted previously, FAA generated $9.58 billion in 2004 taxes, which was less than 3 per cent higher than ATO expenditure. This comparison, however, overstates the correspondence between the two. Both the revenue and cost estimates include international service, and this analysis focuses on domestic services. In addition, the FAA records expenditures but not operating expenses, and the actual operating costs could be significantly different. Accordingly, this analysis proceeds on the assumption that the existing aviation fees pay for the operation of the ATC and the other FAA activities are funded with revenues from the general fund. This chapter’s inquiry, therefore, focuses on the issues surrounding the development of a more sensible and efficient method of collecting the revenues currently derived from the aviation taxes.

12 A person will only take a flight if the value attached to the flight exceeds its price, and the difference between a flight’s value and revenue is equal to its consumer surplus. The value of a flight to consumers, therefore, cannot be lower than the flight’s revenues. The percentage by which the value of the flight will exceed its revenues depends on, among other things, the elasticity of demand and passenger fares. The following discussion assumes that the revenues among flights are highly correlated with the value of the flights.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

67

2.2 Airport Fees In addition to a variety of subsidiary services, airports provide an airfield for aircraft landing and take-offs.13 The United States’ Department of Transportation (DOT) has established a policy with respect to airport rates and charges that, among other things, require airports to establish landing fees that do not exceed the cost of the airfield14 (DOT defines the airfield to include the runways, taxiways, and various other parts of airport properties). Though not required by DOT, airports, almost without exception, recover the cost of operating the airfield with a weight-based landing fee. While weight-based landing fees recover the cost of the airfield, the fee for any given flight does not necessarily reflect the cost of accommodating that flight.15 Instead the fee paid by any flight tends to be tied to the value of that flight. Heavier aircraft generally carry more passengers and more cargo, and consequently they generate more value. In that regard, landing fees resemble FAA taxes. Unlike the FAA taxes, however, aircraft regardless of purpose are subject to the same weight-based fees.

3 THE COST OF AIRPORT AND AIRWAYS Because economic efficiency requires prices to be aligned with costs, an understanding of both the output of the airport and airways system and the effect of variations in output on costs is a prerequisite to the development of a sensible system of charges. The current aviation taxes not only bear little relationships to cost, but the taxes are not even levied on the output of the system, which is an aircraft movement. This section begins by defining an aircraft movement for the purpose of developing an alternative price system and provides cost estimates of ATC services. The airport and airways system produces a service, which cannot be stored. The cost of a service, therefore, very much depends on the number seeking to use the service at any time.

3.1 Defining Output Price allocates goods and services among consumers and signals producers as to how much to produce. Though flight operators presently pay to use the ATC system, these

13

An airfield is only one part of an airport. Airports also facilitate passenger movement to and from flights as well as provide airlines with the space and facilities for needed services. If an airport operator was able to earn sufficient rents from airlines, passengers, and concessionaires for the use of the terminal and parking facilities it could conceivably recover its costs even if it offered use of the airfield at no charge. 14 Airports also establish rates and charges for non-airfield services, but these are not part of providing airport and airway services, and they are not considered here. See, Department of Transportation, “Policy Regarding Airport Rates and Charges,” Federal Register Vol. 61, No. 121, p. 31994. 15 In addition, to landing fees, most major airports also levy a Passenger Facility Charge (PFC) of up to $4.50 per departing passenger to fund capital projects. PFCs, unlike landing fees, are not calibrated to recover a particular set of costs but represent a source of funds for airport capital projects. Some of these projects relate to the airfield and others relate to terminal improvements. PFC payments among flights do not seem to be particularly aligned with the demands those flights place on the infrastructure. It would undoubtedly increase system efficiency if PFCs were incorporated into the proposed pricing system. This, however, is not explicitly considered here.

68

DANIEL P. KAPLAN

payments do not directly correspond to the service the aircraft operator purchases. For example, the ATC system guides flights between airports, yet the largest component of the price depends on the revenues generated by that flight, which is only remotely related to the cost of the service provided. The use of taxes unrelated to output is a common method of financing the provision of public goods, such as police services, with large spillovers. It is not appropriate, however, for essentially private transactions such as an aircraft operator purchasing ATC services. A market is where the purchasers and producers exchange some consideration, usually money, for a good or service. For a market to operate efficiently, the output of the good or service should be clearly defined and the price paid by the consumer and received by the producer should be directly related to the amount being produced. Because the airport and airway system enables aircraft to travel between places, the most appropriate definition of output is an aircraft movement.16 In the present context, an aircraft movement incorporates both a landing or take-off at an airport and an aircraft’s flight between airports.17 En route control can be best measured in terms of time, although mileage can be used as a reasonable approximation. There is an element of imprecision in defining the output of airport and airways system as simply aircraft movements.18 For example, heavier aircraft often require runways that are both longer and have more reinforcement; they may also require greater taxiway clearance. In other words, an airport with a capacity to accommodate 50 regional jet operations per hour may not be able to handle the same number of widebody jet (i.e., Boeing 747) operations. In addition, with the present technology, aircraft travel between airports along designated traffic lanes, and the capacity of these lanes are similarly affected by the size of the aircraft. Factoring in aircraft size, however, would likely affect long-run and short-run costs differently. Despite the higher cost of building an airfield built to accommodate larger aircraft, the variable costs of operating the airport would remain quite small regardless of the size of the aircraft using it. The effect of aircraft size on operating costs is not explicitly factored into this analysis, and including the effect of aircraft size on costs would be unlikely to have a material effect on the welfare gains from adopting a new system.

3.2 Defining Cost The FAA does not report operating costs like firms in the private sector.19 While the FAA distinguishes between expenditures made on “facilities and equipment” and from those on operations, it does not report the useful lives of its assets nor does it incorporate

16 Airport policies with respect to the pricing of non-airfield services are not nearly as uniform. Because

the various airports provide disparate non-airfield services and because of the wide array of contractual

arrangements between airports and flight operators, it would likely be counterproductive to attempt to dictate

a price-setting mechanism in this analysis.

17 As discussed below, ATC includes both control of aircraft in the terminal areas surrounding airports as

well as the control of airports traveling between terminal areas.

18 For some short flights, the terminal area where the flight originates is adjacent to the terminal area of the

flight’s destination. Flights between such terminal areas do not require en route control.

19 Federal Aviation Administration, Air Traffic Organization, “Data Package for Stakeholders”,

November 15, 2005.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

69

an interest charge. Consequently, calculations of fixed costs are imprecise. Most major airports, on the other hand, report financial results consistent with private sector practices. To begin with, most airports rely on capital markets to fund projects, and they must comply with various accounting processes in reporting their results. Moreover, the cost of operating an airfield is little affected by the number of operations, and thus virtually all the operating costs of the airfield are fixed. The efficiency of any cost-based pricing system can be improved by making the data more accurately reflect the cost of service. 3.2.1 ATC Establishing prices to recover costs requires estimates of both the costs to be recovered and the output to be produced. Combined with an understanding of the relationship between how costs change in response to fluctuations in output, these data permit the calculation of prices necessary to recover the costs. Estimates of the relevant rates are determined by using data provided by the FAA. These rates are subsequently used in Section 6 to analyze the costs of service at LaGuardia. Identifying Cost Pools. Table 3 provides an overview of FAA’s costs of operating the ATC system. The FAA has grouped its 615 operating facilities into three functional areas, and for each it has established four categories of costs. For each facility, it distinguishes between capital expenditures and other expenditures, which are for the most part, labor.20 For the purposes of the current analysis, we assume non-capital expenses at a facility to be variable.21 The FAA also allocates its administrative and overhead costs to each operating unit, and this analysis treats these overhead costs along with capital expenditures as fixed. This analysis focuses on establishing prices to recover the costs associated with terminal control and en route control. The costs associated with Table 3 Overview of Air Traffic Control Costs, FY 2004 Facility Type

Number of Facilities

Noncapital Operating Expenses of Facility

Capital Expenditures on Facilities and Equipment

Overhead Expenses

Total

Percentage of Costs which are Variable (%)

4�152 4�626 554 8�777

48�5 50�6

(in millions of dollars) Total en route Total terminal Flight station ATC TOTAL

26 528 61 554

2�016 2�342 375 4�358

1�333 1�462 60 2�795

803 822 119 1�625

Note: Honolulu (en route and terminal) is counted as one facility under terminal.

Other costs are included in overhead.

Source: FAA.

20

Also included with expenditures on equipment and facilities are contract expenditures for weather.

It is likely that at least some of these so-called variable costs do not vary with output and should be

classified as fixed. For example, during off-peak periods a facility that is staffed at a minimum level may not

have to add personnel to handle additional flights.

21

70

DANIEL P. KAPLAN

the operation of Flight Stations as well as other FAA activities, such as licensing and safety expenses, would need to be funded through either a dedicated set of taxes and fees or general fund revenues.22 Terminal control guides aircraft approaches at an airport as well as controlling take­ offs and landings. At many airports, a radar terminal provides both functions. In areas with several airports (including military as well as civilian), terminal radar approach control (TRACON) provides approach control while a limited radar tower handles the landings and take-offs. Of the 500 airports in the United States, 200 have a radar tower or a limited radar tower operated in conjunction with a TRACON.23 Ninety-five per cent of commercial passenger operations occur at these airports.24 Other airports have more limited service. For example 73 airports have VFR Towers, which control landings and take-offs but not approaches. Other towers are staffed by contract employees, while some have no staff at all and provide only automated services. In recovering the costs of terminal control, separate cost pools could be established for each of the more than 500 facilities or, as is done here, all the facilities could be lumped together into a single cost pool. Alternatively, several separate cost pools could be established by combining various facilities with similar characteristics, for example, size or location. Limiting the number of cost pools would make the ratemaking process more transparent to users and easier to administer. En route control involves both domestic and international operations. This analysis does not explicitly consider international services, but in practice one or more separate cost pools could be established for international service.25 The system would likely be most efficiently administered if the costs of domestic en route control were allocated to a single cost pool. Aircraft regularly use tens of thousands of routings, and each flight path requires a unique set of control services. With multiple cost centers, each routing would potentially be subject to a unique charge, increasing administrative difficulties and making price signals unnecessarily opaque to system users. Computing Unit Costs. In FY 2004, ATO spent $8,777 million on Terminal and En route control, with 47.3 per cent of the total dedicated to en route control. For both activities, the non-capital expenditures at each facility were assumed to be variable and to represent one-half of the total cost. Table 4 calculates the unit costs by dividing the costs of operating the relevant ATC service in 2004 by the number of operations. On average, the terminal control cost per landing or take-off was $130.67 and the cost of the en route control was $171.01 per hour.26

22

Flight Service Stations consist of 58 facilities that provide weather briefings, flight plan filing services, and

other assistance to private pilots. The cost of operating these facilities was $554 million in 2004, although

the FAA expects to reduce those costs significantly by turning their operation over to a private contractor.

See Robert W. Poole, Jr, “Outsourced Flight Service Stations Save FAA $2.2 Billion,” Reason Foundation

Commentary, September 1, 2005. (http://www.reason.org/commentaries/poole_20050901.shtml.)

23 See “Data Package for Stakeholders,” p. 8.

24 In contrast, 35 per cent of general aviation operations occur at these airports. Around 60 per cent of general

aviation flights provided in jet aircraft (which for these purposes include fractional ownership aircraft) operate

at these large airports. Ibid, p. 10.

25 For example, separate international cost centers could be established for the Atlantic, Pacific and Latin

America.

26 A flight requires terminal operations at the departing as well as the arriving airport.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

71

Table 4 Unit Costs of Air Traffic Control, FY 2004 ATC Operation

Cost ($)

Landings and Take-offs

Hours En Route

Unit Cost ($)

(in thousands) Terminal control En route control

4,625,900 3,876,500

35,402 22,668

130.67

171.01

Source: Tables 2 and 3.

3.2.2 Airports Though the costs of operating the various ATO facilities providing terminal control (or for that matter, en route control) can be combined into a single cost center, the operating costs of the various airports cannot be similarly aggregated. The various airports are owned and operated by separate entities, each of which have made financing and other commitments tied directly to the operation of the facility.27 Thus, airfield cost pools will continue to be calculated as they are presently. As discussed in detail below, however, the method for recovering those costs will change for some airports.

3.3 Effect of Demand on Cost Because airport and airway services cannot be stored, average cost pricing may not be a reasonable method of recovering costs. Excess capacity at 5 a.m. is of little value to some one wanting to fly at 5 p.m. Faced with insufficient capacity to serve peak period demand, an operator would have to ration the available supply in the short run. In most markets, price is that rationing device, but where the price mechanism is suppressed, other rationing devices evolve. In the case of the airport and airways system, that mechanism is delay.28 It is generally efficient for peak period users of a facility to pay more. Peak period demand determines the size of a facility – a smaller facility could accommodate demand if the use of the facility were spread evenly throughout the relevant time period. Thus a facility operator could profitably expand the size of a facility if two conditions were met: 1) the facility could not otherwise accommodate peak period demand and 2) the revenues generated as a result of the added capacity were sufficient to recover the cost of the added capacity. Because extra capacity is required to accommodate peak period demand, the peak period users should bear the additional cost.29 The addition of a flight during a peak period increases the delays experienced by existing users without affecting the costs of producing the service provided. Thus, a 27 Some entities, for example the Port Authority of New York and New Jersey, own and operate several airports. 28 High prices encourage suppliers to expand output. In the absence of price, mounting delays prompt expansion decisions. Delay statistics, however, do not provide the same quality of information as price. Consider, for example, the case in which delay is caused by low value users. 29 See, for example, W. Kip Viscusi, John M. Vernon, and John E. Harrington, Jr, Economics of Regulation and Antitrust, 2nd Edition (MIT Press: Cambridge, MA. 1995) pp. 399–403.

72

DANIEL P. KAPLAN

transaction involving two parties (an aircraft operator and the airport) imposes costs on those not party to the transaction. This phenomenon, commonly referred to as an externality, is discussed below.

4 BASIC PRINCIPLES UNDERLYING A NEW PRICING SYSTEM Competitive markets generally produce efficient outcomes. They provide consumers with low prices by forcing prices down to producers’ costs and by forcing producers to operate at low costs. Markets for airport and airways services, however, are not competitive. Neither ATC nor airports face much competition, and the facilities are government owned and operated. Relying on the competitive market is, therefore, not an alternative, and the federal government will continue its major role in either establishing or regulating prices of airport and airways services for the foreseeable future. Competitive markets efficiently allocate goods and services by providing both con­ sumers and producers with meaningful signals to guide their consumption and production decisions. The existing prices for airport and airways services bear virtually no relation to the cost of service, and accordingly, do not provide any guidance to producer or consumers as to their efficient allocation. Numerous government agencies have designed regulatory regimes to encourage firms with market power to make price and output decisions consistent with the regulators’ notions of social welfare.30 These efforts, however, have met with only limited success. The pricing mechanism proposed here has much more limited goals; it seeks to create a new set of prices that lack the perverse incentives of the current system. The proposed pricing regime does not treat congestion (and the resulting delays) as an externality, and therefore, an inevitable consequence of providing airport and airways services. The political process – and not the underlying supply and demand conditions – is preventing government agencies from establishing market-clearing prices. The existing economic literature has mistakenly categorized the inability of the political process to craft a satisfactory pricing mechanism as a market failure.31

4.1 Basic Criteria for More Efficient Prices As both a producer and a regulator, government agencies should strive to establish prices consistent with economic efficiency. In the absence of externalities, competitive markets establish efficient prices by establishing prices equal to marginal cost. Moreover, such prices generate sufficient revenues to cover a firms’ operating costs.32 In addition, the

30

For a discussion of the regulation of industries from electric power to trucking, see Viscusi et al. op.cit.,

pp. 377–652.

31 See, for example, Christopher Mayer and Todd Sinai, “Network Effects, Congestion Externalities, and

Air Traffic Delays: Or Why All Delays are Not Evil,” American Economic Review 93 (September 2003),

pp. 1194–1215.

32 Competitive prices cover the costs of efficient firms. The efficiency of government operated enterprises

relative to private firms is not considered here.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

73

new pricing scheme should require prices to be reset periodically to reflect changes in costs and demand. It should also reflect the complementarities between airport and ATC services. 4.1.1 Establishing Prices Equal to Costs Economic efficiency very much depends on the relationship between price and marginal cost.33 With price equal to marginal cost, the value a purchaser attaches to an additional unit of output just matches the cost of the resources needed to produce that added output. A government agency charging such cost-based prices may be unable to generate revenues sufficient to recover fully the costs it incurs in producing an efficient level of output. This is most likely to be a complication where economies of scale are present. Under those circumstances, it is generally most efficient to recover the revenue shortfall through a value-based fee. Variable Cost Recovery. In a competitive market, a firm finds it profitable to expand output so long as price exceeds marginal cost, which is the cost of an additional unit of output. Marginal costs, however, are notoriously difficult to calculate. In the short run, capacity is fixed, and marginal costs consist mostly of labor. Accordingly, average variable cost has generally been accepted as a reasonable proxy, and it serves that function under the proposed pricing scheme.34 Fixed Cost Recovery. By definition, price equal to average variable cost does not recover the fixed costs of providing a service. While it would be inefficient to charge less than average variable cost, short-run efficiency does not require the recovery of fixed costs. Fixed costs are incurred regardless of use, and it would be inefficient to discourage demand during periods of excess capacity by burdening some users with an unnecessarily large share of those costs. Two factors – economies of scale and demand fluctuations – are important in determining how these fixed costs should be recovered. First consider the case where the service is produced subject to constant returns to scale. With constant returns to scale, the size of the facility can be matched with peak period demand. Capacity will be expanded so long as peak users place a sufficiently high value on the added capacity to finance the requisite investment. In the presence of constant returns to scale, therefore, price could vary over time as it balances demand with capacity. All users operating during periods in which there was no excess capacity would bear the fixed costs of the facility, although users during periods of higher demand would pay more. In other words, the charges to recover fixed costs would be established with the goal of reducing variations in demand over time. In no case, however, would a user pay less than average variable cost and the total revenues would not exceed the costs of operating the facility.

33

The output that equilibrates price and marginal cost is efficient, because it would be inefficient to produce either more or less. If too much were produced, users place a lower value on the added output than its cost of production, and thus consumers are unwilling to pay the cost of the extra unit of production. Price above incremental cost is also inefficient, because users place a higher value on additional output than its cost to product, and accordingly output should be expanded. 34 See, for example, Phillip Areeda and Donald Turner, “Predatory Prices and Related Practices Under Section 2 of the Sherman Act,” Harvard Law Review, Volume 86, 1975, pp. 697–733.

74

DANIEL P. KAPLAN

In the presence of economies of scale, a different cost recovery system may be required. Indivisibilities may necessitate investing in too much capacity at any given time.35 As a result, excess capacity may be present even during periods of high demand. In that case, fixed cost recovery would require at least some users to pay more than marginal cost. It has long been recognized that the most efficient method of recovering fixed costs in the presence of excess capacity is through Ramsey prices.36 Ramsey prices allocate fixed costs among users based on their elasticities of demand – those with less elastic demand contribute a disproportionately large share toward the fixed costs. The lower a purchaser’s elasticity of demand, the smaller the impact a given percentage change in price will have on the amount purchased. Accordingly, Ramsey prices are efficient, because they recover the fixed costs and produce the smallest deviation from the output produced by marginal cost prices. A strict implementation of Ramsey prices is technically not possible, because price elasticities of individual purchasers are not observable. As an alternative, variations in prices among purchasers would be based on some proxy The greater the value a consumer attaches to a good or service, the smaller the impact a given percentage change in price has on the quantity demanded. As previously noted, many of the present aviation taxes and fees are value based. The proposed pricing scheme employs aircraft weight and aircraft weight minutes (i.e., an aircraft weighing a thousand pounds flying for 100 miles minutes generates 100,000 aircraft weight minutes) to measure value. The value of a flight is related to its payload, which in turn is likely to be highly correlated with the size of the aircraft. Similarly, longer flights tend to be more valuable.37 Aircraft weight is an equitable and transparent measure of value, because the relationship between aircraft weight and capacity exists regardless of what the flight transports or whether it is a commercial or private service.38 4.1.2 Responsive to Changes in Costs and Demand An added flight during a congested period may significantly degrade the service quality of others. Yet the degradation in service quality reduces demand. A higher price, therefore, might not only be profitable, but could also actually improve the quality of service. (In fact, because of the improved service quality, the higher price would actually increase demand.) Because users would be charged more during periods where congestion resulted in consistent delays, the price at any time depends not only on the costs at the facility but on demand as well. With demand and capacity held constant, prices would be changed periodically to reflect changes in costs of producing the service – a 10 per cent increase in cost would produce a 10 per cent price increase. If, however, increased demand produced increased

35 An investment in too much capacity may be prudent in anticipation of future growth or to employ a lower

cost technology.

36 Viscusi, et. al., op.cit., pp. 365–367.

37 Time is a more accurate measure of value than distance, because the cost per mile of an aircraft’s operation

is lowest when the aircraft reaches cruising altitude.

38 This relationship is not exact. Newer aircraft use lighter materials and are therefore can carry greater

payloads per unit of aircraft weight. In fact, a whole new class of very light jets, which are expected to be

widely employed as air taxis, should begin operating in 2006.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

75

congestion during a particular period, a larger price increase during that period might be appropriate. Specifically, the price for peak period users would be established sufficiently above average variable cost to limit the delay faced by users of the facility at the time. Because the total revenues generated by the fees cannot exceed cost, an increase in the price above the percentage change in cost at any given time requires a corresponding lower rate of increase during other periods. Determining the necessary adjustments to be made is not straightforward. There is uncertainty with respect to the effect of price changes on demand as well as their effect on revenues. Increases in peak period fees would be dictated by delay statistics with higher fees generally associated with periods of time experiencing longer delay. Fees in adjacent periods may also need to be raised to limit the likelihood of congestion being shifted into those periods. For example, raising fees for flights between 5 and 5.30 p.m. may lead some users to simply shift their flights to a later or earlier time period. This may simply shift the congestion to a different period. While it might be appropriate to adjust adjacent period prices, a period with a higher fee should not have fewer delays than a period with a lower fee. 4.1.3 Airports and ATC Offer Complementary Service A flight uses both airport and airways services, and a change in the price of one of the services affects demand for the other. An increase in an airport’s peak period fees would affect the timing and the mix of aircraft using ATC services as well. The relationship is especially close in the case of airports and terminal operations, because aircraft employ both in fixed proportions.39 Suppose, for example, an airport experiencing congestion moved to a cost-based movement fee for the use of the airfield. Independently adjusting terminal operations to reflect congestion could produce prices too high and as a result generate excess capacity during peak demand periods. Because the use of airports and ATC services are so closely related, the prices for the services of airports and the ATC should be established in conjunction with one another.

4.2 The Economics and the Politics of Airport Delay A number of economists consider delay at a congested airport to be the result of a market failure – users do not pay the costs associated with the increased delays that is associated with an additional flight at the airport. Advocacy of congestion pricing began in the mid-1960s in response to a significant increase in delays at a number of airports, and the view among some economists that the problem could be best viewed as an externality. These congestion problems were ultimately resolved with Congress limiting aircraft operations at five airports. Congestion has been a recurring problem, even as restrictions at those airports are actually being relaxed. The government has been exploring alternative methods of controlling access including a “market solution”, which is a euphemism for congestion pricing.

39

Flights use the services of terminal control and the airfield in fixed proportions – much like the sale of left and right shoes. Unlike shoe manufacturing, however, the same firm does not produce both terminal control and airfield services, and it would be sensible if the entities coordinated the price-setting process.

76

DANIEL P. KAPLAN

Delay can be considered an externality because the transaction between a flight oper­ ator and, for example, the airport, can affect other operators. Establishing a congestion fee to reflect the delay costs imposed on others would more closely align the price a user had to pay for a service with the full social cost of the service. By increasing the costs of operating during congested periods, such a fee discourages those flights that place a relatively low value of operating at the facility during peak periods. This is economically efficient because a well-functioning market allocates a scarce resource to those valuing it most highly. Thus, even if the congestion fee were established to maintain existing levels of delay, the change in the flight mix at the airport could yield welfare gains. A higher price would reduce delay and could thereby yield even further gains.40 Explicitly incorporating demand into the establishment of prices would yield similar welfare gains at a lower overall cost to users. By charging a higher price, a so-called congestion fee would internalize the externality. While delay is an externality, it is not a natural outcome of producing airport and airways services. As already noted, the present pricing system actually subsidizes system use during high demand periods. In other words, delay is at least partly the result of the peculiar pricing system established by the political process. Users fear that the imposition of congestion prices will increase carrier costs with no assurance of a corresponding improvement in service quality. Congestion fees increase the cost of all the flights operating during peak periods – high-value and low-value flights would both pay higher fees in order to encourage the low-value flights to move to either less congested time periods or different airports. Yet the existing pricing system actually subsidizes these low-valued flights. Thus the imposition of congestion prices would require high-value flights to continue to sub­ sidize airport and airways services received by these low-value flights, and then pay yet another fee as part of an effort to encourage these subsidized flights to operate differently. When combined with the present pricing system, a congestion fee would require high-value flights to pay twice – an above cost fee to subsidize low-value flights and then a congestion fee to undo the effect of the subsidy. Clearly, it would be far more efficient to align prices more closely with cost before even considering imposing congestion fees. Not only may it be a mistake to classify delay as the product of an externality, recent economic research has suggested that, at least at some airports, delay may not even be a reliable indicator of market failure.41 At many major airports, a hub carrier accounts for a large share of the operations. In order to minimize the time passengers must spend on the ground between flights, a hub carrier often bunches its arrivals and departures. An increase in the hub carrier’s flights will increase delays at the airport, but it is the hub carrier’s own flights that will experience most of the increased delay. Delays a carrier imposes on itself are not an externality. With hub carriers typically accounting for more than half of the flights at their hubs, delays at such airports should not be considered as the actions of two parties affecting a third.

40 It would be both efficient and profitable for an airport to raise price in the face of congestion so long as the value of the resulting reduction in delay exceeds the additional revenue generated by the price increase. 41 See, for example, Mayer and Sinai, op.cit.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

77

The proposed pricing system is not congestion pricing per se, although it recognizes congestion as a factor in allocating the costs of the airport and airways system among system users. The higher prices during more congested periods do not reflect the costs of congestion as much as they provide a clearer signal of the value of expanding a facility. Moreover, unlike congestion prices, the charges under the proposed scheme will not generate revenues in excess of the cost of the service provided.

5 DEFINING THE NEW SYSTEM While representing a movement toward cost-based prices, the proposed price system nevertheless maintains important characteristics of the existing system. Most notably, the proposed pricing system generates revenues equal to the costs of providing the relevant services, and it maintains important attributes of the current system of value pricing. It, however, eliminates different prices for different types of users, and it establishes consistent pricing between airports and airways, which are complementary services. Under the proposed system, a flight would be assessed charges for the airfield and for the terminal operations at the airport where the flight originates as well as at the destination airport.42 These charges would be established to recover specific and welldefined costs. In addition, no user would pay less than average variable cost, a proxy for marginal cost. During periods of congestion, however, this base fee would be increased in order to reduce delay at the facility. Flights operating at the airport would, therefore, pay the greater of average variable cost- or a congestion-based fee. If the revenues from these fees failed to generate revenues sufficient to cover the relevant costs, flights would also be assessed a weight-based fee to make up the shortfall.

5.1 Airports In the hypothetical competitive market, an airport would charge all flights uniform fees with the specific fee at any time reflecting the level of congestion. During periods of low demand, at night for example, price would approach zero, because the marginal cost at airports is quite low. Because the levels of congestion may differ between the time an aircraft arrives at an airport and when it departs, it would likely be efficient to institute separate charges for landings and take-offs. In the subsequent discussion, such a charge is referred to as an airport use fee. While efficient, the revenues generated by these congestion-based prices might be less than the operating cost for the airfield Weight-based fees would recover the shortfall. At airports with substantial excess capacity, therefore, virtually the entire airfield costs would be recovered through a weight-based fee, which is precisely how landing fees

42 As discussed in detail below, at uncongested airports the airfield fees will be constant over time. As a result, these airports can continue to charge a single fee to cover both landings and take-offs as is the case with existing landing fees.

78

DANIEL P. KAPLAN

are currently established.43 There are likely to be many such airports. Runways are not scalable: a runway built to handle one operation per hour can also accommodate up to 45 per hour, resulting in excess capacity for large portions of the day, and in some cases, the entire day.44 This two part charging system – a uniform fee based on congestion coupled with a weight-based fee to recover the shortfall – requires large aircraft operators to continue paying fees well in excess of the cost of service at airports with limited or no congestion. As noted earlier, however, weight-based fees are likely to be reasonable approximation of Ramsey prices, and therefore, a relatively efficient means of recovering fixed costs. A shift to more cost-based prices generates an efficiency gain by providing some basis for quantifying the value of the expanded capacity. That portion of the airport use fee that is influenced by the level of congestion at the airport would be an important input into any analysis of the added value. Added capacity benefits peak period users. In a pricing environment with a minimum airfield use fee, additional airfield capacity would lower these minimum fees and encourage operations by lower valued flights. The decision to expand capacity should be based on whether the revenues generated by the additional flights would be sufficient to justify the added cost of expanding the facility. Decisions to expand capacity should not be based on the ability to recover the cost of the expansion from users that would be accommodated by the existing facilities if they were being priced efficiently. In other words, the ability to extract higher fees from heavier aircraft should not be considered in valuing capacity additions.

5.2 ATC Services Under the proposed system, the two components of ATC services, terminal control and en route control, would be priced separately. Terminal control costs, like airfield costs, are very much influenced by take-offs and landings at a particular airport. En route control, on the other hand, is influenced by the amount of time a flight is airborne. This analysis assumes ATC services are subject to constant returns to scale over a wide range of outputs, and unit costs of ATC services are uniform across the country. A cost pool for terminal operations at each airport, therefore, would be derived by applying a nationwide unit cost (e.g., cost per aircraft movement) to the expected activity at the airport.45 5.2.1 Terminal Control Because the cost of terminal control is airport specific, the derivation of the price of terminal control is also airport specific. As in the case of airfield services, each flight (and each take-off and landing) is priced separately.

43 Weight-based landing fees are also justified by larger aircrafts’ requirements for longer runways, greater reinforcement and wider taxiways. 44 The capacity of the airfield, however, can be expanded by constructing more taxiways. By permitting aircraft to spend less time on the runway, an expansion increases the amount of time the runway can be used to for landings and take-offs. 45 In practice it may be advisable to establish more disaggregated cost pools – for example, by metropolitan area or by a characteristic of the metropolitan area such as size.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

79

All flights, regardless of when they operate, would pay at least average variable cost for terminal control.46 Additions to airport capacity can sometimes be only made in relatively large increments – a runway, for example, represents a significant increment to capacity. In contrast, terminal operations at an airport are much more likely to exhibit relatively constant returns to scale, and, as a result, the facilities needed to provide the appropriate level of services can be tailored to match demand. Because of the flexibility in determining the size of the terminal operations at an airport, there is less likely to be excess capacity, which also reduces the need to employ weight-based pricing to recover fixed costs. Unlike airfields, therefore, value pricing would ultimately likely play only a very small role in the recovery of the costs of terminal operations. Nevertheless, because of the importance of value pricing under the current system, the continued use of value as a component in the pricing of terminal services may be warranted in any transition. Some aircraft operators may have made investments in equipment and other resources based on the existing system of charges. Continuing to have a significant – although over time diminishing – component of the terminal price based on value may soften the financial impact of the shift and make implementing a new pricing regime more politically palatable. 5.2.2 En route control En route control provides ATC services between the arriving and departing airports. Unlike the space surrounding an airport, the space between airports is virtually limitless. In practice, aircraft are routed along defined traffic lanes, which effectively limit capacity at any given point in time. Yet because these traffic lanes are not made of brick and mortar, they are generally less a source of chronic congestion than either terminal control or the airports themselves. First, the FAA has some flexibility to determine the route of any given flight. In addition, over time, FAA has substantial flexibility to reconfigure these highways in the sky, and can add traffic lanes to meet demand growth. More significantly, the FAA is working on a system to increase a flight’s flexibility in selecting its own course in moving between airports. Such “free flight” not only reduces travel times between airports, but it further diminishes the possibility of aircraft experiencing en route congestion. The variable cost of providing en route control services to an aircraft depends directly on the length of time it travels between airports.47 Identifying peak travel periods and assessing fixed charges is more difficult in the case of en route control than it is for terminal control. For example, simply defining a traffic peak would be challenging – different areas experience peak traffic at different times. Moreover, the use of a weightbased measure to recover fixed costs is unlikely to introduce significant distortions. 46

Because unit costs of terminal control are the same at all airports, average variable costs are also the same. This assumes that average variable costs are constant over time. In fact, average variable cost may be lower during off-peak periods. 47 During take-off and landing flights not only travel more slowly, but they also do not fly in a straight path because of the need to position themselves. After leaving terminal control, flights are nearing their cruising speeds, and thus the time for which a flight uses en route control is closely related to distance. The variable cost of operating en route control, therefore, can be allocated to a flight based on either the amount of time or the number of miles, although the relationship between mileage and cost would not necessarily be linear.

80

DANIEL P. KAPLAN

Nevertheless, such weight-based prices involve subsidizing low-valued flights, and this may be a problem at airports where congestion remains even after the imposition of cost-based prices for both the airfield and terminal operations. At such airports, therefore, it may be prudent to recover the fixed costs of en route control through prices that are tied to congestion.

6 EXAMPLE OF LAGUARDIA AIRPORT There is substantial demand by aircraft to operate at New York’s LaGuardia Airport. This airport, however, can only use one of its two intersecting runways at a time, and because it is surrounded by water and Queens, expansion is not a realistic option. Congestion has been a major problem at the airport for over 40 years, and the federal government has limited access to it since 1968, when it was one of a handful of airports governed by the High Density Rule (HDR). The HDR set hourly limits on the number of take-offs and landings at the airport, and it allocated operations among various different classes of carriers. The Congress has ordered an end to the use of the HDR at LaGuardia in 2007.48 General aviation accounted for less than 4 per cent of operations at the airport in 2004, but there is considerable unmet demand by private aircraft. Even if private aircraft were subject to the same prices as other aircraft using the airport, their use of the airport would undoubtedly increase substantially unless otherwise constrained. The following analysis assumes general aviation operations at LaGuardia would continue to be capped. Indeed, it is likely that even with the proposed pricing system, additional measures might be needed to limit congestion at the airport. Despite the apparent great demand to serve LaGuardia, nearly half the passenger aircraft serving the airport in 2004 had 50 seats or less.49 The proposed system would increase the amount these aircraft would pay more for the use of the airport and airway system, and thereby have an adverse effect on the profitability of flights using such equipment. In contrast, the cost of using larger aircraft would decline. The ultimate impact of the new pricing system, however, would depend on how carriers change their existing services as a result of these changes in profitability and the new services they introduce.50 The estimates in Tables 5–11 are based on services patterns at LaGuardia in 2004, and they provide a rough approximation of the costs and revenues of the various services offered at the airport. The calculations are merely illustrative. Most notably, no attempt

48 In 2000, the Congress exempted regional jets from small and medium communities from the HDR. This relaxation of the rule produced a flood of new service and dramatically increased delays at the airport. 49 These statistics are derived from DOT May 2004 T-100 statistics and exclude international flights. Passengers on international flights accounted for 5 per cent of the total at the airport during 2004. See Port Authority of New York and New Jersey, “December 2004 Passenger Report” (http://www.panynj.gov/). 50 While carriers can and do sell their operating rights, it is unlikely that the current mix of flights represents the highest valued services that could be offered at the airport. By refusing to sell an operating right to a higher valued user, a carrier preserves its option to realign its schedule in the future while preventing a rival’s introduction of a competitive service in the near term. Such strategic considerations may limit the ability of a free market in operating rights to result in an optimal pattern of service.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

81

has been made to factor a flight’s contribution to the carrier’s network in developing the revenue estimates.51

6.1 Taxes Under the Existing System To understand the impact of a new pricing system at LaGuardia, the revenues generated by the existing tax and fee system must first be estimated. To do this, flights are grouped into five categories based on aircraft size: the largest consists of aircraft with more than 120 seats, and the smallest has aircraft with fewer than 30 seats. Table 5 shows, the number of departures for each size category as well as averages of the number of seats, distance, and number of passengers. The three smallest size categories of aircraft, each with 50 seats or less, are operated by regional carriers. Though regional carriers are increasingly operating jets with more than 50 seats, these represented a very small proportion of the operations at LaGuardia in 2004. Thus, flights with more than 50 seats are assumed to have been operated by mainline carriers. Estimating the taxes and fees for the average flight in each aircraft group requires information on both aircraft weight and average fare. To determine aircraft weight, a representative aircraft was selected for each size class.52 The average fare for each aircraft was based on the statistical relationship between distance and local fares fare for all LaGuardia markets with nonstop service in the second quarter of 2004.53 Table 6 provides an estimate of the revenues generated by the average flight of each of the five aircraft types as well as the taxes and fees such flights would pay. In computing revenues, a flight is assumed to carry the average number of passengers of its respective size class, with each passenger paying the average fare of local LaGuardia passengers traveling that distance. The ticket tax is 7.5 per cent times those revenues, and the segment fee is $3.10 times the number of passengers.

Table 5 LaGuardia Flights by Aircraft Size, May 2004 Aircraft Seats

120 and up 51–119 40–50 30–39 less than 30

Departures Number

Per cent

7,263 788 2,659 4,046 364

48.0 5.2 17.6 26.8 2.4

Average Seats

Average Passengers

Average Minutes

Average Distance

146�5 107�7 49�4 36�8 19�6

105�8 82�0 30�5 20�4 9�1

103�2 106�2 78�5 67�8 52�1

742�0 772�6 496�8 328�9 178�7

Source: DOT, T-100.

51

For example, the economics of a flight between New York and Boston with only local passengers is very

different from a flight between the two cities where a significant proportion of the passengers are connecting

transatlantic passengers.

52 Maximum gross take-off weights were taken from the websites of the various manufacturers.

53 This is likely to overstate the onboard yields because connecting passengers typically have lower yields

than local passengers.

DANIEL P. KAPLAN

82

Table 6 Taxes and Fees Under Existing System, 2004 Aircraft

A-320 B-717 ERJ-145 ERJ-135 Twin Otter

Landed Weight

Yield per Mile ($)

Landing Fees ($)

Aviation Taxes ($)

TOTAL ($)

145.5 102.0 42.5 40.8 12.5

0.216 0.208 0.315 0.464 0.825

578.36 405.45 140.94 135.10 41.41

1,599.79 1,241.98 453.17 296.58 128.53

2,178.15 1,647.43 594.11 431.68 169.93

Source: Manufacturers websites and DOT Origin and Destination Survey.

Most airports assess landing fees for a flight’s landing and its subsequent take-off, although this analysis assumes landings and take-offs are assessed separate airfield use fees. Because the landing fee at LaGuardia is relatively high, it is assumed that the mainline aircraft operate to airports where the appropriate fee is one-half as large, and the smaller aircraft operate at airports where the landing fee is one-fourth the fee at LaGuardia. This analysis does not specifically incorporate the taxes on fuel and cargo or the PFCs charged by most large airports. As already noted, these taxes and fees, like the ones considered here, are for mostly value based and including them would not fundamentally affect the analysis.

6.2 Cost of Service Table 7 estimates the average cost of providing airport and airways system for the five hypothetical flights considered above. The average cost estimates use the ATC estimates derived in Section 3 and are movement based. Airfield use fees were derived by assuming the existing landing fees recover airfield costs. The existing difference in taxes and fees among aircraft is quite large. A flight in an A-320 flight pays fees that are more than 10 times the amount paid by the Twin Otter flight and three times the

Table 7 Average Cost of LaGuardia Flights, 2004 Aircraft

A-320 B-717 ERJ-145 ERJ-135 Twin Otter

Terminal Control

Landing Fees

LaGuardia ($)

Other ($)

LaGuardia ($)

Other ($)

131 131 131 131 131

131 131 131 131 131

249 249 249 249 249

124 124 62 62 62

Source: Author’s calculation.

En Route Control ($)

Total Cost ($)

294 303 224 193 149

929 937 796 766 721

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

83

Table 8 Price–Cost Differential, 2004 Aircraft

A-320 B-717 ERJ-145 ERJ-135 Twin Otter

Current Taxes and Fees ($)

Average Cost ($)

Difference ($)

2,178 1,647 594 432 170

929 937 796 766 721

1,249 710 −202 −334 −551

Note: Aviation Taxes derived by assuming all passengers pay estimate of

average local fare for the appropriate stage length.

Source: Authors calculation.

amount of a flight by a 50-seat EMB-145.54 Because the costs of the airport and airway services are predominately movement based, these price differences are largely unrelated to differences in the cost of the services provided. Accordingly, the A-320 flight pays $1,240 more in taxes and fees than the average cost of the service it receives (Table 8). In contrast, the Twin Otter pays $551 less than its average cost of service, and even the average flight by 50-seat ERJ-145 pays significantly less in taxes than the average cost of the service it receives. Table 9 compares the total taxes and fees generated by LaGuardia flights with the cost of the airport and airways services at the airport. These estimates assume that each of the five representative flights operates with the same frequency as the flights in its representative size class. For example, there are 18,912 operations of flights with between 50 and 120 seats, and these calculations assume there are an identical number of B-717 operations. Table 9 Comparision of Revenues and Costs of Serving LaGuardia Domestic Passenger Flights, 2004 Aircraft

A-320 B-717 ERJ-145 ERJ-135 Twin Otter TOTAL

Current Revenues ($)

Average Cost ($)

Flights

Revenues from Taxes and Fees ($)

Cost of Services ($)

2,178 1,647 594 432 170

929 937 796 766 721

174�312 18�912 63�816 97�104 8�736

379�678 31�156 37�914 41�918 1�485 492�150

161�912 17�727 50�800 74�344 6�299 311�082

Note: The number of flights is annualized from operations in May. Source: T-100 and Table 8.

54

Assuming aircraft at the other airport pay a lower airfield use fee does not have a substantial effect on the differential. If the Twin Otter operate at an airport with the same airfield use fee as the A-320 its taxes and fees would increase by $8. The taxes and fees of the EMB-145 would increase by $28.

DANIEL P. KAPLAN

84

Payments by Commercial passenger carriers at LaGuardia substantially exceed the cost of the airport and airways services they consume. The estimated discrepancy is entirely due to the aviation taxes, because the landing fees cover the airfield costs. The observed discrepancy may be partly the result of relatively high fares and correspond­ ingly high-ticket tax revenues at LaGuardia.55 The disparity, however, also reflects the disproportionate share of the ATC costs paid by passenger carriers and their passengers.

6.3 Imposing Cost-Based Fees Consider a case in which it is assumed LaGuardia is uniformly congested through its operating hours. As shown above, smaller aircraft pay well below the average cost of the services they receive, while the larger aircraft pay well above. Under the proposed system, each flight would pay the sum of a minimum airfield use fee consistent with some acceptable level of congestion and, if needed, a weight-based airfield fee assessed on all flights to cover any revenue shortfall. As Table 10 shows, while the Twin Otter would face a substantial increase in landing fees, the increase to other aircraft would be more moderate. Given the pent-up demand for service at LaGuardia, the imposition of a flat rate airport use fee may not be sufficient to solve the congestion problem. At other airports, however, shifting to a weight-based movement fee during congested periods would likely have a significant salutary effect. Table 11 shows the total taxes and fees paid by the LaGuardia flights if the fees for both airfield use and terminal operations were set at average costs. This analysis assumes the other airports are not congested, and they charge weight-based airfield use fees. It is also assumed the flights pay average variable cost both for terminal operations at the other airport as well as for en route control. The fixed costs of these two components

Table 10 Impact of Instituting Average Cost Airfield Use Fees LaGuardia Airport, 2004 Aircraft

A-320 B-717 ERJ-145 ERJ-135 Twin Otter

Average Cost Fee ($)

249 249 249 249 249

Existing Fee ($)

Change in Fee ($)

Flight Revenues ($)

386 270 113 108 33

−137 −21 136 141 216

16�956 13�170 4�780 3�112 1�338

Change as Per cent of Flight Revenues (%) −0.8 −0.2 2.8 4.5 16.1

Note: Existing fees are one-half of existing landing fees. Source: Author’s calculations.

55

If the same relationship between distance and yields existed at LaGuardia as it does at other major cities, the LaGuardia flights would have generated 15.7 per cent less revenue. This analysis is based on markets with nonstop service, and like the LaGuardia fares does not reflect the lower yields of connecting passengers.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

85

Table 11 Impact of New Fees on LaGuardia Flights, 2004 Aircraft

A-320 B-717 ERJ-145 ERJ-135 Twin Otter

New Fees ($)

Existing Fees ($)

Difference ($)

929 937 796 766 721

2,178 1,647 594 432 170

−1,249 −710 202 334 551

Flight Revenues ($)

16�956 13�170 4�780 3�112 1�338

Change as Per cent of Flight Revenues (%) −7.4 −5.4 4.2 10.7 41.2

Source: Author’s calculations.

of ATC services, however, are value priced.56 Under this scenario, and assuming fares remained the same, the taxes and fees paid by the A-320 would decline by $1,249, while the taxes and fees paid by the 50-seat ERJ-145 would increase by $202. The increases would be substantially greater for the smaller aircraft.57 These cost estimates assume no congestion at the other airport. If there were congestion at those airports, however, the impact on the smaller aircraft would be somewhat greater. Aircraft with fewer than 40 seats account for 29 per cent of LaGuardia operations. The increase in taxes and fees for the 37-seat ERJ-135 accounts for more than 10 per cent of the revenues of those flights. The average increase for the ERJ-145 is 4 per cent, which of course means a greater increase for half of those flights. The impact on profitability would be substantially greater. If the operating margin on the average ERJ-145 flight were 15 per cent, then the profitability of the average flight would fall by more than a quarter, assuming fares remained constant. The effect of an increase in costs on flight profitability also depends on the impact on revenues. Fares increasing in line with the higher costs would reduce the impact of the new pricing system on a flight’s profitability. The price impact on revenue – and profitability – of the new set of prices depends on the forces of supply and demand, which are market specific. Compare, for example, two markets served with regional equipment. In a market served by a single carrier, the increase in the price of using the airport and airways system will be distributed between the carrier and its passengers based on the elasticities of supply and demand. If the elasticity of demand in the market were relatively low, the new price system would have a relatively small effect on flight profitability, because passengers would bear most of any cost increase. The effect would be greatly magnified, if the regional equipment service were provided in competition with another carrier’s mainline service. In contrast to flights provided in

56

Using value prices for the fixed component is part of a transition process. Because terminal and en route control operations are likely to exhibit constant returns to scale, value pricing would not be efficient in the long run. 57 In practice, the minimum fee might vary over time, and the impact of the new system may be greater on some flights during those periods. This would most likely be the case when the peak demand on a route fails to match the peak demand for the airport overall.

86

DANIEL P. KAPLAN

regional equipment, the switch to cost-based prices would actually reduce the costs of operating mainline equipment on a route. The lower costs could put downward pressure on fares in the market, in which case the flight using regional equipment would be hard pressed to recover of any of the added cost from the new tax and fee system. If the carrier offering the regional service needed to reduce its fares to remain competitive, it would be effectively absorbing more than 100 per cent of the higher fee.

7 CONCLUSION The current system for pricing the airport and airways system certainly contributes to the strains associated with its operation. The system subsidizes low-valued operations when congestion is a problem and suppresses use of price to signal the value of capacity additions. Indeed, a cursory review of the existing set of prices reveals a complex web of subsidies and cross-subsidies, which impose substantial costs but have no clear public policy objective. The present system should be replaced. This chapter has advocated requiring all flights to pay at least the marginal cost of the service it receives, with that minimum price increasing as warranted by congestion. In no case, however, would the prices be permitted to generate revenues in excess of the cost of the service. On the other hand, any revenue shortfall would be recovered through a weight-based fee. The new fee system, therefore, encourages efficient operation while maintaining an element of value pricing, a cornerstone of the existing prices for use of the airport and airways system. While this proposal represents a significant improvement over the current system, a number of factors complicate its adoption. While this analysis has focused on the subsidies presently received by regional aircraft, the existing system provides even greater subsidies to general aviation jet operations. Moreover, a switch to the new system could both put upward pressure on small community airfares as well as discouraging some service. This would likely generate significant political opposition. Because the airport and airways system is both operated and regulated by government agencies decisions concerning pricing and output decisions are fundamentally political ones. A move to a more rational pricing system might therefore require compromise. Crafting a political solution would involve a continuation of at least some of the existing subsidies. Continuing those subsidies in their current form, however, would be counter­ productive. Most notably, there should be a concerted effort to limit subsidies during periods of congestion. Moreover, any subsidies that are provided should be carefully targeted, and they should be discounts off the normal prices.

BIBLIOGRAPHY Areeda, Phillip and Donald Turner. 1975. Predatory prices and related practices under section 2 of the Sherman Act. Harvard Law Review 86, pp. 697–733. Brueckner, Jan K, 2002, “Airport Congestion When Carriers Have Market Power,” The American Economic Review 92, 1357–1375.

TOWARD RATIONAL PRICING OF THE US AIRPORT AND AIRWAYS SYSTEM

87

Federal Aviation Administration, Air Traffic Organization: Airports Data for Stakeholders, November 15, 2005a. Federal Aviation Administration, Air Traffic Organization: Data Package for Stakeholders, November 15, 2005b. General Accounting Office, Airport and Airway Trust Fund, GAO-03-979, September 2003. Kahn, Alfred E. 1970. The Economics of Regulation, Volume 1, Wiley, New York. Levine, Michael E. 1969 Landing fees and the airport congestion problem, Journal of Law and Economics XII, 79–108. May, James C. Smart – and Fair – Skies: A Blueprint for the Future, Speech to the International Aviation Club, Washington, DC, April 18, 2006. Mayer, Christopher and Todd Sinai. 2003 Network effects, congestion externalities, and air traffic delays: Or why all delays are not evil, The American Economic Review 93, 1194–1215. Poole, Jr. and Robert W. Outsourced Flight Service Stations Save FAA $2.2 Billion, Reason Foundation Commentary, September 1, 2005. Transportation Research Board, 1999. Entry and Competition in the U.S. Airline Industry: Issues and Opportunities, National Academy Press, Washington, DC. Viscusi, W. Kip, John M. Vernon, and John E. Harrington, Jr., 1995, Economics of Regulation and Antitrust, 2nd Edition, MIT Press, Cambridge, MA.

This page intentionally left blank

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

4 An Interpretative Survey of Analytical Models of Airport Pricing∗ Leonardo J. Basso† and Anming Zhang‡

ABSTRACT In this chapter, we review analytical models of airport pricing, from 1987 onward. We argue that articles in the literature can be grouped into two approaches, the traditional approach and the vertical structure approach. In the traditional approach, the demand for airports depends on airport charges and on congestion costs of both passengers and airlines; the airline market is not formally modeled under the assumption that airline competition is perfect. In the vertical structure approach, airports are recognized as providing an input for the airline market, which is modeled as an oligopoly where firms have market power. It is the equilibrium of this downstream market that determines the airports’ demand: the demand for airports is thus a derived demand. We present and discuss both approaches and the papers within each of them, highlighting how they have analyzed different aspects of airport pricing such as the efficiency of weight-based airport charges, the effects of concession revenues on pricing and capacity investments, or the effects of airlines’ market power on optimal runway congestion pricing. We study the connection between the approaches and the transferability of results, and also discuss a handful of articles that have looked at the pricing of airport networks, i.e., three or more connected airports, as opposed to airports in isolation. We conclude by providing what we think should be the lines of future research.

∗ Acknowledgement: We would like to thank Darin Lee and Monica Hartmann for helpful comments. Finan­ cial support from the Social Science and Humanities Research Council of Canada (SSHRC) is gratefully acknowledged. † Corresponding author. Sauder School of Business, The University of British Columbia. Department of Civil Engineering, Universidad de Chile. Contact information: 2053 Main Mall, Vancouver BC, Canada V6T 1Z2, Tel.: 1-604 822 0288, Fax: 1-604 822 9574, [email protected]/Casilla 228-3, Santiago, Chile, Tel.: 56-2 978 4380, Fax: 56-2 689 4206, [email protected]. ‡ Sauder School of Business, The University of British Columbia. Contact information: 2053 Main Mall, Vancouver BC, Canada V6T 1Z2, Tel.: 1-604 822 8420, Fax: 1-604 822 9574, [email protected].

90

LEONARDO J. BASSO AND ANMING ZHANG

1 INTRODUCTION Airport pricing has attracted the attention of economists for some time now, starting with Levine (1969) and Carlin and Park (1970). Most of the attention has been devoted to the efficiency of pricing practices by airport authorities and the need to take into account congestion which, even in the early 1970s, was afflicting passengers and airlines. The alleged inefficiencies of actual pricing practices plus the increase in delays at airports around the world have made understanding the models of airport pricing very germane in today’s world. Airport delays in the United States have grown dramatically in recent years. In 2004, 20 per cent of flights arrived more than 15 min late, with Chicago’s O’Hare airport being last with 30 per cent. The US Department of Transportation in its “National Strategy to Reduce Congestion on America’s Transportation Network” (2006) has estimated that aircraft delays cost passengers $9.4 billion. Congestion is perhaps even more acute at some of the major European, Japanese, and Chinese airports. Furthermore, the recent trend of airport privatization and/or commercialization induced, in addition, a focus on the effects of privatization and the efficiency of different regulatory schemes (this trend started in the late 1980s throughout the world following the examples in the United Kingdom). More specifically, privatized airports would pur­ sue maximization of profits. On the other hand, it has usually been accepted that airports enjoy a local monopoly position because they have a captive market. Besides, sizeable economies of scale on airport infrastructure provision and airport operations may exist (Doganis, 1992). Out of the concern that private airports would exert market power in user charges, many private (and public) airports are under some type of economic regulation such as rate-of-return or price caps. The work on airport pricing has been considerable. Some old questions, such as how we should use the price mechanism to signal congestion problems, have persisted in the literature. New questions, such as whether privatization would induce better capacity investment, have appeared. As far as we know, there has been no paper that is devoted to putting together all the questions and answers that have been obtained in the literature since the late 1980s. We attempt to do that in this chapter. Specifically, we review the airport pricing literature, with a focus on analytical papers. Indeed, we are narrowing the scope of our work, by leaving aside a number of important empirical papers. By this we do not mean that the empirical work is irrelevant, but as it will be seen, a comprehensive survey of the analytics of airport pricing easily use up the space in a paper, and we believe that a good command of theoretical and analytic results helps to better grasp empirical findings. Also, we will focus on papers in the last 20 years. We believe that this is enough to understand what is known today about the theory of airport pricing, since earlier contributions such as Levine (1969), Carlin and Park (1970), and Morrison (1983) have been incorporated into the papers we will review. While there are many survey papers on airlines, the survey work on airports is relatively rare. One exception is a recent survey paper by Forsyth (2000), in which he focused mainly on the pre-1990 airport-pricing papers and on models of airport costs and production efficiency. We shall summarize the findings and provide directions of what we think should be future research. In order to do this in an orderly manner, we group the papers into two broad “approaches”. Papers within one approach share many features regarding the analytical modeling, which makes it easier to explain what characterize them, while

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

91

also enabling a better description of the contributions of each of the individual articles. Therefore, Sections 2 and 3 will be devoted to explain what we have called “the tra­ ditional approach” and “the vertical structure approach” to airport pricing respectively and, within each approach, what we have learned from individual articles. Because an obvious question is whether results from one approach can be transferred to the other, in Section 4 we discuss connections between the approaches as a means to better understand how the results stemming from the two approaches relate to each other. In what follows, Sections 2 and 3 deal with a single airport’s decision or, at most, two (complementary) airports, given the complexity of the economics of airport pricing. However, there is a handful of articles that have looked at the pricing of airport networks, i.e., three or more connected airports. We discuss these papers in Section 5, noting that the previous classification may still be applied. We conclude in Section 6 by providing what we think should be the lines of future research.

2 THE TRADITIONAL APPROACH TO AIRPORT PRICING The main characteristic of the traditional approach is that it typically follows a “partial equilibrium” analysis in which an airport’s demand is directly a function of the airport’s own decisions. As will be explained below, since airlines’ decisions (and airline com­ petition) are not directly considered, the derived characteristic of the airport’s demand is not formally recognized. In this section, we consider papers by Morrison (1987), Morrison and Winston (1989), Oum and Zhang (1990), Zhang and Zhang (1997, 2001, 2003), Carlsson (2003), Oum et al. (2004), Lu and Pagliari (2004), and Czerny (2006). Most of these papers follow essentially the same model: the demand for an airport is assumed to be a function of a “full price”. This full price includes the airport charge and, in an additive fashion, some cost measure of the delays caused by congestion. Delay functions have always been measured through some non-linear function of traffic and capacity, although the modeling has not been unique: the main discrepancy has been whether the function should (or not) be homogenous of degree one in the traffic to capacity ratio. Delay is assumed to affect both airlines and passengers, and consumers’ surplus is measured by integration of the airport’s demand. When the airport capacity is variable, the cost function has been usually assumed to be separable in operating and capacity costs. This approach has been used to analyze a variety of issues regarding airport pricing and capacity decisions and under many different sets of assumption, as can be seen in Table 1. Initially, the focus was on deriving optimal prices and capacities in the presence of congestion but, lately, it has been used to assess the effects of privatization and regulation as well. The basics of the traditional approach may be synthesized in a fairly concise analytical manner, which we present below.1 In order to provide aviation services, an airport

1 Certainly, not all the papers can directly be assimilated to this presentation – particularly Lu and Pagliari (2004) and Czerny (2006) may seem more distant – but most of them fit through some adjustments, which will be indicated where relevant.

Table 1 Summary of Papers Using the Traditional Approach (from 1987 on) Author

Goal of the Paper

Objective Functions

Capacity

Delay

Observations

Morrison (1987)

Uncover the importance regulators give to each type of aircraft when they max SW

Max SW st BC

Fixed

NHDO

Many periods with independent demands

Morrison and Winston (1989)

Efficient pricing and capacity with congestion

Max SW

Variable and continuous

HDO

Many periods with independent demands

Oum and Zhang (1990)

Analyze budget adequacy under congestion pricing when capacity investments are lumpy

Max SW

Variable and lumpy

NHDO

Many periods with dependent demands

Zhang and Zhang (1997)

Effects of concessions. Should the BC be common to both concessions and airside activities or separate?

Max SW st BC

Variable and continuous

NHDO

Many periods, independent demands. First model to formally incorporating concessions

Zhang and Zhang (2001)

Analyze whether public airport should have a strict (short run) brake-even constraint or a longer run constraint

Max SW st BC

Lumpy

NHDO

Many periods, independent demands

Carlsson (2003)

Efficient pricing and capacity with congestion and emissions

Max SW

Variable and continuous

NHDO

One period. Social cost of emissions added to SW

Zhang and Zhang (2003)

Analyze privatization and the effects of concessions on pricing and capacities

Max SW Max profits (private case) Max SW st BC

Variable and lumpy

NHDO

One period. They include concession operations. BC is in the long run

Lu and Pagliari (2004)

Regulation and concessions: single-till versus dual-till cap

Max profits st two different forms of regulation

Fixed

No delays

Rather than having delays, they assumed that capacity is a restriction on feasible output: potential for excess demand

Oum, et al. (2004)

Efficiency implications of alternative forms of regulation

Max SW st BC Max profits (private case) Max profits st four different forms of regulation

Variable and continuous

NHDO

One period. They include concession operations. BC is in the long run

Czerny (2006)

Effects of concessions on aeronautical charges. Regulation: single-till versus dual-till cap

Max SW Max Profits st two different forms of regulation

Fixed but large: no excess demand

No delays

Both airside and concession charges determine the number of consumers.

SW: social welfare; BC: budget constraint; NHDO: the delay function is non-homogenous of degree one in the traffic to capacity ratio; HDO: the delay function is homogenous of degree one in the traffic to capacity ratio.

LEONARDO J. BASSO AND ANMING ZHANG

94

incurs both operating and capital expenses. It collects user charges to cover these costs and, in the private-airport case, to make a return on capital investments. For a given capacity, congestion will start to build up at the airport as demand grows, inducing delays and therefore extra costs on passengers and airlines. It is usually assumed that airlines fully pass airport charges to passengers; the same is assumed for airlines delay costs.2 Therefore, passengers will perceive a full price consisting of the airport charge, the flight delay cost, travel-time costs plus other airline charges (e.g., air ticket). It has been argued that since other airline charges are exogenous as far as the airport is concerned, the demand an airport faces may be considered to be a function only of the airport charge P and the flight delay cost D, which includes the delay costs to both airlines and passengers. The variables in the model would be3 Q��� Demand for airport facilities measured by the number of flights, which is a function of the full price � perceived by passengers � = P + D the full price that determines the airport’s demand P airport charge per flight D = D�Q� K� flight delay cost experienced by each flight, which depends on traffic Q and airport capacity K K capacity of the airport C�Q� operating costs of the airport r cost of capital. The capacity may be lumpy or continuously adjustable. The assumption of adjustable capacity has been justified based on the observation that capacity would be defined not only by the number of runways – which can only be increased discretely – but also by air traffic control technology, air navigation systems and other infrastructures, which can be increased or enhanced continuously. One of the first issues that was analyzed using the traditional approach is the nature of the airport’s choices of user charge P and capacity K, for the benchmark case in which social welfare is maximized subject to a budget constraint – the public airport case. The problem the public airport faces is given by �� max P�K

Q���d� + PQ − C�Q� − rK

(1)

�

s�t�PQ − C�Q� − rK = 0

2

(2)

Morrison (1987, p. 48) makes this assumption by equating the airlines’ elasticity of demand for airport services to the elasticity of passengers’ demand with respect to full price times the fraction that airport charges and congestion costs represent in total flight costs (see also Raffarin, 2004, p. 115). Oum et al. (2004) make this assumption explicitly, arguing that this will be the case under perfect competition. 3 Here, for notational simplicity, we present a model with no intraday variations in demand, i.e., a “single period” model. The model can be extended in a straightforward fashion to the case of many periods so long as the demands in these periods are independent. The independence assumption has been made in most papers that deal with multiple periods; the only exception is Oum and Zhang (1990).

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

95

The first term in the objective function would correspond to consumer surplus, the remaining terms are the airport’s profit. Forming the Lagrangean and taking derivatives with respect to P and K, first-order conditions are obtained. From them, the following pricing and capacity investment rules follow: �D � ��� P = C� + Q + (3) �Q 1 + � � −Q

�D =r �K

(4)

where � denotes the Lagrange multiplier of the budget constraint, and � is the (positive) elasticity of demand with respect to the full price. According to Morrison (1987) and Zhang and Zhang (1997, 2001), the interpretation of the pricing rule is as follows: The first two terms on the right-hand side (RHS) of Equation (3) represent the social marginal cost (SMC) of one flight (operational marginal cost plus the marginal cost of congestion), whereas the third term represents a markup that is inversely related to � and depends on the severity of the budget constraint. Hence, the difference with the usual Ramsey-Boiteux pricing is that the pricing rule needs to take into account the congestion that a new flight imposes on others. Regarding the optimal capacity rule – Equation (4) – Zhang and Zhang (1997) note that it does not depend on � and hence it is identical to the one obtained when a budget constraint is not imposed, as in Morrison and Winston (1989). Therefore, airport authorities that adopt Ramsey pricing should still pursue the same optimal policy of capacity investment. In this policy, the socially optimal level of capacity is set such that the marginal benefit of capacity in terms of reduction in delays, equates the marginal cost of capacity (Morrison and Winston, 1989; Zhang and Zhang, 1997). This concludes the explanation of the basic setup of the approach. In what follows then, we will look at how this approach has been used – and modified when needed – to analyze issues other than second best pricing and capacity investment. Authors have used the traditional approach to (i) study the efficiency of weight-based airport charges, (ii) analyze the effects that lumpy investments in capacity may have on bud­ get adequacy, (iii) examine the effects of concession revenues on pricing and capacity investments, (iv) derive efficiency implications of alternative forms of regulation, and (v) study how environmental costs could be incorporated into airports’ charges.

2.1 On Weight-Based Airport Charges Because in general, aircraft are not charged by the contribution they make to congestion but by their weight, Morrison (1987) wanted to uncover the importance regulators give to each type of aircraft when choosing the runway landing fees. For this, he assumed that the demand is Qi , where i denotes a class of airport users, that is, different types of aircraft. Then, assuming that capacity is fixed, he put weights on the contribution of each class of users to the social-welfare function. Hence, the objective function (1) becomes � i

�� �i �

Qi ��i �d�i + Pi Qi − C�Qi �

(5)

96

LEONARDO J. BASSO AND ANMING ZHANG

where �i is the weight of user i. With this social-welfare function, and still considering the budget constraint in Equation (2), the optimal pricing rule Equation (3) changes to � � �D � + 1 − �i �i � P i = C + Qi (6) + �Qi 1+� �i Morrison then asked the following question: what set of weights is implied by actual airport charges? To uncover the weights, he followed Ross (1984) and solved for the weights �i in Equation (6). Using actual data, those weights can be obtained up to a multiplicative constant. Morrison’s main result was that when the airport is uncongested, weight-based landing fees imply welfare weights (the �i ) that are very similar. But when congestion increases, the dispersion in the weights also increases, implying that the weight-based landing fees would be less appropriate when there is congestion. He argued that this happens because, though weight is a reasonable proxy for elasticity of demand, it is a poor proxy for congestion costs.

2.2 Lumpy Capacity and Cost Recovery Oum and Zhang (1990) and Zhang and Zhang (2001) were interested in how budget adequacy would be affected if capacity can be increased only in discrete lumps. The conjecture was that the lumpy nature of capacity expansions would make social marginal congestion pricing lead to alternating periods of airport surplus and deficits. Oum and Zhang (1990) incorporated a positive time trend to the airport’s demand to capture the fact that the aviation demand would increase with the overall economy. By considering lumpy capacity expansions – that is, K can be increased only by a minimum amount �K – they focused on the timing of capacity expansions rather than the capacity investment in a steady state as discussed above (budget constraint was not considered, however). They concluded that, when capacity is indivisible, the optimal congestion pricing – given by Equation (3) with � = 0 – and optimal capacity expansion would lead to alternating periods of excess capacity and capacity shortage. During capacity shortage, the congestion toll would exceed annualized capacity costs but during excess capacity, the congestion toll would fall short of annualized capacity costs. This implies that budget adequacy would depend entirely on the number of shortage/excess capacity periods between capacity expansions. And the number of periods in each case depends on the pattern of traffic growth. Oum and Zhang (1990) concluded that when capacity is indivisible, the cost recovery status of an airport cannot be predicted without reference to the time path of the traffic growth and, therefore, the cost recovery theorem for investment in transportation infrastructure would not hold. This important theorem states that (see, e.g., Mohring, 1976) when operational costs are separable from capacity costs, the latter exhibit constant returns to scale, and the delay function is homogenous of degree one in the traffic to capacity ratio, optimal congestion pricing and capacity provision leads to exact cost recovery of capacity investments and operational costs. This is not the only way in which the cost recovery theorem would fail for airports though. Even if capacity is divisible, as in the basic model shown in Equations (1) and (2), Zhang and Zhang (1997) showed that, without a budget constraint, social-marginal-cost pricing would always give rise to

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

97

a financial deficit to the airport because the delay function D would not be homogenous of degree one in the traffic to capacity ratio (Lave and De Salvo, 1968; US Federal Aviation Administration, 1969; see Horonjeff and McKelvey, 1983). Furthermore, the deficit would increase with congested at the airport. Given all this, Zhang and Zhang (2001) were interested in the case where delays are non-homogenous of degree one, capacity is indivisible, traffic grows over time, but airports are required to recover their costs from both operations and capacity investments. The question they asked was – Should public airports be asked to break even in the short run, or in the long run, which may involve taking losses in early years of a capacity investment but surplus in later years? For this, they modified the airport’s problem (1)–(2), so as to consider that the airport would now maximize social welfare over a period of time S, while achieving cost recovery over the entire period. Capacity was assumed to be fixed during the period, owing to its indivisibility. The new long-run problem faced by the public airport is � � �S �� Q��� s�d� + PQ − C�Q� − rK e−rs ds max P�K 0

s�t�

�

�S

(7) �PQ��� s� − C�Q� − rK�e−rs ds = 0

0

Now, the airport’s demand increases with time, that is, �Q/�s > 0, and future revenues are discounted using the cost of capital, r. The short-run problem is as in Equations (1)–(2). Not surprisingly, Zhang and Zhang found that the short-run financial break-even constraint leads to a lower level of social welfare than a long-run break-even constraint. This increase in welfare is expected, since short-run budget adequacy implies long-run budget adequacy but not vice versa. In fact, Zhang and Zhang showed that the two will be equal only when the airport’s demand remains constant over time, that is, �Q/�s = 0. This directly speaks of the importance of the time path of the traffic growth, as pointed out by Oum and Zhang (1990): To maximize social welfare, airports should be allowed to take losses or make profits at different times, seeking cost recovery only in the long run. What is perhaps more interesting in Zhang and Zhang (2001)’s finding is that under the short-term cost recovery, airport charges are high when the demand is low and there is excess capacity. However, when the demand is high, and there is congestion, airport charges would be low. This seems to be undesirable. On the other hand, under the long-term cost recovery, airport charges grow together with the demand.

2.3 Airport Concessions and Pricing Effects: Public and Private Airports Given the increasing pressure on public airports to self-finance their operations, airports have been increasingly depending on revenues generated by non-aeronautical businesses,

LEONARDO J. BASSO AND ANMING ZHANG

98

such as airport parking, in-airport stores and so on.4 The demand for these concession services is complementary to the demand for aeronautical services, in that the more people there are using the airport, the higher the concession revenues. Zhang and Zhang (1997) wanted to analyze what would be the socially optimal balance between aeronautical revenues and concession revenues given the cost recovery constraint, and how the associated pricing practices would look like. For this, they modified the public airport’s problem Equations (1)–(2) by incorporating the fact that concession demand is complementary to aeronautical demand: � � �� �� max Q���d� + PQ − C�Q� − rK + Q X�p�dp + pX − c�X� P�K�p � p (8) s�t�

PQ − C�Q� − rK + Q �pX − c�X�� = 0

In Equation (8), p represents the price for concession goods or non-aeronautical services provided in the airport, X�p� is the demand for concession services per flight, and c�X� are the costs of providing the concession services, which are assumed to feature constant returns to scale. There are two important things to note in the above setup. First, the complementarity between the demands is unidirectional, that is, consumers’ decision to fly or not is based on the full price of the aeronautical service; they do not take into account the price of the concessions in their travel decisions. Only after arrival at the airport, passengers observe concession prices and make purchasing decisions. Second, note that the budget constraint in Equation (8) includes the revenues from both aeronautical and concession services, which effectively enables cross-subsidies between the two services. Without the budget constraint, the (first best) optimal solution obviously involves marginal cost pricing on the concessions side, i.e., p = c� �X�. On the aeronautical side, the social-marginal-cost pricing of Equation (3) would have an additional markdown; this happens because, now, a smaller aeronautical charge increases the demand for both aeronautical services and concessions services. Hence, the optimal aeronautical charge is smaller. This would, however, lead to deficits if the delay function is non-homogenous of degree one in the traffic to capacity ratio, as discussed above. With the budget constraint, and assuming that the delay function is non-homogenous of degree one, Zhang and Zhang showed that at the (second best) optimal solution of problem (8), the price of concession services would be such that p > c� �X�, showing that profits would be made in concession services. Therefore, concession operations would subsidize aeronautical operations. If the airport were not allowed to make profits from its concessions, but was

4

For the last two decades, concession revenues have grown faster than aeronautical revenues; as a result, they have become the main income source for many airports. At medium to large US airports, for instance, commercial business represents 75–80% of the total airport revenue (Doganis 1992). Furthermore, concession revenues have grown faster than aeronautical revenues. For example, in 1979, Hong Kong International Airport generated similar amounts of revenue from its aeronautical and non-aeronautical (mostly concession) operations. In the late 1980s and 1990s, however, its concession revenue accounted for 66–70% of total revenue (Zhang and Zhang, 1997). More importantly, concession operations tend to be more profitable than aeronautical operations (see e.g., Jones et al., 1993; Starkie, 2001).

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

99

still asked to self-finance its operations, then this would obviously lead to a smaller level of social welfare. Furthermore, Zhang and Zhang showed that the cross subsidy from concessions does not in general restore social-marginal-cost pricing on the aeronautical side – Equation (2) – unless the demands and costs fulfill a very particular condition. The attention to concession revenues, however, does not stop at the pricing and costrecovery issues of public airports. It has also been suggested that the complementary nature of the concessions demand would give incentives for private airports to reduce the price they charge for aeronautical services in order to maximize the number of travelers in the airport using the concessions. This may imply that ex ante price regulations may be unnecessary (see, e.g., Condie, 2000; Starkie, 2001). In order to assess whether the argument holds, Zhang and Zhang (2003) and Oum et al. (2004) used Zhang and Zhang (1997)’s model to look at the decisions a private unregulated airport would make. The profit-maximization problem faced by a private unregulated airport is max PQ − C�Q� − rK + Q �pX − c�X�� P�K�p

(9)

Zhang and Zhang (2003) and Oum et al. (2004) found that, while airside private prices diminish as it was conjectured by Condie (2000) and Starkie (2001), they decrease less than the prices in a public airport that also has concessions, and that this is the case for both the first-best pricing (unconstrained public airport) and second best-pricing (budget-constrained public airport). Therefore, concession revenues would not be a valid argument for de-regulation once an airport is privatized. The intuition of the result is simple: a private airport would care about the extra profits it can make from concession activities; a public airport maximizing social welfare, however, would care about concession profits but also about the consumer surplus induced. Consequently, the decrease in the aeronautical charge would be larger in the public case: concession revenues would actually increase the gap between private and public airside charges. As for the effects of privatization on capacity decisions, Oum et al. (2004) obtained that the capacity investment rule of the private airport would be the same as the one a public airport follows, as in Equation (4). Hence, they argued that, if the capacity can be adjusted continuously, the capacity investment decision of the private unregulated airport would be efficient from a social viewpoint. However, since the price and capacity decisions are jointly determined, and the pricing rules of the two airport types are different, so will be the actual levels of traffic and capacity. In fact, since a private airport charges more, its actual capacity would be smaller. But, their main point is that, conditional on traffic level Q, the capacity K determined by Equation (4) would be efficient because marginal benefit equals marginal cost. In line with the actual capacity of private airports being smaller when capacity can be adjusted continuously, Zhang and Zhang (2003) found that, when capacity is indivisible, a private airport would make the (lumpy) addition of capacity later than a public airport. Note that none of these two results imply anything about the level of actual delays, because traffic levels will be different as well. Czerny (2006) also looked at the effects of concession revenues on airside charges. There are two important differences between his and Oum et al.’s model (2004): First, he considers an airport that is non-congestible and has spare capacity, making the reasons for cross-subsidization discussed above vanish. Second, in Oum et al. (2004) the number of actual flyers would depend only on the full price � and not on the price for concession

LEONARDO J. BASSO AND ANMING ZHANG

100

services. The price of concession services would only determine how many of those who are already flying buy concession services. Czerny (2006), however, considered that both airport and concession charges affect the number of flyers, and that the complementarity arises because only people who are actually flying will be able to purchase concession goods. Hence, in Czerny’s setting it may happen that the airport charge is higher than a consumer’s willingness to pay for flying, but that negative payoff is compensated by positive benefits arising from consumption of commercial services. These differences are material. Czerny showed that in this setting, the monopoly charge for aeronautical activities is actually higher with concession revenues than without concession revenues, thus rejecting the conjecture of Condie (2000) and Starkie (2001). The intuition is as follows: when the airport has concession services, and since these influence the number of flyers, the airport may increase its revenues in two ways. It may increase the price for aeronautical services, using a low concessions charge to mitigate the decrease in demand, or it may decrease its aeronautical charge, hoping to make revenues on the concessions side. But since only passengers can buy commercial services, the demand for the latter is a subset of the demand for flights. Therefore, an increase in aeronautical charges increases revenue more than an increase in the concession services charge.

2.4 Efficiency Implications of Alternative Forms of Regulation Traditionally airports have been owned by governments (national or local). Privatization of major airports started in the late 1980s, and airport privatization has now become an important phenomenon around the world.5 Most of the privatized airports have been regulated out of the market-power concern given the monopoly nature of airports. Oum et al. (2004), Lu and Pagliari (2004) and Czerny (2006) analyze the effects of alternative mechanisms of regulation on the performance of private airports, with a particular focus on how revenues from concession services should be dealt with. Oum et al. (2004) have considered four different regulation mechanisms: single-till rate of return (ROR), dual-till ROR, single-till price cap and dual-till price cap. Under the single-till ROR, airport charges (for both airside and concession operations) are set for cost recovery plus a fair return on the invested capital. If u is the allowed ROR, then the new problem the private airport solves is max PQ − C�Q� − rK + Q �pX − c�X�� P�K�p

s�t� PQ − C�Q� + Q �pX − c�X�� = uK

(10)

The well-known problem with ROR is that, if the allowed return is greater than the cost of capital, i.e., u > r, the airport has an incentive to over-invest in capital, a problem known as the Averch–Johnson effect. However, if the regulators get the allowed return 5 In 1987, the British government privatized the BAA, which owned and operated the three London airports (Heathrow, Gatwick, and Stansted), among other airports in the UK. Since then, many airports around the world have been or are in the process of being privatized. The majority stakes of Copenhagen airport, Vienna airport, Rome’s Leonardo Da Vinci Airport and 49% of Schiphol airport have been sold to private sector owners. Many other European airports are in the process of being privatized. Major airports in Australia and New Zealand have been privatized as well. As a way to partially privatize airports, six Chinese airport companies including seven airports have been listed on stock exchanges.

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

101

right, the problem vanishes. It has been argued though that, even if the allowed return is chosen correctly, the single-till ROR would still misplace the incentives on terms of the productive efficiency, because it is essentially a cost-based mechanism. While the argument is sensible and has been detected empirically in several industries, it does not flow analytically from model (10). Under the dual-till ROR, the allowed return applies only to aeronautical operations. If the regulators get the allowed rate right, the new restriction is PQ − C�Q� = rK. In this case, the airport would make no profits in airside operations and, therefore, would try to maximize Q�pX − c�X��, the profits coming from concession operations. Given the complementary nature of the concessions demand, the airport will, in fact, try to maximize traffic, which is equivalent to minimize the full price �. Hence, this regulation mechanism would lead to a capacity rule as in the public case, that is Equation (4), and to average cost pricing, that is P = �C�Q� + rK�/Q. Note, however, that if u > r, the Averch–Johnson effect re-appears. We now turn to the price-cap regulation, a mechanism in which the regulator sets a ceiling for the aeronautical charge, that is P ≤ P ∗ . Theoretically, the cap is set to limit the airport’s market power, while ensuring its financial viability (this may include a fair rate of return on capital investment). The difference between the single-till and dual-till price-cap regulations is, again, related to whether concession revenues will be lumped together with airside revenues or not; to be perfectly clear, the debate is not about regulating concession activities. Under the single-till price-cap regulation, the cap P ∗ will be set considering that the airport will likely make profits from concession activities. This would imply, according to Oum et al. (2004), a cross-subsidy, just as in the case of a public airport subject to budget constraint (Zhang and Zhang, 1997). However, a problem is that the more profit the airport makes from concessions, the smaller the allowed aeronautical charge would be in future revisions of the cap, even if traffic grows and congestion builds. Because of this, the single-till cap regulation for the case of congested airports has been criticized (e.g., Starkie, 2001): the airport charge would not be a useful signal to users regarding congestion. Moreover, Oum et al. (2004) also showed that a price cap (either single-till or dual-till) induces underinvestment in capacity, worsening the problem. Here, the airport is unable to recoup fully from its investment in capacity – which reduces congestion and hence increases the users’ willingness to pay – because the price is capped. Under the price-cap regulation, therefore, while the market-power distortion is alleviated, the service-quality provision is sub-optimal, suggesting an interesting trade-off between with and without the regulation. This result is in fact very robust. Spence (1975) showed that if a monopolist who initially can choose both price and quality of its product is constrained to charge below some price ceiling, the quality it chooses will be always below what is socially optimal for that price. It is noted that under the dual-till price cap, that is, when concessions revenues are not considered in establishing the cap, Oum et al. (2004) showed that the cap would not be set as low as in the single-till, something that seems desirable. Hence, overall, Oum et al. (2004) concluded that the presence of the concession rev­ enues make the dual-till ROR approach a quite interesting mechanism as it would induce the airport to invest optimally in capacity, while minimizing its costs and congestions delays, since it would try to minimize the full-price. Indeed, Spence (1975) suggested that ROR has nice properties when regulating both quantity and quality.

102

LEONARDO J. BASSO AND ANMING ZHANG

Like Oum et al. (2004) and Lu and Pagliari (2004) have looked at the effects of singletill and dual-till price cap regulations. They used a social-welfare function as the one maximized in Equation (1), but considered that more traffic caused no congestion, that is D = 0. The difference is that in a model with a delay function being non-homogenous of degree one, congestion is essentially a cost. And given that the cost increases more pronouncedly as the traffic gets closer to capacity, equilibrium levels of traffic would never surpass capacity (e.g., when D�Q� K� = Q�K�K − Q��−1 , delays approach infinity when output approaches capacity). In Lu and Pagliari’s case, however, if the aeronautical charge is too low, demand may well exceed capacity, particularly because in their model, capacity is assumed to be fixed. Lu and Pagliari found that a single-till price cap would be appropriate when the average cost of the airport is greater than the market clearing price (for the given capacity), because cross-subsidies from concession revenues would be needed to reduce the airside charge and restore full capacity use. In other cases, however, they found that a dual-till price cap would be better: under the single-till the price cap may be set “too low,” owing to the cross subsidy from concessions, and hence dead-weight losses would occur because of excess demand. Czerny (2006) also compared the single-till and dual-till price-cap regulations. As discussed previously, he examined an airport that is non-congestible and has spare capacity, and considered that both airport and concession charges affect the number of flyers. Under these conditions, he found that the single-till dominates the dual-till in terms of social welfare, a result similar to what Lu and Pagliari found when the airport does not suffer from excess demand. The intuition is that with the single-till price cap, the regulator has better control of the overall profits of the airport, which is not the case with the dual-till regulation. Thus, the single-till helps to limit market power. Hence, overall, when the airports are not congested, a single-till price cap seems like a reasonable approach to control market power. However if congestion actually occurs, the single-till would induce incorrect signals regarding congestion, while the dual-till would distort capacity investments. Furthermore, if there are delays as traffic levels approach capacity (as in the original setup), the socially optimal pricing structure would require cross-subsidization (Zhang and Zhang, 1997), but this is precluded in the dualtill. Hence, in congested airports, the dual-till ROR regulation may be a better option: the incentives for capacity investments would be well placed, while the regulated airport would pursue average cost pricing.

2.5 Airport Pricing Considering Environmental Costs Carlsson (2003) developed a model of airport pricing that, in addition to congestion, also includes environmental damages (noise, emissions).6 For this, he modified the social-welfare function in Equation (1) to include environmental costs, as follows:

6 Air travel is considered a rapidly growing source of greenhouse gases (GHGs), something that has sparked concern. The problem is that, while airport delays result in aircraft’s holding/circling in the air waiting for landing and hence cost to airlines, the circling also burns extra fuel increasing GHG emissions. Furthermore, the possibility of being held up induce airlines to carry extra fuel in their aircrafts, which increases the aircraft’s weight and, consequently, its consumption of fuel and GHG emissions (see Economist, 2006).

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

�� max P

Q���d� + PQ − C�Q� − rK − QE �D�Q� K��

103

(11)

�

where E is the average environmental cost per flight. It depends on the level of conges­ tion because, for example, the delay increases fuel consumption and hence emissions. Carlsson considered many periods throughout the day and allowed the environmental costs to vary according to the type of aircraft. For simplicity we do not do so here; the intuition of the results remains unchanged. The optimal pricing obtained has two more terms than the congestion-only social-marginal-cost pricing in Equation (3) when � = 0. He gets P = C� + Q

�D �E �D +E+Q �Q �D �Q

(12)

The last two terms in the RHS of (12) represent the marginal environmental cost: In addition to the airport’s marginal cost and the marginal cost of congestion, each aircraft would have to pay the environmental cost it produces, plus another sum owing to the fact that the extra delay a new flight imposes on existing flights, increases the average environmental cost of all flights.7 These last two terms are obviously positive, which shows that, when environmental costs are considered, the airside charge is higher. Carlsson then pointed out that, if the proceeds from the environmental charge accrue to the airport, then cost recovery may be feasible. Whether this is the case or not, however is an empirical matter, as it depends heavily on the shapes of the delay function and the average environmental cost. As for the capacity decision, although Carlsson did not look into it, it is fairly evident the direction in which it would change with the added environmental costs. Since now more capacity is beneficial not only because smaller delays decrease the full price, but also because smaller delays reduce average environmental costs, the socially efficient capacity investment rule would induce a larger investment in capacity.

3 THE VERTICAL STRUCTURE APPROACH TO AIRPORT PRICING The vertical structure approach is newer and, hence, there are fewer papers. Here we review Brueckner (2002), Pels and Verhoef (2004), Raffarin (2004), Basso (2005) and Zhang and Zhang (2006). In this approach, the airline market is formally modeled as an oligopoly, which takes airport charges and congestion taxes as given. Airports, however, are not always considered integrally; in some cases, only airport authorities, who need to

7

The optimal charge is differentiated between types of aircraft and times of the day when these are differentiated.

104

LEONARDO J. BASSO AND ANMING ZHANG

set a tax to be paid in addition to the airport charge – implicitly assumed to be marginal cost – are considered. In these cases, airport profits do not enter the social-welfare function. This airport–airlines approach to airport pricing was driven by the policy need to respond to an increasing level of delays at hubs throughout the world. An important characteristic of hub airports is that usually a few major airlines dominate the airports: these are not atomistic carriers and hence they are not price takers. The focus of the approach has mainly been on the characterization of optimal (public) runway pricing under congestion and airline market power, as can be see from Table 2. Thus, the idea has been to highlight the differences between the airport congestion pricing and the road congestion pricing, where decision makers (individual drivers) are atomistic. Until most recently, capacity was assumed to be fixed, and hence was not a decision variable of the airport or the airport authority, in the vertical structure approach. Brueckner (2002) should undoubtedly be credited for starting this stream of literature. In this very influential paper, he considers N airlines that are seen as homogenous by consumers and that compete in a Cournot fashion. He allows for peak and off-peak demands, which are interrelated, and where the peak period consists of a set of relatively short time intervals containing the daily most desirable travel times. Only the peak is congested. From this setup, it would seem that the peak and off-peak travel are vertically differentiated in that, other considerations such as income and congestion levels being absent, consumers would prefer traveling in the peak period to traveling in the off-peak period. In fact, Brueckner does not directly assume downward sloping demands, but starts with a continuum of consumers who would decide to use the peak or the off-peak periods (or not traveling by air at all) depending on the full prices they face: airfare, plus congestion costs caused by delays at the airports. However, Brueckner also adds a “tendency to fly in business,” which correlates to travel in the peak, as a device that would enable simpler (non-corner) solutions. The problem with this is that it actually imposes that in terms of pure utility, with no income or congestion effect whatsoever, some consumers would prefer traveling in the off-peak period. This seems to contradict the idea of the peak period being “the most desirable travel times.” The airlines, observing the demands and understanding how consumers’ decisions are made, choose their quantities in the output market. An important aspect here is that congestion also affects airlines: There are externalities in production in that, the more a rival produces, the higher a firm’s marginal and average costs will be. The delay function is not necessarily linear in traffic. In equilibrium then, the sorting of consumers towards peak and off-peak occurs through the airlines’ quantity decisions (for given airport charge and capacity). Brueckner then looks at what should be the optimal additional tax that should be charged to airlines in the peak period, in order to adequately account for the congestion externality. Since the off-peak period is assumed to be non-congested, no congestion toll would be needed. Thus, he looks at the regulator case in the sense that the airport is not formally incorporated into the analysis: its profits do not enter the social-welfare function, which is composed of only consumer surplus and airlines’ profits, and there is no consideration of cost recovery, something that has drawn important attention within the traditional approach (see Section 2). Brueckner’s main conclusion – the one that has since driven research in the area – is that with Cournot oligopoly, each airline will internalize the congestion imposed on its flights

Table 2 Summary of Papers Using the Vertical Structure Approach Author

Goal of the Paper

Oligopoly model

Objective Function and airport modeling

Observations

Brueckner (2002)

Optimal tax (additional to airport charges) to account for congestion

N airlines in homogenous Cournot

Max SW = CS + � No formal modeling of the airport, only a regulator

There are peak and off-peak periods (peak-load pricing). Sorting to periods is endogenous through airlines decisions. Only the peak is congested Congestion is a non-linear function of traffic and affects both airlines and passengers

Pels and Verhoef (2004)

Optimal tax (additional to airport charges) to account for congestion and market power

Duopoly in homogenous Cournot

Two airports not formally modeled, only two regulators. Max SW = CS + � Also analyze Individual Max SW

One period (congestion pricing). Delay is a linear function of traffic and affects both airlines and passengers

Raffarin (2004)

Efficient congestion pricing

Differentiated duopoly competing in prices and frequencies

Max SW = CS + � + �− congestion costs. Single airport.

One period (congestion pricing). Three stage game: airport pricing, frequencies, prices. Congestion does not affect airlines nor demand. They are only an external social cost. Delay is a linear function of traffic (Continued )

Table 2 Summary of Papers Using the Vertical Structure Approach—Cont’d Author

Goal of the Paper

Oligopoly model

Objective Function and airport modeling

Observations

Basso (2005)

Effects of ownership on prices and capacity

N airlines in differentiated Cournot

Two airports (round trips) Max SW = CS + � + � Max airports’ profits Max airport–airlines joint profits Max SW st BC Max individual airport profits

One period (congestion pricing). Congestion is a non-linear function of traffic and affects both airlines and passengers. Consumers are also affected by schedule delay cost

Zhang and Zhang (2006)

Optimal pricing to account for congestion and market power when there are N airlines and capacity is variable

N airlines in homogenous Cournot

Max SW = CS + � + � Max airports’ profits Max SW st BC

One period (congestion pricing). Congestion is a non-linear function of traffic affecting only the passengers. The demand function is general

SW: social welfare; CS: consumer surplus; �: airlines’ profits (industry wide); �: airport profits; BC: Budget constraint.

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

107

(and passengers) while ignoring the congestion externality imposed on other airlines’ flights, which enables a limited role for congestion pricing by the airport authority.8 In a “symmetric airlines” case, the optimal toll that should be charged during congested periods is equal to the congestion cost from an extra flight times one minus a carrier’s share. In particular, a monopoly airline would perfectly internalize all the congestion it produces and hence there would be no room for congestion pricing. This shows the difference with the road case: with market power, the degree of internalized congestion is usually sizeable. Pels and Verhoef (2004) attempted to expand Brueckner’s work in two directions: first, they explicitly considered the market power distortion and its effect on the optimal congestion toll. Second, they addressed the issue that, at an origin–destination (OD) pair, the airports may not collaborate to maximize overall social welfare; instead, each airport may maximize a local measure of welfare. Their model is as follows. Consider an OD pair in which the airports decide charges prior to airline competition. The capacities of the airports are assumed to be fixed. In this OD pair, two homogenous and symmetric airlines compete in Cournot fashion, taking airport charges and taxes as given when they choose their quantities (frequencies). Congestion delays affect airlines costs; the delay function is a linear function of total traffic at an airport. Passengers choose airlines based on a generalized cost which is the sum of the air ticket and congestion delay costs, and their demand for air travel is roundtrip-based. The model is solved by backward induction to obtain sub-game perfect equilibrium. Hence, the first step is to solve the airlines’ oligopoly, in order to obtain a sub-game equilibrium which will be parametrically dependent on the congestion tolls charged at each airport. With that sub-game equilibrium at hand, the authors looked for the optimal taxes that should be charged at each airport in order to adequately account for congestion. Initially, they consider that a single authority handles both airports and, consequently, maximizes the sum of consumer surplus and airlines’ profits. Hence, like Brueckner (2002), Pels and Verhoef looked at the regulator case, in that the airports’ profits do not enter the social-welfare function. Their main result indicates that the optimal toll would have two components: a congestion effect (which is positive) and a market power effect (which is negative). The first part is the one identified by Brueckner: since airlines only internalize the congestion they imposed on themselves, the uninternalized congestion should be charged. The second term, which decreases the toll, arises because of the market power at the airline level. What happens is that the regulator, in maximizing social welfare, would need to subsidize the airlines to induce them to produce more. The sign of the optimal toll is therefore undetermined; in particular, when the market-power effect exceeds the congestion effect, a subsidy would be the result. The toll would be positive if the congestion effect dominates. They pointed out, for example, that this would undoubtedly be the case for a monopoly airline.

8

As indicated above, Brueckner obtained the result by developing a model that explicitly recognizes the congestion’s effect on airfares. It is noted that Daniel (1995) first raised the internalization issue and developed a detailed simulation model to analyze carriers’ self-internalization and calculate congestion tolls that exclude the internalized congestion.

108

LEONARDO J. BASSO AND ANMING ZHANG

Pels and Verhoef compared their toll to the pure congestion toll suggested by Brueckner (2002). They found that, when the market-power effect is strong, a pure congestion toll may actually be harmful for social welfare, since airlines are charged with a tax when in fact they should be receiving a subsidy. Brueckner did acknowledge this, though, by stating in his proposition that, “since congestion pricing corrects one distortion but leaves the residual market-power effect in place, tolls are guaranteed to be welfare improving only if that effect is sufficiently small. Otherwise, a negative welfare effect is possible” (p. 1367). Pels and Verhoef argued that, if a negative toll (subsidy) is optimal but unfeasible (for example for political reasons) the regulator should charge a zero toll. As indicated above, Pels and Verhoef also considered the case in which, at each airport, different regulators only maximize consumer surplus of passengers that live in the airport’s region, plus the profits of the home airline. The non-cooperative behavior of airports obviously implies that the result will be inferior to the single-regulator case. In fact, the authors showed, both numerically and analytically, that in the non-cooperation case, tolls at each of the two airports would always be positive. Raffarin (2004), like Brueckner (2002) and Pels and Verhoef (2004), was interested in the optimal airport toll. But rather than considering a two-stage model, she considered a three-stage model. In the first stage, the airport chooses its price. But then, conditional on the airport charge, duopoly airlines sequentially decide frequencies and then prices. The difference with Brueckner and Pels and Verhoef is that, in their case, airlines only decided frequencies; the price is determined in equilibrium by the Cournot assumption. Raffarin, however, has a system of differentiated demands (obtained from a representa­ tive consumer framework) such that an airline’s demand increases when its frequency increases or price decreases, and decreases when its rival’s frequency increases or price decreases. Raffarin’s model has three key assumptions that determine her results: first, she assumes that, even though frequencies are airlines’ decisions, any demand will always be fulfilled. And this is not ensured by the airlines’ choice of aircraft size, k, because k is an exogenous parameter in the model (i.e., equilibrium results will be dependent on k). Hence, there is no real connection between the number of passengers and the number of flights, other than the assumption that there will be enough space. Both Brueckner (2002) and Pels and Verhoef (2004) made a “fixed proportions” assumption, by which the number of passengers in a flight is a fixed constant. This assumption makes it easier, yet transparent, to transform the demand in terms of passengers, into an airport’s demand in terms of flights. The second assumption is that congestion delays – which as in Pels and Verhoef (2004) increase linearly with total traffic – do not affect consumers’ or airlines’ decisions. Instead, congestion costs are subtracted in the social-welfare function, which, interestingly, explicitly includes the airport’s profits. Hence, in this case, airlines do not internalize any of the congestion they cause because it does not directly affect them (it is not a cost to them), and passengers do not care about congestion either. Finally, the third important assumption is that an airline’s operational cost per flight, z, depends on the aircraft size in an increasing fashion, that is dz�k�/dk > 0. Hence, even though using larger aircraft means fewer flights, which saves on costs, each of those flights will be individually more costly. Aircraft size, however, is not a decision variable

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

109

but a parameter. Hence, the implication is that, for given airport charges, equilibrium frequencies increase as the aircraft size diminishes. Rafffarin then maximizes social welfare – which is the sum of the airport’s profits, airlines’ profits and consumer surplus, minus congestion costs – in order to find what the optimal frequencies are, that is, the optimal level of airport’s demand. The optimal airport charge is then obtained as the price that would induce the optimal frequencies. The optimal charge she obtained has three components (which she did not recognize): the airport’s marginal cost, plus the cost of congestion (recall that airlines do not internalize any fraction of congestion in this model), plus a third term. This third term is negative, and could be assimilated to Pels and Verhoef’s market-power effect. The interesting twist, however, is that this term depends on the aircraft size, k, and diminishes the higher k. That is, the airport charge should be larger for smaller aircraft. And since aircraft size and weight are positively correlated, this implies that the airport charge should decrease with the aircraft weight, rather than increase as it is usually the case. The airport would reward airlines that use larger aircrafts because that implies smaller frequencies and hence smaller congestion costs. The choice of k, however, is not endogenous for the airlines in the model. The three papers we have reviewed so far have in common two important features: they all consider maximization of social welfare and in all the three cases, the airport capacities are fixed. In closely related but independent work, Basso (2005) and Zhang and Zhang (2006) generalized these two aspects. Both papers considered that the airport decides on price and capacity in the first stage, and in the second stage N airlines choose quantities (frequencies) in the output market. The airlines have identical cost functions; they are insensitive to congestion costs in Zhang and Zhang (2006) while they do bear extra costs owing to congestion in Basso (2005). Passengers, as usual, are sensitive to the full price of travel, that is, the airline ticket plus congestion delay costs.9 Both used congestion delay functions that are not homogenous of degree one in the traffic to capacity ratio, that is, congestion increases more than linearly with total traffic (for a given level of capacity). Other differences between the two papers are Zhang and Zhang considered that airlines are homogenous in the eyes of the consumers, while Basso allowed them to be horizontally differentiated (in a “non-address” fashion). Basso also considered in the full price perceived by the passengers another time cost, namely, schedule delay cost. This time cost arises because flights do not depart at a consumer’s will but have a schedule. Hence, schedule delay costs are a sort of waiting time, which decreases with higher airline frequencies. On the other hand, Zhang and Zhang considered a general demand function (of the full price) while Basso considered a more restrictive system of demands: linear in the full-prices of airlines. Both Basso (2005) and Zhang and Zhang (2006) solved the airport–airlines game by backward induction, characterizing the shape of the derived demand for the air­ port through comparative statics. Then, they both considered three different objective functions (Basso considered two more which are discussed later): unregulated profit

9 This last point is enough for the internalization of own congestion by an airline to arise in oligopoly, as discussed earlier. It is not needed for both, airlines and consumers, to be sensitive to congestion costs in order to derive the result.

110

LEONARDO J. BASSO AND ANMING ZHANG

maximization, unconstrained social-welfare maximization, and social-welfare maximiza­ tion subject to cost recovery. Let us first discuss the pricing rules they obtained. In the case of unconstrained maximization of social welfare, they both considered a welfare function in which the airport’s profit is included. They found that, in their more general settings, Pels and Verhoef’s insight goes through: the optimal pricing rule is the sum of airport’s marginal cost, plus a congestion effect (positive) and a market-power effect (negative). When capacity is fixed, this pricing rule shows that with large values of N , the congestion effect is large while the market-power effect is weakened. Smaller values of N , on the other hand, imply a weaker congestion effect but a stronger market-power effect. With this pricing rule, the airport manages to obtain a “first best” outcome (subject to the market structure of the airline market which may be of monopoly or oligopoly) in the airline market. Note that, in this setting, rather than a regulator setting the toll, it is the airport that would distort marginal cost pricing to account for both uninternalized congestion and market power. Since the optimal airport charge may be below marginal cost and even below zero, the airport may run a deficit. In the case of unregulated profit maximization, Basso (2005) and Zhang and Zhang (2006) clearly found, in the pricing rule of the airport, the “double marginalization” problem that affects an uncoordinated vertical structure of airport and airlines. For a given capacity, the airport charge will decrease with the number of airlines downstream. On the other hand, and in a somewhat expectable result, an airport that maximizes social welfare subject to cost recovery will have a charge that is in between the unconstrained welfare-maximizing charge and the profit-maximizing charge. The balance will be given by the severity of the budget constraint. Turning to capacity decisions, Basso (2005) and Zhang and Zhang (2006) found that an unconstrained welfare-maximizing airport will provide capacity until the marginal cost of capacity equates the marginal benefits in reducing delays (to airlines and passengers in the case of Basso, to passengers only in the case of Zhang and Zhang). Interestingly, Zhang and Zhang (2006) proved that when both price and capacity are decision variables, in their setting, the market structure (i.e., N ) has no impact on airport’s actual demand and capacity. Consequently, delay levels will be independent of market structure. This however does not hold in Basso’s setting, in which airlines are differentiated and/or passengers care about schedule delay cost. The explanation has to do with the “preferred N ” of a welfare-maximizing airport. Basso showed that there are two opposing effects. With the congestion and market power effects being controlled, as it is the case here, fewer airlines in oligopoly would provide – each of them – higher frequencies than more airlines, thus delivering smaller schedule delay costs which increases social welfare. Smaller N would be preferable. On the other hand, differentiation brings about new demand when N increases, so a larger N is preferable. An unregulated private airport, however, would increase its capacity until the marginal revenue of doing so equates its marginal cost. Clearly, this capacity rule is different from the previous one. Basso (2005) noted then, that this is different than what happened in the traditional approach (e.g., Oum et al., 2004), in which the capacity rules of unregulated private airports and unconstrained public airports were the same. However, when N goes to infinity, i.e., airlines become perfectly competitive, the capacity rules become the same. The explanation for this is given in the next section. Further, Zhang and Zhang, and Basso, showed that conditional on the level of traffic, a private airport

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

111

would over-supply capacity. However, that capacity would most likely be too small in a second-best sense. That is, a public airport that is forced to charge using the private airport pricing rule, would most likely supply more capacity than the actual capacity offered by the private airport (Basso, 2005). As with price, a budget constrained airport would, conditional on the level of traffic, choose a level of capacity that is in between the private capacity and the unconstrained public capacity. Basso looked at two other types of ownership as well. First, he investigated the case in which airports and airlines vertically integrate. The reason to look at this is because it has often been argued that more strategic collaboration between airlines and airports would solve incentive problems, particularly regarding capacity expansions. Basso found that the airport charge would include marginal cost and a term equal to the uninternalized congestion cost of each carrier, but would also include a third term, which is positive. This mark-up is put in place to fight against the business-stealing effect, a horizontal externality typical of oligopoly: a firm does not take into account profits lost by competitors when expanding its output. By increasing the airlines’ marginal cost with a higher airport charge, the airport would be able to induce a profitable (for the combined vertical structure) contraction of total output. In fact, the final outcome is indeed that of cooperation between competitors in the airline market. The intuition is that airlines would “capture” an input provider to run the cartel for them, given that they are unable to collude on their own. As for capacity, the vertically integrated structure would have the same capacity rule as the unconstrained public airport. The actual capacity however would be below the second-best capacity (i.e., a public airport that is forced to charge using the vertical integration pricing rule would supply more capacity). Basso also showed that, depending on how differentiated airlines are, and how strong schedule delay effects are, profits may be higher when the airports integrate with a single airline. A non-integrated private airport though will always prefer a larger N .10 Basso (2005) also looked at the case in which two distant airports are privatized separately. Social-welfare wise, the results worsen because the airports’ demands are perfect complements: in his setup with only two airports, a trip that starts at one airport necessarily ends at the other. Therefore “competition” between the airports induces a horizontal double-marginalization problem. This horizontal double marginalization arises in both the unintegrated and integrated vertical structures.

4 RELATIONSHIP BETWEEN APPROACHES It is clear that the questions examined in the two approaches – which we have called the traditional approach and the vertical structure approach – have not perfectly over­ lapped, and the two approaches appear rather different. This raises questions about the transferability of results, something that seems quite important to clarify if one is to apply to policy making what has been learned from analytical models of airport pricing.

10

Both Brueckner (2002) and Zhang and Zhang (2006) had N airlines downstream. However, public airports and vertically integrated airports would have no particular preference for N in their settings, because airlines are homogenous and there are no schedule delay effects.

112

LEONARDO J. BASSO AND ANMING ZHANG

We address the issue of the connection between the two approaches in this section, based on results and discussions in Basso (2005). In the traditional approach, the airline market is not formally modeled, under the assumption that the airport charge would be completely passed to consumers, and that airline tickets and other charges would be exogenous to the airport. Oum et al. (2004) argue that this would be the case under perfect competition. In the vertical structure approach, on the other hand, it is recognized that airports provide an essential service that is required by airlines to move passengers; therefore, airports are viewed as providing a necessary input for the production of an output: travel. In fact, some authors using the vertical structure approach have been somewhat critical of the traditional approach on the grounds that it does not properly consider all the actors involved. For example, Raffarin (2004) said that it is rather strange that the pricing rules obtained from the traditional approach do not consider passengers’ utility. However, this is not completely accurate. Passengers are indeed somehow considered in the approach, as delay costs affect them as well, something that Raffarin missed.11 On the other hand, a view of the problem that recognizes that (i) airlines may have market power and (ii) airports provide an input for the production of an output sold at another market, appears more complete. Using the notations of Section 2, what the papers in the vertical structure approach have shown is that for any given airport charge, P, and airport capacity, K, the airline market – the downstream market – will reach some equilibrium. This equilibrium is constituted not only by equilibrium traffic but also by equilibrium delays and air ticket prices. By stressing this fact, three things become apparent. First, as far as the airport is concerned, its demand will be some direct function of P� K and of the (exogenous) airline market structure, which in most papers is represented by the number of airlines N . Hence, the airport’s derived demand would be Q�P� K� N�. Delays enter the picture through the equilibrium of the downstream market. How this demand faced by the airport responds to changes in P and K is something that a formal analysis of the airline market can unveil. Second, how airport charges and airlines’ delay costs are passed to consumers is built inside the demand faced by the airport and hence depends in general on the nature of the equilibrium reached in the airline market. In this sense, it would seem that a full price model pertains more to the airline-market stage than the airport-market stage. And third, other airline charges may not be exogenous to the airport because the downstream equilibrium – that is, the airport demand – depends on P and K, which are decided by the airport. Airport managers with foresight will take this into account and decide user charges and capacity investment accordingly. Thus, we can go back to the traditional approach and contrast its basic setting with what we have described above. Two important questions arise: 1. Is it reasonable to use the full-price idea at the airport level, rather than at the airlinemarket level? That is, under what conditions would it be legitimate to assume that the airport demand can be written as Q��� – with � = P + D�Q� K� – rather than as Q�P� K� N�?

11

The problem might lie in that Morrison (1987) states that the final consumers of airports services are airlines, even though in his model congestion explicitly affects passengers. In Oum et al. (2004), passengers are said to be the final consumers.

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

113

2. If under some conditions the airport demand can reasonably be written as Q���, would its integration give a correct measure of consumer surplus? We have learned that the consumers of airports are both airlines and passengers. Hence, a socialwelfare function should include both airlines’ profits and passenger surplus. This is in fact what is explicitly done in the vertical structure approach when analyzing the maximization of social welfare. In the traditional approach, however, consumer surplus has been obtained through integration of the airport demand function with respect to a full price. Under what conditions does the derived demand for the airport carry enough information about the downstream market so that its integration gives a correct measure of airlines’ profits and passenger surplus? In short, these two questions attempt to clarify how the two approaches are related to each other. Basso (2005) analyzed this by using a vertical structure model to derive the demand for the airport. Details of his modeling were presented in Section 3 but, in short, he considered an airline oligopoly featuring N symmetric airlines, facing (linear) differentiated demands, which are dependent on the vector of full prices. These full prices are the sum of the airfare plus congestion delay costs. The important thing to note here is that, as opposed to the traditional approach case, the full-price is used at the airline market level rather than at the airport market level. Solving the airline subgame, Basso found an equation which implicitly defined the airport’s derived demand function Q�P� K� N�. Examination of the equation allowed Basso to show that, in general, Q would depend not only on � = P + D�Q� K� but also on DQ ≡ �D/�Q and N . That is, in general, Q ≡ Q��� DQ � N�. However, in the “perfect competition” case, i.e. when N → � under the Cournot conjecture, it is true that Q��� DQ � N → �� ≡ Q���. Thus, the answer to the first question above is: Under perfect competition, a full price as defined by � can in fact be used directly at the airport-market level. It does summarize well the equilibrium of the downstream market. Now we turn to the second question: If we assume that there is perfect competition, would the integration of Q��� correspond to the sum of airlines’ profits and passenger surplus? This question is relevant because, if it is not the case, then even under perfect competition the traditional approach would be maximizing a function that is not total social welfare. This second question is related to the more general subject of the relation between input and output market surplus measures (Jacobsen, 1979; Quirmbach, 1984; Basso, 2006). Results from that literature, however, do not apply directly to this case because, in the traditional approach, the integration of the airport’s demand is with respect to the full price �, rather than the airport charge P. To answer the question, Basso (2005) computed, in subgame equilibrium, the surpluses of airlines and passengers. He then showed that, when N → �, and therefore one can reasonably write the airport demand as Q���, the integration of the airport demand with respect to � would give ��

Q���d� = � + PS

(13)

�

where � is the aggregate airlines’ profits, and PS is passenger surplus. Therefore, the answer to the second question would be this: When there is perfect competition, such

LEONARDO J. BASSO AND ANMING ZHANG

114

that using Q��� is justified, the integration of the airport demand with respect to the full price � will deliver a correct measure of consumer surplus, i.e., airlines’ profits plus passenger surplus. Perfect competition in the airline market was in fact the maintained assumption of Oum et al. (2004). Hence, Basso (2005) provided a theoretical support for their claim. But he also provided boundaries for the use of the traditional approach: it would be reasonable to use it only if market power at the airline level is absent.12 If airlines have market power, modeling the demand for the airport as Q��� would be incorrect. Furthermore, Basso (2005) showed that its integration with respect to � would actually fall short of giving the sum of airlines’ profits and passenger surplus. In this case, a full model that formally considers the airline market, as in the vertical structure approach, would be necessary. Lastly, since Q��� cannot be used when there is market power downstream, one may wonder whether by using the demand function Q�P� K� N� – which may be estimated empirically for instance – and by integrating it with respect to P, one can adequately capture airlines’ profits plus passenger surplus. This is not the case, unfortunately. Using results in Basso (2006) it can be shown that the integration of the airport demand with respect to P would give: �� P

N −1 �N − 1� � �Q Q�P� K� N�dP = � + PS − Q DQ dP N N �P �

(14)

P

Thus, there is no value of N for which the integral of the airport demand with respect to P would be equal to airlines’ profits plus passenger surplus (not even if N is very large).

5 PRICING OF AIRPORT NETWORKS The papers we have reviewed, in both the traditional and vertical structure approaches, do not really deal with airport networks. In most cases they deal with an airport in isolation. The exceptions, so far, have been Pels and Verhoef (2004) and Basso (2005) who consider a “network” of two airports. Yet, real air networks are obviously more complex than that, and it is fairly clear that in these real airport networks other issues arise. We review here three papers – namely, Oum et al. (1996), Brueckner (2005), and Pels et al. (1997) – that have dealt with a network of airports, that is, three or more airports. Oum et al. (1996) argue that in hub and spoke (HS) networks, airports’ demands are complementary because any take-off at a spoke airport will generate a landing at the hub. This complementarity is of different nature than the complementarity that arises in

12 An important qualification here is that these results hold for the specific set-up that Basso (2005) used which, for example, featured linear demands, Cournot competition and symmetric airlines with constant operational marginal costs. An open research question is how these findings change under more general demand and/or cost specifications, and other types of airline competition.

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

115

two-airports networks because the presence of a hub introduces asymmetries. As in the two-airport cases, failure to consider the complementarities when looking for optimal pricing policies will result in social welfare losses. But in a HS network, congestion at the hub will build up more rapidly than at spoke airports. And when budget adequacy is an issue, this may imply the need for cross subsidizations between airports. Depending on the type of ownership however, cross subsidies may be unfeasible. Oum et al. study how ownership and cost recovery constraints affect airport pricing in a HS network and, consequently, social welfare. More specifically, Oum et al. (1996) consider n airports in a HS system: n-1 airports are spoke airports and there is one hub. All the airports have constant operational marginal costs and fixed capacity, but their capacity maintenance costs are positive. The demands for these airports depend on the charges at both the hub and spoke airports. All the airports are congestible, but congestion is an external cost that the airport authority will include in the social-welfare function; it does not affect the demands (as in Raffarin, 2004; see Section 3). This setup shows two things: First, the spoke airports’ demands are indeed complementary with the hub’s demand, but the demands are not directly complementary among the spoke airports. Second, that this paper is ascribable to the traditional approach, since the airline market is not for­ mally included. Indeed, consumer surplus is measured as the integrals of the airports’ demand. Oum et al. first analyze the case in which all the airports are publicly owned and under the control of a single authority: this is the “federal” case. The authority will maximize the airports’ profits plus consumer surplus – the sum of the integrals of airports’ demands – minus external congestion costs. The optimal pricing policy would have all the airports charging SMC, that is, the operational marginal cost plus the external costs of congestion. Since the hub is more likely to be heavily utilized, congestion will be greater there than at the spoke airports. Hence, they assume that SMC pricing would lead to cost recovery at the hub but to deficits at the spoke airports. The first-best federal case then would require cross subsidies from the hub to the less utilized spoke airports. If a budget constraint is set in place, the question becomes whether the hub makes enough profits to cover for the spoke airports’ deficits. If it does, we are back in the first-best case. If it does not, then Ramsey pricing is called for: the charge at the hub will increase. Cross subsidization will be, obviously, still needed and this alternative will be welfare inferior to SMC pricing. They then look at the case in which each airport is under the control of a different authority who, subject to cost recovery, maximizes its local social welfare, that is, the integral of own demand plus own profits, minus congestion costs. This is the “de-federalized” or “local government” case. Given the assumption about SMC not covering costs in spoke airports, in this case, the hub will price at SMC, but the spoke airports would charge average costs to ensure cost recovery. Since individual cost recovery implies overall cost recovery, this case will be inferior, social-welfare wise, to the previous Ramsey pricing case. In general, in the federal case, and independent of whether SMC or Ramsey prices are used, charges at the hub will be larger and charges at the spoke airports will be smaller than in the local government case. Oum et al. (1996) conclude that de-federalization of airports may imply social welfare losses: by not jointly pricing the airports, the local airport authorities will not take into account

LEONARDO J. BASSO AND ANMING ZHANG

116

that demands are complementary and cross-subsidies will likely become unfeasible. The welfare losses, though, would have to be balanced against possible X-inefficiencies gains that de-federalization may bring about.

5.1 Network Airport Pricing with Airline Competition Recall, however, that the conclusion of Section 4 was that a traditional approach would be justified only when air carriers are atomistic. What would happen if carriers have market power? In this case, we would need a vertical structure type of approach. This is what Brueckner (2005) analyzes. The main point here has to do with the meaning of market power. One of the conclusions in Section 3 was that congestion tolls would decrease in an airline’s share of flights at the airport, because an airline only internalizes the congestion caused on own flights. Since in that section, only one or two airports were considered, the share of flights at the airport was identical to the share of flights at the city-pair market level. However, when one considers even a simple network of airports in which airline competition exists, it is no longer true that the share of flights at the airports will necessarily be equal to the share of flights at the city-pair market level. Hence, the relevant question becomes – what is the relevant flight share for congestion internalization? Brueckner considers the following network in which two airlines compete. In this network, airport H is airline 1’s hub, while airport K is airlines 2’s hub. Airline 1 serves four city-pair markets (depicted by the solid lines in Figure 1): AH, KH, BH, and AB (two legs). Airline 2 also serves four city-pair markets (dashed lines). The airlines compete in two markets, KH and AB, while each is a monopolist in its two other markets. It can be easily recognized – for example under full symmetry – that airline 1’s share of departures and take-offs at its hub H is larger than airline 2’s share. Similarly, airline 2 dominates hub K in terms of departures and take-offs. However, in the two markets where the airlines compete, they would both have a 50% share of flights under symmetry. This nicely shows the difference between the shares of flights at airports and the share of flights in city-pair markets, which justify the research question.

H

A

B

K

Airline 1 Airline 2

Figure 1 Network Structure and Airline Competition (Brueckner, 2005).

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

117

To analyze what would be the optimal congestion toll, Brueckner uses a setup which essentially is the same as in his single-airport paper (Brueckner 2002; see Section 3 for a description) but considers each of the various markets. Airports are assumed to have a fixed exogenous capacity with only the hubs being prone to con­ gestion. The derivation of optimal congestion tolls is quite involved so it is omitted here, but the conclusion is simple and important: Regardless of the degree of market power that an airline has in the city-pair markets it serves, the amount of conges­ tion it internalizes depends only on its flight share at the congested airport. Hence, “the appropriate airport congestion tolls are carrier-specific and equal to the conges­ tion damage from an extra flight times one minus the carrier’s airport flight share” (Brueckner, 2005, Proposition 1, p. 612). An important final point that Brueckner (2005) raised has to do with the marketpower effect we discussed in Section 3. There, we saw that, while a congestion toll is justified when carriers are oligopolistic, from a first-best point of view a subsidy was also justified as a means to fight against market power at the airline level and hence reduce allocative inefficiencies.13 In the simple settings of one or two airports, both the congestion effect and the market-power effect depended on a carrier’s flight share. But in that case the airport share and the city-pair market share were the same. Brueckner (2005) showed that, in a network setting, whilst the congestion tolls are airport-specific, the subsidies required are city-pair specific. Hence, an airport regulator would need to calculate appropriate airport-specific congestion tolls together with city-pair specific subsides to obtain, finally, the optimal charge, which would be positive if the congestion effect dominates the market-power effect. Brueckner argues that, since market-level subsides are impractical to implement, only airport congestion tolls would be used, an approach that would be welfare improving, yet not first-best, if congestion effects dominate.

5.2 Network Airport Pricing with Variable Route Structure Now, both Oum et al. (1996) and Brueckner (2005) have assumed that the route structure of airlines, that is, the way airlines move passengers between origins and destinations, remain unchanged and is independent of the pricing practices of airports. But, what would happen in the long run if the route structure is be changed? For example, it has been often argued that economies of density drive the adoption of HS networks. But if congestion at hubs is too important, airlines may decide to by-pass them, offering direct connections in some city-pair markets.14 Would congestion pricing affect the timing of such a decision? May airports use their pricing practices as a way to compete for connecting passengers, that is, may they compete to become hubs? A model including all these elements would be indeed very complicated and has, as far as we know, not yet been proposed. However, there is one paper that, even though in a context of non-congestible airports and a monopoly airline, does look at how airport pricing and

13

On the other hand, we do not normally think of solving the market-power problem by subsidizing the firms,

for several good reasons. The subsidy may alternatively be interpreted as an imperfect proxy for some kind

of antitrust policy in its effect on price reduction.

14 For a paper related to this issue, see Basso and Jara-Diaz (2006).

LEONARDO J. BASSO AND ANMING ZHANG

118

B

Fully connected Network (FC)

Hub and Spoke Route Structure (HS)

A

C

Figure 2 Possible Route Structures.

airline’s choice of route structure are related. Specifically, Pels et al. (1997) consider a model with three non-congestible public airports (hence three OD city-pair markets) and a monopoly airline. Airport charges are directly made to passengers. Thus, the demands for airports and for the airline depend on both airfares and airport charges. The airports and the monopoly airline play a simultaneous game in which each airport chooses its per-passenger charge, while the airline chooses a route structure and its airfares. The objective function of the airports is to maximize own social welfare (as in the de-federalized case of Oum et al., 1996), which is measured as the integral of the airport’s demand, subject to a budget constraint. The airline seeks to maximize its profit. There are some key assumptions in the model, which are more easily explained using Figure 2. First, it is assumed that node (airport) A has more passenger generating capacity. That is, if airports’ and airline charges were zero, the demands in the AB and AC pairs would be �, while in the BC pair it would be ��, where � < 1. Second, consumers only care about the monetary charges (from the airports and the airline) but would not care about travel times (which are higher in a HS route structure) or whether they have to make connections or not. Third, the marginal cost of carrying a passenger is constant and equal across links; hence, in a HS route structure, a passenger traveling from B to C would cost the airline 2c whilst with a FC route it would cost only c. Finally, if a link is used, it has a fixed cost c0 . Hence, a HS route structure is cheaper in terms of fixed costs, as it only uses two links (vs. three links in a FC structure), but is more expensive in term of operational costs. Pels et al. (1997) show that, in this setup, if the airport charges are zero (or, if they are equal but are chosen non-strategically, i.e., without considering what the airline does) a HS route structure will be preferred by the monopoly airline if � < �, i.e. the demand in the BC market is much smaller than the demands in the AB and AC markets. The limit � increases in both c and c0 , and decreases in �. Further, they show that the airline will always choose to place its hub at the node with the highest level of demand, in this case, node A.15 When the airports choose their prices simultaneously with the airline’s choice of route structure and airfares, Pels et al. show that the airport charges will increase in fares, but the fares will decrease in airport charges. The “dynamics” of equilibrium would be:

15 For more discussions on the choice of route structure, see, e.g., Oum et al. (1995), Hendricks et al. (1999), Pels et al. (2000) and Jara-Diaz and Basso (2003).

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

119

The monopoly airline, which is a profit maximizer, would increase its prices depressing demands. Since the public airport must break even, it would raise its own charges, but that would induce the airline to decrease its prices. This in turn would increase demand, inducing a decrease in the airports’ charge, which in turn would induce the airlines to increase price. Eventually, this loop may reach an equilibrium, although Pels et al. show that non-existence of equilibrium is a possible outcome. Since analytical solution of the equilibrium is unfeasible, they rely on a numerical simulation to extract more conclusions. They found that, only if � is small enough, the airline would choose a HS route structure. The higher the �, however, the better for the hub. More importantly, price competition between the airports seems to have little effect on the airline’s choice of a hub; the choice would still be made based on passenger generating capacity. Obviously, one can foresee that the actual geographic position of the airports would be important as well. A hub would not be placed really far away from all its spoke airports. But in this model, distances, that is the topology of the network, does not play a role. This is reasonable under the assumption that all airports are located fairly close or equidistant from each other.

6 CONCLUSIONS AND FURTHER RESEARCH Airport pricing has been widely analyzed in the economics literature. In this survey paper, we have focused on analytical models of airport pricing from 1987 on. We have grouped the models in the literature into two broad approaches. Roughly, the traditional approach has used a classical partial equilibrium model where the demand for airports depends on airport charges and on congestion costs of both passengers and airlines; the airline market is not formally modeled, in several cases under the assumption that airline competition is perfect and hence airport charges and delay costs are completely passed to passengers. The vertical structure approach was motivated initially by the increasing and acute congestion at major hub airports in the United States and around the world. Since hub airports usually have only a few dominant airlines, the airline market there is better characterized as oligopoly: air carriers may possess market power. Thus, the airline market was considered in the analysis of airport pricing. Furthermore, the vertical structure approach has recognized that airports provide an input for the airline market – which is modeled as a rather simple oligopoly – and that it is the equilibrium of this downstream market that determines the airports’ demand: the demand for airports is therefore a derived demand. The questions investigated with the two approaches have not perfectly overlapped. The traditional approach has been used to analyze a variety of issues such as optimal capac­ ity investments, effect of concession revenues, privatization, efficiency of alternative regulation mechanisms, cost recovery when capacity cannot be increased continuously, and efficiency of weight-based airport charges. On the other hand, the vertical structure approach has focused mainly on calculating the additional toll that airlines should be charged to attain maximization of social welfare. It is only recently that vertical structure models have been used to assess such issues as optimal capacity levels, or the effects of privatization on airport charges. Drawing from results in Basso (2005), we indicated here that abstracting from the airline market, as is done in the traditional approach, is a

120

LEONARDO J. BASSO AND ANMING ZHANG

reasonable approximation only when airlines behave competitively, but it is not when airlines have market power. In the latter case, the derived demand for the airport would not be dependent only on its full price, as it is assumed in the traditional approach. As a result, the integration of the airport demand with respect to the full price, which is said to capture consumer surplus, would not adequately capture the surpluses of passengers and airlines because market power and congestion effects preclude it. Therefore, in that case, the function that is maximized in the first-best scenario would not correspond to total social welfare.16 The fact that the airline market cannot be ignored if airlines have market power implies, on one hand, that future research would need to use vertical structure models if the airline market structure is an important factor for the issues to be investigated. This may include re-examination of some of the questions that have been addressed only with the traditional approach, including, for example, the effect of concession revenues on airport charges, the efficiency of regulation mechanisms for congested hub airports, and congestion pricing with lumpy airport capacity. But on the practical side, the fact that the airline market has to be included in the models is also bad news for managers of public airports and regulators: to implement optimal decisions, the amount of information required would be massive even in simple settings, which undoubtedly complicates the problem. In the models we have reviewed in this survey, authors have resorted to a number of simplifications, which was the price to pay to preserve analytical tractability. In the airline market of vertical structure models, two usual simplifications are the assumption of fixed proportions and the assumption of symmetric airlines. The former was made when the authors assumed, as constant, the product between aircraft size and load factor (or both). Yet, it has been widely accepted that airlines enjoy what is called “economies of traffic density” – decreasing average cost on nonstop connections – owing largely to the economies of aircraft size. These economies are not considered under the fixedproportions assumption (which precludes the endogenous choice of aircraft types by airlines). A variable-proportions case would arise because, if the charge per flight is too high, airlines would have an incentive to change to larger airplanes, independently of existing or exhausted economies of airplane size. So, with privatization for example, not only capacities and traffic levels would be distorted downwards, but aircraft size would be distorted as well. Modeling this effect is an interesting area of future research albeit a complex one, as larger aircrafts imply smaller frequencies, which directly affects congestion and demand through schedule delay costs.17 Regarding the assumption of asymmetric airlines, certainly insights would be gained if the analysis could be extended to the case of asymmetric airlines, as the model would depict a more realistic case. Brueckner (2002, p. 1368) stated that “cost differences

16 This result in fact applies not only to airports but to any other types of transport terminal, or even railroad tracks, since the situation is essentially the same. 17 Note that Raffarin (2004) is not an analysis of variable proportions case because, although she did considered different aircraft sizes, the airlines where not free to decide about their preferred aircraft size. Rather, the aircraft size was exogenously given through a parameter, which thus showed up in the final pricing rules of the airport. Also, in her model congestion did not directly affect passengers or airlines but was an external cost to be minimized by the airport authority and capacity was fixed.

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

121

across firms may not be a useful source of asymmetry, however, because a planner would not allow high-cost firms to operate at the social optimum.” In Basso (2005) and Zhang and Zhang (2006), however, there was no social planner but rather managers of public airports maximizing social welfare, who probably would not have the power to preclude less efficient airlines to operate. But they did not consider asymmetries. It seems, to us, that characterizing the properties for the case of asymmetric airlines would be unfeasible analytically; instead, numerical simulations would be required. Pels and Verhoef (2004) did present some numerical simulation results for the case of an asymmetric duopoly (see Table 2 for a description of their setting). The papers in Section 3 have looked at either a single airport in isolation or, at most, round-trip travel between two airports. In the latter case, the airports have complementary demands; consequently, public airports that are priced independently would not achieve a first-best outcome (Pels and Verhoef, 2004), and separate private airports would end up with a horizontal double marginalization (Basso, 2005). However, airport networks are more complex than that; and on this subject, the papers presented in Section 5 represent good progress in understanding the main issues. Nevertheless, we believe that there is still much work to do. In Oum et al. (1996) and Brueckner (2005) there were no route structure decisions on the part of airlines, but it is through route structure decisions that airports may actually compete: they would be competing for connecting passengers. On the other hand, although considering route structure decisions, Pels et al. (1997) do not include congestion, capacity choices, or airline competition. Further work on the pricing of network airports – including effects of privatization and regulation mechanisms – is, in our view, a clear line of future work. A related aspect is geographic competition: airports competing for costumers in the same origin, i.e., with overlapping catchment areas, as in the case of New York and San Francisco Bay Area. There has been some empirical work on this issue (e.g., Ishii et al., 2005, and the references cited there), but not too much work on the analytical side. Some papers have looked at competition between congestible Bertrand facilities (e.g., De Borger and Van Dender, 2006) but they overlooked the intermediate carrier market in vertical structures discussed above. A simple model of geographic competition between two airport-airline structures is Gillen and Morrison (2003). But they considered only the case of one airline per airport and the joint airport–airline profit maximization, and they did not consider the issues of airport congestion and airport capacity choices. We think that competition in multiple-airport regions with congestion is another interesting area of future research.18 Another important aspect is the issue of peak-load pricing, in addition to just con­ gestion pricing. Most of the models we have reviewed are about congestion pricing rather than peak-load pricing, in the sense that even if there is more than one period in those models, the demands between periods are not interdependent. Hence, the only way to fight against excess usage is to dampen the demands. When the periods are interdependent, however, pricing can be used not only to dampen the demands, but also to redistribute consumers and flights across different periods, “flattening” the demand

18

As this survey was being completed, a paper dealing with the specific issue of geographic competition was accepted for publication. Time precluded the presentation of the main results in this review. For details see Basso and Zhang (2007).

122

LEONARDO J. BASSO AND ANMING ZHANG

curve – the case of peak-load pricing. Brueckner (2002) allowed for endogenous sorting to peak and off-peak periods, but the sorting was done mainly through airlines’ decisions. If the airlines use peak-load pricing, then that would deliver a different demand pattern to the airport, which would probably also have peak and off-peak periods. The airport would then have an incentive to choose prices for its own peak and off-peak periods, probably using peak-load pricing as well, in order to maximize its objective function. The objective function would be dependent on the type of ownership and regulation. Hence, we would be in a situation of sequential peak-load pricing, which represents particularly well the case of airports and airlines (Basso and Zhang, 2006). We have highlighted some of the issues that we think should be examined in the future, but perhaps one of the most important aspects of future research has to do with actual policies. It is seldom true that airports are priced as in a system, and it is seldom true that airport managers have access to all the information that they would need to do what is best. Hence, how should public airports be priced when they are not in a system, and when information is incomplete? And given this, what are the costs and gains of airport privatization, and what would be a good and feasible regulation mechanism for privatized airports? Investigating these questions will advance our understanding of the subject and produce useful guidance to policy formulations.

REFERENCES Basso, L.J. (2005) Airport ownership: Effects on pricing and capacity. Working Paper, Sauder

School of Business, The University of British Columbia (http://ssrn.com/abstract=849584).

Basso, L.J. (2006) On input markets surplus and its relation to the downstream market game.

Working Paper, Sauder School of Business, The University of British Columbia. Basso, L.J. and Jara-Diaz, S.R. (2006) Distinguishing multiproduct economies of scale from economies of density on a fixed-size transport network. Networks and Spatial Economics, 6, 149–162. Basso, L.J. and Zhang, A.M. (2007) Congestible facility rivalry in vertical structures. Journal of Urban Economics, 61, 218–237. Basso, L.J. and Zhang, A.M. (2006) Sequential peak-load pricing: The case of airports and airlines. Working Paper, Sauder School of Business, The University of British Columbia. Brueckner, J.K. (2002) Airport congestion when carriers have market power. American Economic Review, 92(5), 1357–1375. Brueckner, J.K. (2005) Internalization of airport congestion: A network analysis. International Journal of Industrial Organization, 23(7–8), 599–614. Carlin, A. and Park, R.E. (1970) Marginal cost pricing of airport runway capacity. American Economic Review, 60, 310–319. Carlsson, F. (2003) Airport marginal cost pricing: Discussion and an application to Swedish airports. International Journal of Transport Economics, 30, 283–303. Condie, S. (2000) Whither airport regulation? In: Smith, H.L. and Bradshaw, W. (Eds.), Privati­ zation and Deregulation of Transport, St. Martin’s Press, Inc., New York, 364–393. Czerny, A.I. (2006) Price-cap regulation of airports: Single-till versus dual-till. Journal of Regu­ latory Economics, 30, 85–97. Daniel, J.I. (1995) Congestion pricing and capacity of large hub airports: A bottleneck model with stochastic queues. Econometrica, 63, 327–370.

INTERPRETATIVE SURVEY OF ANALYTICAL MODELS

123

De Borger, B. and Van Dender, K. (2006) Prices, capacities and service quality in a congestible Bertrand duopoly. Journal of Urban Economics, 60, 264–283. Doganis, R. (1992) The Airport Business, Routledge, London. Economist (2006) Global warming, The Economist, 8 June 2006, Economist Intelligence Unit. Forsyth, P. (2000) Models of airport performance, In: Hensher, D.A. and Button, K.J. (Eds.), Handbooks in Transport, Vol. 1, Pergamon, Amsterdam, 597–608. Gillen, D.W. and Morrison, W.G. (2003) Bundling, integration and the delivered price of air travel: Are low cost carriers full service competitors? Journal of Air transport Management, 9, 15–23. Hendricks, K., Piccione, M., and Tan, G.F. (1999) Equilibria in networks. Econometrica, 67(6), 1407–1434. Horonjeff, R. and McKelvey, F.X. (1983) Planning and Design of Airports. McGraw-Hill. Ishii, J., K. Van Dender and S. Jun (2005) Air travel choices in multi-airport markets, Working Paper, Department of Economics, University of California at Irvine. Jacobsen, S.E. (1979) Equivalence of input and output market Marshallian surplus measures. American Economic Review, 69(3), 423–428. Jara-Diaz, S.R. and Basso, L.J. (2003) Transport cost functions, network expansion and economies of scope. Transportation Research Part E-Logistics and Transportation Review, 39(4), 271–288. Jones, I., Viehoff, I. and Marks, I. (1993) The economics of airport slots. Fiscal Studies, 14, 37–57. Lave, L.B. and De Salvo, J.S. (1968) Congestion, tolls and the economic capacity of a waterway. Journal of Political Economy, 76, 375–391. Levine, M.E. (1969) Landing fees and the airport congestion problem. Journal of Law and Economics, 12, 79–108. Lu, C.C. and Pagliari, R.I. (2004) Evaluating the potential impact of alternative airport pricing approaches on social welfare. Transportation Research Part E-Logistics and Transportation Review, 40(1), 1–17. Mohring, H. (1976) Transportation Economics. Ballinger, Cambridge MA. Morrison, S.A. (1983) Estimation of long-run prices and investment levels for airport runways. Research in Transportation Economics, 1, 103–130. Morrison, S.A. (1987) The equity and efficiency of runway pricing. Journal of Public Economics, 34(1), 45–60. Morrison, S.A. and Winston, C. (1989) Enhancing the performance of the deregulated air trans­ portation system. Brookings Papers on Economic Activity, 61–112. Oum, T.H. and Zhang, Y.M. (1990) Airport pricing – Congestion tolls, lumpy investment, and cost recovery. Journal of Public Economics, 43(3), 353–374. Oum, T.H., Zhang, A.M., and Zhang, Y.M. (1995) Airline network rivalry. Canadian Journal of Economics, 28, 836–857. Oum, T.H., Zhang, A.M., and Zhang, Y.M. (1996) A note on optimal airport pricing in a huband-spoke system. Transportation Research Part B-Methodological, 30(1), 11–18. Oum, T.H., Zhang, A.M., and Zhang, Y.M. (2004) Alternative forms of economic regulation and their efficiency implications for airports. Journal of Transport Economics and Policy, 38(2), 217–246. Pels, E., Nijkamp, P., and Rietveld, P. (1997) Substitution and complementarity in aviation: Airports vs. airlines. Transportation Research Part E-Logistics and Transportation Review, 33(4), 275–286. Pels, E., Nijkamp, P., and Rietveld, P. (2000) A note on the optimality of airline networks. Economics Letters, 69(3), 429–434. Pels, E. and Verhoef, E.T. (2004) The economics of airport congestion pricing. Journal of Urban Economics, 55(2), 257–277.

124

LEONARDO J. BASSO AND ANMING ZHANG

Quirmbach, H.C. (1984) Input market surplus – The case of imperfect competition. Economics Letters, 16(3–4), 357–362. Raffarin, M. (2004) Congestion in European airspace – A pricing solution? Journal of Transport Economics and Policy, 38, 109–125. Ross, T.W. (1984) Uncovering regulators social-welfare weights. Rand Journal of Economics, 15(1), 152–155. Spence, A.M. (1975) Monopoly, quality, and regulation. Bell Journal of Economics, 6(2), 417–429. Starkie, D. (2001) Reforming UK airport regulation. Journal of Transport Economics and Policy, 35, 119–135. US Department of Transportation (2006) National Strategy to Reduce Congestion on America’s Transportation Network, May 2006, Washington DC. US Federal Aviation Administration (1969) Airport Capacity handbook. Government Printing Office, Washington DC. Zhang, A.M. and Zhang, Y.M. (1997) Concession revenue and optimal airport pricing. Trans­ portation Research Part E-Logistics and Transportation Review, 33(4), 287–296. Zhang, A.M. and Zhang, Y.M. (2001) Airport charges, economic growth and cost recovery. Transportation Research Part E-Logistics and Transportation Review, 37, 25–33. Zhang, A.M. and Zhang, Y.M. (2003) Airport charges and capacity expansion: Effects of conces­ sions and privatization. Journal of Urban Economics, 53, 54–75. Zhang, A.M. and Zhang, Y.M. (2006) Airport capacity and congestion when carriers have market power. Journal of Urban Economics, 60, 229–247.

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

5 What if the European Airline Industry had Deregulated in 1979?: A Counterfactual Dynamic Simulation∗ Purvez F. Captain† , David H. Good‡ , Robin C. Sickles§ , and Ashok Ayyar¶

ABSTRACT Studies in industrial organization predict rapid consolidation following deregulation to seize economies of scale. The European airlines, while witnessing some strategic movement, have remained remarkably stable in the wake of deregulation. By contrast, the US industry underwent deregulation beginning in late 1978 and experienced a vigorous shakeout. This begs the question: if Europe had deregulated in 1979 alongside the US, how would have the European industry fared without the American experience in hindsight? We developed a dynamic industry model to answer this question, simulating for optimal levels of operational variables, namely level of employment, network size, and fleet size for the period 1979–1990. The study reveals which European airlines were operating most inefficiently by comparing the simulation results with the actual numbers. Our findings point to several sources of forgone profits, in particular to the need for the European carriers to adopt policies which allow them to take advantage of returns to density by network reconfigurations brought about by code-sharing arrangements.

∗ The findings and interpretations reflected in this article do not reflect in any way those of Ernst and Young, LLP, Houston, Texas. † Ernst & Young, LLP, Houston, TX ‡ Indiana University, Bloomington, IN

§ Corresponding author. Rice University, Houston, TX, 6100 Main St Houston TX 77005. e-mail:

[email protected].

¶ Chicago Partners, LLC, New York

126

PURVEZ F. CAPTAIN et al.

1 INTRODUCTION The European airline industry was traditionally sheltered from competition due to its state-owned national carriers and inflexible bilateral agreements. Consequently, the mar­ ket structure developed brought with it market distortions and inefficiencies. The airfares proved it – fares were consistently higher than those charged for equidistant routes in the US. Case in point: When the Federation of European Consumers planned a conference in 1984, they calculated it would be cheaper to fly all their delegates to Washington, DC than to convene anywhere in Europe (Sampson, 1984). The liberalization movement, achieved in three reform packages between 1987 and 1992, successively eased the airline industry’s straight jacket, creating a competitive mar­ ketplace centered on a profit-maximizing business model rather than the old rent-seeking one. History has shown us that rapid consolidation follows on the heels of deregulation, as firms exploit economies of scale (Bannerman, 2002). Yet, while witnessing some strategic movement, the European industry has remained remarkably stable in the wake of deregulation. By contrast, the US industry underwent deregulation beginning in late 1978 and experienced a vigorous shakeout. This begs the question: if Europe had dereg­ ulated in 1979 alongside the United States, how would have the European industry fared without the American experience in hindsight? We developed a dynamic model of the industry in response, simulating for optimal levels of operational variables, namely employment, network size, and fleet size for the period after the US deregulatory initiatives took hold and before the European deregulatory transition began, the period from 1979 to 1990. The study reveals which European airlines were operating most inefficiently by comparing the simulation results with the actual levels of input use. A number of dynamic industry models have been proposed and estimated. Early work by Jovanovic (1982) modeled a perfect foresight equilibrium industry structure in which efficient firms grow and survive, while the inefficient firms decline and exit the industry. In this model, firms learn about their efficiency as they operate in the industry. Firms decide to enter or exit the industry based on a comparison of the value of staying in the industry and behaving optimally with the discounted present value of the opportunity cost associated with the firm’s fixed factor, such as managerial ability or advantageous location. The latter example of a fixed factor is clearly applicable to the European airline industry, where congestion at most major airports has made gates and landing slots coveted fixed factors. Research on industrial evolution has since focused on the relationship between firm size and growth (Evans, 1987a,b; Hall, 1987), endogenous learning (Pakes and Ericson, 1998), and in endogenizing firm strategies (Berry, 1992). This chapter’s approach builds on the intertemporally nonseparable model introduced by Hotz et al. (1988) and utilized and extended Sickles and Yazbeck (1998) and Sickles and Williams (2006). It is organized as follows. Section 2 gives an international regula­ tory history, for the relevant years of the study. Section 3 outlines the dynamic model and discusses the specification and estimation of its relevant components: demand, produc­ tion, and cost. Section 4 discusses the data sources; Section 5 interprets the simulation results and section 6 concludes.

A COUNTERFACTUAL DYNAMIC SIMULATION

127

2 INTERNATIONAL REGULATORY HISTORY The Paris Convention of 1919 first gave rise to the regulation of international aviation. There it was decreed that states have the sovereign rights over the air space of their territory, which immediately involved national governments in the regulation of the industry. Fifty-two countries met at the Chicago conference of 1944 to spar over Five Freedoms of Air, the fundamental set of rights in airline economics. The First Freedom gave the right to fly over a third country’s airspace while on an agreed service and the Second Freedom permitted the airline to land in a third country for fuel and maintenance but not pick up or discharge traffic. The Third Freedom allowed an airline to carry traffic from its own country to a second country in a bilateral. The Fourth Freedom permitted an airline to carry traffic back from that country to its own country. The Fifth Freedom permitted the transportation of traffic by the first country’s airline between the second country and a third country not party to a bilateral (Taneja, 1988). The key parties at the conference, the US and the UK, were at opposite ends of the economic spectrum. The US, whose civil aviation industry emerged from World War II unscathed, sought operating freedom for its airlines under a multilateral “open skies” agreement. Smaller European countries like the Netherlands and Sweden flanked this policy because they would depend heavily on Fifth Freedom traffic. The UK and other large European countries, devastated by the war, proposed the formation of an international authority, which would regulate capacity and fares on routes, thereby giving their aviation industries a chance to rebuild. These opposing views could not be reconciled at the conference, and the convention ended with concordance only on the first two Freedoms. The US and the UK met in Bermuda in 1946 in an effort to resolve differences on the next three freedoms. The two countries agreed to these freedoms in a bilateral agreement (Bermuda I) on flights to and from the US and the UK. This bilateral became a model for the other countries and their respective aviation partners. It also assured that the aviation industry would be heavily regulated and quagmire in political uncertainty (Williams, 1994). Meanwhile, the other participants of the Chicago conference created the International Air Transport Association (IATA) in Havana in 1945. The proposed plan was to fix fares jointly and submit it to governments for approval, instead of either multilateral or unilateral government imposition of fares on airlines. These fares required a unanimous vote from all members and were binding to all of these members. The US Civil Aeronau­ tics Board (CAB) reluctantly agreed to this fare-setting environment, which remained an international fixture for the next 30 years. The system worked fairly smoothly in Europe. The airlines were government-owned and strongly opposed to any form of competition; the fare submission procedure amounted to little more than a formality. The CAB was never comfortable with this arrangement and often protested fare structures set by IATA. In the late 1970s, frustrated by its efforts to liberalize the structure through IATA, the CAB actively began its attempt to liberalize the transatlantic market by forming bilateral agreements with European nations. CAB’s strategy of penetrating one national market at a time and then forcing liberal agreements on others through the threat of traffic diversion was successful in opening the transatlantic market. The level of competition increased substantially with the entry of new airlines into the market.

128

PURVEZ F. CAPTAIN et al.

While the CAB pried the transatlantic market open, the internal European market remained strictly protected until the mid-1980s. The European airlines were mostly public airlines or majority government-owned; they enjoyed the duopolistic situation created by the bilateral agreements and prevented new entry in the intra-European market. Pooling revenue and sharing capacity, the airlines eliminated any competition among themselves in the internal market. The European Commission (EC) recommended opening aviation to competition as early as 1972, but strong objections from the European governments tabled discussions until 1979 when the EC published Civil Aviation Memorandum Number l. The memo recommended that (1) airlines offer cheaper fares; (2) there was a need to develop new cross-frontier services connecting regional centers within the community; (3) a clear universal policy on government subsidies was required; and (4) full freedom of access to all markets was desirable. The transportation ministers adopted these measures in limited form in the early 1980s, which did marginally improve competition and lower fares (Balassa, 1985). The larger European nations, however, were very reluctant through the mid-1980s to abandon the protected status of their national carriers by advocating more liberal competition policies. These governments directly or indirectly subsidized their carriers, the extent of which varied from country to country. Financial assistance was provided to (l) compensate airlines for the imposition of a public service obligation; (2) develop and operate domestic services; (3) provide service to economically underdeveloped regions; (4) encourage the acquisition and operation of specific airplanes (Airbus); or (5) simply cover an airline’s operating loss (Taneja, 1988). EC commissioner Peter Sutherland provided the catalyst for change, threatening to take the airlines to the European Court in 1987 for violation of the competition rules of the Treaty of Rome. The European transport ministers met thereafter in Brussels to negotiate for flexibility in setting fares. The deal allowed airlines to offer discount fares – ranging between 65 and 90 per cent of the economy class fares – provided this was accepted by the member states. It also allowed for an increase in capacity shares on a route provided that the shares split between two countries were not outside the range of 55 to 45 per cent up to 1 October 1989, and 60–40 per cent thereafter. The next round of liberalization talks ended in 1992 in Luxembourg where after 10 years of hard negotiations, the European Union finally agreed on issues that would establish a more competitive environment in European skies. The five major provisions in the deal were the following: 1. Fares: Airlines would be able to set their own prices, subject to two major controls. Brussels was empowered to limit excessive prices from being charged, following notification from national aviation authorities. It would also be able to set a baseline under fares on a specific route if prices free-fall, foisting losses on all carriers. These mechanisms were designed to obviate predatory pricing. 2. Routes: Consecutive cabotage rights to add a domestic leg onto a flight originating from a carrier’s home base to a foreign destination, provided that the load factor on the domestic leg did not exceed 50 per cent of the total on the main flight. Thus, a KLM flight from Amsterdam to Paris could pick up passengers in Paris and fly to Nice provided that the 50 per cent rule was satisfied.

A COUNTERFACTUAL DYNAMIC SIMULATION

129

3. Flights: Agreement to the Sixth Freedom (which had been in dispute since the Chicago Convention) where airlines could fly passengers to two destinations while stopping at a third country, which was the airline’s base. With this in place, Air France, on a flight from Rome to London, could stop in Paris en route and pick up passengers. The Seventh Freedom was introduced, whereby any carrier could fly between any two EC states without the need to start or end in the home country. For example, British Airways could fly between Paris and Frankfurt, with the flight originating and ending at the two destinations. 4. Domestic services: Starting on 1 April l997, any carrier from any EC country could operate internal flights in any of the 12 member states. 5. Licensing: Common rules governing safety and financial requirements on capital adequacy for new entrants to the market. Once satisfied, they would be able to fly on any EC route under the above package (Schipper et al., 2002). The final accord of 1992 established a beachhead in the gradual deregulation of the airline industry. Conducting reform in gradual packages was Europe’s attempt to avert the “big bang” of US reform (Button and Johnson, 1998). Our dynamic industry model attempts to explain how European airline firms would have operated from 1979 to 1990 had they transitioned to deregulation in 1979, as did the US airlines.1

3 THE DYNAMIC INDUSTRY MODEL Our dynamic model analyzes the long-run strategies of the firms and simulates the optimal profit-maximizing levels of the operational variables for different scenarios. We assume that the airline chooses the level of employment (L), network size (N ) and capital (K) to maximize the flow of expected profits Max Et

T �

−t t Lt  Nt  Kt 

t=

subject to a per-period asset accumulation constraint At+1 = t At + Pt Qt − wt Lt − rt It  where Qt = FKt  Lt  Nt   The output price is set by the inverse demand equation that is specified below. At are the firm’s real assets in the beginning of period t,  is the discount factor, t = 1 + rt  where rt is the real interest rate, Pt is the price of output, and It is the level of investment. Other inputs such as materials are assumed to be state variables in our simulations and

1

For an extensive study of airline deregulation in Europe, see Button (1990, 2003).

PURVEZ F. CAPTAIN et al.

130

are thus not directly introduced through the production function. We assume that T is finite and Qt = 0 when the firm exits the industry. Capital accumulation is written in terms of a perpetual inventory model: Kt = It + at where the law of motion for at is at = 1 − at−1 + Kt−1 Here  measures the rate of depreciation of past levels of capital stock to its current level, while is the constant capital depreciation rate. Temporal nonseparability in the dynamic optimization problem comes in through the distributed lag of current and past investment decisions. The dynamic programming problem is characterized by the value function at time t: Vt At  at  Pt  wt  rt  = MaxLNK t Lt  Nt  Kt  +  Et Vt+1 At+1  at+1  Pt+1  The use of standard solution techniques for maximizing the value function with respect to the control variables labor (L), network size (N ), and fleet size (K) provides us with a set of three highly nonlinear equations – Euler equations. The first-order conditions expressed in the Euler equations are L t −  t Et L t + 1 wt + Lt wL t − Pt QL t − Qt PQ t QL t/ wt+1 + Lt+1 wL t + 1 − Pt+1 QL t + 1 − Qt+1 PQ t + 1 QL t + 1 = 0

(1)

N t + L t Pt QN + Qt PQ t QL t/wt + Lt wL t − Pt QL t − Qt PQ t QL t = 0

(2)

k t − L trt + Kt rK t − Pt QK t − Qt PQ t QK t/wt + Lt wL t − Pt QL t − Qt PQ t QL t +  Et K t + 1 + 1 −  +  × Et rt+1 + Kt+1 rK t + 1 − Pt+1 QK t + 1 − Qt+1 PQ t + 1 QK t + 1/ wt+1 + Lt+1 wL t + 1 − Pt+1 QL t + 1 − Qt+1 PQ t + 1 QL t + 1 × Et L t + 1 − Et K t + 1 = 0

(3)

The production function is specified as a Cobb–Douglas stochastic frontier (Cornwell et al., 1990) of the form: ln Qkt = ln Xkt  + ln Zk  + ln Wkt K + kt K = 0 + ukt

A COUNTERFACTUAL DYNAMIC SIMULATION

131

where the subscripts k = 1   N and t = 1  T refer to firm and time, respectively. Xkt is a vector of inputs, Wkt is a vector of other firm characteristics, and Zk is a vector of explanatory variables, which have different effects for different firms. The unobservable effects, k , can be correlated with other explanatory variables and can interact with selected slope and intercept terms. This allows for the endogeneity of variables such as load factor and network size with respect to the firm specific statistical error. The disturbance term ukt is assumed to be an independent and identically distributed (i.i.d.) zero mean random vector with covariance matrix u . The disturbances kt are taken to be i.i.d. with zero mean, constant variance  2  , and uncorrelated with both the regressors and ukt . Total revenues can then be calculated at time t by specifying the factor market demand equation while total profits at time t can be obtained by specifying a total cost function. To close our dynamic model, we must specify the demand and cost equations. We use the approach adopted by Captain and Sickles (1997). For an alternative dynamic two-stage game for the European industry, see Roeller and Sickles (2000). First, consider the cost function. Suppose an industry in which N firms produce a differentiated output, q, using n inputs, x = x1    xn . The market demand function facing firm k at time t is of the form2 : qkt = qk pt  pmt  Yt   edt  where pmt is an index of all the other firms’ prices, Yt are the other variables (measured on the country level) shifting demand,  are unknown parameters of the demand function and edt are the disturbances. Perceived marginal revenue is PMR = pt + D1 qkt The cost function facing firm k is Ckt = Ck qkt  Wlt  Zt   ect  where Wlt is the vector of factor prices paid by firm k at time t Zt are the other industry variables shifting cost,  are unknown parameters of the cost function, and where D1 = pkt /qkt . Marginal cost is written as: MC = C1 qkt  Wt  Zt   The firm chooses optimal output where MC is equal to perceived marginal revenue in an oligopolistic industry (PMR = p in a perfectly competitive setting). Thus, the quantity-setting condition is C1 qkt  Wt  Zt   = pt + D1 pkt  pmt  Yt   edt qit  The parameter  is an index of the competitive nature of the firm. If  = 0, price equals marginal cost and the industry is perfectly competitive, while  = 1 is consistent with 2

For different forms of this model, see Bresnahan (1989).

PURVEZ F. CAPTAIN et al.

132

Nash behavior. In a price-setting game, the first-order conditions for profit maximization imply  q  qkt pt + qkt − Ckt kt = 0  qkt  pkt  pkt Summing over the N firms, we have Qt = k qkt and thus �  Ckt  qkt  Qt p + Qt − =0 q p  pt t k  kt  kt Q  pt = Ckt − t   Qt  qkt  pt The market demand function is specified as semilogarithmic, ln q = d0 + d1 P + d2 Pindex + d3 GDP + D4 GASP + d5 GCONS + d6 PRAIL + ed where q is the output of firm k P is the price of firm k Pindex is an index of the other N − 1 firms’ prices, GDP is Gross Domestic Product, GASP is the retail price of gasoline (inclusive of taxes) and PRAIL is the price of rail travel. The behavioral equation which identifies the degree of competition is P = MC − /D1 + eB . The costs are specified using the translog cost function: ln Cp q = ln a0 +

� i

ai lnpi  +

1 �� b lnpi lnpj  + bq lnq 2 i j ij

1� 1 + bqq lnq2 + b lnqlnpi  +  + ec 2 2 i qi Here, the inputs are capital (K), labor (L), and materials (M). The prices of the inputs are PK  PL , and PM , respectively. The term  contains heterogeneity controls for service and capital characteristics, which are added linearly and include the (natural logarithm) ln(average stage length), ln(network size), ln(load factor), percentage of planes that are wide-bodied, and percentage of planes that are turbo prop. Applying Shephard’s Lemma, the factor share equations are linear functions in the parameters. Since the sum of the cost shares over all equations always equals 1, and only two of the three share equations are linearly independent, for each observation the sum of the disturbances across equations must always equal zero. Linear homogeneity and symmetry are imposed parametrically. The system of five equations – translog cost, labor share, capital share, demand and behavior – are estimated by iterative nonlinear three-stage least squares, treating output price and quantity (p q), cost (C), labor share, capital share, and the price of labor (pL )

A COUNTERFACTUAL DYNAMIC SIMULATION

133

as endogenous and all others as exogenous (the standard panel data firm fixed effects has been specified in the cost equation). Endogeneity of the labor’s price is due to the strong national carrier status of the European carriers over the sample period and the use of the national carriers to pursue macroeconomic employment stabilization policies. Based on the parameter estimates obtained from these production, cost, and demand equations, the Euler equations were simulated with the Gauss–Newton algorithm in the SAS system, for optimal levels of labor, network, and fleet. Data sources are discussed in the next section.

4 DATA This study uses a panel of seven European carriers with their ticket codes: Air France (AF), Alitalia (AZ), British Airways (BA), Iberia (IB), Royal Dutch Airline, KLM (KL), Lufthansa (LH), Scandinavian Airlines System, SAS (SK), and Sabena (SN), with annual data from 1976 to 1990. The series follows these carriers during the period just following the deregulation of airlines in the US and prior to the beginning of deregulation in Europe. Network alliances in Europe were just beginning to take shape in 1989 and 1990 (e.g., the Northwest Airlines KLM alliance). These alliances have become a standard in the international airline industry (see Table 1). Our measures for system size based solely on the carrier’s physical network begins to lose validity as the alliance provides benefits of network size (passenger feed) without the accounting for the resources necessary to produce it.

Table 1 Airline Alliances in 2006 oneworld

SkyTeam

Star Alliance

Aer Lingus American Airlines British Airways Cathay Pacific Finnair Iberia LAN Chile Qantas

AeroFlot AeroMexico Air France KLM Alitalia Continental CSA Czech Airlines Delta Korean Air Northwest Airlines

Air Canada Air New Zealand Asiana Austrian bmi British Midland LOT Polish Airlines Lufthansa SAS Scandanavian Airlines Singapore Airlines South African Airlines Spanair SWISS TAPPortugal Thai US Airways United

Source: oneworld, Sky Team, Star Alliance websites May 31, 2006

134

PURVEZ F. CAPTAIN et al.

The primary source for the input, output, expense, and revenue data was the Digest of Statistics from the International Civil Aviation Organization (ICAO). This was aug­ mented using output characteristic data from IATA World Air Transport Statistics, asset valuation data from the Avmark Newsletter, purchasing power parity information from the Penn World Table, and demand data from the Organization for Economic Cooperation and Development (OECD) publication Historical Statistics. The data is sketched in this section with readers interested in reconstructing or extending this series directed to Good et al. (1993a). The data can be organized into three broad cate­ gories: inputs and expenses, outputs and their characteristics, and demand side market conditions.

4.1 Input and Expense Data The primary source for the input data was the Digest of Statistics from the ICAO. It is important to note that ICAO data is voluntarily reported rather than being an artifact of international regulatory requirement. When carriers decline to submit their information, we obtained data from alternative sources, often carrier annual reports. The model assumes airline production is a function of three inputs: labor, materials, and aircraft fleet. The labor input is an aggregate of five separate categories of employment used in the production of air travel. Included in these categories are all cockpit crew, flight attendants, mechanics, sales and promotional personnel, and other employees including general administration and aircraft and passenger handlers. Expenses for these categories included fringe benefits in addition to salaries. Quantity and implicit price indices, L and PL, were constructed based on these five subcomponents using a Divisia multilateral index number procedure (Caves et al., 1982). So that our simulations are more interpretable as number of employees, these indices have been rescaled so that the average quantity index is equal to the average number of employees. We are primarily interested in the portion of capital that comprises the carrier’s fleet. Ground-based capital is incorporated into the aggregate materials indices described later. The number of aircraft by type is obtained from the Digest of Statistics for the beginning and end of year. Our quantity measure is the average of these two values. An effective rental price for this fleet is constructed by valuing each type of aircraft at its used equipment price (the average for each year of the Avmark Newsletter), and using the Jorgenson–Hall user price formula, the carrier’s home country’s short-term commercial paper interest rate, and a declining balance depreciation schedule with a remaining asset life of 20 years. In addition, two characteristics that summarize the potential productivity of the fleet are provided: the per cent of the fleet, which is wide bodied, and the per cent using turboprop propulsion. The proportion of fleet that is wide-bodied, PWIDEB, provides a crude measure of average equipment size. We define wide-bodied aircraft as those having two aisles. It is generally accepted that there are economies of equipment size as resources for flight crews, passenger and aircraft handlers, landing slots, and so on do not increase proportionately. The per cent turboprops, PTURBO, provides another measure of the mix of capital available to the carrier. Together, our three capital variables describe both the quantity of capital and the kinds of missions they are suited to serve: turboprop aircraft are ideal for low-density short haul routes, wide-bodied aircraft ideal for high-density long haul routes, and narrow-body jets are ideal for medium-haul

A COUNTERFACTUAL DYNAMIC SIMULATION

135

routes. We should note that by the beginning of our time frame, long-haul narrow-bodied jets (1950s vintage aircraft like the Boeing 707, Douglas DC-8, SUD Caravelle and de Havilland Comets) were in the process of being phased out, and regional jets had not yet been widely adopted (the smallest jets in our sample being the BAC-111 and Fokker F28). The purchase of equipment over the study period was dominated by strong brand loyalty: SAS, Iberia, and Alitalia continued to purchase mostly Douglas aircraft while Air France, British Air, KLM, and Lufthansa continued to purchase predominately equipment from Boeing. It is important to note that Airbus was essentially a one aircraft type manufacturer (the A300) over the bulk of our study period. The A310 introduced in 1985 was essentially a modified version of the same plane. In that regard, Airbus was much more like Lockheed than it was like Douglas or Boeing. It was not until the mid-1990s with the introduction of the A320, A330, and A340 families of equipment that they spanned the range of small narrow bodied to large wide-bodied equipment and became the across the board competitors that they now are. Even given this severe limitation, they made significant inroads in European fleets. When one considers all acquisitions (purchases or leases) compared to retirements (sales, retirement, or returns to the leasing company), Airbus was able to add 127 aircraft to the fleets of these eight carriers (155–28). At the same time, Boeing was able to add net 206 aircraft (547–341). Douglas added only 6 aircraft (269–263), while there was a loss of 27 from all other manufacturers (185–212). The materials component is summarized as price and quantity indices that aggre­ gate several subcomponents. The source for expense information is ICAO’s Digest of Statistics, Financial Data. This is supplemented with either physical quantity or price information from another source to identify price quantity pairs for each material’s sub­ component at each year for each carrier. The largest component of materials is aviation fuel with price information provided by ICAO’s Regional Differences in Fares and Costs Report, under the presumption that a carrier will purchase fuel at many differ­ ent countries in the European region. Expenses for landing fees and en route traffic control facilities are paired with aircraft departures from ICAO’s Commercial Airline Traffic Series. The resulting prices can be considered rental expenses for this publicly owned capital. Expenses for carrier owned ground-based capital services are based on a Jorgensen–Hall user price using depreciated book value, for nonflight capital from ICAO’s Digest of Statistics, Financial Data, a seven per cent annual depreciation rate, and the individual carrier’s interest rate on long-term debt. The remaining materials and services including passenger food, maintenance materials, and outside services including commissions and other services are pooled into a residual materials category using the carrier’s home country purchasing power parity (Summers and Heston, 1991) from the Penn World Table Mark 5.2 as a price deflator. The price index for materials, PM, is normalized to one for the sample average and consequently the implicit quantity index, M, is normalized for average materials expenditures.

4.2 Output, Revenue, and Output Characteristics The airline services actually sold (revenue output) are based on three subcomponents: scheduled passenger and excess baggage, scheduled freight and mail, and nonscheduled

PURVEZ F. CAPTAIN et al.

136

services. Sources for revenue and physical output are based on ICAO’s Digest of Statistics, Financial Data and Commercial Airline Traffic Series. Unfortunately, data availability leaves us with aggregate revenues for small amounts of cargo services (e.g., excess baggage or charter cargo) with passenger traffic for some carriers. This aggre­ gation is carried out under the widely used convention that one revenue passenger kilometer is equivalent to 0.090 t km (or one passenger and standard baggage averages approximately 200 lbs). This has the effect of combining the three subcomponents into the same physical units, which are then aggregated using a multilateral index process normalized to an average price of 1 across the sample. Three characteristics of output are also used in our analysis. The load factor, LOADF, is the ratio of passenger output sold to total passenger output produced. In the American context, low load factors are a traditional indication the level of service is too high. Since the structure of European competition is more collusive, one might expect that load factors might be higher than optimal and that the price is too high and level of service is too low. As Figure 1 points out, trends for load factor among European carriers closely follow that for their American counter parts. Among US carriers, load factor increased from approximately 52 per cent in the beginning of our study period to roughly 67 per cent in 1990. This is as one would expect given that the European system had no competition on inter-European routes with revenue sharing, resulting in, few flights, high fares, and relatively full planes. Stage length, STAGEL, is the ratio of aircraft miles flown to aircraft departures. Typically, longer routes require fewer resources per amount of output produced. Finally, a measure of overall network size, NETSIZE, is the number of route kilometers and is provided by the International Air Transport Association (IATA) World Air Transport Statistics. NETSIZE is the only systematic measure across carriers and over time that

0.75

Load Factor (%)

0.70

0.65

0.60

0.55

0.50 76

77

78

79

80

81

82

83

84

85

86

87

88

89

Year

Figure 1 European Airlines’ Load Factor (%) (1976–1990).

90

A COUNTERFACTUAL DYNAMIC SIMULATION

137

we access from publicly available data sources. The measure is the sum of the distances for all unique routes in the carriers network. When used in an estimated equation that incorporates both lnQ and lnNETSIZE, it has the implicit effect of including network density in the model.

4.3 Demand Data Data important for describing the demand for travel was collected for the home countries of each of our carriers. A weighted sum of the three Scandinavian countries, Denmark, Sweden, and Norway, was used to represent the home country of SAS with GDP used to form the weights. The Gross Domestic Product, GDP, was obtained from the Main Economic Indi­ cators publication of the Economics and Statistics Department of the Organization for Economic Co-Operation and Development (OECD) and provides an overall scale for economic activity in the demand equation. They were reported for the above countries in billions of dollars. The OECD Economic Outlook publication Histori­ cal Statistics was the source of the growth in private consumption expenditure data. They are reported as an implicit price index with year-to-year percentage changes. The annual short-term interest rates, INTRATES, were also obtained from this publi­ cation. The rates are reported by the respective countries on the basis of the follow­ ing financial instruments: Belgium (3-month Treasury certificates), Denmark (3-month interbank rate), France (3-month Pibor), Germany (3-month Fibor), Italy (interbank sight deposits), Netherlands (3-month Aibor), Norway (3-month Nibor), Spain (3-month interbank loans), Sweden (3-month Treasury discount notes), and the UK (3-month interbank loans). The European airline industry differs from the US industry in that the continent’s small size makes autos and rail a feasible alternative to air travel (Captain and Sickles, 1997).3 Jane’s World Railways was the source of the rail data. The rail price, PRAIL, was cal­ culated as the ratio of passenger (and baggage) revenue to passenger tonne-kilometers. The retail gasoline price (prices plus taxes), PGASP, was obtained from the International Energy Agency’s publication, Energy Prices and Taxes. Finally, to capture the effects of competition from other airlines, an index of the “other” airlines’ prices was com­ puted by weighting the individual prices their respective revenue shares in the market, PINDEX. Summary statistics for different carriers/countries are provided in Table 2.

5 SIMULATION RESULTS The results of the dynamic simulation are presented in graphical form in Figures 2a–c. The simulations were run with two values for (constant capital depreciation rate), 0.12 and 0.08,  (discount factor) of 0.95, and  (rate of depreciation of past levels of capital stock to its current level) of 0.08 to solve for optimal levels of operational 3 For a discussion of the history of US airline competition and the industry’s response to deregulation see Morrison and Winston (1990) and Borenstein (1992).

Table 2 Carrier Specific Sample Mean Values for Model Variables Air France AF Inputs: L PL (000) M (000000) PM K PWIDEB PTURBO Outputs: Q (000000) P LOADF NETSIZE (000) Demand: PRAIL PINDEX GASP GDP INTRATES

34877000 21266

Alitalia AZ 18974000 29480

British Air BA

Iberia IB

KLM KL

Lufthansa LH

SAS SK

Sabena SN

45725000 17160

25127000 19666

20238000 27266

34013000 26253

17761000 31953

8971000 23253

19012 09718

7606 08738

26317 09419

12447 07955

11005 09930

17277 10629

8066 11821

5868 09388

109067 04703 01366

76667 02612 00275

172733 02628 01100

87200 01729 00326

52800 04628 00381

109533 03206 00155

91467 01418 00365

27200 02771 00217

256766 10992 06586 748210

126944 11777 06143 332511

356325 11037 06641 621877

143613 10693 06351 350562

181102 09152 06444 370913

253433 12503 06185 509879

109533 14942 06390 225556

74675 10982 06302 215300

00474 11645 06901 653549 01044

00367 11517 07995 495572 01459

00742 11696 05575 527681 01155

00254 11621 06067 224405 01348

00461 11789 06459 162287 00744

00517 11322 05474 819314 00624

00689 11169 06558 259728 00972

00369 10192 06292 108503 01035

A COUNTERFACTUAL DYNAMIC SIMULATION

120

AZ Employees (000)

AF Employees (000)

120

139

100 80 60 40 20

100 80 60 40 20 0

0 79

80

81

82

83

84

85

86

87

88

89

79

90

80

81

82

83

120

120

100

100

80 60

85

86

87

88

89

90

40 20

86

87

88

89

90

86

87

88

89

90

80 60 40 20 0

0 79

80

81

82

83

84

85

86

87

88

89

79

90

80

81

82

83

84

85

Year

Year 120

SK Employees (000)

120 100 80 60 40 20

100 80 60 40 20 0

0 79

80

81

82

83

84

85

86

87

88

89

79

90

80

81

82

83

84

85

Year

Year 120

SN Employees (000)

LH Employees (000)

84

Year

IB Employees (000)

BA Employees (000)

Year

100 80 60 40 20 0 79

80

81

82

83

84

85

86

87

88

89

90

Year

Figure 2a Actual and Simulated Numbers of Employees for Each Carrier.

PURVEZ F. CAPTAIN et al.

250 225 200 175 150 125 100 75 50 25 0

79

AZ Planes

AF Planes

140

80

81

82

83

84

85

86

87

88

89

90

250 225 200 175 150 125 100 75 50 25

0 79

80

81

82

83

250 225 200 175 150 125 100 75 50 25 0 79

80

81

82

83

84

85

86

87

88

89

90

250 225 200 175 150 125 100 75 50

25

0

79

80

81

82

83

SK Planes 81

82

83

84

85

86

87

88

89

90

84

85

86

87

88

89

90

86

87

88

89

90

79

80

81

82

83

250 225 200 175 150 125 100 75 50 25 0 79

80

81

82

83

84

85

86

87

Year

Year

SN Planes

LH Planes

250 225 200 175 150 125 100 75 50 25 0 80

85

Year

Year

79

84

Year

IB Planes

BA Planes

Year

250 225 200 175 150 125 100 75 50 25 0 84

85

86

87

88

89

90

Year

Figure 2b Actual and Simulated Number of Aircraft for Each Carrier.

88

89

90

1500

1400

1300

1200 1100 1000 900 800 700 600 500

400

300

200

100

0 79

AZ NetSize (000 km)

AF NetSize (000 km)

A COUNTERFACTUAL DYNAMIC SIMULATION

80

81

82

83

84

85

86

87

88

89

90

141

1500

1400

1300

1200

1100

1000

900

800

700

600

500

400

300

200

100

0

79

80

81

82

83

1500

1400

1300

1200

1100

1000

900

800

700

600

500

400

300

200

100

0 79

80

81

82

83

84

85

86

87

88

89

90

1500

1400

1300

1200

1100

1000

900

800

700

600

500

400

300

200

100

0

79

80

81

82

83

81

82

83

84

85

85

86

87

88

89

90

84

85

86

87

88

89

90

Year

SN NetSize (000 km) 80

86

87

88

89

90

1500

1400

1300

1200

1100

1000

900

800

700

600

500

400

300

200

100

0

79

80

81

82

83

84

85

86

Year

Year

SN NetSize (000 km)

LH NetSize (000 km)

Year 1500

1400

1300

1200

1100

1000

900

800

700

600

500

400

300

200

100

0 79

84

Year

IB NetSize (000 km)

BA NetSize (000 km)

Year

1500 1400 1300 1200 1100 1000 900 800 700 600 500 400 300 200 100 0 79

80

81

82

83

84

85

86

87

88

89

90

Year

Figure 2c Actual and Simulated Network Sizes for Each Carrier.

87

88

89

90

142

PURVEZ F. CAPTAIN et al.

variables during the time period 1979–1990.4 The chosen parameters were consistent with industry estimates. Solid lines indicate sample observations, small dashed lines indicate simulations for = 0.12, and large dashed lines for = 0.08. The simulation exposes Sabena, Alitalia, and Iberia as the carriers least primed for a deregulated airline market. Both Iberia and Alitalia have recently flirted with bankruptcy. Moreover, Iberia has also had a recent spate of what would appear to be predatory pricing, pushing partners like Viasa into bankruptcy. Sabena sold a large minority position to Air France in 1993. But shortly thereafter Air France, itself struggling mightily, divested its interest, and Swiss Air bought a 49 per cent interest in Sabena in 1995. Swissair liquidated in 2001 partially because it was unable to halt Sabena’s trail of red ink.5 Iberia struggled with poor management and financial performance until it privatized by selling 49 per cent, including a 9 per cent stake to British Airways in 1999. Alitalia foundered in the twenty-first century, undergoing major restructuring despite compacting with Air France–KLM (itself an agreed-upon acquisition by Air France, creating the world’s largest airline by revenues). It remains to be seen if Alitalia can return to profitability; more broadly, if smaller, state-owned carriers can survive in a unified European market. To analyze the results of the simulation in-depth, the airline market in Europe should be divided according to scale of operation. Air France, British Airways, and Lufthansa were larger with similar scales of operation, while Alitalia, Iberia, Sabena, and SAS were smaller. The airlines with levels closest to the simulation results were best prepared for the competitive milieu ahead. The main stylized fact from the simulation was that the larger carriers were better prepared for deregulation than the smaller ones. All airlines, excepting Lufthansa, employed too few workers. At first blush, grow­ ing the workforce hardly seems the way to maximize profits. A possible justification relates to powerful labor unions negotiating wages above competitive levels reducing employment below optimal levels (Captain and Sickles, 1997; Good et al., 1993b). The McGowan and Seabright (1989) study evinced this phenomenon, finding labor costs for many European carriers to have been more than double the US rates.6 The simulation solution for fleet size suggested Air France, British Airways, Lufthansa, and SAS possessed a sizeable fleet relative to the optimal solutions, at times even exceeding the values. Conversely, Iberia and Sabena purchased few or no planes during the period studied, but should have purchased more. As for network size, Lufthansa operated at levels close to optimal. However, for all other airlines, networks were suboptimal because they were too small over much of the sample period. Increasing the size of the network, ceteris paribus, lowers the total

4

A note about the solutions: the solutions predict optimal levels of the operational variables with the assumptions that planes, people, and networks can be increased and decreased without costs. 5 See http://www.sabena.com/EN/Historique_FR.htm 6 As pointed out by a referee, staffing of flight personnel is based on regulatory requirements for particular aircraft types. To address this further institutional fact, we could have allowed labor also to be quasi-fixed but this would add substantial complexity to an already complex modeling scenario. Our labor input is an aggregate of five separate categories of employment used in the production of air travel. Our finding that there is generally understaffing is consistent with the need for European carriers to expand their operations and thus their labor requirements in general. Our model is not detailed enough to point to specific classes of labor that should expand nor is it detailed enough to allow differentiation of demand for own and outsourced labor and/or endogenous wage outcomes of union/firm negotiations.

A COUNTERFACTUAL DYNAMIC SIMULATION

143

cost of the airline. This is the sine qua non of operating a viable airline, for the following reasons: Hub-and-spoke operations allow airlines to concentrate traffic on certain routes, allowing both larger, more efficient planes and more frequent service. In addition, hub-and-spoke operation allows for a greater range of destinations and city-pair combinations to be served, including city-pair combinations, which would not normally generate enough traffic to justify a regular service. The addition of a new spoke to a hub-and-spoke network signif­ icantly increases the city-pair combinations served by the network, at minimal additional cost (OECD, 2000).

Achieving scale economies has to be through alliances because outright acquisition is largely proscribed by further restrictions on foreign ownership (Staniland, 1996). Integrating networks through alliances offers efficiency gains from passenger pooling agreements and fungible airport gate and slot rights. Large networks also exploit cost advantages, as airlines discard linear route systems in favor of the hub and spoke network configuration. This adjustment derives economies of density, and higher load factors on spoke routes radiating from the hub (OECD, 1988). Without alliances, deregulation in Europe has the effect of reducing load factors drastically, as it did in the US failing an acquisition or alliance, and deregulated markets can sink an airline, for example, Pan-Am (Brueckner, 2003; Brueckner and Whalen, 2000; Levine, 1987).78 As an example of Europe’s first intercontinental alliance, in December 1986, British Airways, with its equity wiped out by a debt burden reaching over £1 billion at one time, was sold to the private sector, thus joining Swissair as the only privately owned airlines at the time. To stave off its declining profitability, BA signed an alliance with United Airlines. The agreement integrated United’s flight schedules and networks in America with BA’s transatlantic services to American cities. The agreement enabled the airlines to share passengers and increased the quality of service for time conscious (and high margin) business travelers. As noted in the simulation, 1988 was a watershed year for British Airways, as privatization quickly resuscitated the airline. Other airlines followed suit and formed alliances to brace themselves for the onset of competition, learning from the experience of American deregulation. Despite deregulating, barriers remain in the aviation sector. The march towards com­ plete deregulation in both the US and Europe is hindered by three factors: (1) limitations to existing “open sky” agreements, (2) ownership restrictions, and (3) and barriers to entry. While “open skies” means increased international competition, domestic markets

7 For further research on US Domestic codesharing that closely parallels the experiences of intra-European codesharing, see Ito and Lee (2005) and Bamberger et al. (2004). 8 Substantial variation in the dynamic simulations occurs because the Euler equations are highly nonlinear. We didn’t feel “adjustment factors,” such as those commonly used in dynamic nonlinear forecasts from large macro models (e.g., the WEFA Quarterly Forecasts), were appropriate since they are difficult to justify on any other than ad hoc grounds. That said our results make economic sense because they point out that most European airlines suffered in their ability to maximize the present value of discounted profits because their networks and operational capacity were too limited during the period we studied. European airline networks (excluding those for carriers that exited the industry) expanded substantially after accelerating industry reforms that began around 1990. Lost profits for many of the European airlines in our sample appeared to be most pronounced during the early and middle part of our sample period and by in large were trending toward equilibrium at the end of the 1990s.

144

PURVEZ F. CAPTAIN et al.

remain off-limits to foreign carriers. In other words, British Airways cannot fly from New York to Los Angeles, even as a continuing flight from London (Staniland, 1996). US law dictates that foreign citizens may not own more than 25 per cent of voting stock, and Europe permits no more than 49 per cent foreign ownership (Economist, 2005). Lastly, landing rights and gates at airports often are not traded freely, preventing access for new entrants (Captain, 1993). Further liberalization in these areas is needed to attain more perfect competition (Postert and Sickles, 1998). The expressed concern during the early liberalization talks was that the rush to acquire and ally could lead to the development of mega-carriers that would dominate the market – a reversion to oligopoly, without the stability needed from a vital transportation service. Nearly 10 years removed from 1997, the three factors – “open skies” or lack thereof, ownership restrictions, and barriers to entry – still impede full deregulation. Taken together with firm anti-trust laws in Europe, a reversion to oligopoly is an improbable outcome.

6 CONCLUSIONS, LIMITATIONS AND FUTURE RESEARCH This chapter has focused on an integrated dynamic model of the European airline industry. We use the dynamic structural model to examine the extent to which the European industry allocated its factor inputs during the period 1979–1990, beginning with US airline deregulation and ending with the period of transition to deregulation of carriers in the European Union in keeping with a goal of long-run profitability. We have allowed for a fairly rich menu of strategic decision-making among the carriers and for relatively general production and cost structures. Our findings point to several sources of forgone profits, in particular, the need for European carriers to adopt policies for expansion of their networks. This would allow them to take advantage of returns to density by expanding and reconfiguring their networks and were realized in the years subsequent to our after the sample period, in part by forming the alliances summarized in Table 1. Interestingly, just these sorts of changes characterized the competitive policies undertaken by European carriers in their code-sharing agreements and in their oftenbitter union confrontations as the carriers transitioned from national flag carriers to competitive international companies. This chapter presented a methodology and modeling approach that can be used in other settings to better understand the potential impacts of regulatory changes in an industry. As with any such new approach to study such an issue, our model does have limitations. For example, the use of relatively simple functional forms such as the Cobb–Douglas imposes a degree of substitutability that might exaggerate the swings in our dynamic and may be a reason for such temporal patterns in our simulations. Another limitation is that we applied this model in the European context where there are data limitations and inconsistencies in reporting protocols across time. These are more severe than with US data from the Department of Transportation. Future work could focus on utilizing our methodology and modeling approach for US, Canadian or Australian carriers where prior regulation made for more extensive and consistent data. To that end, one might be

A COUNTERFACTUAL DYNAMIC SIMULATION

145

able to use better measures of network size, such as cities served, or measures of network structure. Our methodology also places an increasing burden of complexity for adding more details – adding more control variables implies adding more Euler equations.

REFERENCES Balassa, B. (1985). European Economic Integration. New York: North Holland Publishing. Bamberger, G., D. Carlton, and L. Neumann (2004). “An empirical investigation of the competitive effects of domestic airline alliances,” Journal of Law and Economics, XLVII, 195–222. Bannerman, E. (2002). The Future of EU Competition Policy. London: Center for European Reform, pamphlet. Berry, S. (1992). “Estimation of a model of entry in the airline industry,” Econometrica, 60, 889–917. Borenstein, S. (1992) “The evolution of U.S. airline competition,” Journal of Economic Perspec­ tives, 6(2), 45–73. Bresnahan, T. (1989) “Empirical studies of industries with market power,” Handbook of Industrial Organization Volume 2, edited by Schmalensee, R., and Willig, R., Amsterdam/New York: North Holland Publishing Company. Brueckner, J. (2003): “International airfares in the age of alliances: The effects to codesharing and antitrust immunity,” Review of Economic Statistics, 85, 105–118. Brueckner, J., and T. Whalen (2000): “The price effects of international airline alliances,” Journal of Law and Economics, 43, 503–545. Button, K. (1990). Airline Deregulation: International Experiences. New York: New York Uni­ versity Press. Button, K. (2003), Recent Developments in Transport Economics, edited by Kenneth Button, Northhampton, MA: Edward Elgar Publishing. Button, K. J. and K. Johnson. (1998) “Incremental versus trend-break change in airline regulation,” Transportation Journal, 37, 25–34. Captain, P. (1993) “Competition and efficiency in the European airline industry,” Unpublished PhD dissertation, Rice University. Captain, P., and R. Sickles (1997) “Competition and market power in the European airline industry: 1976–1990,” Managerial and Decision Economics, 18, 209–225. Caves, D., L. Christensen, and W. Diewert. (1982) “Output, input and productivity using superla­ tive index numbers,” Economic Journal, 92, 73–96. Cornwell, C., P. Schmidt, and R. C. Sickles (1990). “Production frontiers with cross-sectional and time-series variation in efficiency levels,” Journal of Econometrics, 46, 185–200. Economist, “Half-open skies,” November 24, 2005. Evans, D. (1987a). “Tests of alternative theories of firm growth,” Journal of Political Economy, 95, 657–674. Evans, D. (1987b) “The relationship between firm growth, size and age: Estimates for 100 manufacturing industries,” Journal of Industrial Economics, 35, 567–580. Good, D., I. M. Nadiri, L-H Roeller, and R. C. Sickles (1993a). “Efficiency and productivity growth comparisons of European and US air carriers: A first look at the data,” Journal of Productivity Analysis, special issue edited by J. Mairesse and Z. Griliches, 4, 115–125. Good, D., L-H. Roeller, and R. C. Sickles (1993b) “U.S. Airline Deregulation: Implications for European Transport,” Economic Journal, 103, 1028–1041. Hall, B.H. (1987) “The relationship between firm size and firm growth in the U.S. manufacturing sector,” Journal of Industrial Economics, 35, 583–606.

146

PURVEZ F. CAPTAIN et al.

Hotz, J.V., F. E. Kydland, and G. L. Sedlacek (1988). “Intertemporal preferences and labor supply,” Econometrica, 56, 335–360. International Air Transport Association (various years). World Air Transport Statistics Montreal: IATA. International Civil Aviation Organization (various years). Digest of Statistics, Montreal: ICAO Ito, H., and D. Lee (2005). “The impact of domestic codesharing on market airfares: Evidence from the U.S.,” in Advances in Airline Economics, Volume 1, Competition Policy and Antitrust, edited by Darin Lee, Amsterdam: Elsevier. Jovanovic, B. (1982). “Selection and the evolution of industry,” Econometrica, 50(3), 649–70. Levine, M. E. (1987). “Airline competition in deregulated markets: Theory, firm strategy, and public policy,” Yale Journal on Regulation, 4, 434–36. McGowan, F. and P. Seabright (1989). “Deregulating European airlines,” Economic Policy, 9, 284–343. Morrison, S. and C. Winston (1990) “The dynamics of airline pricing and competition,” American Economic Review Papers and Proceedings, 80, 389–393. OECD (1988). Deregulation and Airline Competition, Paris: OECD. OECD (2000). “Airline mergers and alliances,” OECD Journal of Competition Law and Policy, 2(2), 122–220. Pakes, A. and R. Ericson (1998). “Empirical implications of alternative models of firm dynamics,” Journal of Economic Theory, 79, 1–46. Postert, A. and R.C. Sickles (1998). “Air liberalization: the record in Europe,” in Taking Stock of Air Liberalization, edited by M. Gaudry and R. Mayes, Boston: Kluwer Academic, 39–59. Roeller, L-H., and R. C. Sickles (2000). “Capacity and product market competition: Measuring market power in a ‘puppy-dog’ industry,” International Journal of Industrial Organization, 18, 845–865. Sampson, A. (1984). Empires of the Sky (The Politics, Contests and Cartels of World Airlines). London: Hodder and Stoughton. Schipper, Y., P. Rietveld, and P. Nijkamp. (2002) “European airline reform,” Journal of Transport Economics and Policy, 36, 189–209. Sickles, R. C. and Williams, J. (2006) “A intertemporal model of rational criminal choice, panel data econometrics: Theoretical contributions and Empirical Applications,” edited by Badi Baltagi, Elsevier Science, Amsterdam, 135–166. Sickles, R. C. and A. Yazbeck (1998) “On the dynamics of demand for leisure and production of health: Evidence from the retirement history survey,” Journal of Business and Economic Statistics, 16, 187–197. Staniland, M. (1996) “Open skies – fewer planes?: Public policy and corporate strategy in EU–US Aviation Relations, European Policy Papers #3, European Union Center, University Center for International Studies, University of Pittsburgh. Summers, R. and A. Heston (1991). “The Penn World Table (Mark 5): An expanded set of international comparisons, 1950–1988,” Quarterly Journal of Economics, 106(2), 327–368. Taneja, N. (1988). The International Airline Industry: Trends, Issues and Challenges. Lexington, MA: Lexington Books. Williams, G. (1994). The Airline Industry and the Impact of Deregulation, 2nd Edition, Avebury Aviation Ashgate Publishing Limited, UK: Aldershot.

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

6 State Aid to European Airlines

A critical Analysis of the Framework and its Application∗

Pietro Crocioni† and Chris Newton‡

ABSTRACT This chapter provides a critical assessment of European state aid policy as applied to the airline sector. Within the policy framework, competitive distortions and negative spillovers between Member States are typically presumed to result from the granting of aid rather than being the subject of analysis. Whilst such an approach might be justified where the central concern of state aid policy is to prevent national governments from favoring their own flag carrier airlines, it is less likely to be appropriate in relation to the types of state aid case that have begun to emerge in the sector in recent years. The application of state aid rules in the Charleroi decision illustrates the potential weaknesses in the current approach.

1 INTRODUCTION European airline markets were gradually opened up to competition during the 1990s by allowing entry and competition on intra-European Union (EU) and domestic routes. Competition on many routes has increased and liberalization would appear to have had generally beneficial effects on fares, service quality, and choice. Market entry, especially from low cost airlines, has also resulted in new services being created and increased use of secondary or regional airports. With liberalization, competition (antitrust under the US terminology) law has been increasingly applied to the airline sector. Whilst the mainstay of antitrust analysis is the ∗

We wish to thank Darin Lee and Claudio Piga for useful comments. However, the content of this article reflects only the opinion of the authors, who are solely responsible for any remaining errors. † Senior Economist, Chief Economist Team, Office of Communications (Ofcom), UK; e-mail: [email protected]. ‡ Director, Frontier Economics, UK; e-mail: [email protected].

148

PIETRO CROCIONI AND CHRIS NEWTON

competitive distortions that can arise from mergers or the conduct of firms with market power, the granting of subsidies by national Governments to specific firms can also raise significant concerns. In the light of this, the European Commission is given the responsibility to control subsidies (termed state aids) granted by EU Member States and ensure that such subsidies do not serve to distort competition. Such subsidies have been a particular concern in the airline industry given the prominent position of flag carriers and the potential incentive on the part of Member States to promote the interest of their own flag carrier. This contribution provides a critical assessment of the state aid framework in the Treaties of the European Communities and the European Commission’s state aid policy in the airline sector, examining the rationale for supranational control of state aid and identifying a number of possible concerns with the existing approach. In particular, we argue that, whilst a primary justification for supranational control is the potential for state aid to generate negative externalities or distort competition across Member States, under the current approach little emphasis is placed on identifying whether an aid is in fact likely to distort competition or generate such negative externalities. In order to highlight these issues and concerns, we provide an overview of state aid decisions by the European Commission in the airline sector. We consider the state aid cases to failing airlines in the 1990s and also focus on the more recent Charleroi decision and briefly examine its implications. This decision is of particular interest because it concerned an alleged subsidy to an entrant airline rather than a state-owned incumbent.

2 AIR TRANSPORT SECTOR AND STATE AID IN EUROPE Until the mid 1980s, air transport was effectively exempt from the application of EU competition law including the state aid rules (Adkins, 1994; Starkie, 2002). Following three packages of measures implemented in 1988, 1990, and 1993, the EU airline industry gradually moved away from the system of bilateral agreements that still regulates most international airline markets. It is only as this process of deregulation progressed that the European Commission (henceforth Commission) started to apply competition policy and the state aid rules to the sector. Prior to deregulation, international routes within the EU were served by one carrier designated by each of the two Member States at each end of the route. This meant that each route was served by a maximum of two flag carriers with airfares being regulated by bilateral agreements. Domestic routes were also often the sole preserve of the flag carrier. This is in stark contrast with the current situation where fares, capacity, route access, designation, and licensing of airlines are all, to a great extent, deregulated – with full deregulation having come into force in 1997. In response to deregulation, flag carriers reorganized themselves into hub-and-spoke networks. At the same time, new point-to-point low cost carriers emerged whose oper­ ations were based on the business model used by Southwest Airlines in the US. These carriers achieved cost savings by simplifying their organizations and logistics, using secondary airports (with lower taxes and landing and handling fees), cutting out travel agent commissions by, for example, distributing via the Internet and avoiding some of the “legacy costs” (e.g., high manpower costs) faced by flag carriers.

STATE AID TO EUROPEAN AIRLINES

149

Low cost carriers first emerged in the early 1990s in the UK and Ireland with carriers like Ryanair (set up in 1985 but relaunched in 1991), Easyjet (founded in 1995), Go and Buzz. These initially established themselves at secondary UK and Irish airports before expanding, together with other carriers, into airports in continental Europe in recent years. The impact of new entry and increased competition has been particularly evident in the UK where the market has expanded significantly (Gil-Molto and Piga, 2005). Figure 1 shows how UK all passenger growth has been higher in the 1990s than previously (the fall in 1991 was due to the first Gulf War and that in 2001–2002 to the 9/11 terrorist attacks). In particular, while between 1982 and 1990 air passengers have increased on average by just less that 5.5 million a year, and this grew to 9.2 million a year between 1991 and 2000. State aid to European carriers reached its highest point in the mid-1990s and has remained relatively low since. Figure 2 shows that total state aid across all sectors has been declining over the 1990s and has followed a trend different to that followed by the levels of aid to the airline sector. The increase in state aid to airlines in 1994–1997 was largely due to the combined effects of a drop in air traffic following the first Gulf War and the restructuring process that most flag carriers went through in response to or ahead of deregulation. The high levels of state aid that were seen in the mid-1990s were, however, unlikely to have led to significant distortions to competition (although they may have impacted on economic efficiency) since, at this time, the recipient flag carriers faced no or only limited competition and entry was in most cases not allowed in the routes where they operated. The growing number of carriers active in each route and the more intense competition that has emerged in recent years mean that the provision of state aid (granted to either flag carriers or new entrants) now has substantially greater potential to distort competition – even if the actual level of aid granted is lower.

200 180 160

(Million)

140 120 100 80 60 40 20 54 56 19 58 19 60 19 62 19 64 19 66 19 68 19 70 19 72 19 74 19 76 19 78 19 80 19 82 19 84 19 86 19 88 19 90 19 92 19 94 19 96 19 98 20 00 20 02 19

19

19

52

0

Source: Department for Transport.

Domestic passengers are counted both at airport of departure and arrival.

Figure 1 Terminal Passengers UK Airports (All flights).

PIETRO CROCIONI AND CHRIS NEWTON

150

Air Transport

Total

3,000

80,000 70,000

Total Aid

2,500

60,000 2,000 50,000 1,500

40,000 30,000

1,000

Aid to Air Transport 20,000

500 10,000 0

0 1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

Source: European Commission, DG Competition.

Figure 2 State Aid in the European Union (million E).

3 STATE AID POLICY IN THE EUROPEAN UNION 3.1 The Framework The original Treaty of the European Communities contains among its competition law provisions not only general antitrust rules but also a set of rules specifically aimed at limiting the amount of subsidies and assistance granted by Member States to their national firms. These provisions arose from fears that the European Community could witness subsidy competition between Member States reminiscent of the beggar-thy­ neighbor industrial policies that afflicted Europe in the 1930s. The Commission was therefore assigned a supervisory role over the decisions by Member States to grant state aids. Article 87(1) of the Treaty provides a general prohibition on state aid. This apparently very strict policy does, however, in practice allow state assistance to be provided to firms. Firstly, a measure is only defined as state aid if it satisfies a number of criteria. It must confer a benefit or an advantage to one or more firms – i.e., be selective1 – and be granted by a state or through state resources. Furthermore, only state aid that “distorts or threatens to distort competition (    ) insofar as it affects trade between Member States” is prohibited. In the practical application of the rules, where measures are found to meet the definitional criteria – i.e., are selective and granted through state resources – there is generally a presumption that they distort competition and affect trade. Secondly, under Articles 86, 87(2), and 87(3), there is scope for aid to be declared “compatible with the common market” where it is aimed at achieving a number of

1

General measures, such as general tax exemptions, are not classified as state aids.

STATE AID TO EUROPEAN AIRLINES

151

economic and, more often, social objectives. For example, Article 86 provides scope for exemption for firms that provide “services of general economic interest” (SGEI), and this has often been used to exempt aid to firms entrusted with public services obligations (PSO) – subject to a number of conditions being met. Articles 87(2) and (3) list further cases where aid is or could be deemed compatible. The Commission has issued a number of guidelines that set out the types of aid that in principle may be exempt from this general prohibition and the circumstances under which such exemptions are likely to apply. Guidelines cover three broad categories of state aid: horizontal (Research and Development (R&D), environmental protection, small- and medium-sized enterprises (SMEs), training, employment programs, rescue and restructuring, risk capital and undertakings operating in deprived urban areas), sectoral (steel, car, shipbuilding and synthetic fibers), and regional. The Commission generally takes a negative view of granting state aid to ailing firms, given the perverse incentives that this generates in terms of dynamic efficiency and the higher likelihood of generating competitive distortions. However, the Commission is willing to approve aid to ailing firms in certain circumstances – either for social or for economic reasons – as long as (i) distortions are limited, (ii) the aid measures form part of a coherent restructuring plan aimed at restoring firms to financial health, and (iii) the level of aid does not go beyond that needed to achieve this objective.2 In the past, concerns have arisen in relation to hidden or implicit state aid granted to state-owned firms. As a result, the “Market Economy Investor Principle” (MEIP) was introduced and originally confined to state-owned undertakings, though more recently its application has been extended beyond this. The MEIP holds that if a government were to invest on the same basis as the private sector would, under the same circumstances, this would not amount to state aid. This has often been criticized.3

3.2 The Economics of State Aid State aid can be used as an effective tool for remedying market failures and, for example, ensuring the provision of public goods or correcting externalities. For example, firms’ decisions on where to locate generate external benefits and costs which might differ across locations, but the firms’ private decision might fail to reflect these.4 However, even in those circumstances when aid appears justified on such grounds, it can nevertheless generate negative outcomes due to the potential to create significant distortions to competition. The fact that the provision of aid can have both positive and negative effects is implicitly recognized in the framework of the Treaty that involves a general prohibition on aid combined with exemptions when aid is targeted at meeting particular objectives.

2

European Commission, Community guidelines on State aid for rescuing and restructuring firms in difficulty,

OJ C 244, 01.10.2004, pp. 2–17 (Rescue & Restructuring Aid Guidelines henceforth).

3 Nicolaides and Bilal (1999) claim that it is wrong to expect that a government should obtain a financial

return similar to that of any private investor. If this was the case, there would be no reason to grant a subsidy

as private investors would supply capital. On the other hand, if the measure corrects a market failure, returns

can be expected to be lower because of the resulting welfare benefits.

4 There could be production externalities where firms fail to appropriate all the benefits of their production,

for example, with R&D. There could also be agglomeration externalities which might arise when physical

proximity increases the efficiency of all firms in a particular market.

152

PIETRO CROCIONI AND CHRIS NEWTON

It should be borne in mind, however, that although the presence of market failure may be viewed as a necessary condition for justifying state aid, it is not a sufficient condition. State aid can often be a relatively indirect way of addressing market failure and, as such, may not be the most efficient way of addressing it (Nicolaides and Bilal, 1999, p. 104). Ideally, the first best policy is to address market failure directly, and only when it is not feasible to do so, state aid could be considered as the second best. First best policies are very specific and therefore generate minimal distortion. For example, there are negative externalities from congestions at major airports. Airports and airlines might not take into account the negative externalities (noise and pollution), which affect residents in surrounding areas. The Commission policy for addressing congestion concerns at main airports has among its tools the use of subsidies to attract airlines to smaller regional airports. In this case, state aid is unlikely to be a first best policy tool. The most effective way to address that is by internalizing it – i.e., through a surcharge on airport landing charges in congested airports. There are also a number of “non-economic” justifications for state aid, reflecting social or public policy concerns rather than market failures in an economic sense. For example, state aid may be used to redistribute income or for providing universal services.5 In the case of airlines, state aid could, at least in principle, be granted in order to enable services to be run on “thin” routes that would otherwise not be economic to serve. This might be justified on the grounds of market failure – i.e., the airlines could fail to internalize the benefits to the local economy – and/or social grounds – i.e., geographically marginal areas should not be left isolated.6 Despite the presence of potential justifications, state aid can harm efficiency and distort competition. State aid tends to reduce dynamic efficiency because it softens the recipient’s budget constraint as firms that could reasonably expect to receive state aid will have reduced incentives to achieve productive efficiency. It could also provide incentives for firms to invest in wasteful rent seeking rather than invest in productive activities. There are also well-known risks in terms of potential distortion of competition. One can distinguish between two main types of distortionary effects that might coexist, as recognized by Article 87 itself, which is concerned with any state aid scheme that distorts competition “insofar as it affects trade between Member States.” A state aid could first distort competition in the country where it is granted. Therefore, state aid could generate serious inefficiencies for the economy as a whole and if granted to a firm or firms with market power it might also distort competition in the market(s) in which the firm(s) operate(s). Second, when state aid granted by one government distorts competition beyond its national borders – i.e., the market is wider than national – it could generate negative spillovers affecting countries other than the one where the aid was granted. This is a critical consideration and justification for a supranational system of state aid control, as discussed in the next section. 5 However, as in the case of aid aiming at correcting market failure, state aid might not be the most efficient or optimal way even if the purpose is not an economic one. For example, in order to ensure that a service is universally available, it might be more efficient to grant a subsidy to consumers that otherwise would not be supplied and that can therefore select their provider, rather than granting aid to a specific firm. 6 This is, however, potentially prone to abuse and be used to create legal entry barriers. This was allegedly the case of the Italian government’s recent decision to impose PSO on routes between Sarninia and the mainland that some argue are commercially viable.

STATE AID TO EUROPEAN AIRLINES

153

3.3 The Rationale for a Supranational Control State aid might lead to two types of inefficiencies, often but not always coexisting. It might distort competition in the country of the recipient firm, and in some cases, this distortion could also extend to other countries. Both could justify the need for regulatory oversight; however, there is a significant difference between the two. In the case of a domestic-only distortion, there is an implicit assumption of regulatory failure at the national or local government level. This could arise because the government fails to internalize the distortion of competition when granting a state aid. This could happen, for example, because the government is captured by interest groups or because of political economy considerations.7 The critical question in this case is whether the national government or a supranational authority would be better placed to avoid this type of regulatory failure.8 The case for a supranational control is not clear-cut. The latter could perhaps, but not necessarily, be less influenced by interest groups but at the same time might have reduced information about the existence or extent of national or local market failure that the state aid is meant to address. Therefore, national government might still be better placed at addressing national or local externalities and at selecting the optimal tool.9 A stronger rationale for a supranational control of state aid is the presence of negative externalities which national government could fail to internalize because they negatively affect consumers in foreign countries. However, there is an even more serious concern based on a stream of economic literature known as “Strategic Trade” which claims that countries might have an interest to grant aid in order to appropriate rents that arise in imperfectly competitive markets. While this might be in the “private” interest of individual countries, it might not be in the “public” interest of the EU as this aid is expected to lower overall welfare because it imposes negative externalities on other countries.10 This literature concludes that each government has an incentive to grant subsidies to firms located in their jurisdiction when the market is imperfectly competitive potentially leading to a subsidy war/race with all countries being worse-off as a result.

7 A state aid confers a large benefit to the recipient(s) who therefore have a strong interest in the decision. However, the state aid might have to be financed by a (distortionary) tax and therefore have a negative impact on all consumers. While the benefits from the aid are concentrated, the negative impact on consumers is usually diffused. Therefore, even if overall the costs outweighed the benefits, the former might not be large for each consumer who therefore has a reduced incentive in influencing the state aid decision. 8 This question was also recently raised by the former Commission Chief Economist Röller, see Friederiszick et al. (2006). 9 Besley and Seabright (1999) argue that the ability to cast a sceptical eye on Member States’ unrealistic judgments on how to use public funds should not be lightly dismissed. Besley and Seabright (pp. 16–18), also refer to economic geography as providing further justification for national or local government granting subsidies to correct local market failures. Governments compete to attract firms because their decision might have important external effects in the country where they locate. The magnitude of these externalities depends not just on the type of activity but also on where it occurs however, unless government intervened firms would fail to internalize these effects. The basis for government intervention is based on efficiency, but it could also have an equity justification. 10 The main model on which this strand of literature is based on is Brander and Spencer (1985). See also a review and criticism by Corden (1990). The conclusions of the strategic trade literature are not strong to changes in assumptions. For example, if firms play Bertrand with differentiated products rather than Cournot, domestic welfare is increased by the government imposing an export tax rather than subsidy.

154

PIETRO CROCIONI AND CHRIS NEWTON

The implication is that a strict control on state aid is desirable. The important insight is that the distortive impact of state aid can only be assumed when there is evidence pointing to the aid inflicting a strong negative externality on other countries because markets are imperfectly competitive and their geographic scope spans beyond national boundaries. The basic setting of the “Strategic Trade” literature has firms located in separate countries engaging in Cournot competition in a supranational market. Governments compete with each other, shifting rents that arise in imperfectly competitive markets. The tool used to compete is the granting of an export subsidy.11 The type of subsidy assumed has the important consequence that an action by a government is assumed to produce no net benefit in terms of productive efficiency to its domestic consumers – i.e., no increase in domestic output. However, it inflicts a negative externality on other countries and therefore has a negative impact on overall welfare. An export subsidy allows the recipient to capture a greater share of a slightly smaller total industry profit and reduces overall welfare but it also alters its distribution in favor of the recipient of the aid. Recent work (Collie, 2003) extends the conclusions of the “Strategic Trade” literature to the case of a production rather than export subsidy. This is a more benevolent measure as it also expands domestic output of the recipient firms, thus offsetting the domestic distortion from imperfect competition.

3.4 Problems with Current Framework The above discussion shows that in some circumstances national governments should be left with the freedom to grant state aid not only for social or equity but also for economic reasons. The key question for a supranational authority such as the Commission is under which circumstances it should intervene. A supranational authority could either ban state aids that generate substantial cross-border negative spillovers or only do so when such a spillover is larger than any benefits from remedying a domestic market failure.12 At present, state aid policy operates a number of presumptions in order to assess whether competition is likely to be distorted and/or a negative spillover can be expected. This, however, is an imperfect substitute for the rigor, which is required by other provision of competition law, namely those applied to firms rather than Member States. The differ­ ences in approach between state aid and competition law are particularly important in three areas: market definition, distortion of competition and assessment of effect on trade. In principle, in order to assess the potential distortion of competition that a state aid measure might generate, it could be sensible to adopt a similar approach to that followed in abuse of dominance or merger cases. This would require an analysis of the relevant market. State aid can distort competition in one or more markets – either final or input products markets where the recipient of the aid operates. Logically, it would appear useful to define the relevant product and geographic markets. However, for state aid,

11

An export subsidy is a direct or indirect compensation provided by government to private commercial firms to promote exports of domestic products. It makes foreign competitors react less aggressively. 12 This has led some commentators to propose a system based on a welfare assessment whereby state aid should perhaps be allowed when the effect on domestic welfare is expected to be positive and outweigh the negative spillover effect (Nicolaides and Bilal, 1999, p. 102).

STATE AID TO EUROPEAN AIRLINES

155

this is currently not the case, as markets are identified not on the basis of economic analysis but on product classification used for statistical purposes and covering implicitly the whole EU or the European Economic Area (EEA).13 The extent of the geographic market appears an especially critical first step in assessing the impact on trade. Limited or no market or competitive analysis is undertaken, and the likely effects of the subsidies on the position of the firm(s) are rarely analyzed. This would be critical to assess the degree of market power and, hence, the potential impact of the state aid. The impact in turn will depend on the type of state aid. The current policy of the Commission rests heavily on the distinction between generic (general measures) and ad hoc state aids, which selectively favors some firms. This is based on the presumption that generic aids are more likely to be targeted at genuine market failures rather than consist of attempts to strategically shift rents from other countries. Indeed, a generic aid scheme, because it allows all firms to be eligible if they meet the preset conditions, is an ineffective and expensive instrument to shift rents from other countries.14 A full market or competitive analysis would lead to better analysis and policy decisions. State aid granted to firms in a competitive market is unlikely to significantly distort competition. However, in imperfectly competitive markets, state aid can be distortionary. First, the more concentrated a market, the more likely a selective state aid is to affect a larger proportion of output (Garcia and Neven, 2004). Second, when the recipient has market power, state aid could facilitate anticompetitive abuse. The competition distortion also depends on the type of state aid. State aid that affects the firm’s marginal or variable cost and therefore its pricing is most likely to distort competition.15 This is recognized by the current practice of prohibiting aid that affects operating costs. However, state aid may affect or harm competitors even when it does not affect pricing directly (Möellgaard, 2004). The harm could arise because the recipient could use the aid to invest, for example, in R&D and become able to provide a higher quality product. Competing firms will be harmed and forced to reduce their price, output and investment. These considerations are important as most state aids do not directly affect marginal costs. Furthermore, state aid might affect entry and exit decision by both recipients and competitors. One would expect the analysis of the effect on trade between Member States to be central in the Commission’s state aid decisions. However, this is not the case as a negative spillover is presumed when there is even very limited or no trade. The presence of trade does not necessarily mean that a negative spillover will occur as a result of the granting of a state aid. State aid could bring about a negative spillover if the geographic market is wider than national – or encompasses an area which covers at least two Member States. Therefore, both geographic market definition (Fingleton et al., 1999) and market analysis are critical. Even when the market is defined to cover two or more Member States, the granting of state aid might not necessarily result in a

13

The EA consists of the European Union and EFTA (Iceland, Liechtenstein and Norway) Member States. Ad hoc aid may also be looked at suspiciously because they might increase the firms’ ex post bargaining power vis-à-vis the government. However, in this case, state aid control would not be a means to control the negative externalities imposed on other countries, but a way for governments to overcome their own weak bargaining position, once state aid has been granted. 15 In practice, in some cases, it could be difficult to establish whether state aid is likely to affect variable costs and if so for which products. 14

156

PIETRO CROCIONI AND CHRIS NEWTON

negative spillover. Only where markets are imperfectly competitive, a concern arises as aid might be used to capture rents that only exist if the market is not competitive. The extent or existence of a negative profit-shifting effect as a result of a subsidy depends on the degree of product differentiation (Collie, 2003). Suppose that products produced by the domestic and foreign firms are differentiated but not to such an extent to make them part of different product markets. When products are close substitutes, the impact of a subsidy to a domestic firm on foreign firms is maximum. Conversely, at the other extreme (very high product differentiation), a production subsidy is beneficial to other countries because they can now obtain the product cheaper but has no effect on the profit of their domestic firms (as in this case would be in a separate market). This means that when products are relatively homogeneous, state aid is likely to be most harmful. The Commission has recently attempted to partly counter this criticism introducing the Significant Impact Tests (SIT) which attempt to screen those state aids that merit a deeper analysis. The SIT includes the Lesser Amounts of State Aid test (the LASA test), a type of de minimis rule, and the Limited Effect on intra-Community trade Test (the LET test). The LET is a way to identify state aid that is unlikely to raise significant concerns about negative spillovers and which therefore should not be a cause of concern at Community level. The Commission in its Communication summarizes the Community courts’ jurisprudence, which interprets broadly the concept of effect on intra-Community trade. Therefore, if the other elements of state aid exist, then the measure will most likely be considered to affect trade between Member States and thus qualify as state aid under Article 87(1) EC Treaty. Nevertheless, the Community courts have recognized that in particular economic sectors that are not exposed to such intense competition at Community level, a small amount of aid to an undertaking over a given period would not affect trade between Member States. The Commission, on the basis of “economic rationale,” concluded that LET should only apply to a limited number of activities that, by their nature, do not produce significant cross-border effects or do not appear to be characterized by high concentration and barriers to entry. More recently, the Commission has undertaken a wider process of reform of state aid.16 However, because of the constraint of legal precedents, one should not expect radical changes in the Commission’s practice, unless the Treaty itself is modified.

4 STATE AID POLICY IN THE AIRLINE SECTOR The major state aid cases in the air transport sector since the 1990s have involved direct financial interventions by national governments aimed at the rescue or restructuring of state-owned flag carriers. In other words, these measures amounted to subsidies to ailing airlines. The main concerns of the Commission in relation to these cases were (i) whether the financial interventions in fact constituted state aid – i.e., they would not be under­ taken by a private investor according to the MEIP test – and, if so, (ii) to ensure measures

16 The European Commission has recently issued a consultation document entitled State Aid Action Plan, 7 June 2005, which highlights a number of potential reforms in the approach to state aid policy (available at http://europa.eu.int/comm/competition/state_aid/others/action_plan/saap_en.pdf). See also Friederiszick et al. (2006).

STATE AID TO EUROPEAN AIRLINES

157

were in place to limit the amount of aid to that strictly necessary to achieve restructuring objectives and to minimize any potential distortions to competition arising from the aid. Close scrutiny of these measures was particularly warranted, given that the develop­ ment of competition between the flag carriers and other carriers was during the 1990s in its infancy and the measures were selective and aimed at maintaining ailing firms in business. At the same time, the Commission saw this period as an opportunity for the flag carriers to get their houses in order before full liberalization. The desire of the Commission to take a tough line on state aid whilst recognizing the need for restructuring led to the single aid principle policy for restructuring (i.e., “one time, last time” rule). This was a clear statement of the Commission’s position that restructuring was accept­ able, and perhaps (politically) necessary, but wanted to avoid repeated restructuring and rescue aid measures. Although rescue and restructuring measures continued to be a feature of the sector into the latter part of the1990s, the “one time, last time” principle was in most but not all cases (the exception being Air France) largely abided by. This ensured that major restructuring was completed by the time airline markets were fully opened up to competition. Given the Commission’s policy position, the rescue and restructuring measures that have so dominated state aid, the airline sector in the 1990s are likely to be very much a thing of the past. The Commission’s recent Charleroi decision involving aid to Ryanair to set up at Charleroi airport raised entirely different issues.17 In contrast to the state aid decisions of the past, it involves a low-cost entrant rather than an incumbent flag carrier. Moreover, the aid did not involve a flow of capital from a national government to its national airline. Rather, the case involved indirect assistance by a regional government and a publicly owned regional airport to a foreign-based and/or owned airline. Although, so far, an isolated case, the policy questions raised are in many ways as fundamental as those relating to the restructuring of the flag carriers. As such, the Charleroi decision may provide a good indication of the type of issues that will arise in the sector in the future. Below, we highlight the potential deficiencies in the Commission’s approach to the analysis of state aid using the rescue and restructuring state aid cases and, in particular, the Charleroi decision. The latter would appear to have potential implications for the commercial strategies that can be pursued by regional airports within Europe (the vast majority of which are public undertakings) in attracting carriers.

4.1 Aid to Flag Carriers in the 1990s Prior to the 1990s, there were only a few minor state aid cases relating to the aviation sector. However, when demand in the sector fell significantly at the beginning of the 1990s, many airlines sought to restructure and looked to public funds to help them do so. Table 1 provides a summary of the main state aid cases to ailing flag carriers, for which

17 European Commission, Commission Decision of 12 February 2004 concerning advantages granted by the Walloon Region and Brussels South Charleroi Airport to the airline Ryanair in connection with its establishment at Charleroi (notified in Number C(2004) 516), (Charleroi decision henceforth), 2004/393/EC. The decision is currently under appeal by Ryanair.

Table 1 Main Rescue and Restructuring State Aid to European Airlines Case

Member State

Amount of assistance provided

Type of assistance

Undertakings and compensatory measures

Sabena, 1991 (OJ No L300/48, 31.10.91)

Belgium

c. Euro 1,600 million

Capital injections and cancellation of equity and debt

Approved subject to undertakings relating to (i) implementation of an agreed restructuring plan; (ii) ensuring no further direct or indirect state support including privileged access to slots or airport services

Air France, 1991 and 1992 (Commission press releases IP/91/1024 and IP/92/587)

France

c. Euro 900 million

Capital injections

Capital injections determined not be aid upon application of the MEIP

Air France, 1994 (OJ No L254/73, 30.9.94)

France

c. Euro 3,200 million

Capital injection and bond subscription (the latter by CDC a state-owned company)

Aid relating to CDC bond subscription to be repaid. Capital injection approved subject a range of commitments including: (i) implementation of a restructuring plan and meeting of performance targets; (ii) ring-fencing of the aid; (iii) limitations on size of fleet, number of route and services operated, and level of fares; (iv) non-preferential treatment in respect of air traffic rights

Iberia, 1992 and 1996 (Commission press release IP/92/606 and OJ No L104/25, 27.4.96)

Spain

c. Euro 1,300 million

Two waves of capital injection

First capital injection approved subject to commitments regarding the ring-fencing of the aid and absence of further assistance in the future. Second capital injection determined not to be aid following a reduction in amount to be granted

Aer Lingus, 1994 (OJ No L54/30, 25.2.94)

Ireland

c. Euro 200 million

Capital injection

Approved subject to wide ranging commitments including (i) implementation of restructuring plan and meeting of performance targets; (ii) ring-fencing of the aid; (iii) limitations on size of fleet and services operated; (iv) no acquisitions of other carriers within the Community (Continued)

Table 1 Main Rescue and Restructuring State Aid to European Airlines—(Cont’d) Case

Member State

Amount of assistance provided

Type of assistance

Undertakings and compensatory measures

TAP, 1994 (Commission press release IP/94/609)

Portugal

c. Euro 1,800 million

Capital injection, loan guarantees and tax exemptions

Approved subject to conditions including (i) payment of aid conditional on meeting financial targets; (ii) ring-fencing of the aid; (iii) limitations on expansion of services operated

Olympic Airways, 1994, 1998 and 2005 (Commission press releases IP/94/700 and IP/05/1139, and OJ No L128/1, 21.5.99)

Greece

c. Euro 2,000 million

Capital injection, loan guarantees, conversion of debt to equity an debt write-off

Originally approved subject to undertakings regarding: (i) meeting agreed restructuring plan; (ii) limitations on expansion of services; (iii) ending of exclusive rights to scheduled services to Greek islands. Following subsequent investigation by the Commission, a proportion of the aid was ordered to be repaid in light of certain undertakings not having been met

Alitalia, 1997 (OJ No L322/44, 25.11.97)

Italy

c. Euro 1,500 million

Capital injection

Approved subject to undertakings including (i) implementation of agreed restructuring plan; (ii) absence of further aid; (iii) ring-fencing of aid; (iv) non-discrimination in favor of Alitalia in respect of traffic rights, slotting and access to airport services; (v) limitation on fare reductions.

Source: Relevant Commission’s decisions. This is unlikely to be complete as it is based on publicly available information, while sometimes state aid cases are closed informally if a Member State modifies the state aid measure. No significant state aid concerns arose in the sector until the terrorist attack of 11 September 2001. The Commission stressed that these events should not be used to justify substantial increases in aid in the sector. However, compensation to airlines for closure of US airspace and assistance in relation to insurance was permitted under state aid rules in order to make good damage caused by an exceptional occurrence.

160

PIETRO CROCIONI AND CHRIS NEWTON

information is publicly available, and the types of remedies imposed by the Commission to approve the measures. All the examples in Table 1 involve subsidies to ailing flag carriers which come under the label of rescue and restructuring aid. The Commission is aware of the problems and inefficiencies raised by this type of aid.18 Benefits are difficult to frame in terms of either market failure or social benefits. Competitive distortions can be substantial and reductions in incentives for productive efficiency can arise as a result of the softening of the firm’s budget constraint. These effects are more likely to be greater and wider in scope where the recipient firm has market power which was often the case for flag carriers. Furthermore, there might be legitimate reasons for smoothing the social impact of closure, but the same funds might be more efficiently used in other ways (e.g., to retrain unemployed workers). The recent Rescue & Restructuring Aid Guidelines explicitly recognize that aid to ailing firms is potentially very distortive.19 However, they conclude that there might be some economic and social justifications for granting state aid to ailing firms. The former are justified on the basis of social or regional considerations and other policy considerations.20 The Rescue & Restructuring Aid Guidelines also suggest that there might be an economic justification for avoiding bankruptcy and exit by claiming that there is a need to maintain a “competitive market structure when the demise of firms could lead to a monopoly or a tight oligopolistic situation.”21 This is, however, contentious as firms might fail, but this may not necessarily mean that this per se will lead to exit or to a reduction in the number of firms in the market, other than perhaps in the short term. The assets of the failed firm could remain in the market and therefore no reduction in competition might occur, if taken over by an entrant. This was the case of Swissair which went bankrupt in October 2001. Its assets remained in the market as the company was bailed out, although the latter was largely led by the Swiss government. Another perhaps more pertinent example is that of Sabena which was dragged into bankruptcy as well by the collapse of its major shareholder Swissair. Most of its assets were purchased by a group of investors under the brand name of SN Brussels. Therefore, the Commission’s approach seems based on a static view of competition, whereas new entry could occur following exit. Although in some cases it might be argued that state aid to ailing firms could prevent a reduction of competition in a particular market, this has to be based on a case-by-case analysis. For example, state aid might be justified

18

See for example Friederiszick et al. (2006). They recognize that “the exit of inefficient firms is a normal part of the market. It cannot be the norm that a company which gets into difficulties is rescued by the State. Aid for rescue and restructuring operations has given rise to some of the most controversial State aid cases in the past and is among the most distortive types of State aid. Hence, the general principle of the prohibition of State aid as laid down in the Treaty should remain the rule and derogation from the rule should be limited.” Furthermore, they mention that restructuring aid could “shift and unfair share of the burden of structural adjustment and the attendant social and economic problems onto other producers who are managing without aid.” Rescue & Restructuring Aid Guidelines, paras 4 and 31. 20 Rescue & Restructuring Aid Guidelines, para 8. Aid might be thought as legitimate if it does not distort trade and this could be “where the aid is necessary to correct disparities caused by market failures or to ensure economic and social cohesion” (para 19). 21 Rescue & Restructuring Aid Guidelines, para 8. 19

STATE AID TO EUROPEAN AIRLINES

161

where there are significant barriers to entry or when markets exhibit characteristics that might lead to tipping after the ailing firm exited. These, however, seem to be the exception rather than the rule and not particularly relevant in the case of aid to airlines in the 1990s. The Commission, when it considered state aid as justified but had concerns that it may have led to competitive distortions, has often imposed compensatory measures to limit such distortion. During the 1990s, these included among others output restrictions affecting fleet size, routes served, and even prices (Table 1). The range of compensatory measures contemplated in the Rescue & Restructuring Aid Guidelines also includes asset divest­ ments, capacity reductions, or reduction or removal of entry barriers.22 However, some of these compensatory measures are not necessarily pro-competitive. For example, impos­ ing a reduction of capacity might exacerbate rather than reduce the anticompetitive effects of state aid. These measures are defined as an attempt to maintain the status quo which, however, should not be maintained if the measure aims at correcting a market failure. There is another potentially relevant consideration in relation to aid to ailing airlines. In the period in which this aid was granted and largely approved by the Commission, competition on most intra-EU routes was limited. Therefore, the recipients were likely to have market power. As argued earlier, when this is the case, the risk that competition will be distorted is more acute and correspondingly the justification for the state aid weaker. Conversely, because of the nature of this type of aid, the potential and precise impact on the market(s) is often difficult to predict – i.e., how is the aid likely to be used and in which routes?

4.2 The Charleroi Decision 4.2.1 Main Facts In 2004, the Commission ruled against what it determined to be state aid provided to Ryanair, a low-cost carrier, by the Walloon Region and the Brussels Charleroi Airport (BSCA). It concluded that, in part, this aid could be compatible with the common market – in particular, subject to certain conditions, aid to support the launch of new routes. However, it did not authorize other forms of aid received by Ryanair and consequently repayment was required. The measures that were the subject of the decision were contained in two contracts. The first was between Ryanair and the Walloon Region (the owner of the airport infrastructure) and signed in November 2001. Under this contract, Ryanair received a reduction in airport landing charges amounting to approximately a 50% discount against published rates.23 Under a second contract with BSCA (a public company managing the airport under a long-term concession agreement), Ryanair received discounts on fees for ground-handling services (at approximately 10% of the published rates), contributions to promotional and marketing activities, and financial incentives relating to the opening of new routes. In return, Ryanair made guarantees to base its aircrafts at Charleroi and to operate a minimum number of flights over a period of 15 years. These conditions

22

Rescue & Restructuring Aid Guidelines, paras 39–40.

As part of this contract, the Walloon Region also committed to pay compensation should Ryanair suffer

losses as a result of regulatory changes regarding airport taxes and opening hours.

23

162

PIETRO CROCIONI AND CHRIS NEWTON

were only granted as part of the private commercial contract with Ryanair as a departure from the conditions generally available to airlines at Charleroi. The Commission, there­ fore, concluded that Ryanair was the only actual and potential airline to receive these benefits. 4.2.2 Was It State Aid? Following its standard approach, the Commission assessed whether a transfer of state resources took place and conferred an advantage. The former was assessed by applying the MEIP test, whilst the latter was based on the selectivity criterion according to the Commission’s standard practice. The Commission distinguished between Ryanair’s contractual arrangements with the Walloon Region and those with BSCA. The Commission maintained that the Walloon Region is, in effect, a public undertaking that acts as the regulator of airport charges for the Walloon airports. It found that the determination of landing charges “falls within the legislative and regulatory competence of the Walloon Region and that the principle of private investor in a market economy is not applicable in these circumstances.”24 It argued that as the Walloon Region acts as a regulator rather than a company in relation to the setting of these charges,25 it would be inappropriate to apply the MEIP test in this case. In reaching an agreement with Ryanair, the Walloon Region had deviated from the framework of rules it had itself laid down for the determination of airport charges. Since the reductions to landing charges were provided on a selective basis to Ryanair, conferring an advantage in terms of reduced operating costs, they must be considered as state aid. The Commission seems to argue that because of the nature of the economic relationship between the Walloon Region and Charleroi airport, the decision to offer Ryanair reductions on landing charges could not be viewed as acceptable on the basis that it was motivated by normal commercial considerations. In challenging the Commission’s conclusions, Ryanair argued strongly that it was common practice to negotiate terms with airports and that the terms that it had secured at Charleroi were no more favorable than those it had achieved in normal commercial negotiations with a number of private airports across Europe. Moreover, it argued that the Commission had mischaracterized the nature of the relationship between the Walloon Region and Charleroi. In its view, in deciding to reach an agreement with Ryanair, the Walloon Region (the airport owner and main shareholder in BSCA) was acting as economic agent, not as a regulator. Given this, the MEIP test should be applied. The Commission rejected this line of argument. Although it seems to accept that the agreement may have been reached out of a “commercial need” to attract Ryanair (and indeed any other airline) to Charleroi – i.e., the Walloon Region’s behavior may have been motivated, as claimed by Ryanair, by its role as airport owner and shareholder in BSCA – the Commission argued that this was irrelevant. As long as the Walloon Region

24

Charleroi decision, para 144.

The Commission also notes that the income derived from the airport charges was split between BSCA and

an environmental fund, suggesting that the Walloon Region has no (direct) interest of a commercial nature in

the level of airport charges.

25

STATE AID TO EUROPEAN AIRLINES

163

acts as an airport charges regulator, whatever its motivation, a decision to step outside its normal framework and offer selective discounts to a carrier must constitute state aid. Indeed, the Commission’s key concern was that the Walloon Region had “placed itself in a situation of confusion of powers.” Irrespective of whether or not this was a “confusion of powers” the Wallon region was an “investor” in BSCA and, therefore, could in principle have acted in an economically rational way – i.e., according to the MEIP test – in deciding to grant state aid to Ryanair. Regarding the agreement with BSCA, the Commission determined that it was appropri­ ate to apply the MEIP and consider whether a private operator in the same circumstances would have offered these terms to Ryanair. It argued that the evidence put forward by Ryanair comparing the terms offered by Charleroi with those that Ryanair had agreed with a number of European private airports was of limited relevance. The differences in circumstances and/or the potential presence of an element public subsidy at other airports limit the validity of such comparisons. Instead, the Commission relied on an analysis of the business plans and assumptions underlying the agreement with Ryanair. It concluded that the business case underlying the agreement did not take account of relevant costs and was based on what the Commission believed to be inappropriate assumptions. Moreover, according to the Commission, the agreement involved commer­ cial risks for BSCA of a magnitude that would not have been acceptable to a private operator.26 As a result, the Commission concluded that the terms of the agreement with Ryanair failed the MEIP test and both sets of measures were state aid. While the decision is largely based on a formalistic approach on whether a measure is state aid, there might be some additional considerations which will be more fully explored in the next sections. It might be that in the absence of the state aid no airline would have settled at Charleroi. As a consequence, the only one which was open to such option – i.e., Ryanair – could act as monopsonist and extract all the rent. Had Charleroi been privately owned, this could have forced airport charges down to marginal cost. Because Charleroi was state owned, charges could have been forced below marginal costs as the Commission claims. If the Walloon Region agreed to this in order to correct market failure, then the measures could be justified. Lastly, it is unclear whether competition and/or trade were distorted. Airports compete in the provision of ground and landing services to attract airlines. As discussed below, however, competition between airports could only occur if they are sufficiently close to each other that the airlines consider them close substitutes. This is because the demand for airport services is a derived demand from passenger air transport services and passengers demand is for route-specific services – i.e., a flight from Charleroi to Prague is not a substitute for one on the Charleroi–Madrid route. Therefore, it is at least debatable that the function of the aid was to attract airline demand that otherwise would have benefited airports elsewhere in Europe. If this was the case, no distortion of competition would arise.

26

It argued (para 237) that the “BSCA’s financial structure is based on that of the Walloon region, and that without the implicit guarantee and assurance that the Region procures for its public sector companies [   ] and because of the commercial hazards inherent in the business plan, BSCA would never have committed itself to Ryanair.”

164

PIETRO CROCIONI AND CHRIS NEWTON

4.2.3 Was Competition Distorted? Under current practice for a measure to constitute state aid, it must both distort compe­ tition and affect trade between Member States. The Commission’s analysis of the extent of any advantage conferred on Ryanair by the agreements with the Walloon Region and BSCA relies essentially on a presumption that the selective nature of agreements must imply Ryanair had secured an advantage. Ryanair argued that the agreements were not exclusive, and similar reductions to charges could be offered to other users of the airport who wished to enter into comparable arrangements. The Commission argued that because the agreements were made privately, similar terms could not be considered automati­ cally available to other users and this was sufficient for the agreements to be considered selective. While the confidential nature of the agreement might be an important legal consideration, it is at least doubtful whether it had any practical consequences. Indeed, the same (selective) effect could be achieved by structuring the charges Ryanair was offered as conditional to reaching the same traffic volumes that Ryanair guaranteed to Charleroi. Furthermore, because of the large fixed costs in running airports, it is efficient to recover them by charging based on the demand price elasticity of each airline.27 As in all state aid cases, once the financial transfer and the selectivity have been established, the distortion of competition is currently presumed. Nonetheless, Ryanair and the Belgian State argued that the aid could not have an impact on competition in the relevant market because there were no direct competitors on the routes served by Ryanair out of Charleroi.28 Ryanair stressed that its services were not in direct competition with those of other users of Charleroi – i.e., charter carriers operating a limited number of flights mainly in the summer months. Ryanair flights out of Charleroi amounted to about 90% of all flights (including charter flights). Ryanair also pointed out that its direct low-cost rivals – Virgin Express and EasyJet – had refused to operate from Charleroi. As a result, to the extent that Ryanair competed with these operators’ services from other airports, it should not be assumed that the reductions in charges it had achieved at Charleroi placed it at a competitive advantage. There were competitive advantages associated with operating from other airports, and in Ryanair’s view, the reductions it achieved at Charleroi served simply to offset the disadvantages of operating from what, in their absence, would be a relatively unattractive airport. Initially, the Commission appeared to suggest that a detailed analysis of the compet­ itive effects of the measures was required.29 However, the final decision contains very

27

The Commission recognizes that differences in charges are potentially justifiable as established in its Manchester Airport decision. 28 Charleroi decision, paras 70–72 and 247. 29 In its initial assessment in its decision to initiate proceedings, the Commission had concluded: “the granting by the Walloon Region of a reduction in airport taxes to one airline company only, on all its for a period of fifteen years, in a bilateral agreement that was given no publicity and deviates from the stipulations of the law, amounts to giving a tax exemption to the company. This places the company at a more advantageous position than competitors flying out of Charleroi” (para 13). With regard to the agreement with BSCA it similarly concluded: “the bearing by BSCA of certain air service costs also had the effect of putting Ryanair in a more advantageous situation than its competitors, whether they were companies operating out of Charleroi or other companies operating out of other airports” (para 14).

STATE AID TO EUROPEAN AIRLINES

165

limited reasoning either on which markets could be affected or the potentially distor­ tionary effects of the measure. Indeed, the Commission made its view clear that “the analysis of the impact of competition in the State aid sector is not the one applied when the Commission examines and alliance or concentration between airlines. Extending the definition of the relevant geographic market, which is governed by competition law, to the State aid sector would involve ignoring specific nature of these two separate areas of Community economic law.”30 While there might be some justification as to why the approach to market definition might differ under state aid law, the Commission did not elaborate further as to why this is the case. What is striking in this case is that the Commission appears to put forward two implicit markets that might be affected which are very different from those normally defined in the sector under competition law. First, the Commission implicitly argued that airlines compete in a potential market for “air transport services”31 or more precisely “the advantage granted through the bearing by the State of operating costs normally borne by an airline does not only distort competition on one or more routes and on a particular market segment. The advantage gained by the airline strengthens its economic position on its network as a whole in relation to competing companies, whether these are traditional low-cost, charter or regional compa­ nies.”32 Under competition law, the traditional view for short haul (and point-to-point) services such as those provided by Ryanair at Charleroi is that potentially there are as many relevant markets as the routes or city pairs served by the airline (depending on the substitutability of airports that are not too far away from each other (Crocioni, 2000)).33 The second implicitly defined market is that for the provision of airport services to airlines.34 The Commission appears to put forward the notion that regional airports such as Charleroi do not offer the same services as national airport, and evidence shows that “when an airline sets up at a secondary airport, the passenger traffic will not necessarily be diverted from the main airport to the secondary airport.” Therefore, secondary airports such as Charleroi according to the Commission are implicitly in a separate market from traditional national airports. The issue of whether all regional airports or only those that are closely located are in the same market was left open.35 While this second potential market appears sensible at first sight, it is in contrast to the fact that the aid was granted not to the airport but to the airline, given that according to the Commission Charleroi airport would be at an economic disadvantage as a result of inducing Ryanair to establish a hub at its airport. The Commission appears to conclude that Charleroi airport was in a better financial position prior than after the aid. Indeed, it concluded that if BSCA was a private investor it would have not offered such terms to Ryanair. However, one could examine airport services as being an input market affected, instead. The aid is still

30

Charleroi decision, para 248.

Charleroi decision, para 298.

32 Charleroi decision, para 249. This wide definition could be somewhat relevant for rescue and restructuring

aid where the effects of generic measures of state aid could not be easy to determine on a route-by-route basis.

33 For long haul air transport services, markets are defined more widely. This is for two reasons. First, the

relative dis-benefit of traveling to airports which are further away is likely to decline with the distance or

duration of the flight. More airports at both departure and arrival could be part of the same market. Second,

flights which involve an intermediate stop might also be in the same market as non-stop flights.

34 Charleroi decision, para 299–301.

35 Charleroi decision, para 301.

31

166

PIETRO CROCIONI AND CHRIS NEWTON

granted to Ryanair but because it is conditional to the airline committing to Charleroi, the latter benefits from increased demand from its services. The decision contains no analysis of the potentially distortionary impact of the state aid on competition. This case appears different from those of aid to ailing flag carriers. In the cases in Table 1, aid was not tied to specific services, and therefore, this made it difficult to identify the potential specific effects. In the Charleroi decision, most of the measures that were deemed to be state aid were route or flight specific and affected Ryanair’s marginal costs – discounts on landing charges, contribution to the opening of new routes, etc. – and their effect on pricing and route entry decisions could in principle be properly assessed. Because of the nature of the aid measures, a direct impact on fares could be presumed or deemed likely. Yet, the Commission does not seem to reply directly to Ryanair’s argument that because the relevant markets are the routes out of Charleroi where it was the sole provider of scheduled passenger services, there could be no impact on competition. Despite the lack of formal market definition, implicitly the decision concluded that all the routes out of Charleroi are in separate markets than those out of the closest airport, Brussels Zaventem, although the two are only about 30 miles apart.36 If they were not, competition could have been affected. In this case, an analysis of the city pairs served by Ryanair out of Charleroi – including all possible airports at either end – shows that in the 11 routes served by Ryanair, there could have been between zero and five actual competitors (Barbot, 2004).37 When aid is granted to the sole company currently active in the market, it could still distort potential rather than actual competition. For example, potential entrants in the route could be discouraged from entering because the incumbent’s services are subsidized. This raises an important consideration. State aid analysis is necessarily forward-looking as in most cases, the exception being aid which was not notified and later discovered, the Commission does have to assess whether the measure, if allowed, is likely to distort competition. In this case, it appears difficult to claim that this would have been the case. The airport struggled to attract airlines as other low-cost carriers had responded negatively to offers by the BSCA to set up a hub at Charleroi in the past. 4.2.4 Was There a Negative Spillover? As argued above, in the presence of negative spillovers there is a stronger case for a supranational control of state aid. This is reflected by the law that in principle bans only state aid that affects trade between Member States. We argued that, contrary to the current practice based on presumptions, this should be interpreted as whether the aid generates a negative externality on other Member States. The Charleroi decision did

36 In two merger decisions involving the Belgian national carrier Sabena, the Commission appears to have concluded that the merger would have affected city pairs from Brussels. While this appears to suggest that the Commission concluded that airports in the same urban area are in the same market, these decisions relate to a period when in Brussels there was effectively only one national airport. See European Commission, Case No. IV/M.616, Swissair/Sabena, 20 July 1995 and Case No. IV/M.157, Air France/Sabena, 05 October 1992. 37 However, some of the airports of these city pairs are at a considerable distance from each other – i.e., Glasgow Prestwick and Edinburgh or Pisa and Florence are about 60 miles apart – raising some questions as to whether these could be seen as close substitutes by consumers. Barbot also shows that Ryanair’s fares are not influenced by the presence in the same city pair of another low-cost airline – i.e. a competitor offering services which are closest to those of Ryanair.

STATE AID TO EUROPEAN AIRLINES

167

not investigate whether a negative spillover might arise. While airline services across the EU and especially those originating in a small Member State such as Belgium have an international dimension this does not necessarily mean that a negative externality will arise from subsidizing such services. Using the implicitly defined product markets one could start drawing some conclusions. First, let’s consider the provision of airline passenger services. If the relevant market is the route, while all routes out of Charleroi are international it appears difficult to argue that there is a negative externality as the production of the services appears not “based” in Belgium but on both Member States at either end of the route. Second, assume that the relevant market is that for the provision of airport services to airlines. We discussed above that it is difficult to argue that one of the affected markets is that for the provision of airport services when the supposed aid would make the recipient (BSCA) worse off. However, assuming that the aid was granted to Charleroi airport rather than to Ryanair one could assess whether a negative externality could arise. This would require that airports located in different Member States compete with each other in offering services to airlines. This is alluded to in the Charleroi decision which mentions the example of Ryanair taking advantage of competition between the two closely related airports of Tarbes and Pau in France and concluded that it could also occur “between airports located in separate Member States”.38 Whether this is the case it is at least debatable and a question largely and ultimately depending on whether a sufficient proportion of consumers (and therefore airlines) viewed two destinations far from each other as substitutable. Moreover, in this case it seems unlikely that a negative externality could arise in the absence of substitutability between airports located in different Member States. If Charleroi airport received a subsidy and this allowed it to attract airlines such as Ryanair to use it as a base, airports at the end of each route out of Charleroi would also benefit by the increase in traffic in the form of increased demand. This is a positive rather than negative externality. In other words, airports at both ends of a “Charleroi” route are complements rather than substitutes. 4.2.5 Was There a Market Failure or Social Justification? The Commission found that the measures in question could not be exempted under Article 87(2) and 87(3) of the Treaty. However, the Commission recognized the growing role that regional airports may have to play in promoting the economic development of regions (the subject of Article 87(3)(c)). The Commission concluded: operational aid measures intended to help the launch of new airlines or strengthen certain frequencies may be a necessary tool for the development of small regional airports. The measures may indeed persuade the interested companies to take the risk of investing in new routes. However, in order to declare such aid compatible on the basis of Article 87(3)(c) of the Treaty, it should be determined whether this aid is necessary and in proportion to the objective sought, and whether it affects trade to an extent that is contrary to the common interest.39

38 39

Charleroi decision, para 301. Charleroi decision, para 279.

168

PIETRO CROCIONI AND CHRIS NEWTON

Although both the concessions made by the Walloon Region and BSCA were found to constitute aid, the Commission determined that certain elements of this aid were compatible with the common market to the extent that they foster the development of and use of underutilized secondary airport infrastructure. This included aid relating to “one-shot incentives” to develop routes (including BSCA’s contribution to promotion and publicity and sums received over the 2001 to 2003 period).40 Nevertheless, for this aid to be declared compatible the Commission required a number of conditions to be met, namely that (i) the aid is necessary for opening of new routes and is proportional to meeting this objective; (ii) principles of transparency and nondiscrimination between operators are followed; (iii) there are appropriate penalties for carriers who fail to meet commitments; and (iv) the aid is of limited duration (5 years), applying only to genuinely new, rather than replacement, routes. The Commission found the other elements of aid incompatible with the common market.41 In determining that the start-up aid (i.e., intended to promote the development of new routes) provided to Ryanair could be deemed to be permissible, the Commission clearly had accepted that such aid could be justified on either social grounds or on the basis of the existence of a market failure. The Commission’s decision, however, is not entirely explicit regarding the precise nature of the justification for the aid. Two possible types of justifications could be relevant. First, to the extent that the development and growth of a regional airport may have spill-over effects – for example, through the creation of employment – which benefit the wider regional economy, start-up aid might be viewed as having a social justification, especially in less developed areas of the EU. Alternatively, such aid could equally be viewed as addressing a market failure by helping internalize the positive externalities that users of the regional airport may confer on the local economy but that they would not consider in making private decisions. The Commission does not provide evidence of a link between the development of regional airport infrastructure and regional economic development. Second, the Commission makes reference to underutilized capacity at secondary airports. Given the extent of congestion that exists at many major airports, encouraging airlines to use secondary airport infrastructure may generate positive externalities for other users by reducing congestion costs. Although this could provide a market failure justification of the type of assistance provided to Ryanair, it is unlikely to be a first best. Above we argued that a more efficient measure could be a tax on the use of congested airport. 4.2.6 Promoting the Development of Regional Airports? The Commission has argued that its decision provides guidance and clarity regarding the financial incentives that can be offered to attract carriers by airports that are under

40

This accounted for approximately three-quarters of the aid the Commission had identified. In particular, the Commission required aid to be recovered in relation to (i) the reduced airport charges agreed with the Walloon Region – that could only be permissible if granted in a non-discriminatory manner and were time limited; (ii) the reduced ground handing charges agreed with BSCA – the Commission determined that such reduction are unlikely to be compatible in any circumstances; (iii) one-shot incentives to develop new routes which do not reflect the costs of doing and/or that are not tied to meeting this objectives; and (iv) aid relating the Ryanair’s Dublin-Charleroi – this route had been opened in 1997 and therefore in the Commission’s view could not be considered as a ‘new route’. 41

STATE AID TO EUROPEAN AIRLINES

169

public ownership or control. It also claimed that the decision will serve to promote the activities of low-cost airlines and the development of regional airports whilst ensuring a level playing field for competition.42 The decision undoubtedly provides clarity regarding the measures the Commission is likely to view as acceptable for airports to attract carriers. It is clear that Commission considers that such measures should be limited to the development of new routes and should not relate to existing routes. In line with its state aid policy in other areas, the Commission has also placed great emphasis on factors including the need for a coherent set of objectives for the aid, the requirement for long-term profitability and the presence of time limits on the duration of any aid, transparency and nondiscrimination. However, whether the decision will serve to promote the activities of low-cost carriers and regional airports is less clear. The decision would appear to limit significantly the scope of agreements that can be reached between low-cost carriers and regional airports. Agreements of a similar nature elsewhere in the sector may not meet the criteria set out by the Commission in Charleroi. Moreover, it is unclear whether the business model developed by Ryanair and other low-cost operators – or at least further growth of this model – would be a viable proposition if the incentives that can be offered by airports cannot go beyond those set out in Charleroi. It is arguable that the Commission’s decision places a significant regulatory burden on publicly owned or controlled airports wishing to offer carriers start-up incentives. It would appear that, at least in principle, an airport offering such incentives must be in position to demonstrate clear and direct benefits and that the aid is necessary and does not go beyond that needed to achieve these objectives. The stringent nature of the tests to apply to these types of assistance and the potential penalties that may be faced if the Commission deems these criteria are not met and the aid is incompatible, may serve to deter airports and carriers from agreeing even financial incentives that are limited in scope.

5 CONCLUSIONS We have critically reviewed the Commission state aid policy and highlighted the prob­ lems with the current system which is based on a set of presumptions rather than based on the effects, and in particular the distortions, that state aid can bring about. We used state aid decisions in the airline sector to illustrate some of the problems caused by the current approach to state aid. State aid to ailing flag carriers was frequently granted in the 1990s. This type of state aid is one of the most problematic from an economic efficiency point of view and it is equally difficult to justify as pursuing social objectives. While the Commission was successful in largely confining this type of aid to the period of transition to liberalization, it approved most of the state aid measures. We also argued that some of the compensatory conditions imposed and mentioned in the recent Rescue & Restructuring Aid Guidelines could exacerbate rather than reduce the distortionary effects of state aid.

42

Commission press release IP 04/157.

170

PIETRO CROCIONI AND CHRIS NEWTON

The Charleroi decision is of a different nature as it involved a new entrant rather than an incumbent. It highlights the problems caused by the lack of a clear identification of the markets affected by the measure, whether the measure could distort actual and potential competition, whether it generates negative spillovers (and therefore justifies a supra-national control), and when it could be justified in spite of the distortions and spillovers brought about. We argue that a better supranational system that aims at controlling subsidies granted by EU Member States should be based on a more rigorous economic approach and analysis looking at the likely effects of state aid measures.

REFERENCES Adkins B. (1994), Air Transport and EC Competition Law, Sweet & Maxwell, London. Barbot C. (2004), “Low cost carriers, secondary airports and State aid: an economic assessment of the Charleroi affair”, FEP Working Papers No. 159. Besley T. and Seabright P. (1999), “The Effects and Policy Implications of State Aids to Industry: An Economic Analysis”, Report to DG-III of the Commission, mimeo, available at http://europa.eu.int/comm/enterprise/library/lib-competition/libr-competition.html. Brander J. and Spencer B. (1985), “Export Subsidies and International Market Share Rivalry”. Journal of International Economics, 18, 83–100. Collie D. (2003), “Prohibiting State Aid in an Integrated market: Cournot and Bertrand Oligopolies with Differentiated Products”, Journal of Industry, Competition and Trade, 2, 215–231. Corden M. (1990), “Strategic Trade Policy. How New? How Sensible?”, World Band Working Papers, WPS 396. Crocioni P. (2000), “Defining Airline Markets: A Comparison of the U.S. and EU Experiences”, Antitrust Bulletin, 45 (spring), 1–45. Fingleton J., Ruane F. and Ryan V. (1999), “Market Definition and State Aid Control”, mimeo, available at http://europa.eu.int/comm/images/language/lang_en3.gif. Friederiszick H.W., Röller L.H. and Verouden V. (2006), “European State Aid Control: An Economic Framework”, forthcoming in Buccirossi P. (ed), Advances in the Economics of Competition Law. Garcia J.A. and Neven D.J. (2004), “Identification of sensitive sectors in which State aids may have significant distorting effects”, Report to HM Treasury. Gil-Molto M.J. and Piga C.A. (2005), “Entry and exit in a liberalised market”, Discussion Paper Series, Loughborough University Economics Department, available at http://papers.ssrn.com/ sol3/papers.cfm?abstract_id=916505. Möellgaard P. (2004), “Competitive Effects of State Aid in Oligopoly”, mimeo, available at www.econ.ku.dk/cie/Seminars/pdf%20%20seminar/stateaid.pdf. Nicolaides P. and Bilal S. (1999), “An Appraisal of the State Aid Rules of the European Com­ munity – Do they Promote Efficiency?”, Journal of World Trade, 33(2), 97–124. Starkie D. (2002), “Airport Regulation and Competition”, Journal of Air Transport Management, 8, 62–72.

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

7 The Implications of the Commercialization of Air Transport Infrastructure Kenneth Button∗

ABSTRACT Following the on-going liberalization of airline markets from the 1970s, there has begun a movement to inject more commercialization into the provision and operation of airports and air navigation systems. These latter developments have been gradual and taken a variety of forms. The perception of success in these endeavors depends on the particular objective of each case. In some situations, such as those of South American airports, the aim has been to retain state ownership while injecting private capital and expertise into systems. In countries such as the UK, the privatization of most airport capacity and the establishment of a private/public enterprise that runs the air navigation system have been influenced more by a quest for economic efficiency irrespective of ownership. The overall conclusion is that, unlike airlines where the flexibility of the industry allows for relatively rapid responses to institutional change, the reforms to air transportation infrastructure will take time to work there way through, especially with continual interruptions to the process as policy continually redefines itself in the light of such things as new security considerations.

1 INTRODUCTION Globally, air transport has traditionally been heavily regulated. The rationale for this has differed somewhat over time and between countries. In the early years there were generic legal constraints on the sector that applied across industries more generally – the Wright Brothers, for example, held patents on some aspects of airplane design. From the 1920s many nations began to appreciate the role that air transport can play in economic and political integration. The US subsidized domestic airmail services and the UK fostered its longer haul services as a means of Imperial integration. Concerns about market

∗ University Professor and Director of the Aerospace Policy Research Center, School of Public Policy, George Mason University, Fairfax, Virginia. E-mail: [email protected]

172

KENNETH BUTTON

stability, and in some cases potential monopoly power, led to nations imposing specific controls over market entry and fares, and nationalization was common. The potency of air power as a military instrument added to the interest of governments to control their own industries. Wider economic regulation was initiated after the Second World War as international air transportation was seen as a potential and strategically important growth sector and the United Nations, at the 1944 Chicago Convention, set out ground rules for countries to negotiate bilateral agreements in the provision of services. Overriding these market interventions for economic reasons have been the long-standing social regulations designed to meet safety, environmental and strategic needs that recently have included a particular focus on security. Since the late 1970s there have been major shifts in the regulatory environment under which civil air transport services are provided. There has been a move towards what is known as deregulation in the US and market liberalization in much of the rest of the world1 . This has seen governments allowing market forces exert a greater influence in the determination of airlines fares, the services that can be offered, and the carriers that can offer them. The results have been a general lowering of fares, an increase in the number of airlines, the initiation of new types of service, and a significant increase in both passenger and cargo traffic (Morisson and Winston, 1995; Button and Stough, 2000). These developments, however, initially almost exclusively involved changes in the airline services market and there has been much less and a much slower liberalization of air transport infrastructure (Table 1). Only 2% of the world’s commercial airports, for example, are fully managed or owned by the private sector, although where this has taken place the results seem to have been sufficiently encouraging to stimulate further interest by the private sector. Similarly, air navigation services (ANSs) have been regulated, with the majority being state owned, with changes coming slowly. One might add to this though, that the organization and management structure of many airports and ANSs that remain in public control has changed significantly with moves away from them being treated as public utilities (DeNuefville and Odon, 2003). This particular situation, with institutional changes affecting the users of air transport infrastructure preceding that of the infrastructure providers themselves is not unique to the air sector; it can be seen in many countries with regard to railways and maritime transport and is particularly pronounced in road transportation where public ownership and operation of infrastructure remains the norm despite widespread liberalization of the trucking and bus industries. The pattern is also to be found in the energy and telecommunications sectors. Nevertheless, despite the tardiness of change in the way air transport infrastructure is provided and regulated, there has been change in many countries; and gradually evidences of its effects are beginning to emerge. Unlike airlines, where rapid changes were seen relatively, it takes time for reforms in airport and air navigation systems (ANSPs) to be felt, and anything approaching a long-term equilibrium, because of the longevity of the hardware, may not emerge for many years. Infrastructure is also less immediately visible and thus less prone to come under the scrutiny of the public.

1

The variation in jargon is partly due to the different meaning of “liberalization” on either side of the Atlantic. Traditionally, in the US liberalization means more government involvement whereas in Europe it means less. We use the terms interchangeably here.

COMMERCIALIZATION OF AIR TRANSPORT INFRASTRUCTURE

173

Table 1 Major Air Transportation Liberalization Initiatives Year

Action

Airlines/infrastructure

1944 1977 1978 1979 1984 1987 1987 1989 1989 1992 1992 1995 1996 1997 1999 2001 2001

Chicago Convention US Domestic cargo deregulation US Domestic passenger deregulation Term “Open Skies” used UK-Netherlands liberal ASA bilateral EU’s “First Package” UK Airports Act (privatization) EU’s “Second Package” Northwest-KLM strategic alliance EU’s “Third Package” US–Netherlands Open Skies Agreement GATS NAV Canada established Banjul Accord “Single European Skies” initiative MALIAT NATS UK established

Airlines/infrastructure Airlines Airlines Airlines Airlines Airlines Infrastructure Airlines Airlines Airlines Airlines Airlines Infrastructure Airlines Infrastructure Airlines Infrastructure

Note: GATS: General Agreement on Trade in Services; MALIAT: Multilateral Agreement on the Liberalization of International Air Transportation.

2 PRESSURE FOR COMMERCIALIZATION Commercialization is a somewhat vague term. Dictionary definitions largely focus on the seeking of profits, but in many cases it is more generally taken to mean the introduction of some notion, often equally vaguely defined, of market forces, whilst in other narrower contexts, it is seen more as forcing an institution to meet the demands of the user more closely, in particular in terms of allocating facilities according to consumer’s willingnessto-pay. A common feature is seen to be a general tightening of the link between the costs and revenues of an undertaking so that outside finance is reduced and users are made more aware of the opportunity costs of their actions. Many tie commercialization directly to privatization2 , but this need not be the case if a public entity has to be self-sustaining financially and has any monopoly power it may enjoy constrained by countervailing powers or institutional control. The notion of commercialization can also have diverse connotations. At one extreme it is a derogatory term indicating that something, often a service that was previously provided as a social activity, is now sold on the market to the detriment of its quality and the numbers who have access to it. Alternatively, others see commercialization as taking an activity out of the hands of state bureaucrats and allowing its provision at an economic price by

2

The term privatization is itself open to many interpretations. In its broadest sense, some interpret it to mean no state involvement at all in the market (DeNuefville, 1999). Here it is taken to mean the non-state ownership of assets; a more widely used definition.

174

KENNETH BUTTON

individuals and private firms that fosters more efficient use of resources and opens wider avenues for investment funding. We make no firm judgment here, other than to consider any move that increases economic efficiency as beneficial, and, taking the Robbins (1932) line, if this results in a socially undesirable distributional impact then it is up to politicians to initiate remedies through taxes and subsidies. We take a neutral view of commercialization in the sense that per se it is neither seen as a good or bad thing and look at it from the perspective of it being an institutional mechanism for achieving particular objectives. In most cases this has involved the introduction of greater market power into decision making either through the relaxation of public regulation or through the reduction in the direct involvement of government in the ownership of air transport infrastructure.3 The commercialization of airline markets came about through a combination of eco­ nomics and political factors. The changes began in the mid-1970s in the US domestic market where a mounting body of empirical academic evidence comparing regulated inter-state routes manifestly demonstrated higher fare levels than comparable unregu­ lated, intra-state fares (Levine, 1965). These coincided with the emergence of new ideas of how competition may influence firms’ behavior, and in particular how ultra-free entry to and exit from a market will reduce the potential for monopoly exploitation even when there is only one supplier; contestable market theory. It was also at a time when the nature of regulation itself began to be challenged and some economists, particularly from the University of Chicago, started questioning whether economic regulation served the pubic interest or the interests of those being regulated and the regulators themselves; capture theory. Overriding these largely academic debates was the macro-issue of the day; rising prices and unemployment (“Stagflation”). Deregulation was seen as a way of reducing prices and thus easing the cost-push pressures linked to Stagflation. It was this macro concern that was the catalyst for bringing about reform at that particular time. The pressures for change in the airline market within the Europe, and notably the European Union (EU)4 , have been somewhat different. The European air transport market, because of the lengths of domestic routes, is primarily international in its nature and thus bound by international agreements, such as those established by the 1944 Chicago Convention, that make regulatory reform more difficult. In addition most of the major airlines within Europe in the 1970s were state owned and often subsidized, in part to facilitate the reaching of international agreements, but also in many cases for doctrinal reasons; countries like France favoring the “Continental Philosophy” of regulation whereby the onus is to demonstrate intervention failure before resorting to the market rather than giving the market primacy.5 Again while there was considerable academic evidence of the failings in the European regulatory structure, highlighted following the use of new powerful statistical and programming techniques showing the

3 Some moves to privatize transportation infrastructure have actually led to the introduction of new economic regulations such as price-capping in the case of BAA (the company owning the main airports) and NATS (the ANSP in the UK). But these regulations are less stringent than those when the assets were state owned. 4 The European Union has enjoyed several titles during its existence but EU will be used throughout for convenience. 5 Additionally, while the founding Treaty of Rome had a Common Transport Policy (CTP) as one its two cornerstones, air transport was explicitly excluded.

COMMERCIALIZATION OF AIR TRANSPORT INFRASTRUCTURE

175

positive effects of reform in the US, it was a macro issue that really galvanize reform In this case the move to a Single European Market following legislation in 1987. The Single European Act initiated a general, phased removal of government regulation across most sectors of the European economy, including airlines. By 1997 the intra-European airline market had, in some ways, less residual economic regulation than the US domestic market regarding, for example, foreign ownership rules. Air transport infrastructure was almost universally, however, largely left out of this liberalization process. Most of it was government owned, either by the national authority, or in the case of some airports by local authorities. The rationale for this varied. In some cases it was simply a legacy effect; the facility had been initiated by a state agency, perhaps for local economic development reasons, and had remained in its hands subsequently. In some cases, and most notably in countries of South America and Africa, airports and air traffic control were seen as of strategic importance and came under the control of the military. Governments also often saw air transport as a political and social integration mechanism and wanted to ensure adequate access even if maintaining particular airports was not commercially viable. There were also arguments, albeit with a rather thin veneer of rigor, with an economic content put forward for public ownership. Some maintained that efficient airport and air traffic control services required large, periodic lump-sum investment that could not be guaranteed by the private sector. Invoking, generally implicitly, one of Adam Smith’s justifications for public ownership, the argument was that state financing, and with it direct control, was needed to meet the strategic investments of large pieces of air transport infrastructure. In some cases it has also been argued that airports and air traffic control services should be allocated on the basis of “need” rather than effective demand to counteract the poor access enjoyed by remoter parts of a country. But perhaps more strongly, there were concerns that airports and ANSPs are natural monopolies and that intervention was needed to prevent them exercising their market power. In itself, this is not an argument for state ownership; many public utilities have this feature and have been directed through regulation rather than ownership.6 Other arguments, such as that air transport infrastructure is a “public good,” are found in debates, both academic and political, but can hardly be taken seriously because it is clear that it is neither non-excludable nor non-rival; the necessary conditions for publicness. The very success of the deregulation and, in many parts of the world, privatization of airlines and other industries has been one of the reasons for the current interest in fostering more commercialization in airport and ANS provision. Demonstration effects can be strong, but more practically the world’s air traffic has been growing rapidly since regulator reforms of airlines were initiated and with this has come pressure on the capacity of infrastructure. There has been an increased appreciation that it has not been provided or used to its best effect and that there are mounting capacity problems (Button and Reynolds-Feighan, 1999). The existing, state dominated systems finding it difficult to manage existing facilities and to finance new ones at a time when tax increases have become politically unpopular.

6

Indeed, the United Nations’ International Civil Aviation Organization has used fare setting and other regulations since its inception in the 1940s to control international air transport.

176

KENNETH BUTTON

The pressure for change has also come as part of a wider move for more commercial­ ization in the provisions of many forms of infrastructure. This generic push has neither, however, been even across countries nor sectors. Intellectually, new analysis of how to best handle infrastructure where there is potentially monopoly power and, often at the same time, the perception of public need for its services has been somewhat slower to be assimilated than developments more germane to competitive and contestable situations. For example, there has been renewed interest involving “competition-for-the-market,” rather than just “competition-in-the-market,” that has involved ideas of bidding for tem­ porally limited monopoly rights to provide and/or operate air transport infrastructure – build, own, operate, and transfer (BOOT) schemes and concessions being examples. This is designed to attract the most efficient suppliers and to extract excess economic rent from these limited monopolies that can then be used by government for transfer purposes. Where regulation of a private monopoly is seen as still needed, lighter handed approaches have been developed, most noticeably price-cap regulations. The challenge in these cases is to establish mechanisms that meet the objectives set – auctions of monopoly rights, for example, can take a variety of forms as can concessionary contracts. There are also stronger and more concentrated vested interests involved in infrastructure provision than in airlines that can stymie immediate reform and institutional structures have to be devised to counter these. From a more direct public policy perspective, infrastructure always tends to be seen as more remote and of less immediate interest to the electorate and thus actions may be delayed. This is particularly true of maintenance and up-grading where there is little political capital to be earned. Many of the problems in the public provision of air transport infrastructure are associated with its efficient use, but the impact of changing institutional structures to embody more commercially oriented stimuli are slow to materialize and less immediately transparent to users of the system. In some cases, the move towards more commercial provision of air transport infras­ tructure service provision has been a logical continuation of developments in the airline and other transport markets. This has effectively been the situation in the UK where the airline, airport, and ANS industries have all largely been moved into the private sector, albeit in some cases with concomitant sets of economic regulations largely to limit any abuse of monopoly power. But in other cases, different and peculiar arrangements have been favored that have left the facilities in state hands but involved mechanisms to pass operations and development to the private sector.

3 NATURE OF COMMERCIALIZATION To understand the exact nature of commercialization of air transport infrastructure and national variations, it is helpful to review some recent developments in economic thinking about institutional structures. There has been a recent up-surge of interest in New Institutional Economics (NIE) that considers economic outcomes within the broader informal and formal structures within which they take place7 . This type of approach

7

See Williamson (2000) for a review of the new institutional economics.

COMMERCIALIZATION OF AIR TRANSPORT INFRASTRUCTURE

177

provides a useful basis for exploring the nature of some of the changes that have taken place regarding air transport. In particular, the NIE places emphasis not only on the shortterm market outcome, which is seen as embodying conventional neo-classical market clearing concepts such as the pricing of landing slots in the airport context, but also on the informal environment in which these contracts are reached – embodying such things as local transactions costs, methods of bargaining, degree of trust, and the de facto interpretation of commercial laws – but also the longer term formal legal constraints that set the boundaries for deal-making that change relatively infrequently. Overshadowing all of this are the norms of the society involved – for example, the Muslim approach to banking is considerably different to that found in western economies and this influences the way that undertakings such as airports are financed and operated, but even within what are normally seen as market economies there are differences in the ways that various countries view private ownership of large scale assets. This latter, long-term, cultural effect is one factor that has led to a relatively faster move towards commercialization of air transport infrastructure in many western societies than in large parts of South America or Asia. Even where commercialization and privatization have taken place there are variations. Whilst many western style economies such a the UK, Germany, Canada, and Australia have engaged in various programs of divestiture of assets to the private sector, sometimes as profit-oriented entities and sometimes as not-for-profits, South American and Sub-Saharan countries have preferred concessions with the state retaining ownership of the air transport infrastructure. Other countries, such as the US that otherwise has a moderately strong market ethos, prefer to operate large parts of their air transport infrastructure in a manner almost akin to the former Soviet Union with state ownership of, for example, its ANS, with finance coming from taxation and rationing by queues rather than price.8 This cultural effect thus influences the degree to which commercialization has taken place and the channels trough which it as been introduced.

3.1 Airports The national approaches aimed at injecting more commercial pressures into the provision of airport services have varied. The complete privatization of airports, or at least the vast majority of their components, is the norm in most countries. Much of the interest that is now emerging is either the privatization of some aspects of airport activity or engaging the private sector in some partnership arrangement with the state. In part this is because an airport is effectively a composite entity comprising of units offering a variety of services – land access, parking, concessions, terminals, runways, ground handling, fire and response units, security, etc. Commercialization does not have to be applied to all these activities, and may be pursued in a piece-meal way if politics or economics dictate. Approaches often differ, for example, according to the state of the local air trans­ port market which is in turn often linked to the stage in economic development of the

8

The FAA has recently begun moving away from this approach in terms of tower control whereby some facilities have been outsourced to private providers.

KENNETH BUTTON

178

Traffic

Growth

Simple economic regulation

Developing countries • Increased capacity of the airport system • Large share of revenues from airside charges

Developed countries

Complex

economic

regulation

• Maximum revenue base with limited passenger growth • Large share of revenues from commercial services

Share of Commercial Revenues

Figure 1 A Generalization of Airport Trends in Developing and Developed Countries. Source: Adopted from Juan (1995).

country concerned. For example, airports can vary in terms of their potential revenue flow from different sources and this can affect the degree of privatization or deregu­ lation that is possible and the form it is most likely to take. Figure 1 offers a fairly simply representation of what seems to be going on regarding developed and developing countries. Much depends on the state of the regional air transport market. The forecasts of relatively slow longer term growth in air traffic in and between the developed countries (e.g., estimated at about 3.6% a year to 2005 within North America, 3.4% within Europe, and 4.5% between North America and Europe by Boeing Commercial Airplane) means that their major airports will increasingly become dependent on commercial or nonaeronautical revenues to enhance their revenue stream.9 This in turn can pose problems in terms of regulation as has, for instance, already been seen in the debates over the imposition of the price-capping regime used in the UK. While still relatively small, the protected growth of many air markets involving developing countries (e.g., 6.9% within Latin America, 8.8% between Latin America and the Asia-Pacific region, and 8.7% between Latin America and Africa) offers the potential for increased airside revenue in situations where there are potentially fewer social constraints involving such things as noise and land-take on building additional capacity or where there is already adequate capacity. The scope for raising significant commercial income is much less, however, because of the lower initial traffic base.10 It also suggests, though, that the regulatory regime overseeing a privatized airport system needs to be less sophisticated because it only has to deal with airside issues. The potential for various forms of regulatory capture, a phenomenon not unknown in many developing countries, is thus smaller.

9 Additionally, there are capacity issues in many developed markets that are unlikely, for a variety of reasons,

including environmental concerns, to be resolved through the provision of additional facilities.

10 Many large international airports generate up to 60% of the income from non-aviation activities.

COMMERCIALIZATION OF AIR TRANSPORT INFRASTRUCTURE

179

The problem for many of the poorest developing countries, however, is that even though their air traffic flows may in aggregate be growing, this is from a low base and they still seldom generate sufficient revenue to cover the full costs of operations – airports are essentially decreasing cost entities where cost recovery can be difficult especially in a situation where there is competition from other airports. This makes pure privatization options less tenable and the need for outside assistance from aid agencies or from government more relevant. Private/public partnerships offer another alternative. The wide variety of circumstances around the world has led to a diversity of approaches to commercialization of airports. Some have involved complete divestiture of former state assets to the private sector, albeit normally with some continuing oversight of how the airport is operated, but in other instances the withdrawal of the state has been less complete. Table 2 provides some insights into the various options that are available and gives examples of changes in governance that have occurred. The management contract approach retains government control but contracts out air­ port activities for periods, normally by some form of auction, specified elements of airport services; parking, hotels, retail concessions, etc. The system is very much in line with the notion of “competition for the market.” Long-term contracting involves giving over, following a tendering process, the operational side of an airport, sometimes including investment commitments in additional capacity, for an extended period with the authorities retaining a degree of strategic control. The degree of financing required normally entails bringing specialized international companies with the expertise to man­ age an airport, or system of airports, together with financial houses that can provide Table 2 Types of Airport Governance Form of governance Control device

Management contract Periodic tendering

Long-term contracting Rate-of-return regulation

Rate-of-return regulation plus market for corporate control

Management responsi­ bility Time frame

Operational

Operational and financial

Operational, financial, and strategic

5–10 years

15 plus years

Examples

Management concessions • Burbank (5 years) • Indianapolis (10 years) • Westchester, New York (10 years)

Build-operate-transfer • Toronto Terminal 3 • Ataturk Long-term lease • Bolivia (3 airports for 25 years) • Argentina (33 airports for 30 years) • JFK New York Termi­ nal 4 (30 years) • Macao (33 years)

99 year lease or indefinite Initial public offerings (IPOs) • UK – BAA (100%) • Vienna (27%) • Copenhagen (25%) Trade sales • Sydney (100%) • Auckland (25%) • Naples (30%)

Source: Carney and Mew (2003).

Full/part privatization

180

KENNETH BUTTON

the necessary support for large scale service activities. These types of concessions are widespread in South America where local expertise and finance is limited but there is a reluctance for the state to divest itself of aviation assets. Many developed countries have also pursued similar philosophies when expanding the involvement of private enterprise, but falling short of the complete divestiture favored by the UK. Many US airports have adopted various concessionary schemes – for example, Boston, Pittsburgh, and Reagan National Airport, Washington have entered into con­ cessionaire agreements, for the entire operations at Pittsburgh and for specific terminal buildings at the other airports. At Chicago O’Hare airport, parking has been contracted out, and the Port Authority of New York and New Jersey that own a number of airports have a variety of agreements covering such things as the operation of terminal buildings and the supply of heating and cooling at some of its facilities. US airports in general also practice considerable unbundling of activities and there is, for example, a tradition of significant airline involvement in providing check-in facilities and baggage systems. Complete privatization of major airports is uncommon, although some of the larger facilities are now in private hands. In most cases there is concern that a privately owned airport will exercise its monopoly power to extract rent from customers and thus regulatory controls. The challenges are to device and operationalize appropriate regulatory regimes to monitor and direct these large companies in the public interest – often price-capping is deployed.11 The on-going debates about full privatization concern such things as whether single airports, or as with the UK’s BAA, systems of airports, should be privatized and when they are privatized what should be regulated; should it be all airport activities or just those directly aviation related? In their overview of these various governance options, Carney and Mew (2003) focus correctly on the government being involved in both seeking to improve the efficiency of their airports but at the same time trying to direct the gains to particular groups rather than leaving management with full autonomy in their actions.12 This involves complexities that, while common to businesses in developed countries, are unfamiliar in many parts of the world. As a result, this has added to the growth in international firms specializing in airport management, including ownership, to allow the development of common, best-practice methods of operation while at the same time being innovative in creating bespoke models for different circumstances.13

11

For expositional reasons, a more general discussion of the theory behind regulation of monopoly infras­ tructure suppliers is reserved until we deal with ANSP commercialization. The evidence regarding airports is that rate-of-return regulation has not proved successful in Germany (Kunz and Niemeier, 2000), while the price-capping regime is seen as inappropriate for London (Beesley, 1999). A particular issue in the UK is whether the price-cap should be applied to the entire set of activities of the BAA (the “one-till approach”) or whether airside activities should be separated out and be regulated (the “two-till approach”). For details see Starkie (2001). 12 The authors do not, however, spend much time on situations where state-controlled activity of an existing activity is retained but is opened up to private sector competition – for example, the requirement in the EU for airport ground handling to be open to competition. Nor do they consider situations where a large state system, such as that former found in Canada, is divided up between smaller units, largely municipalities in the Canadian case with competition allowed to develop between them. 13 Freathy (2004) offers a similar type of breakdown but focuses more on the role of commercial activities at airports in his analysis.

COMMERCIALIZATION OF AIR TRANSPORT INFRASTRUCTURE

181

3.2 Air Navigation Systems Provision The majority of the world’s ANSPs have traditionally been state owned and operated with objectives akin to those of public utilities; safety, common carrier obligations, allocation by need, etc. Recent trends from the mid-1990s have seen moves to engender more commercial approaches to the supply of ANSs (Charles and Newman, 1995). Table 3 provides details of the existing ownership structures of 11 major ANSPs together with some information of when structures have been created and modified. The economic institutional structures in which they operate vary by form of ownership and by the nature of prevailing rate controls – see Button and MacDougall (2006) for more details. From a financial risk-taking perspective, there are also differences as to whether State Table 3 Basic Features of Selected Air Navigation Service Providers Country

ANSP Name

Ownership

Rate Regulation#

Australia1

Airservices Australia

Government corporation

Commission oversight

Canada∗

NAV CANADA

Not-for-profit private corporation

Legislated principles/appeals

France∗∗

Direction des services de la navigation Aérienne (DSNA)

State department

Approved by transport ministry

Germanyç

Deutsche Flugsicherung Government GmbH (DFS) corporation

Approved by transport ministry

Ireland§

Irish Aviation Authority

Government corporation

Regulatory commission

Netherlands¶

Luchtverkeersleiding Nederland (LVNL)

Not-for-profit Approved by transport government corporation ministry

New Zealandy

Airways Corporation of Corporation New Zealand

Self-regulating/appeals

South Africa

Air Traffic and Navigation Services Ltd

Not-for-profit joint-stock corporation

Transport ministry committee

Switzerland¥

Skyguide

Not-for-profit Approved by transport government corporation ministry

United Kingdom� National Air traffic System, Ltd

Public/private partnership

Price capping

United States

State department

Financing from taxation

FAA’s Air Traffic Organization

Corporatized in 1988; ∗ Corporatized in 1996; ∗∗ Consolidated in 2003; ç Established in 1993 and was to be privatized in 2006 but since aborted; § Corporatized in 1993; ¶ Corporatized in 1993; yCorporatized in 1987; ¥ Incorporated in 2001, predecessor established in 1921 � Public/private partnership in 2001; # Excluding national, generic anti-trust and similar regulations. ALL ANSPs are financed by user fees except for the US Federal Aviation Administration that is funded by taxation. Source: Button and Dougall (2006). 1

182

KENNETH BUTTON

debt guarantees are offered; they are not except in the cases of Australia, France and Switzerland – the issue is not applicable in for the US where taxes are the current mode for finance. There has been a gradual but perceptible shift away from state ownership with a variety of institutional structures emerging. Ownership is clearly, from the political economic perspective, not seen as a neutral matter (Button and McDougall, 2006). Corporatization has become a popular option to distance government from the operations of ASNPs and their financing, although the formulation varies; in some case it involves a the creation of a free-standing entity whereas in others the undertaking is linked to the state. This can be seen as a reflection of wider national approaches to the topic with some countries having a tradition of preferring particular types of corporate entities and have experiences of their workings. In some cases there are explicit rate-of-return conditions – namely non-profitability – built into the terms under which the corporation is established. Rate-of-return regulation in various guises was a characteristic of US industrial policy and that of many other countries until the 1980s. It largely became discredited because of manifest evidence of X-inefficiency when it has been applied and the longer terms effects of over capitalization that accompanied it – the so-called Averch–Johnson effect.14 The X-inefficiency arises largely because of the ability of the regulated enterprise to capture the system through control over information flows about costs and allowed them to enjoy significant inert areas without the pressure for full cost efficiency. The Averch–Johnson effect comes about because rate-of-return regulation creates a bias that leads to excessive capital intensity. Nevertheless, a number of countries now practice an explicit form of rate-of-return regulation (e.g., Canada, Netherlands, South Africa, and Switzerland) over their ANSPs by adopting either private or public “not-for-profit” regulation. New Zealand also practice rate-of-return regulation whereby the ANSP returns money to users once costs have been recovered and an agreed “profit” has been paid to the government. Price-capping, as developed in the UK by Littlechild (1983) for the newly priva­ tized telecommunications sector has been preferred more recently because of its lower informational needs and because it is directly aimed at minimizing X-inefficiency, both static and dynamic. It entails the setting of a maximum average price across a bundle of outputs that is related to changes in general price levels. In the transport context it is the preferred method for regulating the BAA in the UK, and is now to be applied to the privatized NATS (Goodliffe, 2002; Majundar and Ochieng, 2003) and was to be used as the tool to regulate, the now aborted public-privatization of the German ANSP provider, (Classen, 2007). Whether price-capping is appropriate for regulating ANSPs (or for that matter airports) depends on a number of factors. The simplicity of price-capping diminishes as X-inefficiency is driven from the system; as this occurs it effectively converges on a rate-of-return regulation. But it is also most effective when supply is highly flexible. If there are shortages, as does occur at airports and in air-space because of invisibilities or other rigidities in supply, price should be used as a “con­ gestion” charge. A price-cap regime to bring down the cost of use over time is not the instrument to use in these conditions and its rigidity reduces the ability to allocate scarce

14

In the particular context of its use for airports, Tretheway (2001) also points to its complexity, high administration costs, and lack of responsiveness.

COMMERCIALIZATION OF AIR TRANSPORT INFRASTRUCTURE

183

supply effectively. Indeed, most of the early advocates of price-capping seem to have viewed it as a short-term expedience where there was rapidly expanding capacity and while genuine competition built up; hence its use in telecommunications and energy. Other countries have adopted less structured forms for regulating their ANSPs. Some have commissions that monitor the rates that are levied; a government ministry fulfills this role in a number of cases. In the case of Canada, there are appeals procedures that effectively lead to regulation by a judicial process that can review changes in the rates levied by ANSPs to assess whether they violate a set of specific charging principles. The Australian Competition and Consumer Commission that regulates the prices of Airservices Australia does not have powers of price-capping, but only of giving opinions as to the appropriateness of the price increases. In South Africa the ANSP is prohibited from levying or increasing an air traffic service charge unless it has permission from the Economic Regulatory Committee. New Zealand and Ireland largely leave it to the market. Financing investment in expanding and modernizing ATC systems also takes a variety of forms. The US’s Federal Aviation Administration (FAA) is largely funded from a variety of taxes, notably the federal ticket tax and the federal flight-segment tax, with no explicit user charges being levied. It has no access to the private capital market.15 The French ANSP provider operates in a more commercial way by levying user charges and by having recourse to the private capital market, although there is oversight as to investment levels. NATS, DFS, NAV Canada, and Airservices Australia, for example, borrow extensively in the market.

4 IMPACTS OF COMMERCIALIZATION The extent of commercialization of air transportation infrastructure, albeit in a variety of ways, has been seen to be growing but it is still limited in many countries. Perhaps more of a problem from an analytical perspective is that most of the change is relatively recent, and economic data to study it is sparse and not always of prime quality.16 Unlike airline markets it takes time for changes in infrastructure institutions to have their full effects felt; there is thus a need for fairly long and consistent sets of data. There has also been a lot of experimentation that where initial problems have arisen, it has resulted in further institutional reforms being initiated. Separating out the effects of individual elements of such iterative actions is challenging. And overriding all of this, the air transportation market has, or at least one hopes it has, not been typical in recent times due to the unprecedented collapse in demand after the events in the US of September 2001.

15

Gloaszewaki (2002) outlines the nature of the US ANS and also gives a comparative analysis of the interactions between ANSP and airport regulations in the US and Europe. 16 Although privatization often forces providers to adopt standard accountancy conventions, rather than the ad hoc systems used by public entities, commercial confidentiality generally means that companies release less data. Some databases are being developed. For example the Air Transport Research Society does a worldwide survey of airports but participation is far from complete and the data itself poses problems – for example, there are erratic changes in the ranking of airports.

184

KENNETH BUTTON

4.1 Airports The diverse ways in which commercialization has been introduced to airport activities, and the numerous objectives motivating them, combined with the large number of elements that make up airport services makes it difficult to come to firm conclusions about the success of recent changes. Added to this, there has always been considerable variation in the levels of efficiency even among the state owned airports and in their reported financial positions. The subject of the relative efficiency of different governance models has, however, attracted quite a lot of recent interest from academics and policymakers, in part because economic data, albeit not always completely sound, are now becoming available for analysis and because techniques for looking at relative efficiency of multi-product undertakings, such as data envelopment analysis (DEA), are now much easier to apply with the advent of appropriate computer software suites. Table 4 provides brief details of some of the main academic work that has been done looking at various efficiency aspects of airport commercialization around the world. It is not an exhaustive listing, there is inevitably a large gray literature on the subject in the hands of the financial and management interests participating in the commercialization process, and the brief table inevitably does not do justice to the various studies that are reported. Most of the work focuses on privatization, either in a time-series context of before-and-after analysis or in a cross-sectional framework comparing privatized with state-owned airports. Much less work has been done on changes in regulatory regimes, outsourcing of some airport activities, or on the transitional effects of moving to a more commercialized governance structure. While the quality of the analysis inevitably varies because of data constraints (airports use a diversity of methods for cost accounting), the statistical or programming method­ ology adopted, the detailed nature of the commercialization measures, and the simple quality of scholarship applied, the general picture that emerges is that commercialization often does lead to the more efficient provision of airport services, albeit with a range of caveats. In many cases, for example, changes in governance are important but are overshadowed by other changes that are taking place such as increased traffic levels and the amount of traffic hubbing at an airport.17 Taking a conservative position, what the studies do show is that the introduction of commercial pressures, almost irrespective of their nature, certainly does not decrease the efficiency of airports and would, on average, seem to act to increase it. What perhaps this type of table misses are the nuances of individual situations; as the saying goes “The devil often lies in the detail.” The privatization of the UK’s main airports, and especially those around London, for example, seems to have met the main criteria of generating revenues for capacity enhancement, but issues remain about the role of regulation (whether it should continue to only cover air operations charges) and over the single authority ownership of three large airports in the area. Concessions, in their diverse forms, are now a popular form of injecting private finance into airports and of stimulating a more commercial approach to their management. They

17

From a commercialization perspective, findings regarding the underlying economic nature of airports should not be discounted. Studies, for example, do indicate economies of scale exist and this, in itself, is important to public authorities seeking the appropriate way to structure a commercialization strategy.

COMMERCIALIZATION OF AIR TRANSPORT INFRASTRUCTURE

185

Table 4 Studies of the Impacts of Airport Institutional Reforms Study

Case

Topic

Findings

Juan (1995)

10 airport case studies

Institutional comparison of approaches being adopted towards privatization including 6 lower income countries

Privatization is more efficient that corporatization; the need for different approaches to profitable and unprofitable airports; the specification of concession contracts needs to be carefully thought through

US General Accounting Office (1996)

Airports in 50 countries

Looking for lessons for US privatization

Limited evidence regarding performance although privatized facilities generated more revenue

Parker (1999)

BAA

Pre- and postprivatization DEA analysis

Found privatization had no impact on technical efficiency

Serebrisky and Presso (2002)

Argentinian airports

Considered the vertical integration between airport and airline in a concession arrangement

Vertical integration can lead to market distortions as there are not appropriate regulations to control monopoly powers

Hooper (2002)

Asian airports

Financial aspects of privatization

There is a need for an appropriate set of controls to be established before airports are privatized

Forsyth (2002)

Australia and New Zealand airports

Institutional analysis of reforms

Australia: Problems of handling investment decisions when there is price capping. New Zealand: Despite lack of regulation acting as if there were rate-of-return controls

Niemeier (2002)

Hamburg airport

Examination of efficiency of regulatory reforms

Price-capping is superior to rate-of-return regulation

Bosch and García-Montalvo (2003)

Latin America airports

Review of issues of nondiscriminatory access to airports using secondary sources

The problems of Latin American airports are ultimately similar to many of those being encountered in the European Union

Forsyth (2003)

Australian airports

Replacement of price caps by price monitoring

Unclear as to the effects on long-term changes

Holvad and Graham (2003)

UK airports

DEA of relative efficiency of privatization

Privatization increased efficiency

(Continued)

KENNETH BUTTON

186

Table 4 Studies of the Impacts of Airport Institutional Reforms—Cont’d Study

Case

Topic

Findings

Hanaoka and Phomma (2004)

Airports in Thailand

Comparison of fully state owned and partially privatized airports

No clear-cut difference in productivity due to ownership

Oum et al. (2004)

Airports from around the world

Assessment of economic regulation regimes using factor productivity analysis

Rate-of-return regulation leads to excess capacity; price-capping leads to under investment

Pacheco et al. (2006)

Brazilian airports DEA to look at efficiency of managerial changes in the lead-up to privatization

The performance of Ifraero improved as it prepared for privatization

Lin and Hong (2006)

20 large airports from around the world

DEA analysis to examine importance of ownership on efficiency

Ownership has little impact on an airport’s efficiency

Andrew and Dochia (2006)

Global privatization initiatives

Statistical analysis of airport privatization in low and medium income countries

The preferred institutional structure involves concessions rather than divestiture

Vogel (2006)

European airports

Analysis of cost efficiency and financial performance of privatized airports

Generally, privatized airports are more cost efficient but do not earn a higher return

Vogel and Graham (2006)

31 European airports

Examination of factors affecting performance using DEA analysis

Ownership status affects economic performance

Oum et al. (2006)

Large airport around the world

Regression analysis to look at productivity differences

Evidence that airports with majority government ownership leads to less efficiency

Lipovich (2007)

Argentinean airports

Concession issues

Difficulties emerged in the concession allocation process that led to an unrealistic financial structure

Low and Tang (2006)

Asian airports

Degree of factor substitution as outsourcing takes place

Outsourcing has allowed airports to become more adaptive to cost changes

COMMERCIALIZATION OF AIR TRANSPORT INFRASTRUCTURE

187

have generated different types of challenges that have been the subject not only of qualitative analysis as seen in Table 4, but also of wider institutional work. The number of different forms that concessions take make generalizing problematic; not only are they for different time periods but they may allow or disallow for public sector or foreign participation, may be for a single or for a group of airports, may involve commitments for investment, etc. Nevertheless, experiences do provide insights. There are often problems in meeting the terms of concessions. In 1998, for example, Argentina gave a 30-year concession, with a possible 10 extension, to a single consor­ tium, Aeropuertos Argentinia 2000 to run 32 of its main airports. A commitment to major investments in the system was part of the agreement but also was a large annual payment ($171.121 million in 1998 prices and up-dated periodically for inflation) that has proved difficult to meet at approved fee levels. This has led to renegotiations and suggestions that the initial concession arrangements were flawed and expectations were unrealistic. There is also significant political risk in some cases. In Venezuela, the responsibility for airports has been transferred from central to state governments and the latter that have then engaged in trying to bring in private finance. In 1992 three airports in Zulia state were privatized but then subsequently taken back by the governor after a change in the state government.

4.2 Air Navigation Service Providers Compared to airports, there have been relatively few attempts at full privatization of ANSPs, although efforts to make them more commercially accountable are more widespread. The experiment in the UK with the NATS public–private partnership is too recent to pass substantive comments on, although the need to refinance it almost imme­ diately after its initiation following the September 11 attacks does suggest that quite high levels of capitalization may be required to avoid major, unexpected financial “hits.”18 The same problems arise with the various efforts to corporatize airports as not-for-profit entities; essentially a form of government or private sector business subject to zero rate of return regulation. The main ones are relatively new and found it difficult to balance their books after 2001 when with very limited reserves they were confronted with falling traffic but significant fixed costs to recover. Recent political and academic interest in making ANSPs more commercially oriented has come largely from the US, where the system under FAA control has met with criticism and is, in 2007, due for review, and from the European Union, where there are policies to bring together the disparate national providers within a more structured framework – a Single European Skies. The main concern in EU and its associates has only marginally involved commercialization issues – there are more basic challenges in developing a coordinated ANS currently embracing 34 ANPS. Indeed, this focus is reflected in the detailed benchmarking done by EUROCONTROL (2006) that involves considerable analysis of the complexities confronting various ANSs and provides a four

18

NATS initially had a 115% debt gearing.

188

KENNETH BUTTON

level ranking of cost-effectiveness but makes no comment on the implications of the various forms of governance covering the different systems. Much of the quantitative work on ANSP commercialization has been cross-sectional in approach – there is simply not enough long-term data to conduct time series analysis. It has also tended to lack any real statistical rigor because of data constraints. Studies of five ANSPs by the US Government Accountability Office (2005a,b) suggests that commercialization was particularly effective in allowing systems to finance and carry through modernization programs. Findings supported in a larger study of 11 systems from around the world by Button and McDougall (2006). Indeed, it was this need for financing modernization that motivated the formation of the NATS public–private partnership structure in the UK (UK House of Commons Committee of Public Accounts, 2002/3); the state owned undertaking could only cover about half of its investment needs from operating revenues. One reason for this improved financing is the greater flexibility non-state owned can enjoy in gaining funds but still at a relatively low cost – privatized and corporatized systems typically enjoy very high credit ratings, although often less than for state owned entities going to the capital market. Government debt guarantees can reduce the costs of borrowing for privatized and corporatized ANSPs (Magdalena, 2005). DFS the German provider and Airservice Australia, for example, have their debt guaranteed and enjoy exceptionally high credit ratings (although they do also have very conservative financial portfolios).19 The impact of commercialization on costs incurred by the ANSPs and, subsequently on users of the systems, provides the ultimate test for economic efficiency. There is, however, little rigorous econometric analysis exploring links between governance structure and the fees paid for ANSs. Button and McDougal, albeit limited to basic data analysis, find some evidence suggesting that there is less “gold-plating” when there is commercial pressure on suppliers, with investment tending to be reigned in. As seen in Figure 2, the pattern of rates charged over time by various ANSP providers reveals falling generally rates until 2001 when rates rose to permit fixed cost recovery from a considerably reduced traffic volume.20 There is little consistency in the patterns that emerge for the various types of governance regime; perhaps an inevitable situation over has been a relatively short and volatile time period. Other indicators of performance suggest that commercialization has not led to any deterioration in service quality. All the providers examined by the US Government Accountability Office and by Button and McDougall significantly increased the traffic that they handled in the early years of the twenty-first century, and in the case of the commercialized entities, without any increase in air traffic management induced delays; indeed the state owned FAA seems to have encountered the biggest problems in this area since traffic has picked up in the aftermath of the events of September 11th and the SARS fears.

19

Strong and Oster (2007) looking at the US’s FAA, NATS, and NAV Canada, and find that although the governance models are very different they all “appear to have more sustainable business models and organizational structures to meet the challenges of air traffic management in coming years.” 20 This, of course, goes against what one wants in these circumstances when there is a need to counteract falling demand with lower input prices to the air transportation market.

COMMERCIALIZATION OF AIR TRANSPORT INFRASTRUCTURE

189

140 130 SkyGuide

120

Airservices Australia

110

Airways NZ

100

DFS

90

NATS U.K.

80

Irish Aviation DSNA

70

LVNL

60 50 1997

1998

1999

2000

2001

2002

2003

2004

Year Note: All data based on 1997 except for NATS UK 2001 = 100. (a) En route unit rates 180 160 SkyGuide

140

Irish Aviation

120

LVNL Airservices Australia

100 80

DFS

60 40 1997

1998

1999

2000

2001

2002

2003

2004

Year Note: Initials as defined in Table 3. (b) Terminal unit rates

Figure 2 Charges for En route and Terminal Services in Constant 2004 Prices.

5 CONCLUSIONS Air transport is now an important mode for the international and, in many countries, domestic carriage of people and cargo. Reforms in the airline services market, to inject more commercial vitality into the sector, have largely proved beneficial; but devel­ opments regarding airports and ANSs have been slower to materialize, piece-meal in nature, and often only partial in their coverage. This is not atypical in the sense that in many other sectors where commercialization has been fostered, the infrastructure elements have been the last to be reformed. As we have seen, changes are now taking place and they have taken a variety of forms. One may see this as experimentation but

190

KENNETH BUTTON

equally, it may be seen as tailoring institutional structures to local politico-economic realities. Studies are still not conclusive as to the verdict offered on the reforms. Infrastructure is by its nature fixed and changing the way it is provided and operated inevitably has limited short-term impacts. The airport experiences suggest that there are numerous ways to make providers more commercially oriented, and that this may be done in a bigbang approach by throwing the entire enterprise directly into the market or piece-meal and gradually by unbundling services over time and allowing these to be subjected to market in turn. The network nature of ANSs makes this more difficult, although some countries do employ concessionary arrangements for putting commercial pressures on the provision of tower services and the supply of equipment. Perhaps one last point to make is that much of the public and political debate about commercialization of air transportation, both airlines and infrastructure, has focused on the potential safety implications of relaxing economic regulation or of privatization. The one solid thing that has emerged from a wide range of studies is that safety has not been adversely affected by any of the institutional changes – the privatized and corporatized systems have no worse safety record than those in government ownership. Indeed, by most measures air transportation has become safer over time because of improved technology and any recent commercialization initiatives do not seem to have affected this trend. One reason for this has been parallel initiatives to improve safety oversight and to initiate the introduction of new, safer technologies as they have come on-line. The policy-makers seem to have been very successful in separating the impacts of economic reforms from those of a more social nature. Another factor is that privatized or commercialized providers have market incentives to ensure safety standards are met; any accident will almost inevitably affects their financial returns.

REFERENCES Andrew, D and Dochia, S. (2006) The growing and evolving business of private participation in airports: new trends, new actors emerging, Gridlines Note 15, World Bank, Washington, DC. Beesley, M.E. (1999) Airport regulation, in M.E. Beesley (ed) Regulating Utilities: A New Era? Institute of Economic Affairs, London. Bosch, A. and García-Montalvo, J. (2003) Free and Nondiscriminatory Access to Airports: Proposal for Latin America, Inter-American Development Bank, Washington DC. Button, K.J. and McDougall, G. (2006) Institutional changes in air navigation service-providing organizations, Journal of Air Transport Management, 12, 236–252. Button, K.J. and Reynolds-Feighan, A. (1999) An assessment of the capacity and congestion levels at European airports, Journal of Air Transport Management, 5, 113–134. Button, K.J. and Stough, R. (2000) Air Transport Networks: Theory and Policy Implications, Edward Elgar, Cheltenham. Carney, M. and Mew, K. (2003) Airport governance reform: a strategic management perspective, Journal of Air Transport Management, 9, 221–232. Charles, R.A. and Newman, H.K. (1995) Public policy and technology management: changing the role of government in the operation of air traffic control, Transportation Journal, 35, 39–48. Classen, A.B. (2007) Privatization of Air Navigation Services in Germany – an up-to-date case study, paper to the 11th World Conference on Transport Research, Berkeley.

COMMERCIALIZATION OF AIR TRANSPORT INFRASTRUCTURE

191

DeNuefville, R. (1999) Airport privatization issues for the United States, Transportation Research Record, 1662, 24–31. DeNuefville, R. and Odon, A. (2003) Airport Systems: Planning, Design, and Management, McGraw-Hill, New York. EUROCONTROL (2006) ATM Cost-effectiveness (ACE) 2004 Benchmarking Report, EURO­ CONTROL, Brussels. Forsyth, P. (2002) Privatisation and regulation of Australian and New Zealand airports, Journal of Air Transport Management, 8, 19–28. Forsyth, P. (2003) Regulation under stress: developments in Australian airport policy, Journal of Air Transport Management, 9, 25–35. Freathy, P. (2004) The commercialization of European airports: successful strategies in a decade of turbulence, Journal of Air Transport Management, 10, 191–197. Gloaszewaki, R. (2002) Reforming air traffic control: an assessment from the American perspec­ tive, Journal of Air Transport and Management, 8, 3–11. Goodliffe, M (2002) The new UK model for air traffic services – a public private partnership under economic regulation, Journal of Air Transport and Management, 8, 13–8. Hanaoka, S. and Phomma, S. (2004) Privatization and productivity performance of Thai airports, paper to the 8th Air Transport Research Society World Conference 2004, Istanbul. Hooper, P. (2002) Privatization of airports in Asia, Journal of Air Transport Management, 8, 289–300. Holvad, T. and Graham, A. (2004) Efficiency measurement for UK airports: an application of data analysis, Empirical Economics Letters, 3. Juan, E.R. (1995) Airport Infrastructure: The Emerging Role of the Private Sector, Cofinancing and Financial Advisory Services, World Bank, Washington DC. Kunz, M. and Niemeier, H.M. (2000) regulating airports: some lessons from the UK and Germany, paper to the 4th Air transport Research Group Conference, Amsterdam. Levine, M.E. (1965) Is regulation necessary? California air transportation and national regulatory policy, Yale Law Journal, 74, 1416–1447. Lin, L.C. and Hong, C.H. (2006) Operational performance evaluation of international major airports: an application of data envelopment analysis, Journal of Air Transport Management, 12, 342–351. Lipovish, G.A. (2007) The privatization of Argentinean airports, Journal of Air Transport Man­ agement (forthcoming). Littlechild, S. (1983) Regulation of British Telecommunication’s Profitability, Department of Industry, London. Low, J.M.W. and Tang, L.C. (2006) Factor substitution and complementarity in the Asian airport industry, Journal of Air Transport and Management, 12, 261–266. Magdalena, R. (2005) Peer Comparison: Global: Air Traffic Controllers, Standard and Poors, New York. Majundar, A. and Ochieng, W. (2003) The part privatization of NATS: the experience one year after, Journal of Air Traffic Control, June, 28–34. Morisson, S.A. and Winston, C. (1995) The Evolution of the Airline Industry, Washington DC, Brookings Institution. Niemeier, H.M., 2002, Regulation of airports: the case of Hamburg airport – a view from the perspective of regional policy, Journal of Air Transport Management, 8, 37–48. Oum, T., Zhang, A. and Zhang, Y. (2004) Alternative forms of economic regulation of airports, Journal of Transport Economics and Policy, 38, 217–246. Oum, T.H., Adler, N. and Yu, C. (2006) Privatization, corporatization, ownership forms and their effects on the performance of the world’s airports, Journal of Air Transport Management, 12, 109–121.

192

KENNETH BUTTON

Pacheco, R.R., Peixoto de Sequeire Santos, M. and Fernades, E. (2007) The performance of Brazilian airports based on management style, Journal of Air Transport Management (forthcoming). Parker, D. (1999) The performance of BAA before and after privatization: a DEA study, Journal of Transport Economics and Policy, 33, 133–146. Pels, E., Nijkamp, P. and Rietveld, P. (2003) Inefficiencies and scale economies of European airport operations, Transportation Research E, 39, 341–361. Poole, R. (2005) Commercializing air traffic control: a new window of opportunity to solve and old problem, Regulation, 20, 1–12. Robbins, L. 1932, An Essay on the Nature and Significance of Economic Science, Macmillan, London. Serebrisky, T. and Presso, P. (2002) An Incomplete Regulatory Framework? Vertical Integration in Argentine Airports. 37th Meeting of the Argentine Political Economy Association. Starkie, D. (2001) Reforming UK airport regulation, Journal of Transport Economics and Policy, 35, 119–135. Strong, J. and Oster, C (2007) Air Traffic Management under Stress: The Performance of Air Navigation Providers in Canada, Britain, and the United States, paper to the 11th World Con­ ference on Transport Research, Berkeley. Tretheway, M.W. (2001) Airport Ownership Management and Price Regulation, Report to the Canadian Transportation review Committee, Ottawa. UK House of Commons Committee of Public Accounts (2002/3) The Public Private partnership for National Air Traffic Services Ltd, The printing Office, London. US General Accounting Office (1996) Airport Privatization: Issues Related to the Sale or Lease of US Commercial Airports, GAO/T-RCED-96-82 GAO, Washington DC. US Government Accountability Office (2005a) Air Traffic Control: Preliminary Observations on Commercialized Air Navigation Service Providers, GAO-05-542T, GAO, Washington DC. US Government Accountability Office (2005b) Air Traffic Control: Characteristics and Perfor­ mance of Selected International Air Navigation Service Providers and Lessons Learned from their Commercialization, GAO-05-769, GAO, Washington DC. Vogel, H.-A. (2006) Airport privatization: ownership structure and financial performance of European commercial airports, Competition and Regulation in Network Industries, 2, 139–162. Vogel, H.-A. and Graham, A. (2006) A comparison of alternative airport performance measurement techniques: a European case study, Journal of Airport Management, 1, 59074. Williamson, O. (2000) The New Institutional Economics: taking stock, looking ahead, Journal of Economic Literature, 38, 595–613.

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

8 The Role of Regional Airlines in the US Airline Industry Silke Januszewski Forbes∗ , Mara Lederman†

ABSTRACT We describe the role of regional or commuter airlines in the US airline industry. Most major carriers subcontract service on low-density short and medium-haul routes to regional airlines which operate as separate companies. We describe the history of these regional airlines and their growing role for passenger traffic in the last decade. We explore the organizational relationships between majors and regionals and the contribution of the regional jet to the growth of regional airline service.

1 INTRODUCTION In 2005, US regional airlines – carriers which operate aircraft with fewer than 90 seats – carried almost 135 million passengers or approximately one in five domestic travelers. They completed over 14,000 daily departures and had a combined fleet of over 2,700 aircraft. Regional airline service has increased steadily over the past decade and there are no indications that this trend is likely to slow. Yet, despite the increasingly important role played by regional airlines, this segment of the industry has received surprisingly little attention from airline economists. Indeed, the academic literature has traditionally focused on the roles of large network carriers and, more recently, the so-called “low-cost carriers”, leaving regional airlines virtually untouched. This chapter documents the role

∗

Department of Economics, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093­ 0508, USA. E-mail: [email protected] † Rotman School of Management, University of Toronto, 105 St. George Street, Toronto, Ontario, Canada, M5S 3E6. E-mail: [email protected]

194

SILKE JANUSZEWSKI FORBES AND MARA LEDERMAN

and extent of regional air service in the US commercial aviation industry and introduces some of the key economic issues that affect regional airlines. The organization of the chapter is as follows. In Section 2, we discuss the role of regional airlines and document the extent of regional airline participation in the industry. In Section 3, we discuss the history of regional airlines, tracing their origins to the small, unregulated air taxi operations that served small communities during the pre-deregulation era. Section 4 discusses the nature of the relationship between regionals and the major car­ riers with which they partner and highlights some of the organizational economics issues that these relationships raise. Section 5 describes the emergence and diffusion of the regional jet (RJ). A final section briefly speculates on the future role of regional airlines.

2 THE ROLE OF REGIONAL AIRLINES In the United States, regional airlines operate short- and medium-haul scheduled airline service, often connecting smaller communities with larger cities. Almost all regional airlines operate under codeshare agreements with one or more major carriers.1 Under these agreements, the regional operates flights on behalf of the major carrier, who markets and tickets these flights under its own two-letter flight designator code. Typically, no tickets are sold under the regional’s own code. In addition to using the major’s code, the regional’s flights also share the major’s brand. For example, the regional’s planes are painted in the major’s color schemes, the regional’s flight attendants wear the uniforms of the major, passengers traveling on the regional earn the major’s frequent flyer points, and the regional uses the logos, trademarks and even the name of the major (e.g., regional airline Comair operates for Delta under the name Delta Connection).2 To facilitate passenger connections between the regional and the major, the schedules of the regional and its partner are coordinated – in fact the regional carrier’s schedule is often dictated by the major carrier. Check-in and baggage handling are also coordinated so that passengers need only check-in and check their luggage once, at the start of their trip. Majors subcontract service to regional airlines because regionals have a cost advantage on the types of routes that they serve. Table 1 compares several characteristics of routes served by majors themselves with characteristics of routes served by majors via their regional partners. We only use routes under 1,500 miles for this comparison because the type of aircraft flown by regionals have shorter ranges than the type of aircraft flown by majors. As the data in the table suggest, majors tend to use regionals to serve routes that involve at least one very small endpoint, measured either by the number of flights the major operates from the endpoint or by the population of that endpoint. In addition, majors tend to use regionals on low-density routes. Indeed, the average number of passengers flying a regional’s route in a quarter is about one-fifth of the average number flying a major’s route. Finally, even limiting the sample to routes less than

1

In 2003, 99% of regional airline passengers traveled on flights that were codeshared with a major carrier. Note that this is different from the type of codeshare arrangement typically negotiated between two major carriers, such as United Airlines and Lufthansa. Under that type of agreement, both carriers will sell tickets under their own codes for a given flight and the flight carries the identity of the operating carrier. 2

THE ROLE OF REGIONAL AIRLINES IN THE US AIRLINE INDUSTRY

195

Table 1 Characteristics of Routes (<1500 miles) Served by Majors Versus Regionals Served by Major (N = 1,017) Is either endpoint a hub for the major? # of daily flights the major operates out of the larger endpoint airport # of daily flights the major operates out of the smaller endpoint airport Population of the larger endpoint airport Population of the smaller endpoint airport # of passengers traveling route per quarter Distance of the route

Served by Regional (N = 994)

81% 388

70% 204

30.3

6.1

6,790,970 1,994,610 26,620 730

6,744,559 1,238,190 5,220 314

Source: Authors’ construction using Official Airlines Guide schedule data and Department of Transportation Databank 1A data from the second quarter of 2000. Sample includes flights between top 300 US airports operated by American, Continental, Delta, Northwest, TWA, United and US Airways or their regionals.

1,500 miles, we see that routes served by regionals are less than half the distance of routes served by majors themselves. Thus, the routes that majors subcontract to regionals are typically short, low-density routes which are most efficiently served by a small number of daily flights on a small aircraft. Regionals’ trip cost advantage in serving these types of routes results primarily from the lower salaries paid to regional airline employees, relative to the major’s own employees, and the regional employees’ more flexible work rules.3 As we explain in Section 3, regional airlines’ lower labor costs can be traced to their origins as nonunionized and non-regulated operators of small aircraft. Like majors, many regionals now have unionized workforces; nonetheless, this labor cost differential has persisted.4 Codeshared regional airline service has allowed majors to increase their flight fre­ quency on routes which they serve themselves, and to offer service – through their regional partners – on lower-density routes which the majors themselves would not be able to serve profitably. Regionals have had an important role in maintaining and expanding the major carriers’ hub networks. About 70 per cent of the routes served by regionals have the major’s hub at one endpoint. Since hub networks rely on a large num­ ber of routes being served out of the hub, as well as on high flight frequencies to allow passengers to make connections, regional airlines have been important in maintaining – or at least improving – the profitability of the major carriers’ hub networks in the face of competition from low-cost airlines.

3 Salaries are not directly comparable because major airlines fly larger aircraft than regional carriers, but hourly pilot salaries for the smallest equipment flown by major airlines are about twice as high as hourly pilot salaries for the largest equipment flown by regional carriers, controlling for the years of experience that the pilot has. 4 In addition to the advantage provided by regionals’ lower labor costs, there may be gains (in the form of lower maintenance and training costs) to having the major and the regional each specialize their fleet to include only a small number of different aircraft types.

196

SILKE JANUSZEWSKI FORBES AND MARA LEDERMAN

Table 2 Regional Airline Statistics 2004 Carriers operating US airports served US airports served exclusively by regional airlines Daily departures Passengers enplaned (millions) Passengers enplaned as % of total US traffic Revenue passenger miles (billions) Total fleet size Regional jets as % of total fleet

74 655 479 14,400 134.7 20% 56.21 2,757 59%

Source: Regional Airline Association.

In addition, regional airlines provide the sole means of scheduled air transportation at more than two-thirds of all US airports. Table 2 provides some descriptive data about the activity of US regional airlines in 2004. In that year, there were 74 regional airlines in operation. These regionals served a total of 655 US airports, 479 of which were served exclusively by regional airlines. Regionals completed over 14,000 daily departures and carried a total of 134.7 million passengers (or approximately 1 in 5 domestic passengers). Regional carriers combined to operate 2,757 aircraft, 59 per cent of which were RJs. The average seating capacity of one of these aircraft was 40. The high level of regional airline service suggested by these statistics reflects the substantial growth that the regional airline industry has recently experienced. Table 3 provides some evidence of the growth that has taken place in this industry over the past decade. The number of passengers flying on regional airlines more than doubled over this period, increasing from 57 million to almost 135 million. During the same time, the average trip length for a regional airline passenger also doubled from 210 miles to 417 miles. The combination of more passengers and increased trip length has resulted in a 367 per cent increase in revenue passenger miles over this 10-year period. Much of the increase in average trip length can be attributed to the introduction and diffusion of RJs which have a longer range than the turbo-prop planes traditionally flown by regional airlines.5 In addition to having longer ranges, RJs also tend to be larger and faster than turbo-prop planes. As the fourth and fifth rows of Table 3 show, this has resulted in an increase in the average seating capacity of regional aircraft from 23.7 seats in 1994 to 39.9 seats in 2004 and an increase in available seat miles on regionals from 23.7 billion to 82.6 billion, an increase of almost 250 per cent. Interestingly, as the final row of the table indicates, the number of departures by regionals grew only modestly over the decade, from 4.6 million to 5.25 million. This suggests that much of the expansion in service over this period seems to come from changes in the type (and therefore size and distance) of aircraft flown, from turbo-props to RJs. Since 2001, the reduction in passenger demand and the cost pressures facing all major carriers have further contributed to the growth in regional airline service. An analysis 5

We discuss the role of regional jets in greater detail in Section 5.

Table 3 Growth in Regional Airline Service Year Enplaned passengers (millions) Average passenger trip length (miles) Revenue passenger miles (billions) Average seating capacity (seats per aircraft) Available seat miles (billions) Departures completed (millions)

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

Growth, 1994–2004 (%)

571

572

619

663

711

781

846

828

984

1130

1347

1359

384

417

210

223

230

231

245

267

299

311

333

986

120

128

142

153

174

208

253

257

328

433

562

3676

237

246

251

259

277

298

317

335

351

377

399

684

237

255

269

278

304

358

426

442

526

662

826

2481

463

Source: Regional Airline Association.

469

446

438

433

438

446

420

441

488

525

134

198

SILKE JANUSZEWSKI FORBES AND MARA LEDERMAN

by the Regional Air Service Initiative6 shows that between January 2001 and January 2002, regionals replaced service by majors on 108 routes. In addition, 90 routes that were served exclusively by majors in January 2001 were served by a combination of majors’ and regionals’ flights in January 2002. Since then, the regionals’ expansion has continued, while majors are only slowly recovering from their reduction in service after 9/11.7

3 HISTORY The role of regional airlines as providers of air service to small communities can be traced back to the regulated era.89 In 1926, the Air Commerce Act charged the Secretary of Commerce with the task of promoting air commerce and empowered him to issue and enforce air traffic rules, license pilots, certify the airworthiness of aircraft, establish airways, and operate various aids to air navigation. The Civil Aeronautics Act of 1938 added economic authority to this operational and safety authority, thus establishing the Federal Government as the economic regulator of the air transportation industry. The Act created the Civil Aeronautics Board (CAB) and required every air carrier to obtain a certificate from the CAB that authorized it to serve a particular point or route. The 16 carriers in operation when the CAB was formed were given certificates to continue the service they were already providing. These 16 carriers became known as the “trunk carriers”. In keeping with its objectives to grow commercial air service and at the same time protect the economic stability of the trunk carriers, the CAB often granted trunk carriers exclusive access to newly authorized routes and, at least initially, refused to issue operating certificates to any new airlines. Much of this changed, however, following World War II. The war affected the early commercial aviation industry in two important ways. First, the war accelerated the advancement of aviation technology, expanding both the number and size of available aircraft. Second, the war accelerated the expansion of air service by increasing the demand for air service from individuals living in smaller communities. In response to this increase in demand for service, in 1944, the CAB created a new category of experimental “feeder airlines”. The CAB recognized that air service to small communities would likely require subsidization since many small communities could not generate sufficient traffic to cover costs. Reluctant to jeopardize the trunk carriers’ evolution towards financial self-sufficiency, the CAB chose to create a new category of airlines to serve small communities, rather than allocate these additional routes to the trunk carriers. Between 1944 and 1950, the CAB awarded temporary operating certificates to 17 new or existing interstate carriers. In 1955, these temporary certificates were made permanent and these carriers became the “local service” airlines. The local service airlines were 6

www.rasi.org

Another factor that has contributed importantly to the recent growth of regional airlines is the renegotiation

of scope clauses, which are now much less restrictive than they were until 2001. We address scope clauses in

Section 5.

8 See Borenstein and Rose (2005) for a thorough account of regulatory reform in the US airline industry.

9 This section draws heavily on Office of Technology Assessment (1982) and Levine (1987).

7

THE ROLE OF REGIONAL AIRLINES IN THE US AIRLINE INDUSTRY

199

given authority to operate only on low-density routes serving smaller communities or on heavier routes on which they were required to make intermediate stops at smaller cities. These requirements were explicitly for the purpose of keeping the local service airlines from competing directly with the trunk carriers. Coinciding with the emergence of the local service airlines, a third category of commercial air service was beginning to take shape. This category included fixed-base on-demand air taxi service. In 1949, the CAB recognized this third category of airlines and created another experimental class of airlines for “non-certified irregular route” carriers. This category was confirmed in 1952. Unlike the trunk and local service airlines, this new class of carriers – known as scheduled air taxis and eventually as “commuter airlines” – did not require operating certificates from CAB. Regulations did, however, prevent them from operating aircraft of more than 12,500 lb takeoff gross weight and from offering scheduled service between certified points. The weight limitation and the exclusion from certified points were specifically imposed to prevent the commuters from competing directly with the trunk and, especially, with the local service carriers. During the early 1960s, the subsidies required to support the local service airlines increased dramatically. While most of the trunk carriers were, at this time, financially self-sufficient (due largely to them being allowing to terminate service to 211 small communities in favor of local service carriers), the total subsidy required by the local service airlines was almost $67 million in 1962. Anxious to reduce this high level of subsidization, the CAB began to allow the local service carriers to modify their route structures. Specifically, the CAB allowed them to replace the trunk carriers at some points, relaxed the requirement that the locals stop at every intermediate certified point on every flight and allowed locals to drop service to places that generated less than five passengers per day on average. These route modifications did improve the financial performance of the locals (total subsidy payments fell to $34 million by 1970); however, they also resulted in the elimination of 108 small communities from the local service route map. Service to small communities was further reduced as both the trunk and local service airlines transitioned to faster, larger aircraft. Between 1968 and 1978, an additional 125 cities were removed from the local service route map. Once again, there was a major gap in air service to small communities. This role was eventually filled by the commuter airlines. Regulatory and economic changes allowed for significant growth of the commuter airlines in the 1960s and 1970s. In 1965, the CAB amended its regulations to allow commuter airlines to carry mail and to provide service between certified points, often replacing trunk or local service airlines. In 1964, American Airlines contracted for Apache Airlines to replace it in serving Douglas, Arizona – this was the first “air taxi replacement agreement”. In 1967, Allegheny Airlines (which eventually became US Airways) established the Allegheny Commuter program, contracting unprofitable destinations to 12 independent commuter airlines operating under the name Allegheny Commuter. This arrangement marked the first codeshare relationship between a major carrier and a commuter. These types of codesharing agreements increased throughout the 1970s, contributing to commuter airlines’ growth over this period. By 1978, 26 commuter airlines were providing replacement service for trunk and local service airlines at 50 points, mostly without government subsidization.

200

SILKE JANUSZEWSKI FORBES AND MARA LEDERMAN

Deregulation in 1978 more firmly established the role of commuter airlines in the commercial aviation industry. The Deregulation Act of 1978 provided a means for commuter airlines to replace trunk and local service carriers at certain points of their networks by mandating the pro-ration of fares for connecting service. In addition, the Act allowed commuters to offer scheduled service of up to 30 seats, with this limit sub­ sequently being raised to 50 seats. After deregulation, commuters (like larger carriers) were required to obtain operating certifications [now from the Federal Aviation Adminis­ tration (FAA)] and were required to comply with more stringent operating and reporting restrictions. Finally, deregulation established the Essential Air Service program which provided subsidies to airlines providing “essential air service” to small communities. The emergence of hub-and-spoke systems shortly after deregulation further increased the extent of codesharing between large carriers and commuter airlines. Large airlines operating hub-and-spoke systems realized that passengers traveling to or from small communities on commuter airlines were usually flying those short-haul trips as part of a longer itinerary (which typically involved travel on a major carrier). As such, these passengers could provide an important source of feeder traffic for airlines at their hubs. The airlines further realized that, since there was typically little competition for shorthaul portion of the flight and since passengers preferred easy connections to the long-haul portion, coordinating their flights with the commuter airline’s flights could help them capture these passengers on the longer leg of their trip. This left the large airlines with two options: serve the short-haul routes themselves or establish arrangements with the existing commuter carriers. Even with the improved efficiency brought by deregulation, the larger airlines’ costs of serving these small towns were still well above those of the specialized commuter airlines. Partnerships between majors and commuters (eventually renamed regionals) flourished.

4 RELATIONSHIPS BETWEEN MAJORS AND REGIONALS The codeshare partnerships between major carriers and their regionals are governed by one of two types of organizational forms. A regional may be wholly owned by the major with which it partners. Or, a regional may be independently owned and contract with one or more major carriers.10 In the case of an independent regional, the contract between the major and the regional will generally take one of two forms. Historically, most contracts have been revenuesharing agreements (also known in the industry as pro-rate agreements). Under these agreements, the regional agrees to serve a set of routes on behalf of the major and to coordinate its schedule on (and allocation of aircraft to) those routes with the major’s own schedule. In exchange, the major permits the regional to use its service marks and logos and lists the regional’s flights in computer reservation systems under its two-letter designator code. The regional receives an allocated portion of the revenue from each passenger that flies the regional as part of an itinerary that connects with one of the

10

In some cases, a major will take a minority equity position in its regional partner. Unfortunately, we do not have systematic data identifying these cases.

THE ROLE OF REGIONAL AIRLINES IN THE US AIRLINE INDUSTRY

201

major’s flights. Fares are set by the major and marketing and ticketing are carried out by the major. More recently, the industry has shifted towards fixed-fee or capacity-purchase agree­ ments. Under these types of contracts, the regional receives a fixed payment (usually based on block hours flown) for each departure that it operates on behalf of the major. This fixed payment is calculated to cover the regional’s operating costs and to guarantee a reasonable rate of profit. In addition, the regional may receive incentive payments based on operational performance, such as on-time performance and baggage handling. Under a capacity purchase-agreement, the major retains all revenue from flights oper­ ated by its regional. Our conversations with industry participants and examination of the trade presses suggest that the switch to fixed-fee contracts was motivated by two factors. First, these contracts eliminate almost all of the risk faced by the regional. The fixed fee payment with a guaranteed profit margin insulates the regional from both demand risk (since its revenue is independent of the number of passengers onboard) and cost risk (since most costs, including fuel, are passed on to the major).11 Second, fixed fee contracts provide the major with a greater level of control over the regional, in particular over its schedule. The switch to fixed fee contracts began in the late 1990s and, interestingly, largely coincides with regionals’ adoption of RJs. Table 4 lists the major–regional partnerships that were in place in 2000 and 2005 for the largest US network carriers. Regional carriers that appear in bold are ones that are fully owned by their major partner. As the table clearly indicates, there is heterogeneity both across and within majors in the extent to which regional partners are owned. As well, there have clearly been changes over time in a given major’s use of owned and independent regionals. In 2000, Continental, Delta, Northwest and US Airways each used a mixture of owned and independent regionals, while American used only owned regionals and United used only independent regionals. By 2005, Continental and Northwest had each sold the regional partners that they owned, leaving them (as well as United) using independent regionals exclusively. Delta and US Airways continued to use a mixture of owned and independent regionals, while American supplemented the service provided by its owned regionals with contract service provided by several independent regionals. Overall, while we continue to see variation both across and within majors in their use of owned and independent regionals, the data does suggest a trend over time towards greater use of independent regionals.12 While both owned and independent regionals operate as subcontractors for majors, our research suggests that there are a number of important differences between the two types of regionals. First, independent regionals usually own or lease their own aircraft and hire, fire and manage their own employees. Wholly owned regionals – though operating as a separate entity within the major – ultimately have their aircraft and employees included as part of the major’s own fleet and workforce. As such,

11

However, even these contracts do not protect the regionals from financial risks when the major carrier is in Chapter 11 bankruptcy. Under those conditions, the major can reject the prior contract and renegotiate a new contract with substantially lower fixed fee payments to the regional. 12 This may partly reflect the current economic climate in the industry which has forced many majors to sell whatever assets they could, including their owned regional partners, but it is likely also due to the lower costs of many independent regionals.

SILKE JANUSZEWSKI FORBES AND MARA LEDERMAN

202

Table 4 Majors and Regional Partners in 2000 and 2005 Major

Regional Partners – 2000

Regional Partners – 2005

American Airlines

American Eagle Airlines Business Express

American Eagle Airlines Chautauqua Airlines Executive Airlines Regions Air Trans States Airlines

Continental Airlines

Continental Express Gulfstream International Airlines

Cape Air Colgan Airways Commutair ExpressJet Gulfstream Int’l Airlines

Delta Air Lines

Atlantic Coast Airlines/ACJet Atlantic Southeast Airlines Comair SkyWest Airlines Trans States Airlines

American Eagle Airlines Atlantic Southeast Airlines Chautauqua Airlines Comair Mesa Airlines SkyWest Airlines

Northwest Airlines

Express Airlines, I Mesaba Aviation

American Eagle Airlines Big Sky Airlines Pinnacle Airlines Gulfstream Int’l Airlines Mesaba Airlines

United Airlines

Air Wisconsin Atlantic Coast Airlines Great Lakes Aviation Gulfstream International Airlines SkyWest Airlines

Air Wisconsin Chautauqua Airlines Great Lakes Aviation Gulfstream Int’l Airlines Mesa Airlines Shuttle America SkyWest Airlines Trans States Airlines

USAirways

Mesa Air Group/Air Midwest Allegheny Airlines Mesa Air Group/CCAir Chautauqua Airlines Colgan Airways Commutair Mesa Air Group/Mesa Airlines Piedmont Airlines PSA Airlines

Air Midwest Chautauqua Airlines Caribbean Sun Airlines Colgan Airways Mesa Airlines MidAtlantic Airways PSA Airlines Piedmont Airlines Trans States Airlines

Regionals in bold are fully owned. Source: Regional Airline Association.

THE ROLE OF REGIONAL AIRLINES IN THE US AIRLINE INDUSTRY

203

independent regionals retain residual rights of control over their aircraft and workforce while, for wholly owned regionals, these rights ultimately rest with the major. In addition, ownership of a regional carrier allows the major airline to select and replace the regional’s management, while contracting with an independent regional does not. The implication of this is that the managers of a wholly owned regional are ultimately accountable to the major. Second, ownership affects the way in which majors and regionals respond to unantic­ ipated schedule disruptions. These disruptions occur most frequently in adverse weather conditions, but may also result from air-traffic control problems or airline mechanical problems. When an airport experiences adverse weather, the FAA will determine – several hours in advance – the number of flights which will be allowed to land dur­ ing each hour. When weather necessitates a reduction in flights relative to the original schedule, each airline will receive a number of take-off and landing slots in proportion to its original share of scheduled flights. The airline then decides which of its flights to delay or cancel. When a regional is owned by a major, the major and the regional receive a common allocation of slots and the major carrier decides which of its own and which of the regional’s flights to delay or cancel. In fact, rescheduling decisions for wholly owned regionals are done by the major carrier’s Airline Operational Control Center (AOCC). In contrast, when a regional is independent, it receives its own slot allocation and makes its own decision (in its own AOCC) about delays and cancellations, possibly in coordination with the major carrier for which it operates. Third, there may be operating cost differences between owned and independent region­ als.13 The lower salaries paid to regional airline employees have led these employees to seek compensation that is closer to that earned by their counterparts at the mainline. Regional employees’ demands for higher wages may be harder for management to resist when a regional is wholly owned by a major airline.14 In addition to the difference in wages, phone conversations with industry executives and analysts have suggested that owning a regional may lead to costs associated with managing two distinct labor forces, such as more frequent labor disputes. The fact that there are differences between owned and independent regionals – combined with the fact that we observe the same major using different types of regionals on different routes – suggests that majors appear to be making explicit decisions about what type of regional to use and when. Indeed, this decision between using an owned or independent regional has the flavor of a classic “make-versus-buy” decision where a firm (in this case, an airline) must decide whether to carry out a given transaction in-house (“make”) or through the market (“buy”).15 In Forbes and Lederman (2006), we explore this question of why majors may choose to vertically integrate with some or all of their regional partners. Building on the growing body of empirical work in the organizational economics literature, we analyze a major’s decision whether to use an owned or independent regional on a particular route. We develop a simple framework

13

Both types still have a substantial cost advantage over majors.

For example, after Delta acquired the previously independent regional carrier Comair, pilots at Comair

demanded higher salaries and went on strike. The pilots used the now common ownership of Delta and Comair

as the main argument for a salary increase.

15 Williamson (1971) and (1985).

14

204

SILKE JANUSZEWSKI FORBES AND MARA LEDERMAN

that illustrates the costs and benefits of using an owned – versus independent – regional and we test that framework using data on flights served by regional carriers operating for the seven largest US network carriers on city pairs between the 300 largest US airports in the spring of 2000. Our framework illustrates how the benefits and costs of owning a regional result from operational and institutional characteristics of the airline industry. Specifically, we argue that majors interact with their regionals during two types of operational decisions – ex ante scheduling decisions and real-time adjustments to schedule disruptions. Contracts between majors and regionals generally cover the first set of decisions; however, they do not – and likely could not – cover the second set of decisions. Thus, there is an incentive problem between majors and their regionals which results from the incompleteness of contracts with respect to real-time schedule adjustments. When unanticipated disruptions (such as adverse weather or mechanical problems) create the need for adjustments to the major’s planned flight schedule, the major and its regional may disagree on what adjustments should be made. In particular, while the major will attempt to internalize the impact of the disruption on its entire network, the regional, who is compensated only based on the routes it serves for the major, will not. Ownership of a regional mitigates this incentive problem by giving the major residual rights of control over how the regional’s physical assets and labor force are used. This, we argue, is the primary benefit of owning a regional. However, as described above, there are also costs associated with ownership of a regional. Majors subcontract service to regional airlines because regionals have a cost advantage that results primarily from the lower salaries paid to regional airline employees, relative to the major’s own employees. Ownership of a regional has the potential to erode this labor cost savings that regionals afford majors. The lower salaries paid to regional airline employees have led these employees to seek compensation that is closer to that earned by their counterparts at the mainline. Thus, the framework developed in Forbes and Lederman (2006) predicts that a major’s optimal choice of organizational form will reflect the tradeoff between its incentive to exercise control over its regional and its incentive to maximize the labor cost savings that its regional provides. To test this framework, we develop two propositions that relate an airline’s likelihood of using an owned regional on a city pair to airline-specific characteristics of that city pair which proxy for the magnitude of the incentive problem. Our first proposition relates to the extent to which a regional’s flight is integrated into the major’s network. The more integrated a regional’s flight, the more likely it is to experience disruptions and the more costly it will be for the major to have these disruptions resolved by its regional who will not internalize the impact of its decision on the major’s network. Our second proposition relates to the frequency of unforeseen schedule disruptions that result from adverse weather. Adverse weather increases the amount of time that is needed in between consecutive takeoffs or landings, thus forcing airlines to delay or cancel flights. As a result, adverse weather forces majors and regionals to make more frequent adaptation decisions. We therefore expect that wholly owned regionals are more likely to be used on city pairs that are more integrated into the major’s network and on city pairs that are more likely to be affected by adverse weather. We test these two propositions and find strong empirical support for both.

THE ROLE OF REGIONAL AIRLINES IN THE US AIRLINE INDUSTRY

205

5 THE INTRODUCTION OF THE REGIONAL JET As alluded to in Section 2, one of the most important recent developments in the regional airline industry has been the introduction of the RJ. The RJ is a small jet-powered aircraft that holds between 30 and – in the most recent models – 100 passengers. RJs were first introduced in Europe at the end of 1992 by Lufthansa Cityline and in the US in early 1993 by Comair, a regional partner of Delta Air Lines. The introduction turned out to be a commercial success for both airlines, and these and other airlines followed in adopting the RJ for many of their routes over the following years. Figure 1 provides some evidence of the extent of RJ adoption between 1996 and 2000. The figure plots the average daily number of RJ flights operated by the largest six US network carriers (combined). As the figure suggests, the number of RJ flights increased from slightly less than 500 at the beginning of 1996 to over 2,800 by the end of 2000. Moreover, the figure shows that, over the same period, there is no overall increase in the number of daily jet flights operated by these carriers and there is a decrease in the number of turboprop flights. This suggests that the dramatic increase in RJ flights is not simply reflecting an overall trend in increased service and, furthermore, that RJ flights are, at least partially, replacing turboprop flights.16 The technological appeal of the RJ results from the fact that it combines the “best” features of a turboprop with the “best” features of a jet. In particular, RJs have the capacity of a turboprop but the range, speed and comfort of a jet. To see this in the context of an actual airline’s fleet, in Table 5, we compare the capacity, range and cruising speed of two full-size jets, two RJs, and two turboprop planes which were in use

Average # Daily Departures

25000

20000

15000

10000

5000

0 1996

1997

1998

1999

2000

Year TP

RJ

JET

Source: Authors’ calculation using Official Airlines Guide data, 1996 – 2000.

Figure 1 Aggregate Patterns of RJ Usage, 1996–2000 AA, CO, DL, NW, UA and US.

16

We discuss the various uses of RJs over this period in greater detail below.

SILKE JANUSZEWSKI FORBES AND MARA LEDERMAN

206

Table 5 Comparison of RJs to Turboprops and Full-Size Jets, Delta’s Current Fleet Jets

RJs

Turboprops

B-737-800

MD-88

CRJ-700

CRJ-200 Aerospatiale 72–210

EMB120

150 (2 classes)

142 (2 classes)

70 (1 class)

40–50 (1 class)

66 (1 class)

30 (1 class)

Range (miles)

2789

1740

1939

1265–1496

1318

945

Cruising speed (mph)

531

509

544

530

322

292

Seats

Source: http://www.delta.com.

in Delta’s mainline fleet as of October 2006. As the table indicates, while the capacities of the RJs in Delta’s mainline fleet are similar to or slightly larger than the capacities of the turboprops in its fleet, the range and, in particular, the cruising speed of the RJs are similar to that of the 100 and 100-plus seat jets in Delta’s mainline fleet. From an economic perspective, RJs do not necessarily have lower costs per available seat miles than larger jets and in most cases may actually have higher costs per available seat mile. Rather, the benefit of the RJ is that, on thinly traveled routes, the revenue generated from the small number of passengers flying the route could actually cover the costs of a RJ but would be unlikely to cover the costs of a full-sized plane. Put another way, while the costs per available seat mile may be higher on an RJ, the revenue per available seat mile is like to be substantially higher on an RJ relative to a full-sized jet since the number of available seat miles is so much lower. This means that airlines may be able to profitably serve thinly traveled routes with RJs but not with full-sized jets.17 Given these technological and economic advantages of the RJ, it is clear that airlines can benefit from the adoption of RJs in several different ways. In particular, our research suggests that there are four distinct motives that airlines can have for introducing RJs onto specific routes. First, airlines may use RJs to replace larger jets in an effort to reduce over-capacity and improve load factors. Second, airlines can use RJs to supplement existing jet (or turbo-prop) service in order to increase flight frequency during off-peak times of the day and adjust the airline’s capacity on the route. Third, RJ’s can be used to replace existing turbo-prop service if airlines want to offer higher quality jet service and, potentially, a slightly larger seating capacity. Finally, RJ’s can be used to introduce new service on routes that would not be profitable for airlines to serve with the other types of aircraft available to them. We suspect that the impact of RJ introduction on consumer welfare will depend on which of these motives drive the RJ adoption decision. For example, RJs used for new or supplemental service would likely lead to higher consumer surplus, as may the replacement of existing turbo-prop service with higherquality aircraft. On the other hand, RJs used to reduce existing capacity on a route 17 Most travelers prefer RJs over turbo-props because of their greater travel comfort. With low fuel prices, RJs were also more cost-efficient than turbo-props on many routes, but more recently, with rising fuel prices, some carriers are considering to increase their use of the more fuel-efficient turbo-props again.

THE ROLE OF REGIONAL AIRLINES IN THE US AIRLINE INDUSTRY

207

previously served by a larger jet aircraft could lead to lower consumer surplus. Of course, consumer surplus would also depend on how, in equilibrium, other airlines respond to a carrier’s RJ introduction (in terms of changes to their own schedule, capacity and aircraft mix). Using detailed flight schedule data from the Official Airlines Guide from 1996 to 2000, we analyze airlines’ motives for RJ adoption between 1996 and 2000. During this five year period, we observe 537 introductions of RJs by the six largest US carriers on city pairs on which they did not previously use RJs. To operationalize the motives described above, we classify RJ introductions into four categories: (1) Capacity Reduction – which we define as capacity on the route falling with frequency remaining the same or falling or capacity on the route falling by more than 20 per cent with increase in frequency; (2) Frequency Supplement – which we define as frequency increasing with capacity falling less than 20 per cent; (3) Turboprop Replacement – which we define capacity increasing and the airline’s number of turboprop flights on the city pair decreasing; and (4) New Service – which we define as RJ service on a route on which the airline did not previously offer service. We find that, between 1996 and 2000, 23 per cent of introductions were capacity reduction, 23 per cent were frequency supplement, 22 per cent were turboprop replacement, and 32 per cent were new service.18 We also find that the motive for RJ introduction varied with route characteristics, in a fairly intuitive way. For example, on very small spoke routes, over 40 per cent of RJ introductions replaced turboprop service. In contrast, on large spoke routes, almost half of RJ introductions were for the purpose of supplementing frequency on the route. Finally, on point-to-point routes, 56 per cent of introductions were new service. While the continued production and adoption of larger and more sophisticated RJs suggests that the benefits to airlines of RJ adoption continue to be substantial, it is important to note that the diffusion of the RJ has not come without costs. In particular, the emergence of the RJ has exacerbated existing tensions between management and pilot unions at the major carriers. Because of their longer range and greater appeal to business travelers, RJs are much more substitutable with mainline jets than are turboprops. This raises the concern among mainline pilots that management will be tempted to “outsource” an increasingly large number of mainline routes to regional partners to serve with RJs. In order to limit the extent to which management can replace mainline service with lowercost regional service, mainline pilot unions have negotiated so-called “scope clauses” into their labor agreements. These clauses restrict the number of small planes (RJs or turbo-props) that may be operated for the major airline by regional partners, either placing on absolute limit on the number of small planes and/or tying the number of allowed small planes to increases in mainline jet flying. In the late 1990s, scope clauses became a major source of conflict between management and pilots at all major carriers. During that time, scope clauses substantially limited the growth of regional carriers. After 9/11, however, with most major carriers in or near bankruptcy, the scope clauses have been renegotiated and relaxed. While scope clauses still impose some constraints on majors today, they do so to a much lesser degree than before 9/11.

18

13 of the 537 RJ introductions do not fall into any of these four categories.

208

SILKE JANUSZEWSKI FORBES AND MARA LEDERMAN

6 LOOKING FORWARD Looking forward, we expect that regionals will continue to play a large and important role in the commercial aviation industry. However, we also suspect that this segment of the industry will face a number of challenges. First, as RJs increase in size, the distinction between the types of aircraft flown by majors and the types flown by their regional partner will further blur. Both Embraer and Bombardier, the two main producers of RJs, have already introduced 70-seat RJs, which are in use in several airlines’ fleets. As well, both have developed RJs in the 100-seat range and have received orders for these from several airlines. It will be interesting to observe how majors and their pilots address the issue of whether 100-seat RJs should be flown by mainline pilots or regional pilots. Indeed, the close substitutability between the 100-seat RJ and a traditional wide body aircraft is exemplified by the fact that, to date, the largest customer of the 100-seat Embraer jet is not even a regional (or a major on behalf of its regional), but rather JetBlue Airlines, a low-cost carrier. Second, as majors continue to evolve in the face of low-cost competition, they are increasingly using regionals to serve routes on which they compete with low-cost carriers. This is a very different role for regionals than their traditional role of providing feeder service to and from a major’s hub airports. Again, it will be interesting to observe how relationships between majors and regionals change as regionals increasingly take on this role. Interestingly, competition from low-cost carriers may ultimately lead to a reduction in the cost advantages that regionals provide. As majors continue to feel pressure to lower their own costs, they will continue to renegotiate labor contracts with their unions. As they do so, the labor cost advantage that regionals possess will shrink, challenging the very motivation for their use by major carriers.

REFERENCES Borenstein, S. and N.L. Rose (2005), “Regulatory Reform in the Airline Industry”, mimeo, Massachusetts Institute of Technology. Forbes, Silke J. and Mara Lederman (2006), “Control Rights, Network Structure and Vertical Integration: Evidence from Regional Airlines”, mimeo, University of California, San Diego. Levine, M.E. (1987), “Airline Competition in Deregulated Markets: Theory, Firm Strategy and Public Policy”, Yale Journal on Regulation, 4, 393–494. Office of Technology Assessment, US Congress (1982), “Air Service to Small Communities”, (at http://www.wws.princeton.edu/ota/disk3/1982/8201/8201.PDF) Williamson, Oliver (1971), “The Vertical Integration of Production: Market Failure Considera­ tions”, American Economic Review, 63, 112–123. Williamson, Oliver (1985), The Economic Institutions of Capitalism, New York, NY: Free Press.

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

9 Airport Substitution by Travelers: Why Do We Have to Drive to Fly?∗ Gary M. Fournier† , Monica E. Hartmann,‡ and Thomas W. Zuehlke§

ABSTRACT This article explores aspects of the determination of airline fares in selected mediumsized US airports subject to competition from alternative airports within driving distance. Passengers in these markets often face substantial discounts at distant airports, in exchange for the time costs of driving there. Spatial linkages in airport competition are not well studied. A panel of 16 quarters is constructed in order to investigate models of spatial error correlation and spatial autoregression in overall fare levels in adjacent airports. We find that fare differentials between local and nearby alternative airports can lead to lower load factors and other indicators of poor performance in smaller local airports. Fare differentials at nearby airports often provide substantial incentives to travelers and are an important determinant of poor performance at medium-sized airports.

1 INTRODUCTION A common allegation made in many airline markets involving small-to-medium airports is the lack of effective competition and apparent failure of the process of competitive ∗

Authors thank Farasat Bokhari, John Kwoka, Ralph Sandler, Jiyoung Kwon, Donald Lacombe, and Nick Rupp for their comments, as well as the participants in worshops as FSU. An earlier darft was presented at the International Industrial Organization Conference, April 2005. † Corresponding author. Department of Economics, Florida State University, Tallahassee, FL 32306-2180,

USA. Tel.: +1-850-644-5001, Fax: +1-850-644-4535.

‡ Economics Department, St Thomas University, 2115 Summit Avenue, St Paul, Minnesota, 55105, USA.

Tel.: +1-651-962-5681.

§ Department of Economics, Florida State University, Tallahassee, FL 32306-2180, USA. Tel.: +1-850-644­

7206.

Email addresses: [email protected] (Gary M. Fournier), [email protected] (Monica

E. Hartmann), [email protected] (Thomas W. Zuehlke).

GARY M. FOURNIER et al.

210

entry to mitigate persistent high fares out of some local airports. A 1999 study by the Transportation Research Board suggests that fares are substantially higher in shortdistance markets without low-fare competition, often at small-to-medium sized airports. Indeed, the problem can also occur in reverse; travelers in large cities such as Boston may sometimes obtain better options than at Logan Airport by flying out of Providence, RI. This chapter attempts to develop new evidence of the extent of this problem for a set of 65 such airports, by assessing how fare differentials across alternative airports affect performance. Table 1 lists some examples of adjacent airport pairs in which the alternative airport choice is one of the largest 20 airports in the US. To get a more representative picture of the mean airfares, the price index for a fixed bundle of tickets originating out of the smaller airport to its 50 most-frequently chosen destinations is calculated. The mean fare difference is then calculated, for the second quarter of 1998, as the difference between the (passenger-weighted index) mean airfare to these top destinations from each origin airport and the index of mean airfare from the alternative airport to the same destinations. As Table 1 shows, Providence, RI is an attrac­ tive alternative to Boston, saving about $78 on the typical itinerary, before including the economic costs of driving. In many instances, however, the biggest gains are found for travelers originating near small-to-medium airports, such as Melbourne, Fl, where driving to Orlando yielded a gross gain of about $270. Passengers in smaller markets, in partic­ ular, often face substantial discounts at larger distant airports, in exchange for their time

Table 1 Fare Differences at Alternative Airports for a Fixed Set of Destinations From the Origin Airports Origin Airport Providence, RI Atlantic City Colorado Springs Newburgh, NY Islip long island Manchester, NH Allentown, PA Westchester-White Plains Burmingham, Al Chattanouga Albany, NY Melbourne, Fl Jacksonville Greenville-Spartanburg Columbia, SC

Alternative Airport PVD ACY COS SWF ISP MHT ABE HPN BHM CHA ALB MLB JAX GSP CAE

Boston New York–JFK Denver New York–JFK New York–JFK Boston Newark New York–JFK Atlanta Atlanta New York–JFK Orlando Orlando Atlanta Charlotte

BOS JFK DEN JFK JFK BOS EWR JFK ATL ATL JFK MCO MCO ATL CLT

Mean Fare Difference ($)

Travel Miles

−77�791 −64�799 −51�392 −22�841 −20�057 −14�487 8�23 83�272 174�441 204�303 206�591 270�355 277�934 361�518 368�972

60 132 91 85 44 55 75 37 149 125 174 62 169 167 101

Note: The mean fare difference is calculated, for the second quarter of 1998, as the difference between the (passenger-weighted index) mean airfare to the top 50 destinations from each origin airport and, for the same destinations, the index of mean airfare from the alternative airport.

AIRPORT SUBSTITUTION BY TRAVELERS

211

costs of driving there. In this study, we consider localities where the larger, nearby airport provides fare advantages as well as localities where the small nearby airport is cheaper. Carriers originating from adjacent airports directly compete on common destinations when travelers find it attractive to drive to the alternative airport in return for lower fares. These arbitrage actions of travelers selecting among adjacent airports shape airline competition. Carriers set ticket prices optimally recognizing not only local travelers’ demands but also the demand induced from travelers who are willing to drive some distance to purchase a cheaper flight. Spatial linkages in competition at small-to-medium sized airports are not well studied. For this study, a panel of 16 quarters is constructed in order to investigate, for the first time, temporal attraction of spatial fare differentials on performance in smaller airports and the subsequent adjustment in fare differentials following entry. Carriers may find that their costs are raised when travelers are drawn away from the local airport, reducing passenger volume and load factors. We find that fare differentials at alternative airports often reflect substantial incentives to travelers, but the siphoning of price-sensitive passengers through airport substitution leads to lower performance at the local airport in terms of load factors, passenger volume, and revenue. The layout of this chapter is as follows. Section 2 summarizes the related literature. Section 3 describes the construction of a Laspeyres index across space, a weightedaverage cost of a representative bundle of flights between the focal airport and the lowest-fare alternative. These weighted fare indexes proxy how substitutable the two airports are and influence market competition and performance. Section 4 describes the data used, the assumptions made, and the covariates controlled for in our empirical analysis. Section 5 extends the empirical model to allow for spatial autocorrelation in the dependent variable and the error term when estimating the fare differential indexes. The empirical model of how fare differentials affect the performance and entry decisions of airlines operating at small-to-medium airports are described in Section 6. The empirical results for each of the models are reported in Section 7. Our conclusions follow in Section 8.

2 LITERATURE REVIEW In order to analyze how fare differentials across alternative airports affect economic performance, one must understand two issues: (1) the source of the fare differentials and (2) the factors that drive carrier entry decisions and thus their expectations about the potential profitability. First, if alternative airports were perfect substitutes, then substitution by travelers would tend to equalize fares across the alternatives. Second, one can view economic performance issues from the vantage point of a potential entrant to understand the viability of service offerings in the context of existing competition among carriers. There is an extensive literature that examines each of these issues. Starting with the fare differential literature, Borenstein (1989) established that an airline’s airport and route dominance determine the degree in which a carrier can mark-up over cost. Furthermore, their ability to mark-up does not extend to nondominate carriers on that route. Borenstein and Rose (1994) observe fare differentials increasing on routes with more competition or lower flight density. Evans and Kessides (1993) attribute this

212

GARY M. FOURNIER et al.

variation in prices to multimarket contact, that is amount of overlapping network of flights by the carriers servicing that route. These findings suggest that the distribution of airfares available to entrants will differ across alternative airports. The literature on carrier entry provides insights as well on how entrants decide on which routes to offer service. The first strand of this literature identifies factors that determine entry. Berry (1992), Brueckner et al. (1992), and Reiss and Spiller (1989) among others find that entry on a given route increases with a carrier’s airport presence at the endpoints. Morrison and Winston (1990) determine that trunk carriers are less likely to enter routes with relatively high fares. These markets tend to be characterized by relatively high barriers to entry, relatively high costs of service, or aggressive response by incumbent firms to entry of a new carrier. Several articles focus solely on low cost carrier’s (LCC) entry decisions. Ito and Lee (2003a) ascertain that premarket density – the average number of daily passengers transported on that route – is the most important market characteristic to induce nonstop entry by LCCs. Boguslaski et al. (2004) find evidence that Southwest’s entry decisions have evolved over the years. Initially, Southwest entered medium-haul markets to service people who did not previously travel. By the second half of the 1990s, Southwest began to offer longer-haul service on routes previously avoided by major carriers. The second strand of the entry literature examines incumbents’ responses to these entry decisions. Dresner et al. (1996) and Morrison (2001) show that LCC’s influence on airfares that extend beyond a particular airport route they enter. Ito and Lee (2003b) find that incumbent hub-spoke carriers are fairly accommodating when LCCs begin to service their hubs. They rarely undercut entrants’ average fares and usually only match their prices. Goolsbee and Syverson (2005) examine incumbent pricing behavior on routes where Southwest does not operate, but offer service at both endpoints. As expected, incumbents drop fares in anticipation of entry; the decline is the greatest for the more concentrated routes and higher priced business fares. Incumbents do not, however, cut fares on alternative routes, only on threatened routes where Southwest Airlines already operates at both endpoints. Finally, Windle and Dresner (1999), Whinston and Collins (1992), and Bamberger and Carlton (2006) are other studies that document incumbent fare responses to LCC entry. While market power problems at large hub airports with a dominant carrier have been well studied by these articles, less attention has been shown to the price determination at smaller sized airports. An early study by the US GAO (1996) documents regional differences in airfares at small-to-medium sized airports. Following deregulation, airfares in western states were generally found to be declining, while airfares in eastern states were generally increasing. The GAO attributes these differences to a greater frequency of entry by low-cost airlines in the western states. While lacking a formal model, the GAO suggests that low-cost carriers had avoided the eastern states during this earlier period because of “slower growth, airport congestion, and harsher weather” and because “one or two relatively high cost carriers have dominated the routes.” Much has changed since 2000, however, because the east has become the area where most of the new LCC capacity is being deployed.1

1

See, for example, http://www.darinlee.net/data/lccnewcapacity.html.

AIRPORT SUBSTITUTION BY TRAVELERS

213

Passengers’ willingness to trade-off price and travel time is reported in Hartmann’s (2000) study of consumer preferences for airline travel.2 In this study, carriers set ticket prices to maximize market share regarding not only local travelers but also those travelers who are willing to drive longer distance to purchase discounted fares. Using a discrete choice model, Hartmann estimates consumers’ preferences for flying. Utility is defined as a function of flier and flight characteristics, for example, flier income, ticket price, airport size, as well the distance and distance square traveled between the flier’s MSA and the airport. This captures the tradeoff fliers’ face in their purchasing decisions, trading off cheaper flights for longer driving time to airports. The average flier is found to be willing to drive up to 8.6 miles to lower the ticket price by $10.3 Given that fliers are willing to drive some distance for cheaper flight options, it is hypothesized that airports do not operate in isolation, but rather are linked. This linkage is to be expected, given the complicated network of flights carriers operate. Russon and Riley (1993) find evidence that airport substitutability determines passenger flows in the short-haul market. Dresner et al. (1996) observe that Southwest’s presence on a route lowers prices on routes entered as well as on adjacent routes at nearby airports. Morrison (2001) quantifies the aggregate spillover effect of Southwest entry on ticket prices. He distinguishes between actual, adjacent, and potential entry by Southwest. Finally, Ishii et al. (2005) find evidence that passengers substitute between alternative San Francisco airports based upon the total travel time from their home or office to their destination. While business travelers view the time lost to flight delays and driving to an airport as equivalent, airport choices by tourist fliers are more dependent on the travel time to alternative airports. Although the literature has examined how the existence of alternative airports depress passenger volume and ticket prices, it has yet to explore how the differentials in overall fare levels across airports influence the performance of the affected airports in an explicit spatial econometric framework. We postulate an alternative hypothesis concerning the regional differences in fares. We examine how fare differentials across alternative airports affect performance adversely at small-to-medium sized airports. Using ticket price and quantity data from the US Department of Transportation’s Origin and Destination Survey, a panel of entry decisions for 16 quarters is constructed to investigate this link between performance and fare differentials.

3 THE IMPLICATIONS OF AIRPORT SUBSTITUTION FOR COMPETITION AND PERFORMANCE A key indicator of the attraction of a local airport for airline entry is manifested in the tendency of travelers to seek out lower fares in neighboring airports. To illustrate the approach taken here to identify if this is true, consider the situation facing travelers from Birmingham, Al (BHM), two-and-a-half hours drive from Atlanta, Ga (ATL). There are

2 Hsu and Wu (1997) analyzed another trade-off, the trade-off between passenger travel costs and carrier

operating costs in determining the optimal market size for airports.

3 The results are based on a 10% sample of all flights at the top 138 airports between 1986 and 1996.

GARY M. FOURNIER et al.

214

potential cost savings to Birmingham travelers who eschew the local airport and buy lower-priced tickets out of Atlanta, a larger nearby hub airport.4 To get an appropriate comparison of fare options across the two airports, we first analyze the travel destinations of Birmingham passengers, identifying the most frequent destinations for trips exceeding 500 miles and which are among the largest 50 American cities. We then construct a weighted average cost of flying out of Birmingham to this set of destinations, FIBHM�t , weighted by the share of Birmingham passengers traveling to each destination. The weighted cost of this bundle of airline tickets to the same desti­ nations starting from the alternative airport (Atlanta), FIATL�t , is also calculated and the average differential, between airports. In general, for any airport r and its best alternative airport k, FDrt = FIr�t−1 − FIk�t−1 denotes the lagged fare differential at airport r. To get a better sense of what this measure implies, for the moment let us express the average fare differential in terms of per hour of driving time, although in the empirical models, we do not make this adjustment. FDrt is shown in Figure 1 for these airports over a period of 16 quarters (1996–1999). Thus, from Figure 1, one can see that Birmingham passengers on long trips are typically rewarded with savings of about $60–$80 per hour for the extra time spent driving to the Atlanta airport.5 This large average fare differential provides substantial incentives for passengers with elastic demands and relatively low value of time to drive to Atlanta. It is not Mean Fare Differential

Median Fare Differential

80

60 60

40 40

20

20

0

0 1

3

5

7

9 11 13 15

Quarter

1

3

5

7

9 11 13 15

Quarter

Figure 1 Airline Fare Differentials, Per Driving Hour, Birmingham to Atlanta.

4 Fares out of Atlanta may be subject to market power distortions because it is a major hub and has a dominant carrier, Delta (Borenstein, 1989). Nonetheless, ATL still provides pricing advantages relative to options available at BHM. As shown in Table 1, the mean fare index on the top 50 destinations out of BHM was $174 higher than the same bundle of itineraries out of ATL. 5 The line drawn in the figure and in those that follow is a Loess line plot, a nonparametric smoothing technique attributed to Cleveland and discussed in Härdle (1990) and implemented in axum 6.0 software.

AIRPORT SUBSTITUTION BY TRAVELERS

215

difficult to see how the presence of fare differentials might siphon travelers from local airports to distant ones. If this pattern occurs more generally than BHM to ATL, it raises the question as to why fares are high in small-to-medium size airports, relative to large airports or hubs. First, if fares at the alternative airport are competitively determined, but the local airport is imperfectly competitive, fares may reflect market power. This explanation begs the question of why local market power could persist, since the local carriers should view fares at the distant airport as a limit price that they could effectively negate by pricing low enough to retain the local travelers. In addition, high fare differential should be a strong attraction for new carriers offering service out of the local airport. Indeed, the Birmingham airport depicted in Figure 1 saw entry in the form of a substantial expansion of routes served by Southwest Airlines beginning in the 11th quarter (Quarter 3, 1998). It appears that this development had a dampening effect on fare differentials for that airport. One would expect, moreover, that decreases in fare levels and fare differentials following entry would be accompanied by an expansion in carrier performance indicators, such as local passenger volume and revenues as well. This would be an interesting test of the market-expanding effect of entry, in contrast to other industries where the business-stealing effect predominates. Local carriers’ failure to respond to fare discounts prevailing in the alternative airport can adversely affect their costs of providing air service. Local carriers have an incentive to avoid this problem. Allowing travelers to be siphoned away from the local airport may make the scale of operations too small to achieve minimum efficient scale with the residual population. Nevertheless, differences in the cost of service at the local airport may explain some portion of the observed fare differentials. For example, costs of roundtrip fares to a given destination may be naturally higher at the local airport if service entails an additional connecting flight in the itinerary, compared to direct flights available out of the alternate, large hub. Thus, even in a zero profit competitive equilibrium, fare differentials can exist in which local patrons receive the added convenience of a “door-to-door” round-trip by paying a premium to cover the higher costs. Fare differentials like the example described above may not alone capture the potential profitability of a local market to potential carriers. It would be useful to control in the model for other determinants that cause fares to diverge in pairs of nearby airports. For instance, higher fare differentials are likely to prevail in northern states during winter season because the danger of driving in winter weather would deter the kind of airport substitution envisioned here. Moreover, it is unclear whether fare differentials are high enough to justify the inconvenience of adding an extra driving leg to a trip. In communities with high-income demographics, it is reasonable that the normal income elasticity effects will limit the number of passengers willing to accept additional inconveniences to obtain low or modest levels of cost savings. Moreover, empirical fare differential estimates such as FDr�t may be the downward biased measures of the true time cost savings. If one drives to a large, nearby airport, there may be additional savings in terms of total time in transit in the air transport corridor. For example, the large hub may offer direct service that avoids waiting time en route, thus offsetting some of the driving time.

216

GARY M. FOURNIER et al.

A more complete assessment would factor in parking costs and potential alternatives like limousines6 that may be important. Parking fees could be an appreciable factor at congested airports. For the current study, many of these variables are not available in the data and will be proxied by the use of airport-specific fixed effects. Timing issues are difficult to predict in this context. These issues include assumptions about how long is the delay in responses by passengers to changes in fare differentials, as well as the timing of adjustments that carriers might make to local traffic loads. A key assumption will be that there are substantial, unmeasured fixed determinants of the airlines’ performance at an airport. These may include the adjustment of flight schedules and aircraft deployment, information costs, and a variety of location-specific entry costs, such as the rental of counter space, baggage handling, and landing gates. Consequently, we expect to observe a period of time in which high fare differentials have persistent effects on performance. Thus, in any given quarter, measures of performance, such as load factors and passenger volumes, will depend on lagged average fare differentials and other demand and supply side determinants.

4 EMPIRICAL ANALYSIS The current study uses price and quantity data from the Origin and Destination Survey and other data from the Domestic T100 Segment data administered by the Department of Transportation (DOT). The Origin and Destination data contain a 10 per cent sample of domestic air passenger fares with pertinent information about the passenger’s itinerary, carrier and airports. The first step in our analysis is to select a sample of airports meeting certain criteria. First, the airports are located in areas designated “medium or small air traffic hubs”.7 The DOT constructs geographic areas into “air traffic hubs” based on the percentage of US passengers enplaned within. These areas may contain several airports. Large hubs encompass 1 per cent or more of US enplaned revenue passengers, or at least 5.9 million passengers in the local area. The medium or small airports record at least 1.5 million and 294,000 enplaned revenue passengers, respectively. These latter airports would rank below the top 50 in size. Second, to be included, airports must have a nearby alternative airport, larger than itself and within 200 miles driving distance from it. Given these criteria, there are 65 focal airports in the final sample. A complete list of airports is shown in Appendix A. The construction of the fare differentials in this sample also require a number of operational assumptions. First, noting that business and first class travelers would be less likely to be price sensitive regarding alternative airport opportunities, we limit our

6

For example, Van Galder, Inc. operates an economical service between Chicago O’Hare and Madison, Wisconsin. The one-way trip takes 3 hours instead of the estimated 2 hours of driving but may be a preferred choice for some travelers out of Madison. 7 Although the FAA uses the terminology large-, small-, and medium-sized “hub,” those distinctions are somewhat confusing and misleading. If hubs are taken to be airports where passengers make connections (DFW, ATL, ORD, etc.), the FAA’s notion of a “small-to-medium hub” includes airports where in fact no connections take place. We will refer to the latter as simply “small-to-medium airports.”

AIRPORT SUBSTITUTION BY TRAVELERS

217

Mean Fare Differential

Median Fare Differential

100 120 80 80

60

40 40 20

0

0

1

3

5

7

9 11 13 15

Quarter

1

3

5

7

9 11 13 15

Quarter OMADSM

MCIDSM

Figure 2 Mean and Median Fare Differentials Between Des Moines Airport (DSM) and Its Two Alternatives, Omaha (OMA) and Kansas City (MCI).

sample to economy-class fares.8 Moreover, we find that the distribution of fares is such that the average fares are generally higher and more dispersed than median fares. An example is shown in Figure 2 for the two alternative airports for Des Moines (DSM), Omaha (OMA), and Kansas City (MCI). Finally, we find that for our focal airports, there are up to four alternative airports falling within the (admittedly arbitrary) 200 miles distance adopted here.9 Thus, our initial analysis pairs up the focal airport with the alternative airport providing the best reported fare differential in the data. For the example above, if it appears that Omaha yields a higher return as an alternative, it is chosen for our empirical analysis. To be precise, in determining which airport is the best alternative, we impute a full cost measure of the bundle of flight destinations from the focal market by adding a travel time cost to the calculated cost of the bundle at each alternative. The value of time is assumed to be $30 hours. The best alternative is chosen as the airport that minimizes the full cost of the bundle, fare plus travel costs. 8

The fares used in the empirical analysis to follow are not separated into “restricted” versus “nonrestricted” fares, although there may be some advantages to consider separating them in the analysis. An exception is made for Southwest Airlines since the data for this carrier reports all tickets as first class. 9 Morrison (2001) found that defining markets including alternative airport origins or destinations using a 75 mile radius (as opposed to a 25, 50, 100, and 125 mile radius) had the best fit. Given that we are focusing solely on small-to-medium sized airports where consumers may drive farther distances to alternative airports, we believe it was best to increase the radius. Furthermore, extending the market radius to 200 will introduce little error to our estimation given that we identify the best alternative airport based on the lowest FD plus travel costs.

GARY M. FOURNIER et al.

218

4.1 Measuring Entry It is interesting to examine episodes of new carrier entry into these smaller airports, since perhaps these occurrences are competitive responses to profit opportunities reflected in the fare differentials. Pure entry, the beginning of service by a carrier not currently operating at an airport, occurs relatively infrequently during the sample period for our selected localities. We focus on the share of total airport passenger revenues rather than the market shares on particular routes. Multimarket entry via route expansions is the usual pattern of airlines. Evidence of multiple, coordinated route expansions may be measured by the extent of routes serviced from the origin. For example, Southwest Airlines supposedly made a big expansion in its flight offerings out of Birmingham during this period. We refer Wall Street Journal articles to find evidence of route expansion episodes and search the data to verify that there were threshold changes in the airport market share of a major carrier, that is, a share less than 1 per cent rising in a single quarter to a level of more than 3 per cent which is sustained in at least three successive quarters. Between 1996 and 1999, we identify 25 instances in which a major carrier entered a new airport or a national/regional carrier entered and achieved a substantial market share (3% or more) of overall passenger revenue.10 Mean Fare Differential MHT To BOS

Median Fare Differential MHT To BOS

20

BOSMHT

BOSMHT

20

–20

–20

–60

–60

–100

–100

1

3

5

7

9 11 13 15

Quarter

1

3

5

7

9 11 13 15

Quarter

Figure 3 Effect of Entry on Fare Differentials at Manchester (MHT) Relative to Boston (BOS).

10 There are several instances in our data where a carrier expands its route offerings at an airport and becomes a “beefed-up” competitor. During the period 1996–1999, Southwest and Northwest Airlines are especially active in expanding their route offerings at a number of cities included in our sample. We also calculate the number of destinations serviced by carriers at each focal airport as an alternative basis for identifying this route-expansion form of entry. The method is intractable, however, because of the complexity of the service networks affected by expansion via indirect connections.

AIRPORT SUBSTITUTION BY TRAVELERS

219

While there are too few events to do an extensive empirical analysis of the entry process, we hypothesize that fare differentials will decrease at airports following the introduction of new carrier entry. To illustrate with one pair of airports, Figure 3 gives the plot of changes in fare differentials between Manchester, NH and Boston, MA. The focal airport at Manchester (MHT) experiences entry in the 10th quarter, and this was followed by a rapid and substantial elimination of the differential in subsequent quarters.

4.2 Market Structure, Demographic and Control Variables The variables used in our empirical analysis are defined in Table 2, and the summary statistics are shown in Table 3. Local market structures at the medium-to-small airports display a high degree of variation in market shares because of the presence of numerous, small commuter lines and the incidental presence of charter flights by major carriers outside the market. We calculate the presence, at the airport, of a major carrier by including only those that achieve a market share of 3 per cent or higher in a given quarter. This variable is used for carrier-specific fixed effects and to identify the presence of a low-cost carrier at the airport, low cost.

Table 2 Variable Definitions FI-mean FI-median FD-mean FD-median jet fuel gas prices load factor market revenue low cost HHI pop income Delta American Continental America West Northwest Trans World United US Airways Southwest

Lespeyre index of mean fares at the local airport Lespeyre index of median fares at the local airport Difference between mean fare indexes at the local and alternative airport, previous quarter Difference between median fare indexes at the local and alternative airport, previous quarter Index of jet fuel prices Index of gasoline prices Weighted load factor for flights originating at airport Total revenue for flights originating at airport Binary for the presence of a low cost carrier at airport Herfindahl index divided by 1000 SMSA population (in thousands) at the local airport SMSA per capita personal income (in thousands) at the local airport Binary indicating Delta serves airport Binary indicating American serves airport Binary indicating Continental serves airport Binary indicating America West serves airport Binary indicating Northwest serves airport Binary indicating Trans World serves airport Binary indicating United serves airport Binary indicating US Airways serves airport Binary indicating Southwest serves airport

GARY M. FOURNIER et al.

220

Table 3 Summary Statistics Variable FI-mean FI-median FD-mean jet fuel gas prices load factor market revenue low cost HHI pop income passengers Delta American Continental America West Northwest Trans World United US Airways Southwest

Mean

Standard Deviation

Minimum

Maximum

324.175 266.255 122.952 57.088 100.850 0.632 4153177 0.390 3.201 2550.180 25.428 811.610 0.824 0.725 0.484 0.194 0.646 0.400 0.635 0.632 0.180

110.857 88.771 118.755 10.647 7.950 0.072 4626167 0.488 1.730 4888.418 4.624 1425.411 0.381 0.447 0.500 0.395 0.478 0.490 0.482 0.482 0.384

41.509 39.810 −140.478 37.500 84.500 0.422 24221 0.000 1.130 113.823 11.190 10.313 0 0 0 0 0 0 0 0 0

566.449 485.030 393.041 74.400 110.800 0.875 32600000 1.000 10.000 21200.000 42.376 9397.473 1 1 1 1 1 1 1 1 1

The variation in market shares suggests that structure may be best captured with the Herfindahl index calculated from revenues reported by every carrier with tickets originating at the local airport. Note that reported revenues are an estimate of actual revenues; they are constructed from the 10 per cent Origin and Destination sample. It is not known whether biases may be introduced by nonrandomness in the reporting. For example, all major carriers have established affiliations with regional and local airlines to service routes to and from smaller airports for them. We construct the Herfindahl index, HHI, while noting that, for commuter/regional carriers that are affiliated with others, all revenues for the carrier are attributed to the major carrier (if present in the market) in the affiliation. Additional control variables include an index of the price of jet fuel, jet fuel,11 an index of gasoline prices, gas prices, the logarithm of a quarterly population estimate for the metropolitan statistical area, pop, and an estimate of the personal income per capita in the MSA, income.12 Finally, time-specific fixed effects are constructed for the panel models.

11

Note that the average price paid for jet fuel is a function of long-term contracts, spot market prices, and point

of sale (i.e., from which airport purchased). There are difficult timing issues involved in determining what

relevant cost (the price of fuel today or the contracted price from a year ago) is affecting fares. Unfortunately,

the jet fuel price index is the best publicly available data to proxy airline fuel expenses.

12 Income data are linear extapolations from data available at the REIS website at http://www.ciesin.

org/datasets/reis/reis-home.html.

AIRPORT SUBSTITUTION BY TRAVELERS

221

5 THE ECONOMETRIC ANALYSIS OF SPATIAL CORRELATION IN FARE INDEXES We begin with estimates of a fare index model that allows for spatial correlation and spatial autoregressive errors between each local airport and its nearby alternative. Traditional panel data models that allow temporal and cross-sectional fixed effects may be extended to include spatial correlation in the dependent variable or errors. A model allowing spatial autocorrelation in the dependent variable, the Spatial Autoregressive (SAR) model may be written as: FIt = �W FIt + � + �t + Xt � + �t

(1)

where FIt denotes the R × 1 vector of observations on the dependent variable, the fare index, for each of the R airports at time period t, Xt denotes the R × k matrix of observations of k regressors for each of the R airports at period t, and �t denotes the R × 1 vector of errors at time t. The R × R matrix W is the spatial contiguity matrix that indicates the correspondence between focal and alternative airport pairs in this study. The coefficient � is the spatial correlation coefficient. The R × 1 vector � contains the airport-specific fixed effects, and the scalar � is the temporal fixed effect. A similar model allowing spatial error correlation, the Spatial Error Model (SEM), may be written as: FIt = � + �t + Xt � + �t

(2)

�t = �W�t + �t

(3)

where

As with the SAR model, the matrix W is the spatial contiguity matrix, and � is the spatial correlation coefficient. In both models, �t is assumed to have zero mean and covariance matrix � 2 IR . Elhorst (2003) considers estimation of spatial panel data models. The usual properties of ML estimates are dependent on the form of the contiguity matrix. The ML estimator of � is consistent and asymptotically normal for the common cases of a binary contiguity matrix and an inverse distance contiguity matrix. Furthermore, spatial panel models have the usual incidental parameters problem. That is, if the number of time periods is fixed, allowing the number of cross-section units to go to infinity is not sufficient to insure the consistency of the estimate of �. Consistency of the estimates of � and � requires both the number of cross-sections and the number of time periods to go to infinity. Fortunately, as with the standard linear panel models, this inconsistency does not spill over to the estimates of the coefficients �.13

13

Matlab programs for estimation of both the SAR and SEM models are available on Professor James P. LeSage’s web site at www.spatial-econometrics.com.

GARY M. FOURNIER et al.

222

6 EFFECTS OF AIRPORT SUBSTITUTION ON PERFORMANCE We turn next to the effect of fare differentials on the cost performance of airlines operating at our 65 focal airports. The issue is whether the siphoning of passengers by alternative airports results in reduced efficiency at the focal airport.

6.1 Cost Performance at the Smaller Airport Market Our hypothesis is that the load factors, LFrt , are depressed in areas most affected by “drive-to-fly” behavior. Airlines may face adjustment costs in reconfiguring the deployment of aircraft and be saddled with larger planes than needed, for some period of time. In our model, we transform the load factor into logits and estimate logistical regressions with panel features. Alternatively, the performance effect may be reflected in reduced passenger volume from the airport, Paxrt , or in reduced total passenger revenue, TRrt . In summary, to get at the issue of how cheaper alternative airports affect the cost performance in the focal airport, we estimate these alternative models: LFrt = �r + �t + �0 FDrt + �Xrt + �rt

(4)

Paxrt = �r + �t + �0 FDrt + �Xrt + �rt

(5)

TRrt = �r + �t + �0 FDrt + �Xrt + �rt

(6)

for r = 1� � � � � R and t = 1� � � � � T� The time period for this study includes 16 quarters from 1996 to 1999. The �r are the airport-specific effects and the �t are the time-specific effects. For any airport r and its best alternative airport k, FDrt = FIr�t−1 − FIk�t−1

(7)

denotes the lagged fare differential at airport r, and Xrt denotes regressors with both airport and time variation. The typical estimation procedure is to difference each obser­ vation from its temporal sample mean in order to eliminate the airport-specific effects from the estimating equation and to include a set of temporal dummy variables to capture the time-specific effects. This estimating equation produces the familiar “difference-in­ difference” estimator for the restricted case of two time periods and no regressors.

7 RESULTS 7.1 Spatial Econometric Model of Airport Substitution The first general question is what determines the level of fares observed in the data? We approach this first question because it may be that market structure or other variables are driving the observed fare indexes. Table 4 reports panel regression model based on three models of what drives fare levels: ordinary least squares and two different models

Table 4 Estimates of Spatial Panel Models of the Fare Index Ordinary Least Squares Coefficient jet fuel gas prices load factor low cost pop income HHI Delta American Continental America West Northwest Trans World United US Airways Southwest �

Estimate −2�094 1�680 −49�828 −6�699 −0�028 3�137 5�428 7�006 9�835 −3�578 7�615 −1�393 −2�857 3�936 −20�222 3�671 0

Spatial Autoregressive

Spatial Error Correlation

t-Statistic

p-Value

Estimate

t-Statistic

p-Value

Estimate

t-Statistic

−4�96 4�280 −4�510 −3�410 −3�780 2�870 6�220 2�690 4�580 −1�560 2�520 −0�550 −0�690 1�420 −6�720 1�730 –

0�001 0�001 0�001 0�001 0�001 0�004 0�001 0�007 0�001 0�119 0�012 0�583 0�490 0�156 0�001 0�084 –

−0�620 0�537 −71�938 −6�163 −14�596 3�533 5�050 8�796 7�420 −6�430 8�348 −6�621 −4�028 3�365 −13�671 0�454 0�571

−7�84 5�661 −7�536 −3�180 −2�007 9�098 5�939 3�411 3�605 −2�829 2�810 −2�639 −0�981 1�227 −4�548 0�227 20�146

0�001 0�001 0�001 0�001 0�045 0�001 0�001 0�001 0�001 0�005 0�005 0�008 0�327 0�220 0�001 0�820 0�001

−0�858 0�692 −88�443 −5�357 −28�076 4�761 5�951 12�448 7�172 −3�280 10�135 −5�278 −2�791 2�914 −15�534 −2�134 0�538

−9�821 6�546 −9�022 −2�869 −3�788 11�440 6�689 4�739 3�511 −1�490 3�590 −2�132 −0�729 1�083 −5�295 −1�105 24�535

Note: Dependent variable is FI-mean for all specifications. Estimates of temporal and cross-sectional fixed effects have been suppressed.

p-Value 0�001 0�001 0�001 0�004 0�001 0�001 0�001 0�001 0�001 0�136 0�001 0�033 0�466 0�279 0�001 0�269 0�001

224

GARY M. FOURNIER et al.

incorporating the spatial econometric features. We include the OLS estimates to have a baseline comparison for the spatial econometric specifications. Indeed, while the point estimates vary to some degree, the reported signs of the coefficients are remarkably constant across the three estimation procedures. Covariates in the model are mostly significant. The mean fare is found to be negatively and significantly related to the Herfindahl index. The coefficient of the weighted load factor is negative and significant in both models. Airport market density, as proxied by the local population, is negatively related to the fare index, and we find that the presence of a low cost carrier is associated with a significantly lower level of fares. Both these results reflect the local cost shifters in providing service out of the airport and are consistent with earlier studies. The positive effect of gasoline prices suggests that fares are higher in the face of high fuel prices, although the coefficient of the jet fuel price index is unexpectedly negative and significant. The negative coefficient may be a statistical artifact owing to its high correlation with the gasoline price index.14 The proper interpretation of the coefficient of jet fuel is the marginal effect of an increase in jet fuel prices, given constant gas prices: Factors that have a differential impact on jet fuel relative to gas prices are being measured by the coefficients.15 Gas prices capture variations in the implicit travel costs to an alternative airport and would affect passenger demand, reflecting changes in their disposable income or confidence in the economy and consequently their willingness to pay. Fuel prices, on the other hand, represent an input cost driver for the carriers and capture changes in production. In order to include a complete set of carrier, temporal and airport fixed-effects dum­ mies, there are 159 parameters, while the sample contains 2,080 observations. Never­ theless, the fixed effects are also highly statistically significant and an important control for time invariant sources of unobserved variation in the airports. The spatial econometric models are reported in Table 4 as well. These results are important because they confirm a tight economic linkage between the 65 local airports and best alternative near-by airports. The estimated spatial correlation coefficient is approximately 0.5 in both the SAR and SEM models. Recall that the SAR model is one allowing spatial autocorrelation in the dependent variable explicitly, via the contiguity matrix that identifies the spatially lagged influence of fare levels at the larger airports on its neighboring small-to-medium airport. The SEM model works by allowing spatial error correlation.

14 As one might expect, there is a very strong correlation (0.87) between jet fuel and gas prices. Goldberger (1991) argues that “Researchers should not be concerned with whether or not ‘there really is collinearity.’ They may well be concerned with whether the variances of the coefficient estimates are too large – for whatever reason – to provide useful estimates of the regression coefficients.” By this standard, the high correlation does not induce a “multicollinearity problem,” since the coefficient estimates of these variables are extremely precise (p-values below 0.001). In any event, the estimated impacts associated with the focus variables in this study were robust to alternative specifications involving various combinations of the fuel variables. 15 The fact that the average price paid for jet fuel is a function of long-term contracts, spot market prices, and point of sale could explain why the coefficient on jet fuel price index is negative. Unfortunately, the jet fuel price index may not capture the correct timing of when costs are incurred and their effect on fares; it is, however, the best publicly available data to proxy airline fuel expenses.

AIRPORT SUBSTITUTION BY TRAVELERS

225

No matter how one models the process that generates fare differentials between alternative airports, the direction of the relationship is positive. The results suggest that a $1 decrease in fares at the alternative airport results in less than a $1 decrease in fares at the local airport. As the fares at the alternative airport become relatively cheaper, travelers have a greater incentive to drive to the alternative airport. Perhaps because of relatively low load factors and inefficient scale of operations prevailing at many smaller airports, the decreases in passenger volume will result in a less than proportional decrease in fares at the local airport. At least for price-sensitive travelers, driving-to-fly helps to drive equilibrium fares close together, as the high correlation between these fare indexes shows, be it in the SAR model or the SEM model. We interpret this result to show that passengers are cognizant of travel discounts in the distant market and being siphoned away from the local airport when favorable differentials are available. Table 5 reports the panel regression models aimed at measuring the cost performance effects of the fare differentials.16 In these models, we look at the effect of the onequarter lagged fare differential directly on performance, under the assumption that price sensitive passengers respond to this differential by substituting the alternate airport in response. We find that a relatively high FDrt in the previous quarter is highly significant in reducing the passenger volume and total revenue out of the focal airport, and it reduces the average load factors achieved by carriers as well. This result is striking and may explain in part the fact that these airports often remain unattractive for entrants, despite the high fares prevailing. Low passenger volume and load factors may imply added profit risks of entry, and slow the adjustment. We turn next to the effect of carrier airport entry on the fare index itself. Of the 65 airports in our data, 25 airports present the opportunity to observe performance before and after entry has occurred. We restrict our next sample to quarterly observations on airports where entry is observed, as well as observations on their alternative airports, and estimate again the models presented in Table 4. Now we can examine the procompetitive effect on the fare index by adding a indicator variable for quarters after entry has occurred, postentry. Because the sample is smaller (800 observations), we do not include airport fixed effects. Thus, there is some sacrifice of detail in the model, but the results are indicative of what would likely be found in a much larger sample. Key results are reported briefly in Table 6. Across all the three models we employed earlier to explain fare index, the effect of entry is highly significant. Entry leads to a reduction in fare indexes, effects that are especially large after controlling for spatial effects that link the local airport with its alternative.

8 CONCLUSIONS The concepts of mean and median fare differentials between pairs of alternative airports were introduced into our analysis for several purposes. First, they provide a ready measure of the attractiveness of distant airports to travelers who are willing and able to

16 Reported results are based on models with the mean fare differential; other results, not reported here, were based on median fare differentials and are qualitatively similar.

Table 5 Panel Regression Estimates of Performance Models Logit of Load Factor

Passengers

Coefficient

Estimate

t-Statistic

p-Value

FD-mean jet fuel gas prices low cost pop income HHI Delta American Continental America West Northwest Trans World United US Airways Southwest

−0�001 −0�011 0�015 0�032 0�000 0�000 0�000 0�013 0�096 0�034 −0�145 0�180 −0�195 0�050 0�143 −0�098

−2�690 −4�560 3�890 0�960 −0�350 1�180 0�820 0�360 3�080 0�910 −2�430 4�880 −2�670 1�270 2�940 −2�370

0�007 0�000 0�000 0�339 0�728 0�239 0�412 0�719 0�002 0�364 0�015 0�000 0�008 0�204 0�003 0�018

Market Revenue

Estimate

t-Statistic

p-Value

Estimate

t-Statistic

−0�54 −0�20 0�34 11�44 0�05 −6�16 −1�89 6�56 2�93 21�67 −20�22 14�09 −17�86 16�36 −13�91 31�46

−9�660 −0�440 0�460 1�860 1�690 −1�800 −0�800 1�000 0�510 3�120 −1�840 2�080 −1�340 2�270 −1�560 4�160

0�000 0�657 0�643 0�063 0�091 0�072 0�426 0�318 0�607 0�002 0�066 0�038 0�182 0�023 0�119 0�000

−1�789 −2�380 9�782 48�407 −0�001 0�150 0�004 −78�571 11�363 44�363 34�841 14�16 322�570 0�870 5�111 467�231

−6�130 −1�040 2�610 1�520 −6�930 8�470 0�300 −2�310 0�380 1�230 0�610 0�400 4�660 0�020 0�110 11�920

Note: Dependent variable of each model listed in header. Estimates of temporal and cross-sectional fixed effects have been suppressed.

p-Value 0�000 0�297 0�009 0�129 0�000 0�000 0�762 0�021 0�701 0�218 0�540 0�688 0�000 0�981 0�912 0�000

AIRPORT SUBSTITUTION BY TRAVELERS

227

Table 6 Spatial Panel Models of the Fare Index in Airports where Entry Occurred

Fare Index Models: OLS panel with carrier fixed effects OLS panel with temporal and carrier fixed effects SAR with carrier fixed effects SAR with temporal and carrier fixed effects SEM with only carrier fixed effects SEM with temporal and carrier fixed effects

Postentry Period Estimated Effect

t-Statistic

p-Value

−4�075 −5�045 −8�245 −8�477 −18�671 −18�21

−1.62 −1.94 −1.899 −2.329 −4.913 −4.828

0�105 0�053 0�0575 0�02 0 0

Note: Dependent variable is FI-mean for all specifications. Models are specified with explanatory variables as reported in Table 4, except as noted for fixed effects. There are 25 entry events and 800 observations used.

accept the time costs to receive lower fares. Likewise, these fare differentials indicate opportunities for new entrants to enter profitably. Our empirical analysis suggests that, in medium-to-small airports, there are important competitive effects of alternative airports. The fact that consumers are willing to drive some distance to purchase cheaper flight options links airports in price competition. Operation decisions at each airport are not made in isolation of pricing decisions made at alternative airports. When fare differentials are high, entry is more likely to occur in one form or another. Moreover the cost performance in the focal market suffers in terms of the passenger volume achieved and the load factors on flights emanating from the airport. Higher production costs and consequently higher ticket prices lead to a siphoning of business from the local airport toward the alternative larger airports. Airline carriers find themselves in a bad equilibrium. These conditions probably retard the rate of competitive entry because of the additional risks of entry that may be perceived. If the fare differential becomes sufficiently large, then entry is encouraged, suggesting earnings are expected to be sufficient to offset the risk of low passenger volume. Our results also shed light on why today we observe the opposite phenomenon, travel from large to smaller local airports. Carriers have begun to enter and/or expand their presence in secondary airports. For example, between 1996 and 2002, passengers flying out of Manchester Airport nearly quadrupled from 0.5 million to 1.85 million, while passengers flying out of Boston Logan Airport dropped 10 per cent, down to 11 million (Estabrook, 2003). Congestion and higher landing fees at large airports have raised the full cost, ticket price plus driving costs, reducing the fare differential. Increased passenger volume has raised load factors and lower production costs, making these airports more attractive for carrier entry. A final aspect of fare differentials worth considering is that they may be a symptom of entry barriers that are not specified in the model. For example, Dresner et al. (2002) found that carrier yields rise with airport congestion and allocation of airport gates through exclusive use. Unfortunately, information on these entry barriers is publicly available only for medium and large airports. It would be useful to extend this analysis to the question of what determines the entry and the equilibrium fare differentials in a cross-section of airports similar to those included in our sample. If it were possible to extend the panel to a longer time frame,

GARY M. FOURNIER et al.

228

the effects of entry on these measures of fare differentials may be assessed using spatial panel econometric methods. Our short panel would not suffice to test the hypothesis that entry erodes the differential, for it requires a richer sample of entry events and enough periods before and after the entry event to estimate the model. Nevertheless, consistent with competitive theory, we find, in a limited set of observations on carrier airport entry, some evidence that entry is having the procompetitive effect of reducing fare indexes in periods following carrier entry.

APPENDIX A

Complete List of Airports in Sample Focal Airport

Alternate Airport

Albany, NY Albany, NY

ALB ALB

Albany, NY Albany, NY Allentown, PA

ALB ALB ABE

Allentown, PA Allentown, PA Allentown, PA Amarillo, TX Asheville, NC Asheville, NC

ABE ABE ABE AMA AVL AVL

Asheville, NC Atlantic City Atlantic City Atlantic City

AVL ACY ACY ACY

Atlantic City Bangor, ME Baton Rouge Bufffalo, NY Bufffalo, NY Burlington, VT Burmingham, AL Cedar Rapids, IO Charleston, SC Chattanouga Chattanouga

Travel Time

Travel Miles

BOS LGA

164 155

182 163

EWR JFK LGA

187 167 100

197 174 96

EWR JFK PHL LBB CLT GSP

73 108 72 129 127 66

75 104 73 124 111 53

TYS EWR JFK LGA

146 101 133 128

136 107 132 129

ACY BGR BTR BUF BUF BTV BHM

Boston New York-La Guardia Newark New York-JFK New York-La Guardia Newark New York-JFK Philadelphia Lubbock, TX Charlotte GreenvilleSpartanburg Knoxville, TN Newark New York-JFK New York-La Guardia Philadelphia Portland New Orleans Cleveland syracuse Albany, NY Atlanta

PHL PWM MSY CLE SYR ALB ATL

56 142 82 195 121 175 148

56 135 80 308 140 158 149

CID

Des Moines

DSM

124

132

CHS CHA CHA

Columbia Atlanta Knoxville, TN

CAE ATL TYS

104 129 106

101 125 105

AIRPORT SUBSTITUTION BY TRAVELERS

229

Complete List of Airports in Sample—Cont’d Focal Airport

Alternate Airport

Travel Time

Travel Miles

Colorado Springs Columbia, SC Columbia, SC

COS

Denver

DEN

89

91

CAE CAE

CLT GSP

107 111

101 107

Corpus Christie Dayton, OH Dayton, OH Daytona Beach, FL Daytona Beach, FL Daytona Beach, FL Des Moines Des Moines Eugene, OR Fresno, CA Fresno, CA Fresno, CA Ft Meyers Ft Meyers Ft Meyers Ft Meyers

CRP DAY DAY DAB

Charlotte GreenvilleSpartanburg San Antonio Cincinnati Columbia Jacksonville

SAT CVG CMH JAX

134 77 75 99

148 78 79 107

DAB

Melbourne

MLB

81

87

DAB

Orlando

MCO

69

65

DSM DSM EUG FAT FAT FAT RSW RSW RSW RSW

MCI OMA PDX OAK SFO SJC FLL MIA TPA PBI

179 132 120 173 193 165 134 148 135 161

189 143 128 174 189 157 133 147 143 129

Ft Wayne, IN Grand Rapids

FWA GRR

IND MDW

147 191

138 189

Grand Rapids Greensboro, NC Greensboro, NC GreenvilleSpartanburg GreenvilleSpartanburg GreenvilleSpartanburg Harlingen, TX HarrisburgMiddletown

GRR GSO

Kansas City Omaha Portland Oakland San Francisco San Jose Ft Lauderdale Miami Tampa, FL West Palm Beach, FL Indianapolis ChicagoMidway Detroit Charlotte

DTW CLT

135 104

151 101

GSO

Raleigh-Durham

RDU

76

79

GSP

Atlanta

ATL

167

167

GSP

Charlotte

CLT

94

91

GSP

Columbia

CAE

112

107

HRL MDT

CRP BWI

163 98

129 96

HarrisburgMiddletown

MDT

Corpus Christie Baltimore Washington International Philadelphia

PHL

112

111

(Continued)

GARY M. FOURNIER et al.

230

Complete List of Airports in Sample—Cont’d Focal Airport

Alternate Airport

Huntsville, Al

HSV

Huntsville, Al Islip long island Islip long island Islip long island Jackson, MS Jackson, MS Jackson, MS Jacksonville Knoxville Lansing Michigan Lansing Michigan Lexington, KY Lexington, KY Little Rock Louisville, KY Louisville, KY Lubbock, TX Lubbock, TX Madison, WI Madison, WI Madison, WI Manchester, NH Manchester, NH McAllen, TX McAllen, TX Melbourne, FL Melbourne, FL Melbourne, FL Midland, TX Mobile, AL Mobile, AL Myrtle Beach Myrtle Beach Newburgh, NY

MAF MOB MOB MYR MYR SWF

Travel Time

Travel Miles

BHM

94

96

HSV ISP

Burmingham, AL Nashville Newark

BNA EWR

124 81

126 72

ISP

New York-JFK

JFK

52

44

ISP

LGA

54

45

JSA JAN JAN JAX TYS LAN

New York-La Guardia Baton Rouge Memphis New Orleans Orlando Chattanooga Detroit

BTR MEM MSY MCO CHA DTW

173 187 173 163 110 94

183 211 186 169 106 94

LAN

Grand Rapids

GRR

54

56

LEX LEX LIT SDF SDF LBB LBB MSN

CVG SDF MEM CVG IND AMA MAF MDW

95 78 132 98 136 133 193 144

89 75 145 104 130 126 163 153

MSN MSN MHT

Cincinnati Louisville, KY Memphis Cincinnati Indianapolis, IN Amarrillo Midland, TX ChicagoMidway Chicago-Ohare Milwaukee Boston

ORD MKE BOS

122 79 60

135 83 55

MHT

Portland

PWM

106

96

MFE MFE MLB MLB MLB

Corpus Christie Harlingen, TX Ft Lauderdale Orlando West Palm Beach, FL Lubbock, TX New Orleans Pensacola Charleston, SC Columbia New York-La Guardia

CRP HRL FLL MCO PBI

193 48 153 65 111

161 40 161 62 114

LBB MSY PNS CHS CAE LGA

198 174 88 126 188 83

167 150 70 103 159 75

AIRPORT SUBSTITUTION BY TRAVELERS

231

Complete List of Airports in Sample—Cont’d Focal Airport

Alternate Airport

Newburgh, NY Newburgh, NY Norfolk, VA

SWF SWF ORF

Oklahoma City Oklahoma City Oklahoma City Omaha Omaha Palm Springs Palm Springs Palm Springs Palm Springs Pensacola Portland, ME Portland, ME Providence, RI Providence, RI

OKC OKC OKC OMA OMA PSP PSP PSP PSP PNS PWM PWM PVD PVD

Richmond

RIC

Richmond Richmond

RIC RIC

Roanoake, VA Roanoake, VA Rochester Rochester Sarasota Sarasota

ROA ROA ROC ROC SRQ SRQ

Sarasota Savanna, GA Savanna, GA South Bend, IN

SRQ SAV SAV SBN

South Bend, IN Springfield, MO St Petersburg, FL St Petersburg, FL Syracuse Syracuse

Travel Time

Travel Miles

EWR JFK DCA

86 95 186

77 85 190

DFW DAL TUL DSM MCI BUR LAX SAN SNA MOB BGR BOS BOS LGA

184 182 119 128 159 111 116 133 94 85 142 108 64 161

200 204 127 142 172 121 125 144 100 70 136 103 60 167

IAD

126

129

ORF DCA

88 111

86 113

GSO RIC BUF SYR RSW PIE

124 197 59 83 99 57

102 193 61 93 94 44

TPA CHS JAX MDW

68 125 133 93

53 110 128 93

SBN SGF

Newark New York-JFK Washington National Dallas-Ft Worth Dallas-Love Tulsa, OK Des Moines Kansas City Burbank Los Angeles San Diego Santa Anna Mobile, AL Bangor, ME Boston Boston New York-La Guardia Dulles International Norfolk Washington National Greensboro Richmond Buffalo Syracuse Ft Meyers St PetersburgClearwater Tampa FL Charleston, SC Jacksonville ChicagoMidway Chicago-Ohare St Louis

ORD STL

105 198

109 224

PIE

Orlando

MCO

104

102

PIE

Tampa, FL

TPA

24

17

SYR SYR

Albany, NY Buffalo

ALB BUF

128 123

145 141 (Continued)

GARY M. FOURNIER et al.

232

Complete List of Airports in Sample—Cont’d Focal Airport Syracuse Tallahassee, FL Tucson Tulsa WestchesterWhite Plains WestchesterWhite Plains WestchesterWhite Plains Witchita, KS Witchita, KS Witchita, KS

Alternate Airport SYR TLH TUS TUL HPN

Travel Time

Travel Miles

ROC JAX PHX OKC LGA

84 182 112 119 36

93 184 120 127 30

HPN

Rochester Jacksonville Phoenix Oklahoma City New York-La Guardia Newark

EWR

55

49

HPN

New York-JFK

JFK

45

37

ICT ICT ICT

Kansas City Oklahoma City Tulsa OK

MCI OKC TUL

193 150 173

215 173 189

REFERENCES Bamberger, G. and D. Carlton 2006. “Predation and the entry and exit of low-fare carriers,” in Darin Lee (Ed.) Advances in Airline Economics, Vol. 1: Competition Policy and Antritrust, Elsevier Science, Amsterdam. Berry, S. T., 1992 “Estimation of a model of entry in the airline industry.” Econometrica. Vol. 60: No. 4, 889–917. Borenstein, S. 1989. “Hubs and high fares: Dominance and market power in the U.S. airline industry,” The RAND Journal of Economics, Vol. 20, No. 3, 344–368. Borenstein, S. and N. L. Rose, 1994. “Competition and price dispersion in the U.S. airline industry,” Journal of Political Economy, 653–683. Boguslaski, C., H. Ito, and D. Lee, 2004. “Entry patterns in the Southwest airlines route system,” Review of Industrial Organization 25: 317–350. Brueckner, J. and N. Dyer, and P. Spiller, 1992. “Fare determination in airline hub and spoke networks.” The Rand Journal of Economics, Vol. 23, No. 3, 309–333. Dresner, M., J.-S. C. Lin, and R. Windle 1996. “The impact of low-cost carriers on airport and route competition,” Journal of Transport Economics and Policy September, pp. 309–328. Dresner, M., R. Windle, and Y. Yao, 2002. “Airport barriers to entry in the U.S.,” Journal of Transport Economics and Policy, 36(2), pp. 389–405. Elhorst, J. P. 2003. “Specification and estimation of spatial panel data models,” International Regional Science Review, pp. 244–268. Estabrook, B. 2003. “Smaller airports are growing in stature,” The New York Times, December 21. Evans, W. N. and I. N. Kessides, 1993. “Localized market power in the U.S. airline industry,” Review of Economics and Statistics, 66–75. Goldberger, A. S, 1991. A Course in Econometrics, (Cambridge, MA: Harvard University Press). Goolsbee, A. and C. Syverson, 2005. “How do incumbants respond to the threat of entry? Evidence from major airlines,” National Bureau of Economic Research Working Paper 11072. Härdle, W. Applied Nonparametric Regression. Econometric Society Monographs. (Cambridge University Press: Cambridge UK) 1990.

AIRPORT SUBSTITUTION BY TRAVELERS

233

Hartmann, Monica E 2000. “A structural model of airline safety.” PhD Dissertation, Department of Economics, University of Virginia. Hsu, C. and Y.-h. Wu, 1997 “The market size of a city-pair route at an airport,” Annals of Regional Science Vol. 31, pp. 391–409. Ishii, J., S. Jun, and K. Van Dender 2005. “Air travel choices in multi-airport markets.” Unpublished Department of Economics and Institute of Transportation Studies Working Paper UCI-ITS-WP-05-3 University of California, Irvine. Ito, H., and D. Lee (2003a). “Low cost carrier growth in the U.S. airline industry: Past, present, and future,” Unpublished, Department of Economics Working paper. Brown University, April 9, 2003. Ito, H., and D. Lee (2003b). “Incumbent responses to lower cost entry: Evidence from the U.S. airline industry,” Unpublished Department of Economics Working Paper, Brown University, December 11, 2003. Morrison, S. A. “Actual, adjacent, and potential competition: Estimating the full effect of Southwest Airlines,” Journal of Transport Economics and Policy Vol. 35, Part 2, May 2001 pp. 239–256. Morrison, S. A. and C. Winston, 1990. “The dynamics of airline pricing and competition,” American Economic Review Papers and Proceedings Vol. 80, Part 2, pp. 389–393. Reiss, P. and P. Spiller, 1989. “Competition and entry in small airline markets.” Journal of Law and Economics Vol. 32, No. 2, 179–202. Russon, M. G. and N. F. Riley, 1993. “Airport substitution in a short haul model of air transporta­ tion,” International Journal of Transport Economics Vol. XX, No. 2, 157–173. U.S. General Accounting Office, 1996. Airline Deregulation Changes in Airfares, Service, and Safety at Small, Medium-Sized, and Large Communities, GAO/RCED-96-79, April 1996. Whinston, M. D. and S. C. Collins, 1992. “Entry and competitive structure in Deregulated airline markets: An event study analysis of people express,” Rand Journal of Economics Vol. 23, pp. 445–462. Windle, R. and M. Dresner 1999. “Competitive responses to low cost carrier entry,” Transportation Research Part E, Vol. 35, 59–75.

This page intentionally left blank

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

10 Assessing the Role of Airlines and Airports in Multi-airport Markets∗ Jun Ishii, Sunyoung Jun, and Kurt Van Dender Department of Economics, University of California, Irvine, CA 92697-5100♠

ABSTRACT We study how airline and airport choice interact for air travel consumers departing from a multi-airport region. Consumers do not separately choose an airline and an airport but rather choose among imperfect airline–airport substitutes. The interaction differs by end destination, as airline offerings vary by destination, with some airlines offering no service to the destination from a particular regional airport. We investigate travelers’ choices empirically, using data on passengers departing from the San Francisco Bay Area to four distinct destinations: greater Los Angeles, Phoenix, Portland, and Seattle. The results show that the estimated value of particular airline and airport characteristics to business and leisure travelers differs depending on the available airline–airport offerings for the studied market. We discuss some implications of these findings, including the consequences of further vertical disintegration between airports and airlines.

1 INTRODUCTION Many large metropolitan areas are served by several airports, which may provide sub­ stitutable services. For example, de Neufville (1995) lists 26 metropolitan areas, nine of which are in the US, with more than one airport. He suggests that multi-airport systems

∗

We would like to thank Severin Borenstein for providing summarized DB1A data, and Chuck Purvis and Marc Roddin for help with interpreting the 1995 MTC airline survey. Thanks to Leonardo Basso, Darin Lee, and Anming Zhang for comments on an earlier version. Sunyoung Jun and Kurt Van Dender gratefully acknowledge financial support from the US Department of Transportation and the California Department of Transportation through the University of California Transportation Center. ♠ Tel.: +1 949 8249698; [email protected], [email protected], [email protected]. Corresponding author: Kurt Van Dender.

236

JUN ISHII, SUNYOUNG JUN, AND KURT VAN DENDER

need about 10 million originating passengers per year, to be “successful” in the sense that the available capacity is effectively used. The 24 US airports included in the nine US multi-airport regions serve approximately 46% of all passengers departing from the 61 largest airports in the US.1 In order to study the demand and the supply for air travel in such a context, the market for air travel needs to be defined at the level of the multi-airport region, not at the level of an airport-pair. This chapter reviews literature (this section) and provides data that supports the relevance of defining markets at the level of a region (Section 2). Section 2 also contains a detailed analysis of four travel markets in the Western US, on the basis of the 1995 airline passenger survey of the San Francisco Bay Area. In Section 3, we discuss the results, connecting them to ongoing research on competition in the air transportation industry. The presence of multiple airports in a region does not in itself imply that the airports provide substitutable services from passengers’ point of view, as there are incentives for airlines to concentrate services in a single airport. de Neufville (1995, p. 102) argues that airlines tend to concentrate flights at particular airports because passengers value frequent departures. The availability of multiple airports does not necessarily induce an airline to offer service from each of them, and different airports may serve different market segments rather than provide substitute services. To the extent that such segmentation implies that consumers have no choice of airports, it limits competition among airports (at least, it limits competition in the market). Borenstein (1989), Berry (1990), Morrison and Winston (1989), Evans and Kessides (1993), and Berry, Carnall and Spiller (2006) also emphasize that a larger airport presence increases the value of frequent flier programs and other airline marketing programs, and this enables airlines to charge higher fares profitably as passengers pay the higher airfare to earn the greater marketing reward offered by the dominating airline. Airlines that operate hub-and-spoke networks also concentrate services in their hub airports. However, since the early 1990s, two developments have increased the extent to which airports do provide substitutes, in the sense of offering service to the same destination with the same or a different airline. First, due to the growth of air travel demand, the density of demand has become sufficiently high to attract entry by airlines at different airports. Second, low-cost carriers have gained a large market share in many of the densest markets2 and have often chosen to use a region’s “secondary” airports. Consequently, in order to study the market for air travel from multi-airport regions, a regional perspective should be taken, instead of focusing on travel between airports.

1 Share is calculated on the basis of T-100 segment data, not origin–destination data; cf. Van Dender (2006). If all airports included in T-100 are considered, the 24 airports represent approximately 40% of all departing passengers. The share of multi-airport regions declines slightly between 1996 and 2004, but it is not clear whether this is due to a relatively big slump in metropolitan traffic after 9/11, or to a long-term trend of faster growth outside of multi-airport regions. The T-100 segment data supply information on domestic and international flight segments by US and foreign air carriers, by air carrier, origin and destination airport, and by service class, for enplaned passengers, freight, and mail. The data are available from the Bureau of Transportation Statistics. 2 The domestic market share of low cost carriers increased from 5.5% in 1980 to 30.6% in 2004; the share of passengers exposed to competition from low cost carriers increased from 9.5% in 1980 to 75% in 2004 (http://www.darinlee.net/data/).

AIRLINES AND AIRPORTS IN MULTI-AIRPORT MARKETS

237

Borenstein (2005) provides suggestive evidence that airport competition may reduce the impact of airport dominance on airfare. Table 2 of his paper shows calculated hub premiums at the 50 largest US airports, from 1995 to 2004. Hub airports in metropolitan areas served by multiple airports (e.g., San Francisco Bay Area, Los Angeles, Chicago, New York, Washington DC) seem to be associated with lower hub premiums than hub airports in single airport markets (e.g., Charlotte, Cincinnati, Detroit, Memphis Minneapolis), consistent with the notion of airport competition restricting hub premi­ ums. Morrison (2001) shows that the presence of Southwest airlines in nearby airports disciplines carriers at major airports. Focusing on service quality, Januszewski (2004) finds that longer delays imply lower prices and that the size of the effect depends on the availability of substitutes: When substitute flights are available at the same or at competing airports, changes in service quality have larger effects on prices (the overall effect is estimated at $1.16 per minute of delay, and this increases to $1.55 when there is competition). We have argued that it is useful to assume at the outset that travelers perceive airports in multi-airport regions as substitutes, even if substitution turns out to be modest ex post. In Section 2, we investigate passenger choices on the basis of the 1995 survey of air travelers departing from the Bay Area, which is served by three large airports (San Francisco, Oakland, and San José). The 1995 survey is arguably the best available data for studying air travel choices in multi-airport regions, and it has been used in many studies, because it contains rich detail on passengers’ characteristics, including the actual point of departure at the level of a neighborhood (specifically, a “travel analysis zone”). The main contributions in the airport choice literature are by Skinner (1976), Harvey (1987), Ashford and Bencheman (1987), Pels et al. (2001, 2003), Basar and Bhat (2002, 2004) and Hess and Polak (2005); cf. Ishii et al. (2006) for a more detailed discussion of some of these studies. The choice for an airport may be modeled in combination with ground access mode choice or with airline choice, and the models are estimated using discrete choice techniques of varying complexity and generality. There is wide consensus that ground access-times and quality of service (often measured by flight frequencies) strongly affect the choice for a particular airport. A common trait of these airport choice studies is that they focus on the choice of the departure airport, paying little or no attention to the particular destination to which travel takes place. In particular, no distinction is made between different travel destinations. While such pooling of travel destinations is of less concern for analysis that is supposed to inform strategic airport decisions concerning accessibility and service quality, pooling is harder to justify when the ultimate goal is to understand competition between airlines and airports in a multi-airport context. Instead of pooling destinations, we choose to estimate models for separate travel markets. This is the relevant market definition from a traveler’s point of view; in addition, it may allow for better insight on the role of differences in market structures (e.g., airline presence) across markets. Specifically, we restrict attention to four regional markets, for which sufficient obser­ vations are available: the market for travel from the Bay Area to Greater Los Angeles, Seattle, Portland, and Phoenix. Considering these regional markets has the additional advantage that nearly all flights are direct, so that airline network considerations are relatively unimportant. The survey does not allow distinguishing among key market characteristics for long distance flights, for example, to the metropolitan areas in the

238

JUN ISHII, SUNYOUNG JUN, AND KURT VAN DENDER

Midwest and on the East Coast. Even for some of the selected markets, the available number of travel choices does not always adequately cover the relevant product space, leading to identification problems. The small number of product choices should be seen as a market characteristic, caused by the low density of demand in those markets, and not as a shortcoming of the survey. Such limited variation does not imply that there is no competition between airports, but just that it is harder to identify. The results for the Bay Area to Los Angeles market are in line with expectations, and with the earlier literature. Passengers care about all aspects of time and money costs associated with a particular airline–airport combination, but airline and airport dummy variables are important as well; business passengers’ choices are not affected by fares, while leisure travel demand is fare-elastic. The results for the other markets clarify that an airports’ market share depends on service characteristics like delays and access times, but also on which airlines serve the airport. This implies that airport performance depends on airline behavior. Consequently, the effects of airport policy should be analyzed taking airlines’ responses into account, where possible.

2 EMPIRICAL ANALYSIS Section 2.1 provides descriptive statistics to support the relevance of the multi-airport approach for the travel markets under consideration. They also facilitate interpretation of the econometric results discussed in Section 2.2

2.1 Market Definitions and Descriptive Statistics Table 1 shows market shares by departure airport and by airline for each of the four markets analyzed in this chapter. The travel origin in each market is the San Francisco Bay Area, where three airports are available: San Francisco airport (SFO), Oakland airport (OAK), and San José airport (SJC).3 The four travel destinations are Phoenix (PHX), Portland (PDX), Seattle (SEA), and the Los Angeles region. For the latter, we consider two models, one where only LAX is considered, and one where four destination airports in greater LA are distinguished: LAX, Ontario (ONT), Burbank (BUR), and Orange County airport (SNA). The included destinations cover 38% of all passengers surveyed in the October wave of the 1995 Bay Area survey; the greater LA market represents 22% of the sample.4 First, the tables suggest that restricting attention to departures from only one Bay Area airport is potentially misleading for all of the markets, as the three Bay Area airports each have a large share of total travel. Second, there are differences among the

3

SFO is owned and operated by the City and County of San Francisco, the port of Oakland ownes and operates OAK, and SJC is managed by the City of San José airport commission, which is an advisory body to the San José City Council. The airports compete, at least in the sense that there is no joint governance. 4 The October wave of the survey contains 59 destinations with at least 25 departing passengers. We use airline passenger survey data for October exclusively. We omit the first wave, which took place in August, because of shortcomings in the data for greater LA in August (cf. Ishii et al., 2006) and because the fare data are likely to be (even) less representative of transaction prices in August compared to October.

AIRLINES AND AIRPORTS IN MULTI-AIRPORT MARKETS

239

Table 1 Airline–Airport Combinations and Market Shares (%) for Travel Departing from the San Francisco Bay Area, 1995/Q4 San Francisco (SFO) Greater Los Angeles LAX

Ontario (ONT)

Burbank (BUR)

Santa Ana (SNA)

Portland (PDX)

Seattle (SEA)

Phoenix (PHX)

Southwest Airlines United Airlines Other Total Southwest Airlines United Airlines Other Total Southwest Airlines United Airlines Other Total Southwest Airlines United Airlines Other Total Southwest Airlines United Airlines Other Total Southwest Airlines United Airlines Other Total Southwest Airlines United Airlines Other Total

31�1 4�6 35�7

Departure airport San Jose (SJC)

Oakland (OAK)

Total

16�2

28�7 11�8 40�5

44�9 42�9 12�1 100

26�8

47�1 0�2

73�9 26�0

7�5 23�8

25�8 25�8

26�8

47�3

24�0

41�3 10�8

24�0

52�1

100

28�0

28�4

18�2 46�3

28�4

56�4 25�3 18�2 100

16�6

27�7

15�1 31�7

27�7

44�3 38�4 17�3 100

26�3 9�4 0�6 36�3 24�1 1�0 2�3 27�4

36�8 52�4 10�8 100 68�7 22�7 8�6 100

23�9 23�9 25�3 25�3 38�4 2�2 40�6

10�5 43�0 0�7 43�7 21�4 21�7 3�0 46�1

9�5 20�0 23�2 3�3 26�5

100 65�3 34�7

Source: Own calculations on DB1A (summary provided by S. Borenstein).

destinations regarding airline presence, although United Airlines (UA) and Southwest Airlines (WN) dominate in each of them. Specifically, United Airlines flies to Portland and to Phoenix only from San Francisco, but in the Seattle and Greater Los Angeles markets it provides service from Oakland as well. Southwest Airlines is present in the two secondary Bay Area airports (OAK and SJC) in each market, but in the case of Phoenix it also flies from San Francisco. By and large, the patterns suggest that there is market segmentation in the sense that not all airlines are present at each airport, but at the same time passengers do have the choice between various airport-airline combinations to get to their destination. In

JUN ISHII, SUNYOUNG JUN, AND KURT VAN DENDER

240

other words, the airports provide substitute services, but not perfect substitutes. As there is scope for substitution between at least two of the Bay Area airports in each of the analyzed markets, the relevant dimension for any analysis of pricing and market shares is the Bay Area airport region. Tables 2 and 3, and Figures 1 and 2, provide further suggestive evidence from the air passenger survey. Table 2 shows that travelers departing from the Bay Area do not always choose to travel from the nearest airport, implying that the airports’ “catchment areas” overlap. However, the table also shows that patterns differ across destinations. Specifically, passengers flying to the Los Angeles area whose closest departure airport is San Francisco often chose a different airport, while those closest to San José or Oakland Table 2 Percentage of Passengers Departing From Closest Bay Area Airport, by

Departure Airport and Destination, October 1995

San Francisco (SFO)

San Jose (SJC)

Oakland (OAK)

Business travelers Greater Los Angeles LAX Portland (PDX) Seattle (SEA) Phoenix (PHX)

61.5 68.8 96.4 91.4 97.4

94.8 91.9 84.3 81.4 93.4

88.1 85.3 39.4 78.1 48.0

Leisure travelers Greater Los Angeles LAX Portland (PDX) Seattle (SEA) Phoenix (PHX)

46.8 59.5 94.2 84.6 94.3

90.5 89.5 72.2 81.2 82.0

91.2 88.4 69.6 69.2 45.6

Source: Own calculations on the Bay Area airline passenger survey of 1995.

Table 3 Market Shares (%) of Bay Area Departure Airports by Destination, October 1995

San Francisco (SFO)

San Jose (SJC)

Oakland (OAK)

Business travelers Greater Los Angeles LAX Portland (PDX) Seattle (SEA) Phoenix (PHX)

14 18 72 48 59

58 41 18 22 24

28 41 10 31 17

Leisure travelers Greater Los Angeles LAX Portland (PDX) Seattle (SEA) Phoenix (PHX)

14 27 47 53 52

36 40 26 22 40

50 32 26 25 8

Source: Own calculations on the Bay Area airline passenger survey of 1995.

AIRLINES AND AIRPORTS IN MULTI-AIRPORT MARKETS

241

N W

E S

Departure Airport SFO SJC OAK SFO SJC OAK

Figure 1 Initial Location of Passengers Departing for Greater Los Angeles, October 1995. Source: Own calculations on the Bay Area airline passenger survey of 1995.

choose their closest airport in 9 out of 10 cases. The pattern for the other destinations is nearly opposite: travelers located closest to San Francisco airport most often depart from there, but passengers closest to San José or Oakland often chose a different airport. A partial explanation is that San Francisco dominates in all considered markets, except for the case of travel from the Bay Area to Los Angeles; cf. Table 3. This in turn suggests that the strongest airport substitution patterns are likely to be found in the Los Angeles market. Figure 1 displays the initial trip origin for passengers flying from the Bay Area to greater Los Angeles, by departure airport; Figure 2 explains the same for travelers to Seattle. Comparison of both figures shows that demand for trips to greater LA is higher (or at least that more passengers are surveyed), and that there is more overlap of airports’ catchment areas, than in the case of Seattle. In both cases, however, it would be misleading to assume that spatial proximity is the only determinant of airport choice. Table 4 displays the components of the generalized price of travel that are taken into account in our model. Consider first differences across destinations. Fares generally increase with distance and possibly decline with the competitiveness of the market under consideration, explaining part of the difference between the LA markets and the other markets. Another difference between the LA markets and the smaller markets is that

242

JUN ISHII, SUNYOUNG JUN, AND KURT VAN DENDER

N W

E S

Departure Airport SFO SJC OAK SFO SJC OAK

Figure 2 Initial Location of Passengers Departing for Seattle, October 1995. Source: Own calculations on the Bay Area airline passenger survey of 1995.

the frequency of flights in the former is much higher. This indicates that the density of demand in the Bay Area to LA market is particularly high, a feature that can be expected to induce more intense competition than in the other markets. Next, notable differences between the departure airports are (a) that flights out of San Francisco tend to be more expensive and are subject to longer delays, (b) frequencies are lower and access times shorter in San José. Overall, the generalized cost of travel from San Francisco seems at least as high as that from other airports, which suggests that – as long as several alternatives are available and are actually chosen – travelers perceive departing from San Francisco as a superior alternative.

2.2 Econometric Analysis In order to further investigate what determines passengers’ travel choices, we combine the passenger survey with other datasets. The final data describe the chosen departure airport, arrival airport (where relevant), carrier, peak or off-peak departure, and early or

AIRLINES AND AIRPORTS IN MULTI-AIRPORT MARKETS

243

Table 4 Time and Money Costs Per Airline Market, October 1995 Business

Leisure

SFO

SJC

OAK

SFO

SJC

OAK

Fare ($) Greater Los Angeles LAX Portland (PDX) Seattle (SEA) Phoenix (PHX)

77�04 67�65 63�83 72�37 79�35

57�38 53�81 61�33 61�34 89�64

53�5 50�71 79�43 58�5 79�14

68�65 66�35 64�03 72�61 78�14

55�62 52�48 62�11 58�18 83�65

53�48 50�69 94�27 58�31 80�2

Frequency (flights/h) Greater Los Angeles LAX Portland (PDX) Seattle (SEA) Phoenix (PHX)

1�91 2�49 0�37 0�59 0�51

0�67 0�89 0�31 0�33 0�45

1�34 1�85 0�51 0�41 0�64

0�97 1�06 0�37 0�57 0�47

0�38 0�48 0�25 0�28 0�54

0�71 0�95 0�41 0�44 0�84

Access-time (min) Greater Los Angeles LAX Portland (PDX) Seattle (SEA) Phoenix (PHX)

25�79 28�23 31�93 29�6 32�18

15�98 16�92 15�42 15�26 15�72

39�88 37�23 36�00 37�77 24�54

32�84 33�74 26�51 29�84 34�36

17�88 17�24 16�9 17�24 16�21

37�45 33�75 43�37 37�89 34�83

Total delay (min) Greater Los Angeles LAX Portland (PDX) Seattle (SEA) Phoenix (PHX)

20�49 22�29 26�59 26�36 15�25

7�01 10�3 9�36 12�03 1�75

12�25 14�77 3�99 10�61 7�21

21�78 22�63 26�55 27�28 16�46

7�4 10�29 10�4 12�78 2�78

13�63 16�53 5�59 11�00 7�15

Source: Own calculations on the Bay Area airline passenger survey of 1995.

late flight.5 The time cost components include the driving time from the initial origin to the airport, the expected flight delays at the departure airport and at the arrival airport, and the schedule delay cost as approximated by the frequency of flights per airline per airport per hour. The money cost is an approximation of the flight fare. Lastly, we have information on a set of socio-demographic variables, including whether passengers travel for business or leisure purposes, the exact location of departure in the Bay Area, whether travelers are residents of or visitors to the Bay Area, and the travelers’ income group. Our primary data source is the Airline Passenger Survey as conducted in August and October 1995 by the Metropolitan Transportation Commission (MTC) in collaboration with SFO, SJC, OAK, and Sonoma County Airport (STS).6 According to the survey 5 Consistent with the road network model data, peak hours are from 6–9 a.m. and 3–6 p.m.; all other hours are off-peak. Early flights are those arriving within a time interval defined by the earliest arriving flight at that airport, plus 30 min. Late flights are those arriving at an airport in a time interval defined by the latest arriving flight, minus 30 min. 6 http://www.mtc.dst.ca.us/datamart/airpass1.htm.

JUN ISHII, SUNYOUNG JUN, AND KURT VAN DENDER

244

description, a purposely large number of interviews was conducted at San José, but otherwise the sampling is random. We use airport–airline weights in order to correct for explicit choice based sampling and to align the survey market shares with those observed in the Origin and Destination Survey (DB1A). The airline passenger survey is combined with several secondary sources. First, the 1998 car travel times from a passenger’s initial origin in the Bay Area to the airports are derived from the MTC’s transportation network model.7 Second, DB1A provides aggregate fare data. Third, Airline Online Performance Data from the Bureau of Transport Statistics provide information on delays at the level of the origin and destination airports.8 Fourth, the Worldwide Through Flight Schedules Database obtained from OAG is used to construct passengers’ choice sets. We assume that the complete set of flights was actually available to passengers at the time they purchased a ticket. The basic specification used in each the markets is the same, in that (1) we correct for choice based-sampling through airport–airline weights and using a WESMLE estimator, to estimate a weighted conditional logit model (Manski and Lerman, 1977), and (2) the left-hand side variable and the set of explanatory variables are the same in all markets, except for the exclusion of the frequency variable in the non-Los Angeles markets.9 This allows us to compare the impact of airline and airport characteristics on passengers’ choices, across markets. Formally, a passenger’s choice maximizes her indirect utility. Let the following sets denote departure airports, arrival airports, airlines, and time periods i ∈ �OAK� SFO� SJC� j ∈ �BUR� LAX� ONT� SNA� or j = PDX or j = PHX or j = SEA k ∈ �UA� WN� Other�� where Other = �AS� DL� HP� QQ� US� t ∈ �Peak� Offpeak� �j = 1if j ∈ �BUR� LAX� ONT� SNA� and 0 otherwise The indirect utility of a specific alternative is Vi�j�k�t =

� i�=SFO

�i D i +

�

� j Dj +

j=LAX �

�

�k Dk

k=Other �

+ �1 Peak + �2 Fare + �j �3 Freq + �4 Access + �5 Delay + �6 Inc_group_2 + �7 Inc_group_3 � � + �early�j Dearly�j + �late�j Dlate�j + �ijkt j

7

j

ftp://ftp.abag.ca.gov/pub/mtc/planning/forecast/RVAL98/.

http://www.transtats.bts.gov.

9 The frequency variable was omitted for these markets because the relatively low number of observations

combined with the limited number of available airline–airport combinations prohibits distinguishing between

the impact of frequency and airport or airline fixed effects. Estimated coefficients of other key variables,

notably access-time, are hardly affected by this change in specification.

8

AIRLINES AND AIRPORTS IN MULTI-AIRPORT MARKETS

245

The model is estimated separately for business and for leisure travelers, as was found to be relevant in Ishii et al. (2006).10 Table 5 shows estimation results for the different travel markets, and Table 6 contains the marginal effects. The results for the greater Los Angeles area are the most satisfactory.11 Like in Ishii et al. (2006), we find positive fixed effects for United Airlines and Southwest airlines; San Francisco seems to be the preferred departure airport and LAX the preferred arrival airport. The peak dummy variable is positive for business travelers, and negative (though not significant) for leisure travelers. Fares have no impact on business travelers’ choices, but do affect leisure travel choices negatively, and the fare effect declines as income rises. Choice options with higher frequencies, shorter access times, and with shorter delays are more likely to be chosen. Travelers prefer early flights and dislike late flights, although the effect is insignificant in a number of cases. Restricting attention to flights to LAX leads to a loss of precision, but does not fundamentally alter key results. One difference is that business trip decisions now also are sensitive to fares. More troublesome is that delay effects take the “wrong” sign for business travelers. The unsatisfactory result concerning delay effects occurs in other markets with a single destination airport as well, and seems to follow from multi-collinearity between airline dummies and delays. The coefficient of access-time is precisely estimated in the LAX market as well as in the other smaller markets. In general, however, the estimates are less precise than in the greater LA case, because fewer observations and fewer airport–airline combinations are available.12 In order to compare estimation results across markets, we turn to the marginal effects reported in Table 6.13 A first observation is that the marginal effects of a change in access-time appear to be smallest in the greater Los Angeles market, they are larger when LAX is considered separately, and still larger in the other markets. The large

10

Focusing exclusively on the greater Los Angeles market, Ishii et al. (2006) compare the conditional logit specification to a mixed logit model, and find little difference between both. For this reason, and because it is not a priori clear whether airline choice precedes airport choice (because of airline loyalty) or vice versa (because of proximity), we use the conditional logit model here, rather than a nested logit specification. 11 The numerical results for greater Los Angeles differ from those in Ishii et al. (2006) because of the weighting scheme and inclusion of the peak dummy. Instead of airport-based weights (which correct for the known source of sampling bias, i.e., oversampling at SJC), here we use airport–airline based weights, because of apparent bias in airline shares in several smaller markets. The weighting scheme does not strongly alter results for greater LA. We note that the greater LA market represents a large share of the entire passenger survey, suggesting that the greater LA results may be similar to results from models that pool across multiple destinations (the standard practice in the airport choice literature). 12 For the cases of Portland and Seattle, we have experimented with specifications that include a dummy variable for Alaska Airlines (AS), on the grounds that it is an important player in these West Coast markets. The results for these specifications are qualitatively similar to those presented here, and the AS dummy variable is not precisely estimated (in line with results for the other airlines). We conclude that the data do not allow distinguishing between the current model, where AS is included in the set of “other airlines”, and a model where AS is treated separately. It is reasonable to consider the omitted airline in the model without the AS dummy as Alaska Airlines. 13 The marginal effects for continuous variables are the derivatives of the probability of the actual choice with respect to a change in that choice’s characteristic; for dummy variables, they are the probability differences. Cf. Ishii et al. (2006) appendix 2, for a detailed explanation.

Table 5 Conditional Logit Estimates of Travel Choices, October 1995 Table 5.1

San Francisco Bay Area to Greater Los Angeles

Variable

San Jose airport dummy Oakland airport dummy Southwest Airlines dummy United Airlines dummy Burbank airport dummy Ontario airport dummy Santa Ana airport dummy Peak period dummy Fare Flight frequency Access-time Total delay Income group 2 Income group 3 Early flight Burbank Early flight LAX Early flight Ontario Early flight Santa Ana Late flight Burbank Late flight LAX Late flight Ontario Late flight Santa Ana Log L N

Business

Leisure

Variable

Coeff.

SE

Coeff.

SE

−0�0580 −0�2750 0�8694 0�6825 0�0908 −0�0586 0�0943 0�7399 −0�0177 0�7714 −0�0875 −0�0437 0�0026 0�0054 0�1011 0�2879 0�2656 0�3972 0�1165 −0�2918 −0�3467 1�4801

0�3358 0�2501 0�1961 0�2271 0�1884 0�2137 0�2212 0�1267 0�0110 0�1237 0�0050 0�0159 0�0090 0�0089 0�2972 0�2801 0�3172 0�2677 0�3219 0�2473 0�2976 0�4406

−0�9113 −0�4094 1�8072 1�0901 −0�5672 −0�9569 −0�9553 −0�0546 −0�0287 0�9042 −0�0762 −0�0261 0�0243 0�0123 0�1348 0�4811 0�8220 0�9826 0�1306 −0�6314 0�0373 −0�3938

0�3646 0�2450 0�2738 0�2397 0�2066 0�2554 0�3236 0�1786 0�0123 0�2432 0�0048 0�0188 0�0106 0�0126 0�3412 0�2778 0�3212 0�3650 0�2721 0�2290 0�3022 0�6587

−2652.9048 935

Table 5.2 San Francisco Bay Area to LAX

−2480.1961 817

Business

Leisure

Coeff.

SE

Coeff.

SE

San Jose airport dummy Oakland airport dummy Southwest Airlines dummy United Airlines dummy

0�7091 −0�4519 −0�2939 0�6828

0�6312 0�3957 0�3560 0�2837

−0�9761 −0�7922 1�4484 1�1774

0�6051 0�3727 0�4303 0�2941

Peak period dummy Fare Flight frequency Access-time Total delay Income group 2 Income group 3

1�1440 −0�1066 1�1668 −0�0881 −0�0348 −0�0011 0�0272

0�2684 0�0271 0�2172 0�0072 0�0322 0�0177 0�0181

−0�3161 −0�0727 1�1323 −0�0778 −0�0424 0�0273 0�0680

0�2999 0�0244 0�4067 0�0063 0�0306 0�0145 0�0207

Early flight LAX

0�7621

0�3854

0�6564

0�3133

Late flight LAX

0�0367

0�3462

−0�6568

0�2937

Log L N

−642.7972 385

−741.4634 417

Table 5.3 San Francisco Bay Area to Portland, Seattle, Phoenix Variable

Portland Business Coeff.

San Jose airport dummy Oakland airport dummy Southwest Airlines dummy United Airlines dummy Peak period dummy Fare Access-time Total delay Income group 2 Income group 3 Early flight Late flight Log L N

Seattle Leisure

SE

Coeff.

−0�5879

3�6028

−2�2214

−0�9783

4�1331

0�4574

−0�9330

0�9914

−0�1276

2�8295

2�1050

0�0888 −0�5563 −0�1063 −0�0244 0�0238 0�2392 −4�1484 1�2073

Business

Business

Leisure

Coeff.

SE

Coeff.

SE

Coeff.

SE

4�3384 −11�6870

7�3828

7�4959

8�1222

1�1350

1�5002

0�5673

1�5039

4�7011 −10�4488

7�8089

7�9956

8�2050

0�0763

0�9981

0�7486

0�9283

1�5756

−0�1460

1�2670

9�8758

8�2664

1�3012

1�9753

3�3714

1�9957

2�9991

2�2762

1�0708

1�3454

6�8828

5�7964

0�3679

1�7689

1�7488

1�8588

0�9343

−0�0089

1�1485

1�4233

3�6755

1�1842

2�6175

0�8931

1�0771

1�2111

1�0591

0�4146 0�0279 0�1907 0�2015 0�2389 4�2045 3�7549

−0�5945 −0�1684 0�0662 0�2524 0�2214 −5�0836 −1�0016

0�5325 0�0456 0�1613 0�2053 0�4108 5�3481 2�5013

−0�5438 −0�1603 −0�3471 −0�0235 0�0381 −1�1953 3�6656

0�2379 0�0346 0�4726 0�0853 0�0948 4�0029 2�1329

0�5015 −0�1087 0�1763 0�0554 0�1250 −1�5873 1�3002

0�8176 0�0144 0�2553 0�0395 0�0765 2�5724 1�0051

0�0005 −0�1636 0�1773 −0�0034 −0�0164 1�0412 1�3166

0�0869 0�0323 0�0966 0�0376 0�0407 0�6069 0�6881

0�0837 −0�1361 0�1980 −0�0401 0�0485 0�3639 1�4495

0�0891 0�0236 0�1007 0�0428 0�0565 0�5368 0�6774

−156.7490 178

Coeff.

Leisure

SE

−231.3493 140

SE

Phoenix

−325.6681 228

−413.8042 367

−248.3500 179

−332.7815 177

Table 6 Marginal Effects Calculated from Conditional Logit Estimates, October 1995 Table 6.1

San Francisco Bay Area to Greater Los Angeles

Table 6.2 San Francisco Bay Area to LAX

Variable

Business

Leisure

Variable

Business

Leisure

San Jose airport dummy Oakland airport dummy Southwest Airlines dummy United Airlines dummy Burbank airport dummy Ontario airport dummy Santa Ana airport dummy Peak period dummy Fare Flight frequency Access-time Total delay Income group 2 Income group 3 Early flight Burbank Early flight LAX Early flight Ontario Early flight Santa Ana Late flight Burbank Late flight LAX Late flight Ontario Late flight Santa Ana

0�1126 −1�5192 2�8909 1�0596 0�5164 −0�5458 0�5302 1�3303 −0�0909 4�7513 −0�5388 −0�2689 0�6389 1�3885 0�0122 0�2335 0�0327 0�0405 0�0179 −0�5267 −0�0684 0�2498

−4�5163 −1�7248 7�0475 −0�5582 −2�3926 −4�6911 −4�0902 −0�1015 −0�0924 6�3920 −0�5386 −0�1846 9�8909 0�0298 0�0101 0�4037 0�0920 0�0472 0�0383 −1�7556 0�0085 −0�0056

San Jose airport dummy Oakland airport dummy Southwest Airlines dummy United Airlines dummy

12�3385 −7�5291 −7�0736 11�6221

−6�9647 −8�8635 10�3127 3�4662

Peak period dummy Fare Flight frequency Access-time Total delay Income group 2 Income group 3

4�7793 −1�3594 15�6562 −1�1826 −0�4670 −4�0731 25�1770

−1�3884 −0�6767 15�3062 −1�0516 −0�5734 15�2858 47�4600

Early flight LAX

2�1522

1�5780

Late flight LAX

0�2431

−4�8168

N

935

817

N

385

417

AIRLINES AND AIRPORTS IN MULTI-AIRPORT MARKETS

249

Table 6.3 San Francisco Bay Area to Portland, Seattle, Phoenix Portland

Variable

San Jose airport dummy Oakland airport dummy Southwest Airlines dummy United Airlines dummy Peak period dummy Fare Access-time Total delay Income group 2 Income group 3 Early flight Late flight N

Seattle

Phoenix

Business

Leisure

Business

Leisure

Business

−10�1119

−32�1626

−46�0469

25�6700

17�8395

6�4878

−14�4622

13�3576

−48�4414

19�7371

−4�0950

12�8398

−37�0955

−18�3572

−15�1054

25�2275

15�6749

34�4820

61�6568

42�4427

22�3467

15�2954

−6�5724

−10�4938

0�5101

−0�0335

12�2862

0�0756

4�0444

3�4427

−16�9308 −2�0411 −0�4687 16�5458 57�0993 −0�0093 11�4478

−11�4865 −2�5336 0�9961 41�7689 6�5258 −0�0029 −4�4293

−13�4037 −3�3051 −7�1553 −29�7281 50�6901 −12�2352 23�8249

4�1099 −1�1244 1�8240 15�7720 27�4920 −0�3424 10�0123

1�8291 −2�5090 2�7181 −1�4731 −13�5330 4�3192 7�8194

2�4639 −2�7136 3�9475 −44�9019 70�1515 1�0106 14�3755

140

178

228

367

179

Leisure

177

marginal effects for the Phoenix market may be related to the availability of Southwest flights from each of the Bay Area airports, which implies there are no constraints on airport choice for passengers preferring to fly with Southwest. Similarly, the small effects for greater Los Angeles may follow from the fact that not all 12 airport-pair combinations are served by each airline. The apparent difference between the marginal effects for LAX only and those for the smaller markets may indicate that airline loyalty is particularly strong in the LA market, or it may follow from the relative imprecision in the estimates for the smaller markets (as marginal effects depend on all coefficients). A different hypothesis is that changes in a generalized cost component have a bigger impact on airport choice when the initial generalized price is higher, that is, there are decreasing benefits from reductions in the generalized price (Table 4 shows that the different components of generalized cost, in particular fares, frequencies and delays, tend to be lower for the LA markets than for the other markets). It seems reasonable to conclude that the effects of changes in access-time differ across markets and are conditional on the available airport–airline combinations. In that case, however, it is clear that the assumption of exogeneity of these combinations is tenuous: Airlines may very well wish to adapt their service in some markets, when access-times or other airport characteristics change. The airline responses will in turn provoke demand reactions, so that the ultimate impact of a change in an airport characteristic on the size

250

JUN ISHII, SUNYOUNG JUN, AND KURT VAN DENDER

and the composition of passenger traffic at any particular airport may be different from what the demand model predicts. Second, with the notable exception of the greater Los Angeles market, fare effects are not precisely estimated, and the coefficients as well as the marginal effects take positive signs in some cases. The fare data are averages per combination airline and airport-pair, so fares are measured at a much higher level of aggregation than other variables, and possibly very different from individual passengers’ transaction prices. Third, we note a pattern in the marginal effects of airline and airport dummies, with the caveat that many coefficients are not estimated very precisely. Restricting attention to business travelers, the marginal effect of the United Airlines dummy is positive in all markets except for Phoenix. Similarly, the marginal effects of the Southwest dummy and of the San José dummy are mostly negative, but not for Phoenix. As was noted before, a key difference between the Phoenix market and the other markets is that Southwest provides service to Phoenix from San Francisco, whereas in the other markets Southwest only flies from Oakland and San José. Consequently, the difference between the markets may help us identify better the separate impact of airport and airline characteristics. First, United Airlines is usually preferred because it dominates at San Francisco – if it does not, like in the Phoenix market, it is not the preferred airline. Second, San Francisco is usually preferred because United Airlines dominates that airport; San Francisco is not preferred over San José in the Phoenix market. The results suggest that airport dominance may favor both the airport and the airline, in the sense that consumers prefer the bundle formed by the dominating airline and the airport. This bundling effect becomes weaker when airlines compete directly in an airport, but seems sufficiently strong to withstand competition from airlines at adjacent airports.

3 DISCUSSION We have analyzed passengers’ choices of airport – airline combinations in four travel markets in the Western US, by estimating weighted conditional logit models on the basis of a dataset constructed around the 1995 passenger survey for the Bay Area airports. The results for the market for trips to greater Los Angeles are as expected, and are in line with earlier airport choice studies. For the other markets, qualitative results concerning access time are similar to those for greater LA, while other coefficients often take reasonable values but are imprecisely estimated. The reasons for this lack of precision are that the survey contains barely enough observations to allow estimating models for separate markets, and the limited availability of different airport–airline combinations in those markets. The former issue is one of sample size, the latter is related to the density of demand: more airline–airport combinations may emerge when more passengers wish to travel to the selected destinations. On a general level, our results show that airports in a multi-airport region compete, even if airlines do not often trade one airport for another, because consumers perceive the different airline–airport combinations as imperfect substitutes. Furthermore, despite the limited statistical power of some results, the approach of distinguishing markets on the basis of travel destinations is useful. One insight is that changes in airport characteristics, for example, access-time, affect passengers’ choices, but that the effects

AIRLINES AND AIRPORTS IN MULTI-AIRPORT MARKETS

251

differ across markets. Moreover, the differences between markets seem to be related to the available airport–airline combinations. It makes intuitive sense that a reduction in an airport’s access time will have a bigger effect on its market share when many airlines provide service from that airport. But then a comprehensive analysis of air transport markets should also take into account that such a reduction in airport access time may induce airlines to enter that airport, so that airline availability (which in turn affects consumers’ choices) cannot be treated as an exogenous variable. Such supply side issues have received relatively little attention in the multi-airport literature, however. Research on multi-airport markets is relevant to policy concerning the relation between airlines and airports. At present, the allocation of airport capacity14 to airlines in the US is often guided long-term contractual arrangements, but there is a tendency to weaken the ties between airlines and airports (FAA/OST, 1999).15 One reason is that airports relying on a single carrier for revenues find it increasingly difficult to finance new capacity, possibly because of bankruptcy risk. Van Dender (2006) finds that current airport charges in general are related to market structure, but the impact of airline concentration at the airport is not easily explained. This finding may be determined by strong vertical integration (which affects airport charges through vari­ ous channels) or by the absence of clear objectives determining airport behavior. An empirical study focusing on the effects of more vertical separation between airlines and airports is Hartmann (2006), who finds that increased flexibility in the arrangements between airports and airlines positively affects airlines’ probability of offering a non­ stop airport connection, keeping non-routing service characteristics (including prices) exogenous. More generally, one can ask what could be the consequences of different vertical relations between airlines and airports. There are reasons to be skeptical about the merits of full vertical separation and deregulation. A general argument, cf. Borenstein (1988), is that unregulated allocation of access on the basis of willingness-to-pay does not necessarily lead to efficiency when there is imperfect competition. de Neufville (1999) argues that US airports are effectively run like private organizations, despite being publicly owned, largely because of strong competition between airlines and airline control over key airport decisions. Oum et al. (2006) indeed find no evidence that US airport are less well run than privately owned airports in other parts of the world. This view highlights that ownership does not fully determine behavior, and by extension implies that formal vertical disintegration between airlines and airports need not necessarily result in factual vertical separation, nor in strong changes of the current structure of airport access pricing. A small but quickly growing theoretical literature (reviewed by Basso and Zhang, 2006a, in this volume) focuses on the interaction between airport congestion, market power, and price and capacity decisions. The standard assumption is that airports are

14

Access to airports is broader than runway access; it also includes terminal and handling access, and so on. Airports assign slots within an exclusive use, preferential use, or common pool arrangement. As a result, potential entrants may face difficulties obtaining access to many airports, airports rely on revenues from one or a few carriers, and airlines may have veto power concerning decisions on airport development. This situation affects competition among airlines, as well as airport prices and airport capacity decisions.

15

252

JUN ISHII, SUNYOUNG JUN, AND KURT VAN DENDER

monopoly providers of access to a given region. For example, Basso (2006) shows that vertical integration may outperform separation, when airports maximize profit. However, private ownership tends to lead to underprovision of capacity and traffic volumes below surplus-maximizing levels. Some contributions address similar issues in a multi-airport context. De Borger and Van Dender (2006) consider sequential capacity and price com­ petition between congestible facilities like airports, and find that a duopoly leads to lower prices and higher levels of congestion than either a monopoly or the surplus-maximizing solution. In addition, they find that asymmetric outcomes (a small, low quality, cheap air­ port and a large, high quality, expensive airport) are possible, even if the facilities are ex ante identical. Basso and Zhang (2006b) extend this model, amongst others by explicitly accounting for vertical relations with airlines. They find that monopolists will provide higher levels of service than would be obtained in the surplus-maximum, when there is imperfect competition among airlines at an airport.16 However, when the airlines and the airports are vertically integrated, the monopolist will provide the level of service that would be obtained in the surplus-maximum. Note that this service level is not optimal in a second-best sense, as surplus-maximization under the monopoly prices would call for higher levels of service. These models warn us that the consequences of vertical sep­ aration of airports and airlines are not straightforward, and that vertical separation does not necessarily increase total surplus. In particular, the degree of competition between airlines affects the benefits of the “commercialization” of airports, and airport capacity decisions are likely to be distorted both in a multi-airport context and in the case of single airports.17 Lastly, we note that airport choice has implications for the empirical literature on airline entry. Much of the literature, for example, Reiss and Spiller (1989) and Berry (1992), has focused on airport-pairs (or city pairs, but without distinguishing among the different city airports). If city airports are imperfect substitutes, then the airport itself serves as a form of product differentiation for the airline that needs to be accounted for. Increased attention has been brought to the issue of low-cost carriers, most notably Southwest, entering the markets of incumbent, hub-and-spoke carriers. Recent research includes Bogulaski et al. (2004) and Goolsbee and Syverson (2005). Given that low cost carriers often enter the adjacent, non-hub airport in major metropolitan areas (e.g., Southwest using Baltimore Airport to serve the Washington DC), current studies may be underestimating the change in market share and incumbent response effected by the entry of a low cost carrier at the hub airport.

16 There clearly is imperfect competition between carriers at many major airports. If airport decisions are abstracted, the analysis of Brueckner (2002) and Pels and Verhoef (2004) may apply. These authors observe that Cournot interaction between airlines leads to partial internalization of marginal congestion costs, as nonatomistic carriers internalize marginal congestion costs that their marginal flight imposes on their other flights. However, Harback and Daniel (2005) find that non-atomistic carriers do not partially internalize congestion costs, and offer potential explanations. 17 The literature on the deregulation and privatization of airports notes that abuse of market power may be prevented by the complementarity between the demand for aviation services and the demand for concession services (Starkie, 2000). Concessions represent some 50% of revenue in large US airports, and probably a larger share of profits. Oum et al. (2004) find that the complementarity indeed reduces airside charges, but there is no guarantee that they would be set at the socially optimal level.

AIRLINES AND AIRPORTS IN MULTI-AIRPORT MARKETS

253

REFERENCES Ashford N. and M. Bencheman, 1987, Passengers’ choice of airport: An application of the multinomial logit model, Transportation Research Record, 1147, 1–5. Basar G. and C.R. Bhat, 2002, A Parameterized Consideration Set Model for Airport Choice: An Application to the San Francisco Bay Area, Technical paper, Department of Civil Engineering, University of Texas at Austin. Basar G. and C.R. Bhat, 2004, A parameterized consideration set model for airport choice: An application to the San Francisco Bay area, Transportation Research Part B, 38 (10), 889–904. Basso, Leonardo J., 2006, Airport Ownership: Effects on Pricing and Capacity, Sauder School of Business Working Paper (http://ssrn.com/abstract=849584). Basso L. and A. Zhang, 2006a, Congestible Facility Rivalry in Vertical Structures, mimeo, Sauder School of Business, University of British Columbia. Basso L. and A. Zhang, 2006b, A survey of analytical models of airport pricing, In: D. Lee (Ed), Advances in Airline Economics – Volume 2: The Economics of Airline Institutions, Operations and Marketing, Elsevier. Berry S., 1992, Estimation of a model of entry in the airline industry, Econometrica, 60 (4), 889–917. Berry S., 1990, Airport presence as product differentiation, American Economic Review Papers and Proceedings, 80 (2), 394–399. Berry S., M. Carnall and P. Spiller, 2006, Airline hubs: Markups and the implication of customer heterogeneity, In: D. Lee (Ed), Advances in Airline Economics, Vol.1: Competition Policy and Antitrust, Elsevier, 183–214. Bogulaski C., H. Ito and D. Lee, 2004, Entry patterns in the Southwest Airlines route system, Review of Industrial Organization, 25 (3), 317–350. Borenstein S., 1988, On the efficiency of competitive markets for operating licenses, Quarterly Journal of Economics, 103 (2), 357–385. Borenstein S., 1989, Hub and high fares: Dominance and market power in the U.S. airline industry, Rand Journal of Economics, 20 (3), 344–365. Borenstein S., 2005, U.S. Domestic Airline Pricing, 1995–2004, Competition Policy Center Work­ ing Paper 05’048, U.C. Berkeley. Brueckner J., 2002, Airport congestion when carriers have market power, American Economic Review, 92 (5), 1357–1375. De Borger B. and K. Van Dender, 2006, Prices, capacities and service levels in a congestible Bertrand duopoly, Journal of Urban Economics, 60, 264–283. de Neufville R., 1995, Management of multi-airport systems – a development strategy, Journal of Air Transport Management, 2 (2), 99–110. de Neufville R., 1999, Airport Privatization: Issues for the United States, mimeo, MIT. Evans, W. and I. Kessides, 1993, Localized market power in the U.S. airline industry, Review of Economics and Statistics, 75 (1), 66–75. FAA/OST, 1999, Airport Business Practices and Their Impact on Airline Competition, FAA/OST Task Force Study. Goolsbee, A. and C. Syverson, 2005, How do Incumbents Respond to the Threat of Entry? Evidence From Major Airlines, NBER Working Paper 11072. Harback, K.T. and J.I. Daniel, 2005, (When) Do Hub Airlines Internalize their Self-imposed Congestion Delays?, Economics Working Paper 2005–08, University of Delaware. Hartmann M., 2006, Access to airport facilities: Its impact on market competition, In: D. Lee (Ed), Advances in Airline Economics – Volume 1: Competition Policy and Antitrust, Elsevier. Harvey G., 1987, Airport choice in a multiple airport region, Transportation Research, 21A (6), 439–449.

254

JUN ISHII, SUNYOUNG JUN, AND KURT VAN DENDER

Hess S. and J.W. Polak, 2005, Mixed logit modeling of airport choice in multi-airport regions, Journal of Air Transport Management, 11 (2), 59–68. Ishii J., S. Jun and K. Van Dender, 2006, Air Travel Choices in Multi-airport Markets, UCI Econ Working Paper 05-06-22. Januszewski S., 2004, The Effect of Air Traffic Delays on Airline Prices, mimeo, University of California at San Diego. Manski C.F. and S.R. Lerman, 1977, The estimation of choice probabilities from choice based samples, Econometrica, 45 (8), 1977–88. Morrison S.A. and C. Winston, 1989, Enhancing the performance of the deregulated air trans­ portation system, Brookings Papers on Economic Activity, Microeconomics, 1989, 61–123. Morrison S.A., 2001, Actual, adjacent, and potential competition – Estimating the full effect of Southwest Airlines, Journal of Transportation Economics and Policy, 35 (2), 239–256. Oum T.H., A. Zhang and Y. Zhang, 2004, Alternative forms of economic regulation and their efficiency implications for airports, Journal of Transportation Economics and Policy, 38 (2), 217–246. Oum T.H., N. Adler and C. Yu, 2006, Privatization, corporatization, ownership forms and their effects on the performance of the world’s major airports, Journal of Air Transport Management, 12, 19–121. Pels E., P. Nijkamp and P. Rietveld, 2001, Airport and airline choice in a multi-airport region: An empirical analysis for the San Francisco Bay area, Regional Studies, 35 (1), 1–9. Pels E., P. Nijkamp and P. Rietveld, 2003, Access to and competition between airports: A case study for the San Francisco Bay area, Transportation Research, 37A, 71–83. Pels E. and E. Verhoef, 2004, The economics of airport congestion pricing, Journal of Urban Economics, 55 (2), 257–277. Reiss P.C. and P.T. Spiller, 1989, Competition and entry in small airline markets, Journal of Law and Economics, 32(2), 179–202. Starkie D., 2000, Reforming UK airport regulation, Journal of Transportation Economics and Policy, 35, 119–135. Skinner R.E. Jr., 1976, Airport choice: An empirical study, Transportation Engineering Journal, 102, 871–883. Van Dender K., 2006, Determinants of fares and operating revenues at U.S. airports, Journal of Urban Economics, doi:10.1016/j.jue.2006.09.01.

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

11 Airline Ticket Taxes and Fees in the United States and European Union∗ Joakim Karlsson† , Amedeo Odoni‡ , Célia Geslin§ , Shiro Yamanaka¶

ABSTRACT The US and the European Union (EU) are currently witnessing a vigorous debate on public funding of air transportation and the role of taxes and fees levied on airline tickets. Yet, there is practically no economic literature on taxation in the airline industry. Here, we assemble the results of an ongoing research project on airline ticket taxes and fees, for both the US and the EU. An analysis of large samples of tickets shows that the effective tax rate on US domestic tickets was approximately 16% in 2004, compared to 11% for intra–EU travel. However, the value of this comparison is limited, as air navigation services in the EU are billed directly to the airlines and are not recovered through ticket taxes. A preliminary correction for this difference results in an equivalent intra-EU tax rate of 18%. Both in the US and the EU, tax rates can vary considerably, depending for example on the level of the base fare and, in the EU, the national destination and origin. We also discuss tax

∗ This work was supported by the MIT Global Airline Industry Program, the Alfred P. Sloan Foundation, and Amadeus, S.A. The authors wish to thank many individuals at the US Bureau of Transportation Statistics, TSA, Amadeus, Lufthansa, and SAS for valuable assistance, Professor Severin Borenstein of the University of California at Berkeley for providing the US databases, Amadeus, S.A. for providing the EU databases, as well as Professors Richard England and Robert Mohr at the University of New Hampshire Whittemore School of Business and Economics for their insightful comments and suggestions. † Corresponding author. Division of Aviation, Daniel Webster College, 20 University Drive, Nashua, New

Hampshire 03063-1300; Phone: (603) 577-6428; Fax: (603) 577-6001; E-mail: [email protected].

‡ Massachusetts Institute of Technology, Room 33-219, 77 Massachusetts Avenue, Cambridge, Massachusetts

02139; Phone: (617) 253-7439; Fax: (617) 452-2996; E-mail: [email protected].

§ Department of Civil and Environmental Engineering, Massachusetts Institute of Technology. E-mail:

[email protected].

¶ Department of Civil and Environmental Engineering, Massachusetts Institute of Technology. E-mail:

[email protected].

256

JOAKIM KARLSSON et al.

incidence – the question of whether ticket taxes and fees are borne by the airlines or the passengers. We address the theory of incidence under the assumption of perfect competition, discuss the limitations of the theory, and conclude that empirical analysis is required to quantify tax incidence in the airline industry.

1 INTRODUCTION The topic of air transportation taxes has become increasingly controversial due to the many challenges currently facing the airline industry. Also a factor is a series of proposals for new taxes and fees, both in the US and the European Union (EU). The goal of this chapter is to provide descriptive statistics on ticket taxes, as well as an introductory discussion on the incidence of ticket taxes – the question of who bears the economic burden of air transportation taxation. This is motivated, in part, by the near-complete lack of descriptive statistics on ticket taxes and fees and economic literature on the incidence of these taxes. Knowledge of the level of taxation and the distribution of the tax burden also has important implications for US and EU policy makers and air carriers. The main objective of existing aviation taxes is to collect revenues to fund aviation infrastructure and security needs. There are also clear indications, especially in the EU, that future taxes are likely to be introduced to address environmental externalities, including global warming: “Airlines are accused of having a free ride in terms of air pollution because they pay no tax on the fuel they use for international flights” (The Economist, 2006, June 10, p. 67). The majority of existing taxes and fees are added to tickets, and thus represent indirect taxes collected by the airlines. Descriptive statistics are needed to provide a fundamental understanding of the size and impact of these taxes, for the benefit of both policy makers and industry professionals. This is particularly important, since tax rates reported by industry organizations, which in turn are cited by the media and policy makers, tend to be much higher than the tax rate computed from a representative sample of tickets (Burey, 2005, April 11). An analysis of the incidence of ticket taxes is needed to ensure that policy makers have an accurate set of assumptions when proposing changes to the tax structure. The implications of a tax increase may differ substantially if the passengers bear the burden of the change or if the burden falls on the airlines, and, by extension, industry investors and labor. Several ongoing developments offer further motivation for the study of taxation in the airline industry. In the US, the Federal Aviation Administration (FAA) is reviewing the federal funding system for aviation, as the Congressional mandate for the existing tax structure is set to expire in 2007 (FAA, 2005a,b). The importance of this issue is evidenced by testimony provided by the FAA, the Air Transport Association (ATA), Airports Council International – North America, the American Association of Airport Executives, the Air Carrier Association of America, the National Air Traffic Con­ trollers Association, the National Air Transportation Association, the Aircraft Owners and Pilots Association, and the National Business Aviation Association (US House of Representatives Committee on Transportation and Infrastructure, 2005). The reautho­ rization process is also shaping up as a battle between the air carrier industry and gen­ eral/business aviation over their respective tax burdens (Meckler, 2006, June 1, p. A1).

AIRLINE TICKET TAXES AND FEES

257

Finally, the political importance of the process was amplified, as it coincided with con­ tentious contract negotiations between the FAA and its air traffic controllers (Barr, 2006, May 31, p. D4). In the EU, several developments have increased the interest in aviation taxation issues. The Swedish government, for example, included a controversial new ticket tax in its 2006 budget proposal. This would have added SEK 50–100 (∼$7–$14) per ticket in order to account for the environmental impact of air transportation (Swedish Ministry of Finance, 2005). This amount represents 2–4% of the average base fare for intraEU travel originating in Sweden (Yamanaka et al., 2006, p. 49). The tax proposal was part of a “green tax shift” policy, and would have been offset by cuts in broadbased taxes such as the income tax. There is also a proposal for an EU-wide fuel tax, which is supported by environmental groups (European Federation for Transport and Environment, 2005). Finally, an ambitious proposal issued jointly by France, Germany, Spain, Brazil, and Chile calls for “a domestically applied and internationally coordinated levy on air transport travels” in order to combat global hunger and poverty in support of the United Nations Millennium Development Goals (Doland, 2005, August 29; French Ministry of Foreign Affairs, 2005). This proposal has now been approved by several countries (Schroeder, 2006, p. 3). The recent financial crisis of much of the airline industry adds another dimension to the question of taxation. The industry has developed a seemingly unstable pattern of profit and loss cycles, with losses outweighing profits (Hansman, 2005). This has forced the airlines to focus on the extent to which taxes and fees affect their net revenues. At the same time, fares have been declining, at least until very recently. This is due to increased competition, especially from low-cost carriers, and the erosion of high-value business fares. As a whole, this has resulted in taxes and fees becoming a larger proportion of the total ticket price (Karlsson et al., 2004, pp. 291–292). This trend has caused the industry to voice increasing concern over ticket taxes and fees, explaining the following strongly worded joint statement by the ATA and the Association of European Airlines (2005): Aviation taxes and fees have outpaced inflation and fares, and the taxes and fees on an airline ticket purchased either in Europe or the US are higher percentage-wise than the so-called ‘sin-taxes’ on things like alcohol and tobacco. This tax and fee burden threatens the very fiber of the air transportation industry and the economies that rely on it.

It is likely that the growing share of the total ticket price that is made up by taxes and fees has also increased passengers’ sensitivity to air transportation taxation. However, it should be kept in mind that the airlines and their passengers receive airport infrastructure and security services, as well as air traffic control services in the US, in exchange for the payment of these taxes. This chapter provides a basic overview of ticket taxes and fees in two large air transportation markets: Domestic travel inside the US, as well as travel within the so-called EU-151 member nations of the EU. Section 2 defines necessary terms and

1 The EU-15 nations are Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, The Netherlands, Portugal, Spain, Sweden, and the UK. This subset of the current member nations of the EU is used for data availability reasons.

258

JOAKIM KARLSSON et al.

describes the existing ticket tax structure in qualitative terms. Section 3 provides descrip­ tive statistics for both markets, with taxes expressed both in absolute terms and as tax rates. Section 4 summarizes economic theory on tax incidence as applied to the air transportation industry, and ends with a call for empirical research in this area. Section 5 provides concluding remarks and a summary of implications for policy makers.

2 DESCRIPTION OF TICKET TAXES AND FEES The terms “taxes” and “fees” are often used interchangeably. A tax is defined as “a compulsory levy made by public authorities for which nothing is received directly in return” (James and Nobes, 1999). Examples of pure taxes exist in aviation: Denmark levies a general-purpose transportation tax, which will, however, be phased out by 2007 (Mandsberg, 2005). The new “green shift tax” proposed by the Swedish government is also classified as a tax, although it is motivated by environmental policy. However, it can be argued that most forms of ticket taxation should rightly be referred to as “fees”, since infrastructure and services are provided in return. Passenger facility charges, in particular, are specific to individual airports and each airport must apply individually to the FAA for authority to use the resulting revenues for one or more clearly identi­ fied projects (General Accounting Office, 1999, p. 4). Here, we shall generally use the collective term “taxes and fees” without further distinction.

2.1 Ticket Taxes and Fees in the US Four types of taxes and fees are currently levied on domestic airfares in the US: the federal ticket tax (FTT), the federal flight segment tax (FST), the passenger facility charge (PFC), and the federal security service fee (FSSF). Since the FTT and FST are essentially two components of one tax, they are described together. 2.1.1 Federal Ticket and Segment Taxes The FTT and the FST are paid into the Airport and Airway Trust Fund. This fund finances congressional appropriations to cover “those obligations of the United States    which are attributable to planning, research and development, construction, or operation and maintenance of air traffic control, air navigation, communications, or supporting services for the airway system” (Internal Revenue Code, 1986). Together they accounted for $6.4 billion in 2004 (or 66% of the total revenue of the Airport and Airway Trust Fund), supporting FAA operations, facilities and equipment, and federal grants-in-aid for airports (US House of Representatives Committee on Transportation and Infrastructure, 2005). The FTT is equal to 7.5% of the base fare (the total fare less any taxes and fees). As shown in Table 1, the segment tax was $3 per flight segment in 2002 and 2003 (Internal Revenue Code, 1986). A built-in inflation adjustment raised the segment tax to $3.10 in 2004, $3.20 in 2005, and $3.30 in 2006 (ATA, 2005a). Table 1 also shows that the federal segment tax did not exist prior to October 1, 1997 (ATA, 2005a). Domestic air travel was taxed at a flat rate that peaked at 10% during

AIRLINE TICKET TAXES AND FEES

259

Table 1 History of US Infrastructure-Related Taxes and Fees on Domestic Airline Fares Year

FTT (%)

FST ($)

PFC (max.) ($)

FSSF ($)

1941 1942 1943 1955 1956 1970 1980 1982 1990 1992 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006

50 100 150 100 50 80 50 80 100 100 90 80 75 75 75 75 75 75 75 75

– – – – – – – – – – 1.00 2.00 2.25 2.50 2.75 3.00 3.00 3.10 3.20 3.30

– – – – – – – – – 3.00 3.00 3.00 3.00 3.00 4.50 4.50 4.50 4.50 4.50 4.50

– – – – – – – – – – – – – – – 2.50 2.50 2.50 2.50 2.50

Note: A dash (–) indicates the tax or fee was not applicable. Years with no changes in the tax and fee structure, rates, or levels are omitted. Federal authority to collect ticket taxes lapsed during the periods 1 January–26 August, 1996 and 1 January–6 March, 1997. The federal security fee was suspended from 1 June to 30 September, 2003. Only the maximum allow­ able PFC collection value is shown. The descriptive statistics presented in this chapter use the specific PFC collection value in effect at each airport in an itinerary.

the period 1990–1996. The FTT rate was reduced to 7.5% by 1999, in conjunction with a gradual increase of the segment tax from $1 in 1997 to $3.30 in 2006. 2.1.2 The Passenger Facility Charge The PFC was instituted as a means of assisting airports with air carrier service to “finance eligible airport-related projects, including making payments for debt service” (AIR-21, 2000). When the collection of PFCs began after June 1, 1992, airports so authorized by the FAA could charge $1, $2, or $3 per enplanement. Higher PFC levels up to $4.50 were introduced for certain airports effective April 1, 2001 (AIR-21, 2000; ATA, 2005b). PFCs are only collected for up to two boardings per each one-way trip, resulting in a maximum collection of $18 per round-trip (AIR-21, 2000). PFCs are charged by airlines at the time a ticket is purchased and are then transferred directly to the appropriate airport(s).

260

JOAKIM KARLSSON et al.

2.1.3 The Federal Security Service Fee The federal security service fee is the most recently adopted tax on US domestic airline tickets. It was created by the Aviation and Transportation Security Act (2001), and collection began on February 1, 2002. It consists of a $2.50 tax per enplanement, limited to a maximum of two segments per one-way trip (i.e., a maximum of $10 for a round-trip ticket). Congress temporarily suspended the fee from June 1 to September 30, 2003 in order to provide war-time relief to the airline industry (ATA, 2006). 2.1.4 Other Taxes A number of other federal infrastructure and security taxes and fees are assessed on air carriers. These are outside the scope of this study as they either apply only to international travel or are not directly added to the price of an airline ticket. In addition, foreign nations impose taxes and fees on US carriers engaged in international operations. These can be numerous and varied, but do not apply to domestic travel and are not covered here. Finally, air carriers also pay other infrastructure-related charges such as landing fees and airport leases, but these are not added directly to the price of an airline ticket and, thus, also fall outside the scope of this study.

2.2 Ticket Taxes and Fees in the European Union The EU ticket tax structure is considerably more complex than that of the US. Each member nation is entitled to implement its own ticket taxes and fees. Consequently, there is a much higher number of possible combinations of tax codes on an intra-EU ticket than there is on a US domestic ticket. In the EU-15 ticket sample used for this study, a total of 43 unique ticket taxes and fees were found. Moreover, the tax rates and levels often vary by boarding point, destination country, or even the choice of airline. Finally, EU carriers can collect ticket fees of their own, for example for fuel surcharges, separate from the base fare. This practice is not permitted on US domestic tickets. Another important characteristic of EU ticket taxes and fees is that carriers make separate payments to both Eurocontrol and local governments for air navigation services (ANS). In the US, the operations of the FAA are funded primarily by ticket taxes and secondarily by general government tax revenue. Finally, in the EU, there have been examples of ticket taxes where the revenue is directed to the general fund. This differs from the US, where all domestic ticket taxes are paid either into the Airport and Airway Trust Fund, directly to the airports, or to the TSA for security services.

2.3 Classification 2.3.1 Direct Versus Indirect Ticket taxes and fees are considered indirect taxes in that they are not paid directly by the passenger to the treasury, but rather are collected by the airline at the time of booking. All four US ticket taxes and fees are collected by the airlines (or their agents) at the time of purchase. Outside the US, for instance in some Latin American nations, there are examples where passengers pay a tax directly at the airport.

AIRLINE TICKET TAXES AND FEES

261

2.3.2 Unit Versus Ad Valorem A unit (or “specific”) ticket tax is a fixed monetary amount assessed to each ticket or segment. An ad valorem ticket tax consists of a percentage of the base fare. Both types are assessed on airline tickets in the US. The FTT is a 7.5% ad valorem tax. The other three US ticket taxes and fees are all unit taxes. 2.3.3 Progressive Versus Regressive Progressive taxes are those that “take an increasing proportion of income as the income rises” (James and Nobes, 1999). Regressive taxes have the opposite effect: their pro­ portion of income declines as income rises. As is demonstrated in Section 3, the use of unit taxes and fees on airline tickets generally results in effective tax rates that decline as the total fare increases (although the effective tax rate also depends on the number of connections). However, without formally establishing a link between ticket prices and income it cannot be conclusively demonstrated that airline ticket taxes are regressive. The Congressional Budget Office (CBO) has established that excise taxes are gener­ ally regressive. In 2000, total excise taxes constituted 2.2% of the bottom quintile of household income, but only 0.6% of the highest quintile (CBO, 2003, p. 25). However, while the CBO includes federal aviation taxes in its analysis, these represent a small portion of federal excise taxes. For example, federal excise taxes on tobacco and alcohol, which are known to be regressive, are double those collected on air travel (The Tax Foundation, 2006). The regressive nature of a tax does not depend on the type of tax, but rather the type of good being taxed. Also, there is evidence that excise taxes are less regressive than normally thought, when considering the lifetime burden instead of any single year of income (Poterba, 1989, pp. 325–326). While it is not the goal of this study to conclusively demonstrate that ticket taxes and fees are regressive, there is compelling evidence that air travel is not a luxury good (Adrangi and Raffiee, 2000, p. 493). The income elasticity of demand is estimated around unity (Adrangi and Raffiee, 2000; Gillen et al., 2002). If income and fares scale proportionately, this would suggest that ticket taxes are in fact regressive, since the use of unit taxes in the US ticket tax structure places a higher proportional burden on low fares.

3 THE EFFECTIVE TAX RATE IN THE US AND EU-15 We now turn to a summary of the impact of taxes and fees on the cost of airline tickets for domestic travel in the US and for intra-EU and domestic travel in the EU-15. Our principal measure of impact is the effective tax rate, i.e., the proportion of taxes and fees relative to the base fare, expressed as a percentage.

3.1 United States: Domestic Tickets The total fare for a domestic air trip consists of the sum of two parts: the base fare, BF, which is the total fare less any applicable taxes and fees, and TTF, which is the sum of the four ticket taxes and fees described in Section 2.1. TTF = FTT + FST + FSSF + PFC

(1)

JOAKIM KARLSSON et al.

262

For any set of air tickets, the effective tax rate, ETR, is defined as ETR =

ETTF EBF

(2)

where E(TTF) and E(BF) represent the expected values of TTF and BF, respectively. The expected value operator is used to emphasize that these are passenger-weighted averages. In this section, we are interested in estimating ETR for a representative sample of US domestic passengers. We used the US DOT’s Origin and Destination Data Bank 1A Ticket Dollar Value (DB1A) survey to obtain a large sample of domestic ticket records. This database provides the complete itinerary and fare information for a 10% sample of all air tickets used in the US. One shortcoming of the database is that it only records total fares. However, a relatively straightforward procedure can be used to reveal the base fare and the individual taxes and fees. This is because there are only four types of taxes and fees levied on domestic airfares in the US and three of them can be estimated precisely given the ticket itinerary. Data from the second quarters of 1993, 2002, and 2004 show that the average ETR on roundtrip tickets was 10.9, 15.9, and 16.5%, respectively (see Table 2). Although the tax rate appears to be increasing over time, the absolute amount of taxes and fees has actually decreased when expressed in constant dollars. Thus, it is valid to argue that the principal cause of the tax rate increase is the significant decrease in average base fare, which declined by $175 or 40% between 1993 and 2004. In 2Q 2004, the FTT, the only ad valorem tax for domestic air travel in the US, accounted for slightly less than half of the average total taxes and fees. The three unit taxes and fees (i.e., FST, PFC, and FSSF) vary only with the passenger’s itinerary, irrespective of the base fare. It is therefore not surprising that the average ETR decreases as the base fare increases, as shown in Figure 1. Noteworthy, however, is the fact that the ETR on the least expensive tickets, those with a base fare of $200 or less, is greater than 20%, i.e., much higher than the average ETR. With roughly 50% of all domestic tickets in 2Q 2004 having a base fare of less than $200, this provides further insight as to why there is a widespread impression that taxes and fees on airline tickets are excessively high. The impact of taxes and fees on tickets of low-cost carriers (LCC) is smaller than one might expect from the fact that the average base fare for LCC is about $185, compared Table 2 Effective Tax Rate Comparison Quarter

No. tickets in Sample

Base Total Fare ($) Taxes and Fees ($)

Federal Federal Ticket Segment Tax ($) Tax ($)

2Q 1993 2Q 2002 2Q 2004

2,164,162 3,559,912 3,893,783

44489 29174 26829

4449 2188 2012

4840 4626 4425

Note: Expressed in 2004 dollars; includes roundtrip tickets only.

– 8.35 8.06

Passenger Facility Charge ($) 391 907 958

Federal Security Service Fee ($)

ETR (%)

– 6.96 6.50

109 159 165

AIRLINE TICKET TAXES AND FEES

263

30%

ETR (2Q 04)

25% 20% Avg ETR: 16.1% 15% 10% 5% 0% $0

$200

$400

$600

$800

$1,000 $1,200 $1,400 $1,600 $1,800 $2,000

Base Fare Note: Results for 2Q 2004; includes roundtrip tickets only.

Figure 1 Distribution of ETR as a Function of Base Fare. Table 3 Legacy Versus Low-Cost Carriers (2Q 2004, Roundtrip Tickets Only) Carrier Type

Base Fare ($)

Total Taxes and Fees ($)

Federal Ticket Taxes ($)

Federal Segment Taxes ($)

Passenger Facility Charge ($)

Federal Security Service Fee ($)

ETR (%)

Low-cost Legacy

185.91 305.33

34.37 47.00

13.94 22.90

6.99 7.99

7.81 9.66

563 645

1850 1540

Note: Results for 2Q 2004; includes roundtrip tickets only.

to about $305 for legacy carriers2 , as shown in Table 3. One of the reasons is that LCC routes often bypass the most congested airports in favor of secondary ones. As the most congested airports are also the ones that tend to impose passenger facility charges, the average PFC paid by LCC passengers is considerably smaller. A second reason is the fact that the average number of segments in an itinerary from our sample on a LCC is 2.25, as opposed to 2.58 for legacy carriers. This reflects two aspects of low-cost versus legacy carrier operations: first, the average origin–destination distance flown by LCC passengers is 844 miles versus 1,186 miles for legacy carrier passengers; second, the route networks of legacy carriers rely more heavily on connections at hub airports. This means that the FST, PFC, and FSSF costs are smaller, on average, for LCC passengers than those of legacy carriers.

2 The legacy carriers in the sample are American, Continental, Delta, Northwest, United, and US Airways; the low-cost carriers are ATA, jetBlue, and Southwest. These are meant to be samples of typical legacy and low-cost carriers and not a complete set of either category.

JOAKIM KARLSSON et al.

264

Perhaps surprisingly, ETR varies little with the distance between the origin and destination in each itinerary. One of the reasons is that the average base fare increases less than linearly with the origin–destination distance. For example, the average base fare for a distance of between 1,000 and 2,000 miles is only about 35% greater than for a distance of less than 200 miles in our 2004 sample. A second reason is that longer distances are more likely to be associated with itineraries that include a connection at an intermediate airport. This, in turn, means a greater likelihood of a high FST, FSSF, and PFC.

3.2 EU-15: Intra-EU and Domestic Ticket Taxes and Fees The second major subject area of this study was air travel between and within the EU-15 member countries. This part proved far more challenging because no database comparable to the US DB1A data bank is available, publicly or otherwise, in the EU. Moreover, the air ticket taxes and fees structure in the EU-15 is complex, with 43 different primary taxes and fees and occasionally complicated provisions for their application. Our analysis thus relied on a large sample of ticket data supplied by Amadeus Global Travel Distribution, S.A. The data covered 15 days of approximately evenly spaced days spanning the period from January 16, 2004 to February 15, 2005, and contained a total of 3,032,209 intra- and domestic EU-15 tickets. Several important items should be noted concerning our EU-15 data. First, LCC tickets are not included in the ticket sample, because they are not sold through global travel distribution systems. With LCCs currently accounting for a roughly 20% market share in the EU, this means that the ETR we have computed for the EU-15 probably underestimates the true ETR, as tax rates tend to be higher for lower base fares. Second, the EU tickets frequently include two tax codes, YQ and YR (referred to here as “YQYR”), which are reserved for airline surcharges (e.g. fuel and security). To the uninformed consumer, YQYR charges are indistinguishable from the true taxes and fees that appear on a ticket, but these charges do, in fact, represent additional revenues for the airlines. In our analysis, the YQYR charges are added to the base fare, instead of being treated as taxes and fees. In other words, the effective tax rate for this sample is defined as ETR =

ETTF − YQYR EBF + YQYR

(3)

The total amount of YQYR charges observed in the EU-15 ticket sample consists of 27.8% of the total amount of taxes and fees, and is thus significantly affecting the ETR estimates. Finally, it is noted that, unlike the US, ticket taxes and fees in the EU do not cover the cost of ANS. Therefore, comparisons between the EU and US effective tax rates should be performed cautiously. Table 4 summarizes the results of the EU analysis. The overall effective tax rate for the 15-day sample of 2004 tickets was found to be 11.2%. If YQYR were treated as part of the total taxes and fees, the overall effective tax rate would be equal to 16.2%. There is no legitimate justification in treating YQYR as part of the taxes and fees, but this “apparent ETR” may be a better indicator of how

AIRLINE TICKET TAXES AND FEES

265

Table 4 Average Effective Tax Rate (15 day EU-15 sample) Ticket Category All One-way Roundtrip

No of Tickets in Sample

Base Fare (US$)

TTF (US$)

YQYR (US$)

ETR (%)

3,032,209 691,841 2,340,368

272.27 166.74 303.47

44.19 19.12 51.60

1226 450 1456

112 85 116

the size of taxes and fees is perceived by air transportation consumers. Remarkably, the average YQYR charged to air tickets more than doubled over the sample period, from an average of $8 per ticket in January 2004 to $19.6 in February 2005. This trend, if it continues, has the potential of badly distorting public perceptions regarding the impact of taxes and fees on air travel costs. The aggregated results, as presented above, mask the fact that there is great variability in the ETR across the EU-15 countries. This is demonstrated in Table 5, where the ETR ranges from 20.4% for the UK to 6.9% for Spain and 6.3% for Luxembourg. Note that a high tax rate can result from high taxes or from low fares. As pointed out in Section 3.2, one major difference between the US and European add-on taxes and fees is that the former fund ANS cost, while the latter do not. Any comparison of the US and EU-15 ETR values must take this difference into consideration, because ANS costs are substantial on both sides of the Atlantic. Unfortunately, there is no European data collection mechanism comparable to the US carriers’ monthly DOT Form 41 filings, which can readily identify the amounts paid by individual airlines for ANS.

Table 5 Effective Tax Rate by Origin Country Origin Country

No. of Tickets in Sample

Base Fare (US$)

TTF (US$)

YQYR (US$)

ETR (%)

UK Greece Denmark Ireland Finland Sweden The Netherlands Belgium Austria France Portugal Germany Italy Spain Luxembourg

277268 25550 81951 10881 117728 138691 30 690 39683 43020 616651 20850 537799 127859 953583 10005

19101 19957 28224 20617 30165 29738 36476 34926 39444 32957 26384 34390 28988 19806 37577

5618 4659 7288 4473 5824 6689 6242 5726 7647 5606 3382 5991 3977 1481 4410

1430 968 2125 1301 1278 2171 1262 1226 2657 1547 447 2215 1351 115 1916

204 176 170 145 145 142 132 124 119 118 109 103 87 69 63

JOAKIM KARLSSON et al.

266

Table 6 Average Effective Tax Rate for Different Base Fare Ranges in the EU-15 Nations BF Range (US$)

No. Tickets in Sample

% of Tickets

BF ($)

TTF ($)

YQYR ($)

ETR (%)

Segments per Ticket

BF ≤ 100 100 < BF ≤ 250 250 < BF ≤ 500 500 < BF ≤ 1000 1000 < BF ≤ 2000 2000 < BF

738,217 1,104,774 800,166 300,704 85,631 2,717

24 36 26 10 3 0

5657 16651 35581 66496 125987 269272

3026 4193 5094 6138 6871 8253

760 1126 1434 1878 2277 2652

353 173 99 62 36 21

164 194 214 224 255 338

All

3,032,209

100

27227

4419

1226

112

197

Note: Includes one-way and round-trip tickets.

The analysis also looked at how the EU-15 ETR varies with base fare. Table 6 shows that taxes and fees have a much larger impact on low-fare tickets, as ETR values vary between 2.1% and 35.3% for different ranges of the base fare. This is because the total taxes and fees only increased by a factor of three between the lowest and highest base fare range, whereas the base fare increased by a factor of 48. As in the US, passengers with a base fare of less than $250 (60% of the passengers) face a significantly higher ETR than the overall mean of 11.2%. In fact, the impact on less expensive tickets is even stronger than in the US due to the fact that the great majority of EU ticket taxes and fees are unit taxes, not ad valorem.

3.3 A Preliminary Comparison Between US and EU-15 Effective Tax Rates The study team has obtained detailed data from two of the largest European airlines, Lufthansa and SAS Group, regarding their ANS costs in 2004 for intra-EU flights. This information – independently provided by the two carriers and mutually consistent – indicates that ANS costs would add roughly 7% to the effective tax rate estimated in Section 3.2. Both Lufthansa and SAS have flights throughout Europe and are based in countries (with the exception of Denmark) where the effective tax rate is close to the EU-15 average of 11.2%. This suggests the conjecture that, with ANS costs taken into account, the effective tax rate in the EU-15 would be approximately 18–19% (i.e., 11.2% + 7%). This would be slightly higher than the 16.1% for domestic air travel in the US in the second quarter of 2004. However, this is a very preliminary estimate, which serves primarily as a good launching point for future investigation. For example, this comparison does not take into account that a portion of US air traffic control costs are subsidized by general tax revenues.

4 INCIDENCE OF TICKET TAXES AND FEES Who carries the burden of the ticket taxes and fees – the passengers or the airlines? That is the question of tax incidence. We hasten to note that economists consider the

AIRLINE TICKET TAXES AND FEES

267

tax burden to be carried exclusively by individuals, not firms (Gruber, 2005; Rosen, 2005). Thus, when we speak of the airlines carrying a portion of the tax burden, we are actually referring to individual airline employees and investors. This distinction becomes increasingly important when considering the impact of air transportation taxes on other markets, but for the sake of simplicity we will use the term “airlines” when discussing the portion of the burden not borne by passengers. The purpose of our discussion here is to present the economic theory of tax incidence as applied to air transportation, outline the limits of a theoretical approach, and lay out a framework for future empirical research to study incidence.

4.1 Statutory versus Economic Incidence When discussing the burden of taxation, a distinction must be made between statutory and economic incidence. The statutory incidence of ticket taxes defines who is legally obligated to collect the tax and pay it to the Treasury. In the US, the statutory incidence is clearly on the airline industry (Internal Revenue Code, 1986; AIR-21, 2000; Aviation and Transportation Security Act, 2001). However, the statutory incidence is not the same as the economic incidence, which determines who ultimately bears the burden of the tax (Fullerton and Rogers, 1993, p. 1). Our focus is on economic incidence – the question of whose income or welfare is affected.

4.2 Competition in the Airline Industry We begin our discussion of incidence under a set of simplifying assumptions, which we will later relax. There are two theoretical extremes which considerably simplify incidence theory: perfect competition and pure monopoly. In practice, most industries in the US operate in the range between these two extremes, and the airline industry is no exception. A recent study of airline competition in the US concludes that while fares have declined over the last 10 years, there remain a number of “fortress hubs” where there is a mark-up resulting from lack of competition (Borenstein, 2005). Empirical measures support the existence of oligopolies in the airline industry. The so-called Rosse–Panzar test has been used to rule out that airlines fall either in the category of perfect competition or monopolistic competition (Fischer and Kamerschen, 2003).

4.3 Incidence Under Perfect Competition As a starting point for our incidence discussion, we assume perfect competition with flexible prices. Despite the evidence for oligopolistic behavior, the assumption of a perfect competition is a reasonable starting point when discussing the market as a whole, since the oligopolistic characteristics are driven by spatial differentiation. Under this assumption, tax incidence can be theoretically derived. A second simplifying assumption is that we consider a partial equilibrium analysis, in which the airline industry is treated independently of the rest of the economy. We also assume that there is no asymmetric information between airlines or between airlines and passengers. Finally, we consider only the case of a single unit tax, t.

JOAKIM KARLSSON et al.

268

If t = 0, there is no tax, and the fare level is determined by market forces at some equilibrium level P ∗ . After the application of the tax, the passenger faces the after tax price P d , defined by the demand curve, and the supplier receives P d − t.3 We define the incidence or “pass-through fraction” from the perspective of the passenger: i=

Pd − P∗ P∗

(4)

Hence, the passenger’s portion of the tax burden is i · t and the airline’s portion is 1 − i · t. Under these assumptions, the incidence is determined by the relative levels of the price elasticities of demand and supply. Agents with relatively inelastic responses to price changes carry more of the tax burden, whereas those with elastic responses escape it. If either the passenger or the airline responds perfectly inelastically to a change in price, then that agent bears the full burden of the tax (if both do, the result is indeterminate). For small changes in the vicinity of P ∗ , the incidence is given by (Pindyck and Rubinfeld, 2001, p. 317) i=

sqp dP d = s dt qp − dqp

(5)

Here, we use the convention that price elasticity of demand is reported as a negative number, leading to the condition 0 ≤ i ≤ 1. In other words, under perfect competition, neither the passenger nor the airline can bear more than 100% of the tax burden. Several difficulties arise in trying to apply this result: The first is that price elasticities are likely to vary along several dimensions, including type of travel (i.e., leisure vs. business) and trip length (Gillen et al., 2002). Second, short-run price elasticities are likely to differ from those in the long run. One estimate, although out of date, is that the price elasticity of demand for air travel ranges from −0.1 in the short run to −2.4 in the long run (Gwartney and Stroup, 1997). On the supply side, we expect less elastic behavior in the short run, as airlines’ capital investments in aircraft and airport facilities cannot be changed quickly. Third, because the airline industry has undergone structural changes in recent years, including a multi-year demand shock caused by the events of September 11 (Ito and Lee, 2005), we want to limit ourselves to the most recent estimates of price elasticities. On the demand side, we have large numbers of agents (i.e., passengers) acting essen­ tially as price takers and without consideration of the actions of other consumers, so that the demand curve is clearly defined. On the supply side, the agents (i.e., air car­ riers) form a much smaller group, and they are more likely to react to the behavior of competitors and exhibit various forms of price leadership: in the absence of perfect competition, the notion of a supply curve ceases to exist. For this reason, there are large numbers of empirical assessments of price elasticity of demand, including several recent meta-studies to derive numerical estimates, but practically no estimates of the price elasticity of supply. 3 Note that what is referred to as “base fare” in this chapter is actually an estimate of P d −t, not the equilibrium price P ∗ .

AIRLINE TICKET TAXES AND FEES

269

Table 7 Hypothetical Values of Tax Incidence (on Passengers) Under Perfect Competition

sqp = 074

dqp = −070

dqp = −115

dqp = −150

0.51

0.39

0.33

A MITRE Corporation study estimates price elasticity of demand ranging from −0.56 to −1.82, depending on origin and destination distance (Bhadra, 2002). Two large meta­ analyses have similar results: A study for the Canadian Department of Finance resulted in values ranging from −0.70 to −1.52 for domestic travel (Gillen et al., 2002). Another meta-analysis of 204 studies found a mean elasticity of −1.15 with a standard deviation of 0.62 (Brons et al., 2002). On the basis of these values, we adopt the following range of values for the long-run price elasticity of demand: −0.70, −1.15, and −1.50. On the supply side, Ito and Lee (2005) report an implied price elasticity of supply of 0.74, estimated over a time period of several years. Putting everything together, we can now compute some hypothetical values for i in the long run, under the assumption of perfect competition (Table 7). The value of this analysis is quite limited given the lack of empirical estimates and the assumption of perfect competition. Under different assumptions, the conclusions could change, as discussed below.

4.4 Incidence Under Imperfect Competition Unlike perfect competition, models of imperfect competition (and pure monopolies) allow for the possibility of overshifting (Anderson et al., 2001, pp. 7–12), since prices are set above marginal cost (Delipalla and O’Donnell, 2001, p. 891). In the overshifting scenario, the total fare paid by the passenger after a tax increase is higher than the sum of the original fare and the tax increase. In some, but not all, cases of overshifting, a tax increase can result in higher profits. As shown above, overshifting cannot occur under perfect competition, but “once imperfectly competitive markets are allowed, overshifting becomes a possibility and can be guaranteed in some model specifications” (Fullerton and Metcalf, 2002, p. 1825). Additionally, under imperfect competition the economic effects of an ad valorem tax can be different than those of a unit tax. For example, under certain assumptions, an ad valorem tax can lead to firms exiting the market, which reduces the burden on producers and increases the burden on consumers: “While a change in the excise tax does not affect the equilibrium number of firms, a change in the ad valorem tax does    Ad valorem tax incidence can be decomposed into two components: a direct effect and an indirect effect through the change in the equilibrium number of firms” (Fullerton and Metcalf, 2002, p. 1831). In industries with differentiated products, non-price competition opens up additional pathways in which taxes can manifest their impacts. For example, in addition to direct impacts on fare levels, changes in ticket taxes can affect product quality and variety: “Non-price competition can substantially affect the degree to which output taxes are

270

JOAKIM KARLSSON et al.

passed forward to consumers and can lead to counterintuitive results” (Fullerton and Metcalf, 2002, p. 1832). It is therefore possible that the introduction of a new tax is entirely absorbed by the airlines, or even that fares go down after a tax hike, because of strategizing within an oligopoly of airlines. Under theoretical constructs such as perfect competition and pure monopoly, analytical derivation of tax incidence is relatively straightforward. Oligopolies, however, include a range of possible strategic behaviors with varying consequences for the incidence question. These behaviors are usually modeled by selecting one of a few archetypical models of oligopolistic behavior. Tax incidence can be analytically derived for some of these oligopoly models (see, e.g., Barron et al., 2004; Hamilton, 1999). However, this requires knowing which model is applicable to the market under consideration. Empirical analysis demonstrates that imperfect competition exists in the airline indus­ try, but the evidence is inconclusive as to which specific model best applies (Fischer and Kamerschen, 2003, p. 91). There is mixed support in the literature for both the Cournot and Bertrand models, as evidenced for example by Bilotkach’s (2005) analysis of international markets. In an assessment of the presence of Bertrand, Cournot, or cartel behavior, Brander and Zhang find that the Cournot model is the most consistent with the data (1990). However, this analysis is limited to 33 duopoly airline routes out of Chicago. The lack of consistent evidence for a specific model of oligopolis­ tic behavior makes it difficult to proceed much further with a theoretical incidence discussion under imperfect competition, which strengthens the case for an empirical approach.

4.5 Recommendations for Future Work While the theoretical approach adopted in this chapter provides a useful framework for discussing the incidence of ticket taxes and fees, it is ultimately quite limited in its ability to produce numerical results. This is in no small part due to the theoretical challenges of analyzing incidence within a complex oligopoly. The next logical step is to conduct an econometric study of incidence. A good starting point would be a partial equilibrium analysis using US ticket data, which are more complete than data available for the EU. The goal would be to estimate the impact of variations in ticket taxes and fees on the consumer price. There is no natural experiment which would allow the researcher to isolate the effect of ticket tax changes. One possible approach is to investigate the impact of historical changes in the US ticket tax structure on total fares, while controlling for other factors. As indicated by Table 1, major tax changes within the last two decades include the introduction of the PFC in 1992, the change from a 10% FTT to a 7.5% tax combined with a segment fee introduced in the 1997–1999 period, the increase in the maximum PFC from $3 to $4.50 in 2001, and the introduction of the $2.50 federal security service fee in 2002. Alternatively, one could study the impact of lapses in tax collection authority or approved tax breaks. Lapses in tax collection authority include an eight-month period starting January 1, 1996 and a two-month period starting January 1, 1997 (Fischer, 1998). The most substantial approved tax break was the suspension of the federal security fee from June 1 to September 30, 2003.

AIRLINE TICKET TAXES AND FEES

271

One of the challenges associated with an econometric analysis of tax incidence is to develop a model that controls for all other exogenous variables which cause fares to change over time. In order to better isolate such factors, the problem can be narrowed down further. For example, the 1997–1999 change from a 10% ad valorem tax to a combination of a 7.5% ad valorem taxes and segment taxes affected different types of tickets in different ways. Expensive, nonstop tickets typically associated with business travel faced a lower tax whereas low-cost multi-stop tickets faced a higher tax. Comparing these pools of tickets might clarify the impact of the change in the tax structure, while isolating other effects.

5 CONCLUSIONS Because of the structure of airline ticket taxes and fees, there is considerable variation in effective tax rates. For any one ticket, the tax rate depends on the base fare, itinerary, and national tax structure(s) in place. For domestic travel in the continental US, the average tax rate was approximately 16% in 2004. This represents a significant increase since 1993, but this is due to the decline in base fares, not an increase in taxes. In the EU, the tax rate was approximately 11% for domestic and intra-EU travel for the 15 day sample beginning January 16, 2004 and ending February 15, 2005. However, the EU estimate does not include low-cost and charter carriers, and does not reflect fees paid for ANS. National variations within the EU are significant in our sample, ranging from an average low of 6% to a high of 20%. What do our findings mean for policy makers? First and foremost, we provide an understanding of the current and historical levels of taxation. These descriptive statistics are based on very large samples of actual tickets. This differs from the common approach of computing tax examples using a “typical” ticket, which may not be truly representative of actual ticket purchases. We also emphasize how taxes impact travelers, air carriers, and nations differently. Second, the use of ad valorem taxes and the observed large variations in taxation likely mean that taxes and fees do not reflect the per passenger costs of air transportation infrastructure and security services. This is an important factor in the ongoing debate on the reauthorization of the US Airport and Airway Trust Fund. Decision makers and other participants in this debate need an accurate picture of the level of air transportation taxation, which this chapter attempts to provide. In the EU, the finding that airlines’ use of add-on fees has risen dramatically, as reported in this chapter, has led to public reaction. This includes a proposal by the European Commission Directorate-General for Energy and Transport that would require airlines to “publicize full fares, including taxes, charges and booking and credit card fees on their websites and in adverts” (Gow, 2006, July 18). We also wish to highlight that there is a need for improved collection of data: In the US, there is a large database of ticket records, but it only includes total fare, not individual taxes and fees. In the EU, no equivalent database exists at all. The quality of empirical analyses of air transportation taxation is limited by this lack of data. Finally, our discussion on incidence serves as a reminder that the impacts of air transportation taxation policy may differ based on who carries the tax burden. This chapter also seeks to emphasize that statutory procedures in place for tax collection

272

JOAKIM KARLSSON et al.

have no impact on the economic burden. Presumably, decision makers approach air transportation tax policy with implicit assumptions about who bears the burden. Under the assumptions of perfect competition, the tax burden is shared between passengers and airlines. However, this theoretical result has only limited practical application: An empirical analysis of tax incidence is required to answer the question of how the tax burden is distributed. To date, no such study has been published – this topic is ripe for additional research. Empirical results on tax incidence should help to shape the policy debate both in the US and the EU.

REFERENCES Adrangi, B. and Raffiee, K. (2000). New evidence on fare and income elasticity of the U.S. Airline Industry. Atlantic Economic Journal, 28: 493. Air Transport Association (2005a). Domestic ticket taxes. Retrieved May 11, 2006, from http://www.airlines.org/econ/d.aspx?nid=5382 Air Transport Association (2005b). Passenger facility charges. Retrieved December 2, 2005, from http://www.airlines.org/econ/d.aspx?nid=1277 Air Transport Association (2006). U.S. (or U.S.-approved) aviation excise taxes and user fees. Retrieved May 11, 2006, from http://www.airlines.org/econ/d.aspx?nid=4919 Air Transport Association and Association of European Airlines (2005). ATA and AEA announce joint action on common aviation challenges. Retrieved December 29, 2005, from http://www.aea.be/aeawebsite/DataFiles/Pr05-015.pdf Anderson, S., de Palma, A., and Kreider, B. (2001). Tax incidence in differentiated product oligopoly. Journal of Public Economics 81: 173–192. Aviation and Transportation Security Act. 49 U.S.C. § 44940 (2001). Barr, S. (2006, May 31). A push in congress to act on air traffic controllers impasse. The Washington Post, p. D4. Barron, J., Blanchard, K. H., and Umbeck, J. (2004). An economic analysis of a change in an excise tax. Journal of Economic Education, 35: 184–196. Bhadra, D. (2002). Demand for air travel in the United States: Bottom-up econometric estimation and implications for forecasts by origin–destination pairs. AIAA’s Aircraft Technology, Integra­ tion, and Operations (ATIO) 2002 Technical Forum. Washington, DC: American Institute of Aeronautics and Astronautics. Bilotkach, V. (2005). Price competition between international airline alliances. Journal of Trans­ port Economics and Policy, 39: 167–189. Borenstein, S. (2005). U.S. domestic airline pricing, 1995–2004 (Competition Policy Center Working Paper CPC05–48). University of California at Berkeley. Brander, J. A., and Zhang, A. (1990). Market conduct in the airline industry: An empirical investigation. Rand Journal of Economics, 21: 567–583. Brons, M., Pels, E., Nijkamp, P., and Rietvald, P. (2002). Price elasticities of demand for passenger air travel: A meta-analysis.” Journal of Air Transport Management, 8: 165–175. Burey, J. (2005, April 11). Taxes, fees resurrect debate over privatizing air traffic control. Airline Business Report, pp. 1–2, 4. Congressional Budget Office (2003). Effective federal tax rates: 1997–2000. Washington, DC: Government Printing Office. Delipalla, S. and O’Donnell, O. (2001). Estimating tax incidence, market power and market conduct: The European cigarette industry. International Journal of Industrial Organization, 19: 885–908.

AIRLINE TICKET TAXES AND FEES

273

Doland, A. (2005, August 29). France to help finance fight on poverty. Retrieved September 23, 2005, from http://www.washingtonpost.com/wp-dyn/content/article/2005/08/29/ AR2005082901001.html European Federation for Transport and Environment (2005). Ten reasons why an aviation fuel tax is good for European citizens. Retrieved December 28, 2005, from http://www. te.nu/docs/Press/2005/2005-02-16_aviation_fuel_tax_ten_reasons.pdf Federal Aviation Administration (2005a). Questions on future funding of the air traffic con­ trol system, other aviation system components, and related issues. Retrieved December 28, 2005, from http://www.faa.gov/about/office_org/headquarters_offices/aep/aatf/media/Questions for Stakeholders.pdf Federal Aviation Administration (2005b). Trust fund taxes set to expire in 2007. Retrieved Decem­ ber 28, 2005, from http://www.faa.gov/news/testimony/testimony/2005/Trust_Fund.pdf Fischer, J. (1998). Transportation trust funds: budgetary treatment (Congressional Research Service Report 98–63). Retrieved January 3, 2006, from http://ncseonline.org/NLE/CRS/ abstract.cfm?NLEid=925 Fischer, T. and Kamerschen, D. (2003). Measuring competition in the U.S. airline industry using the Rosse–Panzar test and cross-sectional regression analyses. Journal of Applied Economics, 6: 73–93. French Ministry of Foreign Affairs (2005). Joint statement by Brazil, Chile, France, Germany and Spain. Retrieved December 28, 2005, from http://www.diplomatie.gouv.fr/ actual/pdf/050211Brasilia_eng.pdf Fullerton, D. and Metcalf, G. Tax incidence. In Auerbach, A. and Feldstein, M. (Eds.). (2002). Handbook of public economics (Vol. 4). Amsterdam: Elsevier. Fullerton, D. and Rogers, D. L. (1993). Who bears the lifetime tax burden? Washington, DC: Brookings Institution. General Accounting Office (1999). Passenger facility charges: Program implementation and the potential effects of proposed changes (GAO/RCED-99–138). Washington, DC: Government Printing Office. Gillen, D. W., Morrison, W. G., and Stewart, C. (2002). Air travel demand elasticities: Concepts, issues and measurement. Ottawa: Department of Finance Canada. Gow, D. (2006, July 18). EU proposes crackdown on airlines’ hidden charges. Retrieved July 18, 2006, from http://travel.guardian.co.uk/news/story/0,1822927,00.html Gruber, J. (2005). Public finance and public policy. New York: Worth Publishers. Gwartney, J. and Stroup, R. (1997). Economics: Private and public choice. Cincinnati: SouthWestern College Pub. Hamilton, S. (1999). Tax incidence under oligopoly: A comparison of policy approaches. Journal of Public Economics, 71: 233–245. Hansman, J. R. (2005). Airline industry recent trend update. Presentation to the MIT Global Air­ line Industry Program Advisory Board/Airline Industry Consortium Joint Meeting. Cambridge: Massachusetts Institute of Technology. Internal Revenue Code. 26 U.S.C. § 4261 (1986). Ito, H. and Lee, D. (2005). Assessing the impact of the September 11 terrorist attacks on U.S. airline demand. Journal of Economics and Business, 57: 75–95. James, S. and Nobes, C. (1999). Economics of taxation: Principles, policy, and practice (7th ed.). New York: Prentice Hall. Karlsson, J., Odoni, A., and Yamanaka, S. (2004). The impact of infrastructure-related taxes and fees on domestic airline fares in the U.S. Journal of Air Transport Management, 10: 285–293. Mandsberg, R. (2005, December 16). Passagerafgift på flybilletter fjernes [Passenger fee on airline tickets are being removed]. Retrieved December 29, 2005, from http://www.dr.dk/ Regioner/Nord/Nyheder/Politik/2005/12/16/155521.htm

274

JOAKIM KARLSSON et al.

Meckler, L. (2006, June 1). Collision course: Why big airlines are starting a fight with business jets. The Wall Street Journal, p. A1. Poterba, J. (1989). Lifetime incidence and the distributional burden of excise taxes. The American Economic Review, 79: 325–330. Pindyck, R. and Rubinfeld, D. (2001). Microeconomics. Upper Saddle River, NJ: Prentice Hall. Rosen, H. (2005). Public finance (7th ed.). New York: McGraw-Hill/Irwin. Schroeder, F. (2006). Innovative sources of finances after the Paris Conference: The concept is gaining currency but major challenges remain. New York: Friedrich Ebert Stiftung. The sky’s the limit. (2006, June 10). The Economist, p. 67. Swedish Ministry of Finance (2005). The budget for 2006: Investing in new jobs, growth and welfare. Retrieved December 28, 2005, from http://www.sweden.gov.se/sb/d/5932/a/50219 The Tax Foundation (2006). Federal excise tax collections: Fiscal years 1940–2006. Retrieved June 6, 2006, from http://www.taxfoundation.org/taxdata/show/240.html# federa­ lexcisecollections-20060428 U.S. House of Representatives Committee on Transportation and Infrastructure (2005). Subcom­ mittee on Aviation hearing on financial condition of the aviation trust fund: Are reforms needed? Retrieved December 28, 2005, from http://www.house.gov/transportation/aviation/05-04-05/ 05-04-05memo.html Wendell H. Ford Aviation Investment and Reform Act for the 21st Century [AIR–21]. 49 U.S.C. § 40117 (2000). Yamanaka, S., Karlsson, J., and Odoni, A. (2006). Aviation infrastructure taxes and fees in the United States and the European Union. Airlines, Airports, and Airspace: Economic and Infrastructure Analysis (Transportation Research Record No. 1951), 44–51.

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

12 Are Passengers Willing to Pay More for Additional Legroom?∗ Darin Lee† , María José Luengo-Prado‡

ABSTRACT This chapter investigates whether or not the efforts by two of the largest US airlines to increase seat pitch (i.e., legroom) across their aircraft fleet during 2000 resulted in fare premia relative to the other “full-service carriers.” Using panel data from 1998 to 2002, we estimate fixed-effects regressions in markets with overlapping service between large hub and spoke carriers and find that United’s “Premium Economy” program was more successful than American’s “More Room Throughout Coach” program at generating fare premia.

1 INTRODUCTION The rapid expansion of low-cost carriers and recent bankruptcy filings by both United and US Airways has focused most of the recent attention regarding airline costs and service quality on the differences between the low-cost (i.e., Southwest and JetBlue) and “full-service” (i.e., American and Delta) carriers.1 In contrast, relatively little attention has been paid to the differences in service quality among carriers within either of these two groups. One area of service quality competition that has received some

§

The authors thank Dan Kasper, Nicholas Rupp and Todd Schatzki for helpful comments and Phoenix Kalen for valuable research assistance. The views expressed in this paper are those of the authors and do not reflect those of LECG, LLC. All errors remain ours alone. ∗ Reprinted from Journal of Air Transport Management, Vol 10(6), Darin Lee and María José Luengo-Prado, Are passengers willing to pay more for additional legroom?, pp. 377–383, Copyright (2004), with permission from Elsevier. † LECG, LLC, 350 Massachusetts Ave. Suite 300. E-mail: [email protected] ‡ Department of Economics, Northeastern University, 301 Lake Hall, Boston, MA 02115-5000, USA.

E-mail: [email protected].

1 See, for example, “The Airlines’ New Deal; Its Not Enough,” Fortune, April 28, 2003, and “How to Fix the

Airlines,” BusinessWeek, April 14, 2003.

276

DARIN LEE AND MARíA JOSÉ LUENGO-PRADO

recent attention is flight cancellations and delays (Mazzeo, 2003). Rather than explicitly attempting to link prices and service quality, this literature has focused primarily on the relationship between service quality and market concentration.2 With regards to in-flight service quality, the literature has typically assumed that full-service carriers are, for the most part, fairly homogeneous. In any given city or airport-pair market, there are a number of factors that may account for differences in average fares across full-service airlines. Numerous studies (Borenstein, 1989; Evans and Kessides, 1993) have attempted to identify and assess the degree to which factors, such as market share (both in the market and at the endpoint airports), network size, and the number of destinations served by a carrier from the endpoint airports impact a carrier’s cost, potential market power and/or service quality – and hence its relative fares – in a given market. A unique change in relative service quality that occurred during 2000 among the “full-service” carriers in the US airline industry is examined. In particular, two of the largest US carriers – United and American – reconfigured their aircraft fleet to provide additional seat pitch (i.e., legroom) in their coach class cabins.3 By literally removing seats from their aircraft, these two carriers reduced the seating capacity of their aircraft, improving in-flight service quality, but at the same time, increasing unit operating costs. Therefore, an implicit assumption made by both American and United was that passengers would be willing to pay a premium for what was deemed to be a higher quality service offering. The purpose of this chapter is to test the assumption that passengers are willing to pay more for one particular aspect of service quality – additional seat pitch.

2 SERVICE QUALITY COMPETITION IN THE AIRLINE INDUSTRY Since the deregulation of the US airline industry in 1978, service quality competition, broadly defined, has evolved along two main lines: low-cost carriers (LCCs), such as Southwest and JetBlue, and “full-service” carriers, such as American, Delta, and United. Low-cost carriers primarily serve the most heavily traveled routes and are known for their simple, “no frills” in-flight service and lower average fares (Dresner et al., 1996; Morrison, 2001). Full-service carriers, on the other hand, differentiate themselves from LCCs by offering a number of service characteristics typically unavailable from LCCs, such as extensive national and international route networks, preassigned seats, some degree of in-flight meal service on longer flights, multiple service/cabin classes, and comprehensive frequent flyer programs that permit passengers to earn and redeem miles on a wide range of domestic and international partners (both airline and nonairline).

2 Rupp et al. (2005) provide a closer link between pricing and delay/cancellation probability (their measure of

service quality) in their examination of schedule recoveries following airport closures due to security breaches

following 11 September.

3 TWA experimented briefly with expanded coach class seating in 1993 after it re-emerged from bankruptcy,

but quickly discontinued its “Comfort Class” after it failed to generate price premiums.

Source: “New Coach Seating Configurations – Comfort Class Deja Vu All Over Again?,” Plane Business,

12 February 2000.

ARE PASSENGERS WILLING TO PAY MORE FOR ADDITIONAL LEGROOM?

277

While most passengers can readily distinguish between the service quality of low-cost versus full-service carriers, many travelers would be hard-pressed to identify significant differences in service quality on comparable flights among the competing full-service carriers. Thus, competition among the full-service carriers in markets where their net­ works overlap has typically been in the form of price, flight frequency, and schedule competition (Ross, 1997; Borenstein and Netz). Full-service carriers have also attempted to differentiate themselves by comparing their on-time performance or even their amount of overhead luggage space. Carriers can also compete along less quantifiable service quality dimensions, such as crew friendliness. During 2000, two of the largest full-service carriers in the US, American and United, engaged in an overt (and heavily marketed) form of in-flight service-quality competition by reconfiguring their aircraft fleets to increase the “seat pitch” in their coach class cabins. Seat pitch refers to the horizontal distance between the same part (i.e., front) of two seats in consecutive rows of an aircraft, and thus, greater seat pitch should be weakly preferred by passengers (all other things equal) to less seat pitch. Prior to these changes, each of the large network carriers offered industry-standard seat pitches of 31–32 in. While American and United were the only two full-service carriers to increase seat pitch in their coach class cabins, they adopted very different approaches.4 American’s program, referred to as More Room Throughout Coach increased the seat pitch for all coach class seats across its entire aircraft fleet to between 33 and 35 in. In contrast, United’s “Economy Plus” class increased seat pitch to an industry-leading 36 in., but the increased pitch was limited to the first 6–11 rows of the coach class cabin depending on aircraft type. Thus, while all coach class passengers on American experienced More Room Throughout Coach starting in 2001, only a subset of coach class passengers (in general, high-yielding business passengers) received extra legroom on United. In particular, United’s Economy Plus seats are typically reserved for their top-tier frequent flyers or passengers purchasing full fare or only moderately discounted (Y, B or M class) coach tickets.5 While increased seat pitch – all things equal – would likely please most passengers, it is both costly and risky for an airline to provide, given the competitive nature of the industry. Since increased seat pitch necessarily reduces the number of seats per flight and most operating costs remain constant regardless of the number of seats, a carrier that increases its seat pitch also increases its unit operating costs. (Fewer seats would also result in less weight, which in turn would reduce fuel costs, however, this impact is likely to be negligible.) These higher unit costs can potentially be overcome if passengers value extra seat pitch enough to pay a premium for it.6 There is no guarantee, however, that passengers – even if they are aware of the difference in seat pitch – are willing to pay a fare premium relative to other carriers for this added element of in-flight service quality. 4

JetBlue recently announced it too would increase seat pitch in roughly two-thirds of its seats to 34 in. See “JetBlue Adds More Legroom Across Fleet,” company press release, 13 November 2003. 5 Source: www.ual.com. 6 It is also possible that the higher unit costs could be overcome by increasing load factors. However, since both carriers’ programs reduced seating capacity at a time when load factors were high by historical standards (e.g., American’s domestic load factor in 2000 was 70.4% vs. an average of 62.6% during 1990–1999), it is likely that the carriers’ primary goal was to attract higher paying passengers.

278

DARIN LEE AND MARíA JOSÉ LUENGO-PRADO

3 MODEL AND DATA Our analytical approach is to estimate fixed-effects price equations using a cross-section of airport-pair markets prior to and following the changes in seat pitch. Controlling for factors that are known to impact relative fares and assuming that no other changes in relative service quality occurred over the same period, comparing the difference between coefficients on carrier dummy variables prior to and following the changes should allow us to determine what impact – if any – the change in seat pitch had. Since our goal is to identify the impact of changing one particular element of service quality, it is important that we control for other service quality factors as much as possible. We attempt to do this in two main ways. First, a data set is built using a sample of passengers purchasing as close substitutes as possible. Second, a number of independent variables are included that regarded to impact a carrier’s relative service quality and hence, its relative fares in a particular airline market. The data for the analysis are from the US Department of Transportation’s OD1A database, a 10% sample of all domestic Origin and Destination (O&D) passengers traveling on US scheduled carriers. Only passengers traveling on the full-service airlines, commonly referred to as the “Big Six” carriers: American, United, Delta, Northwest, Continental, and US Airways, are considered. Many studies of the airline industry (i.e., Brueckner and Whalen, 2000) have shown that passengers are typically willing to pay more for nonstop versus connecting service. Likewise, it has been well documented (i.e., Morrison and Winston, 1995) that one-way tickets are priced higher than roundtrip tickets and that fares on routes to and from hubs differ from those which neither originate nor terminate at a major carrier’s hub. To control for any potential price premia associated with these factors, the data are restricted to passengers who, purchased a round-trip coach class ticket, neither originated nor terminated at any of the Big Six carriers’ hub airports,7 and traveled on a one-stop itinerary. Moreover, to control for any price differences that may result from either cost or willingness to pay differences associated with “mainline” versus regional/commuter aircraft, the data is further limited to include only those itineraries in which passengers flew exclusively on large jet (i.e., mainline) aircraft. Both American and United implemented changes in their seating configuration throughout 2000, and the data are constructed as a 5-year panel using annual data for the years 1998–2002.8 Annual rather than quarterly data is used to avoid fluctuations in the data caused by short-term labor disruptions or price wars. Based on this subsample of the raw data, a directional airport-pair market is constructed that considers only those markets where either American or United and at least one other Big Six carrier (i.e., Delta) each served 500 or more passengers during each of the 5 years of our sample. Airport-pairs are used

7 The hubs we include for each carrier are American (DFW, ORD, MIA, and STL), Continental (CLE, EWR, and IAH), Delta (ATL, CVG, and SLC), Northwest (DTW, MEM, and MSP), United (DEN, IAD, ORD, and SFO), and US Airways (CLT, PHL, and PIT). 8 While United’s conversion of its Boeing 777 fleet is still ongoing, this aircraft is used primarily for international service. Moreover, to account for American’s acquisition of TWA in 2001 – and the subsequent conversion of TWA’s fleet to include More Room Throughout Coach – American’s itineraries that connected via St Louis are excluded.

ARE PASSENGERS WILLING TO PAY MORE FOR ADDITIONAL LEGROOM?

279

rather than the city-pair markets to control for differences in willingness-to-pay based on different airports within a metropolitan area. Likewise, directional are used rather than nondirectional markets to control for differences in marketing or frequent flyer loyalty at the different endpoints of a market. In the example above, we would require that Delta served the market in each of the 5 years of our panel. Separate samples of markets for American and United are used to see if changes in their respective seat pitch impacted relative fares in markets where they had overlapping connecting service with other carriers. The American sample consists of 994 unique markets and the United sample of 771 markets. Within the set, American has the most overlapping markets with Delta (712) and the fewest with US Airways (228). Similarly, United has the most overlapping markets with American (542) and the fewest with US Airways (220). The combined revenues of the Big Six carriers for connecting passengers each year is between $1.25 and $2 billion. Finally, during the period of time covered by our analysis – 1998 to 2002 – the use of Internet channels such as Orbitz, Expedia, or Travelocity to instantly compare airfares across different carriers became a widespread phenomena. Consequently, the increased price transparency afforded by the emergence of Internet travel sites would lead one to expect any fare premia that existed at the beginning of our data set to diminish as time passed. Fixed effects fare equations are estimated for both markets that allows us to control for unobservable effects correlated with the observed explanatory variables, lessening possible omitted variable biases. The market-fixed effects control for demand and cost differences that are common for all airlines serving the same market (such as distance, total market size, or competition from low-cost carriers) yet vary across markets. Note that this approach does not permit identifying the effects of variables that do not vary within a market. The equation is Fij = Xij  + uj + ij  where Fij is carrier, i’s passenger-weighted average round-trip fare (net of all taxes and fees) in dollars in market j, Xij is a vector of regressors that varies with the airline’s identity within a market, and uj is the market-fixed effects. The random error ij is assumed to be i.i.d. with zero mean and variance 2 . The vector X includes independent variables that control for carrier identity and time, as well as other elements of service quality and/or potential market power: shareij (market share): The carrier’s share (in percentage points) of all O&D passengers (connecting and nonstop) in market j. To deal with possible endogeneity in the determination of fares and market shares in a given airport-pair market, we instrument for shareij using the previous year’s market share. distij (itinerary distance): The average distance (in hundreds of miles) traveled by pas­ sengers on carrier i in market j, passenger weighted by specific routing. We also include the squared distance. Longer distances resulting from more circuitous routings may be considered less desirable, lowering fares. On the other hand, more circuitous routings also cost more to provide, which could result in higher fares. orgshareij (originating share): Carrier i’s share (in percentage points) of passengers across all markets at the originating airport in market j, among the Big Six carriers.

280

DARIN LEE AND MARíA JOSÉ LUENGO-PRADO

Numerous researches (Evans and Kessides, 1993) have noted that high market shares at endpoint airports may provide carriers with a pricing advantage on markets served from that airport due to frequent flyer program loyalty or other marketing advantages. frequencyij , (schedule frequency): Higher flight frequency in a market represents higher quality service for most passengers. We construct our schedule frequency variable in the following way. Travel between the origin and destination of market j on carrier i can involve routings over a number of potential hubs. The schedule frequency for each of these hub routings is computed as the minimum of the average daily flights from the origin to the hub and the average daily flights from the hub to the destination. For each market and carrier, frequencyij is then computed as the sum of these minimum daily routing values across all of the possible hubs for that carrier. There are two elements of scheduling that are not fully account for – the precise timing of flights throughout the day and elapsed travel time. Passengers may prefer flights that depart during one part of the day more than others. Likewise, routings with longer distances may still have shorter elapsed travel times. businessij (business passengers): Passenger mix can have a significant impact on average fares (i.e., Lee and Luengo-Prado, 2005). businessij is the proportion of carrier i’s passengers in market j purchasing tickets with fares of 60% or more of the market’s 95th percentile fare. Business passengers are proxied using this method since the fare coding definitions in the DOT’s OD1A database may not be comparable across carriers or may have changed over time. nsdumij (nonstop dummy): If a carrier offers nonstop service in market j, it may impact its pricing strategies for its connecting service in this market. For example, a carrier may price its nonstop service less aggressively than it otherwise would if it also offered nonstop service in that market, for fear of cannibalizing its higher quality (i.e., nonstop) service. nsdumij is a dummy variable that takes the value 1 if carrier i served market j with nonstop service and takes the value 0 otherwise. ontimei (on-time performance): Carriers with superior on-time performance may be able to charge higher prices on competitive routes if passengers are aware of such performance. On the other hand, higher on-time performance lowers costs, which in turn may be passed along to consumers in the form of lower fares. ontimei measures the percentage of carrier i’s system-wide flights that arrived on-time, as measured by the Department of Transportation’s Air Travel Consumer Reports. We chose systemwide rather route specific on-time performance since on-time performance tends to be reported in the media on a system-wide, rather than route specific basis. Moreover, passengers who experience poor on-time performance on a given carrier are likely to associate this element of service quality to the carrier as a whole rather than the carrier on a specific route. leveragei (firm financial condition): Busse (2002) indicates that a carrier’s financial condition may impact its proclivity to price more aggressively and to meet its debt payment obligations. To allow for this, the carrier’s leverage ratio (defined as Total Assets/Total Stock Equity) is an indicator of its financial condition using data from the Department of Transportation’s Form 41 database. Since US Airways’ leverage becomes negative in 2002, the 2001 level is used. trend (time trend): A time trend is used to reflect the steady decline in average airfares since 1990 (Lee, 2003).

ARE PASSENGERS WILLING TO PAY MORE FOR ADDITIONAL LEGROOM?

281

D(carrier)pre and D(carrier)post (carrier dummies): Since the primary interest is in deter­ mining the impact American and United’s increased seat pitch programs had on their fares relative to other full-service carriers, a number of carrier dummy variables are used. For each carrier, there are two dummies, one taking the value one (zero other­ wise) if the year is prior to the change in seat pitch (1998, 1999, or 2000) and the other taking the value one (zero otherwise) if the year is after the change occurred (2001 and 2002). Although load factor is another possible candidate for an independent variable, average fares and load factors tend to be endogenous. Since there is no obvious instrument for load factor and it was found to be insignificant when included, it is excluded from the list of independent variables. Summary statistics are presented in Table 1. Fixed effects two-stage least squares using two model specifications are used for estimation. Model 1 pools all carriers – other than the base carrier of interest – while Model 2 uses carrier specific dummy variables (Table 2). The overall fits of the regres­ sions are quite strong, and the estimated coefficients tend to have the expected sign and are typically significant at the 1% or 5% level. The estimated coefficients on share, orgshare, frequency and business are positive and significant at the 1% level in all four regressions, consistent with the previous literature. Likewise, the estimated coefficients on leverage and trend are negative when significant, consistent with our a priori beliefs. The coefficient on ontime is negative, indicating that as a carrier’s on-time perfor­ mance improves, all other things equal, the carrier’s average fares decline. This suggests that superior on-time performance does not provide a carrier with a pricing advantage. Rather, the negative estimated coefficient for ontime suggests that there may be cost savings that are being partially passed along to consumers when carriers experience relatively fewer delays. The estimated coefficient on distance is consistently positive and significant at the 1% level, indicating that all things equal, longer routings are

Table 1 Summary Statistics Variables

fare share dist orgshare frequency business nsdum ontime leverage N Markets

American Airlines Overlap Markets

United Airlines Overlap Markets

Mean

SD

Mean

SD

34589 1497 1992 2110 612 2190 010 7788 446

12353 1479 697 1395 394 1290 030 483 211 16,760 994

34776 1456 2032 2096 641 2104 012 7718 473

12789 1450 668 1392 423 1235 032 544 228 13,275 771

DARIN LEE AND MARíA JOSÉ LUENGO-PRADO

282

Table 2 Estimation Results American

United

Model 1

Model 2

Model 1

Model 2

share

0898∗∗ 0076

0996∗∗ 0076

1245∗∗ 0089

1160∗∗ 0090

business

2380∗∗ 0048

2075∗∗ 0048

2407∗∗ 0056

2240∗∗ 0056

distance

6856∗∗ 1206

7599∗∗ 1185

8295∗∗ 1307

7093∗∗ 1301

orgshare

0420∗∗ 0044

0337∗∗ 0046

0386∗∗ 0051

0380∗∗ 0052

frequency

2456∗∗ 0198

2071∗∗ 0195

2670∗∗ 0208

2540∗∗ 0207

nsdum

4148∗ 1789

4743∗∗ 1759

2224 1963

1766 1947

ontime

−1631∗∗ 0135

−0283 0178

−0390∗∗ 0138

−1368∗∗ 0168

leverage

−1058∗∗ 0274

−005 0654

−4308∗∗ 0324

1277 0659

trend

−10547∗∗ 0733

−8069∗∗ 0742

−13109∗∗ 1086

−16885∗∗ 1068

D(all others)pre

−10468∗∗ 1305

−24946∗∗ 1906

D(itself)post

−16877∗∗ 2754

9290∗∗ 3055

D(all others)post

−18975∗∗ 2448

−27039∗∗ 2830 dropped

1863 2677

D(Continental)pre

−25700∗∗ 2126

−38109∗∗ 3178

D(Delta)pre

−16451∗∗ 1765

−15579∗∗ 3136

D(Northwest)pre

−22229∗∗ 2142

−24623∗∗ 2925

D(American)pre

D(United)pre

17127∗∗ 2362

dropped

D(US Airways)pre

−27453∗∗ 3042

−24789∗∗ 3262

D(American)post

−35122∗∗ 2944

−3266 3131

ARE PASSENGERS WILLING TO PAY MORE FOR ADDITIONAL LEGROOM?

283

Table 2 Estimation Results—Cont’d American Model 1

United Model 2

Model 1

Model 2

D(Continental)post

−35737∗∗ 3309

−10182∗∗ 3799

D(Delta)post

−33062∗∗ 2961

−4901 3732

D(Northwest)post

−50081∗∗ 3120 −11092∗∗ 3529

−15818∗∗ 3669 7916∗∗ 3059

−84114∗∗ 5933

−63495∗∗ 4342

D(United)post D(US Airways)post Observations

16,760

Number of markets  R2

994

994

771

771

07902

0.8008

07985

0.8024

∗

Significant at the 5% level;

∗∗

16,760

13,275

13,275

significant at the 1% level.

relatively more expensive. This indicates that cost considerations prompt carriers to price circuitous routings higher than more direct routings. Looking at the dummy variables, there is evidence that American’s More Room Throughout Coach program failed to yield any price premia. To the contrary, Model 1 indicates that prior to implementing the program, American typically received a $10.47 premium per ticket relative to all other carriers. After implementing More Room Throughout Coach however, its overall premium relative to the other full-service car­ riers fell to $2.10, a net drop of $8.37 per ticket. Likewise, relative to its own service before the change, American’s fares fell by $16.88 after the change, all other things equal, as indicated by the estimated coefficient on D(itself)post . Model 2 confirms that on a head-to-head basis with other carriers, American’s premium declined versus every carrier except US Airways. Prior to the change, for example, American had price premia versus Continental and Delta of $25.70 and $16.45, respectively. Following the change, American’s premium versus Continental was reduced to $0.62 while its premium versus Delta became a small deficit (−$2.06). American maintained a positive premium versus Northwest, but it declined from $22.23 before the change to $14.96 after the change. Finally, prior to the change, American already had a price deficit versus United of −$17.13 and following the change, this deficit increased to −$24.03 per ticket. For United’s “Economy Plus” program, Model 1 indicates that prior to the change, United generated a significant fare premium of $24.95 per ticket versus the other fullservice carriers as a whole. Following the change, United’s fare premium expanded to $36.33, an increase of $11.38 per ticket. The positive and significant estimated coefficient on D(itself)post confirms that United’s Premium Economy program helped it boost its average fare, all other things equal. From Model 2, we see that United had fare premia versus all of its full-service competitors (with the exception of American, which was

284

DARIN LEE AND MARíA JOSÉ LUENGO-PRADO

not significant) prior to the change in seat pitch.9 Following the change, United, unlike American, maintained fare premia versus all five of its full-service competitors. United’s fare premium increased following the change versus American, and US Airways declined modestly versus Northwest ($24.62–$23.74) and Delta ($15.58–$12.82), and fell more significantly ($38.11–$18.10 per ticket) versus Continental.

4 CONCLUSIONS This is no evidence that passengers were willing to pay a premium for the extra legroom offered by American’s More Room Throughout Coach program. To the contrary, the evidence is that the program resulted in lower average fares for American. In contrast, United’s Premium Economy program was effective in attracting passengers willing to pay higher fares for greater seat pitch when offered a choice of otherwise comparable service among competing full-service carriers. Thus, United’s Economy Plus program would seem to have been more effective at generating or maintaining fare premia than American’s More Room Throughout Coach program. Indeed, the relative success of United’s program compared to American’s sheds some insight as to why American recently announced it would discontinue its More Room Throughout Coach program in roughly one-quarter of its fleet.10 That fact that United’s increased seat pitch program aimed squarely at the “business” traveler segment of the market appears to have performed better may be a reflection of the importance of business travelers to the full-service carriers. Many leisure travelers are likely to choose the lowest-priced carrier, regardless of service quality. Business travelers on the other hand, tend to be less price-elastic, and since United’s Economy Plus seats offer the greatest coach class seat pitch of the major carriers, those passengers who value the extra space the most may be willing to pay a fare premium for United’s service. In this sense, the analysis provides empirical evidence to support models of spatial competition (Hotelling, 1929). Nevertheless, the full impact of this aspect of service-quality competition will likely not be known for some time to come. For example, since many leisure passengers travel by air infrequently, it may take time for passengers to experience and learn about differences in the aspect of in-flight service quality we study in this note. Likewise, while the focus of our analysis has been on whether or not the increased seat pitch programs generated fare premia, it is important to note that the carriers themselves are likely to focus more heavily on the overall revenue impact. Thus, while More Room Throughout Coach does not appear to generated fare premiums for American, it is possible that the revenue impact may have been positive if the higher quality service resulted in higher load factors. Finally, some of the differences in relative fares we find may have been

9

The estimated coefficient for D(United)pre in American Model 2 differs from the estimated coefficient on D(American)pre in United Model 2 because the data include different sets of markets where the airlines do not compete. 10 See “American Airlines Charts Course for Brighter Future: CEO Arpey Unveils ‘Turnaround Plan’ at Annual Meeting,” company press release, 21 May 2003.

ARE PASSENGERS WILLING TO PAY MORE FOR ADDITIONAL LEGROOM?

285

caused by other changes in service quality during the period of our analysis that we have not been able to control for, such as perceived safety and crew friendliness.

REFERENCES Borenstein, S. (1989) Hubs and High Fares: Dominance and Market Power in the US Airline Industry, RAND Journal of Economics, 20, 344–365. Borenstein, S. and J. Netz (1999) Why do all the flights leave at 8 am:? Competition and departuretime differentiation in airline markets, International Journal of Industrial Organization, 17, 611–640. Brueckner, J. and T. Whalen (2000) The price effects of international airline alliances, Jounal of Law and Economics, 43, 503–545. Busse, M. (2002) Firm financial conditions and airline price wars, Rand Journal of Economics, 33, 298–318. Dresner, M., J.C. Lin, and R. Windle (1996) The impact of low-Cost carriers on airport and route competition, Journal of Transport Economics and Policy, 30, 309–328. Evans, W., and I. Kessides (1993) Localized Market Power in the US Airline Industry, Review of Economics and Statistics, 75, 66–75. Hotelling, H. (1929) Stability in competition, Economic Journal, 39, 41–57. Lee, D. (2003) Concentration and price trends in the US airline industry: 1990–2000, Journal of Air Transport Management, 9, 91–101. Lee, D. and M. Luengo-Prado (2005) The Impact of Passenger Mix on Reported Hub Premiums in the US Airline Industry, Southern Economic Journal, Volume 72, No. 2, pp. 372–394. Mazzeo, M. (2003) Competition and service quality in the US airline industry, Review of Industrial Organization, 22, 275–296. Morrison, S. and C.Winston (1995) The Evolution of the Airline Industry. The Brookings Institu­ tion, Washington, D.C. Morrison, S.A. (2001) Actual, adjacent and potential competition: estimating the full effect of Southwest Airlines, Journal of Transport Economics and Policy, 35, 239–256. Ross, L. (1997) When will an airline stand its ground? An analysis of fare wars, International Journal of the Economics of Business, 4, 109–127. Rupp, N., G. Holmes, and J. DeSimone (2005) Airline Schedule Recovery after Airport Clo­ sures: Empirical Evidence since September 11th, Southern Economic Journal, Volume 71, No. 4, pp. 800–820.

This page intentionally left blank

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

13 Assessing the Potential Success of the Low-Cost Business Models in Asian Aviation Markets David Gillen∗ , Natthida Taweelertkunthon†

ABSTRACT The low-cost carrier (LCC) phenomenon observed in the US in the early 1970s, which expanded dramatically after deregulation, has been repeated in Europe in the mid-1990s. In Asia, the low-cost challenge is a more recent phenomenon, but it is growing rapidly partic­ ularly in Southeast Asia, a unique region where economics and the aviation environment are unlike that of either the US or EU. This chapter asks a fundamental question of whether LCCs will enter and gain market share in Southeast Asia, in a similar fashion to what has happened elsewhere    We investigate the question how serious a threat do Asian budget carriers pose to traditional state-owned carrier’s core business by examining the degree of demand substitutability among carriers? Discrete choice models, multinomial logit and nested logit (NL) models are applied to estimate the factors affecting carrier choices based on data obtained from a passenger survey conducted at Bangkok International Airport in November 2005. The demand substitutability analysis reveals that for short-haul leisure market, LCCs can easily penetrate this market with low fare no frills offerings. Short-haul leisure passengers do not value differences in product offerings between network carrier and LCC, and among the LCCs. Unlike leisure passengers, business passengers do value the competitive distance between the two different types of products offered by legacy carriers and LCCs.

1 INTRODUCTION Over the last decade the low-cost carrier (LCC) business model has fundamentally changed the level and nature of competition in aviation markets. While the US had

∗ †

Sauder School of Business and Director, Centre for Transportation Studies University of British Columbia Centre of Transportation Studies University of British Columbia

288

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

Southwest Airlines since 1972 other markets, and even the US to some degree have not experienced the LCC model until the mid-1990s. The creation of the single aviation market in the European Union (EU) shaped the liberalized environment within which LCCs can and have flourished. Similar events have occurred in Canada, Australia, Malaysia and Latin America. Even the US has seen a significant growth (and demise) in LCC activity. In the former Eastern Europe LCCs are springing up continuously as are LCCs in India and other developing economies. The LCC has dramatically expanded their share of many markets as they initially captured customers from other modes and other activities and took passengers from established airlines. Interestingly the experience in Europe is different from that of the US. LCCs captured many passengers not only from legacy carriers but also from charter carriers. In the US with LCCs having broad point-to-point networks they captured passengers from other modes and from traditional carriers, there was no charter segment to capture from in North America. The LCCs have broken the legacy carrier model that previously was able to engage in significant price discrimination and prevent fareclass substitution with fences and other restrictions. The LCCs unbundled the airline product, offered one-way fares and provided a simply fare structure which de-emphasized price discrimination and focused on cost efficiency. It took more than 15 years in the US and 10 years in Europe before major network carriers began to take the challenge of this new business model seriously. There have been a number of strategies undertaken by legacy carriers to respond to LCCs including pricing, strategic capacity use and introducing fighting brands or “airlines within airlines.” Subsequently legacy carriers in North America and Europe focused on reducing their fixed costs, attempting to reduce costs on a continuing basis, and strategic seat density of aircraft.1 In 2000s, the low-cost phenomenon has arrived in Asia, and it is growing rapidly particularly in Southeast Asia, a unique region where economics and the aviation envi­ ronment are unlike that of the US or EU. This chapter therefore asks a fundamental question of whether LCCs will enter and gain market share in Southeast Asia in a similar fashion to what has happened in the US and EU. The success and sustainability of LCCs in this region are questionable. With a massive population of 500 million people and rising middle-income class, LCCs are believed to prosper in the countries where ground and surface transportation are limited and not viable. However, the scarcity of secondary airports and the lack of open sky agreements among Southeast Asian countries lead many analysts to conclude that these budget carriers would have a more difficult time to succeed and survive. Not all LCCs can be accommodated by the market. As a result, the rise of Southeast Asian LCCs may not threaten network carriers quite as much as the forerunners such as Southwest and JetBlue, and Ryanair and easyJet in the US and Europe. The growth potential of LCCs is not disputable, but just as the Europe model for LCCs was only an adaptation of what American pioneered, so the Southeast Asian model will find its own way. The review of the literature on LCCs reveals that the study on the challenge of LCCs to Asian major network carriers is in its infancy. Existing analyses on this issue are

1

Use of aircraft with higher seating density can reduce the costs per ASM (available seat mile).

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

289

descriptive and anecdotal, based on aggregate statistical data, and comparative study drawing information from secondary sources. They are unable to shed light on the level of challenge LCCs pose on major network carriers’ core business. The research reported here investigates the level of low-cost challenge through the degree of demand substitutability among network and LCCs. This empirical analysis provides insight into the impact of LCCs on network carriers in Southeast Asia. Like some of the US and European network carriers, Asian legacy carriers have created the airline within an airline model. In response to low-cost competition, they launched low-cost subsidiaries; Tiger Airways by Singapore Airlines, Nok Air by Thai Airways and Citi-link by Garuda Indonesia, for example. Nevertheless, the success of airline within an airline in Asia is still uncertain. A large number of US and EU defunct low-cost offshoots have led to the conclusion that launching a low-cost offshoot is not a sustainable solution for network carriers to respond with the low-cost era. Moreover, there is a likelihood that these Asian low-cost offshoots will unavoidably take away their parent mainlines’ core business, creating an inadvertent self-cannibalization. The demand analysis reveals that for short-haul leisure market, LCCs can easily penetrate this market with low fare no frills offerings. Short-haul leisure passengers do not recognize the differences in product offerings between network carrier and LCC, and among the LCCs. Leisure passengers are more likely to postpone trips to specific locations and time in response to high fare, or to spend more time shop around for more affordable fares. For business market, LCCs have to spend more time and provide more effort to attract business passengers. Unlike leisure passengers, business passengers do recognize the competitive distance between the two different types of products offered by legacy carrier and LCCs. Competitive distance is appreciated through frills including comfort, service level, flight frequency, brand, frequent flyer program (FFP), ticket flexibility and physical product itself. However, our elasticity analysis further shows that business passengers are increasingly price sensitive for short-haul domestic service as well. Frills are becoming less important for short-haul trips. Our findings should be of concern to legacy carriers for a potential loss of their short-haul market presence, first in leisure market and then business market. We therefore anticipate that the low-cost phenomenon would be repeated in Asia, but it may take a longer time and legacy carriers may be able to adjust to response strategy. The chapter is organized as follows; Section 1 provides a brief survey of the LCCs in Southeast Asia, Sections 2 and 3 describe the modeling methodology and data respec­ tively. In section 4 we report the modeling results and draw conclusions in section 5.

2 LCC MODELS IN US, EU, ASIA AND REST OF THE WORLD The first experience with LCCs began with the formation of Southwest Airlines in the early 1970s when they flew intrastate routes in Texas. After the US congress deregulated domestic aviation in the US in 1978, a number of low fares with low frills or no frills carriers entered the airline business such as People Express and Muse Air. Though some were initially successful, it became clear in a matter of a few years that low fares alone were not enough. Most of these new entrants failed, either because they lacked sufficient

290

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

financing to do battle with network carriers, or because they were offering a product that passengers did not want, with infrequent flights using inefficient aging aircraft. The failure rate of the new entrants was much as a result of aggressive competitive response of the incumbent carriers as it was from poorly thought out or executed business plans. Network carriers successfully use their yield management systems to sell spare capacity at equally low fares, a hub and spoke system to take advantage of economy of traffic density (Barkin et al., 1995) and FFP to raise switching costs as well as create brand loyalty. It was not until the early 1990s that a new generation of low-cost, low-fare carriers emerged. The most profitable and successful low-cost airline, Southwest Airlines has proven repeatedly that strict adherence to the low-cost business model that it initiated in 1971 is a key for success. It has inspired many independent low-cost followers around the world to enter the low-cost market. Examples of new low-cost players in the US include ValuJet, AirTran, Frontier and Air South. These carriers came with strong financial backup and younger aircraft than their forerunners. They have rewritten the business model and basis of competition in the commercial airline market in the US; shifting from product differentiation to low cost. Learning from the past, they have served price sensitive passengers with more frequent flights, efficient and new fleets and at low fares. Further, they have avoided head-to-head competition with network carriers, flying routes that network carriers would not. Because of their efficiency they have gained a cost advantage over network carriers. Their operational effectiveness has forced network carriers to bring their own costs down. Network carriers attempted to replicate some of the cost advantage of low-cost com­ petitors by establishing their own low-cost offshoot. Continental was a pioneer of the first generation of airline within an airline; it launched Continental Lite in 1994. After operating for only one year, its low-cost unit was forced to leave the market with tremendous losses. Porter (1996) concluded that Continental has failed to successfully run the offshoot because it attempted to compete in two ways at once. On some routes, operated by its offshoot, Continental eliminated meals and first class cabin, increased departure frequency, lowered fares and shortened turnaround time. But on other routes, it remained in full service, continued to use travel agents, used mix fleets of aircraft, and provided baggage transfer as well as seat assignment. Other mainline offshoots joining the first generation of airline within an airline were Shuttle (1994) by United, Delta Express (1996) by Delta, and MetroJet (1998) by US Airways. With the lesson from Continental Lite, they closely duplicated fundamental attributes of Southwest model by using a single type of aircraft (B737s), eliminating meals, simplifying fare structure and using e-ticketing. At the same time, the offshoots would remain legally part of the mainline carriers. They attempted to fend off other independent low-cost competitors by allowing their passengers to accumulate miles in parent airline FFP. For example, in addition to earning miles in United Mileage Plus FFP, Shuttle’s passengers would still enjoy pre-assigned seating, and transfer seamlessly to and from mainline service. Moreover, passengers could attain Shuttle tickets via the United’s Apollo Computer Reservation System. During the dot-com bubble, the offshoots became profitable. However, they encoun­ tered mounting operation problems and labor unrest. They also found it difficult to avoid cannibalizing their core business. In fact, MetroJet did not improve US Airways’ loss-making record. Moreover, many of MetroJet’s passengers were cannibalized from

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

291

other US Airways operation, such as its major presence at Ronald Reagan Washington National Airport. When air travel demand dropped drastically in 2001, it became apparent that cost savings had not materialized to justify the separate operation of the offshoots. As a result, MetroJet, Shuttle by United and Delta Express were folded back into their mainline by 2002. By the end of 1990s, the impacts of low-cost phenomenon became apparent. Initially, the low-cost, low-fare carriers drew passengers who were disenchanted with the tradi­ tional carriers. Next, they obtained market share from passengers who had switched from non-aviation modes such as car and bus or who were shifting between activities. Only then did they begin to “steal” passengers from the traditional carriers. High-end business passengers, a high-yield target group of network carriers, were no longer willing to pay expensive ticket fares for air travel. LCCs have broken the traditional price discrimation strategy of network carriers with the low, no frills one-way fares and direct to customer Internet ticket distribution. They have gained more market share and done the unthink­ able; they moved beyond the secondary airport, they had conquered and challenged the network carriers in their own way. Today, LCCs in the US operate more and more flights in direct competition with network carriers, even in the heavily traveled region, such as New York and Los Angeles, Philadelphia and Dallas. JetBlue Airways, established in 2000, flies from NY’s Kennedy Airport to a set of cities, targeting high-end leisure and frequent flyer business passengers. It is famous for luxurious in-flight services such as leather seats, and free satellite-television programming from DIRECTV (Blum, 2005). The LCC segment now has approximately 30% of the US market in term of revenue per miles (RPMs). The success of Southwest, JetBlue and AirTrans encouraged network carriers to join the low-cost market again. The second generation of airline within an airline began with Tango (2001) and Zip (2002) by Air Canada. Next, United’s Ted made its debut in Denver with more meticulous branding in 2003, while Delta relaunched Song in the same year.2 This time, United and Delta were significantly concerned with brand strategy, and how to position two potentially different company brands. After conducting customer research, United drew two conclusions. First a new brand, Ted requires a level of uniqueness and distinction from United. Thus, Ted would offer low and simplified fares and fly to where United will not fly. Second, it would maintain the strengths of the mainline brand, United Mileage Plus FFP. Delta also put more distance between itself and Song. The importance of brand strategy for the success of low-cost offshoots became evident when Zip and Tango had to fold back to Air Canada in 2004.

3 THE EUROPEAN LOW-COST EXPERIENCE Across the Atlantic, the glory of Southwest inspired Ryanair, an Irish loss-making full service airline, to transform itself to be the first European low-cost airline in 1991. After the European airline market was deregulated in three stages in the mid-1990s,

2

Song has since left the market.

292

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

the US low-cost evolution has been repeated in Europe. A large number of LCCs, beyond Ryanair and easyJet, entered the market hoping to succeed like their forerunners. However, many of them failed, and went bankrupt as well as went out of business through acquisition. Examples of failures include European mainline offshoots, such as Go (1998) by British Airways (BA) and Buzz (1999) by KLM. While Buzz failed because of its poor business plan and misalignment between its marketing strategy and operational strategy, Go became a liability for BA because it was cannibalizing the airline core business. Both KLM and BA introduced low fares on their main routes, diverting passengers from their own subsidiaries, and creating further confusion in the minds of passengers. Furthermore, BA experienced an internal difficulty in launching Go. It believed that it would be harder to convince staff of the need for radical cost cutting on the existing unprofitable short-haul operations if Go was still part of the group. In 2002, BA decided to sell its low fare airline Go to easyJet, while KLM sold Buzz to Ryanair in 2003. Similar to the US experience, the first generation of airline within an airline in Europe did not last long.3 After 2000s, the number of LCCs grew incredibly fast. It was estimated that there were 54 LCCs operating in Europe in the summer of 2004, compared with just 12 in 2000 (Baker et al., 2005). Some emerging LCCs are from former Eastern Europe, including Wizzair, Air Polonia, Centralwings, SkyEurope and Smart Wings. A prominent player among new low-cost entrants is germanwings by Lufthansa. After failing to launch its first offshoot Lufthansa Express, Lufthansa now has gained sufficient insight from the past failures and the key success factors for new low-cost offshoot. With unique and distinctive brand and management team, germanwings has been set as the JetBlue of Europe, a premium LCC flying mostly to primary airports and offering leather seats with in-flight entertainment. The second generation of airline within an airline in Europe seems to succeed, following their counterpart in the US. In fact, many of these 54 carriers are very small. Only a few of LCCs are operating with outstanding performances in term of traffic and revenue. Ryanair, easyJet, Air Berlin and germanwings have proven to be the winners of low-cost battle for intra-European travel, becoming the four respective largest European carriers in term of network. They have expanded aggressively and adversely affected European network carriers. In Europe, LCCs now have about 25% of the EU market. This is growing as unlike North America, LCCs are garnering market share from charter carriers as the LCCs enter 4+hour flight segments (see Appendix A for a list of LCCs in each area of the world).

4 THE ASIAN LOW-COST EXPERIENCE In the US and Europe, presently LCCs account for roughly 25% of market share by seat capacity (Baker et al., 2005). The current intra-regional traffic of Southwest is not too far behind Delta and American, while Ryanair has overtaken Lufthansa to become the largest intra-European carrier. In Asia, the low-cost phenomenon is in its infancy,

3

Interestingly BA realized a substantial profit when it sold off GO.

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

293

but it is growing rapidly particularly in Southeast Asia (SEA).4 Despite the problem of restrictive bilateral agreements among Southeast Asian countries, Kuala Lumpur based AirAsia became a successful pioneer. AirAsia took to the skies in November 2002. Today it has a 30% domestic share of Malaysian market (O’Connell and Williams, 2005). Hoping to imitate the success of AirAsia, a half dozen LCCs have opened for business or plan to do by the end of 2002. One-Two-Go and Thai AirAsia have launched in Thailand, Valuair is flying out of Singapore, and Lion Air offers daily flights from Jakarta to Singapore, Kuala Lumpur and other destinations in Indonesia. In response, major carriers have set up low-cost subsidiaries: Tiger Airways by Singapore Airlines, Nok Air by Thai Airways, and Citi-link by Garuda Indonesia. However, the future of the first generation of airline within an airline in Asia is still uncertain. Will the new low-cost offshoots take away flag carriers’ mainstream business, an inadvertent self-cannibalization? The boom of LCCs is being facilitated by the governments of Malaysia, Thailand, Indonesia and Singapore, which are easing their aviation regulatory, granting landing rights to the low-cost entrants in hopes of boosting tourism and business travel. Never­ theless, views still vary on whether low-cost carriers will blossom in Southeast Asia. Proponents of LCCs cited reasons for their success. The fact that Southeast Asia has a massive population, more than 500 million people with a rising middle class and a growing propensity for travel. The geography location of surrounding islands without viable and competitive alternative modes of transportation in Southeast Asia will enhance LCCs an enormous competitive advantage over surface and ground modes. For example, an express coach trip on the 690 kilometer trip by road from Bangkok to Chiang Mai in northern Thailand takes about 10–11 hours, while flying time is only 70 min. The journey by coach from Bangkok to Siem Reap, the sight of AngKor Wat in Cambodia, involves a road trip of 6 hours, while a flight takes 60 minuets (Hooper, 2005). Because of the inefficiency and inconvenience of ground and surface modes, LCCs are anticipated to drastically attract passengers from other modes and budget-conscious passengers to fly more often. It is further believed that LCCs would pose serious threat to existing major network carriers in the long run, taking away a proportion of their valuable business passengers just as in Europe and the US. Nevertheless, some aviation industry analysts argue that the rise of Southeast Asian LCCs may not threaten established airlines quite as much as upstarts such as Southwest and JetBlue, and Ryanair and easyJet have in the US and Europe. They do not dispute the growth potential of LCCs, but they do believe that just as the Europe model for LCCs was only an adaptation of what American pioneered, so that Southeast Asia model will fine its own way. The success of Southwest Airlines came after deregulation when state boarders did not matter. Ryanair and easyJet have enjoyed the prosperity after the EU reached the open skies agreement and formed a single market. The difference in Southeast Asia is not only that the boarders are international but also involved with bilateral air service agreements. Customs and immigration procedures would hamper the quick turnarounds that Southwest or Ryanair models rely on to enhance productivity and

4 Southeast Asian countries are composed of Thailand, Indonesia, Malaysia, Philippines, Singapore, Cambodia, Laos, Myanmar and Vietnam.

294

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

keep costs down. It would also appear that the principle sources of lower costs available to US and European LCCs may not be available to aspiring Asian LCCs (Gillen and Lall, 2004). How and where cost differentials between network carriers and their low-cost com­ petitors would be replicated, is still questionable. European and American LCCs enjoy substantial cost advantage over network carriers in term of labor issue and landings fees at secondary airports (Lawton, 2002). Strong labor unions force network carriers to pay expensive labor rates and follow work rule classifications which low-cost players have avoided. But Southeast Asia is known as a region of employment opportunity, skill shortage, labor flexibility and labor union weakness (Thomas, 2005). A scarcity of secondary airports within commuting distance of Southeast Asia’s capital cities means that low-cost airlines there cannot avoid congested and pricey primary airports as easily as US and European counterparts can. This limits their ability to save on landing fees, have quick turnaround times and reach high aircraft utilization rates. Southeast Asia’s major airlines, such as Singapore Airlines and Thai Airways, are more cost competitive than their European and American counterparts. They already enjoy lower average costs per kilometer than other global carriers, even as low as no frills airlines. They can sell a round trip economy class at a discount for short-haul regional routes by keeping a wide body aircraft busy with several daily flights, and then using the same plane for an overnight long-haul international route. Network carriers can match cheap tickets offered by LCCs because they have first and business classes plus cargo. Thus, one of the big distinguishers between network and budget carriers in Europe and the US may not play an important role in Southeast Asia. The new Southeast Asian budget carriers also face more competition than LCCs elsewhere. In Europe or the US, LCCs target routes that are dominated by just one or two carriers. However, in Southeast Asia, competition is more rigorous. In flying between Singapore and Bangkok, Valuair and Thai Air Asia compete with a dozen of other competitors. On the Hong Kong–Singapore route, Valuair has to fight against seven other carriers. The history in other continents is that not all new LCCs can be accommodated by the market. As an industry shakes out, many of low-cost entrants are unlikely to survive. The common prediction is that any Southeast Asia budget carriers would have a more difficult time to succeed and survive. Thus, it is likely to see a wave of mergers and failures in Southeast Asia’s highly competitive low-cost airline industry. The first merger was already started in Singapore between Jetstar Asia and Valuair in July 2005. In light of the experiences in North America and Europe, LCCs are able to deliver 80% of the service quality at less than 50% of the costs of network carriers on con­ tinental travel routes (Franke, 2004). Consequently they at least tackle more than 70% of O&D, taking them far from their origins as niche businesses (Franke, 2003). With the peculiar characteristics of Southeast Asian market, the impact of LCCs posing on major network carriers is still disputable. It is questionable how serious budget carriers can pose threat to traditional state-owned carrier’s core business. The cost strategy of LCCs elsewhere may not be sustainable in Southeast Asia. Therefore one must con­ sider what is the degree of demand substitutability among carriers? Does the major carrier’s launch of low-cost subsidiary provide a viable solution to respond with low-cost competition?

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

295

5 MODELING SUBSTITUTABILITY BETWEEN LCCS AND LEGACY CARRIERS In a liberalized air transportation market, passengers typically have the choice between several fare class products on one or more available flight itineraries offered by the airlines serving the desired markets. Air passengers will choose a particular airline, flight and fare class to satisfy their travel needs, based on their characteristics, needs and preferences. As a result, demand for air travel at the class level is the consequence of the tradeoffs individual air travelers make when they choose among different airlines, flights and fare classes. For the airline managers, the key airline planning decisions, for instance, flight scheduling, pricing, fare class restriction design and seat allocation among classes (yield management) require comprehensive information on passenger demand. It is then important to gain some insight into air travel preferences and to understand the determinants of demand for air travel at the class level. In most airline markets, air travelers have a choice between various travel alternatives. The number of available alternatives varies market by market based on the number of airlines serving that market, the number of flights offered by each airline, and the number of fare class products they actually market. For a given time period, this number is always finite. Each individual air traveler has to make a choice among a finite number of possible travel alternatives. This means the set of alternatives called the choice set is collectively exhaustive. The choice set might vary across decision makers based on their preferences or on their access to information, however, the number of alternatives is always finite. The air traveler’s decision to buy a plane ticket, which airline to take, when to travel, and by what route are all decisions that reflect an “either-or” choice. The air traveler either takes airline A or he does not. If he takes an airline A to Florida, he cannot simultaneously take an airline B. As a result, the set of alternatives is mutually exclusive. A nested logit (NL) model of the form below was used to examine the decision-making process of travelers. The substitution patterns of travelers is illustrated in Figure 1. In such a tree, each branch denotes a subset of alternatives within which IIA holds. Every leaf on each branch denotes an alternative. The tree consists of two branches, labeled “Airline A” and “Airline B,” for the two subsets of alternatives, and each of the branches contains two subsequent branches for the two alternatives within the subset (labeled flight A1, A2, B1, and B2 respectively).5 More formally, the estimation model is constructed as, let the J alternatives be parti­ tioned into K nests labels B1        Bk . The generalized nested logit (GNL) probabilities are derived from

GY1      YJ  =

 K   k=1

5

k jk Yj 

1/k

j∈Bk

There is proportional substitution among various flights offered by the same airline but not across flights from different airlines. In this case, if airline B were to launch a third flight in the market, demand for flight B1 and B2 would decrease by the same proportion but the proportion would be different from the decrease in passenger demand for flights A1 and A2.

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

296

Airline A

Flight A1

Airline B

Flight A2

Flight B1

Flight B2

Figure 1 The Nested Logit Tree Structure. Source: Carrier (2003).

where Bk is the set of alternative in nest k, jk is an allocation parameter which reflects the extent to which alternative j is a member of nest k.6 This parameter must be non­ negative: jk ≥ 0 ∀j k. A value of zero means that alternative is not in the nest at all.  The additional condition k jk = 1 ∀j provides a useful interpretation with respect to allocation of each alternative to each nest. The logsum or parameter k is defined to measure the degree of independence among alternatives within the nest k. The NL model is consistent with utility maximization if the condition 0 ≤ k ≤ 1 is satisfied for all k ; a higher value of k means greater independence and less correlation. A value of k = 1 indicates complete independence within nest k, that is, no correlation. The probability that a decision maker n chooses alternative i is  Pi =





Vi

ik e 

k K  l=1



j∈Bk

 j∈Bl

k −1 Vj 1/k

jk e 

l

jl eVj 1/l

If each alternative enters only one nest, with jk = 1 for j ∈ Bk and zero otherwise, the model become a NL model. In addition, if k = 1 for all nests, then the model becomes standard MNL. Interpretation is facilitated by decomposing the generalized NL probability as follows:

Pi =



PiBk Pk

k

where the probability of nest k is

6

The utility maximizing nested logit model is a special case of the GEV model. McFadden (1978) develops the model to ensure that it is consistent with utility maximization, provided that the log sum parameters are bounded appropriately.

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

 Pk =

 j∈Bk

K 



l=1

297

k Vj 1/k

jk e 

 j∈Bl

l jl eVj 1/l

and the probability of alternative i given nest k is  eVi 1/k PiBk =  ik Vj 1/ jk e  k j∈Bk

Wen and Koppelman (2001) derive direct and cross elasticities of the GNL model. The direct elasticity of an alternative i which appears in one or more nests with logsum, k , less than one, is  P

EXii =

k

   1 Pk PiBk 1 − Pi  + − 1 1 − PiBk  k Xi Pi

The term in the summation collapses to zero for any nest which does not include alternative i. The elasticity reduces to the MNL elasticity, 1 − Pi  Xi , if the alternative does not share a nest with any other alternative or is assigned only to nests for which the logsum value, k , equals one. The corresponding cross elasticity of a pair of alternative, i and j, which appear in one or more common nests, is ⎡ ⎢ P EXji = − ⎢ ⎣Pi +

 k



 ⎤ 1 − 1 Pk PiBk PjBk ⎥ k ⎥ Xi ⎦ Pj

In this case, the term in the summation becomes zero for any nest which does not include both alternative i and j, and reduces to the MNL cross elasticity, – Pi Xi , if the alternatives do not share any common nest. Note that these elasticities are independent of the elasticities for any other alternative or pair of alternatives.

6 DATA DESCRIPTION Data were obtained using interview survey questionnaires applied to leisure and busi­ ness passengers on short-haul domestic trips in Thailand. Face-to-face interviews were conducted to obtain the passenger’s socioeconomic profile and the chosen airline’s

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

298

attributes.7 Similar to a methodology adopted by Mason (2001), and O’Connell and Williams (2005), the data were obtained Bangkok Int’l Airport in Thailand from two large groups of passengers: one flying with local LCCs, and the other flying with an incumbent flag carrier. Each group also consists of leisure and business passengers. The passengers surveyed chose airlines which included Thai Airways (TG), and three local LCCs (LCCs), Nok Air (NOK), Thai AirAsia (TAA), and One-Two-Go (OTG) which are operating in the recently liberalized Thai domestic market. Bangkok Int’l Airport had a throughput approximately 45.11 million passengers in 2004.8 Key pieces of data for modeling were fare and flight frequency. While the data on daily frequency of a chosen airline were obtained from an airline website, the data on fare of a chosen airline were collected through questionnaires. Moreover, attitude statements regarding airline product attributes, various socioeconomic and behavioral data were collected. An attitude rating scale was used of each airline attribute on a nine point ranked continuum scale. Revealed preference data was collected to examine market shares of Thai domestic aviation market and to capture information on real choices. A total of 1,067 responses were collected at the airport, 63% of which were leisure passengers and 37% of which were business passengers. After removing responses from individual passengers who were on award redeeming flights, airline employees and some further data cleaning, a final sample of 910 individual responses was obtained. These divided into 344 business passengers and 566 leisure passengers, with 320 (35%) of which comprised passengers using TG (see Table 1 for data breakdown). Out of total responses, 590 or 65% of which represent passengers of the LCCs: Nok Air (28%), Thai AirAsia (23%), and One-Two-Go (14%). Both groups of respondents using TG and LCCs were traveling from Bangkok to the following destinations: Chiang Mai, Chiang Rai, Phuket, Hat Yai, Surat Thani, Ubon Ratchathani, and Udon Thani. All domestic flights are less than 2 hours flying out of Bangkok, regarded as short-haul services. Popular tourist destinations such as Chiang Mai, Hat Yai and Phuket are served by all carriers surveyed, whereas some routes have either two or three operators serving (see Table 2 for routes used in modeling analysis).

Table 1 Number of Questionnaires for Each Carrier Adjusted data

TG NOK TAA OTG Total

Business

Leisure

All

Business (%)

Leisure (%)

139 86 74 45 344

181 166 137 82 566

320 252 211 127 910

1527 945 813 495 3780

1989 1824 1505 901 6220

All (%)

3516 2769 2319 1396 10000

7 Interviews were conducted at the departure gates for short-haul service in domestic terminal. Passengers at

the domestic terminal were customers of all four airlines surveyed.

8 A detailed description of the data, collection methods and interview instruments as well as data description

can be found in Taweelertkunthon (2006).

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

299

Table 2 Destinations Served by TG and LCCs in Thailand Destination

Operator

Chiang Mai Chiang Rai Hat Yai Phuket Surat Thani Udon Thani Ubon Ratchathani

TG, TG, TG, TG, TG, TG, TG,

Nok Air, TAA, OTG TAA, OTG Nok Air, TAA, OTG Nok Air, TAA, OTG OTG Nok Air, TAA TAA

The passenger survey dataset contains information on the actual choice of a given set of travelers including fare and flight information. However for a modeling analysis, this needs to be complemented by dataset describing attributes of the different alternatives contained in the travelers’ choice set. In particular, important attributes such as daily flight frequency and fares of the non-chosen carrier on a given date of departure, period of booking and route are necessary for modeling. To construct the data on the non-chosen alternative in revealed preference data, we took the average for the fare levels of each observed alternative and substitute these averages as the values for the attribute levels of the non-chosen alternatives for those who did not choose them. Therefore, for each respondent, while retain the information on the individuals’ chosen carrier, we have generated data on the non-chosen carriers by using the averages of the non-chosen carrier’s fare levels as chosen by the other respondents given a period of booking, date of departure and destination. From official airline websites, we obtained the number of daily flights of non-chosen airline operating on the selected routes for the time period of the survey. It should be noted that our strategy of taking the averages of the fare level definitely reduces the variance of the fare level distribution in the sampled population. As a result, these averages would provide a smoother set of fare levels than what would be the levels if we knew the actual levels available to the person who has the alternative as the non-chosen. Our constructed data then may not perfectly represent the carrier fares actually faced by respondents to the extent the traveler may have faced a somewhat different fare for non-chosen alternatives than is contained in the data set. This was controlled for to some degree by observing how far before the flight the reservation was made. Therefore, the procedure used will provide relatively accurate values for fares on non-chosen alternatives.9

9 An alternative way of capturing this information on the non-chosen alternative is to synthesize the data. However, this approach requires expert knowledge as to how the data are to be synthesized. The norm is to use known information such as travel distances or other socioeconomic characteristics and to condition the synthesized data on these (Hensher et al., 2005). But like our approach, synthesizing the data still leaves one open to the criticism that the created data may not represent the carrier information actually faced by respondents and as such the estimation process will be tainted. In the following sections descriptive statistics of survey data are provided for both business and leisure passengers.

300

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

6.1 Business Survey Analysis As expected, LCCs have attracted a large number of young and middle-aged business people, with 5.9% of passengers using LCCs being in the under 25 years age group, and 61.5% being in the 25–40 year age group compared to 5 and 57% of TG passengers respectively. However, older passengers (more than 40 years old) tend to prefer the incumbent flag carrier, accounting for 38% of TG passengers surveyed. Unsurprisingly, a large proportion of high-income business passengers are willing to pay higher fares to travel with Thai Airways. Nearly 65% of TG passengers surveyed earn more than 30,000 baht per month, while the majority of LCC passengers surveyed (42%) earn less than 30,000 baht of monthly income. In addition, most business passengers whose companies paid for their tickets are likely to use Thai Airways compared to LCCs. About 82% of TG passengers’ trips were sponsored by their own companies, while the proportion of self-paid passengers using LCCs was double of those traveling TG. Moreover, the type of accommodation stayed at destination shows distinct difference between business passengers using Thai Airways and LCCs. Nearly 73% of TG passengers stayed in hotels, while approximately half of LCC passengers stayed at their parent, friend, company and own house. Business travelers usually tend to travel singularly. We found that about 70% of respondents using both types of carriers were traveling alone. The price is an area where some LCCs have segmented leisure and business travelers. While TG and One-Two-Go have a simple flat fare per route policy, Thai AirAsia has used a number of different fare classes and ticket restrictions. It sells their lowest fares first and then increases fares as the departure date draws closer. Similar price discrimination is also practiced by Nok Air with their further introduction of business fares that are most suited to meet the ticket flexibility needs of business travelers. In this way, business travelers that can plan ahead can book a cheap one-day return trip. The nearer the departure date, travelers will find the prices significantly higher than the lead in fare particularly on early morning departures. However, our investigation on the period of booking finds that Thai business passengers generally do not tend to seek lower fares through advance booking, with 74% of TG passengers booking tickets less than 1 week before departure, and 80% of LCC passengers doing the same. The passengers using TG indicated that the airline was selected primarily because of flight schedule (70%), FFP award (37%), pre-seat assignment (33%) and in-flight service (25%), while low-fare offering or promotion accounted for only 2.2%. In contrast, nearly 50% of passengers using TAA and OTG stated that low-fare offering was the reason of selecting the airline. A slightly lower proportion (35%) of the passengers using Nok Air pointed out the same reason. This is possibly because Nok Air is positioning itself as a premium LCC, charging 15–20% higher fares compared to other low-cost competitors. Though these results shows that Thai business passengers using LCCs are fairly price sensitive, they are still time sensitive. Flight frequency is the most common reason (65%) the LCC passengers indicated why the carrier was selected. Interestingly, both TG and LCCs had similar proportion of travelers making trips on sales/marketing, internal meeting/visit, and training. However, the main difference between TG and LCCs is in a large proportion of TG travelers going to conferences relative to LCCs; 34% compared to 21% respectively.

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

301

6.2 Leisure Survey Analysis Most LCC passengers surveyed were locals, with more than 92%, while international passengers composed of 20% TG passengers. LCCs attracted a high number of younger to early middle-age leisure passengers (25–40 years old), with more than 60% of LCC passengers surveyed. Looking at the range of monthly income of leisure passengers surveyed, we found that 50% of TG passengers earn more than BHT 30,000 per month, while nearly 65% of LCC passengers surveyed earn less than BHT 30,000 of monthly income. Clearly high-income leisure passengers were more likely to place more value on TG attributes such as in-flight comfort, flight frequency and FFP. Moreover, there is a high proportion of TG passengers whose parents or spouse paid for their trips. This group accounted for 28% of TG passengers while 83% of those using LCCs paid their own trips. From the survey we found leisure passengers using TG selected the airline mainly because of its flight schedule (67%), pre-seat assignment (38%), and FFP award (32%), while low-fare offering accounted for only 14%. However, for the passengers using LCCs, the most common reason is low-fare offering (33%) following by pre-seat assign­ ment (26%). The influence of pre-seat assignment on airline selection making for both business and leisure passengers using TG and LCCs provides a conclusion that Thai people generally do prefer to take a carrier which offers this service. Another interesting finding is the characteristics of Thai AirAsia leisure passengers. Among the LCCs surveyed, Thai AirAsia is the only LCC that strictly adheres to lowcost business model. It does not offer FFP, in-flight service, business lounge and pre-seat assignment. It also has the widest route network. In addition, the airline does not primarily operate yield management but maximizes seat capacity or load factor. As a result, it launches promotional discounts very often just to fill as many seats as possible. The answers from Thai AirAsia passengers show that short-haul domestic passengers were willing to trade off low-fare offering with any unnecessary airline attributes. Overall, the passengers surveyed took Thai AirAsia simply because of its low-fare offering (55%) and flight schedule (62%). The data showed that the biggest leisure market segment comes from those passengers who regularly visit friends, family and relatives (VFR). Low-cost carriers’ VFR traffic accounted nearly 50% of its total respondents, while that of TG fewer than 5%. This segment represented the largest number of leisure passengers carried on every airlines surveyed. The growth of this segment is primarily due to a large number of non-Bangkok born people living and working in Bangkok, and for Thai people time spent with family and friends is culturally a very important leisure activity. Sightseeing travel accounted for the second largest leisure market, with almost the same proportion of passengers using TG and LCCs (33%). The growth of sightseeing market is driven by the notion that the likelihood of leisure travel will go up as the size of the middle income class increases. Similar to their European and US counterparts, apparently Asian LCCs are taking advantage of online distribution and direct sales to generally avoid travel agent. AirAsia is the first airline in Asia to launch Internet booking with online payments and ticketless travel. Thomas (2003) reported that 45% of its bookings were made through the Internet by May 2003. Moreover, the carrier is pioneering, being the first airline worldwide to offer SMS booking, and is processing 2,000–3,000 messages per month (O’Connell and

302

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

Williams, 2005). In Indonesia, Air Asia passengers can also book tickets at nation-wide post offices, another alternative distribution channel. In Thailand, besides direct sales distribution channel, LCCs are competing to attract passengers with variety and convenience of payment methods. After booking ticket via call centers, LCCs enable passengers to pay ticket fares at convenient stores, 7–11 and supermarkets. Nok Air passengers are also able to pay for their tickets at bank’s ATMs. The variety and convenience of ticket payment encourage more passengers without credit cards to travel and book tickets via their call centers.10

7 MODELING RESULTS The basic MNL and NL models described earlier were estimated for different traveler types; two separate models were estimated, dividing business and leisure travelers. The final usable data set from the passenger survey were divided into 344 business passengers and 577 leisure passengers. In each model, the influence of important airline attributes was explored. These attributes were fare and frequency because they were found to have a consistently significant impact, according to air transportation literature. In addition to these attributes, a number of passenger socioeconomic and demographic characteristics (SDG) were included to better understand passenger’s decision-making process. These variables were age, income level, number of people traveling together, TG’s FFP membership, type of accommodation at destination, sponsor of the ticket, and past flying experience on different LCCs in both business and leisure trips. Because it is impossible to capture all information that affects the choice of a given decision maker, the utility of a given alternative is not fully observed, and the unobserved part of utility or error term remains. By adding alternative specific constants (ASC) to the utility of alternatives, the mean of this random distributed error term is added into the observed utility function, such that the remaining error term has a mean of zero. Therefore, these ASCs capture the mean effect of all unobserved variables, including general attitude toward an alternative, while the remaining error term captures the variation in this effect. For identification reasons, one of the ASCs has to be normalized. In the present analysis, the ASC of a LCC, One-Two-Go (OTG) was set to zero in both the MNL and NL models. It should be noted that our sample domestic market shares of the alternatives does not match the actual known market shares. It was not possible to sample passengers in the same proportion as the actual carrier market shares. The sample we wanted was to be random. In order to allow our estimations to reflect the true market shares of the

10 Thus, a significant discrepancy between the booking profiles of TG passengers and those of LCC passengers reflects the legacy incumbent’s inability to implement change particularly in technology and keep pace with its innovative LCC competitors. Though use of the Internet in Thailand is not widespread compared in Singapore, our results provide evidence that Thai passengers will seek out the available booking channels to access lower fares.

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

303

alternatives, we applied endogenous weighting criteria11 so that the actual market shares of the alternatives are represented by the choice variable within the data set. After imposing endogenous weighting scale on the sample data, estimations proceed by maximizing the likelihood of the alternative choices made by the 344 business travelers and 566 leisure travelers in our sample. We test two model structures (MNL and NL) to ascertain whether a correlation structure exits between the alternatives. The estimated models allow us not only to compute individual choice probabilities, which form the basis for demand forecasts, but also to derive own and cross elasticities with respect to fare and frequency for each model under analysis. We analyze the results from derived elasticity values in both leisure and business models. Our fundamental purpose is to investigate the degree of substitutability among carriers with respect to own and alternative fares and frequencies.

7.1 Model Results and Elasticities To investigate whether our models are better fit with NL specification, we estimated two-level NL models based on three different tree structures for both leisure and busi­ ness travels. The first tree structure is tested based on the assumption that unobserved attributes among LCCs are likely correlated. Since NOK is a TG’s low-cost offshoot, a second nested tree structure is based on the assumption that these two airlines’ unob­ served attributes are correlated. The last nested tree structure examined goes further by assuming that unobserved attributes of TAA and OTG are correlated. We found that the values of Pseudo R-square and LL functions of all estimated NL models show that NL model is no better than the MNL model. Moreover, all logsum parameters in every NL model are either greater than one or statistically equal to one, meaning that the all branches should collapse into a single branch, which is equivalent to a MNL model. The NL structures are rejected in favor of the MNL. We report the variables used in the leisure and business models and the estimated parameter values in Appendix B. The overall significance of leisure and business MNL models were established by comparing the log-likelihood (LL) function of these fitted models with the LL function of a model fitted using only information of the market shares (base model) as they exist within the data set. If the fitted model does not improve the LL function then the additional attributes and SDGs do not improve the overall model it beyond the base model. A base model using the market shares within the data are equivalent to a model estimated with ASC only. The base model represents the average utility for each of the alternatives and also represents the market shares present within the data set. We thus conduct a likelihood ratio test with respect to the null hypothesis that the specified model is no better than the base model. The likelihood ratio test indicates that the fitted models possess explanatory variables that provide a significant explanation of the data in relation to the constants only (or market share)

11

The sample data is weighted by the ratio of actual and sample market share fractions so that the estimation process is adjusted to take account of the weights in fitting the model. This approach provides us with more accurate models. After weighting the data, maximum likelihood estimator for our choice-based sample still yields consistent estimates for the parameters (see Ben-Akira and Lerman, 1985).

304

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

models. The likelihood ratio value with respect to this null model is far greater than the criteria value of Chi-square at the 95% confidence level. To understand the variation in choice of alternatives, we include explanatory variables. If an explanatory variable does not add to our understanding of choice, statistically the weight attached to the variable will equal zero. In linear regression, this test is usually performed via a t- or F -test. However, for choice analysis based upon MNL models, neither the t- nor F -statistic are available. The asymptotic equivalent test, the Wald statistic is calculated and interpreted in the same manner as the t-test associated with linear regression model. All the results presented (see Appendix B) show that all explanatory variables in both MNL models have expected signs and are statistically significant at the 95% level, except income variable (INC) in business model. The fare variables in both models are of correct sign, and the estimated fare coefficient in both models indicates that leisure and business travelers attach similar value to cost of travel. However, this result does not necessarily imply that elasticity values with respect to fares for leisure travelers would the same as those for business travelers. Frequency variable characterizing less travel time also has expected positive sign and significant different from zero. The difference in frequency coefficients between the two models also reveals that business travelers attach much higher value to time through flight frequency. An interaction term between fare and frequency was included in both models to capture interaction effects between fare and frequency upon choice. It is very possible that the level of fare charged by a carrier, particularly TG, will have a differential impact upon the choice of carrier when considered in concert with the number of flight frequency. TG operates 13 and 11 daily flights for BKK-Chiang Mai and BKK-Phuket routes while most LCCs operates 2–4 flights a day for the same routes. However, TG average fares are about 30–40% higher than fares offered by LCCs on these routes. Our empirical results bears out this relationship showing the interaction is statistically significant in both models. The business model also includes the interaction effects between flight frequency and destinations the carriers are operating (that were included in the survey). The dummy interaction terms were included to capture interaction effects upon choice of TG. It is very likely that the number of flight frequencies operated by TG will have a differential impact upon the choice of the carrier when considered in accord with popular tourist and business destinations such as Chiang Mai and Phuket. All of these interaction terms are significant supporting previous conclusion that business travelers attach higher value to time and flight frequency when making a carrier choice. High daily flight frequency to primary business destinations of TG is regarded as its core strategy to attract business travelers. Compared to ground and rail modes, air travel is regarded as an expensive transporta­ tion mode. Higher income passengers are likely to have a higher value on time and are willing to pay air fares relative to low-income passengers. As a result, the income level of a passenger is expected to have a positive impact on carrier choice probabilities for both leisure and business travel. Individual passenger’s average income is included in both leisure and business models. We found that the income coefficient of leisure passengers is statistically significant and of correct sign but its impact on carrier choice probabilities is minimal.

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

305

For business travel, unsurprisingly, the income level has no significant impact on carrier choice probabilities at the 95% level. The result might be explained by several reasons. First, the majority of business travelers’ trips surveyed were paid by their firms and clients. Second, business passengers using air transport as a business tool generally place an extra high value of time. They are willing to pay air fares to save time on short trip. Third, their business trips may involve the “must-go” short trips which are not easily planed and can be completed within a day. Because of these, income level possibly has an insignificant impact on business travel. The impact of TG FFP is tested by including a dummy variable for passengers using TG are holding its FFP membership. The dummy variable is then added only on TG utility function on both business and leisure models. Rewards are expected to have positive impact on TG passenger’s utility and we found this in both models. The difference in magnitude between leisure and business models reveals that this loyalty program is more powerful in attracting frequent flyer business travelers when making a choice to use TG, while unsurprisingly leisure travelers place less value on the reward program membership for short-haul domestic trip. To investigate the impact of past flying experience on usage behavior toward LCCs, we included several dummy variables in LCC utility functions whether a traveler had ever used a LCC he was taking before. These past experience variables are expected to have positive impact on a carrier repeated choice. We found that past flying experience on a LCCs which the passengers were taking has a positive impact on its passenger choice probabilities. Further, the larger magnitude of these coefficients in leisure model compared to business model indicates that past experience on any of LCCs is certainly taken into account when leisure passengers make a repeated choice of using LCCs. The ASC terms measure the relative “preference” of a reference group for the alterna­ tive whose indirect utility the constant term enters. Because TG is a full service carrier, we expect passengers to favor TG relative to LCCs, while there are no reasons to expect the alternative preference parameters to favor one or the other of the LCCs. We found these estimated constant terms indicate that, all else constant, there is an equal preference for selecting NOK or OTG, TAA or OTG for leisure and business trips. However, we also found that both types of travelers have a definite preference for TG relative to OTG, NOK and TAA. The relative preference for TG most likely reflects the convenience, seat comfort, ticket flexibility and in-flight services of full service carrier, TG, in comparison with LCCs, as well as the effect of other variables that are not included in the models.

7.2 Demand Elasticities The coefficients obtained from MNL models do not provide direct behavioral interpre­ tation of a parameter estimate of a choice model beyond the sign of the parameter which indicates whether the associated explanatory variable of interest has either a positive or negative impact upon the choice probabilities. In order to obtain a behaviorally meaning­ ful interpretation, we calculated the elasticities of the choice probabilities with respect to fare and frequency attributes. For direct elasticities, the calculated elasticity is interpreted as the percentage change of the choice probability for alternative i given a 1% change in attribute k of alternative i. For cross elasticities, the calculated elasticity is interpreted as the percentage change of the

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

306

choice probability for alternative j given a 1% change in attribute k of alternative i. We follow Louviere et al. (2000) to use the probability weighted sample enumeration (PWSE) technique to calculate the aggregate elasticities. The use of PWSE avoids the pitfalls of the aggregation method or naïve pooling because it calculates the elasticity for each individual decision maker by weighting each individual elasticity by the decision maker’s associated choice probability. The PWSE technique thus recognizes the contribution to the choice outcome of each alternative. It should be noted that the use of PWSE will produce non-uniform cross-elasticities; however, the individual level cross elasticities are strictly identical for the IID assumption. We used NLOGIT software to calculate the elasticity values and its calculation method is the point elasticity method. Tables 3 and 4 present the own and cross point elasticities with respect to fare and frequency derived from the leisure and business Table 3 Own and Cross Elasticity Derived From Leisure Model Leisure

Own and Cross Elasticities TG

NOK

TAA

OTG

−0.853 0952 1057 0973

0183 −0.978 0252 0239

0159 0196 −0.918 0192

0129 0167 0171 −0.962

0.181 −0205 −0189 −0245

−0029 0.154 −004 −0039

−0025 −0038 0.159 −0038

−0028 −0038 −0029 0.202

Fare TG NOK TAA OTG Frequency TG NOK TAA OTG

Note: Own elasticity is highlighted.

Table 4 Own and Cross Elasticity Derived from Business Model

Business

Own and Cross Elasticities TG

NOK

TAA

OTG

−0.69 0747 0818 0682

0151 −0.985 0306 0332

0121 0243 −0.871 0257

01 0234 0223 −0.941

0.656 −0791 −067 −0673

−0123 0.765 −0223 −0251

−0113 −0235 0.823 −0241

−0092 −0231 −0188 0.863

Fare TG NOK TAA OTG Frequency TG NOK TAA OTG

Note: Own elasticity is highlighted.

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

307

models respectively. Using Table 3, we can take the example of the business elasticity for the fare attribute on the Thai Airways (TG) alternative, the direct effect is calculated as –0.69. This indicates that a 10% increase in TG air fare will decrease the probability of selecting the Thai Airways alternative by 6.9%, all else being equal. As would be expected, raising one’s own price is likely to decrease demand for one’s own good or service. The remaining point elasticities represent the cross-elasticity effects in that column. Examining these effects, the business elasticities show that a 10% increase in the fare for the Thai Airways alternative will result in a 7.47% increase in the choice probabilities for the Nok Air alternative and an increase of 8.18% and 6.82% in the choice probabilities of the Thai AirAsia, and One-Two-Go alternatives respectively. The findings conform to our expectations. An increase in a ticket price of Thai Airways is likely to increase the demand for competing low-cost airlines and there is no reason why the increase in demand would be uniformly spread across these carriers. We summarize key findings from our elasticity analysis as follows. First, our estimated own price elasticities for both leisure and business travels fall in the range of short-haul leisure and business elasticity values reported by Gillen et al. (2002). They report the findings of a review of the economics and business literature on empirically estimated own price elasticities of demand for air travel for Canada and other major developed countries. This provides ample evidence that our demand study which segments distinct markets for business and leisure travel, and long-haul and short-haul travel is appropriate. Second, our estimated own price elasticities for leisure travel are higher than those for business travel according with expectation that the demand for air transport for leisure reasons will be more elastic than business travel. Leisure travelers are more likely to postpone trips to specific locations and time in response to high fare, or to spend more time shop around for more affordable fares. Table 3 shows that for leisure travel the magnitude of own and cross price elasticity is far greater than the elasticity with respect to frequency. Leisure travelers will seek the lowest fares possible and thus LCCs will find it easier to capture and penetrate this market with low-fare offerings. The explanation why the short-haul business elasticity is smaller in value than the short-haul leisure elasticity is straightforward. Business people using air transport as a business tool, will place an extra high value of time. Therefore they are willing to pay high fares to save time on short-haul trips. Moreover, there are a number of “must-go” short-haul trips that occur in the course of business dealings. These short-haul business trips are not easily planed and can be completed in the morning or afternoon without requiring scheduling meetings or packing or making family arrangements. The high airfare is low when factored into the overall value of the trip. However, our elasticity results reveal that there are very small differences in the values of leisure and business own price elasticity. This finding points out that either business elasticity is rising or leisure elasticity is falling. Rising business elasticity implies that business travelers are becoming more cost conscious or increasingly price sensitive particularly for short-haul domestic trips. As can be seen for business travel, the minimal gap between the elasticity values with respect to fare and frequency implies that TG business passengers now attribute almost the same value to time and ticket fares for short-haul domestic trip. Perhaps after a careful consideration of fare and flight frequency offered by TG, more and more cost-conscious business travelers will switch to fly with LCCs on domestic routes. This explanation will clearly be of concern to the

308

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

network carrier, TG for its domestic market presence. The declining leisure elasticity can intuitively be accounted for the very low-fare offered by LCCs. After LCCs entered Thai domestic market, domestic fares have fallen significantly and leisure travelers have increased their number of trips. The leisure market has possibly come to the point where further small drops in fare would have little impact on the market. A fare falling from 1,800 baht to 1,700 baht has trivial impact on the cost of a trip expense; as a result short-haul leisure elasticity is reduced. We could also investigate whether the own fare elasticity values for leisure travel are statistically significantly different from those values for business travel. The simplest way is to look at the values of fare variable’s standard error in the leisure and business models. We observed the standard error values of 0.0002 and 0.0004 from fare variables in leisure and business models respectively. These values are very small; we then expect the ranges of own fare elasticity values (after adjusting fare coefficient by adding and subtracting its standard error) for both leisure and business travel to be minimal. It is then unlikely that the range of own fare elasticity values for leisure travel is overlapping with the range for business travel. Therefore, own fare elasticity values for leisure travel are likely to be statistically significantly different from those values for business travel. We can also see in business travel there is a distinct pattern in differences of cross price and frequency elasticity values between TG and LCCs, and between one LCC and other LCCs. Cross elasticity between TG and LCC is about 50% lower than cross elasticity between the two LCCs. For business travel, Table 4, cross elasticity between TG and LCCs are 0.15, 0.12 and 0.1 in response to change in percentage of fares of NOK, TAA and OTG respectively. However, cross elasticity between NOK and other LCCs (TAA and OTG) are 0.22 and 0.25, and cross elasticity between TAA and other LCCs (NOK and OTG) are 0.24 and 0.25. The pattern reveals that business travelers recognize the competitive distance between two different types of products offered by legacy carrier (TG) and LCCs. In simple words, business travelers recognize product and service differentiation offered between the two carriers. Note that competitive distance is the degree of comparability between different products. Sources of distance come from in-flight entertainment and comfort, service level, frequency, brand, FFP, ticket flexibility and physical product itself. Unlike business travel, in Table 3 cross fare and frequency elasticity values between TG and LCCs and between LCC and LCC are almost the same level for leisure travel. It reveals that leisure travelers do not perceive the competitive distance between the legacy and LCC products. In other words, for leisure short-haul domestic trip, legacy’s frills such as in-flight service and entertainment, FFP reward are not well appreciated. For leisure travelers, it is clear that LCCs offer a strong substitute to the full service airline product. Leisure travelers are willing to trade off frills with lower fares if possible (for short-haul trip). As a result, LCCs would be able to successfully and easily attract leisure travelers with low fare no frills offerings, compared to business travelers.

8 SUMMARY In Asia, the low-cost phenomenon is in its infancy, however, it is growing rapidly especially in Southeast Asia. These LCCs were modeled after prominent European and

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

309

US LCCs. There are two types of LCCs operating in the region, independent operators and legacy incumbent’s low-cost offshoots. We have found that Southeast Asian LCCs can be segmented into two groups: low cost no frills and low fare low frills carriers. Low cost no frills carriers are AirAsia and Tiger Airways, because both strictly adhere to the low-cost recipe; no free frills and unbundling every product and process. Low fare low frills carriers include Nok Air, One-Two-Go, Valuair and Jetstar Asia. A Thai Airways’s offshoot, Nok Air is aimed to be a premium LCCs. The carrier charges fares 10–15% higher than other low-cost competitors with a further introduction of business cabins. One-Two-Go provides free on-board drinks and snacks plus ticket flexibility. Valuair serves meals on board, while Jetstar Asia has pre-paid in-flight entertainment kits. To this point, only the success of low cost no frills carriers can be observed. We observe that AirAsia is the most successful LCC even though it is not the first mover in Asian market. Its solid LCC business model is to strictly adhere to easyJet and Ryanair models which are in turn were adapted from the Southwest model. Currently, AirAsia is the largest intra-Asian LCC in terms of network, traffic and aircraft. The success of Tiger Airways in Singapore, a Ryanair clone, has become apparent after the recent merger of its low-cost rivals, Valuair and Jetstar Asia. Therefore our prediction for low fare low frills carriers is that they would have a more difficult time to succeed and survive when an industry shakes out. Deep-pocket low-cost offshoot such as Nok Air and Jetstar Asia are financially better off than independent low fare low frills carriers such as One-Two-Go. Our business model analysis further reveals that Thai Airways and its low-cost subsidiary Nok Air are subject to cannibalization in the airline within an airline model. Besides their overlap in product offerings and customer target groups in short-haul domestic routes, Nok Air is likely growing at the expense of Thai Airways for several reasons. First, by the nature of country’s geography and demography, both can hardly avoid head-to-head competition. In Thailand, the longest domestic flights are less than 2 hours and there are only a few major cities. As a result, they both operate flights out of Bangkok to the same tourist and business destinations including Chiang Mai, Phuket, Hat Yai, Udon Thani. Second, Nok Air was given aircraft from Thai Airways. As a result, the logo of Thai Airways is overwhelmingly covered on seat cabins and accessories. This allows passengers to perceive that without free in-flight service and meal, Nok Air’s product offering are very much the same as those of Thai Airways. It should be noted that in-flight service and meal would become less important in short-haul domestic trips. Apparently, the airline within an airline model seems to create the opportunity for cannibalization simply because the way Thai Airways and Nok Air have structured the business models and organized their short-haul domestic route and service offerings. Unlike Nok Air, Jetstar Asia can refrain from cannibalizing Qantas’s core business because they are offering distinctive products while also operating in different country bases. In Australia, Jetstar is designed to avoid direct competition with Qantas by serving non-overlapping destinations. Singapore Airlines also clearly put distance between itself and Tiger Airways. Tiger is a real LCC offering low-cost no frills products while Singapore Airlines offers full service frills products. Their product differentiation is distinctive and customer target groups are different.

310

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

In the modeling we found several interesting results. First, for both leisure and business air travel, fare and frequency are significant factors influencing a carrier’s choice probabilities. Our estimated results also show that business travelers place a higher value on time via flight frequency. Moreover, fare has differential impacts on carrier choice probability in accordance with its flight frequency available. For business travel, travelers have to trade off between high fare with high flight frequency offered by Thai Airways. For leisure travel, traveler’s utility on LCCs goes up when these carriers increase their daily frequency while offering low fares. We also found Thai Airways frequency flyer program has positive impacts on the possibilities to travel by the carrier for both leisure and business travels. Unsurprisingly, frequent flyer business travelers attach higher value to this loyalty program, compared to leisure travel in short-haul domestic trips. The individual traveler’s past experiences with LCCs create a significant positive impact on the probability of traveling by LCCs and hence lead to an increasing likelihood of choosing a LCC. Interestingly, past experience with LCCs has more influence for leisure travel than for business travel. Moreover, past experiences on business trips by low-cost carriers positively affects the business travelers’ intention and possibilities to travel by LCCs. Last, ASC terms of both models reveal that travelers do not perceive unobservable attributes across LCCs, but they do distinguish the differences in these attribute between Thai Airways and LCCs. In other words, both leisure and business travelers have equal preference for selecting among LCCs. However, they do have definite preference for selecting Thai Airway above LCCs. For leisure travel, travelers do not differentiate, to a significant degree, the differences in product offerings across LCCs. This finding is supported by the estimated results of ASC variables; that is there is an equal preference for selecting LCCs; perhaps this is reflecting the “loyalty to price rather than carrier” talked about in the literature. Furthermore, leisure travelers do not distinguish product and service offerings across Thai Airways and LCCs. Frills offered by Thai Airways are regarded as insignificant for short-haul domestic leisure trips. Our elasticity results also support the findings in the air transportation literature that leisure travelers place a higher value on fare level than flight frequency. With low fare low/no frills offerings, LCCs will find it easier to penetrate fare sensitive, non-business market. Our results should alert Thai Airways for a potential loss of its leisure domestic market shares. It is thus unavoidable that in the near future Thai Airways will need to strategically focus on medium to long-haul international routes, following its legacy peers such as Singapore Airlines and British Airways. For business travel, passengers do perceive the differences in product offerings between Thai Airways and LCCs, unlike leisure travelers. Frills offered by Thai Airways such as frequency flyer program, ticket flexibility, brand and service level are taken into account for business travel. However, our results further reveal that the degree of demand substitutability is getting higher across Thai Airways and LCCs. It implies that business travelers are becoming more price sensitive for short-haul domestic trips. Perhaps after a careful consideration of fare and quality of products, more and more cost conscious business travelers would choose to fly LCCs as travel budgets may deteriorate due to the economic downturn. This phenomenon needs more time to be visible in Asia. However, recent research has shown that business travelers now represents a sizable market for LCCs in the US and Europe.

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

311

APPENDIX A Table A1 Country 1. North America United States

Canada

Mexico

Low cost airline AirTran Airways Allgiant Airways America West Airlines American Trans Air Frontier Airlines Independence Air JetBlue Airways Primaris Airlines Song Southwest Airlines Spirit Airlines Sun Country Airlines Ted USA 300 Airlines Canjet Harmony Airways Tango Airlines WestJet ZIP Click Mexicana

Note

ex HMY Airways now Air Canada now Air Canada

2. Europe Austria Belgium Czech Republic Denmark Finland France Germany

Hungary Iceland

Niki SN Brussels Airlines Virgin Express Smart Wings Maersk Air Sterling Airlines Bluel Flywest Air Berlin Dba Eurowings Germania Germnwings Hapag-Lloyd Express Wizzair Iceland Express

part of the Lufthansa Regional group wholely owned by Eurowings

(Continued)

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

312

Table A1 Country Ireland

Italy

Low cost airline Aer Arann Aer Lingus Ryanair Air Service Plus Alpi Eagles Evolavia Jet X

Netherlands Norway Poland Romania Serbia and Montenegro Slovakia Spain Sweden

Switzerland Turkey United Kingdom

Note

operating with French Axis Airways

Icelandic company operating in Italy

Meridiana Myair Volare Windjet Transavia Airlines Norwegian Air Shuttle Centralwings Blue-Air Air Maxi Skyeurope Air Madrid Vueling Airlines Snowflake FlyMe FlyNordic Helvetic Airways Onur Air Air Southwest Air Wales BMIbaby EasyJet Flybe Flyglobespan Jet2.com Monarch Airlines My Travel Lite Thomsonfly

3. Asia Hong Kong

India

Hong Kong Express Oasis Airways WOW Asia SpiceJet Air Deccan Air India Express

just launched, subsidiary of Air India

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

313

Table A1 Country

Low cost airline Kingfisher Airways Go Air IndiGo Paramount Airways Awair Citilink Lion Air Adam Air Hokkaido International Airlines Skymark Airlines Skynet Asia Airways Al Jazeera Airways Macao Eagle Air Asia Air Phillippines Asian Spirit Cebu Pacific South East Asian Airlines

Indonesia

Japan

Kuwait Macao Malaysia Philippines

Note just launched lauching by Feb 2006 just launched subsidiary of Air Asia subsidiary of Garuda Indonesia

also known as Air Do

known as WOW Macao

known as SEAIR

Table B1 Definition and Description of Explanatory Variables used in Leisure Model

Attribute

Variable

Description and Definition

Mean

Min

Max

FARE

One-way airplane ticket fare Daily flight frequency Interaction between fare and frequency Destination chosen dummy-Chiang Mai Destination chosen dummy-Phuket Destination chosen dummy-Hat Yai Destination chosen dummy-Chiang Rai Destination chosen dummy-Ubon Ratchathani

2009.90

500.40

3595.00

4.78 10581.72

1.00 650.40

13.00 36075.00

FREQUENC FAREFREQ CHMAI PHUK HYAI CHRAI UBON

0.38

0

1

0.15

0

1

0.18

0

1

0.07

0

1

0.06

0

1

(Continued)

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

314

Table B1 Definition and Description of Explanatory Variables used in Leisure Model—Cont’d Variable

Description and Definition

UDON

Destination chosen dummy-Udon Thani Destination chosen dummy-Surat Thani Interaction between frequency and CHMAI Interaction between frequency and PHUK Interaction between frequency and HYAI Interaction between frequency and CHRAI Interaction between frequency and UBON Interaction between frequency and UDON Interaction between frequency and SURAT Average age of passenger Average income of passenger Dummy-1 if stay at hotel; 0 for otherwise Dummy-1 if someone paid for your trip; 0 for otherwise Dummy-1 if hold TG’s FFP; 0 for otherwise Dummy-1 if had domestic flight with NOK; 0 for otherwise Dummy-1 if had domestic flight with TAA; 0 for otherwise Dummy-1 if had domestic flight with OTG; 0 for otherwise Dummy-1 if had leisure trip with LCCs; 0 for otherwise Dummy-1 if had business trip with LCC; 0 for otherwise

SURAT FCHMAI FPHUK FHYAI FCHRAI FUBON FUDON FSURAT SDGs

AGE INC DLODGE SPSOR1

FFPTG DFWNOK

DFWTAA

DFWOTG

LWLCC

BWLCC

Mean

Min

Max

0.15

0

1

0.01

0

1

2.67

0

13

0.73

0

11

0.68

0

5

0.19

0

4

0.15

0

3

0.34

0

3

0.02

0

2

33.43 34636.91

20 5000

65 90000

0.39

0

1

0.21

0

1

0.38

0

1

0.40

0

1

0.36

0

1

0.33

0

1

0.58

0

1

0.30

0

1

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

315

Table B2 Definition and Description of Explanatory Variables used in Business Model

Attribute

Variable

Description and Definition

Mean

FARE

One-way airplane ticket fare Daily flight frequency Interaction between fare and frequency Destination chosen dummy-Chiang Mai Destination chosen dummy-Phuket Destination chosen dummy-Hat Yai Destination chosen dummy-Chiang Rai Destination chosen dummy-Ubon Ratchathani Destination chosen dummy-Udon Thani Destination chosen dummy-Surat Thani Interaction between frequency and CHMAI Interaction between frequency and PHUK Interaction between frequency and HYAI Interaction between frequency and CHRAI Interaction between frequency and UBON Interaction between frequency and UDON Interaction between frequency and SURAT

FREQUENC FAREFREQ CHMAI PHUK HYAI CHRAI UBON

UDON SURAT FCHMAI FPHUK FHYAI FCHRAI FUBON FUDON FSURAT SDGs

AGE INC DLODGE SPSOR1

FFPTG

Average age of passenger Average income of passenger Dummy-1 if stay at hotel; 0 for otherwise Dummy-1 if someone paid for your trip; 0 for otherwise Dummy-1 if hold TG’s FFP; 0 for otherwise

Min

Max

202937

1364

3775

467 1042548

1 1364

13 49075

034

0

1

018

0

1

023

0

1

007

0

1

007

0

1

009

0

1

003

0

1

242

0

13

079

0

11

085

0

5

018

0

8

017

0

3

020

0

3

006

0

4

3821 4220833

20 5000

65 90000

06225

0

1

074

0

1

055

0

1

(Continued)

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

316

Table B2 Definition and Description of Explanatory Variables used in Leisure Model—Cont’d Variable

Description and Definition

Mean

Min

Max

DFWNOK

Dummy-1 if had domestic flight with NOK; 0 for otherwise Dummy-1 if had domestic flight with TAA; 0 for otherwise Dummy-1 if had domestic flight with OTG; 0 for otherwise Dummy-1 if had leisure trip with LCCs; 0 for otherwise Dummy-1 if had business trip with LCC; 0 for otherwise

0.50

0

1

0.48

0

1

0.40

0

1

0.46

0

1

0.63

0

1

DFWTAA

DFWOTG

LWLCC

BWLCC

Table B3 Estimation Results from MNL Models Variable

Attributes FARE FREQ FAREFREQ (TG) FAREFREQ (LCCs) FCHMAI (TG, OTG) FPHUK (TG) FHYAI (TG) FUDON (TG) FCHRAI (TG) FUBON (TG) FSURAT (TG) SDCs INC FFP (TG) DLODGE (TG) SPSOR1 (TG) ENOK (NOK) ETAA (TAA) EOTG (OTG) BWLCC (LCCs)

Leisure model

Business model

Coefficient

Std. Er.

P-value

Coefficient

Std. Er.

P-value

−00008 00643

00002 00265

00003 00153

−00008 03219 −00001

00004 01551 00000

00410 00379 00056

00001

00000

00099 −02222 −04540 −04423 −11321 −06333 −10057 −06834

01203 01318 02621 04553 02977 04755 02815

00647 00006 00916 00129 00334 00344 00152

00000 12170

00000 03073

05969 00001

14375 09914 12487 16890 05406

04104 03687 03846 04442 03187

00005 00072 00012 00001 00899

00000 09844 04948 09013 14305 13493 17492

00000 02015 02086 02525 02792 02584 03021

00206 00000 00177 00004 00000 00000 00000

LOW-COST BUSINESS MODELS IN ASIAN AVIATION MARKETS

317

Table B3 Estimation Results from MNL Models—Cont’d Variable

ASCs ASCTG ASCNOK ASCTAA LL fn Pseudo R Sample size Note:

∗

Leisure model

Business model

Coefficient

Std. Er.

P-value

Coefficient

Std. Er.

P-value

19411 00772 −03855 −4864247 03743 566

05415 03530 03496

00003 08269 02701

53980 03432 −04231 −2692676 04354 344

16722 05131 04875

00012 05035 03855

Not significantly different from zero at the 95 confidence level, Std. Er = standard error.

REFERENCES Baker, C., Field, D., and Ionides, N. (2005). “Global reach.” Airline Business, 60. Barkin, I. T., O. Staffan Hertzell, and Young, S. J. (1995). “Facing low cost competitors.” The McKinsey Quarterly(4), 87–99. Ben-Akira, M. and S. Lerman. (1985). Discrete Choice Analysis: Theory and Application to Travel Demand, MIT Press, Boston, Massachusetts. Blum, A. (2005). “JetBlue’s Terminal Takes Wing.” Business Week Online. Franke, M. (2003). “Dawning of a New Airline Business Model:New Service Offerings for Changed Demand.” The 6th Hamburg Aviation Conference, Hamburg. Franke, M. (2004). “Competition between network carriers and low-cost carriers–retreat battle or breakthrough to a new level of efficiency?” Journal of Air Transport Management, 10(1), 15–21. Gillen, D. W., Morrison, G.W., and Christopher Stewart. (2002). “Air travel demand elasticities: concepts, issues and measurement.” Report to Department of Finance, Canada. Gillen, D., and Lall, A. (2004). “Competitive advantage of low-cost carriers: some implications for airports.” Journal of Air Transport Management, 10(1), 41–50. Hensher, D. A., Rose, M.J., and W.H. Greene. (2005). Applied choice analysis: A primer, Cambridge University Press, Cambridge. Hooper, P. (2005). “The environment for Southeast Asia’s new and evolving airlines.” Journal of Air Transport Management, 11(5), 335–347. Lawton, T. C. (2002). Cleared for take-off: structure and strategy in the low fare airline business, Ashgate Publishing Ltd., Hants. Louviere, J. J., Hensher, D.A., and Swait, J. (2000). State choice methods: analysis and appli­ cations in marketing, transportation and environment valuation, Cambridge University Press, Cambridge. Mason, K. J. (2001). “Marketing low-cost airline services to business travellers.” Journal of Air Transport Management, 7(2), 103–109. McFadden, D. (1978). Modeling the choice of residential location, North-Holland, Amsterdam. O’Connell, J. F., and Williams, G. (2005). “Passengers’ perceptions of low cost airlines and full service carriers: A case study involving Ryanair, Aer Lingus, Air Asia and Malaysia Airlines.” Journal of Air Transport Management, 11(4), 259–272.

318

DAVID GILLEN AND NATTHIDA TAWEELERTKUNTHON

Porter, M. E. (1996). “What is strategy?” Harvard business review (November/December), 61–78. Taweelertkunthon, N. (2006). “The Empirical valuation of the potential success in low cost business model in Asian aviation market,” Unpublished MSc thesis, Sauder School of Business, University of British Columbia. Thomas, G. (2003). “In tune with low fares in Malaysia.” Air Transport World, 45–46. Thomas, G.(2005) “The Undiscovered Country.” Air Transport World, 42. Wen, C. H., Koppelman, F.S. (2001). “The generalized nested logit model.” Transportation Research Part B, 35, 627–641.

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

14 Pricing Strategies by European Traditional and Low Cost Airlines: Or, when is it the Best Time to Book on Line? Claudio A. Piga∗ , Enrico Bachis†

ABSTRACT It is often assumed that the airlines’ fares increase monotonically over time, peaking a few days before the departure. Using fares for about 650 thousand flights operated by both low-cost and full-service carriers (FSCs), we show several instances in which the monotonic property does not hold. We also show that the volatility of fares increase in the last 4 weeks before departure, which is the period when the airlines can formulate a better prediction for a flight’s load factor. Finally, especially within the last 2 weeks, FSCs may offer lower fares than those posted by low-cost carriers.

1 INTRODUCTION The conventional wisdom on the temporal profile of airlines’ prices holds that carriers facing uncertain demand can enhance their profits by assigning a monotonically increas­ ing price to sequential booking classes (Dana, 2001). McGill and Van Ryzin (1999) refer to such a practice as “Low-before-high fares” and argue that it is based on the assumption that booking requests arrive in strict fare sequence, generally from lowest to highest as the flight departure approaches. Belobaba (1987) discusses how such an assumption on booking behavior is generally found to lead to higher revenues. The use of ∗ Corresponding author. Economics Department, Loughborough University, Leicestershire, LE11 3TU, UK, Tel.: +44 (0) 1509 222701, Fax: +44 (0) 1509 223910, e-mail: [email protected]. Piga gratefully acknowledges receipt of the British Academy Larger Research Grant LRG-35378. † Nottingham University Business School, Wollaton Road, Nottingham, NG8 1BB, UK, e-mail: [email protected].

320

CLAUDIO A. PIGA AND ENRICO BACHIS

monotonic fares is further motivated by the fact that airlines have to set their price before demand is known, and that the transaction costs of adjusting prices in response to current information about demand are high. This is because, for example, even in the simple case where a single flight leg is considered in isolation for seat inventory purposes, the evaluation of a booking request is a complex one, as it involves passengers with many different origin–destination itineraries on the same aircraft, all of whom generate different amount of revenues (Belobaba, 1987).1 The development of hub-and-spoke has exacerbated the complexity of the seat inventory problem, even at the single-leg level of analysis. It is thus important to verify if the monotonic property is observed in situations where transaction costs are less relevant, that is, for fares offered on the Internet by some European low-cost carriers (LCCs) that, unlike full-service carriers (FSCs), do not adopt a hub-and-spoke network and only operate airport-to-airport flights. In this chapter, we use price information for about 650 thousands flights, collected from both the most important European LCCs’ web sites and an online travel agent. The latter was used to retrieve prices for flights operated by traditional carriers. An important characteristic of our dataset is that for each flight operated by a LCC, we have up to 13 fares collected at regular intervals before departure, while for FSCs, we could collect up to 9 fares posted before the flight’s departure. For each advertised fare, a strictly positive number of seats was available for purchase, that is, a fare corresponds to the minimum price an airline is willing to accept in exchange for a seat on a given flight. Although other studies have used fares for the same flight posted a number of days prior to departure (see, for instance, Pels and Rietfeld, 2004; Pitfield, 2005), these were usually for a very limited number of routes. A distinguishing feature of our approach is that we retrieved data for a large proportion of routes operated by the main LCCs, where they either are monopolists or face competition by other LCCs or FSCs. Therefore our analysis, by focusing on the features which are common across routes, enables us to draw general conclusions on the different online pricing schemes adopted by the airlines in our sample. To this purpose, we present evidence that we directly relate to some recent contribu­ tions on pricing in electronic markets. More precisely, we try to address the following issues: • Do fares increase monotonically as the date of departure approaches (Klein and Loebbecke, 2003)? • Do fares change frequently, that is, are there menu costs online (Brynjolfsson and Smith, 2000; Smith et al., 1999)? • Is it true that LCCs always offer the cheapest fares, or, consistent with a hit-and­ run strategy, there is significant variation in the identity of the low-price firm in a competitive route (Baye et al., 2004b)? While we observe consistent hikes for fares posted less than a week prior to departure, we also find that for some airlines the early booking fares may be higher than those available from 4 to 2 weeks prior to departure. It would therefore seem that the monotonic

1

Even within the same flight, the yield management system has to decide the desirability (in revenue terms) of selling a seat at a higher price to a single-leg passenger as opposed to selling that seat at a lower price to a multileg passenger, thereby generating a higher revenue (Belobaba, 1987).

PRICING STRATEGIES

321

property does not adequately and fully describe the time profile of many LCCs’ pricing schemes. This is probably related to the easiness with which fares can be changed online due to low menu costs (Smith et al., 1999). We also find that prices tend to remain more stable when departure is further away in the future, while volatility increases as the date of departure approaches. Furthermore, FSCs tend to change their fares more frequently than LCCs. This is somewhat surprising, given the higher transaction costs incurred by the FSCs in managing complex hub-and-spoke networks. However, our findings indicate that high transaction costs are more than offset by the adoption of more sophisticated revenue management techniques. As far as the identity of the low-price firm is concerned, the evidence confirms the notion that a LCC offers the lowest price in a competitive route. However, consider­ able differences are observed across LCCs, especially in conjunction to the distinction between early and late booking fares. Indeed, the same airline may offer the cheapest early booking fares but the dearest late booking fares. Furthermore, we show that a few days before departure, it is not unlikely that a FSC may offer the cheapest fare. The strategy used to collect the data has exploited some of the innovative features of the pricing schemes followed by LCCs (Pender and Baum, 2000; Piga and Filippi, 2002; Shon et al., 2003). For instance, consider some of the forms of price discrimination that have dominated the industry and that are highly documented in the literature: the Saturday night stay-over requirement, the surcharge for one-way tickets, and the advancepurchase discounts (Stavins, 2001; Giaume and Guillou, 2004). The European LCCs have eliminated completely the first two forms: for example, departing on a Monday and returning on a Thursday is likely to cost less than returning on a Sunday. In any case, each leg is priced independently, and the same price would be shown online for the Monday flight if one tried to book a one-way ticket. An important implication of pricing each leg independently without any penalty for one-way tickets is the need to deepen our understanding of the links between airport dominance and market power. Indeed, the European LCCs in our sample do not practice any of the marketing strategies, that is, Frequent Flyer Program, Travel Agents’ Commissions Override program, and so on, indicated as the source of airport dominance (Borenstein, 1989 and 1991). Moreover, the aforementioned invariance of fares when booking a round-trip or two one-way tickets drastically limits the possibility to take advantage of airport dominance by charging a higher fare.2 As far as the advance-purchase discount tactic is concerned, it arises from the air­ lines’ need to derive schemes in situations where demands or travelers’ preferences are uncertain (Gale and Holmes, 1993; Dana, 2001). A central contribution of this chapter is to show that the LCCs follow it in a rather flexible manner. It is also widely known that all European LCCs offer “no-frills” flights with no service distinction for seats, thus excluding any form of discrimination based on quality (Mussa

2

Assume that airline 1 dominates airport A, and airline 2 airport B. Both airlines operate a service on the route A–B. Absent the penalty and all other flights’ characteristics equal, charging each leg independently intensifies competition on each segment of the route, as each company has an incentive to steal customers from the other. A Bertrand equilibrium would then be the outcome. In the traditional case, charging a penalty for one-way tickets effectively segments the two markets, with airline 1 offering the service A–B–A, and airline 2 the service B–A–B. Implicitly, we are assuming that the penalty may facilitate tacit collusion.

322

CLAUDIO A. PIGA AND ENRICO BACHIS

and Rosen, 1978).3 Finally, price variations due to the inclusion of connecting flights are ruled out by the fact that LCCs issue only “point-to-point” tickets (Clemons et al., 2002). In Section 2, we illustrate the data collection strategy. This is followed by three sections where we address the questions listed above and present some supporting evidence. Some business strategy implications are discussed in the concluding section.

2 DATA COLLECTION Most of the empirical contributions on pricing behavior in the Civil Aviation industry have focused, so far, mainly on the US market. Such studies have been mostly conducted relying on different cohorts of the same dataset, namely the Databank of the US Depart­ ment of Transportation’s Origin and Destination Survey, which is a 10 per cent quarterly random sample of all tickets that originate in the US on US carriers (Borenstein, 1989 and 1991; Evans and Kessides, 1993; Borenstein and Rose, 1994; Hayes and Ross, 1998; Stavins, 2001). In these studies, prices are measured as one-way fares and are computed as one-half of the reported fare round-trip tickets. All tickets other than one-way and round trips are excluded. In contrast, our analysis is based on primary data on fares and secondary data on routes traffic. Since the start of this research project in May 2002, fares were collected using an “electronic spider,” which connected directly to the web sites of only the main LCCs (i.e., Ryan Air, Buzz, Easyjet, GoFly) operating in Great Britain at the time. Collection of fares for flights operated by FSCs (i.e., British Airways, Air Lingus, Air France, Lufthansa, KLM, Alitalia, Iberia, SAS, Tap Portugal, Air Europa, and Maersk) started in March 2003: In this case, fares were collected only for flights that had FSCs operated on routes similar or identical to those where a LCC also flew.4 This decision was necessary to reduce the number of routes under study, where each route is identified in this study as an airport pair combination. The dataset includes daily flights information from June 2002 up to, and including, June 2004, for a total of 25 months. Over such a period, a number of important events took place, which are reflected in the dataset. First, a series of takeovers occurred: Easyjet acquired GoFly (December 2002) and Ryan Air took over Buzz (March 2003). Second, new LCCs began their operations: the “spider” was upgraded to retrieve fares from the BMIbaby and MyTravelLite sites. However, due to technical difficulties, fares from Flybe, which was already an established LCC, and Thomson Fly, a new entrant, could not be obtained. Over the 25 months period, fares from UK for flights to and from the following Euroadopting countries were obtained: Austria, Belgium, France, Germany, Greece, Ireland, Italy, Netherlands, Portugal, and Spain. A term of comparison is provided by fares

3 Vertical product differentiation can be found among US LCCs. JetBlue, for example, has positioned itself

as a carrier with LCC prices but a relatively high level of amenities.

4 The airfares of the traditional companies were collected from the website www.opodo.co.uk, which is owned

and managed by British Airways, Air France, Alitalia, Iberia, KLM, Lufthansa, Aer Lingus, Austrian Airlines,

Finnair, and the global distribution system Amadeus. Thus, fares listed on Opodo are the official prices of

each airline, although Opodo may not report promotional offers that each airline may offer on their web sites.

PRICING STRATEGIES

323

for flights to the following countries outside the Euro area: Czech Republic, Norway, Sweden, Switzerland as well as UK, whose domestic routes were also considered. In order to account for the variety of fares offered by airlines at different times prior to departure, every day we instructed the spider to collect the fares for departures due, respectively, 1, 4, 7, 10, 14, 21, 28, 35, 42, 49, 56, 63, and 70 days from the date of the query. Henceforth, these will be referred to as “booking days.” So, for instance, if we consider London Stansted–Rome Ciampino as the route of interest, and assume the query for the flights operated by a given airline was carried out, say, on 1 March 2004, the spider would retrieve the prices for both the London Stansted–Rome Ciampino and the Rome Ciampino–London Stansted routes for departures on 2/3/04, 5/3/04, 8/3/04, 11/3/04, and so on. We repeated this search every day. To continue the previous example, on 4 March 2004, we would retrieve the fare available one day before departure for a flight scheduled on 5/03/04, the fare available 7 days before departure for a flight due to depart on 11/03/04, and so on. By doing so, for every daily flight, we managed to obtain up to 13 prices that differ by the time interval from the day of departure. The main reason to do so was to satisfy the need to identify the evolution of fares – from more than 2 months prior to departure to the day before departure – which has been noted to be very variable for the case of LCCs (Giaume and Guillou, 2004; Pels and Rietveld, 2004). The return flight for both types of directional journey was scheduled one week after the departure. For those routes where an airline operates more than one flight per day, all fares for every flight were collected. While the spider could have retrieved any number of prices, in practice the need to reduce both the number of queries made to an airline server and the time of program execution to a manageable level, led to the design above. Furthermore, given the site characteristics of Opodo, it was impossible to collect FSCs’ fares 1 and 4 days prior to departure: It was also decided to omit collecting fares from these companies for flights due to depart more than 49 days after the query. Thus, for FSCs, up to eight fares per daily flight are available. It is important to stress that for LCCs and FSCs alike, we consider online posted prices that may not have led to any purchase of tickets: This is a main difference from the US studies using the DOT O&D Survey, which includes actual transacted tickets. The collection of the airfares has been carried out everyday at the same time: In addition to airfares, we collected the name of the company, the time and date of the query, the departure date, the scheduled departure and arrival time, the origin and destination airports, and the flight identification code. Fares were collected before tax and handling fees. Furthermore, fares for LCCs were one-way, while those for FSC were for a roundtrip and were therefore halved. It is widely known that fares levels may depend on such ticket restrictions as refundability and changeability status. Opodo offered fares for nonchangeable, nonrefundable tickets. These belong to the cheapest available fare class and were chosen to facilitate comparison with the fares by LCCs. Unlike some US LCCs, the European ones initially published a single homogenous class of fares, which carried equivalent restrictions to those obtained from Opodo. That is, online customers were not offered the choice between different combinations of fares and restrictions. The LCCs in our sample maintained the same display of fares throughout the sample period. However, towards the end, some LCCs, namely Ryan Air and Easyjet, introduced the possibility to change a ticket, subject to a fixed penalty and the payment of any fare difference.

CLAUDIO A. PIGA AND ENRICO BACHIS

324

This new strategy does not generally impinge on the following analysis, except in the case where fares levels by LCCs and FSCs are compared. It is likely, however, that the benefits from a possible change of ticket are balanced against the payment of the fixed fee, leaving the relative comparability of LCCs’ and FSCs’ fares practically unchanged. To complement the price data with market structure characteristics, secondary data on the traffic for all the routes and all the airlines flying to the countries indicated above was obtained from the UK Civil Aviation Authority (henceforth, CAA).5 For each combination of company, route and departure period (i.e., month/year), the CAA provided the number of monthly seats, the number of monthly passengers, and the monthly load factors. These were broken down at the flight identification code level, that is, for each flight each airline operated in a given month on a route. However, in order to create a more balanced panel, fares and traffic statistics were aggregated at the route level for each airline. Table 1 illustrates how the data retrieved from the Internet represent an accurate sample of the activity of each of the LCCs in the markets we consider. It compares the Table 1 Number of Routes by Type of Sample, Airline and Period

Period

Jul 2002 Aug 2002 Sep 2002 Oct 2002 Nov 2002 Dec 2002 Jan 2003 Feb 2003 Mar 2003 Apr 2003 May 2003 Jun 2003 Jul 2003 Aug 2003 Sep 2003 Oct 2003 Nov 2003 Dec 2003 Jan 2004 Feb 2004

5

BMIbaby

Ryanair

Easyjet

Routes Routes Comp. Price CAA Routes Sample Sample CAA Sample

Routes Routes Comp. Price CAA Routes Sample Sample CAA Sample

Routes Routes Comp. Price CAA Routes Sample Sample CAA Sample

– – – – – – 26 26 30 26 31 32 33 34 35 35 37 38 33 36

See www.caa.co.uk.

– – – – – – 35 35 37 37 40 43 45 45 44 48 42 47 49 47

– – – – – – 10 11 12 9 10 10 11 11 11 13 12 15 15 14

34 37 37 37 37 37 49 50 50 56 69 69 69 83 83 84 87 87 42 84

59 59 59 59 60 60 61 64 64 65 88 88 88 89 89 92 93 94 98 94

7 8 7 7 8 8 9 7 7 7 6 6 6 8 6 8 8 8 8 8

19 19 28 28 29 61 61 63 66 66 67 67 67 88 88 89 88 88 46 88

38 38 40 41 41 79 80 82 84 88 89 89 89 92 92 96 95 98 98 98

9 9 9 10 9 20 20 21 22 19 19 20 21 24 23 26 23 25 25 25

PRICING STRATEGIES

325

Table 1 Number of Routes by Type of Sample, Airline and Period—Cont’d

Period

Mar 2004 Apr 2004 May 2004 Jun 2004

BMIbaby

Ryanair

Easyjet

Routes Routes Comp. Price CAA Routes Sample Sample CAA Sample

Routes Routes Comp. Price CAA Routes Sample Sample CAA Sample

Routes Routes Comp. Price CAA Routes Sample Sample CAA Sample

38 34 34 34

43 48 50 55

13 17 16 18

84 87 81 84

Buzz Jul 2002 Aug 2002 Sep 2002 Oct 2002 Nov 2002 Dec 2002 Jan 2003 Feb 2003 Mar 2003 Dec 2003 Jan 2004 Feb 2004 Mar 2004 Apr 2004 May 2004 Jun 2004

21 21 21 21 20 22 22 22 22 – – – – – – –

33 33 33 32 20 22 22 21 26 – – – – – – –

94 99 94 96

8 10 9 9

89 89 89 88

GoFly 3 5 5 5 0 0 1 0 4 – – – – – – –

17 17 30 30 32 32 – – – – – – – – – –

37 37 35 39 38 38 – – – – – – – – – –

101 107 110 114

25 27 27 29

MyTravelLite 11 11 9 11 11 11 – – – – – – – – – –

– – – – – – – – – 13 13 13 13 13 10 9

– – – – – – – – – 14 14 13 11 11 9 9

– – – – – – – – – 5 5 5 4 4 3 3

Source: Price sample is retrieved from the airlines’ web sites, Total routes and competitive routes are from the Civil Aviation Authority dataset.

number of routes for which we have price data with the actual total number of routes by each airline. The latter figure is taken from the CAA dataset, which also provides the number of routes where our LCCs face competition by either a major FSC or another LCC. To test the spider’s functionality, initially we limited the number of surveyed routes. Indeed, in August 2002, the percentage of routes with prices was 63% (37 over 59) of the total number operated by Ryan Air, 50% for Easyjet, 64% for Buzz, and 46% for GoFly. However, thanks to the speed of the program, within a few months such percentages could be increased significantly for all the airlines, to cover 80% or more of the total routes they operated. Considering that the spider took all the prices for all the daily flights, the price dataset provides an exhaustive illustration of the online pricing activity of each airline. Table 1 also shows that the airlines differ in the amount of competition they face. For instance, in about 25% of EasyJet’s routes at least another competitor (FSC or LCC) is also present. At the other extreme, Ryan Air (and Buzz to a lesser extent)

326

CLAUDIO A. PIGA AND ENRICO BACHIS

faced competition in a very limited subset of routes. The other airlines lie somewhere in between, with competitive routes accounting for about 33% of the total. Such differences may be driven by the choice of the arrival destinations. Ryan Air and Buzz chose almost exclusively secondary airports that may be many miles away from the city of arrival, while the other airlines also fly to major airports where FSC also land.

3 THE TEMPORAL PROFILE OF FARES Existing theoretical literature like Gale and Holmes (1992, 1993) and Dana (1998, 2001) state various reasons for the airlines to offer lower-priced seats to earlier purchasers. Gale and Holmes (1993) use a mechanism design approach to explain the adoption of Advance-Purchase Discounts (APD) in a monopoly model with capacity constraints and perfectly predictable demand. They show that firms using APD practice a form of seconddegree price discrimination where travelers self-select according to their preference for a peak or an off-peak flight. The advice they offer is to set a low fare for the off-peak flight at an early stage. Such a result hinges around the assumption of certain demand, implying that the airlines can tell an off-peak flight from a peak one. Moreover, its practical implications are that we should expect a flatter temporal profile of fares for the peak flights, and a monotonically increasing profile for the off-peak flights. With demand uncertainty, Gale and Holmes (1992) show that APD can promote efficiency by spreading consumers evenly across flights before timing of the peak period is known. The implication is that ex-post, both types of flight will exhibit a monotonically increasing time profile. For competitive markets where firms set prices before demand in known, Dana (1998) shows that firms may offer APD because travelers with more certain demand and weaker departure time preferences are better off buying in advance due to the presence of other consumers with higher valuations and more uncertain aggregate demand. To drive this result is the assumption that the airlines commit to a rationing rule that limits the number of cheaper seats and thus reduces the incentive of consumers with more certain demand to postpone purchase. In practice, a central aspect of yield management is to deal with the decision to set the limits of each booking category. Consider, for instance, the simple two-fare, seat inventory rule known as the Littlewood’s rule (Belobaba, 1987; McGill and Van Ryzin, 1999). It states that discount fare bookings should be accepted as long as their revenue value exceeds the expected revenue of future full fare bookings. That is f2 ≥ P1 S1 f1 where f2 and f1 are the lower and higher fare, respectively, and P1 S1  is the probability of selling all remaining seats to high-fare passengers. The smallest value of S1 that satisfies the above condition provides the low-fare booking class limit, S2 = C − S1 , where C is the flight capacity. From an empirical viewpoint, price dispersion may be due to such forms of intertem­ poral price discrimination as the APD, as well as to systematic and stochastic peak-load pricing (Borenstein and Rose, 1994). Using an original dataset with information on seats availability at each advertised fare, Escobari (2006) finds that the difference in prices

PRICING STRATEGIES

327

Table 2 Mean Fares by Airline and Booking Days. Fares in British Sterling Booking days before departure 1 4 7 10 14 21 28 35 42 49 56 63 70

BMIBaby

Ryanair

Easyjet

Buzz

GoFly

N

Fare

N

Fare

N

Fare

N

Fare

N

Fare

24659 25073 25919 25802 25725 25292 24729 24171 23498 22852 18742 21563 21299

68.5 65.3 60.2 58.4 56.7 54.5 54.2 53.0 47.6 49.8 49.8 48.2 45.8

233880 245460 245583 240143 238375 225967 226578 227183 210423 210751 200317 197461 197253

99.4 79.0 63.2 58.5 51.2 45.4 42.7 41.2 42.1 38.2 37.3 37.8 35.9

170810 163589 177324 160884 164496 162229 159942 157437 151369 152408 143147 140777 141639

91.0 72.9 62.9 56.7 59.3 55.7 57.5 56.2 53.1 51.0 50.3 49.6 47.7

6617 5889 6842 6012 6647 6441 6262 5983 5840 5642 5381 5155 4874

84.0 76.8 72.5 68.1 66.6 62.3 59.0 56.5 55.0 53.5 51.7 49.5 48.0

11708 10804 15816 11106 15345 11464 11041 10561 10049 6197 5812 8856 8315

88.5 71.8 62.6 67.3 58.9 64.6 62.4 59.9 57.9 70.5 69.5 52.7 50.6

MTL N

Fare

2841 107.1 2837 82.6 2972 56.1 2923 54.0 2871 52.7 2764 51.6 2665 51.2 2560 50.6 2402 50.8 2210 51.3 1978 52.4 1819 53.6 1610 53.9

Source: Data retrieved from the airlines’ web sites from June 2002 until June 2004.

paid between earlier purchasers and later purchasers is mostly explained by flights’ capacity constraints, that is, peak-load pricing seems to be a highly relevant factor in the determination of the shape of the temporal profile of airlines’ fares. In the present setting, lack of data on a flight’s load factor at the time the fares were retrieved makes it impossible to distinguish the factors behind the temporal price dispersion. The size of our dataset however enables us to address the commonly held belief that fares increase as the date of departure approaches. Recall how for each flight we obtained up to 13 fares, collected at regular intervals from 70 days up to the day prior to departure. Table 2 shows the mean fares across all routes by booking day and airlines. Even when using such a highly aggregate measure, some interesting features arise. For BMIbaby, the mean price offered 42 days before departure is lower than that posted in the three preceding weeks. Subsequently, BMIbaby’s fares increase monotonically, but in a rather flat manner. One could still argue that discounts offered 42 days before departure may still be considered within the advance-purchase category, but in any case, we find the first violation of the monotonic property. Furthermore, considering that mean fares increase only by 5.4 British Sterling between 35 and 10 days prior to departure, it is also possible that for a number of flights within this period the monotonic property did not hold. As far as Ryan Air’s mean fares are concerned, similarly to BMIbaby, we observe marginally lower fares 35 days prior to departure, relative to those posted a week earlier. However, Ryan Air presents a steeper time profile in the week preceding departure, where the fare available 4 days before departure is more than twice as high as that available 2 months before the flight departs. Such a finding is both consistent with an explanation based on the fact that Ryan Air realizes higher load factors (hence, the higher fares are due to peak-load pricing) or simply due to its willingness to implement a second-degree form of price discrimination as suggested in the previous discussion.

CLAUDIO A. PIGA AND ENRICO BACHIS

328

The mean fares shown in Table 2 for Easyjet provide compelling evidence that the monotonic property does not always constitute a precise way to describe the evolution of fares. Indeed, those posted 21 and 10 days are lower than those available in the preceding booking day. Interestingly, this is consistent with similar evidence for the same airline shown in Pitfield (2005) and Pels and Rietveld (2004). The combination of our evidence, which includes a high number of routes, with that collected on specific routes seems to suggest the existence of a company fixed effect playing a crucial role in the shaping of the temporal price profile in a route. GoFly is another airlines whose fares’ time profile violates the monotonic property in various instances. Indeed, fares available 42, 14, and 7 days before departure are lower than those in the preceding periods. The column for MyTravelLite (MTL) exhibits another type of pricing behavior, one that tends to be U-shaped. For this airline, we record fares that decline every week from 70 up to 35 days before departure, and then rise at first rather smoothly but then quite sharply in the last 4 days. Finally, Buzz seems to be the only LCC in our sample whose aggregate fares follow a monotonically increasing, although at a relatively small gradient, path. Figures 1 and 2 show clearly how the conclusions from Table 2 continue to hold when we consider specific routes. Each box in these figures provides a graphical summary of the distribution of the fares for each booking day. We focus the attention on the line inside each box, which represents the median of the distribution (the lower hinge in the box represent the 25th percentile, the top hinge the 75th percentile). It is evident how the monotonic property is often violated. For instance, in Figure 1 the median price available 10 days prior to departure is below than that of the immediately preceding days. Interestingly, the 10 days fare is at the level of the 49 days, 56 days, and 63 days median prices. In Figure 2 the median prices offered 4 and 63 days before departure are of a similar magnitude: Within this time interval, fares fluctuate widely. The evidence presented so far suggests that for the LCCs in our sample the commonly held belief of airlines’ fares increasing over time misrepresents the actual temporal

100 50 0

Fares Distribution

150

Go Fly - London Stansted-Faro

d70 d63 d56 d49 d42 d35 d28 d21 d14 d10 Number of days prior to departure

d7

d4

d1

Figure 1 The Temporal Profile of Fares Offered by Easyjet on a Specific Route.

PRICING STRATEGIES

329

50 0

Fares Distribution

100

Easyjet - Edinburgh Amsterdam

d70 d63 d56 d49 d42 d35 d28 d21 d14 d10 Number of days prior to departure

d7

d4

d1

Figure 2 The Temporal Profile of Fares Offered by Go Fly on a Specific Route.

profile of fares. Although the lack of sales data at each point in time does not allow us to identify the reasons behind price dispersion, our findings suggest a rather complex relationship between fares and load factors. Indeed, exceptions to the monotonic rules often occur between 2 and 1 month before departure: these may be considered advancepurchase discounts. However, they are also observed only a few days before departure, something that is hard to reconcile with the recommendations of theoretical models. In the Section 4 we will further investigate the extent to which fares change over the booking period. Quite interestingly, there seems to be a mismatch between what the airlines preach and what they practice. Consider for instance this quote, taken from http://www.easyjet.com/EN/Book/aboutourfares.html: “In general, our fares increase as the departure date gets nearer, so to get the best deals make sure you book as early as possible.” Ryan Air allows itself more leeway when it states: “Our lowest fares gen­ erally require an advance purchase of 14 days; however this can vary up to 28 days.” Both quotes suggest a preference for early bookings, possibly because of the associated financial benefits as a ticket purchased 60 days before departure is equivalent to an interest-free loan. In the Section 4 we show that such a preference for early booking may be one of the possible reasons behind the frequency with which fares change from one booking day to another.

4 MENU COSTS ON LINE? As discussed in Smith et al. (1999), in an electronic market menu costs, that is, the costs incurred by retailers in making price changes, should be lower, since they are comprised primarily of the cost of making a single price change in a central database. Evidence supporting this hypothesis can be found in Brynjolfsson and Smith (2000),

CLAUDIO A. PIGA AND ENRICO BACHIS

330

Table 3 Percentage of Times in which each Company Changes the Airfare Offered on its Website (Number of Changes in Prices, Flight by Flight, Divided by Total Number of Potential Price Changes). Company

Mean

Standard Deviation

Number of observations

Bmibaby Ryanair Easjet Buzz GoFly MyTravelLite Aer Lingus Air Europa Air France Alitalia BMI British Airways Czech Airlines Finnair Iberia KLM Lufthansa MaerskAir Scandinavian Airlines Swiss TapPortugal Volare

0.44 0.65 0.56 0.50 0.39 0.62 0.57 0.54 0.67 0.81 0.78 0.77 0.80 0.87 0.72 0.90 0.76 0.79 0.90 0.82 0.75 0.85

0.185 0.198 0.191 0.233 0.283 0.215 0.278 0.260 0.274 0.139 0.268 0.264 0.204 0.192 0.302 0.199 0.241 0.224 0.213 0.234 0.240 0.188

29248 268584 181644 6987 12466 3179 8046 296 8061 6448 33068 35683 547 1298 4896 6624 18597 683 12033 4228 126 270

.63

.234

643012

Total

where it is shown that (1) online retailers change prices of books and CDs more fre­ quently than their conventional counterparts, and (2) changes are often of negligible size online. In Table 3, we show the percentage of times in which we observe a fare change from one booking day to the next one. That is, for each flight we first calculate all the fare changes between two consecutive booking days: for example, between the 70 and 63 days fares, between 63 and 56 days fares, and so on. Since we have fares for 13 booking days (see above, Section 2), we can have at most 12 possible fare changes for each flight. For each airline, we then identify the total number of possible changes for all its flights in the database, and work out the averages in Table 3. Among the LCCs, Ryan Air and MyTravelLite are the ones that are more likely to revise their posted fares with, respectively, a probability of 65% and 62% that two consecutive fares may not be identical. For such LCCs as Go Fly and BMIbaby, however, fares tend to change less frequently. Operating a more complex network increases the complexity of the decision of whether a fare for a flight in a given route should be revised. Given the associated transaction costs, the FSCs have to adopt more sophisticated yield management techniques, which may be why in Table 3 most FSCs exhibit a stronger tendency to revise their fares,

PRICING STRATEGIES

331

Percentage of times in which fares change within the period (%)

with probabilities ranging from 77% for British Airways up to 90% for Scandinavian. Although a number of other factors may be responsible for such observed variability, our main aim in this chapter is to present evidence that is consistent with the hypothesis of small menu costs. Table 3 clearly provides support to this hypothesis: what it does not reveal is when fare changes are more likely to occur, something we now address. In this sense, the following analysis is directly related to that presented in the Section 3 on the temporal profile on fares. Figures 3–7 illustrate the percentage of flights in which fares between two booking periods either remained the same or changed by either increasing or decreasing. Flights are for the year 2003. Figure 3 considers the case for fares posted by BMIbaby. About

80 70 60 50 40 30 20 10 0 70–63

63–56

56–49

49–42

42–35

35–28

28–21

21–14

14–7

7–1

Period between consecutive booking days decrease

stable

increase

Percentage of time in which airfares change within the period (%)

Figure 3 BMIbaby, Year 2003: Percentage of Times in which Fares Change or Remain Constant over two Consecutive Booking Days.

100 90 80 70 60 50 40 30 20 10 0 70–63

63–56

56–49

49–42

42–35

35–28

28–21

21–14

14–7

7–1

Period between consecutive booking days decrease

stable

increase

Figure 4 Ryan Air, Year 2003: Percentage of Times in which Fares Change or Remain Constant over two Consecutive Booking Days.

CLAUDIO A. PIGA AND ENRICO BACHIS

Percentage of times in which airfares change within the period (%)

332

120 100 80 60 40 20 0 70–63

63–56

56–49

49–42

42–35

35–28

28–21

21–14

14–7

7–1

Period between consecutive booking days decrease

stable

increase

Percentage of times in which airfares change within the period (%)

Figure 5 EasyJet, Year 2003: Percentage of Times in Which Fares Change or Remain Constant over two Consecutive Booking Days.

70 60 50 40 30 20 10 0 70–63

63–56

56–49

49–42

42–35

35–28

28–21

21–14

14–7

7–1

Period between consecutive booking days decrease

stable

increase

Figure 6 BMI British Midlands plc, Summer 2003: Percentage of Times in which Fares Change or Remain Constant over Two Consecutive Booking Days.

65% of times, there is no change between 70 and 63 days: The average fare from Table 4 is £49.16. About 22% and 13% of times fares increased or decreased in this time interval, when mean decreases amount to £9.75, while increases to £10.38. A characteristic of BMIbaby is that the probability of a price reduction remains relatively stable (and not very high) throughout the whole booking period while price increases become more likely as the date of departure approaches. However, about 35% of flights keep their fares stable in the last week. The relative stability of fares for BMIbaby seems to suggest little

Percentage of times in which fares change within the period (%)

PRICING STRATEGIES

333

90 80 70 60 50 40 30 20 10 0 63–56

56–49

49–42

42–35

35–28

28–21

21–14

14–7

Period between consecutive booking days decrease

stable

increase

Figure 7 British Airways, Summer 2003: Percentage of Times in which Fares Change or Remain Constant over Two Consecutive Booking Days.

role for peak-load pricing and a pricing scheme that is independent of the realization of current load factors. The case for Ryan Air, illustrated in Figure 4, tells a rather different story. The highest percentage of stable fares (about 40%) is obtained in the 63–56 days period. In all other cases, the proportion of stable fares is lower. Up until 28 days before departure, we observe a similar probability for an increase or a decrease, which tend to be of similar magnitude. From 28 days onward, increases become more likely and of larger size than decreases, which however still account for about 25 and 20% of flights in the periods 21–14 and 14–7 days, respectively. This contradicts Ryan Air’s quote in the Section 3 that the lowest fares are available before 14 or 28 days prior to departure. Conspicuous and almost certain price hikes are observed in the last week. Figure 5 shows that EasyJet kept its fares stable about 55–60% of times up to 28 days prior to departure: a huge jump from 20 to 40% in the probability of observing a discount of about £14.8, reported in Table 4, relative to the price in the previous week is found in the 28–21 days period (see also the discussion of Table 2 above). In the last 3 weeks, EasyJet fares are frequently increased, and they hardly stay stable in the last 7 days. Generally, the evidence for the three LCCs aforementioned is indicative of small menu costs. It also suggests that the enhanced volatility observed within 4 weeks from departure may reflect a more intense use of pricing as an yield management tool aimed at increasing a flight’s load factor. The intense variability in posted fares is confirmed in Figures 6 and 7, which consider the Summer timetable fares of two FSCs: BMI British Midland and British Airways. The probability of a fare decrease falls steadily as the date of departure approaches, although, quite surprisingly, for quite a large proportion of BMI flights (about 15%), fares declined in the 14–7 days period. For both FSCs fares tend to remain less stable than those of BMIbaby and Easyjet, and generally the magnitude of changes is smaller than that recorded for the LCCs. Recall how the fares for the FSCs were taken from an online travel agent, which is also owned by Amadeus, the computer reservation service. Thus,

Table 4 Average Fares when Fares Remain Constant Between Two Consecutive Booking Days, and the Average Increase and Decrease Between Two Consecutive Booking Days BMIbaby Consecutive booking days before departure 70–63 63–56 56–49 49–42 42–35 35–28 28–21 21–14 14–7 7–1

Ryanair

Easyjet

Average constant fare

Average decrease

Average increase

Average constant fare

Average decrease

Average increase

Average constant fare

Average decrease

Average increase

4916 5027 5214 5349 5528 5638 5846 6087 6375 6907

975 1048 1116 1113 1266 1401 1625 1761 1702 1790

1038 1097 1114 1196 1222 1412 1433 1545 1710 2086

3923 3938 4075 4059 4111 4278 4442 4499 4850 5990

998 1046 900 1008 1104 1080 1110 1316 1409 1533

830 997 951 1030 1146 1345 1578 1931 2388 5777

4871 5147 5662 5314 5441 5631 5742 5521 5818 6095

771 1000 958 1149 1183 1374 1484 1342 1632 1644

1043 1157 1196 1258 1330 1190 1045 1138 1357 3368

Source: Data retrieved from the airlines’ web sites from June 2002 until June 2003.

PRICING STRATEGIES

335

it is likely that the fares posted online are the same as the ones available on Amadeus. The small and highly frequent fare changes even from an online travel agent supports the hypothesis of small menu costs even in cases where the airlines are feeding their fares through a computer reservation system. The discussion in this and the previous section highlights some important characteris­ tics of airlines’ pricing behavior online. The low menu costs facilitate price changes, an important factor that enables companies to gauge current demand by trying out different fares. Note however, that while fares may vary across booking days, it is not likely that they change frequently within each day. As Ellison and Ellison (2005) discuss, inertia in Internet prices is often observed, suggesting that companies do not continually monitor the market situation and reoptimize.6 Nonetheless, the variability we observe over the booking period is probably one of the reasons why the temporal profile of fares hardly replicates a monotonically increasing curve. This reflects the airlines’ ability to combine recommendations from online pricing with the more traditional schemes typically used in the industry. Indeed, sometimes the airlines may want to reconsider the planned pric­ ing scheme they are adopting, in order to reflect and better adjust to demand conditions. This is highly facilitated by the Internet technology. Although this also suggests the airlines’ preference not to commit strictly to a temporal price discrimination scheme, traditional pricing schemes still play a central role. Indeed, in line with the conventional wisdom fares generally grow over time and last-minute offers are virtually nonexistent. This makes it highly risky for a traveler to choose to postpone purchase hoping for a lower fare to become available. Thus, the prospect of future discounts will not deter low-evaluation, risk-averse consumers with more certain demand from buying at an early date. Therefore the traditional second-degree price discrimination schemes still remain a very effective managerial tool even when applied online.

5 WHICH AIRLINE IS THE CHEAPEST, AND WHEN? The previous sections have revealed a synergetic relationship between the traditional pricing schemes in the airlines’ industry and the opportunities presented by the Internet. It is thus worth looking at our data from the perspective offered by some recent theoretical and empirical contributions on pricing in online markets. Baye et al. (2004a) present a model where firms using a price comparison site (a “clearinghouse”) must try to sell to two types of consumers: “Shoppers” (S), who actively engage in searches on the comparison site and buy at the lowest possible price, and “Loyals” (L), who do not search and pay up to their reservation value for the service offered by their preferred brand. In a clearinghouse model, the Law of One Price does not hold, and persistent price dispersion will be observed (Baye et al., 2006). More relevant for our purposes, the identity of the firm offering the lowest price will vary unpredictably over time (Baye et al., 2004b). This is because randomization arising from a “hit-and-run” strategy will be used to prevent

6 Moreover, we casually noted that after buying tickets online from the LCCs in our study, fares remained unchanged despite the obvious reduction in the seat availability.

336

CLAUDIO A. PIGA AND ENRICO BACHIS

rivals from being able to systematically undercut a fixed high price. Indeed, should all firms set a high price to extract the maximum surplus from Shoppers and Loyals, the predictability of such a scheme would give an incentive to each firm to reduce its price by an arbitrarily small amount to attract the Shoppers. To prevent systematic undercutting by rivals, firms in highly competitive e-retail markets adopt “hit-and-run” sales promotions, which in turn leads to the absence of a persistent “low-price” firm in the market. The clearinghouse model can be related to our analysis. We can think of Shoppers as those travelers who search for the lowest price in many airlines’ web site, even if they offer differentiated routes, while Loyals may be those who do not consider alternative airlines maybe because of their preference for a route or type of service (e.g., they may prefer FSCs to LCCs). It is clear, however, that some aspects of the clearinghouse model do not directly translate into the airlines’ setting.7 A comparison site lists all the available prices for exactly the same product. In the airlines industry, flights are differentiated along many horizontal dimensions, namely, time of departure and endpoints airports, as well as vertical ones (e.g., the FSCs vs. the LCCs business concept).8 We tried to account for such sources of product differentiation in the preparation of Tables 5–8, where we show the percentage of times an airline offers the cheapest or the most expensive fare over different booking days. We assigned flights into eight time bands groups of similar size, and within each time band, we compared fares posted by the airlines operating in a market defined as a city-pair (e.g., London–Rome). While we are aware that to eliminate any effect due to geographical differentiation we should have used the flights within a route (i.e., an airport pair), we have shown in Table 1 that the number of competitive routes is very limited, especially if we intend to analyze competition between FSCs and LCCs. Therefore, in the following analysis, we assume that a flight, say, from London Stansted to Rome Ciampino at 9:30 a.m. can be usefully compared with a flight leaving from London Luton for the same destination at 10:00 a.m. However, imperfect substitutability across routes is likely to drive some of the results we will present. Another limitation of the present analysis is that, although we compare fares for the same booking day, the load factors of each flight at each point in time, and thus the cost conditions, may differ. Tables 5 and 6 show the percentage of times an airline offers the cheapest or the most expensive fare over different booking days in markets where LCCs and FSCs are both present. A first interesting result is that LCCs do not always post the cheapest price: this is even more surprising when we consider that FSCs operate in major airports that are often considered to be able to enhance the quality of a journey’s experience. Most likely, such a finding hinges around the imperfect substitutability of the routes in a city-pair. Furthermore, the likelihood to observe a cheaper fare by a FSC increases as the date of departure approaches. This may be explained by the results in the Section 4, where we show that FSCs’ fares exhibit smaller changes between consecutive booking periods.

7 Nonetheless, search engines, for example, www.traveljungle.co.uk or www.skyscanner.net, are present in the European Airlines’ market but they do not operate as the comparison site illustrated in Baye et al. (2004a,b and 2006). 8 The time of booking is also a source of differentiation, which we account for by using the booking days described above.

PRICING STRATEGIES

337

Table 5 Percentage of Times in which an Airline offers the Cheapest Airfare Company

Days before departure 7

Mean

10

14

21

28

35

42

49

28 75 61 19 75 18

28 77 59 18 65 16

26 79 59 16 66 12

30 84 52 24 68 14

32 86 52 25 68 15

32 85 53 22 65 11

32 84 55 21 66 9

31 80 56 21 69 15

23 12 13 16 13 11 12 9 6 21 29 32

20 15 11 16 14 11 11 8 5 20 26 29

19 10 9 17 15 9 12 7 5 20 25 29

17 9 9 17 14 9 12 5 5 23 23 27

17 10 9 18 14 8 11 6 5 28 23 25

17 9 9 18 14 7 10 5 5 27 23 26

20 11 12 18 14 10 11 8 6 23 26 27

Low-Cost Companies BMIbaby Ryanair Easyjet Buzz MTLl GoFly

37 74 57 20 76 22

Traditional Full-Service Companies Aer Lingus Air France Alitalia BMI BA Finnair Iberia KLM Lufthansa Maersk Air SAS Swiss

30 16 21 17 11 17 13 15 8 28 31 32

25 13 15 17 12 13 12 10 6 28 30 42

Note: These percentages are drawn from competitive markets in which 2–5 different companies, both LCA and traditional carriers, operate. Data for the LCC are for the period running from June 2002 to December 2003. Data for FSC are from April 2003 to December 2003.

In addition to helping explaining why sometimes FSCs are cheaper than LCCs, product differentiation may also be responsible for the fact that such LCCs as Ryan Air, MyTravelLite, and EasyJet to a lesser extent, very often turn out to be the “lowest-price” companies in the markets where they operate, while the other LCCs quite rarely offer the cheapest fare. That is, if differentiation did not matter, the identity of the low- and high-price firm would be more difficult to predict, as easier substitutability across routes would tend to smooth the differences across airlines in Tables 5–8.9 On the contrary, within each booking day in these tables important differences across firms are observed, making the above identities easier to forecast. That is, in market where either Ryan Air or MyTravelLite operate, it would be a safe bet to expect them to be the low-price companies, especially as far as early booking is concerned. Across booking days, the same identities may however vary, thereby further suggesting that product differentiation may matter less than the airlines’ idiosyncratic online pricing behavior. That is, since the same routes are evaluated for the same airlines, the effects of

9

That is, if an airline operates a route with better attributes and that therefore the travellers prefer, then that airline should always be the one with the highest fares across booking days.

CLAUDIO A. PIGA AND ENRICO BACHIS

338

Table 6 Percentage of Times in which an Airline Offers the most Expensive Fare Company

Days before departure 7

Mean

10

14

21

28

35

42

49

66 11 7 81 25 76

67 9 8 82 30 79

69 7 8 84 19 83

64 5 11 76 17 81

57 4 12 75 18 80

58 4 11 78 21 85

57 3 11 79 16 87

62 7 10 79 21 80

22 56 23 21 38 54 73 69 68 57 39 45

25 69 14 21 40 59 72 75 64 65 41 51

28 73 13 20 41 63 75 78 61 59 42 47

31 73 15 15 41 63 74 80 62 54 44 59

33 75 13 18 42 64 77 82 61 53 45 60

34 75 13 21 42 64 74 84 62 52 46 62

26 64 17 20 42 54 70 72 63 57 41 48

Low-Cost Companies BMIbaby Ryanair Easyjet Buzz MTL GoFly

60 13 9 80 22 70

Traditional Full-Service Companies Aer Lingus Air France Alitalia BMI BA Finnair Iberia KLM Lufthansa Maersk Air SAS Swiss

18 40 20 22 47 35 41 48 62 53 36 19

20 53 22 22 43 33 71 60 64 62 38 42

These percentages are drawn from those markets in which operate from 2 to 5 different companies, including both LCA and traditional carriers. Data for the LCA are for the period running from June 2002 to December 2003. Data for traditional carriers are from April 2002 to December 2003. Air Europa, TapPortugal and Volare are not included because of the low number of observations.

product differentiation are fixed: Changes across booking days may be ascribed to the air­ lines’ pricing schemes.10 Consider the cases of BMIbaby and GoFly. Especially for the latter, early fares tend to be among the highest on the market, but the probability to be the “low-price” firm increases as the date of departure approaches. This is even more evident in Tables 7 and 8, where we consider the market where only LCCs compete. The prob­ ability that BMIbaby offers the cheapest fare 1 day before departure is almost 62% (the highest across the companies), but it falls to less than 30% about 2 months before depar­ ture. The same probabilities for GoFly are, respectively, about 40% and 7.5%. A reverse example is given by MyTravelLite, whose probability to be the cheapest company falls to 8.3% 4 days before departure, from more than 83% in one of the previous booking days. Ryan Air behaves similarly to MyTravelLite, although with less drastic variations.

10 To what extent is such behaviour rational? That is, why do airlines stick to a fixed temporal time profile, instead of responding promptly to intervening circumstances? A simple and relatively static rule may reflect an airline’s willingness to keep its investment effort in developing and maintaining an effective yield management unit low. This may have changed, especially for those LCCs that have grown considerably bigger after the period we survey.

Table 7 Percentage of Times in which an Airline Offers the Cheapest Airfare, when Competing with other Low-cost Airlines Company

BMIbaby Ryanair Easyjet Buzz GoFly MyTravelLite

Booking days before departure

Mean

1

4

7

10

14

21

28

35

42

49

56

63

70

61.73 51.54 49.44 35.58 40.30 16.67

47.70 58.87 49.09 26.88 34.18 8.33

25.67 65.43 51.00 16.57 30.22 73.33

19.46 67.54 51.84 14.56 26.23 60.00

17.53 71.17 51.37 16.48 18.48 46.67

14.95 74.10 48.26 13.86 19.22 80.00

22.39 80.18 37.82 24.22 21.79 73.33

32.57 78.63 36.36 27.04 19.69 78.57

31.47 77.30 40.67 20.13 11.97 70.00

33.05 79.76 38.02 16.22 11.82 83.33

29.82 77.21 41.42 15.65 9.84 60.00

29.38 75.89 42.24 20.28 7.51 –

34.62 69.54 47.33 18.12 6.88 –

30.79 71.32 44.99 20.43 19.86 50.02

These percentages are drawn from markets in which at most 2, and not more than 3, LCA operate. Data for the LCA are for the period running from June 2002 to December 2003.

Table 8 Percentage of Times in which an Airline Offers the most Expensive Airfare, when Competing with other Low-cost Airlines. Company

BMIbaby Ryanair Easyjet Buzz GoFly MyTravelLite

Booking days before departure

Mean

1

4

7

10

14

21

28

35

42

49

56

63

70

38.27 47.82 49.85 64.42 55.60 83.33

51.94 40.71 50.40 73.13 61.54 91.67

74.00 34.20 48.36 83.43 64.49 26.67

80.20 32.06 48.16 85.44 67.62 40.00

82.13 28.35 48.63 83.52 75.25 53.33

84.70 25.71 51.74 86.14 74.02 20.00

77.24 19.62 62.18 75.78 71.60 26.67

67.43 21.17 63.64 72.96 72.05 21.43

68.13 22.60 59.33 79.87 79.91 30.00

66.95 20.24 61.98 83.78 79.55 16.67

70.18 22.79 58.58 84.35 82.38 40.00

70.62 24.11 57.64 79.72 86.71 –

65.38 30.46 52.31 81.88 86.88 –

69.01 28.45 54.83 79.57 73.66 34.60

These percentages are drawn from markets in which at most 2, and not more than 3, LCA operate. Data for the LCA are for the period running from June 2002 to December 2003.

PRICING STRATEGIES

341

To conclude, our analysis seems to confirm the hypothesis put forward in Baye et al., 2004b) that the identity of the firm offering the lowest price exhibit some degree of unpredictability over time. Such results seem to be independent of the fact that we compare fares for imperfectly substitutable flights and can be mainly attributed to the way the airlines set their fares over time, a topic we have discussed at length in the previous two sections. Nonetheless, within each booking day the low- and highfare firms are easier to predict, partly because of the imperfect substitutability of the routes in a city-pair. Overall, the two results combined point at the conclusion that the airlines do not seem to adjust their fares to match their closer rivals’ offers. If they did, within each booking day, identifying the low-price firm would be more difficult than it would seem to be. This confirms the importance of product differentiation as a strategic weapon to reduce the incentive to engage in tough price competition (Tirole, 1988). The enhanced unpredictability across booking days suggests that the heterogeneity of the airlines’ idiosyncratic pricing schemes is thus likely to be mostly motivated by the aim to maximize a flight’s load factor and not by concerns about the fares offered by the immediate rivals.

6 CONCLUSIONS Using evidence from about 650,000 flights, for which up to 13 fares were available, the present article addressed three highly interrelated topics on the behavior of fares over time: (1) whether airlines’ fares grow monotonically, as it is often assumed; (2) how often and when fares change; and (3) which companies offer the lowest price, and when. First, we show violations to the monotonic property both when we aggregate fares over all the airlines’ routes and when we consider fares in specific routes. Second, we observe higher volatility of fares in the 4 weeks preceding the departure date. Both findings contrast with the statements on some LCCs’ web sites declaring that the best deals are available before 28 days before departure. It would seem that these announcements might be aimed at consumers with more certain demands to induce them to purchases early. The high variability of fares serves the purpose of discouraging such consumers to postpone the ticket purchase, as the probability of the availability of a discounted price in the future is often more than offset by the probability of ending up paying a higher price. In this respect, the airlines manage to apply the traditional schemes based on a second-degree price discrimination strategy where consumers with more certain demand buy at an early stage, while later purchasers with uncertain demand pay a premium. A contribution of this article is to show that such traditional schemes are adjusted to reflect the innovative features offered by the Internet. Thus, the investigation of the first and the second issue leads to conclude that a potential area for future research is a deeper evaluation of how the airlines’ traditional pricing schemes can be applied to reflect such prevailing conditions on the Internet as low search and transaction costs and high price transparency. A first step in this direction is offered in Bachis and Piga (2006), which shows evidence of a form of online price discrimination entailing the airlines charging, at the same time and for the same flight, fares expressed in different currencies that violate the law of One Price.

342

CLAUDIO A. PIGA AND ENRICO BACHIS

The third part of this study explicitly relates airline pricing to some recent theoretical and empirical contributions on pricing in electronic markets. In line with the predictions in Baye et al (2004b), we find that the identity of the airline offering the lowest price in a market changes with the booking day, although within each booking day it tends to remain the same and is therefore easier to predict. The former result is obviously related to the first two issues analyzed in the paper, that is, the fares’ temporal profile and frequency and timing of their change. In Baye et al. (2004b), the unpredictability of the low-price firm is a consequence of the intense online competition. In our case, the identity of the low-price airline changes because of the different inter-temporal pricing schemes used by the airlines. These different approaches seem to be less driven by competitive pressure, given that the airlines are often protected by product differentiation, and more by distinctive strategies used by the airlines to maximize a flight’s revenue.

BIBLIOGRAPHY Bachis, Enrico and Claudio A. Piga (2006), On-line International Price Discrimination with and without Arbitrage conditions, mimeo. Baye, Michael, John Morgan and Patrick Scholten (2006), Persistent Price Dispersion in Online Markets, in Dennis W. Jansen (ed.), The New Economy and Beyond: Past, Present Future. Edward Elgar, London. Baye, Michael, John Morgan and Patrick Scholten (2004a), Price Dispersion in the Small and in the Large: Evidence from an Internet Price Comparison Site, Journal of Industrial Economics, LII(4), 463–496. Baye, Michael, John Morgan and Patrick Scholten (2004b), Temporal Price Dispersion: Evidence from an Online Consumer Market, Journal of Interactive Marketing, 18(4), 101–115. Belobaba, Peter. P (1987), Airline Yield Management. An Overview of Seat Inventory Control, Transportation Science, 21(2), 63–73. Borenstein, Severin, (1989), Hubs and High Fares: Airport Dominance and Market Power in the U.S. Airline Industry, Rand Journal of Economics, 20, pp. 344–365. Borenstein, Severin, (1991), The Dominant-Firm Advantage in Multiproduct Industries: Evidence from the U.S. Airline Industry, Quarterly Journal of Economics, 106, 1237–1266. Borenstein, Severin, and Nancy L. Rose. (1994), “Competition and Price Dispersion in the U.S Airline Industry”, Journal of Political Economy 102, 653–683. Brynjolfsson Erik, and Michael D. Smith. (2000), “Frictionless Commerce? A Comparison of Internet and Conventional Retailers”, Management Science 46, 563–585. Clemons, Eric K., Il-Horn Hann and Lorin M. Hitt. (2002), Price Dispersion and Differentiation in Online Travel: An Empirical Investigation, Management Science, 48(4), 534–549. Dana, Jr., J. (1998): “Advance-Purchase discounts and price discrimination in competitive mar­ kets,” Journal of Political Economy, 106(2), 395–422. Dana, James D. (2001), Monopoly Price Dispersion under Demand Uncertainty, International Economic Review, 42(3), 649–670. Ellison, Glenn and Sara Fisher Ellison, (2005), Lessons about Markets from the Internet, Journal of Economic Perspectives, 19(2), Spring, 139–158. Escobari, Diego (2006), Are Airlines Price Discriminating? Tourist versus Business Travelers. mimeo, paper presented at the 2006 International Industrial Organization Society, Boston. Evans, William N. and Ioannis N. Kessides. (1993), “Localized Market Power in the U.S. Airline Industry”, Review of Economics and Statistics 75(1), 66–75.

PRICING STRATEGIES

343

Gale, I. and Holmes, T. (1992): “The efficiency of advance-purchase discounts in the presence of aggregate demand uncertainty,” International Journal of Industrial Organization, 10, 413–437. Gale, Ian L. and Thomas J. Holmes. (1993), “Advance-Purchase Discounts and Monopoly Allo­ cation of Capacity”, American Economic Review 83, 135–146. Giaume, Stephanie, and Sarah Guillou. (2004), Price Discrimination and Concentration in Euro­ pean Airline Market, Journal of Air Transport and Management 10(5), 293–370. Goldberg, Pinelopi K. and Michael M. Knetter. (1997). Goods Prices and Exchange Rates: What Have We Learned?, Journal of Economic Literature, XXXV, Sept., 1243–1272. Greene, William H. (2003). Econometric Analysis. Prentice Hall, Upper Saddle River, NJ, 5th edition. Hayes, Kathy J., and Leola B. Ross. (1998), “Is Airline Price Dispersion the Result of Careful Planning or Competitive Forces?” Review of Industrial Organization 13, 523–541. Klein, Stefan. and Claudia Loebbecke. (2003), “Emerging Pricing Strategies on the Web: Lessons from the Airline Industry”, Electronic Markets 13(1), 46–58. McGill, Jeffrey I. and Garrett J. Van Ryzin. (1999), Revenue Management: Research Overview and Prospects, Transportation Science, 33(2), 233–256. Mussa, Michael and Sherwin Rosen. (1978), Monopoly and Product Quality, Journal of Economic Theory, 37(3), 1067–1082. Pels, Eric, and Piet Rietveld. (2004), “Airline pricing behaviour in the London-Paris market”, Journal of Air Transport Management 10, 279–283. Pender, Lesley, and Tom Baum. (2000), Have The Frills Really Left The European Airline Industry?, International Journal of Tourism Research 2, 423–436. Piga, Claudio and Nicola Filippi. (2002), “Booking and Flying with Low Cost Airlines”, Interna­ tional Journal of Tourism Research 4, 237–249. Pitfield, D. E. (2005), Some Speculations and Empirical Evidence on the Oligopolistic Behaviour of Competing Low-Cost Airlines. Journal of Transport Economics and Policy, 39(3), 379–390. Shon, Zheng-Yi, Fang-Yuan Chen, and Yu-Hern Chang. (2003), “Airline e-commerce: the revo­ lution in ticketing channels”, Journal of Air Transport Management 9, 325–331. Smith, M., J. Bailey, and E. Brynjolfsson (1999), Understanding the Digital Economy: Review and Assessment, In E. Brynjolfsson and Kahin B. (eds), Understanding the Digital Economy, MIT Press. Stavins, Joanna. (2001), Price Discrimination in the Airline Market: The Effect of Market Con­ centration, The Review of Economics and Statistics 83(1), 200–202. Tirole. Jean (1988), The Theory of Industrial Organization, MIT Press, Cambridge MA.

This page intentionally left blank

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

15 The Long-Run Distributional Effects of Industry and Carrier Changes in the US Air Transport Market Aisling Reynolds-Feighan∗

ABSTRACT The chapter uses Gini decomposition analysis to evaluate changes in the spatial distribution and industry shares of air traffic, as well as analysing the decomposition components for individual airlines and airports. The chapter develops explicit relationships between the two main decomposition schemes used in the poverty literature and applies them to US air traffic distributions. Problems arising with the application of these schemes in continuous distributions are highlighted and an adjustment mechanism is presented to take account of variations in the number of units of analysis in different time periods. A multi-dimensional Gini index and its decomposition are derived using this adjustment method.

1 INTRODUCTION Since airline deregulation in the US in 1977 and 1978, the volume of air traffic has grown enormously and has become more concentrated around a smaller system of airports. Some of the airports have experienced rapid and sustained growth in their traffic shares, while others have gone through periods of expansion and decline, closely linked to the fortunes of the airlines servicing them. The airline industry has gone through several economic cycles in the last 30 years and changed significantly in terms of the firms, networks/products, technology, strategy and geographical market scope. ∗ School of Economics and Geary Institute, University College Dublin, Belfield, Dublin 4, Ireland. Tel: +353-1-716 8525; Fax: +353-1-283 0068; e-mail: [email protected]. The author wishes to thank the UCD Geary Institute for financial support for this research. Funding through the UCD President’s Research Awards is also gratefully acknowledged.

346

AISLING REYNOLDS-FEIGHAN

The academic literature has been concerned with measuring and modelling the effects of carrier network structures on several aspects of firm behaviour and decision-making. The airline’s network represents its production plan and also its range of products. The network structure gives rise to cost interdependencies among the routes in the carrier’s system. Economies of scope and density associated with hubs yield efficiencies to larger “hubbing” carriers under a variety of circumstances (see e.g. Brueckner and Spiller, 1994). The carrier’s dominance at its hub airports gives rise to fare mark-ups and increased yields compared to carriers with smaller traffic volumes at these airports and has been considered a barrier to entry by new carriers (Borenstein, 1989). The network structure influences demand patterns, as passengers evaluate the generalised travel costs arising from indirect versus direct routing options. In the recent period, the new entrant low-cost/low-fare carriers have had a growing impact on fares and market shares at the larger airports and have generally tended to offer point-to-point direct service in contrast to the legacy “hubbing” carriers (US Department of Transportation, 1996; General Accounting Office, 1999). There is considerable concern in the European Union at present regarding the regional and national implications of changes in ownership and regulatory policies that will facilitate consolidation in the airline industry. Despite liberalisation of the European internal market and recent moves to privatise or part-privatise several of the state-owned “national carriers”, external ownership requirements have resulted in a relatively small degree of consolidation in the European industry so far compared with the US following deregulation. This chapter sets out a framework for examining the long-run spatial and industry distributional trends in air traffic activity. The framework presented is based on Gini analysis usually used by economists to analyse trends in inequality or financial portfolios. In applying this method to air traffic activity, Gini decomposition provides several useful and insightful statistics that describe airline network organisation strategies, market shares and the extent of multi-market contact. Gini decomposition is also useful in providing airport-specific measures of concentration and airline mix. The framework is outlined in the first section of the chapter. The second section describes the application of the Gini framework to the US airways system and several key industry and airport trends are highlighted. The final section presents some conclusions and extensions.

2 GINI ANALYSIS IN THE ASSESSMENT AND MEASUREMENT OF AIR TRAFFIC ACTIVITY The application of Gini analysis in economics has tended to be focused on issues related to poverty, income inequality and financial portfolio evaluation, with its application more recently to the evaluation of spatial concentration of employment (Krugman, 1991; Kim, 1995; Ellison and Glaeser, 1997). Typically in these analyses, per capita household income brackets are defined and the frequency distribution is analysed and decomposed by factor components. The factor components will be distributed across all of the income categories, but usually with a different frequency. Individual component Ginis, income shares and “Gini correlation ratios” are isolated and related to the overall Gini index.

DISTRIBUTIONAL EFFECTS OF INDUSTRY AND CARRIER CHANGES

347

Yitzhaki and Lerman (1991), Yitzhaki (1994) and Milanovic and Yitzhaki (2002) have developed an alternative decomposition, taking account of the fact that some income categories will have no factor components present. This decomposition focuses on the notion of stratification existing within the factor components, and the decomposition scheme proposed captures component Gini coefficients, shares and the degree of over­ lapping between component distributions. In the poverty literature typically, the number and range of income categories do not change over time. The two decomposition schemes will be set out and the relationship between them for individual data or a continuous distribution will be derived. With a transportation system, the number of nodes receiving service can change from period to period and/or the number of firms in the industry can change. A method is presented for isolating the effects of changes in the system size between evaluation periods. Using this method, a new decomposition scheme is derived. This scheme is further extended to take account of multi-dimensional variation in the traffic distribution. A combined spatial and industry concentration measure is put forward for the air transport sector with individual airline network structures and airport traffic distributions related to the overall trends. An overview of the application of this framework to US air traffic distributions for the period 1973–2005 is then presented. The Gini coefficient is used in a variety of circumstances but most frequently in economics to measure inequality in the distribution of income. The Gini coefficient may be written as G = 1−

n 1 S + Si−1 n i=1 i

(1)

where Si is the cumulative income share of all individuals with income less than or equal to that of the ith individual. In the current application, Si will measure the share of air traffic at airport i, where airport i is a single airport in the national airways system. Air traffic can be measured in terms of passengers boarding aircraft (passenger enplanements) or seats available. The Gini coefficient may be computed in a variety of different ways. The formula presented by Pyatt et al. (1980) and elaborated upon by Lerman and Yitzhaki (1984) will be used in this chapter. These formulations may be presented as G=

2 covx Fx 

(2)

where Fx is the cumulative distribution function of air traffic and  is its mean (Lerman and Yitzhaki, 1984; Dorfman, 1979). This is the empirical formulation of the population Gini, as the number of individuals in the sample goes to infinity. The second formulation is the sample Gini, given as G=

2 covx rx  nx

(3)

where n is the number of individual airports sampled,  x is the mean of x, covx rx  is the covariance between the air traffic distribution, x and the ranks of airports according to their traffic shares (rx ) from the smallest (rx = 1) to the biggest (rx = n).

AISLING REYNOLDS-FEIGHAN

348

In decomposing the overall Gini into subgroups, two decomposition schemes have been proposed in the literature (the first in Lerman and Yitzhaki (1984, 1985) and the second in Yitzhaki and Lerman (1991) and Yitzhaki (1994)). These have been used extensively to examine inequality in the distribution of income sources (Garner, 1985) and to examine regional and intercontinental differences in the income inequality (Milanovic and Yitzhaki, 2002). Lerman and Yitzhaki (1985) show that the overall Gini coefficient based upon i subgroup components is 2 Gx =

N 

covxi  Fx

i=1

2 =

x

N 

covxi  rx

i=1

Nx

(4)

The first decomposition is thus  N  N   covxi  Fx 2 covxi  Fxi  xi = Ri G i S i G= xi x i=1 covxi  Fxi  i=1

(5)

where Ri is the rank correlation ratio, Gi is the relative Gini of component i, and Si is component i’s share of total traffic (Lerman and Yitzhaki, 1984). This decomposition requires that each subgroup has a distribution over the same range as x. Thus the number of observations will be the same for each subgroup as it is for x. In applying this decomposition to air traffic distributions, we can decompose the overall air traffic across the system of airports by individual carriers or by groupings of carriers. The second decomposition scheme put forward by Yitzhaki and Lerman (1991) and refined in Yitzhaki (1994) allows subgroups to cover a subset of the range of x. This decomposition is given as Gx =



Si G∗i Oi + Gb

(6)

i

where G∗i is the relative Gini coefficient for carrier i over airports in its network, Si is the traffic share for carrier i as before, Oi is an “overlapping index” and Gb is “between group” concentration. The overlapping index, Oi , is discussed at length in Milanovic and Yitzhaki (2002) and defined as Oi =

covxi∗  Fx∗  covxi∗  Fxi∗ 

(7)

the ratio of the covariance between carrier i’s traffic distribution ranked by the overall air traffic distribution for airports served by carrier i, to the covariance of carrier i’s traffic distribution ranked by its own air traffic distribution across airports in its network. The Oi component for carrier i is the sum of overlaps with all other carriers. This component may be interpreted as a measure of multi-market contact for individual carriers with all other carriers (see Evans and Kessides, 1994; Fournier and Zuelke, 2004). Yitzhaki and Lerman (1991) argue that the Oi component could be further decomposed to yield measures of overlap between pairs of subgroups, yielding measures

DISTRIBUTIONAL EFFECTS OF INDUSTRY AND CARRIER CHANGES

349

of multi-market contact between pairs of individual carriers when applied to air traffic distributions. The “between group” concentration is twice the covariance between the average traffic for each carrier (over all of the airports served) and its mean rank in the overall traffic distribution, divided by the overall average air traffic, ie Gb =

2 covxi ∗  F x∗  x

(8)

The number of carriers or carrier subgroups and the distributional range of the subgroups will influence the size and sign of Gb . As the number of subgroups increases, the Gb effect may be expected to increase. The greater the degree of “stratification” in the range of airports occupied by each carrier, the larger the Gb component will become, and the smaller will be the “within group” concentration. The limiting case would be where the carrier subgroups each occupy one single airport, in which case all of the variability will be between groups, with no within group variation. However, Gb may be positive or negative depending on the skewness in individual subgroup distributions1 . The Gini coefficient has also been used to measure industrial concentration and we wish to consider airline industry concentration as well as air traffic spatial concentration. Deltas (2002) has demonstrated a bias in the Gini coefficient for small samples and argues that the size of the bias is large compared to the standard error. He also shows that this bias varies substantially across different distributions. He proposes an adjustment to the Gini coefficient of (n/n−1) in order to reduce this bias and demonstrates its application among shipping cartels. The Gini coefficient has been applied recently to the measurement of spatial concentration in patterns of employment (Krugman, 1991; Kim, 1995; Ellison and Glaeser, 1997). The application presented for the air transport system in this chapter examines both spatial and industrial concentration patterns in the distribution of air traffic across US airports and among airlines. The units of observation will be individual airports and airlines, thus avoiding the “modifiable areal unit problem” (MAUP2 ). In examining changes in concentration over time, explicit account needs to be taken of changes in the number of firms operating in the market and in the number of airports receiving air services in a given year. An adjustment factor is derived to account for that component of concentration change due to variations in number of firms and airports and that component due to changes in the distributions of traffic among firms and within individual firm networks. The decomposition scheme presented earlier in Equation (5) above assumes that each  subgroup will have observations over the full range of airports, hence M i=1 xi = x, the overall mean. The subgroup Gini computed in Equation (6) however is the carrier Gini for the range of airports present in the carrier’s network. Typically the number of airports in carrier i’s network (ni ) will be less than the total number of airports in the national

1

Indeed Yitzhaki and Lerman cite an example of how this can arise in their 1991 paper.

MAUP consists of both a scale and an aggregation/zoning problem. It is the variability in statistical results

obtained within a set of modifiable units as a function of the various ways these units can be grouped at a

given scale, or as a result of the variation in the size of those areas.

2

AISLING REYNOLDS-FEIGHAN

350

airways system, N . The two Gini coefficients are then related in the following way. The covariance of xi with its cumulative distribution evaluated across the full range of x is   1  ∗ ∗  ∗ N +1 covxi  Fxi  = covxi  rxi /N = xi ri − xi (9) N i 2N i while the covariance of xi with its cumulative distribution evaluated across the range of xi only is covxi∗  Fxi∗  = covxi∗  rxi∗ /ni =

  1  ∗ ∗  ∗ 2N − ni + 1 xi ri − xi 2ni ni i i

(10)

This gives the relationship between the two subgroup Gini coefficients as Gxi  =

n i

N

 Gxi∗  +

N − ni N

 (11)

Substituting for Gi in Equation (5) yields a new decomposition: G=

 i

Si G∗i Ri

ni  N − ni + Si Ri N N

(12)

This breaks down the overall Gini into (i) that part due to the effect of distributions within each individual subgroup and (ii) that part due to differences between subgroups, allowing that subgroups may have a different number of observations. Clearly when ni = N , Equation (12) reduces to Equation (5). The “between group” factor presented by Yitzhaki and Lerman (1991), in contrast does not explicitly take account of the impact of the number of subgroups or the range of the overall distribution occupied by the subgroups, but is clearly influenced by these factors. If all subgroups vary over the full range of x, then Ri and Oi components in Equations (5) and (6) coincide and Gb = 0. If large subgroup distributions overlap to a significant degree, then the within group measure in Equation (6) may exceed the overall Gini index and Gb may be negative. The interpretation of a negative Gb component may be problematic, as shall be shown later. The relationship in Equation (12) can also be used to adjust the Gini index in order to estimate the effect of the changes in the number of observations between different periods. In this case, N represents the number of observations in period t and ni can represent the number of observation in period t + 1. The adjustment to the Gini index then reflects the extent of changes in the number of observations. This “adjusted Gini index” will be used in the US air transport application in the next section to isolate the changes in the Gini index between time periods due to changes in the number of observations (a shift factor) and changes in the distribution of shares. A two-dimensional Gini index can be developed where x is a variate subdivided among M subgroups and these subgroups themselves form a distribution which can be

DISTRIBUTIONAL EFFECTS OF INDUSTRY AND CARRIER CHANGES

351

ranked from the smallest share to the largest. The two-dimensional Gini index further decomposes the variate x as follows: ⎡ G2D = 2

⎤

M  N 

⎡

⎤

M  N 

ij

covxij  rxij 

⎢ covxij  F  ⎥ ⎢ ⎥  ⎦=2  ⎣ ⎣  ⎦ x + x  x + x /MN j i j i j=1 i=1 j=1 i=1 j

i

j

(13)

i

where F ij is the cumulative distribution of x over both i and j entities. Each xij represents the level of x in category i for subgroup j. The Gini Index involves a pairwise comparison of each cell in this (MxN) matrix. The two-dimensional Gini is then a weighted average of the Gini for the i and for the j entities. The Gini Index applied to the distribution of air traffic across the airports measures spatial concentration. The Gini Index across the airline traffic shares measures industry concentration. The two-dimensional Gini Index is a weighted average of industrial concentration and spatial concentration, with the weights being determined by the relative number of airports (N ) and airlines (M). This multivariate Gini index turns out to be equivalent to the “distance Gini” formulation presented in Kosevoy and Mosler (1997). The two-dimensional Gini index further decomposes the variate x as follows: ⎡ G2D = 2

⎤

N M  

covxij  rxij 

N M ⎢ ⎥ Gi + Gj  ⎣  ⎦= + x  x /MN MN MN j i j=1 i=1 j

(14)

i

where r ∗ is the ranking of x over both i and j entities. This distribution of r is derived by adding the column rank and the row rank and this is equivalent to taking the cumulative distribution of x summed over both the rows and columns. Decomposing this twodimensional Gini into subcomponents as in Equation (9) above gives the following relationship: G2D =

 i

Si G∗i Ri

  mj M − mj ni N − ni  + Si R i + Sj G∗j Rj + Sj R j MN MN MN MN j j

(15)

where the four components measure spatial concentration due to variations within air­ ports, concentration due to variations between airports, industry concentration due to within airline concentration and industry concentration due to between airline concen­ tration. The two-dimensional aggregate Gini, G2D , may be adjusted to take account of changes in the number of observations for i and j in the same way that the univari­ ate Gini index was adjusted in Equation (8) earlier. The general equations presented in this section allow for variations in the number of observations within subgroups and can measure concentration in a given distribution across two sets of subgroups. These formulations may be applied in a variety of applications to link spatial or regional aspects of a distribution with variations across groupings of economic agents or sectors.

352

AISLING REYNOLDS-FEIGHAN

3 APPLICATION OF GINI DECOMPOSITION TO US AIR TRAFFIC TRENDS The methodology outlined in the previous section was applied to the US air transport system, using the T100 and T3 databases maintained by the Department of Transporta­ tion.3 These databases record different measures of air traffic activity. The main tables reported below use the total onboard passenger volume (outbound4 ) from all US airports receiving “certificated” traffic in this period. The T100 database covers domestic and international traffic of US and non-US carriers since 1990. For comparison purposes, the T3 database going back to 1973 is also used and the passenger measure here is the total number of enplanements5 . US carriers’ domestic and international passengers are included, though non-US passenger statistics are not. For each year of analysis, a traffic matrix giving air passenger volumes by airport and by airline was generated. The results concentrate on the traffic carried by the larger carriers operating at least 60-seater aircraft and do not include traffic operated by commuter and regional carriers. Regional jet services are also excluded. The analysis focuses therefore on a subset of the US air transportation system, looking at the larger airports and carriers only. Further research could link the trends in regional and commuter air traffic services to the patterns at the large centres and carriers.

3.1 Overall Trends in Spatial and Industry Concentration The US airline industry experienced rapid growth during the 1990s and recovered quickly from the adverse impacts of the first Gulf War in 1990/1991. During the mid­ 1990s, several new carriers entered the market and domestic growth was driven by these predominantly low-cost/low-fare carriers. The legacy carriers focused their expansion on international markets and took advantage of new freedoms offered through the large number of “open-skies” agreements signed from 1992 onwards. Most of the larger carriers entered international airline alliances facilitating more streamlined flows among networks of alliance members. The market experienced an expansion in the number of non-national carriers operating into US airports. A significant number of airports began handling international services during this period [see Brueckner and Whalen (2000), who examine the role and impacts of airline alliances].

3 Under the US Code Title 14, Vol. 4 Parts 200–1199, the reporting requirements for US and non-US air carriers serving US markets are set out under the “Uniform system of Accounts and Reports for Large Certificated Air Carriers” (Title 14, Chapter II, Part 241). Carriers are required to file on a monthly basis, Form 41 Schedules T-100 and T3, US air traffic and capacity data by non-stop segment and on-flight. The schedule T-100 reports give monthly traffic data by origin-destination market and by non-stop segment and are filed by the large ‘certificated carriers’. The schedule T-3 reports give quarterly traffic reports by airport. 4 The total number of revenue passengers transported outbound over a single flight stage, including those already on board the aircraft from a previous flight stage. 5 A count of the number of passengers boarding and tons of cargo loaded on an aircraft. For this purpose, passengers and cargo on aircraft entering a carrier’s system on interchange flights are considered as enplaning at the interchange point; and passengers and cargo moving from one operation to another operation of the same carrier, for which separate reports are required by the Department of Transportation, are considered as enplaning at the junction point.

DISTRIBUTIONAL EFFECTS OF INDUSTRY AND CARRIER CHANGES

353

Comparison of Spatial Gini 1973–2005 0.87 0.86

Gini Index

0.85 0.84 0.83 0.82 0.81 0.8 0.79 0.78 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005

T100 Spatial Gini

T3 Spatial Gini

Adjusted T3 Spatial Gini

Adjusted T100 Spatial Gini

Figure 1 Trends in the Spatial Concentration of US Mainline Air Traffic, 1973–2005.

The impacts of these changes on the spatial distribution of traffic are examined for the period 1990–2005 using the T100 database and for the longer 1973–2002 period using the T3 database. Figure 1 shows the Gini index for the US airports system for the period 1990–2005 (the “T100 spatial Gini”) and for the period 1973–2002 for US carriers only (the “T3 spatial Gini”). The spatial Gini increases between 1991 and 1997 and falls gradually until 2001, rising slightly to 0.837 in 2002 and more substantially in 2003/2004. The standard error of the annual spatial Gini estimates range from 0.01 in 1997 to 0.014 in 2000.6 The T3 spatial Gini shows a similar trend in the 1990s, but has Gini values three percent lower on average than the T100 estimates. This difference is due to the large number of international carriers. The unadjusted spatial Gini trend over time is very similar to the trend in the number of airports, as Figure 2 illustrates. This figure shows the number of airports included in the analysis using the T100 and T3 databases. Many of the airports at the lower end of the airports hierarchy experienced cycles of jet service withdrawal and short-term provision as the industry went through cycles of growth and downturns and carriers experimented with fine-tuning their network structures and extent. The inclusion of these “marginal” airports in the analysis has a statistically significant impact on the spatial Gini coefficient. The adjusted spatial Gini coefficients derived using Equation (11), are also depicted in Figure 1 and show very different trends. The “T3 adjusted Gini” shows a significant increase in spatial concentration after deregulation in 1978 and following a roughly 10-year adjustment period, very little change thereafter. By accounting for changes in the number of airports from period to period, the adjusted Gini coefficient isolates changes in the distribution of traffic across the rest of the hierarchy. The Gini index for the airline market shares is illustrated (“industry Gini”) in Figure 3. Once again the T100 and T3 are depicted for the “raw” Gini scores and for the adjusted

6 The standard errors are calculated to take account of variability in the Gini index associated with reporting errors and missing data.

AISLING REYNOLDS-FEIGHAN

Number of Airports Receiving Certificated Services

354

Comparison of Airport Counts 1973–2005 450 400 350 300 250 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 T100 D+I Airports

T3 Airports

Figure 2 Number of Airports in the T100 and T3 Air Traffic Databases, Mainline Operations Only, 1973–2005.

Comparison of Industry Gini 1973–2005 1

Gini Index

0.9 0.8 0.7 0.6 0.5 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005 T100 Industry Gini

T3 Industry Gini

T3 Adjusted Industry Gini

T100 Adjusted Industry Gini

Figure 3 Concentration in the US Airline Industry, Mainline Operations, 1973–2005.

Gini scores. The T3 adjusted Gini index shows a significant reduction in the years immediately after deregulation. The industry becomes more concentrated in the 1980s with the larger number of consolidations and financial failures, peaking in 1991. In the last 10 years, concentration has decreased once again with the entry of new low-cost carriers and many international carriers. Figure 4 shows the two-dimensional Gini coefficient defined in Equation (14) and this is adjusted to isolate changes due to changes in the number of carriers and airports over time. The combined T3 “spatial and industry” Gini shows an increase in the 10 years following deregulation, and a further increase in the 1986–1988 period (reflecting the increased industry concentration in this period), with very little change in the most recent decade. The adjusted T100 “combined Gini index” measures the extent of concentration among the top carriers and the largest airports in the US airports hierarchy. All US and non-US carriers are included. The trend line shows a reduction in the 1994–1997 period, increasing until 2000 with a decline in the 2003–2005 period. This reflects the increased dispersion in traffic across carriers and airports in the most recent period. The Gini index is sensitive to changes in the number of observations between periods and a statistically significant change can be accounted for entirely by changes in the

DISTRIBUTIONAL EFFECTS OF INDUSTRY AND CARRIER CHANGES

355

Comparison of Combined Spatial and Industry Gini 1973–2005 0.94

Gini Index

0.92 0.9 0.88 0.86 0.84 0.82 0.8 1973 1975 1977 1979 1981 1983 1985 1987 1989 1991 1993 1995 1997 1999 2001 2003 2005

Adjusted T3 Combined Spatial & Industry Gini Adjusted T100 Combined Spatial & Industry Gini

Figure 4 Combined Spatial and Industry Concentration, 1973–2005.

number of observations. By applying the adjustment factor given in Equation (11), changes in the Gini due to shifts in the distribution can be isolated. Decomposition of Spatial Ginis. The decomposition of spatial concentration into “within” and between’ components is illustrated in Figure 5, using the decomposition scheme introduced in Equation (12). Nine categories of carrier grouping were used ini­ tially and compared with a more detailed decomposition by all individual carriers. The nine carrier groupings were derived purely on the basis of aggregate market shares. The “between carrier” concentration component increased between 1994 and 1998 from 0.65 to 0.70, indicating an increased difference between carrier network traffic distributions. In comparing the nine carrier categories to the decomposition using all individual carrier distributions separately, the “between group” concentration increases as more groupings are used. But because this measure of between group concentration weights each cate­ gory by its traffic share (Si ), and both classifications use data for the top five carriers individually, the extent of the increase in the between group component averages .05.

Gini Index & Components

Gini Decomposition - T100 All Airlines, 1990–2005 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 1990

1992

1994

1996

1998

2000

Year Within Group Concentration (All) Between Group Concentration (All) Spatial Gini

Within Group (9 Groups) Between Group (9 Groups)

Figure 5 Within and Between Concentration for the Spatial Gini.

2002

AISLING REYNOLDS-FEIGHAN

356

Gini Coefficient or Component

Comparison of Decompositions Using Different Carrier Groupings 2 1.5 1 0.5 0 –0.5 –1

1991

1994

1997

2000

2002

Year Spatial Gini Within Group (9)

Between Group (9) YL Within (all)

YL Between (9) YL Within (9)

YL Between (all)

Figure 6 Comparison of Gini Decompositions into Within and Between Components, using Different Carrier Categories.

Figure 6 compares this performance to the within and between decomposition for the Yitzhaki–Lerman scheme of 1992 (hereafter YL92) and two significant differences are apparent. First, the “between group” component is negative because of the skewed nature of the traffic distribution. Secondly, this effect becomes far more pronounced when the individual carrier distributions are utilised. The YL92 scheme is far more sensitive to changes in the number of subgroup categories used, as groups are not weighted by their market shares. It is difficult to interpret the meaning of the YL92 decomposition results in these circumstances. Table 1 records the decomposition components for the T100 nine carrier grouping for 2002 for illustration purposes. These components were computed for individual carriers for each of the years 1990–2005. The contribution of each carrier subgroup to overall concentration and the relative contribution to concentration are also derived (see Lerman and Yitzhaki, 1984; Garner, 1985). Figures 7–10 illustrate some of the interesting trends emerging from analysis of these subcomponents. Carrier-specific Decomposition Components. The Gini index (G∗i ) captures the concen­ tration of traffic in individual carrier networks and clearly distinguishes “point-to-point” from “hub-and-spoke” strategies. This measure was applied extensively by ReynoldsFeighan (2001) to assess long-run US carrier network strategy patterns. Figure 7 illus­ trates the trends in the carrier Gini measures for the 1990–2005 period for the top 10 carriers. This figure clearly illustrates the different network strategy operated by South­ west Airlines (WN) compared to all of the other large full-service carriers (FSCs). The significant reduction in USAir’s Gini index between 1991 and 2002 is noted, reflect­ ing the restructuring and contraction in the carrier’s network in the last 3 years and the more dispersed nature of traffic distribution among remaining airports in its network. Figure 8 shows the Gini Index and Concentration Ratio scores for full service networking carriers and low-cost carriers in 2002. The “legacy carriers” are very similar in terms of

Table 1 Decomposition Components for Nine Carrier Grouping using T100 Database for 2002 Carrier Category

DL AA WN UA NW US (CO:HP;AS) Other US Other Nth American International System Total Within Concentration Between Group Concentration

Total Number of Airports Served

Correlation with Rank of Total Traffic

Gini of Gini for Carrier Carrier Category Network

Overlap Index

Traffic Share

YL “Within” Group Concentration

Contribution to Total Concentration

New “Within Group” Concentration

New “Between Group” Concentration

Share of Traffic Concentration

N

Ri

Gi

Gi∗

Oi

Si

Gi∗ SiOi

RiGiSi

SiGi∗ Ri(ni/N)

SiRi(N-ni/N)

Ii = (RiGiSi)/G

124 108 59 87 122 68 113 320 53

0.9543 0.9602 0.8727 0.9651 0.9324 0.8919 0.9410 0.9292 0.9649

0.9113 0.9243 0.9110 0.9350 0.9129 0.9357 0.9264 0.7850 0.9490

0.7531 0.7582 0.4797 0.7423 0.7536 0.6737 0.7753 0.7683 0.6680

1.1748 1.1774 1.1031 1.3031 1.1518 1.1463 1.6320 0.9717 1.5100

0.1435 0.1404 0.1302 0.1033 0.0773 0.0748 0.1131 0.1563 0.0087

0.1272 0.1256 0.0694 0.1002 0.0672 0.0580 0.1434 0.1168 0.0737

0.1247 0.1246 0.1036 0.0932 0.0658 0.0624 0.0986 0.1140 0.0080

0.0371 0.0320 0.0093 0.0187 0.0192 0.0089 0.0270 0.1035 0.0009

0.0877 0.0926 0.0942 0.0745 0.0466 0.0536 0.0716 0.0105 0.0071

0.1486 0.1484 0.1233 0.1110 0.0784 0.0744 0.1174 0.1358 0.0095

142 345

0.9257 1.0000

0.9242 0.8396

0.8158 0.8396

0.0000 0.0000

0.0523 1.0000

0.0000

0.0448 0.8396

0.0163 0.2727

0.0285 0.5669

0.0533 1.0000

0.8815

0.2727

−0.0419

0.5669

AISLING REYNOLDS-FEIGHAN

358

Carrier Gini Index, 1990–2005 0.9 American (AA)

Gini Index Value [0,1]

Continental

0.8

United Delta

0.7 America West

TWA

0.6 US Airways Southwest

0.5

0.4

1990

1992 1991

1994 1993

AA

1996 1995

DL

1998 1997

WN

UA

2000 1999

NW

2002 2001

US

2004 2003

TW

2005

CO

HP

Figure 7 Trends In Carrier Gini Index Scores (Gi∗ ) For the Top 10 Carriers, Mainline Operations, 1990–2005.

1 7

N

Spirit

Full Service Network Carriers

L

AA NW

D

S

U

FL 6 B

Continental (CO) American (AA) United (UA) Delta (DL) Northwest (NW)

AS

SY

Low Costs Carriers

0.7 Southwest

EV

Concentration Ratio (Ci)

O

A

C

U

TZ

J

N

N

W

K

N

0.8

HP

F9

0.9

0.6 0.5 0.4

0.4

0.5

0.6

0.7

0.8

0.9

1

Gini Index (Gi*) 2002

Figure 8 Full Service and Low Cost Carriers Gini Index and Concentration Ratio Scores, 2002.

Gini scores and concentration ratios. The concentration ratio reflects the overall ranking of airports in a carrier’s network. Thus the legacy carrier’s networks are highly concen­ trated and focused on the top of the airports hierarchy. The low-cost carriers by contrast have generally less concentrated networks and tend to be focused lower down the airports hierarchy. This is not the case for America West (HP), National (N7), and Frontier (F9). Carrier market shares are given as “Si ” in Table 1. Southwest Airlines’ share of total passenger traffic (i.e. domestic and international) grew substantially, doubling between 1991 and 2002. All other US carriers outside the top 10 collectively doubled their

DISTRIBUTIONAL EFFECTS OF INDUSTRY AND CARRIER CHANGES

359

0.85 Southwest

Oi Value Scaled to [0,1]

0.8

Continental

0.75

American

USAir

0.7 0.65 Delta

0.6 0.55 0.5 0.45

Oi90 [0,1]

Oi92 [0,1]

Oi94 [0,1]

AA

DL

Oi96 [0,1]

WN

Oi98 [0,1]

UA

NW

Oi00 [0,1]

US

Oi02 [0,1]

TW

Oi [0,1) 04

CO

Figure 9 Trends in Carrier Overlap (Oi), Scaled to [0,1] for the Top Carriers, 1990–2005. 1

L

N JF

International Gateways

K

Concentration Ratio

EN

IA M

X LA

H

0.8

JFK MIA HNL LAX

0.7

0.6 0.6

0.7

Airline Hubs for Full Service Carriers

L D

S LA

W

D

O SF S

BO O

LT C H L D IA R AT O FW PIT D L G ST CV SP TW D R M EW

X

PH

PH

I

C

A

M

SE

L FL

0.9

M

L

ases

ier B

Carr

BW

A LG

ost ow C

New York (Kennedy), NY Miami, FL Honolulu, HI Los Angeles, CA

0.8

0.9

ATL CLT CVG DEN DFW DTW EWR IAH JFK MSP ORD PHL PIT STL

Atlanta, GA Charlotte, NC Cincinnati, OH Denver, CO Dallas/Fort Worth, TX Detroit, MI Newark, NJ Houston, TX New York NY Minneapolis/Saint Paul, MN Chicago, IL Philadelphia, PA Pittsburgh, PA Saint Louis, MO

1

Airport Gini (Gi*) Top 20 Airports, 2002

Figure 10 Airport Gini Index and Concentration Ratios for Top 27 Airports, 2002.

market share in the same period. While the number of international carriers increased significantly over the period, the market share of total traffic handled by this subgroup increased by 1% in the 12 years. The “market overlap” measure Oi captures the extent of market overlap between a carrier’s traffic distribution in the subset of airports in its network, ranked by the carrier’s ranking compared with the aggregate traffic ranking. Figure 9 illustrates trends in Oi for the top 8 carriers between 1990 and 2005. This can be further decomposed to give a pair-wise measure of multi-market contact or overlap for any carriers operating to

360

AISLING REYNOLDS-FEIGHAN

airports in a given carrier’s system, though it is not reported here. The “Oi ” measure varies from 0 to 2, indicating increasing overlap with the ranking of total traffic at the airports. Significant changes in the pattern of the “Oi ” variable are observed in the 1991–1994 and 2002–2005 periods particularly. Southwest’s overlapping index is at 1.58 (or 0.79 on a [0,1] scale) in 1991 and declines significantly in subsequent periods as its network expands. In the 2002–2005 period however, Southwest’s overlapping index rises sharply. During the 1994–2002 period, the airports around which traffic growth had been focused had lower rankings in the national distribution than in Southwest’s system and the carrier faced a lower level of competition in its larger network. This measure of multi-market contact is based on the set of airports served rather than based on routes carriers may have in common. In the most recent 2002–2005 period, Southwest’s growth has increased the ranking of its main airports in the national distribution and now overlaps to a greater degree with all other carriers’ traffic distributions. The airport-based measure will implicitly take account of common routes among carriers. However it also captures the notion of competing networks, where the legacy carriers for example, offer alternate routings between origin–destination pairs and may not necessarily have common routes with other carriers. The measures of airline multi-market contact presented in the literature to date have focused purely on route overlaps (Evans and Kessides, 1994; Gimeno 1999). American Airlines’ overlapping score also increases sharply in the 2002–2005 period. American and Southwest were the top ranked carriers in 2004 and 2005 in terms of onboard passengers on mainline jet operations. Further exploration of the overlap between pairs of carriers can help identify the main competitors for each carrier. The international carriers have consistently the highest market overlap, since their operations focus on the main international gateways and these will tend to be at the top of the US airports hierarchy. The decline in the overlap measure for these carriers in the 2000–2002 period is noted and related to more liberal “open skies” agreements that have facilitated direct international access to a greater number of US airports. Industry Concentration and Airport Measures. The industry concentration measure may be decomposed in the same manner as spatial concentration, yielding a variety of summary statistics for individual airports based on the traffic distributions of carriers serving the airports. Figure 10 illustrates the airport specific Gini index (G∗i ) and con­ centration ratio scores for the top 27 airports in 2002 (corresponding to FAA large hubs, but counting airports rather than communities). The key airport hubs for the full service carriers have very high Gini index scores (G∗i ) reflecting the dominance of the hubbing carriers at those airports. Hartsfield Airport in Atlanta (ATL), O’Hare Airport in Chicago (ORD), Charlotte Airport in North Carolina (CLT) and William Hobby Airport in Hous­ ton (IAH) have the four highest Gini index scores among the top 20 airports. The high value of the concentration ratio indicates the ranking of airlines at the individual airports. The second grouping that is clearly discernible in Figure 10 is the international gateway airports at Los Angeles (LAX), New York (JFK), Miami (MIA) and Honolulu (HNL). These airports have lower Gini index scores and lower concentration ratios, reflecting the much greater influence of the large number of international carriers at the airports. The low-cost carrier bases have high concentration ratio values but generally lower Gini index scores indicating the less concentrated traffic distributions among carriers serving the airports. The airport Gini index scores can be traced over time to track the impact of carrier network changes and carrier entry and exit effects on traffic distributions at

DISTRIBUTIONAL EFFECTS OF INDUSTRY AND CARRIER CHANGES

361

particular airports. There are several examples of dramatic changes in these component scores for particular airports (e.g. significant traffic declines at St Louis, Missouri; Raleigh Durham, North Carolina; Dayton, Ohio and dramatic traffic growth at Las Vegas and Reno, Nevada and Midway Airport in Chicago, Illinois). There is much scope for further regional and carrier specific analysis of these trends. This analysis quantifies the effects that Borenstein and Rose (2003) identified and provides a framework for tracing individual carrier or airports impacts on overall trends in industry or spatial concentration. While the overall trend shows little change in spatial and industry concentration in the US passenger air traffic patterns in the past decade, there can be dramatic and significant changes at individual airports or for individual airlines. It is possible to examine how particular airports have performed over time by using “Gini mobility” statistics.

4 CONCLUSIONS AND EXTENSIONS This chapter examined the two main Gini decomposition schemes used in the poverty and income inequality literature. The relationship between the two schemes was set out and used to develop a general mechanism to take account of changes in the number of observations between periods and to put forward an alternate decomposition of the overall Gini index. The development of a framework for evaluating changes in two sets of subgroups and relating them to overall trends in concentration was put forward. The adjustment mechanism was extended to take account of variations in the number of units of analysis in both sets of subgroups in different time periods. This framework has many possible applications. The framework was applied to the US air transport industry and used to relate individual carrier and airport activity to overall national trends in industry and spatial concentration. It was shown that most of the change in spatial concentration of air traffic in the 1990–2005 period was due to changes in the number of airports served. When account is taken of changes in the number of airports served, it was demonstrated that except for a once-off increase in spatial concentration following deregulation, spatial concentration has remained stable at similar levels for the past 20 years. The carrier decomposition of spatial concentration produced measures of (i) carrier market share (ii) market overlap or multi-market contact for individual carriers and (iii) carrier network organisation strategy. The individual carrier Gini index measures clearly demonstrated the different network organisation strategy employed by Southwest Airlines vis-à-vis the other large network carriers. Further research will examine at a micro level the impact of individual carriers behaviour in terms of market overlap and pricing strategy. A European application could contribute to the debate and analysis of the future direction of the industry under scenarios of consolidation, changes in ownership requirements for international (extra-EU) services and enlargement of the internal market. The statistics relating to carrier behaviour can be related to financial and cost aspects of carrier operations. The two-dimensional Gini approach presented in the chapter may also be applied to decomposition of other economic trends by region or by industrial sector. Using continuous data, explicit account of structural changes in regions or sectors may be accounted for and their impacts isolated from distributional changes.

AISLING REYNOLDS-FEIGHAN

362

APPENDIX Carrier codes used in figures and table: SY CS NJ 9N N7 F9 NK B6 TZ EV FL AS HP CO US NW UA WN AA

Sun Country (New) CO – Micronesia Vanguard Trans States National Frontier Spirit Air Jetblue American Trans Atlantic Southeast Airtran Alaska America West Continental USAIR Nortwest United Southwest American

Airport Codes Used in Figures: SAN SLC TPA PIT FLL MDW HNL BWI LGA CVG CLT BOS PHL STL MCO SEA MIA JFK SFO EWR

San Diego, CA Salt Lake City, UT Tampa, FL Pittsburgh, PA Fort Lauderdale, FL Chicago (Midway), IL Honolulu, HI Baltimore, MD New York (Laguardia), NY Cincinnati, OH Charlotte, NC Boston, MA Philadelphia, PA Saint Louis, MO Orlando, FL Seattle, WA Miami, FL New York (Kennedy), NY San Francisco, CA Newark, NJ

DISTRIBUTIONAL EFFECTS OF INDUSTRY AND CARRIER CHANGES

MSP DTW IAH LAS DEN PHX DFW LAX ORD ATL

363

Minneapolis/Saint Paul, MN Detroit, MI Houston, TX Las Vegas, NV Denver, CO Phoenix, AZ Dallas/Fort Worth, TX Los Angeles, CA Chicago, IL Atlanta, GA

REFERENCES Borenstein, S. (1989) “Hubs and High Fares: Dominance and Market Power in the U.S. Airline Industry,” Rand Journal of Economics, 20(3): 344–365. Borenstein, S. and N. Rose (2003) “The impact of bankruptcy on airline service levels”, American Economic Review Papers and Proceedings, May 2003, 415–419. Brueckner, Jan K and Spiller, Pablo T. (1994) “Economies of Traffic Density in the Deregulated Airline Industry,” Journal of Law & Economics, 37(2): 379–415. Brueckner, J.K. and W.T. Whalen (2000) “The Price Effects of International Airline Alliances,” Journal of Law and Economics, 43, 503–545. Deltas, G. (2002) “The small-sample bias of the Gini coefficient: results and implications for empirical research”, Review of Economics and Statistics, 85(1): 226–234. Dorfman, R. (1979) “A Formula for the Gini Coefficient”, Review of Economics and Statistics, 61(1): 146–149. Ellison, G. and E. Glaeser (1997) “Geographic concentration in US manufacturing industries: a dartboard approach”, Journal of Political Economy 105(5): 889–927. Evans, W.N and I.N. Kessides, (1994) “Living by the “Golden Rule”: Multimarket Contact in the U.S. Airline Industry”, Quarterly Journal of Economics, 109(2): 341–366. Garner, T. (1985) “Consumer Expenditures and Inequality: An Analysis Based on Decomposition of the Gini Coefficient”, Review of Economics and Statistics 75(1): 134–38. Gimeno, J. (1999) “Reciprocal Threats in Multimarket Rivalry: Staking out ‘Spheres of Influence’ in the U.S. Airline Industry,” Strategic Management Journal, 20, 101–128. Fournier, G. and T. Zuelke (2004) “Price effects of reciprocal multi-market contacts among Airline carriers”, paper presented at the 74th Annual Conference of the Southern Economics Association, New Orleans, November 2004. Kim, S. (1995) “Expansion of markets and the geographic distribution of economic activities: the trends in US manufacturing structure, 1860–1987, Quarterly Journal of Economics, 11(4): 881–901. Krugman, P.R. (1991) Geography and Trade., MIT Press, Cambridge, USA. Kosevoy and Mosler (1997) “Multivariate Gini Indices”, Journal of Multivariate Analysis 60, 52–276. Lerman, R.I. and Yitzhaki, S. (1984) A note on the calculation and interpretation of the Gini coefficient. Economics Letters 15, 363–368. Lerman, R.I. and Yitzhaki, S. (1985) “Income inequality effects by income source: a new approach and application to the U.S.” Review of Economics and Statistics 67, 151–156. Milanovic, B. and S. Yitzhaki (2002) “Decomposing World Income Distribution: Does The World Have A Middle Class?” Review of Income and Wealth, 48(2): 155–178.

364

AISLING REYNOLDS-FEIGHAN

Pyatt, G., C.Chen and J. Fei, (1980) “The distribution of income by factor components”, Quarterly Journal of Economics, 95(3): 451–173. Reynolds-Feighan, A.J. (2001) “Traffic distribution in low-cost and full-service carrier networks in the US air transportation market”, Journal of Air Transport Management, 7(5): 265–275. U.S. Department of Transportation (1996) “The Low Cost Service Revolution,” Office of Aviation and International Economics, Washington, D.C. U.S. General Accounting Office (1999): “Airline Deregulation: Changes in Airfares, Service Quality and Barriers to Entry,” March 1999, GAO/RCED 99–92, Washington, D.C. Yitzhaki, S. (1994) “Economic Distance and Overlapping of Distributions,” Journal of Economet­ rics, 61, 147–159. Yitzhaki, S and R. Lerman (1991 September) “Income Stratification and Income Inequality,” Review of Income and Wealth, 37(3): 313–329.

Advances in Airline Economics, Vol 2 Darin Lee (Editor) © 2007 Elsevier B.V. All rights reserved

16 Air Travel Demand Elasticities Concepts, Issues and Measurement David Gillen∗ , William G. Morrison† , and Christopher Stewart‡

ABSTRACT This paper reports the findings of a review of the economics and business literature on empirically-estimated own-price elasticities of demand for air travel for Canada and other developed countries. Insights are provided into key factors affecting this elasticity and judgements are made on elasticity values that may be representative for different markets for air travel in Canada.

1 INTRODUCTION The purpose of this study is to report on all or most of the economics and business literature dealing with empirically estimated demand functions for air travel and to collect a range of fare elasticity measures for air travel and provide some judgement as to which elasticity values would be more representative of the true values to be found in different markets in Canada. While existing studies may include the leisure – business class split, other important market distinctions are often omitted, likely as a result of data availability and quality.1

∗ Corresponding author. Sauder school of Business, University of British Columbia, 2053 Main Mall, Vancouver, BC Canada V6T1Z2. e-mail: [email protected]. † School of Business and Economics, Wilfrid Laurier University, 75 University Avenue West, Waterloo, ON Canada N2L 3C5. e-mail: [email protected]. ‡ School of Business and Economics, Wilfrid Laurier University, 75 University Avenue West, Waterloo, ON

Canada N2L 3C5.

1 In some cases separate equations are estimated for these markets; PODM (The Transport Canada air travel

forecasting model) for example uses different equations and variables for leisure and business markets.

366

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

One of the principal value-added features of this research and what distinguishes it from other surveys, is that we develop a meta-analysis that not only provides measures of dispersion but also recognizes the quality of demand estimates based on a number of selected study characteristics. In particular, we develop a means of scoring features of the studies such as focus on length of haul; business versus leisure; international versus domestic; the inclusion of income and inter-modal effects; the age of the study; data type (time-series versus cross-section) and the statistical quality of estimates (adjusted R-squared values). By scoring the studies in this way, policy makers are provided with a sharper focus to aid in judging the relevance of various estimated elasticity values.2

2 ELASTICITY IN THE CONTEXT OF AIR TRAVEL DEMAND Elasticity values in economic analysis provide a “units free” measure of the sensitivity of one variable to another, given some pre-specified functional relationship. The most commonly utilized elasticity concept is that of “own-price” elasticity of demand. In economics, consumer choice theory starts with axioms of preferences over goods that translate into utility values. These utility functions define choices that generate demand functions from which price elasticity values can be derived. Therefore elasticities are summary measures of people’s preferences reflecting sen­ sitivity to relative price levels and changes in a resource-constrained environment. The ordinary or Marshallian demand function is derived from consumers who are postu­ lated to maximize utility subject to a budget constraint. As a good’s price changes, the consumer’s real income (which can be used to consume all goods in the choice set) changes. In addition the goods price relative to other goods changes. The changes in consumption brought about by these effects following a price change are called income and substitution effects respectively. Thus, elasticity values derived from the ordinary demand function include both income and substitution effects.3 Own-price elasticity of demand measures the percentage change in the quantity demanded of a good (or service) resulting from a given percentage change in the good’s own-price, holding all other independent variables (income, prices of related goods etc.) fixed. The ratio of percentage changes thus allows for comparisons between the price sensitivity of demand for products that might be measured in different units (natural gas and electricity for example). “Arc” price elasticity of demand

2

Previous surveys (e.g. Oum et al., 1992) provide a listing of the elasticities and their ranges but no basis for choosing from among the values within the range. 3 Theoretically an alternative to the ordinary demand function is the compensated demand function, obtained from a resource expense minimization subject to a given level of utility. Elasticity values from the compensated demand function incorporate only substitution effects; however in practice we can estimate only the ordinary demand function. Nevertheless the distinction is important since large price changes may yield significant income effects.

AIR TRAVEL DEMAND ELASTICITIES

367

calculates the ratio of percentage change in quantity demanded to percentage change in price using two observations on price and quantity demanded. Formally this can be expressed as:

arc =

Q p P q

(1)

where: Q and P represent the observed change in quantity demanded and price p¯ and q¯ represent the average price and quantity demanded. The elasticity is unitless and can be interpreted as an index of demand sensitivity; it is measuring the degree to which a variable of interest will change (passenger traffic in our case) as some policy or strategic variable changes (total fare including any added fees or taxes in our case). In the limit (when Q and P are very small) we obtain the “point” own-price elasticity of demand expressed as:

point =

QP S p p q

(2)

where: Q(P, S) is the demand function P = a vector of all relevant prices p = the good’s own-price. q = equals the quantity demanded of the good S = a vector of all relevant shift variables other than prices (real income, demographic characteristics etc.) We expect own-price demand elasticity values to be negative, given the inverse rela­ tionship between price and quantity demanded implied by the “law” of demand, with absolute values less than unity indicating “inelastic” demand: a less than proportion­ ate response to price changes (relative price insensitivity). Similarly, absolute values exceeding unity indicate elastic or more sensitive demand: a more than proportionate demand response to price changes (relative price sensitivity). The ratio of change in quantity demanded to change in price [Equation (1)] high­ lights that elasticity measures involve linear approximations of the slope of a demand function. However, since elasticity is measuring proportionate change, elasticity values will change along almost all demand functions, including linear demand curves.4 Esti­ mation of elasticity values is therefore most useful for predicting demand responses in the vicinity of the observed price changes. As a related issue, analysts need to recognize that in markets where price discrimination is possible aggregate data will not allow for accurate predictions of demand responses in the relevant market segments. In air travel, flights by a carrier are essentially joint products consisting of differentiated service

4

The exception would obviously be the constant elasticity demand function.

368

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

bundles that are identified by fare classes. However the yield management systems employed by full-service carriers (FSCs) also create a complex form of inter-temporal price discrimination, in which some fares (typically economy class) decline and some increase (typically full-fare business class) as the departure date draws closer. This implies that ideally, empirical studies of air travel demand should separate business and leisure travellers or at least be able to include some information on booking times in order to account for this price discrimination, and that price data should be calibrated for inter-temporal price discrimination: for example, the use of full-fare economy class ticket prices as data will overestimate the absolute value of the price elasticity coeffi­ cient. Within the set of differentiated service bundles that comprise each (joint product) flight, the relative prices are important in explaining the relative ease of substitution between service classes. Given the nature of inter-temporal price discrimination for flights, the relative price could also change significantly in the time period prior to a departure time. The partial derivative in (2) indicates that elasticity measures price sensitivity inde­ pendent of all the other variables in the demand function. However when estimating demand systems over time, one can expect that some important shift variables will not be constant. It is important that these shift variables be explicitly recognized and incor­ porated into the analysis, as they will affect the value of elasticity estimates. This will also be true with some cross-sectional studies or panels.5 In particular changes in real income and the prices of substitutes or complements will affect demand. In air travel demand estimations, income and prices of other relevant goods should be included in the estimation equation. Alternative transportation modes (road and rail) are important variables for short-haul flights, while income effects should be measured for both shortand long-haul. The absence of an income coefficient in empirical demand studies will result in own-price elasticity estimates that can be biased. With no income coefficient, observed price and quantity pairs will not distinguish between movements along the demand curve and shifts of the demand curve.6 The slope of a demand function, which affects the own-price elasticity of demand, is generally expected to decrease (become shallower) with: • • • • •

The The The The The

number of available substitutes; degree of competition in the market or industry; ease with which consumers can search and compare prices; homogeneity of the product; duration of the time period analyzed.7

Given the implied relationships above, any empirical demand study should carefully define market boundaries to include all relevant substitutes and complements and

5

A panel is a data set that contains both time-series and cross-sectional information.

This will be true for all factors other than own-price.

7 The exception here is durable goods, for the opposite relationship is expected between long and short run

elasticities.

6

AIR TRAVEL DEMAND ELASTICITIES

369

to exclude products that might be related through income or other more general variables. In air travel, ideally market segment boundaries should be defined by first separating leisure and business passengers and second long- and short-haul flights. The reason is that we expect different behaviour in each of these markets. Within each of these categories, distinctions should then be made between the following: • Connecting and origin–destination (O–D) travel; • Hub and non-hub airports;8 • Routes with dominant airlines and routes with low-cost carrier competition. In addition, for the North American context, long-haul flights should be further divided into international and domestic travel (within continental North America). These mar­ ket segment boundaries are illustrated in Figure 2, which also highlights the relative importance of inter-modal competition for short-haul travel. While distinctions in price and income sensitivity of demand between business and leisure or long- and short-haul travel are more intuitive, other distinctions are perhaps less obvious. If available, data that distinguishes between routes, airlines and airports would provide important estimates of how price sensitivity is related to the number of competing flights and the willingness to pay of passengers utilizing a hub-and-spoke network, relative to those traveling point-to-point, more commonly associated with lowcost carriers. To the extent that existing studies assume that each passenger observation represents O–D travel, they will not be capturing fare premiums usually associated with hub-and-spoke networks and full-service carriers, nor will they necessarily capture the complete itinerary of travellers utilizing a number of point-to-point flights with a lowcost carrier. For example, a passenger who travels from Moncton to Vancouver with Air Canada, and utilizes the hub at Pearson International airport, is being provided with a number of services that includes baggage checked through to the final destination and frequent flyer points as well as a choice in flights and added flight and ground amenities. The fare for Moncton–Vancouver includes a premium for these services. Now consider a passenger who is travelling with WestJet from Moncton to Hamilton, and then with JetsGo from Toronto Pearson Airport to Vancouver. In this case there are no frequent flyer points to be attained and baggage has to be collected and re-checked after a road transfer between Hamilton and Pearson International. Although the origin and destination is the same for these passengers, the itineraries are significantly different. In many cases data used for demand estimates would not able to account for these differences. Route-specific data can also capture competition that may exist between airports and the services they offer as well as airlines. This may be especially true for certain shorthaul routes where inter-modal competition (road and rail) can play an important role in shaping air travel demand (Figure 1).

8

The difference between this point and the previous one is that hub airports will have different service levels and will generally have a hub premium.

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

370

Leisure Travel

Business Travel

Leisure Travel

International Travel

Rail Travel Demand

Business Travel

International Travel

Hub Airport Air Travel Demand

Short-Haul

Hub Airport Long-Haul

Point-to-Point Airport

Point-to-Point Airport Domestic Travel

Leisure Travel

Road Travel Demand

Business Travel

Domestic Travel

Leisure Travel

Business Travel

Figure 1 Market Segments in Air Travel Demand.

3 MEASUREMENT ISSUES Oum et al. (1992) provide a valuable list of pitfalls that occur when demand models are estimated and therefore affect the interpretation of the elasticity estimates from these empirical studies. 1. Price and Service Attributes of Substitutes: Air travel demand can be affected by changes in the prices and service quality of other modes. For short-haul routes (markets) the relative price and service attributes of auto and train would need to be included in any model; particularly for short-haul markets. Failure to include the price and service attributes of substitutes will bias the elasticity. For example, if airfares increase and auto costs are also increasing, the airfare elasticity would be overestimated if auto costs were excluded. 2. Functional Forms: Most studies of air travel demand use a linear or log-linear func­ tional specification. Elasticity estimates can vary widely depending on the functional form. The choice of functional form should be selected on the basis of statistical testing not ease of interpretation. 3. Cross-Section vs. Time-series Information: In the long run demand elasticities for non-durable goods and services are larger in absolute terms, than in the short run. This follows because in the long run there are many more substitution possibilities that can be used to avoid price increases or service quality decreases. In effect there are more opportunities to avoid these changes with substitution possibilities. Data tends to be cross-sectional or time-series although more recently panels have become available. A panel is a combination of cross-section and time-series – information on several routes for a multi-year period is a panel. Cross-sectional information is generally

AIR TRAVEL DEMAND ELASTICITIES

371

regarded as indicating short-run elasticities while time-series data is interpreted as long-run elasticities. In time-series data the information reflects changes in markets, growth in income, changes in competitive circumstances, for example. Policy changes should rely on long-run elasticities since these are long-run impacts that are being modelled. Short-run elasticities become important when considering the competitive position of firms in a highly dynamic and competitive industry. 4. Market Aggregation/Segmentation: As the level of aggregation increases the amount of variation in the elasticity estimates decreases. This occurs because aggregation averages out some of the underlying variation relating to specific contexts. Since air travel market segments may differ significantly in character, competition and dominance of trip purpose, interpreting a reduction in variation through aggregation as a good thing would be erroneous. Such estimates might have relatively low standard deviations but would be also be relatively inaccurate when used to assess the effect of changes in fares in a specific market. 5. Identification Problem: In most cases only demand functions are estimated in attempts to measure the demand elasticity of interest. However, it is well known that the demand function is part of a simultaneous equations system consisting of both supply and demand functions. Therefore, a straightforward estimation of only the demand equation will produce biased and inconsistent estimates. The problem of identification can be illustrated by describing the process by which fares and travel, for example, are determined in the O–D market simultaneously. To model this process in its entirety, we must develop a quantitative estimate of both the demand and supply functions in a system. If, in the past, the supply curve has been shifting due to changes in production and cost conditions for example, while the demand curve has remained fixed, the resultant intersection points will trace out the demand function. On the contrary, if the demand curve has shifted due to changes in personal income, while the supply curve has remained the same, the intersection points will trace out the supply curve. The most likely outcome, however, is movement of both curves yielding a pattern of fare, quantity intersection points from which it will be difficult, without further information, to distinguish the demand curve from the supply curve or estimate the parameters of either.9 Earlier we identified sources of bias that can arise from problems with aggregation, data quality, implicit assumptions of strong separability among others. Almost all demand studies have an implied assumption of strong separability in that they only consider aviation markets in the analysis. Such studies in effect constrain all changes or responses in fares or service to be wholly contained in the aviation component of people’s con­ sumption bundle. The paper by Oum and Gillen (1986) is the one exception where consideration of substitution with other parts of consumption was included in the mod­ elling. It would be difficult to extract a conclusion from this one study as to existence, degree and direction of bias in elasticity estimates when other parts of consumption are

9 Fortunately, several techniques have been developed for the estimation of the structural parameters of an a priori specified system of simultaneous stochastic equations. These include indirect least squares, two stage least squares, instrumental variables, three stage least squares, full information maximum likelihood, and limited information maximum likelihood.

372

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

and are not included in the modelling. However, having said this, an inspection of the elasticity estimates from this study shows they are not significantly different than other time-series estimates.

3.1 Data Issues Elasticity estimates depend critically on the quality and extent of the data available. Currently, the best data for demand estimation is the DB1A 10 percent ticket sample in the US, but even this data has some problems.10 The DB1A sample represents 10 percent of all tickets sold with full itinerary identified by the coupons attached to the ticket. However with electronic tickets, as more and more tickets are being sold over the Internet, there is a growing portion of overall travel that may not be captured in the sample. This means that the proportion is not 10 percent but something less.11 Other important considerations are the amount of travel on frequent flyer points, by crew and airline personnel. In Canada we have poor quality data because it is incomplete, even if it were accessible. Airports collect traffic statistics but these data make it very difficult to distinguish O–D and segment data. Airlines report traffic data to Statistics Canada (or are supposed to) but these data do not include fare information or routing. Knowing the itinerary or routing is important because of differences in service quality and hubbing effects. Fare data is also more useful than yield information since it identifies the proportion of people travelling in different fare classes. Yet, in many cases yield information is used as a weighted average fare. There is also the problem that carriers of different size may have different reporting requirements. Some researchers and consultants have been cobbling together data sets for analysis by using the PBX clearing house information. These data are limited and apply only to those airlines that are members of IATA.12 The current public data available in Canada simply does not permit estimation of any demand models. Besides demand-side data it is also important to have supply-side information. Elas­ ticity estimates should emerge from a simultaneous equations framework. This data is more accessible through organizations like the OAG13 , which provide information on capacity, airline and aircraft type for each flight in each market.14 These data measure changes in capacity, flight frequency and timing of flights. One study, which undertook an extensive survey to collect multi-modal data,15 was the High Speed Rail study sponsored jointly by the Federal, Ontario and Quebec governments. This study, which had three different demand modelling efforts, exam­ ined the potential for High Speed Rail demand, and subsequent investment, in the

10

The term ‘best’ means researchers observe this data source to be the most geographically comprehensive,

detailed and temporally available.

11 The growth of the Internet in booking tickets is being integrated into the DB1A database, as is the growing

use of electronic tickets.

12 IATA is the International Air Transport Association.

13 OAG is the Official Airline Guide.

14 These data are sold and can be expensive.

15 Estimates were that in excess of $1 Million was spent on data collection alone.

AIR TRAVEL DEMAND ELASTICITIES

373

Windsor–Quebec corridor. The analysis included inter-modal substitution between air, rail, bus and car. The study was undertaken in the early 1980s. However, it is not possible for public access to any of the technical documents that would allow an assessment of the study. Attempts in the past to obtain access to the data have proven fruitless.

3.2 Distinguishing Elasticity Measures As we have stated, price elasticity measures the degree of responsiveness to a change in own or other prices (fares). However, care must be exercised in interpreting the elasticity since they differ according to how they have been estimated. Many empirical studies of air travel demand estimate a log-linear model. In evaluating such studies, it is important to keep in mind that the empirical specification implies a certain consumer preference structure because of the duality between utility functions and demand functions. It is equally important to remember that empirically estimated demand functions should contain some measures of quality and service differences or quality changes over time. Failure to include metrics for frequent flyer programs, flight frequency, destination choice or service levels in estimating an air demand function can lead to downward bias in the price elasticity estimates. Price elasticities can be estimated for aggregate travel demand as well as modal demand. Figure 2 illustrates the differences between aggregate and modal elasticities.16 Our interest is in modal elasticities not the aggregate amount of travel but it is important ultimately that any policy analysis take account of the impact of any policy change on aggregate travel as well as modal redistribution. The impact of a change in price on aggregate demand would be measured by the – fi s in Figure 2 whereas the Fii s would measure the impact on air travel demand. The Fii s are a composite or combination of the fi s and the Mii s. The Canadian aviation industry has undergone significant change in the last several years. In 2000 Air Canada completed its takeover of Canadian Airlines, which left it

Elasticity of Aggregate Market Demand for Transport

F

Fii

Fij

Fjj

Mode Specific

Elasticities and Cross-

Elasticities

–fi

–fj

–fj

Elasticity of Aggregate

Market Demand w.r.t.

Price of Mode i

Mii

Mij

Mjj

Mode Choice Elasticities

Figure 2 Aggregate and Modal Elasticities.

16

This figure is adapted from Figure 1 in Oum et al. (1992)

374

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

with an excess of 80 percent market share. Market dominance leads to different fare and service-quality levels. As a result of higher fares, for example, we should find higher absolute values of elasticities of demand simply because with higher fares we have moved further up the demand curve. In 1996 Westjet entered the market and has continued to grow each year. Canada 3000 exited the market in 2001, as did Canjet and Royal (as part of Canada 3000). Roots airline has come and gone but Canjet has reemerged in eastern Canada and JetsGo is offering some level of service on longer haul domestic flights as well as in the Montreal–Toronto market. The entry of low-cost carriers leads to lower fares for a subset of traffic and competitors will offer a supply of seats to match these fares. Lower average fares should lead to lower demand elasticity estimates, while increases in the number of competitors in the market will lead to higher demand elasticity estimates. One should not confuse low-cost carriers with a seeming lack of exploiting monopoly power. High prices or fares are not synonymous with monopoly and low fares with competition. Airlines like Westjet where they are the sole airline serving the market may still act as a monopolist but charge low(er) fares. Profit maximizing monopolists price where marginal cost equals marginal revenue, if marginal cost is low, one should expect to see lower fares but still marginal cost and revenue are equalized. Monopolists are generally viewed as being high price because they are of high costs that are attributable to some degree from a lack of competitive discipline in the market. Full service carriers operating with hub-and-spoke systems have a high-cost business model while low-cost carriers have a low-cost business model.

4 EVALUATION OF ELASTICITY STUDIES Overall we have collected some 254 demand elasticity estimates from 21 studies. Each of these studies is described, using a standardized summary sheet; illustrated in Appendix A. To aid our understanding of how existing elasticity estimates might inform policy makers in forecasting air travel demand, we provide a descriptive meta-analysis of various distributions of estimated values in Section 4.1. We next develop a weighted scoring table with respect to generally desirable data, design and output characteristics of the studies. This allows us to generate a rank ordering of the studies, from which to generate a sub-sample of estimates from studies with a “passing grade” score. A passing grade is simply defined as 50 percent of the maximum score attainable. From these studies we provide suggested ranges of elasticity values in six key market segments: 1. 2. 3. 4. 5. 6.

Short-haul business travel Short-haul leisure travel Long-haul, domestic business travel Long-haul, domestic leisure travel Long-haul, international business travel Long-haul, international leisure travel

Before we discuss our scoring system for the studies, we first present some more general descriptive information on the distribution of estimated elasticity values in various categories.

AIR TRAVEL DEMAND ELASTICITIES

375

4.1 Descriptive Distributions of Elasticity Estimates Here, we present for the aggregate and for several important sub-categories, histograms of the estimates in the studies we have researched. We begin with the most general dis­ tribution: the set of all the studies containing some 254 estimates of own-price elasticity. We next present sub-categories in increasing detail defined in terms of market charac­ teristics. We also present sub-samples of the estimates based on data type (cross-section versus time-series) and the age of the study (less than 5 years old, versus between 5 and 10 years old). In each case we report the median value as a measure of central tendency, along with the kurtosis and skewness of the distributions.17 4.1.1 All Studies 18 We generate a histogram for all own-price elasticities with 254 estimates taken from 21 studies.19 The minimum estimated elasticity value is −3.20.20 The histogram, Figure 3, demonstrates a crowd of estimates between zero and −2.5. The median, or midpoint, of all estimates is −1.122. We use the median as the measure of central tendency, as opposed to the mean, in order to remove the effects of outliers in our data set. The skewness of the histogram is (−0.37). This indicates that our data is not normally distributed (Table 1). 4.1.2 All Long-haul Studies 21 We sub-divide the aggregate data into a subset of long-haul own-price elasticity esti­ mates. The data set includes estimates for distances greater than 1500 miles, or estimates that are reported as “long-haul” or “international” in their respective study. The subset is

17

In computing numerically descriptive measures of data we are generally interested in two indicators: the measure of average or central value of the data and the measure of the degree to which the data are spread out about this average value. The most popular measure of central tendency is the arithmetic mean (known generally as simply the mean) but this can be unduly influenced by extreme observations; which is the case here. We have therefore used the median as the measure of central location. The median is the value that splits the data set exactly in half. The measure of spread is called the variance; how widely the data are dispersed around the measure of central location. Two added descriptors of the distribution of the data are skewness and kurtosis. The first, skewness, describes whether the data are distributed symmetrically around the measure of central location. If they are not, they are said to be ‘skewed’ or have a short tail on one side and a longer tail on the other rather than having equally sized tails in each side. Kurtosis is a measure of whether the data are peaked or flat relative to a normal distribution. That is, data sets with high kurtosis tend to have a distinct peak near the measure of central location, decline rather rapidly, and have heavy tails. Data sets with low kurtosis tend to have a flat top near the mean rather than a sharp peak. Since in many cases, the distribution of estimated values is skewed, the mean is not a useful measure of central tendency and the standard deviation is not useful in providing a range around the mean. 18 The data-set is comprised of all 21 studies located in the appendix. 19 The study by Anderson and Kraus (1981) is not included in this histogram since they do not calculate elasticities directly. Doubt over the quality of estimated positive elasticities in Jung and Fuji (1976) lead us to exclude their estimates also. 20 It is conventional to present own-price elasticities with a negative sign indicating the general negative relationship between price and quantity demanded. Larger values of the elasticity imply greater price sensitivity while lower values imply less price sensitivity. 21 Source: Bhadra (2002), Hamal (1998), Taplin (1997), BTCE (1995), Oum et al. (1986), BTE (1986), Lubulwa (1986), May and Butcher (1986), BIE (1984), Abrahams (1983), Hollander (1982), Taplin (1980).

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

376

90 80

60 50 40 30

Frequency

70

20 10 0.5

0

–0.5

–1

–1.5

–2

–2.5

–3

–3.5

0

Own-Price Elasticities

Figure 3 Histogram of All Own-Price Elasticities for All Studies.

Table 1 Summary Statistics of All Studies-Own Price Elasticities 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−1.967 −1.418 −1.122 −0.633 −0.190 0.785 254 −3.200 0.040 0.312 −0.370 0.177

comprised of 100 estimates with a median elasticity of −0.857. A majority of the values are bunched up between zero and −2 as indicated by the skewness of the histogram at −0.275 (Figure 4, Table 2). 4.1.3 All Short/medium-haul Studies 22 The data set for short/medium-haul own-price elasticity estimates includes estimates for distances less than 1500 miles, or estimates that are reported as “short-haul”, “medium­ haul”, or “regional” in their respective study. The subset is comprised of 109 estimates.

22 Source: Battersby and Oczkowski (2001), Bhadra (2002), Nairn (1992), Oum et al. (1986), BTE (1986), Lubulwa (1986), May and Butcher (1986), Abrahams (1983).

AIR TRAVEL DEMAND ELASTICITIES

377

12

8 6 4

Frequency

10

2

0

–0.1

–0.2

–0.3

–0.4

–0.5

–0.6

–0.7

–0.8

–1

–0.9

–1.1

–1.2

–1.3

–1.4

–1.5

–1.6

–1.7

–1.8

–2

–1.9

–2.1

–2.2

–2.3

–2.4

0

Own-Price Elasticities (Long-Haul)

Figure 4 Histogram of Long-Haul Own-Price Elasticities. Table 2 Summary Statistics of All Long-Haul Own-Price Elasticity Estimates 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−1.851 −1.365 −0.857 −0.495 −0.190 0.870 100 −2.234 −0.010 0.298 −0.275 −0.874

Note that the sum of long-haul and short/medium-haul estimates (100 + 109) does not equal the number of estimates in the aggregate data set. This is a result of the exclu­ sion of elasticity estimates that are not defined by their distance in their respective reports. The median elasticity in this subset is −1.15. A crowd of estimates is located between zero and −1.5. The minimum value (−3.20) represents a Sydney-Brisbane route taken from Milloy et al. (1985). The skewness of the histogram is −0.434 (Figure 5, Table 3). 4.1.4 All Long-haul International Travel Estimates 23 This sub-category of long-haul international travel is comprised of 69 estimates extracted from the aggregate data set. The data set represents estimates for country-to-country

23 Source: Hamal (1998), Taplin (1997), BTCE (1995), Lubulwa (1986), BIE (1984), Hollander (1982), Taplin (1980).

378

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

25

15

10

Frequency

20

5

0.5

0

–0.5

–1

–1.5

–2

–2.5

–3

0

Own-Price Elasticities (Short/Medium-haul)

Figure 5 Histogram of Short/Medium-Haul Own-Price Elasticities.

Table 3 Summary Statistics of All Short/ Medium-Haul Own-Price Elasticity Estimates 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−1.992 −1.520 −1.150 −0.712 −0.112 0.808 109.000 −3.200 0.040 0.329 −0.434 0.710

international travel taken from seven studies. The estimates are distributed between zero and −2.7, with some crowding below −0.5. The median elasticity is −0.79 and the distribution is somewhat skewed (Figure 6, Table 4). 4.1.5 All Long-haul Domestic Studies 24 This subset is comprised of 36 estimates extracted from six studies. The majority of the estimates are bunched between the maximum value (−0.44) and −1.81. The skewness is 0.168 (Figure 7, Table 5)

24 Source: Oum et al. (1986), May and Butcher (1986), Lubulwa (1986), BTE (1986), Bhadra (2002), Abrahams (1983).

AIR TRAVEL DEMAND ELASTICITIES

379

9 8 7

5 4

Frequency

6

3 2 1 0

–0.2

–0.4

–0.6

–0.8

–1

–1.2

–1.4

–1.6

–1.8

–2

–2.2

–2.4

–2.6

–2.8

0

Own-Price Elasticities (Long-haul International)

Figure 6 Histogram of All Long-Haul International Estimates.

Table 4 Summary Statistics of Long-Haul International Travel Own-Price Elasticities 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−1.960 −1.400 −0.790 −0.349 −0.172 1.051 69.000 −2.700 −0.010 0.407 −0.672 −0.456

4.1.6 All Long-haul, International Business Travel Estimates 25 The international business travel subset contains 16 estimates from two studies. A majority of the estimates (15) are calculated by the Bureau of Transport Communi­ cations and Economics (1995) for business travellers to and from Australia. The lowest estimate (−2.0) represents Australian business travellers to the UK The majority of the estimates are bunched between the maximum value (−0.01) and −0.6. The median elas­ ticity estimate is −0.265. The histogram is negatively skewed (−2.405), which indicates a non-normal distribution (Figure 8, Table 6).

25

Source: BTCE (1995), Lubulwa (1986).

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

380

9.00 8.00

6.00 5.00 4.00 3.00

Frequency

7.00

2.00 1.00 0.00

–0.10

–0.20

–0.30

–0.40

–0.50

–0.60

–0.70

–0.80

–0.90

–1.00

–1.10

–1.20

–1.30

–1.40

–1.50

–1.60

–1.70

–1.80

–1.90

0.00

Own-Price Elasticities (Long-haul )

Figure 7 Histogram of All Domestic Long-Haul Own-Price Elasticities.

Table 5 Summary Statistics of All Long-Haul Domestic Travel Own-Price Elasticities 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−1.685 −1.528 −1.150 −0.828 −0.553 0.700 36.000 −1.810 −0.440 0.149 0.168 −1.078

4.1.7 All Long-haul International Leisure Travel Estimates 26 The long-haul leisure travel segment contains a total of 55 estimates, representing seven studies. Nearly 50 percent of the estimates (24) are taken from the Bureau of Transport Communications and Economics (1995) study. The median of the estimates is −0.993 with estimates distributed between −0.14 and −2.7. The minimum values (−2.7) are taken from Taplin (1980) and represent elasticity estimates calculated by Jud and Joseph (1974) (for travel from the US to Latin America), and from Straszheim (1978) (for high discount travel). The skewness of the histogram is −0.555 (Figure 9, Table 7).

26 Source: Hamal (1998), Taplin (1997), BTCE (1995), Lubulwa (1986), BIE (1984), Hollander (1982), Taplin (1980).

AIR TRAVEL DEMAND ELASTICITIES

381

7 6

4 3

Frequency

5

2 1

0

–0.2

–0.4

–0.6

–0.8

–1

–1.2

–1.4

–1.6

–1.8

–2

0

Own-Price Elasticities (Long-haul)

Figure 8 Histogram of All Long-Haul International Business Travel Own-Price Elasticities. Table 6 Summary Statistics of All Long-Haul

International Business Travel Own-Price Elasticities

5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−1.423 −0.475 −0.265 −0.198 −0.093 0.278 16.000 −2.000 −0.010 0.251 −2.405 6.095

4.1.8 All Long-haul Domestic Business Estimates 27 The long-haul domestic business travel subset is comprised of 26 estimates from two studies. The estimates are bunched up between the −0.5 and −1.6. The median of the histogram is −1.15. The skewness (0.270) indicates a non-normal distribution (Figure 10, Table 8). 4.1.9 Long-haul Domestic Leisure Histogram 28 The long-haul domestic leisure travel subset is comprised of seven estimates from two studies. The estimates are distributed between −0.44 and −3.20. The median elasticity is −1.120 (Figure 11, Table 9).

27 28

Source: Oum et al. (1986), Lubulwa (1986). Source: Lubulwa (1986), Abrahams (1983).

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

382

10 9 8

6 5 4

Frequency

7

3 2 1 0

–0.2

–0.4

–0.6

–0.8

–1

–1.2

–1.4

–1.6

–1.8

–2

–2.2

–2.4

–2.6

–2.8

0

Own-Price Elasticities (Long-haul)

Figure 9 Histogram of All Long-Haul International Leisure Travel Own-Price Elasticities. Table 7 Summary Statistics of All Long-Haul International Leisure Own-Price Elasticities −2.070 −1.650 −0.993 −0.535 −0.220 1.115 55.000 −2.700 −0.140 0.423 −0.555 −0.393

5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

7 6

4 3

Frequency

5

2 1 0

–0.1

–0.2

–0.3

–0.4

–0.5

–0.6

–0.7

–0.8

–0.9

–1

–1.1

–1.2

–1.3

–1.4

–1.5

–1.6

–1.7

0

Own-Price Elasticities (Long-Haul)

Figure 10 Histogram of Long-Haul (Domestic-Business) Own-Price Elasticities for All Studies.

AIR TRAVEL DEMAND ELASTICITIES

383

Table 8 Summary Statistics of All Long-Haul Domestic Business Own-Price Elasticities −1.670 −1.428 −1.150 −0.836 −0.780 0.591 26.000 −1.700 −0.543 0.113 0.207 −1.119

5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

4

2

Frequency

3

1

0

–0.5

–1

–1.5

–2

–2.5

–3

–3.5

0

Own-Price Elasticities (Long-haul)

Figure 11 Histogram of Long-Haul (Domestic-Leisure) Own-Price Elasticities for All Studies. Table 9 Summary Statistics of All Long-Haul Domestic Leisure Own-Price Elasticities 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−2.744 −1.472 −1.120 −0.887 −0.514 0.585 7.000 −3.200 −0.440 0.821 −1.640 3.265

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

384

4.1.10 All Short-haul Business Travel Estimates 29 The short-haul business travel subset is comprised of 18 estimates taken from four studies. The median elasticity is −0.73. The histogram demonstrates some crowding of values between −0.5 and −0.8 (Figure 12, Table 10). 4.1.11 All Short-haul Leisure Travel Estimates 30 This subset is comprised of 19 estimates from five studies. The median elasticity is −1.52 with estimates distributed across the range of values with little crowding. The

6

4 3 2

Frequency

5

1

0

–0.1

–0.2

–0.3

–0.4

–0.5

–0.6

–0.7

–0.8

–0.9

–1

–1.1

–1.2

–1.3

–1.4

–1.5

–1.6

–1.7

0

Own-Price Elasticities (Short-haul)

Figure 12 Histogram of All Short-Haul Business Travel Own-Price Elasticities.

Table 10 Summary Statistics of All Short-Haul Business Travel Own-Price Elasticities 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

29 30

−1.169 −0.798 −0.730 −0.608 −0.126 0.190 18.000 −1.500 −0.100 0.106 −0.151 1.509

Source: Battersby and Oczkowski (2001), Nairn (1992), Lubulwa (1986), Oum et al. (1986). Source: Nairn (1992), Oum et al. (1986), BTE (1986), May and Butcher (1986), Abrahams (1983).

AIR TRAVEL DEMAND ELASTICITIES

385

7 6

4 3

Frequency

5

2 1

0

–0.2

–0.4

–0.6

–0.8

–1

–1.2

–1.4

–1.6

–1.8

–2

–2.2

–2.4

0

Own-Price Elasticities (Short-haul)

Figure 13 Histogram of All Short-Haul Leisure Travel Own-Price Elasticities.

Table 11 Summary Statistics of All Short-Haul Leisure Travel Own-Price Elasticities 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−2.307 −1.745 −1.520 −0.885 −0.688 0.860 19.000 −2.370 −0.400 0.307 0.158 −0.704

histogram is positively skewed (0.158), which indicates that the number of estimates decrease as we approach zero (Figure 13, Table 11). 4.1.12 All Cross Section Study Estimates 31 The subset of all cross-sectional studies is comprised of 85 estimates, of which 80 estimates are taken from Oum et al. (1986) and represent US city-pair routes. All of the estimates are taken from studies between 1981 and 1986. The median elasticity is −1.33. The histogram is positively skewed (0.314) (Figure 14, Table 12).

31 Source: Oum et al. (1986), Talley (1988), Agarwal and Talley (1985), Morrison and Winston (1985), Ippolito (1981).

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

386

18 16

12 10 8 6

Frequency

14

4 2

0

–0.1

–0.2

–0.3

–0.4

–0.5

–0.6

–0.7

–0.8

–0.9

–1

–1.1

–1.2

–1.3

–1.4

–1.5

–1.6

–1.7

–1.8

–1.9

–2

0

Own-Price Elasticities (Cross-section)

Figure 14 Histogram of Own-Price Elasticities for All Cross-Section Studies. Table 12 Summary Statistics of All Cross-Section Study Own-Price Elasticities 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−1.766 −1.520 −1.330 −0.810 −0.606 0.710 85.000 −2.010 −0.181 0.158 0.314 −0.563

4.1.13 All Time-series Study Estimates 32 This subset is comprised of 136 estimates, 28 of which are taken from studies published within the last 5 years. The histogram is negatively skewed with a crowd of estimates between zero and −2. The median elasticity is −0.847 (Figure 15, Table 13). 4.1.14 Studies 5–10 Years Old 33 Two subsets have been created based on the age of the studies. The first subset is comprised of estimates calculated in studies published between 1992 and 1997. This subset contains 45 estimates from two studies (Figure 16, Table 14).

32 Source: Bhadra (2002), Battersby and Oczkowski (2001), Hamal (1998), BTCE (1995), Fridstroom (1989),

BTE (1986), May and Butcher (1986), BIE (1984), Abrahams (1983), Andrikopoulos (1983), Oum and Gillen

(1982), Hollander (1982), Taplin (1980).

33 Source: BTCE (1995), Nairn (1992).

AIR TRAVEL DEMAND ELASTICITIES

387

35 30

20 15

Frequency

25

10 5

0.5

0.25

0

–0.25

–0.5

–0.75

–1

–1.25

–1.5

–1.75

–2

–2.25

–2.5

–2.75

0

Own-Price Elasticities (Time-series)

Figure 15 Histogram of Own-Price Elasticities for All Time-Series Studies. Table 13 Summary Statistics of All Time-Series Study Estimates −1.870 −1.196 −0.847 −0.470 −0.138 0.726 136.000 −2.540 0.040 0.313 −0.542 −0.227

5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

8 7

5 4 3 2

0

–0.1

–0.2

–0.3

–0.4

–0.5

–0.6

–0.7

–0.8

–1

–0.9

–1.1

–1.2

–1.3

–1.4

–1.5

–1.6

–1.7

–1.8

–1.9

–2

–2.1

–2.2

–2.3

–2.4

1 0

Own-Price Elasticities

Figure 16 Histogram of Aggregate Own-Price Elasticity for Studies 1992–97.

Frequency

6

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

388

Table 14 Summary Statistics of Estimates for All Studies 1992–1997 −1.972 −1.160 −0.560 −0.290 −0.124 0.870 45.000 −2.300 −0.010 0.369 −0.907 −0.152

5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

The median elasticity is −0.56 with the majority of estimates residing between −0.1 and −1.1. The histogram is negatively skewed with a skewness of −0.907, which indicates a non-normal distribution. The second subset of estimates based on the age of the study is comprised of estimates calculated in studies published between 1997 and 2002.34 Four studies qualify for this subset resulting in 30 estimates. The histogram demonstrates no crowding around a small range of values. Instead, there is a wide distribution of values between zero and −2.3. The median elasticity is −0.847 (Figure 17, Table 15). In comparing the median elasticity value for 1997–2002 studies (−0.847) with the median elasticity for studies produced between 1992–97 (−0.56), it would appear that own-price elasticity of demand has become more price sensitive (elastic) over time.

4.5 4

2.5 2 1.5 1

0.1

–0.3

–0.7

–1.1

–1.5

–1.9

–2.3

0.5 0

Own-Price Elasticities

Figure 17 Histogram for Own-Price Elasticities for All Studies 1997–2002.

34

Bhadra (2002), Battersby and Oczkowski (2001), Hamal (1998), Taplin (1997).

Frequency

3.5 3

AIR TRAVEL DEMAND ELASTICITIES

389

Table 15 Summary Statistics of Own-price Elasticity Estimates for All Studies 1997–2002 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−1.978 −1.368 −0.847 −0.484 −0.084 0.883 30.000 −2.234 0.040 0.407 −0.426 −0.731

However, interpreting this change is not straightforward. The completion date of a study does not map directly to the age of the data employed. For example, the Nairn (1992) study utilizes data from 1983–1984, while the Hamal study (1998) uses time series data from 1974–96. Furthermore, the comparison becomes less informative when we examine the range between first and third quartiles for each distribution. The range for 1997–2002 studies is −0.5 to −1.4 while the range for 1992–97 studies is −0.3 to −1.2. 4.1.15 All Income Elasticities 35 The subset of all income elasticities contains 132 estimates from 14 studies. The min­ imum estimated elasticity value is −1.21, which represents inbound pleasure travel to Australia from the US, as calculated by Hollander (1982). The maximum value is 11.58, which is calculated in the Bureau of Transport Communications and Economics (1995) report for leisure travel by Australian residents to Taiwan. The median estimate is 1.39. There is a crowd of estimates bunched up between 0.5 and 2.5 (Figure 18, Table 16). 4.1.16 Summary Table 17 summarizes the median values of estimated own-price elasticities by market segment and study characteristics (data type and age). The table indicates that there are significant differences between some market segment elasticities (long-haul international business and short-haul leisure in particular) and the median value for all estimates (−1.122). Time-series estimates indicate relatively less price sensitivity than those derived from cross-section studies. Moreover, recent studies have returned relatively elastic values compared with older studies.

35 Source: Battersby and Oczkowski (2001), Hamal (1998), Taplin (1997), BTCE (1995), Fridstroom (1989), Oum et al. (1986), BTE (1986), Lubulwa (1986), BIE (1984), Abrahams (1983), Oum and Gillen (1982), Hollander (1982), Ippolito (1981), Taplin (1980).

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

390

30

Frequency

25 20 15 10 5

Own-Price Elasticities

Figure 18 Histogram of Aggregate Income Elasticities for All Studies. Table 16 Summary Statistics of All Studies Income Elasticities 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness

0.249 0.840 1.390 2.169 4.640 1.329 132.000 −1.210 11.580 2.506 2.671

Table 17 Summary of Median Elasticity Values by Type Category All All All All All All All All All All All All All All All

estimates long haul estimates long-haul international estimates long-haul international business estimates long-haul international leisure estimates long-haul domestic estimates long-haul domestic business estimates long-haul domestic leisure estimates short/medium haul estimates short/medium haul business estimates short/medium haul leisure estimates cross-section study estimates time-series study estimates estimates from studies 1992–1997 estimates from studies 1997–2002

Median Own-price Elasticity value −1.122 −0.857 −0.790 −0.265 −0.993 −1.150 −1.150 −1.120 −1.150 −0.730 −1.520 −1.330 −0.847 −0.560 −0.847

11

11.5

10

10.5

9

9.5

8

8.5

7

7.5

6

6.5

5

5.5

4

4.5

3

3.5

2

2.5

1

1.5

0

0.5

–1

–0.5

–1.5

0

AIR TRAVEL DEMAND ELASTICITIES

391

5 SCORING THE STUDIES To improve the level of confidence regarding the practical use of elasticity values in forecasting air travel demand, we developed a scoring system based on desirable input and output characteristics of empirical demand studies. Following on from our earlier discussion of theoretical and measurement issues, we have identified 11 characteristics that contribute to the quality of elasticity estimates. In each case, the point scores represent our assessment of the relative importance of either the inclusion or exclusion of the characteristic in question. We readily acknowledge that the subjective assignment of point scores cannot provide definitive scientific results. Nevertheless we feel that in the absence of time or resources for more sophisticated analysis (meta regression analysis, risk analysis and bootstrapping techniques for example), the scoring rule provides a useful rule of thumb for comparing the reliability of estimates we feel should be of higher quality with the overall set of estimates from all the studies surveyed. Specifically, we have rated the studies based on the following characteristics (Table 18): i. Separation of business and leisure travel We expect business travel to be more price insensitive than leisure travel. Consequently studies that do not distinguish between these market segments are likely to provide elas­ ticity estimates that would be biased if applied in any detailed analysis whether applied to specific business or leisure market segments, or to routes, which are predominantly business or leisure oriented. Scoring rule: Estimates for both business and leisure = 3 points Estimates for either business or leisure = 2 points No separation of business and leisure = 0 points ii. Separation of Long-haul vs. short-haul travel We expect less price sensitivity for long-haul flights than for short-haul flights (where more inter-modal substitution is possible). In a similar fashion to the business/leisure Table 18 Set of Characteristics for Weighting Studies i. ii. iii. iv. v. vi. vii. viii. ix. x. xi.

Separation of business and leisure travel Separation of long-haul vs. short-haul travel Inclusion of an income coefficient Inclusion of intermodal substitution Data type: panel vs. time-series vs. cross-section Country focus Route-specific estimates Hub vs. non-hub airports Connecting vs. O–D passengers Age of the study Adjusted R-squared values

392

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

distinction, studies that do not distinguish market segments by flight length will provide elasticities that underestimate price sensitivity for short-haul flights and over-estimate it for long-haul flights. Scoring rule: Estimates for both long and short-haul = 3 points Estimates for either long or short-haul = 2 points No separation of long and short-haul = 0 points iii. Inclusion of an income coefficient Without an income coefficient, demand studies will confuse a shift of the demand curve with movements along the demand curve. With a positive income elasticity for air travel, and increasing per-capita real income, demand studies with no income coefficient will overestimate the absolute price elasticity of demand for price decreases and underestimate it for price increases. Scoring rule: Income coefficient = 2 points No income coefficient = 0 points iv. Inclusion of inter-modal substitution ( for short-haul flights) The shorter the distance comprising a trip, the more road and trail transportation become effective substitutes for air travel. Therefore we would expect the price and other char­ acteristics of alternative modes to have a more significant (shift) impact on the demand for short-haul air travel, ceteris paribus. Studies of short-haul flights that do not include inter-modal effects are likely to provide bias estimates if the shadow prices of alternative modes change. The scoring rule in this case attempts to award short-haul studies that incor­ porate inter-modal effects, without penalizing studies of longer-haul air travel. Scoring rule: Inter-modal substitution in short-haul study = 2 points Not a short-haul study = 1 point No inter-modal substitution in short-haul study = 0 points v. Data: panel vs. time-series vs. cross-section Policy analysis should not be guided by immediate or short-term reactions to prices that result from policy changes. Consequently, policies that impact air travel demand should rely more on long-term elasticity measures. While panel studies are ideal as they capture cross-section and time-series effects, studies from time-series data that are sufficiently long in duration will also capture longer-term elasticities.

AIR TRAVEL DEMAND ELASTICITIES

393

Scoring rule: Use of panel data or time-series = 2 points Use of cross-section data = 0 points vi. Country focus There are likely to be many structural details of price sensitivity that relate to the specific national context of the airline industry, including the degree of competition, the size of the market and the regulatory environment. The impact of policies on air travel prices in Canada can be more readily related to some countries more than others. The close geographical proximity of international hubs and agreements within the EU make European studies somewhat less relevant to the Canadian context. US studies are more relevant given the geographic proximity of the US to Canada and the number of US cities to which Canadians travel. Australia on the other hand, provides reasonably comparable demographic, urban, geographical, governmental and regulatory structures. Scoring rule: Study relates directly to Canada = 2 points Study relates to similar foreign country (US or Australia) = 1 point Study relates to non-similar foreign country = 0 points vii. Route-specific estimates Studies that aggregate the effects of price changes on multiple routes will not capture the effects of market competition in which certain airlines enjoy significant market power on some routes but not others. A well-known example of this in the US is the effects of low-cost competition by Southwest Airlines on routes flown by full-service carriers. A related issue is that studies, which focus on multiple short-haul routes, run the danger of aggregating effects of routes that are predominantly used by business travellers with routes that are more leisure-oriented. This latter category often constitutes a significant portion of business for low-cost carriers, who offer cheap short-haul flights in competition with alternative leisure activities and entertainment. An example of this is the market for special event parties in Dublin (wedding stags for example) that was created by flights offered by RyanAir from various locations in the UK. Scoring rule: Study provides route or airline-specific estimates = 1 point Study does not provide route or airline-specific estimates = 0 points viii. Hub vs. non-hub airports Studies that do not separate out hub from connecting airports will not be able to distinguish “hub premium” effects. Passengers with an itinerary that utilizes a hub airport may be willing to pay a “hub premium” for the integrated service that hubs provide,

394

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

including sequenced flight segments that minimize waiting time, and baggage that is checked through to the final destination. The existence of a hub premium effect is supported by research in the US. Scoring rule: Study identifies hub airports = 1 point Study does not identify hub airports = 0 points ix. Connecting vs. O–D passengers There is a difference between an itinerary and the measurement of traffic volumes between city pairs. If a passenger is travelling from Moncton to Vancouver via Toronto, then their willingness-to-pay and their price sensitivity relates to the trip from Moncton to Vancouver. However, such a passenger could be included in the data that is measuring price sensitivity on the city pair Toronto–Vancouver. Scoring rule: Study identifies connecting vs. O–D passengers = 1 point Study does not distinguish connecting vs. O–D passengers = 0 points x. Age of the study The airline industry is a dynamic and changing industry, in the evolution of business models (full-service versus low-cost carriers for example), infrastructure (airport business practices) and government regulation. Studies conducted in the US prior to 1978 would not incorporate the effects of deregulation. A similar argument applies to studies that predate 1984 in Canada. Further the National Airport Policy in Canada has led to a grad­ ual devolution of airports from Transport Canada to independent local airport authorities throughout the 1990s. This devolution has led to important infrastructure and pricing decisions. Only the most recent studies would capture system-wide effects of this evolu­ tion as some local airport authorities have only come into being in the last year or two. Scoring rule: Studies completed during 1997−2002 = 3 points Studies completed during 1990−97 = 2 points Studies completed prior to 1990 = 1 point xi. Adjusted R-squared coefficient values This last item addresses the quality of output in the studies rather than the quality of inputs. In regression results, a low R-squared value indicates that only a small portion of variation in the dependent variable (O–D passengers), is explained by the independent

AIR TRAVEL DEMAND ELASTICITIES

395

variables. The adjusted R-squared value is a weighted measure that penalizes the addition of a large number of independent variables with low explanatory power.

Scoring rule: Adjusted R-squared value over 08 = 3 points Adjusted R-squared value between 0.6 and 08 = 1 point Adjusted R-squared value lower than 06 = 0 points

The highest possible score under the criteria we have developed is 23 points. Table 19 summarizes the scores of each study, from which we have generated histograms in six sub-categories using only those studies with a “passing grade” of 12 points or higher. The categories provide separation of long- and short-haul, international and domestic travel and business and leisure travel. Note that the column headings in the table refer to the numbered characteristics discussed above (Figure 19a and b). 5.1.1 Short-haul Business Travel 36 This subset is comprised of 16 estimates taken from three studies, the most recent of which is Battersby and Oczkowski (2001). The median elasticity for the data set is −0.70 (Figure 20). 5.1.2 Short-haul Leisure Travel 37 Three studies scoring 12 or more points in our scoring system generate 16 estimates of short-haul leisure travel. The estimates are distributed between a range of −0.4 and −2.37. The minimum value is taken from the Bureau of Transport Economics (1986) Table 19 Summary Statistics of Short-Haul Business

Travel Own-Price Elasticities: Studies Scoring ≥ 12 Points

5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

36 37

−1.103 −0.783 −0.700 −0.595 −0.123 0.188 16.000 −1.110 −0.100 0.072 0.697 1.396

Source: Battersby and Oczkowski (2001), Oum et al (1986), Lubulwa (1986). Source: Oum et al. (1986), BTE (1986), Abrahams (1983).

396

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

Study Characteristics Study Title Author(s) An Econometric Air Travel Demand Model For The Lasse Fridstrom and Harald Entire Conventional Domestic Thune-Larsen Network: The Case of

i

ii

iii

iv

v

vi

vii

viii

ix

x

xi

Score

0

0

2

1

2

0

1

0

0

1

1

8

Norway A Service Quality Model of Air Travel Demand: An Empirical Study

Michael Abrahams

3

3

2

2

2

1

1

0

0

1

1

16

Demands for Fareclasses and Pricing in Airline Markets

Tae H. Oum, David W. Gillen and S.E. Noble

2

3

2

0

0

1

2

1

0

1

0

12

The Structure of Intercity Travel Demands in Canada: Theory Tests and Empirical Results

Tae H. Oum and David W. Gillen

0

0

2

2

2

2

1

0

0

1

1

11

The Demand For International Air Passenger Service Provided by U.S. Air Carriers

Vinod Agarwal and Wayne K. Talley

2

0

0

0

0

1

1

0

0

1

3

8

Estimating Airline Demand With Quality of Service Variables

Richard A. Ippolito

0

0

2

1

0

1

1

0

0

1

0

6

The Demand For Air Services Provided By Air PassengerCargo Carriers In A Deregulated Environment

Wayne K. Talley and Ann Schwarz-Miller

0

0

0

0

0

1

1

0

0

1

3

6

An Abstratc Mode Model: A Cross-Section And TimesSeries Investigation

Andreas A. Andrikopoulos and Theophilos Terovitis

0

0

2

2

2

0

2

0

0

1

0

9

A Coherence Approach To Estimates of Price Elasticties In The Vacation Travel Market

John H. E. Taplin

2

2

2

1

2

1

1

0

0

1

0

12

Quality of Service and the Demand for Air Travel

James E. Anderson and Marvin Kraus

3

3

2

0

2

1

2

0

0

1

1

15

Tourist Expenditure in Australia

Bureau of Industry Economics

2

2

2

1

2

1

2

0

0

1

0

13

Determinants of Demand for Travel to and from Australia

G. Hollander

2

2

2

0

2

1

1

0

0

1

0

11

Tourism Related Movement Study Final Report

Nairn, R.J. and Partners and Hooper, P.

3

2

0

0

0

1

1

0

0

2

0

9

Demand for Australian Domestic Aviation Services Forecasts by Market Segment

Bureau of Transport Economics

2

3

2

2

2

1

1

0

0

1

0

14

Brandow Demand Functions For Australian Long Distance Travel

A.S.G. Lubulwa

3

3

2

2

2

1

2

0

0

1

0

16

Independent review of economic regulation of domestic aviation: Consumer Responsiveness to Changes in Air Fares

T.E. May, E.W.A. Butcher, and G. Mills

2

3

0

0

2

1

1

0

0

1

0

10

Demand Elasticties for Air Travel to and from Australia

Bureau of Transport Communications and Economics

3

2

2

0

2

1

1

0

0

2

3

16

The Price Elasticity of Demand for Air Travel

J.M. Jung and E.T. Fujii

0

0

0

0

2

1

2

0

0

1

0

6

An Econometric Analysis of the Demand for Domestic Air Travel in Australia

B. Battersby and E. Oczkowski

0

0

2

2

2

1

2

0

0

3

1

13

Demand for Air Travel in the United States: Bottom-Up Econometric Estimation and Implications for Forecasts by O&D Pairs

Dipasis Bhadra

0

3

2

0

2

1

1

0

0

3

0

12

An Econometric Analysis of the Demand for Intercity Passenger Transportation

Steven A. Morrison and Clifford Winston

3

0

0

2

0

1

1

0

0

1

0

8

A Generalized Decomposition of Travel-Related Elasticities Into Choice and Generation Components

Taplin, J.H.E

2

3

2

2

0

1

1

0

0

3

0

14

Australian Outbound Holiday Travel Demand: Long-haul Versus Short-haul

K. Hamal

2

3

2

2

2

1

1

0

0

3

0

16

Figure 19 Summary of Study Scores.

AIR TRAVEL DEMAND ELASTICITIES

397

6

4 3 2

Frequency

5

1

0

–0.1

–0.2

–0.3

–0.4

–0.5

–0.6

–0.7

–0.8

–0.9

–1

–1.1

–1.2

–1.3

–1.4

–1.5

–1.6

–1.7

0

Own-Price Elasticities (Short-haul)

Figure 20 Histogram of Short-Haul (Business) Own-Price Elasticities for Studies Scoring 12+ Points.

and represents winter vacation travel in Australia. The median estimate for all values is −1.520 (Figure 21, Table 20). 5.1.3 Long-haul International Business Travel 38 The subset of international business travel estimates provides 16 estimates taken from two studies. The median elasticity is −0.265, which is the same value derived prior to applying the scoring model to the aggregate data set. This occurred because both data sets are comprised of the same estimates (Figure 22, Table 21).

5

3 2

Frequency

4

1

0

–0.1

–0.2

–0.3

–0.4

–0.5

–0.6

–0.7

–0.8

–1

–0.9

–1.1

–1.2

–1.3

–1.4

–1.5

–1.6

–1.7

–1.8

–1.9

–2

–2.1

–2.2

–2.3

–2.4

0

Own-Price Elasticities (Short-haul)

Figure 21 Histogram of Short-Haul Leisure Own-Price Elasticities for Studies Scoring 12+ Points.

38

Source: BTCE (1995), Lubulwa (1986).

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

398

Table 20 Summary Statistics of Short-Haul Leisure Travel Own-Price Elasticities: Studies Scoring ≥ 12 Points −2.100 −1.743 −1.520 −1.288 −0.640 0.455 16.000 −2.370 −0.400 0.278 0.485 −0.116

5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

7 6

4 3

Frequency

5

2 1

0

–0.2

–0.4

–0.6

–0.8

–1

–1.2

–1.4

–1.6

–1.8

–2

0

Own-Price Elasticities (Long-haul)

Figure 22 Histogram of Long-Haul International-Business Own-Price Elasticities for Studies Scoring 12+ Points. Table 21 Summary Statistics of Long-Haul International Busi­ ness Travel Own-Price Elasticities: Studies Scoring ≥ 12 Points 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−1.423 −0.475 −0.265 −0.198 −0.093 0.278 16.000 −2.000 −0.010 0.251 −2.405 6.095

AIR TRAVEL DEMAND ELASTICITIES

399

5.1.4 Long-haul International Leisure Travel 39 There are 49 international leisure travel price elasticity estimates from six studies with at least 12 points in our scoring system. A majority of the estimates (31) are taken from studies published after 1995. The median elasticity is −1.040 with a large proportion of the estimates bunched up between −0.14 and −1 (Figure 23, Table 22). 5.1.5 Long-haul Domestic Business Travel 40 The domestic long-haul business travel subset consists of 26 estimates from two studies. The median elasticity is −1.15. The estimates are taken from Lubulwa (1986) and Oum et al. (1986) (Figure 24, Table 23). 7 6

4 3

Frequency

5

2 1

–0.1 0

–0.4 –0.3 –0.2

–0.5

–0.8 –0.7 –0.6

–1.1 –1 –0.9

–1.2

–2.7 –2.6 –2.5 –2.4 –2.3 –2.2 –2.1 –2 –1.9 –1.8 –1.7 –1.6 –1.5 –1.4 –1.3

0

Own-Price Elasticities (Long-haul)

Figure 23 Histogram of Long-Haul International and Leisure Own-Price Elasticities for Studies Scoring 12+ Points. Table 22 Summary Statistics of Long-Haul International Leisure Travel Own-Price Elasticities: Studies Scoring ≥ 12 Points 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

39 40

−2.140 −1.700 −1.040 −0.560 −0.254 1.140 49.000 −2.700 −0.140 0.420 −0.465 −0.474

Source: Hamal (1998), Taplin (1997), BTCE (1995), Lubulwa (1986), BIE (1984), Taplin (1980). Source: Oum et al. (1986), Lubulwa (1986).

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

400

7 6

4 3

Frequency

5

2 1

0

–0.1

–0.2

–0.3

–0.4

–0.5

–0.6

–0.7

–0.8

–0.9

–1

–1.1

–1.2

–1.3

–1.4

–1.5

–1.6

–1.7

0

Own-Price Elasticities

Figure 24 Histogram of Long-Haul Domestic-Business Own Price Elasticities for Studies Scoring 12+ Points. Table 23 Summary Statistics of Long-Haul Domestic Business Travel Own-Price Elasticities: Studies Scoring ≥ 12 Points 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−1.670 −1.428 −1.150 −0.836 −0.780 0.591 26.000 −1.700 −0.543 0.113 0.207 −1.119

5.1.6 Long-haul Domestic Leisure Travel 41 There are six long-haul domestic leisure travel price elasticity estimates taken from two studies. The median elasticity is −1.104. The histogram demonstrates no crowding around a range of values (Figure 25, Table 24). 5.1.7 Income Elasticities 42 A subset of 103 income elasticity estimates is generated from the “passing grade” studies. In similar fashion to the histogram for all studies, a crowding of estimates around the values of 0.5 to 2.5. The median value of the subset is 1.14 (Figure 26, Table 25).

41

Source: Lubulwa (1986), Abrahams (1983).

Source: Battersby and Oczkowski (2001), Hamal (1998), Taplin (1997), BTCE (1995), Oum et al. (1986),

BTE (1986), Lubulwa (1986), BIE (1984), Abrahamas (1983), Taplin (1980).

42

AIR TRAVEL DEMAND ELASTICITIES

401

1

Frequency

2

0

–0.1

–0.2

–0.3

–0.4

–0.5

–0.6

–0.7

–0.8

–1

–0.9

–1.1

–1.2

–1.3

–1.4

–1.5

–1.6

–1.7

–1.8

–2

–1.9

–2.1

–2.2

–2.3

–2.4

0

Own-Price Elasticities

Figure 25 Histogram of Long-Haul Domestic-Leisure Own-Price Elasticities for Studies Scoring 12+ Points. Table 24 Summary Statistics of Long-Haul Domestic Business Travel Own-Price Elasticities: Studies Scor­ ing ≥ 12 Points −1.576 −1.228 −1.104 −0.787 −0.502 0.441 6.000 −1.680 −0.440 0.191 0.015 −0.171

5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

30

20 15 10 5

Own-Price Elasticities

Figure 26 Histogram for Income Elasticities for All Studies 12+ Points.

11.5

11

10.5

10

9.5

9

8

8.5

7.5

7

6

6.5

5.5

5

4.5

4

3.5

3

2.5

2

1.5

1

0.5

0

–0.5

–1

0 –1.5

Frequency

25

402

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

Table 25 Summary Statistics of Income Elasticities for All: Studies Scoring ≥ 12 Points 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

0.242 0.807 1.140 2.089 4.636 1.282 103.000 −1.039 11.580 2.642 3.051 14.139

5.1.8 Studies that Account for Inter-modal Effects 43 The data set for studies that include the effects of inter-modal competition (e.g. auto, rail, bus, ship) is comprised of 104 own-price elasticity estimates taken from 13 studies (including short-, medium- and long-haul routes). The histogram does not demonstrate any bunching around a set of values as supported by the skewness (−0.641). The median elasticity value is −1.113 (Figure 27, Table 26).

20 18 16

12 10 8

Frequency

14

6 4 2 –3.3 –3.2 –3.1 –3 –2.9 –2.8 –2.7 –2.6 –2.5 –2.4 –2.3 –2.2 –2.1 –2 –1.9 –1.8 –1.7 –1.6 –1.5 –1.4 –1.3 –1.2 –1.1 –1 –0.9 –0.8 –0.7 –0.6 –0.5 –0.4 –0.3 –0.2 –0.1 0 0.1

0

Own-Price Elasticities

Figure 27 Histogram of Own-Price Elasticities for Studies that Include Inter-modal Effects.

43

Source: Battersby and Oczkowski (2001), Hamal (1998), Taplin (1997), Fridstroom (1989), BTE (1986), Lubulwa (1986), Morrison and Winston (1985), BIE (1984), Abrahams (1983), Andrikopoulos (1983), Oum and Gillen (1982), Ippolito (1981), Taplin (1980).

AIR TRAVEL DEMAND ELASTICITIES

403

Table 26 Summary Statistics of Own-Price Elasticities: Studies with Intermodal Effects 5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−2.085 −1.290 −1.113 −0.588 −0.138 0.703 104.000 −3.200 0.040 0.389 −0.641 0.614

A subset of estimates is extracted that includes estimates for short/medium-haul elas­ ticities.44 The data set is comprised of 34 estimates taken from four studies including those elasticities calculated by Battersby and Oczkowski (2001). The median estimate (−0.720) is lower than the median elasticity calculated for all inter-modal studies (−1.113). However, some details of these studies make the interpretation of this result difficult. First, the subset includes discount, economy, and business fare-class estimates. The elasticities reflect both the nature of travel on the routes (business or leisure) and the fare class. For example, the Sydney–Melbourne route is a significant business route in Australia with relatively low–elasticity estimates (Discount = −007, Economy = −081, Business = −01). The dataset contains several estimates from routes that are historically business travel city pairs. Secondly, 13 of the estimates are from city-pair routes with distances of approximately 870 to 1000 km (the high end of the short-haul distance condition). Only five out of 34 estimates are explicitly defined as short-haul routes of less than 750 km. These elasticity estimates are: Melbourne-Adelaide (−0.46); Australia short-haul <500 km (−0.728); New South Wales, Australia, routes <200 km (−2.54); Short-haul Western and Mid-western, US routes <500 miles (−0.08); Short-haul Eastern city pairs, US <500 miles (−0.36). Two of the estimates (Short-haul Eastern US, and Western and Mid-Western US) are likely capturing business travel. Lastly, 28 of the 34 estimates are taken from studies comprised of Australian citypair routes. These studies do not provide sufficient information about the city-pair characteristics, such as whether or not a specific route has one or more (or possibly no) competing transportation modes. If the use of a competing mode is infeasible or highly unlikely then the elasticity estimate is not capturing inter-modal effects (Figure 28, Table 27).

44

Source: Battersby and Oczkowski (2001), BTE (1986), Lubulwa (1986), Abrahams (1983).

404

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

4.5 4

3 2.5 2 1.5

Frequency

3.5

1 0.5 –3.3 –3.2 –3.1 –3 –2.9 –2.8 –2.7 –2.6 –2.5 –2.4 –2.3 –2.2 –2.1 –2 –1.9 –1.8 –1.7 –1.6 –1.5 –1.4 –1.3 –1.2 –1.1 –1 –0.9 –0.8 –0.7 –0.6 –0.5 –0.4 –0.3 –0.2 –0.1 0 0.1

0

Own-Price Elasticities

Figure 28 Histogram of Short/Medium-Haul Own-Price Elasticities for Studies.

Table 27 Summary Statistics of Short/Medium-Haul

Own-Price Elasticities: Studies with Intermodal Effects

5th percentile First quartile Median Third quartile 95th percentile Interquartile range Number of estimates Minimum Maximum Variance Skewness Kurtosis

−2.176 −1.108 −0.720 −0.415 −0.077 0.693 34.000 −3.200 0.040 0.508 −1.524 2.925

6 DISCUSSION In a number of cases, studies that are focused on the impact of price changes or fees on demand use a single elasticity measure to compute the quantity, revenue and profit change for a route, market, airline or entire economy (see for example, PODM (Transport Canada) which uses one elasticity of business and one for leisure; economic impact studies for airports often use this approach as well). Using a single-value implicitly assumes that the elasticity measure is transferable across markets and time. There is a rich and extensive literature that explains the conditions under which such estimates

AIR TRAVEL DEMAND ELASTICITIES

405

Table 28 Summary of Absolute Elasticity Values Category

Median (1st quartile) (3rd quartile) Elasticity values All studies

Own-price: Long-haul international business

0.475

Own-price: Long-haul international leisure

1.65

Own-price: Long-haul domestic business

1.428

Own-price: Long-haul domestic leisure

(1st quartile)

Elasticity values Studies scoring ≥ 12 points

0.265

0.265 0.198

0.475

0.535

1.700

0.836

1.428

0.993

1.745

Own-price: Short/medium-haul business

0.798

0.887

1.228

0.787 1.520

0.885

1.743

0.608

0.783

2.169

0.807

0.730

1.288 0.700

1.390 0.840

0.836 1.104

1.520

Income elasticity

0.560 1.150

1.120 1.472

0.198 1.040

1.150

Own-price: Short/medium-haul leisure

Median (3rd quartile)

0.595 1.140 2.0489

are transferable.45 The properties or characteristics of the data in different markets should meet statistical tests in order to have statistical validity in having a common elasticity. We have shown that elasticity values can and do differ significantly between travel distance, type of traveller and even domestic and international routes. This is illustrated in Table 28 in which we report elasticities for three different route types and two passenger types. We have argued that the usefulness of estimates should be based upon other criteria such as the inclusion of income coefficients and distinctions between types of passengers and airports. We have also argued that route-specific data is especially valuable in capturing competitive, geographic and market differences. In the report we show that for the entire set of studies as well as for categories of studies the distribution of elasticity estimates is highly skewed. Such a distribution makes the use of the mean or average tenuous at best. The mean may turn out to be a value that was yielded by none of the studies. The variance is also large which makes the level of confidence we can place in a “mean” value particularly low. Therefore, we have used the “median” value of the elasticity estimates as an indicator of what elasticity value might be used in forecasting changes in revenue, passengers and profit in markets where the elasticity is appropriate – short- or long-haul, business or leisure etc. as a result of 45 This literature grew out of the early demand modeling efforts. See for example, Watson, Peter L. and Richard Westin. Transferability of Disaggregate Mode Choice Models, Amsterdam: North-Holland, 1975 and Frank S. Koppelman and Eric I. Pas, Multidimensional Choice Model Transferability,. Transportation Research. Part B, Methodological. Vol. 20B, no. 4 (Aug. 1986) p. 321–330.

406

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

a policy change.46 In addition to the reported median values in the various categories, we have also reported quartile information from the distributions of elasticity values; dividing the observations into quartiles simply means we have divided it into quarters, so the first quartile would be the first 25 percent of observations. In particular, Table 28 draws attention to the first and third quartiles (25 percent of the values in a distribution fall below the first quartile and 75 percent fall below the third quartile).47 The first and third quartiles form a useful range around the median that widens (narrows) as the tails of the distribution grow thicker (thinner). This is illustrated in Figures 29 and 30. Table 2 below shows median absolute values (meaning we have dropped the negative sign in front of the elasticity value) of estimated demand elasticities along with the first

1st Q

Median

3rd Q Distribution

Inter-quartile range

Figure 29 Inter-quartile Range with Wide Tails in the Distribution.

3rd Q Median 1st Q Distribution

Inter-quartile range

Figure 30 Inter-quartile Range with Narrow Tails in the Distribution.

46

We remind the reader the median is the value that divides the sample in half so 50 percent of observations will lie above the median value and 50 percent will lie below it. 47 When using mean values as a measure of central tendency, the standard deviation of the distribution can be used to create confidence intervals of plus and minus one standard deviation around the mean. Since we are using the median as a measure of central tendency, we cannot use standard deviations (which assume a normal distribution which by definition is not skewed).

AIR TRAVEL DEMAND ELASTICITIES

407

and third quartiles of the distribution for all studies and for “passing grade” studies in six categories.48 In the long-haul international market, there is no apparent difference between the elasticity values from all studies and the group regarded as having a “passing grade”, based on our scoring system.49 The median values are low (0.265) for business travel and close to unity for leisure travel. This seems reasonable, since long-haul international business travel demand has relatively few close substitutes, making demand insensitive to fare changes. On the other hand, international leisure travellers are more likely to postpone trips to specific locations in response to higher fares, or shop around for thoselocations offering more affordable fares. In the vacation market, international travel competes more directly with domestic travel for vacation destinations. The long-haul domestic business segment elasticities are the same whether looking at all studies or the sub-set of “passing grade” studies. The value of 1.15, being close to unity indicates that domestic_business travellers will have higher elasticities (in this case about four times the value) than international business travellers. In domestic markets, alternatives such as telecommunications are more substitutable than in international markets due to common culture, laws, contracts etc. International trips are typically planned well in advance, with the travel spread over more time we would expect the airfare to be a lower proportion of overall trip costs. The median long-haul domestic leisure elasticity values do not differ significantly between all studies and those rated superior, however the range of elasticity values in the passing grade studies (as defined by the first and third quartiles of the distribution) is narrower and slightly lower. The value of 1.104 does not seem unreasonable in comparison to the domestic business travel elasticities. The median elasticity value from all studies versus “passing grade” studies for short/medium-haul leisure are identical and characteristically elastic at 1.52. Notice how­ ever that the range of values around the median is narrower in the passing grade studies, excluding the possibility of inelastic demand at the lower bound of the range. The estimated median fare elasticity for short/medium-haul business travel is moder­ ately inelastic at 0.73 with a very tight range around the median. Generally, the price or fare elasticity will decline with distance other things held constant. However, we can see in Table 2 that the short/medium haul business elasticity is smaller in value than the long-haul domestic business elasticity. The explanation is quite straightforward, people using aviation as a business tool, will have an especially high value of time. They will therefore willingly pay high fares to save time on short-haul trips. Secondly, there are numerous “must-go” short-haul trips that arise in the course of business dealings, trips that are not easily planned and can be completed in a morning or afternoon without requiring scheduling meetings or packing or making family arrangements. These shorthaul business trips will have low elasticities; for example, in flying to Montreal from

48 Absolute values mean we have dropped the negative sign before the elasticity value. We had included the sign in section 4 because price elasticities were negative while income elasticities were positive. In a few cases studies reported positive price elasticities, a clear error. Here we are interested in only the degree of difference in price sensitivity as reported by the magnitude of the elasticity value. 49 These are not the same set of studies but the superior group is a subset of the total.

408

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

Toronto to close a multi-million dollar deal and sign the contact, the high air fare is low when factored into the overall value of the trip. Table 2 provides ample evidence that using a single elasticity for all market segments is inappropriate just as a single elasticity will not reflect impacts on the aggregate market. Furthermore, simply segmenting markets by business and leisure is insufficient to provide any degree of accuracy to forecast changes in passengers with changes in fares. For example, given the differences between short- and long-haul market elasticities, using long-haul values to evaluate impacts on short-haul markets would generally provide an underestimate. It may be the case that elasticity values reported in the empirical literature are under­ estimated for both leisure and business travel, particularly in short- to medium-haul markets. The reasoning is based on research, which shows that the entry of low-cost carriers into markets leads to a reduction in the average fares on those routes (Windle and Dresner, 1999).50 For example, if a low-cost carrier like Southwest enters a market, the effect has been a reduction in fares of almost 50 percent. Table 3 compares very recent revenue yields in various US markets that are not only delineated based on length of flight but also on whether there is competition from Southwest Airlines, and the form of that competition. The table shows, quite dramatically, that the more direct the competition from Southwest, the lower the yields (which translates into lower average fares). (Table 29) There are few studies that have included, as a time-series, the growth in markets where low-cost carriers have concentrated their activity. In empirical studies, routes are usually aggregated so an average elasticity is estimated across short-, medium- and long-haul routes. Thus, in studies using detailed US data markets served by Southwest Airlines and more recently by other low-cost carriers such as JetBlue and Air Tran are aggregated with those served by other full-service carriers even though the growth in traffic in these markets is quite different. In order to look at the potential underestimation of demand response in markets where low-cost carriers participate we used US data from 1999–2000 second Quarter. The year-after-year changes in passengers and fares are used to calculate arc-elasticities for routes of different length and for fare increases and fare decreases. We found the Table 29 Revenue Yields of Other Airlines (OA) and Southwest (SW) Market Type OA – no SW presence OA–SW connecting competition OA–SW direct competition SW-connect SW-nonstop

Yields (cents per passenger mile) 500 Miles 1000 Miles 51 31 26 21 18

26 20 19 14 12

Source: US Department of Transportation (2002).

50 Robert Windle, Martin Dresner (1999) Competitive Responses to Low Cost Carrier Entry, Transportation Research. Part E, Logistics and Transportation Review. Vol. 35E, no. 1 (Mar. 1999) p. 59–75.

AIR TRAVEL DEMAND ELASTICITIES

409

calculated arc elasticities did not differ in any significant way from the values we have found from our survey of the literature. This applies to values for long-haul as well as short to medium-haul markets; short/medium-haul markets are more price sensitive than long-haul markets. We therefore feel the values reported in Table 2 fairly reflect the sensitivity of markets including those served by low-cost and low-fare carriers.51 As we note in the discussion of Table 2, there is no single elasticity value that is representative of air travel demand. There are several distinct markets and several different elasticities should be used when exploring the impact on markets from changes to the aviation environment. Furthermore, even given the elasticity for a market segment, there is a range around this elasticity that should be considered in using the elasticity to forecast the impact of fare changes. The aggregate elasticities for the market segment reflect the combined effect of demand relationships in each component market. Each market will typically exhibit different elasticities than that considered for the aggregate market level.

BIBLIOGRAPHY Abrahams, M., A Service Quality Model of Air Travel Demand: An Empirical Study, Transportation Research, 17A(5), 385–393, 1983. Agarwal, V. and W. Talley, The Demand for International Air Passenger Service Provided by U.S. Air Carriers, Interntaional Journal of Transport Economics, 12(1), 63–70, 1985. Anderson, James E. and Marvin Kraus, Quality of Service and the Demand for Air Travel, The Review of Economics and Statistics, 63(4), 533–540, 1981. Andrikopoulos A.A. and T. Terovitis, An Abstract Mode Model: A Cross-section and Time-series of Investigation, International Journal of Transport Economics, 10(3), 563–576, 1983. Battersby, B. and E. Oczkowski, An Econometric Analysis of the Demand for Domestic Air Travel in Australia, International Journal of Transport Economics, 28(2), 193–204, 2001. Bhadra, Dipasis, Demand for Air Travel in the United States: Bottom-Up Econometric Estima­ tion and Implications for Forecasts by O&D Pairs, Center for Advanced Aviation System Development – The Mitre Corporation, 2002. Brons Martijn, Eric Pels, Peter Nijkamp and Piet Rietvela, Price Elasticities of Demand for Passenger Air Travel: A Meta-Analysis, Tinbergen Institute, Amsterdam, 2001. Bureau of Industry Economics, Research Report 16, Bureau of Industry Economics, Canberra. BIE, 1984. Bureau of Transport and Communications Economics, International Aviation, Report 86, 19–22, 1995. Bureau of Transport Communications and Economics, Demand Elasticities for Air Travel to and from Australia, Working Paper 20, Department of Transport and Communications, 1995. Bureau of Transport Economics, Demand for Australian Domestic Aviation Services Forecasts by Market Segment, AGOS, Canaberra. Occasional Paper 79, 1986. Fridstroom, L. and H. Thune-Larsen, An Econometric Air Travel Demand Model for the Entire Conventional Domestic Network: The Case of Norway, Transportation Research, 23B(3), 213–224, 1989.

51

We have not carried out any analysis on Canadian data since is there is no comparable data set to that available in the US.

410

DAVID GILLEN, WILLIAM G. MORRISON, AND CHRISTOPHER STEWART

Gillen D. W. and W.G. Morrison, Airport Financing, Costing, Pricing and Performance, Report to the Canadian Transportation Act review Committee, April 2001. Hamal, K., Australian Outbound Holiday Travel Demand: Long-haul Versus Short-haul, Bureau of Tourism Research, Canberra, BTR Conference Paper 98.2, 1998. Hollander, G., Determinants of Demand for Travel to and from Australia, BIE Working Paper no. 26, Bureau of Industry Economics, Canberra, 1982. Ippolito, R.A., Estimating Airline Demand With Quality of Service Variables, Journal of Transport Economics and Policy, 15(1), 457–464, 1981. Jud, G.D. and H. Joseph, International Demand for Latin American Tourism, Growth and Change, 5(1), 25–34, 1974. Jung, J.M. and E.T. Fujii, The Price Elasticity of Demand for Air Travel, Journal of Transport Economics and Policy, 10, 257–262, 1976. Lubulwa, A.S.G, Brandow Demand Functions for Australian Long Distance Travel, Forum Papers, 11th Australian Transport Research Forum, 2, 200–218, 1986. May, T.E., E.W.A. Butcher and G. Mills, Consumer Responsiveness to Changes in Air Fares, Indpendent Review of Economic Regulation of Domestic Aviation, 2(Appendix L), 1986. Milloy, H.B., L. Douglas and S.M. Sullivan, Market Response to Discounted Airfares on Selected Domestic Trunk Routes, paper presented at 10th Australian Transportation research Forum, 1985. Morrison, Steven A. and Clifford Winston, An Econometric Analysis of the Demand for Intercity Passenger Transportation, Research in Transportation Economics, 2, 213–237, 1985. Nairn, R.J. and P. Hooper, Tourism Related Movement Study Final Report, Roads and Traffic Authority, NSW, Sydney, 1992. Oum T.H. and D.W. Gillen, The Structure of Intercity Travel Demands in Canada: Theory Tests and Empirical Results, Transportation Research, 17B(3), 175–191, 1982. Oum T.H., D.W. Gillen and S.E. Noble, Demand for Fareclasses and Pricing in Airline Markets, Logistics and Transportation Review, 22(3), 195–222, 1986. Oum, T.H., W.G. Waters and J.S. Yong, A Survey of Recent Estimates of Price Elasticities of Demand for Transport, World Bank Working Paper, WPS 359, 1990. Oum T.H., W.G. Waters and J.S. Yong, Concepts of Price Elasticities of Transport Demand and Recent Empirical Estimates, Journal of Transport Economics and Policy, 26(2), 139–154, 1992. Oum T.H., A. Zhang and Y. Zhang, Inter-Firm Rivalry and Firm-Specific Price Elasticities in Deregulated Airline Markets, Journal of Transport Economics and Policy, 27(2), 171–192, 1993. Talley, W.K. and Schwarz-Miller, The Demand for Air Services Provided by Air Passenger-Cargo Carriers in a Deregulated Environment, International Journal of Transport Economics, 15(2), 159–168, 1988. Taplin, J.H.E., A Coherence Approach to Estimates of Price Elasticities in the Vacation Travel Market, Journal of Transport Economics and Policy, 14(1), 19–35, 1980. Taplin, J.H.E., A Generalized Decomposition of Travel-Related Elasticities Into Choice and Gen­ eration Components, Journal of Transport Economics and Policy, 31(2), 183–192, 1997. Straszheim, M., The International Airline Industry, Brookings Institution, Washington, D.C., 1978. Windle, R. and Dresner, M., “Competitive Responses to Low Cost Carrier Entry”, Transportation Research E: The Logistics and Transportation Review, 35, 59–75, 1999.

Index

Ailing firms, 151, 157, 160 Air navigation system providers (ANSP), 182, 183, 188 Air navigation systems (ANS), 74, 172, 181–3 Air traffic control, 9, 17, 61, 65, 66, 69, 71, 94, 175, 203, 256, 257, 273 Air transportation taxation, 256, 257, 271 Air travel demand, 236, 291, 365–409 Airfare differentials, 210, 211, 213, 214, 215, 216, 217, 218, 219, 222, 225, 227–8 Airline deregulation, 2, 10, 15, 18, 19–23, 29, 144 Airport choice, 210, 213, 237, 249, 252 Airport pricing, 22–3, 89–119 Airport privatization, 3, 22, 90, 100, 122 Airport privatization and regulation, 3–10 Airport substitution, 209–32 Alliances, 3, 10–14, 16, 133, 143, 144 Antitrust immunity, 12, 13, 14 Aviation infrastructure costs, 256 Bankruptcy, 16, 17, 19, 23, 28, 29, 47, 142, 275 Bargaining, 28, 50, 53, 54, 177 Bilateral agreement, 2, 5, 6, 9, 10, 11,14, 18, 126, 127, 128, 148, 172, 293 Booking day, 235, 327, 335, 336, 337, 338, 341 Build-own-operate-transfer (BOOT), 176 Business, 10, 15, 22, 43, 47, 66, 104, 111, 126, 143, 157, 245, 280, 287–317 Business passenger, 238, 277, 280, 289, 291, 293, 298, 300, 302, 305, 369 Capacity investment, 23, 95, 96, 97, 99, 102, 103, 112, 119 Chicago Convention, 129, 172, 174 City-pair market, 4, 5, 6, 8, 13, 15, 24 Codesharing, 199, 200 Commercialization, 90, 171–90

Common Atlantic Aviation Area, 3, 17–19 Commuter airlines, 199–200 Comparability, 42, 50, 308, 324 Concession revenue, 95, 98, 99, 100, 101, 102, 120 Concessions, 16–17, 28, 29, 30, 31, 32, 37, 47, 53, 54, 97–100, 101, 102, 120, 138, 168, 176, 177, 180, 184, 187, 190 Congestion delays, 20, 91, 107, 108, 109, 113 Congestion pricing, 3, 21, 22, 23 Cost recovery, 73, 74, 96–7, 98, 100, 103, 104, 110, 115, 119, 179, 188 Cross-border merger, 2, 3, 11, 14, 15, 17, 18–19 Cross-elasticity, 307 Current Population Survey (CPS), 30, 31, 34, 35, 36, 37, 38, 40, 42–3, 49, 50, 52, 56 Data envelopment analysis, 184 Delay, 20, 75–7, 91 Demand substitutability, 289, 294, 310 Deregulation, 2, 3–10, 15–17, 19–23, 24, 28, 46, 126, 133, 142, 143, 144, 148, 174, 175, 178, 200, 212, 251, 276, 346, 354, 394 Discrete choice models, 237 Distribution, 36, 43, 45, 154, 174, 212, 256, 264, 345–61 Dynamic structural model, 213 E-commerce, 343 Economic regulation, 90, 172,174, 175, 176, 190 Efficiency, 2, 3, 4, 5, 8, 16, 17, 22, 62, 67, 69, 73, 78, 90, 95, 101, 119, 120, 126, 149, 151, 154, 174, 184, 188, 288, 346 Effective tax rate, 261–6 Electronic spider, 322

INDEX

412

En route control, 62, 68, 70, 71, 78, 79–80 Entry barriers, 161, 227 Euler equations, 130, 133 EUROCONTROL, 187, 260 European Union (EU), 128, 144, 146, 150, 174, 187, 255–71, 288, 346 Externalities, 72, 104, 148, 151, 152, 153, 168, 256 Federal Aviation Administration (FAA), 62–7, 68, 69, 79, 183, 200, 251, 256, 257, 259, 360 Feeder airlines, 198 Flag carrier, 2, 3, 5, 6, 7–8, 9, 10, 14, 15–17, 18, 19,148, 149, 157–61, 293, 298, 300 General aviation, 21, 65, 80, 86 Gini decomposition, 346, 352–61 High density rule, 80 Hub and spoke, 3–5, 115, 143, 148, 200, 236, 290, 320, 321, 356, 369, 374 Income elasticity, 215, 261, 400–402 Interline, 11–15 Intermodal, 16 International Air Transport Association (IATA), 5, 6, 11–12, 16, 127, 134, 136, 372 International Civil Aviation Organisation (ICAO), 134, 135, 136 Internalization of congestion, 5, 15, 19, 20–2, 116 Kurtosis, 375 LaGuardia, 62, 63, 69, 80–6 Leisure passenger, 17, 284, 289, 298, 301, 302 Leisure, 10, 17, 22, 243, 245, 268, 289, 301–302, 303, 308, 382, 384, 385, 391, 395, 399, 400, 407, 408 Logit, 222, 244, 250, 287, 290, 306 Long haul, 134–5, 294, 375–6, 377–8, 379–83, 397, 399, 400, 407, 408 Low cost airlines, 169, 195, 212, 294, 307, 319–41

Low-cost carrier (LCC), 2, 3, 15, 16, 19, 29, 47, 53, 161, 169, 193–4, 208, 212, 219, 236,252, 257, 262, 263, 264, 275, 276, 288, 289–91, 293, 300, 301, 302, 305, 308, 309, 310, 320, 321, 322, 328, 354, 358–9, 360–1, 374, 393, 394, 408 Market definition, 154, 155, 165, 166, 237, 238–42 Market failure, 72, 75, 76, 151, 152, 153, 154, 155, 160, 161, 163, 167–8 Market power, 3, 20, 22, 23, 72, 89, 90, 100, 101, 102, 104, 107, 108, 109, 110, 112, 114, 116, 117, 119, 120, 148, 155, 160, 174, 175, 212, 215, 279, 321, 393 Market share, 16, 29, 47, 117, 218, 219, 220, 238, 244, 251, 278, 288, 291, 292, 298, 303, 346, 353, 355, 356 Median, 36, 217, 375, 377, 380, 384, 389, 395, 402, 403, 405, 407 Medium haul, 134–5, 194, 212, 376, 403, 404, 407, 408, 409 Menu costs, 329–35 Merger, 3, 11, 14, 15, 17, 18–19, 32, 148, 294, 309 Monopoly power, 172, 173, 176, 180, 374 Monotonic fares, 320 Movement fee, 75, 84 Multi-airport region, 236, 237, 250 Multinomial logit, 287 Nash Behavior, 132 NATS, 182, 183, 187, 188 NAV Canada, 183 Nested logit (NL) model, 295 Network, 3–10 Network size, 126, 129, 131, 136, 142, 276 New carrier entry, 218, 219 New institutional economics, 176, 177 On-line pricing, 190, 329–35 Open skies, 3, 14–15, 18 Panel data, 133, 221 Passenger survey, 236, 240, 242, 243, 244, 287, 299, 302 Perceived marginal revenue, 131 Persistent price dispersion, 335 Point-to-point, 3–5, 16, 17, 165, 207, 288, 322, 346, 356, 369

INDEX

Price-capping, 178, 180, 182, 183 Price discrimination, 288, 300, 321, 326, 335, 341, 367–8 Price elasticity, 164, 268, 269, 307, 366–7, 368, 373, 375, 389, 392, 399 Privatization, airport, 3, 22, 90, 100, 122 Product differentiation, 156, 252, 290, 309, 336, 338, 341, 342 Public goods, 68, 151 Public-private partnerships, 187, 188 Quality competition, 276–7, 284 Ramsey pricing, 95, 115 Rate-of-return regulation, 182 Regional airlines, 30, 46, 47, 49, 50, 51, 52, 193–208 Regional airports, 147, 152, 157, 165, 167, 168–9 Regional jets, 51, 68, 135, 205–207, 352 San Francisco Bay Area, 121, 236, 237, 239, 246, 247, 248, 249 Scope clauses, 207 Seat pitch, 276, 277, 278, 279, 281, 284 Short-haul, 200, 213, 289, 293, 294, 301, 307, 310, 369, 384, 385, 391, 392, 393, 395, 398, 403, 408 Simulation., 119, 121, 125–44 Single European Skies, 187 Skewness, 349, 375, 377, 380–1, 388, 402 Slot auction, 3, 20, 21, 22, 23 Small-to-medium airport performance, 238 Spatial concentration, 346, 351, 353, 360, 361

413

Spatial econometrics, 213, 224 State aid, 147–69 Stochastic frontier, 133 Strategic alliances, 173 Strategic trade, 153, 154 Subsidies, 5, 22, 86, 98, 102, 115, 116, 117, 128, 148, 150, 152, 153, 155, 160, 174, 199, 200 Supranational control, 148, 153–4, 166 Tax incidence, 256, 258, 267, 269, 270 Telecommunications, 172, 182, 183, 407 Terminal control, 70, 71, 78–9 Ticket taxes and fees, 255–71 Treaty of Rome, 128 Turbo-props, 196, 207 Unions, 28, 29, 31, 47, 53, 54, 142, 207 US Federal Aviation Administration, 97 Value pricing, 77, 79, 86 Vertical integration, 111, 251–2 Vertical relations, 251, 252 Vertical structure approach, 91, 103–12, 113, 114, 116, 119, 120, 121 Wage cycles, 29 Wages, 17, 29, 30, 40, 46–52, 53, 142, 203 X-inefficiency, 182 Yield management, 290, 295, 301, 326, 333, 368

This page intentionally left blank

This page intentionally left blank

This page intentionally left blank