ECMT Round Tables No. 136 Estimation and Evaluation of Transport Costs (Ecmt Round Table)

Page 1

13 6 EUROPEAN CONFERENCE OF MINISTERS OF TRANSPORT

ESTIMATION AND EVALUATION OF TRANSPORT COSTS Cost data for the construction and operation of facilities are essential for the evaluation of infrastructure services supplied by private or public providers.

Which methods can measure the efficiency of service provision and effectively benchmark providers? How do regulatory regimes impact on operators’ and infrastructure service providers’ cost levels? How can regulators counter the asymmetry of information as well as the incentives for data providers to selectively serve business rather than user interests?

www.oecd.org/publishing (74 2007 06 1 P1) ISBN 978-92-821-0151-3

-:HSTCSC=VUVZVX:

136

R O U N D TA B L E

These were the main questions discussed by the Round Table. Background papers were provided by Antonio Estache (World Bank) and Lourdes Trujillo (Universidad de Las Palmas de Gran Canaria), Piet Rietveld et al. (Free University of Amsterdam), Carlos Barros (Portugal) as well as Marc Ivaldi and Philippe Gagnepain (Université de Toulouse).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS

What essential data do regulators need to ensure that transport serves the user best?

2007

C E N T R E

14:07

R E S E A R C H

20/08/07

T R A N S P O R T

fullTab136GB

ESTIMATION AND EVALUATION OF TRANSPORT COSTS

ROUND TABLE

13 6

Cover_f.fm Page 1 Wednesday, April 7, 2004 11:00 AM

TRANSPORT RESEARCH CENTRE

REPORT OF THE ONE HUNDRED AND THIRTY SIXTH ROUND TABLE ON TRANSPORT ECONOMICS

ESTIMATION AND EVALUATION OF TRANSPORT COSTS

EUROPEAN CONFERENCE OF MINISTERS OF TRANSPORT

ORGANISATION FOR ECONOMIC CO-OPERATION AND DEVELOPMENT The OECD is a unique forum where the governments of 30 democracies work together to address the economic, social and environmental challenges of globalisation. The OECD is also at the forefront of efforts to understand and to help governments respond to new developments and concerns, such as corporate governance, the information economy and the challenges of an ageing population. The Organisation provides a setting where governments can compare policy experiences, seek answers to common problems, identify good practice and work to co-ordinate domestic and international policies. The OECD member countries are: Australia, Austria, Belgium, Canada, the Czech Republic, Denmark, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Japan, Korea, Luxembourg, Mexico, the Netherlands, New Zealand, Norway, Poland, Portugal, the Slovak Republic, Spain, Sweden, Switzerland, Turkey, the United Kingdom and the United States. The Commission of the European Communities takes part in the work of the OECD. OECD Publishing disseminates widely the results of the Organisation’s statistics gathering and research on economic, social and environmental issues, as well as the conventions, guidelines and standards agreed by its members.

This work is published on the responsibility of the Secretary-General of the OECD. The opinions expressed and arguments employed herein do not necessarily reflect the official views of the Organisation or of the governments of its member countries.

© OECD 2007 No reproduction, copy, transmission or translation of this publication may be made without written permission. Applications should be sent to OECD Publishing [email protected] or by fax 33 1 45 24 99 30. Permission to photocopy a portion of this work should be addressed to the Centre français d’exploitation du droit de copie (CFC), 20, rue des Grands-Augustins, 75006 Paris, France, fax 33 1 46 34 67 19, [email protected] or (for US only) to Copyright Clearance Center (CCC), 222 Rosewood Drive Danvers, MA 01923, USA, fax 1 978 646 8600, [email protected].

EUROPEAN CONFERENCE OF MINISTERS OF TRANSPORT (ECMT)

The European Conference of Ministers of Transport (ECMT) is an inter-governmental organisation established by a Protocol signed in Brussels on 17 October 1953. It comprises the Ministers of Transport of 44 full Member countries: Albania, Armenia, Austria, Azerbaijan, Belarus, Belgium, Bosnia-Herzegovina, Bulgaria, Croatia, the Czech Republic, Denmark, Estonia, Finland, France, FRY Macedonia, Georgia, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Liechtenstein, Lithuania, Luxembourg, Malta, Moldova, Montenegro, Netherlands, Norway, Poland, Portugal, Romania, Russia, Serbia, Slovakia, Slovenia, Spain, Sweden, Switzerland, Turkey, Ukraine and the United Kingdom. There are seven Associate member countries (Australia, Canada, Japan, Korea, Mexico, New Zealand and the United States) and one Observer country (Morocco). The ECMT is a forum in which Ministers responsible for transport, and more specifically the inland transport, can co-operate on policy. Within this forum, Ministers can openly discuss current problems and agree upon joint approaches aimed at improving the use and ensuring the rational development of European transport systems. At present, ECMT has a dual role. On one hand it helps to create an integrated transport system throughout the enlarged Europe that is economically efficient and meets environmental and safety standards. In order to achieve this, ECMT assists in building bridges between the European Union and the rest of the European continent at a political level. On the other hand, ECMT also develops reflections on long-term trends in the transport sector and, more specifically, studies the implications of globalisation on transport. In January 2004, the ECMT and the Organisation for Economic Co-operation and Development (OECD) brought together their transport research capabilities by establishing the Joint Transport Research Centre. The Centre conducts co-operative research programmes that address all modes of inland transport and their intermodal linkages to support policy-making throughout Member countries. Ministers at their Dublin Council in May 2006 agreed a major reform of ECMT designed to transform the organisation into a more global body covering all modes of transport. This new international transport Forum will aim to attract greater attention to transport policy issues, and will hold one major annual event involving Ministers and key sectoral actors on themes of strategic importance. 2007 is a transitional year for the setting up of the Forum. The new structure will be fully operational as of 2008.

Also available in French under the title: Estimation et évaluation des coûts de transport

Further information about the ECMT is available on Internet at the following address: www.cemt.org © ECMT 2007 – ECMT Publications are distributed by: OECD Publishing 2, rue André-Pascal, 75775 PARIS CEDEX 16, France

Cover_f.fm Page 1 Wednesday, April 7, 2004 11:00 AM

TABLE OF CONTENTS -

5

TABLE OF CONTENTS

SUMMARY OF DISCUSSIONS.......................................................................................................... 7 INTRODUCTORY REPORTS Measuring Inefficiencies in Transport Systems: Between Technology and Incentives by P. GAGNEPAIN (Spain) and M. IVALDI (France) ........................................................... 29 Introduction............................................................................................................................ 33 1. The preliminary frontier................................................................................................... 35 2. Incentives ......................................................................................................................... 37 3. Regulation of public transit in France.............................................................................. 38 4. The deregulation of European airlines ............................................................................. 43 Conclusion ............................................................................................................................. 46 Infrastructure Maintenance Costs: A Comparison of Road, Rail and Inland Navigation, and Implications for User Charges – by P. RIETVELD, F. BRUINSMA and M. KOETSE (The Netherlands)................................................................................................. 49 1. Introduction...................................................................................................................... 53 2. Pricing in freight transport: base line in The Netherlands ............................................... 54 3. Marginal cost pricing and its problems............................................................................ 56 4. User dependent infrastructure maintenance costs ............................................................ 57 5. Alternative pricing structures for fixed maintenance costs.............................................. 61 6. Effects of alternative pricing measures on the transport sector ....................................... 67 7. Conclusions...................................................................................................................... 71 Annex 1: Emission factors of euro classes in road freight transport...................................... 74 Annex 2: Imposing a minimum tariff under a differentiated fixed cost charge ..................... 76 Technical Efficiency in the Portuguese Airports with a Stochastic Cost Frontier Model by C. BARROS (Portugal) .......................................................................................................... 81 1. 2. 3. 4. 5. 6. 7. 8.

Introduction...................................................................................................................... 85 Institutional setting........................................................................................................... 87 Literature review.............................................................................................................. 89 Theoretical framework..................................................................................................... 92 Data .................................................................................................................................. 94 Discussion ........................................................................................................................ 98 Contribution, limitation and extensions of this study .................................................... 100 Conclusions.................................................................................................................... 100

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

6 – TABLE OF CONTENTS

Transport Cost Levels, Productivity and Efficiency Measures: Some Theory and Main Policy Uses – by A. ESTACHE (United States) and L. TRUJILLO (Spain)........................ 105 1. 2. 3. 4. 5. 6. 7.

Introduction.................................................................................................................... 109 What’s so special about transport?................................................................................. 110 Why should policymakers care about transport costs? .................................................. 111 From policy concerns to measures................................................................................. 113 How to decompose the sources of cost efficiency changes ........................................... 117 From theory to practice .................................................................................................. 119 Concluding comments.................................................................................................... 120

LIST OF PARTICIPANTS............................................................................................................... 127

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

SUMMARY OF DISCUSSIONS -

SUMMARY OF DISCUSSIONS

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

7

SUMMARY OF DISCUSSIONS -

9

SUMMARY

1.

INTRODUCTION ......................................................................................................................... 11

2.

COST INFORMATION AND TRANSPORT POLICY OBJECTIVES ...................................... 12 2.1. Cost estimation in practice: social costs and cost functions................................................... 14 2.2. Cost function estimation and regulatory regimes................................................................... 17

3.

CONCLUSIONS ........................................................................................................................... 18

ANNEX: SKETCH OF COST ESTIMATION INSTRUMENTS ........................................................ 20 BIBLIOGRAPHY ................................................................................................................................. 24

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

SUMMARY OF DISCUSSIONS -

11

1. INTRODUCTION

The issues examined at the Round Table on the “Estimation and Evaluation of Transport Costs” were introduced by four background papers. Antonio Estache (World Bank) and Lourdes Trujillo (Universidad de Las Palmas de Gran Canaria) contributed a paper that discussed the importance of cost estimation in evaluating transport policies and reviewed the principal methods employed. Piet Rietveld et al. presented a paper on the estimation of costs in different transport modes and their use in setting regulatory limits to infrastructure charges. Carlos Barros’ paper set out an econometric approach to measuring the efficiency of airports and applied it to airports in Portugal. Finally, Philippe Gagnepain (Universidad Carlos III, Madrid) and Marc Ivaldi (Université de Toulouse) extended this econometric approach to the estimation of cost functions by taking into account the effect of incentives in regulatory regimes on the efforts of public transit firms and airlines to reduce their costs. The Round Table was chaired by Tae Oum (University of British Columbia), a leading researcher on transport costs, and on policies for the aviation sector in particular. Other leading experts from Europe, Japan and the USA, with backgrounds in engineering as well as economic approaches to the topic, participated in the discussions. The motivation for the Round Table was several-fold: −

Despite the almost universal commitment of transport policy to achieve efficiency, the empirical basis for assessing the productivity of the sector, either in the provision of infrastructure or the operation of transport services is generally weak. Relevant data are not available or they lack credibility due to their poor quality.

−

Costs need to be understood in order to be able to set pricing rules for public and private providers of transport services. For this, information is required on the resource requirements for the technically efficient provision of infrastructure, i.e. the least cost for providing a unit of additional service of a certain quality. The efficient cost can differ from currently observable costs for at least two reasons: absence of competition, which provides scope for administrative rents; over-specification of prestige projects.

−

More generally, as not all transport operations and infrastructure facilities can be exposed to competition in the market (Knieps, 2005), regulators critically depend on information on costs and cost functions to determine the scope and substance of regulation. Even regulation through yardstick competition (Bouf and Leveque, 2005), which rewards transport and infrastructure firms for being close to best practice, requires empirical identification of the best-practice cases.

−

Even firms that currently have the lowest level of costs might fail to adopt the lowest cost strategies and innovations available. Up-to-date engineering information is therefore essential for regulating prices and providing incentives for adopting new technologies. As local prices for inputs, including labour, may differ widely, an economic or econometric analysis will have to complement engineering information in determining least costs. The

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

12 – SUMMARY OF DISCUSSIONS Round Table therefore sought also to provide for an exchange between experts with differing professional perspectives on the assessment of costs and efficiency in the transport sector. Discussions at the Round Table were structured around the four papers, as follows: 1.

After the presentation by Antonio Estache and Lourdes Trujillo, the pros and cons of the different methods of estimating and evaluating transport costs were examined. The usefulness of each method depends on the purpose of the cost analysis and the costs of its application.

2.

As a concrete example of the difficulties that arise in studies that include the estimation of social costs, and the problems that arise in practice from the lack of data and resources available, the Round Table discussed a proposal for a charging mechanism to recover maintenance costs in all land transport modes in the Netherlands (Rietveld et al.).

3.

As an application of a formal cost benchmarking analysis, the participants discussed the comparative analysis of the efficiency of Portuguese airports, presented by Barros.

4.

This type of analysis was extended to include policy variables reflecting the regulatory regime in force, on the basis of the paper by Gagnepain and Ivaldi. The merits of benchmarking were discussed on the basis of its application to public transport regimes in French cities and the liberalisation of European airlines.

2. COST INFORMATION AND TRANSPORT POLICY OBJECTIVES

Perhaps the most fundamental objective of transport policymakers is to achieve an efficient transport sector, or a transport sector that contributes as much as possible to overall economic development. This implies that transport policy should aim at achieving mobility for passengers and goods with a minimum of inputs and a minimum sacrifice in the consumption of other goods and services. More specifically, higher than minimum transport costs reduce the competitiveness of national economies by increasing the prices of imported goods, reducing the net returns from exporting goods and reducing real incomes by increasing the prices of domestic products (e.g. Clarke, Dollar and Micco, 2004). At the subnational level, high transport costs distort the division of labour between different regions and remove part of the agglomeration benefits of cities and metropolitan areas. These agglomeration effects are considered to be central to knowledge accumulation and long-run economic growth (Black and Henderson, 1999).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

SUMMARY OF DISCUSSIONS -

13

At the local level, high transport costs impede the functioning of goods and factor markets. An example of such a malfunctioning of markets that has recently received much attention in research is the spatial mismatch in labour markets resulting from the high costs of intra-urban passenger transport (Patacchini and Zenou, 2005). In general, a reduction in production costs increases the real income of the population. Pressures to reduce costs are normally brought about by market forces. However, major parts of the transport sector are not subject to the market forces that guide firms and households to adapt to the supply of inputs and to adopt least-cost technologies (Round Table 129 – Transport Services: The limits of (de)regulation). In this context, transport policy measures are required to signal relative scarcities to producers and consumers and to help the implementation of least-cost solutions in providing infrastructure and transport services. These incentives can only be provided if the costs of infrastructure services and transport operations are assessed and monitored. Effective regulatory and pricing policies depend on reliable and timely cost information. The fact that cost data will be used to regulate firms that are at the same time data providers may lead to resistance against data collection. In such cases, data collection may be associated with high costs to monitor its quality. Moreover, data on the actual costs of infrastructure facilities and transport services may be rather uninformative with respect to the overall efficiency objectives. If providers enjoy monopolistic powers through ownership of an essential facility, some of the costs may reflect monopoly rents. These may result from overstaffing or higher salaries than those paid for similar jobs in private industry. In other cases monopoly service providers may have little difficulty in obtaining government subsidies to cover deficits. The reported cost levels might then, for example, reflect a high degree of unionisation of the workforce, which can be used to negotiate high wages and high levels of employment (cf. Laffont and Tirole, 1986). As a first step, measuring cost levels observable in the market allows the identification of best practice infrastructure facilities or transport firms. Knowing about best practice can be used to reduce relative inefficiencies in the market by providing information for regulatory incentives to imitate best-practice technologies and behaviours. Even accepting this reduced ambition, it is not obvious how costs should be measured and what are the most useful accounting methods. Regulation is helped substantially by establishing separate cost accounting entities for those parts of transport activities and infrastructure services which are or can be exposed to market pressure to reduce costs (Knieps, 2005), and those which enjoy the powers of a natural monopoly or an essential facility (Estache and Trujillo). Such a separation should reduce the opportunities to diffuse regulatory pressures by arbitrarily assigning cost items to different activities. Discussions on the relative performance of transport firms and infrastructure facilities suffer from the fact that there is no unique way to measure performance. If partial performance measures are used there are naturally a number of indicators that might give conflicting signals on performance. A simplified illustration is provided in the table below. Table 1. Partial Performance Indicators Operator A B

Labour (L) 200 400

Capital (K) 2 1

Output (Y) 2000 2000

Source: Estache/Trujillo.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

Y/L 10 5

Y/K 1000 2000

14 – SUMMARY OF DISCUSSIONS Looking at labour productivity in column 5, operator A would rank higher. Choosing capital productivity as the ranking index, operator B would perform better than operator A. The ambiguity of the partial performance measure can only be resolved if output per “bundle” of inputs is measured. The reference bundle for performance measurement can, unfortunately, be defined in different ways, too. The measurement of total performance of a firm or facility requires the use of index numbers to express total factor productivity (Coelli et al. 1998, ch. 4). In most empirical applications, the Toernquist Index formula is used for purposes of output and input index calculations (see Annex). For this index, as for most other applications when aggregating the influence of individual inputs on output, a functional relationship is assumed which implies that a certain percentage increase in individual inputs leads to a constant percentage increase in output. Measuring the efficiency of a facility or firm has three dimensions, which the formal cost estimation methods try to disentangle (see Annex). − Is the least-cost technology used? That is, are there other technologies that require smaller amounts of one or more inputs per unit of output? Answers to these questions determine whether the firm is technologically efficient. − Does the provider adequately respond to the prices of the inputs needed? In other words, is the right combination of inputs used? Allocative efficiency is achieved if no cost savings could be made by changing the shares of the different inputs in total costs. − Scale efficiency is achieved if the costs per unit of service cannot be reduced by changing the size of the firm or the facility. The Round Table discussed to what extent the standard production function and cost function estimation methods neglected engineering information. That engineering data can be usefully integrated into economic simulation models to help regulators has recently been demonstrated for the telecommunications sector (Gasmi et al., 2002). However, a research programme on engineering production functions in economics (cf. the review in Wibe, 1984; and Chenery, 1992), that included some remarkable applications to the transport sector, found very little impact on cost estimation and evaluation. 2.1. Cost estimation in practice: social costs and cost functions Even if it is clear in principle what cost information is needed and how cost estimations should be undertaken, applications to solve concrete policy problems often follow very different tracks. The reasons for this are many: − First of all, shortage of cost data seems to be a particular problem for the transport sector. − As already mentioned, data quality is imperfect in cases where the interests of primary data providers depend on the reported data. − Data collection costs or the costs of monitoring data may be such that the benefits of collection do not justify the expense. − The application of cost estimation and cost evaluation methods may be difficult to communicate in the policymaking process and imperfect surrogate methods are applied instead. ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

SUMMARY OF DISCUSSIONS -

15

The imprecision of quantitative information obtained from employing rough ad hoc measures can be high. The Round Table discussed these problems in relation to the background paper of Rietveld et al., on the introduction of infrastructure charges for road, rail and inland navigation to recover infrastructure maintenance costs in the Netherlands. The starting point for the paper is the concept of social marginal cost pricing: charging infrastructure users according to the additional costs of infrastructure use is the pricing rule that should induce optimal use of infrastructure facilities. A price higher than marginal costs would induce an underutilisation of the facility and lower prices an overuse, with high levels of congestion. It is acknowledged that in principle not only additional infrastructure costs should be charged, but also external costs in the form of environmental damage, increased safety risks, increased congestion costs, etc. at the margin. These concepts for pricing of infrastructure services are, however, unrelated to cost recovery. In general, a self-financing charging policy would require marginal cost pricing per unit of service plus a fixed access charge to cover the full costs. Moreover, incentives to contain the external costs of transport are not necessarily best provided by charges per unit of infrastructure use. Other instruments to internalise some of them are in place (such as fuel taxes to contain CO2 emissions). These might have to be revised should charging for external costs be transferred to charges for infrastructure use. Against the backdrop of these difficulties, a relatively simple model of cost allocation has been proposed. It starts from the observation that not all maintenance costs vary with infrastructure use and that there are hence fixed and variable maintenance costs. For the cost recovery scheme, both the variable and fixed infrastructure costs are allocated to different vehicle types, leading to recommendations on charges per distance unit. The core marginal costs are identified by engineering information. For roads, they are computed on the basis of axle load factors, which are calculated using the AASHO 4th Power Rule. Similar engineering information has been used to calculate marginal infrastructure costs for rail and inland waterways. Vehicle-km charges to cover marginal vehicle-km costs in terms of wear and tear, noise and increased accident risks will not cover the full average costs per vehicle-km. This holds in particular if congestion costs – the external costs directly linked to infrastructure use – are disregarded in the price calculation. If we exclude non-linear pricing schemes, fixed costs have then to be allocated to vehicle types and translated into per-km charges. As is shown in the cost figures of the background papers, even if we restrict the cost recovery scheme to maintenance costs, a substantial share of the costs are fixed costs, for example 55 per cent of the Dutch highway maintenance costs are fixed costs. The allocation of fixed costs to vehicle-km prices often follows ad hoc rules, referring to technical characteristics of vehicles and/or external costs which are expected to be associated with certain vehicle types. The Dutch proposal for the recovery of maintenance costs allocates fixed maintenance costs to vehicle types according to average vehicle sizes. By doing so, the proposal follows the recommendation of the European Commission to impute fixed costs on the basis of kilometre equivalence factors (EU Directive 1999/62/EC). The base rate calculated in this way is differentiated according to the social external costs of transport, in particular the contribution of vehicles of different classes to environmental damage. The Round Table discussion revealed the problems of partial cost recovery schemes, with the restriction of cost recovery to maintenance costs as a particular example. If two different modes had identical total costs but different shares of construction and maintenance costs, basing pricing decisions on maintenance costs only discriminates against the maintenance-intensive mode. The cost ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

16 – SUMMARY OF DISCUSSIONS estimates presented for a maintenance cost recovery scheme suggest that charges per rail passenger-km or rail tonne-km would be far higher than for roads. This result might or might not be confirmed by an analysis of total costs (including construction costs). Translating fixed costs into per unit charges for services would in general conflict with a rational use of infrastructure. It could be inferior to a two-part combination of fixed access fees and marginal, cost-based, per-km charges (see Round Table 135; and ECMT, 2005). For non-congested infrastructure facilities, full cost pricing will in general lead to an overpricing and underutilisation of the infrastructure. Moreover, over- or underutilisation of infrastructure can easily be induced by accounting for external costs of transport in an ad hoc manner. Of particular importance for the modal split are congestion costs which should be an integral part of pricing schemes. Imprecise estimation methods - such as basing the estimate of additional congestion costs just on the size of a vehicle, which is then used for fixed maintenance cost distribution to vehicle types - are likely to lead to an under-pricing of the use of congested facilities. They might contribute to a failure to achieve congestion reduction targets and, by over-pricing of uncongested facilities, to an under-utilisation of uncongested infrastructure. A more complete collection of data, a precise estimation of costs and cost functions and the implementation of more complete cost allocation schemes may pay off by helping the optimisation of infrastructure capacities and inducing use levels that lead to prices closer to the minimum costs per unit of infrastructure service. Barros’ paper proposed to measure the efficiency of Portuguese airports by means of a stochastic cost frontier, presented in detail in the Annex. The stochastic cost frontier estimation aims at the identification of the minimum costs for all output levels, taking into consideration that there might be random measurement errors or omitted variables. The estimation model extends the standard approach set out in the Annex by accounting for “inefficiency effects”, effects that influence the distribution of the error term of the estimation function. The most important result of the exercise is a strong indication of the size advantages of airports. Given the public policy concern about the fiscal burden arising from the deficits of airports, the result might suggest an indication of airport overcapacity. For conclusions on national airport policies, further analysis of airport operations might be required. The standard cost frontier estimation does not take into account that different airports might serve different sub-markets in a differentiated overall market for airline services. For example, network effects make it desirable for international carriers to have a single, large national hub. The scale effects shown in the cost estimation may not be replicable for other airports. Moreover, the expected increase in the specialisation of airports (cargo, low-cost carriers) might lead to decreased average costs, or the increased productivity of smaller airports in the future. It might appear tempting to conclude from the results that a reduction in the number of airports and an increase in their average size should in general lead to efficiency gains. This would disregard congestion costs which form an important cost component of nodal infrastructure facilities. Adding congestion costs to the core infrastructure costs would fundamentally change the cost function of infrastructure facilities like airports. Instead of a cost function which implies continuously falling costs per unit of service, a u-shaped cost curve would result, with the increasing congestion costs of additional traffic leading to increasing total average costs. The minimum costs per unit of airport service could serve as an indicator of the desirable airport capacity. The results of the study on the costs of Portuguese Airports confirm the need to include external costs, in particular congestion costs, if the aim is to identify optimal airport sizes. ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

SUMMARY OF DISCUSSIONS -

17

2.2. Cost function estimation and regulatory regimes Gagnepain and Ivaldi (2007) extend the standard stochastic cost function approach by introducing information on regulatory regimes and associated incentives to reduce costs. Differences in the incentives created by regulatory regimes arise from differences in dealing with the asymmetries in information available to firms and regulators. The implications are explored in a study of public transport in France and deregulation of European airlines. Thirteen urban areas were included in the study of 59 networks, and data collected over a period of eight years, from 1985 to 1993. In these areas, urban public authorities are in charge of regulating transit systems, with services provided by single operators. Two regulatory regimes are observed in practice. First, under a cost-plus contract, all costs of the public transit firm are reimbursed ex post. As such a scheme provides at best very weak incentives to reduce costs, the estimation approach assumed that in such a context firms do not try to cut costs. The second regulatory regime is a fixed price contract, leaving the operator with the responsibility for insufficient revenues and cost overruns. As 60 per cent of the costs of public transit are labour costs, the cost reduction effort will mainly focus on labour, through training and organising work to avoid conflict. Results in two-thirds of the sample clearly confirm expectations: a first group with the highest productivity operates under a fixed-price contract, suggesting that the regulatory regime has a decisive influence on the efforts to reduce costs. The vast majority of a group of 20 operators in the middle range of productivity operate under a cost-plus regulatory regime, in line with the expectation that firms whose costs are reimbursed take less action to reduce costs. Complicating this clear picture, the productivity of a third group, with the worst performance, contains firms operating under both regulatory regimes. This shows that factors other than the regulatory regime can play a determining role in performance. The result can, for example, be interpreted as indicating that some cities suffer from infrastructural and institutional legacies that prevent fixed-price regulation achieving the same productivity-enhancing effects realised for the best performers. On the liberalisation of the European airline industry, alternative scenarios were compared for the impact of the deregulatory packages introduced by the European Commission: a) Firms realise efficiency independently of de-regulation; i.e. the estimation model does not need an effort and inefficiency term; b) Firms are inefficient but do not react to deregulation by cutting costs; c) Firms start efforts to cut costs after the implementation of the third EU package in 1992; d) Deregulation changes the behaviour of firms affected by the introduction of bilateral agreements (British Airways, KLM, Lufthansa and Sabena after 1985, and other airlines in 1993). Cost estimation allows testing of these scenarios against each other, with clear implications for regulatory policies. −

Scenario (d) is rejected in favour of scenario (c). This suggests that the third aviation package was far more effective than bilateral agreements in affecting the cost reduction efforts of the airlines.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

18 – SUMMARY OF DISCUSSIONS −

Scenario (a) rejected in favour of scenario (c): that is, the inclusion of an inefficiency term and, accounting for the possibility that deregulation changed the airlines’ cost management, led to superior estimation results.

−

The standard approach of cost estimation suggests that the European airline industry is characterised by increasing returns to scale. By contrast, Scenario (c), which fared better in the tests, led to the result that it is characterised by constant returns to scale.

Overall, the results confirm the positive impact of airline deregulation on the adoption of cost-reducing technologies and its effectiveness in inducing efforts to increase labour productivity. To obtain a better understanding of the potential for cost reductions, and the implications this has for market structure, combining engineering information and econometric cost estimations to construct simulation models for transport sub-sectors may be helpful. Research along these lines has so far been restricted to the market for telecommunications services (Gasmi et al., 1999, 2002).

3. CONCLUSIONS

The Round Table addressed the following questions: −

Do planners and researchers need a broader statistical data base for transport policymaking?

−

Are the cost estimation methods used in practice adequate to generate the information required for policymaking?

−

What are the limits of currently available cost assessment methods? What should guide the selection of methods?

−

Are there deficiencies in currently available approaches to estimating and evaluating costs; is there a need to develop them further?

Based on the background papers, the Round Table discussion arrived at the following answers: −

There was a consensus that there are deficiencies in the statistical information available, for example, in comparison with the data available for other infrastructure sectors. Current resource allocations to transport data collection in some member countries causes concerns that the situation is not improving.

−

Decisions on the scope and design of data collection efforts face a number of problems. Extending the scope of data collection requires a careful estimation of the benefits for transport planning, and an answer to the question of whether they justify the additional costs. Data quality problems arise from the informational asymmetries between data users and data providers. At least in some instances, the anticipation of data use, for example for regulatory purposes, will invite the production of distorted data. When this concerns technological information, engineering data may help to check data quality. ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

SUMMARY OF DISCUSSIONS -

19

−

Cost studies often take inadequate account of external costs. To avoid an ad hoc inclusion or exclusion of external cost data, a careful theoretical analysis must precede cost estimation. This can serve to identify categories of external costs that might be largely irrelevant to regulatory policy, but would be particularly costly to analyse. For example, road damage and congestion costs are essential for road pricing decisions, while other types of external costs may not be.

−

Lack of data might make it difficult to avoid the application of ad hoc methods to assess costs in practice. Examples of infrastructure cost recovery mechanisms showed that the application of crude cost estimation approaches is likely to have negative effects on the design of regulation, leading to inappropriate incentives and unintended policy consequences.

−

The Round Table discussed the pros and cons of partial cost indicators, indices of overall factor productivity and methods to estimate cost functions. The estimation of cost functions allows for the disentangling of the technological factor price and scale dimensions of costs. However, as moving from partial to general cost estimates implies increasing costs for the analysis, a critical assessment of what information is necessary for what policy or planning problem is required. Participants reported that policy discussions often suffered from reading more into simple cost indicators than is justified.

−

Clear progress has been made in cost assessment methods by accounting for asymmetric information and consequent incentive structures in the cost estimation approaches. Extension by including engineering information could help to address the fact that least-cost technologies may not be observable in the market and address data quality problems arising from self-interest of data providers to distort the data reported. An example of the feasibility of such an extension exists for the telecommunications sector.

Without information about cost levels, it is impossible to identify resource requirements for transport policy. Knowledge of best practice technologies and their associated costs provides the basis for setting operational productivity targets and, as was discussed during Round Table 129, substantial data collection and cost estimation efforts are essential for the implementation of new regulatory mechanisms such as yardstick competition.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

20 – SUMMARY OF DISCUSSIONS

ANNEX: SKETCH OF COST ESTIMATION INSTRUMENTS

Index numbers try to avoid the deficiencies of partial cost indices. They aim at indicating the ratio between output variables and a bundle of inputs. The most frequently used productivity index is the Toernquist index. The Toernquist total factor productivity index is defined in its simpler logarithmic form as follows, comparing two entities, s and t :

ln TFPst = ln

OutputIndexst = ln Output Indexst − ln Input Indexst InputIndexst

1 K 1 N ( ) = ∑ (ω it + ω is ) ln yit − ln yis − ∑ (υ jt + υ js )(ln x jt − ln x js ), 2 j =1 2 i=1 where y denotes outputs, indexed by i, and x denotes inputs, indexed by j1. The ω and υ denote the shares of goods i in total real output, and the shares of inputs j in total costs. Total factor productivity indices indicate the overall cost effectiveness of projects or firms without, however, giving any indication about its sources. Of particular interest is whether inefficiencies are due to the use of technologies other than least-cost (technical efficiency) and/or whether providers fail to respond to the appropriate price signals (allocative efficiency). Moreover, inefficiency can result from mistakes in the choice of the capacity of firms or infrastructure facilities. In general, infrastructure capacity, and often transport operations, can only be changed in large, discrete steps. An important dimension of average cost levels in transport is therefore scale economies. Indivisibilities and network economies which beset capacity choice in transport imply that average costs are decreasing for some potential levels of operation. To achieve scale efficiency, i.e. to choose the size of a firm or infrastructure facility such that it operates close to the minimum average costs, requires a precise forecasting of demand and the possible degree of congestion. To make empirical statements about the different aspects of efficiency, the quantitative relationship between inputs and output(s) of the fully efficient firm or facility must be known. Such a production function is not known in practice. The bulk of the literature on the identification of an empirical production function follows the suggestion in Farrel’s (1957) seminal article, to estimate such a function from sample data using either a non-parametric piece-wise linear technology or a parametric function. The first suggestion has developed into the Data Envelopment Analysis. The Data Envelopment Analysis tries to identify the set of minimal combinations of inputs required to produce

1

There are other index numbers which can be used for productivity measurement, which

differ by certain mathematical properties. Which index number should be used depends on the purpose of the productivity study and the characteristics of the index. The discussion on ideal index numbers has been inconclusive (Diewert, 1992). ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

SUMMARY OF DISCUSSIONS -

21

one unit of service (the unit isoquant) by linear programming techniques, as illustrated by Figure 1 (Charnes et al., 1995). Figure 1. Piece-wise linear convex unit isoquant in case of two inputs and one output

x2/y S

•

• •

S’

0

Source: OECD/ECMT (2007).

x1/y

All combinations of the inputs x1/y and x2/y which lie northeast of the convex hull SS’ are inefficient. One unit of the service could be supplied with a smaller (at least one) input. The approach used to identify the efficient relationship between inputs and output, implying the duality of minimum costs, is the frontier estimation approach (Färe et al., 1994). The frontier estimation approach tries to distil, from observed input and output figures, the least-cost values in the market. To illustrate the discussion on what is actually estimated, Figure 2 assumes that there is only one input x and one output y.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

22 – SUMMARY OF DISCUSSIONS Figure 2. Estimation of a frontier cost function y I

•

II

• •

•

•

• • •

0

•

• •

x Source: OECD/ECMT (2007).

Originally, the approach to determine a parametric production function started out from the following model:

ln( yi ) = ln ( xi )β − ui , with i = 1,..., N and ui ≥ 0, (graph I) where xi is a (K+1)-row vector of the inputs of firm i, yi the (scalar) output of the firm, β the (K+1) of parameters to be estimated and ui a non-negative variable, indicating the inefficiency at the firm level. Aigner and Chu (1968) proposed in their pioneering article to determine the parameters by linear or quadratic programming, minimising the sum of absolute residuals ui or the sum of squared residuals, respectively. This deterministic frontier model was criticised for being unable to take account of measurement errors, omitted variables or unpredictable, random human responses (Schmidt, 1976). As a response to this criticism, Aigner, Lovell and Schmidt (1977) as well as Meeusen and van den Broeck (1977) independently proposed the stochastic frontier production function, adding a random error εi :

ln( yi ) = ln( xi )β − ui + ε i , with i = 1,..., N and ui ≥ 0, (graph II) The εi’s are assumed to be independent and identically distributed normal random variables with a mean of zero and a constant variance. The inefficiency variables, required to be non-negative, are assumed to be exponential or half-normal variables. The stochastic frontier model permits the estimation of standard errors and hypothesis testing, using maximum likelihood methods, which was not possible with the early deterministic models. ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

SUMMARY OF DISCUSSIONS -

23

In the background paper of Gagnepain and Ivaldi (2005), the level of inefficiency depends on the effort of the management of firm i, denoted by ei, to produce efficiently, i.e. the effort to reduce the ui’s. With the effort level included in the estimation equation, we have:

ln( yi ) = ln( xi )β + ε i + g (ei − ui ) , with i = 1,..., N and ui ≥ 0. The estimation of a cost frontier is obtained from minimising expenses for variable inputs for a given output level. The cost frontier is then a function of input prices, the output level, the inefficiency level and the effort to reduce the inefficiency.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

24 – SUMMARY OF DISCUSSIONS

BIBLIOGRAPHY

Aigner, D.J. and S.F. Chu (1968), On estimating the industry production function, American Economic Review, 58: 826-839. Aigner, D.J., C.A.K. Lovell and P. Schmidt (1977), Formulation and estimation of stochastic frontier production models, Journal of Econometrics, 6: 21-37. Barros, C.P. (2007), Technical efficiency in Portuguese airports with a stochastic cost function. In OECD/ECMT Transport Research Centre (ed.), Estimation and Evaluation of Transport Costs, Round Table 136, OECD, Paris. Black, D. and V. Henderson (1999), A theory of urban growth, Journal of Political Economy, 107: 252-84. Bouf, D. and J. Leveque (2005), Yardstick competition for transport infrastructure services, in: OECD/ECMT Transport Research Centre (ed.), The Limits of (De-)Regulation of Transport Infrastructure Services, Round Table 129, OECD, Paris. Charnes, A., W.W. Cooper, A.Y. Lewin and L.M. Seiford (1995), Data Envelopment Analysis: Theory, Methodology and Applications, Boston. Chenery, H.B. (1992), From engineering to economics, Banca Nazionale del Lavoro Quarterly Review (183): 369-405. Clark, X., D. Dollar and A. Micco (2004), Port efficiency, maritime transport costs and bilateral trade, Journal of Development Economics, 75: 417-450. Coelli, T., D.S.P. Rao and G.E. Battese (1998), An Introduction to Efficiency and Productivity Analysis, Dordrecht. De Salvo, J.S. (1969), A process function for rail line haul operations, Journal of Transport, Economics and Policy (3-27). Diewert, W.E. (1992), Fisher ideal output, input and productivity indexes revisited, Journal of Productivity Analysis, 3: 211-48. ECMT (2005), Rail Reform and Charges for the Use of Infrastructure, OECD, Paris. European Union (1999), Directive 1999/62/EC of the European Parliament and the Council concerned with the Charging of the Use of Certain Infrastructure Services to Heavy Goods Vehicles, Brussels.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

SUMMARY OF DISCUSSIONS -

25

Estache, A. and L. Trujillo (2007), Transport cost levels, productivity and efficiency measures: some theory and main policy conclusions, in: OECD/ECMT Transport Research Centre (ed.), Estimation and Evaluation of Transport Costs, Round Table 136, OECD, Paris. Farrell, M.J. (1957), The measurement of productive efficiency, Journal of the Royal Statistical Society, Series A, 120, Part 3: 253-290. Färe, R., S. Grosskopf and C.A.K. Lovell (1994), Production Frontiers, Cambridge, Mass. Gagnepain, P. and M. Ivaldi (2007), Measuring inefficiencies in transport systems: between technology and incentives, in: OECD/ECMT Transport Research Centre (ed.), Estimation and Evaluation of Transport Costs, Round Table 136, OECD, Paris. Gasmi, F., D.M. Kennet, J.J. Laffont and W.W. Sharkey (2002), Cost proxy models and telecommunications policy, Cambridge, Mass. Gasmi, F., J.J. Laffont and W.W. Sharkey (1999), The natural monopoly test reconsidered: An engineering process-based approach to empirical analysis in telecommunications, International Journal of Industrial Organisation, 20: 43-59. Hildenbrand, W. (1981), Short run production functions based on microdata, in: R.J. Aumann, J.C. Harsanyi, W. Hildenbrand, M. Maschler, M.A. Perles, J. Rosenmüller, R. Selten, M. Shubik and G. L. Thompson (eds.), Essays in Game Theory and Mathematical Economics in Honor of Oskar Morgenstern, Mannheim. Knieps, G. (2005), The limits of competition in transport markets, in: OECD/ECMT Transport Research Centre (ed.), The Limits of (De-)Regulation of Transport Infrastructure Services, Round Table 129, Paris. Laffont, J.-J. and J. Tirole (1986), Using cost observation to regulate firms, Journal of Political Economy, 94: 614-41. Meeusen, W. and J. Van den Broeck (1977), Efficiency estimation from Cobb-Douglas production functions with composed error, International Economic Review, 18: 435-44. Patacchini, E. and Y. Zenou (2005), Spatial mismatch, transport mode and search decisions in England, Journal of Urban Economics, 58: 62-90. Rietveld, P., F.R. Bruinsma and M.K. Koetse (2007), Infrastructure maintenance costs: a comparison of road, rail and inland navigation and implications for user charges, in: OECD/ECMT Transport Research Centre (ed.), Estimation and Evaluation of Transport Costs, Round Table 136, Paris. Schmidt, P. (1976), On the statistical estimation of parametric frontier production functions, Review of Economics and Statistics: 238-239. Smith, V.L. (1957), Engineering data and statistical techniques in the analysis of production and technological change: Fuel requirements in the trucking industry, Econometrica, 25: 281-301. Wibe, S. (1984), Engineering production functions: A survey, Economica, 51: 401-11.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INTRODUCTORY REPORTS -

INTRODUCTORY REPORTS

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

27

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES -

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES

Philippe GAGNEPAIN Universidad Carlos III Madrid Spain

Marc IVALDI Universite de Toulouse Toulouse France and CEPR London United Kingdom

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

29

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES -

31

SUMMARY

INTRODUCTION ................................................................................................................................. 33

1.

THE PRELIMINARY FRONTIER............................................................................................... 35

2.

INCENTIVES................................................................................................................................ 37

3.

REGULATION OF PUBLIC TRANSIT IN FRANCE ................................................................ 38

4.

THE DEREGULATION OF EUROPEAN AIRLINES ................................................................ 43

5.

CONCLUSION ............................................................................................................................. 46

BIBLIOGRAPHY ................................................................................................................................. 47

Toulouse and Madrid, September 2005

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES -

33

INTRODUCTION

In modern microeconomics, a firm associates various inputs and a specific technology to reach a particular level of production. The theory of frontiers has defined the notion of a maximum level of production that can be reached, given the inputs and the technology available. Such a frontier becomes a reference for the producers, in the sense that all firms are affected during the production process by productive inefficiencies. These inefficiencies have been studied by the econometricians interested in estimating frontiers of cost. Originally proposed by Aigner, Lovell and Schmidt (1977) and Meeusen and van den Broeck (1977), a frontier model consists of a function of the usual regression type with an error term comprising various parts. The first part corresponds to the usual white noise process, while the second part represents inefficiency. As is well established in the literature, global inefficiency of individual sample firms can be predicted on the basis of cross-sections or panel data sets on these firms. The method based on cross-sectional data suffers from one serious difficulty: the estimation procedure must assume that inefficiency is independent of regressors. This might be incorrect since input and output quantities are together determined at the equilibrium, and since firms may know something about their level of inefficiency when they choose input quantities. Now, consider the incentives for cost reduction faced by a producer during the production process. This article provides two examples to show that the cost frontier of this producer involves a global inefficiency term that comprises two terms. The first component is a purely exogenous random term; the second is endogenous in the sense that it depends on the producer’s actions and hence on observable variables in an indirect and imbricated way. This decomposition of global inefficiency resembles the specification commonly used in the literature on stochastic frontiers. However, it is here endogenously derived while, in the tradition of stochastic frontiers, it is imposed in an ad hoc way. This resemblance finds its source in the economic literature. On one side, a tradition initiated by Leibenstein (1966) motivates the specification of stochastic frontiers. Without referring explicitly to the notion of frontier, Leibenstein clearly mentions the existence of a global inefficiency that depends on the will of managers and workers in a production process. A low-powered incentive environment, due to workers’ lack of productivity, induces the inefficiency, while appropriate incentives can lead to significant operating cost reductions. On the other hand, since the emergence of the new theory of regulation, economists admit that, in industries where a producer is regulated by an authority, the principal-agent relationship is at the core of the question of assessing the performance of a firm (see Loeb and Magat, 1979; Baron and Myerson, 1982; and Laffont and Tirole, 1986). Hence, technical inefficiency and effort are two unobservable parameters, which characterise the incentives faced by a firm to reduce costs and define the source of global inefficiency. Likewise, in industries where several producers enjoy a local monopoly power, a sudden opening of the market to perfect competition may create a new pressure, in terms of incentives for carriers to reduce costs and improve efficiency. ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

34 - MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES In this perspective, incentive models provide a relevant framework for the analysis of cost frontiers. In addition, because such models are able to elicit the structural relationship between the observable variables and the inefficiency term, they directly provide a way to deal with the endogeneity of the inefficiency term, i.e. with the stumbling block of the econometrics of stochastic frontiers. The analysis presented below is based on two examples of cost reduction incentives. The first one focuses on the regulatory structure drawn from the French urban transport industry. There are two types of contract that regulate the activity of transport operators, and these types of contract provide operators with different incentives to reduce costs. This analysis is based on the work by Gagnepain and Ivaldi (2002). A second example is the deregulation of European aviation. Until the beginning of the 1980s, flag operators had enjoyed monopoly power and had kept costs and prices high. The introduction of several waves of deregulation from 1985 to 1993 has allowed competition in the European industry, forcing firms to be more efficient and to reduce costs if they want to stay in the market. This analysis is based on the work by Gagnepain and Marin (2005). Chapter 1 presents the preliminary cost frontier to be estimated using a Cobb-Douglas technology. Chapter 2 is devoted to the way incentives for cost reduction are introduced in this cost frontier. Chapters 3 and 4 present two applications: on urban transportation in France, and deregulation of European airlines, respectively. Chapter 5 concludes.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES -

35

1. THE PRELIMINARY FRONTIER

The aim of this chapter is to build the structural cost and production frontiers of a transport operator. First, preliminary frontiers, conditional on effort, are obtained. Second, using the regulatory or the competitive constraints, we construct structural frontiers to be estimated. We specify a stochastic production function as: (1)

ln Y = ln f ( X ,α ) + g (e − θ ) + ε Y ,

where f ( X , α ) is the production function, i.e. the locus of technically efficient production levels. Here, the levels of output and input are given by Y and X, respectively. The parameter ε Y is a symmetric statistical noise which accounts for measurement errors. Moreover, θ ≥ 0 is the exogenous technical inefficiency, and e ≥ 0 is the effort of productivity exerted by the firm. Technical inefficiency represents the amount of knowledge, the experience of the transport operator and its ability to associate inputs with the technology available. A cost-reducing effort is exerted in order to reduce the technical inefficiency θ . The function g (e − θ ) then provides a measure of the global inefficiency. The nature of this function depends on the source of inefficiency and effort activity in the production process. We assume that g (.) is strictly increasing with θ and strictly decreasing with e. A measure of the total distortion under the production frontier is naturally defined by the ratio: (2)

Y [ f ( X ,α )exp(ε Y )] = exp[g (e − θ )] ,

which tells us how far is actual output from the most efficient production level, represented by the stochastic production function ⎡⎣ f ( X , α ) exp ( ε Y )⎤⎦ . The regulator knows neither θ nor e. In order to provide the level of output Y, a transport operator requires quantities of input X j , j = 1 ,... n , from a set X of inputs. The utilisation and management of the set of inputs are affected by technical inefficiency θ . This leads to an over-consumption of the inputs. By exerting a significant level of effort, the monopoly can reduce this excess of factors demand. * Then we distinguish X i from X i . On the one hand, X i is the physical amount of input i used by the producer in the process. This amount is observable by the regulator. On the other hand, X i* is the efficient level of input i which is not observable by the regulator. Hence, operating costs depend on the quantity X i whereas the actual level of production depends on X i* . The relationship between observed and efficient quantities of input i is given by:

X i* =

(3)

Xi

exp(θ − e ) ,

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

36 - MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES that is to say, the efficient quantity is expressed as a percentage [measured by exp (θ − e ) ] of observed quantity. Note that θ is expected to be greater than e. However, we do not impose this constraint when we estimate the model. Consider now a Cobb-Douglas technology: n

Y = A∏ X j k α exp[g (e − θ ) + ε Y ]

(4)

*α j

k

j =1

,

with one output and n variable inputs, where k stands for capital and A, α k , and the α j ’s are parameters describing the technology. Capital is considered as a fixed input. We turn to the construction of the dual cost frontier. Assume that the producer seeks to minimise the cost C of producing its desired rate of output Y under technical inefficiency. The associated allocation of inputs sets the factor demand X j , j = 1,... n . The programme of the producer is then: n

min ∑ w j X j Xj

(5)

j =l

,

under the technological constraint (4), where w j is the price of input j. The excess of factor demand above its frontier prevents the producer from reaching the theoretical level of production. Such a distortion of factor demands leads to a rise of operating costs. In logarithmic form, the associated stochastic cost frontier, C (Y , w, e, θ ) , is given by: n

ln C (Y , w, e,θ ) = Κ + ∑

(6)

j =1

(θ − e ) + ε 1 α ln w j + ln Y − k ln k + C , r r r r

αj

n

r = ∑α j

j =1 where measures returns to scale, K is a constant and w is the vector of input prices. Note that the term (θ − e ) r is the total cost distortion above the cost frontier. Global inefficiency is less significant when the industry enjoys large returns to scale r.

Both production and cost frontiers, in equations (4) and (6) respectively, allow the econometrician to estimate a technology in a similar way. The choice between estimating one functional structure or the other depends upon exogeneity assumptions. A cost function should be rather considered if output quantities are exogenous. Nevertheless, in any case, both frontiers in equations (4) and (6) are preliminary, since the unobservable structure is partially endogenous. The cost-reducing effort e is endogenous and depends on regulatory schemes or the competitive environment set by the regulator. We turn now to the cost-reducing activity aspect of the problem.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES -

37

2. INCENTIVES

We consider a transport operator whose profit function is: (7)

π = R (Y ) − C (Y , w, e,θ ) −ψ ( e ) ,

where R(Y) denotes its revenue. Exerting effort is costly and leads to internal cost ψ ( e) . A profit-maximising operator determines the optimal effort level in equation (7). The first order condition is given by: (8)

ψ ′(e ) = −C e ,

which states that the optimal effort level is chosen to equalise the marginal disutility of effort and the marginal costs savings. Let us define a specific functional form for the internal cost of effort: (9)

ψ (e ) = exp(τ e ) − 1,

where τ > 0 . Assume that ψ ′(0) = 0 . The first order condition (equation 8) can then be expressed using the cost frontier (equation 6) and the internal cost of effort (equation 9). The level of endogenous effort exerted by the operator is obtained as:

(10)

⎡ ⎤ α 1 n ln w j 1 θ Κ + +∑ + ln Y − k ln k + − ln τ ⎥ ln ⎢ r j =1 r r r r ⎥ e=⎢ ⎢ ⎥. τ +1 r ⎢ ⎥ ⎣ ⎦

The optimal effort level depends on input prices w j , production level Y, capital stock k, the inefficiency θ and the technology available α . The optimal effort level (equation 10) is now reintroduced in the preliminary frontiers (equation 6) in order to derive the final structural cost frontier to be estimated:

(11)

⎡ n ln w j 1 α 1 ⎤ + ln Y − k ln k + θ ⎥ + ε C , ln C = Η C + ξ ⎢∑ r r r ⎦ ⎣ j =1 r

ξ=

τ

τ + 1 r . Note that, when the effort of the producer is nil, the structural where Η C is a constant and cost frontier is given by the expression:

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

38 - MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES

(12)

n

ln w j

j =1

r

ln C = Κ + ∑

α 1 1 + ln Y − k ln k + θ + ε C , r r r

We now propose two applications of this model.

3. REGULATION OF PUBLIC TRANSIT IN FRANCE

We use a database created in the early 1980s. It assembles the results of an annual survey conducted by the Centre d'Étude et de Recherche du Transport Urbain (CERTU, Lyons) with the support of the Groupement des Autorités Responsables du Transport (GART, Paris), a nationwide trade organisation that gathers most of the local authorities in charge of an urban transport network. For our study, we have selected all urban areas of more than 100 000 inhabitants, for the purpose of homogeneity. However, the sample does not include the largest networks of France, i.e. Paris, Lyons and Marseilles, as they are not covered by the survey. The result is that the panel data set covers fifty-nine different urban transport networks over the period 1985-93. In each urban area, a public authority is in charge to regulate the transit system, which is provided by a single operator. The authority chooses the regulatory scheme that defines cost reimbursement rules and the final owner of commercial revenue at the end of the reference period, usually a year. Two types of contract are observed in practice. The first corresponds to the so-called cost-plus contract. This contract represents a very low-powered incentive scheme, as firms under this regime have no incentives to produce efficiently, since all operating costs are reimbursed ex post. With the second type of contract, the so-called fixed-price contract, the operator is a residual claimant for effort. This time it obtains a transfer equal to the expected balanced budget, which is the difference between expected costs and expected revenue. This contract is a very high-powered incentive scheme, as the operator is now responsible for insufficient revenues and cost overruns. In the public transit industry, the network operator is better informed on labour inefficiency than the regulator. Note that labour costs represent more than 60% of total costs in this industry. Bus drivers play a decisive role in operating the network, especially with respect to the flexibility and punctuality of operations in peak periods. First, bus drivers permanently meet the end users. Their behaviour vis-à-vis the customers may perceptibly affect the quality of service during high peak periods. Indeed, the driver has to perform several tasks at the same time, selling tickets, monitoring the passengers’ up-and-down movements, managing the use of bus seats and space. Clearly, these tasks are much harder to accomplish in periods of traffic congestion. Moreover, drivers have to deal with social and security problems, particularly in areas where the underprivileged population is large. There is an additional feature worth mentioning. The network structure may affect driving conditions. On the same network, each bus route has its own characteristics of traffic lanes, route length and road access, which complicates the evaluation of drivers’ skills. All these remarks have the same implication: appraising efficiency by just looking at the observed quantity of physical input is more difficult in the case of labour than in the cases of materials and soft capital, whose consumption can be more easily observed and monitored by the regulator. As a ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES -

39

result, we distinguish between observed and efficient labour forces, i.e. between the physical amount of labour that is the source of cost distortions and which is observable by the authority, and the efficient level of labour required to produce the output. Since drivers’ behaviour is the source of cost distortions, we assume that managers spend time and effort in monitoring drivers, providing them with training programmes, solving potential conflicts, etc. Both the technical inefficiency of labour and cost-reducing activity are unobservable to the regulator and to the econometrician. The empirical work involves fitting the stochastic cost functions presented in equation (11) and equation (12) to this panel data set of French urban transport networks. We assume that the production process requires four inputs. These inputs are labour L, materials M, soft capital S and capital k. To identify the cost reducing activity through effort, we need to consider that effort is nil under cost-plus regimes, while it is optimal under fixed-price schemes. Hence, under fixed-price regimes, we would estimate equation (11), while we would estimate equation (12) under cost-plus contract. In the same database where both regulatory contracts are present, we consider the following function:

(13)

{

}

′ ln C = ρ β 0 + ξ [β L ln wL + β M ln wM + β S ln wS + β Y ln Y + β k ln k + β Lθ ] +

(1 − ρ ){ln β 0 + β L ln wL + β M ln wM

+ β S ln wS + β Y ln Y + β k ln k + β Lθ } + ε C .

where ρ takes value ρ = 1 for a fixed-price contract and ρ = 0 for a cost-plus contract. Note that only labour is considered as the potential candidate for primal inefficiency and asymmetries of information. The parameters in (13) are all functions of the production frontier parameters. Thus, ln β 0 = Κ , β j = α j r , j = L, M , S , β y = 1 r , β k = − α k r , ξ = τ (τ + β L ) , and β 0′ = ln β 0 + β L ( ln τ − ln β L − ln β 0 ) (τ + β L ) .

Computations are available from the authors. Estimating the Cobb-Douglas cost function requires measures on the level of operating costs, the quantity of output and capital and the input prices. Total costs C are defined as the sum of labour, materials and soft capital costs. Output Y is measured by the number of seat-kilometres, i.e. the number of seats available in all components of rolling stock, times the total number of kilometres travelled on all routes. Capital k, which plays the role of a fixed input in our short-run cost function, includes rolling stock. Since the authority owns capital, the operators do not incur capital costs. The average wage rate w L is obtained by dividing total labour costs by the annual number of employees. Materials include fuel, spares and repairs. As the number of buses actually used mainly determines these expenditures, one derives an average price of materials w M by dividing material expenditures by the number of vehicles. Soft capital includes commercial vehicles, computer services and office supplies. These charges are induced by the activity of network management. By dividing investment charges by the number of customer trips per year, one obtains the price wS of managing single consumer travel. Summary statistics on the variables used in the analysis are given in Table 1. Table 2 presents the estimation. Table 3 lists the estimated technical inefficiency, effort levels and cost distortions over the frontier for the biggest networks included in our dataset. The other networks are available upon request. A distortion equal to 1.015, as in Toulouse for example, suggests that the observed operating costs are, on average, 1.5% above the frontier.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

40 - MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES Table 1. Descriptive statistics on the cost structure Variable

Mean

Total cost (103 FF)

Standard deviation

117500.000

137731.000

174.940

28.384

26.311

31.386

8.000

5.918

143

134

151302.680

367805.920

Labour share

0.573

0.128

Material share

0.296

0.117

Soft capital share

0.129

0.078

Wage (103 FF) 3

Material price (10 FF) 3

Soft capital price (10 FF) Capital (# vehicles) Production (103seat-kilometres)

Table 2. Estimation results Standard model

Parameters

Estimation

Standard error

βO

Asymmetric information model Estimation

Standard error

0.3068

0.150

βL

0.4285

0.041

0.4491

0.048

βS

0.1027

0.011

0.0824

0.006

βY

0.0400

0.037

0.1825

0.022

βK

0.7063

0.092

0.7010

0.048

ln τ

4.2827

0.257

ν

0.5931

0.035

μ

1.8007

0.287

0.0834

0.007

σε

0.1300

0.012

Log-likelihood

0.549

0.594

Sample size

531

531

Note:

The fifty-nine firm specific constant terms β i of the standard model are not reported here. They are available upon request.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES -

41

Table 3. Technical inefficiency, effort level and cost distortion for some networks Network

Technical inefficiency

Effort

Distortion

Aix

0.067

0.089

0.990

Besançon

0.318

0.000

1.153

Bordeaux

0.086

0.000

1.039

Caen

0.749

0.103

1.337

Cannes

0.646

0.000

1.337

Clermont

0.155

0.000

1.072

Dijon

0.120

0.000

1.055

Grenoble

0.083

0.114

0.986

Le Havre

0.266

0.000

1.127

Lille

0.180

0.126

1.024

Montpellier

0.131

0.110

1.009

Nantes

0.104

0.117

0.994

Nice

0.489

0.113

1.184

Nîmes

0.035

0.097

0.972

Rennes

0.484

0.000

1.243

Strasbourg

0.806

0.117

1.363

Toulon

0.064

0.000

1.029

Toulouse

0.158

0.124

1.015

Valence

0.111

0.000

1.051

Note: Under cost-plus regulation, the effort level is equal to zero.

Consider Figure 1, where we present our set of fifty-nine networks, ranked according to their cost distortions. Figure 1 provides, for each network, the level of the inefficiency parameter and indicates the type of contract used to regulate it. Note that three groups of networks are easily detected. The first group, with the lowest levels of cost distortion, gathers sixteen networks, all of which are managed under a fixed-price contract. The next twenty can be collected in a second group, as all of them (except four networks) are regulated through a cost-plus contract. Finally, the last twenty networks are assembled in a third group, almost equally shared between the two types of contract. Concerning the third group, we just conclude that technical inefficiency is so high that even a high-incentive scheme, such as a fixed-price contract, cannot cure the problem. These results show that, because we account explicitly for the effect of each type of contract on the cost function and that our sample covers a large spectrum of existing networks, we are able to fully recover the distribution of the efficiency parameter.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

42 - MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES Figure 1. Inefficiency and regulatory schemes

0.95 0.85 Inefficiency and Cost Distortion Level

0.75 0.65

Inefficiency

Cost distortion

0.55

Fixed-price

Cost-plus

0.45 0.35 0.25 0.15 0.05 -0.05

Note:

To each network are associated three data: the inefficiency level (white bar), the cost distortion (black bar) and the type of contracts (a black diamond refers to a fixed-price contract and an empty circle indicates a cost-plus contract).

Note that there are networks for which e > θ . Since we did not impose θ > e in the course of the estimation, we obtained cases where the effort is slightly greater than inefficiency, implying negative cost distortions. In fact, these estimated negative cost distortions are not significantly different from 0. The estimated variances of the cost distortions are available upon request.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES -

43

4. THE DEREGULATION OF EUROPEAN AIRLINES

In this chapter, we study the effect of liberalisation on costs in the European airline industry, accounting for inefficiency and cost-reducing efforts. Inefficiency and cost-reducing efforts are of particular importance when comparing industries subject to different incentives, or analysing changes in firms’ behaviour after a structural change in the rules governing the market, as is the case here. At the beginning of the 1980s, European aviation was regulated by restrictive, bilateral air service agreements between the countries concerned. Most routes were served by a duopoly, operating under perfect collusion, and the industry was characterised by a lack of incentives to improve efficiency. This situation allowed firms, in many cases subsidised by their governments, to increase costs inefficiently. Under the pressure of the US, several changes took place in the European market. First, in 1984-86, several governments started renegotiating their bilateral agreements, allowing for entry and price reductions on a few international routes. Second, the European Community introduced three packages of measures in 1987, 1990 and 1992, respectively, leading to freedom to set frequencies, capacities and prices, and free entry by European carriers onto any international European route. This process of gradual liberalisation left the industry open to international competition, introducing a significant variation in firms’ incentives. Simultaneously, European flag carriers became privatised, and explicit permission by the EU authorities became necessary in order to receive any form of public subsidy. The new competitive pressure provided the strongest incentive for carriers to reduce costs and improve efficiency. Additionally, during the second half of the 1990s, European carriers organised themselves around code-sharing agreements and international alliances, which emerged after long and complex processes of negotiation. Here, we test several scenarios of incentive pressures against each other, in order to identify the one that best fits the data. Before deregulation, European airline carriers were mainly public entities, regulated by bilateral service agreements. Subsidies would generally allow these firms to completely cover costs. It is therefore assumed that before deregulation, any operator would behave as a non-residual claimant firm and would not provide any effort at all. After deregulation, as already mentioned, the new competitive pressure, as well as the abandonment of subsidising practices, would provide the operating firms with perfect incentives for cost and inefficiency reduction. We consider then that the optimal effort provided by a deregulated firm is given by the condition (10). The cost function to be estimated is then:

{

}

′ ln C = ρ β 0 + ξ [β1 ln wL + β 2 ln wM + β 3 ln Y + β 4 ln k + β 5 NET + β 6 ASL + θ ] +

(14)

(1 − ρ ){ln β 0 + β1 ln wL + β 2 ln wM + β 3 ln Y + β 4 ln k + β 5 NET + β 6 ASL + θ } + ε C ,

where ρ takes value 1 if the firm operates in a deregulated industry, and 0 if the firm operates in a regulated industry. Moreover, ASL denotes the average stage length, and NET is the size of the network. In the course of the estimation, several combinations will be assumed, depending on the nature of the various deregulatory measures introduced in the European airlines market. ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

44 - MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES The dataset has been constructed for the period 1985-99 from raw data included in Digest of Statistics, published by the International Civil Aviation Organization (ICAO), World Air Transport Statistics, published by the International Air Transport Association (IATA), and Economic Outlook, published by the Economics and Statistics Department of the Organisation for Economic Co-operation and Development (OECD). The companies under study are the flag carriers from the largest European countries affected by the European liberalisation process, namely, Alitalia, Air France, Air Portugal, British Airways, Iberia, KLM, Lufthansa, Sabena and SAS. The variables have been constructed as follows. Production, wages, capital and average stage length correspond to total operating expenses (ICAO), seat-kilometres available, flight crew salaries and expenses and maintenance and overhaul expenses over number of employees, fleet’s total number of seats and total aircraft kilometres over total aircraft departures, respectively. Finally, the price of materials has been constructed as, respectively, the average fuel price for the carrier’s home country and the OECD (published by OECD), weighted by the company’s domestic and international operations (ICAO). This equation is estimated under alternative scenarios related to the deregulatory packages introduced by the EU and the liberal bilateral agreements signed by the UK with other countries. The following distinctions are made: 1)

Model with no effort and no inefficiency term;

2)

Firms do not make any effort to reduce inefficiency after the introduction of deregulatory measures, i.e. the effect of deregulation is not accounted for;

3)

Deregulation affects firms’ behaviour after the third EU package of measures in 1992; and

4)

Deregulation influences the behaviour of the firms affected by the introduction of liberal bilateral agreements, which are British Airways, KLM, Lufthansa and Sabena after 1985, and the remaining companies in 1993.

The comparison of scenarios (3) and (4) allows us to identify whether the liberal bilateral agreements have any effect on firms’ behaviour. Finally, given that some new competitors, like easyJet and Virgin, not included in the sample, started operating a significant number of international European routes during the period 1997-99, and that this could bias our measure of rivals’ prices, we construct scenario (3''), which is recovered from scenario (3) after having excluded the last two years of observations, namely, 1998 and 1999. The results are presented in Table 4. The variables are significant and have the expected sign. Costs are seen to increase with wages and production, while they decrease with the size of the network and the average stage length. The alternative scenarios are tested against each other, by applying a non-nested hypothesis test (see Vuong, 1989). The test shows that scenario (4) is rejected in favour of scenario (3). This suggests that liberal bilateral agreements had a limited effect on firms’ behaviour, probably because they affected only a reduced number of routes. In addition, the results for scenario (3'') are consistent with those for scenario (3).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

β6 βt

μ)

ASLi

T

ei

2.777

0.474 (0.036) 0.023 (0.026)

(2) -0.090 (0.411) 0.393 (0.037) 0.933 (0.063) -0.228 (0.088) -0.381 (0.053) 0.068 (0.032) -

(3) 0.049 (1.013) 0.318 (0.071) 1.009 (0.067) -0.360 (0.079) -0.422 (0.097) 0.271 (0.057) 3.754 (0.214) 0.270 (0.077) 0.164 (0.040) 1.996

(4) -0.292 (0.747) 0.375 (0.076) 1.028 (0.060) -0.368 (0.078) -0.345 (0.083) 0.097 (0.040) 5.088 (0.821) 0.404 (0.069) 0.098 (0.057)

(3’’) -0.276 (0.637) 0.326 (0.073) 1.051 (0.057) -0.398 (0.074) -0.405 (0.101) 0.269 (0.061) 3.969 (0.256) 0.313 (0.054) 0.121 (0.054)

45

Standard deviations in parentheses. Values for the Vuong test below –2 favour the alternative model against model (3), and above 2 favour model (3) against the alternative model. Scenarios: (1) Deregulation has no effect (ei=0), and the model does not account for one-side inefficiency (θi=0). (2) Deregulation has no effect. (3) Deregulation affects firms’ behaviour after 1992. (4) Deregulation affects firms’ behaviour after 1985 for British Airways, KLM, Lufthansa, and after 1992 for the remaining companies. (3’’) As scenario (3) but dropping the observations for the last two years (1998-1999). Note that (3’’) and (3) cannot be tested against each other since they consider two samples of different sizes. In all scenarios but (1) the model accounts for one-side inefficiency term (θi ≥ 0).

3.121

0.257 (0.015) 0.87

-

(1) 1.095 (0.724) 0.437 (0.065) 0.864 (0.065) -0.242 (0.084) -0.400 (0.088) 0.071 (0.040) -

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

Notes:

R2 Vuong test. Scenario (3) against alternative scenarios

Standard Deviation of the error term

Standard Deviation of θ

β5

NETi

ln(

β3

Yi

Coeff. β0 β1

Variable

wLi

Constant

Table 4. Cost function airlines

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES -

46 - MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES Scenarios (1) and (2) are rejected against scenario (3), which includes an inefficiency measure and assumes that deregulation affects firms’ behaviour after the introduction of the third EU package of deregulatory measures in 1992. Given that scenario (1) represents the standard approach proposed by the literature focusing on oligopolistic competition, its rejection advocates the construction of models including these components and indicates that we have to be cautious when interpreting the results derived from other models. For instance, the results for scenario (3) suggest that the European airline industry is characterised by constant returns to scale, while scenarios (1) and (2) suggest the existence of increasing returns. More in particular, rejection of scenario (2) shows the importance of accounting for the effects of deregulation on firms’ technology and inefficiency.

5. CONCLUSION

This report provides evidence indicating that a structural analysis is needed to properly identify productive inefficiency. Indeed, the global inefficiency of a production unit results from a technical inefficiency that is exogenous and an endogenous effort that depends on technological and regulatory conditions in a very specific way. From a policy perspective, the main lesson to be drawn is that compensation for inefficiency in public or private firms may be found in any system of incentives and institutional constraints.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

MEASURING INEFFICIENCIES IN TRANSPORT SYSTEMS: BETWEEN TECHNOLOGY AND INCENTIVES -

47

BIBLIOGRAPHY

Aigner, D.J., C.A.K. Lovell and P. Schmidt (1977), “Formulation and Estimation of Stochastic Frontier Production Models”, Journal of Econometrics, 6, 21-37. Baron, D. and R. Myerson (1982), “Regulating a Monopolist with Unknown Costs”, Econometrica, 50, 911-930. Gagnepain, P. and M. Ivaldi (2002), “Incentive Regulatory Policies: The Case of Public Transit in France”, Rand Journal of Economics, 33. Gagnepain, P. and P. Marin (2005), “Regulation and Incentives in European Aviation”, The Journal of Law and Economics, forthcoming. Laffont, J.J. and J. Tirole (1986), “Using Cost Information to Regulate Firms”, Journal of Political Economy, 64, 614-641. Leibenstein. H. (1966), “Allocative Efficiency Versus ‘X-Efficiency’”, American Economic Review, 56, 392-415. Loeb, M. and W. Magat (1979), “A Decentralized Method of Utility Regulation”, Journal of Law and Economics, 22, 399-404. Meeusen, W. and J. Van den Broeck (1977), “Efficiency Estimation from Cobb-Douglas Production Functions with Composed Error”, International Economic Review, 18, 435-444. Vuong, Q. (1989), “Likelihood Ratio Tests for Model Selection and Non-Nested Hypotheses”, Econometrica, 57, 307-334.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

Cover_f.fm Page 1 Wednesday, April 7, 2004 11:00 AM

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION, AND IMPLICATIONS FOR USER CHARGES

Piet RIETVELD Frank R. BRUINSMA Mark K. KOETSE Free University Amsterdam Amsterdam The Netherlands

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

49

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

51

SUMMARY

1.

INTRODUCTION ......................................................................................................................... 53

2.

PRICING IN FREIGHT TRANSPORT: BASE LINE IN THE NETHERLANDS...................... 54

3.

MARGINAL COST PRICING AND ITS PROBLEMS............................................................... 56 3.1. 3.2. 3.3. 3.4.

Uncertainty on the measurement of marginal social costs ..................................................... 56 Cost coverage problems ......................................................................................................... 56 Equity issues........................................................................................................................... 57 Other issues ............................................................................................................................ 57

4.

USER DEPENDENT INFRASTRUCTURE MAINTENANCE COSTS..................................... 57

5.

ALTERNATIVE PRICING STRUCTURES FOR FIXED MAINTENANCE COSTS............... 61 5.1. The base rate tariff .................................................................................................................. 61 5.2. Differentiation of the fixed cost charge based on social external costs ................................. 63

6.

EFFECTS OF ALTERNATIVE PRICING MEASURES ON THE TRANSPORT SECTOR..... 67

7.

CONCLUSIONS ........................................................................................................................... 71

NOTES .................................................................................................................................................. 72 ANNEX 1: EMISSION FACTORS OF EURO CLASSES IN ROAD FREIGHT TRANSPORT....... 74 ANNEX 2: IMPOSING A MINIMUM TARIFF UNDER A DIFFERENTIATED FIXED COST CHARGE ............................................................................................................................ 76 BIBLIOGRAPHY ................................................................................................................................. 77

Amsterdam, September 2005

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

53

1. INTRODUCTION

Many countries experienced a rapid increase in their infrastructure stocks after World War II. However, following the evolutionary path described by Gruebler and Nakicenovic (1989), the infrastructure growth rates gradually decreased. This implies a shift away from new infrastructure construction towards maintenance and renewal. The increasing importance of maintenance has led to research questions on ways of charging infrastructure users, and mobilisation of funds for maintenance activities. An interesting development is that technological progress helps to bring advanced and differentiated schemes of infrastructure charging closer. For example, in 2000, Switzerland introduced a system of road pricing for trucks, and several other European countries are nowadays considering systems of pricing trucks for road use (Ubbels et al., 2002). This leads to the issue of fair competition between modes and possibly also between countries. For long-distance transport, charging for road use in one country may well lead to changes in the competitive positions of other countries (Den Borger and Proost, 2004). For example, a system of road user charges introduced in the Netherlands might change the competitive position of the port of Rotterdam compared with competing ports in neighbouring countries, such as Antwerp and Hamburg. Charging for infrastructure use may lead to fiscal competition (Ubbels, 2004) that not only changes competitive positions but also has adverse effects on total welfare. This may call for international co-ordination between fiscal authorities. The other dimension of competition is between transport modes. Depending on the various origin-destination submarkets, there may be substantial competition between transport modes. This calls for an integrated approach to infrastructure charges in order to avoid equity problems. It should be noted, however, that also in the status quo there may be substantial differences in the treatment of different transport modes. In the present report, we will carry out an integrated empirical analysis of maintenance costs for three transport modes in freight transport: road, rail and inland waterways. The aim of our analysis is threefold: 1. 2. 3.

To assess the extent to which maintenance costs are user dependent for these three infrastructure types; To develop charging structures for marginal costs and for full cost recovery; To assess the consequences of these strategies for modal choice and transport demand at large.

This paper will take the Netherlands as a special case, but the methodology is in principle also applicable to other countries. The main difference between the Netherlands and its neighbouring countries is the strong reliance on inland navigation.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

54 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION The task set for this paper must be considered as rather ambitious. The size constraints of this paper do not allow us to discuss the three modes with equal degrees of detail. Therefore, we have chosen to give a detailed treatment of one of the three modes (road) in the text, and give very concise reports on the other modes. Details on the underlying studies can be found in CE/VU (2004). Only in the last part of the paper, where we give an integrated treatment of the effects of the infrastructure charges, will the three infrastructure types be treated at equal levels of detail.

2. PRICING IN FREIGHT TRANSPORT: BASE LINE IN THE NETHERLANDS

In our study we focus on infrastructure that is under the supervision of the national government. For inland navigation and rail this is essentially the whole freight transport network, but for roads this means that the local and regional roads are excluded. Since freight transport, compared with passenger transport, mainly takes place over longer distances and the spatial density of national highways is high, this means that here also we cover a substantial share of total transport1. Some key indicators are given in Table 1. Table 1. Comparison of transport infrastructures in the Netherlands (2002) National highways Transport volume (bill. tonne-kms) Transport volume (bill. passenger-kms) Number of firms (freight) Employment (freight) Length of main infrastructure (kms) Maintenance costs (mill. euros)

Inland waterways

Rail

40.6

34.7

4.3

70

0.1

15

10 100

3 600

4

125 000

14 000

3 000

3 000

1 400

2 800

590

300

810

Source: (IBO, 2005).

Table 1 shows that the share of trucks in tonne-kilometres is highest, but also that inland navigation has a very high share. This high share is owing to the splendid natural waterways that penetrate substantial parts of the country. They allow the use of inland waterways for bulk products in many origin-destination combinations, in particular for long-distance transport. As a result, the share of rail in total transport is rather small compared to other countries. When we compare these shares with passenger transport, it appears that inland waterways are rarely used. Here rail has a higher share than in freight transport. It is also interesting to compare the market structures for the three transport modes. Regulatory reform in the rail sector has led to a situation where competition is possible, but the real number of operators on the Dutch rail system is still only four. This market is still dominated by Railion, the former freight branch of the former Dutch national railways, which was sold to the German DB. On the other hand, the number of suppliers in inland navigation and road transport is still very large and strongly dominated by small firms. The low barriers to entry have led to a high degree ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

55

of competition and low profit rates, obviously to the benefit of the users. The figures for employment underline this result. The average number of workers per firm in the rail freight sector is about 750, whereas it is only twelve in road transport and four in inland navigation. The length of infrastructure for the three modes is remarkably similar. The inland waterways extend over about 1 400 kilometres. They vary substantially in the size of ships allowed. Some parts are mainly used for smaller barges, but some of the main waterways between Rotterdam and the German inland port of Duisburg allow large barges, up to a capacity of 2 000 tonnes, as well as large inland container ships with 300 or more TEU. This indeed makes inland waterways a formidable competitor of rail transport. The lengths of the railway system and the national highway system are both close to 3 000 kilometres. With a surface area of about 40 000 kilometres2 this may be considered as a high-density network. Of course, in addition to the 3 000 kilometres of national roads, there are still about 100 000 kilometres of local and regional roads; but these are left out of consideration in this study, where the focus is on freight transport with its longer distance orientation. The last row in Table 1 addresses infrastructure maintenance costs. This appears to range between €€ 300 million per year for inland waterways to €€ 600 million for national roads to €€ 800 million for rail. This result underlines that when maintenance costs are compared with transport flows2, the national highways are performing most favourably, whereas rail has much higher maintenance costs. Waterways assume an intermediate position. After this sketch of the competitive position of the three transport infrastructures, we now address the situation of the three transport modes with respect to the current infrastructure-related charges. Road transport currently does not have a real system of infrastructure-related charges for infrastructure use. However, several other taxes are paid, in particular the diesel tax, plus a tax for truck ownership, the latter being much less important than the former. In addition, there is the Eurovignette, whereby trucks are charged for the use of the national highways in a selection of European countries: Sweden, Denmark, Germany and the Benelux (Belgium, the Netherlands and Luxembourg). The vignette is not dependent on the degree of use, and its level is moderate. This situation will probably change, as since 1-1-2005 Germany introduced a system with an infrastructure charge that depends on the distance covered by trucks, and which is comparable to the Swiss charge for road use. This step will induce some of the neighbouring countries to follow suit. For inland waterways the situation is rather different. There are nowadays no charges for the use of infrastructure. The only exception is that inland navigation may pay for the use of ports, but these charges, if they are paid, tend to be low. Since these are not owned by the national government, there is no direct money flow from the sector to the government. Note that there is also no tax on fuel use in this sector. This favourable situation for the sector has strong historical roots. The Mannheim Convention of almost 150 years ago (1868) declared that there should be no user fees imposed in the barge sector on the Rhine and its tributaries. It essentially extrapolated the principle of free navigation on the sea to the use of rivers. The countries through which the Rhine flows all agreed to this convention. Later on, in the 20th century, these countries also agreed that there should be no excise taxes on fuel for inland navigation. In the rail sector, the situation is again rather different. Here, the European rules have led to infrastructure charges that do depend on the intensity of use. Countries have gradually started to introduce these charges, but at different speeds. The Dutch charges are still low, but steadily increasing. The 2004 official level of the charge per train-kilometre is €€ 1.08, which is lower than in the neighbouring country of Germany. Besides, various reductions have been allowed thus far that ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

56 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION lead to actual levels of €€ 0.86 per train-kilometre for passenger trains and €€ 0.60 for freight trains. These reductions are expected to be removed in the future.

3. MARGINAL COST PRICING AND ITS PROBLEMS

An important benchmark in the discussion on pricing for infrastructure use is the concept of social marginal cost pricing. This pricing principle means that users pay the additional costs they impose on infrastructure providers (when the infrastructure is owned by the public sector), as well as the external costs of transport. In the present report, we will focus in particular on the first aspect. The social marginal cost pricing principle has been devised as the cornerstone for pricing in the EU, as formulated in, among others, policy documents on fair and efficient pricing (EC, 1995). It is well known that marginal cost pricing presents a number of problems which obviate a simple application. A recent survey of these problems has been undertaken by Rothengatter (2003), with a response by Nash (2003). Some important issues in the debate will be shortly discussed below. 3.1. Uncertainty on the measurement of marginal social costs During the last decade, progress has been made in the valuation of various cost components, such as accident costs, various types of environmental costs and the costs of infrastructure use. Nevertheless, substantial problems remain. The nature of the estimation problems remains varied: in some cases it concerns the valuation aspect, in other cases it is the estimation of marginal vs. average costs, or long-run vs. short-run costs. It should be admitted that substantial uncertainties still do exist, but the progress made in the meantime has been considerable, and there are promises of further advancement. 3.2. Cost coverage problems Transport infrastructure is characterised by increasing returns to scale. Hence, marginal costs are lower than average costs, and this implies that marginal cost pricing will lead to incomplete cost coverage. Policies to improve cost coverage imply that pricing starts to deviate from the simple marginal cost pricing rule. Well-known examples are the use of mark-ups via Ramsey pricing, or two-part pricing with a fixed and a variable component. An important issue is to what extent the potential inefficiencies – due to departures from marginal cost pricing in transport in order to achieve cost coverage – are larger than those imposed on the economy by general taxation. A related issue is at what level the cost coverage is to be imposed: at the level of each transport mode separately, or across all transport modes? Inefficiencies can be reduced by using the broadest possible formulation, but this leads to issues of cross-subsidisation and equity.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

57

3.3. Equity issues Fairness of pricing is an issue that tends to attract decisionmakers’ attention more than efficiency. However, equity is characterised by a good number of, sometimes conflicting, interpretations (Rietveld, 2003). Equity can be applied to the individuals affected, but also to collectives of transport mode users, and this may not lead to identical results. Examples of statements that express equity concerns are: “the polluter pays”, “each transport mode should pay its way” and the principle that there should be a “level playing field between transport modes”. A closer examination of the latter term is carried out by CPB (2004). It appears that, basically, two interpretations of the level playing field exist: first, the rule-based interpretation that all firms are treated the same in equal circumstances with regard to legislation, taxes, subsidies, etc.; second, the outcome-based interpretation, that all firms have the same expected profit, possibly leading to subsidies to disadvantaged firms. It is clear that the former interpretation remains closer to the efficiency principle than the latter. 3.4. Other issues There are some other concerns on marginal cost pricing that deserve attention. For example, when there are evident distortions in other markets, the application of marginal social costs in transport may have counterproductive effects. Further, marginal cost pricing may lead to very complex pricing structures that are too demanding for the infrastructure users, so that the intended effect is not realised. In addition, the implementation costs of strongly differentiated charges based on marginal cost pricing may be high compared with simple charges. Therefore, a simplistic use of marginal cost pricing should be avoided. Of particular relevance is the notion of second-best pricing. Given the various distortions, implementation costs and constraints on pricing, the first-best principle would lead to inefficiencies. Therefore, these aspects should be explicitly taken into account in order to derive adjusted decision rules that are nevertheless still based on marginal cost principles (see, for example, Verhoef, 1996).

4. USER DEPENDENT INFRASTRUCTURE MAINTENANCE COSTS

In this and the next chapter, we analyse the structure and magnitude of highway network maintenance costs in the Netherlands, and pay special attention to the ways in which these costs are or can be related to traffic. As such, we make a distinction between user-dependent (variable) and user-independent (fixed) maintenance costs. The former costs are imputed to road users according to the way in which they depend on traffic characteristics, such as vehicle-kilometres and vehicle weight. Usually, maintenance costs are calculated by analysing government expenditures. A problem here is that these costs may be and often are biased, since government maintenance expenditures may have been suboptimal. A method applied by DWW (2002) takes care of this particular problem by calculating maintenance costs under an optimal maintenance level, where optimal is defined as the level of maintenance that is needed to preserve the current state of the infrastructure, physically as well as functionally. Using the costs estimated by DWW, highway maintenance costs in the Netherlands were about €€ 634 million in 2002.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

58 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION Regarding the DWW figures, the first distinction is made between fixed (traffic independent) and variable (traffic dependent) costs, while the second distinction holds solely for variable costs and is based on the specific type of variability. The latter distinction was possible because the DWW study provides bottom-up figures; for each of these figures we can determine in what way and to what extent they were dependent on traffic. The costs are summarised in Table 2. Observe that a large share of the costs are user-independent, but that user-dependent maintenance costs are substantial and amount to approximately 40% of total costs3. Table 2. Fixed and variable highway maintenance costs in the Netherlands in 2002 Cost category

Costs (million €€ )

Fixed (traffic independent)

372.5

Variable (traffic dependent):

261.3

ƒ

Dependent on vehicle-kilometres and vehicle weight

205.0

ƒ

Dependent on vehicle-kilometres

36.4

ƒ

Dependent on vehicle-kilometres and noise production

11.0

ƒ

Dependent on number and severity of traffic accidents

8.9

Total

633.8

Source: DWW (2002) (see also CE/VU, 2004).

Observe that, regarding variable costs, the bulk is due to road wear and tear, and is thus dependent on vehicle-kilometres as well as vehicle weight. Costs dependent solely on vehicle-kms are costs from traffic regulation measures, boarding, cleaning of roads and disposal of garbage, etc. Variable costs, dependent on the noise production of a vehicle4, are the costs of maintaining noise prevention measures alongside roads. These costs are traffic dependent, because if traffic volume or the number of relatively noisy vehicles increases, so will the number and size of noise prevention measures (and thus also their maintenance). Finally, a small part of variable costs are dependent on the number and severity of traffic accidents. Unfortunately, no data are available on this variable, so we used the number of fatal traffic accident victims as a proxy, since this variable can be expected to be correlated with both traffic accident numbers and severity5. The methodology of the imputation of the different types of variable costs is fairly straightforward. Costs that are imputed on the basis of vehicle-kilometres only will produce identical costs per vehicle-km for each road user. Costs that are imputed on the basis of vehicle weight factors, e.g. passenger car equivalents (PCEs), and on the basis of noise weight factors, display a structure that is identical to the structure of vehicle weight factors and noise weight factors, respectively. Presentation and discussion of the data that are needed for imputation is beyond the scope of this paper; they are given in CE/VU (2004). Suffice it to say that the results are rather dependent on axle load factors. As is clear from Table 3, most user-dependent costs are imputed on the basis of axle load factors, which are generally computed using the AASHO 4th Power Rule (see AASHO, 1962). However, due to ongoing changes in vehicle characteristics, an important question arises as to whether using the 4th Power Rule for the ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

59

imputation of infrastructure maintenance costs is still realistic (see Johnsson, 2004). The potential impact of using the wrong power can be considerable. As shown in VU (2002), using an identical approach as in this report, the cost structure changes substantially when, for instance, a 2nd power instead of a 4th power computation is used. In contrast, using simulation in a general equilibrium framework, Johnsson (2004) finds that erroneously using the 4th power in designing a kilometre charge, when the true power is equal to 2, 3 and 5, generally has a small effect on road wear and government revenue. The generalised costs are higher when the true power is 1, basically because such a power relation implies a weight-distance charge rather than an axle-weight-distance charge. Table 3. Variable highway maintenance costs per road user (in eurocents per vehicle-kilometre) in 2002 Vehicle category A Freight transport Truck solo < 12 tonne Truck solo > 12 tonne Truck combination > 12 tonne Truck trailer > 12 tonne Passenger transport Car Touring car Delivery truck

Cost categories dependent on: B C

D

Total

0.17 2.22 6.52 5.39

0.08 0.08 0.08 0.08

0.05 0.07 0.09 0.09

0.13 0.03 0.02 0.02

0.43 2.40 6.71 5.57

0.00 4.81 0.01

0.08 0.08 0.08

0.02 0.05 0.02

0.02 0.06 0.03

0.11 5.01 0.13

Source: CE/VU (2004). A = Number of vehicle-kilometres and vehicle weight B = Number of vehicle-kilometres C = Noise production D = Number and severity of traffic accidents

Keeping in mind the uncertainty regarding the power rule and its potential impact, we stick with the 4th Power Rule in computing axle load factors and impute the relevant costs accordingly. Variable costs per vehicle-kilometre for different road users are presented in Table 3. Observe that total variable costs per vehicle-kilometre for light trucks are low, while costs for heavy trucks and passenger transport per bus are high. This is as could be expected, since costs of road wear and tear are relatively large and axle loads for light vehicles are relatively low (or axle loads for heavy vehicles are relatively high). For railway and inland navigation infrastructure, similar cost figures have been computed; see Tables 4 and 5, respectively. Due to lack of space, we do not discuss the computation methods used: for these methods the reader should consult CE/VU (2004). Observe that costs per vehicle-kilometre are much higher for these two modalities. However, note that the loads carried per truck are much smaller.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

60 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION Table 4. Total costs (in million euros) and costs per kilometre (in eurocents per kilometre) for user-dependent maintenance of railway infrastructure in 2002* Variable costs imputed on the basis of: Gross Train-kilometres kWh tonne-kilometres Freight transport Train electric (eurocents per kilometre) Train diesel (eurocents per kilometre) Total costs (mill. euros) Passenger transport Train electric (eurocents per kilometre) Train diesel (eurocents per kilometre) Total costs (mill. euros)

0.38

46

0.03

0.38

46

-

24.0

4.6

1.3

0.38

46

0.03

0.38

46

-

106.4

53.6

31.6

Source: CE/VU (2004).

*

Costs imputed on a certain measure are presented in that same measure, e.g. costs imputed on the basis of train-kilometres are presented as costs per train-kilometre. Table 5. Total costs (in million euros) and costs per ship-kilometre (in eurocents per kilometre) for user-dependent maintenance of inland navigation infrastructure in 2002

Load capacity (tonnes) <250

Total costs (mill. euros) 0.2

Costs per ship-kilometre (in eurocents) 53

250-400

1.4

53

400-650

3.8

53

650-1000

6.2

53

1000-1500

7.0

53

1500-2000

3.3

53

2000-3000

4.8

53

>3000

2.3

53

10.0

27

Recreational shipping Source: CE/VU (2004).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

61

5. ALTERNATIVE PRICING STRUCTURES FOR FIXED MAINTENANCE COSTS

Fixed costs of infrastructure maintenance are costs that are not related to use of the infrastructure by traffic; they consist of costs of operation and exploitation, of research and consultancy, and trafficunrelated costs of maintaining the infrastructure. For highways they amounted to €€ 372.5 million in 2002 (see Table 2). One might claim that these costs should be recovered through general government receipts, since otherwise the price of using infrastructure would not be equal to marginal costs. However, payment via general government receipts would lead to distortions elsewhere in the economy, and therefore there are good reasons to also consider the case where road users pay the full costs of infrastructure maintenance. Since there is no relation between fixed costs and use of the infrastructure, the method of imputation is fairly arbitrary. In the remainder of this chapter we calculate a base rate tariff; a tariff which simply differentiates between the road users given in the previous chapter. Subsequently, we divert from the base rate tariff by allowing further differentiation within freight transport. 5.1. The base rate tariff There are several possibilities for calculating a base rate tariff. A first possibility is treating every vehicle category alike, equating the base rate tariff for every road user. As we will see further on, this method is relatively beneficial for heavy vehicles. The advantage of this method is that it is transparent and in line with EU propositions. A disadvantage is that the tariff could be too low for some vehicle categories to allow a further differentiation, which may be true especially for freight transport, as will be explained below. A second option is to impute fixed costs in such a way that it least disturbs the transport market. This basically implies that lower costs are imputed to the segments of the transport market that are most price sensitive (as often happens with air transport). However, a disadvantage of this method is that no reliable estimates are available on the price elasticities of different market segments, in both passenger and freight transport. A third possibility is to impute the fixed costs on the basis of the relative amount of external costs caused by a certain vehicle. In this case we calculate the average social external costs per kilometre and per vehicle category, calculate a percentage per vehicle category on the basis of these figures and impute the fixed costs according to these percentages. An advantage of this method is that socially optimal vehicle choice is already stimulated in the base rate tariff. Moreover, since this tariff is based on external costs, there is always enough room for further differentiation on the basis of external costs. A disadvantage is that this method will lead to discussions on the magnitude of external costs per vehicle category. Therefore, the option chosen in this paper is imputation of fixed costs according to the size of the vehicle. Advantages are that the method is easily explained, that it appeals to a certain logic (as discussed below) and that costs are already differentiated to some extent. Obviously, vehicle size can be measured in several ways, such as maximum weight, actual average weight, passenger car ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

62 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION equivalents, and capacity factors during peak hours. The latter can be argued to be in some way related to the production of new infrastructure, since as roads get more congested new infrastructure will be built. European Commission propositions adhere to imputation on the basis of kilometre-equivalence factors (see EU, 1999). In these propositions, the passenger car and the delivery truck have equivalence factors equal to one, while other, heavier vehicles have an equivalence factor equal to three. However, since these factors are rather arbitrary, we decided to calculate the base rate tariff on the basis of capacity factors during peak hours, especially since these factors are in some way related to the production of new infrastructure. We collected these data from a study by HCG (1996). The second column in Table 6 presents capacity factors during peak hours. Table 6. Capacity factors during peak hours and the base rate tariff (in eurocents per kilometre) of fixed highway operation and maintenance costs Capacity factors

Base rate tariff (eurocents per kilometre)

Truck solo < 12 tonne

1.85

1.30

Truck solo > 12 tonne

2.20

1.54

Truck combination > 12 tonne

2.20

1.54

Truck trailer > 12 tonne

2.20

1.54

Car

1.00

0.70

Touring car

1.85

1.30

Delivery truck

1.00

0.70

Vehicle category Freight transport

Passenger transport

Source: HCG (1996) and CE/VU (2004).

The HCG (1996) study tests the distance between two subsequent cars (capacity factors), for both passenger and freight transport, using traffic data from the A2 highway between Utrecht and Amsterdam in the Netherlands. It presents several results based on different measurement units and situations. We followed the recommendations in the report, and chose to use the results based on the space measured in front of the vehicle. The traffic situation chosen was the one in which the distance between vehicles was limited, since this most resembles the situation during peak hours. Since the study only mentions capacity factors for the passenger car and freight transport, we had to make some assumptions regarding the other vehicle categories. For the capacity factor for touring cars we took the capacity factor of light trucks (< 12 tonne), and the PCE of the delivery truck was set equal to the capacity factor of the car (see CE, 1999, and VU, 2002). Base rate tariffs per road user are presented in the third column of Table 6. The cost structure presented in this column is identical to the structure in the capacity factors. For comparison, the results for inland navigation infrastructure and rail infrastructure are presented in Tables 7 and 8, respectively. For details on these modalities, we again refer to CE/VU (2004).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

63

Table 7. Total fixed costs (in million euros) and the base rate tariff (in eurocents per kilometre) based on fixed operation and maintenance costs of inland navigation infrastructure Load capacity (in tonnes) <250

Total fixed costs (million euros) 0.5

Base rate tariff (eurocents per kilometre) 1.42

250-400

5.0

1.95

400-650

19.0

2.66

650-1000

43.7

3.73

1000-1500

55.7

4.18

1500-2000

32.5

5.20

2000-3000

50.9

5.58

>3000

34.9

8.11

Recreational shipping

18.7

0.51

Source: CE/VU (2004).

Table 8. Total fixed costs (in million euros) and the base rate tariff (in eurocents per kilometre) of fixed operation and maintenance costs of railways Total fixed costs (million euros) Fixed costs O&M

490.7

Freight transport (eurocents per km) Electric Diesel

Passenger transport (eurocents per km) Electric Diesel

3.82

3.82

3.82

3.82

Fixed costs stations

32.4

-

-

0.27

0.27

Fixed costs energy

32.6

0.31

-

0.31

-

555.7

4.1

3.82

4.40

4.10

Total Source: CE/VU (2004).

5.2. Differentiation of the fixed cost charge based on social external costs The base rate tariff calculated in the previous section exactly covers total fixed costs of infrastructure maintenance under a constant transport volume. However, next to the recovery of total costs, an important goal of transport pricing is to achieve economic efficiency. In the situation where social external costs of transport are not recovered through transport pricing, economic welfare can be enhanced by differentiating fixed costs of infrastructure maintenance on the basis of these external costs. The argument is that through such a differentiation in tariff structure, social external costs are included in transport choices, thereby stimulating transport efficiency and reducing the negative social externalities of transport. The magnitude of external costs depends on several parameters, such as vehicle characteristics (e.g. pollution characteristics associated with Euro classes), load characteristics (transport of ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

64 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION dangerous goods), location (e.g. for some polluting gases it matters whether they are emitted in a city or in nature), time (noise production is more harmful at night than during daytime), driver (driving faster implies more traffic accidents and higher external costs of pollution). We will consider differentiation on the basis of vehicle characteristics because these are identifiable. Differentiation on the basis of other characteristics would be either impossible or require data that are currently unavailable. Furthermore, the choices we make regarding the external effects used for differentiation depend on three criteria. First, our main attention goes to the external effect cost that is most dominant in total social external costs. Second, we take account of the extent to which the external effect actually varies within one vehicle category. For example, although external costs of traffic noise are substantial, differences in noise production between different truck categories are fairly limited (the same holds for external costs of congestion). Finally, differences in external effects within a specific vehicle category must be related to vehicle characteristics. For external costs related to victims of traffic accidents, this is only the case to a very limited extent. Given these considerations, we base the differentiation on external costs of environmental pollution, both because this is one of the most dominant sources of social external costs in road transport and because these costs are related directly to Euro classes. Since for each Euro class there are strict regulations regarding polluting emissions, and data are available on the shadow prices of these emissions, we can found our differentiation of the base rate tariff on Euro classes. At the time of research, truck motors are divided into six Euro classes, i.e. Pre Euro to Euro 5. Each Euro class has certain regulated emission characteristics, with regulations getting stricter with each Euro class (Euro 5 is the cleanest vehicle). We can therefore base our differentiation on an existing norm structure6. The emission factors we use for differentiation are actually composite emission factors, based on emission factors and shadow prices of three air pollutants, i.e. VOC, NOx and PM10. The composite factors are presented in Table 9 (see Annex 1 for details on calculating the composite emission factors). Table 9. Composite emission factors for freight transport per truck weight category per Euro class in 2002 Euro class

Truck solo < 12 tonne 1.01

Truck solo > 12 tonne 1.85

Truck combi > 12 tonne 1.81

Truck trailer

Euro 0

0.94

1.57

1.80

1.70

Euro 1

0.56

1.08

1.12

1.20

Euro 2

0.52

1.02

1.04

1.11

Euro 3

0.38

0.77

0.78

0.86

Euro 4

0.20

0.40

0.40

0.44

Euro 5

0.12

0.24

0.24

0.27

Pre Euro

1.81

Source: CBS and TNO (2003).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

65

The ratio in fixed maintenance costs per kilometre for the different Euro classes within each truck weight category should be identical to the ratio in composite emission factors (see Table 9). Furthermore, total fixed costs should be exactly covered under the resulting tariff structure, ceteris paribus. The resulting differentiation of charges is summarised in Table 10. Table 10. Charge per kilometre for different Euro classes Euro class Pre Euro

Market share (vehicle-kilometre) KM1

Emission factor

Charge

EF1

X * EF1/EF7 * S

Euro 0

KM2

EF2

X * EF2/EF7 * S

Euro 1

KM3

EF3

X * EF3/EF7 * S

Euro 2

KM4

EF4

X * EF4/EF7 * S

Euro 3

KM5

EF5

X * EF5/EF7 * S

Euro 4

KM6

EF6

X * EF6/EF7 * S

Euro 5

KM7

EF7

X

In Table 10, X represents the charge per kilometre for the least polluting vehicle, i.e. the Euro 5 vehicle, while S is a scaling factor that we will pay more attention to below. The table shows that the structure in the charge is indeed identical to the structure in emission factors as long as S=1, and that the charges per kilometre for each Euro class other than the Euro 5 category are directly related to X. Therefore, we have to calculate X. Since total fixed costs have to be exactly recovered, we get: TC = X * {KM7 + EF6/EF7*S*KM6 + EF5/EF7*S*KM5 + EF4/EF7*S*KM4 + EF3/EF7*S*KM3 + EF2/EF7*S*KM2 + EF1/EF7*S*KM1},

(1)

with TC representing total costs. When S=1, it follows from (1) that: X = TC / {KM7 + EF6/EF7*KM6 + EF5/EF7*KM5 + EF4/EF7*KM4 + EF3/EF7*KM3 + EF2/EF7*KM2 + EF1/EF7*KM1}

(2)

It is evident from (2) that X = TC / (KM7 + KM6 + KM5 + KM4 + KM3 + KM2 + KM1) for Euro 5 vehicles if all vehicle-kilometres are being produced by Euro 5 vehicles. In this case the charge exactly equals the base rate tariff calculated in section 5.1. Furthermore, the use of parameter S is related to possible governmental restrictions to the minimum tariff. Its use and calculation under such restrictions is explained in Annex 2. Below we set S equal to one, i.e. no restrictions are in place. Market shares for the relevant Euro classes for different trucks are presented in Table 11. The distribution of market shares of the different truck categories does not display large variations in general, with the market shares of the truck trailer as a notable exception. As the table shows, vehicles within this truck category are generally cleaner.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

66 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION Table 11. Market shares of six different Euro classes for several truck weight categories in 2002 Euro class

Pre Euro

Truck solo < 12 tonne 12%

Market shares Truck solo > Truck combi > 12 tonne 12 tonne 10% 10%

Truck trailer 3%

Euro 0

12%

8%

8%

3%

Euro 1

12%

12%

12%

10%

Euro 2

48%

52%

52%

63%

Euro 3

16%

17%

17%

20%

Euro 4

0%

0%

0%

0%

Euro 5

0%

0%

0%

0%

100%

100%

100%

100%

Total

Source: RIVM (2004) and own calculations.

In Table 12, the costs per Euro class per truck category are presented without restrictions on X, such that the cost differences exactly reflect the differences between Euro classes presented in Table 9. The costs per kilometre do not differ much between the four truck weight categories because the differences in market shares are small. This holds even for the truck trailer, for which differences in market shares are relatively large. Changes in market shares of Euro classes in freight transport apparently lead to only marginal changes in costs per kilometre for the distinguished Euro classes. Table 12. Fixed highway maintenance cost charge of freight transport in the Netherlands in 2002 after differentiation of the base rate tariff (in eurocents per kilometre) Euro class

Truck solo < 12 tonne 2.13

Truck solo > 12 tonne 2.57

Truck combi > 12 tonne 2.45

Truck trailer

Euro 0

2.00

2.19

2.43

2.38

Euro 1

1.18

1.50

1.52

1.67

Euro 2

1.11

1.42

1.40

1.55

Euro 3

0.82

1.07

1.06

1.20

Euro 4

0.43

0.55

0.54

0.62

Euro 5

0.25

0.33

0.33

0.37

Pre Euro

2.53

Furthermore, as more kilometres are being driven in cleaner vehicles, costs per kilometre go up at first for each Euro class. The most extreme situation occurs when all vehicle-kilometres are driven by the Euro 5 class, in which case costs per kilometre for this class are 1.96 eurocents, i.e. the base rate tariff. The result can be explained by the fact that as soon as less kilometres are being driven in ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

67

relatively polluting vehicles, tariffs for other categories go up, since their original kilometre prices were smaller than those of the more polluting vehicles. However, as soon as this happens, total social external costs of air pollution will go down (ceteris paribus). This brings us to the fact that such a tariff system is obviously dynamic on multiple levels. As mentioned, next to an autonomous development, the introduction of a tariff has an impact on market shares of Euro classes as well. Moreover, as can be seen from (1), relative emission factors of Euro classes are important and can change over time, especially when a new Euro class is introduced. Therefore, periodical updates have to be made of both market shares and composite emission factors in order to make the system viable over time. For inland navigation, the most important external costs are also costs of air pollution. However, differentiation of the base rate tariff according to distinctions between motor characteristics, i.e. similar to the differentiation in road freight transport, produces only tiny variations in the cost structure (see CE/VU, 2004, p. 59). We therefore do not present the differentiated costs. Something similar holds for differentiating the base rate tariff of railways. The most important external social costs for trains are those of noise production, but no accepted noise categories for trains exist. Differentiation is therefore not attempted for railway infrastructure (see CE/VU, 2004, p. 62).

6. EFFECTS OF ALTERNATIVE PRICING MEASURES ON THE TRANSPORT SECTOR

The introduction of a new system of infrastructure charges across transport modes will lead to various adjustments within transport companies, shifts between transport modes, and adjustments by customers. The adjustment possibilities within transport companies are described in Figure 1 (see also Van den Bossche et al., 2005). The figure indicates that transport companies can respond to changes in infrastructure charges by adjusting load factors and empty kilometres driven. In addition, the composition of their fleets may change by using, for example, larger vehicles. Furthermore, infrastructure charges will be transferred to the customers. The extent to which this happens will depend on the degree of competition in the transport markets. In highly competitive markets, such as in road transport and inland navigation, it is safe to assume that the extra costs will be imposed on the customers, but in the rail sector this transfer might be incomplete. Given this change in transport fares, a market response may take place, leading to changes in modal split as well as in total transport. In terms of Figure 1, this means that the customer of the transport company responds by adjusting his transport flows in terms of transport distances by, for example, changing the origins of inputs or the spatial structure of distribution systems. Another possible response would be the adjustment of routes. This is of special relevance when national routes would be affected by the charging system, whereas other routes would not. Or – in the case of long-distance transport – when countries would differ in their use of infrastructure charging, making a certain country cheaper as a transit country compared with others.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

68 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION Figure 1. Relationship between transport demand in terms of tonne-kilometres and supply in terms of vehicle-kilometres Tonnes

* Average transport distance

= Tonne-kilometres

/ Average load capacity per vehicle

Load factor

=

*

Average load transported per loaded vehicle

= Empty vehiclekilometres

+

Loaded vehiclekilometres

=

Vehiclekilometres

It is clear that the present generation of operational freight transport models in Europe does not allow a complete simulation of all network effects mentioned here. Van den Bossche et al. (2005) use a freight transport demand model based on aggregate values of direct and cross price elasticities taken from the literature (Oum et al., 1992, Goodwin, 1992 and PRC, 2002) for four sub-markets: general cargo, dry bulk, liquid bulk and containers. In the policy analysis, two alternatives are considered. In the first one, only the user-dependent maintenance costs are charged to the infrastructure users, as computed in Chapter 4. This variant will be coined MC (after marginal costs). In the second, the full costs are charged in the way indicated in Chapter 5. This variant will be termed AC (average costs). Since the marginal costs are considerably smaller than the average costs, the effects in the MC case tend to be much smaller. An additional policy dimension is that two methods of introduction are considered: one where the infrastructure charges only apply in the Netherlands, and one where they are introduced all over Europe. The most important findings are given in Table 13. The bottom line in the table shows that the effects on total transport flows through the Netherlands are small. For example, the decrease in tonnes transported in the Netherlands due to the introduction of a marginal cost based infrastructure charge, is 0.4%. In the case of an average cost-based charge, the decrease is 1.1%. These figures for total transport volumes are not very different when the infrastructure charge is introduced EU-wide. There are two countervailing forces here. On the one hand, the international introduction of an infrastructure charge has a negative effect on transit flows going through the Netherlands (in particular the Port of Rotterdam), since the total transport costs – including the international part – are increased more strongly. The other effect is that the one-sided introduction of an infrastructure charge in the ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

69

Netherlands will divert some of the transport flows via ports in neighbouring countries, such as Antwerp in Belgium and Hamburg, Germany. The European-wide introduction of infrastructure charges would therefore have a positive effect on freight volumes transported via the Netherlands compared to the situation where only the Netherlands would introduce such a charge. A closer look at Table 13 reveals that the modal shifts due to the infrastructure charges are more substantial. For example, in inland navigation the maintenance costs are hardly user-dependent and hence barges would improve their competitive position compared with road and rail in the case of MC pricing. For rail, on the other hand, the user-dependent costs are relatively high and the introduction of user charges for this transport mode hence leads to a relatively strong decrease in demand. The positive effects of infrastructure charges on road transport in the AC variants can be explained by the relatively low increase in road-related costs compared with its competitors. Table 13. Changes in freight volumes transported in the Netherlands due to the introduction of an infrastructure user charge (in mill. tonnes and in percentage changes between brackets); reference year 2002 Total 2002 before introduction

Change after introduction of infrastructure charge MC -NL

AC-NL

MC-EU

AC-EU

Road

572.4

-5.0 (-0.9%)

4.0 (0.7%)

-5.7 (-1.0%)

10.7 (1.9%)

Rail

28.1

-1.1 (-3.9%)

-2.8 (-10.0%)

-2.3 (-8.2%)

-6.2 (-22.1%)

Inland waterways

314.0

2.5 (0.8%)

-11.6 (-3.7%)

4.0 (1.3%)

-17.1 (-5.4%)

Total

914.5

-3.6 (-0.4%)

-10.4 (-1.1%)

-4.0 (-0.4%)

-12.6 (-1.4%)

Source: Van den Bossche et al. (2005).

After having considered the total transport volume in terms of tonnes transported in the Netherlands, we now turn to the total transport activity in terms of vehicle-kilometres (Table 14). The effects on vehicle-kilometres are stronger than the effects in terms of tonnes transported. This is due to a combination of shifts indicated in Figure 1: an increase in load factors, decrease in empty vehicle movements and a small increase in the average size per vehicle. Again, the effects of the introduction of user charges on transport flows are small for road transport, whereas they are substantial for rail. Effects on inland water transport are intermediate. Apart from the effects on transport flows, the introduction of infrastructure charges can of course be evaluated from a broader perspective. First of all, the question is to what extent the introduction leads to a decrease in total costs of maintenance. The answer is that the effects are small. The reason is simply that, in the end, the shifts in tonne-kilometres are limited, as is the share of user-dependent costs in total maintenance costs (see Chapter 4).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

70 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION Table 14. Changes in vehicle kilometres on the Netherlands network due to the introduction of an infrastructure user charge (in million vehicle kilometres and in percentage changes between brackets); reference year 2002 Reference year 2002

MC-NL

AC-NL

MC-EU

AC-EU

Road

4 512

4 459 (-1.2%)

4 507 (-0.1%)

4 450 (-1.4%)

4 550 (+0.8%)

Rail

10.2

9.7 (-5.2%)

8.6 (-15.6%)

9.2 (-10.2%)

7.0 (-31.5%)

56

56 (-0.0%)

50 (-10.7%)

56 (-0.0%)

47 (-16.1%)

Inland waterways

Source: Van den Bossche et al. (2005).

Another perspective concerns the government receipts. These increase substantially up to about 500 million euros per year in the case of the AC pricing. Given the international nature of transport, a charge on infrastructure use is, of course, attractive as a tax source in a small country like the Netherlands, since foreign consumers finally pay a substantial part of the charges. A final remark has to be made on the environment. A remarkable result of our analysis is that it implies a modal shift away from rail to inland water and road. This is at variance with the current policy trends in Europe that aim at increasing the share of rail transport and decreasing the share of road transport. There are two main arguments for the current policy: congestion on the roads calls for a shift towards less congested modes where possible; and the environmental performance of the transport system calls for a shift away from roads. We did not explicitly study the effects on road congestion of the present policy alternatives, but they are probably small. First, the effect on modal shift toward the road is minimal. In addition, the contribution of freight towards total road congestion seems to be overestimated (Gerondeau, 1997). Another issue is the environmental performance of the road sector compared with rail and water. Although there is a potential environmental benefit to be gained by a modal shift away from the road, this shift is not as large as is often thought, since technological progress in road transport has been substantial during the past decades. Further, there are a number of mechanisms that have a favourable effect on the environmental performance of the infrastructure charge. These mechanisms are the reduction in total volume of freight transported, the increase in load factors and the move towards larger vehicles. Also, the way in which the AC charges have been formatted stimulates the adoption of environmentally friendly versions of vehicles. It appears that these countervailing forces dominate: Van den Bossche et al. (2005) find that, in the present case, the final effect of the infrastructure charging mechanism is positive for the environment.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

71

7. CONCLUSIONS

The infrastructure maintenance costs of rail are relatively high, considerably higher than the maintenance costs of highways. This gives the road system an advantage that is often overlooked in discussions on modal shift in freight transport. The inland waterways’ maintenance costs are at an intermediate position. The contribution of freight transport to the costs of construction and maintenance is relatively low in the various sectors. The charges paid by inland navigation are almost zero, due to the well-known Convention of Mannheim, and fuel taxes are absent here. In most European countries freight transport is gradually starting to pay for its user-dependent infrastructure maintenance costs, but the contribution to total infrastructure-related costs are still low. Taxes are highest in road transport, where fuel taxes and fixed costs of truck ownership apply. Marginal cost pricing would imply that transport sectors pay for the user-dependent costs of infrastructure maintenance. Marginal cost pricing is in harmony with the basic principles of economic theory, but since marginal costs are considerably lower than average in the domain of maintenance, this may still lead to substantial cost coverage problems. Given the possible distortions due to other sources of taxation – the maintenance costs have to be covered after all – it makes sense to investigate possibilities for full cost coverage that minimise the distortions induced by departures from the marginal cost rule. We apply engineering formulas to find that the share of the user-dependent part of highway maintenance costs is about 40%. In inland navigation this is only 10%, and rail has an intermediate position. In addition to charges equal to marginal costs, we also develop a system of charges based on the full costs but with variations, in order to accommodate issues of environmental costs that are not yet internalised. Simultaneous introduction of the system of charges to the three modes of freight transport in the Netherlands leads to the conclusion that the overall decrease in transport demand is limited, but that substantial modal shifts may take place. The most robust outcome is that rail transport’s modal share will decrease. This is at variance with the general EU policy to increase the share of rail in freight transport. This outcome is obviously a consequence of the focus on maintenance costs – where rail performs unfavourably – in this study. An important conclusion is therefore that in the debates on pricing policy and modal shift in freight transport, more attention should be paid to the role of maintenance costs.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

72 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION

NOTES

1. This does not mean that we want to imply that freight transport on local and regional roads is unimportant. A certain part of total freight traffic takes place via local and regional roads, although surprisingly little is known about the exact share. In addition, adverse effects of freight transport in terms of congestion, air quality and accidents are pertinent on local and regional roads. In the present context we focus on long-distance transport, since it is here that it makes sense to compare the three transport modes. We also do not address the usually short-haul pick-up and delivery movements of trucks in the case of intermodal transport. 2. For convenience, we have added tonne-kilometres and passenger-kilometres. This is obviously a crude approach, but in the context of an analysis of maintenance costs that obviously depend on the weight of loads transported, it may be defendable, given the average weight of a passenger car. 3. Previous research in the Netherlands on maintenance costs (see DWW, 2000 and KOAC/WMD, 2001) used substantially lower percentages. However, these studies did not have access to the detailed, bottom-up information in the study provided by DWW (2002). 4. Data on noise production factors are obtained from CE (1999). 5. The actual data used for imputation of these costs provide information on whether a traffic accident caused a death and on which vehicle caused the accident. 6. We distinguish between six Euro classes, which is three more than the number of classes in the German “LKW-Maut” (see IWW/Prognos, 2002).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

73

ACKNOWLEDGEMENTS

An earlier version of this paper was presented at the International Workshop on Infrastructure Measurement and Management at the Jönköping International Business School, Jönköping, Sweden, 2004. Chapters 4 and 5 are based on a study carried out for the Ministry of Finance, and the Ministry of Transport and Public Works in the Netherlands (2004-5). For details, see CE/VU (2004). Chapter 6 is mainly based on Van Den Bossche et al. (2005). These studies have been part of a larger research project, initiated by The Netherlands Ministry of Finance and the Ministry of Transport and Public Works, on infrastructure charges in freight transport based on maintenance costs (IBO, 2005).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

74 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION

ANNEX 1

EMISSION FACTORS OF EURO CLASSES IN ROAD FREIGHT TRANSPORT

The composite emission factors per truck category and per Euro class (see Table 9) are calculated by taking a weighted average of emission factors for three types of pollutants, i.e. VOC (HC), NOx and PM10. Weight factors are the respective shadow prices, since differentiation of the base rate tariff in the main text is based on social external costs. As shown in De Blaeij et al. (2004), official shadow prices of local emissions vary widely between countries. Moreover, differentiation between areas on the basis of population density has a substantial impact on the magnitude of the shadow price. Our interest in this paper is mainly to give an example of how fixed costs can be differentiated on the basis of social external costs, not to give an exact estimate. We therefore take average shadow prices for VOC, NOx and PM10 from CE (1999). These are €€ 3.36, €€ 7.84 and €€ 78.36 per kilogramme, respectively. Emission factors for VOC, NOx and PM10, representing minimum standards, are presented in Table A1 per truck weight category for each Euro class. Data were available for almost all vehicle types distinguished in our study. However, as with market shares, truck categories in Table A1 are not fully compatible with the categories used in this study (except for the truck trailer). We therefore had to make some assumptions in order to transform the data. Data for light trucks (< 10 tonnes) are used for the truck solo < 12 tonnes. Regarding the truck solo > 12 tonnes (truck combination > 12 tonnes), we assumed that approximately 80% (20%) of the vehicle-kilometres are driven by middle weight trucks and 20% (80%) by heavy trucks; figures are based on CBS (1996) data. For these two truck categories, the composite emission factors are therefore a weighted average with both the number of vehicle-kilometres by middle weight and heavy trucks and shadow prices of the three different emissions as weight factors. For instance, the calculation of the composite emission factor of a Euro 4 vehicle within the truck combination category (see Table 9) reads as {20%*(3.36*0.31 + 7.84*3.87 + 78.36*0.02)} / {3.36 + 7.84 + 78.36} + {80%*(3.36*0.40 + 7.84*5.44 + 78.36*0.03)} / {3.36 + 7.84 + 78.36} = 0.40.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

75

Table A1. Emission factors for VOC (HC), NOx and PM10 on highways in freight transport per truck weight category per Euro class (in grams per vehicle-kilometre) VOC (HC) Solo Combi Light trucks (<10 tonne) Pre-Euro 1.07 Euro 0 0.89 Euro 1 0.24 Euro 2 0.14 Euro 3 0.17 Euro 4 0.16 Euro 5 0.17 Middle weight trucks (10 – 20 tonne) Pre-Euro 1.03 0.79 Euro 0 0.86 0.79 Euro 1 0.44 0.46 Euro 2 0.26 0.27 Euro 3 0.32 0.32 Euro 4 0.31 0.31 Euro 5 0.32 0.31 Heavy trucks (> 20 tonne) Pre-Euro 0.75 0.51 Euro 0 0.51 0.51 Euro 1 0.56 0.60 Euro 2 0.33 0.35 Euro 3 0.40 0.42 Euro 4 0.39 0.40 Euro 5 0.40 0.41 Truck trailers Pre-Euro 0.62 Euro 0 0.49 Euro 1 0.53 Euro 2 0.30 Euro 3 0.36 Euro 4 0.34 Euro 5 0.35 -

NOx Solo

PM10 Combi

Solo

Combi

6.68 6.81 4.83 5.20 3.75 2.11 1.15

-

0.44 0.36 0.15 0.07 0.06 0.01 0.01

-

12.39 11.30 6.52 7.31 5.30 2.97 1.66

14.53 14.53 8.47 9.10 6.90 3.87 2.19

0.50 0.361 0.22 0.11 0.10 0.02 0.02

0.51 0.51 0.33 0.19 0.13 0.02 0.02

15.36 13.87 9.37 10.32 7.66 4.36 2.45

17.27 17.27 11.39 12.07 9.30 5.44 3.08

0.62 0.46 0.35 0.19 0.15 0.03 0.03

0.62 0.56 0.46 0.26 0.18 0.03 0.03

14.87 14.31 9.74 10.39 8.09 4.60 2.60

-

0.56 0.49 0.37 0.22 0.16 0.03 0.03

-

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

76 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION

ANNEX 2

IMPOSING A MINIMUM TARIFF UNDER A DIFFERENTIATED FIXED COST CHARGE

In policymaking it is possible that other issues play a role in the final tariff structure, for instance, public perceptions of fairness. In such situations a minimum value could be imposed, in which case the system does not exactly cover total costs (TC) any longer. For this reason we need the parameter S. Consider, for instance, the case that we impose a minimum on X of 0.5*B. The background for such an approach would be that a vehicle’s good environmental performance should indeed lead to a lower fixed maintenance cost charge, but that it is considered as unfair when the reduction brings the charge too far below the base rate tariff. Since in this case X is no longer an unknown, we can calculate S from equation (1). Rewriting (1), we get: S = {TC – 0.5*B*KM7} / {0.5*B*( EF6/EF7*KM6 + EF5/EF7*KM5 + EF4/EF7*KM4 + EF3/EF7*KM3 + EF2/EF7*KM2 + EF1/EF7*KM1)}. We repeat the procedure until costs per kilometre for all truck categories are larger than or equal to 0.5*B. Note that when the restriction is binding for one or more categories, the cost structure is no longer fully identical to the structure in emission factors presented in Table 9.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

77

BIBLIOGRAPHY

AASHO (The American Association of State Highway Officials) (1962), The AASHO Road Test, Proceedings of a Conference held May 16-18, National Academy of Sciences, St Louis, MO, National Research Council, Publication 1012, or in Special Report/Highway Research Board, ISSN 0077-5622, p. 73. Blaeij, A. de, M.J. Koetse, Y.-Y. Tseng, E.T. Verhoef and P. Rietveld (2004), Valuation of Safety, Time, Air Pollution, Climate Change and Noise: Methods and Estimates for Various Countries, Free University, Final report to ROSEBUD Thematic Network, Amsterdam. Borger, B. den and S. Proost (2004), “Transport pricing when several governments compete for transport tax revenue”, Paper presented at STELLA meeting, Athens. CE (1999), Efficiënte Prijzen voor Verkeer: Raming van Maatschappelijk Kosten van het Gebruik van Verschillende Vervoermiddelen (Efficient Transport Prices: Estimation of Social Costs of Using Different Transport Modes), Centre for Energy Saving and Clean Technologies (CE), Delft. CE/VU (2004), The Price of Transport: Overview of the Social Costs of Transport, CE publication, No. 04.4850.40, Delft. CPB (2004), Equal Rules or Equal Opportunities?, CPB, The Hague. DWW (2000), Gebruikerskosten (User costs), Road and Hydraulic Engineering Institute (DWW), Delft. DWW (2002), Basisonderhoudsniveau 2001 (Base Level Maintenance 2001), Road and Hydraulic Engineering Institute (DWW), Delft. EC (Commission of the European Communities) (1995), Towards Fair and Efficient Pricing in Transport, Brussels. Ecorys (2005), Effecten Gebruiksvergoedingen in het Goederenvervoer (Effects of User-Cost Charges in Freight Transport), forthcoming. EU (1999), Directive 1999/62/EG of the European Parliament and the Council Concerned with Charging the Use of Certain Infrastructural Services to Heavy Goods Vehicles, European Communities, Brussels. Gerondeau, C. (1997), Transport in Europe, Artech House, London. Goodwin, P.B. (1992), A review of new demand elasticities, Journal of Transport Economics and Policy, 26, 155-169.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

78 - INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION Gruebler, A. and N. Nakicenovic (1991), The Evolution of Transport Systems: Past and Future, IIASA Research Report 91-008, Laxenburg, Austria. HCG (1996), LMS Basismatrices ’94; PAE-Factor Vrachtverkeer (Passenger Car Equivalents in Freight Traffic), The Hague (study commissioned by AVV, Ministry of Transport). IWW/Prognos (2002), Wegekostenrechnung für das Bundesfernstrassennetz (Calculating Road Costs for the Public Road Network), IWW and Prognos, Basel/Karlsruhe. IBO, Interdepartementaal Beleidsonderzoek (2005), Beprijzing van het gebruik van rijksinfrastructuur door het goederenvervoer (Pricing the Use of Highways by Freight Transport), The Hague. Johnsson, R. (2004), The Cost of Relying on the Wrong Power – Road Wear and the Importance of the Fourth Power Rule (TP446), Transport Policy, 11, 345-353. KOAC-WMD (2001), Onderzoek naar de Jaarlijkse Onderhoudskosten aan het Wegennet, Veroorzaakt door Overbelading van Vrachtauto's in Nederland (Research into the Annual Maintenance Costs of Road Infrastructure Caused by Overloaded Trucks in The Netherlands), for the Road and Hydraulic Engineering Institute (DWW), Delft. Nash, C. (2003), Marginal Cost and other Pricing Principles for User Charging in Transport, Transport Policy, 10, 345-348. Policy Research Corporation (2002), Economische Impact Studie Goederenvervoer, Rotterdam. Oum, T., W. Waters and J. Yong (1992), Concepts of price elasticities of transport demand, Journal of Transport Economics and Policy, 26, 139-154. Rietveld, P. (2003), Winners and Losers in Transport Policy: On Efficiency, Equity, and Compensation, in: D.A. Hensher, K.J. Button (eds.), Handbook of Transport and the Environment, Handbooks in Transport Series, 4, pp. 585-602, Amsterdam: Elsevier. RIVM (2004), Methoden voor de Berekening van de Emissies door Mobiele Bronnen in Nederland t.b.v. Emissiemonitor; Jaarcijfers 2001 en Ramingen 2002 (Methods for Calculating Emissions by Mobile Sources in The Netherlands for the Emission Monitor; Yearly Figures for 2001 and Estimations for 2002), Report Series MilieuMonitor, No. 13, Bilthoven. Rothengatter, W. (2003), How Good is First Best? Marginal Cost and other Pricing Principles for User Charging in Transport, Transport Policy, 10, 121-130. TNO (2003), Emissiefactoren Heavy Duty Wegvoertuigen ten behoeve van EmissieRegistratie 2003 (Emission Factors Heavy Goods Vehicles on the Road for Emission Registration 2003), TNO, Delft. Ubbels, B., P. Rietveld and P. Peeters (2002), Environmental Effects of a Kilometre Charge in Road Transport, Transportation Research D, 7, 255-264. Ubbels, B. (2004), Institutional Barriers to Efficient Policy Intervention by Different Levels of Government in the European Port Sector, Vrije Universiteit, Amsterdam.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

INFRASTRUCTURE MAINTENANCE COSTS: A COMPARISON OF ROAD, RAIL AND INLAND NAVIGATION -

79

Van den Bossche, M., J. Bozuwa, W. Spit and K. Vervoort (2005), Effecten gebruiksvergoeding in het goederenvervoer, Ecorys, Rotterdam. Verhoef, E.T. (1996), The Economics of Regulating Road Transport, Edward Elgar, Cheltenham. VU (2002), Infrastructuurkosten van het Goederenwegverkeer: Een Verkenning op basis van Beschikbare Gegevens (Costs of Road Infrastructure of Freight Transport: An exploration on the basis of available data), Department of Spatial Economics, Free University, Amsterdam.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

Cover_f.fm Page 1 Wednesday, April 7, 2004 11:00 AM

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL

Carlos Pestana BARROS Instituto Superior de Economia e Gestao Technical University of Lisbon Lisbon Portugal

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

81

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

83

SUMMARY

ABSTRACT .......................................................................................................................................... 85 1.

INTRODUCTION ......................................................................................................................... 85

2.

INSTITUTIONAL SETTING ....................................................................................................... 87

3.

LITERATURE REVIEW .............................................................................................................. 89

4.

THEORETICAL FRAMEWORK................................................................................................. 92

5.

DATA ............................................................................................................................................ 94 5.1. Results.................................................................................................................................... 95 5.2. Efficiency rankings ................................................................................................................ 97

6.

DISCUSSION................................................................................................................................ 98

7.

CONTRIBUTION, LIMITATIONS AND EXTENSIONS OF THIS STUDY .......................... 100

8.

CONCLUSIONS ......................................................................................................................... 100

BIBLIOGRAPHY ............................................................................................................................... 101

Lisbon, July 2005

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

85

ABSTRACT

This paper uses an econometric frontier model to evaluate the technical efficiency of Portuguese airports from 1990 to 2002, combining operational and financial variables. A Cobb-Douglas cost specification of the technical efficiency effects model is used to generate airports’ efficiency scores, allowing for non-discretionary variables which may affect inefficiency. We conclude that the efficiency scores are mixed. A policy is devised for the management of this sector. Keywords: Portugal, airports, Cobb-Douglas cost frontier model, technical efficiency, policy implication.

1. INTRODUCTION

In this paper, we measure the technical efficiency of the Portuguese airports with a Cobb Douglas cost frontier model, using data obtained in Statistics of Transportation, published yearly by INE (the Portuguese statistical agency) from1990 to 2002. Previous research into these airports has made use of DEA (Data Envelopment Analysis), e.g. Gillen and Lall (1997), Parker (1999), Murillo-Melchor (1999), Yoshida and Fujimoto (2004). Moreover, Pels, Nijkamp and Rietveld (2001, 2003), have analysed the airports with an Error Component Model. This paper analyses the efficiency of Portuguese airports by using a technical efficiency model, allowing for non-discretionary, as well as designed contextual variables, which may affect the performance of the airports. Non-discretionary variables are outside management control. In fact, the airport economics literature gives rise to the expectation that airport location will have a major effect on performance. As the airport context is composed of the population and income in the airport area, these contextual variables have to be included in the cost function. The motivation for the present research is the following. Portuguese airports have grown in the period under analysis, from 1990 to 2002, but face a number of threats in the ever more globalised contemporary world. Firstly, the public nature of airports makes them prone to the principal-agent relationship (Jensen and Meckling, 1976). This relationship concerns the difficulty of controlling those empowered as managers to act on behalf of the owner (i.e. the government), and is a prevalent issue in public enterprises. The job tenure of the airport managers may encourage the development of principal-agent problems, since managers are always linked to friends, associates or patrons in the incumbent political party. However, recognising that the principal-agent theory - which assumes that a benevolent principal has full control over the legal framework and over rewards and penalties - is too restricted, we also consider the rent-seeking theory (Mueller, 1979).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

86 - TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL Rent-seeking implies that politicians are motivated by their own self-interest and that their subsequent behaviour is often against the public interest (Tullock, 1967). In this context, the actions by airport directors that are designed to obtain personal gains, such as a better appointment, from the authorities (government), may imply that they lack commitment and dedication to the position they hold. Consequently, the time and effort they devote to improving the internal efficiency of their operations may be reduced. A second motivation is the increase in air traffic and thus congestion, which creates a pressing need for the construction of a new airport in place of the present Lisbon Airport. This new construction will destroy a competitive advantage held by the present airport at European level, which is its convenient position near the city centre. Moreover, the row observed in the media between different interest groups, each vying to construct the airport in their chosen zone, signifies that this project will be affected by the interest group theory, Olson (1965). A third motivation is based on the deficit in management skills observed in Portugal, according to the perceptions of Portugal-based business managers from overseas, for whom incompetence and inefficiency is rife among their Portuguese counterparts. This perception has emerged from an exhaustive survey carried out jointly by Ad-Capita Executive Recruitment and Research and the Cranfield School of Management, UK (see report in pdf: “Can Portuguese Managers Compete?” at www.adcapita.com). Being a generic finding, it can be applied to the Portuguese airports. The final motivation for the present research is based on the growing outsourcing observed for the different airport services of the main Portuguese airports, such as luggage handling and shops. Despite management appeals for outsourcing, the monitoring of airport operations has to increase in this context, and there is an observed tendency towards long waiting periods for luggage and congestion in the access areas to planes, which is inclined to be handled with complacency by the public managers. On the other hand, Lisbon’s European capital airport has no pharmacy as yet, and no train or metro linking it to the city centre, a characteristic it shares with the other airports analysed. The relative efficiency of airports may also be affected by differences in their structural characteristics within the industry, leading to variations in performance due to different strategies (Caves and Porter, 1977) or resources (Barney, 1991). This paper extends previous research into airport efficiency, adopting a stochastic frontier model, alongside Pels, Nijkamp and Rietveld (2001, 2003), to evaluate the technical efficiency of Portuguese airports. However, this paper adopts the technical efficiency effects model, found in Coelli et al. (1998), which allows for contextual variables in the cost function. A sample of the main, representative Portuguese airports is used. The paper is organised as follows: Chapter 2 describes the institutional setting; Chapter 3 surveys the literature on the topic; in Chapter 4, the theoretical framework is presented; Chapter 5 shows the data and results; in Chapter 6, the results are discussed; Chapter 7 discusses the limitations and possible extensions of the paper and, finally, Chapter 8 concludes.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

87

2. INSTITUTIONAL SETTING

There are 37 Portuguese airports, most of which are small, regional airports operating in a discontinuous fashion. The main airports are run by public enterprises, regulated by a public body, the Instituto Nacional de Aviação Civil (INAC) (national institute of civil aviation), and this same public organisation comes under the direct control of the Ministry of Transport. This public authority oversees the airports, which are situated throughout mainland Portugal and the archipelagos of Madeira and the Azores. The airports are owned by public enterprises: ANA owns the airports of mainland Portugal and the Azores (Lisboa, Faro, Ponta Delgada, Santa Maria, Horta and Flores), whilst ANAM owns the Funchal and Porto Santo airports, Barros and Sampaio (2004). The Portuguese Government has shareholdings of 68.111% in ANA and Parpublica-Participações Pública owns 31.89%. Parpublica is a holding company of public enterprises, owned by the Government. The Madeira Regional Government owns 20% of ANAM, the rest being held by ANA (70%) and by the Portuguese Government (10%). In 1998, the public service of support to civil air transport was taken out of ANA and organised as an independent enterprise (Navegação Aérea de Portugal, E.P. (NAP). Table 1 below lists the 13 main airports and the annual freight handled by each of them, arranged in geographical order from north to south, followed by the Islands. Table 1. Movement of planes and passengers through Portuguese airports 1999

1 2 3 4 5 6 7 8 9 10 11 12 13

Mainland: Porto Lisbon Faro Azores: Santa Maria Ponta Delgada Lajes Horta Graciosa Pico Saint George Flores Madeira: Funchal Porto Santo

2000

Planes

Passengers

Planes

Passengers

11 574 30 862 11 252

1 355 682 5 282 553 2 757 549

22 446 54 260 16 797

293 582 9 394 532 4 704 780

634 2 954 3 801 1 237 402 661 576 281

64 783 388 562 344 329 101 339 28 142 49 221 42 896 16 705

682 4 514 4 676 1 951 418 703 554 487

70 716 681 552 477 561 169 296 29 099 48 825 42 347 29 353

6 475 2 403

1 010 603 111 081

10 274 2 570

2 019 535 195 261

Source: Financial Reports of ANA and ANAM. ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

88 - TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL We can see that the sample includes all airports that allow heavy load planes (with loads of more than 350 tonnes) to land, as well as those allowing medium load planes (with loads of between 201-350 tonnes) to operate. The thirteen airports listed in Table 1 represent almost 100% of the traffic passing through Portuguese airports, indicating that they are highly representative of the national sector. Mainland Portugal has three main airports, two situated in the cities of Lisbon and Porto and another situated in the main tourist area, Faro. The other airports are on the islands of the Azores and Madeira. Based on the number of planes, we can see that Lisbon is the main airport, followed by Porto, Faro and Funchal. The importance of Faro and Funchal in terms of air traffic is based on the flow of tourism they receive, to which the local traffic is also added. The fifth position attained by Lajes airport is due to the American air force base located there, which causes the local flow to be higher than that found in the capital of the Azores (Ponta Delgada). However, this airport also has civil traffic, justifying its inclusion in the present analysis. Portuguese airports are faced with various threats in our increasingly global contemporary world: • • • • •

• •

•

Firstly, Lisbon Airport is set to reach its full capacity in the near future, and there are plans to build a new airport. Secondly, Funchal has enlarged its previous capacity in order to allow bigger planes to land. Thirdly, certain problems, such as the excessively long time needed to reclaim baggage on arrival in Lisbon, are evident when comparisons are made with other European airports. Fourthly, all airports are currently operating with financial deficits, which creates a climate of uncertainty as to their declared possible privatisation. Fifthly, Portuguese airports used to apply quantity discounts to those airlines cross-subsidising the national airline, TAP; however, this subsidy was condemned by the European Court in the Summer of 2001, and INAC has adopted another pricing policy in line with the demands of the European Union. Sixthly, Portuguese airports usually charge higher landing fees than Paris, Rome or Madrid, attracting complaints from tourist authorities and low-cost airline companies. However, in the past year there has been much progress on this issue. Seventhly, the appointment of airport managers is prone to the influence of politicallydominant interest groups and their private agendas (Olson, 1965), which impedes modernisation and development on the basis of economic, commercial and professional prerogatives. Moreover, there is the continued existence of corporate nepotism, exemplified recently by the concession of works for the airport to an enterprise belonging to a friend of a senior manager of Lisbon Airport (Journal Publico, 29 April 2005, page 22). Finally, there is an historically determined capital deficit that has been partially offset by European funds.

The Government bears the cost of the airport authorities’ financial deficits. The allocation of public resources to each airport authority is based on its declared deficit. However, despite the follow-up inspection procedure carried out by the Regulatory Agency, auditing and feedback about the airports’ activities are still imprecise, as can be inferred from the financial reports presented by each airport authority. These reports focus mainly on the financial performance of the unit, which is presented in a standard format, in accordance with legal stipulations. However, despite being standardised, information on activity does not include the breakdown of some airports’ financial accounts, with the result that aggregated data are frequently presented that do not allow for an accurate assessment of their performance. Moreover, the costs of some airports are not broken down in the financial reports. This, in turn, implies that the Government is, at best, under-informed as to the effects of its policy. From this, it can be inferred that airports are free to set their own private agendas, bypassing the public objectives which they are assumed to pursue and which would be in keeping with ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

89

the Government’s declared policy goals. Barros and Sampaio (2004) and Jessop and Barros (2004) have analysed Portuguese airports.

3. LITERATURE REVIEW

Whilst there is extensive literature on benchmarking, applied to a diverse range of economic fields, the scarcity of studies with regard to airports bears testimony to the fact that this is a relatively under-researched topic. In relation to the two scientific methods used to analyse efficiency in quantitative terms, namely, the econometric frontier and DEA methods, several papers were found that applied at least one of these. There are four basic DEA models: the CCR model (Charnes et al., 1978), the BCC model (Banker et al., 1984), the multiplicative DEA model of Charnes et al. (1982) and the additive DEA model of Charnes et al. (1985). The main differences between the models stem from several factors, such as whether they do or do not take into account the existence of economies of scale, the geometric form of the efficiency frontier, and the way in which inefficient units are projected onto the frontier. Besides these basic DEA models, there also exist other tools of analysis, such as the Malmquist index (Malmquist, 1953) and the allocative and technical disentangling of efficiency (Coelli, Rao and Battese, 1998). Related to stochastic frontier models, as already said, there are two basic models (Coelli et al., 1998): the error component model, adopted in this paper, and the technical efficient effects model adopted in Pels, Nijkamp and Rietveld (2001, 2003). Besides these two basic models there are also other tools of analysis, such as the decomposition of technical change, the input distance function, Yoshida and Fujimoto (2004), and the disentangling of technical and allocative efficiency. Among the papers using DEA, Parker (1999) analyses the performance of the British Airport Authority before and after its privatisation, with data from financial reports for the period 1979-80 to 1995-96, using the CCR and BCC models. Murillo-Melchor (1999) estimates a Malmquist index to analyse Spanish airports between 1992-94, whilst Gillen and Lall (2001) estimate a Malmquist index for 22 major USA airports, using DEA for the period 1989-93. Pels, Nijkamp and Rietveld (2001) use DEA to analyse the efficiency of a sample of European airports between 1995 and 1997, using the BCC model. Fernandes and Pacheco (2002) use DEA to analyse 16 Brazilian domestic airports, using the BCC model. Pels, Nijkamp and Rietveld (2003) compare the efficiency rankings of a sample of European airports with the DEA-BCC model and an econometric stochastic frontier, whilst Adler and Berechman (2001) use a principal component analysis, derived from a survey based on a DEA-BCC model to explain airport quality with the use of a questionnaire. In a related paper, Hooper and Hensher (1997) analyse the total factor productivity of US airports with an index number approach. For surveys on airport performance, see Humphreys and Francis (2002). In Table 2 we present the models, inputs and outputs used in the various papers referred to.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

90 - TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL Table 2. Research into airport efficiency Papers Method Gillen and Lall DEA-BCC model (1997) and a Tobit model

Units 21 USA airports,

Inputs Outputs i) Terminal services model: I) Terminal services 1) Number of runways model: 2) Number of gates 1) Number of passengers 3) Terminal area 2) Pounds of cargo 4) Number of baggage ii) Movements mode:l collection belts 1) Air carrier movements 5) Number of public parking 2) Commuter movements spots ii) Movement model: 1) Airport area 2) Number of runways 3) Runway area 4) Number of employees Parker (1999) DEA-BCC and CCR 32 UK regulated 1) Number of employees, 1) Turnover, models airports, 1979/1980 to 2) capital input estimated as an 2) passengers handled, 1995/1996. In a second annual rental based on a real rate 3) cargo and mail model, uses 22 airports of return of 8% each year business. from 1988/89 to 1996/97 applied to net capital stock, and 3) other inputs defined as the residual of total operating costs. MurilloDEA-Malmquist 33 Spanish civil 1) Number of workers, Number of passengers Melchor (1999) airports, 1992 to 1994 2) accumulated capital stock proxied by amortisation, 3) intermediate expenses Gillen and Lall DEA-Malmquist 22 major USA airports, i) Terminal services model: i) Terminal services (2001) 1989 to 1993 1) number of runways, model: 2) number of gates, 1) number of passengers, 3) terminal area, 2) number of pounds. 4) number of employees, i) Movement model: 5) number of baggage collection 1) air carrier movements, belts, 2) commuter movements. 6) number of public parking places. ii) Movement model: 1) airport area, 2) number of runways, 3) runway area, 4) number of employees Pels, Nijkamp DEA-BCC model. 34 European airports, 1) Terminal size in square i) Terminal model: and Rietveld 1995 to 1997 metres, 1) Number of (2001)* 2) number of aircraft parking passengers. positions at the terminal, ii) Movement model: 3) number of remote aircraft 1) aircraft transport parking positions, movement. 4) number of check-in desks, 5) number of baggage claims. Pels, Nijkamp Stochastic frontier 34 European airports, 1) Constant, i) Terminal model: and Rietveld model. 1995 to 1997 2) number of baggage claim 1) Number of (2001)* units, passengers. 3) number of parking positions ii) Movement model: at the terminal, 1) aircraft transport 4) number of remote parking movement. positions. Adler and DEA-BCC with 26 European airports 1) Passenger terminals, 1) Principal components Berechman Principal Component runways, obtained from a (2001) Analysis. 2) distance to city centres, questionnaire on airlines. 3) minimum connecting times in

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

91

Table 2 (continued) Fernandes and DEA. Pacheco (2002)

16 Brazilian airports, 1998

Pels, Nijkamp and Rietveld (2003)**

33 European airports, 1995 to 1997

DEA-BCC model.

Pels, Nijkamp Stochastic frontier and Rietveld model (2003)** Sarkis (2000) Several DEA models, including the CCR and BCC models. Sarkis and DEA-CCR and Talluri (2004) cross-efficiency DEA model from Doyle and Green (1994)

As above.

1) Airport surface area in m2, 2) departure lounge in m2, 3) number of check-in counters, 4) curb frontage in metres, 5) number of vehicle parking spaces, 6) baggage claim area in m2. i) Terminal model: 1) Airport surface area, 2) number of aircraft parking positions at terminal, 3) number of remote aircraft parking positions, 4) number of runways; 5) dummy z variables for slotcoordinated airport and 6) dummy z variable for time restrictions ii) Movement model: 1) number of check-in-desks, 2) number of baggage claim units; 3) annual number of domestic and international movements. As above.

43 US airports from 1990-1994.

Operating costs, employees, gates and runways.

43 US airports from 1990-1994.

Operating costs, employees, gates and runways.

Domestic passengers.

i) Terminal model: 1) annual number of domestic and international movements ii) Movement model: 1) annual number of domestic and international passengers.

As above.

Operating revenues, aircraft movements, general aviation, total passengers, total freight. Operating revenue, aircraft movements, general aviation, total passengers, total freight.

Barros and DEA - allocative Sampaio (2004) Model.

10 Portuguese airports Number of employees, capital Number of planes, 1990-2000. proxied by the book value of number of passengers, physical assets, price of capital general cargo, mail cargo, and price of labour. sales to planes and sales to passengers. Yoshida (2004) Endogenous-Weight 43 Japanese airports, Runway length and terminal Passenger loading, cargo method 2000. size. handling, aircraft movement. Yoshida and DEA-CCR, DEA- 43 Japanese airports, Runway length, terminal size, Passenger loading, cargo Fujimoto (2004) BCC and Input 2000. monetary access cost, time handling, aircraft distance function. access cost, number of movement. employees in terminal building.

* **

The paper by Pels, Nijkamp and Rietveld (2001) presents two methods for analysing efficiency. We therefore present the paper in two rows in order to explain the techniques. The paper by Pels, Nijkamp and Rietveld (2003) presents two methods for analysing efficiency. We therefore present the paper in two rows in order to explain the techniques.

We can see that there is a tradition of analysing airports by separating activities into terminals and movements (Gillen and Lall, 2001; Pels, Nijkamp and Rietveld, 2001; Pels, Nijkamp and Rietveld, 2003). Several papers compare the DEA model with the frontier model (Pels, Nijkamp and Rietveld, 2001; Pels, Nijkamp and Rietveld, 2003), while others mix principal component analysis with a DEA ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

92 - TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL model (Adler and Berechman, 2001). Therefore, the use of a stochastic frontier model allowing for contextual variables is innovative in this context.

4. THEORETICAL FRAMEWORK

In this paper, we adopt the stochastic cost econometric frontier approach. The frontier approach, first proposed by Farrell (1957), was based on cost functions and came to prominence in the late 1970s as a result of the work of Aigner, Lovell and Schmidt (1977), Battese and Corra (1977) and Meeusen and Van den Broeck (1977). The adequacy of a cost or production function depends on the environment in which the units analysed operate. In an environment where the ultimate objective is to minimise costs, the producers face endogenously determined input prices and exogenously determined output prices, and attempt to allocate inputs and outputs so as to minimise costs. Assuming this is the main strategy of airport management, the cost frontier is the most adequate model for analysing efficiency (Kumbhakar, 1987). The general frontier cost function, which is dual to the production function proposed by Aigner, Lovell and Schmidt (1977) and Meeusen and Van den Broeck (1977), is as follows: Cost it = β 0 t + α it Pit + γ it Y it + (V it + U it )

i = 1,2,…N; t = 1,2,…N.

(1)

where Cit represents a scalar cost of the i decision-making unit under analysis in the t-th period, Pit is a vector of input prices, and Yit is a vector of output descriptors used by the i-th airport in the t-th period. The error term Vit is the traditional error term of econometric models, assumed to be independently and identically distributed, which represents the effect of random shocks (noise) and is independent of Uit. The inefficient term Uit represents technical inefficiencies and is assumed to be positive and distributed normally with zero mean and variance σ U2 . The Uit positive disturbance is reflected in a half-normal independent distribution truncated at zero, N (mit,σU2 ) , signifying that each airport’s production must lie on or above its cost frontier, but above the level of one. This implies that the two effects, the V effect, which is a random shock, and the U effect, which is a management shock controlled by the airport, cause any deviations from the frontier. The total variance is defined as σ 2 = σV2 + σU2 . The contribution of the error term to the total variation is as follows: σV2 = σ 2 /(1+ λ2) . The contribution of the inefficient term is as follows: σU2 = σ 2λ2 /(1 + λ2 ) , where: σV2

is the variance of the error term V,

2 σU

is the variance of the inefficient term U,

and λ is defined as

σ λ= U σV

,

providing an indication of the relative contribution of U and V to ε (Battese and Corra, 1977). ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

93

The mean inefficiency (mit) of the technical efficiency effects model, in Coelli et al. (1998) is a deterministic function of p explanatory variables:

mit = zit δ + w it

(2)

where: zit is a vector of factors affecting airport i inefficiencies in period t; δ is a vector of parameters to be estimated; and wit is defined by the truncation of the normal distribution with mean zero and variance σ2 and -zitδ<wit, in order to conform with the assumption that Uit is non-negative and truncated at zero. Using this parameterisation, a test can be constructed to determine whether the estimated frontier is actually stochastic: λ = 0 implies that the variance associated with the one-sided (efficiency) errors, 2 , is zero, meaning that these deviations from the frontier are better represented as fixed effects in the σU production function. Therefore, a test of the null hypothesis that λ = 0 against the alternative hypothesis that λ is positive is used to test whether deviations from the frontier are stochastic, and whether one should proceed with the estimation of parameters related to the sources of inefficiency within the context of a stochastic production frontier. Failure to reject the null hypothesis suggests that the determinants of inefficiency, Zit should be included in the cost function. The parameters of the model (β, δ, σ and λ) are estimated using the maximum-likelihood estimator; the likelihood function can be found in Battese and Coelli (1988). Thus, the technical inefficiency of the i-th airport at time t is: TEit = ex(−Uit ) = exp(− zitδ − Wit )

(3)

The conditional expectation of TE is defined under the half-normal assumption:

[

E Ui / εi1,...ε it

]

⎡ ⎢ ⎢ ⎢ * * = μi + σ i ⎢ ⎢ ⎢ ⎢ ⎣⎢

⎤ ⎥ ⎥ ⎥ ⎥ μi* ⎥ (− ) ⎥ σ i* ⎥⎦⎥

μ* φ( i ) σ i*

φ

where: 2 μi* = γ i μ + (1 − γ i )(−εi ) , γ i = 1/(1 + λ ) and σ i* = ( σU ) ; Ti 1 + λTi μ is the mean value of the distribution and T is the time period of the panel;

φ is the standard normal distribution and φ is the respective cumulative distribution function. (Coelli, Rao and Battese, 1998; Kumbhakar and Lovell, 2000).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

(4),

94 - TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL

5. DATA

To estimate the production frontier, we used balanced panel data on Portuguese airport authorities in the years 1990 and 2002 (13 years × 13 airports = 169 observations). The airports considered in the analysis are those listed in Table 1. Given the lack of consensus in airport literature about the appropriate production model to be employed, we considered the airports as a service industry that operates by enabling the landing of planes and the access of passengers and cargo to these same planes. The chosen variables were based on two criteria: the literature survey and the data available. Therefore, using a stochastic production function framework and data from the Transportation Statistics on airport authorities, we measured the costs as total operating costs. We measured output through three indicators: sales to planes, sales to passengers and aeronautical fee. Input prices are not directly observed and so must be constructed from the available information, by dividing flows of expenditure by stocks. The price of labour equals salaries and benefits divided by the number of employees. The price of physical capital premises is expenditure on equipment and premises, divided by the book value of the physical assets. The price of capital stock is the ratio of earnings to the stock. A time trend is included to capture time effects. The contextual variables are the population and income in the district area of the airport and a dummy variable which is one for national airports and zero for regional airports. The characteristics of these variables are shown below in Table 3.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

95

Table 3. Descriptive statistics of the data Variable Log C Trend

Log PL

Log PK

Log PM

Log Planes Log Passengers Log Fee Log Population Log Income National

Description

Minimum

Maximum

Mean

Logarithm of operating costs (euros) at constant prices 1999=100 Variable that ranges from 1 at the beginning of the period to 11 at the end of the period, for each airport. Logarithm of price of labour, measured by dividing total salary expenditure by the number of equivalent employees at constant prices Logarithm of price of capital-premises, proxied by expenditure on equipment and premises divided by the book value of physical assets Logarithm of the price of materials, proxied by the ratio of other costs estimated as the operational cost less wages by the book value of premises Logarithm of the sales to planes at 1999=100 Logarithm of the sales to passengers at 1999=100

5.00

10.71

6.534

Standard deviation 0.783

1.00

11.00

5.766

3.513

2.69

4.06

3.404

0.348

0.16

7.51

2.573

0.914

0.05

3.86

2.384

0.251

4.74 4.81

8.93 9.14

6.855 6.929

1.151 1.182

Logarithm of aeronautical fees at constant prices 1999=100 Log of the population in the district where the airport is situated Log of the income in the district where the airport is situated Dummy variable which is one for national airports and zero for regional airports

2.25

4.23

3.825

2.523

5.830

7.016

6.649

6.618

4.144

4.507

4.342

3.742

0

1

0.35

⎯

Source: Transportation statistics. INE (Portuguese Statistical Agency). Years 1990-2001.

We verify that the airports analysed are relatively homogeneous, with a standard deviation always lower than the mean, which validates the benchmark exercise undertaken on it. 5.1. Results In this study, we estimated a Cobb-Douglas stochastic cost function with three input prices (one price of labour and two prices of capital), three outputs (sales to planes, sales to passengers and non-aeronautical fee), trend and contextual variables (population and income in the airport area and a dummy for national airports).

⎛ C

Log ⎜

it ⎜ PM it ⎝

⎞ PK PL it ) + β LogPlanes ⎟ = β + β Trend + β Log ( it ) + β Log ( 0 1 3 4 ⎟ 2 it PM PM it it ⎠

+ β LogPassengers + β 6 LogFee + (Vit + U it ) it it 5 Ui = δ 0 + δ1 log Populationit + δ 2 log Incomeit + δ 3Nationalit

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

(5)

96 - TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL This is the cost frontier model found in Coelli et al. (1998), known as the technical efficiency effects model because it accounts for the causes of efficiency that are controlled by management (labour, capital, sales to planes, sales to passengers, non-aeronautical fee) and for the contextual factors that are beyond managerial control (population, income, national). The trend variable is introduced into the cost function to capture exogenous change. The variables were defined and characterised in Table 3. Linear homogeneity in input prices is imposed by dividing monetary values by the input price of capital-premises, Cornes (1992). Table 4 presents the results obtained for the stochastic frontier using Frontier 4.1 from Coelli (1996), with a half-normal distribution specification. Table 4. Stochastic Cobb-Douglas panel cost frontier model. (Dependent variable log of total cost) Variables Constant (β0) Trend(β1) Log PL (β 2) Log PK1 (β3) Log Planes (β4) Log passengers (β5) Log Fee (β6) Constant (δ0) Log Population (δ1) Log Income (δ2) National (δ3)

2 σ 2 = σV2 + σU

γ= μ

2 σU

σ2

Log (likelihood) Lagrange test Observations

Coefficients (t-ratio) -0.075 (-5.355) 0.325 (2.996) 1.751 (15.062) 0.101 (2.777) 0.543 (1.840) 0.301 (2.085) -0.400 (-3.064) 0.899 (0.190) -0.308 (-1.260) -0.303 (-3.553) -0.046 (-1.172) 0.369 (4.939) 0.252 (3.868) 0.66794 (7.614)* 13.693 0.366 169

Note: t Statistics in parentheses are below the parameters, those followed by * are significant at the 1% level. ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

97

We can see that the Cobb-Douglas cost function specified above fits the data well, as the R-squared from the initial ordinary least-squares estimation that was used to obtain the starting values for the maximum-likelihood estimation is in excess of 85% and the overall F-statistic is 282.02. We can also see that the variables have the expected signs, with the operating cost increasing with the price of labour and the price of capital-stock. Moreover, the total cost decreases with sales to planes, sales to passengers, and non-aeronautical fee. Finally, relative to the contextual variables, the coefficients δ are the inefficient coefficients, so a negative coefficient indicates a positive impact on efficiency. The frontier parameters are all statistically significant and the inefficient error term (γ) is 66% of the total variance. Technical inefficiency exists in the sample, based on the statistical significance of miu (μ). 5.2. Efficiency rankings Table 5 presents the results of the time-invariant efficiency scores computed from the residuals. Technical efficiency is achieved, in a broad economic sense, by the unit which allocates resources without waste, and thus the concept refers to a situation on the frontier. A unit with a score equal to one is on the frontier and those with a score lower than one are above the cost frontier of best practices. The value of waste is measured by the difference between one and the score, so that, for example, the waste of Porto Santo is 1-0.725= 0.275. This is a relatively high degree of waste. Table 5. Average technical efficiency scores for Portuguese airports, 1990-2000 Airports Lisbon Porto Faro Funchal Lajes Graciosa Flores Santa Maria Saint George Ponta Delgada Porto Santo Pico Horta Mean Median Standard Deviation

Efficiency scores 1.000 0.978 0.975 0.972 0.912 0.850 0.872 0.866 0.859 0.837 0.725 0.627 0.528 0.846 0.866 0.142

We can see that the mean score is 84%. This score suggests that Portuguese airports could reduce their output cost by 16% without decreasing their input, which, in this case, is the price of labour and the price of capital stock. The maximum airport score was naturally 1, which was achieved by Lisbon Airport, while the minimum efficiency score was 52.8% and was achieved by Horta. The median was 86.6%, which was higher than the mean. Therefore, there are more airports above than below the ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

98 - TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL mean. The standard deviation was 1.42%. These efficiency scores are average in comparison with those found elsewhere in the same activity (Pels, Nijkamp and Rietveld, 2001).

6. DISCUSSION

This paper has proposed a simple framework for the evaluation of Portuguese airports and the rationalisation of their management activities, taking into account the traditional input and output descriptors of airport activity. The analysis is based on a stochastic cost frontier model. Benchmarks are provided for improving the performance of inefficient airports. The stochastic frontier model used identifies the factors that cause the inefficiency and directs attention to the airports in which inefficiency exists. The cost increases with the price of labour and capital, and decreases with planes, passengers and aeronautical fees. This is an intuitive result, with the cost increasing with the production factors, but decreasing with the output. Relative to the contextual variables, the coefficients δ are the inefficient coefficients, so a negative coefficient indicates a positive impact on efficiency. Therefore, the population of the airport area contributes positively to efficiency, alongside income and national classification of airports. The scale in a Cobb-Douglas function is defined as the sum of the parameters and, in the present case, the output elasticity is equal to 0.444 in the sample mean, signifying increasing returns to scale. Thus, a 10% increase in outputs leads to a decrease in costs of approximately 44.4%. The reverse of this is higher than unity, indicating increasing returns to scale on production. This result means that scale is a major issue for Portuguese airports, and explains the position of the major airports in the ranking. Confronting these results with the threats faced by the airports, how do we interpret these findings? With traffic increasing because of the generalised preference for air transport, the Portuguese airports have many choices available to upgrade their efficiency. There are some priorities to handle the situation from a public policy perspective: firstly, to increase the quality of services; secondly, extracting value from the traffic; and finally, to relate the subsidy allocation to the performance of the airports. This is a desirable policy, because the country currently runs public deficits and an increasing public debt, which is against the rules of the European Union. In this context, the opportunity cost of a subsidy increase is high. Other theoretical reasons that can be advanced as causes of inefficiency in airports stem from strategic-based groups and differences in resource. Firstly, strategic-based groups (Caves and Porter, 1977) refer to differences in structural airport characteristics within an industry, which cause differences in performance. In the airports sector, companies with a similar asset configuration pursue similar strategies with similar performance results (Porter, 1979). While there are different strategic options among sectors of an industry, because of mobility impediments, not all options are available to each airport, inducing a spread of the efficient scores in this industry. Second, the differences in resource among airports (Barney, 1991; Rumelt, 1991; Wernerfelt, 1984) hold that airports are heterogeneous in relation to the resources and capabilities on which they base their strategies. These resources and capabilities may not be perfectly mobile across the industry, which results in ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

99

competitive advantage for the best-performing airports, and the dispersion of the efficient scores in the sample. Purchasable assets cannot constitute sources of sustainable profits. Indeed, critical resources are not available in the market. Instead, they are built and accumulated on the airport premises, their non-imitability and non-substitutability being dependent on specific traits of their accumulation process. Differences of resources thus result in barriers to imitation (Rumelt, 1991) and airports’ inability to alter their accumulated stock of resources over time. In this context, unique assets are seen as exhibiting inherently differentiated levels of efficiency; sustainable costs are ultimately a return on the unique assets owned and controlled by the airports (Teece et al., 1997). Accordingly, airports may achieve high levels of competitiveness by utilising vast amounts of resources, and thus perform inefficiently. Considering the difficulty in handling the current market situation, the airports have room to adjust their managerial decisions in order to upgrade their efficiency. These decisions should first of all be based on decreasing costs. Second, they should upgrade the quality of their management practices, responding to the criticisms made by the AdCapita report. Giving greater value to educational achievement among the workforce is an example of what can be done in this field. Third, they should adopt a benchmark management procedure in order to evaluate their relative position and to adopt appropriate managerial procedures for catching up with the frontier of “best practices”. Finally, they should pursue market-oriented strategies, which increase outputs and decrease inputs. What should the public policy be in this context? Despite being planned for a long time by different incumbent parties, the privatisation of Portuguese airports has not yet been implemented. The Government should first privatise the airports, recognising that competition stimulates growth and innovation (Jones, Tandon and Vogelsang, 1990). Moreover, this policy would immediately eliminate interest groups and rent-seeking procedures. Alternatively, the Government should adopt human resources policies that limit the principal-agent relationship, as well as eliminating collective action problems (Olson, 1965), in which workers can free-ride on the back of the management’s own efforts to improve performance, or public managers can free-ride government policy. Secondly, coherent transport policies should be adopted in constructing the new airport, in order to diminish congestion, pollution and noise in Lisbon, without completely eliminating the existing convenient location.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

100 - TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL

7. CONTRIBUTION, LIMITATIONS AND EXTENSIONS OF THIS STUDY

In the light of the literature on productivity in airports, it is useful to consider the potential contributions of the current research. Based on the literature survey, a stochastic frontier model for airport efficiency is estimated. Moreover, the Error Components Model proposed by Coelli et al. (1998) is adopted, which takes into account efficiency that is due to the variables in the cost function. Finally, our stochastic cost frontier model lends support to similar works by Pels et al. (2001), but it is worth noting that our construct has a stronger theoretical foundation, since it takes different variables and a cost function with homogeneity of input prices. The limitations of the paper are as follows: firstly, the homogeneity of the airports used in the analysis is questionable, since units of different sizes are compared. However, it can always be claimed that the units are not comparable and that, therefore, a ratio analysis could not be carried out. Moreover, the data set is short, so that the conclusions are limited. In order for them to be generalised, more extensive panel data would be needed. A variety of extensions can be made to this paper. Firstly, a non-parametric free-disposal hull analysis can be used to assess the efficiency scores. However, previous research has shown that DEA scores are lower in value than econometric scores, but that the ranking is preserved when the same variables are used (Bauer et al., 1998). Secondly, and finally, comparisons with different airports operating in the other European markets represent the way forward for benchmarking purposes.

8. CONCLUSIONS

This article has proposed a simple framework for the evaluation of Portuguese airports and the rationalisation of their performance. The analysis is based on a stochastic cost frontier model. Benchmarks are provided for improving the operations of sub-optimal performing airports. Several interesting insights and implications from the study are raised. For the least efficient airports, adjustment is needed in order to achieve the efficient frontier. The general conclusion is that the Portuguese airports have different efficiency scores. Scale is a main driver in cost control, together with control of the cost factors, confirming the importance of a sound managerial strategy. More investigation is needed to clarify the determinant factors of inefficiency among Portuguese airports.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

101

BIBLIOGRAPHY

AdCapita and Cranfield University (2003), “Can Portuguese Managers Compete?”, Mimeo, available at www.adcapita.com Adler, N. and J. Berechman (2001), Measuring airport quality from the airlines’ viewpoint: an application of data envelopment analysis, Transport Policy, 8, 171-181. Aigner, D.J., C.A.K. Lovell and P. Schmidt (1977), Formulation and estimation of stochastic frontier production function models, Journal of Econometrics, 6, 21-37. Bauer, P.W., A.N. Berger, G.D. Ferrier and D.B. Humphrey (1998), Consistency conditions for regulatory analysis of financial institutions: a comparison of frontier methods, Journal of Economics and Business, 50, (2) 85-114. Banker, R.D., A. Charnes and W.W. Cooper (1984), Some Models for Estimating Technical and Scale Inefficiencies in Data Envelopment Analysis, Management Science, 30,1078-1092. Barney, J. (1991), Firm resources and sustained competitive advantage, Journal of Management, 17, 99-120. Barros, C.P. and A. Sampaio (2004), Technical and Allocative Efficiency in Airports, International Journal of Transport Economics, 31, 3:355-377. Battese, G.E. and T.J. Coelli (1988), Prediction of firm-level technical efficiencies with a generalised frontier production function and panel data, Journal of Econometrics, 38, 387-399. Battese, G.E. and G.S. Corra (1977), Estimation of a Production Frontier Model: With application to the pastoral zone of Eastern Australia, Australian Journal of Agricultural Economics, 21, 169-179. Caves, R. and M.E. Porter (1977), From entry barriers to mobility barriers: Conjectural decisions and contrived deterrence to new competition, Quarterly Journal of Economics, 91, 241-261. Charnes, A., W.W. Cooper, B. Gollany, L. Seiford and J. Stutz (1985), Foundations of Data Envelopment Analysis for Pareto-Koopmans Efficient Empirical Production Functions, Journal of Econometrics, 30, ½: 91-107. Charnes, A., W.W. Cooper, L. Seiford and J.A. Stutz (1982), Multiplicative Model of Efficiency Analysis, Socio-Economic Planning Sciences, 16 (5), 223-224. Charnes, A., W.W. Cooper and E. Rhodes (1978), Measuring the Efficiency of Decision-Making Units, European Journal of Operations Research, 2, 429-444.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

102 - TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL Coelli, T.J., P. Rao and G.E. Battese (1998), An Introduction to Efficiency and Productivity Analysis, Boston, Kluwer Academic Press. Coelli, T.J. (1996), A Guide to FRONTIER Version 4.1: A Computer Program. For Stochastic Frontier Production and Cost Function Estimation, Working Paper No. 7/96, Centre for Efficiency and Productivity Analysis, University of New England, Armidale, Australia. Cornes, R. (1992), Duality and Modern Economics, Cambridge, Cambridge University Press. Fernandes, E. and R.R. Pacheco (2002), Efficient Use of Airport Capacity, Transportation Research: Part A - Policy and Practice, 36 (3) 225-38. Farrell, M.J. (1957), The Measurement of Productive Efficiency, Journal of the Royal Statistical Society, Series A, 120 (3), 253-290. Gillen, D. and A. Lall (2001), Non-Parametric Measures of Efficiency of US Airports, International Journal of Transport Economics, 28 (3), 283-306. Gillen, D. and A. Lall (1997), Developing Measures of Airport Productivity and Performance: An Application of Data Envelopment Analysis, Transportation Research Part E, 33 (4), 261-273. Hooper, P.G. and D.A. Hensher (1997), Measuring Total Factor Productivity of Airports: An Index Number Approach, Transportation Research Part E - Logistics and Transportation Review, 33 (4), 249-59. Humphreys, I. and G. Francis (2002), Performance Measurement: a review of airports, International Journal of Transport Management, 1, (2), 79-85. Jensen, M.C. and W. Meckling (1976), Theory of the firm: managerial behaviour, agency costs and capital structure, Journal of Financial Economics, 3, 305-60. Jessop, A. and C.P. Barros (2004), Ranking Airports’ Efficiency: Comparing DEA and Blockmodels, Working Paper CIEF No. 10/2004. Jones, L.R., P. Tandon and I. Vogelsang (2000), Selling public enterprises: A cost-benefit methodology, Cambridge, MA, MIT Press. Kumbhakar, S.C. (1987), Production frontiers and panel data: an application to US Class 1 railroads, Journal of Business and Economic Statistics, 5, 249-255. Khumbhakar, C. and C.A.K. Lovell (2000), Stochastic frontier analysis, Cambridge, UK, Cambridge University Press. Malmquist, S. (1953), Index Numbers and Indifference Surfaces, Trabajos de Estatistica, 4, 209-42. Meeusen, W. and J. Van den Broeck (1977), Efficiency estimation from a Cobb-Douglas production function with composed error, International Economic Review, 18, 435-444. Mueller, D.C. (1979), Public Choice, Cambridge, Cambridge University Press.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TECHNICAL EFFICIENCY IN PORTUGUESE AIRPORTS WITH A STOCHASTIC COST FRONTIER MODEL -

103

Murillo-Melchor, C. (1999), An Analysis of Technical Efficiency and Productive Change in Spanish Airports using the Malmquist Index, International Journal of Transport Economics, 26 (2), 271-92. Olson, M. (1965), The logic of collective action: Public goods and the theory of groups, Cambridge, MA: Harvard University Press. Parker, D. (1999), The performance of BAA before and after privatisation, Journal of Transport Economics and Policy, 33, 2, 133-146. Pels, E., P. Nijkamp and P. Rietveld (2003), Inefficiency and scale economics of European airport operations, Transportation Research Part E, 39, 341-361. Pels, E., P. Nijkamp and P. Rietveld (2001), Relative Efficiency of European Airports, Transport Policy, 8, 183-192. Porter, M.E. (1979), The structure within industries and companies’ performance, The Review of Economics and Statistics, 61: 214-227. Rumelt, R. (1991), How much does industry matter?, Strategic Management Journal, 12, 2, 167-185. Sarkis, J. (2000), Operational Efficiency of Major US Airports, Journal of Operation Management, 18, (3), 335-251. Sarkis, J. and S. Talluri (2004), Performance-based clustering for benchmarking of US airports, Transportation Research Part A, 38, 329-346. Teece, D., G. Pisano and A. Shuen (1997), Dynamic capabilities and strategic management, Strategic Management Journal, 18, 7, 509-533. Tullock, G. (1967), The Welfare Costs of Tariffs, Monopolies, and Theft, Western Economic Journal, 5, 224-232. Yoshida, Y. and H. Fujimoto (2004), Japanese airport benchmarking with DEA and endogenousweight TFP methods: testing the criticism of overinvestment in Japanese regional airports, Transportation Research Part E, 40, 533-546. Yoshida, Y. (2004), Endogenous-weight TFP measurement: methodology and its application to Japanese airport benchmarking, Transportation Research Part E, 40, 151-182. Wernerfelt, B. (1984), A resource-based view of the firm, Strategic Management Journal, 5, 2, 171-180.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

Cover_f.fm Page 1 Wednesday, April 7, 2004 11:00 AM

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES: SOME THEORY AND MAIN POLICY USES

Antonio ESTACHE World Bank Washington D.C. USA and ECARES Université Libre de Bruxelles Belgium

Lourdes TRUJILLO Universidad de Las Palmas de Gran Canaria Las Palmas Spain

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

105

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

107

SUMMARY

ABSTRACT ........................................................................................................................................ 109 1.

INTRODUCTION ....................................................................................................................... 109

2.

WHAT’S SO SPECIAL ABOUT TRANSPORT?...................................................................... 110

3.

WHY SHOULD POLICYMAKERS CARE ABOUT TRANSPORT COSTS?......................... 111

4.

FROM POLICY CONCERNS TO MEASURES........................................................................ 113

5.

HOW TO DECOMPOSE THE SOURCES OF COST EFFICIENCY CHANGES ................... 117

6.

FROM THEORY TO PRACTICE .............................................................................................. 119

7.

CONCLUDING COMMENTS ................................................................................................... 120

NOTES ................................................................................................................................................ 121 BIBLIOGRAPHY ............................................................................................................................... 123

Washington/Gran Canaria, December 2005

This paper was prepared for the December 2005 ECMT Round Table on the Estimation and Evaluation of Transport Costs in Paris. It has benefited from comments made during the meetings by participants. We are particularly grateful to Andreas Kopp and Professors Madanat, Oum and Quinet for useful discussions ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

109

ABSTRACT

This paper provides a brief overview of key dimensions of costs and of the measurement of their changes in the context of major economic policy decisions in the transport sector. It emphasizes the policy relevance of the fact that the general concept of costs includes dimensions of performance that are under the control of operators as well as dimensions which cannot be controlled by them. The ability of policymakers to separate these two types of dimension is essential when trying to provide incentives to operators to improve their efficiency performance, while protecting the interests of users and taxpayers. There are useful techniques to assess changes in efficiency but their implementation can be a challenge, as discussed in this paper. These challenges can be overcome. Indeed, the technical ability to generate aggregate and unbundled cost efficiency estimates can and has been used by policymakers to shift the burden of proof for the justification of poor performance onto the operators in imperfect markets, as part of standard interactions in the context of regulatory or competition policy proceedings.

1. INTRODUCTION

In any transport economics textbook or journal, costs are likely to feature pre-eminently, both conceptually and empirically1. Costs have indeed long been and continue to be central to most of the crucial transport policy debates. They guide the discussions on financing or on subsidy requirements from the OECD countries to developing countries. They also guide many of the debates on the limits to growth or to competitiveness. For instance, underinvestment in road networks, due to limited fiscal capacity to meet road construction and maintenance costs, is widely viewed as one of the impediments to Africa’s growth and to that of some countries in Central Asia. High logistics costs are also seen as one of the sources of comparative disadvantage in Latin America, when compared to East Asia for instance. In addition to these links between growth and competitiveness, transport costs also have a major social dimension. In many countries transport costs are an essential source of inequity among users. The poor tend to spend much more time in transport to access their jobs. Many policies aiming at reducing transport costs thus end up being quite progressive, with high social payoffs. This also means that policies which tolerate high cost solutions, when more cost-effective ones are available, tend to be quite regressive. For instance, in developing countries, the health costs associated with high pollution effects tend to harm poor children significantly more than the children of the rich, who can afford to travel out of polluted cities on weekends to recover. In view of this central role of costs and their various dimensions, it is somewhat surprising to see the many disagreements that continue to prevail in the transport community. There is indeed no clear ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

110 - TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES consensus on the actual policy use of good conceptual knowledge, in particular the correct measurement of costs. From an academic viewpoint, this may be illustrated as follows. While many academics suggest that there is a need for a major review of the power rules obtained by statistical analysis of 1960s data, these rules are still used in today’s publications and inform policy decisions. At the policy level, one of the most persistent failures in Europe may be seen in the port sector. Many policymakers are quite familiar with the fact that port costs are central to international competitiveness, and yet these costs are a “black box”2. In particular, it is surprising to see that the ports sector is the only one that has not yet received a clear policy decision from the European Commission3. Indeed, the incentive to cut transport costs by relying more on short sea shipping is still far from what it should be for many transport operators, in spite of the mode’s strong potential for many products. Perhaps even more shocking is the fact that, in most countries, there is not enough data to make a fair and convincing assessment of the potential financial, congestion and pollution payoffs associated with the shift from road and rail to maritime transport. The rest of the paper is organised as follows. First, we compare some stylised facts on transport costs data with the equivalent facts in other infrastructure sectors. Second, we review the main policy motivations for the measurement of costs. Chapter 3 discusses how policymakers try to address their concern for cost measurement in practice. In Chapter 4, we briefly summarise the main techniques available to policymakers to assess cost levels, productivity and efficiency. In the process, we show how some current practice can be misleading in policy decisionmaking. Chapter 5 discusses how changes in the performance of a service provider can be unbundled into its various sources. Chapter 6 explains how policymakers have moved from theory to practice in the field of cost level and cost efficiency measurement. Chapter 7 concludes.

2. WHAT’S SO SPECIAL ABOUT TRANSPORT?

Why is the transport sector so different from telecoms or electricity distribution when it comes to getting operators to generate costs data relevant to fundamental, strategic policy decisions? The usual argument is that intermodal competition in transport tends to be strong enough to keep costs down. This sort of argument ignores captive shippers and collusion, for instance, which are so well documented in some sectors – e.g. maritime transport and trucking. Such risks are much less ignored in other infrastructure sectors that also suffer from certain restrictions to competition. Cost measurement is a major concern in the telecom sector, which is at least as competitive as the transport sector. Already about twenty years ago, the concern for cost measurement in that sector led to a major reform of US telecoms which forced the regulators and operators to rely on costs models to assess the costs of universal service obligations or to assess access costs. The US model has since been copied or adapted in a large number of OECD and developing countries. These costs models have resulted in cost cuts from between 10% to 40% in a wide range of countries quite independent of the cost savings associated with technological progress4. Cost measurement is also a major part of the reforms in energy utilities, in particular, when regulatory regimes that are intended to cut costs are introduced as part of the reform. Regulatory cost benchmarking is common in these sectors in most Anglo-Saxon countries and in many northern European countries. It is now also starting to be carried out in the water and sanitation sector, although much less so than in the other sectors.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

111

There is thus no obvious explanation for the poor measurement of costs and policy use of the information available in the transport sector, in particular when compared to the practice of other public services. There are a number of interesting correlations – we did not write causality! – between the poor measurement of costs and the odds of conflicts in infrastructure. In a study of about one thousand infrastructure concessions contracts signed in Latin America, Guasch (2004) shows that over 50% of the transport contracts were renegotiated, much more than for telecoms and electricity, where very few contracts were subject to trouble between 1990 and 2002. Estache et al. (2004) show that in most cases, these renegotiations resulted in cost increases. Flyvbjerg et al. (2002, 2004) find equivalent results for a large number of transport contracts in OECD countries. Costs overruns are much more common than expected in large transport projects. Are they justified? Flyvbjerg and his various co-authors argue that failed procurement rules and over-optimistic demand, in a sector with significant scale economies, are to blame – i.e. the larger the promised demand the lower the promised average unit costs, but when demand does not materialise, the average unit costs end up higher than expected. All these studies fuel the case for a debate on transport infrastructure and service costs and the strategic use of measures by providers with some market power. We will return to this concern later in the paper. But before addressing the process issues, it is perhaps useful to revisit and take stock of the most common policy uses of cost measures, and of the challenges faced in bridging the gap between theory and practice in this core field of transport economics and policy.

3. WHY SHOULD POLICYMAKERS CARE ABOUT TRANSPORT COSTS?

The policy uses of transport costs data can be found in three broad contexts: • • •

Regulation; General competition policy; Yardstick competition.

The regulatory use of cost data is probably the best known and best documented. In spite of the widespread deregulation of services, the sector continues to count many activities provided by monopolies, and hence tariff regulation is still an important responsibility of the state. Regulating monopolies is needed to avoid situations in which operators charge too much or provide too little service. Both concerns require an assessment of costs. These costs can be simply financial costs, when captive shippers are threatened by potentially abusive transport service providers, but also economic costs, when congestion or pollution is a major concern. Costs may also have to be measured by a regulator when subsidies are needed to cover the expenses associated with service obligations imposed on operators. In the case of passenger services, for instance, service obligations – i.e. maintaining traffic where demand is not sufficient to cover costs – are quite common in the railway and bus industries, and are often financed by direct or indirect subsidies. To avoid abusing the taxpayers or other users’ contributions to the financing of these services, costs measurement is once again essential. Returning to the textbook coverage of these issues, it is interesting to note that the concept of costs to be recovered is sometimes, but not often, addressed in much empirical detail5. Of particular interest here is the fact that the cost to recover or subsidise should be the efficient cost level. As widely ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

112 - TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES documented in the more technical literature, observed costs and efficient costs are not necessarily the same6. This is why some regulators allocate so many resources to assess what these efficient cost levels should be. This concern for efficiency has in fact, for some ten to fifteen years now, been formalised by many countries in their choice of regulatory regimes. Incentive-based regimes, aimed at specifically promoting efficiency such as price or revenue caps, have been adopted in a wide range of countries and sectors. These regimes are associated with recurrent reviews of cost levels to assess their efficiency and ensure, on a regular basis, that the efficiency gains and cost savings achieved by monopolies are eventually shared with the users rather than simply transformed into a monopolistic rent7. In that context, regulators also try to ensure that scale, scope, density economies or technological progress are internalised by operators. While they are still a much more common practice in Anglo-Saxon countries (the UK or Australia in particular) and in certain sectors (electricity and telecoms), there are many examples around the world for the transport sector: railways in Argentina, Brazil and Mexico or ports in Mexico, Peru or most of the largest Asian actors are relatively well-documented examples in the literature. Many privatised airports or ports are subject to some sort of incentive-based regulatory regime. So are many toll roads. Furthermore, it would seem that the maximum price regulation of commodities in railway regulation is a system which could be used to push incentives, as with the more standard price caps. However, in many countries, these maximum prices in the railways sector are never binding, simply because the long-run marginal cost calculations on which they are based rely on outdated cost levels and technological options. This is a common issue in the pricing of railway services in developing countries, for instance8. The use of cost measurement in the context of competition is well recognised among practitioners as well as their more standard regulatory uses. The evolution of costs is central to any assessment of risks associated with reduced competition in a market. Mergers are probably the most commonly known example of such a risk. A recent illustration within a sector is provided by Argentina’s thorough assessments of the mergers between two port terminals in Buenos Aires, based on extensive yet confidential costs data9. But there is also a risk of reduced intermodal competition. In Brazil, trucking companies have become co-owners of some of the railway operators with a view to stimulating modal integration. While the policy seems rational, some observers expressed concerns about the reduction in competitive options that this could represent for some shippers. It could also be a way of developing sources of cross-subsidies, allowing predatory pricing on some business lines. Finally, reduced competition in the market can also come from collusion. Cost assessments are then an essential component of any analysis aimed at assessing the likelihood of collusion between operators or between an operator and a government agency. There are many cases of concern about collusion between trucking companies to override the railway operators, which cannot be assessed simply because the costs data is not available. But competition issues not only arise in the market but also for the market. The best example of competition issues in which costs have been the main concern is provided by the assessment of the effectiveness of public procurement rules. This concerns the assessment of the extent to which the competition for the market is consistent with the delivery of a given service of a given quality level at the most cost-effective level. It is to protect against the risk of future under-provision of quality and cost effectiveness that the Mexican Competition Authority is asked to assess the main bids for new transport projects in that country. More generally, yardstick competition or comparative benchmarking of the cost performance of operators is increasingly tempting for policymakers. Indeed, if the productivity gains of a specific monopolistic operator allowed by a regulator are based mostly on gains achieved by this firm in the past, this operator may not have strong incentives to improve efficiency to cut costs. Comparative ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

113

performance assessments provide additional information on the potential efficiency gains each operator could conceivably achieve, while recognising that not all operators face the same conditions. It can also be used to generate average, sector-specific performance improvement estimates, which provide good benchmarks and can certainly be used in negotiations with each individual operator. Thus, in practice, costs are at the core of the implementation of policies to promote competition in and for the markets in rail, ports and airports in many countries. Increasingly also, there is an interest in relying on cost data to promote competition between markets. Progress on this front is slow, however, at least in the transport sector. Indeed, while yardstick competition has many supporters among academics and has been used in the water and energy sector to compare the performance of various regional operators within a country or national operators across countries, there are not many examples of countries which have adopted the policy in practice. The Mexican port sector has relied on yardstick competition assessments to inform its port infrastructure tariff reviews10. Similar monitoring efforts have been conducted in Brazil to compare the performance of the various regional railway operators11. A recent proposal was made to introduce yardstick competition for regional passenger railway operators in France, to reveal the high costs of the sector and to stimulate their reduction12. Overall, the main message of this chapter is that costs are the core ingredient of the most common policy issues that governments need to be concerned with. Whether countries are developed or developing is quite irrelevant. It may be worth pointing out that these are not the only policy areas in which transport costs are essential. Costs are at the interface of transport policy and environmental policy. They are also essential in the design of any policy aimed at addressing congestion problems, as clearly illustrated by the recent introduction of a congestion charge in London. While these are important areas in which costs are central, they go beyond the scope of this paper and we will thus limit ourselves to highlighting the fact that any full costs analysis needs to address these interfaces.

4. FROM POLICY CONCERNS TO MEASURES

The discussion so far has suggested that policymakers are faced with two main types of measurement concerns: • •

Measuring actual costs levels; Assessing their efficiency.

To get to costs data, it would be naïve to believe that accounting rules generate the kind of data needed to address policy issues such as those discussed in Chapter 3. Even if international accounting standards are starting to converge, they are not designed to allow assessment of the extent to which market failures or violation of competition result in unfair or inefficient outcomes. In the context of regulated industries, it is quite standard to impose on monopolistic operators the adoption of regulatory accounting rules. These allow a separation between the cost centres associated with monopolistic activities to be monitored and those associated with competitive activities, which should 13 be left alone by regulators . This is probably the most cost-effective solution to generate observable data on current cost levels. The widespread use of analytical accounting is a huge step in the right direction. The recognition that the data associated with the delivery of monopolistic activities should ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

114 - TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES be made transparent to regulators and competition agencies is the next step to take. However, the transport community has not been willing to take it. Consider as basic evidence the fact that the international statistics on government finances produced by the IMF are at present essentially unable to generate international comparisons of public expenditures in the transport sector. Most of what is published on that front suffers from a basic measurement flaw! To get to efficiency measures, policymakers need to know a lot more than accounting data. They need to know something about the physical side of the business; about the physical outputs of the operators (volume and type of freight or passengers); the types and levels of inputs (employment, equipment, intermediate inputs such as energy purchases); and the pricing of both outputs and inputs. They need to monitor the quality of the services used and purchased. To be able to compare the efficiency performance across operators (and often across time as well), policymakers must be able to account for the relevance of the physical environment in which the services are being delivered, which are likely to drive costs and engineering or economic options (e.g. rocky mountains or flat land; based on an island or in a landlocked country?). They finally need to know the regulatory environment in which the service is being delivered (vertically integrated or not, subject to what type of regulation, etc.). Based on that information, any policymaker should be able to generate an indicator allowing an assessment of the changes in costs while recognising that these changes can be influenced by many factors, some controllable by the operator, some not. While a lot of the necessary information is missing, as a result of the poor accounting practices mentioned earlier, a relatively robust assessment of the evolution of the productivity of the operator can be obtained from data on the production side of the business, generated by standard engineering monitoring processes. In other words, with relatively little data on cost, a policymaker can get a sense of the cost performance and the costs savings achieved or achievable by simply generating a synthetic indicator from the production side of the business. Increasingly, the new generation of policymakers is recognising that these synthetic productivity indicators are much more robust and reliable than the partial performance indicators historically used in the profession. Box 1 shows how partial indicators can mislead the ranking of performance. The most common synthetic indicator is Total Factor Productivity (TFP). It measures how the outputs of an operator are associated with the combination of inputs used by that operator. With one output and one input, TFP is simply the usual formula used to assess partial performance indicators such as labour productivity (labour divided by output). When there is more than one input (and/or output), this calculation requires weights to be specified for each input and output. The standard practice is to use the price of these inputs and outputs as weights. The most common and easiest way of measuring TFP is to rely on index number approaches. Malmquist, Tornqvist or Fisher formulae are usually relied on14. Additional transformations are needed when these approaches are applied to cross-sectional data to ensure transitivity of comparisons. The methods are quite simple to implement. They generate a measure of TFP with only two data points – say, data on two firms or data on one firm at two points in time. Their main drawback is that they require information not only on quantities but also on prices (or values) of inputs and outputs. This can be quite challenging when multi-country comparisons are considered.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

115

Box 1: Who’s the best performer? How Partial Factor Productivity Can Mis-inform Policy Decisions While the use of partial productivity indicators is quite common in the literature and on the ground, it may be useful to point out its limitations. They may for instance result in overcapitalisation in order to increase an indicator such as labour productivity. The partial approach ignores the relevance of other inputs and of the possibilities of substitution between inputs. To illustrate these issues, it may be useful to rely on an example. Consider two port operators, A and B. They are operating container terminals and they handle the same volume of output, say 2 000 containers per day. Operator A is more capital intensive while operator B relies more on workers. Which one is the most efficient? Table B1 compares the partial performance indicators of these two operators with respect to labour and to capital. Operator A is the most efficient if the assessment is made on the basis of the use of labour, as is the usual case. However, operator B is the most efficient if the assessment is made in terms of the use of capital. In sum, relying on partial indicators does not generate a clear performance ranking. Operator A B

Labor (L) 200 400

Table B1. Partial indicators Capital (K) Output (Y) 2 2000 1 2000

Y/L 10 5

Y/K 1000 2000

This is not to say that partial performance indicators are not useful. They are, and will continue to inform decisions very effectively under many circumstances for a given operator. But they are a weaker tool to inform regulatory decisions, in particular decisions which require an integrated vision of the sector and comparison of performance across operators.

While TFP provides a reliable synthetic assessment of the cost or production performance of an operator, for some policy analysis needs, it may be too synthetic. It is thus important to be able to unbundle the sources of changes in TFP. The relevance of this decomposition should not be underestimated. Indeed, the broad impression that efficiency should be one of the main focuses of transport sector performance assessments may, for some more casual users of the information, hide the fact that efficiency is in fact a multidimensional concept. Each concept represents a specific dimension of the transport business and can be influenced by very different types of policy instruments. There are five main concepts of efficiency in the literature: • • • • •

Technical efficiency Scale efficiency Technological change Allocative efficiency Cost efficiency15.

Technical efficiency (TE) is the ability of a firm to achieve maximum output, given its input set. TE scores vary between 0 and 1. A value of 1 indicates full efficiency and operations are on the production frontier. A value of <1 reflects operations below the frontier. The wedge between 1 and the ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

116 - TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES value observed measures technical inefficiency. This is an output-orientated efficiency measure. An input-orientated technical efficiency measure reflects the degree to which a firm that must produce a particular output level, y, could proportionally reduce its input usage, and still remain within the feasible production set. This is the efficiency that operators should be able to act upon most effectively and is typically the main focus of sector regulators. In standard regulatory procedures, this is how the efforts of operators to “catch up” with the average or best performance of the field is measured. Changes in technical efficiency are said to measure the “catching up” by an operator. Technological change (TC) (or technological progress) is the increase in the maximum output that can be produced, given an input vector, x, reflected in a shift in the production function/frontier over time. This is often slow for utilities and transport, with the exception of the telecommunications sector where progress has been, and continues to be, dramatic. Ensuring that operators make the most of its evolution is also a common concern of sector reformers and regulators. Changes in TC are used in regulatory proceedings to measure the “shift” in the production or cost frontiers, and need to be distinguished from the “catching up” effects. Scale efficiency (SE) is a measure of the degree to which a firm is optimising the size of its operations; a firm can be too small or too large, resulting in a “productivity penalty” associated with not operating at the technically optimal scale of operation. This is one of the best known and yet trickiest dimensions from a policy viewpoint. First, it is not unusual to yield on scale economies to try to achieve significant increases in other sources of efficiency by promoting competition between the unbundled profit centre of an historically integrated monopoly. This is common in railways, ports or airports, with the replacement of a national monopoly by regional or local monopolies. Second, scale efficiency is often driven by demand. Demand shocks often cannot be controlled by operators (e.g. the East Asia crisis) and hence penalising operators for reductions in efficiency associated with this sort of shock is not a desirable policy decision16. Allocative efficiency (AE) is of two types. The first is “input-mix allocative efficiency”. This corresponds to the ability of a firm to select the correct mix of input quantities so as to ensure that the input price ratios equal the ratios of the corresponding marginal products (i.e. the additional output obtained from an additional unit of input). The second type is “output-mix allocative efficiency”. This corresponds to the ability of a firm to select the output mix that would prevail without interference from, for example, price controls, output tax, direct or cross-subsidy-driven distortions. The AE score varies between zero and one, with a value of 1 indicating full allocative efficiency. In most microeconomics textbooks, it is assumed that all firms are technically efficient – in that special case, full allocative efficiency equates to full cost efficiency or cost minimisation. This is one of the dimensions which is not necessarily controlled directly by an operator. Wage and employment agreements are often signed at the sector or even national level rather than at the firm level and are thus taken as an input by operators rather than a variable they can be expected to adjust in order to improve efficiency. The cost of allocative inefficiency is a key component of any approximation of the costs associated with the fact that some transport sector users are held hostage by certain interest groups in many countries. The poor cost data on many of the monopolistic transport sector activities is probably not unrelated to the control of the sector by these interest groups and their unwillingness to generate the information that would make them accountable. Cost efficiency (CE) is the ability of the firm to produce a particular output, y, at minimum cost, given the input prices it faces. Note that CE=AE×TE, and hence that CE varies between zero and one, with a value of one indicating full cost efficiency. This is the main focus of most regulators and competition agencies. It is crucial to keep in mind that it is composed of two dimensions which are subject to very different degrees of direct control by the operators. ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

117

The policy upshot of this multidimensionality may be summarised in three observations. The first is that the TFP of two firms facing the same operating environment (at one point in time) can differ because of TE, AE or SE differences. Over time, TFP can in addition vary due to changes in TC. The second is that operators do not have full control over each of the dimensions of the business and that it can be irrational to demand improvements for dimensions that operators cannot control. The third is that the restructuring of a sector often impacts on the various efficiency dimensions in very different ways, a fact often ignored by reformers.

17

5. HOW TO DECOMPOSE THE SOURCES OF COST EFFICIENCY CHANGES

The main measurement methods used to unbundle productivity and conduct sources of inefficiency studies are: • •

data envelopment analysis (DEA); and stochastic frontier analysis (SFA).

DEA is a linear programming (LP) method which constructs a non-parametric production frontier by fitting a piece-wise linear surface over the data points. SFA is an econometric method which estimates a cost or a production frontier18; alternatively, a distance function can be used. The basic intuition is simple: these methods tell the analyst the maximum output or the minimum cost that could be achieved. Comparing observed costs or production to these optima provides a sense of the inefficiency of the operators. The methods generate information which can also be used to decompose these inefficiencies into their various sources19. Table 1 offers a brief summary of these methods, of their advantages and their drawbacks. There are a few rules-of-thumb used in practice to pick among these methods. SFA and DEA do not require price data (except for the case of SFA cost frontiers), however, they do require information on a large number of firms to obtain statistically reliable estimates of the frontiers involved. If SFA or DEA is used to measure TFP, the regulator is making the implicit assumption that the shadow prices20 are equal to the observed prices. That is, one assumes all firms are allocatively efficient. If the assumption is wrong, the SFA and DEA methods will not pick up any change in allocative efficiency and may lead to misleading decompositions of the sources of TFP change21. In the early stages of a new regulatory regime, allocative efficiency improvement could be a significant component of the realised TFP improvements. In sum, the choice between these various methods to assess TFP differences or changes, as well as their composition, leaves some flexibility to the policymakers when facing different constraints. This explains why in the applied literature, the same topic can be treated in many different ways and why it is crucial for the regulator to be quite explicit about its methodological preferences. Creativity in methodological choice is in fact quite common among theorists who often use these methods in combination22. DEA could be used instead of SFA; however, for various reasons, including the ability of SFA to deal with noise and accommodate standard statistical tests, we recommend that SFA be used in most cases.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

118 - TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES Table 1. Comparing DEA and SFA measures of cost efficiency

Data needs

Advantages

Drawbacks

Data Envelopment Analysis (DEA) Quantity data on inputs and outputs and their costs for a sample of firms – ideally over a number of years.

Identifies a set of peer firms (efficient firms with similar input and output mixes) for each inefficient firm. Can easily handle multiple outputs. Does not assume a functional form for the frontier or a distributional form for the inefficiency error term. May be influenced by noise. Usual hypothesis tests are impossible. Requires large sample size for robust estimates –this may not be available early on in the life of a regulator.

Stochastic Frontier Analysis (SFA) For a long-run cost frontier: total costs, input prices and output quantities. For a short-run cost frontier: variable costs, variable input prices, fixed input quantities and output quantities. If cost data is not available, an approximation of the operator’s performance can be obtained from a production frontier or distance function which requires quantity data on inputs and outputs for a sample of firms – ideally over a number of years. Attempts to account for noise – i.e. poor data quality. Environmental variables easier to deal with. Allows for the conduction of traditional statistical tests of hypotheses. Easier to identify outliers. Cost frontiers and distance functions can deal with multiple outputs. The decomposition of the error term into noise and efficiency components may be affected by the particular distributional forms specified, and by the related assumption that error skewedness is an indication of inefficiency. Requires large sample size for robust estimates – this may not be available early on in the life of a regulator.

In the end, whatever the method used, any TFP decomposition will be an approximation. The information problems are likely to be so large that the best a policymaker can hope to do is to come up with a set of figures that can serve as a discussion opener. The correct cost efficiency measure and its decomposition is likely to be influenced by many factors. These will be raised by the operators during their discussion with regulators or competition agencies. But the computation of this cost efficiency estimate by a policymaker has the effect of shifting the burden of proof onto the operator. This is a major change in policy approach since, under more traditional regulatory processes, operators functioning in imperfect markets have tended to benefit from policymakers’ limited access to information on their performance.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

119

6. FROM THEORY TO PRACTICE

Now that it is clear that cost efficiency and specific cost levels are often going to be approximations, often rough, it is important to keep in mind the sources of “roughness”. On the cost side, arbitrary rules are typically relied upon to assign joint costs across the various business units of an operator. On the production side, it is almost impossible to consider all inputs and often all outputs in the assessment of the evolution of productivity. For outputs, most studies focus on variables such as tonnes-km, passengers-km, containers movers, bulk load, etc. For inputs, most studies focus on the key inputs which tend to drive the largest share of the costs of an operator. Typically, this means that labour, capital and energy inputs are the centre of attention and, when needed, any other input that is believed to be a major cost driver. Why does theory sound simpler than it is in practice? First, most of the inputs are fairly heterogeneous. Labour can cover a wide range of skills and hence of wages and the relative weight of these various types of employees may be important information which is not always available. In most cases, analysts have an idea of people’s total wages and of the total employment level and not much else. Similarly, the capital variable reflects quite a wide diversity of assets. From rolling stock to tracks in the railway industry, from cranes and their types to storage facilities and their sizes, from docks to the types of port terminal, the possible refinements seem quite numerous. The ages, composition and history of vehicle use by operators, for instance, can all be revelant. The ideal disaggregation is driven by the type of issues to be assessed by the analysis. Second, operators do not function in a vacuum. There are many more dimensions to take into account and a good diagnosis of the market structure and its characteristics is often the first step to take in any assessment of an operator’s efficiency. The diversity of factors that can be important can sometimes be overwhelming, and analysts need to be able to make judgement calls on diverse matters: the relative importance of domestic and international traffic, the geography (landlocked vs. coastal, island vs. continental, etc.) and the composition of users (industrial, residential, etc.). Quality and its dimensions can also be relevant, in assessing the efficiency of an operator and because of its multidimensionality, often ignored in cost assessments (see Box 2 for an illustration of the relevance of quality in efficiency measures). It is quite well known from economic theory that quality is an adjustment variable which can be oversupplied under cost-plus or rate-of-return regulatory regimes, while it can be undersupplied under incentive-based regimes23. The point is that under either regulatory regime, cost efficiency estimates may be biased if the adjustment in quality (i.e. delays, interruptions, security, safety/accidents/incidents) are not accounted for24. The effectiveness of researchers and policymakers in dealing with these issues and in generating policy-relevant data varies significantly across sectors. The rail sector and the bus industry25 have been quite well covered and, increasingly, ports and airports, although still very much behind the coverage available for rail or buses. For the road sector, most of the useful information comes from engineering analyses rather than from efficiency studies. Ultimately, the quality of the costs assessments and of operators’ performance, whether public or private, greatly depends on the quality of the data available. With poor data, the specification of the ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

120 - TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES cost or productions functions tends to be imperfect and most of the analyses conducted thus tend to be biased. Once more, the point for practitioners here is not to look for academic perfection in the short run but to force operators, with residual monopoly power and with a control of information not shared with the regulators or other policymakers, to reveal some of the information they control. The reduction in information asymmetry allowed by these techniques can be very successful in ensuring that monopolistic rents are at most temporary and eventually shared with the users. The interest of the users, with a fair treatment of the operators – and of the taxpayers who often end up subsidising excess costs – is the main reason why costs need to be assessed and analysed when monopolies are in charge of a transport service. Competition helps but is often not sufficient to eliminate all rents in the sorts of market structures characterising the airport, rail, road and port industries.

7. CONCLUDING COMMENTS

The overview provided by this paper is definitely broad-brushed, and possibly somewhat unfair to some of the very good work being done in academia and by some policymakers. It does however point quite clearly to a major gap between theory and practice on average. At the core of the problem is data, and the willingness of governments to generate the data needed to increase the accountability of the operators. It seems that this willingness has been decreasing in recent years, particularly in the transport sector. Ignorance about actual costs and their efficiency is a major problem for the users, since they tend to bear the bulk of the impact through quality and user fee adjustments. This is also a problem for taxpayers in general, since such a large share of costs in this sector end up being financed by fiscal revenue and hence they share in the costs of inefficiency. Finally, this is a serious problem for society as a whole in a sector known for its association with serious corruption risks, as documented by Flyvbjerg and his colleagues for a wider range of countries, both OECD and developing countries. Solutions to the problem are possible. They start with a willingness to consider policy relevant information as a priority by governments. They continue with the willingness to rely more systematically on increasingly sophisticated accounting practices, including non-intrusive regulatory accounting guidelines. This is not only about financing information to allow researchers to better contribute to improve societies’ options. It is about being fair to today’s and tomorrow’s citizens, as users and as taxpayers, while not penalising the operators who adopt fair costing practices. These concerns are well known among most academics, they are less well recognised in policy circles and when they are, they tend to lead to interesting discussions but little increase in commitments to address the issue. Looking ahead, the main challenge for the transport community in terms of awareness of the relevance of costs may be the willingness to move from good intentions to actions. In deciding whether to take on the challenge, it may be useful to keep in mind that information is necessary to shift the burden of proof for wrong cost levels onto the monopolistic operators and onto interest groups holding the transport sector users in hostage. This shift is much needed to ensure that the interest of users and taxpayers, in both the short and long runs, can be protected effectively.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

121

NOTES

1.

Good examples of textbooks characterised by a wide range of coverage of cost and related issues include Boyer (1998), Hensher and Brewer (2001), McCarthy (2001) or Quinet and Vickerman (2004).

2.

See Clarke, Dollar and Micco (2004) for a recent assessment of the competitiveness and trade costs of the poor management of maritime costs.

3.

See Tovar de la Fé, B., L. Trujillo and S. Jara-Díaz (2004); or Gonzalez, Tovar and Trujillo (2005) for recent discussion.

4.

Gasmi et al. (2002), for instance.

5.

This is in spite of the fact that one of the core references in the field, Waters, W.G. II (1976), started his seminal papers by accounting aspects of the analysis of costs.

6.

As pointed out by Profs. Madanat and Quinet in private conversations, the engineering literature on efficiency costs in the road sector, for instance, is probably well ahead of the curve compared to the economic literature.

7.

For a very useful discussion of the measurement of efficiency in a regulatory context, see Bernstein and Sappington (2002).

8.

In developing countries, the issue is however often irrelevant in competitive segments of the markets, simply because predatory pricing policies adopted by the trucking industry avoid the need for regulators to ensure that maximum prices have the right incentive effects on costs management behaviour.

9.

Serebrisky and Trujillo (2005).

10.

Estache, Gonzalez and Trujillo.

11.

Estache, Gonzalez and Trujillo.

12.

Leveque.

13.

Although this practice is much more widespread in the world for regulated utilities than it is for transport.

14.

For an excellent and very clear introduction, see Coelli , Rao and Battese (1998).

15.

For a much more detailed discussion, see Coelli et al. (1998 or 2003).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

122 - TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES

16.

See Estache, Tovar and Trujillo (2004) for a recent assessment of the impact of the East Asia crisis on the efficiency of Mexico’s Pacific ports. But the assessment of scale efficiency has a much longer history, including the seminal papers by Caves et al. (1980 and 1981) for the railways industry.

17.

This chapter is based on Coelli, Estache, Perelman and Trujillo (2003), to where the interested reader is referred for more details.

18.

Ordinary least squares (OLS) estimation of a frontier can be viewed as a special case of SFA, where one assumes that there is no inefficiency. Corrected OLS (COLS) estimation, where the OLS intercept is shifted, so that the frontier envelopes all data points, is also a special case of SFA, where one assumes that there is no noise.

19.

A useful recent survey of how the econometric methods are applied to the transport sector is provided by Braeutigam (1999) or McCarthy (2001).

20.

Shadow prices are a very important concept in this document. Essentially, the (input) shadow prices for a particular firm are that set of input prices that would ensure that the firm is operating at full allocative efficiency. The shadow prices are derived from the shape of the estimated frontier (at the point where the firm is operating). When observed prices and shadow prices are equal, the firm is allocatively efficient. When they differ, the firm will have some allocative inefficiency. That is, it is not operating at the minimum possible cost.

21.

This statement applies to the estimation of production frontiers. However when cost frontiers are involved, the issue becomes quite complex. Some allocative efficiency could be captured, but perhaps not all if there is systematic deviation from allocative efficiency by the industry.

22.

For instance, Kumbhakar and Lovell (2000) and Orea (2000) use PIN methods to measure TFP differences, and SFA methods to decompose these differences into the above components.

23.

See Tirole (1988) or Laffont-Tirole (1993), for instance.

24.

See, for instance, all the work done by Hensher and his various co-authors, as summarised in Hensher and Brewer (2001). More recently, see Estache, Perelman and Trujillo (2004).

25.

See Oum, Yu and Fu (2003) for a recent survey on airports; De Borger, Kerstens and Costa (2002) for the bus industry; Oum, Waters II and Yu (1999) for railways; Gonzalez and Trujillo (2006) for the port sector; and Estache, Perelman and Trujillo (2006) for a survey of the evidence from developing countries.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

123

BIBLIOGRAPHY

Bernstein, J.I. and D.E.M. Sappington (1999), “Setting the X Factor in Price-Cap Regulation Plans”, Journal of Regulatory Economics, 16:5-25. Boyer, K.D. (1998), Principles of Transportation Economics, Addison Wesley Publisher, Reading. Braeutigam, R.R. (1999), “Learning about transport costs”, in Gomez-Ibanez et al., Essays in Transportation Economics and Policy: Essays in Honor of John R. Meyer, The Brookings Institution, pp. 57-98. Caves, D.W. and L.R. Christensen (1980), “The Relative Efficiency of Public and Private Firms in a Competitive Environment : The Case of Canadian Railroads”, Journal of Political Economy, 80, 958-976. Caves, D.W., L.R. Christensen and J.A. Swanson (1981), “Productivity Growth, Scale Economies and Capacity Utilization in US Railroads, 1955-74”, The American Economic Review, 71, 994-1002. Clarke, X., D. Dollar and A. Micco (2004), “Port efficiency, maritime transport costs and bilateral trade”, Journal of Development Economics, 75, pp. 417-450. Coelli, T., A. Estache, S. Perelman and L. Trujillo (2003), A Primer on Efficiency Measurement for Utilities and Transport Regulators, World Bank Institute Publications, “Studies in Development” Series. Coelli, T.J., Prasada Rao, D.S., and Battese, G.E. (1998), An Introduction to Efficiency and Productivity Analysis, Kluwer Academic Publishers, Boston. De Borger, B., K. Kerstens and A. Costa (2002), “Public transit performance: What does one learn from frontier studies?”, Transport Reviews, 22 (10), 1-38. Estache, A., S. Perelman and L. Trujillo (2006), “Infrastructure reform in developing countries: evidence from a survey of economic performance measures”, in: T. Coelli and D. Lawrence (eds.), Performance Measurement and Regulation of Network Utilities, Edward Elgar Publishing, Cheltenham. Estache, A., B. Tovar and L. Trujillo (2004), “Sources of Efficiency Gains in Port Reform: A DEA Decomposition of a Malmquist TFP Index for Mexico”, Utilities Policy, December, Vol. 12, 4, pp. 231-230. Estache, A., J.L. Guasch and L. Trujillo (2003), “Price caps, efficiency pay-offs and infrastructure contract renegotiation in Latin America”, in: Bartle, I. (2003), The UK Model of Utility

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

124 - TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES Regulation: a 20th anniversary collection to mark the “Littlechild Report” retrospect and prospect, CRI Proceedings, 31, pp. 173-198. Estache, A., S. Perelman and L. Trujillo (2004), “Measuring Efficiency-Quality Trade-offs: An Application to Brazil’s Freight Railways”, VIIIth European Workshop on Efficiency and Productivity Analysis, Oviedo, September 2003. Flyvbjerg, B., M.K. Skamris Holm and S.L. Buhl (2004), “What Causes Cost Overrun in Transport Infrastructure Projects?”, Transport Reviews, Vol. 24, No. 1, pp. 3-18. Flyvbjerg, Bent, M.K. Skamris Holm and S.L. Buhl (2002), “Underestimating Costs in Public Works Projects: Error or Lie?”, Journal of the American Planning Association, Vol. 68, No. 3, Summer, pp. 279-295. Gasmi, F., D.M. Kennet, J.J. Laffont and W.W. Sharkey (2002a), Cost Proxy Models and Telecommunications Policy: A New Empirical Approach to Regulation, The MIT Press, Cambridge, MA. González, M. and L. Trujillo (2006), “La medición de la eficiencia en el sector portuario: revisión de la evidencia empírica”, Documentos de trabajo conjuntos ULPGC-ULL. Gonzalez, M., B. Tovar and L. Trujillo (2005), “Eficiencia de los puertos Europeos”, mimeo, ULPGCULL. Guasch, J.L. (2004), Granting and Renegotiating Infrastructure Concessions: Doing it right, WBI Development Studies, Washington, DC. Henscher, D.A. and A. Brewer (2001), Transport: an Economics and Management Perspective, Oxford University Press, New York. Kumbhakar and Lovell (2000), Stochastic Frontier Analysis, Cambridge University Press, Cambridge. Laffont and Tirole (1993), A Theory of Incentives in Procurement and Regulation, The MIT Press, Boston. McCarthy, P.S. (2001), Transportation Economics Theory and Practice: A Case Study Approach, Blackwell Publishers, Oxford. Orea, L. (2000), “A parametric decomposition of the generalised Malmquist-type productivity index”, paper presented at the Second Oviedo Workshop on Efficiency and Productivity, June 2000, Oviedo, Spain. Oum, T.H., C. Yu and X. Fu (2003), “A comparative analysis of productivity performance of the world’s major airports: Summary report of the ATRS global airport benchmarking research report-2002”, Journal of Air Transport Management, Vol. 9, No. 5, September, pp. 285-297 (13).

Oum, T.H., W.G. Waters II and C. Yu (1999), “A survey of productivity and efficiency measurement in rail transport”, Journal of Transport Economics and Policy, 33, No. 1:9, 42.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

TRANSPORT COST LEVELS, PRODUCTIVITY AND EFFICIENCY MEASURES -

125

Quinet, E. and R. Vickerman (2004), Principles of Transport Economics, Edward Elgar Publishing, Cheltenham. Tovar de la Fé, B., L. Trujillo and S. Jara-Díaz (2004), “Organization and regulation of the port industry: Europe and Spain”, in: P. Coto-Millan (ed.), Essays on Microeconomics and Industrial Organisation, Second Edition, Springer-Verlag Publishing. Tovar de la Fé, B., S. Jara-Díaz and L. Trujillo (2004), “Funciones de Producción y Costes y su Aplicación al Sector Portuario. Una Revisión de la Literatura”, Documentos de Trabajo Conjuntos. ULL – ULPGC. V 2004-06. Tirole, J. (1988), The Theory of Industrial Organization, MIT Press, Boston. Waters II, W.G. (1976), “Statistical Costing in Transportation”, Transportation Journal, pp. 49-62.

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT 2007

Cover_f.fm Page 1 Wednesday, April 7, 2004 11:00 AM

LIST OF PARTICIPANTS -

127

LIST OF PARTICIPANTS

Prof. Tae Hoon OUM University of British Columbia Faculty of Commerce and Business Administration 2053 Main Mall, Henry Angus 452 CND- VANCOUVER, BC, V6T 1Z2 CANADA

Chairman

M. le Professeur Philippe GAGNEPAIN Universidad Carlos III de Madrid Department of Economics Calle Madrid, 126 E-28903 GETAFE (MADRID) ESPAGNE

Co-Rapporteur (excused)

Prof. Marc IVALDI École des Hautes Études en Sciences Sociales IDEI Manufacture des Tabacs 21 Allée de Brienne bat. F F-31000 TOULOUSE CEDEX FRANCE

Co-Rapporteur

Prof. Piet RIETVELD Free University Amsterdam Faculty of Economics and Econometrics De Boelelaan 1105 NL-1081 HV AMSTERDAM PAYS-BAS

Rapporteur

Prof. Carlos PESTANA BARROS Institute Superior de Economica e Gestao Universidade Tecnica de Lisboa 20 Rua Miguel Lupi P-1200 LISBOA PORTUGAL

Rapporteur

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT, 2007

128 – LIST OF PARTICIPANTS Mr. Thorsten BECKERS Berlin University of Technology (TU Berlin) Workgroup for Infrastructure Policy (WIP) Secretariat H 33 Strasse des 17. Juni 135 D-10623 BERLIN ALLEMAGNE Monsieur Christophe DEBLANC CE/EC Office DM24 08/46 rue de la Loi 200 B-1049 BRUXELLES BELGIQUE Mme Christiane DELEPIERE-DRAMAIS Directeur de recherche Université Libre de Bruxelles Centre Interuniversitaire d'Étude de la Mobilité (CIEM) Avenue F. Roosevelt 50 (CP 194/7) B-1050 BRUXELLES BELGIQUE Mr. Knut Sandberg ERIKSEN Senior Research Economist Institute of Transport Economics (TOI) Postboks 6110 Etterstad N-0602 OSLO NORVEGE Mr. Antonio ESTACHE Banque Mondiale 1818 H Street, NW USA- WASHINGTON DC 20433 ÉTATS-UNIS Mr. Antonio ESTEVAN Ingenieur Industriel Gea21, S.L. Puerta del Sol 13, 5° E-28013 MADRID ESPAGNE

ESTIMATION AND EVALUATION OF TRANSPORTCOSTS – ISBN 92-821-0151-3 - © OECD/ECMT, 2007

LIST OF PARTICIPANTS -

Mr. Tom FERRIS Senior Economist and Head of Public Transport Planning Division Department of Transport Transport House 44 Kildare Street IRL- DUBLIN 2 IRLANDE Mrs. Ganna IVANOVA Chief Expert Foreign and Economic Relations Department Ministry of Transport and Communications 14 Peremohy Avenue UKR-KIEV 01135 U KRAINE Mr. Bruno JACQUES Director – Economic and Environmental Analysis & Research Transport Canada Place de Ville 330 Sparks Street CDN- OTTAWA K1A 0N5 CANADA Prof. Börje JOHANSSON Jönköping University Jönköping International Business School PO Box 1026 S-551 11 JÖNKÖPING SUEDE Mr. Charlie KARLSSON Jönköping University Jönköping International Business School PO Box 1026 S-551 11 JÖNKÖPING SUEDE Prof. Kiyoshi KOBAYASHI Graduate School of Urban Management Kyoto University Yoshida-Honmachi, Sakyo-ku J-606-8501 KYOTO JAPON

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT, 2007

129

130 – LIST OF PARTICIPANTS Mr. Niels Buus KRISTENSEN Director Danish Transport Research Knuth-Wintherfeldts Allé Building 116 Vest DK-2800 Kgs LYNGBY DANEMARK Mr. Samer MADANAT Director University of California at Berkeley Institute of Transportation Studies 109 McLaughlin Hall USA- BERKELEY, CA 94720-1720 ÉTATS-UNIS Prof. Silvia MAFFII TRT Trasporti e Territorio SRL Via Rutila, 10/8 I-20146 MILANO ITALIE Dr. Lance NEUMANN President Cambridge Systematics, Inc. 100 CambridgePark Drive, Suite 400 USA-CAMBRIDGE, MA 02140 ÉTATS-UNIS Dr. Jozsef PALFALVI Head of Division Institute for Transport Sciences (KTI) Than Karoly ut. 3-5 PO Box 107 H-1518 BUDAPEST HONGRIE M. le Professeur Emile QUINET Chef du Département École Nationale des Ponts et Chaussées Département d'Économie et des Sciences 28 rue des Saints-Pères F-75007 PARIS FRANCE

ESTIMATION AND EVALUATION OF TRANSPORTCOSTS – ISBN 92-821-0151-3 - © OECD/ECMT, 2007

LIST OF PARTICIPANTS -

Mr. Jeppe RICH Center for Traffic and Transport Technical University of Denmark Anker Engelunds Vej 1 Building 101A DK-2800 Kgs LYNGBY DANEMARK Prof. Robert RIVIER École Polytechnique Fédérale de Lausanne Institut des Transports et de la Planification Laboratoire d'Intermodalité des Transports EPFL - LITEP, Bât. GC CH-1015 LAUSANNE SUISSE Mrs. Iryna SADLOVSKA Deputy Director, Department for Financial Regulation and Social and Economic Policy Ministry of Transport and Communications 14 Peremohy Avenue UKR-KIEV 01135 UKRAINE Dr. Kuppusamy THIRUMALAI RSPA-Innovation, Research and Education U.S. Department of Transportation Room 7108 400 Seventh Street, SW USA- WASHINGTON, DC 20590-0001 ÉTATS-UNIS Prof. Lourdes TRUJILLO CASTELLANO Universidad de Las Palmas de Gran Canaria Faculdad de Ciencias Economicas y Empresariales Campus de Tafira E-35017 LAS PALMAS DE GRAN CANARIA ESPAGNE Mrs Pauline WORTELBOER-VAN DONSELAAR Ministry of Transport, Public Works and Water Management AVV Transport Research Team P O Box 1031 NL-3000 BA ROTTERDAM PAYS-BAS

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT, 2007

131

132 – LIST OF PARTICIPANTS Mr. Jerzy WRONKA Ass. Prof. D. Econ Research Institute for Transport Economics Korzeniowskiego 1 P. Box 996 PL-70-952 SZCZECIN 2 POLOGNE

OECD/ECMT SECRETARIAT

JOINT OECD/ECMT TRANSPORT RESEARCH CENTRE Dr. Andreas KOPP – Chief Economist JTRC/ECMT Dr. Michel VIOLLAND - Administrator JTRC/ECMT Ms. Françoise ROULLET - Assistant JTRC/ECMT Mrs Julie PAILLIEZ - Assistant JTRC/ECMT

ESTIMATION AND EVALUATION OF TRANSPORTCOSTS – ISBN 92-821-0151-3 - © OECD/ECMT, 2007

LIST OF PARTICIPANTS -

133

ALSO AVAILABLE

16th International Symposium on Theory and Practice in Transport Economics. 50 Years of Transport Research (2005) (75 2005 05 1 P) ISBN 92-821-2333-2 Time and Transport. Series ECMT – Round Table 127 (2005) (75 2006 01 1 P) ISBN 92-821-2330-8 National Systems of Transport Infrastructure Planning. Series ECMT – Round Table 128 (2005) (75 2005 10 1 P) ISBN 92-821-2341-3 Transport Services: The Limits of (De)Regulation. Series ECMT – Round Table 129 (2006) (75 2006 02 1 P) ISBN 92-821-2345-6 Transport and International Trade. Series ECMT – Round Table 130 (2006) (75 2006 13 1 P) ISBN 92-821-1338-8 Transport and Decentralisation. Series ECMT – Round Table 131 (2006) (75 2006 12 1 P) ISBN 92-821-1342-6 Transport Infrastructure Investment and Economic Productivity. Series ECMT – Round Table 132 (2007) (74 2007 04 1 P) ISBN 978-92-821-0124-7 (De)Regulation of the Taxi Industry. Series ECMT – Round Table 133 (2007) (74 2007 02 1 P) ISBN 978-92-821-0112-4 Market Access, Trade in Transport Services and Trade Facilitation. Series ECMT – Round Table 134 (2007) (74 2007 05 1 P) ISBN 978-92-821-0146-9 Transport Infrastructure Charges and Capacity Choice : Self-financing Road Maintenance and Construction. Series ECMT – Round Table 135 (2007) (74 2007 02 1 P) ISBN 978-92-821-0108-7

To register for information by email about new OECD publications: www.oecd.org/OECDdirect For orders on line: www.oecd.org/bookshop For further information about ECMT: www.cemt.org

ESTIMATION AND EVALUATION OF TRANSPORT COSTS – ISBN 978-92-821-0151-3 - © OECD/ECMT, 2007

Cover_f.fm Page 1 Wednesday, April 7, 2004 11:00 AM

OECD PUBLICATIONS, 2, rue André-Pascal, 75775 PARIS CEDEX 16 PRINTED IN FRANCE (74 2007 06 1 P) ISBN 978-92-821-0151-3 – No. 55783 2007

Page 1

13 6 EUROPEAN CONFERENCE OF MINISTERS OF TRANSPORT

ESTIMATION AND EVALUATION OF TRANSPORT COSTS Cost data for the construction and operation of facilities are essential for the evaluation of infrastructure services supplied by private or public providers.

Which methods can measure the efficiency of service provision and effectively benchmark providers? How do regulatory regimes impact on operators’ and infrastructure service providers’ cost levels? How can regulators counter the asymmetry of information as well as the incentives for data providers to selectively serve business rather than user interests?

www.oecd.org/publishing (74 2007 06 1 P1) ISBN 978-92-821-0151-3

-:HSTCSC=VUVZVX:

136

R O U N D TA B L E

These were the main questions discussed by the Round Table. Background papers were provided by Antonio Estache (World Bank) and Lourdes Trujillo (Universidad de Las Palmas de Gran Canaria), Piet Rietveld et al. (Free University of Amsterdam), Carlos Barros (Portugal) as well as Marc Ivaldi and Philippe Gagnepain (Université de Toulouse).

ESTIMATION AND EVALUATION OF TRANSPORT COSTS

What essential data do regulators need to ensure that transport serves the user best?

2007

C E N T R E

14:07

R E S E A R C H

20/08/07

T R A N S P O R T

fullTab136GB

ESTIMATION AND EVALUATION OF TRANSPORT COSTS

ROUND TABLE

13 6