New Directions in Regional Economic Development (Advances in Spatial Science)

Advances in Spatial Science Editorial Board Manfred M. Fischer Geoffrey J.D. Hewings Peter Nijkamp Folke Snickars (Coordinating Editor)

Charlie Karlsson Paul C. Cheshire

l

l

˚ ke E. Andersson A Roger R. Stough

l

Editors

New Directions in Regional Economic Development

Editors Professor Charlie Karlsson ˚ ke E. Andersson Professor A Jo¨nko¨ping University Jo¨nko¨ping International Business School Department of Economics Gjuterigatan 5 55318 Jo¨nko¨ping Sweden [email protected] [email protected]

Professor Paul C. Cheshire London School of Economics Geography & Environment Department Houghton St. London WC2A 2AE United Kingdom [email protected]

Professor Roger R. Stough George Mason University 4400 University Drive, MS 6D5 Fairfax VA 22030 USA [email protected]

Advances in Spatial Science ISSN 1430-9602 ISBN 978-3-642-01016-3 e-ISBN 978-3-642-01017-0 DOI: 10.1007/978-3-642-01017-0 Springer Dordrecht Heidelberg London New York Library of Congress Control Number: 2009926181 # Springer-Verlag Berlin Heidelberg 2009 This work is subject to copyright. All rights are reserved, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilm or in any other way, and storage in data banks. Duplication of this publication or parts thereof is permitted only under the provisions of the German Copyright Law of September 9, 1965, in its current version, and permission for use must always be obtained from Springer. Violations are liable to prosecution under the German Copyright Law. The use of general descriptive names, registered names, trademarks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. Cover design: SPi Publisher Services Printed on acid-free paper Springer is part of Springer Science+Business Media (www.springer.com)

Preface

This book is based on papers presented at an international workshop organised in Jo¨nko¨ping, Sweden, in June 2005 to celebrate the 60th birthday of Professor Bo¨rje Johansson – a dear friend and admired colleague of ours. The book provides a limited sample of Bo¨rje Johansson’s broad ranging research interests. In this volume, some of his friends and colleagues have contributed chapters on the theme of “Innovation, Dynamic Regions, and Regional Dynamics”. This is a field of research in which Bo¨rje Johansson has been a great inspiration to us all, and to which he him-self has contributed with characteristic enthusiasm and insight as part of his prodigious output. The workshop and the creation of this book were sponsored by the Alfa Savings Bank Foundation in Jo¨nko¨ping, Jo¨nko¨ping International Business School, and the School of Public Policy, George Mason University, Fairfax, VA. We thank them for their generous support. The authors and the editors thank Kerstin Ferroukhi for all her efforts to organise the workshop and Ulla Forslund-Johansson and Uma Kelekar for working tirelessly to get the papers refereed and revised, to put together multiple edits of this book and for preparing it for the publisher. It would have been impossible to produce this book without their dedicated work. Sweden Sweden UK USA

Charlie Karlsson Åke E. Andersson Paul Cheshire Roger R Stough

v

Contents

1

Innovation, Dynamic Regions and Regional Dynamics . . . . . . . . . . . . . . . . . . . 1 ˚ ke E. Andersson, Paul Cheshire, and R.R. Stough Charlie Karlsson, A

2

The Pure Theory of Spatial Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Martin Beckmann

3

Smith–Ricardo Specialization in the Presence of Tiring Effects . . . . . . . . 47 Tonu Puu

4

Dynamics of Innovation Fields with Endogenous Heterogeneity of People . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 Masahisa Fujita

5

Economics of Creativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 ˚ ke E. Andersson A

6

Simple Memes and Complex Cultural Dynamics . . . . . . . . . . . . . . . . . . . . . . . . 97 David Batten and Roger Bradbury

7

The Fashioning of Dynamic Competitive Advantage of Entrepreneurial Cities: Role of Social and Political Entrepreneurship . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 Lata Chatterjee and T. R. Lakshmanan

8

The Social Capital of Regional Dynamics: A Policy Perspective . . . . . . 121 Hans Westlund

9

Hidden Order in Traffic Flows Using Approximate Entropy: An Illustration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143 Kingsley Haynes, Rajendra Kulkarni, and Roger Stough

vii

viii

Contents

10

Regional Input–Output with Endogenous Internal and External Network Flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161 John R. Roy and Geoffrey J.D. Hewings

11

Regional Unemployment and Welfare Effects of the EU Transport Policies: Recent Results from an Applied General Equilibrium Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 177 Artem Korzhenevych and Johannes Bro¨cker

12

Infrastructure Productivity with a Long Persistent Effect . . . . . . . . . . . 197 Tsukai Makoto and Kobayashi Kiyoshi

13

Science Parks and Local Knowledge Creation: A Conceptual Approach and an Empirical Analysis in Two Italian Realities . . . . . . 221 Roberta Capello and Andrea Morrison

14

The Low Participation of Urban Migrant Entrepreneurs: Reasons and Perceptions of Weak Institutional Embeddedness . . . . . 247 Enno Masurel and Peter Nijkamp

15

The Location of Industry R&D and the Location of University R&D: How Are They Related? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 267 Charlie Karlsson and Martin Andersson

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291 Paul Cheshire and Stefano Magrini

17

Urban–Rural Development in Sweden . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 317 Johan Klaesson and Lars Pettersson

18

Patents, Patent Citations and the Geography of Knowledge Spillovers in Europe . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331 Manfred M Fischer, Thomas Scherngell, and Eva Jansenberger

19

Co-authorship Networks in Development of Solar Cell Technology: International and Regional Knowledge Interaction . . . . . . . . . . . . . . . . . . 347 Katarina Larsen

20

Off-shoring of Work and London’s Sustainability as an International Financial Centre . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373 Ian Gordon, Colin Haslam, Philip McCann and Brian Scott-Quinn

21

The Genesis and Evolution of the Stockholm Music Cluster . . . . . . . . 385 Pontus Braunerhjelm

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 409

Contributors

˚ ke E. Andersson A Jo¨nko¨ping International Business School, Jo¨nko¨ping University, Ho¨gskoleomra˚det, Gjuterigatan 5, 553 18 Jo¨nko¨ping, Sweden, [email protected] Martin Andersson Jo¨nko¨ping International Business School, Jo¨nko¨ping University, Ho¨gskoleomra˚det, Gjuterigatan 5, 553 18 Jo¨nko¨ping, Sweden, [email protected] David Batten The Temaplan Group and CSIRO, CSIRO Marine and Atmospheric Research, Private Bag 1, Aspendale, Victoria 3195, Melbourne, Australia [email protected] Martin Beckmann Economics Department, Brown University, 64 Waterman Street Providence, RI 02912, USA, [email protected] Roger Bradbury Tjurunga and the Australian National University, 9 Scott Street, Narrabundah, ACT 2604, Canberra, Australia Pontus Braunerhjelm Department of Economics , The Royal Institute of Technology, Drottning Kristinas Va¨g 30, 100 44 Stockholm, Sweden, [email protected] Johannes Bro¨cker Institute for Regional Research, University of Kiel, Olshausenstraße 40, 24098 Kiel, Germany Roberta Capello Department of Management, Economics and Industrial Engineering, Politecnico di Milano, Via Giuseppe Colombo 40, 20133 Milano, Italy [email protected] ix

x

Contributors

Lata Chatterjee Center for Transportation Studies, Boston University, One Sherborn Street, Boston, MA 02215, USA, [email protected] Paul Cheshire London School of Economics, Houghton Street, London WC2A 2AE, United Kingdom, [email protected] Masahisa Fujita Konan University, 8-9-1 Okamoto, Higashinada-ku, Kobe 658-8501, Japan [email protected] Manfred Fisher Institute for Economic Geography & GIScience, Vienna University of Economics and Business Administration, Nordbergstr. 15/4/Sector A, 1090 Vienna, Austria [email protected] Ian Gordon Geography Department, London School of Economics, Houghton St, London WC2A 2AE, United Kingdom, [email protected] Colin Haslam Center for Research in Finance and Accounting, University of Hertfordshire, Hatfield AL10 9AB, United Kingdom Kingsley Haynes The School of Public Policy, George Mason University, 4400 University Drive, Fairfax, VA 22030, USA, [email protected] Geoffrey J.D. Hewings REAL, University of Illinois, 607 S. Matthews, Urbana, IL 61801-3671, USA [email protected] Eva Jansenberger Institute for Economic Geography & GIScience, Vienna University of Economics and Business Administration, Nordbergstr. 15/4/Sector A, 1090 Vienna, Austria Charlie Karlsson Jo¨nko¨ping International Business School, Jo¨nko¨ping University, Ho¨gskoleomra˚det, Gjuterigatan 5, 553 18 Jo¨nko¨ping, Sweden, [email protected] Kobayashi Kiyoshi Graduate School of Management, Kyoto University, Yoshida-hommachi, Sakyoku, Kyoto, 606-8501, Japan, [email protected]

Contributors

xi

Johan Klaesson Jo¨nko¨ping International Business School, Jo¨nko¨ping University, P.O. Box 1026, 551 11 Jo¨nko¨ping, Sweden Artem Korzhenevych Institute for Regional Research, University of Kiel, Wilhelm-Seelig-Platz 1, 24118 Kiel, Germany, [email protected] Rajendra Kulkarni Senior Research Analyst, School of Public Policy, George Mason University, 4400 University Drive, Fairfax, VA 22030, USA T.R. Lakshmanan Center for Transportation Studies, Boston University, One Sherborn Street, Boston, MA 02215, USA, [email protected] Katarina Larsen KTH - The Royal Institute of Technology, Valhallava¨gen 79, 100 44 Stockholm, Sweden, [email protected] Stefano Magrini Dipartimento di Scienze Economiche, University of Venice, Fondamenta S Giobbe Cannaregio, 873 30121 Venezia, Italy Tsukai Makoto Graduate School of Engineering, Hiroshima University, 1-1-1, Noji Higashi, Kusatsu, Shiga 525-8577, Japan Enno Masurel Centre for Innovation and Sustainable Entrepreneurship, Free University, Habelschwerdter Allee 45, 14195 Berlin, Germany, [email protected] Philip McCann Economics Department, University of Reading, Whiteknights, Reading RG6 6BA, United Kingdom Andrea Morrison Department of Economics, Universita` del Piemonte Orientale, Via Bellini, 25/G, 15100 Alessandria, Italy CESPRI, Universita` Bocconi, Via Sarfatti, 25 20136 Milano, Italy Peter Nijkamp VU University, Department of Spatial Economics, room 4A-33, De Boelelaan 1105 1081 HV Amsterdam, The Netherlands

xii

Contributors

Lars Pettersson The Swedish Board of Agriculture, Jo¨nko¨ping 551 82 Jo¨nko¨ping, Sweden, [email protected] Tonu Puu CERUM, Umea˚ University, 90187 Umea˚, Sweden, [email protected] John R. Roy ETUDES, 3 Scenic Ct, PO Box 96, Mallacoota, Victoria 3892, Australia [email protected] Thomas Scherngell Institute for Economic Geography & GIScience, Vienna University of Economics and Business Administration, Nordbergstr. 15/4/Sector A, 1090 Vienna, Austria Brian Scott-Quinn ISMA Centre, University of Reading, Whiteknights RG6 6BA, Reading, United Kingdom Roger Stough Vice President for Research and Economic Development, George Mason University, 4400 University Dr. MS3A2, Fairfax, VA 22030, USA Hans Westlund ¨ stersund, Sweden National Institute for Working Life, Studentplan 1, 831 40 O [email protected]

Chapter 1

Innovation, Dynamic Regions and Regional Dynamics ˚ ke E. Andersson, Paul Cheshire, and R.R. Stough Charlie Karlsson, A

1.1

Introduction

The development of economic theory after World War II has focused on clarifying the necessary and sufficient conditions for the existence of an idealized general equilibrium. Debreu (1956), Arrow and Hahn (1971), and Scarf and Hansen (1973) established these conditions, building on earlier attempts by Cassel (1917) and Wald (1933–34, 1934–35). A key assumption in the formulation and proofs of the existence of a general equilibrium of a competitive economy is a large (or even infinite) number of buyers and sellers (Aumann 1964), which ensures anonymous markets and mutual independence of agents. Another assumption is the convexity of preference and production technology sets (Uzawa 1962). A third assumption is flexible pricing of goods and production factors. The flexibility of prices is the assumption that economists first called into question. Keynes formulated the most influential early criticism of the realism of assuming flexible prices in his General Theory of Employment, Interest and Money (1936). In his macroeconomic analysis, Keynes questioned the downward flexibility of the price of labor services and interest rate (i.e., the price of loanable funds), implying the possibility of equilibrium without full employment. Later, Uzawa (1976) and Benassy (1975) included such Keynesian macroeconomic fixed-price assumptions in a new general equilibrium theory and proved the existence of an even more general class of equilibrium theorems that does not depend on complete price flexibility. Frank (1969) formulated the first successful attempt to relax the assumption of a convex production technology set. He proved the existence of a set of prices that can sustain both a structure of production in general equilibrium and increasing returns to scale. Andersson and Marksjo¨ (1972) extended Frank’s analysis by assuming

R.R. Stough ð*Þ Vice President for Research and Economic Development, George Mason University e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_1, # Springer‐Verlag Berlin Heidelberg 2009

1

2

C. Karlsson et al.

continuous increasing returns of the technology sets. In both studies it was shown that sellers of each good must price-discriminate between consumers in order to sustain a general equilibrium. One of the core characteristics of Bo¨rje Johansson’s research is the development of theories and models in which increasing returns to scale are compatible with economic equilibrium. Another characteristic is his questioning of the independence of economic agents. The starting point of his research on the consequences of agent interdependence was his doctoral dissertation defended in 1978; Contributions to Sequential Analysis of Oligopolistic Competition. That game theoretic study not only assumes statically interdependent agents as in prisoners’ dilemmas and other suboptimal equilibrium games, but also takes into account strategic interactions that are truly dynamic. Interdependencies among agents take on a deeper significance for applied work when agents are distributed in continuous space or on some discrete network. Such interdependencies were almost completely disregarded by American economists, with only a few exceptions such as Hotelling (1929), Chamberlain (1936), Isard (1956), and Greenhut (1971). In Europe, there is however a separate tradition of focusing on such interdependencies, as is exemplified by von Thu¨nen (1826), Alfred Weber (1929), Launhardt (1872, 1882), Palander (1935), Lo¨sch (1954), Beckmann (1952, 1956), as well as Beckmann and Puu (1985). The role of spatial interdependence in the determination of a spatial general equilibrium with assumptions of convex production technology and preferences has been most thoroughly developed in the contributions by Beckmann (1952, 1956) and Beckmann and Puu (1985). Building on this European theoretical heritage, Bo¨rje Johansson has explored spatial and dynamic interdependencies in models where the assumption of convex production technologies is discarded in favor of assumptions of internal and external increasing returns. He has also refocused the modeling of interdependencies toward explicit dynamic economic mechanisms, instead of the simple additions of time subscripts, which is typical of static theories and models. Bo¨rje Johansson superbly follows the theoretical advice formulated by Schumpeter: This distinction [between statics and dynamics] is crucial. Statics and dynamics are two totally different areas. Not only do they deal with different problems, but they use different methods and they work with different materials. They are not two chapters in the same theoretical construction – they are two totally different buildings. (Schumpeter 1908, pp. 182–183)

1.2

Innovations and Innovation Networks

Innovation is the fundamental factor behind the development and renewal of firms, markets, regions, and entire economies. According to Schumpeter (1934), an innovation can be a new (1) product, (2) production technology, (3) market, (4) organization or (5) input. We focus on the first three types of innovation, since they

1 Innovation, Dynamic Regions and Regional Dynamics

3

usually constitute the majority of innovations. Similarly to production and economic growth, innovations are always unevenly distributed across countries, regions, as well as across localities within regions.1 Spatial differences result from the unequal attributes of each location. Consequently, Johansson (1998a) calls such attributes location attributes. For each type of economic activity, one can identify certain combinations of location attributes that support it better than other combinations. Some location attributes are gifts of nature, while others are created by investments in physical and human capital with low spatial mobility. Still others are the result of the behavior of economic agents with spatial preferences, such as households or firms. Standard economic theory has devoted little attention to regional differences concerning location, innovation, productivity, and growth. Research with a regional focus has therefore been forced to create its own platform and conventions, which specify relevant and challenging research problems. It is possible to identify a selforganized research program in Sweden, which is based on the work of economists, geographers and other regional scientists since the early 1950s. The inspiration for that research program harks back to the interwar period and the contributions by, in particular, Ohlin (1933) and Palander (1935) (Johansson 1998a). One economist and regional scientist who has played a central role in the research program since the 1970s is Bo¨rje Johansson. This introductory chapter has as its main purpose to provide an overview of his engagement with – and contributions to – the research field within spatial economics that deals with innovation, regional specialization, and dynamic systems of functional regions. One way to understand and analyze innovation processes is to study the increased formation of economic networks among producers, subcontractors, and buyers of final products (Johansson 1990b; Johansson and Westin 1994). Such networks consist of nodes and links (Karlsson et al. 2005). Johansson (1991b) outlines some of the fundamental elements of the emerging theory of economic networks by providing an economic model which explains the creation of linkages and networks, and which also attempts to explain the durability of such relations. The network approach recognizes the importance of repeated mutual investments in the links that connect customers and suppliers (Johansson 1990a; Teubal and Zuscovitch 1994). Investments in links between suppliers and customers create and expand networks. The amount of investment that is required to establish and strengthen a link between two economic agents is a negative function of the existing affinities between the two nodes and a positive function of the spatial friction. The dominant flows of a specific product or type of information will use links, which have the most appropriate attributes, while at the same time being constrained by barriers and other types of friction.

1

Since people and firms are highly concentrated in space of course we would not expect innovation to be randomly distributed across space. The problem is that we need a priori to formulate a null hypothesis about, what would constitute an ‘‘even distribution’’ (see Glaeser and Ellison 1997; Duranton and Overman 2005).

4

C. Karlsson et al.

The links in an economic network must be analyzed as immobile capital goods, which have incurred sizable sunk costs. Existing linkages therefore impose rigidity and inertia on firms’ interaction patterns such as trade flows, deliveries of current inputs and capital equipment, and exchanges of technological knowledge. Normally, a link between a supplier and customer will not be broken unless a new supplier can offer a new input, which is clearly superior to the current input, since the new supplier has to overcome the sunk cost advantages of an established link. Emphasizing the network aspects of the economy, rather than using the traditional price-oriented view of the market, implies that link attributes increase in importance relative to node attributes as explanations of trade patterns, service networks, spatially distributed production networks, and innovation networks. The archetypical model of a market economy with independent actors, in which a quantity of a product is bought from the seller who offers the lowest price at the point of delivery, focuses in a way upon production costs in nodes and, rarely, if at all, on transport and transaction costs. Thus, it disregards the dynamic interplay between market actors, which is not only typical of the market but also shapes its development trajectory.

1.2.1

Innovation Networks

Innovations never occur in splendid isolation. Instead, it is natural to describe product development and renewal of production processes as a natural part of the interaction between a firm and its customers and suppliers through its customer and supplier networks (Johansson 1993a). As part of its research and development, a firm also buys R&D results and knowledge support through its network of knowledge channels. The opportunities for an individual firm to improve its production process are dependent upon the conditions for buying new equipment and new knowledge from the suppliers in the firm’s supplier network. Suppliers of new techniques and sellers of new equipment frequently try to use established economic networks as a means to access potential technology customers (Johansson 1991b). This is why networks within large corporations often function as arenas for innovation diffusion (Karlsson 1988). Established networks have two distinct roles. First, the seller of technical systems and production knowledge must supply products, which are either designed specifically for each customer, or which can be adapted to fit the demands of the buyer. Hence, the seller needs existing links as channels through which it is possible to find customers, who also have sufficient purchasing power to pay for the necessary customization. One should emphasize that the customers are, in fact, carrying through their own innovations – although a lot of imitation may be involved. Second, the delivery of new equipment and installation of new systems are processes that frequently take a long time to perform and require frequent interactions between the delivering and receiving firms. Both parties need a reliable link for their co-production, which may include joint development and learning.

1 Innovation, Dynamic Regions and Regional Dynamics

5

Firms also receive knowledge about how its products ought to be redesigned through information from their customer networks. In addition, many firms have specialized knowledge links, which were created to generate better conditions for research and development within the firm. Thus, we can talk about innovation networks as a sub-structure of a firm’s general economic network. Of course, the strength of the innovation network varies among firms due to factors such as size, age, and industry. We may combine the above observations into a model of innovation behavior in economic networks: l

l

Established networks for economic interaction are important vehicles for the diffusion of technological solutions. The delivering and receiving parties make contact via direct and indirect links in such networks. The networks therefore facilitate the transmission of knowledge. Networks may play this role regardless of their initial use and rationale. The ability of a firm to improve its production, distribution and other techniques depends on its capacity to build new links to suppliers of knowledge and equipment. Network formation is equally important for a firm that tries to establish cooperative ventures with other firms in order to renew and develop products.

Knowledge plays a critical role in innovation processes. Karlsson and Johansson (2006) argue that it is meaningful to make a distinction between three types of knowledge: l

l

l

Scientific knowledge consists of basic scientific principles that can form a basis for the development of technological knowledge. Technological knowledge comprises implicit and explicit blueprints in the form of inventions (or technical solutions) that may be transformed into new products or production processes. Entrepreneurial knowledge consists of economic knowledge about potentially profitable entities such as products, business concepts, markets, customers, and suppliers.

The different types of knowledge flow from ‘‘sources’’ to ‘‘sinks’’ using links in different types of knowledge networks.

1.2.2

Knowledge ‘‘Sources’’ and Knowledge ‘‘Sinks’’

Links that connect nodes are the conduits for flows in networks. The direction of a flow is always from a ‘‘source’’ to a ‘‘sink’’. If the flow represents an economic transaction, the ‘‘source’’ is a supply node while the ‘‘sink’’ is a demand node. The concepts ‘‘source’’ and ‘‘sink’’ include but are not limited to ‘‘supply’’ and ‘‘demand,’’ and are general starting points for analyzing transmissions of knowledge and experiences among individuals, organizations, and over space.

6

C. Karlsson et al.

Scientific knowledge is disseminated in open scientific networks with universities and research institutes as permanent ‘‘sources’’ and with courses, conferences and scientific publications as links to the ‘‘sinks,’’ which are students and scientists as well as firms that are interested in transforming scientific knowledge into inventions and innovations. Technological knowledge includes knowledge about production methods as well as technical solutions about the design and construction of goods and services. Technological knowledge usually differs from scientific knowledge in that intellectual property rights in the form of patents and copyrights prevent general use of the knowledge. This implies that new technological knowledge is traded for a price or – if the knowledge creating firm decides to use it as a strategic resource – is simply unavailable. As a technology ‘‘sink’’ we can imagine a firm with an intention to start new production or to improve on its current production methods. To make this possible, it needs two types of technological knowledge: (1) knowledge about alternative designs of the planned product, which is the same as knowledge about product outcomes, and (2) knowledge about available production technologies or processes for producing the product. There are many ‘‘sources’’ of new technological knowledge. They include the firm’s own experiments, surveys of and contacts with customers, imitation of other firms’ technological knowledge, purchases of patents and licenses, employment of other firms’ employees and new university graduates, as well as in some cases industrial espionage. Technological knowledge is transmitted from ‘‘sources’’ to ‘‘sinks’’ in three ways: l l l

As individuals (human capital) As books or software (information) As equipment (physical capital)

When new technology is embodied in individuals, technology transfer takes place when individuals move from one organization to another or when individuals from different organizations come together in face-to-face meetings. After technological knowledge has ‘‘matured,’’ knowledge workers may codify and transfer it by using drawings, software, and texts or by structured education. When firms buy patents and licenses, they buy codified knowledge. The third form of technology transfer emerges when a firm buys physical capital such as technical equipment or machines, which embody new technological knowledge. It is not unusual for technology transfer to be a complex process, which may involve a combination of hiring individuals which embody critical human capital, training, acquisition of patents, and acquisition of capital goods. The third type of knowledge – entrepreneurial knowledge – is also critical for innovation processes. It includes knowledge about the demand for products with varying characteristics and the willingness among customers to pay for such products. Entrepreneurial knowledge also includes knowledge about competitors such as their strategies to attract various types of customer. The ‘‘sources’’ of knowledge are customers and competitors, both actual and potential. The links

1 Innovation, Dynamic Regions and Regional Dynamics

7

are whatever connects a producer with its customers and competitors, such as information and transport networks. Inventions and innovations are acts of creation2 with elusive ultimate causes. It is difficult to go beyond the distinction between inventions and innovations. An invention is the solution to a technical problem. To transform the invention into an innovation it is necessary that the innovator expects the technical solution to be economically viable. Economic viability is determined by production costs (including development costs) and revenue generated from the potential customers. Innovation processes often involve a combination of developing new production methods and new product characteristics. However, there are innovations that only introduce new production methods for producing existing products without any new characteristics, and there are also innovations that only concern the introduction of new product attributes with negligible process innovation. Maillat et al. (1993) distinguish between three types of product innovation. The most modest as well as the most common type entails the incremental addition of new elements to an already existing product. In this case, the aim may be to make the product more reliable and versatile. A transformation of the functionality of the product implies a more far-reaching product innovation. Now the product not only fulfills the needs of customers better, it offers new and unexpected functions. Most radical are those innovations that not only create new functions but also new markets. During the post-war period, many studies analyzed the innovation intensity of firms by measuring their patent frequency. These studies have been conducted even though there is a general agreement that patents only reflect a small part of all innovations. One question that has interested many economists is the extent to which the market and developments on the demand side stimulate product innovation, and to what extent the internal forces within companies together with the technological conditions for each product group generate new products. A large study by Scherer (1984) relates patent frequencies to: l l l

The size of the market for a firm. Differences in technological opportunities for different kinds of goods. The renewal readiness of the individual firm. In Scherer’s study the market explains a little more than 40%, technological opportunities explain about 30%, and the individual renewal readiness explains a little more than 10% of the variability in patent frequencies.

Energy and skills in knowledge ‘‘sources’’ and knowledge ‘‘sinks’’ govern the diffusion of technological knowledge (Johansson 1993a). The diffusion of knowledge and technology does not depend on the volume and intensity of the flow from the knowledge ‘‘source.’’ Technology transfer also results from the demand from

2

The same is true of new scientific knowledge.

8

C. Karlsson et al.

the knowledge ‘‘sink.’’ The implication is that innovation networks primarily contain links between strong knowledge ‘‘sources’’ and strong knowledge ‘‘sinks.’’

1.2.3

Cost and Innovation of Product Characteristics

It is common in analyses of innovation and technology diffusion to make a schematic distinction between innovations that focus on improving production techniques (i.e., process innovation) and those that focus on improving existing products or introducing totally new products (i.e., product innovation). Conventionally, process innovation denotes all changes of production techniques that are used in the production of a given product in a given firm. However, the term process need not exclusively imply a narrow conception of technology but may also imply ‘‘non-technological’’ activities in a firm (Fischer and Johansson 1994). This more inclusive interpretation of ‘‘process’’ corresponds to its use by Nelson and Winter (1982). They argue that a firm embodies a set of interdependent production routines, which combine to form a complex process. Nelson and Winter’s definition implies that a complex production process includes the following sub-processes: l l l l

Distribution Production Routine design and construction Management, administration and commercial activities

Improvements to any of these sub-processes are process innovations. They primarily refer to changes that lead to more efficient resource use, which, for example, reduce production or distribution costs. In such cases, process innovation equals cost-reducing technical changes. But process innovation also includes those changes in the production processes which increase a product’s quality and reduce the proportion of defects, while preserving the original functions of the product. Process innovations are therefore all innovations that are not product innovations. What is then a product innovation? To answer this question we need a systematic way of describing products. Lancaster (1971) offers one such approach. He suggests a product description, which specifies the various attributes that characterize the product. He calls the attributes ‘‘characteristics’’ and assumes that it is possible to measure the quantity of each such product characteristic. As a consequence, each good or service becomes a specific combination of characteristics. Lancaster’s approach is closely related to Schumpeter’s analysis of innovation. Schumpeter (1934) treats innovation as the result of a process of new combinations. When a firm develops a new capital good, we can distinguish between two cases. In the first case, the firm intends to use the good itself and will for that reason attempt to prevent competitors from learning about it. In the second case, the firm will market the new good with the goal of making a profit. The goal is thus to find as

1 Innovation, Dynamic Regions and Regional Dynamics

9

many customers as possible with a sufficient willingness to pay for the new capital good with its various attributes. In this case, the firm has made a product innovation. When the buyers of the new capital good start using it in their production process they are making a process innovation.3 A need to cut production costs usually causes a firm’s efforts to improve its production process. This need is most obvious and persistent for products that are exposed to price competition from rival producers. The impetus to improve the efficiency of the production process recurs every time a competitor has succeeded in improving its production methods, and it also recurs at the onset of each cyclical downturn. The ability to manage continual improvements to the production process requires a continuous supply of new technology in the form of new technological knowledge. This includes imitating rivals, taking up suggestions from consultants and suppliers, and adapting information that has been gathered through the firm’s intra-regional and inter-regional innovation networks. We should also note that there are interdependencies between product and process innovation. For mature products, there is often a choice between old and new production processes, but new products normally require new production processes.

1.2.4

Innovation at the Industry Level

At the industry or sector level, economists study both product and process innovation as entry and exit processes (Johansson 1987; Johansson and Holmberg 1982). This approach builds on an important insight in Schumpeter’s theory of economic development, which is that the original entrepreneurs receive a premium in the form of greater profits for being pioneers (Schumpeter 1934). This ‘‘extra profit’’ to innovators is a temporary monopoly, which results from the specific knowledge that they do not (yet) share with their competitors or only share with a few of them. Irrespective of whether one assumes that such innovations occur continuously or continually and irrespective of the character of the imitative diffusion process, one should expect an uneven distribution of productivity and profits among the firms in an industry. We should expect pioneering entrepreneurial firms to earn greater profits and have greater productivity than their imitating followers. Empirical data confirm that economic rewards are ‘‘Schumpeter-distributed,’’ and that such distributions have a characteristic form (Johansson and Marksjo¨ 1984; Johansson and Stro¨mquist 1981). Moreover, not only does the general form of such reward distributions persist in each industry, but the specific parameters of the distributions exhibit long-term stability. 3

This implies that product innovations in one industry often show up as process innovations in one or several other industries. What is a product or a process innovation depends upon the perspective taken in the analysis. In the case of consumer goods there is normally no need to make this type of distinction.

10

C. Karlsson et al.

For innovations among firms in an industry, it is important to observe that innovations appear in two distinct forms: each firm may renew its production technique, but it may also adjust its old technique in order to develop new products. One may use Lancaster’s (1971) approach to analyze the effects of introducing new products. It is possible to combine the substitution of new for old products with the dynamic substitution of new for old production techniques. The dynamic processes of entry and exit generate specific distributions of process and product vintages that are associated with observable profit and productivity distributions. Different assumptions about the entry and exit dynamics generate different forms of the productivity and profit distributions in an industry. Product changes with logistical substitution processes explain the steepness of empirically observed productivity and profit distributions. In the absence of product evolution, technical change generates productivity and profit distributions, which are quite flat.

1.3

Regional Specialization

In the previous section, we analyzed innovation processes from the perspective of the firm, without considering the fact that innovation processes tend to locate in certain regions, in particular, large urban regions. In this section, we turn to the question of which factors determine the specialization of regions. Before considering these factors, however, we need to consider what a region is. In a functional economic region, one can identify one or (often) several spatial economic nodes, for example population centers, which physical infrastructure networks and established economic interaction networks jointly connect (Johansson 1993a). Of special importance are labor market networks, where the links between employees and employers create a tentative structure. Every employment relationship presupposes a contract, which also (indirectly) connects a dwelling to a workplace. A region’s accessibility patterns decide how these contract links generate geographically contiguous labor markets of various sizes. The links in the labor market constitute one of many networks, which integrate a regional economic system. Another such network is the communication network which job-seekers use to find suitable jobs and employers use to find workers with suitable skills. A functional economic region becomes an integrated economic system through the interaction, which takes place in established networks and includes communication, decision-making, and distribution of goods and services. A functional region has greater mobility of production factors within its interaction borders than with areas outside. Commuting and all other forms of interaction, even within functional regions, give rise to interaction costs. The size of these costs determines the spatial extent of the region. The heterogeneity of natural conditions and historical development paths means that functional regions differ from one another in their economic milieus, and thus

1 Innovation, Dynamic Regions and Regional Dynamics

11

offer dissimilar conditions for economic specialization. The regional economic milieu comprises those location attributes that are durable (fixed or slowly changing), that the individual firm cannot control, that are not traded other than as land attributes, and that influence firms’ production activities (Johansson 1998a).

1.3.1

The Infrastructure as a Set of Durable Location Attributes

A special type of durable location attributes is that part of the built environment in a region that qualifies as material infrastructure. The material infrastructure is durable capital that generates location attributes services, which influence the regional economic milieu (Johansson and Snickars 1992). It comprises of three parts: networks that convey people, goods, and messages; facilities that supply public goods; neighborhoods that provide access to housing and workplaces. Johansson (1991a) maintains that one may envisage the infrastructure as a landscape of interaction possibilities for resource flows as well as inter-personal and inter-firm contacts. Infrastructural changes are slow in comparison with the fast adjustments of most social and economic activities, which mean that in the short term the material infrastructure provides an arena for rapidly changing social and economic processes. The material infrastructure supplies services to a collective of users, but the spatial extent of the services is limited. It satisfies at least one of the three following criteria (Johansson and Snickars 1992): polyvalence; inter-temporal consistency; a systemic or network function that generates accessibility. It is also possible to identify a non-material regional infrastructure that consists of collective, durable, and relatively immobile location attributes, for example agglomerations of human capital and regional institutions (Andersson 1985). For both the material and the non-material infrastructure, the slow time scale is essential. The durability of location attributes implies that the allocation of other more mobile production factors has sufficient time to adjust to persistent spatial differences (Johansson 1998a). Johansson (1993b) recognizes that the material infrastructure, with its associated networks, functions as a set of systems for economic interaction. He claims that the development of prototypes, the adaptation of novel products, and the routine processing of mature products each constitutes a distinct type of activity. Each such type has specific interaction characteristics and needs. Thus, each type demands particular combinations of infrastructure attributes from its regional economic milieu. A network is an infrastructure, which facilitates interaction within and between regions. The interaction between intra-regional and inter-regional networks determines the long-term evolution of spatial economic systems (Johansson 1993b). Intra-regional networks make it possible for economic actors to benefit from the proximity of dense urban structures and to develop and restructure interpersonal networks. Such development and restructuring of links between economic partners and between buyers and sellers constitute the most basic mechanism for the

12

C. Karlsson et al.

evolution of every market. These link-shaping activities are almost exclusively hosted by urban environments with suitable infrastructure attributes (Johansson 1989a). They are investments in more or less durable links for communication and the exchange of information and knowledge, where the formation and maintenance of the links require personal face-to-face contacts. Frequent contacts require appropriate intra-regional and, in particular, urban infrastructure. Such infrastructure combines accessibility in local networks with a dense environment of meeting places and multi-purpose facilities. A productive urban economic milieu offers a variety of opportunities for personal contacts among people with diverse experiences, competences and skills (Johansson 1993b). In a city with general and polyvalent characteristics, maturing activities often migrate to peripheral parts of the city region, while new activities benefit from a central location. Production that benefits disproportionately from a certain location can force out other activities by offering higher land rents. In this way, new and alert economic actors can use the same infrastructure over and over again. This implies that the market does not treat the infrastructure as a sunk cost. The urban infrastructure instead displays ‘‘hotel attributes’’ (Johansson 1993b).

1.3.2

Regional Economic Milieus and the Economic Specialization of Regions

The dynamic processes that over time reshape a region’s economic milieu are driven on the one hand by external forces, and on the other hand by adjustment, development and investment processes within the region. The dynamics of these processes are often extended in time, due to the inertia associated with the transformation of regional resources. This inertia gives functional regions their identity and implies that their economic structure only changes gradually and at a slow pace. The economic milieu of functional regions influences economic agents and their behavior in three ways: l

l

4

The production capabilities of regions differ between industries. This implies that a specific set of infrastructural location attributes influences the productivity and cost structure of firms in a non-uniform fashion. (Johansson 1998a).4 The attractiveness of regions regarding different activities, for example the inand out-migration of households and firms, and the expansion and contraction of firms (Johansson 1998a).5

In Johansson (1993c), a quasi-dynamic model is applied to estimate how the economic milieu in municipalities influences the production in different manufacturing industries (see also, Johansson et al. 1991; Johansson and Karlsson 1994; Forslund and Johansson 1995). 5 The study by Holmberg and Johansson (1992) indicates, for example, that service sectors, such as wholesale, transportation, consulting and financial services are concentrated in municipalities in which the infrastructure facilitates interpersonal contacts and mobility.

1 Innovation, Dynamic Regions and Regional Dynamics l

13

The innovative capabilities of regions, such as the creation of new knowledge, inventions, and innovations.

Regional scientists have employed two types of models to explain location patterns and regional specialization, both of which can be extended to include dynamic change processes. The first type consists of models with a central place system. Central place models focus on demand-driven specialization, in the sense that regions that are large and dense can host a richer variety of output than smaller and sparser regions (Beckmann 1958, 1996; Tinbergen 1967). In such models, it is the size of the set-up costs for each product that determines the size a region’s market area which a product must have. If the market area is too small, the region will not host the activity in question. At a given point in time, it is possible to identify products, which are only produced in those regions where the regional demand is large enough. The location advantages offered by a region’s economic milieu may also determine its specialization. Location advantages are relative characteristics of regions. It is only possible to evaluate a region by comparing the location advantages offered by different regions. Every functional region’s profile of location advantages has its basis in the region’s relative supply of resources. Lasting location advantages can only derive from resources that are immobile and change slowly. This builds on the assumption that it is possible to classify economic adjustment processes according to their speed (Johansson 1985; Johansson and Karlsson 1987). Johansson (1989b) presents results from mathematical models of dynamic systems, with the aim of identifying the importance of separating processes that operate on significantly different time scales. The formation of a network infrastructure and network flows constitute, a slow and a fast process, respectively. In the second type of model – location advantage models – the relative supply of trapped resources determine the specialization patterns of regions in a multi-regional system (Johansson and Karlsson 1987; Johansson 1997). The assumption of trapped resources has long been important for explaining regional specialization and trade within a Heckscher–Ohlin framework (Ohlin 1933). Certain economic activities use natural resources, which producers have to extract or harvest within the region of production. A standard location advantage model will predict where, among available regions, such resource production will take place (Moroney and Walker 1966; Smith 1975). Location advantages are not limited to the spatial distribution of natural resources, but also include various localized (i.e., regionally trapped) non-land production factors, such as infrastructural and human capital. These resources are not as immobile as natural resources but their potential relocation (‘‘speed of adjustment’’) is slow relative to other economic adjustment processes. A starting point for analyzing how location advantages influence regional specialization is that at each point in time, the various types of trapped resources are unevenly distributed over functional regions. Moreover, certain trapped resources are highly concentrated in functional regions with specific characteristics, such as their positions in networks for communication and transportation. If we assume that the spatial density of certain trapped resources is changing at a much slower pace than technology, it is possible that a technological change induces a relocation of

14

C. Karlsson et al.

production and a corresponding change in interregional trade patterns. As a consequence, slowly adjusting resources govern the emergence of new patterns of regional specialization. The structural economic development in a system of functional regions is the outcome of various interlinked adjustment processes that operate at different time scales. Both the central place and the location advantage approach stress the role of durable regional characteristics. Central place models focus on the accessibility to local and external markets, while location advantage models focus on durable trapped characteristics. Nevertheless, both these types of traditional models have limited explanatory power. If we assume durable regional characteristics as the only explanation of trade patterns, it becomes impossible to explain why regions that produce an almost identical set of goods trade with each other. The traditional approaches are also unable to explain how the behavior of economic agents may change the specialization of regions.

1.3.3

Spatial Transaction Costs and Endogenous Specialization

By combining assumptions about internal market potentials, increasing returns and spatial transaction costs, Johansson and Karlsson (2001) provide a framework for analyzing the endogenous specialization of functional regions. Both internal and external economies of scale can generate increasing returns. External economies of scale (i.e., agglomeration economies) consist of localization economies and urbanization economies. Localization economies are specialized external economies of scale, and are common in both large and small functional regions. An abundance of general positive supply externalities cause urbanization economies, and they are therefore associated with large urban regions (Vernon 1960). While large regions can specialize in diversity, Johansson and Karlsson (2001) argue that localization economies provide an opportunity for small and mediumsized functional regions to develop competitive specialization clusters, even though the internal market potential of such regions is much smaller than that of a large metropolis. They therefore elaborate on the role of internal and external scale economies in combination with product-specific spatial transaction costs in the economic development of small and medium-sized functional regions. Spatial transaction costs comprise both transportation and general transaction costs, which vary with the geographical distance between seller and buyer, and the properties of each specific spatial interaction link. Using the two concepts of functional (urban) regions and spatial transaction costs as their starting point, Johansson and Karlsson employ the following assumptions in order to generate a framework for analyzing endogenous regional specialization: l

The overall pattern of spatial transaction costs delimits functional regions. For contact-intensive transactions, the spatial transaction cost level is much higher across than within regions.

1 Innovation, Dynamic Regions and Regional Dynamics l

l

l

15

A region’s population size and total purchasing power determines its internal market potential. Internal and external markets make up the total market potential of a functional region. Networks for trade and other economic interactions connect each functional region to its external markets. The interaction intensity varies across such networks, and makes it possible to identify a hierarchy of sequentially widening transaction areas for each region, so that transaction costs rise in a stepwise sequence. A region’s location of activities and specialization is a process, which is influenced by two basic conditions: technology and scale effects; and durable regional characteristics.

Using this framework, Johansson and Karlsson (2001) explain internal and external scale economies theoretically, by showing how these phenomena combine and interact to generate cumulative specialization processes in functional regions. In particular, they focus on the specialization of small and medium-sized regions. An insightful contribution is their development of the spatial transaction cost concept, which is essential for understanding both the specialization opportunities of regions of different sizes and scale-based specialization. In relatively small regions, they show that the development of localization economies is indispensable in the absence of natural resource endowments. Combinations of three phenomena cause external scale effects: specialized labor markets, specialized neighborhood firms, and information spillovers. The first two phenomena give rise to intra-market effects, whereas information spillovers among firms are collective extra-market effects. They also illustrate how it is possible to order contact-intensive goods and services with respect to their dependence on the size of the internal market potential. Generally speaking, the flow intensity of longdistance inter-regional trade drops discontinuously at the borders of affinity-classified transaction areas, where such borders act as affinity barriers. It is important to observe that spatial transaction costs do not remain constant over time. A general development path is the seemingly unlimited extension of markets until they become global. Two network phenomena explain this (Hacker et al. 2004): The first phenomenon, which usually involves multinational corporations, is the development of economic links that allow transactions to be carried out over long distances at reduced cost. The second associated phenomenon is the development of networks for conveying information, services, goods, and people. The evolution of such networks reflects ambitions of making transactions less distance-sensitive (Andersson 1986). External economies play a key role in current explanations of location advantages and regional economic specialization. However, the literature is not always unambiguous in its use of this concept. Johansson (2005) suggests that it is possible to avoid such ambiguity by making three distinctions: the ‘‘source’’ of the externality (proximity versus network externalities); the economic nature of externalities (pecuniary versus non-pecuniary externalities); and the consequence of the externalities (efficiency versus development externalities).

16

C. Karlsson et al.

1.3.4

Combining Resource-Based and Scale-Based Models of Regional Specialization

The discussion of regional specialization in the preceding sections has focused either on resource-based or scale-based specialization. However, Holmberg et al. (2003) shows that it is possible to combine resource-based and scale-based assumptions into an integrated theoretical framework of endogenous regional specialization and growth. They do this for each sector in the regional economy by associating resource-based advantages with input-market potentials and scale-based advantages with customer-market potentials. Input-market and customer-market potentials tend to vary with the economic size of functional regions. This makes it possible to combine resource-based with scale-based regional specialization and growth processes. Modern resource-based models emphasize the supply of knowledge-intensive labor as a primary specialization factor. Thus, Holmberg et al. (2003) focus on the interaction between population changes and the development of economic activity in functional regions, paying special attention to the knowledge intensity of the labor force. This includes labor location dynamics relating to housing and job opportunities as well as the supply of household services. A major concern is to combine two conflicting assumptions, which are: l l

People follow jobs Jobs follow people

Holmberg et al. (2003) assume the self-generating processes that change regional specialization over time to have the form of interdependent dynamic processes that involve economic activities and the population size. The literature contains a number of empirical models that emphasize the exact form of the dynamic interdependence (Mills and Carlino 1989; Holmberg and Johansson 1992; Johansson 1996). In this theoretical framework, the infrastructure for interaction functions like an arena that links resource-based and scale-based models of regional specialization. The market potential of a firm refers to its accessibility to customers and input suppliers, including suppliers of labor services. The infrastructure facilitates the development and growth of the market potential as well as its density.6 The location factors for households include accessibility to jobs, household services, and amenities. Again, the same infrastructure helps to create accessibility and density. A basic idea in this approach is that not only physical infrastructure but also market potentials are slowly adjusting variables. Holmberg et al. (2003) illustrate how a set of self-reinforcing processes contributes to the growth (decline) of the market potential of a region that is experiencing a process of endogenous change. The development of firms interacts with the development of customer-market potential, input-market potential, and 6

A number of recent empirical studies illustrate the importance of ‘‘economic density’’ in functional regions (Ciccone and Hall 1996; Johansson 1996; Johansson et al. 1998; Karlsson and Pettersson 2005).

1 Innovation, Dynamic Regions and Regional Dynamics

17

labor-input market potential. Households interact with job-market, housing-market, and consumption-market potentials. The input-market labor-input-market potentials are core variables in resource-based models of regional specialization and growth. The customer-market potential refers to the opportunities of firms to benefit from both internal and external scale economies. The job-market potential is a measure of the friction households with a given location face when they search for jobs with acceptable commuting conditions. A combination of large job-market and housingmarket potentials increases a household’s opportunities of finding an efficient match between its job and housing locations. The size of the consumption-market potential determines the opportunities for households to benefit from variety in consumption. What can these self-reinforcing processes look like? We can imagine a functional region whose market potential has increased due to improvements in the transportation infrastructure. This will stimulate firms with internal scale economies to locate in the region and existing firms to expand their activities. Inmigration of firms to the region and expansion of the region’s native firms will increase the market potential of the region, generating further in-migration and expansion. As production grows the cost per unit of output falls, due to scale economies. This allows the price of interregional exports to fall, which stimulates growing export flows. In such a process, the external market potential grows as a share of the total market potential. When firms with similar activities locate and expand in the region they generate external economies, which induce more firms in the same industry to locate and expand in the region. A growing demand for inputs stimulates input suppliers to locate and expand in the region as long as their deliveries are distance-sensitive, which in turn stimulates the in-migration and expansion of customer firms. A growing demand for inputs increases the opportunities of input suppliers to take advantage of their internal economies of scale but also to develop their own external economies. When the internal market potential expands this may induce falling output prices, which in turn further stimulates exports to other regions. In this way, the external market potential increases its impact on the cumulative growth trajectory. The demand from export markets may also by itself generate self-reinforcing growth (Johansson and Lo¨o¨f 2006). What about the location of labor? The assumption here is that functional regions with attractive location characteristics for consumers attract households, especially households with a lot of human capital. A region’s attractiveness depends on the infrastructure, which comprises the region’s housing market and the accessibility from dwellings to the supply of household services, the supply of amenities of various kinds, and job opportunities (i.e., to household market potentials). This implies that regional labor markets must increasingly adjust through a process where firms follow the location of the supply of labor, rather than the opposite. The location of households and jobs forms a self-reinforcing dynamic process. The process is affected by the formation of regional infrastructure, which gradually improves or deteriorates, from the economic actors’ points of view. Naturally, the job-location process partly shapes the economic milieu. However, the assumption is that the infrastructure changes at a much slower pace than the location of jobs. In

18

C. Karlsson et al.

the short run it is therefore possible to treat the infrastructural characteristics as approximately fixed. The same argument applies to the relation between location characteristics and the dynamics of household location. The overall regional change process is dynamic in which jobs and households mutually adjust to each other. This formulation is in sharp contrast to the well-known export-base model. According to that model, economic activities have fixed locations while the labor supply of households adjusts to the demand for labor through a process where households follow jobs.

1.3.5

Economic Specialization in Small and Large Regions

When analyzing a functional region and its location advantages, it is useful to make a distinction between two dimensions as in Table 1.1. The table highlights the differences between large and small regions, and therefore their specialization opportunities. A region with a clear and narrow specialization is quite different from a region that has a diversified economy with many specializations. Smaller regions may rely on the availability of a particular natural resource, on economies of scale, or on localization economies, which are always combined with a limited intra-regional market potential. In small regions, the material and non-material infrastructure are less general and diversified than in a large regional economy (Johansson and Karlsson 1990a). According to Marshall (1920), localization economies derive from a pooled market for labor with specialized skills, the provision of non-traded inputs of a collective nature, and spillovers of entrepreneurial and technological knowledge, which can spread more easily in a local environment. Localization economies may develop when firms with similar activities locate together, whereby they form a ‘‘cluster.’’ This implies that cluster formation is a cumulative process (Johansson 2006). At each point in time, one may analyze a static cross-section of co-located industries and firms. It is common to interpret such location patterns as equilibrium outcomes. However, it is also possible to conceive of such a cross-section as a momentary image of a dynamic process, where an attractor drives the dynamics, and where this attractor may have (implicit) equilibrium properties. A small region can specialize in the exploitation of natural resources (to the extent that they are available) and in the development of a limited number of Table 1.1 The infrastructure of functional regions Demand conditions Supply conditions Intra-regional Intra-regional accessibility to general Intra-regional accessibility to labor with infrastructure customers, specialized customers, varying skills, natural resources and purchasing power amenities, housing and consumer services, producers in different industries, knowledge resources Inter-regional Inter-regional accessibility to general Inter-regional accessibility to suppliers, infrastructure customers, specialized customers, competitors, knowledge resources purchasing power

1 Innovation, Dynamic Regions and Regional Dynamics

19

industries serving distant markets. If it is successful in harnessing its location advantages, one or a few specialized clusters may emerge. Early phases of cluster development often build on notable innovative successes. Large and dynamic urban regions offer agglomeration economies which provide a creative milieu (Andersson 1985), a diversified supply of producer services, a diverse supply of human capital, as well as intra-regional and inter-regional information flows. For the most part, large urban regions offer a more diverse supply of markets than smaller regions (Hacker et al. 2004). This reflects differences in geographic transaction costs among goods and services. Profit-seeking firms cannot supply distance-sensitive goods or services in functional regions where the demand is too small to cover fixed costs. The theoretical background is as follows: Diversity in the set of regionally produced consumer goods or producer inputs can yield external scale economies, even if all individual competitors and firms earn normal profits. The size of a functional region in terms of aggregate purchasing power determines the number of specialized local consumer goods and producer inputs, given the degree of substitutability among the specialized local goods in consumption and among specialized inputs in production: A larger city will have a greater variety of consumer products and producer inputs. Since the greater variety adds to consumer well-being, it follows that larger cities are more productive, and the well-being of those living in cities increases with their size. This is true even when all firms in these cities earn a normal rate of profit. (Johansson and Quigley 2004, p. 170)7

There are two well-known models, which deal with the advantages of a diversified urban economy. The first model focuses on urbanization economies in general such as consumers’ taste for variety and, in addition, the productivity of specialized production factors. The second model is quite different: the proximity and linkages of firms in an agglomeration enhance their productivity. The perspective here is forward and backward linkages among economic agents such as firms. Thus, large functional regions have quite different specialization opportunities compared with smaller regions, since the demand for diversity and variety favors location of activities and households in large functional regions. Large home markets in conjunction with high accessibility to external markets enable many large urban regions to develop specializations (i.e., clusters) in many different industries. Firms in the same cluster may represent different stages in the production chain and also industries offering supplementary services. In many large regions, services predominate. Because of their great market potential, large regions also make it possible for firms with a ‘‘thin’’ but distance-sensitive demand to find sufficient demand to earn a profit, even with substantial fixed costs. Large urban regions are especially attractive to such firms which imply that one

7

However, there are factors, which limit the growth of cities. Otherwise, cities would grow continuously. There are costs which rise with city size, most obviously prices (space in particular), and some external costs like congestion and pollution. Also probably, crime rises with city size.

20

C. Karlsson et al.

important characteristic of large urban regions, besides hosting many clusters in different industries is the diversity of goods and services offered to consumers as well as to other firms. Another characteristic of large urban regions is that they host great concentrations of knowledge in the form of human capital, both in the form of labor and infrastructural facilities such as universities. Head offices of multinational corporations are nearly always in the downtown areas of major cities, and so are often their research and development divisions. Large regions are almost always well connected to the global air transportation network and – in Europe and Japan – to high-speed rail networks. Their access to global transportation networks makes large cities attractive meeting places in which to stage conferences, trade fairs and the like. Taken together, these conditions imply that large cities often perform gateway functions, which means that they functions as import nodes for new ideas, inventions, and innovations, which are then disseminated to their low-accessibility hinterlands (Andersson and Andersson 2000). Most large urban regions have high per capita incomes, relative to smaller regions in the same part of the world. Their relative affluence implies a greaterthan-average demand for income-elastic goods and services. Such regions also tend to host a number of firms that are demanding customers in their own right, such as hospitals and high-technology firms with a high demand for new advanced technology. There is consequently a large demand for new advanced products from both consumers and producers. The demand for new advanced products in large urban regions has two implications. First, there is an especially great demand for imported products, since most new advanced products are first produced somewhere else. Second, most large regions offer good conditions for the development and introduction of new products since there is a spatial concentration of customers with a sufficient willingness to pay. We see a general pattern emerging. The larger the region is, the better are the conditions for innovation. And the better the conditions for innovation are, the more dynamic the region becomes. Of course, large urban regions are not equally dynamic, but there is nonetheless a strong tendency for the largest regions to be the most dynamic in a global sense. The names of the most successful dynamic regions are well known in both the scholarly and popular literature, and they coincide with those functional urban regions that have the greatest aggregate purchasing power. Many of these large urban regions have been dynamic and innovative for a long time. Still, historically there are plenty of examples of urban regions, which have lost much of their creative potential. One colorful description of how a dynamic region lost much of its innovative power can be found in Jacobs (1969), where she describes the fate of Manchester, which for a long time was the world’s leading innovative milieu in the textile industry. An interesting aspect of Jacobs’ analysis is that she contrasts Manchester with Birmingham, where Birmingham fared better because it was less dependent on (standardization-prone) large-scale manufacturing. We will not discuss any more examples of large urban regions, which over time seem to have become less dynamic. Instead, we turn to the problem of keeping regions innovative.

1 Innovation, Dynamic Regions and Regional Dynamics

1.4

21

Regional Dynamics

The international economic system contains large metropolitan regions, which serve as spatial focal points (Johansson and Karlsson 1990b). In small countries, it is common for a single urban region to develop into the only international gateway. In other cases, several large regions share the role of being gateways to a system of functional regions. Gateways specialize in importing recent ideas, inventions, and innovations. Besides being ‘‘sinks’’ for novel inputs from the world economy, gateways function as incubators for new products. Imports are important stimuli for product innovation. On the one hand, they may stimulate direct imitation. More importantly, imports may induce the product innovation, both in the form of incremental adaptations of the design and new complements. Given their dual nature as ‘‘sinks’’ and incubators, gateways are normally the most dynamic regions in each country. All gateways coordinate spatial customer networks, which link a set of peripheral nodes to the central gateway node. Peripheral nodes usually specialize in production for export. They receive information about new ideas and trends from the central gateway node. Export nodes sometimes have strong linkages to several more central nodes, including nodes in other countries. Multinational corporations are especially notable as facilitators of such international links. Not all large urban regions are gateways, and some gateways have mediumsized populations. For example, one group of large regions specializes in largescale industrial production for long-distance export. If we take a snapshot of a country and its system of functional urban regions, we can observe how the different regions have acquired specialized roles in the national and global economy by pursuing different development paths. The development path of an individual region is the result of a dynamic interplay between internal and external forces. Different paths additionally have path-dependent risks and uncertainties, since some – but not other – regional specializations may become technologically obsolete or uncompetitive due to new low-cost competitors in other parts of the world. At the same time, accumulated investments in specialized skills, capital, and institutions may create rigidities which make necessary restructuring both slow and difficult (Johansson and Karlsson 1990a). To understand how the regional specializations and patterns of interaction change over time, it is necessary to adopt a time scale that is long enough to accommodate cyclical changes in production patterns. The production of most goods and services evolves through a cyclical pattern where an initial expansion yields to standardization and, in the long run, obsolescence. In the production of standardized goods, only those firms survive that manage to cut production costs, either through relocation to regions with cost advantages or through process innovation. In the long run, however, it is only through the substitution of new goods for obsolete ones that the relative wealth of a region can endure (Johansson and Karlsson 1990b). Economic development depends on the pace and coordination of such renewal processes. In successful cases, the expansive phase gives rise to

22

C. Karlsson et al.

product and process innovations that cause temporary monopoly profits. Successful innovations not only cause temporary profits but also give rise to long-term productivity and employment gains. Regions without natural resource advantages must develop knowledge advantages to achieve high living standards. In order to observe the product life cycle it is necessary to devise a method that can distinguish between different products in such a way that each product has a definite market entry (birth) and market exit (death), to the extent that the latter occurs (Batten and Johansson 1989). With such a method, it also becomes possible to superimpose individual life cycles that result in aggregated cycles, which describe long-term economic waves and their associated spatial relocation waves. Such aggregation generates an image of long waves for entire product groups and technological families. The importance of this observation becomes significant when we note that during recent centuries it is possible to distinguish periods with identifiable technological shifts. Every such period of restructuring has intensified the initiation of new product cycles. What are then conditions for successful innovation, leading to a take-off and expansion of production as well as the initiation of a new product cycle? The simple answer is that production must be profitable enough to redirect resources from existing production, within the region and sometimes from other regions. This implies that entrepreneurs bid up the prices for land and labor. When the prices of production factors increase, firms that produce mature products discover that they suffer losses even if they introduce process innovations. Before its long, they face the choice between terminating and relocating production. Whatever they decide, the result is that resources become available for newer products. Recurrent structural change is a precondition for a region’s long-term viability. Even if higher wages induce a flow of labor to the innovative regions, the termination or relocation of older production is necessary in order to release land to the new, higher-valued, and thus more efficient use. In addition, improvements to the physical infrastructure make an increased density possible and leads to greater overall accessibility. The key point in this section is the importance of the out-migration of mature products from dynamic, innovative regions for the development of both innovative and imitating regions in a multi-regional system. In the next section, we will present two theories, which both offer dynamic explanations of the location behavior of firms in a system of functional urban regions.

1.4.1

Location Dynamics in a System of Functional Urban Regions

The filtering-down theory and the spatial product cycle theory provide alternatives to the neo-classical convergence theories. They both offer dynamic explanations of the location behavior of firms in a system of functional urban regions, employing – in

1 Innovation, Dynamic Regions and Regional Dynamics

23

the first case – a central place system or – in the second case – the concept of location advantage. Both theories assume that the development of a product or an industry follows a sequence with an introduction, a growth, and a maturation phase, that takes the form of an S-shaped growth curve. The life-cycle perspective makes it possible to see patterns in the continuously adjusting spatial structure of both intraregional and inter-regional development. In particular, the life-cycle concept seems to be a useful device for explaining location dynamics, especially in the case of inter-regional relocation processes for new products and industries (Aydalot 1984).8 However, we should note that some activities do not exhibit cyclical behavior. These activities are mainly non-standardizable activities such as customized delivery of goods and services, where each delivery has new and individual attributes (Forslund 1997). Even if there are many similarities between the two theories there is also one major difference: while the filtering-down theory stresses that products and industries filter down through the system of functional urban regions in a hierarchical manner from regions with larger market areas to regions with smaller market areas (Thompson 1969; Moriarty 1991), the spatial product cycle theory does not present any similar strict hypothesis concerning the spatial diffusion and relocation pattern as products age (Karlsson 1988). Both theories distinguish between the development of new (young) products and the production of mature standardized products with routine production. Both theories further assume that a high proportion of all new products are initiated or imitated (at an early stage) in the leading functional urban regions, with opportunityrich economic milieus and with substantial concentrations of knowledge resources (Johansson et al. 2006). Non-standard goods and services comprise customized deliveries as well as young products. Firms with such products find it advantageous to locate in large urban regions with good accessibility to diverse customer segments, R&D resources, and other suppliers of knowledge services. Other desirable location characteristics include a high purchasing power and good contact opportunities. When a product and its market mature, it often becomes possible to standardize its design and automate its production process. At this stage, its production depends less on metropolitan market, making production in other regions possible. If a relocation or diffusion of production takes places, it can take many different forms (Johansson and Karlsson 2003). Firms may relocate part or all of their production to other regions, but they may also outsource part or all of their production to one or several firms located in such other regions. Firms with production units in several regions may change its inter-regional division of labor. Chain-type firms may gradually expand production in different regions or franchise their business concept. In addition, firms in smaller regions may imitate products developed in large urban regions. According to the filtering-down theory, technical and demand changes induce firms to shift the location of the production of existing products (or product groups) 8

The continual self-reorganizing and evolution of the global spatial economy at a macro scale can also be analyzed by applying the ‘‘new economic geography approach’’ (Fujita and Mori 1998).

24

C. Karlsson et al.

over time, thereby transforming the specialization pattern of regions (Camagni et al. 1986). This makes the filtering-down process both market-driven and technologydriven. On the one hand, standardization of products and production may lower both the set-up costs and variable production costs. On the other hand, demand may increase due to increasing real incomes, changing preferences, and outsourcing of activities from firms and households. Consequently, the production of different products may gradually filter down or diffuse downwards in the hierarchy of functional urban regions. In this way, the filtering-down theory refers to products that spread from one level of the hierarchy to all functional regions at the next lower level. If we consider a new product with high spatial transaction costs that has been introduced in the largest region in an economy, it will start to filter down the system of functional urban regions when the real income and hence demand increases in smaller regions or when the set-up costs for production have decreased sufficiently. This process stops at some level in the hierarchy of urban regions, if it is inconceivable to mobilize enough demand to make production profitable. The spatial product cycle theory similarly assumes a relocation of production from the leading urban regions. However, the number of followers is limited, since economies of scale normally prevent decentralization to many regions, except in cases with very high spatial transaction costs. Changes in location are dependent on location advantages, even at later stages of the product cycle (Vernon 1960; Hirsch 1967; Andersson and Johansson 1984a). Hence, this theory stresses the importance of external economies for the location of production. When relocation does take place, it is limited to a small set of specialized regions. Localization economies are decisive and provide individual regions with their most important location advantages (Marshall 1920; Krugman 1991). Andersson and Johansson (1984a,b) use microeconomic models to show how product cycle assumptions generate location and relocation processes (see also Johansson and Karlsson 1986, 1987). Both papers demonstrate how clusters of product cycles can be observed empirically in the form of aggregate specialization patterns, which describe a time-space hierarchy. In a later contribution, the two authors reformulate their results into a more coherent framework and emphasize new directions for this type of model formulations (Johansson and Andersson 1998). In their later model, knowledge intensity, product standardization, and process routinization are key notions. Along a product cycle path, Johansson and Andersson assume that the knowledge intensity is high when a product is nonstandardized and the production process is non-routinized. Standardization and routinization imply reduced knowledge intensity. Andersson and Johansson proceed to present a class of models that explains this regularity. They then derive interdependencies between location dynamics and product cycles that incorporate elements from models of monopolistic competition. Moreover, they use notions of product and process vintages to classify structural properties and the associated markets. Products with relocated production usually have low spatial transaction costs and in this case production may be relocated for defensive as well as offensive reasons. The occurrence of new locations indicates that the product is no longer

1 Innovation, Dynamic Regions and Regional Dynamics

25

new, but the production process may continue to be renewed (Johansson 1998b). When the product is in its growth phase it may be too expansive to expand production in large urban regions and thus new locations are sought for the organization of large-scale production. Since the scale of production increases, there are strong preferences for locations offering good accessibility in the national and international logistical systems. At later stages of the product cycle, cost considerations become more important and production relocates to more peripheral regions or to regions abroad. Comparative advantages in the case of the spatial product cycle are often in regions that have lower land and property prices as well as lower costs of unskilled labor inputs (Andersson and Johansson 1984a,b; Johansson and Stro¨mquist 1986; Johansson 1993b).

1.4.2

Lead–Lag Models

Johansson (1993b) emphasizes the dynamics of product vintages as the force that drives the behavior of filtering-down and spatial product cycle models – an assumption that forms the foundation for empirical lead–lag models9 (see also, Forslund and Johansson 1995; Forslund 1996, 1997; Johansson and Karlsson 2003). The lead–lag model has the specific objective of generating hypotheses, which can be tested empirically. It classifies economic activities in such a way that it is possible to refer to them as clusters of products with synchronized location dynamics.10 For a given system of functional urban regions, the model specifies – for each type of product group (industry) – its average share of all product groups (industries) in the system of functional regions (measured as employment or value added). The model identifies a specific leading region, for a given system of functional regions. The relative industry shares for the leading region are predictive indicators. The basic hypotheses in lead–lag models are associated with the location leadership of the leading region. The first hypothesis states that product groups (industries) with high relative shares in the leading region should be expected to grow in other regions in the system of functional regions. The second hypothesis states that product groups (industries) with low relative shares in the leading region should be expected to decline in other regions in the system of functional regions. The first hypothesis implies that new product groups (industries) originate in the leading region. The second hypothesis implies that leading regions lose employment in mature product groups (industries) before other regions. Hence, they are leading regions also as regards the decline of products.

9

For example, the ‘‘flying geese model’’ proposed by Fujita and Mori (1998) can be considered as a special case of the more general lead–lag model. 10 The lead–lag model does not apply to activities, which have to be harvested in the region where they are located. The location of such production is analysed by standard location advantage models, where the comparative advantages are of Ricardo type.

26

C. Karlsson et al.

The above basic hypotheses yield a number of sub-hypotheses. For example, industries with both high relative shares and fast growth rates in the leading region are non-routinized activities and have non-standardized products that compete on the basis of product rather than price. They also tend to involve research and other knowledge production. Industries with low relative shares are on the other hand routinized activities that compete on the basis of price that aim at reducing the labor input coefficient. Lead–lag models assume that a high proportion of new products are initiated or imitated (at an early stage) in the leading region in the system of functional regions. As the production expands, products are frequently standardized and production techniques routinized, which is referred to as product vintage dynamics. As new product vintages are introduced, the pertinent activities are relocated or diffused within the system of functional regions. Analyzing vintage process dynamics as the driving force in spatial product-cycle and filtering-down models, it is possible to show that a gradual change in location takes place. The follower regions host industries for which the vintage renewal is dominated by standardization and routinization (Forslund and Johansson 1998). The main point we want to stress here is that the economy of nations and regions is rejuvenated when production from large urban regions relocates to successively smaller regions. Of course, the establishment of new production in these regions will also generate structural change. The new production units will tend to bid up prices for land and labor in these regions, thereby making the previous marginal production activities unprofitable. To the extent that labor relocates to large regions when new product cycles emerge, the structural changes in medium-sized and smaller regions may be even more pronounced.

1.5

Content

The subsequent chapters of this book are arranged in a sequence running from theoretical to empirical perspectives and ending with a chapter written by Bo¨rje Johansson.

1.5.1

Theoretical Contributions

In Chap. 2, Martin J. Beckmann develops a systematic theory of spatial markets. It is expositional, drawing on previous work by T.C. Koopmans, the author himself and others of the ‘‘efficient allocation school.’’ The underlying theme is the familiar one of showing how prices (indexed by location) in competitive markets can serve to guide the allocation of resources in space, and the deviations from this optimum that occur under various types of institutions. To¨nu Puu in Chap. 3 reconsiders the Smith–Ricardo paradigm of complete specialization and comparative advantage. He shows how minute effects arising from the monotony of repetitive work shatter linear theories.

1 Innovation, Dynamic Regions and Regional Dynamics

27

The ultimate concern of Chap. 4 by Masahisa Fujita is the further development of the New Economic Geography towards a more comprehensive theory of spatial economics in the age of the brain-power society, in which the dynamics of the spatial economy arise from the dual linkages in the economic and knowledge fields. ˚ ke E. Andersson and David Emanuel Andersson the mechanisms In Chap. 5 by A of creativity are discussed as well as individual and social aspects of creative organizations. They show that because of the public nature of creativity there are increasing returns to scale in many creativity-based production systems, and that economic incentives to promote creativity are complicated by the importance of uncertainty and the importance of ‘‘star performers.’’ David F. Batten and Roger Bradbury in Chap. 6 present the view that the building blocks of regions and regional policies such as ideas, actions, habits, skills, and inventions are akin to selfish Darwinian entities – memes – that, like genes, interact and replicate in complex ways with humans to shape our culture. Whether good or bad, a single, omnipotent meme can dominate a local region of meme-space. The objective of Chap. 7 by Lata Chatterjee and T.R. Lakshmanan is to briefly describe the process by which entrepreneurial cities fashion or socially create their dynamic competitive advantages, which underlie their ability to function and thrive in the new global economy. The authors argue that three autonomous and independent urban sectors – economic, political and social – are involved in the joint production and maintenance of urban dynamic competitive advantage. Chapter 8, by Hans Westlund, deals with social capital as an extra-market externality, and its role for innovation and growth. The chapter provides analyses of the changes in innovation activity over time from early industrialism to the global knowledge economy, how the relations between the actors of today’s innovation systems have developed and the role of social networks for innovations. Kingsley E. Haynes, Rajendra Kulkarni and Roger Stough in Chap. 9 use information methods to describe the patterns in urban freeway traffic flows in order to analyze ‘‘hidden order’’ in such high volume congested systems. They introduce and develop a method for measuring order in linear flow patterns based on Kolmogorov entropy. Chapter 10 by John R. Roy and Geoffrey J.D. Hewings deals with multi-regional input–output analysis where five sets of component flows are jointly determined, thus ensuring that observed flows, which contain all these components, are consistent with the flows being modeled. In addition, rather than assuming just a single abstract path between each pair of regions, feasible multiple paths are assumed.

1.5.2

Empirical Contributions

In Chap. 11, Artem Korzhenevych and Johannes Bro¨cker study the implications of factor mobility and wage rigidity assumptions for the evaluation of infrastructure policy effects using a multiregional computable general equilibrium model. Their

28

C. Karlsson et al.

modeling results suggest that the introduction of wage rigidity matters for inference, expanding the range of values notably. Makato Tsukai and Kiyoshi Kobayashi in Chap. 12 stress the lack of studies that focus on the persistent (lagged) production effect of infrastructure. To remedy this lack, the authors present an approach where the lagged effects of infrastructure productivity and technological innovation are specified as a multiple time series model with a long persistent effect. Chapter 13, by Roberta Capello and Andrea Morrison, aims to measure the effectiveness of science parks in fostering knowledge transfer processes at the local level. Their empirical analysis provides prima facie evidence that Science Parks play a role in creating relationships among local actors while they perform their gate-keeping function rather poorly. In Chap. 14, Enno Masurel and Peter Nijkamp address the lack of institutional collaboration among urban ethnic (or migrant) firms as a reason for their low innovation profile. Poor communication, a low chance to be accepted by the external party, and economic market factors appear to be important reasons why ethnic entrepreneurs do not join franchise organizations. The purpose of Chap. 15 by Charlie Karlsson and Martin Andersson is to analyze the location relationship between industry and university R&D in Sweden using a simultaneous equation approach. Their results indicate that the location of industrial R&D is quite sensitive to the location of university R&D and that the location of university R&D is sensitive to the location of industrial R&D. Chapter 16 by Paul Cheshire and Stefano Magrini investigates growth differences in the urban system of the EU12 over the last decades of the twentieth century. It contrast the set of factors associated with population growth with those associated with growth in output per head. There are several factors common to both types of growth but while economic growth is associated with factors driving innovation and productivity such as highly skilled human capital and concentrations of R&D, population growth responds strongly to differences in climate but only to differences within countries. Johan Klaesson and Lars Pettersson in Chap. 17 analyze the influence of urban size on the development of neighboring rural population and employment. Employing a Carlino-Mills type of model they find that working age population is the most significant factor for explaining changes in the working-age population. Generally, their analysis seems to support the hypothesis that ‘‘jobs follow people.’’ In Chap. 18, Manfred M. Fischer, Thomas Scherngell and Eva Jansenberger analyze patent citation data pertaining to high-technology firms in Europe to test the extent of knowledge spillovers. Using the case-control matching method they find strong evidence of geographic localization at two different spatial levels (country and region), even after controlling for the tendency of inventive activities in hightechnology sectors to be geographically clustered. The aim of Chap. 19 by Katarina Larsen is to increase our understanding of science output, structure, and impact within the area of nano-structured dyesensitized solar cells. The results are based on co-authorship data and interpretative

1 Innovation, Dynamic Regions and Regional Dynamics

29

analyses of the Swedish network hub and point to the importance of factors such as citation window and journal impact factor for citation eminence. Ian Gordon, Colin Haslam, Philip McCann and Brian Scott-Quinn in Chap. 20 first discuss the current activity and employment base of London’s financial center in relation to the kinds of capacity that is developing in offshore centers (particularly in India), and then examines the approaches which City investment banks are currently adopting to these issues. The objective of Chap. 21 by Pontus Braunerhjelm is to shed new insights on the forces that prompt the emergence of clusters, and how these forces interact with more well-known mechanisms to enforce and sustain existing clusters, using the Stockholm music cluster as an example.

References ˚ E (1985) Kreativitet: Storstadens framtid. Prisma, Stockholm Andersson A ˚ E (1986) The four logistical revolutions. Pap Reg Sci Assoc 59:1–12 Andersson A ˚ E, Andersson DE (eds) (2000) Gateways to the global economy. Edward Elgar, Andersson A Cheltenham ˚ E, Johansson B (1984a) Knowledge intensity and product cycles in metropolitan Andersson A regions, WP-84-13. IIASA, Laxenburg ˚ E, Johansson B (1984b) Industrial dynamics, product cycles, and employment Andersson A structure, WP-84-9. IIASA, Laxenburg Andersson AE, Marksjo¨ B (1972) General equilibrium models for allocation in space under interdependence and increasing returns of scale. Reg Urban Econ 2:133–158 Arrow KJ, Hahn F (1971) General competitive analysis. Holden-Day, San Francisco Aumann R (1964) Markets with a continuum of traders. Econometrica 32:39–50 Aydalot P (1984) Reversal of spatial trends in French industry since 1974. In: Lambooy JG (ed) New spatial dynamics and economic crisis, IRPA Yearbook 1984. Finn, Tampere, pp 41–62 ˚ E et al (eds) Batten D, Johansson B (1989) Dynamics of product substitution. In: Andersson A Advances in spatial theory and dynamics. North-Holland, Amsterdam, pp 23–44 Beckmann M (1952) A continuous model of transportation. Econometrica 20:643–660 Beckmann M (1956) Studies in the economics of transportation. Yale University Press, New Haven Beckmann M (1958) City hierarchies and the distribution of city sizes. Econ Dev Cult Change 6:243–248 Beckmann M (1996) The location of market oriented industries in a growing economy. Paper presented at the 5th World Congress of the RSAI in Tokyo Beckmann M, Puu T (1985) Spatial economics: density, potential and flow. Springer, Berlin Benassy J-P (1975) Neo-Keynesian disequilibrium theory in a monetary economy. Rev Econ Stud 42:503–524 Camagni R, Diappi L, Leonardi G (1986) Urban growth and decline in a hierarchical system – a supply oriented dynamic approach. Reg Sci Urban Econ 16:145–160 Cassel G (1917) Dyrtid och sedelo¨fverflo¨d. PA Norstedt, Stockholm Chamberlain E (1936) The theory of monopolistic competition. Harvard University Press, Cambridge, MA Ciccone A, Hall RE (1996) Productivity and the density of economic activity. Am Econ Rev 86:54–70 Debreu G (1956) Market equilibrium. Cowles Foundation Discussion Papers, 10, vol 42. Cowles Foundation, Yale University, New Haven, CT, pp 876–878

30

C. Karlsson et al.

Duranton G, Overman HG (2005) Detailed location patterns of UK manufacturing industries. London School of Economics, London (mimeo) Fischer MM, Johansson B (1994) Networks for process innovation by firms: conjectures from observations in three countries. In: Johansson B, Karlsson C, Westin L (eds) Patterns of a network economy. Springer, Berlin, pp 261–274 Forslund UM (1996) Industrial location – interregional leads and lags in Sweden and Norway, ¨ stersund Working Paper No. 1. Swedish Institute for Regional Research, O Forslund UM (1997) Studies of multi-regional interdependencies and change, Licentiate Thesis. Division of Regional Planning, Royal Institute of Technology, Stockholm Forslund UM, Johansson B (1995) Assessing road investments: accessibility changes. Cost benefit and production effect. Ann Reg Sci 29:155–174 Forslund UM, Johansson B (1998) Specialization dynamics in a system of functional regions. Paper presented at the western regional science meeting in Monterey, CA Frank CR (1969) Production theory and indivisible commodities. Princeton University Press, Princeton Fujita M, Mori T (1998) On the dynamics of frontier economies: endogenous growth or selforganization of a dissipative system. Ann Reg Sci 32:39–62 Glaeser E, Ellison G (1997) Geographic concentration in US manufacturing industries: a dartboard approach. J Polit Econ 105:889–927 Greenhut ML (1971) A theory of the firm in economic space. Austin, Austin, TX Hacker RC, Johansson B, Karlsson C (2004) Emerging market economies in an integrating europe – an introduction. In: Hacker RC, Johansson B, Karlsson C (eds) Emerging market economies and European economic integration. Edward Elgar, Cheltenham, pp 1–27 Hirsch S (1967) Location of industry and international competitiveness. Oxford University Press, Oxford Holmberg I, Johansson B (1992), Growth of production, migration of jobs and spatial infrastructure, Working Paper 1992-6, Regional Planning, Royal Institute of Technology, Stockholm Holmberg I, Johansson B, Stro¨mquist U (2003) A simultaneous model of long-term regional job ˚ E, Johansson B, Andersson WP (eds) The economics and population changes. In: Andersson A of disappearing distance. Ashgate, Aldershot, pp 161–189 Hotelling H (1929) Stability and competition. Econ J 39:41–57 Isard W (1956) Location and space-economy. Wiley, New York Jacobs J (1969) The economy of cities. Vintage, New York Johansson B (1978) Contributions to sequential analysis of oligopolistic competition. Dissertation, Department of Economics, University of Gothenburg Johansson B (1985) Dynamics of metropolitan processes and policies. Scand Housing Plann Res 2:115–123 Johansson B (1987) Technological vintages and substitution processes. In: Batten D, Casti J, Johansson B (eds) Economic evolution and structural adjustment. Springer, Berlin, pp 145–165 Johansson B (1989a) Metropolitan nodes in the innovation networks of the Nordic economies. In: The long term futures of regional policy – a Nordic View, Report on a joint NordREFO/OECD Seminar, Borga˚, pp 98–118 Johansson B (1989b) Economic development and networks for spatial interaction, CERUM Working Paper 1989:28. Umea˚ University, Umea˚ Johansson B (1990a) Innovation processes in the urban network of export and import nodes: a Swedish example. In: Nijkamp P (ed) Sustainability of Urban Systems. Avebury, Alderhot, pp 219–245 Johansson B (1990b) Det osynliga handslaget. Ekonomiska na¨tverks organisation och sja¨lvorganisation. In: Karlqvist A (ed) Na¨tverk. Teorier och begrepp i samha¨llsvetenskapen. Gidlunds, Va¨rnamo, pp 60–96 Johansson B (1991a) Transportation infrastructure, productivity and growth. In: Thord R (ed) The future of transportation and communication. Swedish National Road Administration, Borla¨nge, pp 163–176

1 Innovation, Dynamic Regions and Regional Dynamics

31

Johansson B (1991b) Economic networks and self-organization. In: Bergman E, Maier G, To¨dtling F (eds) Regions reconsidered. Economic networks, innovation, and local development in industrialized countries. Mansell, London, pp 17–34 Johansson B (1993a) Ekonomisk dynamik i Europa. Na¨tverk fo¨r handel, kunskaps-import och innovationer. Liber-Hermods, Malmo¨ ˚E Johansson B (1993b) Economic evolution and urban infrastructure dynamics. In: Andersson A et al (eds) The cosmo-creative society. Springer, Berlin, pp 151–175 Johansson B (1993c) Infrastructure, accessibility and economic growth. Int J Transport Econ 20:131–156 Johansson B (1996) Location attributes and dynamics of job location. J Infrastructure Plann Manage 530(IV-30):1–15 Johansson B (1997) Regional differentials in the development of economy and population. In: So¨rensen C (ed) Empirical evidence of regional growth: the centre–periphery discussion. The Expert Committee of the Danish Ministry of the Interior, Copenhagen, pp 107–162 Johansson B (1998a) Infrastructure and economic milieu: Swedish contributions 1960–1995. In: Lundqvist L, Mattsson L-G, Kim TJ (eds) Network infrastructure and the urban environment. Springer, Berlin, pp 19–52 Johansson B (1998b) Economic milieu, specialization and location dynamics. In: Reggiani A (ed) Accessibility, trade and location behaviour. Ashgate, Aldershot, pp 121–152 Johansson B (2005) Parsing the menagerie of agglomeration and network externalities. In: Karlsson C, Johansson B, Stough RR (eds) Industrial clusters and inter-firm networks. Edward Elgar, Cheltenham, pp 107–147 Johansson B (2006) Spatial clusters of ICT industries. In: Johansson B, Karlsson C, Stough RR (eds) The emerging digital economy. Entrepreneurship, clusters and policy. Springer, Berlin, pp 137–167 ˚ E (1998) Schloss Laxenburg model of product cycle development. Johansson B, Andersson A In: Beckmann MJ et al (eds) Knowledge and networks in a dynamic economy. Springer, Berlin, pp 181–219 Johansson B, Holmberg I (1982) A regional study of the distribution of vintages and profits of ˚ E, industrial establishments: a stochastic transition model. In: Albegov M, Andersson A Snickars F (eds) Regional development modelling: theory and practice. North-Holland, Amsterdam, pp 215–227 Johansson B, Karlsson C (1986) Industrial applications of information technology: speed of introduction and labour force competence. In: Nijkamp P (ed) Technological change, employment and spatial dynamics. Springer, Berlin, pp 401–428 Johansson B, Karlsson C (1987) Processes of industrial change: scale, location and type of job. In: Fischer MM, Nijkamp P (eds) Regional labour markets. North-Holland, Amsterdam, pp 139–165 Johansson B, Karlsson C (1990a) Stadsregioner i Europa, SOU 1990:34. Allma¨nna fo¨rlaget, Stockholm Johansson B, Karlsson C (1990b) Evolving technological patterns in a Nordic perspective: an introduction to the volume. In: Johansson B, Karlsson C (eds) Innovation, industrial knowledge and trade. A Nordic perspective. CERUM and Institute for Futures Studies, Umea˚, pp 1–17 Johansson B, Karlsson C (1994) Transportation infrastructure for the Ma¨lar region. Reg Stud 28:169–85 Johansson B, Karlsson C (2001) Geographic transaction costs and specialisation opportunities of small and medium-sized regions: scale economies and market extension. In: Johansson B, Karlsson C, Stough RR (eds) Theories of endogenous regional growth – lessons for regional policies. Springer, Berlin, pp 150–180 Johansson B, Karlsson C (2003) Va¨xande branscher – om Stockholmsregionens samspel med o¨vriga landet, Storstadspolitik 8:2003. Regionplane- och trafikkontoret, Stockholm Johansson B, Lo¨o¨f H (2006) Globala FoU-fo¨retag I Sverige. Vilken betydelse har de fo¨r ekonomins utvecklingskraft? In: Johansson D, Karlsson N (eds) Svensk utvecklingskraft. Ratio, Stockholm, pp 127–153

32

C. Karlsson et al.

Johansson B, Marksjo¨ B (1984) Interactive system for regional analysis of industrial sectors. In: Nijkamp P, Rietveld P (eds) Information systems for integrated regional planning. NorthHolland, Amsterdam, pp 231–249 Johansson B, Quigley J (2004) Agglomeration and networks in spatial economics. Pap Reg Sci 83:165–176 Johansson B, Snickars F (1992) Infrastruktur, BT33:1992, Byggforskningsra˚det, Stockholm Johansson B, Stro¨mquist U (1981) Rigidities in the process of structural economic change. Reg Sci Urban Econ 11:336–375 Johansson B, Stro¨mquist U (1986) Teknikspridning och importsubstitution – Stockholmsregionens roll fo¨r svensk teknikfo¨rnyelse, Rapport 1986 nr 7:2. La¨nsstyrelsen i Stockholms la¨n, Stockholm Johansson B, Westin L (1994) Revealing network properties of Sweden’s trade with Europe. In: Johansson B, Karlsson C, Westin L (eds) Patterns of a network economy. Springer, Berlin, pp 125–141 Johansson B et al (1991) Infrastruktur och produktivitet, Expertutredning Nr 9 till Produktivitetsdelegationen. Allma¨nna Fo¨rlaget, Stockholm ˚ berg P (1998) Regioner, handel och tillva¨xt. RTK, Stockholms la¨ns Johansson B, Stro¨mquist U, A landsting, Stockholm Johansson B, Karlsson C, Stough RR (2006) Entrepreneurship, clusters and policy in the emerging digital economy. In: Johansson B, Karlsson C, Stough RR (eds) The emerging digital economy. Entrepreneurship, clusters and policy. Springer, Berlin, pp 1–19 Karlsson C (1988) Innovation adoption and the product life cycle. Umea˚ economic studies no. 185. University of Umea˚, Umea˚ Karlsson C, Johansson B (2006) Towards a dynamic theory for the spatial knowledge economy. In: Johansson B, Karlsson C, Stough RR (eds) Entrepreneurship and dynamics in the knowledge economy. London, Routledge, pp 12–46 Karlsson C, Pettersson L (2005) Regional productivity and accessibility to knowledge and dense markets. CESIS working paper 32. Jo¨nko¨ping International Business School, Jo¨nko¨ping Karlsson C, Johansson B, Stough RR (2005) Industrial clusters and inter-firm networks – an introduction. In: Johansson B, Karlsson C, Stough RR (eds) Industrial clusters and inter-firm networks. Edward Elgar, Cheltenham, pp 1–25 Keynes JM (1936) The general theory of employment interest and money. Macmillan, London Krugman P (1991) Geography and trade. MIT, Cambridge, MA Lancaster K (1971) Consumer demand – a new approach. Columbia University Press, New York Launhardt CWF (1872) Kommercielle Tracirung der Verkehrswege. Zeitschrift des Architectenund Ingenieur-Vereins Hannover 18:515–534 Launhardt CWF (1882) Die Bestimmung des Zweckma¨ssigsten Standortes einer Gewerblichen Anlage. Zeitschrift des Vereines Deutscher Ingenieure 26:105–116 Lo¨sch A (1954) The economics of location. Yale University Press, New Haven Maillat D, Crevoisier O, Lecoq B (1993) Innovation networks and territorial dynamics: a tentative typology. In: Johansson B, Karlsson C, Westin L (eds) Patterns of a network economy. Springer, Berlin, pp 33–52 Marshall A (1920) Principles of economics. Macmillan, London ˚ E, Batten DF, Mills ES, Carlino G (1989) Dynamics of county growth. In: Andersson A Johansson B (eds) Advances in spatial theory and dynamics. North-Holland, Amsterdam, pp 195–206 Moriarty B (1991) Urban systems, industrial restructuring and the spatio-temporal diffusion of manufacturing employment. Environ Plann A 23:1571–1588 Moroney JR, Walker JM (1966) A regional test of the Heckscher–Ohlin theorem. J Polit Econ 74:573–586 Nelson RR, Winter SG (1982) An evolutionary theory of economic change. Harvard University Press, Cambridge, MA Ohlin B (1933) Interregional and international trade. Harvard University Press, Cambridge, MA Palander TF (1935) Beitra¨ge zur Standortstheorie. Almqvist and Wicksell, Uppsala

1 Innovation, Dynamic Regions and Regional Dynamics

33

Scarf HE, Hansen T (1973) The computation of economic equilibria. Yale University Press, New Haven Scherer FM (1984) Innovation and growth – Schumpeterian perspectives. MIT, Cambridge, MA Schumpeter JA (1908) Das Wesen und der Hauptinhalt der Theoretischen Nationalo¨konomie. Duncker&Humblot, Leipzig Schumpeter JA (1934) The theory of economic development. Oxford University Press, New York Smith B (1975) Regional specialization and trade in the UK. Scott J Polit Econ 22:39–56 Teubal M, Zuscovitch E (1994) Demand revealing and knowledge differentiation through network evolution. In: Johansson B, Karlsson C, Westin L (eds) Patterns of a network economy. Springer, Berlin, pp 15–31 Thompson WR (1969) The economic base of urban problems. In: Chamberlain NW (ed) Contemporary economic issues. Pichard Irving, Homewood, IL, pp 1–47 Tinbergen J (1967) The hierarchy model of the size distribution of centres. Pap Reg Sci Assoc 20:65–68 Uzawa H (1962) Aggregative convexity and the existence of competitive equilibrium. Econ Stud Q 12:52–60 Uzawa H (1976) Disequilibrium analysis and Keynes’s general theory Vernon R (1960) Metropolis 1985. Harvard University Press, Cambridge, MA Von Thu¨nen JH (1826) Der isolierte Staat in Beziehung auf Landwirtschaft und Nationalo¨konomie. Perthes, Hamburg ¨ ber die Eindeutige positive Lo¨sbarkeit der neuen Produktionsgleichungen. Wald A (1933–34) U Ergebnisse eines Mathematischen Kolloquiums 6:12–20 ¨ ber die Produktionsgleichungen der o¨konomischen Wertlehre. Ergebnisse Wald A (1934–35) U eines Mathematischen Kolloquiums 7:1–6 Weber A (1929) Theory of the location of industries. The University of Chicago Press, Chicago

Chapter 2

The Pure Theory of Spatial Markets Martin Beckmann

2.1

Introduction

Spatial phenomena which had been topics in theoretical Geography and Location Economics were seen as common objects in the new Regional Science of the 1950s. A point of crystallization was the notion of spatial markets going back to Wilhelm Launhardt (1886), who perceived them, as market (and supply) areas. Market areas are territories in which a given firm is the nearest and ceteris paribus the cheapest and hence exclusive supplier. This essay develops a systematic theory of spatial markets. It is thus an expository, drawing on previous work of T.C. Koopmans, Martin Beckmann and others of the ‘‘efficient allocation’’ school. The underlying theme is the familiar one of showing how prices (indexed by location) obtained in competitive markets can guide allocation of resources in space, and the deviations from this optimum that occur under various types of institutions. While thus backward looking, spatial markets can still provide an organizing framework to our contemporary interest in innovation. When buyer and seller are not in the same place, distance intervenes and transaction costs for transportation and/or communication arise, ordinary market theory no longer applies, e.g., ‘‘law of the single price’’ is violated. We are then, faced with spatial markets. This is actually among the oldest subjects treated in location theory. In Von Thu¨nen’s ‘‘Isolated State’’ land use, production and sales to a metropolitan market are investigated as functions of distance (Von Thu¨nen 1826). Wilhelm Launhardt’s location theory is grounded on market areas as sets of exclusive sales in a territory surrounding a firm or localized industry (Launhardt 1886). In this paper we offer a modern perspective of the theory of spatial markets in perfect competition. We exclude spatial price policies under monopoly, which have a rich literature of their own (Greenhut and Ohta 1975; Beckmann 1976). Market

M. Beckmann Senior Academy Secretary, Brown University

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_2, # Springer‐Verlag Berlin Heidelberg 2009

35

36

M. Beckmann

strategies of oligopolists aimed at defence or penetration of market areas, in which pricing is confounded with locational choice are also beyond the scope of this paper. The pure theory of spatial markets considers only those questions that exist when locations are given. In essence, it is location theory without locational choice.

2.2

The Transportation Problem of Linear Programming

Traditionally, spatial markets have been classified as either market or supply areas centered on an exporting or importing location. But these do not exhaust all conceivable market configurations. At any rate spatial structures like these should be discovered, not assumed, by spatial economic theory. The modern approach to spatial markets takes off from the Transportation Problem of Linear Programming (Koopmans 1949). We reformulate it in terms of given constant excess supplies q. In a set of locations i, j, k ¼ 1, . . . , n, allowing transhipments in the presence of necessary restrictions of flows to a given transportation network N, so that in all summations it is understood that indices run only over locations in the network. Our interest is in flows xij of a commodity from point location i to point location j at unit transportation costs rij that will satisfy a commodity balance condition to meet excess demand qi X xji  xij ¼ q: ð2:1Þ j

Competitive market equilibrium, achieving a Pareto optimum, must then minimize total transportation costs X min rij xij : ð2:2Þ xij 0

i;j

This linear program (2.1) and (2.2) is feasible provided X qi ¼ 0;

ð2:3Þ

i

i.e., aggregate excess demand must be zero. An optimal solution, i.e., a market equilibrium is characterized by necessary and sufficient ‘‘efficiency conditions’’ (Koopmans 1949). The equilibrium involves ‘‘dual variables’’ pi that can be interpreted as competitive market prices. The efficiency conditions are     ¼ < x^ij ð2:4Þ 0 , pj  pi r :  ¼ ij In equilibrium the transactions x^ij are ‘‘efficient’’ and thus recover their transportation cost rij, while any of the excluded inefficient transactions do not. The

2 The Pure Theory of Spatial Markets

37

remarkable thing is not that this should be necessary, but that this is sufficient and together with constraints (2.1) will determine all valid equilibria. The efficiency conditions (2.4) now imply pi ¼ max pj  rij ;

ð2:4aÞ

pj ¼ min pi þ rij :

ð2:4bÞ

j

i

These equations state that sellers in i export to buyers j in order to maximize prices pi received in i. Buyers in j look for the cheapest sources i after transportation costs rij. A trading post j may both buy from cheapest sources i and sell to best buyers k min pi þ rij ¼ pj ¼ max pk  rjk : i

ð2:4cÞ

k

Geographically the areas containing sellers (sources) or buyers (sinks) may, but need not, overlap. In the fur trade the sources were Indians in the wilderness and the buyers, residents of Europe. Observe, that since only price differences occur in (2.4) the price level is indeterminate, a consequence of the feasibility equation (2.3). The ‘‘primal variables’’ x^ij satisfying (2.1)–(2.4) need not be unique. A theorem assures the existence of an optimal trade pattern with at most n1 flows, where n is the number of locations. This means that the flow system can form a tree or a set of trees, each being an independent market by itself. A set of demand locations j receiving from a single supplier i is i’s market area and a set of supply locations i for a single demand location j is j’s supply area. In addition there may be locations that will both import and export (forward). These may be called transit points or trading posts: they are common on given transportation networks. An interesting identity relates the minimal transportation cost T to the price system X X X X   X T¼ rij x^ij ¼ pj  pi x^ij ¼ pi pi qi ð2:5Þ x^ji  x^ij ¼ i; j

i; j

i

j

i

using (2.1) and (2.4). This identity is a part of the Duality Principle of LP (Dantzig 1959) X X T ¼ min rij xij ¼ max pj qj pj

Xij

s.t:

X

j

xji  xij ¼ qi

j

s.t: pj  pi  rij :

ð2:5aÞ

38

2.3

M. Beckmann

Braess’s Paradox

Expanding a program by simultaneously adding an amount a to supply at some location i and to demand at some location j can reduce total transportation cost T. Proof: Change qi at a supply location qi <0 with pi >pj to qi a and at a demand location qj >0 to qj +a to obtain T¼api +api ¼a (pj pi)<0. Figure 2.1 shows an example. Increasing supply at location d by one unit and demand in a by one unit increases ‘‘cheap’’ shipments from d to c and reduces (expensive) shipments from b to c by one unit while raising ‘‘cheap’’ shipments from b to a for a total saving of 2. Conversely decreasing supply and demand simultaneously, even to the point of eliminating locations can increase system transportation cost.

2.4

Relaxation

The rigid relationship between excess demand and net imports may be relaxed so that demand requirements are always covered and supply availabilities are never exceeded by setting X xji  yij  qi : ð2:1aÞ j

The feasibility condition (2.3) is now relaxed to X qi  0:

ð2:3aÞ

i

Aggregate excess demand must be non-positive. This causes (efficiency) prices to be non-negative, and zero in the case of supply under-utilization through an additional efficiency condition     X ¼ > xji  xij ð2:6Þ 0, q: pi  ¼ i j

b

a

5

Fig. 2.1 Point a¼1, point b¼0, point c¼5, point d¼4

c

1

d

2 The Pure Theory of Spatial Markets

39

^ following holds If ‘‘<’’ in (2.3a), this means that at some location ithe X

xji^  xij^ > qi^:

ð2:6aÞ

j

In view of (2.4b) and rij >0 this cannot happen at locations of positive (excess) demand; it must occur in a location of excess supply. The event (2.6a) will fix the price level pi^ ¼ 0 and make prices uniquely determined. Because of the ‘‘complementary slackness’’ in conditions (2.4) and (2.6) the identity (2.5) of (minimal) transportation cost and priced excess demand remains valid. The inclusion of production cost X

min xij 0

rij xij þ

i;j

X

hi ðzi Þzi ;

ð2:7Þ

i

zi 0

X

xji  xij  qi  zi ;

ð2:7aÞ

j

where zi is the production in location and hi(zi) is the cost function, or more simply, with constant unit cost hi in the cost minimization description of spatial market equilibrium (2.2) is straightforward. The results are included in the next section.

2.4.1

Flexible Demand

As long as demand is given, independent of price, competitive market equilibrium can be described as achieved by aggregate cost minimization. When demand is flexible, viz. price dependent, we must resort to welfare maximization. There are two ways of constructing an appropriate welfare function: either using the inverse of a price dependent excess demand function, or more directly as total utility, in money terms, minus aggregate costs. That utility may be expressed in money units and hence made interpersonally comparable while perhaps objectionable to economic purists, is accepted practice in applied micro-economics. We sketch the first and elaborate the second approach. Let qi ðpÞ

ð2:8Þ

be the excess demand function for location i, assumed to be strictly decreasing with price p. Its inverse exists, say, pi ðqÞ

ð2:8aÞ

40

M. Beckmann

and is also strictly decreasing. The integral Z v i ð qi Þ ¼

qi

pi ðqÞdq

ð2:9Þ

0

can then be shown to represent the sum of consumers’ and producers’ surplus vi when excess demand equals qi. Purists’ objections to the consumers’ surplus notwithstanding, it is this welfare measure max xij  0

X

vi

i

X

! xji  xij



X

rij xij

ð2:10Þ

j

which is maximized with respect to the flows xij in spatial market equilibrium (Samuelson 1952). In fact (2.10) yields x^ij

    ¼ < 0 , pj  pi r ; < ¼ ij X

x^ji  x^ij ¼ qi ðpi Þ:

ð2:4Þ ð2:11Þ

j

In the second approach utility ui(q) of consuming a quantity q in location i of the good under consideration is considered, with the usual assumptions u0i > 0;

u00i  0:

ð2:12Þ

Together with convex costs (say) we consider the welfare function X X   X uj qj  hi ð z i Þ  rij xij : j

i

ð2:13Þ

i;j

Maximizing welfare function with the linear constraints qi ¼ z i þ

X

xji  xij

ð2:1bÞ

j

is a well-behaved concave nonlinear program (NLP) yielding the Kuhn–Tucker conditions in terms of dual variables or prices pi     ¼ 0 < 0 , ui p; qi  ¼ i

ð2:14Þ

    ¼ < 0 , h0i p;  ¼ i

ð2:15Þ

zi

2 The Pure Theory of Spatial Markets

x^ij

    ¼ < 0 , pj  pi 0:  ¼

41

ð2:4Þ

For simplicity and in the tradition of location theory let utility be a square of consumption q and thus demand a linear function, qi ¼ai pi, and costs be linear h(zi)¼bi + hizi     ¼ > 0 , hi p zi  ¼ i

ð2:16Þ

(for standardized quantity and price units). Then (2.14) becomes     ¼ < 0 , a i  qi p qi  ¼ i

ð2:14aÞ

qi ¼ maxð0; ai  pi Þ:

ð2:14bÞ

or simply

Demand is a (piece-wise) linear function of price. Depending on climatic or other regional conditions, or by longstanding local custom, low values ai of utility may prevail which exclude local demand for a good even when its price is low. By contrast in places of poor accessibility even desirable goods of high (local) utility ai may not be consumed due to high prices pi. When production is subject to capacity limits zi  ci say (2.7a), then (2.15) is modified to 8 <

9 8 9 zi ¼ 0 = <> = 0  z i  c i , hi ¼ pi : ; : ;  zi ¼ ci

ð2:16aÞ

production is a step function of price, and (ph) is a capacity rent. The study of spatial markets in discrete locations for supply and demand has also been studied for supply and demand in continuously extended market areas, characterized by isotims (lines of equal price) and fields of flow (Beckmann 1952; Puu 1977; Beckmann and Puu 1985). The flow variable xij becomes the vector v, the commodity balance equation (2.1) the divergence equation div vðxÞ þ qðxÞ ¼ 0 the efficiency condition (2.4) became the gradient equation

ð2:1cÞ

42

M. Beckmann

k

v ¼ grad p: jvj

ð2:4cÞ

But this will not be pursued here.

2.5

Uniform Pricing

Monopolistic price strategies include, besides mill pricing (or f.o.b.) uniform delivered pricing and discriminatory pricing such as zonal tariffs or basing point (Pittsburgh plus) pricing. Under uniform pricing, buyers pay the same price inclusive of transportation costs, if they live within a specified area. Under competitive conditions, which we consider here, only mill or uniform pricing are viable. Uniform pricing is viable only if within the specified area (or radius) all customers must be served within the specified area (or radius). Otherwise price cutting will cause the areas of free delivery to shrink, ultimately to a point, which is the suppliers’ (common) location. The special character of uniform pricing is apparent also from the fact that market equilibrium can no longer be derived from maximization of welfare (or minimization of cost). Uniform pricing, although a form of price discrimination, has the advantage of simplicity and attractiveness and thus is the most common type in consumer product markets (Greenhut and Ohta 1975). Under mill pricing pj ¼ hi þ rij

ð2:4bÞ

the market radius R is limited by the consumers’ willingness to pay. Assume the same demand function f in all locations. Then at the market limit R 0 ¼ aj  pj ¼ a  h  RM :

ð2:17Þ

Under uniform pricing p the market radius R is determined by the suppliers’ ability to cover their production and transportation costs p  h þ r

ð2:18Þ

and perfect competition will cause the ‘‘¼’’ sign to hold p  h þ Ru :

ð2:18aÞ

Under mill pricing, competition drives the mill price to marginal (or constant) production costs, say at i¼0

2 The Pure Theory of Spatial Markets

43

po ¼ h:

ð2:19Þ

From (2.4b), (2.17), (2.18a), (2.19) a  h  RM ¼ 0 < a  p ¼ a  h  Ru

ð2:20Þ

Ru < RM

ð2:20aÞ

it follows that

since f is decreasing. The market radius under uniform pricing cannot be larger than under mill pricing. When marginal production cost¼average production cost¼h ¼constant, mill pricing in equilibrium does not allow firms to recover any fixed cost F, but uniform pricing under competitive pressure does. For profits in a market of radius R from prices p set to cover marginal production and transportation costs (2.18a) are Z

R

Z

R

ðR  r ÞmðrÞdr ¼

0

Z MðrÞdr > 0;

where

0

MðrÞ ¼

r

mðrÞdr;

ð2:21Þ

0

where m(r) is the density of demand in a ring of width dr at distance r. Thus for uniform population density m mðrÞ ¼ 2pmr

ð2:22Þ

one has MðrÞ ¼ pmr 2 ; pmR2 ¼ F;

ð2:23Þ

If the area which suppliers must serve can be expanded beyond the minimal radius R needed to recover fixed cost, rffiffiffiffiffiffiffi F R¼ pm

ð2:23aÞ

both p and each firm’s profits are raised until free entry reduces the latter by lowering the representative firm’s demand density m so that (2.23a) holds once more.

44

2.6

M. Beckmann

Heterogeneous Products

Let the varieties of a good be identified with the locations of their production i. The set of product varieties is then a subset of all locations i. We assume an additive logarithmic utility function uj ¼

X

aij log qij :

ð2:24Þ

i

We denote consumption of i in location j by qij and standardize X

aij ¼ 1:

ð2:24aÞ

pij qij ¼ yj ;

ð2:25Þ

i

Also, assume a budget constraint X i

where yj is the budget allocated to the class of goods i in location j, where pij are prices. A straightforward calculation for this well-known type of utility function yields the nice demand functions yj : pij

ð2:26Þ

pij ¼ pi þ rij :

ð2:27Þ

aij yj pi þ rij

ð2:26aÞ

qij ¼ aij As before, assume mill pricing so that

Demand functions of the type qij ¼

are close to a gravity model of spatial interaction, except for the addition of mill prices pi to the distant terms rij. Equation (2.26a) shows that sales are approximately proportional to inverse distance. The ratio of sales of two rival product i, k in location j is approximately qij ai rkj ffi  ; qkj ak rij

ð2:28Þ

2 The Pure Theory of Spatial Markets

45

where ‘‘attractions’’ aij and akj are independent of buyer locations j and prices are small compared to distances (transportation costs). Attractions are all the same aij akj, this implies that the closest seller has the dominant market share. If we redefine conventional market areas – the set of points closest to a given supplier – as areas of dominant market share, they will once more cover the entire region as mutually exclusive and exhaustive market areas. When attractions differ, these market areas are still exhaustive and exclusive, but distorted from the case where distance alone matters. The logarithmic utility function leads to a particularly nice and simple resolution of the budget constraint. When the heterogeneous good is inexpensive enough to leave out any budgetary restrictions, additive power functions as utility functions will also generate a gravity distance effect. Only utilities of the entropy type will generate a (negative) exponential distance effect in spatial markets.

2.7

Innovation

Spatial markets are a convenient vehicle to study the regional impact of innovations. In particular they offer a natural classification of the spatially relevant types of innovation.

2.7.1

Demand

An innovation in demand can take the following forms: A known product is introduced in new locations or market areas. Inventive sales managers in search of new outlets will eventually sell refrigerators even to the Eskimos. Secondly, new uses may be found for a product in some locations (perhaps in conjunction with price reductions). Most importantly, new products may be introduced, which of course will have to be advertised. It is a well established practice to test the acceptance of a new product in some test location(s) that are considered to be normal enough to serve as good market predictors.

2.7.2

Supply

New (and better) locations – with better access to labor or other inputs, better climate, etc. – may be discovered for the production of a known good, or better methods may be found in the existing locations. The locations of a firm may also be restructured to generate new varieties of a product. The invention of a new product will always require locational choice. New products with an initially small national demand are best started from the metropolis.

46

M. Beckmann

An innovation improving supply over extended areas was the so-called green revolution, the introduction of higher yielding and better resistant crops. Another important innovation is the discovery of new resource deposits (oil, coal, ores) or sources of labor supply in overlooked settlements.

2.7.3

Distribution

While geographical distances remain unchanged, their impact through transportation and communication cost has been strongly reduced through innovations. Transportation costs have shown a secular trend to fall. This has been due to both technological innovations and to changes in organization (regulation). In our times the greatest changes have occurred in communication, particularly through the internet. As a result market areas no longer need to be contiguous or close to the point of supply. The limitation on demand that is imposed by information deficits about a product has been lifted and this has vastly expanded the potential markets of many products.

References Beckmann MJ (1952) A continuous model of transportation. Econometrica 20:643–660 Beckmann MJ (1976) Spatial price policies revisited. Bell J Econ 7(2):619–630 Beckmann MJ, Puu T (1985) Spatial economics: density, potential and flow. North Holland, Amsterdam Dantzig G (1959) Linear programming. Princeton University Press, Princeton Greenhut ML, Ohta H (1975) Theory of spatial pricing and market areas. Duke University Press, Durham, NC Koopmans TC (1949) Optimum utilization of the transportation systems. Econometrica 17: 136–146 Launhardt W (1886) Mathematische Begru¨ndung derVolkswirtschaftslehre. Wilhelm Engelmann, Leipzig Puu T (1977) A proposed definition of traffic flow in continuous transportation models. Environ Plan 9:559–567 Samuelson PA (1952) Spatial price equilibrium and linear programming. Am Econ Rev 42: 283–303 Von Thu¨nen JH (1826) Der Isolirte Staat. Perthes, Hamburg

Chapter 3

Smith–Ricardo Specialization in the Presence of Tiring Effects Tonu Puu

3.1

Introduction

One of the best selling ideas economics ever had was the Smith–Ricardo paradigm of specialization, division of labor, and comparative advantage. Adam Smith (1776) wrote almost lyrically about the advantages of specialized pin-making. Through dividing the process in many tiny operations, combined with education to very specialized moments, such as polishing the nails, or wrapping them in paper, based on natural talent, Smith reported huge increases in productivity. David Ricardo (1817) then launched his theory of comparative advantage, among nations as well as among individuals, and pointed at the advantage for total productivity through total specialization when each one carried out only one particular activity. As a result, everybody should, for individual and common benefit, specialize in one special activity and be a consumer of all the other activities. The theory blows up traditional ideals of educated humanity, dissipated by for instance Baldassare Castiglione (1507) in ‘‘The Courtier’’, on which most of Western education was based for centuries. If you are a brain surgeon, you should operate brains and not paint or make music yourself, because it is better to go to a concert or a gallery in your free time to enjoy the production of other specialists, who, of course, perform their specialized tasks better. Traditional craftwork, with its alternation between very diverse and sometimes enjoyable operations, is superseded by work at the endless conveyor belt. Some periods of multiple occupation seem unenlightened in the light of the specialization paradigm, and it is not understandable how some periods in history, such as Florence of 1500 or Vienna of 1900, could be so productive, despite notorious violation of the paradigm of comparative advantage. Surely, Alberti would have been more productive if he had concentrated on building churches, or, even better, writing on moral philosophy for which he was educated, rather than

T. Puu CERUM, Umea˚ University, SE-90187 Umea˚, Sweden e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_3, # Springer‐Verlag Berlin Heidelberg 2009

47

48

T. Puu

wasting time on writing treatises on the principles of painting? And, Leonardo should above all have decided to be a draftsman or a scientist? Who knows how ingenious a philosopher Wittgenstein would have been if he had not wasted time on sculpture, and what compositions could Schonberg have achieved if he had not spent time on painting? Yet, it is surprising how productive these periods of non-specialization were. In this brief paper we want to give a partial answer to how such things are possible, by taking into account just one tiny little factor. Even if initially productivity is different among individuals, there is no doubt that repeating the same operation over and over after a while becomes boring and as a consequence productivity differences diminish, in which case it may turn out to be better not to specialize completely. The Smith–Ricardo paradigm is linear by the bulk of mainstream economic theory and it is sufficient to introduce just any tiny tiring effect to blow it up. This is the purpose of this paper. It is not to deny that there is much in the theory of comparative advantage, neither is the intention to claim that tiring is the most significant factor working against it. The paper intends to show how easily any kind of nonlinearity fundamentally alters the conclusions of any linear model.

3.2

The Individual

First, consider one individual worker who divides total working time into n different activities in shares denoted by ti . Total working time for the period we consider is taken as given, and, for convenience, normalized to unity. Hence we also take total leisure time as fixed, but this does not constrain anything, as the substitution of leisure for consumption is not an issue at present. Hence i¼n X

ti ¼ 1:

ð3:1Þ

i¼1

Next denote efficient work done in the ith activity xi and define xi ¼ ai  bi ti ; ti

i ¼ 1; . . . ; n:

ð3:2Þ

The quotient is average productivity, which starts at ai , the initial productivity, but is subject to a linear tiring effect bi ti , the longer the time spent on this particular kind of work. When ti ¼ ai =bi , productivity goes down to zero. If ai =bi 1, this never quite happens, even if all time available is spent on only this particular activity. In the classical (linear) Smith–Ricardo world bi always equals zero. Hence this never happens. The obvious thing to do with (3.2) is to multiply through by ti , to obtain xi ¼ ai ti  bi t2i ;

i ¼ 1; . . . ; n:

ð3:3Þ

3 Smith–Ricardo Specialization in the Presence of Tiring Effects

49

Even if productivity goes down to zero first at ti ¼ ai =bi , nobody ever works more than ti ¼ ð1=2Þai =bi for which production is maximal. This is so because (3.3) is quadratic, so a given labor achievement can be obtained with both an effort smaller or larger than this maximizing effort. If we disregard the pleasure of work, nobody chooses the same achievement with a larger effort. Accordingly, if we solve (3.3) for ti , we get the unique smaller root vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi )ffi u (  2 1 ai 1 u a x i i : ti ¼  t 4 2 bi 2 bi bi Now (3.1) and (3.3) provide n þ 1 equations in 2n variables so that we could in principle eliminate all the time shares ti and any one of the work variables xi . As we do not want to give any particular activity preference, we prefer to state the result as an implicit function (a kind of efficiency frontier): Fðx1 ; x2 ; x3 . . . . . . . . . . . .Þ:

ð3:4Þ

From (3.3) and (3.1) it is obvious that when bi = 0, (3.4) is flat, i.e., a plane. In general, we easily realize that the exact form obtained from the solutions of the quadratic equation (3.3) and summing up to unity in accordance with (3.1), renders a much too complex expression to be of any use as given below: vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi )ffi u( i¼n i¼n u  2 X X 1 ai 1 ai xi t ¼ 1:  4 2 i¼1 bi 2 i¼1 bi bi We therefore study the properties of (3.4) for a much simplified case, n ¼ 2 and ai ¼ a; bi ¼ b to be precise. Defining b

B ¼ ð2a  bÞ;

ð3:5Þ

Aðx1  x2 Þ2 þ 2ðx1 þ x2 Þ ¼ B;

ð3:6Þ

A¼

ða  bÞ2

;

we can write,

which is a parabola in terms of coordinates (x1  x2 ) and (x1 þ x2 ), which amounts to a rotation by 45 . In Fig. 3.1 we display a family of curves (3.6), where we fix the initial efficiency to a ¼ 1, and let b the ‘‘tiring coefficient’’, range in the interval [0, 1]. As we see from (3.5), with b ¼ 0; A ¼ 0 (3.6) loses its quadratic term, and the efficiency frontier becomes a straight line. Likewise, with b ¼ 1, A becomes infinite and (3.6) collapses to a segment of the positive diagonal.

50

T. Puu

x2

x1

Fig. 3.1 Efficiency frontiers with different tiring coefficients

There is an additional feature in terms of shading. Through implicit differentiation from (3.6) we obtain dx2 1 þ Aðx1  x2 Þ : ¼ 1  Aðx1  x2 Þ dx1

ð3:7Þ

The rising sections of the efficiency frontiers in Fig. 3.1 do not make any sense, so we have to require dx2 =dx1 0, which, using (3.7) states A2 ðx1  x2 Þ2 < 1:

ð3:8Þ

The dark shaded area in Fig. 3.1 indicates where (3.8) does not hold. Recall that we are dealing with one individual worker whose objective is to choose a mix of two different work activities. Once we also know the wage rates for the two activities, we can find an optimal mix of activities for this individual. Note that, if the tiring coefficients are zero, i.e., bi ¼ b ¼ 0, so that we deal with the straight line, the worker will always choose one of the endpoints of this line, i.e., specialize completely. In the presence of tiring we are dealing with one of the parabolas, and for any wage ratio we can then find a point of tangency with the given wage ratio line. Further note that we assumed no particular advantage in any of the two activities, by assuming not only bi ¼ b, but also ai ¼ a. This is of no importance as long as we deal with the single worker, but, once we are considering division of labor in a

3 Smith–Ricardo Specialization in the Presence of Tiring Effects

51

group of workers, it becomes important to introduce asymmetries. Only then can we evaluate the Smith–Ricardo argument.

3.3 3.3.1

Two Individuals Work Sharing Vs. Specialization

To this end consider two workers combined with two types of labor, a setup international trade specialists would approve of. Let us keep x1 ; x2 as work contributions for the first individual, and use y1 ; y2 for the second. We then have a companion to (3.6): Aðy1  y2 Þ2 þ 2ðy1 þ y2 Þ ¼ B

ð3:9Þ

still keeping symbols A; B as determined from (3.5). Later we assume asymmetry between the individuals. As a preliminary, let us compare two extreme cases: complete specialization vs. no specialization at all, i.e., 50/50 sharing of time for each individual. With complete specialization we will for instance have x2 ¼ y1 ¼ 0, and hence from (3.6), (3.5), and (3.9): x1 ¼ y2 ¼ a  b:

ð3:10Þ

As in the other extreme case, suppose there is no specialization at all. Then, we put x1 ¼ x2 in (3.6) and y1 ¼ y2 in (3.9). Obviously: 1 x1 ¼ x2 ¼ y1 ¼ y2 ¼ a  b: 2

ð3:11Þ

In the non-specialization case we have x1 þ y1 ¼ x2 þ y2 ¼ 2a  b as we see from (3.11), whereas in the specialization case, from (3.10), x1 þ y1 ¼ x2 þ y2 ¼ a  b. Hence, in the presence of tiring, b > 0, two identical individuals produce more of both commodities by sharing their time between activities, rather than specializing. In the sequel we will see to what extent the conclusion holds in the presence of asymmetry.

3.3.2

Maximizing Labor Income

Let us so consider working time optimization for each of the individuals given wage rates for the activities. For the first individual the value of total labor time spent on n activities is, using (3.3): i¼n X i¼1

wi xi ¼

i¼n X i¼1

wi ðai ti  bi t2i Þ:

ð3:12Þ

52

T. Puu

In our setup, wages are, of course, associated with efficient work done, not with time spent. Maximizing (3.12) subject to (3.1) yields wi ðai  2bi ti Þ ¼ l;

ð3:13Þ

where l denotes a Lagrangian multiplier associated with the constraint (3.1). We can easily solve for the time shares:   1 l ai ti ¼  2 bi wi bi

i ¼ 1; . . . ; n:

ð3:14Þ

Finally substituting from (3.14) in (3.1), we can also solve for the Lagrangian: P ai 2  i¼n i¼1 b l ¼ Pi¼n 1 i :

ð3:15Þ

i¼1 bi wi

Substituting back from (3.15) into (3.14), we obtain the time shares of each individual spent on various jobs as expressed alone in the wage rates wi and the coefficients that describe the working faculties in the different occupations ai , bi .

3.4

Two Identical Individuals

We already stated the efficiency frontiers for two identical individuals in (3.6) and (3.9) respectively. Let us now maximize the value of their combined work: w1 ðx1 þ y1 Þ þ w2 ðx2 þ y2 Þ:

ð3:16Þ

By associating Lagrangian multipliers m with (3.6) and v with (3.9) we find w1 ¼ 2mð1 þ Aðx1  x2 ÞÞ ¼ 2vð1 þ Aðy1  y2 ÞÞ

ð3:17Þ

w2 ¼ 2mð1  Aðx1  x2 ÞÞ ¼ 2vð1  Aðy1  y2 ÞÞ:

ð3:18Þ

and

Together with (3.6) and (3.9), (3.17)–(3.18) boil down to m¼v¼

w1 þ w2 4

ð3:19Þ

and x1 ¼ y1 ;

x2 ¼ y2 :

ð3:20Þ

3 Smith–Ricardo Specialization in the Presence of Tiring Effects

53

x2 + y2

x1 + y1

Fig. 3.2 Efficiency frontier for two identical individuals

As the activity mixes for the two individuals are identical, we conclude that the joint efficient labor mix is twice that for each individual. The facts can easily be presented in terms of a diagram of the type well known from international trade theory. In Fig. 3.2 the efficiency frontier for one individual is turned the right way, whereas that for the other is rotated by 180 along with its coordinate system. By letting the rotated system slide down at tangency with the fixed one, we take care of all different wage ratios, and in this sliding motion the origin, marked by black dots, of the movable system describes the total combination of efficient work for the two individuals. The motion of the origin of the second coordinate system then traces the curve which is an exactly scaled up copy (by the factor 2) of the individual efficiency frontier.

3.5

Two Asymmetric Individuals

More interesting is the case where the individuals indeed have a different comparative advantage. We are not going to focus on differences in tiring coefficients in their relation to initial efficiency, so we put the efficient labor done as follows: x1 ¼ kðat1  bt21 Þ;

1 x2 ¼ ðat2  bt22 Þ; k

ð3:21Þ

54

T. Puu

y1 ¼

 1 at1  bt21 ; k

  y2 ¼ k at2  bt22 :

ð3:22Þ

In this way we can keep the same symbols a; b as before and hence A; B as defined in (3.5) above. It is a bit inadequate to use the symbols ti as time shares for both individuals, but as we only want to see the structure of (3.21)–(3.22), we will not make any use of the ti themselves. This structure is that in terms of the variables x1 =k, kx2 or ky1 , y2 =k. Equations (3.21)–(3.22) look exactly like (3.3). Provided k1, the first individual is k2 times more efficient in the first activity (at each level of effort), whereas the second individual is k2 times more efficient in the second activity. Given our above digression, we can hence rephrase (3.6) and (3.9) as x 2 x  1 1  kx2 þ 2 þ kx2 ¼ B k k

ð3:23Þ

  y2  2 y2  A ky1  þ 2 ky1 þ ¼ B: k k

ð3:24Þ

A and

Figure 3.3 shows a case of high asymmetry, k2 ¼ 16. In this figure, we put both efficiency frontiers in the same coordinate system. The picture also displays a wage

x2,y2

x1,y1

Fig. 3.3 The case of two asymmetric individuals

3 Smith–Ricardo Specialization in the Presence of Tiring Effects

55

ratio line (assuming equal wage rates for the two activities), and we see very different choices made by the individuals. However, despite the 16-fold comparative advantage, there is not any complete specialization. Before concentrating on formal detail, we display a companion to Fig. 3.2 for the asymmetric case as seen in Fig. 3.4. Again, we keep one of the efficiency frontiers with its coordinate system fixed, turn the other 180 around, and let it slide down at tangency. We note the total efficiency curve for the two individuals. It is noteworthy that, despite the asymmetry, the total efficiency curve is almost as round as that of Fig. 3.2. We will consider very extreme asymmetries at the end of this paper. So, let us return to formal detail. Again we associate Lagrangian multipliers m; v with the constraints (3.23)–(3.24), assume arbitrary wage rates w1 , w2 and maximize the value of total work effort for the two individuals according to (3.16). The result, corresponding to (3.17)–(3.18) above is x  2m 1  kx2 w1 ¼ 1þA k k

!

 y2  ¼ 2vk 1 þ A ky1  k

! ð3:25Þ

and x  1  kx2 w2 ¼ 2mk 1  A k

!

!  2v y2  1  A ky1  ¼ : k k

x2+y2

x1+y1

Fig. 3.4 Aggregate efficiency frontier for two asymmetric individuals

ð3:26Þ

56

T. Puu

We can eliminatese, n between (3.25) and (3.26), obtaining: x

 1 w1  12  kx2 ¼ ww21 k1 k A w2 þ k 2 1

ð3:27Þ

and 

ky1 

w 2 y2  1 w12  k : ¼ w1 k A w2 þ k 2

ð3:28Þ

In this way we obtain the composite rescaled new variables (x1 =k  kx2 ) and (ky1  y2 =k), as dependent on known coefficients alone, the wage ratio, the asymmetry coefficient, and the rest of the technical coefficients as determined by (3.5). In combination with (3.23)–(3.24), (3.27)–(3.28) enable us to derive the efficiency frontiers for different asymmetry coefficients. The derivation is messy without being instructive, so we let the computer do the job. The result is shown in Fig. 3.5. We see that the higher the asymmetry factor, the more Leontief like the possibility frontier becomes, which means that an optimum close to the ‘‘knee’’ is chosen for almost all wage ratios. However, note that the outmost frontier is obtained for the enormous asymmetry factor of k2 ¼ 625.

x2+y2

x1+y1

Fig. 3.5 Efficiency frontiers for different asymmetry coefficients

3 Smith–Ricardo Specialization in the Presence of Tiring Effects

57

Another fact shown is that the higher the asymmetry, the further out the efficiency frontier is located. This can illustrate Smith’s argument that complementary specialization among manpower enhances overall productivity. Yet, with the tiring factor accounted for here, there is never complete specialization.

References Castiglione B (1507) Il Cortegiano, English translation: The book of the courtier. Penguin, London, 1976 Ricardo D (1817) Principles of political economy and taxation. Everymans Library Reprint, 1912 Smith A (1776) An inquiry into the nature and causes of the wealth of nations. Everymans Library, 1910

Chapter 4

Dynamics of Innovation Fields with Endogenous Heterogeneity of People Masahisa Fujita

4.1

4.1.1

Introduction: Towards the New Economic Geography in the Brain Power Society Welcome to the Brain Power Society

According to Lester Thurow at MIT, advanced countries are shifting from capitalism based on mass production of commodities to the brain power society in which creation of knowledge and information using brain power plays the central role (Thurow 1996). The concept of brain power society is essentially the same as that of ˚ ke Andersson who maintains that advanced countries the C-society advocated by A are leaving the industrial society (with its reliance on simplicity of production and products and the heavy use of natural resources and energy) and entering the C-society with and increasing reliance on creativity, communication capacity, and complexity of products (Andersson 1985). In this paper, the term ‘‘brain ˚ ke Andersson. power society’’ is synonymous with the ‘‘C-society’’ of A The ultimate concern of this paper is the further development of the New Economic Geography (NEG) towards a more comprehensive theory of geographical economics in the age of brain power society, in which the dynamics of the spatial economy arise from the dual linkages in the economic and knowledge fields. Before elaborating this ultimate objective, let me explain briefly what is the socalled the New Economic Geography.

M. Fujita Konan University e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_4, # Springer‐Verlag Berlin Heidelberg 2009

59

60

4.1.2

M. Fujita

The New Economic Geography and Its Future: Incorporating Dual Linkages in Economic and Knowledge Fields

Since about 1990 there has been a renaissance of theoretical and empirical work on economic geography. Among others, the pioneering work of Paul Krugman (1991) on the core–periphery model has triggered a new flow of interesting contributions to economic geography. The work represented by this new school of economics is called the New Economic Geography (NEG).1 The hallmark of the NEG is the presentation of a unified approach to modeling a spatial economy characterized by a large variety of economic agglomeration – one that emphasizes the three-way interaction among increasing returns, transport costs (broadly defined), and the movement of productive factors – in which a general equilibrium model is combined with nonlinear dynamics and an evolutionary approach for equilibrium selection. Figure 4.1 represents the basic conceptual framework of the NEG. The observed spatial configuration of economic activities is considered to be the outcome of a process involving two opposing types of forces, that is, agglomeration (or centripetal) forces and dispersion (or centrifugal) forces.2 As a complicated balance of these two opposing forces, a variety of local agglomeration of economic

Agglomeration forces

Dispersion forces

balance

emergence of local agglomerations and self-organization of the spatial structure slow changes in environments evolution through a sequence of structural changes

Fig. 4.1 The basic framework of the New Economic Geography

1 See Fujita et al. (1999) for a comprehensive manifestation of this approach. See also Fujita and Thisse (2002) and Baldwin et al. (2003) for the recent development of the NEG. For an overview of the NEG, see Fujita and Krugman (2004), Fujita (2005), Fujita and Mori (2005). 2 This hypothesis is not entirely new, of course. For, e.g., Zipf (1949) conjectured that the changing spatial configuration of economic activities was the outcome of the two sets of centripetal (unifying) and centrifugal (diversifying) forces.

4 Dynamics of Innovation Fields with Endogenous Heterogeneity of People

61

activity emerges, and the spatial structure of the entire economy is self-organized. And, with the gradual changes in technological and socioeconomic environments, the spatial system of the economy experiences a sequence of structural changes, evolving towards an increasingly complex system. In this framework, then, the two questions of obvious importance are: Question 1: how to explain the agglomeration forces? Question 2: how to explain the dispersion forces? The answer to Question 2 is rather easy, for the concentration of economic activities at a location will naturally increase factor prices (such as land price and wage rate) and induce congestion effects (such as traffic congestion and air pollution as well as more severe competition among similar firms), which can be readily explained by the traditional economic theory. Thus, the principal concern of the NEG is Question 1, i.e., how to explain the agglomeration forces behind the formation of a large variety of spatial agglomeration such as cities and industrial districts. In most models of the NEG so far, agglomeration forces arise solely from pecuniary externalities through linkage effects among consumers and industries, neglecting all other possible sources of agglomeration economies such as knowledge externalities and information spillovers. This has led to the opinion that the theories of the NEG have been too narrowly focused, ignoring as much of the reality as old trade theory. I fully understand the concern. But, such a narrow focus of the NEG was designed in order to establish a firm micro-foundation of geographical economics based on modern tools of economic theory. It does not necessarily mean that the NEG is limited to such a narrow range of models and issues. On the contrary, its framework is widely open to further development. Indeed, recently many of such possibilities are being explored vigorously by many young scholars.3 That much said, however, I admit that there still remains a big room for further development of the NEG. In particular, there remains one type of agglomeration forces of which micro-foundations have seen little development so far, i.e., the linkages among people through the creation and transfer of knowledge, or in short, the K-linkages. (Hereafter, ‘‘knowledge’’ is defined broadly to include ideas and information.) Traditionally, K-linkage effects have either been called ‘‘knowledge spillovers’’ or ‘‘knowledge externalities’’. However, the term, ‘‘spillovers’’, tends to have a connotation of passive effects. And, the term, ‘‘externalities’’, tends to imply too many different things at once. So, in the remaining discussion, instead of knowledge spillovers or externalities, let me use the term, K-linkages, in order to emphasize that they represent the agglomeration forces resulting from the activities related to both the ‘‘creation of knowledge’’ and the ‘‘transfer of knowledge’’ or ‘‘learning’’ (either in an active way or a passive way). In contrast to the K-linkages, the traditional linkages through the production and transactions of (traditional) goods and services may be

3

See those articles reviewed in Fujita and Mori (2005).

62

M. Fujita

called the E-linkages (where ‘‘E’’ represents the economic activities in the traditional economics). Using such a terminology, we may imagine that the agglomeration forces in the real world arise from the dual effects of E-linkages and K-linkages. In this context, we conjecture that the role of K-linkages has been becoming increasingly more dominant recently. Yet, developing the micro-foundations of K-linkages seems to be the most challenging task, largely left for young scholars in the future. This paper represents my modest efforts with my colleagues towards this objective. Needless to say, there has been a great amount of conceptual studies on knowledge externalities/spillovers in a spatial context, starting with Marshall (1890), and including more recent pioneering work such as Jacobs (1969), Andersson (1985) and Lucas (1988) in an urban context, and Porter (1998) in the context of industrial clusters. Yet, it would be fair to say that there is a lot of room left for advancing the micro-foundations of K-linkages in space. Particularly, in developing the microfoundations of K-linkages, ‘‘creation of knowledge’’ must be clearly distinguished from ‘‘transfer of knowledge’’ or ‘‘learning’’. Furthermore, for the creation of new ideas, cooperation among heterogeneous people is essentially important. Yet, through communication and migration, the degree of heterogeneity of people in a region changes over time. Thus, the nature of K-linkages is essentially dynamic, and hence their full-fledged treatment requires a dynamic framework as elaborated in the next section.

4.1.3

Dynamics of Innovation Fields Through the Endogenous Heterogeneity of Brains

Figure 4.2 represents abstractly the cooperative process of knowledge creation by two persons, i and j, when they meet and collaborate to create new ideas (or new knowledge) together. The left circle, Ki , represents the state of knowledge, or just knowledge, of person i (at the time of meeting), whereas the right circle, Kj represents the knowledge of person j. The overlapping area, Cij , represents their knowledge in common, or just common knowledge,4 whereas the left area, Dij ¼ Ki  Cij , shows the differential knowledge of person i from j, the right area Dji ¼ Kj  Cij the differential knowledge of person j from i. Through mutual communication and discussion based on the common knowledge Cij , the two persons endeavor to develop new ideas by combining their differential knowledge Dij and Dji . This joint process of knowledge creation can

4

Here, ‘‘common knowledge’’ represents simply the short expression of ‘‘the knowledge in common’’ or ‘‘mutual knowledge’’. It is not the term used in game theory.

4 Dynamics of Innovation Fields with Endogenous Heterogeneity of People

Ki Dij

Kj

Cij

Differential

Common

Knowledge

Knowledge of person

63

i

Dji Differential Knowledge of person

j

Fig. 4.2 Cooperative process of knowledge creation

be expected to be most productive when the proportions of the three components, i.e., the common knowledge (Cij ), the differential knowledge of person i (Dij ), and the differential knowledge of person j (Dji ), are well balanced. A sufficient amount of common knowledge is necessary for effective communication between two persons. Furthermore, if one person does not have a sufficient amount of differential knowledge, there is little motivation for the other person to meet and collaborate. In other words, too much common knowledge means little heterogeneity or originality in the collaboration, unable to yield enough synergy. Therefore, in general, for a cooperative process of knowledge creation by a group of people to be productive, both a sufficient heterogeneity and a sufficient common base in their states of knowledge are essential. When such a delicate balance in their states of knowledge holds, an unexpected synergy may be created from their close collaboration. Actually, this observation is not entirely new. We have, e.g., an old Chinese saying, ‘‘San ge chou pi jiang, Di ge Zhuge Liang’’ which roughly means ‘‘With three ordinary persons getting together, splendid ideas will come out’’. However, any nice saying must be taken with caution, for it may imply an antinomy. Concerning the previous Chinese saying, we may continue: ‘‘But, after three ordinary persons meeting for three months, no more splendid idea will come out’’. Likewise, returning to Fig.4.2 even when the two persons have initially a sufficient heterogeneity in their states of knowledge, if they continue a close cooperation in knowledge creation, their heterogeneity may keep shrinking. This is because the very cooperative process of knowledge creation results in the expansion of their common knowledge through both the sharing of newly created ideas and the transfer of differential knowledge to each other. Thus, unless some additional complementary mechanisms are not working, the cooperative process of knowledge creation among the same group of people tends to become less productive eventually.

64

4.2

M. Fujita

The Model

Building upon what has been discussed above, in this section, I present a micromodel of knowledge creation through the interaction of a group of people, which has been developed by Berliant and Fujita (2007).5 In describing the model, the analogy between partner dancing and working jointly to create and exchange knowledge is useful, so we will use terms from these activities interchangeably. We assume that it is not possible for more than two persons to meet or dance at one time, though more than one couple can dance simultaneously. When agents meet, they create new, shared knowledge, thus building up knowledge in common. When agents are not meeting with each other, their knowledge base grows more different. The fastest rate of knowledge creation occurs when common and differential knowledge is in balance.6 Specifically, suppose that there exist N persons in the economy. Consider a given time t, and focus on two persons i and j. And, let in terms of Fig.4.1, ndij ðtÞ be the size of Dij , the differential knowledge of person i from j; ncij ðtÞ be the size of Cij , the common knowledge for person i and j; ndji ðtÞ be the size of Dji , the differential knowledge of person j from i. And let ni ðtÞ ¼ ncij ðtÞ þ ndij ðtÞ;

ð4:1Þ

nj ðtÞ ¼ ncij ðtÞ þ ndji ðtÞ;

ð4:2Þ

so that ni ðtÞ represents the size of Ki , the knowledge of person i at time t; nj ðtÞ the size of Kj , the knowledge of person j at time t. Knowledge is a set of ideas that are possessed by a person at a particular time. However, knowledge is not a static concept. New knowledge can be produced either individually or jointly, and ideas can be shared with others. But all of this activity takes time. Now we describe the components of the rest of the model. To keep the description as simple as possible, we focus on just two agents, i and j. At each time, each faces a decision about whether or not to meet with others. If two agents want to meet at a particular time, a meeting will occur. If an agent decides not to meet with anyone at a given time, then the agent produces separately and also creates new knowledge separately, away from everyone else. If two persons do decide to meet at a given time, then they collaborate to create new knowledge

5

See Berliant and Fujita (2007) for the further elaboration of the following model. For simplicity, we employ a deterministic framework. It seems possible to add stochastic elements to the model, but at the cost of complexity. It should also be possible to employ the law of large numbers to a more basic stochastic framework to obtain equivalent results.

6

4 Dynamics of Innovation Fields with Endogenous Heterogeneity of People

65

together. Here we limit the scope of our analysis to knowledge creation as opposed to knowledge transfer.7 What do the agents know when they face the decision about whether or not to meet a potential partner j at time t? Each person knows both Ki ðtÞ and Kj ðtÞ. In other words, each person is aware of his own knowledge and is also aware of others’ knowledge. Thus, they also know ni ðtÞ, nj ðtÞ, ncij ðtÞ ¼ ncji ðtÞ, ndij ðtÞ, and ndji ðtÞ (for all j 6¼ i) when they decide whether or not to meet at time t. The notation for whether or not a meeting of persons i and j actually occurs at time t is: dij ðtÞ ¼ dji ðtÞ ¼ 1 if a meeting occurs and dij ðtÞ ¼ dji ðtÞ ¼ 0 if no meeting occurs at time t. For convenience, we define dii ðtÞ ¼ 1 when person i works in isolation at time t, and dii ðtÞ ¼ 0 when person i meets with another person at time t. Next, we must specify the dynamics of the knowledge system and the objectives of the people in the model in order to determine whether or not two persons decide to meet at a particular time. The simplest piece of the model to specify is what happens if there is no meeting between person i and anyone else, so i works in isolation. Let aii ðtÞ be the rate of creation of new ideas created by person i in isolation at time t (this means that i meets with itself). Then we assume that aii ðtÞ ¼ ani ðtÞ when dii ðtÞ ¼ 1;

ð4:3Þ

where a is a positive constant. So we assume that if there is no meeting at time t, individual knowledge grows at a rate proportional to the knowledge already acquired by an individual. If a meeting occurs between i and j at time t (dij ðtÞ ¼ 1), then joint knowledge creation occurs, and it is governed by the following dynamics:8 h i13 aij ðtÞ ¼ b ncij ðtÞndij ðtÞndji ðtÞ

when

dij ðtÞ ¼ 1 for j 6¼ i;

ð4:4Þ

where b is a positive constant. So when two people meet, joint knowledge creation occurs at a rate proportional to the normalized product of their knowledge in common, the differential knowledge of i from j, and the differential knowledge of j from i. The rate of creation of new knowledge is highest when the proportion of ideas in common, ideas exclusive to person i, and ideas exclusive to person j are split evenly. Ideas in common are necessary for communication, while ideas exclusive to one person or the other imply more heterogeneity or originality in

7

In an earlier version of this paper, Berliant and Fujita (2004, available at http://econpapers.hhs. se/paper/wpawuwpga/0401004.htm), we have worked out the details of the model with both knowledge creation and transfer when there are only two persons, and found no essential difference in the results. However, in the N person case, it is necessary to keep track of more details of who knows which ideas, and thus the model becomes very complex. This extension is left to future work. 8 See Berliant and Fujita (2007, Sect.4.6) for a more general form of joint knowledge creation.

66

M. Fujita

the collaboration. If one person in the collaboration does not have exclusive ideas, there is no reason for the other person to meet and collaborate. Whether a meeting occurs or not, there is production in each period for both persons. Felicity in that time period is defined to be the quantity of output.9 Define yi ðtÞ to be production output (or felicity) for person i at time t. Normalizing the coefficient of production to be 1, we take yi ðtÞ ¼ ni ðtÞ:

ð4:5Þ

So, y_i ðtÞ ¼ n_ i ðtÞ: By definition, y_i ðtÞ n_ i ðtÞ ¼ yi ðtÞ ni ðtÞ

ð4:6Þ

which represents the rate of growth of income. We now describe the dynamics of the system, dropping the time argument. Let us focus on agent i, as the expressions for the other agents are analogous. y_i ¼ n_ i ¼

N X

dij aij ;

ð4:7Þ

j¼1

n_ cij ¼ dij aij n_ dij ¼

X

for

dik aik

all j 6¼ i;

for

all j 6¼ i:

ð4:8Þ ð4:9Þ

k6¼j

Equation (4.7) means that the increase in the knowledge of person i is the sum of the knowledge created in isolation and the knowledge created jointly with someone else. Equation (4.8) means that the increase in the knowledge in common for persons i and j equals the new knowledge created jointly by them. This is based on our previous assumption that there is no transfer of existing knowledge between agents even when they are meeting together. Finally, (4.9) means that all the knowledge created by person i either in isolation or jointly with persons other than person j becomes a part of the differential knowledge of person i from person j. By definition, it is also the case that N X

dij ¼ 1:

j¼1

9

Given that the focus of this paper is on knowledge creation rather than production, we use the simplest possible form for the production function.

4 Dynamics of Innovation Fields with Endogenous Heterogeneity of People

67

Furthermore, on the equilibrium path it is necessary that dij ¼ dji

for all

i and j:

Concerning the rule used by an agent to choose their best partner, to keep the model tractable in this first analysis, we assume a myopic rule. At each moment of time t, person i would like a meeting with person j when the increase in their rate of output while meeting with j is highest among all potential partners, including himself. Note that we use the increase in the rate of output y_i ðtÞ rather than the rate of output yi ðtÞ since in a continuous time model, the rate of output at time t is unaffected by the decision made at time t about whether to meet. As we are attempting to model close interactions within groups, we assume that at each time, the myopic persons interacting choose a core configuration. In order to analyze our dynamic system, we first divide all of our equations by the total number of ideas possessed by i and j: nij ¼ ndij þ ndji þ ncij

ð4:10Þ

and define new variables mcij  mcji ¼ mdij ¼

ndij n

; ij

ncij nij

¼

mdji ¼

ncji nij ndji nij

; :

By definition, mdij represents the percentage of ideas exclusive to person i among all the ideas known by person i or person j. Similarly, mcij represents the ideas known in common by persons i and j among all the ideas known by the pair. From (4.10), we obtain 1 ¼ mdij þ mdji þ mcij :

ð4:11Þ

Then, using (4.7)–(4.9) and ( 4.11), we can rewrite the income growth rate, (4.6), as follows:10 y_i n_ i ¼ ¼ dii a þ y i ni

10

X j6¼i

dij

h i13 b ð1  mdij  mdji Þmdij mdji 1  mdji

;

For details of the analyses in the rest of this paper, see Berliant and Fujita (2007).

ð4:12Þ

68

M. Fujita

where h   i h  i13 m_ dij ¼ að1  mdij Þ dii 1  mdji  djj mdij  dij mdij b 1  mdij  mdji mdij mdji    X b1  md  md md md 13 ik ki ik ki d d þ 1  mij 1  mji dik d 1  m ki k6¼i;j h  i13   X b 1  mdjk  mdkj mdjk mdkj  1  mdij mdij djk for i; j ¼ 1; 2; . . . ; N : 1  mdkj k6¼i;j ð4:13Þ At time t, since yi ðtÞ is a state-variable, maximizing y_i ðtÞ is equivalent to maximizing the growth rate, y_ i ðtÞ=yi ðtÞ. Hence, at each moment of time, the equilibrium values of dij (i; j ¼ 1; 2; . . . ; N) are to be determined as the core of the game in which each agent wishes to maximize the growth rate of income given by (4.12). Thus, the dynamics of the system are described in terms of mdij (i; j ¼ 1; 2; . . . ; N) only.

4.3 4.3.1

Equilibrium Dynamics The General Framework

The model with only two people is very limited. Either two people are meeting or they are each working in isolation. With more people, the dancers can be partitioned into many pairs of dance partners. Within each pair, the two dancers are working together, but pairs of partners are working simultaneously. This creates more possibilities in our model, as the knowledge created within a dance pair is not known to other pairs. Thus, knowledge differentiation can evolve between different pairs of dance partners. Furthermore, the option of switching partners is now available. We limit ourselves to the case where N is divisible by 4. This is a square dance on the vertices of the Hilbert cube. When the population is not divisible by 4, our most useful tool, symmetry, cannot be used to examine dynamics. Although this may seem restrictive, when N is large, asymmetries apply only to a small fraction of the population, and thus become negligible. In the general case, we impose the assumption of pairwise symmetric initial heterogeneity conditions for all agents. The initial state of knowledge is symmetric among the dancers, and given by ncij ð0Þ ¼ nc ð0Þ

for all i 6¼ j;

ð4:14Þ

ndij ð0Þ ¼ nd ð0Þ

for all i 6¼ j:

ð4:15Þ

4 Dynamics of Innovation Fields with Endogenous Heterogeneity of People

69

At the initial state, each pair of dancers has the same number of ideas, nc ð0Þ, in common. Moreover, for any pair of dancers, the number of ideas that one dancer knows but the other does not know is the same and equal to nd ð0Þ. Given that the initial state of knowledge is symmetric among the four dancers, it turns out that the equilibrium configuration at any time also maintains the basic symmetry among the dancers. When all dancers are pairwise symmetric to each other, that is, when mdij ¼ mdji

for all i 6¼ j

ð4:16Þ

the income growth rate (4.12) is simplified as X y_i n_ i ¼ ¼ dii a þ dij gðmdij Þ yi ni j6¼i

ð4:17Þ

and the dynamics (4.13) can be rewritten as m_ dij 1  mdij

h   i ¼ a dii 1  mdij  djj mdij  dij mdij gðmdij Þ  X X þ 1  mdij dik gðmdik Þ  mdij djk gðmdjk Þ; k6¼i; j

ð4:18Þ

k6¼i; j

where the function gðmÞ is defined as

gðmÞ ¼ b



1 m  m 2 3 1 1m 1m

ð4:19Þ

which represents the growth rate when the two persons meet. Figure 4.3 illustrates the graph of the function gðmÞ as a bold line for b ¼ 1. Differentiating gðmÞ yields, we can readily see that

 <2 1 for m 2 0; : ð4:20Þ >5 2   Thus, gðmÞ is strictly quasi-concave on 0; 12 , achieving its maximal value at mB ¼ 25; we call the latter the ‘‘Bliss Point’’. It is the point where the rate of increase in income or utility is maximized for each person. Next, taking the case of N ¼ 4, we illustrate the possible equilibrium configurations, noting that the equilibrium configuration can vary with time. Figure 4.4 gives the possibilities at any fixed time for N ¼ 4. Given that the initial state of knowledge is symmetric among the four dancers, as noted above, the equilibrium configuration at any time also maintains the basic symmetry among dancers. g0 ðmÞ

> 0 <

as m

70

M. Fujita . y y

B

g (m)

0.5

I

J

a

mB 0

mJ

0.4

m d (0)

mJ 0.5

m

Fig. 4.3 The g(m) curve and the bliss point when b ¼ 1

(a) solos

(b–1)

(b–3)

(b–2)

1

2

1

2

1

2

1

2

3

4

3

4

3

4

3

4

(c–1) 1

d12

(c–2)

(c–3)

(d)

2

1

2

1

2

1

2

4

3

4

3

4

3

4

d13 3

Fig. 4.4 Possible equilibrium configurations when N=4

Panel (a) in Fig.4.4 represents the case in which each of the four dancers is working alone, creating new ideas in isolation. Panels (b-1)–(b-3) represent the three possible configurations of partner dancing, in which each of the two couples dance separately but simultaneously. In panel (b-1), for example, 1 and 2 dance together. At the same time, 3 and 4 dance together.

4 Dynamics of Innovation Fields with Endogenous Heterogeneity of People

71

Although panels (a)–(b-3) represent the basic forms of dance with four persons, it turns out that the equilibrium path often requires a mixture of these basic forms. That is, on the equilibrium path, people wish to change partners as frequently as possible. The purpose is to balance the number of different and common ideas with partners as best as can be achieved. This suggests a square dance with rapidly changing partners on the equilibrium path. Please refer to panels (c-1)–(c-3) in Fig. 4.4. Each of these panels represents square dancing where a dancer rotates through two fixed partners as fast as possible in order to maximize the instantaneous increase in their income. In panel (c-1), for example, dancer 1 chooses dancers 2 and 3 as partners, and rotates between the two partners under equilibrium values of d12 and d13 such that d12 þ d13 ¼ 1. Dancers 2, 3 and 4 behave analogously. In order for this type of square dance to take place, of course, all four persons must agree to follow this pattern. Finally, panel (d) depicts square dancing in which each dancer rotates though all three possible partners as P fast as possible. That is, for all i 6¼ j, dij 2 ð0; 1Þ, and for all i, dii ¼ 0 and j6¼i dij ¼ 1. At this point, it is useful to remind the reader that we are using a myopic core concept to determine equilibrium at each point in time. In fact, it is necessary to sharpen that concept in the model with N persons. When there is more than one vector of strategies that is in the myopic core at a particular time, namely more than one vector of joint strategies implies the same, highest first derivative of income for all persons, the one with the highest second derivative of income is selected. The justification for this assumption is that at each point in time, people are attempting to maximize the flow of income. Now we are ready to investigate the actual equilibrium path, depending on the given initial composition of knowledge, which is common for all pairs i and j (i 6¼ j). In Fig. 4.3, let m J and mI be defined on the horizontal axis at the left intersection and the right intersection between the gðmÞ curve and the horizontal line at height a, respectively. mdij ð0Þ ¼ md ð0Þ ¼

nd ð0Þ : þ 2nd ð0Þ

nc ð0Þ

In the remainder of this paper, we assume that a < gðmB Þ

ð4:21Þ

so as to avoid the trivial case of all agents always working in isolation. Figure 4.5 provides a diagram explaining our main result. The top horizontal line represents the initial common state md ð0Þ, while the bottom horizontal line represents the final common state or sink point, md ð1Þ. There are four regions of the initial state that result in four different sink points, which are explained in turn below. Case 1: 0 < md ð0Þ  2=5 ¼ mB

72

M. Fujita

0

mB = 2/5

mJ

ˆ m

m1 1/2

md (0)

(iii) (i)

(ii)

(iv)

0 1/3

mJ

mB = 2/5

1/2

md (¥)

Fig. 4.5 Correspondence between the initial point md ð0Þ and the long-run equilibrium point md ð1Þ

First suppose that the initial state is such that mJ < md ð0Þ  mB : Then, since gðmdij ð0ÞÞ ¼ gðmd ð0ÞÞ > a for any possible dance pairs consisting of i and j, no person wishes to dance alone at the start. However, since the value of gðmdij ð0ÞÞ is the same for all possible pairs, all forms of (b-1) to (d) in Fig. 4.4 are possible equilibrium dance configurations at the start. To determine which one of them will actually take place on the equilibrium path, we must consider the second derivative of income for all persons. In general, consider any time at which all persons have the same composition of knowledge: mdij ¼ md

for all i 6¼ j;

ð4:22Þ

where gðmd Þ > a: Focus on person i; the equations for other persons are analogous. Since person i does not wish to dance alone, it follows that dii ¼ 0

and

X

dij ¼ 1:

j6¼i

Substituting (4.22) and (4.23) into (4.17) yields y_i ¼ gðmd Þ: yi

ð4:23Þ

4 Dynamics of Innovation Fields with Endogenous Heterogeneity of People

73

Likewise, substituting (4.22) and (4.23) into (4.18) and arranging terms gives     m_ dij ¼ ð1  md Þgðmd Þ 1  2md  1  md dij :

ð4:24Þ

_ above is independent of the values of dij Since the income growth rate y=y (j 6¼ i), in order to examine what values of dij (j 6¼ i) person i wishes to choose, we must consider the time derivative of y_i =yi . From this second-order condition, we can show that each person, say i, chooses the optimal strategy such that dij ¼

1 N1

for all j 6¼ i:

ð4:25Þ

The vector of optimal strategies is the same for all persons. Thus, all persons agree to a square dance in which each person rotates through all N  1 possible partners while sharing the time equally. The intuition behind this result is as follows. The condition md < 2=5  mB means that the dancers have relatively too many ideas in common, and thus they wish to acquire ideas that are different from those of each possible partner as fast as possible. That is, when mJ < mdij ¼ md < mB in Fig. 4.3, each dancer wishes to move the knowledge composition mdij to the right as quickly as possible, thus increasing the growth rate gðmdij Þ as fast as possible. This means that when mJ < md ð0Þ ¼ mdji ð0Þ < 2=5 ð¼ mB Þ for all i 6¼ j, on the equilibrium path, the square dance with dij ¼ 1=ðN  1Þ for all i 6¼ j takes place at the start. Then, since the symmetric condition (4.22) holds thenceforth, the same square dance will continues as long as mJ < md < 2=5 ð¼ mB Þ. The dynamics of this square dance are as follows. Setting mdij ¼ md and dij ¼ 1=ðN  1Þ in (4.24), we obtain m_ d ¼ ð1  md Þgðmd Þ

ðN  2Þ  ð2N  3Þmd : N1

ð4:26Þ

Setting m_ d ¼ 0 and considering that md < 1, we obtain the sink point md ¼

N2 : 2N  3

ð4:27Þ

Surprisingly, when N ¼ 4, md ¼ 2=5 ¼ mB . The value of m_ d is positive when m < mB ¼ 2=5, and zero if md ¼ 2=5. Hence, beginning at any point md ð0Þ ¼ 2=5, the system moves to the right, eventually settling at the bliss point mB . Since the right hand side of (4.27) is increasing in N, d

md ¼

N2 > 2=5  mB 2N  3

when N > 4:

ð4:28Þ

Hence, when N > 4 and N is divisible by 4, beginning at any point m < md ð0Þ < 2=5, the system moves to the right and reaches mB ¼ 2=5 in finite J

74

M. Fujita

time. When N agents reach the bliss point mB , they break into groups of 4 to maintain heterogeneity at the bliss point. Next, when 0  md ð0Þ < mJ , it is obvious that the four persons work alone until they reach mJ . Then they follow the path explained above, eventually reaching mB . Case 2: ^ mB < md ð0Þ  m Next, let us consider the dynamics of the system when it begins to the right of ^ < mI .11 The equilibrium process takes the following mB ¼ 2=5 but to the left of m three phases. Phase 1: Since the initial state reflects a higher degree of heterogeneity than the bliss point, the dancers want to increase the knowledge they have in common as fast as possible, which leads to couple dances. Thus, person i wishes to choose any partner, say k, and set dik ¼ 1, whereas dij ¼ 0 for all j 6¼ k. The situation is the same for all dancers. Hence, without loss of generality, we can assume that N persons agree at time 0 to form the following combination of partnerships: P1  ff1; 2g; f3; 4g; f5; 6g; . . . ; fN  1; Ngg

ð4:29Þ

and initiate pairwise dancing such that dij ¼ dji ¼ 1

for fi; jg 2 P1 ;

dij ¼ dji ¼ 0

for fi; jg 2 = P1 :

ð4:30Þ

The same pairwise dance, however, cannot continue too long by the following reason. On one hand, the proportion of differential knowledge for each couple, say f1; 2g, decreases with time, making the partnership less productive eventually. On the other hand, the proportion of the differential knowledge increases for any pair of persons, say f1; 3g, who are not dancing together. Thus, eventually, the shadow partnership f1; 3g 2 = P1 becomes more productive than the actual partnership f1; 2g 2 P. Thus, there exists a switching time t0 at which each dancer switches to a new partner. Phase 2: One example of new equilibrium partnerships at the switching time t0 is given by P2  ff1; 3g; f2; 4g; f5; 7g; f6; 8g; . . . ; fN  3; N  1g; fN  2; Ngg

ð4:31Þ

meaning that the first four persons form a group and exchange partners, the next four persons form another group and switch partners, and so on. (There exist many other possibilities for equilibrium partnerships to be chosen by N dancers at time t0 . It turns out, however, that the essential characteristics of equilibrium dynamics are ^ see Berliant and Fujita (2007). For the determination m,

11

4 Dynamics of Innovation Fields with Endogenous Heterogeneity of People

75

not affected by the choice at time t0 . Hence, let us assume that N persons agree to choose the new partnerships P2 at time t0 .) It turns out, however, that these new partnerships last only for a limited time. To examine this point, let us notice that in the dance form P2 , each group of four persons is isolated from everyone else. Thus, in the sequel, we focus on the dynamics of a four-person group, 1, 2, 3 and 4. Under the partnership P2 , since md12 ðtÞ is increasing with time while md13 ðtÞ is decreasing, there exists a time t00 at which md12 ðtÞ and md13 ðtÞ become the same, md12 ðt00 Þ ¼ md13 ðt00 ÞmB

ð4:32Þ

which can be shown to occur in the left of the bliss point mB . Thus, if partnerships f1; 3g and f2; 4g were maintained beyond time t00 , then it would follow from (4.20) that gðmd12 ðtÞÞ > gðmd13 ðtÞÞ for t > t00 : This implies that the same partnerships cannot be continued beyond t00 . To see what form of dance will take place after t00 , first note that dancers cannot go back to the previous form of partnerships f1; 2g and f3; 4g. If they did so, then the proportion of the knowledge in common for the actual partners f1; 2g would increase, while the proportion of the differential knowledge for the shadow partnership f3; 4g would increase. This means that the following relationship, md12 ðtÞ < md ðt00 Þ < md13 ðtÞ < mB holds immediately after t00 , and thus gðmd12 ðtÞÞ < gðmd13 ðtÞÞ which contradicts the assumption that f1; 2g is the actual partnership. Furthermore, relation (4.32) implies that under any possible partnership, the following inequality gðmd13 ðtÞÞ > gðmd14 ðtÞÞ holds immediately after t00 . Thus, immediately after time t00 , the equilibrium dance cannot include partnerships f1; 4g and f2; 3g. Hence, provided that gð1=3Þ > a, we can see from Fig.4.4 that the only possible equilibrium configuration immediately after t00 is a square dance in the form (c-1) involving a rapid rotation of non-diagonal partnerships, f1; 2g, f1; 3g, f2; 4g and f3; 4g. That is, for dancer 1, d11 ¼ 0 and d1j ¼ 12 if j ¼ 2 or 3, d14 ¼ 0. Analogous expressions hold for the other dancers. Phase 3: The dynamics for this square dance under the form (c-1) are as follows. We set mdij  md

for fi; jg 2 P2 :

76

M. Fujita

Then, since conditions (4.22) and (4.23) hold also in the present context, setting dij ¼ 1=2 in (4.24), we get m_ d ¼ ð1  md Þgðmd Þ

1  3md ; 2

which is negative when md > 13, and zero if md ¼ 13. Thus, beginning at any point md ðt00 Þ > 13, the system moves to the left, eventually settling at md ¼ 13. Case 3: ^ < md ð0Þ  mI m ^ < md ð0Þ  mI . As in Case 2, dancers are Next suppose md ð0Þ is such that m more heterogeneous than at the bliss point, so they would like to increase the knowledge they hold in common through couple dancing, for example using configuration (b-1) in Fig. 4.4. The initial phase of Case 3 is the same as the initial phase of Case 2. However, since gðmd12 ðtÞÞ > gðmd13 ðtÞÞ for all t before md12 ðtÞ reaches mJ , whereas gðmd12 ðtÞÞ > a > gðmd13 ðtÞÞ when md12 ðtÞ reaches mJ . So each dancer keeps their original partner as the system climbs up to B and on to J. When the system reaches md ðtÞ ¼ mJ , each dancer uses fractional dij to attain mJ by switching between working in isolation and dancing with their original partner. Case 4: mI < md ð0Þ  1=2 Finally, suppose md ð0Þ > mI . Then, gðmd ð0ÞÞa, and hence there is no reason for anyone to form a partnership. Thus, each person dances alone forever, and eventually reaches md ¼ 1=2. Compiling all four cases, we obtain the result summarized in Fig. 4.5. There are important remarks to be made about the result. First, the sink point changes discontinuously with changes in the initial conditions. Second, from each set of initial conditions, the N persons eventually divide into many separate groups between which no interaction occurs. Thus, from an initial state that is symmetric, we obtain an equilibrium path featuring asymmetry. Third, concerning the welfare properties of the equilibrium path, the most surprising result is with Case 1. That is, whenever md ð0Þ < mB , the equilibrium path either approaches (when N ¼ 4) or reaches in finite time (when N > 4) the most productive state, mB . Clearly, initial heterogeneity plays an important role in the efficiency properties of the equilibrium path. What distinguishes Case 1, aside from a relatively homogeneous beginning, is that the dancers can switch partners rapidly enough to increase heterogeneity while at the same time maximizing the increase in output. That is because each agent spends 1=ðN  1Þ of the time dancing with any particular agent, and

4 Dynamics of Innovation Fields with Endogenous Heterogeneity of People

77

ðN  2Þ=ðN  1Þ of the time dancing with others. This is what leads to the most productive state.12 Bearing in mind the limitations of the model, it may have empirical relevance. The main result may explain the agglomeration of a large number of small firms in Higashi Osaka or in Ota ward in Tokyo, each specializing in different but related manufacturing services. Another example is the third Italy, which produces a large variety of differentiated products. Yet another example is the restaurant industry in Berkeley, California. In each case, tacit knowledge accumulated within firms plays a central role in operation of the firms.

4.4

Conclusion

We have presented a micro-model of knowledge creation through the interaction of a group of people. Our model incorporates two key aspects of the cooperative process of knowledge creation: (1) heterogeneity of people in their state of knowledge is essential for successful cooperation in the joint creation of new ideas, while (2) the very process of cooperative knowledge creation affects the heterogeneity of people through the accumulation of knowledge in common. The model features myopic agents in a pure externality model of interaction. Surprisingly, in the general case for a large set of initial conditions we find that the equilibrium process of knowledge creation may converge to the most productive state, where the population splits into smaller groups of optimal size; close interaction takes place within each group only. This optimal size is larger as the heterogeneity of knowledge is more important in the knowledge production process. Equilibrium paths are found analytically, and they are a discontinuous function of initial heterogeneity. However, what we have done so far is, in effect, to open Pandora’s box, scattering around a great number of new problems to be investigated further. Indeed, to take our model more realistic and interesting, we must extend it by considering/introducing various new elements such as knowledge transfer, knowledge structures and hierarchies, side payments and the markets for ideas, foresights and strategic behavior, and uncertainty and stochastic elements. In particular, we must return to our original motivation for this model, as stated in the introduction. That is, location seems to be an essential feature of knowledge creation and transfer, so regions and migration are important, along with urban economic concepts more generally. Thus, incorporating locations/regions in our model, we may be able to move one step closer to our ultimate objective of developing a comprehensive 12

Here, it is natural ask why the optimal group size in knowledge production is four. Actually, using a more general functional form of joint knowledge production, Berliant and Fujita (2007) shown that when differential knowledge is relatively more important than common knowledge in knowledge production, the optimal group size is larger.

78

M. Fujita

theory of geographical economics in the brain power society, in which the dual linkages in the economic and knowledge fields work in unison. As the model becomes more realistic and hence more complex, however, its analytical tractability reaches the limit soon. Eventually, thus, we must appeal to computer simulations. In particular, the evolutionary process of knowledge creation and transfer may be simulated with the help of multi-agent-based simulation. Acknowledgments The author is grateful to the three anonymous referees and to David Batten, ˚ ke Andersson and other participants in the workshop on ‘‘Innovation, Dynamic Regions and A Regional Dynamics’’ for valuable comments on the earlier drafts of this paper. The author is also grateful for Grants Aid for Scientific Research Grants S 13851002 and A 18203016 from the Japanese Ministry of Education and Science.

References ˚ E (1985) Kreativitet: storstadens framtid. Prisma, Stockholm Andersson A Baldwin R, Forslid R, Martin P, Ottaviano G, Robert-Nicoud F (2003) Economic geography and public policy. Princeton University Press, Princeton Berliant M, Fujita M (2007) Knowledge creation as a square dance on the Hilbert cube. Institute of Economic Research, Kyoto University, Kyoto (mimeo) Fujita M (2005) Spatial economics. Edward Elgar, Cheltenham Fujita M, Krugman P (2004) The new economic geography: past, present and the future. Pap Reg Sci 83:149–164 Fujita M, Mori T (2005) Frontiers of the new economic geography. Pap Reg Sci 84(3):377–405 Fujita M, Thisse J-F (2002) Economics of agglomeration: cities, industrial location, and regional growth. Cambridge University Press, Cambridge Fujita M, Krugman P, Venables AJ (1999) The spatial economy: cities, regions and international trade. MIT, Cambridge, MA Jacobs J (1969) The economy of cities. Random House, New York Krugman P (1991) Increasing returns and economic geography. J Polit Econ 99:483–499 Lucas RE Jr (1988) On the mechanics of economic development. J Monet Econ 22:2–42 Marshall A (1890) Principles of economics. Macmillan, London Porter ME (1998) On competition. A Harvard business review book. Harvard Business School Press, Boston, MA Thurow LC (1996) The future of capitalism. Leighco, New York Zipf G (1949) Human behavior and the principle of least effort. Addison-Wesley, New York

Chapter 5

Economics of Creativity ˚ ke E. Andersson A

5.1

Division of Labor by Comparative Advantage or Creativity

Most of us have got an education adapted to the demands for specialized labor emanating in industry or public administration. Most of the jobs have been decided according to the basic principle of division of labour, generating productivity of the work. According to this principle the worker should be specialized to perform certain highly specialized tasks without any greater space for improvisation or change of work routines. Adam Smith (1776, 1904) argued strongly in favour of a far-going division of labor (or specialization of the workforce) as a way of achieving growth of productivity. However, Adam Smith clearly saw the potential conflict between creativity and productivity by division of labor and specialization of the work force: In the progress of the division of labor, the employment of the far greater part of those who live by labor, that is, of the great body of the people, comes to be confined to a few very simple operations; frequently to one or two. But the understandings of the greater part of man are necessarily formed by their ordinary employments. The man whose whole life is spent in performing a few simple operations, of which the effects to are, perhaps, always the same, or very nearly the same, has no occasion to exert his understanding, or to exercise his invention in finding out expedience for removing difficulties which never occur. He naturally loses, therefore, the habit of such exertion and generally becomes as stupid and ignorant as it is possible for a human creature to become. (Wealth of Nations, II)

The industrial society became based on a far-going division of labor and a hierarchical organization of the firms. Research and development became a sort of tinkering, oriented to improvement of the techniques for producing a given set of goods. Creativity was looked upon as a social side-activity for artists, scientists and inventors. The first stage of an upgrading of creativity was to occur during the Second World War, when decision-makers realized that at least chemists and physicists ˚ .E. Andersson A Jo¨nko¨ping International Business School, Jo¨nko¨ping University e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_5, # Springer‐Verlag Berlin Heidelberg 2009

79

˚ .E. Andersson A

80

were of value in military projects. The largest example was the Manhattan project, within which scientists were organized into secret research groups with a mission to transform the knowledge of theoretical physics into an atom bomb (Fermi 1954/1994). On the basis of this experiment in organized creativity American think tanks became a way of improving the cooperation between creative scientific research and the development and innovation of new products in the post-war American industry. A real integration of creative research and technological development was, however, not realized before the end of industrialism in USA and Western Europe. In the early 1970s Daniel Bell (1973) formulated a scenario describing a new postindustrial society. It was based on the observation that manufacturing industry in USA and Western Europe had already seen its employment stagnating and even declining. It became obvious that the highly industrialized societies could no longer expect an increasing employment in the production of material goods. Many of the analysts of the 1970s expected service industries to become the new guarantee of full employment. Few analysts expected creativity in science, technological research and development, design, entertainment and arts to become an important factor explaining growth of real income, employment and general welfare in the postindustrial society. Real developments in the structures of some regions, e.g. San Francisco Bay with Silicon Valley, Route 128 around Boston, Massachusetts and Cambridge, UK, saw a new type of interaction between creative scientists and industry, indicating a new role for creativity in the economic system. In recent decades the role of creativity as a factor of economic development has been realized in somewhat surprising directions. First, there has been a rapid increase in resources allocated to scientific research. The number of science articles published has been increasing at approximately 7% annually since 1975 (Andersson and Persson 1993). Second, industrial research and development (R&D) has become a strategic factor of growth policies among firms and governments of OECD-countries since the 1960s. This development has triggered numerous scientific papers on the interdependencies between R&D and economic growth (see, e.g. Uzawa 1965; Shell 1966; Romer 1986). Third, there has been a remarkable growth of the entertainment and arts activities, called Creative Industries by Richard Caves (2000). According to recent statistics consumption of such goods and services has risen to more than 15% of total household consumption in Sweden.

5.2

Mechanisms of Creativity

Creativity is a process based on a capacity. As a process it is dynamic, because creativity always means the emergence of something genuinely new. Discoveries and inventions are outcomes of a creative process. Discovery is based on a capacity to find patterns in a seemingly chaotic world. The real creative capacity lies in the

5 Economics of Creativity

81

Fig. 5.1 This implies that the brain has the tendency to be anchored in the original perception and needs a certain excess supply of information before it can give up the initial interpretation in favor of a new. There is certain stabilization in the already known. Expressed differently, creativity requires a certain degree of instability of the brain. Such instability is evidently there in all of us

ability to comprehend and explain the mechanisms generating such patterns. The detection of a hidden pattern and its transformation into something meaningful is often something suddenly occurring in the brain. The mathematician I. N. Stewart and the psychologist P.L. Peregoy (1983) have, shown by a series of experiments, how the brain can discover a hidden structure. With this experiment they can support the claim that the brain ought to be represented as a non-linear dynamical system. Using Fig. 5.1 they were able to show that the perception of a man is suddenly changed into a clear perception of a woman after three to six steps from left to right and the perception of a woman is suddenly transformed into a perceived man, when starting from the right and moving three to five steps to the left. Inventions and discoveries are different names of the created ideas. An invention mostly starts by perceiving a structure and later suddenly realizing that below this

˚ .E. Andersson A

82

surface structure there is a more important deep structure that can be used in the formation of a new principle of composition to be used as an instrument of generating inventions. Margaret Boden (1990) has proposed a subdivision of creativity into different classes. The first class of creativity implies the invention of a completely new principle of construction, composition or set of concepts, providing a new structuring of some problem area. This type of creativity is fundamental or infrastructural. The other type of creativity is built on variations of themes given by a given creative infrastructure. A few examples suffice to clarify the differences. Schoenberg was the creator of the most important principles of composition of 12-tone music and would consequently be seen as the creator of the infrastructure of modernist music. In contrast Anton Webern and Alban Berg would give examples of variational creativity in their application and further development of the original new principles of composition as created by Schoenberg. When applied to painting, the same principle would imply that Cezanne is the infrastructurally creative artist within modernist painting, while Braque and Picasso would be the most important painters in terms of variational creativity on this basis. In science an example is Inequalities by Hardy et al. (1934). Reformulating many mathematical equations as inequations they formed a basis for much of the developments in mathematical programming developed and innovated in the 1940s. In this context George Dantzig and Harold Kuhn with their formulations of linear and non-linear programming would be examples of variational creators. It ought to be stressed that there is no obvious qualitative distinction between infrastructural and variational creators, except in terms of the potential of further developments on the basis of the infrastructural creators.

5.2.1

Creative Capacity: Acquired or Inherited?

Are all people born creative? There are certain indicators that creativity is not a genetic deviation from the normal but rather a general human capacity. One indicator is the development of the capacity to speak. Already in small children completely new spoken sentences are created. Even the smallest child can create completely new linguistic constructions in their communication with other children and adults. Sometimes they even seem surprised at their own linguistic discoveries and inventions. The capacity of linguistic creation seems to develop by social interaction throughout the life span. The concentration of musical and pictorial creativity onto a minority of the population might just be a consequence of too little of daily training during the early years of childhood. Most of the creative musicians and other artists have had the benefit of an education in the arts from their earliest years. Surprisingly many artists have grown up in an environment rich in artistic activities. Simonton (1984) has used extensive empirical material to show that the early exposure to scientifically or artistically creative personalities is of importance for creativity of young people.

5 Economics of Creativity

83

The formal schooling of children does not generally compensate for the lack of artistic and other creative inspiration in the homes of children. In contrast most educational systems in the old and new industrialized countries have been oriented on diffusion of already well established knowledge later to become useful in manufacturing firms or bureaucracies. This implies that schools have primarily been oriented to the development of discipline and adaptation and to the need for cooperation in groups with specified problems to be solved as rapidly as possible. The education before the university level is rarely oriented to formulation of problems independently and to the generation of different ways of solving such problems. Rather, most education is oriented to learning techniques of how to solve already formulated problems in a way, pleasing to the teacher. The learning of already developed techniques has been favored at the expense of a loss of creativity already during the elementary school years. Gudmund Smith (1990) has in his studies of the psychology of creativity found that the development of creativity during the years of childhood and adolescence follows a typically cyclical pattern. During some of these cyclical periods learning is favored and absorption of education is easy, while in other periods creativity develops rapidly. The ages of development of creativity seem to be between 5 and 7 years, 10–12 years and 17–19 years of age. In most industrial countries the latter two creativity peaks seem to be used by the schools for intensive teaching and examinations, curbing the development of creativity. Smith has even claimed that a school where development of creativity has a priority might need to be free of fixed curricula.

5.3

Creative Personalities

The transformation from an industrial society towards a society based on the exploitation of knowledge, creativity in the arts, design, and entertainment and with an increasing complexity of products will need a better development of as well as use of human creativity. Finding and supporting people, suitable for creative work has become much more important than during the industrial era. Gudmund Smith (1990, 1995) has oriented some of the research of his team towards investigations of creative personality traits. Some of the results can be summarized. First, a typical characteristic of a creative personality is a capacity to formulate and energetically work on the solution to the formulated problem. Sometimes the problem is not conceived as especially interesting by anyone else and is often looked upon as somewhat strange or even bizarre by others. Second, a part of a creative personality is a subjective and quite an emotional relation to the problem which is developed during the period of problem solving. The solution to such an independently formulated problem is often not obviously profitable for anyone. Third, another personality trait is an orientation towards aesthetic solutions to the generated problem.

˚ .E. Andersson A

84

Fourth, a general characteristic of creative personalities, according to Smith, is the oceanic capacity. This is a capacity to get a feeling of almost infinite possibilities, when a new creative solution turns out to be correct. This implies an instant and yet sustainable reward of greater importance than external rewards in the form of money or fame. Fifth, creative persons tend to be victims of angst, which according to Smith is the natural companion of creative activities. Sixth, creative persons tend to have – in comparison with the non-creative – a strong interest of their childhood. They often think about it and it is prevalent in their dreams which are more frequent than among non-creative persons. One of the surprising properties of these dreams is that they are described in the interviews of creative persons as dreams in intensive colors. For these and possibly other reasons there is a tendency among creative persons to combine childish behavioral traits with a capacity to concentrate and be quite serious in the process of formulating and solving problems. It does not seem to be the case that very goal-oriented, wealthy homes are the best breeding grounds for the development of creativity among children. Remarkably often creative persons seem to have come from disadvantaged homes.

5.4

Different Capacities of the Creative Mind

In his book How to Solve It, the mathematician George Polya (1945) claims that the most important approach to creative problem formulation and solving is by heuristics or the use of proper analogies: ‘‘Analogy pervades all our thinking, our everyday speech and our trivial conclusions as well as our ways of expression and the highest scientific achievements’’ (Polya 1945, p.37). This is obvious in mathematics but seems to be of relevance also in creative writing and composing. Belonging to some style or genre of literature essentially means that a certain degree of similarity of composition exists. Such formulations are often analogous at least in terms of deep structure. Production requires predictability and structural stability of the process in order to be efficient. Creation is an almost contrary process. The creator has to accept fundamental uncertainty and its companion, structural instability. This implies that creativity can only be achieved by individuals, who have accepted a career with an embedded uncertainty of production and the corresponding uncertainties of income and wealth.

5.5

The Pecuniary Rewards of Creativity

In the scientific world, income is normally secured for the creators by a combination of subsidies and payment for other work than creation of scientific research. In universities much of the salaried time is used for elementary teaching, administration

5 Economics of Creativity

85

and other non-creative activities. The financial rewards for creativity are primarily determined by decisions in public or private funds based on earlier research records and an estimate of the likelihood of success as evaluated by some more or less credible peer group. In the R&D world the financial rewards are calculated with methods similar to the ones used in the evaluation of the returns to material investments, i.e. an estimate is made of the expected net present value and risk. Because of the public nature of knowledge the risk is very large and different procedures to protect the inventions are necessary. The common procedure of protecting a new material product is patenting that exists in all countries prone to imitate new knowledge. Patent rights are regulated by international treaties and give the property right to the proceeds from the new product for a time period of 20 years. However, in reality the rights can normally be executed for approximately 15 years. Because of the delays in production, rights are executed after the patenting has been granted. In the arts world there is a situation somewhat similar to the scientific world. Composers and other creative musicians are often hired to do non-creative work such as teachers, administrators or regular employees of subsidized orchestras. Painters and authors can rarely live from their creative work and have to live from incomes as teachers, postmen and other non-creative jobs. Economies of scale are of great importance in the entertainment world. Making a film normally requires 200–400 man-years and large amounts of studio equipment and other material capital with large fixed costs as a consequence. This has led to a number of organizational responses, such as conflicts about quality and economic rewards among composers and script writers, reliance on performance stars, spatial concentration of production and syndication of the outputs.

5.6

Variable Probabilities and the Importance of Stars

In industrial R&D the probability of success of a particular project has been estimated to be in the range of 7–12%. This means that the majority of projects will be financial fiascoes. To compensate for the losses, most of the industrial research and development costs are borne by large firms in a limited number of manufacturing sectors. These firms are large enough to run a substantial number of parallel R&D projects to compensate for the low success probabilities of most of these projects. The substantial returns of a few of these projects must then compensate for the losses of most of the projects. This is partially true for entertainment firms, such as Disney, Sony or MTG. While most painters and authors are struggling in the first hand market to achieve a reputation a few, often dead colleagues, have become important suppliers in the second hand market of originals and reproductions. Many art and entertainment goods – books, magazines, movies or amusement games – are only sold to final users as copies and the markets for these reproductions are quite different from the markets for originals. Most reproduction

86

˚ .E. Andersson A

processes – apart from forgery and other hand-copying – are multi-stage processes with complicated rules of interaction between stages. One example is the music industry (see Table 5.1). There are distinct probabilities of success in the interaction between agents within and between the different stages of such a production and reproduction process and associated problems of negotiating the reward structure. Assuming the probability of success to be the same everywhere and equal to 50%, the probability of success for the whole 4 by 2 process is (1/2)8, which is approximately equal to 0.4%. In this case, the popular music publisher would accordingly need to judge thousands of music proposals from unknown creative music composers to be reasonably sure of a success in the market. Raising the probability of success within and between stages to 90% would lead to a probability of success of the whole four–stage process to 43%. There have consequently been efforts to design individual and institutionalized procedures to increase these probabilities within and between all stages. It is for example often the case that artists compose music and write lyrics themselves. Publishing and recording can be vertically integrated and the owners of record companies can also own television stations, and so on. In film production, these problems are further reinforced by the complexity of production of film negatives (Vogel 1998). Composers and directors often have their contract income based on revenues and therefore they tend to be oriented to the maximization of quality and quantity with potentially detrimental consequences for the profitability of the whole process. Economic efficiency in music and film making would gain from contracts based on profit-sharing for the creators. However, there are several problems associated with profit-sharing that are especially relevant in the complex structures of modern music and film-making. Substantial parts of the fixed costs are unknown to the creators and can easily be redistributed between different products (and their creators). The heterogeneity of arts and entertainment products associated with the dependency of consumer taste on the individual characteristics of a few star performers is especially important in this context. Certain consumers may have a strong preference for individual performers, such as the pianist Glenn Gould, the singer Ella Fitzgerald or the actor Julia Roberts. Such stars do in fact have an almost monopolistic negotiating position at each first recording of a piece of music or a film Table 5.1 The music industry as a multi-stage production process Stage 1 Composition of music Artist’s first performance: innovation (including lyrics): creation Stage 2 Music publisher: production Diffusion to reproducers Stage 3 Recording on CDs and Diffusion to radio and TV stations, DVDs: production and record distributors Stage 4 Purchases by consumers: Collection of royalty incomes by ASCAP, BMI, diffusion SESAC, etc., for distribution between upstream agents

5 Economics of Creativity

87

manuscript. In a way the appearance of such an artist increases all the probabilities of success discussed above and all of the agents have to yield to this fact. The complex production technology of most reproductive art and entertainment goods leads to high fixed costs of production and globally concentrated industries. The film industry is one such globally concentrated industry. Most countries rely on imports of films from the global centers of production and especially from Hollywood. This is a consequence of the complexity of production, which causes high fixed and irreversible costs for each film. These scale economies are further reinforced by the low probability of success of each individual film. The organizational result has been an increase in the size of firms, which makes it possible to diversify production in order to reduce the risk of bankruptcy. Table 5.2 gives the size of film production in a number of countries, measured as the number of film negatives produced from 1991 to 1995. The rank size distribution of film production in different nations is as follow Film production ¼ e7:22 ðRankÞ1:3 ;

R2 ¼ 0:95:

An alternative approximation form of the distribution is Film production ¼ eð5:90:12ðRankÞÞ ;

R2 ¼ 0:95:

These equations imply that the distribution is highly skewed, which is also indicated by the fact that the mean of the number of films produced is more than twice as large as the median production. Vogel (1998) collected financial data for the production of films in the United States from 1976 to 1996. While some of these films were profitable, others suffered disastrous losses. The mean cost of production was US $34 million with a standard deviation of US $23 million, while the mean revenue was US $91 million with very large standard deviation of US $81 million. There was no correlation between revenues and costs. Table 5.2 Production of film negatives in the top ten countries in the period 1991–1995 Rank Country Number of film negatives 1 India 838 2 United States 420 3 Hong Kong 315 4 Japan 251 5 Thailand 194 6 China 154 7 France 141 8 Italy 96 9 Brazil 86 10 United Kingdom 78 Source: UNESCO (1998), World Culture Report

˚ .E. Andersson A

88

Using regression analysis, we estimated the effect of top-ranking directors or actors on assessed revenue. The result is as follows: ln ðRevenueÞ ¼ 2:88 þ 1:41T;

ðn ¼ 23Þ;

in which T=1 if a top-ranked director or actor is involved in the making of the film (otherwise T=0). The t-value of the slope parameter estimate is 2.3, indicating that the estimated value is significantly different from zero at the 5% significance level. The regression equation implies that a Hollywood-produced film without a topranking director or actor can be expected to generate US $18 million in revenue, while the revenue figure for a film with a top-ranking director or actor is US $73 million. For production costs, there is no corresponding statistically significant ‘‘celebrity impact’’. This impact gives these artists a strong bargaining position, which should enable them to obtain substantial shares of revenues or profits. The contract variations are almost endless, but it is not unusual that the leading actor, actress, and the director together obtain more than 10% of the total revenue when the total exceeds US $150 million.

5.6.1

Lining up Behind Giants

Most labor markets are similar to markets for standardized goods. The price of the good itself and the prices of substitutes and complements determine the supply. Similarly, different prices determine the demand and the supply and demand simultaneously determine the equilibrium price and quantity. In the labor markets there are deviations from this simple competitive principle. Some occupations require many years of education and training and the movement toward equilibrium is consequently slow. Institutional safety constraints regulate other types of labor, as for instance airline pilots or medical doctors, which therefore constrain the supply. For some occupations, unionization works as a barrier to entry, which prevents the attainment of a competitive equilibrium. These factors to some extent are also relevant for artists. However, more important are the combined effects of the number of gatekeepers that block entry and advancement and the uncertain success of the final, creative product. Market success depends on the impact of the most visible artist who is involved in production. Because of the intangibility of created ideas, when innovated as a piece of music or a new film, expectations are of great importance for the demand on the day of the premiere. Expectations of a rewarding experience derive from the probabilities of success, as the potential audience perceives them. These perceptions in turn depend on the rank of the artist among the group of comparable artists. There is in most artistic and entertainment occupations a continuous inflow of new entrants, owing to the attractiveness of many artistic careers to young people. Most of these new entrants fail when attempting to get on

5 Economics of Creativity

89

the first rung of the career ladder (i.e. through the first ‘‘gate’’), and the probability of failure becomes greater at each further step on the career ladder to stardom. Let us assume that the probability of advancing from one rung of the ladder to the next is 5%. The probability of succeeding to the nth level is then 0.05n. If there are five rungs, the probability of reaching the fifth level is 3 in 10 million. If we instead assume that a person has talent enough to have a probability of 20% to climb each rung of the ladder, the probability will equal 0.32 in 1,000 attempts to reach the top. If we assume that there are one million aspiring young entertainers and there is a probability of 10% (i.e. probability is 0.1) to reach local recognition, there will be 100,000 local successes. If there is an additional 10% probability to reach regional recognition, it means that 10,000 will continue to that level in their career. Let us assume that the probability is 20% that they will reach national recognition, given that they are already regionally recognized, then that would imply that 2,000 will reach that stage of their career. To reach recognition on a continental scale might have a very low probability of, say, 1%, so that only 20 will reach that level of recognition and finally maybe only five will have a substantial global impact. There are many ways to measure the impact of an artist. In science, it is common to use global citations in scientific journals to measure the impact of a scientist on the public (in this case, other scientists). To an artist, recognition by other artists is often pleasing, but of little importance in the markets for artistic products. We therefore need some other, a more general measure of impact. We have chosen to use the number of Google (an internet search engine) hits as such a general measure of the impact of various kinds of artists. Tables 5.3, 5.4 and 5.5 reveal the impact of different creative artists, as measured by Google hits in early 2005. The average year of birth of the top ten composers is 1789. This implies that the average age of the top ten compositions is almost two centuries. This is also reflected in the current programming strategies among concert houses and symphony orchestras. The importance of the English language for global success is clear from these rankings. Six out of the top ten Nobel laureates have English as their mother tongue. No such language effect is discernible for the other art forms (except for films). Table 5.3 Top ten composers of classical music Rank Composer Year of birth 1 J.S. Bach 1685 2 L. van Beethoven 1770 3 W.A. Mozart 1756 4 G. Verdi 1813 5 F. Schubert 1797 6 P. Tchaikovsky 1840 7 J. Brahms 1833 8 D. Shostakovich 1906 9 F. Chopin 1810 10 A. Vivaldi 1678 Sources: Larousse Encyclopedia of Music and Google, January 2005

˚ .E. Andersson A

90 Table 5.4 Top ten laureates in literature Rank Nobel laureate 1 J.P. Sartre 2 T.S. Eliot 3 B. Russell 4 W.B. Yeats 5 G.B. Shaw 6 T. Mann 7 S. Beckett 8 A. Camus 9 W. Faulkner 10 A. Gide Source: Google, January, 2005

Table 5.5 Top ten jazz musicians Rank Musician Year of birth 1 M. Davis 1926 2 C. Parker 1920 3 L. Armstrong 1900 4 B.B. King 1925 5 B. Webster 1909 6 L. Young 1909 7 King Oliver 1885 8 E. Fitzgerald 1919 9 D. Ellington 1899 10 B. Holiday 1915 Source: Larousse Encyclopedia of Music and Google, January 2005

The average year of birth of the top ten jazz musicians is 1910. All except one have passed away and can only be heard on recordings. One way of analyzing the citation rates of the ranking lists of artists is by using the following equation: Citations ¼ eðabðRankÞÞ : We have used least-squares regression analysis to estimate the parameters a and b. The parameter estimate b refers to the percentage decline in the number of citations of the artists when their ranking is increased by one unit. The estimated equation for the 40 highest-ranked composers is Citations (composersÞ ¼ eð7:50:07ðRankÞÞ ;

R2 ¼ 0:98:

Increasing the number of observations does not influence the equation to any considerable degree. The estimated equations for the other groups of artists are as follows: Citations ðNobel laureatesÞ ¼ eð5:570:10ðRankÞÞ ;

R2 ¼ 0:98;

5 Economics of Creativity

Citations ðJazz musiciansÞ ¼ eð8:040:10ðRankÞÞ ;

91

R2 ¼ 0:96:

These equations indicate exponential decline of rank-ordered citation rates and are remarkably good at accounting for the variability in the number of citations. A conversion of these citation rates into probabilities of recognition gives a similar rapid decline of recognition as we move down the rankings of the artists. The estimates also show that jazz musicians and Nobel laureates have greater estimated b coefficients in absolute values, which possibly reflect the lower age of their works. The average birth year of the top ten creative artists varies considerably among the different categories, as shown in the above tables. In literature and music there are incredible numbers of ‘‘giants’’ who implicitly compete with new entrants aspiring for positions of global fame. A young painter, poet or composer therefore has to compete for recognition with artists who died a long time ago. This competition with the dead generates incentives for creative artists to develop new styles, niches or even completely new rules of composition. The extreme durability of great art is an advantage to the general public but an obstacle to recognition among all aspiring artists. The skewed distribution of recognition among creative artists leads to a correspondingly skewed distribution of revenues, which inevitably leads to a skewed distribution of artists’ material assets and incomes. By way of example, assume that the price of a painting by the highest-ranked artist is $100 million. If the price distribution corresponds to an estimated citation function, this would imply that an artist at global rank 100 would receive $33,000 per painting, while the painter who is ranked as number 150 in our global ranking would receive only $614 for a painting. The top ten would generate most of the total wealth derived from the sale of paintings in these circumstances, as long as the supplied quantities do not increase dramatically with increasing rank number (i.e. decreasing number of citations). Our example conforms in its general pattern to the markets for paintings and compositions in classical music, but it does not conform to the markets for films and popular music, where the rankings change rapidly. However, even in these more changeable markets a similar pattern persists at each short period of time. During their much shorter stable ranking periods, the rent and income distribution should be expected to be extremely skewed in favour of only the top-ranking segment or sometimes even just one giant.

5.7

Syndication

A special form of vertical and horizontal integration – syndication – is typical of arts and entertainment industries (Werbach 2000). The basic preconditions for economic advantage of syndication are the following: 1. The product must have the property of a public good, i.e. it should be possible to be used by many at the same time or consecutively, i.e. the same unit of

˚ .E. Andersson A

92

2. 3. 4.

5.

6.

a product can generate utility to many users. This is typical of information and knowledge. A concert by the Vienna Philharmonics on n radio and television stations does not decrease the quantity or quality to the listeners of the concert, even if n goes towards infinity. However, aggregate utility and thus aggregate willingness to pay increases with an increasing n and thus the potential revenue is an increasing function of the number of radio and TV stations allowed to relay the concert. The product must be based on information only so that Internet can be used for transmission of the product. The product must be modular, i.e. capable of being cut into pieces – modules – and reassembled together with other modules. The product must be easily adaptable to different consumer groups. For example, the puns and jokes of an entertainment program should be capable of translation. Language free jokes as in the old Chaplin or Mr Bean movies are ideal from this point of view. Transaction costs (other than transport costs) should be limited to allow for syndication. A radio or TV program that only contains music could easily be syndicated, even globally, as there are small language and culture barriers to be overcome in the transfer of the program from country of origin to a country of destination. Syndicating a movie is more costly. It might require dubbing and cutting to suit a specific public. Sometimes a syndicated TV program needs to become a part of some coherent programming strategy, which gives rise to to adaptation costs. Distributors must be independent of each other. If distributors can organize themselves in some cartel or resale network, advantages of syndication would drop. Either the number of paying distributors would drop or the revenue from each distributor would be constrained to be below the resale price within the cartel or resale network. With Internet distribution these resale prices could approach zero if there are inefficient copyright rules and regulations. Essentially syndication contains the following agents: Agents

Example

Creator

Author

Producer

Scriptwriter and innovation team

Syndicator

TV program syndicator

Distributors

TV stations

Consumers

TV audiences

5 Economics of Creativity

5.7.1

93

Integration by Syndication

Examples of syndicated entertainment products are Robinson, Jeopardy, the Oprah Winfrey and Jerry Springer shows, sports arrangements like Olympic Games and other global championships. Examples of syndicated arts products are films by independent filmmakers (e.g. Wim Wenders or Ingmar Bergman), classical music programs on radio and TV, novels suitable for conversion into film and photographic art. With the development of the size and quality of internet, syndication advantages will determine production–distribution system for entertainment and arts.

5.7.2

Global Creative Networks or Big Is Interactive

With the growing efficiency of communication of new ideas, there is an obvious increase in the economic advantages of interaction among creators of arts, entertainment and science. Assuming the value of a creation to a creator living in region, i.e. to be dependent on the interact ion with other creators, living in regions j (=1,. . .,n), we have the following optimal interaction problem: max vðiÞ ¼ SpðIði; jÞÞQðiÞ  Scðdði; jÞÞIði; jÞ; where v(i) is the profits (or recognition) accruing to the creator of region i, p(I(i, j) is price (unit value) of interaction with creators of region j, Q(i) is the predetermined level of creative activity in region i, and c(d(i, j) is unit cost of interacting from region i with region j. The p-functions are assumed to be concave and differentiable everywhere (at least twice), while the unit cost of interaction is a given to be constant for any pair of regions. The conditions of optimal interactions are thus: dp/dI(i, j) ¼ c(d(i, j)/Q(i); for all interacting pairs of regions. The implications are the following: l l

l

Interactions would increase with increasing impact of synergies upon creativity. Interactions would increase with decreases in the transactions, transport and communication costs. Interactions would be larger for creative activities operating at a large scale.

In an earlier paper by Andersson and Persson (1993) it was shown that under an assumption of a Cobb–Douglas production function, the interactions would follow a gravity equation.

˚ .E. Andersson A

94

5.8

Conclusions

There has been a slow and steady transformation of the advanced market economies from a focus on productivity towards a focus on creativity and innovation. This refocusing means a greater importance of economic organization based on synergy and interactions than on division of labor and occupational specialization. A creative focus implies a change in the working of the labor market. Because of the great uncertainties in creative multi-stage production systems, there are great advantages of employing internationally renowned creators. These can often demand substantial ‘‘celebrity rents’’, leading to highly skewed income and wealth distributions. The large uncertainties also cause an increase in the optimal scale of production. This is further reinforced by the increasing possibilities of syndication of the created products. Syndication essentially means that the same idea can be sold to many users in separated markets after adaptation to the specific user preferences. This has been used since long in the news media and among consultants, who have developed production processes, repackaging and users adapting the creative ideas of scientists. Syndication advantages have increased by orders of magnitude with the increasing efficiency of Internet. The advantages of creative synergy will increase the tendency to interact globally among scientists and artists. Optimal global interaction conditions are deduced. They indicate that interactions should be driven to the point where the unit cost of interaction divided by the scale of operations equals the marginal increase in the value of the created idea (eventually innovated as a product).

References ˚ E, Persson Olle (1993) Networking scientists. Ann Reg Sci 27:11–21 Andersson A Bell Daniel (1973) Coming of post-industrial society: a venture in social forecasting. Harvard University Press, Cambridge, MA Boden M (1990) The creative mind. Weidenfeld/Abacus & Basic Books, London Caves RE (2000) Creative industries contracts between art and commerce. Harvard University Press, Boston Fermi L (1954/1994) Atoms in the family: my life with Enrico Fermi. University of Chicago Press, Chicago Hardy GH, Littlewood JE, Po´lya G (1934) Inequalities. Cambridge University Press, Cambridge Polya G (1945) How to solve it: a new aspect of mathematical method. Princeton University Press, Princeton Romer PM (1986) Increasing returns and long-run growth. J Polit Econ 94(5):1002–1037 Shell K (1966) Towards a theory of inventive activity and capital accumulation. Am Econ Rev 56 (2):62–58 Simonton DK (1984) Genius, creativity and leadership – historiometric inquiries. Harvard University Press, Cambridge, MA Smith Adam (1904) An inquiry into the nature and causes of the wealth of nations, 5th edn. Methuen and Co. Ltd, London (edited by Edwin Cannan)

5 Economics of Creativity

95

Smith GJW (1990) The Creative Process: A functional model based on empirical studies from early childhood up to middle age. International Universities Press, Madision, Connecticut Stewart IN, Peregoy PL (1983) Catastrophe theory modeling in psychology. Psychol Bull 94 (2):336–362 Uzawa H (1965) Optimum technical change in an aggregative model of economic growth. Int Econ Rev 6(1):18–31 Vogel HL (1998) Entertainment industry economics: a guide for financial analysis. Cambridge University Press, Cambridge Werbach K (2000) Syndication: a new model for business relationships in the Internet Era. Harv Bus Rev 78(3):84–93

Chapter 6

Simple Memes and Complex Cultural Dynamics David Batten and Roger Bradbury

Abstract Regions and their policies are built on many things, such as ideas, actions, habits, skills, inventions, songs and stories, to name a few. This paper views all of these as selfish Darwinian entities – memes – that, like genes, interact and replicate in complex ways with humans to shape our culture. Perniciously, simple memes can exploit our limited capacity to deal collectively with complex problems. Whether good or bad, a single, omnipotent meme can dominate a local region of meme-space. Most arguments in this paper originated at a workshop on ‘‘Memes as Complex Systems’’ held in Canberra from 13–17 August, 2004 and funded by CSIRO’s Centre for Complex Systems Science (see Batten et al. 2007).

6.1

Introduction

Public policy sets the framework for the conduct of human affairs. Whether grandly enshrined in law and treaty, or more humbly promulgated as municipal regulation, the intent is always the same – the civilising of interactions between and among individuals, communities, regions and nations. Irrespective of whether policy concerns basic human needs like food, shelter and sex, or the arcana of intellectual property rights arising from new ideas, actions and artefacts, it always deals with the nature of Homo sapiens – that peculiar animal imbued with culture. Can we develop a science of good public policy? Such a question has been at the centre of the public policy literature for at least three centuries. However, this literature has been heavily influenced by certain notions from economic theory, like restoring the Pareto-efficiency of the competitive mechanism or achieving Paretooptimality through planning. Static notions are clearly irrelevant when a society is

D. Batten (*) The Temaplan Group and CSIRO, Melbourne, Australia e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_6, # Springer‐Verlag Berlin Heidelberg 2009

97

98

D. Batten and R. Bradbury

far-from-equilibrium. Furthermore, the public policy literature pays scant attention to two other crucial factors: l

l

The public policy ‘‘arena’’ has a very slowly-changing dimension, founded upon the co-evolution of tangible and intangible cultural capital (Batten 1993). This slowly-changing arena, on which the faster games of Homo sapiens are played, places constraints on society’s ability to achieve ‘‘good’’ public policy.

In order to understand what makes good policy, and what makes some policies fail while others succeed, this paper suggests that we must embed it in a theory of human culture that is consistent with evolution. Dennett (2000) calls this ‘‘minimal Darwinism’’. Because culture is the playground of other evolving entities besides humans, any proper understanding of policy must be built along Darwinian lines. We need, in fact, to understand how policy depends on the complex interplay between humans, genes and memes. The contention in this paper is that a science of co-evolving ideas, habits, actions and artefacts – in fact, all the elements of human culture – could be built on Dawkins’ notion of the meme (Dawkins 1989) using the analytical tools of complex systems (Anderson et al. 1988; Epstein and Axtell 1996; Holland 1998). Genes appear to be selfish (Dawkins 1989). That is, the interests of the genes and the interests of the organisms in which they live may not always coincide. Although genes are not conscious, purposeful agents, blind natural selection makes them behave as if purposeful. This Darwinian idea of purpose or self-interest is only a metaphor, of course, ensuring that genes will grow and spread blindly through the world. Genes that encourage parents to take (sometimes fatal) risks to protect their young, for example, serve the interest of the genes, not the organism, by increasing the chance that copies of the genes survive (in the offspring) even as they decrease the chance that the parent survives. Genes treat organisms as vehicles to protect them from the vagaries of the environment and increase their chances of survival. Dawkins extended this notion of vehicle to things that organisms create in their environment (Dawkins 1982), whether beaver dams or human culture. He coined the term memes to describe how human ideas, objects and artefacts can be thought of as agents evolving separately from their human hosts, owing their existence as entities to contingent facts about brains and their interactions. Their dynamics are governed by the principles of Universal Darwinism (Dennett 1995). This is the idea that if the conditions for Darwinian evolution are met then it will, indeed, must occur. The conditions are that the entities should exhibit heredity or replication (be copied from one generation to the next), variation (an abundance of different elements, because copying is not perfect), and natural selection (some of the exhibited variation is associated with conditions of existence). Genes, based on DNA and RNA, fulfil these conditions, and so life has evolved. But other entities in the world – prions (Aguzzi and Haass 2003), the computer programs of artificial life (Langton 1995), and memes (Blackmore 1999) – also fulfil the conditions of Universal Darwinism. Thus they must evolve and they must act as if they are selfish. They will differ from each other and from genes in how they evolve, but that they evolve and have interests that appear to be selfish can hardly be in doubt.

6 Simple Memes and Complex Cultural Dynamics

99

The crucial idea about memes – what makes them uniquely Darwinian – is that if a meme can get itself copied it will. The new Oxford English Dictionary defines meme (n. Biol.): ‘‘An element of a culture that may be considered to be passed on by non-genetic means, esp. imitation’’. When we imitate someone else, something gets passed on. Copying and imitation come naturally to human beings. In fact, what makes humans different from other animals is our ability to imitate (Blackmore 1999). That which gets passed on can be passed on again, and again, and thus takes on a life of its own. And that is what makes the meme a replicator and gives it replicative power. Education and imitation are ways of transmitting ideas, habits and behaviours from one vehicle to another. It has been said that a chicken is just an egg’s way of making another egg. As Dennett suggests, perhaps a scholar is just a library’s way of making another library (Dennett 1995). A meme’s existence depends on physical embodiment in one medium or another. Like genes, memes are potentially immortal. But, also like genes, they depend on a continuous chain of physical vehicles for their existence. Books, buildings and music are relatively permanent, as are inscriptions on monuments. But unless all of these are under the protection of human conservators, they tend to dissolve over time. Manfred Eigen makes a similar point: Consider, for instance, one of Mozart’s compositions, one that is retained stably in our concert repertoire. The reason for its retention is not that the notes of this work are printed in a particular durable ink. The persistence with which a Mozart symphony reappears in our concert programmes is solely a consequence of its high selection value. In order for this to retain its effect, the work must be played again and again, the public must take note of it, and it must be continually re-evaluated in competition with other compositions. (Eigen 1992)

6.2

Cui Bono?

Lawyers often ask (in Latin), Cui Bono? Who benefits from this matter? – A question that strikes at the heart of important policy issues. In the case of evolutionary theory, by and large the fate of a body and the fate of its genes are closely linked. But when push comes to shove, the evidence suggests that what’s good for the genes determines what the future will hold. A ‘‘gene’s-eye-point-of-view’’ explains things in terms of the complex interactions between long-range, largescale ecological facts, long-term historical facts, and local, molecular-level facts (Dawkins 1982). Many people feel threatened by this gene’s-eye-point-of-view. They want to be in charge of their own destinies – or at least feel that they are. After all, none of us want our interests playing second fiddle to something else! But soon the threat may come more from non-genetic sources. Memes are outpacing biological change. Cultural evolution operates many orders of magnitude faster than genetic evolution, so it may not be long before what’s good for the memes may determine what the future will hold, partly at the expense of our genes. What better example is there of this phenomenon than the Internet – a meme itself and a lightning-fast transmitter of memes.

100

D. Batten and R. Bradbury

If memes, as well as genes, build humans and their culture to further their own interests, then even bigger questions loom. Where is human purpose or free will in this description (Hull 2000; Dennett 2003; Blackmore 2003). Modern genetics have undermined the belief that humans are born with the freedom to shape their individual destinies. Scientists recognize that genes shape our minds as well as our bodies. Evolutionary psychologists place personal qualities – like altruism and aggression – squarely in the Darwinian camp of random mutation and natural selection (Dennett 2003). If memes have a hand in shaping our minds as well, then who is really in charge – ourselves or our memes? Can humans possibly survive as the ruling vehicles in the face of such a complex mix of memetic influences operating at vastly different speeds? There are no simple answers to these questions. You may be appalled by the idea of your brain being: a sort of dung heap in which the larvae of other people’s ideas renew themselves, before sending out copies of themselves in an informational diaspora. (Dennett 1995)

It does seem to rob your mind of its importance as both author and critic. Most of us would like to think of ourselves as godlike creators of ideas, manipulating and controlling them as our whim dictates. But, even with the most masterful and creative minds, this is seldom, if ever, the reality. As Mozart observed of his own ‘‘brainchildren’’: When I feel well and in a good humour, or when I am taking a drive or walking after a good meal, or in the night when I cannot sleep, thoughts crowd into my mind as easily as you would wish. Whence and how do they come? I do not know and I have nothing to do with it. Those which please me I keep in my head and hum them; at least others have told me that I do so. (Dennett 1995)

6.3

Genes, Memes and Replicators

One challenge for understanding genes and memes is how to transfer some of the deep knowledge and understanding of genetics to the domain of memes without imposing the particulars of genetics that are the result of the way living things are built. In fact, the field of memetics has probably been held back by attempts to map memetic to genetic phenomena too precisely (Hull 2000). Perhaps a more useful framework for memetics may be found in complex systems science. It provides a rich array of theory and practice for gaining insight into the emergent properties of systems whose dynamics range from adaptive to evolutionary in the strict Darwinian sense (Holland 1998). Embryonic and cultural development can be looked upon partly as the evolution of cooperation (Axelrod 1984). Axelrod’s computer tournaments among different strategies in ceaseless games of the Prisoner’s Dilemma serve as useful metaphors for the way we can think of animals, plants, and even genes (Dawkins 1989). It is natural to ask whether his optimistic conclusions – about the emergent success of nice, forgiving, non-envious strategies like Tit for Tat – also apply in the world of

6 Simple Memes and Complex Cultural Dynamics

101

nature. Dawkins’ answer is yes, of course. So long as the shadow of the future is long, and games are the non-zero sum variety, embryonic development can be viewed as a cooperative venture – jointly run by thousands of genes together. In natural selection, genes are selected for their capacity to flourish in the environment in which they find themselves. We usually think of this environment as the outside world, that world of predators and climate. But, from the gene’s-eye-point-of-view, possibly the most important part of its environment is all the other genes that it encounters. A gene encounters other genes mostly in the cells of the successive individual bodies in which it finds itself. Each gene is selected for its capacity to cooperate successfully with the population of other genes that it is likely to meet in other bodies. Could the same be true of memes? From the meme’s-eye-point-of-view, possibly the most important part of its environment is all the other memes that it encounters. Memes encounter one another mostly in the brains of the successive individual bodies in which they find themselves. Could it be that each meme is selected for its capacity to interact successfully with the population of other memes that it is likely to meet? Then, doing well in such environments would correspond to collaborating with these other memes. Another interesting parallel between genes and memes is their informational nature. As an evolutionary unit, a long-lived gene is not any particular physical structure but the textual, archival information that is copied on down the generations. This textual replicator is widely distributed in space among individuals, and widely distributed in time over many generations. The population of genes is not just the temporary collection that happens to come together in the cells of any particular body, but the set of all genes in the population of inter-breeding individuals – the gene-pool. Just as genes propagate themselves in the gene-pool by passing archival information on from body to body via sperm or eggs, memes propagate themselves in the meme-pool by passing information on from brain to brain via an imitative process. The above gives rise to a technical question: how do memes and genes interact to create the vehicles that allow them to replicate? Perhaps it is through the emergence of autonomous structures by modularisation and hierarchical organization. Perhaps self-organization may be a key architect. These are well-described complexity phenomena in other problem domains. We shall steer around this Scylla and Charybdis for the moment, charting a more pragmatic course by examining some particulars of how public policy as a purposeful domain of ideas may be driven by memes.

6.4

Vignette Number One

Consider, first, international development aid. It is an area where public policy has failed spectacularly, and in a way that continues to confound conventional explanations. Despite transfers of vast amounts of money over many decades from the rich world to the poor, poor countries are getting poorer while the rich get richer.

102

D. Batten and R. Bradbury

Memetics allows us to propose a testable scientific explanation for this continuing policy failure. Our model suggests that aid is an expression of a biologically-based altruistic drive modulated in Western culture by two sets of memes entrenched at least since the Enlightenment: ideas of universal human values and ideas of progress and technological control. Together these encourage simplistic intervention in the complex system that is some other (non-Western) culture – where such memes are not strongly established. But these memes never need to become established in poor countries for them to continue to prosper in rich ones. All that is needed is that memes emanating from the poor countries reinforce those in the rich. Simple memes carried in pictures of starving children – regardless of their truthvalue – created in poor countries replicate well in rich countries. Simplicity serves the interests of this new cluster of memes, and encourages continued simplistic intervention, regardless of the actual effects – good or bad – of the aid. We may predict confidently that development aid will continue to be a naı¨ve and failing intervention in a complex system, and that it will continue to fail for so long as the interests of these simplistic clusters – meme complexes – are satisfied.

6.5

Vignette Number Two

The War on Terror is another case where public policy is in trouble. Historically tested policy settings failed to either anticipate or cope with this new international security emergency. The intelligence and security services of key members of ‘‘the coalition of the willing’’ (USA, UK and Australia) have each been reviewed to discover why the West was caught napping. The usual bureaucratic reasons – lack of cooperation and coordination – and the usual policy nostrums – creation of new organisational structures – have been aired. But a memetic perspective offers a different insight, as well as new directions. We suggest that fundamentalist Islamic terrorism is an emergent property of the complex adaptive system that is our strongly-interacting cultural world. It was created from the historical interplay of humans and memes. It was never entirely predictable that it would come into being, for the emergent properties of complex adaptive systems are not generally predictable, even in principle (Anderson et al. 1988). And in that sense, the intelligence services did not fail, in failing to predict it. But now that it exists, it can be understood, and that understanding can guide policy. And the key lies in understanding the balance of interests among the memes in play in the brains of terrorists. Clearly, terrorism memes have found a very successful strategy for their replication and spread, especially by associating with powerful and long-lived religious memes and by using new channels such as television and the internet to spread from brain to brain. We predict that no policy based only on stopping acts of terrorism or locking up or killing terrorists can be successful, since the terrorism memes are not affected by this. However a policy to change the selective pressure on terrorism

6 Simple Memes and Complex Cultural Dynamics

103

memes – perhaps on their linkage with religious memes – could drastically reduce the spread of these memes. The global outbreak of terrorism could collapse or, more likely, evolve to something like local gangsterism (as in the Irish situation) that can be handled by normal policing. And this could occur despite the absence of any other social or political reforms, such as democratisation or market reforms.

6.6

Vignette Number Three

Our third example concerns recreational drug use. Many such drugs are intensely pleasurable to human beings. Some of these drugs are harmful to health. But drug policy often creates social and economic problems that are, arguably, out of all proportion to that harm. These include the huge black market, the cost of policing, the size of prison populations, the criminalisation of children, and the financing of terrorism and organised crime. Existing models of the drug user – sick person, social victim, and sinner – fail to adequately describe the situation or produce better policy. We suggest that drug policy is stymied because simple memes – such as ‘‘drugs are bad’’ – are often more successful than complex ones, and successful memes attract other memes – ideas of criminality, morality, sociality and so on – to form even more powerful complexes. The interests of this meme complex – or memeplex (Aguzzi and Haass 2003) – are served by the actions of all human players in our society – licit and illicit, users and non-users, victims and victors – and it has become so powerful that it excludes all other competing memes. But complexity theory allows us to imagine a richer universe of policy possibilities built on biological predispositions and pharmacological effects, and to see the current policy as a deep basin of attraction in a policy landscape. We know, for example, that Amazonian Indians using the drug ayahuasca have a core meme: drug use is dangerous but can lead to spiritual experiences. This allows the memes associated with ritual, social control, art and creativity to form memeplexes not found in other societies. Exploration of the policy landscape in the vicinity of such memes could provide the opportunity for the evolution of memes that are both successful as memes and beneficial to their human hosts.

6.7

Vignette Number Four

Our final example focuses on American economic policy, where fiscal irresponsibility seems to be politically seductive. Once again, the world’s largest debtor is touting huge tax cuts as stimulants to economic growth and massive increases to defence spending (Davies 2004). In the early 1980s, the Reagan administration did the same. Its reasoning was simple: middle-class Americans are overregulated and overtaxed, groaning under the weight of Big Government. Reagan illustrated his

104

D. Batten and R. Bradbury

point with striking images such as welfare queens driving Cadillacs and huge rooms full of bureaucrats each taking care of a single Indian (Rivlin 2004). Yet all these images were fantasies and Reagan’s stance highly questionable. Middle-class Americans pay lower taxes than residents of other advanced nations and most of those taxes go to pay for their own social programs. Why did Reagan build his political stance on a compelling mythology so far from reality? Was he simply too frightened to do the necessary? Raising taxes and cutting spending are extremely painful. A political leader needs to be convinced that the pain is worth it. A memetic perspective, however, offers a different explanation. First, we must ask where ideas about economics come from? Most come from economists. But not all economists are alike. Krugman (1994) notes that the genus includes two radically different species: professors and policy entrepreneurs. A professor writes for other professors. Lurking behind his words – no matter how simple – are concepts too complicated for a broad audience to understand. On the other hand, a policy entrepreneur writes and speaks in simple terms, largely for that broader audience. In the late 1970s, a powerful group of policy entrepreneurs – the ‘‘supply-siders’’ – came upon the political scene. Mostly journalists and political staffers, they shunned demand-side policies and proclaimed that sharp tax cuts will produce a huge surge in economic growth. Reagan loved this meme and based his campaign on these supply-side cranks. Supply-side memes flourished and spread. ‘‘Voodoo economics’’ ruled the American roost for the next twelve years, despite consistently failing to live up to its various promises. Today, other misguided policy entrepreneurs prevail. Some understand even less about the economy than supply-siders. Is there a memetic version of Gresham’s Law at work, in which bad ideas tend to drive out good ones? Whether they are good or bad, simple memes propagate more effectively than complicated ones. They can be copied faithfully by politicians and populations alike. One attractively packaged meme can become omnipotent, dominating any local region of political meme-space.

6.8

Concluding Remarks

Our four vignettes have several features in common. First, each shows that memes can exploit our limited capacity to deal collectively with complex problems. They undermine our efforts to grapple with the complexity of the situation. Simple memes propagate better for the mechanical reason that something simple can be copied with greater fidelity than something complicated. And our own complexitybased work suggests that there is room for only one powerful meme in any local region of meme-space – and such memes will usually be simple. Thus Universal Darwinism presents a special challenge to public policy, since policies are built from ideas in action. We need to imagine a more complex and organic process where humans and their cultures have limited agency. We need to

6 Simple Memes and Complex Cultural Dynamics

105

understand how policy is constructed by and for often short-lived, relatively simple memes, each with their own selfish interests, within a complex framework of culture built by relatively longer-lived genes and memes, again each with their own selfish interests. In short, we need to bring memes and complex systems into the arena of public policy – whether regional, national or international. Then, perhaps, our limited abilities to address complex problems collectively might improve.

References Aguzzi A, Haass C (2003) Science 302:814–818 Anderson PW, Arrow KJ, Pines D (eds) (1988) The economy as an evolving complex system. Addison Wesley, New York Axelrod R (1984) The evolution of cooperation. Basic Books, New York Batten D (1993) Pap Reg Sci 72:103–112 Blackmore S (1999) The meme machine. Oxford University Press, Oxford Blackmore S (2003) Consciousness: an introduction. Hodder and Stoughton, London Davies P (2004) Foreign Policy 144:36–38 Dawkins R (1982) The extended phenotype: the gene as the unit of selection. Freeman, Oxford Dawkins R (1989) The selfish gene. Oxford University Press, Oxford Revised edition with additional material Dennett DC (1995) Darwin’s dangerous idea: evolution and the meanings of life. Simon and Schuster, New York Dennett DC (2000) In: Aunger R (ed) Darwinizing culture: the status of memetics as a science. Oxford University Press, Oxford, pp vii–ix Dennett DC (2003) Freedom evolves. Viking Books, New York Eigen M (1992) Steps towards life. Oxford University Press, Oxford Epstein JM, Axtell RL (1996) Growing artificial societies: social science from the bottom up. Brookings Institution, Washington Holland JH (1998) Emergence: from chaos to order. Oxford University Press, Oxford Hull DL (2000) In: Aunger R (ed) Darwinizing culture: the status of memetics as a science. Oxford University Press, Oxford, pp 43–67 Krugman P (1994) Peddling prosperity: economic sense and nonsense in the age of diminished expectations. Norton, New York Langton G (1995) Artificial life: an overview. MIT, Cambridge, MA Rivlin M (2004) Foreign Policy 144:45–46

Chapter 7

The Fashioning of Dynamic Competitive Advantage of Entrepreneurial Cities: Role of Social and Political Entrepreneurship Lata Chatterjee and T. R. Lakshmanan

7.1

Introduction and Overview

There has been a major change, over the last three decades, in the functions, policy mechanisms, and the spatial forms of many urban regions in the highly industrialized countries in North America and Europe. These transformations reflect these cities’ roles as key actors and sites of change in the contemporaneous process of globalization, and the constituent economic, social and spatial restructuring. The term ‘‘Entrepreneurial City’’ pertains to this emerging urban entity. Lakshmanan and Chatterjee (2003, 2004, 2006; Chatterjee and Lakshmanan 2005a, b) have argued that a variety of change processes have converged in recent years to create a new global environment in which three types of change agents have collaborated to effectuate a major economic and spatial evolution in the form of a global production system and the rise of the entrepreneurial city (Fig. 7.1). Such change processes comprise of three types: (a) multiplicity of knowledge-rich material (transportation, communications and production) technologies and infrastructures which have made economically feasible production systems spanning the globe; (b) the advent of neoliberal ideologies which have spawned many nonmaterial (institutional and organizational) technologies and infrastructure which have dropped institutional barriers to and promoted freer cross-border flows of goods, services, finance and knowledge; and (c) secular economic changes such as the rise of quality competition and demand for variety, and the weakening of earlier macroeconomic management apparatus (e.g., Keynesian). These change processes collectively facilitate a global ‘‘space of flows’’ of goods, services, capital, knowledge and technology, and enable a globally distributed production system. In effect, these three classes of change forces create a new context or stage or arena for action by the economic, political, and social actors of the emerging global system. L. Chatterjee (*) Boston University e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_7, # Springer‐Verlag Berlin Heidelberg 2009

107

108

L. Chatterjee and T.R. Lakshmanan Change Factors

Change Agents

Material Technologies (Knowledge-rich Transport Communications & Production Technologies)

Economic and Spatial Evolution

A. Global Network Corporations, Dynamic Small and Medium size (SME) Enterprises Non-Material Technologies & Infrastructures(Neo-liberal Ideologies, Open Trade Regimes, Logistical and Financial Innovations, Entrepreneurship as a pervasive model)

Weakening of the Earlier “Economic Regime” (Rise of customized production and quality competition & demand for variety; the weakening of the National Keynesian apparatus)

Outcomes

B. Public Sector Entrepreneurial Agents C. Social Sector Entrepreneurial Agents

A. Global Transformation Global organization of production systems (economic volatility) B. Rise of Dynamic “Learning Regions”

Rise of the Entrepreneurial City (Emphasis on Wealth Creation) A. The production of Urban Dynamic Competitiveness B. Innovations in Governance in Policies in Institutions C. De-emphasis of Redistributive Functions

Fig. 7.1 Convergent forces leading to the rise of the entrepreneurial city

In this new global context or arena, the relevant socioeconomic actors come from three interdependent and complementary sectors – market, government and social sectors – and have become major agents of change, shaping the structure, geography and composition of the world economy and its component urban regions. In the market sector, the global network corporations utilize their economies of scale in knowledge, and the economies of scope of their corporate (finance, marketing, etc.) networks, and take advantage of spatial differences in the costs of labor and of other factors by creating and maintaining production units around the world; and small and medium enterprises (SMEs) create and commercialize new knowledge in dynamic urban regions. The second class of change agents, in the public sector, embraces national, regional, and urban levels of governance of the global economy and the constituent urban regions. These agents’ roles, as elaborated later, vary with the level. At the national level, these actors: l

l l

1

Have a market enabling role – enabling efficient factor, asset, and product markets and upholding legal and commercial institutions underpinning these markets. Promote national interests at international negotiations (WTO, IMF, etc.). Have a responsibility to minimize the social disruptions of structural change. At the regional and urban levels, the public actors design and implement economic policies in support of the economic goals of the regional and urban constituents they represent.1

At the supranational level, a variety of entities have arisen to deal with some types of market or extra-market failures in cross-border activities of global network corporations. These are of several types: formally constituted supra-national bodies like the European Union, resource providers (World Bank. IMF), rule or standard setters (World Trade Organization), and a focal point for information assembly, research, exchange of views, etc.

7 The Fashioning of Dynamic Competitive Advantage of Entrepreneurial Cities

109

The third class of change agent is the actor in the social sector, comprising of nonprofit organizations and nongovernmental organizations, and has been a part of governance of democratic capitalist societies for over a century. These social sector organizations (engaging in economic, social and environmental issues) have several functions: policy activism identifying unmet goals and demanding new policies, supplementing and facilitating markets for targeted services, promotion of increased transparency in governance, and engagement in socioeconomic coordination jointly with the public and the private sector agents. The joint outcome of the entrepreneurial actions over the last quarter century of these economic, political, and social agents has been a major worldwide economic and spatial restructuring – marked by a globally organized production system (and value chain), competition between local clusters and dynamic urban regions and the rise of the Entrepreneurial City, engaged in intense interurban global competition (Fig. 7.1). As globalization creates a new geography of competitive advantage and restructured webs of power, an urban region’s successful participation in this global division of labor depends upon its ability to promote economic and extra-economic capacity in that region, that supports sustainable endogenous urban development. Such capacity derives from the acquisition and maintenance by the city of dynamic competitive advantage or structural competitiveness. As elaborated below, such cities which socially create dynamic competitive advantage develop configurations of institutions and practices, which provide a favorable environment for firms or networks of firms to compete entrepreneurially in the global economy (Lundwall and Johnson 1994; Jessop 1997; Lakshmanan and Chatterjee 2003). These cities introduce new economic, social, and political innovations to enhance productivity and other attributes governing the dynamic competitive advantage of local and mobile capital. The appellation ‘‘entrepreneurial’’ is appropriate for this type of city, which exhibits the various traits associated with the entrepreneurship – discovery, risk taking, and many types of innovations – artifacts, processes, organizations, etc. (Hebert and Link 1982; Kirzner 1973; Knight 1921; Schumpeter 1928, 1939, 1961). The objective of this paper is to describe briefly the process by which entrepreneurial cities fashion or socially create their dynamic competitive advantage, which underlies their ability to function and thrive in the new global economy. We argue that three autonomous and interdependent urban sectors – economic, political and social – are involved in the joint production and maintenance of urban dynamic competitive advantage. The common attributes among entrepreneurs in all three sectors are their ability to anticipate change, assess social needs based on perceptions and identify innovative solutions to urban opportunities and problems, undertake risks and take proactive actions to intuited opportunities. Entrepreneurial actors from the three sectors have, however, different motivations, attributes, activities, resources, instruments, and networks, but they coordinate their actions to produce a mutually beneficial joint outcome (Chatterjee and Lakshmanan 2005a, b). To address the complex problems of urban adjustment and reinvention in a volatile global environment, entrepreneurial actors from these three sectors are jointly fashioning new economic and political capacities which lead to the city’s structural competitiveness and support of its wealth creation role. These capacities

110

L. Chatterjee and T.R. Lakshmanan

include: new economic roles and functions; an economic ambience comprising of an enterprise culture, permanent innovation, and labor market flexibility aided by human capital investments; a mix of strategic vision and performance-oriented activities; and institutional innovations. At the same time, there is a shift from the mode of (hierarchical) government to governance in these interrelationships between the urban public, social, and public sector actors. Governance pertains to any mode of coordination of interdependent activities. In the urban case this leads to an interorganizational coordination of mat‘erially interdependent but formally autonomous organizations, each of which controls important resources in order to secure a joint beneficial outcome. In short, urban Economic/Political/Social entrepreneurs are fashioning a multisectoral model of urban governance with self-organization of intersectoral coordination and exploitation of complementarities among the three sectors to create a competitive urban milieux. In view of space limitations, this paper focuses on the motivations, activities and networks of just two – economic and social – urban entrepreneurial agents as they collaborate and fashion dynamic urban competitiveness. This paper is part of a trilogy. In two earlier papers we focused on the complementarities between public and economic entrepreneurs (Lakshmanan and Chatterjee 2004) and social and economic entrepreneurs (Chatterjee and Lakshmanan 2005a) in entrepreneurial cities. The central argument in this, and the related papers, is that entrepreneurs in each of the three sectors – private, public and civil – play complementary roles in urban development and redevelopment. They play complementary roles because of a set of universal attributes common to them and a set of singular attributes that reflect their varying institutional contexts. Section 7.2 of the paper surveys the motivations, attributes, and activities of urban social entrepreneurs and how they deploy their networks and other resources to contribute to and collaborate on urban development and regeneration. Section 7.3 describes the evolution of urban public sector actors from government to governance and the development of new modalities of economic coordination between economic and social entrepreneurial agents. Section 7.4 analyses the complementarities and innovations in economic coordination between urban economic and social entrepreneurs, with illustrations from the experience of Chicago, Boston, New York, Tulsa and Santa Fe. Section 7.5 concludes the paper.

7.2

Social Entrepreneurs in Urban Development and Regeneration

All economic, political and social activity occurs in a place; space and time are critical elements in the entrepreneurial decision process. Urban entrepreneurs take three interrelated but distinct decisions – what to produce, where to produce and when to produce. The difference between success and failure in the innovative efforts of entrepreneurs of what to produce depends on their correct assessment of the locational and timing decisions. History is replete with examples of entrepreneurs who failed as

7 The Fashioning of Dynamic Competitive Advantage of Entrepreneurial Cities

111

they were ahead of their time. Failure also results from the choice of inappropriate locations and successful entrepreneurs have good intuition of where to apply their innovations. The literature on locational issues in entrepreneurial decision making has focused on the behavior of market actors, and on the impacts of such activity on the regional economy (e.g., Stohr 1989; Malecki 1994; Acs et al. 2002). In this paper we extend the discussion of urban entrepreneurship to the roles of nonmarket – civil society and public sector – entrepreneurs and the manner in which all three types of actors influence each other in their entrepreneurial decisions in urban space. Urban social entrepreneurs (SEs) respond to existing urban conditions and use innovative strategies to change social realities, and the urban environmental context broadly defined, for the collective good. First, social entrepreneurs take a strategic view – visualizing and judging the potential of urban localities and aiming to bring about urban transformation. SEs alter perceptions that members of civil society, the public sector and the business community have of development potentials of urban localities. Second, they focus on improved service delivery in specific localities and of the role that improved service delivery plays in promoting development potentials of those localities. Thus, SEs recognize an urban social need and relevant innovative solutions, promote and market their ideas changing the perceptions and attitudes among public, private and social sector actors, and marshal personal and community resources through institutional innovation, risk taking, and performance-oriented implementation. Social entrepreneurs have generated, for well over a century, a variety of innovative solutions (that have improved the life chances and mobility of urban residents and fostered social change in urban areas) and have been primarily viewed as humanitarians. Such late nineteenth and early twentieth century social innovators as Jane Addams (Settlement House, Housing), Horace Mann (promotion of public education), Chamberlin (first public provision urban water, sewer, and power provision in Birmingham, UK), Olmstead (urban parks and environment), and Patrick Geddes (English New Towns) are viewed as visionaries. Given the laissez faire ambience of those times (and the recent resurgence of neo-liberal views), it is not surprising that the economic, political and societal impacts of the reforms of SEs, and their role as entrepreneurs of altered urban realities, were less emphasized. Consequently, most of the literature on SEs, then and now, discusses the actions of individual social entrepreneurs or their motivations, and their achievements (Bornstein 2004). The translation of innovative ideas of a few SEs or visionaries into socioeconomic change in urban space is, however, contingent. Not all visionaries with creative ideas are social entrepreneurs. Their ideas about new urban development orientations and supportive institutions need to be promoted and marketed in an ambience where agents of change confront defenders of the status quo.2 When we say the ‘‘time for an idea has come’’ what we really imply is that there is some

2

We can draw a parallel here between the differences between invention and innovation in the economic domain and the ideas of visionaries and implementation of their ideas in the social domain.

112

L. Chatterjee and T.R. Lakshmanan

individual or a collective which has successfully implemented a radical idea – promoting an idea, overcoming social resistance, taking risk, investing money and nonmonetary resources and other elements of entrepreneurial behavior. Thus, SEs change the behavior of other decision makers in society, by building on the ideas and research of those before them, and by their own truly original ideas. They make societal change possible through their innovative solutions and persistent strategies. Their efforts input into the perceptions and behavior of other individuals, and collectives at large, and spread through the body politic as the desirable and the doable. A major impact that SEs have are in creative institution building for service provision. The success of SEs depends on their possessing two critical, and related, attributes – namely their capacity to network and to understand the social milieu in which their innovative actions have to be embedded. Networks are ‘‘Interconnected dyadic relationships where the nodes maybe roles, individuals or organizations.’’ Networking remains crucial since entrepreneurship is a continuous process of innovative activity requiring information on new opportunities and constraints. While entrepreneurship is a continuous process, entrepreneurs are not continuously entrepreneurial. They act in the entrepreneurial role when new opportunities for intervention are identified. Networking provides information about risks, uncertainties, peer evaluation for successful venture creation and growth (see Table 7.1). It allows successful entrepreneurs to mobilize social resources and increase their stock of social capital. Networks are socially embedded relationships and network ties can be of three types – information networks, exchange networks and networks of influence. All three types of networks are common to entrepreneurs in general, though the relative weighting of these network ties varies by the type of entrepreneurial activity. Information networks are critical for SEs since they lack monetary resources (relative to business and public entrepreneurs) to implement their ideas. They need to convince a larger citizenry of the benefit and feasibility of their innovative solutions to alter existing urban realities. Their networks primarily work through informal channels of mentoring and social contact and their ability to effect change is based on people power rather than on monetary power even though access to resources are facilitative. One of the many constraints faced by SEs is the lack of funding for innovative projects. Networks of influence and exchange with public and business entrepreneurs are also critical for the success of SEs and their networks extend to entrepreneurs of the other two sectors. These networks are addressed in Sect. 7.4, while discussing complementarities between SEs and Political Entrepreneurs. SEs focus on institutional development, primarily through establishing innovative start ups or radical modification of existing not-for-profit institutions. The activities of SE are implemented through nongovernmental organizations that start small and are community based. However, these start ups have the potential to grow in their original location and, through processes of diffusion, spread to other cities in the nation and internationally. Thus there are both innovative and imitative entrepreneurs in urban transformation through SEs. The success of innovative SEs

7 The Fashioning of Dynamic Competitive Advantage of Entrepreneurial Cities

113

Table 7.1 Urban, political and social entrepreneurs: attributes, motivations, composition and networks Political entrepreneurs (PE) Social entrepreneurs (SE) Attributes and Seek political payoffs. Risk takers Aim is social value creation to motivations with strategic vision, detecting improve quality of life; potentials in localities; flexible combine innovations, and patient in order to change resourcefulness and attitudes of different types of opportunity to transform people; PEs can influence attributes of urban space and to generative allocations of urban improve social efficiency and entrepreneurial resources; (NY equality; focus on COMPSTAT), PEs can also accountability to urban cause rent seeking activities constituencies Activities Change reward structures (e.g., Problem identification, rules for entrepreneurial consciousness raising; large activities); leverage social, input of vision, determination, public funds from private and community support but limited civil society sectors; allocate money; demanders of new revenues for innovative policies; focus on risk reduction solutions; knowledge transfers; and on (change of urban various partnerships with SEs location values); providers of and EEs; co-production of targeted services; consensus urban development with building; communicate to and economic and social sectors. gain support from clients Fosters opportunities and removes barriers for SEs and EEs Composition Political and elected leaders, Not-for-profit, nonprofit, private administrators, special voluntary sector commissions, etc. Networks Node in the flow of knowledge Networks and connectivity within, linking SEs and EEs; node for and between communities resource transfers through Networks based on ‘‘trust’’ with grants, loans, loan guarantees, PEs and EEs; leverages fostering SE and EE networks, community and market power and their cooperation for locality improvement

is interlinked through knowledge networks with imitative entrepreneurs often brokered through the public sector or other resource rich SEs such as private foundations. Institution building is the major instrument in their efforts to bring about change in perceptions of what and how of social innovations. Endogenous growth theory posits the perpetuation of positive trends through dynamic knowledge accumulation. For urban welfare, this implies endogenous growth in a neighborhood will be sustained if the community continues to innovate and to make it more attractive for propulsive firms, diverse service providers and optimistic people to locate there, either through immigration or retention of local residents. SEs perform a catalytic role in propelling growth through stemming a downward spiral resulting from processes of deskilling, widening social pathologies, loss of confidence in the area, and disinvestments by public and private

114

L. Chatterjee and T.R. Lakshmanan

sectors – by propelling quality of life improvements, they are instrumental in causing an upward spiral through positive feedbacks. Successful urban transformation occurs only when there are economies of scope in which a variety of social entrepreneurs focus on complementary areas of activities such as housing, employment generation, crime reduction, environmental improvement through parks and recreation facilities, establishment of health clinics, drug rehab centers and so on in a specific location. While entrepreneurial communities do not simultaneously start all these innovative activities, activities confined to one or two types of service modifications will not be able to bring about urban regeneration in a neighborhood or city. Communities need to be viewed through an evolutionary prism. The combination of economies of scale and scope allow communities to move into a self sustaining path of a virtuous cycle over time as noted earlier. The more rapid the spread of complementary innovative activities, the more dynamic the neighborhood becomes.3

7.2.1

Public Entrepreneurs’ Roles in Urban Development/Regeneration

The recognition of the failure of the invisible hand of the market in the urban economy brought into focus the role of the public sector in urban public goods provision. Concepts of market failure, negative externalities and the theory of public goods provided a theoretical grounding for the involvement of the public sector in urban service provision. However, the involvement of the public sector in urban service provision (infrastructure, health, police, indigent shelter and like activities) began in late nineteenth century – under pressure from the social entrepreneurial agents such as Jane Addams, Joseph Chamberlain, and Patrick Geddes, whose interests in urban development and regeneration predated those of the public sector actors. The acceptance of the distributive role of service delivery, where the market failed to provide services particularly to indigent peoples and poor neighborhoods, fostered the professional development of urban planning and management. Thus the public sector became involved in service delivery as producers of distributive and redistributive services – e.g., housing, education, etc. (Lakshmanan and Chatterjee 2003). However, not all agents in the public sector are entrepreneurial. Most are engaged in routine activities. Public entrepreneurs (PE) are creative individuals who find innovative solutions to problems. They possess a special set of attributes discussed below. 3

When social sector agents make mistakes in judgment about resources available from governmental and nongovernmental sources, overestimate the forces of change, underestimate local urban dynamics and so on, they fail to stimulate change through new institutional development. Institutions die in their nascent stage and their efforts are lost or are adopted and modified by SE at a later time.

7 The Fashioning of Dynamic Competitive Advantage of Entrepreneurial Cities

115

Schumpeter (1984) claimed self interested motivated actors in both the public and the market spheres, were thus analogous. Just as market agents are primarily motivated by profit making considerations when producing specific goods, public officials in favoring projects are primarily motivated by the desire to get elected, for enjoying political power and prestige and the other perks of office. Business entrepreneurs create products that benefit society in the long run but these are instrumental and subservient to their profit making logic. Similar to the firms’ production of socially useful goods, the public sectors’ production of socially useful services are also instrumental means for attaining self-interested goals. Society benefits from innovations in both these sectors. Downs in the Economic Theory of Democracy (1997) further elaborates on these ideas. We can argue that both these types of entrepreneurs engage in strategic action in the sense described by Habermas (1984) in which actions are taken for self interested goals. Increase in societal resources and social welfare is a byproduct of the ambitious, self-interested actions of innovative individuals in both sectors. While this is a rather sweeping generalization and there are examples of socially well meaning entrepreneurs in both sectors who have taken innovative decisions guided primarily by their social conscience, rather than self interest, they are by and large exceptions to prove the rule.4 William Baumol (1990) pointed out that the productive contribution of society’s entrepreneurial activities varies much because of differences in the society’s incentive structures and the resultant allocation of social resources between productive activities (such as innovation) and largely unproductive activities (such as rent seeking behavior or organized crime). He argued that it was not the total pool of available entrepreneurial talent that differentiated between entrepreneurial societies in space and time. The allocation of entrepreneurial resources was heavily influenced by the relative payoffs society offers to different activities arguing ‘‘what is required in society is the adjustment of rules of the game to induce a more felicitous allocation of entrepreneurial resources’’ (William Baumol 1990, p.894).5 We extend Baumol’s central thesis to: 1. The role of public entrepreneurs in urban areas 2. Introduce spatial variables explicitly in entrepreneurial behavior Public policies can influence the allocation of entrepreneurship more effectively in urban areas than it can influence its supply. There are examples of productive and unproductive entrepreneurship in cities. For example, the selective use of arson for land clearance in urban sites and the resulting acquisition of capital from insurance claims may be entrepreneurial in nature but private gains are made at the cost of social loss. Gains from drug trafficking, pimping and other forms of social pathologies

4

Downs’ arguments in the Economic Theory of Democracy provide an insightful discussion of this issue.

116

L. Chatterjee and T.R. Lakshmanan

can be entrepreneurial but it is of the unproductive type. It is not uncommon to find corrupt officials and police aiding such non-beneficial entrepreneurial activities in the market with a partnership between creative corrupt officials in the public sector and illegal entrepreneurs in the private sector. In depressed, underserved urban areas entrepreneurial talents can be devoted to unproductive uses due to economic and socio-political rewards for such activities. Nevertheless, there are numerous creative individuals in local, state and federal governments who desire positive urban change for efficiency and equity reasons. PEs are able to design innovative approaches to deal with unproductive allocation of resources in urban space. With regulatory and financial powers at their command, they have the ability to change the allocation of entrepreneurial resources in urban space from destructive to constructive purposes through innovative policies. They can stop parasitical and destructive actions through changes in and the creative enforcement of rules and regulations, thereby, altering the reward structures. They can implement their creative and innovative ideas using fiscal resources, channeling general revenues, targeting of seed monies, leveraging limited city funds to change the reward structures in favor of productive activities.

7.3

Complementarities Between Urban Social and Political Entrepreneurs

Table 7.1 compares and contrasts the characteristics, objectives, composition, activities, and the networks used by contemporary urban social and political entrepreneurial agents. Urban social entrepreneurs are motivated by several objectives: social value creation; the improvement of urban quality of life; the transformation of the attributes of urban space in order to enhance urban efficiency and equality; and enhancement of the accountability of different urban constituencies. By contrast, political entrepreneurs (PEs) in urban areas seek political payoffs and direct their actions accordingly. They are risk takers with strategic views of the urban area and development potentials of localities; they are flexible and patient in order to market their ideas about these potentials and change relevant attitudes of other social, political, and economic actors. PEs can either cause rent-seeking activities or influence generative allocations of urban entrepreneurial resources [in the William Baumol (1990) sense], as noted in the NY COMPSTAT case described below. In terms of activities, urban SEs identify urban problems and opportunities, engage in consciousness raising, and demand new urban policies; they offer large input of vision, determination, and community support, but limited money; SEs engage in consensus building, communicating and gaining support from clients; SEs provide services targeted to residents and localities and reduce risks in specific localities (thereby enhancing development potential in those localities). Political entrepreneurs (PEs) change reward structures (e.g., rules for entrepreneurial activities), and foster

7 The Fashioning of Dynamic Competitive Advantage of Entrepreneurial Cities

117

opportunities and removes barriers for SEs and EEs; they leverage social and public funds from public, private and civil society sectors; allocate revenues for innovative solutions; engage in knowledge transfers, and various partnerships with SEs and EEs; PEs engage in co production of urban development with economic and social sector entrepreneurial actors. Both SEs and PEs build, maintain, and use respective networks in their activities to advance their objectives. The networks urban SEs use in interacting with PEs is based on ‘‘trust.’’ They are often deployed to leverage community power for locality improvement. Political entrepreneurial actors serve as nodes in the flow of knowledge linking PEs and SEs, and for resource transfers through grants, loans and loan guarantees, and for fostering SE and EE networks and their cooperation. We argue that entrepreneurial activities of both the public and civil society sectors are complementary even though the social and public entrepreneurs have different bottom lines. It is the convergence of these dual bottom lines (in actuality three bottom lines if we include the economic entrepreneurs) in a specific urban location that confers on that a location dynamic competitiveness and helps promote endogenous growth in the entrepreneurial city. While all types of entrepreneurs act in environments of uncertainty and ambiguity, PEs are interested in reducing development uncertainties through place specific investments in infrastructure, housing, transport and the like. PEs often partner with SEs in provision of services, since SEs are commonly early movers in sectors such as initiating health clinics, crime watch programs, drug rehabilitation, and housing for the homeless. Creating transitional housing for the homeless, and drug rehab programs have place-based benefits which are equity motivated, but also generate development payoffs and efficiency gains. By reducing uncertainty and risk in certain urban localities, the activities of SEs can attract economic entrepreneurs to locate physically their ventures in (prior) risky areas (through reduction of transaction costs). Dynamic efficiencies can be realized when creative ways of achieving equity are realized through entrepreneurial skills. As areas begin to change their attributes, alert economic entrepreneurs, with ability to use existing information on activities of SEs and PEs, perform arbitrage – thus linking all three types of entrepreneurs in urban regeneration. Such innovations usually involve the creation of new institutions or radical transformation of existing ones. For example New York city developed a crime reduction management system called COMPSTAT, which helped to reduce crime dramatically, combining innovative computer technology developed by the private sector to identify and target high crime areas with a new style of police management. The city of Chicago is an entrepreneurial city rich in all three classes of entrepreneurs. For example, the mayor’s office in cooperation with city departments of Environment, Planning, Housing, and Transportation designed and implemented The Chicago Brownfield Initiative in 1993 with an investment of $2 million (using general obligation bonds). The City then leveraged these funds to acquire $74 million of Section 108 funds from the Federal governments HUD program. Section 108 is a loan guarantee program of CDBG (Community Development Block Grants). With increasing shift from the goods producing city to a service

118

L. Chatterjee and T.R. Lakshmanan

oriented city, (common in advanced industrial countries), large areas of redundant, often abandoned industrial sites occur in close proximity to waterfronts, railroad depots and the like. These sites are often environmentally polluted and unproductive sites. The Chicago Brownfield Initiative recycled neglected properties and transformed blighted land with new construction of industrial parks, green spaces, affordable housing. It created 3,000 jobs and increased the tax base by more than $1 million annually. From the initial five brownfield sites the program expanded to 30 sites through leveraging USEPA money to acquire private investments for development. Chicago’s Rooftop Garden is another example of (development enhancing) creative environmental conservation, that saves $3,600 annually in energy costs of one building, improves air quality, absorbs rain water, decreases stormwater runoff and provides an urban park in a congested site. Initially, the program started on the rooftop of City Hall, an 11 storey building with a garden on 29,300 sq.feet containing 20,000 plants of 150 varieties. As of June 2004, there are more than 80 municipal buildings and countless private roof top gardens in Chicago. Since there is a 50 F difference between a garden roof and a black roof, a roof garden reduces the urban heat island effect. Roof top gardens require innovative construction technologies and the use of new materials to save roofs from water damage and to bear the load of soil and moisture. Thus the public sector had an important role in transferring knowledge to the new adoptees of the roof top garden idea, which required a combination of public, social, and private entrepreneurship. There are other examples of urban public entrepreneurship in the USA that has been stimulated by the Ford Foundation (a nonprofit Social Entrepreneur) through its Innovation in American Government program. Innovation awards honor public sector innovations from the Federal, State and local sectors. Five such innovative programs of intersector partnership and co production of services, area improvement and locality competitiveness for new economic growth in urban areas are illustrated: l

l

l

l

Santa Fe Affordable Housing Roundtable builds affordable housing for low income households through a public–private partnership of local governments, nonprofit agencies, builders and lenders. San Diego’s Single Room Occupancy Residential Hotel program promoted the development of low cost, permanent, private rental units through preservation, rehabilitation and construction incentives. The Quincy (Boston) Model Domestic Abuse program helps battered women through a two pronged effort that sanctions abusers and provides a wide variety of services to the abused. Seattle instituted a Community Voice Mail program where clients have personal telephone numbers and access codes to receive messages. Using private or public phones they can connect to potential employers, landlords and social service providers.

Tulsa, Oklahoma received an Innovation award for their interdisciplinary Sexual Assault Nurse Examiners Program that combined police, health and legal agencies

7 The Fashioning of Dynamic Competitive Advantage of Entrepreneurial Cities

119

with forensic nurses to provide treatment and follow up care. Twenty communities adopted/adapted this model within two years and at least three states had developed legislation or administrative procedures based on their program. The Ford Foundation conducted a survey of the 95 awards to urban innovators and found that over 85% of these innovations had been replicated and hundreds of communities had adapted these innovative programs to their own communities. Many of the Innovations awards provided the basis of legislation – for example the Quincy program was a model for the Federal Violence Against Women Act (1993), Seattle’s program was replicated in 15 cities and formed the basis of the Community Technology Institute which helps launch and support similar new initiatives. Over 75 cities in eight states had replicated the San Diego program. These illustrations of complementarities, partnership and joint production of urban regeneration and development among entrepreneurial actors in the urban political and social sectors highlight the importance of knowledge transmission, of network activity, and partnership between political entrepreneurs (city hall departments and city hall) and a variety of social entrepreneurs. Key elements of such partnerships are networking, information flows and mentoring potential adoptees. The consequence of such activities is the creation of urban competitiveness understood in Schumpeterian terms as possessing a ‘‘structural’’ or ‘‘systemic’’ character.

7.4

Concluding Comments

This paper argues that an entrepreneurial city functions and thrives in the global economy by creating and maintaining a favorable environment and bundles of institutions and practices, which enables the different enterprises in that city to compete entrepreneurially in the global economy. In this environment, entrepreneurial actors from the market, public, and social sectors – materially interdependent but formally autonomous organizations each of which controls important resources – coordinate their actions to secure a joint beneficial outcome in that city. They jointly fashion new economic and political capacities which support the city’s structural competitiveness. These capacities include: new economic roles and functions, an economic ambience comprising of an enterprise culture, permanent innovation, a mix of strategic vision and performance-oriented activities, and institutional innovations. Entrepreneurial actors from the three sectors, while each having different motivations, activities, resources, and networks, exploit their complementarities to capture the ‘‘added value’’ of interorganizational coordination and the creation of a competitive urban milieux (Chatterjee and Lakshmanan 2005a, b). This paper has focused on the role of two classes of urban entrepreneurial agents – political and social – in the above process of adaptation and reinvention of the city in order to be dynamically competitive in the evolving global economy. It has characterized the objectives, activities, and networks of these two sectors and how they engage in positive coordination in the larger economy of interfirm

120

L. Chatterjee and T.R. Lakshmanan

networks, multilateral partnerships between the three (public, social, and market) sectoral organizations, and joint production of urban regeneration and development. These ideas are illustrated from the recent urban experience of cities such as Chicago, Boston, New York, Tulsa, and Santa Fe.

References Acs ZJ, de Groot HLF, Nijkamp P (eds) (2002) The emergence of the knowledge economy. Springer, New York William Baumol J (1990) Entrepreneurship: productive, unproductive, and destructive. J Polit Econ 98:893–921 Bornstein D (2004) How to change the world: social entrepreneurs and the power of new ideas. Oxford University Press, New York Chatterjee L, Lakshmanan TR (2005a) The dual bottom line: complementarities between urban social and economic entrepreneurs. Paper presented at the Tinbergen Conference, George Mason University, July 10–11 Chatterjee L, Lakshmanan TR (2005b) Urban social and political entrepreneurship: attributes and complementarities. Paper presented at the Special Workshop at Jo¨nko¨ping International Business School, Jo¨nko¨ping, Sweden, June 16–18 Downs A (1997) Economic theory of democracy. Harper and Row, New York Habermas J (1984) Reason and rationalization of society. Theory of Communicative Action, vol1. Beacon, Boston (English translation by Thomas McCarthy) Hebert RF, Link AN (1982) The entrepreneur: mainstream views and radical critique. Praeger, New York Jessop B (1997) The entrepreneurial city: reimaging localities, redesigning economic governance, or restucturing capital. In: Jewson N, Macgregor S (eds) Transforming cities. Routledge, London, pp 29–41 Kirzner I (1973) Competition and entrepreneurship. University of Chicago Press, Chicago Knight F (1921) Risk, uncertainty, and profit. Houghton Mifflin, New York Lakshmanan TR, Chatterjee L (2003) The entrepreneurial city and the global economy. Paper presented at the International Workshop on Modern Entrepreneurship, Regional Development and Policy, The Tinbergen Institute, Amsterdam, May 23–24 Lakshmanan TR, Chatterjee L (2004) Entrepreneurship and innovation-led regional growth: the case of the entrepreneurial urban place. Paper presented at the 51st North American Regional Science International, Seattle, November 11–13 Lakshmanan TR, Chatterjee L (2006) The entrepreneurial city in the global marketplace. Int J Entrepreneurship Innov Manage 6(3):155–172 Lundwall BA, Johnson B (1994) The learning economy. J Ind Stud 1(2):23–42 Malecki E (1994) Entrepreneurship in regional and local development. Int Reg Sci Rev 16 (1–2):119–154 Schumpeter JA (1928) The instability of capitalism. Econ J 38:361–86 Schumpeter JA (1939) Business cycles. McGraw-Hill, New York Schumpeter JA (1961) The theory of economic development. Oxford University Press, New York Schumpeter JA (1984) Capitalism, socialism and democracy. Harper Collins, New York Stohr W (1989) Local development strategies to meet local crisis. Entrepreneurship Reg Dev 1(3):293–300

Chapter 8

The Social Capital of Regional Dynamics: A Policy Perspective Hans Westlund

Parts of this paper are also published in Westlund H (2006) Social capital in the knowledge economy: theory, and empirics from the United States, Japan and Sweden. Springer, Berlin, Heidelberg, New York.

8.1

Introduction

Creating something new, improving the quality and characteristics of existing products or producing things in a more cost efficient manner are three of the ways responsible for economic growth. Of these three, it is only the last one that can be considered connected to neoclassical theory, in the form of optimum combination of the given production factors under a given technology. A change in technology, as well as the sources to the other two ways contributing to the economic growth do not occur through variations in the quantities of production factors, but by the setting up of new production functions through different types of innovations, or what Schumpeter (1934, 1950) denominated as new combinations of production factors. In this expression also lies an understanding of the heterogeneity of the concepts of labor and capital and the possibility of combining an infinite number of labor and capital in an infinite number of combinations. Thus, studying innovations and economic change requires other approaches than those of traditional mainstream economics. In the last few decades, a number of such approaches have emerged: clusters and innovation systems being the most well-known. Even if the approaches often are connected to Marshall’s (1880/1920) notion of industrial districts, their theoretical base lies outside traditional economics. Also, these alternative approaches to a large

H. Westlund ¨ stersund, Sweden National Institute for Working Life, Studentplan 1, SE-831 40 O e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_8, # Springer‐Verlag Berlin Heidelberg 2009

121

122

H. Westlund

extent lack the rigor of formal theory and can be considered as a conceptual framework in their early stages of development (Fischer and Fro¨hlich 2001). However, important contributions to the formalization of these approaches are those by Krugman (1991, 1995). Within the discipline of economics, the concept mostly connected to the new approaches is that of externalities. The concept of externalities dates back to Marshall (1880/1920) and has been considered one of the most intangible and hard-formalized in the economic literature (Scitovsky 1954). Sraffa (1926) considered externalities as the only source of increasing returns under perfect competition. Based on Scitovsky’s (1954) classification of externalities into pecuniary and technological, Johansson (2004) has made a fundamental distinction between firms’ intra-market and extra-market externalities. Intra-market externalities are mediated through the formation of prices, while extra-market externalities consist of links, agreements, networks and other arrangements of club type, and also information and knowledge spillovers.1 The two types of externalities have an impact on different activities of a firm. While, according to Johansson (2004) intra-market externalities affect the firm’s transaction costs and productivity, extra-market externalities affect the firm’s access to information and knowledge spillovers, i.e., its innovation potential. In both cases, space forms an important restriction implying agglomeration economies. Both intra- and extra-market externalities are distance-sensitive (as most contact-intense activities are) and space forms a restriction for the spatial reach of such externalities as transportation across space is connected with a cost. This paper focuses on the second type of externalities, the extra-market externalities, and how they contribute to the emergence of new combinations of production factors in a regional context. More precisely, it investigates the role of social capital for regional dynamics and what role policies can play in these processes. Social capital is here defined as social, non-formalized networks that are used by the networks’ nodes/actors to distribute norms, values, preferences and other social attributes and characteristics. An important feature of this definition is that it distinguishes between the networks and the norms, etc., that are distributed in the networks. Social capital can be seen as a type of infrastructure with nodes and links. The nodes consist of individuals and organizations, which establish links between each other. The creation of links is governed by the individuals’ and/or organizations’ norms, preferences and attitudes, which can prevent emergence of links between individuals or organizations as well. In the links, different types of information and knowledge are distributed between the nodes. From an infrastructure perspective, this distribution of information and knowledge can be compared with traffic in the transport infrastructure. The impact of social capital on society depends on both its quality and quantity. The norms, preferences and attitudes of the nodes, and thereby the kind of information and knowledge being distributed in

1

Information and knowledge spillovers are by Fujita and Thisse (2002) denominated communication externalities.

8 The Social Capital of Regional Dynamics: A Policy Perspective

123

the links, are at least as important as the number of links. ‘‘Strong’’ social capital can thus have preservative as well as progressive effects, depending on its qualitative characteristics.2 A starting point for this paper is that it is necessary to distinguish between a general, social capital on societal level and social capital specified for the needs of organizations (groups, firms, public sector bodies). In the latter case, social capital has characteristics of sunk costs, i.e., it often cannot be used for other purposes than it was formed for and that it might become useless or even detrimental when the organization changes its activities. Analogous with this, we can make a distinction between public social networks, which in principle everyone with certain skills have access to, and private networks, formally or informally controlled by certain groups. Section 8.2 analyzes the changes of innovation activity over time, from early industrialism to the global knowledge economy, how the relations between the actors of today’s innovation systems have developed and the role of social networks for innovations. Section 8.3 discusses the different kinds of networks built by the three constructers of social networks: the individual, the organizations and the (public and civic) society. Section 8.4 analyzes the role of public policy in building social capital for innovations and growth. Section 8.5 contains some concluding remarks and suggestions for future research.

8.2 8.2.1

Innovations and Social Capital New and Old Concepts

A number of concepts have been formulated to describe and analyze the proximityor link-based interaction between individual firms and other actors producing externalities. Industrial districts – the term coined already by Marshall – are normally defined as spatial agglomerations of SMEs in one or a few complementary industries (Paniccia 2002). In particular, the term has been used for agglomerations of SMEs in Italy. Cluster, a concept with a number of slightly different interpretations, has received, through Michael Porter’s book The Competitive Advantage of Nations (1990), an enormous amount of attention in both research and policy circles. Clusters are often defined as spatially delimited industrial systems regardless of the size of the enterprise (Paniccia 2002), but it should be noted that Porter (1990) has also considered clusters as being functional industrial systems without a proximity dimension (Malmberg 2002). Another ambiguity is that much of the cluster literature, based on Porter (1990) treats clusters as a purely a spatial concentration of related firms (see, e.g., Enright 1998), while Porter (1998, 2000) later explicitly includes public institutions, such as government educational institutions and support services, in the definition of clusters. The vast popularity of the 2

See Westlund (2004) for a more extended discussion.

124

H. Westlund

concept, not least in industrial policies, has resulted in ‘‘cluster’’ becoming a possible denomination of almost any agglomeration of economic activity. Even if clusters are thus sometimes regarded as consisting of firms as well as public institutions, both the cluster and the industrial district approach have their main focus on inter-firm relations. While the terms industrial districts and clusters have mainly been used for local and regional relations between firms, the concept of innovation systems was originally formulated for systems at a national level and denoted not only interfirm relations but also links between firms and government, firms and research institutions or between all three of them. It was used for the first time by Freeman (1987) in his analysis of the economic development of Japan after World War II, where government, especially the Ministry of Industry and Trade (MITI) played a crucial role. Leading scholars of this tradition (Lundwall 1992; Nelson 1993) have regarded the nation as the evident level of analysis as ‘‘. . . the policies and programs of national government, the laws of a nation, and the existence of a common language and shared culture define an inside and outside that can broadly affect how technical advance proceeds’’ (Nelson 1993, p.16). In the last decade the concept of regional innovation systems (RIS) has yielded a rapidly increasing literature (see, e.g., Cooke 1992, 2001, 2003; De la Mothe and Paquet 1998; Asheim and Gertler 2004; Doloreux and Parto 2004, etc.). The regional approach on innovation systems, according to Doloreux and Parto (2004) is a normative and descriptive approach, which is based on two main bodies. The first is the national innovation systems approach, based on evolutionary, nonequilibrium theories and in which innovation is a result of processes both internal and external to the firm. These processes are not only technical and economic but also social. Learning, through interaction, is a key concept in the innovation processes. The second body of literature is that of regional milieu, embeddedness and the role of proximity. According to its analysts the concept of regional innovation systems has increasingly become an all-embracing term for firms’ interaction with each other and other actors at regional level. A fourth concept, strongly linked to the abovementioned is that of triple helix, which: . . . is a spiral model of innovation that captures multiple reciprocal relationships at different points in the process of knowledge capitalization. . . . . . The triple helix denotes the university–industry–government relationship as one of relatively equal, yet interdependent, institutional spheres which overlap and take the role of the other. (Etzkowitz 2002, p. 2)

It is no coincidence that university is the actor named first. According to Etzcowitz, an important difference between the innovation system and triple helix approaches is that the former has its focus on the firm and views innovation as primarily occurring within the firm. In contrast, the view of the triple helix approach is that ‘‘Innovation is increasingly likely to come from outside of the individual firm or even from another institutional sphere such as the university. . .’’ (Etzkowitz 2002, p. 1). Triple helix processes are possible at regional, national as well as multinational level.

8 The Social Capital of Regional Dynamics: A Policy Perspective

125

The four approaches, very briefly summed up above, have one thing clearly in common: the focus on interaction where firms are involved. Apart from that, the approaches show differences between each other but also between different interpretations of the same approach, when it concerns, e.g., spatial level, included actors, their size and sectoral scope. The industrial district approach is the most limited as it only comprises interaction at local level between SMEs in one or a few closely related industries. The different interpretations of clusters – from pure industrial districts with only firms involved, to non-spatial, sectoral systems of innovation with several types of actors – is an illustration of the concept’s popularity, but also of the concept’s weakness as an analytical tool (Markusen 1999). Similar criticism has been raised against the regional innovation systems concept (Doloreux and Parto 2004), which, as shown, has also been considered as a still wider concept than the cluster. Finally, the triple helix approach is a more delimited normative approach which not only states that three types of actors should interact but also that their activities partly overlap. Moreover, triple helix’ prime focus is not on the firm’s knowledge input and innovation process but on the interaction as such and how it transforms the actors. Although not always explicitly expressed, the four approaches also have something else in common, namely their acknowledgment of externalities in the form of transfer of (tacit) knowledge or knowledge spillovers, emergence of new knowledge and (collective) learning as a primary outcome of the interaction. It is in these knowledge creating and transfer processes that social capital constitutes a ubiquitous but multifaceted factor. The ‘‘right’’ social capital facilitates or even spurs these spillovers, learning and innovation processes, whereas ‘‘wrong’’ social capital is like sand in a complicated machinery.

8.2.2

From the Lonely Genius to Innovation Nodes

The theories of (national and regional) innovation systems, clusters, industrial districts and triple helix have in common the focus on interaction between a number of key actors. The industrial district approach, as well as many other cluster approaches, concentrates solely on firms’ interaction, while other cluster approaches, the innovation systems approaches and the triple helix approach underline the interaction between at least two of, but often the three key actors of innovation: companies, public sector bodies and universities. However, this view of innovation, as a result of interaction of actors with different tasks and different principles of production and exchange (see below), is a relatively new standpoint. Historically, innovation activities seem to have had quite other characteristics than the complex systems of today. The history of technology and economic applications of technology is full of examples of individual inventors that came up with path-breaking prototypes and methods, which rapidly were commercialized into successful products. Even if we perhaps can find some examples of lonely great geniuses in the computer industry, there is no doubt that the individual inventor belonged to a certain economic era; an

126

H. Westlund

early industrial era that lasted until about World War I. Most of the world-leading corporations of today stem from that era – an era where a single innovator could build up a company from a prototype or a method. The interwar years can be viewed as a transition period from ‘‘individual inventor capitalism’’ to ‘‘corporate innovation capitalism’’. Figure 8.1 shows that albeit the gap between patents assigned to corporations and patents assigned to individuals in the USA increased slowly after 1900, the former increased rapidly after 1945, while the number of patents issued to individuals remained practically constant during the rest of the century. After World War II innovation activities seemed to have entered a new stage. With a larger public sector after the war and raised demand for, among others, military security and transportation infrastructure, governments of the developed world began to act as a qualified customer of private corporations. The most far-reaching example of this is probably what was denominated the space- and military industrial complex in the USA which had its counterparts in other countries. Some Swedish examples of this symbiosis between government and state-owned companies on the one hand and private companies on the other are: Vattenfall (hydroelectricity) and Asea (today ABB, generators and other electrical equipment); Televerket (former state telephone monopoly, now TeliaSonera) and Ericsson (switchboards and other telephone equipment); the state railways and Asea (engines) and the air force and SAAB (combat aircrafts). The common denominator in these so-called ‘‘development couples’’ was a state monopoly (complete or partial) that through its safe position could make long-term, costly R&D investment and act as qualified customer for the (at that time Swedish-owned) private companies (So¨rlin and To¨rnqvist 2000). Another example of intimate collaboration between government and private companies is the Japanese system after World War II; the system of collaboration for which the concept of innovation system was coined. Freeman (1987) noted important differences between the Japanese national system of innovation and

Fig. 8.1 Patenting in the United States 1900–2001 Source: Suarez-Villa 2004; U.S. Patent and Trademark Office

8 The Social Capital of Regional Dynamics: A Policy Perspective

127

industrial policies in other countries. Like Johnson (1982) and Lakshmanan (1994) he stressed the role of MITI, the Ministry of Industry and Trade, in identifying strategic future key technologies and actively promoting company R&D in theses technologies. Even if there were important differences between the American, the Swedish and the Japanese innovation systems during this period, they had in common the intimate cooperation between government and industry in certain key technologies. Government was a qualified customer with strong resources for R&D, which was performed in cooperation with private companies. Only certain, special fields of university research were involved in this cooperation. The innovation systems of late industrialism were mainly a system with two actors: government and private companies. The industrial crisis of the 1970s can also be considered a crisis for ‘‘corporate innovation capitalism’’. According to Fig.8.1, the number of patents assigned to corporations in the USA diminished between 1975 and 1990. Thereafter, a still much sharper increase in the number of corporations’ patents took place. It is possible to interpret this ‘‘patent explosion’’ as a new stage of innovation activity, connected to the theories of knowledge society (Andersson and Stro¨mquist 1988), Mode-2 society (Gibbons et al. 1994; Nowotny et al. 2001) and triple helix (Etzkowitz and Leydesdorff 1996). In spite of different perspectives and focuses, the three theories have in common a stress of the new role of knowledge and knowledge-producing organizations in society.3 Knowledge has been transformed from one of several resources in production to ‘‘the predominant part in the creation of wealth (. . .) in all manner of economic activity’’ (DTI 1998). While the main value of the typical manufacturing firm resided in its physical capital, the value of a knowledge intense firm is in its intellectual property. Whereas the manufacturing firm sells tangible products for consumption or refinement, the knowledge intense firm’s products consist of R&D products, including patents, with a potential for being commercialized and profitable. Innovations, defined as new combinations of production factors have become the core of knowledge society. However, the innovation activities of the knowledge society differ fundamentally from those of the early industrial period. Innovation activity in the knowledge society is a collective process in which people and organizations have to cooperate. This is the circumstance brought up in the ‘‘macro’’ theories of innovation systems with three actors and that of triple helix. On micro level, innovation activity in the knowledge society seems to require a permanent flow of new information and knowledge, which in practice means a flow and exchange of people in the innumerable innovation processes of everyday (see, e.g., Kobayashi and Takebayashi 2000). As been pointed out by many scholars this gives the great cities a special role as knowledge and innovation nodes. Their size

3 Here it should be noted that we in line with North (1990) make a distinction between organizations (firms, governmental organizations, universities and NGOs) and institutions (laws and regulations, formal and informal rules of the game).

128

H. Westlund

creates a diversity that makes specialized supply and demand – and new combinations of both categories – possible. This means that the knowledge economy not only creates another type of innovation than the late industrial society but also changes the spatial allocation of production factors. As the great cities become centers for the increasingly important production factor, i.e., human capital, they also emerge as stabilizing factors in the global economy where knowledge is footloose but human beings and the organizations of knowledge are more strongly rooted. To partly quote Markusen (1996), the great cities are sticky, innovation nodes in a space where information and knowledge ‘‘slip’’ around.

8.2.3

Why Care About Social Links?

From what is said above it seems as innovation over time has become an increasingly complex process. It is an exaggeration to say that innovation in the early industrial period was only a process of merging technology and capital in the form of providing the innovator with financial resources to start production. Access to capital, finding the right customer and getting the innovation accepted was also essential in the nineteenth century.4 However, in the knowledge society, innovation activities can be divided in a large number of stages from basic research, via, e.g., development, testing, licensing, marketing and sales to final use, each of them requiring a certain partner for example financing. One way to express the differences between innovation activities in the two periods is to say that they differ substantially in the number of actors involved, in the number of links between them, and in the amount of knowledge and information being distributed between the actors. The emergence of spatial clusters and regional innovation systems can be viewed as an expression of the intra- and extra-market externalities and their distance-dependency. Following Johansson (2004), it can be assumed that knowledge transfers take place through two types of processes: 1. Deliberate, formalized transaction-links, agreements, networks and other clublike arrangements between firms and between firms and other actors. 2. Unintended knowledge spillovers between firms or between firms and other actors, caused by non-formalized interactions. These kinds of interactions consist of: l

4

Vertical technical/economic interactions between firms and their suppliers and/or customers.

The problems with getting an innovation accepted can be exemplified with John Ericsson’s steam fire-engine, which was successfully demonstrated in London in the year 1829 and caused such an anxiety in ‘‘The London Fire Brigade’’, practically a guild with monopoly on fire-fighting, that it was rejected by the committee in charge (Goldkuhl 1961).

8 The Social Capital of Regional Dynamics: A Policy Perspective l

l

129

Spin-offs of new firms from existing ones and turnover and exchange on the labor market. Horizontal interaction in the form of informal exchange of information and knowledge in the (local/regional) civil society, between individuals connected to firms or other actors.

In both these two types of processes, social links and the norms and values connected to them play an important role. The reason is simply that good social relations facilitate knowledge transfers while lack of relations or bad relations do not. Regions have different prerequisites to deal with this circumstance. Diversified metropolitan regions consist of a number of co-located sectoral clusters that often do not have more in common than the use of the regional infrastructure and certain regional markets. Apart from that, each cluster has its own links, those external to the firm but internal to the cluster, and those between the clustered firms and the rest of the world. The relations of each cluster are formed in accordance with the stages of innovation, types of production, positions in the product life cycles, etc. In this way a metropolitan region can accommodate competitive clusters in both expanding and declining sectors. If small regions contain any clusters, it is with few exceptions only one cluster. Regardless of the sector of the cluster – expanding or declining – the small regions’ development is highly dependent on the quality of the cluster’s social relations. Well functioning internal and external social relations facilitate acquiring of knowledge and information about changes in demand, new methods, etc., as well as credits. Smaller regions, being dependent on one cluster are in general more vulnerable and more dependent on good relations between all relevant regional actors.5 Who are these actors that build and maintain this essential social capital? This question is dealt with in the next section.

8.3

Social Capital on Three Levels

The theories of innovation systems and triple helix concentrate on the interplay between different types of organizations. However, organizations represent only one of three levels in which social capital can be analyzed. Individuals build organizations and together those levels form a society. This section discusses the social capital built by the actors on the three levels and how these forms of social capital are based on the fundamental needs and aims of these actors.

5

In a study of determinants of economic growth in the Swedish municipalities, Eliasson et al. (2005) found that the importance of business-related social capital decreased with municipality size.

130

8.3.1

H. Westlund

Organizations and Their Social Capital

As analyses and policies for innovations are primarily focused on organizations it might be appropriate to start the discussion on this level. According to the policies based on modern innovation theories, the three types of organizations – firms, universities and government – should cooperate in order to meet the needs of the knowledge economy. Universities should provide the knowledge, government provides favorable institutions and development resources and firms provide resources and know-how for commercialization. However, a dilemma is that the three actor blocs are based on different principles of exchange, which are reflected in different rules of the game. Following Polanyi (1944) it can be argued that a firm bases its activities on a market principle where profit is a necessary ingredient. For public government, which has the power to collect individuals’ and organizations’ resources and redistribute them, the basic principle is redistribution. The third type of organization, the academy (or university), is for its part historically predominated by a third principle, viz. reciprocity – a mutual exchange of knowledge and ideas. Academy-produced knowledge is by tradition neither sold on a market nor taken from one actor and given to another, but exchanged and valued by equals (peers) without any losses. It goes without saying that organizations with such principal differences build social capital with very dissimilar networks, which connect different types of actors and are based on different norms and attitudes. The activities of the firm are executed with the aim of making profit. The firm builds technical and economic links internally and to external actors. These links are established and maintained if they are assessed to bring net revenues. The social networks of a firm are based on more compound motives. Creation and maintenance of social links that the firm makes deliberate investment in – e.g. corporate culture, personal customer relations, etc. – are in principle controlled by the same net revenue principle as economic links.6 But many social networks are unintended by-products of other interactions (Putnam 1993). Thus, many social links of the firm are by-products of its economic networks. To the extent to which human beings are involved, social links/relations develop as a consequence of the economic links. Consequently, a firm makes certain deliberate investment in social networks, but many of the social networks of a firm are by-products of technical and economic networks. Accordingly, companies’ social networks have two sources: deliberate, formal investment decisions by management on different levels, in accordance with the firm’s basic mission, and spontaneous, informal investment decisions by individuals, originally connected through the economic links, based on a volition to interact, to socialize. The volition to interact is connected to the ‘‘affinity’’ – here defined as attraction, liking or feeling of kinship – between the actors (cf. Johansson 6

However, according to modern managerial theory of the firm, managers might have personal goals that include other things than profits, i.e., managers might benefit from social capital independently of their firms’ profits.

8 The Social Capital of Regional Dynamics: A Policy Perspective

131

and Westin 1994). For a social link to be established, the nodes/actors should have something in common (e.g., some norms, values or preferences, cultural similarities, or some minimum degree of mutual trust). Moreover, economic interaction can to some extent be governed by the ease of formation of social capital between actors. Rauch (1996, 1999, 2001) has underlined the role of social capital and networks for international trade and for example shown the significant impact of common language/colonial ties on trade between countries. The second type of organization, public government, is run by political objectives, but a fundamental need for public government is to legitimize itself. For this reason it builds social links to the citizens and organizations of society, beside the necessary economic and technical networks it needs to develop in order to fulfill its objectives. As in the case of the firm, public government’s activities also create ‘‘uncontrolled’’ social networks as by-products. However, as the basic mission of public government is to redistribute the resources of society, both the intended and unintended social networks of government, and the norms and values distributed in them, fulfill other objectives than the social networks of the firm. The third type of organization is the academy. In spite of the fact that it is financed in a number of different ways, it has an international, joint identity with missions, objectives and norms. This academy-internal social capital is an important reason behind the academy’s relative independence vis-a`-vis other actors in society. It is on the other hand a potential obstacle to collaboration with organizations having other missions and social capitals. These three types of organizations build social capital deliberately and contribute to unintended, spontaneous social capital-building as well. Table 8.1 describes the different component parts of organizations’ social capital. Depending on the organization’s mission, certain norms, values and attitudes are developed, which in their turn govern the extension and allocation of the organization’s internal and external links. Analogous to the increased complexity of innovation activities over time, discussed in Sect. 8.2, it can be argued that organizations’ social capital has Table 8.1 Social capital of organizations broken down into different component parts Organization-internal The organization’s external social capital social capital Activity-related Environment-related Market-related Links/relations filled Links/relations to Links/relations to the General relations to the with attitudes, norms, suppliers, customers, local/regional anonymous mass of traditions, etc., that are clients, partners in environment, to (actual and potential) expressed in the form cooperation and organizations of the customers and clients, of internal ‘‘spirit’’; development two other types, (non- built through climate for activity-related links to) marketing, customer/ cooperation; and other organizations of client clubs, programs, methods for codifying the same type etc., and expressed in, knowledge, product e.g., trademarks development, conflict resolution, etc.

132

H. Westlund

become more and more complex. The assembly line–the archetypical symbol for manufacturing industrialism – required few social skills of its workers, not even a common language. In contrast, work in a consultancy company of today requires ability to cooperate, build networks and even certain attitudes. People without this social competence do not get access to the social capital of these companies of the knowledge economy. In the period when public government was small, it was held together by a strong social capital, expressed in the ideology of the public official, standing above the interest groups of society. The increased involvement of government in different areas of society has made its mission much more complex and consequently also the economic, technical and social networks of government and the values distributed within the social networks have become more composite. The same can be said about the academy. As long as the university employed small elite of researchers and students, it was easy to keep its identity, values and networks. With increased resources and increased demands from the resource-providers, university’s tasks have multiplied, and also its networks. Thus, the fact that the three types of organizations discussed are based on different missions result in different forms of social capital. These forms are an outcome of both intended and unintended investment. Over time, along with the evolvement of a knowledge society from an industrial society, these forms of social capital of organizations have become more complex. Without considering the different missions and the differences in social capital of the three types of organizations, modern innovation policies prescribe that they should interact and create innovations. The problem is described in Table 8.2. The traditional activities of the types of organizations are market with an O. The consequence of innovation policies is that actors of the three organizational types partly should expand their activities to the fields traditionally upheld by the other types of actors. A successful fulfillment of these expectations demands new strategies for combining the organization’s core activity, O, with the new activities (o) that with few exceptions has not been involved with previously. The theories behind the modern innovation policies are most likely based on empirical observations of an expansion of the organizations’ activities outside their traditional fields. There is, for instance, some evidence that government in many countries is acting less redistributional and more growth-oriented. Universities are increasingly facing a situation where they either have to cut down or act more entrepreneurial. As the knowledge economy expands, companies get stronger incentives to collaborate with universities. However, the traditional Table 8.2 The traditional activity of the three types of organization O, and the activities expected by modern innovation policies (o) Activity Type of organization University Government Firm Education and research O (o) (o) Public infrastructure and service (o) O (o) Product development and production for profit (o) (o) O

8 The Social Capital of Regional Dynamics: A Policy Perspective

133

norms, values and networks, i.e., the existing social capital of each of the type of organization, are formed in accordance with their traditional activity and not changed from one year to another. Thereby, the established social capital of the organizations constitutes intangible obstacles to the implementation of the modern innovation policies.

8.3.2

Social Capital of the Individual

The needs of an individual are other than those of an organization. A primary need of a human being is some form of safety. This affects many acts: work is not only an activity for a pay but also something that colleagues contribute to achieve a high degree of social safety; socialize with friends, and raise a family. In short: individuals construct social capital with relatives, friends and workmates on a day-to-day basis. Relations are built; values and norms are formed to create the necessary stability and safety in a world of uncertainty. The social capital formed by individuals at their workplace falls under the category of spontaneously created organizational social capital. This social capital is not controlled by the organization, but as it is built on workplace relations it has, in varying degree an (positive or negative) impact on the organization’s innovation potential. From a traditional view of economics, it is harder to find any arguments for the impact on innovations of social capital individuals build on their leisure time. By definition, working time is production but leisure time is consumption and for that reason there are more reasons to expect innovations occurring on working time. However, as stated in the introduction of this paper, traditional economics may not be the best tool for analyzing innovations and economic transformation. Moreover, it can be hypothesized that the sharp dichotomy between production time and consumption time (work and leisure) of the industrial economy increasingly is being dissolved in the knowledge economy. Informal discussions, information exchange, evaluations, negotiations, etc., connected to production activities are going on during peoples’ leisure time. This would mean that individuals’ social activities during their leisure time contribute to the forming of a place surplus (Bolton 2002; Westlund and Bolton 2003) which indirectly may have an impact on the development on innovations, their commercialization and diffusion. Concerning the individual’s social capital, we also should note that some of the social capital created by groups of individuals indeed is destructive for innovations and growth. One obvious example is the social capital of individuals in criminal gangs. Another example is, what is referred to ‘‘unemployment cultures’’ in deindustrialized or low developed areas. Both are examples of social capital that have emerged from fundamental needs of safety and which in the given situation are experienced as positive for the concerned individuals – without contributing to positive innovations or economic growth.

134

8.3.3

H. Westlund

Society’s Social Capital

We have established that social capital is built by organizations and individuals, i.e., units with some kind of autonomous decision power. In organizations, public government is also included. However, society in its general meaning, consisting of all individuals and organizations, has no decision power of its own. In what sense is it then possible to talk about society’s social capital? The answer is that society’s social capital can be described as the lowest common denominator of all the networks, norms and values existing among all individuals and organizations in society. Thus, a society with many separate networks and few common norms and values can be characterized as a very heterogeneous society with a ‘‘weak’’ social capital, while a society with few and overlapping networks and many common norms and values can be characterized as a very homogeneous society with a ‘‘strong’’ social capital. But – it is not necessarily that ‘‘weak social capital’’ is always ‘‘bad’’ and ‘‘strong social capital’’ is always ‘‘good’’. One example of a strong social capital on societal level is the Swedish ‘‘local industrial community spirit’’ (bruksanda) which is characterized by small and midsized places with one dominant manufacturing industry during the industrial epoch. A spirit of common interest, formed through demands and counter-demands, resulted in the local factory taking responsibility for the welfare of their employees and their families in exchange for the loyalty of the families to the local factory. Other enterprises, apart from the necessary local service businesses, were potential competitors for the labor force which were regarded as unnecessary. As a consequence, entrepreneurship and establishment of new enterprises were not supported by the norms and values of the local industrial community spirit. The factory and the workers opposed consciously or subconsciously the emergence of new economic actors. During Sweden’s late industrial era, the local industrial community spirit was a local expression of the ideology behind the successful Swedish Model of stable growth and national understanding. On the other hand, during the structural adjustment since the 1970s, this spirit has been a critical problem for these communities. When the context changed, the communities needed actors to renew both the local industry structure and the local social capital. However, to a large extent, the local industrial community spirit blocked the emergence of such renewers. The Swedish Model, of which bruksandan was one component part had its great days from the 1930s to the 1960s. Since then, Sweden has become globalized and the knowledge society has replaced the manufacturing-industrial society. Sweden has also become much more diversified in a number of respects, not least concerning lifestyles. There is a huge new formation and inflow of social capital, among youth, immigrants and people in new professions. In this respect, there is certainly no shortage of social capital in Sweden. However, on societal level, be it a city, region of the whole nation, the social capital is ‘‘weakened’’, with less common denominators than during the days of the Swedish Model. This conclusion is well in line with Putnam’s (2000, 2001) results that the social capital in the USA – and probably also other parts of the developed world – is

8 The Social Capital of Regional Dynamics: A Policy Perspective

135

‘‘weakened’’. But Putnam’s measurements of social capital in the USA are per se a good reason to question his earlier claims (Putnam 1993) of a general correlation between social capital on societal level and economic development. In spite of several decades of weakened social capital (in Putnam’s measures), the USA experienced a remarkably strong economic growth in the 1990s (when the knowledge economy expanded) – a circumstance that stands in complete contrast to what Putnam (1993) found in his study of Italy up to the 1970s (i.e., the industrial epoch). The reason is probably that Putnam’s measures of social capital are focused on the homogeneity of society. Measured in this way, the American regions that in the 1990s scored highest were homogeneous, stagnating, depopulous regions with limited immigration. Consequently, expanding metropolises like Los Angeles showed a very low social capital in his measures. A reasonable hypothesis could be that the homogeneous social capital that Putnam (1993, 2000) focuses on, in general stood in a positive, mutual, selfreinforcing relationship with economic growth during the late industrial period, which in most developed countries lasted up to the 1970s but in Japan lasted until about 1990. During this period, economic growth was built on mass production based on improvement of old innovations through increased capital intensity of production, without any need for new, path-breaking innovations. The decline of industrial society and emergence of knowledge society has changed these conditions dramatically. Computerization and other applications of digital technology have together with other emerging technologies brought innovations back as an essential ingredient for growth. In other words: new combinations of production factors have once again emerged as important – and a social capital that facilitates and promotes these new combinations is needed. In this way, the formation of social capital, the forces for continuity and for change of the content of social capital are processes that evolve in response to the changing societal conditions. It can be assumed that the quantity of ‘‘new combinations’’ is dependent on the quantity and quality of production factors, including the bearers of human capital. This would mean that societies with a certain grade of diversity would promote new combinations. A social capital of certain degree of heterogeneity would in that case be best suited for the current stage of knowledge society. As metropolitan regions often are the most diversified, this can explain why they normally are the centers of growth in the knowledge economy. However, diversity without coherent forces would end up in anarchy. Other characteristics, such as mutual tolerance, are needed to utilize diversity. This line of reasoning corresponds to that of Florida (2002, 2005). Albeit Florida (2002, 2005) avoids using the term social capital – in order to distance himself from Putnam (1993, 2000) – Florida’s contributions center on the role of social norms and values, the networks that are based on them and their impact on regional dynamics. However, Putnam’s and Florida’s theories on the homogeneous social capital and on the importance of diversity and tolerance for regions’ growth respectively, have in common that a large number of links in the cause-and-effect-chain are only assumed but not investigated. Moreover, a weakness in both Putnam’s and Florida’s hypotheses is that they only deal with the social

136

H. Westlund

capital of civil society. The social networks and norms of companies and the other actors of clusters and innovation systems are remarkably absent in their hypotheses.

8.4 8.4.1

Public Policies for Economic and Social Innovations Policies on Different Spatial Levels

In Sect. 8.2, innovations were treated solely from an economic perspective. However, analogous to Chatterjee and Lakshmanan’s contribution to this volume where they distinguish between economic, social and political entrepreneurship, we can also make a difference between economic, social and political innovations. In the remainder of this paper we deal with the first two types of innovation: economic and social. In the former section, three sources of social capital were discerned: the individuals, the organizations and society. If social capital has come to play an increasingly important role for innovations as the innovation system has become more complex, there are obvious reasons to ask what public policies can do to contribute forming social capital with as advantageous characteristics as possible for innovations and growth. Starting with the social capital of individuals, it can be argued that the individual as member of a family, neighborhood and leisure clubs in general gets connected to and forms his/her own links to get connected to social networks in accordance with his/her basic preferences. From the perspective of innovations and growth there are often no motives for public policies improving the social capital of individuals. However, there are many examples of social networks and values that diminish the potential for innovations and growth. Networks based on ethnicity, religion, neighborhood, etc., may on the one hand act as critical support structures for its members’ economic activities. Businesses based on ethnicity can often exploit certain niches and have often low transaction costs. On the other hand, these networks with their particular norms and values may simultaneously lead to lock-ins in low-productive activities and non-efficient utilizing of resources. Thus, there may be good arguments for policies aiming at creating new links, improving access for individuals in certain groups, to new networks. When it comes to organizations, we have already shown that organizations are the prime builders and maintainers of their own social capital. Concerning public sector organizations’ social capital, it is self-evident that it is governed by public policies. Regarding the social capital of firms and other organizations independent of government, the influence of public policies is much smaller, but laws and regulations affect the activities of organizations, their social capital-building included. From the perspective of innovations and growth, what would be the motives for public policies aimed at influencing organizations’ social capital? The answer lies in the increased complexity of innovation processes, discussed above. Innovation is

8 The Social Capital of Regional Dynamics: A Policy Perspective

137

no longer dependent on only combination of production factors but also on actors. The role of policies in general terms is to facilitate actors’ interaction – individuals’ and organizations’ interaction with government included. The only question is: how can this interaction be facilitated? On a national level, the role of governmental policies is mainly indirect. Government can establish good relations with national organizations and leading individuals; change and adapt laws, impose regulations and taxes; create platforms and gather actors, etc., thereby contributing to improvements of the ‘‘innovation climate’’. More direct effects on innovations might come out of direct initiatives and projects with selected actors.7 On the other hand, such direct initiatives from above have a higher risk for failure, due to lack of information and (tacit) knowledge, possessed by other actors. On regional and local level, government can play a much more direct role in getting the actors of innovation together and promoting good relations between them, i.e., to ‘‘create’’ and support clusters. However, a problem is that the leading actors of today not necessarily are those of tomorrow. Companies that have their expansion phase behind them might also have their most innovative phase behind them. The same holds for established, large organizations. Thus, governmental innovation policies at regional and local levels might easily become a victim of path dependencies and promote a social capital that opposes new innovations. Although Schumpeter did not use the concept of path dependencies, he was clearly aware of the problem when he described the problem of social environment’s dislike of changes which might go as far as ‘‘. . . social ostracism and finally to physical prevention or to direct attack’’ (Schumpeter 1934, p.87). Schumpeter’s arguments were of course based on the fact that innovations often bring creative destruction that strikes certain actors. This circumstance makes governmental innovation policies at regional and local level more complicated than normally are taken into account. However, a perhaps bigger problem with current policies for economic and social innovations is the abovementioned fact that the three types of actors have different missions, different core-activities and consequently different norms and values. Innovation policies are normally based on the assumption that the actors of the desired cooperation have a common denominator large enough to motivate investing resources in long-term cooperation. Investment of these resources may in itself be seen as a proof of that common denominator and innovation policies seem to be built on the assumption that the projects strengthen this common denominator automatically. The issue of the cooperating actors’ social relations and norms and values are normally not considered in innovation and cluster policies. Instead, these issues are mainly paid attention to in social and welfare (i.e., redistribution) policies. This problem is illustrated by three Swedish examples.

7 The post-war Japanese National System of Innovation was according to Johnson (1982) and Freeman (1987) a successful example of national innovation policies with such direct effects on innovations.

138

8.4.2

H. Westlund

Three Swedish Examples

VINNOVA, the Swedish agency for innovation systems, was launched in 2001 and has an annual budget of approximately $140,000,000. A substantial part of the ¨ XT, resources are used in different R&D programs. One of these programs, VINNVA offers 50% support to selected long-term (10 years) regional innovation projects. Among the selection criteria are a couple of factors that can be connected to relations and values: the regional leadership should support renewal and there should exist a shared vision. A fundamental idea for the program is triple helix-cooperation between the three actors: business, government and academy. However, the problems connected to the different missions, norms and values of the three actors are not considered. In the projects that so far have been supported, the perceived common denominator is reflected in the actors’ investment of own resources. Apart from investment in the projects’ own trademarks and information, there are very few features of relation-building and other activities that can be compared with individual firms’ investment in corporate culture. One exception is the project for the biotech cluster in Uppsala, which contains ideas about pub evenings for actors in the biotech sector. The program is simply based on the idea that the common denominator exists, that it is sufficiently strong in itself and that there is no need for particular investment in social relations, joint norms and values for the actors in the projects. A second example, a Metropolitan Policy Program for deprived urban neighborhoods in the three biggest cities was launched by the Swedish Government in 1998. The overall goals of the policy are to increase the prospects of the Swedish metropolitan regions for long-term sustainable growth, primarily by contributing to new job opportunities, and to stop social, ethnic and discriminatory segregation. In order to achieve these goals, it could be expected that actions for local innovations and entrepreneurship would be taken. However, hardly any such projects have been started. Instead a large number of other projects in a number of areas were launched. A substantial part of the projects was concentrated on issues related to social capital, aiming at strengthening the cohesion of the neighborhood by changing attitudes and building links between different groups and individuals. Instead of building social capital connected to production, the metropolitan policy has focused on social capital connected to consumption, i.e., peoples’ leisure, living and culture. Instead of building links between the deprived neighborhoods and their inhabitants and the rest of the metropolitan regions, activities were mainly concentrated on the pure local neighborhoods. The third example is the National Delegation for Regional Cooperation on Higher Education, active from 2002 to 2004. The delegation gave economic and supervisory support to projects where universities, public sector bodies and companies collaborated on developing new education projects, adapted to the regional labor market, and to a more general ‘‘platform-building’’ for possible future collaboration between the three actors. Even if social capital-building was not an explicit aim for the delegation, the official evaluation of the delegation concluded that the

8 The Social Capital of Regional Dynamics: A Policy Perspective

139

delegation in practice supported the forming of new social capital through creating new relations between regional actors. The evaluation found that this implicit aim was successfully fulfilled in many projects, but that in many cases it was highly uncertain whether the collaboration would continue when the project grants were finished (Westlund et al. 2005). The three examples show the dilemma of current sectoral policies: on the one hand growth policies through cluster and innovation policies, without understanding of the role of social networks, norms and values; on the other hand policies aiming at growth and social equalization through building local social networks and joint norms and values, without understanding the role of innovation, entrepreneurship and the intraregional labor market; and the support for short-term projects without any strategy as to how to develop the newly established networks of collaboration.

8.5

Concluding Remarks

Innovation has become an increasingly complex process with an increasing number of interacting actors involved. One of the things that facilitate this interaction is positive social relations between the actors. In the wake of the emergence of the knowledge economy, new theories, as those of clusters and regional innovation systems, have stressed region as the spatial level where innovation processes take place. The actors of the economy mainly form their social capital themselves. Whereas most actors solely act in accordance with their own needs, government is the only actor that must take the ‘‘public interest’’ into consideration. This means that governmental policies have a central role in the forming and reforming of regions’ social capital. This circumstance has so far mainly had an impact on social and welfare policies, but very little influence on policies for economic transformation and growth. Thus, research on the social capital of the actors’ of innovation would shed new light on critical aspects of these processes. One of these critical aspects is that the three actor blocs of innovation systems and triple helix have different missions and base their activities on different principles. The fact that government already is launching policies for clusters, innovation systems and triple helix, is in itself a strong argument in favor of conducting further research in the area.

Acknowledgments The author has benefited from comments from Martin Andersson, Kiyoshi Kobayashi, Takashi Omori and three anonymous referees.

References ˚ , Stro¨mquist U (1988) K-Samha¨llets Framtid. Prisma, Stockholm Andersson A Asheim B, Gertler M (2004) Understanding regional innovation systems. In: Fagerberg J, Mowery D, Nelson R (eds) Handbook of innovation. Oxford University Press, Oxford

140

H. Westlund

Bolton R (2002) Place surplus, exit, voice, and loyalty. In: Johansson B et al (eds) Regional policies and comparative advantage. Edward Elgar, Cheltenham Cooke P (1992) Regional innovation systems: competitive regulation in the new Europe. Geoforum 23:365–382 Cooke P (2001) Regional innovation systems, clusters, and the knowledge economy. Ind Corp Change 10:945–974 Cooke P (2003) Regional innovation and learning systems, clusters, and local and global value chains. In: Bro¨cker J, Dohse D, Soltwedel R (eds) Innovation clusters and interregional competition. Springer, Berlin De la Mothe J, Paquet G (eds) (1998) Local and regional systems of innovation. Kluwer, Amsterdam Doloreux D, Parto P (2004) Regional innovation systems: a critical synthesis. Discussion paper series #2004–17. Institute for New Technologies, United Nations University, Maastricht DTI (1998) Our competitive future: building the knowledge-driven economy, CM4176. Department of Trade and Industry, London Eliasson K, Westlund H, Fo¨lster S (2005) Does social capital contribute to regional economic growth? Swedish experiences. In: Kobayashi K, Westlund H, Matsushima K (eds) Social capital and development trends in rural areas. MARG, Kyoto University, Kyoto ¨ , Hagstro¨m P Enright MJ (1998) Regional clusters and firm strategy. In: Chandler AD Jr, So¨lvell O (eds) The dynamic firm: the role of technology, strategy, organization and regions. Oxford University Press, Oxford, pp 315–342 Etzkowitz H (2002) The triple helix of university–industry–government: implications for policy and evaluation. SISTER working paper 2002:11. SISTER, Stockholm Etzkowitz H, Leydesdorff L (1996) Universities and the global knowledge economy: a triplex of university–government relations. Pinter, London Fischer MM, Fro¨hlich J (2001) Knowledge, complexity and innovation systems: prologue. In: Fischer MM, Fro¨hlich J (eds) Knowledge, complexity and innovation systems. Springer, Berlin, pp 1–17 Florida R (2002) The rise of the creative class: and how it’s transforming work, leisure, community and everyday life. Basic Books, New York Florida R (2005) The flight of the creative class: the global competition for talent. HarperBusiness, New York Freeman C (1987) Technology and economic performance: lessons from Japan. Pinter, London Fujita M, Thisse JF (2002) Economics of agglomeration. Cambridge University Press, Cambridge Gibbons M, Limoges C, Nowotny H, Schwartzman S, Scott P, Trow M (1994) The new production of knowledge: the dynamics of science and research in contemporary societies. Sage, London Goldkuhl Ca (1961) John Ericsson, Mannen och uppfinnaren. Bonniers, Stockholm Johansson B (2004) Parsing the menagerie of agglomeration and network externalities. CESIS electronic working paper series, Paper no. 2. www.infra.kth.se/cesis/research/workpap.htm Johansson B, Westin L (1994) Affinities and frictions of trade networks. Ann Reg Sci 28:243–261 Johnson C (1982) MITI and the Japanese miracle: the growth of industrial policy 1952–1975. Stanford University Press, Stanford Kobayashi K, Takebayashi M (2000) Exploiting gateway externalities: the future of Kansai. ˚ E, Andersson DE (eds) Gateways to the global economy. Edward Elgar, In: Andersson A Cheltenham, pp 235–253 Krugman P (1991) Increasing returns and economic geography. J Polit Econ 99(3):483–499 Krugman P (1995) Development geography and economic theory. MIT, Cambridge, MA Lakshmanan TR (1994) State market networks in Japan: the case of industrial policy. In: Johansson B, Karlsson C, Westin L (eds) Patterns of a network economy. Springer, Berlin, pp 99–112 ˚ (ed) (1992) National systems of innovation: towards a theory of innovation and Lundwall B-A interactive learning. Pinter, London Malmberg A (2002) Klusterdynamik och regional na¨ringslivsutveckling – begreppsdiskussion och ¨ stersund forskningso¨versikt. A2002:008. ITPS, O

8 The Social Capital of Regional Dynamics: A Policy Perspective

141

Markusen A (1996) Sticky places in slippery space: a typology of industrial districts. Econ Geogr 72:293–313 Markusen A (1999) Fuzzy concepts, scanty evidence, policy distance: the case for rigour and policy relevance in critical regional studies. Reg Stud 33:869–884 Marshall A (1880/1920) Principles of economics: an introductory volume, 8th edn. Macmillan, London Nelson R (ed) (1993) National innovation systems: a comparative analysis. Oxford University Press, Oxford North D (1990) Institutions, institutional change and economic performance. Cambridge University Press, Cambridge Nowotny H, Scott P, Gibbons M (2001) Re-thinking science: knowledge and the public in an age of uncertainty. Polity, Cambridge Paniccia I (2002) Industrial districts: evolution and competitiveness in italian firms. Edward Elgar, Cheltenham Polanyi K (1944) The great transformation. Beacon, Boston Porter M (1990) The competitive advantage of nations. Macmillan, Basingstoke Porter M (1998) Clusters and the new economics of competition. Harv Bus Rev 76:77–90 Porter M (2000) Location, clusters and company strategy. In: Clark GL, Feldman MP, Gertler MS (eds) The Oxford handbook of economic geography. Oxford University Press, Oxford, pp 253–274 Putnam RD (1993) Making democracy work. Civic traditions in modern Italy. Princeton University Press, Princeton Putnam RD (2000) Bowling alone. The collapse and revival of American community. Simon & Schuster, New York Putnam RD (2001) Social capital community benchmark survey: community result matrix. http:// www.ksg.harvard.edu/saguaro/communitysurvey/results_matrix.html (2004-01-16) Rauch JE (1996) Trade and search: social capital, sogo shosa, and spillovers, Working paper 5618. National Bureau for Economic Research, Cambridge, MA Rauch JE (1999) Networks versus markets in international trade. J Int Econ 48:7–35 Rauch JE (2001) Business and networks in international trade. J Econ Lit XXXIX:1177–1203 Schumpeter JA (1934) The theory of economic development. Harvard University Press, Cambridge, MA Schumpeter JA (1950) Capitalism, socialism, and democracy, 3rd edn. Harper, New York Scitovsky T (1954) Two concepts of external economies. J Polit Econ 62:143–151 So¨rlin S, To¨rnqvist G (2000) Kunskap fo¨r va¨lsta˚nd: Universiteten och omvandlingen av Sverige. SNS, Stockholm Sraffa P (1926) The laws of returns under competitive conditions. Econ J 40:79–116 Suarez-Villa L (2004) Technocapitalism and the new ecology of entrepreneurship. In: de Groot HLF, Nijkamp P, Stough R (eds) Entrepreneurship and regional economic development: a spatial perspective. Edward Elgar, Cheltenham Westlund H (2004) Social capital and the emergence of the knowledge society: a comparison of ¨ stersund Sweden, Japan and the USA. ITPS, O Westlund H, Bolton R (2003) Local social capital and entrepreneurship. Small Bus Econ 21:77–113 Westlund H, Decaio E, Johansson M (2005) Utva¨rdering av Delegationen fo¨r Regional Samverkan ¨ stersund om Ho¨gre Utbildning. ITPS, O

Chapter 9

Hidden Order in Traffic Flows Using Approximate Entropy: An Illustration Kingsley Haynes, Rajendra Kulkarni, and Roger Stough

9.1

Introduction

The dynamic nature of traffic flows on urban freeways is self-evident. The plots of workday traffic on segments of major roads against time of day display the familiar contours of lumpy, peaked curves. Over the years the peaks have become blunt and the valleys filled, suggesting nearly day long high-volume traffic. At the same time that the average travel speed on congested freeways has gone up, average commute time has either remained steady or increased marginally and the number of accidents per 100 million VMTs has gone down or remained constant (Gordon et al. 1991; BTS 2006). Traffic at high volumes and high speeds or under designed roads should result in more accidents and slower travel times. This has not occurred but traffic has continued to increase. Congested traffic patterns suggest an inherent disorder or randomness. Could it be that there is a hidden order in the congested traffic patterns? It would be helpful to analyze and understand these linear spatial patterns to see the degree to which order/disorder associated with these patterns can be determined.

9.2

Level of Service

Currently, congestion is measured as either a ‘‘Level of Service’’ (LOS) category (HCM 1998a) or as a volume to capacity (V/C) ratio (Meyer 1994). However, both of these measures are found increasingly to be inadequate in describing the nature of high volume congested traffic in urban regions with growing travel demand.

R.R. Stough (*) Vice President for Research and Economic Development, George Mason University e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_9, # Springer‐Verlag Berlin Heidelberg 2009

143

144

K. Haynes et al.

Usually, traffic flow data are gathered by direct observation using permanent and/or seasonal traffic sensor counters installed on roads, and in some cases, aerial surveys (SkyComp 1996). These data are used by transportation agencies to plan for land use changes, design road patterns and capacity levels and schedule activity such as construction and maintenance work. Analysis of such data is also useful to real time traffic management centers in controlling for the effects of congestion, and ultimately it is useful to the users of road networks if the data can be used to produce the appropriate kinds of information to support decentralized decision making. Figure 9.1 shows traffic patterns for various linear spatial segments of highvolume urban freeways in terms of level of service (LOS). The letters ‘‘A’’ through ‘‘F’’ (HCM 1998b) are proxies for the number of vehicles per lane per mile in a given time period. The Highway Capacity Manual (HCM) defines LOS as a ‘‘qualitative measure describing operational conditions within a traffic stream, and their perception by motorists and/or passengers’’. To quote TRB Special Report 242, ‘‘Congestion for any facility using LOS approach depends on the quality of service ‘expected’, which may vary between designers and users and even among users. . . .The judgment of the designer and analyst plays a large part in what is defined as congestion (p.21)’’. From this explanation it is obvious that, there is an inherent fuzziness associated with LOS definitions of A–F. For example, according to the Highway Capacity Manual, the spectrum of LOS from ‘‘A’’ to ‘‘F’’ grades the traffic flows in terms of a certain number of cars per lane per mile. The LOS ‘‘A’’ A C

A

A

B C A A

C B

B C

B C C D C F

C F E B B

D E C C B

A A

Fig. 9.1 LOS on multiple links of a freeway network

B

A

9 Hidden Order in Traffic Flows Using Approximate Entropy: An Illustration

145

Table 9.1 Average separation distance and LOS expressed as alphabets Level of service (LOS)~ Avg. distance in feet Level of service (LOS) number of cars/lane-mile between two consecutive expressed as alphabets (C/lm) cars and number of cars (C) A~12 ~440 (~22) a~b+c B~20 ~260 (~13) b~c+d C~30 ~175 (~09) c~d+e D~42 ~125 (~06) d~e+f E~67 ~080 (~04) E F>67 ~020 (~01) F

pattern has thirteen or fewer cars representing free flow condition, while LOS ‘‘F’’ at the other end of the spectrum represents near breakdown in flow with more than 67 cars per lane per mile (see Table 9.1). Another way to look at LOS is in terms of the average space – or separation distance – between two vehicles in a lane segment. Column 2 in Table 9.1 shows the corresponding separation distances for each level of service (A through F) measured in terms of the number of car spaces. We represent these in small case letters ‘‘a’’ through ‘‘f’’. The representation of the separation distances (‘‘a’’, ‘‘b’’, ‘‘c’’, ‘‘d’’, ‘‘e’’ and ‘‘f’’) will be used to develop the production rules for a flow of traffic. These rules will allow us to express complex traffic patterns in a single comparable metric. Below we discuss computational (Kolmogorov) entropy as it relates to traffic flow. We do this in the most minimalist way possible so that operational background issues are covered in order to move to a statistical specification of approximate entropy.

9.3

Kolmogorov’s Entropy and Traffic Flow

Consider a coin tossing experiment. For each toss, if the outcome is a head, we write ‘‘1’’, otherwise we write ‘‘0’’. The following is one of the possible outcomes of a large number of coin tosses s ¼ 001010001110100010:::000110101000:

ð9:1Þ

Though the frequency of number of tails and heads (zeros and ones) turns out to be nearly 1/2 as the number of coin tosses tends to infinity, the series can never be predicted to reproduce itself exactly in the same way as is shown above. Hence the randomness of this series is measured in terms of the length of the series ~| s |, or the number of bits needed to specify the series ‘‘s’’. On the other hand, if with a biased coin we get only heads (all ones) or only tails (all zeros), obviously a non-random sequence, its randomness is simply the number of bits needed to specify the number ‘‘n’’. For example, a series with random heads and tails of length 100 would need 100 (binary) bits to specify that series, while a series of 100 ones (or zeros) can be specified by a maximum of seven binary bits (27 >100). Kolmogorov’s entropy ‘‘k’’

146

K. Haynes et al.

of a sequence ‘‘s’’ is measured in terms of the number of bits of the smallest program that can regenerate the sequence. Thus, kðsÞ ¼ js j;

ð9:2Þ

where the right hand side of the above equation measures the length of the program s*. Of course, if there is no such program then the length of the entire sequence ‘‘s’’ in bits becomes the entropy of the sequence. Thus the number of binary bits needed to specify a sequence can be used as a measure of the entropy of the sequence (Zurek 1989a,b). Note that, it is customary to represent the entropy of an integer ‘‘n’’ simply as kðnÞ  lnðnÞ:

ð9:3Þ

If there are two sequences ‘‘s’’ and ‘‘t’’ then the following relations hold: kðs þ tÞ  kðsÞ þ kðtÞ if s > t such that s > t ; then

ð9:4Þ

kðs Þ > kðt Þ:

ð9:5Þ

Next, let us apply the Kolmogorov’s entropy concept to the following traffic situation on a free way. Imagine a single lane spatial segment of a freeway with a traffic sensor installed somewhere in the middle of this segment. Let the output of traffic sensor be fed into a processor (computer) for further analysis of traffic patterns, just as the outputs from all other sensors from different sections of a freeway road network are fed to the processor. At any instant the traffic sensor detects the presence or absence of a vehicle. Let the presence of a vehicle be coded as ‘‘1’’ and the absence as ‘‘0’’. If this spatial segment has a near free flow traffic, then we may observe a series of ‘‘1’’s and ‘‘0’’s such as the one shown below: . . . 0000000000110000100100001. . . 00001:

ð9:6Þ

The series in (9.6) shown above has no pattern and appears as a random sequence of zeros and ones. Next let us imagine extremely congested traffic on the same one lane link. One of the possible series of observations is given by: 1111111111111111111111 . . . 11111:

ð9:7Þ

The series of ‘‘1’’s is clearly not random. To describe the series in (9.7), all one needs to do is to count the number of ‘‘1’’s¼n1¼n, an integer number. On the other hand there is no way to describe the series in (9.6), but to reproduce the entire sequence as it is. To get maximum information from series in (9.7), all we need to know is the number ‘‘n’’, representing the number of ones, while for series in (9.6), the only way to gain information is to look at the entire sequence. The example above is analogous to the description of Kolmogorov randomness. To quote

9 Hidden Order in Traffic Flows Using Approximate Entropy: An Illustration

147

Chaitin, ‘‘. . .A series of numbers is random if the smallest algorithm capable of specifying it to a computer has about the same number of bits of information as the series itself’’ (Chaitin 1975). Thus the computer would process series in (9.7) in a single step, while the series shown in (9.6) would need as many steps as there are ones and zeros in the series. In fact, the series in (9.6) appears similar to the outcomes of a series of un-biased coin tosses, while series in (9.7) is similar to a biased coin toss that invariably produces a series of heads (ones). As was stated earlier, the traffic patterns have been described in terms of levels of service ‘‘A’’ through ‘‘F’’. Our aim here is to transform these into numeric patterns that could be used to compute degree of randomness based on Kolmogorov entropy. One of the ways this can be achieved is by devising transformation rules based on Table 9.1. 1. Given below is a set of transformation rules: Rule I a ! bc | cb Rule II b ! cd | dc Rule III c ! de | ed Rule IV d ! ef | fe Rule V af ! a, bf ! b, cf ! c, df ! d, ef ! e, ff ! f; where, ‘‘a’’, ‘‘b’’, ‘‘c’’, and ‘‘d’’, ‘‘e’’, ‘‘f’’ are service levels and ‘‘f’’ denotes a null character, and ‘‘|’’ and ‘‘!8’’ are meta-characters. Note that ‘‘a’’ is equivalent of combinations of either ‘‘b’’ and ‘‘c’’ or ‘‘c’’ and ‘‘b’’ (see Figs. 9.3 and 9.4 for details.) Similarly, ‘‘b’’ is equivalent to a combination of ‘‘c’’ and ‘‘d’’ and so on. The above rules are applied as follows, all occurrences of ‘‘a’’ are replaced by ‘‘bc’’ or ‘‘cd’’ (Rule I); all occurrences of ‘‘b’’ are replaced by ‘‘cd’’ or ‘‘dc’’ (Rule II) and so on. Further Rule V can be decomposed into two sub-rules based on the separation distance between vehicles in a lane. For example, level of service ‘‘e’’ means the separation distance between two cars in a lane is about 80 ft, equivalent of four cars’ lengths including headway. Thus ‘‘e’’ can be decomposed as follows: Rule V(1) e ! 100001 Similarly, a level of service ‘‘f’’ is equivalent of two vehicles separated by a distance of one car length plus the headway. Thus ‘‘f’’ can be decomposed as: Rule V(2) f ! 101 2. Next a series ‘‘W’’ of levels of service on a stretch of road can be represented numerically in terms of the length of such a series as: | W |, and for a series with no vehicles can be represented as | |=0. 3. Within a series ‘‘W’’, a frequency of service element such as ‘‘e’’ is given by |W|e. 4. If U and V are two elements on a stretch of road such that U6¼V then, UV¼VU; |U+V|¼|U|+|V|, and |U+V| f ¼|U| f +|V| f. Thus, the above transformation rules 1 through 4 can be used to describe a typical traffic pattern expressed in levels of service.

148

9.4

K. Haynes et al.

Randomness and Order in Traffic Patterns

Next, let us consider an urban freeway with N links and L levels of service (LOS). The total number of possible traffic patterns in terms of LOS would be given by NL. For example, a freeway of N=10 links with L=6 would have 106 possible traffic patterns in terms of LOS. Let this 10 link freeway have a traffic pattern represented by LOS ‘‘A’’, i.e. AAAAAAAAAA:

ð9:8Þ

When expressed in terms of the formal traffic language as defined above, it would be equivalent to aaaaaaaaaa:

ð9:9Þ

Without loss of generality let us consider link 1, with LOS ‘‘A’’, and coded as ‘‘a’’ (see Table 9.1). Then using the production rules I–V successively we would get the following: A ¼ bcjcb ¼ efðcdcjdccÞjðdebjedbÞg;

ð9:10Þ

where, each of the terms inside the curly brackets can be further expanded by recursively applying rules I–V. An example of decomposition of ‘‘a’’ into its components is shown in Fig. 9.2. Although there are multiple ways of decomposing each of the LOS as shown in Figs. 9.3, 9.4 and 9.5, if each rule is applied to the left most symbol at a time, then the total number of possible combinations is given by jwjeþf ! je þ f j! ; qffi jej!  jf j! jwje !  jwjf !

Fig. 9.2 One of several ways of decompositions of ‘‘a’’ into its sub-parts

ð9:11Þ

9 Hidden Order in Traffic Flows Using Approximate Entropy: An Illustration efcc fecc ddec dedc dcde dced

dcc

efced feced

bc

dedc cdc

a cb

deb

eddc feeb efeb

ccd

effeeed efefeed

efdeed efeded efeefed fedeed fefeeed feeded feefeed feefeeef feeefed feefeefe feeefefe feeefeef

149

effeeeef effeeefe efefeeef efefeefe efeefeef efeefefe fefeeeef fefeeefe

decd edcd

Fig. 9.3 Partial decomposition of LOS ‘‘a’’

ded

cd edd b dc

efeef efefe feeef feefe efefe efeef eefef eeffe

100001,010,100001,100001,010

efefe efeef effee efefe

efed

efc

efde

100001,010,100001,010,100001

fefee fec

fede

feefe feefe

feed

feeef

Fig. 9.4 Partial decomposition of LOS ‘‘b’’

where ‘‘!’’ stands for the factorial, ‘‘w’’ is a combination, ‘‘e’’ and ‘‘f’’ are the levels of service, ‘‘q’’ represents the tiny number of combinations prohibited by the production rules, and from axiom four, | w |e is the frequency of ‘‘e’’ in ‘‘w’’. It is obvious that when ‘‘a’’ is decomposed using the rules of production, the number of possible combinations are many more than for ‘‘b’’, ‘‘c’’ and ‘‘d’’. In terms of number of possible combinations, the following relation holds a  b  c > d > e ¼ f:

ð9:12Þ

150

K. Haynes et al.

e

e

100001

f

f

010

ef

100001,010

d fe

010,100001

efe

100001,010,100001

fee

010,100001,100001

de c

eef

100001,100001,010

efe

100001,010,100001

ed

cd d dc

Fig. 9.5 Partial decomposition of LOS ‘‘c’’

Specifically, we have the following relations: total number of combinations expressed in terms of service levels ‘‘e’’ and ‘‘f’’ for combined service level ‘‘a’’: ffi

je þ f j! 8! ¼ ¼ 56 jej!j f j! 5!3!

ð9:13Þ

ffi

je þ f j! 5! ¼ ¼ 10 jej!j f j! 3!2!

ð9:14Þ

combined service level ‘‘b’’:

combined service level ‘‘c’’: and je þ f j! 3! ¼ ¼3 jej!j f j! 2!1!

ð9:15Þ

je þ f j! 2! ¼ ¼2 jej!j f j! 1!1!

ð9:16Þ

combined service level ‘‘d’’:

9 Hidden Order in Traffic Flows Using Approximate Entropy: An Illustration

151

while level ‘‘e’’ is ‘‘e’’ and level ‘‘f’’ is ‘‘f’’. Note that, for each of these combinations, the neighboring links have their own set of combinations. Next let us consider a freeway consisting of ten links with traffic flow patterns under different conditions described by the LOS. For illustration purposes, let us assume that all ten links are of same length (about 500 ft.). Suppose that, each spatial segment has a free flow traffic pattern corresponding to LOS ‘‘A’’, which could be represented in terms of the separation distance corresponding to LOS ‘‘A’’, namely ten ‘‘a’’s, as a combination sequence ‘‘aaaaaaaaaa’’. Next we apply the production rules I–IV from axiom 2, to each link, then one of the possible combinations is given by RðAÞ ¼ efeffeeeffefeefeef . . . fefefefeef;

ð9:17Þ

where ‘‘f’’, is a null character that separates each link. Note that ‘‘f’’ does not contribute anything towards the free flow traffic pattern. Alternately, coding ‘‘e’’ as ‘‘100001’’ and ‘‘f’’ as ‘‘010’’ we get the following: RðAÞ ¼ 100001f 010f 100001f010f010f100001f100001f100001f 010f100001f 010f100001f100001f010f100001f100001f . . . : f100001f010f100001f010f100001f010f100001f100001f ð9:18Þ If we want to determine Kolmogorov’s entropy k for the sequence in (9.18), the best we could do is to rewrite the entire sequence as is. Hence from (9.2), k for such a sequence is just the length of the sequence, given by kðRðAÞÞ ¼ jRðAÞj ffi 80:

ð9:19Þ

When each of these ten links have patterns corresponding to LOS ‘‘D’’, then from the production rule IV, we could rewrite corresponding ‘‘d’’ as either ‘‘ef’’ of ‘‘fe’’. Rewriting ‘‘100001’’ for ‘‘e’’ and ‘‘010’’ for ‘‘f’’, one of the possible combinations (~12) is given by RðDÞ ¼ 100001f010f100001f010f100001f010f100001f010f 100001f010f . . . f100001f010f100001f010f100001f010f ð9:20Þ 100001f010f100001f010f: The best one can do to get a measure of Kolmogorov entropy ‘‘k’’ is to compute the length of the sequence, just as shown in (9.19). Of course, there is a possibility that all ten links will have exactly the same pattern only of ‘‘ef’’ (‘‘fe’’), an exceedingly small probability that such an event would occur.

152

K. Haynes et al.

On the other hand we consider the following situation, where all ten links have traffic pattern corresponding to LOS ‘‘E’’, rewriting the pattern in terms of alphabet ‘‘e’’ we get the following: RðEÞ ¼ eeeeefeeeeef . . . feeeeef eeeeef

ð9:21Þ

Alternately, when ‘‘e’’ is replaced by ‘‘100001’’ we get RðEÞ ¼ 100001f100001f100001f100001f100001f100001f 100001f100001f100001f100001f . . . 100001f100001f

ð9:22Þ

100001f100001f100001f: The pattern in (9.22) can be described using (9.2) and (9.3) by its Kolmogorov entropy as kðRðEÞÞ ¼ lnð10Þ  kðeÞ ¼ lnð10Þ  j10001j;

ð9:23Þ

where the first term corresponds to the number of links and the second term corresponds to the number of bits needed to describe the pattern. Next, let us consider the situation when all ten links have a traffic pattern corresponding to LOS ‘‘F’’, which can be represented in terms of alphabet ‘‘f’’ as given below: RðFÞ ¼ fffffffffffffffffffffff ffffffffffffffffffffffff f . . . ffffffffffffffffffffffff f:

ð9:24Þ

Note that there can be 255 f’s corresponding to 20-foot separation distances between cars over one lane-mile. Relation (9.24) can be coded as ‘‘010’’ in place of ‘‘f’’ and we get RðFÞ ¼ 010f010f010f010f010f010f010f010f010f010f010ff010f010 f010f010f010f010f010f010f010f010f010f010f010f010f010 f010f010f010f010f010f010f010f010f010f010f010f . . . 010f 010f010f010f010f010f010f010f010f010f010f010f010f: ð9:25Þ Series R(F) can be described as ‘‘n’’ links with ‘‘010’’ pattern and hence the Kolmogorov entropy of the sequence using (9.2) and (9.3) is given by kðRðFÞÞ ¼ kðn  ðf ÞÞ ¼ lnðnÞ j010j:

ð9:26Þ

where, the first term refers to the log of the number of links and second to the number of bits needed to describe the traffic pattern of type ‘‘f’’. It is clear from (9.18)–(9.25), that although Kolmogorov’s entropy is high for LOS ‘‘A’’ through ‘‘D’’, it goes to a minimum when the traffic patterns correspond to LOS ‘‘E’’ and

9 Hidden Order in Traffic Flows Using Approximate Entropy: An Illustration

153

LOS ‘‘F’’. Alternately, LOS ‘‘F’’ has the least amount of randomness in traffic flow patterns, and hence has the maximum order resulting in maximum throughput. It is clear that Kolmogorov’s entropy measure ‘‘k’’ can be used to measure the hidden order in the traffic flows.

9.5

Approximate Entropy

Algorithmic entropy described above can be used to distinguish between regular sequences and those that are irregular. However, it may happen that, there can be a number of sequences with enough variations in 1 and 0s, but have the same value of algorithmic entropy (Pincus and Singer 1996, 1998). So, which of these sequences are more random or more irregular or have more algorithmic entropy? How does one measure the degree of randomness? Pincus (1991) offers a statistical tool called Approximate Entropy (ApEn) to help determine degree of randomness of sequences of arbitrary length. Here we provide a brief description of this measure. According to Pincus (1991), the following definitions are applicable to real number sequences. Given a time-series of data t(1), t(2), . . . ,. t(N) from measurements equally spaced in time, let x(i) and x(j) represent two subsequences, where xðiÞ ¼ ½tðiÞ; tði þ 1Þ; . . . ; tði þ m  1Þ

ð9:27Þ

xðjÞ ¼ ½tðjÞ; tðj þ 1Þ; . . . ; tðj þ m  1Þ

ð9:28Þ

such that m is positive integer and mN. Then the distance between x(i) and x(j) is defined as d½xðiÞ; xðjÞ ¼

max ðtði þ k  1Þ  tðj þ k  1ÞÞ:

k¼1;2;...;m

ð9:29Þ

Next, define, 1 Ym ðrÞ ¼ ð N  m þ 1Þ

Nmþ1 X

! log Pm i ðrÞ ;

ð9:30Þ

i

where r 0 pm i ðrÞ ¼

1 ½number of subsquences  ðN  m  1Þ jd ½xðiÞ; xðjÞ  r : ðN  m  1Þ ð9:31Þ

Now from (9.27)–(9.30), the approximate entropy is defined as

154

K. Haynes et al. mþ1 ApEnðm; r; NÞðSÞ ¼ ðYm Þ; i  Yi

when m 1:

ð9:32Þ

And by definition when m=0 ApEnð0; r; N ÞðSÞ ¼ Y1 ðrÞ;

ð9:33Þ

where r 0. It follows from the above definition that, when subsequences of length m and m+1 are similar (not random), i.e., the distance between x(i) and x(j) is smaller than r, then ApEn is small. When subsequences are dissimilar (random), the distance between subsequences x(i) and x(j) is large and therefore ApEn is large. Note that m specifies a sliding window that travels over the sequence and generates subsequences, which are tested against each other for similarity with the distance function (9.29), subject to the upper limit of value r. Thus, the value of ApEn indicates the regularity or randomness of a sequence. Pincus refers to ApEn as a measure of logarithmic frequency of subsequences (Pincus and Kalman 1997). Variations of above definitions are applied to binary sequences, when r [0, l] and r
9.6

Illustration

Once a formal set of rules for traffic are described, it would be possible to divide/ separate the LOS strings of traffic flows into sequences that can be studied for their entropy (disorder) content. This helps in recognizing whether a segment of highway shows random or regular traffic patterns. However, as mentioned above, the Kolmogorov entropy (disorder) measure is quasi-quantitative. Even though it does allow one to classify traffic patterns in broader categories based on a randomness (non randomness) measure, it still leaves much to be desired in terms of finding a degree of randomness in such patterns. The methodology by Pincus and Singer (1998) will do just that. Hence, the logical next step in the research would be to compute statistically the approximate entropies (ApEn) of the traffic patterns. ApEn is a more refined measure of order/randomness. A simplified example, given below, will illustrate the difference between and usefulness of Kolmogorov’s computational and Pincus’ statistical or approximate (ApEn) measures. Example 1: Consider a four-lane freeway with a vehicle detection station. Without loss of generality, (for illustration only), let us assume that the station output is summed for the four lanes and we will count the data for short duration, (again without loss of generality), for example, just for short duration of eight time periods. Then the possible output of the detection station is: no vehicles (0),

9 Hidden Order in Traffic Flows Using Approximate Entropy: An Illustration

155

1 vehicle, 2 vehicles, 3 vehicles or 4 vehicles. One of the possible scenarios could be high-speed congested traffic that is detected by vehicle detection sensors for eight time periods. One such output sequence is: A ¼ 44444444 Yet another possible sequences is: B ¼ 23043424 or a sequence such as: C ¼ 34343434 or D ¼ 12341324 And E ¼ 123043 42 Clearly, sequence A is not random and hence from the definition (9.2) has the least amount of Kolmogorov entropy. Sequence C also appears non-random, however, it is Kolmogorov entropy is more than sequence A. Finding Kolmogorov entropy of sequences such as B, D and E appears a little difficult. In that case Pincus’ statistical estimate of approximate (ApEn) entropy could be used to rank these sequences by their randomness. To compute, ApEn, let us assume r is a nonnegative integer and r<2. Further, let m¼2. Then for m¼2, 0r>2 and 1>r>3 Next compute the ApEn from (9.9) for each of the sequences, the output of which is shown in Table 9.3. It is clear that the ranking of the sequences with increasing ApEn values (increasing randomness) is D<E
Y(m= 2) (0
Y(m=2) (1
156

K. Haynes et al.

Table 9.3 Sub-sequences when m=3 Pattern Subsequence for m=3 B=2 3 0 4 3 4 2 4 2 3 0,3 0 4,0 4 3,4 3 4,3 4 2,4 2 4 D=1 2 3 4 1 3 2 4 1 2 3,2 3 4,3 4 1,4 1 3,1 3 2,3 2 4 E=1 2 0 3 4 3 4 2 1 2 0,2 0 3,0 3 4,3 4 3,4 3 4,3 4 2

Table 9.4 Approximate entropy (ApEn) of the sequences ApEn of sequence ApEn(0
Y(m= 3) (0
Y(m=3) (1
ApEn(1
well as long sequences, and holds independent of r. Note that one could use similar methodology to compute Kolmogorov and ApEn entropies for a single or multiple lanes, either separately or in combination. This may help in understanding the formation, growth and/or dispersion of vehicle platoons. It could also help in determining the entropies associated with mixing of vehicle types. Appendix shows two types of examples, non-binary and binary. The results were computed using a commercial software product (Simulnet-Pro, # Edward J. Rzempelouck, 1993–2001).

9.7

Conclusion and Future Research

As was mentioned in the introduction, increase in VMTs on urban freeways and the apparent lumpiness of traffic flow contours occurring round the clock (Fig. 9.1) suggest that, now and in future the urban freeway traffic would be characterized more by LOS ‘‘C’’, ‘‘D’’, ‘‘E’’, and ‘‘F’’ than by ‘‘A’’ and ‘‘B’’. But even with moderate to heavy congestion levels experienced on urban freeways, the commute times have remained steady and there is a decline in the number of accidents per 100 million VMTs. This apparent paradox can be partly explained by the inherent order hidden in the congested traffic. The traffic flow patterns defined in this paper when combined with the computational entropy concepts (Kolmogorov entropy ‘‘k’’) helps us to quantify and measure the inherent order associated with the congested traffic. If we are able to keep the Kolmogorov’s entropies for congested traffic low we might be able to increase the throughput of freeway systems without increasing capacities of freeway systems. A more complex pattern of traffic based on information gathered by advanced ITS technologies would help us to identify the optimal levels of congestion and increase the throughputs by being able to reduce Kolmogorov’s entropy. Some researchers have suggested and experiments are underway on electronically coupled trains of motor-vehicles for freeway systems that could in theory travel at high speeds, with practically no accidents and very high throughput, i.e., the so called Automated Highway Systems. This paper

9 Hidden Order in Traffic Flows Using Approximate Entropy: An Illustration

157

suggests the first few analytical steps, based on Kolmogorov entropy, towards building the information management infrastructure to support operation of such systems. What could be more appropriate in the information age than to be able to utilize such information using the methods stated in this paper towards achieving the goal of efficient transport? Since, none of the measurements made on traffic conditions are exhaustive or complete, we could enunciate the following relationship between the physical entropy of urban traffic freeway system with its Kolmogorov entropy Physical entropy Q ¼ Ignorance I þ Kolmogorov entropy k: The above relationship (Zurek 1989a,b) could serve as a basis to reduce the Ignorance ‘‘I’’ by getting better information to characterize the traffic patterns by using more advanced ITS technologies which in turn would decrease the total physical entropy thereby enhancing the overall efficiency of urban freeway traffic system. Acknowledgments The authors appreciate the support of the NSF/EPA Grant #SES-9976483 ‘‘Social Vulnerability Analysis’’ and NSF Grant #ECS-0085981 ‘‘Road Transportation as a Complex Adaptive System’’ as well as the School of Public Policy’s USDOT Center of Excellence in Evaluation and Implementation funded under DOT Grant #DTRS98-G-0013. Any errors are the responsibility of the authors.

Appendix Example 1: Non-binary sequence* 1, 2, 4, 0, 4, 2, 3, 4, 1, 2, 0, 1, 2, 1, 3, 4, 2, 1, 0, 1 1, 2, 3, 4, 1, 3, 2, 4, 1, 2, 0, 3, 4, 3, 4, 2, 1, 4, 2, 0 1, 2, 3, 4, 1, 2, 3, 4, 1, 2, 3, 1, 2, 3, 4, 1, 2, 3, 4, 1 3, 0, 3, 0, 3, 0, 3, 0, 3, 0, 3, 0, 3, 0, 3, 0, 3, 0, 3, 0 4, 3, 4, 3, 4, 3, 4, 3, 4, 3, 3, 3, 4, 4, 3, 4, 3, 3, 4, 3 M=1, r=2: Approximate entropy Row

ApEn

1 2 3 4 5

0.2376717 (4) 0.2277839 (3) 0.1243837 (2) 0 (1) 0.3837251 (5)

M=2, r=2: Approximate entropy Row

ApEn

1 2 3 4 5

0.2538802 (4) 0.2369515 (3) 0.1523855 (2) 0 (1) 0.4008884 (5) (continued)

158

K. Haynes et al. Example 1: (continued) M=3, r=2: Approximate entropy Row

ApEn

1 2 3 4 5

0.258574 (4) 0.2432303 (3) 0.1967103 (2) 0 (1) 0.298493 (5)

Example 2: Binary sequence* 0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1,0,1 0,1,1,0,1,1,0,0,1,1,0,1,1,1,1,0,0,0,1,0 M=1, r=2: Approximate entropy Row

ApEn

1 2

0.104261 (1) 0.6832949 (2)

M=2, r=2: Approximate entropy Row

ApEn

1 2

0 (1) 0.5252663 (2)

M=3, r=2: Approximate entropy Row

ApEn

1 0 (1) 2 0.5232481 (2) *The numbers in parentheses show the ranking (or order) of the sequences. M=sliding window and r=length of the filter

References BTS (2006) Pocket guide to transportation statistics research and innovative technology administrator. Bureau of Transportation Statistics, U.S. Department of Transportation, Washington, D.C Chaitin G (1975) Randomness and mathematical proof. Sci Am 235(5):47–51 Gordon P, Richardson H, Hjung-Jin (1991) The commuting paradox: evidence from the top twenty. J Am Plann Assoc 57(4):416–420 HCM (1998a) National Research Council TRB Highway Capacity Manual, special report 209, 3rd edn. TRB, Washington, D.C HCM (1998b) Metric analysis reference guide, supplement to 1997 update of special report 209 Highway Capacity Manual. TRB, Washington, D.C Meyer MD (1994) Alternative methods for measuring congestion levels. TRB special report 242(II). TRB, Washington, D.C., pp 32–63 Pincus S (1991) Approximate entropy as a measure of system complexity. Proc Natl Acad Sci USA 88:2297–2301 Pincus S, Kalman R (1997) Not all (possibly) ‘‘random’’ sequences are create equal. Proc Natl Acad Sci USA 94:3513–3518

9 Hidden Order in Traffic Flows Using Approximate Entropy: An Illustration

159

Pincus S, Singer B (1996) Randomness and degree of irregularity. Proc Natl Acad Sci USA 93:2083–2088 Pincus S, Singer B (1998) A recipe for randomness. Proc Natl Acad Sci USA 95:10367–10372 SkyComp (1996) Draft report on traffic quality on the Metropolitan Washington area freeway system spring. Prepared by SkyComp Inc., Rockville, MD for the Metropolitan Washington Council of Governments TRB Special Report 242 (II), pp 1–31 Zurek WH (1989a) Algorithmic randomness and physical entropy. Phys Rev A 40(8):4731–4751 Zurek WH (1989b) Thermodynamic cost of computation, algorithmic complexity and the information metric. Nature 341:119–124

Chapter 10

Regional Input–Output with Endogenous Internal and External Network Flows John R. Roy and Geoffrey J.D. Hewings

10.1

Introduction

Regional I–O analysis has a long history, including seminal works by Chenery (1953) and Leontief and Strout (1963), with the latter analysis seen as a disaggregation of the Leontief–Strout (L–S) approach. Other significant contributors include Isard et al. (1960), Polenske (1980), Hewings (1985), Miller and Blair (1985) and Oosterhaven (1988). An overview is provided in Roy (2004a). As more and more regional survey data became available, such analysis was approached with more confidence. In fact, regional I–O has become one of the most widely practiced techniques in the field of regional science. Before proceeding further, we need to clarify the terminology. For this, we turn to Isard et al. (1998). The first class of regional model which they define is the interregional model where both the flows and the I–O coefficients have four indices, that is, the flow of sector i into sector j from region r to region s. As it is extremely difficult to implement a full interregional model, most developments have concentrated on devising multi-regional models with less stringent data requirements. Although the dimensionality of these approaches reduces from four to three, different indices are absorbed in the flows compared to the I–O coefficients. The flows relate to the total flow of sector i as input to all other sectors between regions r and s, with the aggregation over destination sectors j. These flows are more likely to be available within freight statistics. The I–O coefficients relate to the amount of the sector i product being supplied as intermediate inputs to sector j in region s per unit of output of sector j in region s, aggregated over the different regions r which supply the inputs. It is precisely this different nature of the aggregation of the flows versus that over the I–O coefficients which creates the main challenge to development of sound multiregional methods.

J.R. Roy (*) ETUDES, Australia e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_10, # Springer‐Verlag Berlin Heidelberg 2009

161

162

J.R. Roy and G.J.D. Hewings

In conventional regional input output analysis, merely total flows are considered, without any attempt at disaggregation. In Roy (2004a,b) two advances were made (1) internally generated flows direct to final demand were determined and (2) regional flows of external exports to outside the regional system, as well as external imports from outside the regional system to satisfy both internal intermediate and final demand were recognized in the flow relations, but not yet determined endogenously.1 Although this was a useful advance on the classical analysis, it was not yet very practical, as multi-regional intermediate input flows arising from external imports needed to be available as data. This anomaly was recognized by the second author, who realized that a truly useful approach needed to yield endogenous multi-regional flows disaggregated into all of their five components: 1. Regional flows of internal intermediate inputs (as in the classical analysis) 2. Regional flows of internally generated final demand (as in Roy 2004a, b) 3. Regional flows of external exports transshipped internally to exit the regional system 4. Regional flows of external imports from outside the regional system transshipped internally to provide internal intermediate inputs 5. Regional flows of external imports from outside the regional system transshipped internally to contribute to internal final demand Note here that, although we have endogenized the transshipment flows within our regional system caused by exports to and imports from outside our regional system, we still need external export and import totals for each sector and region to be provided exogenously. Thus, our model does not directly contribute to the debate surrounding the concepts of Cole (1997). In fact, our approach can be aligned with those of the numerous critics of Cole (see, e.g., Jackson and Madden 1999; Oosterhaven 2000). In particular, the modeling framework proposed here accords with Oosterhaven’s (2000) comment that to handle endogenous interregional feedbacks, a full interregional model is required. Part of the motivation for the need for our disaggregation may be found in an analysis of the Japanese interregional system over time. The results revealed that changes in interregional components were far more important that changes in technology in accounting for changes in output over time (see Hitomi et al. 2000). A further improvement is to identify each of the flows on possible multiple paths between each set of regions. Once such disaggregated flows are available, a path-link transformation can evaluate the flows on each link of the network, allowing a future consideration of congestion (not included in this paper; on this issue, see Sohn et al. 2004; Kim et al. 2004). Whereas Leontief and Strout (1963) developed balance relations to incorporate technology into the flow determination, Chenery (1953) used matrix inversion, with 1

We remain indebted to an anonymous reviewer for stimulating the recognition of internal final demand flows, as well as to Suahasil Nazara, formerly at Regional Economics Applications Laboratory (REAL), for suggesting the inclusion of the external import and export transshipments in the flow totals.

10

Regional Input–Output with Endogenous Internal and External Network Flows

163

neither explicitly modeling the transport network (instead, using trade coefficients). However, in principle, the influence on the pattern of flows of the transport network and of the technology should be jointly determined. As there are insufficient equations to solve such a system deterministically, Wilson (1970) enhanced the deterministic L–S procedure, introducing uncertainty into the analysis via entropy, with his model being constrained in estimation to reproduce base period values of both a transport cost constraint and the right-hand sides of the L–S balance relations. By this means, he also bypassed the need for trade coefficients. Thus, the flows were co-determined by the transport cost information together with the technological information embodied in the L–S representation of input–output. The entropy approach represents one useful procedure to estimate models to determine flows which simultaneously satisfy certain observed base period quantity and price/cost relations, whilst at the same time implicitly accounting for variability in the behavior of the individual agents within the market segments being modeled. It is a generic statistical technique possessing very useful asymptotic properties in the presence of large ‘‘populations’’, defined and formalized by Smith (1990) as Most-Probable-State Analysis. It must be interpreted differently depending on the particular field of application. Many of these interpretations have been made in the field of regional science by the first author, as illustrated in Roy (2004a). Another alternative is to replace the entropy framework by one from information theory based on historical trade patterns (Snickars and Weibull 1977), yielding models such as in Batten (1983). Finally, in the following analysis in terms of the five sets of disaggregated component flows, the L–S relations are clearly insufficient in number in themselves to uniquely determine these flows, again requiring the introduction of an overarching objective such as entropy maximization, even when transport influences are neglected. In any short run model of regional supply, it is desirable to include output capacity constraints. Of course, an obvious way to achieve this is to introduce
inequality constraints on production with respect to regional capacity for each sector. However, such constraints, when inactive, maintain separability and have no influence whatsoever on the flows – they only factor into the analysis once they become active. Intuitively, this is not very plausible. In most sectors, a ‘‘vintage’’ distribution over regional capacity exists, and when the capacity limits of a sector are being hard pressed in a certain region, one would expect some spillovers into adjacent regions. In fact, the generic logistic forms of regional supply functions motivated by Hotelling (1932) and discussed by Johansson (1991) demonstrate this property. In this paper, an additional entropy term recognizing heterogeneity within the available capacity is shown to yield a logistic supply function. This generates a further enhancement to the Wilson framework, as already included in Roy (2004a, b). In addition, the special information theory method of Roy (1987) allows the model here not only to perfectly reproduce the base period flows (as for conventional information theory models), but to be simultaneously responsive to future changes in the input–output coefficients, regional output capacities, freight prices and the transport network itself, with the technological balance relations being imposed anew in the projection time period.

164

10.2

J.R. Roy and G.J.D. Hewings

Multi-regional I–O Model with Endogenous Internal and External Component Flows

In Roy (2004a, b) a deterministic interregional I–O model was developed, with the flows including both internal and external components. The surprising result from this analysis was that the conventional interregional I–O model (the ‘‘Isard’’ model) cannot be made consistent with externally provided final demand, thus undermining its special equilibrium assumptions. However, when we turned to a multi-regional analysis, the fundamental balance relations turned out to be generally consistent with those in the L–S approach, reinforcing the foundations of the latter. In the following development, L–S ideas are used again to develop the balance relations, but the flows are now endogenously determined into their five internal and external components, rather than as merely the aggregation of these five components, as implied by L–S.

10.3

Basic Definitions and I–O Relations

Firstly, we define total aggregated flows x
rs im of sector i on route m (strictly route mrs) between regions r and s in terms of their internal and external components as x
rs im ¼

X  j

 xijmrs þ iijmrs þ yimrs þ eimrs þ iimrs

8imrs:

ð10:1Þ

These pool the conventional intermediate internal flows xijmrs, the intermediate external (import) flows iijmrs of product i entering the regional system at region r and transshipped on route m to region s for sector j, the internal flows yimrs direct to final demand, the external export flows eimrs of product i transshipped on route m from region r to exit the system at region s and the import flows direct to final demand iimrs of product i entering the system at region r and transshipped on route m to region s. Note that, in this form, the sectoral aggregation over the destination sectors of the flows of a good between each pair of regions is a more general form of the supply pool assumption of Leontief–Strout. Then, we represent the overall output technology by defining multi-regional coefficients aijs, denoting the total number of units xijs =S rm xijmrs of internally produced and iijs =S rm iijmrs of externally imported intermediate inputs of sector i going into sector j along routes m from all regions r to region s required to produce a unit of sector j output in the same region s, given as aijs ¼

hX  rm

i xijmrs þ iijmrs =Xj s

8ijs:

ð10:2Þ

10

Regional Input–Output with Endogenous Internal and External Network Flows

165

In order for the intermediate flows to be consistent with multi-regional I–O, they ~rs are defined [see (10.1)] as x~rs im and iim , aggregated over all destination sectors j, via x~rs im ¼

X

rs ~ rs j xijm ; iim

¼

X

iijmrs

8imrs

ð10:3Þ

rs in terms If these are now substituted back into (10.1), we have the total flows x
im of all the component flows which are to be endogenously determined rs rs rs ~rs ~rs x
rs im ¼ x im þ iim þ yim þ eim þ iim

8imrs:

ð10:4Þ

The next step is to express the usage relations of the output Xir of sector i in region r X rs rs Xri ¼ xrs ð10:5Þ sm ð~ im þ yim þ eim Þ 8ir: Here, the usage of the output provides internal intermediate inputs, internal flows direct to final demand and transshipped flows for exports outside the regional system. Before proceeding further, we provide some basic definitions. The total exports Eir of sector i abroad out of region r and total imports Iir of sector i into region r from abroad are expressed in terms of the component flows as Eri ¼ Ssm eimsr ;

Iir ¼ Ssm ðiimrs þ iimrs Þ

8ir:

ð10:6Þ

These denote transshipments2 to exit ports r from producing regions s for the exports, and transshipments from the entry ports r to consuming regions s for the imports, which include components direct to final demand and those to supply intermediate inputs. Also, the final demand Yir satisfied for sector i in region r, including that of its own region and that flowing in from all other regions s, is Yir ¼ Ssm ðyimsr þ iimsr Þ

8ir:

ð10:7Þ

These contain both the internal and externally imported components.

10.4

Fundamental Relationships to Be Satisfied

In this section we look at the base period data, re-establishing the L–S relations in an enhanced form, as well as specifying the row and column sum totals which the base period component flows must satisfy. 2

We could further augment the above flow structure to include outside imports that flow right through our regional system without entering the internal production or consumption systems. Also note that the external flows could be disaggregated into two parts, rest of the country and rest of the world.

166

J.R. Roy and G.J.D. Hewings

Because the demand-driven relations (10.5) relate to usage of the output in terms of outflows, we follow the L–S approach by developing a viable way to eliminate the unknown outputs from the analysis by transposing the basic I–O relations (10.2), reversing the r and s indices and summing over j to yield the following inflow relations of goods i to all sectors j in region r via   Sjsm xijmsr þ iijmsr ¼ Sj aij r Xj r

8ir:

ð10:8Þ

In terms of the multi-regional analysis quantities in (10.2), this yields r r ~sr xsr Ssm ð~ im þ iim Þ ¼ Sj aij Xj

8ir:

ð10:9Þ

The next step is the elimination of the outputs Xjr via (10.5), simply reversing i and j h i r rs rs ~sr xsr xrs Ssm ð~ im þ iim Þ ¼ Sj aij Ssm ð~ jm þ yjm þ ejm Þ

8ir:

ð10:10Þ

These represent the balance relations in terms of endogenous quantities needing to be satisfied to represent the technology. Although (10.10) is derived from the L–S philosophy, the implementation of this analysis in terms of the five component quantities rather than the aggregated flows, yields a form where all terms are endogenous. As such, the relations (10.10) here represent consistency conditions, rather than equations with exogenous right-hand sides, as in L–S. They must be satisfied at both the base period and the projection period, with aijr0 as the base period I–O terms. Looking now at the base period totals which the individual component flows must satisfy, we express the first of (10.6) as a condition for the external base period export totals Eir0 of sector i out of the system at region r in terms of the transshipments, and (10.7) for base period final demand Yir0 for sector i in region r via Ssm eimsr ¼ E r0 i ;

Ssm ðyimsr þ iimsr Þ ¼ Y r0 i

8ir:

ð10:11Þ

Then, we express the external base period imports Iir0 of sector i at their port of entry r in terms of the corresponding intermediate input and final demand transshipment flows via the second of (10.6) as r0 rs Ssm ði~rs im þ iim Þ ¼ I i

8ir:

ð10:12Þ

Let m be the number of sectors, n the number of regions and p the average number of paths between each regional pair. From (10.4) it is seen that we have (5mn2p) unknowns, but only (4mn) in (10.10)–(10.12). This is in strong contrast to the deterministic L–S approach, where the flows in (10.10) are aggregated over s and the number of equations then equals the number of unknowns. Thus, it is necessary to introduce some sort of plausible objective function which can be

10

Regional Input–Output with Endogenous Internal and External Network Flows

167

maximized with relations (10.10)–(10.12) included as constraints. Finally, the model can be further enhanced by the addition of discretionary information in terms of constraints to be satisfied in the base period, as follows.

10.5

Addition of Discretionary Information via Constraints

The form of our entropy objective (see later) implies that out-flow constraints on output of each sector for each region should be omitted. However, we can enrich the model by applying the following output constraints, aggregating the regional outputs Xir0 in (10.5) over either region r or sector i 0 rs rs Srsm ð~ xrs im þ yim þ eim Þ ¼ Xi

8i;

r0 rs rs Sism ð~ xrs im þ yim þ eim Þ ¼ X

8r: ð10:13Þ

A further very interesting enhancement is to equate the total base period output Xs0 in region s in terms of the column-based purchase relations of the I–O table, including the base period value-added quantities Vis0, with the endogenous import flows now disaggregated in the form s0 s0 rs ~rs Sirm ð~ xrs im þ i im þ iim Þ ¼ X  Si Vi

8s:

ð10:14Þ

This constraint provides a connection with the supply-driven approach, even though we retain the demand-driven I–O coefficient matrix A. Finally, the influence of the transport network on the component flows is recognized by inclusion of a transport cost constraint, as an extension of the ideas of Wilson (1970). Let c
0 be the average transport cost (or distance) per value unit of commodity shipped and cimrs0 be the internal transport cost3 per value unit of sector i between regions r and s along path m. We must also include the average extra transport costs cir0 to bring the external imports into their entry port r from outside our regional system and the average extra transshipment costs cis0* to transport the external exports from their exit region (port) s to their final destination outside our regional system. Thus we apply an average total internal and external transport/ transshipment cost constraint over the entire set of flows as rs rs rs rs0 rs r0 ~rs ~rs Sirsm ð~ xrs im þ iim þ yim þ eim þ iim Þcim þ ðiim þ iim Þci j þ eimrs cis0 ¼ c
0 X
0 :

3

ð10:15Þ

In cases where transport costs are a relative small proportion of the total transaction costs, such as for a set of small densely-populated regions, or for cases where the other logistic costs of the transactions are quite high, transport costs and the associated constraint [see (10.15)] can be omitted.

168

J.R. Roy and G.J.D. Hewings

All the constraints (10.10)–(10.15) [(4mn+m+2n+1) conditions] must be simultaneously satisfied at the base period in the following entropy maximization procedure.

10.6

Formulation of the Objective Functions

Armed with the above balance relations (10.10) and the constraint sets (10.11)– (10.15), all in terms of the five unknown sets of quantities, we can now proceed further with development of a probabilistic model. A considerable enhancement of the Wilson (1970) approach is made by ensuring that the flows reflect the current technology by disaggregating the analysis, inserting the enhanced technological balance relations (10.10) and setting logistic constraints on regional output capacity. Whereas the fundamental form of the conventional entropy model in the absence of economic information would yield equal sectoral production in each region, the enhanced model yields equal relative capacity utilization, a more plausible hypothesis. Also, a strong distinction is made between the estimation form of the model and a transformed version for projection, as fully illustrated in Roy (2004a). The first major point is that we will need two separate but linked entropy models. These are necessary because the capacity usage quantities (10.5) to be allocated contain no import terms. Thus, the first entropy term relates to division of the available capacity into that which is utilized and that which is non-utilized, with the former then being subdivided into the individual component flows. The simpler second entropy relates to expressing the total regional imports into their component flows. The key linkage constraints are the balance constraints (10.10) (containing the import flows i~sr im of sector i which are supplied as intermediate inputs to all sectors), the second of (10.11) on final demand (containing the import flows iisr direct to final demand), the column-sum purchase conditions (10.14) on total regional output (containing both the iisr and the i~sr im import terms) and the transport cost constraint (10.15) containing all flow quantities. These constraints provide a rich linkage between the two models. Note that, an alternative procedure would be to maximize a weighted sum of the two entropies, with the weights being determined endogenously, as illustrated by Roy and Lesse (1985). However, we consider it more expedient to iterate between the two models to obtain consistent flows.

10.6.1 Entropy Related to the Productive Capacity Consider that we have base period data, both on the total output capacity X~ir0 of sector i to supply intermediate inputs, final demand and external exports from region r, as well as (optionally) the total in-flow X
s0 of all sectors into regions s. The incorporation of the output capacity in the entropy yields a logistic form of the flow function, consistent with Hotelling (1932), as suggested by Johansson (1991).

10

Regional Input–Output with Endogenous Internal and External Network Flows

169

Also, let X
0 be the total value of all inputs in the system. The number of microstates Z can now be given as the number of ways X~ir0 distinguishable output capacity units may be divided into Xir which are utilized and (X~ir0  Xir ) which remain unutilized, with the former then being allocated to the three categories of flows contained in the usage relations for Xir in (10.5)       rs rs rs rs Z ¼ Pir X~ir0 != X~ir0  Ssm x~rs !: Psm x~rs : ð10:16Þ im þ yim þ eim im !yim !eim ! As usual, P denotes the product sign. Setting the entropy S as the natural log of Z and applying the Stirling approximation, we obtain the entropy maximization objective    rs rs S ¼ Sir X~ir0  Ssm x~rs im þ yim þ eim   r0    rs rs  1  Sirsm log X~i  Ssm x~rs im þ yim þ eim  rs    rs rs rs rs x~im log x~rs im  1 þ yim ðlog yim  1Þ þ eim ðlog eim  1Þ :

ð10:17Þ

rs Applying Lagrangian theory, (10.17) is maximized in terms of x~rs im , yim and rs0 eim under (10.10) with multiplier firl , (10.11) with multipliers air and gir1 respectively, (10.13) with multipliers r and ti respectively, (10.14) with multipliers rs rs ls1 and (10.15) with multiplier b1. Differentiation with respect to x~rs im , yim and eim to and equating to zero gives

 r0  rs     ~im þyimr þeimrs exp ls1 þr þti þb1 cimrs0 þfir1 1Sj aijr0 ; x~rs im ¼ Xi Ssm x       rs rs yim rs ¼ X~ir0 Ssm x~rs exp b1 cimrs0 þr þti þgis1 fir1 Sj aijr0 ; im þyim þeim    rs rs eim rs ¼ X~ir0 Ssm x~rs im þyim þeim n   o r0 exp b1 cimrs0 þcs0 Þþ þt þa f S a 8imrs: i is j r ir1 i ij ð10:18Þ Calling the exponential terms on the right of the three equations (10.18) Aimrs, Bimrs and Dimrs respectively, summing these equations over (sm) and subtracting from X~ir0 , we finally obtain explicit relations for the unknowns as x~imrs ¼ Xir0 Aimrs =½1 þ Ssm ðAimrs þ Bimrs þ Dimrs Þ; yimrs ¼ X~r0 Bimrs =½1 þ Ssm ðAimrs þ Bimrs þ Dimrs Þ; i

ers im

ð10:19Þ

¼ X~ir0 Dimrs =½1 þ Ssm ðAimrs þ Bimrs þ Dimrs Þ 8imrs:

In many ways, it is tempting to estimate the non-linear equations obtained by substituting (10.19) into the constraints entirely by Newton–Raphson iteration. However, we may alternatively use Newton–Raphson for the multipliers b1 and

170

J.R. Roy and G.J.D. Hewings

firl associated with the economic constraints and successive substitution for the pure summation constraints.

10.6.2 Entropy for External Imports As discussed earlier, the values of the intermediate and final demand external import flows need to be determined via a linked entropy model, with rich linkages existing within the constraints. Firstly, note that there is no natural capacity measure for regional external imports such as we had in the previous section for regional productive capacity. Thus, let us consider the number of ways Z0 that the observed external imports Iir0 may be allocated using the definition in (10.6) into the component flows i~rs im to supply intermediate inputs and those iimrs to supply final demand, yielding   rs  rs ð10:20Þ Z0 ¼ Pir I r0 i !=Psm i im !:iim ! : Upon taking the natural log of both sides of (10.20), applying the Stirling approximation and removing constant terms, the external import entropy S0 comes out as     rs rs ~rs S0 ¼ Simrs i~rs ð10:21Þ im log i im  1 þ iim ðlog iim  1Þ : The entropy S0 is to be maximized in terms of the import flows under various constraints, including the basic import constraints (10.12) with multipliers cir. The inflow balance constraints (10.10) are applied with multipliers fir2 , the final demand constraints (10.11) with multipliers gir2, the column sum output constraints (10.14) with multipliers ls2 and the travel cost/distance constraints (10.15) with multiplier b2. Upon differentiation of (10.21) and the nominated constraints with respect to the external import flows and equating to zero, we obtain     þ fir2 iimrs ¼ exp  cir þ ls2 þ b2 cimrs0 þ cr0 i ð10:22Þ     iimrs ¼ exp  cir þ ls2 þ b2 cimrs0 þ cr0 þ gir2 8imrs: i Upon elimination of cir via (10.12), the results are expressed as   rs0     r0 þ cr0 i~rs þ ls2 þ fir2 =Ssm fexp ½b2 cimrs0 þ cr0 im ¼ Ii exp  b2 cim i i   þ ls2 þ fir2  þ exp ½b2 cimrs0 þ cr0 þ gir2 g; i     þ ls2 þ gir2 =Ssm fexp ½b2 cimrs0 þ cr0 iimrs ¼ Iir0 exp ½b2 cimrs0 þ cr0 i i   þ ls2 þ fir2  þ exp ½b2 cimrs0 þ cr0 þ gir2 g 8imrs: ð10:23Þ i As before, (10.23) can be solved for the unknown Lagrange multipliers using Newton–Raphson iteration.

10

Regional Input–Output with Endogenous Internal and External Network Flows

171

10.6.3 Coordination of the Two Models In our combined model, (10.19) and (10.23) are estimated successively, passing the rs rs current estimated values of x~rs im , yim and eim as temporary known values into the rs constraints associated with (10.23), estimating new values of i~rs im and iim from reestimation of (10.23) and substituting these values in turn back into the constraints associated with (10.19), and so on. As both objective functions are strictly concave, convergence would be obtained if the data are consistent. Once all the sets of the five component flows are obtained, substitution into (10.4) yields the total observable multi-regional flows x
rs im . As we satisfy the balance relations (10.10) which insert the usage relations (10.5) into the fundamental input output relations (10.9), the regional outputs Xir can be evaluated directly from (10.5) and should automatically satisfy (10.9) when the Xjr values are substituted, reversing indices i and j. With the flows on alternative paths m between regions r and s being available above, we now examine link flows. Let us define a {0,1} matrix dmars, with entries equal to 1 if link a occurs on path m between regions r and s and zero otherwise. From this we can obtain the total flows for each link a on the network. For example, for the flows x~ rs im of intermediate internal inputs in (10.19), the associated link flows x~ai for sector i on any link a are simply given as x~ai ¼ Srsm dma rs x~ rs im

8ai:

ð10:24Þ

This is a useful output, even if the model does not yet consider congestion.

10.7

Use of Models for Projection

The main task in transforming the estimated models in (10.19) and (10.23) into a form suitable for projection is to choose the Lagrange multipliers which should be treated as parameters in projection, and those which must be evaluated anew. The formalism of adapting the Lagrangian procedure to handle this change is illustrated copiously in Roy (2004a). The main issue is which information should reasonably be treated as exogenous input and which should be treated as endogenous output. In terms of the structure of the model, it is considered that the exogenous input should include any quantity changes, such as changed regional output capacities X~ ri , changed external exports Eir, changed external imports Iir and changed final demand Yir. If transport costs change to cimrs, the availability as a parameter of the gravity Lagrange multipliers b1 and b2 allows the influence of these changes to be assessed by the model. Also, not expecting the total output Xs to region s to be available in the projection period, its Lagrange multipliers ls1 and ls2 should also be treated as parameters. The same applies to the multipliers r and ti on the aggregated output constraints (10.13). As with the classical analysis, any new multiregional input–output coefficients aijr must be provided exogenously. Visually, the

172

J.R. Roy and G.J.D. Hewings

projection equations will be of identical form to the base period relations (10.19) and (10.23). The main differences are: l

l

That the multipliers r, ti, b1, b2, ls1 and ls1 are now knowns rather than unknowns, with their associated right-hand sides now becoming outputs rather than inputs. We now enter potentially new values of the exogenous data, including output capacities X~ ri , final demand Yir, external exports Eir, external imports Iir, unit travel costs cimrs, cir and cis and regional I–O coefficients aijr.

10.8

An Adjustment from Information Theory

As stressed by Batten (1983), the more general information theory approach may give improved prediction ability for flow models of this type. If the goodness of fit of the estimated models is not satisfactory, the projection model is likely to yield improved results if information bias terms are computed and inserted into the models, as demonstrated via a new information theory procedure in Roy (1987). One interesting feature of the current formulation is that we are unlikely to routinely have survey values of the full five sets of component flows – the most we can expect is to have survey values x
rs0 im of the total flows for each sector along each path between each pair of regions. Although this limits the potential power of the procedure in Roy (1987), it still allows the bias terms to be accommodated in a more aggregate sense. For example, if the result of the base period estimation of the 0 0 rs
rs0
rs total flows in (10.3) is given as x
rs im / x im , we define bias terms qim = x im which are normalized and applied as prior probabilities to the entropy of (10.17). For each component flow, such as that of the intermediate internal inputs x~ rs im , this would yield the first of (10.17) in the revised form rs ~r0 rs rs rs rs rs x~rs im ¼ qim Xi Aim =½1 þ Ssm qim ðAim þ Bim þ Dim Þ 00

8imrs:

ð10:25Þ

If we make the same corrections to the other four components in (10.19) and (10.23), then sum the results to the total values x
rs} im in (10.3), we would ensure that rs0
¼ x , that is, the total estimated flows are identical to the total observed flows x
rs} im im in the base period. Such bias factors are then applied to the models when used in projection.

10.8.1 What if We Do Not Have Sector Output Capacity Data? r0 Data on the net sector output capacities X~im for each sector in each region may be quite difficult to acquire. Also, the concept of capacity may be ‘‘fuzzy’’ in situations of high demand, where some dormant or outmoded plants may well be called into

10

Regional Input–Output with Endogenous Internal and External Network Flows

173

emergency production. In such cases, base period constraints should be introduced on the value Xir0 of the observed output of sector i from each origin region r, writing (10.5) in the form   rs rs Ssm x~rs ¼ Xir0 im þ yim þ eim

8ir:

ð10:26Þ

Clearly, these regional sectoral output constraints make the aggregated output constraints on both total sectoral production and total regional production in (10.13) redundant. In addition, the capacity entropy is omitted in the first term in (10.12), which now deals just with the allocation of utilized capacity, and (10.26) is attached with multipliers ir, yielding the results rs0 x~rs þ fir1 ð1  Sj aijr0 Þg; im ¼ exp  fls1 þ  ir þ b1 cim

yimrs ¼ exp  fir þ b1 cimrs0 þ gis1  fir1 ðSj aijr0 Þg;   eimrs0 ¼ exp  fir þ b1 cimrs0 þ cs0 þ ais  fir1 ðSj aijr0 Þg 8imrs: i

ð10:27Þ

This model is estimated rather similarly to the capacity-constrained model in (10.19) and can be solved via a combination of the Newton–Raphson method and iterative adjustments. With the total output Xs to region s being not available as input in the projection period, its Lagrange multiplier ls1 in the first of (10.27) must be treated as a parameter in projection. However, constraints (10.11) are applied anew in the projection period, with new export totals Eir and final demand quantities Yir being required as input in the projection period. Thus, their associated Lagrange multipliers are unknowns rather than parameters, and can be eliminated from the second and third relations in (10.27), yielding   rs r yimrs ¼ Yis  S rm i~rs im exp  fir þ b1 cim  fir1 ðSj aij Þg =½Srm exp  fir þ b1 cimrs  fir1 ðSj aijr Þg; eimrs ¼ Esi Fsi exp  fir þ b1 cimrs  fir1 ðSj aijr Þg =½Fsi fSrm exp  fir þ b1 cimrs  fir1 ðSj aijr Þg

ð10:28Þ 8imrs

in which Fis =expcis* and firl are the new Lagrange multipliers on the L–S balance constraints (10.10) when estimated with revised I–O coefficients in the projection period and the intermediate import flows i~rs im are passed up iteratively from the linked import model in (10.23). Although the necessity of (10.27) to satisfy the balance relations (10.10) plus the export and final demand constraints (10.11) in the projection period induces it to remain non-separable, the logistic form (10.19) has a more complex interdependency structure and should be used where reasonable capacity data can be found. It’s important to realize that the multipliers ir are treated as known parameters in projection, allowing the final output Xir to emerge endogenously via direct substitution of the new flows into (10.5). This biases the flows to take the observed base period output into account in projecting final output via (10.5). Thus, this class of

174

J.R. Roy and G.J.D. Hewings

projection model shares some of the properties of the general information theory approach of Batten (1983), whilst allowing the transport costs to be explicitly represented and changed for the projection period. Note that, as (10.26) contains no unknown external import terms, our companion import model (10.23) remains unaffected in structure. Of course, as with the previous model with logistic capacity constraints, it still needs to be estimated iteratively with the new form in (10.27) and (10.28).

10.9

Conclusions

The approaches presented above integrate technology change, output capacity change, changes in regional final demand, endogenous determination of transshipped external export flows to and external import flows from outside the regional system, as well as transport network and cost changes, in the evaluation of multiregional flows. Thus, they are well suited to analysis of dynamic regions, where external flows are important. As probabilistic models, they fit parameters to observations, rather than relying on deterministic optimization. The aggregation from an interregional approach adopts the same pooling assumptions as Leontief–Strout. The logistic form of the supply relationships promotes spillovers into adjacent regions when there is both a vintage distribution of capacity and a high pressure on this capacity in a given region. These models do not attempt to project changes in technology. It is the user’s responsibility to provide any changed I–O coefficients aijr. The key advance with respect to Roy (2004a, b) is the identification and evaluation of the five sets of component flows, including regional flows to satisfy internal intermediate inputs, internal regional flows direct to final demand, as well as transshipment flows, including external export flows, external import flows to provide internal intermediate inputs and external import flows to go directly to internal final demand. Also, one of the constraints relates to the column sums of the I–O table, providing a linkage with supply-driven approaches. As discussed by Roy (2004a) in his critique of the earlier papers, for any route between regions, we can now evaluate the mix of the five component flows which are present. This can be further specialized to evaluation of the component flows on all individual links on the network. For example, we can see which links are especially vulnerable in a situation of changing external imports or exports. In many ways, the above method may appear to resemble a commodity flow model. The main differences are that technology is included by the use of the L–S balance relations (10.10) and we follow the I–O method in which final demand is provided exogenously and we evaluate the values of all regional outputs plus the associated component flows. Much empirical work remains to be done to demonstrate the relevance of the proposed model. In addition, a probabilistic RAS type approach should be devised to determine the full set of component flows, extending the analysis in Roy (2004a),

10

Regional Input–Output with Endogenous Internal and External Network Flows

175

Sect.10.7.2, where changes in technology are inferred from changes in the regional outputs. A further challenge, if data is available, is to include the above ideas within a commodity by sector framework. This is especially important for the above models, where transport costs are included explicitly. Within a sector producing several commodities, the unit transport costs for each commodity can vary quite markedly, and are just averaged out in the sector by sector framework above.

References Batten DF (1983) Spatial analysis of interacting economies. Kluwer, Boston Chenery H (1953) Regional analysis. In: Chenery H, Clark P, Pinna V (eds) The structure and growth of the Italian economy. US Mutual Security Agency, Rome Cole S (1997) Closure in Cole’s reformulated Leontief model: a response to R.W. Jackson, M Madden and H. A. Bowman. Pap Reg Sci 76:29–42 Hewings GJD (1985) Regional input–output analysis. Sage, Beverly Hills, CA Hitomi K, Okuyama Y, Hewings GJD, Sonis M (2000) The role of interregional trade in generating change in the interregional Leontief inverse in Japan, 1980–1990. Econ Syst Res 12: 515–537 Hotelling H (1932) Edgeworth’s taxation paradox and the nature of supply and demand functions. J Poli Econ 40:577–616 Isard W et al. (1960) Interregional and regional input–output techniques. In: Isard W et al. (ed) Methods of regional analysis, chap.8. Wiley, New York Isard W et al. (1998) Methods of interregional and regional analysis. Ashgate, Brookfield, VT Jackson RW, Madden M (1999) Closing the case on closure in Cole’s model. Pap Reg Sci 78: 423–427 Johansson B (1991) Regional industrial analysis and vintage dynamics. Ann Reg Sci 23:1–18 Kim E, Hewings GJD, Hong C (2004) An application of an integrated transport network – multiregional CGE Model I: a framework for economic analysis of highway projects. Econ Syst Res 16:235–258 Leontief W, Strout A (1963) Multi-regional input–output analysis. In: Barna T (ed) Structural interdependence and economic development. Macmillan, London Miller RE, Blair PD (1985) Input–output analysis: foundations and extensions. Prentice-Hall, Englewood Cliffs, NJ Oosterhaven J (1988) On the plausibility of the supply-driven input–output model. J Reg Sci 28:203–217 Oosterhaven J (2000) Lessons from the debate on Cole’s model closure. Pap Reg Sci 79:233–242 Polenske KR (1980) The U.S. multiregional input–output accounts and model. Lexington Books, Lexington, MA Roy JR (1987) An alternative information theory approach for modelling spatial interaction. Environ Plann A 19:385–394 Roy JR (2004a) Spatial interaction modelling: a regional science context. Advances in spatial science series. Springer, Heidelberg Roy JR (2004b) Regional input–output analysis and uncertainty. Ann Reg Sci 38:397–412 Roy JR, Lesse PF (1985) Entropy models with entropy constraints on aggregated events. Environ Plann A 17:1669–1674 Smith TE (1990) Most-probable-state analysis: a method for testing probabilistic theories of population behaviour. In: Chatterji M, Kuenne RE (eds) New frontiers in regional science. MacMillan, London, pp 75–94

176

J.R. Roy and G.J.D. Hewings

Snickars F, Weibull J (1977) A minimum information principle: theory and practice. Reg Sci Urban Econ 7:137–168 Sohn J, Hewings GJD, Kim TJ, Lee JS, Jang SG (2004) Analysis of economic impacts of an earthquake on transportation networks. In: Okuyama Y, Chang S (eds) Modeling spatial and economic impacts of disasters. Springer, Heidelberg Wilson AG (1970) Entropy in urban and regional modelling. Pion, London

Chapter 11

Regional Unemployment and Welfare Effects of the EU Transport Policies: Recent Results from an Applied General Equilibrium Model Artem Korzhenevych and Johannes Bro¨cker

11.1

Introduction

The spatial effects of the European infrastructure policy, especially in relation with the cohesion objective, constitute a research question of particular interest for the policymakers, as can, for instance, be inferred from the significant number of largescale research projects that incorporate this issue. These projects have a general focus on the EU transport policy, but differ in research questions and the indicators required to be produced by the employed models. The common feature is that the actual policy measures, such as building the Trans-European Networks, revitalizing railways, implementing effective road pricing, etc., are first transferred into an economically treatable indicator of interregional transport costs. An economic model is then needed that can explicitly incorporate this information and calculate the policy impact on regional welfare, its efficiency and equity effects. A successful model capable of taking account of changing transport infrastructure is CGEurope model introduced in Bro¨cker (1998). This is a multiregional (up to 1,400 regions covering the whole world) computable general equilibrium model constructed specifically for a transport policy analysis. Following the recent demands from the policymakers in the field of European infrastructure development, several extensions of the model are currently being introduced. It is a well-known fact that inside the expanding European Union some people live in areas of high unemployment, while others are surrounded by little joblessness. One of the biggest political concerns about complex effects generated by implementing certain measures in transport is whether these measures can improve welfare and stimulate employment. In fact, as unemployment rates vary strongly across regions in Europe, the regional (un)employment impacts can be considerable and in some cases undesirable. So, one of the extensions of the CGEurope model

A. Korzhenevych (*) Institute for Regional Research, Kiel University e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_11, # Springer‐Verlag Berlin Heidelberg 2009

177

A. Korzhenevych and J. Bro¨cker

178

aims at estimating regional (un)employment effects resulting from policy scenarios in transport development. The theoretical background and modeling results are presented in this paper. The results are then analyzed to check for consistency with the cohesion objective. The paper is organized as follows. The next section presents the structure of the basic general equilibrium model upon which the analysis of transport policy is based. The basic model assumes flexible prices and full factor employment. In Sect. 11.3, we present the perfect labor market framework, introduce employment response to policy scenario, and allow some degree of factor mobility (but not labor mobility). Welfare measurement issues that arise from introduction of factor mobility are discussed in Sect. 11.4. The calibration procedure and the results of numerical experiments are reported in Sects. 11.5 and 11.6. Finally, Sect. 11.7 concludes the paper briefly.

11.2

The Basic Model

We provide only a verbal description of the basic model here (see Bro¨cker 2002 for a full formal description). CGEurope is a spatial general equilibrium model for a closed system of regions covering the whole world. All the regions are treated separately and are linked through endogenous trade. The results here will be presented for 270 NUTS2 regions,1 but in some projects NUTS3 level of detail is used. The inference method is comparative statics, which means that in each model run, two equilibria (benchmark and scenario) are compared. The basis for comparison is a generally understandable indicator, like real income or real GDP. In every region, a set of households is assumed to dwell, owning a bundle of immobile production factors, which is used by regional firms for production of goods. We distinguish between two types of goods, local and tradable. Local goods can only be sold within the region of production, while tradables are sold everywhere in the world, including the own region. Given the main task of assessing total employment response to transport policy measures, introduction of more sectoral detail would lead to excessive difficulties. Producers of local goods combine primary factor services, local goods and tradables, using Cobb–Douglas technology with region-specific cost share parameters. The output of locals is assumed to be completely homogeneous, and is produced under constant returns to scale. Firms take prices for inputs as well as output as given, and do not make any profits. Instead of directly selling this output to households or other producers, firms can use it as the only input needed to produce tradables. The respective technology is subject to increasing returns to scale. Tradable goods are modeled as being close but imperfect substitutes, following the Dixit and Stiglitz (1977) approach. Different goods stem from producers in different regions. Therefore, relative prices of tradables do play a role. Changes of exogenous variables (transportation costs) 1

NUTS stands for ‘‘Nomenclature of territorial units for statistics’’.

11

Regional Unemployment and Welfare Effects

179

make these relative prices change and induce substitution effects. For producers of tradables, only input prices are given, while the output price can be set under the framework of monopolistic mark-up pricing. Due to free market entry, however, profits are driven to zero, as they are in the market for locals. Households are assumed to act as utility maximizers, taking all prices as given. Utility emerges from consumption (in fixed income shares) of local goods and a CES-composite of tradables, consisting of all, regionally produced and imported variants. Utility is modeled such that households appreciate a higher number of variants of tradables. The same income spent on more diverse variants means higher utility for the households. In other words, they share the ‘‘love for variety’’. For the sake of simplicity, all components of final demand, that is private and public consumption and investment, are subsumed under household demand. There is no explicit consideration of a separate public sector. As a consequence of perfect price flexibility, the regional factor supply is always fully employed (this assumption to be relaxed in the next section). Apart from factor income, disposable income contains a positive or negative net transfer payment from the rest of the world, depending on whether the regional current account with respect to all other regions has a surplus or a deficit in the benchmark situation. These transfers are held constant (in real terms) in our simulations. They are negligible with regard to quantitative results, but are needed for keeping budget constraints closed. Two features that give the CGEurope model its spatial dimension are: l l

The distinction of goods, factors, firms and households by location The explicit incorporation of transaction costs for goods, depending on geography as well as national segmentation of markets

The term ‘‘transaction cost’’ for interregional trade is used to represent any kind of trade-related costs. Usually trade costs are assumed to depend on the quantity of goods traded. However, some costs of interregional transfer, especially costs of information exchange and insurance costs, depend on the value rather than the quantity traded. Letting trading costs depend on the value of trade makes the model much simpler. We therefore assume that the transaction costs for goods to be delivered from region r to region s amount to a share trs
1 in the traded value, trs 1 denoting the transport cost mark-up factor. We introduce two kinds of trade costs: costs related to geographic distance (transport costs), and costs for overcoming impediments to international trade. The first are modeled under the assumption that transport costs are increasing with distance but at a diminishing rate. The change of these costs will constitute the policy scenario. The values for the second type of costs are calibrated within the model and include not only tariffs, but also, and more importantly, all costs stemming from non-tariff barriers, like costs due to language differences, costs for bureaucratic impediments, time costs spent at border controls and so forth. Transport costs are modeled according to the modified iceberg assumption,2 which implies that the value of transported good in the origin and the destination are

2

See Sect. A.4 in Appendix.

180

A. Korzhenevych and J. Bro¨cker

the same. Alternatively, one could model a zero profit transport sector providing the transport service and using the composite tradable good as an input. From a theoretical point of view, the model relies strongly on Shoven and Whalley (1984) approach to general equilibrium modeling, even though it does not assume perfect competition on all markets. In doing so we follow, Venables and Gasiorek (1999), who show that the estimated welfare impact of transaction cost reductions can change dramatically with a deviation from the perfect competition assumption.

11.3

Introducing Wage Rigidity and Factor Mobility

Two stylized facts about the European economies that are not reflected by the basic model, but are of ever growing concern to the policymakers are unemployment and factor mobility. In this section we suggest a way to incorporate these facts into the spatial CGE model. In the basic model labor effort is a part of the regional composite factor service and is assumed to be fully-employed, labor market being cleared by a flexible wage. Of course, this is an extremely unrealistic representation of European regions. A more realistic approach is to introduce imperfect factor markets to reflect stylized facts in most industrialized countries. In order to allow for a gap between supply and demand on the labor market, we need to relax a corresponding assumption of the neoclassical Walrasian model. We have to assume that adjustments are constrained by a certain degree of wage rigidity. A possible measure of wage stickiness was proposed by Blanchflower and Oswald (1994). In their study the empirical responsiveness of workers’ remuneration to the state of the labor market is captured by the coefficient on log unemployment rate in an equation for log real earnings. A way to provide an intellectual rationale for the wage curve is by appealing to the efficiency wage theory. The wellknown characteristic of efficiency wage analysis is that firms set pay in an environment where the wage influences productivity. Shapiro and Stiglitz (1984) is an archetypal case. In equilibrium, firms try to maximize profits and workers choose how hard to work. If the costs of shirking at work are low, employees put in little effort. The outside rate of unemployment plays a role, because it determines the ease with which a fired worker can get another job. In a highly depressed labor market, employees are frightened of losing their jobs, and so put in high effort even if pay is comparatively low. Put differently, a marginal rise in unemployment leads to a corresponding marginal fall in the level of wages. The reason is that firms can reduce pay slightly while maintaining a motivated workforce. Unemployment is a discipline device: when it is high the generosity of workers remuneration can be low. Hence there is an efficiency wage interpretation of a negative dependence between wages and unemployment. Following Blanchflower and Oswald (1994), we replace the conventional labor supply curve with a regional wage-fixing curve:

11

Regional Unemployment and Welfare Effects

181

wr ¼ dr uzr r ; pr where in the left-hand side expression regional wage wr is corrected for regional price index pr to give the real wage; dr is the shift coefficient, corresponding to the regional dummy in the wage curve regression; ur is the regional unemployment rate, and zr is the unemployment elasticity of pay. The country-level values of zr are taken from the rich wage curve literature. Table 11.2 in Appendix B summarizes the results of our literature survey. The values range from
0.52 for Latvia to
0.02 for Denmark. However, a lot of estimates lie close to
0.1, a remarkable similarity discovered by Blanchflower and Oswald. The wage curve tends to be flatter in developed countries with highly regulated contract systems, while allowing for more wage adjustment in the former socialistic countries. To see how the infrastructure policy imposes response in the regional labor market, consider Fig. 11.1. In this example, improved accessibility increases demand for region’s output, causing regional factor demands to increase. Labor demand schedule shifts outwards, and the equilibrium moves from point A to point B. In contrast to fixed employment assumption, in this case, only a part of adjustment is done by wages, the degree of rigidity set by the exponent appearing in the wage curve equation. The rest of adjustment is accomplished by the increase in employment level.

Real wage

Wage curve

Labour demand

B A

Employment

Fig. 11.1 Labor market response

Full

A. Korzhenevych and J. Bro¨cker

182

Recent examples of application of wage curve equation in CGE modelling are papers by Roson (1997) on multisectoral model of Italy, and Annabi (2003) on theoretical modeling of different labor market regimes. Another stylized fact about European economies, not reflected by the basic model, is a certain degree of factor mobility. The theoretical models allowing for commodity movements (trade) as well as some degree of factor mobility date back to a seminal paper by Mundell (1957). Today there exist several widely used methods to account for international capital flows while staying in static CGE modeling. Shoven and Whalley (Shoven and Whalley, 1992) uses an exogenous capital ownership scheme, through which regions get income from the factors located abroad. Hertel (1997) introduces ‘‘investment good’’ which enters regional utility and production functions, and is collected and redistributed by the ‘‘global bank’’. This approach is taken by the researchers using GTAP model. Decaluwe et al. (2004) essentially follow the Ricardian approach by assuming that movement of physical capital is governed by differences in international rates of return, and that total capital stock is fixed. A second innovation, in comparison with the basic model, is to assume that a certain fraction of regional factor stock is mobile internationally. It can move (following higher return) to other regions. We assume that there is a single rental clearing the market for this homogeneous factor. The total stock of mobile factor is assumed to be fixed. The share of this mobile factor will be approximated by the regional capital share, for the lack of other reliable information. Because we allow the amount of mobile factor employed in the region,Kre , to differ from the factor stock owned by the inhabitants of the same region, Kr, we need to take account of international income flows. Thus, when calculating household income, besides the fixed transfer, the amount of ðKr
Kre Þrmust be added to regional GDP Yr (GDP being equal to employed factors income in the absence of taxes). Factor owners themselves, however, are assumed not to move between regions. For the lack of other information, Kr and Kre are assumed to coincide in the benchmark equilibrium. Full formulation of this extended model is given in Appendix A.

11.4

Welfare Measurement

Having introduced an internationally mobile factor, a proper regional welfare measure needs to be defined. The fundamental problem is how to treat the income of domestically owned factors employed abroad. From the point of view of the household owning some factors abroad, it is still a part of his personal income. However, a policymaker in the respective region may well regard this factor outflow as a pure loss if the capitalist moves away together with his ownership. The difference between these two welfare evaluations will of course depend on the amount of factors actually moving internationally. We choose to assume that

11

Regional Unemployment and Welfare Effects

183

income from factors employed abroad is repatriated to the region of ownership, and that capitalists do not move between regions. As the welfare effects of transport cost change are expected to be small, we will use the change of log real income between situations a (scenario) and b (benchmark) to approximate the regional welfare effect: DRr ¼ logðNra =Nrb Þ
logð par = pbr Þ; where nominal income is denoted by Nr, and pr is the same price index as in the wage curve equation. We assume that no additional welfare gains or losses arise due to change in unemployment rate. Most importantly, we assume that there is no considerable welfare loss associated with moving from unemployment to employment (e.g. potential leisure loss).

11.5

Data Sources and Model Calibration

The comparative statics experiment in this paper will be carried out for the year 2020. To describe the future benchmark economic situation we need a consistent database for a recent year, as well as the prediction of relevant growth rates for the period until 2020. The source of national accounts and international trade data for 34 European countries is the GTAP Version 6 Data Package, with base year 2001. This source ensures consistency between data on value-added structure, GDP, and trade structure. Information on regional level is taken from REGIO database of EUROSTAT with the same base year. The important indicators taken from this source are regional GDP, area, population, and unemployment rate. Data on transport distances between all the regions in order to calibrate interregional trade matrix are also needed. These data are calculations of transport costs (in minutes), based on the network database of Spiekermann and Wegener (used in IASON project, see Bro¨cker et al. (2004)), which contains data for all major links in Europe, including their specific characteristics of speed limits and likelihood of congestion. An important piece of data that is not directly available and must be calibrated is the matrix of mark-up factors for international trade. The mark-up factors represent impediments additional to distance-related costs in international trade. These impediments are specific for each country pair and are determined during the calibration of the interregional trade matrix based on the 2001 national accounts data. They are calculated such that the observed international trade flows equal the corresponding aggregates of trade flows between the regions of the two respective countries. Next, the assumption about the development of these calibrated mark-up factors for the period until 2020 needs to be made. In the current situation these impediments

184

A. Korzhenevych and J. Bro¨cker

to trade are considerably lower for the EU15 countries than for the new member states, but it is assumed that after the accession to the EU the new member states will gradually catch-up and their trade integration will increase. It is therefore assumed that the impediments to trade within the group of new member states and between EU15 and the new member states in our scenario year 2020 will on average be equal to the average of the impediment to trade within EU15 in 2001. After this procedure, the model is ready to be calibrated for the scenario year. For calibrating the model for the year 2020, GDP and population growth projections by the European Commission are used for the EU countries, and the growth rate estimates from the PRIMES project Mantzos and Capros (2006) for the non-EU countries.

11.6

Experiment Description and Main Results

In our experiment for year 2020, all projects of the TEN/TINA programme (see Fig. 11.8 and ESPON 2004) for road and rail plus some national infrastructure development will be realized. The scenario also includes the motorways of the sea which are treated as a part of the road network for freight transport. This scenario predicts the regional economic impacts of a full realization of the trans-European network programme plus corresponding infrastructure in Norway and Switzerland. Figure 11.4 (Appendix C) shows the basic model prediction for the spatial welfare changes of the regional households, generated by installing the respective links and thereby reducing the costs for transferring goods between regions. Utility changes of household come from the production side and the consumption side. On the production side, a better access of firms to input and output markets for tradables increases the factor return, thus bringing a higher income to the household. On the consumption side, a better access to the supply of tradables reduces prices and increases the available product diversity. The utility need not increase everywhere, however; a region making only little use of a new link but trading intensively with those regions making much use of it could suffer from the fact that demand for its output will shift away to other, more accessible places. Our empirical results confirm that these effects do exist (in the regions outside the EU), but they are generally small. Introduction of the factor mobility does not seem to change the welfare effects in an important way (see Fig. 11.5). However, important distinction must be made between the effects on real income (welfare) and real GDP (factor income) in the region. In the basic model these are virtually the same (see Fig. 11.2), because the international income transfers are fixed in real terms. In case of the model with mobile factor, however, regional income contains an additional amount ofðKr
Kre Þr, which is not fixed. If the amounts of mobile factor moving internationally are small (as is the case in our experiment), the loss of some domestic factor income and the gain from foreign factor ownership roughly cancel

11

Regional Unemployment and Welfare Effects

185

Fig. 11.2 Comparison of welfare and GDP effects predicted by different model versions

out, and the welfare response stays close to the prediction of the basic model (correlation 99.9%, slope equal to 1.03). The GDP response, however, must be different, reflecting the effect of factor movements. The size of this effect depends on the underlying share of mobile factor income in GDP. Further extending the model to allow for labor market imperfection also appears to be important for welfare effects prediction (see Fig. 11.6). The range of predicted values expands notably in comparison to the basic model, and thus also in comparison to the model with factor mobility only. The estimated slope coefficient is equal to 2.5. Quite importantly, the effects on real income and real GDP are related in a similar way, as in the case of the model with factor mobility only (see Fig. 11.3). The strongest positive results are predicted for the south of Italy (Messina bridge project), with unemployment rate expected to decrease by 6 percentage points until year 2020. This number is an outlier, next biggest value being two times smaller. Strong positive results are also predicted for regions on the Czech–Slovak border, Ireland and southern Sweden. The reason is the introduction of a lot of new infrastructure in these regions. Average unemployment rate response for the study area is
0.3 percentage points (see Fig. 11.7). Average response in the new member states (including Romania and Bulgaria) is stronger than the EU27-average. The negative employment effects are predicted for the zones outside continental Europe. As it should be expected from introduction of wage rigidity, the response of wages (+0.2% on average) is smaller, compared to the results of basic model (+0.4% on average). The results of the extended model run are summarized in Table 11.1. An issue of constant political concern in the EU is the one of maintaining the cohesion. A policy that creates additional welfare in relatively rich regions at the cost of poorer regions would be violating this objective. In our case, correlation analysis suggests that there is no evidence of EU transport policies violating the cohesion objective. For the extended model, the correlation between benchmark

A. Korzhenevych and J. Bro¨cker

186

Fig. 11.3 Welfare and GDP effects from the model with wage rigidity

Table 11.1 Summary of results: model with wage rigidity Indicator Max Mean (EU27) Welfare response (%) 4.8 0.39 Real GDP response (%) 9.9 0.58 Real wage response (%) 2.0 0.21 Employment response (%) 6.2 0.31

Mean (NMS12) 0.58 1.00 0.36 0.51

Min 0.01
12
0.11 0.03

GDP per capita and percentage increase of employment level is equal to
20%. The correlation coefficient between the benchmark GDP per capita and welfare change due to policy implementation is around
0.22 for all three models. Approximately equal correlation coefficients are found if we substitute real income effects with effects on real GDP. This is obvious, given strong correlation between real GDP and real income effects, as presented in the figures above. These results suggest that the implementation of the analyzed policy package slightly improves cohesion.

11.7

Conclusions

In this paper regional employment response to infrastructure development in Europe, as well as a certain degree of factor mobility are introduced. Labor market is modeled by specifying a wage curve for each region in the study area. The results of our numerical experiment suggest that introduction of mobile factor creates a wedge between welfare and real GDP effects, if the income of factor employed abroad is repatriated to the region of origin. The size of the wedge depends on the share of mobile factor in the total value added. Introduction of wage rigidity using

11

Regional Unemployment and Welfare Effects

187

the elasticity estimates from the wage curve literature scales the predicted welfare effects in our experiment upwards by a factor of 2.5. We interpret this value as a sign of importance of imperfect labor market modeling for applied policy analysis. A weak evidence of trans-European network programme favoring cohesion with regard to real income, real GDP and employment level is also found.

Annex A: Formal Description of the General Equilibrium Model Households Preferences are represented by a two-level utility function. The upper nest is of Cobb–Douglas form and determines the choice between the local goods and the composite of all tradable goods, their prices in region r being pr and qr , respectively. The values of country-specific shares of local goods (services) er are taken from GTAP database. The lower nest is a symmetric CES function combining all tradables. The elasticity of substitution s in this nest is assumed to be equal to 12 (see Bro¨cker (2002) for the discussion of this point). Households get their income Nr from two sources. First, they receive all factor income from domestically owned factors, which includes the income from factors employed abroad. Second, each region gets a transfer Gr . In this way we represent the fact that the trade account may be unbalanced. We assume that the surplus countries pay a transfer equal to their benchmark trade surplus to deficit countries, such that each deficit country receives an amount equal to its trade deficit. Transfers are distributed among regions in proportion to their GDP’s Yr . These transfers are kept constant in real terms. For defining transfers in real terms we need a price index. The natural regional r price index corresponding to the CD utility of households is pr ¼ perr q1
e . As an r overall price index we use the weighted average of pr over all regions, the weights being the base year regional incomes. Fixing transfers in real terms is the same as fixing them in nominal terms and scaling prices such that the overall price index remains unchanged.

Firms In the local goods sector the representative firm of region r operates with constant returns-to-scale technology, combining local goods, tradable goods, and primary factors. The upper nest of a two-level production function is Cobb–Douglas with shares of primary factors, local goods, and tradables being ar , br and gr . The output price pr , which equals the minimal unit cost, is then

A. Korzhenevych and J. Bro¨cker

188

pr ¼

1 ar br gr v p q ; mr r r r

mr , being the level of regional productivity, and vr denoting the composite factor price. Introducing new parameters, r ¼

ar ar þ g r

and solving for pr yields r: pr ¼ pr vr r q1
 r

The lower symmetric CES nest for tradables is assumed to be the same as the one for households, with elasticity s. It aggregates the large number of tradables to a single composite tradable with price qr . The primary factors lower nest is Cobb–Douglas with shares of immobile factors (excluding labor), mobile factors, and labor equal to wr , kr , and lr , respectively: vr ¼ lwr r r kr wlr r : where lr, r, and wr are corresponding factor prices, to be discussed in the next section.

Factor Markets The immobile factor is assumed to be fully employed and its market cleared by a flexible rental lr. As the corresponding supply Tr is fixed, the Cobb–Douglas structure implies the following equilibrium condition: Tr ¼

wr Yr : lr

The homogeneous mobile factor market is cleared by the single rental r, which is equalized internationally. The equilibrium condition follows from the assumption of a fixed total factor stock: X r

Kre ¼

X r

Kr :

11

Regional Unemployment and Welfare Effects

189

Equilibrium in the labor market is defined by the intersection of the wage curve and the labor demand curve. Inserting the values of employment Er and fixed labor force Lr into the wage curve equation of Sect. 11.3, we get  zr wr ¼ dr 1
Er=Lr : pr On the other hand, given the Cobb–Douglas functional form, the wage rate is wr ¼

lr Y r : Er

These two equations define the equilibrium employment level and the wage rate.

Trade Costs and the Market for Tradables Benchmark trade costs are made up of costs related to geographic distance, and costs for overcoming international trade impediments (tariff and non-tariff barriers). If region r belongs to country k and region s to country l, the trade costs markup factor is trs ¼ f ðgrs Þdkl : Here dkl  1 denotes international trade impediments, grs is transport distance, and f is the transport cost function, f ð0Þ ¼ 1. Following Bro¨cker (1998) we assume a combination of exponential and power function for f: f ðgrs Þ ¼ exp½xgrso 

 < 1: with x > 0 and 0 < o

The parameter choice ensures concavity for the entire range of grs values we are  working with. We use estimates of x¼0.03 and o¼0.6 from Bro¨cker (1998). In a general equilibrium one must specify where trade costs are going. A way to dispense with explicit introduction of transport sector is to use the ‘‘icebergassumption’’, as in Samuelson (1983). According to this method, a certain share of transported good itself is used up during transportation. In our ‘‘modified iceberg’’ approach, however, a composite of all tradables is used up for transporting every single tradable good. The value of composite goods used up during transportation equals the trade cost. We neglect tariff receipts (which are only present outside the EU), for the sake of simplicity. As shown in Bro¨cker (2002), incorporating transport costs and assuming monopolistic competition as in Dixit and Stiglitz (1977) in the tradables sector implies the following expression for the tradables price:

A. Korzhenevych and J. Bro¨cker

190

qr ¼ c

X

1 !1
s

1
s Ss p
s s tsr

:

s

Here, Sr is the value of tradables supply, which is calculated as total output minus total demand for local goods and is given by Sr ¼

1 Yr
er Nr : r

Corresponding regional demand for tradables, made up by final and intermediate demand, is given by 

   1
1 Yr ¼ Sr þ Gr þ Kr
Kre r: Dr ¼ ð1
er ÞNr þ r The expression for the value of trade from s to r, valued at mill prices is derived as Ss ðps tsr Þ
s tsr ¼ P
s Dr : s Ss ðps tsr Þ

ð11:1Þ

Finally, the equilibrium condition requires equality of supply of tradables and demand for tradables stemming from a region r, both valued at mill prices: Sr ¼

X

trs :

ð11:2Þ

s

Annex B: Results of Literature Survey Table 11.2 Available estimates of the unemployment elasticity of pay for the study area Country Source 1 Austria Blanchflower and Oswald (2000) 2 Belgium Janssens and Konings (1998) 3 Denmark Nicolaisen and Tranaes (1996) 4 Finland Pekkarinen (2001) 5 France Montuenga et al. (2003) 6 Germany Blien (2003) 7 Great Britain Bell et al. (2002) 8 Greece No estimates available 9 Ireland Blanchflower and Oswald (2000) 10 Italy Montuenga et al. (2003) 11 Luxembourg No estimates available 12 Netherlands Blanchflower and Oswald (2000) 13 Portugal Montuenga et al. (2003)

Value
0.12
0.05
0.02
0.04
0.18
0.08
0.12
0.36
0.07
0.17
0.06

(continued)

11

Regional Unemployment and Welfare Effects

191

Table 11.2 Continued Country

Source

Value

14 Spain 15 Sweden 16 Cyprus 17 Czech Republic 18 Hungary 19 Malta 20 Poland 21 Slovakia 22 Slovenia 23 Estonia 24 Latvia 25 Lithuania 26 Albania 27 Romania 28 Bulgaria 29 Croatia 30 Switzerland 31 Norway 32 Iceland 33 Russia 34 Turkey

Montuenga et al. (2003) Blanchflower and Oswald (1994) No estimates available Blanchflower (2001) Blanchflower (2001) No estimates available Blanchflower (2001) Blanchflower (2001) Simoncic and Pfajfar (2004) Blanchflower (2001) Blanchflower (2001) No estimates available No estimates available Kallai and Traistaru (2001) Blanchflower (2001) No estimates available Blanchflower and Oswald (2000) Johansen (2002) No estimates available Blanchflower (2001) Ilkkariacan and Raziye (2003)


0.11
0.06

Annex C: Modelling Results

Fig. 11.4 Prediction of welfare effects for 2020 by the basic model


0.02
0.11
0.13
0.05
0.10
0.29
0.52
0.13
0.21
0.12
0.02
0.18
0.07

192

A. Korzhenevych and J. Bro¨cker

Fig. 11.5 Prediction of welfare effects for 2020 by the model with mobile factor

Fig. 11.6 Prediction of welfare effects for 2020 by the model with rigid wages

11

Regional Unemployment and Welfare Effects

Fig. 11.7 Prediction of unemployment effects 2020 by the model with rigid wages

193

194

Fig. 11.8 TEN and TINA rail and road projects

A. Korzhenevych and J. Bro¨cker

11

Regional Unemployment and Welfare Effects

195

References Annabi N (2003) Modeling labour markets in CGE models. Universite Laval, Quebec Bell B, Nickell S, Quintini G (2002) Wage equations, wage curves and all that. Labour Econ 9:341–360 Blanchflower DG (2001) Unemployment, well-being and wage curves in Eastern and Central Europe. J Japanese Int Econ 15(4):364–402 Blanchflower DG, Oswald AJ (1994) The wage curve. MIT, Cambridge Blanchflower DG, Oswald AJ (2000) International wage curves. In: Freeman R, Katz L (eds) Differences and changes in wage structures. University of Chicago Press and NBER, Chicago Blien U (2003) The wage curve: effects of regional unemployment on the wage level. Mitteilungen aus der Arbeitsmarkt- und Berufsforschung 36(4):439–460 Bro¨cker J (1998) Spatial effects of transeuropean networks: preliminary results from a spatial computable general equilibrium analysis. Technical University Dresden, Dresden (mimeo) Bro¨cker J (2002) Spatial effects of European transport policy: a CGE approach. In: Hewings GJD et al (eds) Trade, networks and hierarchies. Springer, New York Bro¨cker J, Meyer R, Schneekloth N, Schu¨rmann C, Spiekermann K, Wegener M (2004) Modelling the socio-economic and spatial impacts of EU transport policy. IASON (Integrated Appraisal of Spatial economic and Network effects of transport investments and policies) Deliverable 6. Kiel/Dortmund Decaluwe B, Lemelin A, Bahan D, Annabi N (2004) Endogenous labour supply and capital mobility in a bi-regional CGE. Paper presented at the ECOMOD workshop Dixit AK, Stiglitz JE (1977) Monopolistic competition and optimum product diversity. Am Econ Rev 67:297–308 ESPON (2004) Territorial impact of EU transport and TEN policies. Deliverable 2.1.1 Hertel T (ed.) (1997) Global Trade Analysis: Modeling and Applications. Cambridge University Press: Cambridge, UK Ilkkaracan I, Raziye S (2003) The role of unemployment in wage determination: further evidence on the wage curve from Turkey. Appl Econ 35(14):1589–1598 Janssens S, Konings J (1998) One more wage curve: the case of Belgium. Econ Lett 60 (2):223–227 Johansen K (2002) Regional wage curves: empirical evidence from Norway. Norwegian University of Science and Technology, Trondheim Kallai E, Traistaru I (2001) Characteristics and trends of regional labour markets in transition economies: empirical evidence from Romania. LICOS Discussion paper 72/2001. Katholieke Universiteit Leuven, Leuven Mantzos L, Capros P (2006) European energy and transport: trends to 2030 – update 2005. Office for Official Publications of the European Communities, Luxembourg Montuenga V, Inmaculada G, Melchor F (2003) Wage flexibility: evidence from five EU countries based on the wage curve. Econ Lett 78(2):169–174 Mundell RA (1957) Transport costs in international trade theory. Canadian Journal of Econmics 23:331–348 Nicolaisen S, Tranaes T (1996) Wage curves for Denmark. Nationalokon Tidsskr 134 (3):223–237 Pekkarinen T (2001) The wage curve: evidence from the Finish metal industry panel data. Finnish Econ Pap 14(1):51–60 Roson R (1997) Wage curves and capital mobility in a general equilibrium model of Italy. In: Policy simulations in the European Union. Routledge, London, pp 79–98 Samuelson P (1983) Thu¨nen at two hundred. J Econ Lit 21:1468–1488 Shapiro C, Stiglitz JE (1984) Equilibrium unemployment as a worker discipline device. Am Econ Rev 74:433–444

196

A. Korzhenevych and J. Bro¨cker

Shoven JB, Whalley J (1984) Applied general-equilibrium models of taxation and international trade: an introduction and survey. J Econ Lit 22:1007–1051 Shoven JB, Whalley J (1992) Applying general equilibrium. Cambridge University Press, Cambridge Simoncic M, Pfajfar L (2004) An estimate of the wage curve for Slovenia. Institute for Economic Research, University of Ljubljana, Ljubljana Venables AJ, Gasiorek M (1999) The welfare implications of transport improvements in the presence of market failure. SACTRA report

Chapter 12

Infrastructure Productivity with a Long Persistent Effect Tsukai Makoto and Kobayashi Kiyoshi

12.1

Introduction

12.1.1 Production Function Approach Appropriate investment and management of infrastructure is an important issue in national or regional planning. Aschauer (1989) reported two important findings: that infrastructure productivity was significantly higher than expected, and that fewer investments in the USA after 1970 would result in less growth of national productivity. Since then, measurement of infrastructure productivity has become a focal issue in national or regional management policy, and a large number of empirical studies have accumulated; see Sturm (1998). An aggregate production function approach for timeseries applied in Aschauer’s study, however, was criticized by economists, or econometricians. Among the assumptions inherent in the production function approach, ‘‘constant return to scale’’ and ‘‘competitive input factor market’’ are often viewed with suspicion from the viewpoint of the endogenous growth theory by Romer (1986) and Lucus (1988) or from an uncompetitive structure of infrastructure market, respectively. Basu and Fernald (1997) empirically tested these assumptions with the production function approach, based on data from 34 sectors of US industries. They concluded that in some segmented sectors of industry, the constant return to scale assumption was violated but that it held true in most sectors. Holtz-Eakin and Lovely (1996) analyzed the effect of infrastructure on economic activities by using a general equilibrium system explicitly considering scale economies. They showed that infrastructure decreases the input factor cost, increases the variety of industries, and increases the number of newly founded companies. Haughwout (2002) applied the spatial equilibrium theory considering regional monopolistic power over input factor market, and measured infrastructure productivity. This study showed that K. Kiyoshi (*) Graduate School of Management, Kyoto University e-mail: [email protected].

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_12, # Springer‐Verlag Berlin Heidelberg 2009

197

198

T. Makoto and K. Kiyoshi

infrastructure provided significant marginal benefit on regional economic activities. Most studies have reported that there is a positive production/cost reduction effect caused by infrastructure, but unfortunately, some studies have paid little attention to the data generation process that significantly influences findings through model specifications. Recent American, European, and Japanese studies about the production function approach have been widely reviewed by Ejiri et al. (2001). From an econometric standpoint, model misspecification, the possibility of reverse causation from the product to infrastructure, and spurious correlation are often focused upon as nuisance problems in statistical regression analysis. All these faults in modeling result in inconsistent, inefficient, or biased estimates. A typical model misspecification is to dismiss technological improvement or scale economies. Duggal et al. (1999) formulated the time series production function model which explicitly considered technological growth as a function of infrastructure, and nonlinear labor productivity (allowing increasing and decreasing return to scale), and then applied the model to US data. In this approach, it is necessary to estimate factor demand function, due to a violation of the constant return to scale assumption; hence the two stage least square method was adopted. The result obtained in this study supported that of Aschauer. Everaert and Heylen (2001), used the simultaneous equations method based on the error correction mechanism, which uses a differenced data series with first order, in order to test spurious correlation, reverse causation, and endogenous bias problems in the production function approach. In the application to Belgian data, the reverse causation from output product to infrastructure was rejected. To address the spatial correlation problem, many studies using the spatial econometrics approach have been conducted. Tsukai et al. (2002) estimated regional production function with spatial spillover effect, and in fact found a significant spatial spillover effect. The serial autocorrelation problem, however, remains unresolved in their study, since the production function is estimated by cross-sectional data. Unfortunately, our review found no studies focusing on the persistent (lagged) production effect of infrastructure. An analysis of the persistent production effect of infrastructure requires an appropriate model specification for both spatial expansion and temporal persistency. Another requirement for the time-series production function approach is an appropriate consideration of the technological innovation effect. The technological innovation effect is modeled as residuals that cannot be explained in structural terms of the production function, but which might show particular spatio-temporal structure. In other words, the characteristics of spatial/ cross-sectional structure should be explicitly modeled. In this study, the lagged effects of infrastructure productivity and technological innovation are specified as a multiple time series model with a long persistent effect.

12.1.2 Long Persistent Productivity of Infrastructure If the production effect of infrastructure lasts over years after the investment, multiple time series used for regional production function would have the long

12

Infrastructure Productivity with a Long Persistent Effect

199

persistent property, which in the time series data is a kind of statistical property of the data generation process (DGP), mathematically defined in Sect. 12.2.1. Various empirical studies about time series data such as those involving temperature in climatology, voting behavior in political science, currency exchange rate in macroeconomics, and merger and acquisition behavior in management economics, have reportedly found the long persistent property in these time-series data (Box-Steffensmeier and Tomlinson 2000; Barkoulas et al. 2001). The studied series obtained by aggregation over some regions or several durations would have the long persistent property, even if each series does not have long persistency. Davidson and Sibbertsen (2005) gave simulation results to demonstrate what appears to be the long persistent property in aggregated data series from nonpersistent disaggregated series. Suppose a process exhibits short memory fluctuations (i.e., uncorrelated standard disturbance such as with white noise) around a local average which randomly switches such that the durations of the regimes follow a power low. The authors showed that the aggregated process composed of a large number of independent copies of such process results in fractional Brownian motion, known to show long persistent behavior. They stated that an aggregated time series actually does not correspond to the single representative firm, but encompasses an extensive geographical area including many heterogeneous firms or regions that are interdependent. In terms of long persistency in economical data, Michelacci (2004) refers to the delay of innovation to replace capital in the private sector. Even though new private capital embedding the latest technology can make more profit, an immediate catch-up strategy to replace present capital with the latest one may not be optimal for each individual firm. An option to postpone the catch-up may rather be preferred under the rapid growth of technology, or in the case where present capital has been recently replaced. In a situation where some firms immediately adopt new technology but others do not, technological innovation in aggregated series would show long persistency. Michelacci’s consideration about the ‘‘delay of innovation’’ phenomenon may also be applicable to infrastructure productivity measured through private sector output. Suppose there is a change in land use, such that factories are not immediately, but gradually concentrated along newly constructed infrastructure. Then the production effect by new infrastructure would be observed with some delays. The other notable point is that infrastructure data is estimated by aggregation of individual stock with different vintage. If infrastructure productivity depends on its vintage, because of, e.g., gradual change in land use, it is not appropriate to assume an instantaneous productivity effect of infrastructure. Therefore, the long persistent effect of infrastructure on production output seems worth explicitly considering in the econometric measurement of infrastructure production effect.

12.1.3 Time-Series Model with a Long Persistent Effect In the Box–Jenkins approach known as a standard time series analysis, a nonstationary time series should be differenced in order to convert it into stationary.

200

T. Makoto and K. Kiyoshi

Then the converted (stationary) series is used for modeling the stochastic process with AR and MA parameters, called the ARIMA model. While differentiation to remove the non-stationary property of time series data is a popular data handling technique proposed in the standard Box–Jenkins method, Munnel (1992) argued that first differencing destroys long-term relationships in the data, and therefore does not make economic sense. Strum and de Haan (1995) estimated production function, and obtained insignificant estimates of private capital and labor by using first differenced data series (they assumed original data series are integrated). The loss of information by first differencing was originally pointed out by Granger and Joyeux (1980) and Hosking (1981). Instead of first differencing, they proposed fractional integration order between no integration (level) and integration. Fractionally integrated time series exhibits long persistent behavior but still holds the stationary property if fractional integration order is less than 0.5. The ARFIMA (Auto-Regressive, Fractionally Integrated, Moving Averaged) model that combines fractional order of time series integration with the conventional autoregressive/ moving average method has been widely applied to prove its empirical applicability, such as for stock price series or international exchange rate (Smith et al. 1997; Henry and Olekalns 2002; Robinson and Hidalgo 2003). The ARFIMA model can include exogenous variables for deterministic trend, called the ARFIMAX model (ARFIMA model with exogenous variables). In order to measure infrastructure productivity, regional output series should be modeled as stochastic process, which deterministic term is explained by the production function including labor, private capital, and infrastructure term. The purpose of this study is to clarify the long persistent effect of infrastructure on regional productivity. The regional production function with fractional integration is formulated and is applied to Japanese data. This chapter is organized as follows: Sect. 12.2 shows the general formulation of the ARFIMA model with exogenous variables, and specifies the production function to be estimated. Data and the results of estimation with some discussion are presented in Sect. 12.3. Finally, the conclusion and a summary of further issues for study appear in Sect. 12.4.

12.2

Long Persistent Model

12.2.1 ARFIMA Model with Exogenous Variables The formulation of an ARFIMA model without exogenous variables is similar to the ARIMA model proposed by Box and Jenkins, but an ARFIMA model permits fractional differencing/integration order. Suppose that t is discrete time indices (t ¼ 0; :::; T), Yt is an endogenous variable driven by normally distributed disturbances as et □ Nð0; s2 Þ, L is a lag operator defined by L Yt ¼ Yt1 , and d is fractional differencing/integration order. The ARFIMA model is formulated as follows in (12.1).

12

Infrastructure Productivity with a Long Persistent Effect

201

ð1  LÞdY fðLÞYt ¼ ’ðLÞet ; ð1  LÞdY ¼

1 X j¼0

ð12:1Þ

GðdY þ 1Þ ð1Þj Lj ; Gðj þ 1ÞGðdY  j þ 1Þ

ð12:2Þ

fðLÞ ¼ 1  f1 L  f2 L2     ;

ð12:3Þ

’ðLÞ ¼ 1  ’1 L  ’2 L2     :

ð12:4Þ

In (12.1), dY > 0 indicates fractional differencing, dY < 0 indicates fractional integration, and dY ¼ 0 indicates no differencing, no integration. Note that (12.2) is obtained as Maclaurin expansion of ð1  LÞdY which has a structure mathematically identical with f ðxÞ ¼ ð1  xÞd as follows: f ðjÞ ðxÞ ¼ dðd  1Þ    ðd  j þ 1Þð1  xÞdj ð1Þj ¼

d!ð1  xÞdj ð1Þj : ðd  jÞ!

Therefore d

ð1  xÞ ¼

1 j X x j¼0

¼

1 X j¼0

j!

ðjÞ

f ð0Þ ¼

1 j X x j¼0

j!

d!ð1Þj ð1  xÞdj ðd  jÞ!

! ¼ x¼0

1 X d!ð1Þj j x j!ðd  jÞ! j¼0

Gðd þ 1Þ ð1Þj xj : Gðj þ 1ÞGðd  j þ 1Þ

In (12.2), GðaÞ is a gamma function, which is interpreted asR expansion of 1 factorial calculation into real numbers, such as GðaÞ ¼ 0 xa1 ex dx; GðaÞ ¼ aGða  1Þ. If a is an integer, GðaÞ ¼ ða  1Þ!. Note that GðaÞ can also be defined for negative real numbers except for negative integers. Equation (12.2) indicates that fractional differencing can be expanded as an infinite series of lagged Yt s. Here, an invertible condition for dY that dY < 1=2 is assumed (Granger and Joyeux 1980). fðLÞ in (12.3) and ’ðLÞ in (12.4) are called AR, or MA lag polynomials. Assume invertible condition for fðLÞ and stationary condition for ’ðLÞ, which is f1 ;    ; fp < 1 and ’1 ;    ; ’q < 1 (see Box et al. 1994). In this case, (12.1) can be transformed into (12.1a) Yt ¼ ð1  LÞdY ð1  LÞdY ¼

1 X j¼0

’ðLÞ et ¼ HðLÞet ; fðLÞ

1 X GðdY þ jÞ  Lj ¼ pj Lj : Gðj þ 1ÞGðdY Þ j¼0

ð12:1aÞ

ð12:2aÞ

202

T. Makoto and K. Kiyoshi

Equation (12.2a) is again obtained, as Maclaurin expansion of ð1  LÞdY indicates that the series of pj contains infinite terms (see details in Minotani 1995). Since the inverse of AR polynomial f1 ðLÞcan be expanded into an infinite series of lagged parameters, Yt is driven by an infinite series of disturbance terms. When f1 ðLÞ and ð1  LÞdY satisfy stationality, they eventually reach 0. However, two parameter series differ in diminishing speed with a lag increase. Suppose that hj is the composite parameter of (12.1). Gourierouex and Monfort (1997) showed the following approximation for hj if j ! 1: hj 

’1 d1 j : f1 GðdÞ

ð12:5Þ

In (12.5), jd1 =GðdÞ stems from pj converges to 0 under the stational condition d < 1=2. While AR and MA parameters also exponentially converge to 0 such that kj ðjkj < 1Þ, these parameters converge faster than jd1 =GðdÞ, so that they do not appear in (12.5). Lower convergence property with lag in fractionally differenced series is called long persistency, or long memory. The ARFIMA model with exogenous variable, called the ARFIMAX model, is formulated in (12.6) ð1  LÞdY fY ðLÞYt  ð1  LÞdX fX ðLÞf ðXt Þ ¼ ’ðLÞet ;

ð12:6Þ

where Xt are exogenous explanatory variables, f ðXt Þ is a linear function of Xt , dX and dY are fractional integration parameters (i.e., dX > 0 indicates fractional integration), respectively, and fY ðLÞ; fX ðLÞare AR lag polynomials. Stationality and invertibility are assumed to be parameters in (12.6), and the transformation of (12.6) is shown as (12.6a) Yt ¼ ð1  LÞdX dY

fX ðLÞ ’ðLÞ f ðXt Þ þ ð1  LÞdY et : fY ðLÞ fY ðLÞ

ð12:6aÞ

Equation (12.6a) indicates that the long persistent property appears for both deterministic and stochastic processes as f ðXt Þ and et , hence they have a parameter series with infinite lags. Note that if dY ¼ 0, the stochastic process is identical to the ARMA process in the latter case, hence long persistency does not appear in error term, and if dX ¼ dY , long persistency does not appear in infrastructure term. In the ARFIMA model and the ARFIMAX model, differencing parameters are not given but estimated from observed data series. Difficulties in parameter estimation can be summarized into three types. First, because of slow and infinite persistency of fractionally differenced or integrated process, a relatively larger number of samples is necessary to obtain good (i.e., consistent, unbiased, and efficient) parameters (Bhardwai and Swanson 2006). Second, a step-wise estimation procedure ^ then estimates the other AR and MA parameters using d^ to that initially estimates d, filter the series, such as the GPH method, tends to be biased under small samples,

12

Infrastructure Productivity with a Long Persistent Effect

203

because d and the other parameters are dependent (Beran 1992; Igresias et al. 2006). Therefore, differencing parameters and other parameters should be simultaneously estimated. Robinson and Hidalgo proposed the maximum likelihood estimator obtained by frequency domain data series, which has desirable asymptotic properties (2003), and Tanaka showed that the ML estimator by time domain data also has such properties (1999). Finally, when long persistency is observed, it is difficult to distinguish whether the long memory process is truly appropriate as DGP. If regime switch (average drift) occurred in a time series, a misspecified long persistent model should yield a significant fractional differencing parameter (Davidson and Sibbertsen 2005; Dfrenot et al. 2005). Such difficulties should be carefully considered in empirical application. Using composite parameters functioning as HX ðLÞ; He ðLÞ, (12.6a) can be written into (12.6b) Yt ¼ HX ðLÞf ðXt Þ þ He ðLÞet :

ð12:6bÞ

HX ðLÞand He ðLÞ determine the shape of lag distribution of deterministic term and stochastic term. The lagged parameters for deterministic terms are already known as distributed lag models (Almon 1965; Shiller 1973), in which average lag delay, lex , can be calculated by using estimated lag parameters. Suppose that hxti s are a series of lag parameters (hxt ¼ 1), aT is a data aggregation span, and m is lag truncation length. lex shown in (12.7) Pm lex

¼

ðj þ 1Þhtj Pm  aT : j¼0 htj

j¼0

ð12:7Þ

Note that lex of ARFIMAX model does not converge for infinite m, since lex includes the term j  jd1 ¼ jd if d > 0, see and compare (12.5) and (12.7). When finite m is set at empirical parameter estimation, lex can be calculated by (12.7), but the result could be interpreted as a kind of approximated index. Since further discussion about this point is beyond the scope of the present paper, we adopt (12.7) to calculate average lag delay.

12.2.2 Specification of Production Function with a Long Persistent Effect In this paper, we estimate a simple Cobb–Douglas type production function. Unfortunately, the number of available samples is insufficient among annual statistics in Japan, so that cross-sectional and longitudinal data are pooled in order to obtain stable parameter estimates. The data we use is provided by Doi (2003) on his website, available for 46 prefectures from 1955 to 1998 (44 years), details of which are provided in Sect. 12.3.1. Suppose that i is regions

204

T. Makoto and K. Kiyoshi

(i ¼ 1; ::: ; 46; N ¼ 46), t is discrete time indices (t ¼ 1; ::: ; 44  m; T ¼ 44  m, from 1955 þ m to 1998, m is truncated lag length), Yit is gross regional product, Nit is labor input, Kit is private capital stock, and Git is infrastructure stock. The logtransformed production function with a constraint of constant returns to scale, and the deterministic progress of technology is shown in (12.8) log Yit ¼ a0 þ a1 log Nit þ ð1  a1 Þ log Kit þ a2 log Git þ a3 Tt :

ð12:8Þ

Among input factors, long persistency would appear in Git but not in Nit or Kit because labor and private capital can be easily adjusted to maximize productivity, so no long persistent effect is expected. Equation (12.8) is a linear additive function in parameters, hence the long persistent parameters can be set for infrastructure as dG . This setting makes it possible to test the hypothesis that the long persistent effect would be longer for infrastructure than for private capital. These lagged variables are shown in (12.9) log Git ¼ ð1  LÞ ¼

1 X j¼0

dG

 1  X dG þ j  1 j L log Git log Git ¼ j j¼0

GðdG þ jÞ log Gi;tj : Gðj þ 1ÞGðdG Þ

ð12:9Þ

In (12.9), non-lagged variable is included due to GðdG Þ=Gð1ÞGðdG Þ ¼ 1 if j ¼ 0. Note that positive dG indicates differencing and negative dG indicates integration. ðdþjÞ Figure 12.1 plots lag coefficient values given by GðGjþ1 ÞGðdÞ for lags (j  1) at dG ¼ 0:2 and dG ¼ 0:2. In the case of dG ¼ 0:2, all lagged coefficients are positive. Therefore dG is expected to be positive if past infrastructure has a positive production effect on present output.

0.25 0.20

d_G = 0.2 d_G = –0.2

0.15 0.10 0.05 0.00 –0.05

1

2

3

4

5

6

7

8

9

10

11

12

–0.10 –0.15 –0.20 –0.25

Fig. 12.1 Decays in coefficients for lagged variable with long persistency

13

14

12

Infrastructure Productivity with a Long Persistent Effect

205

In addition to lagged variables, also considered is the spatial spill over effect brought by infrastructure in other regions Gi;t . Gi;t are weighted by inverse of distance, then summed up for all regions except for the own region log Git ¼

j6¼i X

wij log Gjt :

ð12:10Þ

j

Equations (12.9) and (12.10) are added to (12.8). The regional production function with stochastic disturbance uit is in (12.11). In this model, it is important to set the structure of the stochastic process appropriately log Yit ¼ a0 þ a1 log Nit þ ð1  a1 Þ log Kit þ flog Git þ a2 log Git þ a3 Tt þ uit :

ð12:11Þ

In (12.6), the long persistent property will also appear in the stochastic error term in the case of dY > 0. In our model, long persistency and first order MA polynomial structure are assumed for the stochastic process, in (12.12) uit ¼ rei;t1 þ ð1  LÞde eit ;

or uit ¼ ð1  ’LÞð1  LÞde eit :

ð12:12Þ

This specification corresponds to the case of fX ðLÞ ¼ fY ðLÞ ¼ 1, and ’e ðLÞ ¼ 1  ’L in (12.6). Parameters to be estimated are a0 ; a1 ; a2 ; a3 ; ’ in (12.11), dG in (12.9), de and MA parameter ’ in (12.12). Here, let us introduce some matrix algebra in order to clarify the difference in differencing order for each term

Y ¼ f. . . ; log Yit ; . . .g0 ; !g ¼ f. . . ; log Git ; . . .g0 ;

X¼

8 > > > < > > > :

1; log Nit ; log Kit ;

.. . P j6¼i

wij log Gjt ; log Tt

.. .

9 > > > = > > > ;

;

e ¼ f. . . ; eit ; . . .g0 ; a ¼ fa0 ; a1 ; ð1  a1 Þ; f; a3 g0 : ð12:13Þ

Using the above algebra, (12.14) or (12.15) is obtained (as to no differencing parameter for error term) Y  Xa  a2 ð1  LÞdG g ¼ ð1  ’LÞð1  LÞde e;

ð12:14Þ

h i ð1  ’LÞ1 ð1  LÞde ðY  XaÞ  a2 ð1  LÞde dG g ¼ e:

ð12:15Þ

206

T. Makoto and K. Kiyoshi

For the error term, e □ Nð0; s2 IÞ is assumed. Parameters are estimated by pseudo maximum likelihood estimation for time domain data series proposed by Chung and Baillie (1993) and Tanaka (1999), called the CSS estimator. In this study, the data set is composed of 46 time series for each prefecture. The log-likelihood function is shown in (12.16) log L ¼ 

n n 1 XX 2 log 2p  log s2e  2 e : 2 2 2se i t it

ð12:16Þ

By using estimated parameters, instantaneous marginal productivity and the sum of persistent marginal productivity of infrastructure are calculated in (12.17), and (12.18) X Yjt @Yt @Yit X @Yjt Yit ¼ þ ¼ a2 þf wij ; @Git @Git @G G G it it it j j j6¼i

m X @Yi;tþj j¼1

@Git

j6¼i

¼ a2

m X j¼1

GðdG þ jÞ Yi;tþj : Gðj þ 1ÞGðdG Þ Git

ð12:17Þ

ð12:18Þ

In (12.17), the first term on the right indicates productivity contribution to own region, while the second term is contribution to other regions. The truncation order m in (12.18) is exogenously set at parameter estimation.

12.2.3 Diagnostic Tests Since the CSS estimator is asymptotically consistent and asymptotically efficient, a likelihood ratio test based on Large-sample theory can be applied as a diagnostic test for parameter significance (Chung and Baillie 1993). Now the parameter vector €  ¼ ðb € ;    ; b € ;   Þ. The null and of the CSS estimator in (12.15) is denoted by b 1 i alternative hypothesis of likelihood test is in (12.19) 

H0 H0

€ ¼ 0; b i € 6¼ 0: b i

ð12:19Þ

The parameter vector of the CSS estimator estimated under the constraint € € ¼ 0 is denoted by b  . A statistics of likelihood ratio of objective parameter as b i i € is xi in (12.20) for b i

n h i h io € ; s € s €2 €2 Þ  ln Sðb xi ¼ 2 ln Sðb; i  Þ :

ð12:20Þ

12

Infrastructure Productivity with a Long Persistent Effect

207

Since xi asymptotically converges to w2 ð1Þ distribution, the critical value to reject null hypothesis H0 is obtained in (12.21) xi  w2ð100aÞ% ð1Þ:

ð12:21Þ

As in (12.11), under the null hypothesis a2 (for infrastructure) should be considered simultaneously with estimates of dG , because these two parameters are dependent. Statistics for simultaneously testing a2 with dG are xG; dG , shown in (12.22), n h i h io € 2 € s € €2 Þ  ln Sðb  ; s Þ ; xG; dG ¼ 2 ln Sðb; G;dG

ð12:22Þ

€  where b G;dG is the parameter vector of CSS estimator under the constraint with a2 ¼ dG ¼ 0. xG; dG follows w2 ð1Þ distribution. In terms of longitudinal correlation, Durbin–Watson statistics (DW) are used to detect first order serial autocorrelation. Suppose ^eit are the residuals at region i and time t. Overall Durbin–Watson statistics are in (12.23a), and Regional DW statistics are in (12.23b). If positive first order autocorrelation remains in residuals then DWis close to 0, and in the case of no first order autocorrelation, DW is close to 2 PN PT t¼mþ1 DW ¼ 2  2 Pi¼1 N PT i¼1

^eit^ei;t1

t¼mþ1

PT DWi ¼ 2  2 Pt¼mþ1 T

ð^eit Þ2

^eit^ei;t1

t¼mþ1

ð^eit Þ2

:

;

ð12:23aÞ

ð12:23bÞ

When the residuals of cross sectional models show unspecified spatial correlation, the spatial correlation left in ^et can be tested by Moran’s I statistics (Moran 1948). Suppose ^et is a residual vector for each cross-section. Based on ^et ,IM is calculated by (12.24) IM ¼

^e0 t

1

0 2 ðW 0 ^e

 þ WÞ ^et ; et t^

ð12:24Þ

where, W is a spatial weight matrix. We assumed it by inverse distance matrix M identical which have wij s in (12.10). The standardized Moran’s I statistics I is defined as follows: I M ¼

  IM  E IM  1=2 ; Var IM

208

T. Makoto and K. Kiyoshi

where,

  trðUÞ ; E IM ¼ nk     trðUPW0 Þ þ tr U2 þ ½trðUÞ2   M 2 Var IM ¼  E I ; ð n  k Þ ð n  k þ 2Þ

ð12:25Þ

and n is a number of samples, k is a number of parameters, P(projection matrix) and M U are defined as P ¼ I  XðX0 XÞX0 ,U ¼ PW, respectively. It is known that I asymptotically follows the standardized normal distribution.

12.3

Empirical Measurement of Infrastructure Productivity

12.3.1 Data The privatization of Japan Railway companies (JRs), Nippon Telephone and Telecommunications (NTTs), and Japan Tobacco (JT) caused a disjoint at 1987 in the longitudinal statistics of infrastructure. The corrections in the data set of infrastructure and private capital are provided by Doi in order to remove the inconsistency before and after 1987 (2000). Both data were depreciated with the identical rule to infrastructure data. Gross Regional Product (GRP) and labor force are also estimated based on the national census. Input and output data are deflated at the 1995 price. These data sets were provided for each prefecture unit, from 1955 through 1999. In order to ameliorate the small sample problem, we pooled all the available data. Okinawa prefecture was excluded because of its distant location and the lack of data availability before 1972.

12.3.2 Results of Estimation In order to estimate (12.11), lag truncation order m needs to be set. After several trials, m is set at 10. Therefore, 34 cross-sections (from 1965 through 1998) and 46 prefectures data are pooled up to be used for parameter estimation. Then totally 1,568 observations are used. Previous studies in the ARFIMA model reported that setting time-trend parameters affect dG ; de and other structural parameters. Considering the economical situation of the data period, we set three durations with different time-trend parameters as follows: (1) 1965–1973 as a rapid economic growth term up to the oil shock in 1973; (2) 1974–1990 as a post oil shock term up to the end of the bubble economy in 1991; and (3) 1992–1998 as the post bubble economy. In the regional production function, time-trend parameter indicates an average growth of total factor products, or of technological innovation rate.

12

Infrastructure Productivity with a Long Persistent Effect

209

Table 12.1 shows the estimation results; the first two columns are the estimates of production function with long persistency and three deterministic trends, referred to as long memory model-1. The estimated value of infrastructure parameter is considerably small compared with that of conventional studies (0.2–0.3). Because the long persistent model measures persistent productivity and spatial spillover effect of infrastructure, the productivity index would separately appear in different terms. As alternate model specifications, two models are estimated. The middle two columns are the estimates of the model without long persistent terms for both as de ¼ dG ¼ 0, referred to as short memory model. The last two columns are the estimates of the model without long persistent term of infrastructure as dG ¼ 0, referred to as long persistent model-2. First, we compare the long memory model-1 with the short memory model. The adjusted R-squared coefficient is slightly high in the long persistent model. The critical values of Durbin–Watson statistics

Table 12.1 Parameter estimates of regional production function Variables Long persistent MA model Long persistent model-1 model-2 Estimates ChiEstimates ChiEstimates Chisquaredc squaredc squaredc a Labor 0.544** 24.9 0.488** 35.69 0.459** 62.69 0.456** 99.94 0.512** 138.45 0.541** 105.73 Private capitala 0.028** 13.36 0.149** 14.73 0.072** 10.73 Infrastructureb Spatial spillover of 0.030** 6.63 0.013** 1.13 0.022** 8.04 infrastructure Long persistency of 0.310** 13.36 – – – – infrastructure Long persistency of 0.438** 57.24 – – 0.487** 150.52 stochastic error MA(1) in stochastic error 0.753** 194.19 0.972** 370.25 0.734** 187.12 Deterministic trend 0.077** 35.78 0.008** 0.06 0.078** 48.64 (1965–1973) Deterministic trend 0.025** 9.57 (0.021) 1.47 0.027** 15.3 (1974–1991) ** Deterministic trend (0.016) 23.9 (0.051) 78.86 (0.014) 38.1 (1992–1998) ** ** ** Constant (2.612) 13.73 (0.320) 0.03 (2.516) 48.3 ** ** ** Variance 0.032 0.034 0.033 – No. of samples 1,564 1,564 1,564 0.999 0.998 0.993 Adjusted R2 REG statistics 1.716 3.935** 2.024* Durbin–Watson statistics 1.93 1.718** 1.924 a Parameters of private capital and labor are constrained by constant returns to scale b In (12.9), infrastructure parameter and long persistency parameter of infrastructure are jointly tested because they are nonlinearly dependent. Critical values of Chi-squared test are 5.99 (2 d.f., 5% significance), 9.21 (2 d.f., 1% significance) c ** Indicates null hypothesis of zero slope is rejected with 1% significance, and * indicates the null is rejected with 5% significance. Critical values of Chi-squared test are 3.84 (1 d.f., 5% significance), 6.63 (1 d.f., 1% significance)

210

T. Makoto and K. Kiyoshi

(1,564 samples, 1% significant level, 11 parameters) are 1.868–1.896 for ’ > 0 and 2.104–2.132 for ’ < 0, so that ’ > 0 is accepted if DW < 1:868, ’ ¼ 0 is accepted if 1:896 < DW < 2:104, and ’ < 0 is accepted if DW > 2:132, respectively. The statistic values calculated by (12.23a) indicate that ’ ¼ 0 is accepted for long persistent model-1, while ’ > 0 is accepted for the MA model. The prefecture’s Durbin–Watson statistics calculated by (12.23b) ranges between 1.576 and 2.189. The critical values of Durbin–Watson statistics (34 samples, 1% significant level, 11 parameters) are 0.610–2.160 for ’ > 0 and 1.840–3.390 for ’ < 0. Table 12.2 shows that only two prefectures can reject ’ < 0, but the others cannot test the autocorrelation. Figure 12.2 shows the standardized Moran’s I for each cross M section. The rejecting range of no spatial correlation is I > 1:96; therefore there is no significant spatial correlation for all cross-sections. Parameters of labor, private capital and infrastructure are significant in long persistent model-1. Compared with the MA model, the labor parameter is slightly large, and private capital parameter is slightly small due to the constant returns to scale constraint. The estimated infrastructure parameter is considerably small compared with that of conventional studies (0.2–0.3). Because the long persistent model measures persistent productivity and the spatial spillover effect of infrastructure, productivity index would separately appear in different terms. Three timetrend parameters in the long persistent model are significant, but in the MA model, only one parameter for 1991–1998 is significant. The spatial spillover parameter is significant in long persistent model-1, but not for the MA model. The MA parameter is significant for both models. Note that the MA parameter is smaller in long Table 12.2 Results of Durbin–Watson test f>0 f<0 Neither reject nor accept (0.610
0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 –0.1

65

Rejected Neither reject nor accept

27

Neither reject nor accept

2

67 69 71 73 75 77 79

Fig. 12.2 Moran’s Index

Number of preferences 17

81 83 85

87 89 91 93 95 97

12

Infrastructure Productivity with a Long Persistent Effect

211

persistent model-1. Both long persistent parameters in long persistent model-1 are positive, and fulfill stationary and invertible conditions as jde j; jdG j < 0:5. Since positive dG is estimated, GRP for each region enjoys a positive production effect from past infrastructure as shown in Fig. 12.1, which shows the persistent production effect of infrastructure. And positive de indicates that production activity is stationary but that exogenous shock will last for a while.

12.3.3 Statistical Tests for Long Persistent Specification The appropriateness of long persistent specification for production function is tested to confirm two hypotheses simultaneously: that long persistent property remains in the model without long persistent specification and that long persistent property does not appear in the model with long persistent specification. In this section we test these hypotheses by applying the REG test for residual series estimated by three specifications of regional production function. The REG test is proposed by Agiakloglou and Newbold (1994), which is an expansion of the Lagrangean Multiplier (LM) test for ARIMA model parameters by Godfrey (1979). The residuals of long persistent model-1, the MA model, and long persistent model-2 are denoted as ^e1t ; ^e2t ; and ^e3t , respectively. The null hypothesis is that the residual series follows the white noise process as shown in (12.26) ^k et

¼ Wtk

ðk ¼ 1; 2; 3Þ:

ð12:26Þ

An alternative model is ARFIMA(0,d,0), as shown in (12.27) ð1  LÞdk ^ekt ¼ Wtk

ðk ¼ 1; 2; 3Þ:

ð12:27Þ

The LM test statistics on long persistent parameter can be seen in (12.28) where, T and R are number of cross-sections, and number of regions, respectively   T k k k0 k 1 k0 k X RW l l l l W t t t t t t Uk ¼ k0 k Wt Wt t¼1 lkt ¼

@Wkt : @dk

;

ð12:28Þ

dk ¼0

ð12:29Þ

Note that in (12.29), partial differentiated vector lkt is evaluated at neighbor of dk ¼ 0. Between (12.26) and (12.27), difference of degree of freedom is 1, hence the test statistics U k follows chi-squared distribution with d.f.1. Agiakloglou and Newbold noted that (12.28) is proportional to the R-squared coefficient of the regression model that lkt is regressed on Wkt (lkt and Wkt are evaluated at dk ¼ 0),

212

T. Makoto and K. Kiyoshi

then proposed the REG test as the T-statistic test of lkt ’s coefficient by estimating the regression model by OLS. Note that ^ekt ¼ Wtk dk ¼0 from (12.27), lkt can be calculated as in (12.30) @Wkt dk k ^ ¼ ð logð1  LÞ Þð1  LÞ e t @dk dk ¼0 dk ¼0 k ¼ logð1  LÞW t dk ¼0

LÞ^ekt

¼ logð1  1 X ¼ j1^ektj ð¼ Skt Þ:

ð12:30Þ

j¼1

Equation (12.31) shows the auxiliary regression model of the REG test ^ekt ¼ tk Stk þ zkt

ðk ¼ 1; 2; 3Þ;

ð12:31Þ

where, zkt is the error term of auxiliary regression model. Therefore, a long persistent property test for the estimated residual is shown in (12.32). 

H0k H1k

^tk ¼ 0; ^tk 6¼ 0:

ð12:32Þ

If H0k is rejected, the long persistent property remains in the residual series. The result of the REG test is shown in Table 12.1 in the second row from the bottom. The REG statistic for the MA model is 3.935. Since H02 is rejected with 1% significance, the long persistent effect remains in the residual series of the MA model. The REG statistic for the long persistent model-1 is 1.716. Since H01 is not rejected with 5% significance, the residual series does not have the long persistent property. The REG statistic for the long persistent model-2 is 2.024. Since H03 is rejected with 5% significance, the residual series has the long persistent property. Summarizing the results of the REG test, the residual series of MA model (without de and dG ) and of long persistent model-2 (without dG ) have the long persistent property, while the residual series of long persistent model-2 do not have the long persistent property. Therefore, we can conclude that the long persistent property of infrastructure productivity cannot be rejected. However, the implications from the above statistical test for the long persistent model have certain limitations. First, the possibility of misspecification for spatial auto-correlation remains in our model because spatial correlation structure in error term was not considered. Second, a problem might arise from an insufficient number of observations. Hosking (1996) asserted that the expected mean of the ARFIMAX process asymptotically distributes with normal distribution, but its parameters converge with n1=2de , while the parameters of a linear regression model with i.i.d. error converge with n1=2 . Note that d^e of our long persistent model-1 is 0.428; therefore the convergence rate of our model is very slow. In

12

Infrastructure Productivity with a Long Persistent Effect

213

other words, a much larger number of observations is required in the estimation of the ARFIMAX model to keep its efficiency as high as that of the corresponding ARIMAX model. Third, structural change of deterministic time-trend might occur. Previous studies reported that misspecification of deterministic time-trend would affect d^e (Davidson and Sibbertsen 2005; Dfrenot et al. 2005). In order to specify these points properly, the long persistent property and spatial auto-correlation should be simultaneously modeled by spatially and temporally correlating with the long persistent effect. The possibility of structural change in deterministic time-trend, which is exogenously given in our production function, should be statistically tested. These points are further issues which need to be addressed in regional production function approach.

12.3.4 Average Growth of Technological Innovation From the estimates of deterministic time-trend, average growth of technological innovation is calculated. Table 12.3 shows a comparison of average growth of technological innovation between long persistent model-1 and the MA model. Figure 12.3 shows graphs of the average growth of technological innovation calculated from long persistent model-1 and growth of GDP. In this figure, the growth of technological innovation is prior to the growth of GDP. The technological innovation calculated from the MA model, however, does not fit with GDP growth. Figure 12.3 indicates that the long persistent property should be considered to estimate appropriate deterministic trends, but even with our approach some problems remain. In conventional studies, the least squared model with i.i.d. error structure or the ARIMA model is widely adopted. Our results indicate that the conventional approach, when assessing growth productivity, cannot separate between technological innovation and long persistent infrastructure productivity, and that the conventional approach cannot properly estimate deterministic time-trend parameters.

12.3.5 Infrastructure Productivity Figures 12.4 and 12.5 show marginal productivity of infrastructure in 1970 and in 1990, respectively, calculated from long persistent model-1. It is decomposed into three types: instantaneous productivity for own region (MPG) calculated by the first term of (12.17); the sum of instantaneous productivity for the other regions (spill-over Table 12.3 Technological innovation rate Term Long-persistent model (%) MA model (%) 1965–1973 8.00 0.80 1974–1991 2.53 2.08 1992–1998 1.59 4.97 Note: Technological innovation rate is calculated by using time trend parameters atk as exp(atk)1

214

T. Makoto and K. Kiyoshi 3.0

17 15

2.5

13 11

2.0

9 7

1.5

5 3

1.0

65 67 69 71 73 75 77 79 81 83 85 87 89 91 93 95 97

1

Fig. 12.3 Longitudinal plots of technological innovation and GDP growth in Japan (red line/left axis: technological innovation, blue line: GDP in Japan)

0.04 0.03 0.02 0.01

Instantaneous productivity (To own region)

0.04 0.03 0.02 0.01

Instantaneous productivity (To the other regions)

0.060 0.045 0.030 0.015

Persistent productivity (To own region)

Fig. 12.4 Marginal productivity of infrastructure in 1970

0.04 0.03 0.02 0.01

Instantaneous productivity (To own region)

0.04 0.03 0.02 0.01

Instantaneous productivity (To the other regions)

Fig. 12.5 Marginal productivity of infrastructure in 1990

0.060 0.045 0.030 0.015

Persistent productivity (To own region)

12

Infrastructure Productivity with a Long Persistent Effect

215

Fig. 12.6 Total marginal productivity of infrastructure

effect, MPGS) calculated by the second term of (12.17); and the sum of persistent productivity for own region (long persistent effect, MPGM) calculated by (12.18). In order to calculate MPGM in 1990, GRP information in 1999 and 2000 is required. In the calculation, we assumed that GRP after 1998 is identical to 1998. In 1970, MPG and MPGM are relatively high in industrialized areas, such as Tokyo, Osaka, Nagoya, Hiroshima, and Fukuoka, while MPGS is high in the surroundings of industrialized areas. In 1990, MPG and MPGM are high in Aichi and Shizuoka, where the car industries (Toyota, Honda and Yamaha) are clustered, and in the prefectures surrounding Tokyo and Osaka. A comparison of 1970s with 1990s results show that prefectures with high MPG are distributed throughout Japan, and that prefectures with high MPGS can commonly be found around high MPG prefectures. MPGM is still high around the three metropolitan regions (Tokyo, Nagoya, and OSAKA), while MPGM in western prefectures between Osaka to Fukuoka is significantly decreased. Figure 12.6 shows the total marginal productivity of infrastructure (TMPG) in 1970, 1980, and 1990, which is the sum of MPG, MPGS and MPGM for each crosssection. In Fig. 12.6, bi-polar distribution of TMPG around Tokyo and Osaka appears in 1970. In 1990, high TMPG prefectures successively appear between Tokyo and Osaka, including Nagoya. Compared to those of conventional studies, values of TMPG of the long persistent model are slightly lower (Ejiri et al. 2001). The estimated marginal productivity obtained by the long persistent model is lower than in conventional studies. However, the decreasing trend of marginal productivity of infrastructure, which is often pointed out in conventional studies, is not observed, but rather the increasing trend is observed. Since diagnostic tests for residuals show that the long persistent model does not seriously suffer from serialand spatial- autocorrelation problems, the obtained results seem to be credible. Such a difference in longitudinal and cross-sectional behavior of infrastructure productivity stems from the difference in model specification. Neglecting the

216

T. Makoto and K. Kiyoshi

persistent effect of past infrastructure in marginal productivity would result in overestimation of present infrastructure productivity. In our result, one third of marginal productivity stems from past infrastructure, hence its contribution should not be neglected. Figure 12.7 shows the share of three types of marginal productivity in total marginal productivity averaged for all regions. The averaged shares are around 0.33–0.34 for all cross-sections, hence the three types of marginal productivity are almost at the same level. The share of MPG is highest in 1968, and continuously decreases up to 1998. On the other hand, the shares of MPGM and MPGS continuously increase from 1965 to 1998. Therefore, long persistent productivity of infrastructure is stable and has increased for whole regions. The growth of MPGS would correspond to the globalization of domestic economic activities, and such trends will not change in the future. The average delay of infrastructure productivity ^ and d^G of long persistent can be calculated from (12.7). Using the estimated ’ model-1, the average delay of infrastructure productivity is about 5.78 years. As discussed in Sect. 12.2.1, infrastructure productivity of the long persistent model under the infinite lags diverges to positive infinite. Note that the calculated 5.78 years implies approximated delay truncated with 10 years. If longer series of infrastructure stock data were available, the average delay would be more largely estimated. Most conventional studies propose to measure infrastructure productivity for own regions (Ejiri et al. 2001). However during the planning stage, infrastructure is expected to have productivity effect extending over a long term and to global regions. An empirical analysis for productivity measurement neglecting the effect to the future and to neighboring regions would result in underestimation of infrastructure productivity. Of course, it is impossible to conclude immediately the

0.34

MPG

MPGS

MPGM

0.339 0.338 0.337 0.336 0.335 0.334 0.333 0.332 0.331 0.33

65

67

69

71

73

75

77

79

81

83

85

87

89

91

93

95

Fig. 12.7 Shares of components in total marginal productivity, averaged for all regions

97

12

Infrastructure Productivity with a Long Persistent Effect

217

extent of such underestimation. More careful studies about the long persistent effect or spatial spillover effect should be undertaken, followed by intense discussion of the productivity of infrastructure.

12.4

Conclusions

In order to measure infrastructure productivity with lasting effects for the future, we formulated a production function with a long persistent effect, and the proposed model was applied to measure infrastructure productivity in Japan from 1965 to 1998. The estimated model showed that a positive and significant long persistent effect was observed for infrastructure and for stochastic error term. Based on the estimated production function with a long persistent effect, the longitudinal change in marginal productivity of infrastructure was estimated. The estimated series of marginal productivity gave us the novel implication that the marginal productivity of infrastructure is longitudinally increasing, which is different from the implication of conventional approaches measuring infrastructure productivity only within the own region. The share of persistent productivity over the total infrastructure productivity composed of instantaneously to the own, instantaneously to the other regions and lasting for the future to the own region was about one-third throughout the data duration. Because of its stable share, the persistent productivity effect of infrastructure should not be neglected. In other words, considerable underestimation of infrastructure productivity would occur if a policy maker neglected the long persistent productivity effect of infrastructure. Our approach is significant in that it shows an approach to measure the long persistent productivity effect of infrastructure by using a long memory model, but there remain further issues to be resolved. First, the long persistent property should expand for spatial correlation structure. Our model considered only longitudinal correlation structure, but the spatial or spatial-temporal correlation structure was neglected. The development of a model specification and estimation procedure is one important issue. Second, points of trend shift in deterministic time-trend were exogenously set in our approach based on retrospective assessment in the economic situation. The estimated time trend parameters were the appropriate signs due to the ad-hoc shift setting. However, in long persistency in empirical analysis, many papers stress the importance of properly specifying regime switching in empirical modeling in order to improve long persistent (fractional integration) parameters (Gonzalo and Lee 1998; Cheung and Lai 2001; Patel and Shoesmith 2004). Much careful treatment is required for deterministic time-trend. A further issue is to develop a statistical test in regime switching applicable for multiple time-series data. Third, about the spatial spillover effect, our model only considered the infrastructure spillover effect. However, the knowledge of spillover effect embedded in regional output would affect social capital spillover. In applied spatial econometrics, a production function approach explicitly considering spatial heterogeneity is proposed as regional specific fixed or random effect based on panel data

218

T. Makoto and K. Kiyoshi

analysis (Baltagi et al. 2001; Jhun et al. 2003). Panel data analysis with the long persistent property is another target issue. Finally, technological innovation is modeled as a deterministic trend in our model. Investment in infrastructure would induce technological innovation. Therefore, investment in infrastructure, long persistent property in stochastic error term, and deterministic trend would have correlations. . Studies on proper estimation procedure about production function based on endogenous growth theory are required. All the above issues are related with an appropriate specification of spatialtemporal process in the long persistent model, so that a theoretical approach for the spatial-temporal stochastic process is important.

References Agiakloglou C, Newbold P (1994) Lagrange multiplier tests for fractional difference. J Time Series Anal 15:253–262 Almon S (1965) The distributed lag between capital appropriation and expenditures. Econometrica 33:178–196 Aschauer D (1989) Is public expenditure productive? J Monet Econ 23:177–200 Baltagi B, Song S, Jung B (2001) The unbalanced nested error component regression model. J Econom 101:357–381 Barkoulas T, Baum F, Chakraborty A (2001) Waves and persistence in merger and acquisition activity. Econom Lett 70:237–243 Basu S, Fernald J (1997) Returns to scale in U.S. production: estimates and implications. J Polit Econ 105:249–283 Beran J (1992) Statistical methods for data with long-range dependence. Stat Sci 7:404–427 Bhardwai G, Swanson N (2006) An empirical investigation of the usefulness of ARFIMA models for predicting macroeconomic and financial time series. J Econom 121:539–578 Box P, Jenkins M, Reinsel C (1994) Time series analysis: forecasting and control, 3rd edn. Prentice hall, New York Box-Steffensmeier M, Tomlinson R (2000) Fractional integration methods in political science. Elect Stud 19:63–76 Cheung Y, Lai K (2001) Long memory and nonlinear mean reversion in Japanese yen-based real exchange rate. J Int Money Finance 20:115–132 Chung F, Baillie R (1993) Small sample biases in conditional sum-of-squares estimators of fractionally integrated ARMA process. Empir Econ 18:791–806 Davidson J, Sibbertsen P (2005) Generating schemes for long memory processes: regimes, aggregation and nonlinearlity. J Econom 128:253–282 Dfrenot G, Guegan D, Peguin-Feissolle A (2005) Long-memory dynamics in a SETAR mode – applications to stock markets. Journal of Financ Mark Inst Money 15:391–406 Doi T (2003) http://www.econ.keio.ac.jp/staff/tdoi/index-J.html (in Japanese) Duggal V, Saltzman C, Klein L (1999) Infrastructure and productivity: a nonlinear approach. J Econom 92:47–74 Ejiri R, Okumura M, Kobayashi K (2001), Productivity of social capital and economic growth: state of the art. J Infrastruct Plann Manage 688/IV-53:75–87 (in Japanese) Everaert G, Heylen F (2001) Public capital and productivity growth: evidence for Belgium, 1953–1996. Econ Model 18:97–116 Godfrey L (1979) Testing the adequacy of a time series model. Biometrica 66:62–72 Gonzalo J, Lee T (1998) Pitfalls in testing for long run relationships. J Econom 86:129–154

12

Infrastructure Productivity with a Long Persistent Effect

219

Gourierouex C, Monfort A (1997) Time series and dynamic models. Cambridge University Press, Cambridge Granger CWJ, Joyeux R (1980) An introduction to long-memory time series models and fractional differencing. J Time Series Anal 1:15–29 Haughwout A (2002) Public infrastructure investments, productivity and welfare in fixed geographic areas. J Public Econ 83:405–428 Henry T, Olekalns N (2002) Does the Australian dollar real exchange rate display mean reversion? J Int Money Finance 21:651–666 Holtz-Eakin D, Lovely M (1996) Scale economies, returns to variety, and productivity of public infrastructure. Reg Sci Urban Econ 26:105–123 Hosking J (1981) Fractional differencing. Biometrika 68:165–176 Hosking J (1996) Asymptotic distributions of the sample mean, auto-covariances, and autocorrelations of long-memory time series. J Econom 73:261–284 Igresias P, Jorquera H, Palma W (2006) Data analysis using regression models with missing observations and long-memory: an application study. Comput Stat Data Anal 50:2028–2043 Jhun M, Song S, Jung B (2003) BLUP in the nested regression model with serially correlated errors. Comput Stat Data Anal 44:77–88 Lucus R (1988) On the mechanics of economic development. J Monet Econ 22:3–42 Michelacci C (2004) Cross-sectional heterogeneity and persistence of aggregate fluctuations. J Monet Econ 51:3121–1352 Minotani C (1995) Differencing and Integral for economic analysis. Taga-Shuppan, Tokyo Moran P (1948) The interpretation of statistical maps. J R Stat Soc B 10:243–251 Munnel H (1992) Policy watch: infrastructure investment and economic growth. J Econ Perspect 6:189–198 Patel A, Shoesmith G (2004) Term structure linkages surrounding the Plaza and Louvre accords: evidence from Euro-rates and long-memory components. J Bank Finance 28:2051–2075 Robinson P, Hidalgo F (2003) Time-series regression with long-range dependence. In: Robinson PM (ed) Time series with long memory. Oxford University Press, Oxford, pp 305–333 Romer M (1986) Increasing returns and long-run growth. J Polit Econ 94:1002–1037 Shiller J (1973) A distributed lag estimator derived from smoothness priors. Econometrica 41:775–778 Smith J, Taylor N, Yadav S (1997) Comparing the bias and misspecification in ARFIMA models. J Time Series Anal 18:507–527 Strum E, de Haan J (1995) Technological change and aggregate production function. Rev Econ Stat 39:312–320 Sturm E (1998) Public capital expenditure in OECD countries. The causes and impact of the decline in public capital spending. Edward Elger, Cheltenham Tanaka K (1999) The non-stationary fractional unit root. Econom Theory 15:549–582 Tsukai M, Ejiri R, Okumur M, Kobayashi K (2002) Productivity of infrastructure and spillover effects. J Infrastruct Plann Manage 716/IV-57:53–67 (in Japanese)

Chapter 13

Science Parks and Local Knowledge Creation: A Conceptual Approach and an Empirical Analysis in Two Italian Realities Roberta Capello and Andrea Morrison

Though the paper is the result of a joint effort of the two authors, R. Capello wrote Sects. 13.1, 13.2 and 13.5, while the remaining sections have been written by A. Morrison.

13.1

Introduction

Thanks to their ability to perform knowledge-related tasks such as diffusing knowledge locally, promoting high-tech firms, establishing links between knowledgecreating bodies (e.g. universities, research centers) and knowledge-exploiting bodies (e.g. public and private firms, local institutions), science parks have long been considered efficient instruments of industrial and regional policy (Jones 1996; Martin 1997). They were expected to enhance the diffusion of new and advanced technologies/knowledge among firms, and consequently boost the competitiveness of firms and regions. At present, a significant number of studies show that most science parks have failed to perform their intended function (Appold 2004; Massey et al. 1992; Quintas et al. 1992; Vedovello 1997). Many reasons have been attributed to this failure. An important reason is the erroneous and misleading belief that simple geographical proximity between sources of knowledge and local firms is sufficient to foster the widespread spatial diffusion of information, technologies and new ideas (Macdonald 1987; Vedovello 1997). Another reason is the peculiar governance structure of science parks: in fact they may address as many objectives as there are main stakeholders, which in turn may lead to inconsistent policies (Monk et al. 1988; Lo¨fsten and Lindelo¨f 2002). Beyond the reasons encountered in the literature, all with a sound scientific validation, we consider that two additional major elements can account for the questionable effectiveness of science parks which has been reported. First, the R. Capello (*) Department of Management, Economics and Industrial Engineering, Politecnico di Milano e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_13, # Springer‐Verlag Berlin Heidelberg 2009

221

222

R. Capello and A. Morrison

existing assessment literature pays little attention to the variety of science parks. They have different structures, missions, aims and functions: these factors have to be taken into account in the evaluation procedure. Second, we feel that relevant factors affecting the effectiveness of science parks have been inadequately addressed in the literature. In particular, the role that science parks can play in the diffusion of knowledge by supporting and stimulating spontaneous local knowledge creation and transfer channels, has been overlooked. Moreover, little is known about the gatekeeping function of science parks – that is their ability to search and scan the external environment in order to connect external knowledge sources to local actors1 – and their ability to connect local actors (i.e. to contribute to strengthen what regional economists have for a decade called ‘‘collective learning’’). This paper intends to fill in this gap. In particular, the paper aims at measuring the effectiveness of science parks; defined and measured as the ability of science parks to support spontaneous mechanisms of local knowledge transfer among local actors. The paper is structured in two parts. In the first part we present a conceptual framework in order to classify science parks according to their ‘‘effectiveness’’ (Sect. 13.2). In the second part, we present the results of the empirical analysis. On the basis of a quantitative approach we test the effectiveness of science parks in helping to foster spontaneous mechanisms of local knowledge creation and transfer. For this purpose, we carry out our analysis on a sample of 160 firms located in two Italian regions where selected science parks operate (Sect.13.3). Findings suggest that our conceptual framework is well equipped to guide investigation into this issue (Sects. 13.4 and 13.5).

13.2

Science Parks and Local Knowledge

13.2.1 Science Parks: A Definition The expression ‘‘Science Park’’ covers a large variety of research centers and innovation incubators. In general terms, a science park is defined as a geographical area in which firms, universities and research centers have a common location in order to exploit proximity advantages, knowledge spillovers and dynamic agglomeration economies. Examples of these kinds of Science Park are the American success stories of Silicon Valley and of Stanford Research Park, replicated in Europe, e.g. in the Sophia Antipolis Park in France and in the Cambridge Science Park in Great Britain. However, very different types of institutional entity can be included under the generic label of science park, such as private/public institutions created with the aim 1

For a comprehensive discussion on the role of gatekeepers see Allen (1977) and Morrison (2004).

13

Science Parks and Local Knowledge Creation

223

of encouraging the formation and growth of innovative (generally science-based) businesses, and actively engaging in transfers of technology or business skills to a ‘‘customer’’ organisation (Colombo and Delmastro 2002). A science park of this kind is not necessarily located in a particular geographical area, but is represented by a formal institution running research activities and hosting research laboratories; Innovation and Technological Centers can be included within this category. Finally, a third typology of science park is represented by public/private institutions whose sole aim is to act as an intermediary body between knowledge creators (e.g. university, research centers) and knowledge users. In this case, most of the literature refers to Business Innovation Centers (BIC).2 The deep differences in the nature of science parks explain the large variety of functions that they may develop with regard to: l

l

l

l

The diffusion and transfer of advanced technologies (e.g. best practice) among firms, sectors and regions, and support to the creative adoption of traditional innovation (a knowledge transfer function) The creation of radical innovation, aiming at contributing to a shift in the technological frontier (a knowledge creation function) The creation of a scientific environment, where firms accrue the benefits of being close (geographical proximity) to different sources of knowledge (a seedbed function) The creation of new technology-based firms (NTBFs), through spin-off processes where scientists move from research laboratories to privately-owned research activities (an incubator function for new firms)

A science park, regarded as a real-estate investment in a given geographical area, where R&D laboratories of public and private firms, research centers and universities are hosted, in principle performs all of the above functions, with the exception of the knowledge transfer function (Fig. 13.1.). A different and opposite case is represented by a science park which does not host research activities; but the most important function is that of knowledge transfer. Science and Research Parks, such as Innovation Centers, which host incubator functions and R&D laboratories, are generally able to fulfil most functions, although with more limited expected performance than: ‘‘geographical’’ science parks. We maintain that measurement of science park effectiveness should take into account their peculiarities. For example, it would be misleading and wrong to measure the ability of a science park which does not host any R&D activity according to its ability to create new knowledge, or be a seedbed for innovation. However, a word of caution has to be sounded. Although many science parks do not host or run R&D activities, they can still play a strategic role in supporting firms’ competitiveness by fostering firms’ capacity for innovation, and processes imitating and adopting best practice.

2

BICs have been set up by the European Union through DG XVI with the aim of supporting the conversion of technological knowledge into commercial knowledge and the creation of new firms.

224

R. Capello and A. Morrison Knowledge creation function

Knowledge transfer function

Public/private institutions with no internal R&D activities such as BICs

Public/private institutions hosting R&D, such as innovation Centers

Geographical areas devoted to host R&D laboratories, like Sophia Antipolis

Seedbed function

Incubator function for new firms

Fig. 13.1 Typology and functions of science parks

13.2.2 Science Parks and Learning Processes Most empirical analyses on the effectiveness of Science Parks have mainly approached this issue by comparing samples of on-park and off-park firms and measuring their different innovative performance (Monk et al. 1988; Westhead and Batstone 1998; Lo¨fsten and Lindelo¨f 2002). The main aim of these studies is to capture the role that science parks play in the creation of knowledge. However, our main idea in this study is that science parks can play a strategic role in the innovative performance of firms by supporting, stimulating and increasing the number of channels through which knowledge develops and cumulates at a local level. In other words, science parks can play a very important role in enabling and reinforcing spontaneous mechanisms of local knowledge diffusion. In doing this, they would indirectly foster the innovative activity and performance of local firms. The literature on local knowledge diffusion has recently been enriched by many conceptual and empirical studies. There is agreement that ‘‘physical proximity’’ among firms plays a crucial role in improving their innovative capacity (Audretsch and Feldman 1996; Jaffe 1989). Space matters due to the existence of ‘‘knowledge spillovers’’. However, the way in which space is conceptualized through the concept of ‘‘knowledge spillovers’’ differs greatly, depending on the approach considered (Capello and Faggian 2005).

13

Science Parks and Local Knowledge Creation

225

Geographers and industrial economists interpret space as simple physical proximity among innovative actors. Solid econometric results prove the importance of physical proximity. In practical terms, this approach sees proximity to university laboratories and other research centers as providing nearby firms with easier access to scientific expertise and research results, thus facilitating transfer of research into commercial application (Colombo and Delmastro 2002). Many authors address this issue in their contributions. Audretsch and Vivarelli (1994), for instance, try to measure the effect of knowledge spillovers on innovation – measured in terms of new patents (using data on Italian firms) – and find a significant effect of these spillovers on small and medium sized firms. However, their definition of spillovers is quite limited, since it only covers physical proximity (physical distance) to universities or research centers.3 Autant-Bernard (1999) extends the definition of spillovers to include sectoral aspects – firms are close if they belong to the same sector. Again, as in Audretsch and Vivarelli (1994), he finds a significant positive relationship between knowledge spillovers – measured in terms of R&D expenditure and researchers of firms in the local area – and the innovative performance of firms. Another flourishing stream of studies has focused on agglomeration economies, in particular on the relative importance of localization versus urbanization economies.4 The two perspectives differ in the respect that the localization economies’ argument, following Marshall, points that specialization matters for the co-localization of firms and in turn for spurring knowledge externalities at local level (Glaeser et al. 1992). Conversely, the latter view, drawing on Jacobs, claims that diversity, as the one observed in cities, is crucial for cross-fertilization among different industries, and in turn for the generation of new ideas (Jacobs 1969). Despite recognizing that proximity to universities, research centers and other firms – belonging to the same or to different sectors – is important, what emerges from a critical review of the literature is that the existence of knowledge spillovers is explained purely in terms of the probability of contacts between economic actors, which increases in a limited geographical space. The second and older approach is linked to regional economic studies. In these studies the concept of proximity, in the form of agglomeration economies, has always been an important element in explaining location choices, and local economic performance.5 During the 1970s and 1980s agglomeration economies explained the performance of new industrial areas; in the literature using this

3

When dealing with knowledge spillovers, we refer to Jaffe (1989) as a seminal work, followed by, among others: Acs et al. (1994), dealing with the capacity of large vs. small firms to exploit knowledge spillovers; Audretsch and Feldman (1996) and Feldman and Audretsch (1999), dealing with the importance of diversified vs. specialised knowledge spillovers; and Anselin et al. (2000), dealing with the definition of the physical distance over which knowledge spillovers disappear. For a recent review on the role of knowledge spillovers on regional development, see de Groot et al. (2001). 4 For a comprehensive review of these debates and some empirical results see (Ejermo 2005). 5 For a comprehensive review of these theories, see Capello (2004, Chap. 8).

226

R. Capello and A. Morrison

approach, territory is analyzed as an active element in economic development, being a productive resource in itself and a source of advantage for firms. In more recent times, territory has been conceived as a support for innovation activity, being able to decrease uncertainty and risks accompanying innovative processes. In this view, space is a complex concept, which does not only refer to geographical proximity, measured in terms of both physical and time distance.6 Space is interpreted in terms of relational proximity, defined as the ability of local firms, institutions and people to put in place strong local relationships – market relationships, power relationships, and co-operation.7 These relationships, alternatively called ‘‘relation capital’’, underlie any process of collective learning. Collective learning is the territorial counterpart of learning in an industrial context; it is thought of as the vehicle for knowledge transmission, both in a temporal and in a spatial dimension. In the former dimension, the transfer of knowledge is guaranteed by continuity; in the latter by interaction among agents, which guarantees transmission among individuals and firms and which becomes, in the case of the milieu, an element for the spatial transfer of knowledge.8 The channels through which knowledge develops locally are considered to be: l

l

6

The local labor market. The local labor market plays an important role within the local production system, as the high internal turnover of specialized labor. Low external mobility assures cross-fertilization processes for firms and professional upskilling for individuals; local know-how grows through a collective and socialized process, but this is subject to the risks of isolation and locking-in unless external energy is also captured through selected external co-operation linkages. Stable linkages between suppliers and customers. Stable input–output relationships generate a codified and tacit transfer of knowledge between suppliers and customers – it accumulates over time and defines patterns of incremental innovation which feed a specific technological trajectory. In this case also, comparison with firms’ technological trajectory is straightforward. As Aydalot (1986) suggested, the innovation process in a territorial entity such as a milieu is a process of ‘‘rupture/filiation’’ (break and continuity): if an innovation is a break

A vast literature exists on the best measurement of geographical distance. This has been for a long time measured in terms of physical distance (in kilometres); in recent times, an additional concept has been studied, that of accessibility , measured in terms of time to overcome a certain physical distance (among others, see Andersson and Ejermo (2004); Spiekermann and Wegener (2006). 7 A vast literature has in recent years underlined that pure physical proximity (or contact probability) is not enough to explain local knowledge spillovers. A large number of contributions have emphasized the role of social elements to explain knowledge transfer at the local level, defined as institutional proximity (Lundvall 1992; Lundvall and Johnson 1994) or relational proximity (Camagni, 1999). 8 When dealing with relational capital and innovation we refer to the works of Camagni (1991); Keeble and Wilkinson (2000); Lawson and Lorenz (1999); Camagni and Capello (2002). For the concept of organizational and cultural proximity, also see the French school on ‘‘la proximite´’’, among others Rallet (1993).

13

l

Science Parks and Local Knowledge Creation

227

with a pre-existing situation, economic creativity and innovation potential have their seeds in the local accumulated knowledge and know-how acquired over time. Intense innovative interactions with suppliers and customers and by mechanisms of local spin-off. Theoretically, a spin-off is defined as a new independent firm fulfilling two criteria (Perhankangas and Kauranen 1996; Dahlstrand 2000): (a) the startup of a new business by an agent previously belonging to another local firm, and (b) the derivation of a new business idea due to the previous employment of the founder. Local milieux provide both the social and the market preconditions for this phenomenon to take place: from the social point of view, high trust and a common sense of belonging to the same cultural community make this process acceptable.9 Local market conditions, such as stable interactions with suppliers known from a previous job, receptive local demand for particular products developed in a previous job, and the presence of external economies, assure locational advantages, guarantee the achievement of profits and thus give rise to chances for survival on the local market.

Beyond this set of mainly informal, ‘‘un-traded’’ relationships – among customers and suppliers, among private and public actors – and a set of tacit transfers of knowledge taking place through individual professional mobility and inter-firm imitation processes, another knowledge acquisition channel has been highlighted in the literature. More formalized, mainly trans-territorial co-operation agreements – among firms, among collective agents and among public institutions – in the field of technological development, vocational and on-the-job training, infrastructure and services provision are important channels for achieving new knowledge. In transterritorial networks, partners are single and selected economic units – enterprises, banks, research centers, training institutions or local authorities – where the locational element is only one of many elements defining the unit. At first glance, therefore, these networks only link together different economic actors, with no necessary relation to space. But when the location of a unit takes on significant meaning, inasmuch as it reveals a set of relations which generate territorial development and identity (e.g. Apple at Cupertino, Silicon Valley) and when these network relations start to multiply, they do become territorial. When carefully observed, the identity of the local milieu often prevails over the identity of the individual partner, stressing the importance of the territorial: the strategic importance of links with a company in Silicon Valley resides more in the opening of a ‘‘technological window’’ in Silicon Valley than in access to that specific company’s know-how. This second kind of network can define a process of ‘‘learning through networking’’. Through strategic alliances, non-equity agreements and technological cooperation, firms are able to capture some of the necessary assets from outside, overcoming the costs of internal development. This model is in a sense intermediate between internal and collective learning, it opens the firm to the general context, but maintains it within a set of selected and targeted relationships.10 The continuity 9

On the social homogeneity of local districts a vast literature exists. See among others, Bagnasco and Trigilia (1984); Becattini (1979, 1990).

228

R. Capello and A. Morrison

element is generated by the relative complexity of processes involved in setting up the terms of the cooperation contract, the clauses and sanctions for excluding opportunistic behavior; this promotes a long-term horizon and relative stability in the agreements. On the other hand, the knowledge transfer element is generally seen as rapid and powerful due to the complementarity of the different partners in the cooperation network. This learning channel helps local firms to avoid being locked into old knowledge, since it allows new and advanced knowledge to be introduced into the area. When science parks are created to perform innovation and knowledge diffusion functions, their primary role should be to support and foster direct, but more importantly indirect channels, i.e. all socialized processes of knowledge creation and diffusion. In other words, science parks should be able to participate in processes of: Creating both vertical and horizontal stable linkages among firms, both at the local and at the international level Helping the transparency and information of the local labor market Giving support to spin-off activities

l

l l

Figure 13.2 reports the role science parks may play in supporting local relational activities, on the one hand, and external trans-territorial networks, on the other. Different combinations of the two give rise to different learning trajectories, namely:

Role of SP in generating external networking for innovative activities

‘Learning through networking’

‘Collective learning’

‘No learning through Science Parks’

‘Localized collective learning’

Role of SP in generating local relationships for innovative activities

Fig. 13.2 The role of science parks in learning processes: possible alternatives 10

On the concept of ‘‘firms’ networking’’ see among others Chesnais (1988); (Gordon 1989).

13 l

l

l

l

Science Parks and Local Knowledge Creation

229

A ‘‘network learning trajectory’’, when science parks are able to strengthen relationships between local actors and agents outside the area, through which new and advanced knowledge can be acquired by local firms. A ‘‘localized collective learning trajectory’’, when, conversely, science parks support solid and long lasting relationships among local actors, both vertically and horizontally. In this way, knowledge will cumulate in a socialized (collective) way around a well defined technological trajectory, giving rise to what industrial economists call ‘‘localized learning’’. Firms will search for new technological solutions around well known technological and geographical boundaries if no new and radical knowledge is inserted locally. A ‘‘collective learning trajectory’’, when science parks are able to strengthen both local and external relationships. In such a case, socialized learning processes also encompass new technological solutions which drive towards paradigmatic change in technological trajectories brought into the area and shared among local actors. A ‘‘no-learning trajectory through science parks’’, when science parks do not play any role either supporting local relationships or external networking.

In our understanding, the effectiveness of science parks can be defined as their ability to enter into and support socialized processes of knowledge creation. The greater their capacity to have an active role in ‘‘collective learning’’ processes – by supporting channels of local knowledge transfer, such as relationships among local actors, and by enhancing long distance relationships – the greater their effectiveness. When science parks play a role in localized collective learning processes, they risk helping firms’ competitiveness only in the short term, but do not help local actors to avoid lock-in mechanisms or to jump to more advanced technological trajectories, which would assure long-term competitiveness. On the contrary, when science parks only act on networking processes of learning, they neglect the innovation transfer function, and are thus not actively involved in socialized mechanisms of knowledge diffusion.

13.2.3 Effectiveness of Science Parks: An Overview of Propositions In the previous sections a definition of the effectiveness of science parks was proposed. From this definition we can easily deduce our first testable proposition. Proposition 1. We expect the innovative capacity of local firms to be influenced by the active involvement of science parks in socialized processes of local knowledge diffusion and in networking activities. This general statement is true under certain conditions. The first one we refer to is firm size. Socialized processes of knowledge diffusion are important mainly for small firms. In large enterprises, large-scale R&D functions and engineering

230

R. Capello and A. Morrison

departments act as information storage agents and select decision-making routines, because they are long-term units. More importantly, R&D functions serve as a firm’s memory, where knowledge is cumulated, embedded in routines, and transferred as tacit knowledge in the process of searching for new technological innovations, giving rise to specific technological trajectories. In these organizational structures, therefore, there are no reasons for processes of knowledge socialization to occur. Instead, in small firms the innovation searching function does not exist due to diseconomies of scale and the unpredictable and relatively short life of small firms. It is in this type of productive system that knowledge cumulates in a socialized way (e.g. in the local labor market; in the network of local customers and suppliers) (Camagni 1995). From these considerations, a second research proposition emerges: Proposition 2. The active involvement of science parks in socialised processes of local knowledge diffusion and in networking plays a greater role in the innovative activity of small firms than large firms. A further important element, which helps to assure the effectiveness of science parks, is the degree of relational capital existing in the area. This stems from a strong sense of belonging and a highly developed capacity for cooperation which is typical of institutions and agents with similar culture (Capello and Faggian 2005). In areas where this attitude is absent, the chances of a science park developing local cooperation is expected to be very limited. In areas where relational capital is very intense and functions efficiently, science parks risk having a superfluous role. From these considerations another testable proposition can be envisaged. Proposition 3. We expect that greater the role science parks play in the socialized processes of local knowledge diffusion and in networking, greater is the degree of relational capital existing in the local area. Last, but not least, a third important condition under which science parks play an active role in socialized processes of knowledge creation/diffusion is the reactive capacity of firms. As has been widely suggested by the literature, a firm’s capacity to exploit new knowledge depends to a significant extent on the level of prior related knowledge stored within the firm, which enables the value of new information to be recognized, assimilated and applied for commercial purposes. These abilities collectively constitute what has been labelled as ‘‘absorptive capacity’’ (Cohen and Levinthal 1990), and more recently reinterpreted in terms of knowledgerelatedness in order to take account of technological diversification processes occurring within firms (Breschi et al. 2003). These considerations bring us to a fourth testable proposition, i.e.: Proposition 4. Greater the involvement of Science Parks in socialized processes of local knowledge diffusion and in networking, greater is the absorptive capacity of a firm. In the following sections we test these research propositions through an empirical analysis.

13

Science Parks and Local Knowledge Creation

13.3

231

Database and Methodology

13.3.1 The Sample The empirical analysis was carried out in two areas where science parks operate, namely Pisa and Genova. In the Pisa area we selected firms from a variety of science parks.11 Overall these technological centers to a large extent internalize the research function. According to our taxonomy they perform most of the functions attributed to the ideal Science Parks (e.g. transfer function, incubator function, – see Sect. 13.2.1). Only one Science Park operates in the Genova area, the Science Park of Liguria.12 As against the Science Park from Pisa, it does not own research facilities, acting mostly as coordinator and promoter of best practices and innovationrelated initiatives.13 A sample of tenant firms was investigated along a vast array of dimensions.14 For this purpose we elaborated a questionnaire covering the following areas:15 l

l

l

11

Measures of a firm’s characteristics: year of establishment, number of employees, a variety of indicators of economic performance (e.g. sales, exports), competitive position, knowledge base, etc. Measures of a firm’s innovative behavior: input (R&D expenditure; licences) and output (patents, percentage of firm’s turnover related to product and process innovation) indicators of innovation Measures of milieu and networking learning processes: information about the importance of local and external knowledge sources (e.g. competitors, providers,

The special agency of the Pisa Chamber of Commerce, the local Innovation Relay Centre (IRC), the Technological Pole of Navacchio, the technological centre ‘‘Firenze Tecnologia’’, the CNA (National Confederation of Micro-entrepreneurs) of Firenze and the technological centre CPR (Centro Pisa Ricerche). 12 This science park was established in 1996 by the local regional authority (Regione Liguria) and the University of Genova. 13 We have to point out that although these Science Parks differ in terms of research capacity and functions, nevertheless all of them share the common goal of diffusing knowledge locally. Hence it makes sense to jointly analyze their effectiveness in performing the bridging and networking functions, which are our main research questions. 14 The sample was selected by the CPR and the Engineering Faculty of the University of Genova using the list provided by PS. It is worth stressing that we used the term clients for firms included in those lists. Hence, contrary to common usage, we also use the term client to define firms that do not operate in the science park precinct, or that may have sporadic contacts with the science park. This is due to the specific features of the knowledge facilitators investigated in this survey. 15 The questionnaire has been built according to the guidelines proposed by the Oslo manual and the CIS (Community Innovation Survey) (OECD-EUROSTAT, Proposed guidelines for collecting and interpreting technological innovation data – Oslo Manual, OECD, Paris, 1997).

232

R. Capello and A. Morrison Table 13.1 Sample distribution by firm size, sector and location Variable Numbers Percent Size <20 121 75.63 Between 20 and 50 18 11.25 >50 21 13.13 Firm’s location Genova Pisa Sectorsa Others Specialised suppliers Supplier dominated Scale intensive Science based Low skill services High skill services a We introduce additional classes to 1984)

l

l

l

80 80

50 50

4 2.52 11 6.92 21 13.21 47 29.56 16 10.06 4 2.52 56 35.22 the Pavitt taxonomy (Pavitt

clients, universities, knowledge facilitators), which contributed to recent product and process innovations Measures of Science Park effectiveness: the role of science parks in connecting clients with relevant actors for developing the innovation (e.g. local and external competitors, providers, clients, universities, knowledge facilitators) Measures of relational capabilities: the percentage of relationships with local actors, in particular with customers and providers and their contribution in terms of relevant knowledge; the characteristics of the local labor market Measures of a firm’s linkages with science parks: frequency, typology of information/knowledge accessed through science parks, obstacles to knowledge acquisition, etc.

The survey was conducted over a period of 3 months covering 160 firms equally distributed between the two areas Genova and Pisa (80 firms each).16 A large number of firms belong to high-tech sectors (Table 13.1), although each geographic area presents some peculiarities. For example, we observed that sampled firms from Genova are skewed towards ‘‘old economy’’ sectors (e.g. oil, metal products, machinery), whereas in Pisa the majority of firms fall into the high-skill service sector. As far as size is concerned, few differences emerge: the vast majority are small or medium-sized firms and most established in the last two decades.

16

Data were collected by the Centro Pisa Ricerche (CPR) and by the Engineering Faculty of the University of Genova. Questionnaires were submitted through face-to-face interviews to firms’ technical staff and owners. The survey was administered during November, 2003 and January, 2004.

13

Science Parks and Local Knowledge Creation

233

13.3.2 Description of Variables Table 13.2 reports the description of the variables included in the empirical analysis. We built a set of conventional explanatory variables, a set of key variables – measuring the role of science parks in socializing knowledge – and an independent variable, which measures a firm’s performance. The following ‘‘explanatory variables’’ are taken into consideration: l

l

l

l

l

A variable of absorptive capacity (AC), constructed as a dummy variable coded 1 if a firm’s R&D expenditure is above the sample mean, 0 otherwise. A variable of firm size (DIM), which is coded 1 if a firm’s size is above 20 employees, 0 otherwise. A variable of relational capital (REL), which assumes value 1 if the share of local linkages of the firm is above the sample mean, 0 otherwise. A variable of local collective learning (RELLOC), measured in terms of collaborations with local firms aimed at developing a specific innovation. It is a binary variable assuming value 1 if firms assign a high score to local firms, 0 otherwise. A variable of external collective learning (RELEXT), measured in terms of collaborations with external research centers aimed at developing a specific innovation. It is a binary variable assuming value 1 if firms assign a high score to external relationships with these centers, 0 otherwise.

As far as the key determinants are concerned, a number of variables interpreting the ability of science parks to connect local firms to either local or external sources of knowledge (the bridging and networking functions respectively) are introduced. Dummy variables are constructed as follows: l

l

l

The role of Science Parks in supporting local or external linkages, PSTLOC and PSTEXT respectively, are coded 1 if firms value science parks as a relevant factor for searching and identifying local or external sources of knowledge respectively (e.g. competitors, clients, universities), 0 otherwise. The role of Science Parks in supporting local or external linkages for small and medium firms, PSTLOCSME and PSTEXTSME respectively, are coded 1 if SMEs value Science Parks as a relevant factor for searching and identifying local or external sources of knowledge respectively (e.g. competitors, clients, universities), 0 otherwise. The role of Science Parks in supporting local or external linkages for firms having a high absorptive capacity, PSTLOCAC and PSTEXTAC respectively, are coded 1 if firms with high absorptive capacity value Science Parks as a relevant factor for searching and identifying local or external sources of knowledge respectively (e.g. competitors, clients, universities), 0 otherwise.

The dependent variable (INNOPD) measures the innovative performance of a firm. It is constructed as a binary variable coded 1 if firms introduced at least one radical product innovation over the last 5 years, 0 otherwise.

234

R. Capello and A. Morrison

Table 13.2 Description of variables Name Description Dependent INNOPD Product innovation variable

Explanatory DIM variables REL

AC

Key RELEXT variables

Firm size Relational capital

Absorptive capacity

External relational capital

RELLOC

Internal relational capital

PSTLOC

Science Park bridging function

PSTEXT

Science Park networking function

PSTLOCSME Science Park bridging function for SMEs

PSTEXTSME Science Park networking function for SMEs

PSTLOCAC

Science Park bridging function for high AC firms

PSTEXTAC

Science Park networking function for high AC firms

Construction Binary dummy coded 1 if firm introduced at least one product innovation over the last 5 years, otherwise 0 Binary dummy coded 1 if firm has more than 20 employees, 0 otherwise Binary dummy coded 1 if firm’s share of vertical (with customers and providers) and horizontal linkages (with other firms) is above the sample mean, 0 otherwise Binary dummy coded 1 if firm’s R&D expenditure divided by sales are above the mean, 0 otherwise Binary dummy coded 1 if firm assigns a high score to external links with research centers as sources for innovation, 0 otherwise Binary dummy coded 1 if firm assigns a high score to local horizontal links as sources for innovation, 0 otherwise Binary dummy coded 1 if firm assigns a high score to the science park’s capacity to connect local firms, 0 otherwise Binary dummy coded 1 if firm assigns a high score to the science park’s capacity to connect local firms with external research centers, 0 otherwise Binary dummy constructed multiplying PSTLOC and DIM. It assumes value 1 if DIM is equal to 0 and PSTLOC is equal to 1, 0 otherwise Binary dummy constructed multiplying PSTEXT by DIM. It assumes value 1 if DIM is equal to 0 and PSTEXT is equal to 1, 0 otherwise Binary dummy constructed multiplying PSTLOC by AC. It assumes value 1 if AC is equal to 1 and PSTLOC is equal to 1, 0 otherwise Binary dummy constructed multiplying PSTEXT by AC. It assumes value 1 if AC is equal to 1 and PSTEXT is equal to 1, 0 otherwise

13

Science Parks and Local Knowledge Creation

13.4

235

The Bridging and Networking Functions of Science Parks: Some Descriptive Results

In this section we report some data describing the role of science parks in carrying out bridging and networking functions. Results show that the effectiveness of science parks is strongly affected by firm’s characteristics; in particular to which we refer to relational capital, absorptive capacity and firm size. Figure 13.3 shows the percentage of respondents divided by firm size. Results undoubtedly show that small firms benefit more from the bridging function of science parks; nevertheless science parks seem to support larger firms in building arm’s-length relationships (networking function). Moving to the second dimension (i.e. relational capability), Fig. 13.4 shows that the greater the firm’s relational capability, the greater is the ability of science parks to connect these firms to either local or external sources of knowledge. Results suggest that the higher the relational capability, the higher the share of firms benefiting from the gatekeeping ability of science parks. Similar results can be seen in Fig. 13.5. This shows that the higher the absorptive capacity, the greater the importance assigned to science parks in carrying out their functions (i.e. bridging and networking).

13.5

Science Parks and Knowledge Transfers: Interpretative Results

13.5.1 The Determinants of Firms’ Innovativeness: The Role of Science Parks The theoretical propositions set out above have been analyzed here through econometric techniques; the aim is to test the impact of science parks on firms’ innovative performance. A logit model is run in order to estimate the probability of a firm introducing an innovation in terms of some explicative variables (e.g. firm size, relational and absorptive capacity, interaction with science parks). The model can be expressed with the help of the following equation: INNOPD ¼ a þ b1 DIM þ b2 AC þ b3 RELOC þ b4 RELEXT þ b5 PSTLOC þ b6 PSTEXT þ e1 ;

ð13:1Þ

where INNOPD is the radical product innovation, RELOC is the local relational capital; RELEXT is the external relational capital, DIM is the firm size, AC is the absorptive capacity, PSTLOC is Science Park’s bridging function between clients and other local firms and PSTEXT is Science Park’s networking function between clients and external research institutions.

236

R. Capello and A. Morrison

30% 25% 20%

bridging networking

15% 10% 5% 0% < 20

between 20 and 50

> 50

Fig. 13.3 The importance of SP in establishing linkages by firm size 35%

25% 20%

SP networking function

30%

15%

cluster 2

cluster 1 10% 0%

10%

20%

cluster 4

Mean 30%

40%

50%

60%

5%

70%

80%

90%

SP bridging function

cluster 3 0%

Fig. 13.4 The importance of SP in establishing linkages by degree of relational capital

35% 30% 25% 20%

bridging

15%

networking

10% 5% 0%

low relational capital

high relational capital

Fig. 13.5 The importance of SP in establishing linkages by degree of absorptive capacity

13

Science Parks and Local Knowledge Creation

237

Results of the estimates are presented in Table 13.3. Initially we test the explicative power of the locational explanatory variables (13.1). We then introduce different key and firm’s explanatory variables one at a time to capture their contribution to an explanation of a firm’s propensity to innovate.17 Local and external relational capital variables (RELLOC, RELEXT) always have a positive and significant effect on a firm’s propensity to introduce a radical product innovation. Our results confirm the established wisdom on this issue: strong relational capital enhances a firm’s innovativeness. Once inserted into (13.1), variables measuring the bridging function of science parks play a positive and significant role in explaining a firm’s propensity to innovate. In particular this is true for a subset of small firms [see PSTLOCSME, (13.2) and (13.3)]. This latter result strongly supports our first proposition. Science parks play a greater role in transferring knowledge to small firms than to large ones. Variables capturing the networking function of science parks (PSTEXT and PSTEXTSME), however, exhibit significant, though negative coefficients [see (13.4) and (13.5)], suggesting that by connecting tenant firms to external sources of knowledge, science parks reduce the probability that they will develop innovative products. Although counterintuitive, such a finding appears quite reasonable if compared with other survey outcomes, showing that firms distrust such collaborations for fear of losing strategic information. Equations (13.6)–(13.8) introduce a last but important explanatory variable, the absorptive capacity (AC), which exhibits significant and positive coefficients (13.6). The same result is obtained with the variable (PSTLOCAC) interpreting the role science parks play in promoting local learning through the bridging function (for firms having a high absorptive capacity); (13.7) in fact shows a positive and significant sign of this variable. When the role of science parks in generating external linkages – for firms having a high absorptive capacity – is examined, the networking function of science parks hinders rather than boosts local learning (PSTEXTAC). On this latter point, we may presume that the networking function can function effectively for a subset of firms. We will test this proposition later in the paper. Overall, our findings can be summed up as follows. The propensity of science park tenants to introduce radical product innovations is positively influenced by a firm’s characteristics (e.g. absorptive capacity). We confirm that the ability of science parks to perform a bridging function is relevant, especially for small firms. Nevertheless we still have to clarify some unresolved points. First, it is still unclear whether firm-specific characteristics, such as relational capital, affect science park mechanisms of transferring knowledge. Second, we have to understand what factors prevent science parks from effectively performing their networking function. In order to shed light on these latter points we carried out a cluster analysis. This aimed to single out homogeneous groups of firms – in terms of innovative

17

DIM and REL were excluded since they did not produce significant estimations.

Eq. 1 Coeff. 1.05a (2.88) 1.12a(2.52)

Internal relational capital External relational capital Science Park bridging function Science Park bridging function for SMEs Science Park networking function Science Park networking function for SMEs Absorptive capacity Science Park bridging function for high AC firms Science Park networking function for high AC firms Constant 0.97a (4.34) LR Chi2 17.82a Pseudo R2 0.082 No. obs. 160 Note: z-value in brackets a Significant at least at 5% level b Significant at 10% level

Variables

Table 13.3 Results from logit estimations

0.87a (3.48) 13.72a 0.074 134

Eq. 2 Coeff. 0.85a (2.16) 0.7 (1.48) 0.52 (1.15)

1.04a (4.54) 21.66a 0.1 160

0.91a (1.95)

0.83a (3.35) 19.16a 0.1 134

1.8a (2.43)

0.97a (4.34) 21.54a 0.1 160

1.39a (1.85)

1.29a (4.86) 24.81a 0.11 160

0.95a (2.64)

Dependent variable: product innovation Eq. 3 Eq. 4 Eq. 5 Eq. 6 Coeff. Coeff. Coeff. Coeff. 0.81a (2.17) 1.25a (3.1) 1.17a (3.14) 0.95a (2.64) 0.94 a (2.04) 1.2a (2.42) 1.43a (2.89) 1.02a (2.24)

1.03a (4.5) 21.26a 0.09 160

0.96b (1.83)

Eq. 7 Coeff. 0.89a (2.44) 0.98a (2.14)

1.62b (1.96) .99a (4.39) 22.03a 0.1 160

Eq. 8 Coeff. 1.23a (3.26) 1.35a (2.84)

238 R. Capello and A. Morrison

13

Science Parks and Local Knowledge Creation

239

performance, degree of relational capital, size and degree of absorptive capacity – which share similar patterns of learning.

13.5.2 Firm Size, Relational Capital and Absorptive Capacity in Science Park’s Processes of Knowledge Socialization In this section we aim to test the influence of firm size, relational capital and absorptive capacity on science park processes of knowledge socialization. We also intend to identify what factors explain how the networking activity of science parks functions properly. Cluster analysis allows homogeneous groups of firms presenting heterogeneous behavior to be singled out. Hence, it allows identification of those groups of firms assuming coherent or distant behavior with respect to our research proposition and their characteristics. Four main clusters are identified in our analysis (Table 13.4):18 Table 13.4 Cluster analysis results 1 2 3 4 Meanb Clustersa Size (1 large firms, 0 small firms) 74.3 28.9 0 13.3 24.4 Product innovation 31.4 52.6 30.6 73.3 40 Absorptive capacity 20.0 44.7 33.3 60.0 35.6 Relational capital 17.1 44.7 31.9 60.0 34.4 Science Park bridging function 8.7 31.4 14.8 80.0 25.4 Science Park networking function 13.0 14.3 0 33.3 9.7 Local relational capital 0 100 0 100 33.1 External relational capital 40 0 0 100 18.1 Science Park bridging function for SMEs 5.7 26.3 11.1 73.3 19.4 Science Park networking function for SMEs 14.3 13.2 5.6 46.7 13.1 Science Park bridging function for high AC firms 2.9 21.1 8.3 46.7 13.8 Science Park networking function for high AC firms 2.9 10.5 0 26.7 5.6 Firm location (GE=0; PI=1) 28.6 65.8 45.8 80 50 Sectorsc Others 5.9 2.6 1.4 0 2.5 Specialised suppliers 5.9 2.6 5.6 26.7 6.9 Supplier dominated 17.6 10.5 13.9 6.7 13.2 Scale intensive 41.2 31.6 29.2 0 29.6 Science based 8.8 10.5 11.1 6.7 10.1 Low skill services 2.9 2.6 1.4 6.7 2.5 High skill services 17.6 39.5 37.5 53.3 35.2 Number of observations within each cluster 35 38 72 15 160 a Percentage of positive answers. In italics values well below the mean; in bold values well above the mean b Sample mean value c We introduce additional classes to the Pavitt taxonomy (Pavitt 1984)

18

We grouped firms with respect to three factors: DIM, RELOC, RELEXT.

240

R. Capello and A. Morrison

The first group includes firms specialized in traditional sectors with networking behavior: this cluster is characterized by low innovative performance and large firms operating in scale intensive or supplier dominated sectors. External linkages, either developed directly or mediated by science parks, appear to be the main channel for accessing knowledge. Although they benefit from the gatekeeping function of science parks, their innovative capacity is poor. The second group is populated by innovative firms with networking and milieu behavior: this cluster matches almost exactly with sample mean values. It includes small innovative firms operating in both high tech and traditional sectors. Local and external networks of linkages with knowledge sources fuel learning mechanisms. Science parks play a relevant role in strengthening these relationships, in particular those at local level. The third group contains firms characterized by weak innovative performance coupled with weak local and external linkages: this cluster identifies a group of firms which are isolated with respect to both local and external sources of innovation. Their knowledge base is also poor, preventing them from seeking and absorbing complementary technological inputs. In addition, their poor ability and propensity to relate with the external environment may prevent them from benefiting from Science Parks services. Firms within this cluster are equally distributed over the two areas (Genova and Pisa) and represent almost half our sample. In the last group we have small dynamic firms specialized in high tech sectors and with milieu and networking behaviour: firms within this cluster define a small group of highly innovative firms specialized in ICT. They have entered a virtuous learning trajectory, where internal competencies enable them to establish fruitful linkages with local and external sources of knowledge. In addition they do not ignore the importance of knowledge facilitators; they have in fact established strong connections with science parks, exploiting more then any other cluster the bridging and networking functions provided by science parks.

l

l

l

l

These results confirm our research propositions for a subset of sample firms. As Table 13.4 clearly shows, a group of small firms (cluster 4) has higher relational and absorptive capacity, accompanied by a strong innovative performance and stable linkages with science parks. These firms are inserted into a fast learning path.19 We may conclude that our research propositions related to science park involvement in knowledge processes are satisfied for at least a small, but significant, subset of firms. Figure 13.6 sums up the main findings: the vertical and horizontal axes represent the science park networking and bridging functions and our four clusters are depicted according to the role science parks play in each of them. Clusters 2 and 4 clearly identify those firms following a collective learning trajectory. Interestingly, in this case we can observe a positive relation between the role of science parks in

19

We are aware that we cannot infer any causal links from this analysis.

13

Science Parks and Local Knowledge Creation

40% 35% 30% 25% 20% 15% 10% 5% 0%

241

bridging networking

low absorptive capacity

high absorptive capacity

Fig. 13.6 Different learning trajectories sustained by SPs

supporting local and external linkages and firm’s characteristics – small firms with strong relational capital, a high absorptive capacity and a high degree of innovativeness. Cluster 1 identifies a group of large firms that benefit from the networking function of science parks, though they are not inserted in collective learning processes. In conclusion, cluster 3 identifies a group of firms with a passive learning behavior. It includes firms having extremely low innovative performance coupled with a poor knowledge base. In this case, science parks fail to enhance any learning process. This passive behavior is somehow a by-product of firm-specific factors, such as low relational capital, coupled with low internal competencies, which also reduce their ability to search for external knowledge inputs. In these cases science parks seem unable to provide feasible alternatives and fail to implement effective measures or design appropriate services for solving firms’ needs.20

13.6

Conclusions

The purpose of this work was to evaluate the effectiveness of science parks, measured as the ability to support and enhance spontaneous mechanisms of local knowledge creation, (i.e. to support networking among local actors and with external agents). The effectiveness of science parks in supporting innovation activities has been widely studied in the literature. Our contribution to the large scientific debate on this subject has been developed in two main research directions. The first line of research suggests it is important to clearly define the typology of a science park when evaluating it. Indeed, science parks may differ to a great extent in terms of their statutory mission, so some have the broad aim of contributing to the breeding of a local economy, while others may have a narrower focus, e.g. that of incubating high-tech start ups. In addition, the motivation behind the establishment 20

Our findings are in line with the existing evidence on this case study (Buratti and Penco 2001).

242

R. Capello and A. Morrison

of science parks has sensibly changed over the last 30 years. In the early stages of their development, they were primarily intended as policy tools to re-industrialize or renew urban or regional areas. More recently, science parks have narrowed their scope of intervention, and specialized in few and more focused functions. For example, many recent experiences were related to the creations of ICT firms. Such a variety of aims entail that any attempt to assess science parks’ effectiveness has to clearly state the functions subject to evaluation. Thus, for instance, if a science park does not develop any research activity, it would be misleading and wrong to investigate its role in generating new knowledge or in creating a seedbed for innovation. The transfer function will be its main objective and it should be evaluated accordingly. The second issue we highlight is that science parks run the transfer function in different strategic ways. On the one hand, science parks are able to support and stimulate spontaneous local knowledge creation and transfer channels; on the other hand, they maintain an arm’s-length relationship with external sources of knowledge. These issues, however are still largely uncovered. Our empirical analysis, by providing prima facie evidence on role of two science parks in creating relationships among local actors, shows that firms’ characteristics are key determinants in explaining science parks effectiveness. We found that only a bunch of clients composed of small firms effectively makes use of science parks activities. Moreover, results also show that firms with high absorptive and relational capacity are those that benefit more from science parks’ bridging and networking functions. The purpose of this study was not to directly assess the policy measures designed to support or implement a science park though the findings from the empirical analysis and the wider investigation over the customers of the science parks provide some useful suggestions in this direction. Overall, it clearly emerges that science parks are far from being easy policy instruments for promoting innovation activities. Science parks in principle should have the aim of making firms aware of their technological needs and scan the environment in order to find solutions for those needs. However, it could be the case that such intervention is superfluous, that is local firms would have, independently from public intervention, satisfied their technological needs. This entails that the public money has been misallocated. It suggests that policy makers, and more broadly the stakeholders and promoters of science park initiatives, should carefully interact with the potential beneficiaries and monitor their needs, and consequently identify the specific kind of services which could match the local demand. However, we should also consider that science parks, as stated above, often pursue several competing goals, and accordingly their members adhere to an initiative for staying motivated in diverse ways. This may explain why our sample includes both clusters of firms nearly disconnected from science parks functions and clusters with strong links. In this case, it is reasonable to say that the former, which mainly include large firms, have become clients for some other institutional reasons other than knowledge transfer (e.g. prestige). All in all, any intervention to promote science parks should take into account the characteristics of the area where they are supposed to be settled and operate, in

13

Science Parks and Local Knowledge Creation

243

addition to the needs of potential customers. Only by tailoring their functions and mission to local needs can they be effective tools of innovation policy and local economic development.

Acknowledgements The authors acknowledge financial support from the Italian Ministry of University and Research (FISR project entitled ‘‘Efficiency and Effectiveness of Science Parks’’) and research contributions from the partners of the project (University of Genova, Pisa Research Centre, Liguria Science and Technology Park, Scuola Superiore Sant’Anna).

References Acs Z, Audretsch D, Feldman M (1994) R&D spillovers and recipient firm size. Rev Econ Stat 76:336–340 Allen TJ (1977) Managing the flows of technology: technology transfer and the dissemination of technological information within the R&D organisation. MIT, Cambridge, MA Andersson M, Ejermo O (2004) Sectoral knowledge production in Swedish functional regions ˚ (eds) Knowledge spillovers and knowl1993–1999. In: Karlsson C, Flensburg P, Ho¨rte S-A edge management. Edward Elgar, Cheltenham Anselin L, Varga A, Acs Z (2000) Geographic and sectoral characteristics of academic knowledge externalities. Pap Reg Sci 79(4):435–443 Appold SJ (2004) Research parks and the location of industrial research laboratories: an analysis of the effectiveness of a policy intervention. Res Policy 33:225–243 Audretsch D, Feldman M (1996) R&D Spillovers and the geography of innovation and production. Am Econ Rev 86(3):630–640 Audretsch D, Vivarelli M (1994) Small firms and spillovers: evidence from Italy. Revue d’Economie Industrielle 67:225–235 Autant-Bernard C (1999) Geographic knowledge spillovers and technological proximity. Paper presented at the international conference of knowledge spillovers and the geography of innovation: a comparison of national systems of innovation’’, University Jean Monet SaintEtienne, 1–2 July 1999 Aydalot Ph (ed) (1986) Milieux Innovateurs en Europe. GREMI, Paris Bagnasco A, Trigilia C (1984) Societa` e Politica Nelle Aree di Piccola Impresa: il Caso di Bassano. Arsenale, Venice Becattini G (1979) Dal Settore Industriale al Distretto Industriale. Alcune Considerazioni Sull’ Unita` di Indagine dell’Economia Industriale. Riv Econ Pol Ind 1:35–48 Becattini G (1990) The Marshallian industrial district as a socio-economic notion. In: Pyke F, Becattini G, Sengenberger W (eds) Industrial districts and inter-firm cooperation in Italy. International Institute of Labour Studies, Geneva, pp 37–51 Breschi S, Lissoni F, Malerba F (2003) Knowledge-relatedness in firm technological diversification. Res Policy 32:69–87 Buratti N, Penco L (2001) Assisted technology transfer to SMEs: lessons from an exemplary case. Technovation 21:35–43 Camagni R (1991) Local ‘‘milieu’’, uncertainty and innovation networks: towards a new dynamic theory of economic space. In: Camagni R (ed) Innovation networks: spatial perspectives. Belhaven-Pinter, London, pp 121–144 Camagni R (1995) Global network and local milieu: towards a theory of economic space. In: Conti S, Malecki E, Oinas P (eds) The industrial enterprise and its environment: spatial perspectives. Avebury, Aldershot, pp 195–214

244

R. Capello and A. Morrison

Camagni R (1999) The city as a milieu: applying the Gremi approach to urban evolution. Rev Econ Reg Urbaine 3:591–606 Camagni R, Capello R (eds) (2002) Apprendimento Collettivo e Competitivita` Territoriale. FrancoAngeli, Milan Capello R (2004) Economia Regionale. Il Mulino, Bologna Capello R, Faggian A (2005) Collective learning and relational capital in local innovation processes. Reg Stud 39(1):75–87 Chesnais F (1988) Technical cooperation agreements between firms. STI Rev OECD 4:51–120 Cohen WM, Levinthal DA (1990) Absorptive capacity: a new perspective on learning and innovation. Adm Sci Q 35:128–152 Colombo MG, Delmastro M (2002) How effective are technology incubators? Evidence from Italy. Res Policy 31:1103–1122 Dahlstrand A (2000) Large firm acquisitions, spin-offs and links in the development of regional clusters of technology-intensive SMEs. In: Keeble D, Wilkinson F (eds) High technology clusters, networking and collective learning in Europe. Ashgate, Aldershot, pp 156–181 De Groot H, Nijkamp P, Acs Z (eds) (2001) Knowledge spillovers, innovation and regional development. Pap Reg Sci Assoc 80(3) (special issue) Ejermo O (2005) Technological diversity and Jacobs’ externalities hypothesis revised. Growth Change 36(2):167–196 Feldman M, Audretsch D (1999) Innovation in cities: science-based diversity, specialisation and localised competition. Eur Econ Rev 43:409–429 Glaeser EL, Kallal HD, Scheinkman JA, Shleifer A (1992) Growth in cities. J Polit Econ 100 (6):1126–1152 Gordon R (1989) Entrepreneurs, firms and the social foundation of innovation. Sociologie du Travail 1:25–48 Jacobs J (1969) The economy of cities. Vintage, New York Jaffe AB (1989) Real effects of academic research. Am Econ Rev 79:957–970 Jones C (1996) The theory of property-led local economic development policies. Reg Stud 30:797–801 Keeble D, Wilkinson F (eds) (2000) High-technology clusters, networking and collective learning in Europe. Ashgate, Aldershot Lawson C, Lorenz E (1999) Collective learning, tacit knowledge and regional innovation capacity. Reg Stud 33:305–317 Lo¨fsten H, Lindelo¨f P (2002) Science parks and the growth of new technology-based firmsacademic-industry links, innovation and markets. Res Policy 31:859–876 Lundvall BA (1992) Introduction. In: Lundvall BA (ed) National systems of innovation. Towards a theory of innovation and interactive learning. Pinter, London, pp 1–19 Lundvall BA, Johnson B (1994) The learning economy. J Ind Stud 1(2):23–42 MacDonald S (1987) British science parks: reflections on the politics of high technology. R&D Manage 17(1):25–37 Martin F (1997) Business incubators and enterprise development: neither tried nor tested? Small Bus Enterprise Dev 4:3–11 Massey D, Quintas P, Wield D (1992) High tech fantasies: science parks in society science and space. Routledge, London Monk CSP, Porter RB, Quintas P, Storey D, Wynarczyk P (1988) Science parks and the growth of high technology firms. Croom Helm, London Morrison A (2004) Gatekeepers of knowledge within industrial districts: who they are how they interact. WP.163, CESPRI-Universita` Bocconi Pavitt K (1984) Sectoral patterns of technical change: towards a taxonomy and a theory. Res Policy 13:343–373 Perhankangas A, Kauranen I (1996) Spin-offs from established corporations – a systematic classification of spin-off firms and a study of their contribution to industry growth. Paper presented at the European Network Meeting on Networks, Collective Learning and Research

13

Science Parks and Local Knowledge Creation

245

and Technology Development in Regionally Clustered High-Technology SMEs, Nice, 27–28 September 1996 Quintas P, Wield D, Massey D (1992) Academic–industry links and innovation: questioning the science park model. Technovation 12:161–175 Rallet A (1993) Choix de Proximite´ et Processus d’Innovation Technologique. Rev Econ Re´g Urbaine 3:365–386 Spiekermann M, Wegener M (2006) Accessibility and spatial development in Europe. Sci Reg – Ital J Reg Sci 5(2):1–25 Vedovello C (1997) Science parks and university–industry interaction: geographical proximity among agents as a driving force. Technovation 17(9):491–502 Westhead P, Batstone S (1998) Independent technology-based firms: the perceived benefits of a science park location. Urban Stud 35(12):2197–2219

Chapter 14

The Low Participation of Urban Migrant Entrepreneurs: Reasons and Perceptions of Weak Institutional Embeddedness Enno Masurel and Peter Nijkamp

14.1

Introduction

Despite the large influx of ethnic minority (or migrant) entrepreneurs of various origins in the Dutch society, members of this group rarely join institutional collaborative business organizations. This is surprising, as institutional embeddedness and anchoring is usually seen as a powerful organizational mechanism for small and medium-sized enterprises (SMEs). In this paper we focus on franchise organizations, which ethnic entrepreneurs hardly ever join. It is, in contrast, noteworthy that the degree of mutual collaboration among native Dutch firms in the SME sector is generally very high. The motivation for our study stems from the fact that it is, from a social cohesion point of view, important to understand why a certain group within society acts very differently from the rest of society. Furthermore, it is also important to know whether this deviation is temporary or structural, and whether it will be solved by itself over time or only with external assistance (Davidsson 2002). Much has been written about the attitude and behavior of ethnic entrepreneurs in Western economies, mostly from a sociological point of view. However, to the best of our knowledge, the subject of this paper, viz. the lack of institutional collaboration in the form of franchise organizations among ethnic small firms, has never been dealt with till date. In this paper, we investigate the reasons for the weak institutional collaboration among small ethnic firms, by means of an empirical analysis based on semi-structured interviews with 40 ethnic (Turkish) entrepreneurs in Amsterdam (the Netherlands). Turks, who have already lived in large groups in the Netherlands since the 1960s, appear to have a strong inclination to engage in entrepreneurial activities, which means that this group is particularly relevant for the present study. The retail sector, with its low entry barriers, is by definition an important sector in this context. E. Masurel (*) Centre for Innovation and Sustainable Entrepreneurship, Free University e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_14, # Springer‐Verlag Berlin Heidelberg 2009

247

248

E. Masurel and P. Nijkamp

An important survival strategy for SMEs in the highly competitive retail grocery sector in the Western world is franchising (‘‘franchising’’ is used in this paper as a generic term that encompasses all institutional collaborative systems in the retail sector and also includes purchase (or buying) groups and voluntary wholesale organizations. Participating in franchise organizations enables entrepreneurs to compete with big chain stores and to negotiate more effectively with wholesalers and manufacturers. This participation can also lead to improvement of their innovative capacity. All in all, (more) participation in franchise organizations may also be beneficial from a macro point of view. In any case, differences in organizational behavior of firms calls for further investigation. In the light of these advantages, the question emerges: Why do Turkish entrepreneurs not join such organizations? This is the key question of our study. The paper is organized as follows. Following a brief and selective review of the most essential elements of ethnic entrepreneurship (Sect. 14.2), we outline the main features of franchising (Sect. 14.3). Next, several research questions are formulated (Sect. 14.4), which are tested by means of interviewing a sample of 40 Turkish entrepreneurs in Amsterdam. After a description of the empirical database (Sects. 14.5 and 14.6), the research questions are addressed empirically. This is done first from the point of view of the ethnic entrepreneurs themselves (Sect. 14.7), and then in terms of their perceived reasons for rejection by the franchise organizations (Sect. 14.8). Finally, our main conclusions and recommendations are formulated (Sect. 14.9).

14.2

Ethnic Entrepreneurship

14.2.1 Prefatory Remarks Ethnic entrepreneurship has been intensively studied in recent years. The roots of these studies may be found in earlier works dating back to the 1950s. Rinder (1958) stressed the fact that all societies are internally differentiated or stratified, as a result of (amongst other factors) their history, level of economic development, societal complexity, cultural values, and political views. Simmel (1950) mentioned that – throughout the history of economics – the stranger everywhere appears as the trade (or the trader as stranger); this obviously places the ethnic entrepreneur in a specific socioeconomic context. Bonacich and Modell (1980) defined ‘‘ethnicity’’ as a communalistic form of social affiliation, depending on the assumption of a special bond between people of like origins or on a disdain for people of dissimilar origins. Besides affiliation, another important reason may be found in solidarity based on shared class interest. Both forms of solidarity (ethnicity and class) cut across each other in complex societies. A set of ethnic and racial minorities share a comparable position in the social structure of the societies in which they reside. They are referred to as ‘‘middlemen minorities’’, and are often active in trade. The aspects of ethnic entrepreneurship that have been most extensively studied in the literature are: the entrepreneurs’ relationships with their clients; their acquisition

14

The Low Participation of Urban Migrant Entrepreneurs

249

of capital and labor; and their motivations (Deakins 1999). These subjects, of course, overlap to some extent and are also interlinked. The ethnic group, or network, appears to play an important role in the behavior of ethnic entrepreneurs, or as Waldinger et al. (1990a) put it: connections and regular patterns of interaction among people sharing a common national background or experience of migration influence the behavior of ethnic entrepreneurs. Their own ethnic or social group plays an important role for ethnic entrepreneurs (see Bru¨derl and Preisendo¨rfer 1998; Sullivan 2000). The core of this group is formed by the (extended) family (Ram 1994). In order to better understand the relation between the entrepreneurs and their group members, this relationship is operationalized into three items: clients, capital and labor. Motivation is an important aspect of any form of entrepreneurship, especially in ethnic entrepreneurship. Waldinger et al. (1990a) aimed to explain both the immigrant groups’ entry into business and their different fates in terms of their access to opportunities, group characteristics, and the embeddedness of opportunities and resources within a specific set of historical conditions encountered by immigrating groups. The opportunity structure may have to do with market conditions (e.g., demand for ethnic consumer products, and undersupplied or abandoned markets). Group characteristics deal with a complex of interacting economic, social and psychological factors, and with resource mobility (due, e.g., to close ties between co-ethnics and ethnic social networks).

14.2.2 Clients Ethnic loyalties, informal networking, and communication patterns within the ethnic community provide an ethnic firm with potential competitive advantages. However, the literature also points to ambivalent relationships among ethnic entrepreneurs and their ethnic clientele. Dyer and Ross (2000), e.g., noted ambivalent signals by business owners in such relationships. On the one hand, these entrepreneurs commented favorably on the loyalty of their co-ethnic clients, expressed a preference for these clients, and had a good relationship with them and also many business connections. On the other hand, the same entrepreneurs appeared to cast doubts on their coethnic clients’ loyalty, their inappropriate demands, their resentment of the entrepreneurs’ financial success, their spreading of unfavorable rumors, and the generally poor image of the ethnic enterprises in the minds of their fellow native countrymen. Donthu and Cherian (1994) also noted specific sentiments within co-ethnic groups. They reported that, in the United States, Hispanics who have strong community identification were more likely to patronize Hispanic vendors than those who have weak community identification. The same was true about loyalty to brands used by family and friends: they were more influenced by targeted media messages and were less concerned with purely economic value. However, the competitive advantage of ethnic loyalty also makes the ethnic entrepreneurs vulnerable. In his study of Moroccan entrepreneurs in Amsterdam, Aakouk (2000) concluded that adjustment of their marketing mix (and especially of

250

E. Masurel and P. Nijkamp

their assortment) might enable ethnic entrepreneurs to go beyond their own ethnic market segments. Paradoxically, this could also imply alienation from their trusted ethnic clients. But, it may be possible that future generations of ethnic entrepreneurs will have less problems with institutional differences between the internal and external orientation of their business.

14.2.3 Capital and Labor Important aspects of the relationships within the ethnic group also concern the input variables of capital and labor. Van Delft et al. (2000) revealed that social networks are ethnic-related attributes that may provide advantages. These networks appear to be multifaceted and flexible, and offer good possibilities for the efficient acquisition of finance and the recruitment of personnel. In general, ethnic businesses rely heavily on labor from their particular ethnic group or, more specifically, their (extended) family. Capital can be more easily borrowed informally in this setting. In addition, within the network of the ethnic group, individuals depend on an informal way of doing business and exchanging information, because there is mutual trust within the network. Lee et al. (1997) called this phenomenon the ‘‘social resources explanation’’: the success of ethnic minority business can in part be explained by the existence of such social resources as rotating credits, a protected market, and a labor source. Deakins et al. (1997) stressed that constraints to successful diversification and development of ethnic businesses center mainly around their ability to access resources (especially finance) and new markets. The use of existing networks can form the bridge to mainstream development within their adopted country. Through their networks of relatives and co-nationals, ethnic entrepreneurs have privileged, flexible access to information, capital and labor (Kloosterman et al. 1998). Basu (1998) found that the nature of ethnic entrepreneurial entry predominantly depends on the access to informal sources of capital and information, as well as on the entrant’s previous experience. Ruiz-Vargas (2000) concluded that non-native (immigrant-owned) businesses in Puerto Rico had better access than natives to credit markets, possibly because of their economic position and power within the ethnic community. Texeira (1998) mentioned not only friends and relatives as being ethnic resources but also written media like newspapers. It thus seems plausible that a complex array of co-ethnic input factors may form a possible stimulus for successful entrepreneurship.

14.2.4 Motivation Motivational factors form a third set of frequently studied issues in ethnic business. Ram (1994) argued that the social networks of immigrants, comprising community and family, play a major role in the operation of ethnic enterprises. Reliance on

14

The Low Participation of Urban Migrant Entrepreneurs

251

these networks may also be a response to the (assumed) presence of racism in the wide environment. Externally, the family is seen as a means of overcoming racial obstacles in the market, while, internally, it provides a flexible source of labor and a means of managerial discipline. According to Deakins (1999), the history of disadvantage and discrimination has led to the concentration of ethnic minority firms and entrepreneurs in marginal areas of economic activity. Johnson (2000) mentioned both culture and disadvantage theory in explaining why immigrants become self-employed. Rafiq (1992) saw socioeconomic status as a better explanation for ethnic minorities entering business: their socioeconomic status is often relatively low, which is in part determined by cultural factors, especially the low participation of women in the labor force. Therefore, Rafiq (1992) argued that culture has an indirect impact on entrepreneurship. Ethnic entrepreneurship has become a popular strategy in developing principles of self-reliance among ethnic groups, because it stimulates and encourages immigrants to take care of themselves, with only limited support from the government (van Delft et al. 2000). Kloosterman et al. (1998) stressed the fact that high levels of unemployment provide the motivation for an increasing number of immigrants to start their own businesses. This issue had already been discussed in the context of a dual labor market by Wilson and Portes (1980), who focused attention on the incorporation of new immigrants into the labor market. These authors confirmed the theories that considered new immigrants as mainly being additions to the secondary labor market, linked with small peripheral firms. They also introduced the possibility of the ‘‘enclave economy’’ in connection with immigrant-owned firms, defining an enclave as a self-enclosed immigrant community (see also Peterson and Roquebert 1993). Flexibility on the labor market is another success factor. Li (1993) mentioned the traditional culture of certain ethnic groups, as well as blocked mobility, as important reasons for the successful development of ethnic entrepreneurship. Yoon (1995) asserted that there are three interacting factors which promote the growth of ethnic entrepreneurship: blocked employment opportunities in the general labor market (because of the language barrier and non-transferable education and occupational skills); resource mobilization (from stable structures and strong family ties); and business structures (social networks).

14.3

Franchising and Other Collaborative Forms

The modern retail sector in the Western world cannot be imagined without SME collaboration, in the form of buying groups, voluntary wholesale organizations and franchise organizations. Although significant differences are present within these three collaborative forms (Stern et al. 1996), this paper will only deal with franchising as a generic term for all forms of commercial collaboration van Witteloostuijn (1995). And, although many different types of franchising exist

252

E. Masurel and P. Nijkamp

(e.g., product, trademark, business format, soft/hard), we will not deal with the differences of these separate types. Furthermore, it is worth mentioning that a complex system of fees may occur, with elements like initial franchise fee, royalty fees, advertising fees, sales of products, rental and lease fees and management fees. Lewison (1997) defined franchising as a continuing relationship in which the franchisor provides a licensed privilege to do business, plus assistance in organizing, training, merchandising and management in return for financial compensation by the franchisee. Services provided by franchisors to their franchisees include market survey, site selection, joint buying, management training, and advertising. There are various forms of franchising, some of them require very intensive forms of collaboration and some less intensive forms (‘‘hard’’ vs. ‘‘soft’’ franchising). Reviews of various choices regarding franchising can be found, inter alia, in Kaufman (1999); Stanworth and Curran (1999). Note that all four aspects of ethnic entrepreneurship which were mentioned in the previous sections (clients, capital, labor, motivation) are involved in the franchise concept. The empirical signals of the success of franchising efforts are ambiguous. Reijnders and Verhallen (1996) noted that membership of a franchise organization is lucrative: allied small retailers perform better than their non-allied counterparts, in that they tend to realize higher profits and to show a more professional, active market approach. Apparently, franchising is a balanced system of pros and cons. On the other hand, Bates (1995) concluded that franchise start-ups exhibit both higher rates of business failure and lower mean profitability than corresponding independent start-ups. In his opinion, the reason for this was that the popular franchising niches were already saturated. In general, collaboration in the market is concerned with the creation of a certain balance among competing firms: the existence of market power creates an incentive to organize another power to counter it (Galbraith 1980; Kent et al. 1982; Kirzner 1997; Shane 2003). This is illustrated by Masurel and Janszen (1998), who found that, as large firms tend to dominate the market, more SMEs participate in franchise organizations. They observed that 29.7% of all SMEs in the Netherlands were members of franchise organizations, and even as much as 79.0% in the grocery sector (1996 figures). This is not, however, the place to deal extensively with the franchise phenomenon. We are primarily looking for explanations as to why the ethnic entrepreneurs do not join these collaborative forms. For more details on franchising, the interested reader may refer to Gru¨nhagen and Dorsch (2003), Michael (1996, 2002, 2003), Hoy and Shane (1998), Shane (1998) and Lafontaine (1992). They also showed that there are many different types of franchising. Empirical evidence of ethnic entrepreneurs in the retail sector joining franchise organizations is hard to find, both in the Netherlands and in other Western countries. The general view is that their degree of participation is almost zero (Detailhandels Magazine 1995). Suyver et al. (1998) calculated that only 2% of ethnic entrepreneurs were members of a franchise organization. Although this information may be somewhat dated, it can be stated that their degree of participation is still very low. The main valid explanation is a mutual lack of familiarity,

14

The Low Participation of Urban Migrant Entrepreneurs

253

i.e., ignorance on both sides. This finding is not a political statement, but based on value-free observation. Information on the relationship between ethnic entrepreneurs and franchise organizations is scarce, and hardly any additional information on different ethnic groups is available in this context. Two reasons for this scarcity of information are that no specific systematic information on the background of the entrepreneur has been gathered, and there is also a great deal of distrust towards interviewers among ethnic entrepreneurs. These feelings of distrust are generally reflected in the low participation in surveys. This low participation is a serious source of concern, and therefore in our empirical fieldwork, as well as doing other things to solve this problem, we worked with a co-ethnic interviewer. In this paper we focus on one specific form of institutional collaboration, viz. franchising. In accordance with the Dutch situation, the phrase ‘‘franchising’’ will be used loosely, also covering more general forms of commercial collaboration, like buying groups. Other relevant forms of institutional business cooperation for ethnic entrepreneurs (e.g., trade associations (see Bennet and Robson 2001), and local shopkeepers’ associations) are not considered here. To date, the issue of weak institutional orientation among ethnic firms has not been dealt with in the literature.

14.4

Research Questions

As the literature on SMEs does not offer much systematic information about the non-participation of ethnic entrepreneurs in franchise organizations or other collaborative institutions, this makes it difficult to formulate in advance any clearly founded and testable hypotheses. It is not known whether it is a matter of unwilling, unable, unknown, or even unwanted. Therefore, the research we present here is of an exploratory character. Ultimately, we try to relate the research outcomes to the specific aspects of ethnic entrepreneurship, in this way linking theory and practice with each other. An empirical investigation into the reasons why ethnic entrepreneurs do not join franchise organizations forms the core of the research in this paper. Some possible or plausible explanatory reasons for this low participation were largely identified by internal brainstorming sessions, by a scan of the scarce literature, and by pilot interviews. Clearly, such reasons are not mutually exclusive and may overlap to some extent. The possible explanations as to why the ethnic entrepreneurs do not join franchise organizations can be formulated as follows: 1. 2. 3. 4. 5. 6.

They think joining such organizations will serve no useful purpose. They think membership will cost too much. Their business is too informal. They are afraid to lose their independence. They are not well-informed about this form of collaboration. It is contrary to their cultural background.

254

E. Masurel and P. Nijkamp

7. It is contrary to their religious background. 8. They think they will not be accepted by such organizations. The first four reasons have to do with the firm itself: these we will refer to as the ‘‘economic factors’’ which are concerned with the market. Reason 5 has to do with communication, reason 6 with culture and reason 7 with religion. The latter two factors come closest to typical ethnic traits. Finally, reason 8 is external to the entrepreneurs, i.e., it depends on the franchise organizations. This all-embracing reason is broken down in greater detail into a number of perceived reasons (given by the entrepreneur) as to why the franchise organization would not accept him (or her) in Sect. 14.8. These research questions have to be tested against empirical facts; this requires field research among ethnic entrepreneurs.

14.5

Turkish Entrepreneurs in Amsterdam

Much literature has been devoted to the diversity in attitude and behavior of ethnic groups. One thing is clear: ethnic entrepreneurship is a multifaceted phenomenon that has at least as many sides as there are different ethnic groups. We may refer to an abundance of studies, by among others Aldrich and Waldinger (1990); Barret et al. (1996); Basu (1998); Boissevain and Grotenbreg (1986); Boraah and Hart (1999); Curran and Blackburn (1993); Deakins et al. (1997); Deakins (1999); European Commission (2003); Johnson (2000); Lee et al. (1997); Ram and Deakins (1996); Waldinger and Aldrich (1990) and Waldinger et al. (1990b). The literature shows convincingly that it is not possible to distinguish one single model of ethnic entrepreneurship or of ethnic business, not even within one given ethnic group (see, e.g., Curran and Blackburn 1993 and Li 1993). As already mentioned, in our investigation we have chosen to study one specific group of ethnic entrepreneurs in Amsterdam, viz. those of Turkish origin. This group is illustrative of ethnic entrepreneurs in the Netherlands and is very recognizable in the Dutch socioeconomic landscape (Choenni and Choenni 1998). The first wave (or generation) of Turkish immigration to the Netherlands took place in the 1960s and 1970s. The shortage of unskilled laborers led the Dutch government to sign a treaty with the Turkish government concerning the immigration of ‘‘guest laborers’’. Following the initial solo male immigration period, the males were joined by their families in the 1970s and 1980s. Today, there are some 300,000 people of Turkish origin residing in the Netherlands (about 2% of the total population). In Amsterdam alone there are about 35,000 Turkish residents (5% of the population) (de Feijter et al. 2001). These figures are not entirely unambiguous and depend on the definition used; here, we deploy the limited, most common definition of ‘‘allochthonous’’: anyone born abroad and who has at least one parent born abroad, as well as anyone born in the Netherlands both of whose parents were born abroad (CBS 1998). It should be

14

The Low Participation of Urban Migrant Entrepreneurs

255

noted that there is a great deal of discussion concerning the definition of the term ‘‘ethnic’’, but this is not the place to deal with that question. Choenni (1997) noted that more than 10% of the Turkish working population in Amsterdam are entrepreneurs, and they constitute about 20% of all ethnic entrepreneurs in Amsterdam. This information is somewhat outdated, and their share has likely increased. The retail sector, with its low entry barriers, is important for Turkish entrepreneurs. Recent research has pointed out that Turks are the main group of ethnic entrepreneurs in the Netherlands (http://www.kvk.nl). They operate mainly in the hospitality sector, the retail sector and the temporary employment sector. Jansen et al. (2003) concluded that, despite certain disadvantages compared with the native Dutch population, immigrants from Turkey show the same rate of entrepreneurship. Although the Turkish immigrants have similar characteristics as other immigrants (from Morocco, Surinam and the Antilles), their rate of entrepreneurship is much higher than the rate of these other groups. Kruiderink (2000) concluded that the percentage of starters is far higher for ethnic groups than it is for domestic groups. However, their failure rate is also much higher. This results in a relatively low survival rate. Clearly, because ethnic entrepreneurship is a multifaceted phenomenon, a made-to-measure policy is imperative. Ethnic entrepreneurship is not a homogeneous phenomenon, and one may wonder to what extent Turkish entrepreneurs are representative of all ethnic entrepreneurs in the Netherlands. However, franchise organizations hardly ever seem to incorporate ethnic entrepreneurs of any kind in their membership. Furthermore, it should be noted that interviewing minorities is always difficult, as often there is suspicion about the use of the outcomes of this type of research. There is also a fear that information collected on informal activities is not in the interest of the ethnic entrepreneur. Werbner (1999) mentioned the occasional nature of their earnings, and the unreported, informal economy as the main reasons for false statistics in the context of ethnic entrepreneurship. We tried to cope with this problem by working with a co-ethnic interviewer. Many Turkish entrepreneurs have an Islamic background. Although this background is increasingly becoming mixed with Western values, it is important to note some of the main characteristics of Islam. Rice (1999) observed that Islam is not generally understood, and frequently misunderstood, but in fact contains essentially a complete socioeconomic system, with its own behavioral codes. In Islam, ethics appear to dominate economics. The teachings of the Koran (revealed by God to Mohamed in the seventh century) and the Sunnah (the recorded sayings of Mohamed) are the basis of the Islamic ethical system: the ‘‘moral filter’’, unity, justice, trusteeship, and the need for balance. Clearly, not all these factors act as manifest explanatory variables for an ethnic entrepreneurial behavioral code of conduct. But it is likely that a blend of such factors may lead to deviant market behavior (for more information, see Lewis and Algaoud 2001 and Igbel and Llewellyn 2001).

256

14.6

E. Masurel and P. Nijkamp

Characteristics of the Interviewees

A sample of 40 Turkish grocers in Amsterdam were interviewed in the period June– July 2001. The interviewer, also Turkish, had mastered both Turkish and Dutch, and occasionally assisted the respondents by providing with additional information. This improved the reliability of the answers given. Nevertheless, due to the cultural context of the respondents, there is always the possibility of misunderstanding. Thus, such information has to be treated with great caution. It should be added that not all respondents answered all questions. In total, 50 Turkish grocers were asked to participate in the interviews. The selection criteria were twofold: the respondents had to be Turkish, and had to be grocers. The main reasons why some of the entrepreneurs refused to cooperate were that they distrusted the interviewer, or that they were too busy in their shop. The selection process was not easy, as there is – for reasons of privacy – no official record in the Netherlands of Turkish entrepreneurs. The Yellow Pages were initially used to select the participants by family name, after which the interviewer visited their respective shops. Some participants were selected merely by walking around those neighborhoods where many Turkish entrepreneurs are concentrated. So some bias might have occurred, although this is offset by the fairly high response rate. The sample is therefore most likely typical of the Turkish grocery sector in the Netherlands. The questionnaire was available in both Turkish and Dutch. Only four respondents chose to use the Dutch version, and the remaining 36 preferred the Turkish version. Of the 40 entrepreneurs, 27 were between 25 and 39 years old, while 4 were younger than 24, and 9 were older than 40. Eight respondents had completed university or polytechnic education.1 Seventeen respondents had started their business themselves, mostly in the 1990s; two entrepreneurs mentioned that they had started their business as an extension of a current shop. The remaining 23 respondents indicated that their business was a takeover, mostly in the 1990s. Of the 40 respondents, 29 were the sole proprietors of the shop. Of the remaining group, the average number of proprietors was 1.7 (for these cases two or three owners). In 10 of the 11 cases where there was more than one proprietor, all of them were Turkish, and generally from the same region in Turkey. The exception was a Turkish entrepreneur from Bulgaria. As the respondents were all business owners, we call them entrepreneurs. The average number of people employed in the

1

The entrepreneur was also asked whether he belonged to the first or second generation of immigrants. There appeared to be a great deal of confusion about this question, and at the end of the project it became clear that a number of respondents interpreted this question differently from the way it was intended. They interpreted generation as referring to the decade in which they immigrated into the Netherlands (the migration of the 1960s and 1970s comprises more or less the first generation, and the migration of the 1970s and 1980s the second generation).

14

The Low Participation of Urban Migrant Entrepreneurs

257

business was 2.6 persons. Twelve of the 40 shops had no employees other than the owner(s). It should be noted that there is, in general, a number of what are referred to as ‘‘non-registered’’ working people in ethnic firms: they are not by definition ‘‘black workers’’, but can also be volunteers or family members (mostly working part-time). The average business had a floor space of 120 square meters, and 35 of the 40 owners had only one shop. The average age of the firms was 12 years. It was striking that, although the interviewees were grocers, most sales were obtained specifically from greengrocery, rather than from general grocery. Furthermore, there were substantial sales in both meat and dairy products. The other categories of sales were mainly cigarettes, soft drinks, and candy. This indicates a blurring of traditional sector differences. All respondents purchased products from the Dutch domestic market. Only three respondents also imported goods, mostly from Turkey. Almost all products were purchased from wholesalers, especially Dutch and Turkish wholesalers in the Netherlands. The respondents dealt with both Dutch and Turkish wholesalers, because they served both native and ethnic clients. The buying practice correlated with the nationality of the wholesaler: for Dutch products, the retailers turned mostly to native wholesalers, whereas for Turkish products they turned to Turkish wholesalers.

14.7

Research Responses and Interpretation

It is striking that none of the 40 interviewed entrepreneurs actually participated in a franchise organization or was a member of any other comparable collaborative organization. Some of them had tried to start some form of collaborative association in the past, but these efforts were not successful. The main reasons for this failure were: lack of interest by other retailers; lack of agreement among the potential members; lack of experience with this form of collaboration; lack of unity among the potential members; and, finally, lack of trust. Furthermore, a number of interviewees were prejudiced against Dutch institutions in general, viz. they believed they were not welcome or that the institutions would work against their interests. We will come back to this later. We will now address the research questions from Sect. 14.4. First, the average scores are presented in Table 14.1. We asked the entrepreneurs whether they agreed with a certain answer, on a five-point scale, which varied from ‘‘completely disagree’’ (score 1) to ‘‘completely agree’’ (score 5). Furthermore, on the basis of the Wilcoxon test, we also determined whether there are significant differences between the importance attached to the different answers. Note that not all questions were answered by all 40 respondents. The Wilcoxon test is a so-called non-parametric test, especially fit for dealing with small numbers. The main reason for not joining a franchise organization was that the respondents were not well-informed about this particular phenomenon. Note that the

258

E. Masurel and P. Nijkamp

Table 14.1 Reasons not to join a franchise organization by Turkish entrepreneurs Numbers Average Insufficient information about franchising 39 3.03 No chance of being accepted 39 2.95 No perceived benefits 38 2.95 Costs too high 38 2.89 Current business too informal 38 2.84 Afraid to lose independence 38 2.42 Tension with cultural background 38 2.26 Incompatibility with religious background 38 2.16 a Does not differ significantly at the 10% level from first item b Differs significantly at the 10% level from first item

P-value 1.000a 0.767a 0.745a 0.611a 0.470a 0.018b 0.002b 0.000b

average score for all the reasons is close to the answer ‘‘neutral’’, and not close to the convincing ‘‘(strongly) agree’’. The second reason given was that the respondents thought it was very unlikely that they would be accepted by franchise organizations. Additional evidence indicated that the market does not offer sufficient benefits to ethnic retailers (see Sect. 14.8). In this respect, it appeared to be a matter of simply believing they were unable to join. This reason does not differ significantly in importance from the first reason. The doubted usefulness of joining a franchise organization for the entrepreneurs was ranked as the third reason for not joining such an organization. However, in terms of significance, this reason appears to be just as important as the first two reasons. So, in this respect, the entrepreneur deliberately decided to choose not to join. The fourth reason given for not joining a franchise organization was the cost factor: it would be too expensive. So, even if the entrepreneur was willing to join a franchise organization, he would still not be able to do so. This response may correlate with scale-disadvantages: Turkish shops are generally small, in contrast with the rest of the Dutch retail sector. Note that this reason is equally as significant as the first one. Next, we found the informal way of doing business by the ethnic entrepreneurs was an obstacle to joining franchise organizations. This may be reflected in, e.g., the varied composition of their assortments, in contrast to the more limited composition of franchisees’ assortments. This latter reason does not differ significantly from the first four reasons but is ranked last of the group of main reasons for not joining (based on its average score). It has an average score which is closer to ‘‘neutral’’ than to ‘‘disagree’’. A gap occurs between these first five reasons and the next three. Fear of losing their independence is significantly less important, as is shown in Table 14.1. The most pronounced disagreement was expressed in the answers to the questions dealing with cultural and religious background. Note that the last three reasons are closer to ‘‘disagree’’ than to ‘‘neutral’’. This pattern may have to do with the fact that franchising is a rather unknown phenomenon to the ethnic entrepreneurs.

14

The Low Participation of Urban Migrant Entrepreneurs

259

In summary, communication, decision making by external parties, and economic reasons are primarily responsible for the fact that ethnic entrepreneurs do not join franchise organizations. Cultural and religious reasons as such hardly play a role. However, it is striking that the highest average of the reasons is still not higher than 3.03, which is close to ‘‘neutral’’. The lowest average is 2.16, which is slightly higher than ‘‘disagree’’. We also looked for differences within the group of respondents. The data set was segmented on the basis of age (with a cut-off point at 35 years old), generation (born in the Netherlands or younger than 6 years old at the time of immigration defines the first generation, the remainder being second generation), and year of immigration (before or after 1988). Only the combination of religious background and the year of immigration appeared to play a role: for respondents who had immigrated before 1988, religious reasons were more important in deciding not to join franchise organizations than they were for respondents who had immigrated after 1988. So we can state, in general, that our group of respondents is rather homogeneous. One crucial question is whether the typical aspects of ethnic entrepreneurship are reflected in this explanation. The answer to this is negative. We may conclude that the lack of institutional collaboration of ethnic entrepreneurs is not so much explained by the typical traits of ethnic entrepreneurship but merely by communication, by the perception of having no chance of being accepted and by economic factors. The typical ethnic traits are at best indirectly important in influencing communication, visualizing external decision making, and economic factors. As such, the typical ethnic traits (culture and religion) end up in the rear. Coming back to Waldinger et al. (1990a): the low degree of institutional collaboration of ethnic entrepreneurs is associated with their opportunity structure (in relation to communication, acceptance, and economic factors); it does not seem to have to do with their group characteristics, except perhaps indirectly.

14.8

Perceived Grounds for Rejection by Franchise Organizations

The uniformity of Turkish entrepreneurs in their reasons for not joining a franchise organization finds its counterpart in another intriguing issue, viz. whether Turkish entrepreneurs see their perceived acceptance by franchise organizations as a major obstacle. And therefore, our respondents were also asked the main reason as to why franchise organizations might not be willing to accept them. Clearly, this perception may differ from the actual reasons why they are not accepted, but it is this perception, which actually influences the daily behavior of the entrepreneur in his attitude to joining franchise organizations. In the end, it is the franchise organization that decides whether a firm is allowed to join or not. Its criteria mainly have to do with the characteristics of the firm and the market situation. The policy of franchise organizations to attract new member varies from passive to active.

260

E. Masurel and P. Nijkamp

We also note that we did not look whether the respondents were member of other organizations as an alternative. It is not unthinkable that co-ethnic networks have such a function. The main reason for not being accepted by franchise organizations is the idea that these organizations think that the market does not offer sufficient benefits to the ethnic firms, as indicated in Table 14.2. Whether this is true, is still an open question in two respects. First, is this really the reason why franchise organizations reject them? And, second, is it really true that the market does not offer sufficient benefits? In addition, there are other important supposed reasons for rejection that do not differ in strength significantly from the principal perception, in particular: l

l l l

l

Language deficiency among ethnic entrepreneurs, which makes communication too complicated and costly Insufficient specific knowledge of ethnic minorities by franchise organizations Presumed lack of records, diplomas or certificates among ethnic entrepreneurs Insufficient knowledge of management, legal procedures and rules, marketing, and technology among ethnic entrepreneurs Inferior educational level of ethnic entrepreneurs, that is too low for them to become a meaningful business partner

So the main perceived reason among Turkish entrepreneurs as to why franchise organizations do not accept ethnic entrepreneurs, viz. market conditions, is actually beyond the control of the entrepreneurs. The next reasons for non-acceptance seem to be rectifiable, i.e., language deficiency of ethnic entrepreneurs, and the knowledge of franchise organizations about ethnic people. The other reasons do not seem to be rectifiable at least in the short run: lack of records, diplomas and certificates; lack of knowledge and skills; and low educational level. We note that the scores of all six perceived reasons as to why franchise organizations would not accept ethnic entrepreneurs center around 3.00, which is the ‘‘neutral’’ score.

Table 14.2 Perceived reasons by Turkish entrepreneurs why franchise organizations would not accept ethnic entrepreneurs Numbers Average P-value No competitive advantages for ethnic entrepreneurs 40 3.23 1.000a Language deficiency of ethnic entrepreneurs 40 3.20 0.913a Insufficient knowledge about ethnic people 40 3.15 0.769a Lack of records, etc., by ethnic entrepreneurs 40 3.00 0.441a Lack of knowledge and skills by ethnic entrepreneurs 40 2.95 0.318a Low educational level of ethnic entrepreneurs 40 2.83 0.147a Tension with cultural background of ethnic entrepreneurs 40 2.70 0.060b Limited experience with the Dutch market 40 2.68 0.057b Inappropriate location of ethnic enterprises 40 2.58 0.018b Low returns of ethnic enterprises 40 2.50 0.013b Lack of solid financial basis for ethnic enterprise 40 2.40 0.003b Incompatibility with religious background 40 2.30 0.000b a Does not differ significantly at the 10% level from first item b Differs significantly at the 10% level from first item

14

The Low Participation of Urban Migrant Entrepreneurs

261

Finally, we address the less important perceptions: tensions with the sociocultural background of the ethnic entrepreneurs; their lack of knowledge of, and experience with the Dutch market; the unattractiveness of their site; insufficient market opportunities for good profitability; lack of a solid financial basis; and, finally, the tension with the religious background of the entrepreneur. All these scores are in the order of magnitude of 2.50, which fall in-between ‘‘neutral’’ and ‘‘disagree’’. Regarding perceived grounds for rejection by franchise organizations, we also looked for differences within the group of respondents. Only a few differences appeared to emerge. Those who immigrated before 1988 more frequently thought that lack of a solid financial basis played a role, as did those who were older than 35. Furthermore, those who were younger than 35 tended to believe more frequently that lack of records, diplomas or certificates would play a role, compared with the opinion of their counterparts who were older than 35. In general, however, the group of respondents appeared to be fairly homogeneous, with regard to their perceptions.

14.9

Conclusions and Recommendations

Ethnic entrepreneurs hardly ever join formal constellations of institutional collaboration in general, and franchise organizations in particular, a situation which may partly explain their high failure rate. The generality of our conclusions is subject to debate; however we are only dealing with one single group (Turks) in one single country (the Netherlands). Nevertheless, our approach has value for future research in this field, not in the least because of our research orientation, which originates from a broad review of the international literature. Poor communication, a low chance of being accepted by the external party, and economic market factors appear to be among the main reasons why ethnic entrepreneurs do not join franchise organizations. Poor communication is reflected in insufficient information about franchising. The perceived reasons as to why franchise organizations would not accept ethnic entrepreneurs is a mixture of economic factors, the difficulty of dealing with each other, and various deficiencies on the side of the ethnic entrepreneurs. The economic factors explaining why ethnic entrepreneurs do not want to join franchise organizations, are reflected in no perceived benefits, costs being too high, current business being too informal and the fear of losing their independence. The low chance of being accepted is reflected in the entrepreneurs’ perception of market conditions, which are beyond their control. But furthermore, there are also rectifiable reasons for rejection, which appear to play a role, especially concerning the development of language skills by the ethnic entrepreneurs, and the improvement of the franchise organizations’ knowledge about ethnic entrepreneurs. The other reasons do not seem to be rectifiable, at least in the short term: lack of records, diplomas and certificates; lack of knowledge and skills; and low educational level.

262

E. Masurel and P. Nijkamp

Altogether, these are the factors, which are responsible for the isolated position of the ethnic entrepreneurs and which would need to be dealt with if the aim is to increase the participation of ethnic entrepreneurs in franchise organizations. Our analysis has demonstrated that – as part of a research agenda – additional research among franchise organizations themselves about their views on ethnic entrepreneurs is necessary. Complementary research is also required to analyze the attitude to franchising of other ethnic groups, as we know that the phenomenon of ethnic entrepreneurship may be quite diverse among different ethnic groups. The process model introduced by Hood et al. (1993) may serve as a basis for this. In conclusion, the main reasons explaining why ethnic entrepreneurs do not join franchises stem from poor communication, the perception of rejection, and economic factors. Concerning the poor communication: this is a phenomenon that unfortunately often emerges in a multicultural society. In order to encourage the participation of minority groups in collaborative systems (not only franchise organizations but also trade associations, local shopkeepers’ associations, etc.), new tailor-made initiatives have to be developed. In the context of the retail sector, this task of encouraging ethnic membership of franchise organizations seems to be a natural role for wholesalers, with whom the ethnic entrepreneurs come together in their dealings. It might begin as a low-profile and light institutional effort, perhaps with bonuses for purchases and low membership fees, etc., for specific groups, but such an effort might have a clear benefit to all in the long run. Additional research among franchise organizations themselves about how they themselves view ethnic entrepreneurs is necessary. The economic factors in this context can hardly be denied: it is a fact that ethnic businesses are in general economically vulnerable and often show an informal way of doing business. This may also have to do with the typical characteristics of ethnic businesses (the role of the extended family and their motivation to start up their own business). Normally, after some time, ethnic business may upgrade and thus become more eligible to join franchise organizations and other native-dominated institutional collaborations so as to avoid lock-in behavior. The perceived preliminary conclusions and recommendations resulting from our survey bring us back to Simmel (1950): the stranger appears as a trader and usually has big problems becoming assimilated in the new society. Improving communication and suppressing prejudices may help to make the stranger a more accepted person in a new society. This will undoubtedly improve his competitive and innovative potential. The previous analysis has shown that institutional support systems may act as a significant support mechanism of economy dynamics and innovation. From an endogenous growth perspective one might argue that such institutional frameworks may become instrumental parameters in dedicated urban growth policy, where communication and participatory strategies may act as key conditions. The degree of flexibility and sharing of interest among ethnic groups is an important handle for new entrepreneurial perspectives among migrant entrepreneurs in the city.

14

The Low Participation of Urban Migrant Entrepreneurs

263

Acknowledgements The authors gratefully acknowledge the fieldwork by Kemal Tasdelen and the data analysis by Jan Holleman and Gabriella Vindigni.

References Aakouk M (2000) Marokkaanse ondernemers in Amsterdam: een inventarisatie van mogelijkheden (Moroccan entrepreneurs in Amsterdam: An inventory of possibilities). Masters Thesis, Free University, Amsterdam Aldrich HE, Waldinger R (1990) Ethnicity and entrepreneurship. Annu Rev Sociol 16:111–135 Barret GA, Jones TP, McEvoy D (1996) Ethnic minority business: theoretical discourse in Britain and North America. Urban Stud 33(4–5):783–809 Basu A (1998) An exploration of entrepreneurial activity among Asian small businesses in Britain. Small Bus Econ 10(4):313–326 Bates T (1995) Analysis of survival rates among franchise and independent small business startups. J Small Bus Manage 33(2):26–36 Bennet RJ, Robson PJA (2001) Exploring the market potential and bundling of business association services. J Mark Serv 15(3):222–239 Boissevain J, Grotenbreg H (1986) Culture, structure and ethnic enterprise: The Surinamese of Amsterdam. Ethn Racial Stud 9(1):1–23 Bonacich E, Modell J (1980) The economic basis of ethnic solidarity. University of California Press, Los Angeles Boraah VK, Hart M (1999) Factors affecting self-employment among Indian and Black Caribbean men in Britain. Small Bus Econ 13(2):111–129 Bru¨derl J, Preisendo¨rfer P (1998) Network support and the success of newly founded businesses. Small Bus Econ 10(3):213–225 CBS (1998) Statistisch jaarboek (Statistical yearbook). CBS, Voorburg/Heerlen Choenni A (1997) Veelsoortig assortiment (Multifaceted assortment). Het Spinhuis, Amsterdam Choenni AO, Choenni CES (1998) Allochtoon ondernemerschap (Allochtonous entrepreneurship). In: Scherjon DP, Thurik AR (eds) Handboek ondernemers en adviseurs in het MKB (Handbook for entrepreneurs and advisors in the SME sector). Kluwer, Deventer Curran J, Blackburn R (1993) Ethnic enterprise and the high street bank: a survey of ethnic business in two localities. Small Business Research Centre and the Midland Bank, Kingston Davidsson P (2002) What entrepreneurship research can do for business and policy practice. Int J Entrepreneurship Educ 1(1):1–20 de Feijter H, Sterckx L, de Gier E (2001) Nieuw Amsterdams Peil : Amsterdamse Sociaalculturele verkenningen (A new standard for Amsterdam: Socio-cultural exploration in Amsterdam). SISWO, Amsterdam Deakins D (1999) Entrepreneurship and small firms, 2nd edn. McGraw-Hill, London Deakins D, Majmudar M, Paddison A (1997) Developing success strategies for ethnic minorities in business: evidence from Scotland. New Community 23(3):325–342 Detailhandels Magazine (1995) Toegang tot ketens verloopt moeizaam (Access to chains is difficult). Detailhandels Magazine 11:36–37 Donthu N, Cherian J (1994) Impact of strength of ethnic identification on Hispanic shopping behavior. J Retail 70(4):383–393 Dyer LM, Ross CA (2000) Ethnic enterprises and their clientele. J Small Bus Manage 38(2):48–66 European Commission (2003) Entrepreneurship in Europe: green paper. Enterprise, Brussels Galbraith JK (1980) American capitalism. Sharpe, Armonk Gru¨nhagen M, Dorsch MJ (2003) Does the franchisor provide value to franchisees? Past, current, and future value assessment of two franchise types. J Small Bus Manage 41(4):366–384 Hood JN, Logsdon JM, Thompson JK (1993) Collaboration for social problem solving: a process model. Bus Soc 32(1):1–17

264

E. Masurel and P. Nijkamp

Hoy F, Shane S (1998) Franchising as an entrepreneurial venture form. J Bus Venturing 13(2): 91–94 Igbel M, Llewellyn DT (2001) Islamic banking and finance. Edward Elgar, Cheltenham Jansen M, de Kok J, van Spronsen J, Willemsen S (2003) Immigrant entrepreneurship in the Netherlands. EIM, Zoetermeer Johnson PJ (2000) Ethnic differences in self-employment among Southeast Asian refugees in Canada. J Small Bus Manage 38(4):78–86 Kaufman PJ (1999) Franchising and the choice of self-employed. J Bus Venturing 14(4):345–362 Kent C, Sexton D, Vesper K (eds) (1982) The Encyclopedia of entrepreneurship. Prentice-Hall, Englewood Cliffs Kirzner I (1997) Entrepreneurial discovery and the competitive market process. J Econ Lit 35(1):60–85 Kloosterman R, van der Leun J, Rath J (1998) Across the border: Immigrants’ economic opportunities, social capital and informal business activities. J Ethn Migr Stud 4(2):249–268 Kruiderink E (2000) Allochtoon ondernemen (Foreign entrepreneurship). ESB 85(18-2):144–146 Lafontaine F (1992) Agency theory and franchising: some empirical results. Rand J Econ 23 (2):263–283 Lee Y, Cameroen T, Schaeffer P, Schmidt CG (1997) Ethnic minority small business: a comparative analysis of restaurants in Denver. Urban Geogr 18(7):591–621 Lewis MK, Algaoud ML (2001) Islamic banking. Edward Elgar, Cheltenham Lewison DM (1997) Retailing, 6th edn. Prentice Hall, Englewood Cliffs Li PS (1993) Chinese investment and business in Canada: Ethnic entrepreneurship reconsidered. Pac Aff 66(2):219–243 Masurel E, Janszen RM (1998) The relationship between SME cooperation and market concentration: evidence from small retailers in the Netherlands. J Small Bus Manage 36(2):68–73 Michael SC (1996) To franchise or not to franchise: an analysis of decision rights and organizational form shares. J Bus Venturing 11(1):57–71 Michael SC (2002) Can a franchise chain coordinate? J Bus Venturing 17(4):325–341 Michael SC (2003) First mover advantage through franchising. J Bus Venturing 18(1):61–80 Peterson MF, Roquebert J (1993) Success patterns of Cuban–American enterprises: implications for entrepreneurial communities. Hum Relat 6(8):921–937 Rafiq M (1992) Ethnicity and enterprise: a comparison of Muslim and Non-Muslim owned Asian business in Britain. New Community 19(1):43–60 Ram M (1994) Unravelling social networks in ethnic minority firms. Int Small Bus J 12(3):42–53 Ram M, Deakins D (1996) African-Caribbeans in business. New Community 22(1):67–84 Reijnders WJM, Verhallen TMM (1996) Strategic alliances among small retailing firms: empirical evidence for the Netherlands. J Small Bus Manage 34(1):36–45 Rice G (1999) Islamic ethics and the implications for business. J Bus Ethics 18(4):345–358 Rinder ID (1958) Strangers in the land: social relations in the status gap. Soc Probl 6(3):253–260 Ruiz-Vargas Y (2000) Small business financing sources between immigrants and natives in Puerto Rico. Q Rev Econ Finance 40(3):387–399 Shane S (1998) Explaining the distribution of franchised and company-owned outlets in franchise systems. J Manage 24(6):717–739 Shane S (2003) A general theory of entrepreneurship. Edward Elgar, Cheltenham Simmel G (1950) The stranger. In: Simmel G (ed) The sociology of George Simmel. Free Press, New York Stanworth J, Curran J (1999) Colas, burgers, shakes and shirkers: towards a sociological model of franchising in the market economy. J Bus Venturing 14(4):323–344 Stern LW, El-Ansary AI, Coughlan AT (1996) Marketing channels, 5th edn. Prentice Hall, Englewood Cliffs Sullivan R (2000) Entrepreneurial learning and mentoring. Int J Entrepreneurial Behav Res 6(3):160–175

14

The Low Participation of Urban Migrant Entrepreneurs

265

Suyver JF, Lie A, Kwie JA (1998) Allochtoon ondernemerschap in de detailhandel (Allochtonous entrepreneurship in the retail sector). EIM, Zoetermeer Texeira C (1998) Cultural resources and ethnic entrepreneurship: a case study of the Portuguese real estate industry in Toronto. Canad Geogr 42(3):267–281 van Delft H, Gorter C, Nijkamp P (2000) In search of ethnic entrepreneurship opportunities in the city: a comparative policy study. Environ Plann C Gov Policy 18(4):429–451 van Witteloostuijn A (1995) Studies in industrial organization. Kluwer, Dordrecht Waldinger R, Aldrich H (1990) Trends in ethnic business in the United States. In: Waldinger R, Aldrich H, Ward R et al. (eds) Ethnic entrepreneurs. Sage series on race and ethnic relations, vol 1. Sage, London, pp 49–78 Waldinger R, Aldrich H, Ward R (1990a) Opportunities, group characteristics and strategies. In: Waldinger R, Aldrich H, Ward R et al. (eds) Ethnic entrepreneurs. Sage series on race and ethnic relations, vol 1. Sage, London, pp 13–48 Waldinger R, Aldrich H, Bradford WD, Boissevain J, Chen G , Korte H, Ward R, Wilson P (1990b) Conclusions and policy implications. In: Waldinger R, Aldrich H, Ward R et al. (eds) Ethnic entrepreneurs. Sage series on race and ethnic relations, vol 1. Sage, London, pp 177–197 Werbner P (1999) What color ‘‘success’’? Distorting value in studies of ethnic entrepreneurship. Sociol Rev 47(3):548–575 Wilson KL, Portes A (1980) Immigrant enclaves: an analysis of the labor market experience of Cubans in Miami. Am J Sociol 86(2):295–319 Yoon I-J (1995) The growth of Korean immigrant entrepreneurship in Chicago. Ethn Racial Stud 18(2):315–335

Chapter 15

The Location of Industry R&D and the Location of University R&D: How Are They Related? Charlie Karlsson and Martin Andersson

15.1

Introduction

At the same time as we can observe strong tendencies of globalization of R&D (Florida 1997; Cantwell 1998), we also observe strong spatial clustering of R&D and related innovation activities (Audretsch and Feldman 1996). The standard explanation in the literature of the clustering phenomenon is that clustering brings about external knowledge economies, typically in the form of knowledge flows,1 which tend to be spatially bounded (Jaffe et al. 1993). R&D is a typical innovation activity; irrespective of whether it is focused on new products or new processes. Industry – i.e., private firms – and universities are the major performers of R&D. It is well established in the literature that both university and industry R&D have a positive effect on innovation output (often measured by patent applications), but that the effect diminishes with distance because knowledge flows are spatially bound and that clustering is consequently an effective spatial configuration. Despite a vast literature on how university and industry R&D affect innovation output, the literature on how university and industry R&D are related to each other is rather limited. This paper focuses on possible spatial interdependencies between the location of university and the location of industry R&D. It seems rather straightforward to assume that industrial R&D might be attracted to locate near research universities doing R&D in fields relevant to industry. As far back as in the 1960s a number of case studies confirmed the important roles played by Stanford University and MIT for commercial innovation and entrepreneurship

C. Karlsson (*) Jo¨nko¨ping International Business School, Jo¨nko¨ping University e-mail: [email protected] 1

For reasons given in Sect. 15.2 we use the general term ‘‘knowledge flows’’ instead of the term ‘‘knowledge spillovers’’ commonly used in the literature.

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_15, # Springer‐Verlag Berlin Heidelberg 2009

267

268

C. Karlsson and M. Andersson

(Teplitz 1965; Wainer 1965; Shimshoni 1966; see also Dorfman 1983). Starting with Nelson (1986) a large number of formal studies presented evidence of a positive impact of university R&D on firm performance. Does industrial R&D function as an attractor for university R&D? There are actually several reasons as to why university R&D may grow close to concentrations of industrial R&D. First of all political decision-makers may decide to start or expand university R&D at locations where industry is already doing R&D. Second, we can imagine that industry doing R&D in a region might use part of their R&D funds to finance university R&D. Thirdly, universities in regions with industrial R&D might find it easier to attract R&D funds from national and international sources due to co-operation with industry. This means that there are relationships between industrial R&D and university R&D and vice versa. This was observed by Jaffe (1989), who modeled these relationships as a simultaneous system. However, we have found few other studies dealing with this problem. The study by Anselin et al. (1997) is an exception. Most studies have concentrated on the unidirectional effect from university R&D to industrial R&D and the outputs of industrial R&D in most cases measured in terms of the number of patents and neglected the possible mutual interaction. Against the background above, the purpose of this paper is to analyze interdependencies between industry R&D and university R&D in Sweden as regards location using a simultaneous equation approach.2 However, it has to be recognized that all decisions on where to start universities are taken by politicians and that many universities in Sweden were founded several hundred years ago. For this reason the paper does not focus on location decisions in terms of establishments of new industrial and university R&D departments or units. Instead, the research question is posed in the following manner: Does industrial R&D tend to expand in locations with high accessibility to university R&D and vice versa? An underlying conjecture in the paper is that the extent of links between industry R&D units and university research departments is a function of the physical proximity3 between the two and that many types of knowledge flows are mediated via links. It is further argued that links, in turn, may give rise to interdependencies between industry and university R&D. Of course, the extent to which link formation is determined by spatial proximity depends on the transaction’s contact intensity, but transactions involving knowledge – such as when a firm purchases R&D

2

Obviously, not all types of university R&D attract industrial R&D. There are reasons to believe that, in particular, university R&D in natural, technical and medical sciences attract industrial R&D but that there are also strong reasons to believe that there are variations between different sectors of industry in terms of how dependent their R&D is on proximity to university R&D. However, although a distinction between different science areas would be interesting, the present paper focuses on the aggregate pattern. 3 Many scholars use very crude measures of physical accessibility. By focusing on physical accessibility and relying on actual travel time distances in the accessibility calculations, we believe that proximity is measured in a coherent fashion, (see also Karlsson and Manduchi 2001; Andersson and Karlsson 2004).

15

The Location of Industry R&D and the Location of University R&D

269

services from universities – are highly contact intensive. The paper is akin to a recent paper by Faggian and McCann (2006) in terms of its content and also in terms of its technique. The outline of this paper is as follows: In Sect. 15.2 we discuss the knowledge concept and the conditions for knowledge flows. In Sect. 15.3 we review some of the literature on the location of university R&D and industrial R&D, respectively. The relationship between researcher network formation, knowledge flows and physical accessibility is discussed in Sect. 15.4. Section 15.5 contains a presentation of the data and the variables used in the empirical analysis as well as our empirical analysis. Our conclusions and suggestions for future research are presented in Sect. 15.6.

15.2

Knowledge and Knowledge Flows

In this paper, knowledge is defined as organized or structured information that is difficult to codify and interpret, generally due to its intrinsic indivisibility.4 As a consequence, knowledge is difficult to exchange5 without direct face-to-face interaction, since human capital is the major knowledge carrier. Loosely speaking, when knowledge is exchanged between two persons they both have to calibrate their explanation and interpretation activities, i.e., the exchange of knowledge needs oral communication and reciprocity.6 Since knowledge exchange requires face-to-face contacts, it requires an extensive amount of somewhat diffused movements throughout various transportation networks.7 Hence, while the costs of transmitting information may be close to invariant with respect to distance, the cost of exchanging knowledge increases together with the distance.8 As Teece (1981) remarked, knowledge is neither shared ubiquitously nor passed around at zero cost. This 4

Von Hippel (1994) persuasively demonstrates that highly contextual and uncertain knowledge, i.e. what he refers to as ‘‘sticky knowledge’’, is best transmitted via (preferably frequent) face-toface interactions. This is in line with the claim by Teece (1998) that knowledge assets are often inherently difficult to copy. Von Hippel’s sticky knowledge is also referred to as tacit knowledge in many studies from the last decade (Kogut and Zander 1992). Tacit knowledge cannot be codified easily in the form of a blueprint or a contract (Mowery and Ziedonis 2001), or a published article (Audretsch and Feldman 1996). 5 Knowledge exchange is defined here as any face-to-face interaction that can contribute to the process of the disclosure, dissemination, transmission, and/or communication of knowledge. 6 In this way face-to-face contacts become a necessary or facilitating condition, though not a sufficient condition, for knowledge transfer. 7 Historically, the transfer/communication of rich information has required proximity and specialized channels to customers, suppliers, and distributors. However, we must acknowledge the possibility that the new developments are undermining the traditional chains and business models, and that new structures – generally less dependent on physical communication channels – might become more and more often an economically viable option (cf. Teece 1998). 8 Interestingly, some authors assume that geography plays no role for the costs of accessing knowledge (Spence 1984; Cohen and Levinthal 1990).

270

C. Karlsson and M. Andersson

implies that geographical proximity matters and that knowledge has the properties of a public good only within a short distance from the source (Harhoff 1997a); Bottazzi and Peri (2003) show, that the costs of accessing and absorbing knowledge are not invariant to geographic location. Several studies show that the capacity to absorb flows of new knowledge is facilitated by geographical proximity (Jaffe et al. 1993; Baptista and Swann 1998). Obviously, there are costs and fundamental difficulties in exchanging knowledge. It explains why markets for exchange of knowledge are rare. Potential buyers may question the value of the knowledge, and sellers cannot easily assuage their concern without revealing their valuable asset – the specific knowledge. The buyer’s and the seller’s transaction information is intrinsically asymmetric. It also explains why companies prefer – in principle – to carry out R&D in-house rather than having it contracted out or licensed (Soete 2001). In view of the above exposition, it seems useful for our purposes to distinguish two knowledge concepts: l

l

Scientific knowledge in the form of basic scientific principles that can form a basis for the development of technological knowledge. Technological knowledge – implicit and explicit blueprints – in the form of inventions (or technical solutions) that either materialise in new products or can be readily used in the production of goods and services.

In concordance with Schumpeter’s analysis, scientific knowledge functions as a background to or platform for technological knowledge in the innovation process (Schumpeter 1934). As suggested by Nelson and Winter (1982), a company’s innovation can be a change in the routine (technique, organization, etc.) of the company and/or a new product (e.g., a change in attributes of a good or a service). In dealing with different concepts of knowledge it is essential to characterize them according to the degree to which they are rivalrous and excludable (cf., Cornes and Sandler 1986). A purely rivalrous good has the property that its use by one company or person precludes its use by another, whereas a purely nonrivalrous good has the property that its use by one agent in no way limits its use by another. Excludability relates to both technology and legal systems (Kobayashi and Andersson 1994). A good is excludable if the owner can prevent others from using it. While conventional goods are rivalrous and excludable, pure public goods are both non-rivalrous and non-excludable. Scientific knowledge has the character of a pure public good, although it is generally only available to those with the relevant scientific training. Hence, access to scientific knowledge can differ between companies and between regions, not only due to an unequal supply of scientifically trained labor but also due to the general costs of transferring knowledge over space. Technological knowledge may be perceived and even deliberately created as a non-rivalrous, partially excludable good (Romer 1990). Its non-rivalrous character stems from the fact that technological knowledge is inherently different from other economic goods. Once the costs of creating new ‘‘technological knowledge’’ have been incurred, this knowledge may be used over and over again at no additional

15

The Location of Industry R&D and the Location of University R&D

271

cost. It is in this sense that technological knowledge is non-rivalrous. The partially excludable character of technological knowledge stems from the fact that companies generally protect new inventions by having patents issued on them. However, patent applications – and therefore patents – must be quite detailed. This opens up opportunities for the competitors to imitate or to ‘‘invent around’’ patents, so that as a matter of fact technological knowledge may be accessible for intellectual purposes. At the same time, investigation and imitation activities consume resources. This implies that there is a cost or friction element in the process of imitating. The processes by which the different types of knowledge may flow from their creators to other individuals or companies take place in spatial networks, i.e., ‘‘knowledge networks’’ (Batten et al. 1989; Kobayashi 1995) consisting of a set of nodes and a set of links connecting them. At a coarse spatial resolution these nodes are represented by human settlements such as towns, cities and metropolitan regions, providing different instances of functional regions.9 At a finer geographical scale we can observe network links between companies and even individuals. The nodes can be characterized by their endowment of knowledge production capacities and related activities, including knowledge infrastructure such as universities, meeting infrastructure, stocks of knowledge and human capital, local knowledge networks, and so on. The links include transportation as well as communication channels. The spatial perspective adds a further dimension to knowledge transfers. Partial excludability of the new knowledge is not only a result of patents, business secrets, and so on but also a consequence of limited physical accessibility. Much of the discussion and analysis of knowledge flows has become contaminated because of unclear and fuzzy definitions of pertinent flows. In particular, many scholars have employed the concept of ‘‘knowledge spillovers’’ in an unfortunate way (Echeverri-Carrol 2001; Gordon and McCann 2000). As a step towards more clarity and precision in the analysis, Johansson (2005) suggests a separation into two groups of knowledge flows: 1. Transaction based knowledge flows 2. Pure knowledge spillover Transaction based flows include pure market transactions and link transactions whereas knowledge spillovers include both spillovers due to spatial proximity and spillovers through links. The distinctions made are important for several reasons. First, when the flows are transaction-based the participating economic agents have – in their own hands – market-like instruments to influence the resource allocation. Second, the mechanisms that generate the flows are different for the three categories which have implications for policy formation. Third, the externalities that can arise in the cases vary in nature (e.g., pecuniary and non-pecuniary) and should not be confused 9

Functional regions are delimited based upon the spatial interaction patterns of the economic agents in a country. A functional region is fundamentally characterized by its size, by its density of economic activities, social opportunities and interaction options, and by the frequency of spatial interaction between the actors within the region (Johansson 1997).

272

C. Karlsson and M. Andersson

with each other. Knowledge flows generate knowledge externalities towards R&D performing companies when the source (a research university or another company) is not fully compensated for the value of the knowledge flow (Harris 2001).

15.3

The Spatial Distribution of R&D: Interdependencies Between University and Industrial R&D

As mentioned previously, there are two major performers of R&D: industry and universities. The subsequent subsections discuss the location of each type of R&D respectively. The discussion focuses on the interdependencies between university and industrial R&D.

15.3.1 The Location of University R&D The first universities10 were founded in medieval times (Karlsson 1994). A second wave of founding of universities came in the late nineteenth and early twentieth century and a third wave in the post-war period culminating during the 1960s. This implies that the decisions of where to locate university R&D were taken a long time ago and long before the rapid increase of total R&D expenditures, which is a phenomenon of recent decades. However, important decisions concerning the location of university R&D have also been taken in recent decades in terms of governmental allocations and private grants to university R&D. In many countries institutions of higher education have been upgraded to university status and new universities have been started. The motivations for these decisions have certainly varied but it is quite natural to assume that some of them have been taken as a response to or as an indirect support to industrial R&D. It is in this connection important to recognize that modern (research) universities are multi-product organizations. The set of functions and outputs include (see, e.g., Luger and Goldstein 1997): l l

l

10

The creation of new basic knowledge though research The creation of human capital through teaching (i.e., knowledge transfer from faculty to students) The transfer of existing know-how (technology) to businesses, governmental agencies, and other organisations

There is no generally accepted definition of a university. We use the term university here as a collective term for institutions of higher education, whether they are major R&D performers or not. Major R&D performing universities are termed as research universities in this paper.

15 l

l

The Location of Industry R&D and the Location of University R&D

273

The application of knowledge to the creation and commercialization of new products and processes, or the improvement of existing ones (i.e., technological innovation) Co-production (with other R&D organisations) of a regional knowledge infrastructure

In this context it should be recognized that universities might pursue both reactive and proactive policies with regard to industrial R&D. Significant industrial R&D as well as lack of such R&D in a region might stimulate the local university to hire more research faculty, to be more active in acquiring R&D funds, to set up new campuses, to start business incubators, to start science and technology parks, etc. Even if lists of functions and outputs of (research) universities, such as the list above, can be helpful in understanding the scope of the activities of a university, they do not provide a basis for an analytical understanding of universities and their behavior. It is obvious that there is a lack of theoretical understanding of the role of the (research) university as an actor in technological change, the innovation process, organizational transformation and (regional) economic development (cf., Florida and Cohen 1999). Without such theoretical understanding, we have great difficulties in understanding the factors driving the localization of university R&D. A (research) university might be defined as an institution that in competition with other similar institutions generates and disseminates knowledge with the objective to achieve eminence, reputation and prestige. These objectives are defined in an objective function that each university tries to maximize under a budget constraint. To achieve its objectives each university competes for highly reputed faculty. Highly reputed faculty is a strategic production factor for a university for several reasons. First, they attract outstanding graduate and undergraduate students. Second, they reduce the budget constraint by attracting R&D funds. However, private and independent universities only make up a limited share of the ‘‘university market’’. Most universities are public and in various ways controlled by the public sector at the national or the regional level.

15.3.2 The Location of Industrial R&D There is plenty of evidence in the literature that industrial R&D is substantially more concentrated spatially than industrial production.11 For example, Kelly and Hageman (1999) show that innovation exhibits strong geographical clustering, independent of the distribution of employment. Sectors locate their R&D not where they are producing but close to where other sectors do their R&D. However,

11

However, there are authors that claim that R&D-intensive and high-tech industries do not necessarily agglomerate (Devereux et al. 1999; Shaver and Flyer (2000); Barrios et al. 2003; Alecke et al., 2003). In her study of Japanese investments in Europe Mariani (2002) found that R&D tends to locate close to production activities.

274

C. Karlsson and M. Andersson

Audretsch and Feldman (1996) found that there are substantial sectoral differences in spatial clustering with some industries like computers and pharmaceuticals displaying a higher degree of concentration compared to all manufacturing. Similar conclusions were drawn by Breschi (1999) after an examination of patent data for the period 1978–1991 from the European Patent Office. Theoretical arguments concerning localized knowledge flows suggest that knowledge production and innovative activities within a company will tend to be more efficient in agglomerations containing research universities and other R&D performing companies, since the access to knowledge flows and thus potential knowledge externalities is greater. The knowledge production and the innovative activities will be more productive and more cost-efficient because in such agglomerations there is a high probability that companies can access potentially useful external knowledge at a cost that is lower than producing this knowledge internally or of trying to acquire it externally from a geographic distance (Harhoff 2000). The cost of transferring such knowledge is a function of geographic time distance and this is why R&D agglomerations give rise to localized knowledge externalities (Siegel et al. 2003). Thus, given the character of knowledge flows, it seems natural to assume that the spatial dimension is a key factor explaining the location of R&D activities of companies. Obviously, the location of R&D activities of companies is influenced by the potential knowledge externalities from knowledge flows, university R&D and R&D in other companies. There is a rich literature regarding various aspects of the relationship between university R&D and industrial R&D and innovation. Some studies focus on the ability of companies to utilize knowledge flows from universities (Cohen and Levinthal 1989, 1990; Cockburn and Henderson 1998; Ziedonis 1999; Lim 2000). Another strand of literature studies the characteristics of universities that generate knowledge flows of interest for industrial R&D and innovation (Henderson et al. 1998; Thursby and Thursby 2002; Feldman et al. 2002; Jensen and Thursby 1998; Di Gregorio and Shane 2000). A third set of studies analyze the channels through which knowledge flows from universities to industry (Cohen et al. 1998, 2002; Agrawal and Henderson 2002; Colyvas et al. 2002; Shane 2002). These channels include: l

l l l

Personal networks of academic and industrial researchers (Liebeskind et al. 1996; MacPherson 1998) Spin-offs of new firms from universities (Stuart and Shane 2002) Participation in conferences and presentations Flows of fresh graduates to industry (Varga 2000)

However, there seems to be fewer studies that explicitly study the influences of university R&D on companies in general and on company R&D, in particular. Zucker et al. (1998) examine the location decisions of companies relative to the location of star university scientists. Mariani (2002) in a study of Japanese investments in Europe showed that geographical proximity to the local science base is an important factor for locating only R&D laboratories compared to R&D and production and production only. Agrawal and Cockburn (2002) use data on scientific publications and patents as indicators of university R&D and industrial R&D and

15

The Location of Industry R&D and the Location of University R&D

275

find strong evidence of geographic concentration in both activities at the level of metropolitan statistical areas (MSAs) in the USA. They also find strong evidence of co-location of upstream and downstream R&D activities. Agrawal and Cockburn (2003) report that high levels of university publishing in metropolitan areas in the United States and Canada tend to be matched by high levels of company patenting in the same technology field and metropolitan area, suggesting co-location of research activities. Other empirical studies suggest a strong correlation between the specialization of the regional R&D infrastructure and the innovative activities conducted by industry (Feldman 1994a; Felder et al. 1997a, b; Harhoff 1997a, b; Nerlinger 1998). These results can be interpreted as indicating that knowledge externalities from R&D infrastructure can be best used in innovation activities in companies in the same or closely related scientific and technological field(s). The correlation tends to increase with the complexity of the R&D and innovation activities and the more specific the demand for technological know-how (Feldman 1994a; Feldman and Florida 1994). Results presented by Bade and Nerlinger (2000) indicate strong correlations between the occurrence of new technology-based firms and the proximity to R&D-facilities comprising universities, technical colleges and non-university R&D-institutes as well as private R&D. Griliches’ ‘‘knowledge production function approach’’ introduced above did not acknowledge that knowledgeable persons and knowledge production activities are spread out in geography and at the same time to a high degree concentrated to agglomerations (Griliches 1979). However, the original ‘‘knowledge production function approach’’ has later been modified to also accommodate the spatial dimension (Jaffe 1989; Audretsch and Feldman 1994, 1996; Feldman 1994a, b). The inputs and outputs considered in these studies vary from study to study and so does the geographic unit of analysis. With a few exceptions (Henderson et al. 1994; Beise and Stahl 1999), empirical research suggests that knowledge flows from public science to companies decline with geographical distance. The input ‘‘federal research funding’’ is related to the output ‘‘new patents issued’’ at the state level in the USA by Jaffe (1989). Acs et al. (1992) correlate the input ‘‘university research spending’’ with the output ‘‘new product announcements’’. Jaffe et al. (1993) use the input ‘‘original patents’’ to explain the output ‘‘patents that cite the original patents’’ at the city level in the USA. They as well as several other studies (Narin et al. 1997; Verspagen 1999; Malo and Geuna 2000) find that academic papers and university patents are more frequently cited than their equivalents from private companies suggesting that public science outputs are an important knowledge source for inventions in companies. However, this method is not entirely accurate because the cited papers and patents may not have contributed to the invention, since the citation may be included only to build the patent claim. This method also underestimates the value of public science since many inventions that are not patented (Arundel and Kabla 1998; Audretsch and Feldman (1996) connect the input ‘‘local university research funding’’ in the USA to the output ‘‘local industry value-added’’ at the state level. The input ‘‘number of local research stars’’ is associated to the output ‘‘number of new local biotech firms’’ at the level of the economic region in the USA by Zucker et al. (1998). Branstetter (2000) links the

276

C. Karlsson and M. Andersson

input ‘‘scientific publications from the University of California’’ to the output ‘‘patents that cite those papers’’ at the state level. The input ‘‘hours of interaction with the MIT professor associated with a particular patented invention’’ is used by Agrawal (2002) to estimate the effect on the output ‘‘likelihood or degree of success of commercializing that invention’’ and he also evaluates the impact of distance on this effect. Irrespective of whether these studies use the production function approach or patent citations they find that knowledge flows from academic research to private companies are highly localized at the regional or state level in the USA. From the theoretical arguments and empirical results summarized above there seems to be clear evidence that the location of industrial R&D is attracted to locations offering good opportunities to take advantage of knowledge flows from universities (and public research institutes). There seems to be less evidences concerning whether concentrations of industrial R&D is an attractor for industrial R&D. Obviously there are both costs and benefits from locating company R&D and other innovative activities close to similar activities of other companies competing in the same market. Adams (2001) surveyed 208 private R&D laboratories in the USA and found that distance is a greater barrier to take advantage from knowledge flows occurring from public science than from companies.

15.4

Network Formation, Knowledge Flows and Physical Accessibility

The preceding sections suggest interdependencies between industrial and university R&D as regards the location across space. This section illustrates the importance of accessibility between these actors for the establishment of contacts and durable links between them. Durable links constitute important means by which knowledge is transmitted. The probability that durable links will be established between actors depends on the conditions for personal interaction. Therefore, economic networks and networks for transportation and infrastructure are complementary (Fischer and Johansson 1994). A link between two economic actors can be established via transactions, e.g., when a supplier and customer specify a delivery contract. In general, the extent to which such a link formation is determined by spatial proximity depends on the transaction’s contact intensity. Transactions involving knowledge – such as when a firm purchases R&D services from universities – are highly contact intensive. The outcomes of R&D projects are often uncertain and the transmission of complex and tacit knowledge often requires face-to-face communication. Because of this, durable links between industrial R&D units and university researchers are likely to be particularly dependent on the physical accessibility between the two. Moreover, many types of knowledge flows are transmitted via durable links.

15

The Location of Industry R&D and the Location of University R&D

277

Against the background above, we now consider an industrial R&D unit k located in municipality i and follow the basic set-up in Johansson et al. (2002). When it comes to establish contacts (links) with university researchers, we assume that a typical R&D unit k faces a set of M alternatives. The set M={1, . . . ,i, . . . , j, . . . ,n} contains all municipalities in the economy. Thus, each alternative pertains to university researchers in a specific municipality. We might now ask: what determines the preference value of R&D unit k regarding contacts with university researchers in location j? It is assumed that this preference value, denoted by pij; k , is a function of: l l l l l

The size of the R&D resources in the university The overall quality of the university R&D The price differential of university R&D services in location j and i The travel costs between i and j Random influence from non-observed factors

This is specified in (15.1): pij; k ¼ yj uj  aðpj  pi Þ  scij  gtij þ eij ;

ð15:1Þ

where yj denotes the overall quality of the university R&D in municipality i, pi,j denotes the price of R&D services in the respective municipality, cij is the monetary cost of travelling between municipality i and j, tij denotes the travel-time distance between municipality i and j and g represents the value of time.12 eij represents the random influence from non-observed factors. In (15.1), yj uj can be interpreted as the attraction factor in municipality j whereas cij, tij and the price differential can be interpreted as factors that pertains to the link between municipality i and j. Letting Pkij ¼ pkij  eij and assuming that eij is distributed independently, identically in accordance with the extreme value distribution, the probability that an R&D-unit located in municipality i will choose to establish contacts with university researchers in municipality j, Pkij , is given by13 k efPij g Pkij ¼ P : k efPij g

ð15:2Þ

j2M

Thus, (15.2) gives the choice probabilities in the multinomial logit model, (cf. Anderson et al. 1992). In (15.2), the numerator is the preference value for contacts with university R&D in municipality j whereas the denominator is the sum of such preference values, (cf. Johansson et al. 2002). This means that, ceteris paribus, the

Since p and c are monetary values, a and s translate these values to a common preference base (cf. Johansson et al. 2002). 13 This condition is derived in several texts, see inter alia Train (1993). 12

278

C. Karlsson and M. Andersson

probability of choosing contacts with university researchers in municipality j increases with the size of the attraction factor (the size of the R&D resources in j) and decreases with the time distance to municipality j. We now consider the denominator in (15.2) and assume that: l l l

The quality of university R&D is equal in all regions. The price differential is equal to zero, aðpj  pi Þ=0. The monetary travel costs are proportional to the time distance, such that cij ¼ gtij .

Moreover, we assume that uj ¼ ln Uj where Uj is the size of university R&D resources in municipality j. Using these assumptions, the denominator in (15.2) can be expressed as AU i ¼

X

Uj efltij g ;

ð15:3Þ

j2M

where l ¼ ðsg þ gÞ. AU i in (15.3) is a standard measure of accessibility with exponential distance decay. Obviously, an industrial R&D unit with high accessibility to university R&D is likely to have more frequent contacts and durable links with university researchers. Both the size of the attractor and time distances in (15.3) are arguments in the preference function in (15.1). Moreover, since durable links are important means by which knowledge is transmitted, knowledge flows between industrial R&D units and university researcher can be expected to be larger, higher the accessibility between the two. A similar set-up can be specified for university researchers who wish to establish contacts with industrial R&D units. In this case, the accessibility to industrial R&D for university researchers in municipality i becomes AIi ¼

X

Ij efltij g ;

ð15:4Þ

j2M

where Ij denotes the size of industrial R&D resources in municipality j. In (15.3) and (15.4), the accessibility measures represent the total accessibility. However, a national economy can be divided into functional regions that consist of one or several municipalities. Functional regions are connected to other functional regions by means of economic and infrastructure networks. The same prevails for the different municipalities within a functional region. Moreover, each municipality can also be looked upon as a number of nodes connected by the same type of networks. With reference to such a structure, it is possible to define three different spatial levels with different characteristics in terms of mobility and interaction opportunities. Because of this, it is also possible to construct three different categories of accessibility. Johansson et al. (2002) make a distinction between:

15 l l l

The Location of Industry R&D and the Location of University R&D

279

Intra-municipal accessibility Intra-regional accessibility Extra-regional accessibility

Letting R denote the set of municipalities belonging to functional region R, the total accessibility to university R&D of municipality i can be expressed as UM UE AU þ AUR i ¼ Ai i þ Ai ;

ð15:5Þ

P ¼ Ui efltii g is intra-municipal accessibility, AUR ¼ j2R;i6¼j Uj efltij g is where AUM i i P intra-regional accessibility and AUE ¼ j 2= R Uj efltij g is extra-regional accessibility. i The subscript of the time-distance sensitivity parameter l is different for each type of accessibility. In the sequel, the decomposition in (15.5) will be applied on both industrial and university R&D to empirically examine the interdependencies between industrial R&D and university R&D. An underlying conjecture is that high accessibility promotes contacts between the actors, which in turn encourage knowledge flows.

15.5

Interdependencies Between University and Industrial R&D: An Assessment Using Swedish Data

This section analyzes the relationship between industrial and university R&D using Swedish data at the municipality level 1995–2001. The section starts by presenting the data and the variables used in the analysis and goes on to analyze the relationship between industrial and university R&D across municipalities in Sweden.

15.5.1 Data Sources and Variables The R&D data used in this paper originates from Statistics Sweden. These data are collected by SCB via questionnaires that are sent out to firms and universities. The R&D data is measured in man-years. One man-year is the amount of work a fulltime employee performs during a year. This means that a full-time employee who only spends 50% of her work on R&D counts as 0.5 man-years. The data used in this paper cover the years 1995 and 2001. To calculate the accessibilities in (15.5) with respect to each type of R&D, we employ data on travel time distances by car between Swedish municipalities. These data are provided by the Swedish Road Administration (SRA). The data reports the travel time distance by car between each of the municipalities in Sweden. Moreover, a pre-specified value of the time distance sensitivity parameter, l, has to be used in order to calculate each accessibility value. In this paper we set l to 0.017,

280

C. Karlsson and M. Andersson

which is the estimated time distance sensitivity for business trips between regions in Sweden (Hugosson and Johansson 2001).

15.5.2 University and Industrial R&D: Description and Empirical Analysis of Interrelationships on Swedish Data Sweden is among the most R&D-intensive countries in the world. Figure 15.1 compares Sweden’s R&D expenditure as a share of GDP with a set of advanced industrialized countries during the twentieth century. As evident from the figure, Sweden surpassed Japan in the early 1990s and has then shown a steady increase in R&D expenditure as a share of GDP. In figures, R&D/GDP has increased from about 2.7% in 1991 to well over 4% in 2001. Moreover, relative to other countries Sweden has a very high level of R&D expenditure relative to its GDP. Which are the major performers of R&D? In vast majority of countries, the major performers of R&D are universities and private firms. Figure 15.2 presents Sweden’s R&D man-years – the data source that will be used in the empirical analysis – by four performers 1995–2003: l l l l

Industry Universities Private non-profit organizations Public authorities

4.5

R&D/GDP

4 3.5 3 2.5 2 1.5 1 0.5 0 1991 Sweden

1993 US

1995 Japan

1997 Finland

France

1999

2001

Norway

Great Britain

Year

Fig. 15.1 R&D expenditure as a share of GDP in selected countries 1995–2001

15

The Location of Industry R&D and the Location of University R&D

281

R&D man-years 70000 60000 50000 40000 30000 20000 10000 0 1995

1997

1999

Industry

Universities

Private non-profit organizations

Swedish total

2001

2003

Public authorities

Year

Fig. 15.2 R&D man-years 1995–2003 in Sweden by performer

Table 15.1 Descriptive statistics for the aggregate industrial and university R&D in Sweden 1995 and 2001 1995 2001 D man-years D share of total 1995–2001 (%) 1995–2001 Man- Share of Man- Share of years total (%) years total (%) Industrial 41,647 66.5 49,192 68.6 18.11 2.1 R&D University 18,246 29.1 19,715 26.6 8.05 2.5 R&D Swedish 62,635 – 72,190 – 15.26 – total Data source: Statistics Sweden (SCB)

It is evident from Fig. 15.2, that industry and universities are the major performers of R&D. These two performers of R&D constitute over 90% of the Swedish total R&D. Moreover, the relative contribution of each performer tends to be stable over the period considered. Private non-profit organizations and public authorities have very limited R&D man-years. Thus, focusing on industrial and university R&D does not imply the exclusion of any significant R&D performer. In addition to the above figure, Table 15.1 provides descriptive statistics for the aggregate industrial and university R&D in Sweden 1995 and 2001. Industry R&D is more than twice as large as university R&D and the distance between them has increased between 1995 and 2001. However, both industry and university R&D man years have increased during the period. Turning to the spatial distribution of university and industrial R&D it is clear that both university and industrial R&D are highly concentrated in space.

282

C. Karlsson and M. Andersson 100 90 Cumulative share (%)

80 70 60 50 40 30 20 10 0 0

50 Population

100

150 University R&D

200

250

Industrial R&D

Municipalities ranked in ascending order according to population share

Fig. 15.3 R&D man-years 1995–2003 in Sweden by performer

Figure 15.3 compares the spatial concentration of industrial and university R&D with population in the year 2001. Municipalities were ranked in ascending order according to their share of the total population. Then, the cumulative percentage of population, industrial R&D and university R&D were calculated. As is evident from the figure, both industrial and university R&D are much more concentrated than population. To complement the above figure, Table 15.2 presents descriptive statistics for university and industrial R&D in 1995 and 2001. As is evident from the table, the distribution is highly skewed. The standard deviations are large compared to the means, and the statistics for both skewness and kurtosis are high. Both university and industrial R&D are highly concentrated to specific municipalities. As an example, only 11 municipalities individually hosted more than 1% of the total university R&D in Sweden in 2001. Yet, these 1014 municipalities hosted approximately 90% of the total university R&D. The corresponding figures for industrial R&D are 1715 municipalities with a cumulative share of industrial R&D that amounts to about 78%. Moreover, many municipalities have zero university and industrial R&D man-years.

14

These municipalities are Stockholm, Uppsala, Go¨teborg, Lund, Umea˚, Solna Linko¨ping, Huddinge, Malmo¨ and Lulea˚. 15 These municipalities are Stockholm, Go¨teborg, Mo¨lndal, Linko¨ping, Lund, So¨derta¨lje, Trollha¨ttan, Malmo¨, Va¨stera˚s, Uppsala, Ja¨rfa¨lla, Karlstad, Karlskoga, Lulea˚, Sandviken, Jo¨nko¨ping and Solna.

15

The Location of Industry R&D and the Location of University R&D

283

Table 15.2 Descriptive statistics for industrial and university R&D in Sweden 1995 and 2001 across 286 municipalities University R&D Industrial R&D 1995 2001 1995 2001 Min 0 0 0 0 Max 3,572.77 3,452.03 10,135.41 11,912.35 Mean 63.8 68.94 145.62 171.87 Std. deviation 388.2 380.57 769.45 894.58 7.62 (0.14) 7.08 (0.14) 10.18 (0.14) 10.45 (0.14) Skewnessa 60.38 (0.29) 52.31 (0.29) 117.79 (0.29) 123.27 (0.29) Kurtosisa No. obs. 286 286 286 286 Data source: Statistics Sweden (SCB) a Standard errors presented within brackets Table 15.3 Top five municipalities in terms of R&D man-years of university and industrial R&D 2001 University R&D Industrial R&D Top 5 R&D manShare of Top 5 R&D manShare of municipalities years 2001 Swedish total municipalities years 2001 Swedish total (%) (%) Stockholm 3,452.0 17.5 Stockholm 11,912.4 24.2 Uppsala 3,116.3 15.8 Go¨teborg 7,850.3 16.0 Go¨teborg 2,891.7 14.7 Mo¨lndal 2,632.5 5.4 Lund 2,487.3 12.6 Linko¨ping 2,561.3 5.2 Umea˚ 1,529.7 7.8 Lund 1,962.8 4.0 Sum 13,852.6 68.4 Sum 26,919.2 54.8 Data source: Statistics Sweden (SCB)

Table 15.3 lists the top five municipalities in terms of the number of R&D manyears as regards both industry and universities in the year 2001. As is evident from the table, despite the cumulative percentage being larger for the university R&D, industrial R&D shows a larger concentration to specific municipalities. Stockholm and Go¨teborg alone hosts more than 40% of Sweden’s total industrial R&D. In order to analyze the spatial interdependencies between university and industrial R&D, we start by simply regressing industrial R&D in municipality i, IiR&D , on the university R&D in the same municipality, UiR&D , (and vice versa) for the year 2001. This gives an overall picture of the relationship as regards the location. The result of this undertaking is presented in (15.6): IiR&D ¼ 47:81 þ 1:8 UiR&D þ ei ; ð1:38Þ

ð20:06Þ

UiR&D ¼ 12:89 þ 0:33 IiR&D þ ei ; ð0:87Þ

ð20:06Þ

ð15:6Þ

where it is apparent that the coefficient for the independent variable is significant and positive in both equations. The R2 of both equations equals to 0.59. Thus, university and industrial R&D tend to coincide in space.

284

C. Karlsson and M. Andersson

However, in this paper the major hypothesis is that there are interdependencies between the location of university and industrial R&D. In order to test this hypothesis we analyze the aggregate pattern of the change in industrial and university R&D across Swedish municipalities from the period 1995–2001 in a simultaneous setting. The formulation of the simultaneous two-equation model follows the work by Holmberg et al. (2003), which in turn is a variation of the model in Mills and Carlino (1989). The equations that are estimated simultaneously are presented in (15.7a) and (15.7b): UM UR UE R&D Ii;R&D þ b5 DLi tþt ¼ a þ b1 Ai; tþt þ b2 Ai; tþt þ b3 Ai; tþt þ b4 Ii; t

þ b6 DSL i þ ei ;

ð15:7aÞ

IM IR IE R&D Ui;R&D þ ’5 DLi tþt ¼ m þ ’1 Ai; tþt þ ’2 Ai; tþt þ ’3 Ai; tþt þ ’4 Ui; t

þ ’6 DSL i þ ei ;

ð15:7bÞ

where t þ t refers to 2001 and t refers to 1995. Table 15.4 explains each of the variables in the above equations. The calculations of accessibility follow the derivation of accessibility in Sect. 15.4. Observe that the accessibility variables are measured in the year 2001. This is to reflect that they are themselves determined by the change in university and industrial R&D respectively. Parameter estimates of the variables in the simultaneous two-equation system in (15.7a) and (15.7b) are presented in Table 15.5.16 Two alternative estimation techniques were considered, 2SLS and 3SLS. As 2SLS estimators tend to be very unstable depending on their specification17 the equation system is estimated by Table 15.4 Explanations of the variables in (15.7a) and (15.7b) Variable Explanation Industrial R&D in year t in municipality i Ii;R&D t R&D University R&D in year t in municipality i Ui; t M Intra-municipal accessibility of municipality i in time t Ai; t R Intra-regional accessibility of municipality i in time t Ai; t Extra-regional accessibility of municipality i in time t AEi; t Dummy which takes the value 1 if municipality i is the central municipality in the region DLi it belongs to; 0 otherwise Dummy which takes the value 1 if municipality i is not the central municipality in the DSL i region it belongs to but the region is large; 0 otherwise Superscripts U University R&D (i.e., AUM i; t means intra-municipal accessibility to university R&D) I Industrial R&D (i.e., AIM i; t means intra-municipal accessibility to industrial R&D)

16

Lagged values are used as instruments. We are grateful to an anonymous referee for emphasizing this point.

17

15

The Location of Industry R&D and the Location of University R&D

285

Table 15.5 3SLS estimation of (15.7a) and (15.7b) Variable

Ii;R&D tþt

Ui;R&D tþt

Ii;R&D tþt **

a 12.84 (0.96) – 13.82 (1.04) m – 0.15 (0.02) – 0.20* (7.80) – 0.21* (8.14) AUM i; tþt 0.006 (107) – 0.003 (0.65) AUR i; tþt UE 0.009 (1.24) – 0.01 (1.35) Ai; tþt R&D 1.09* (101.90) – 1.09* (100.03) Ii; t IM – 0.01 (1.72) – Ai; tþt – 0.001 (1.21) – AIR i; tþt IE – 0.002 (0.15) – Ai; tþt R&D – 0.95* (70.56) – Ui; t 16.07 (1.04) 15.23 (1.60) 15.25 (1.00) DLi 5.87 (0.40) 1.72 (0.18) 5.46 (0.71) DSL i R2 0.99 0.98 0.99 N 286 286 285 t-Values are presented within brackets * Denotes significance at the 0.05-level ** Denotes that one municipality (Solna) is excluded from the sample

Ui;R&D tþt ** – 0.45 (0.06) – – – – 0.02* (3.20) 0.0005 (0.46) 0.0003 (0.23) 0.93* (78.37) 16.27* (1.98) 4.20 (0.52) 0.98 285

using the 3SLS technique. Moreover, the 3SLS estimator is more efficient than the 2SLS in the presence of correlation between the disturbances in the structural equations. The Table presents the parameter estimates of the variables in (15.7a) and (15.7b) using the iterated 3SLS estimator. The table presents estimates obtained on the full sample – i.e., all municipalities in Sweden – and the estimates obtained by excluding one municipality (Solna) which comes out as an extreme outlier in the estimations. The results in Table 15.5 suggest that university R&D tend to increase in location offering high accessibility to municipal R&D and that industrial R&D tend to increase in locations offering high accessibility to university R&D. Thus, the aggregate results support the hypothesis set out in the paper. Interestingly, regional accessibility does not have any statistically significant effect on the change in either university or industrial R&D. Hence, the effect between industrial and university R&D seems to be highly local in scope.

15.6

Conclusions and Suggestions for Future Research

During the years, a large number of formal studies have presented evidences of a positive impact of university R&D on firm performance in general and on the location of industrial R&D, in particular. The question is does it also work the other way around? Does industrial R&D function as an attractor for university R&D? What are the relationships between industrial R&D and university R&D as regards

286

C. Karlsson and M. Andersson

location and vice versa? The fact that knowledge flows seem to be spatially bound implies that proximity matters for the relationships between industrial and university R&D. We argue in this paper that spatial proximity should be measured using accessibility measures. Furthermore, accessibility measures can be used to model interaction opportunities at different spatial scales: local, intra-regional and interregional. Against this background, the purpose of this paper has been to analyze the locational relationship between industry R&D and university R&D in Sweden using a simultaneous equation approach. First of all, our empirical results show that there is a strong persistence or path-dependence in the location of both industrial and university R&D. Concerning the interdependencies between the location of the two types of R&D, our results indicate that the location of industrial R&D is quite sensitive to the location of university R&D, and that the location of university R&D is sensitive to the location of industrial R&D. However, the latter result is achieved only when we take away one outlier in the data – the municipality of Solna which is the location for the largest medical university in Sweden. Having established the general and overall interdependencies between industrial and university R&D, it is important to continue to investigate this kind of interdependencies between industrial R&D in different sectors and university R&D in different faculties and disciplines. We expect to find quite different interdependencies in the different cases. We believe that analyses of this kind have strong policy relevance since they will show the extent to which the public university funding is allocated in an optimal way given the location of industrial R&D. They will also show to what extent that public investments in R&D changes the location of industrial R&D and then show the extent to which the Swedish university policy also functions as a regional industrial policy. Another avenue for future R&D is to analyze the flows of R&D funds between industry and universities in more detail and to see how these flows change over time. Similarly, it would be interesting to study the extent to which trained researchers move from universities to industry. How important is this mobility and to what extent is it local and interregional? It would also be useful to explore extent to which new innovations are transferred from universities to industry and to what extent is this transfer dependent upon spatial proximity.

References Acs ZJ, Audretsch DB, Feldman MP (1992) Real effects of academic research. Am Econ Rev 82:363–367 Adams JD (2001) Comparative localisation of academic and industrial spillovers. NBER Working Paper Series 8292, National Bureau of Economic Research, Cambridge, MA Agrawal A (2002) Technology acquisition strategy for early-stage innovations: measuring the importance of engaging the inventor. Queen’s University (mimeo) Agrawal A, Cockburn IM (2002) University research, industrial R&D and the anchor tenant hypothesis. NBER Working paper 9212, National Bureau of Economic Research, Cambridge, MA

15

The Location of Industry R&D and the Location of University R&D

287

Agrawal A, Cockburn IM (2003) The anchor tenant hypothesis: exploring the role of large, local, R&D intensive firms in regional innovation systems. Int J Ind Organ 21:1227–1253 Agrawal A, Henderson R (2002) Putting patents in context: exploring knowledge transfer from MIT. Manage Sci 48:44–60 Alecke B et al (2003) New empirical evidence on the geographic concentration of German industries: do high-tech clusters really matter? Ann Reg Sci Anderson SP, de Palma A, Thisse J-F (1992) Discrete choice theory of product differentiation. MIT, Cambridge, MA Andersson M, Karlsson C (2004) The role of accessibility for regional innovation systems. In: Karlsson C, Flensburg P, Ho¨rte SA (eds) Knowledge spillovers and knowledge management. Edward Elgar, Cheltenham, pp 283–310 Anselin L, Varga A, Acs Z (1997) Local geographical spillovers between university research and high technology innovations. J Urban Econ 42:422–448 Arundel A, Kabla I (1998) What percentage of innovations are patented? Empirical estimates for European firms. Res Policy 27:127–141 Audretsch DB, Feldman M (1994) Knowledge spillovers and the geography of innovation and production. Discussion Paper 953, Centre for Economic Policy Research, London Audretsch DB, Feldman M (1996) R&D spillovers and the geography of innovation and production. Am Econ Rev 86:630–640 Bade FJ, Nerlinger EA (2000) The spatial distribution of new technology-based firms: empirical results for West-Germany. Pap Reg Sci 79:155–176 Baptista R, Swann P (1998) Do firms in clusters innovate more? Res Policy 27:527–542 Barrios S et al (2003) Agglomeration economies and the location of industries: a comparison of three small European countries. CORE Discussion paper, 2003-67 ˚ E (1989) Knowledge, nodes and networks: an analytical Batten DF, Kobayashi K, Andersson A ˚ E, Batten DF, Karlsson C (eds) Knowledge and industrial perspective. In: Andersson A organization. Springer, Berlin, pp 31–46 Beise M, Stahl H (1999) Public research and industrial innovation in Germany. Res Policy 28: 397–422 Bottazzi L, Peri G (2003) Innovation and spillovers in regions: evidence from European patent data. Eur Econ Rev 47:687–710 Branstetter L (2000) Measuring the link between academic science and industrial innovation: the case of California’s research universities. Paper presented at the NBER Summer Institute Breschi S (1999) Spatial patterns of innovation: evidence from patent data. In: Gambardella, Malerba (eds) The organization of economic innovation in Europe. Cambridge University Press, Cambridge, pp 71–102 Cantwell J (1998) The globalisation of technology: what remains of the product cycle model? In: Chandler AD, Hagstro¨m P, So¨lvell O (eds) The dynamic firm: the role of technology, strategy, organization, and regions. Oxford University Press, Oxford, pp 263–288 Cockburn I, Henderson R (1998) Absorptive capacity, coauthoring behaviour and the organisation of research in drug discovery. J Ind Econ XLVI:157–182 Cohen DW et al (1998) Industry and the academy: uneasy partners in the cause of technological advance. In: Noll (ed) Challenges to research universities. Brooking Institution Press, Washington Cohen DW, Levinthal DA (1989) Innovation and learning: the two faces of R&D. Econ J 99: 569–596 Cohen DW, Levinthal DA (1990) Absorptive capacity: a new perspective on learning and innovation. Adm Sci Q 35:128–152 Cohen DW, Nelson RR, Walsh J (2002) Links and impacts: the influence of public research on industrial R&D. Manage Sci 48:1–23 Colyvas J et al (2002) How do university inventions go into practice? Manage Sci 48:61–72 Cornes R, Sandler T (1986) The theory of externalities, public goods and club goods. Cambridge University Press, Cambridge

288

C. Karlsson and M. Andersson

Devereux MP, Griffith R, Simpson H (1999) The geographic distribution of production activity in the UK. IFS Working Paper W99/26, Institute for Fiscal Studies, London Di Gregorio, Shane S (2000) Why do some universities generate more spillovers than others? University of Maryland, Baltimore (mimeo) Dorfman NS (1983) Route 128: the development of a regional high technology economy. Res Policy 12:299–316 Echeverri-Carrol E (2001) Knowledge spillovers in high technology agglomerations: measurement and modelling. In: Fischer, Fro¨hlich (eds) Knowledge, complexity and innovation systems. Springer, Berlin, pp 146–161 Faggian A, McCann P (2006) Human capital flows and regional knowledge assets: a simultaneous equations approach. Oxford Econ Pap 58(3):475–500 Felder J, Fier JA, Nerlinger E (1997a) High Tech in den neuen La¨ndern: Unternehmensgru¨ndungen und Standorte. Zeitschrift fu¨r Wirtschaftsgeographie 41:1–17 Felder J, Fier JA, Nerlinger E (1997b) Im Osten nichts Neues? Unternehmensgru¨ndungen in HighTech Industrien. In: Harhoff (ed) Unternehmensgru¨ndungen – Empirische Analysen fu¨r die alten and neuen Bundesla¨nder, vol 7. Nomos, Baden-Baden, pp 73–100 Feldman M (1994a) The geography of innovation. Kluwer, Dordrecht Feldman M (1994b) Knowledge complementarity and innovation. Small Bus Econ 6:363–372 Feldman M et al (2002) Equity and the technology transfer strategies of American research universities. Manage Sci 48:105–121 Feldman MP, Florida R (1994) The geographic sources of innovation: technological infrastructure and product innovation in the United States. Ann Assoc Am Geogr 84:210–229 Fischer MM, Johansson B (1994) Networks for process innovation by firms: conjectures from observations in three countries. In: Johansson B, Karlsson C, Westin L (eds) Advances in spatial and network economics series. Springer, Berlin, pp 261–274 Florida R (1997) The globalisation of R&D: results from a survey of foreign-affiliated R&D laboratories in the USA. Res Policy 23:85–103 Florida R, Cohen WM (1999) Engine or infrastructure? The university role in economic development. In: Branscomb LM, Kodama F, Florida R (eds) Industrialising knowledge. University– industry linkages in Japan and the United States. MIT, Cambridge, MA, pp 589–610 Gordon IR, McCann P, Gordon IR (2000) Industrial clusters: complexes, agglomeration and/or social networks. Urban Stud 37:513–532 Griliches Z (1979) Issues in assessing the contribution of research and development to productivity growth. Bell J Econ 10:92–116 Harhoff D (1997a) Innovation in German manufacturing enterprises: empirical studies on productivity, externalities, corporate finance and tax policy. Universita¨t Mannheim, Mannheim Harhoff D (ed) (1997b) Unternehmensgru¨ndungen – Empirische Analysen fu¨r die alten and neuen Bundesla¨nder, vol 7. Nomos, Baden-Baden Harhoff D (2000) R&D spillovers, technological proximity, and productivity growth – evidence from German panel data. Schmalenbach Bus Rev 52:238–260 Harris RG (2001) The knowledge-based economy: intellectual origins and new economic perspectives. Int J Manage Rev 3:21–41 Henderson RM, Jaffe A, Trajtenberg M (1994) Numbers up, quality down? Trends in university patenting 1965–1992. Presentation to CEPR/AAAS Conference University Goals, Institutional Mechanisms and the Industrial Transferability of Research, Stanford University, 18–20 March 1994 Henderson RM, Jaffe A, Trajtenberg M (1998) Universities as a source of commercial technology: a detailed analysis of university patenting, 1965–1988. Rev Econ Stat 80:119–127 Holmberg I, Johansson B, Stro¨mqvist U (2003) A simultaneous model of long-term regional job ˚ E, Johansson B, Andersson WP (eds), The economics and population changes. In: Andersson A of disappearing distance. Ashgate, Aldershot, pp 153–161

15

The Location of Industry R&D and the Location of University R&D

289

Hugosson P, Johansson B (2001) Business trips between functional regions. In: Hugosson P (2001) Interregional business travel and the economics of business interaction. JIBS Dissertation Series No. 009, Jo¨nko¨ping Jaffe AB (1989) Real effects of academic research. Am Econ Rev 79:957–970 Jaffe AB, Trajtenberg M, Henderson R (1993) Geographic localisation of knowledge spillovers as evidenced by patent citations. Q J Econ 108:577–598 Jensen R, Thursby M (1998) Proofs and prototypes for sale: the tale of university licensing. NBER Working Paper No. 6698, National Bureau of Economic Research, Cambridge, MA Johansson B (1997) Regional differentials in the development of economy and population. In: So¨rensen (ed) Empirical evidence of regional growth: the centre–periphery discussion. The Expert Committee of the Danish Ministry of the Interior, Copenhagen, pp 107–162 Johansson B (2005) Parsing the menagerie of agglomeration and network externalities. In: Karlsson C, Johansson B, Stough R (eds) Industrial clusters and inter-firm networks. Edward Elgar, Cheltenham, pp 107–147 Johansson B, Klaesson J, Olsson M (2002) Time distances and labor market integration. Pap Reg Sci 81:305–327 Karlsson C (1994) From knowledge and technology networks to network technology. In: Johansson B, Karlsson C, Westin L (eds) Advances in spatial and network economics series. Springer, Berlin, pp 207–228 Karlsson C, Manduchi A (2001) Knowledge spillovers in a spatial context – a critical review and assessment. In: Fischer, Fro¨hlich (eds) Knowledge, complexity and innovation systems. Springer, Berlin, pp 101–123 Kelly M, Hageman A (1999) Marshallian externalities in innovation. J Econ Growth 4:39–54 Kobayashi K (1995) Knowledge network and market structure: an analytical perspective. In: Batten DF, Casti J, Thord R (eds) Networks in action. Communication, economics and human knowledge. Springer, Berlin, pp 127–158 ˚ E (1994) A Dynamic input–output model with endogenous technical Kobayashi K, Andersson A change. In: Johansson B, Karlsson C, Westin L (eds) Advances in spatial and network economics series. Springer, Berlin, pp 243–259 Kogut B, Zander U (1992) Knowledge of the firm, combinative capabilities, and the replication of technology. Organ Sci 3:383–397 Liebeskind J et al (1996) Social networks, learning and flexibility: sourcing scientific knowledge in new biotechnology firms. Organ Sci 7:428–443 Lim K (2000) The many faces of absorptive capacity: spillovers of copper interconnect technology for semiconductor chips. MIT, Cambridge, MA (mimeo) Luger M, Goldstein H (1997) What is the role of public universities in regional economic development? In: Bingham R, Mier R (eds) Dilemmas of urban economic development. Issues in theory and practice. Sage, London, pp 104–134 MacPherson AD (1998) Academic–industry linkages and small firm innovation: evidence from the scientific instruments sector. Entrepreneurship Reg Dev 10:261–276 Malo S, Geuna A (2000) Science–technology linkages in an emerging research plattform: the case of combinatorial chemistry and biology. Scientometrics 47:303–321 Mariani M (2002) Next to production or to technological clusters? The economics and management of R&D location. J Manage Gov 6:131–152 ˚ et al (eds) Advances in Mills E, Carlino G (1989) Dynamics of county growth. In: Andersson A spatial theory and dynamics. Elsevier, London Mowery DC, Ziedonis AA (2001) The geographic reach of market and non-market channels of technology transfer: comparing citations and licences of university patents. NBER Working Paper No. 8568, National Bureau of Economic Research, Cambridge, MA Narin F, Hamilton KS, Olivastro D (1997) The increasing linkage between U.S. technology and public science. Res Policy 26:317–320 Nelson RR (1986) Institutions supporting technical advance in industry. Am Econ Rev 76:186–189

290

C. Karlsson and M. Andersson

Nelson RR, Winter SG (1982) An evolutionary theory of economic change. Harvard University Press, Cambridge, MA Nerlinger E (1998) Standorte und Entwicklung junger innovativer Unternehmen: Empirische Ergebnisse fu¨r West-Deutschland. ZEW-Schriftenreihe 27. Nomos, Baden-Baden Romer PM (1990) Endogenous technological change. J Polit Econ 98:71–102 Schumpeter J (1934) The theory of economic development. MIT, Cambridge, MA Shane S (2002) Selling university technology: patterns from MIT. Manage Sci 48:122–137 Shaver JM, Flyer F (2000) Agglomeration economies, firm heterogeneity, and foreign direct investments in the United States. Strateg Manage J 21:1175–1193 Shimshoni D (1966) Aspects of scientific entrepreneurship. Unpublished doctoral dissertation. Harvard University, Cambridge, MA Siegel DS, Westhead P, Wright M (2003) Assessing the impact of university science parks on research productivity: exploratory firm-level evidence from the United Kingdom. Int J Ind Organ 21:1357–1369 ˚ (eds) Soete L (2001) The new economy: a European perspective. In: Archibugi D, Lundvall B-A The Globalising learning economy. Oxford University Press, Oxford, pp 21–44 Spence MA (1984) Cost reduction, competition and industry performance. Econometrica 52: 101–121 Stuart TE, Shane S (2002) Organisational endowments and the performance of university startups. Manage Sci 48:151–170 Teece DJ (1981) The market for know-how and the efficient international transfer of technology. Ann Am Assoc Polit Soc Sci 458:81–86 Teece DJ (1998) Capturing value from knowledge assets: the new economy, markets for knowhow and intangible assets. Calif Manage Rev 40:55–79 Teplitz P (1965) Spin-off enterprises from a large government-sponsored laboratory. Unpublished doctoral dissertation. Sloan School, MIT, Cambridge, MA Thursby J, Thursby M (2002) Who is selling the ivory tower? sources of growth in university licensing. Manage Sci 48:90–104 Train K (1993) Qualitative choice analysis – theory econometrics and an application to automobile demand. MIT, Cambridge, MA Varga A (2000) Local academic knowledge transfers and the concentration of economic activity. J Reg Sci 40:289–309 Verspagen B (1999) Large firms and knowledge flows in the Dutch R&D system: a case study of Philips Electronics. Technol Anal Strateg Manage 11:211–233 Von Hippel E (1994) Sticky information and the locus of problem solving: implications for innovation. Manage Sci 40:429–439 Wainer HA (1965) The spin-off technology from government-sponsored research laboratories: Lincoln laboratory. Unpublished doctoral dissertation. Sloan School, MIT, Cambridge, MA Ziedonis A (1999) Inward technology transfer by firms: the case of university technology licences. University of California, Berkeley (mimeo) Zucker L, Darby M, Brewer M (1998) Intellectual capital and the birth of US biotechnology enterprises. Am Econ Rev 88:290–306

Chapter 16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences Paul Cheshire and Stefano Magrini

16.1

Introduction1

In this chapter we investigate growth differences in the urban system of the EU12 over the last decades of the twentieth century, defined in two distinct ways: as growth in population, off setting for natural change so proxying for net migration; or as growth in real GDP percent. Each of these growth processes is investigated using a family of related models. We do not give substantial technical details of the two families of models used since these are available in Cheshire and Magrini (2006a, b). Rather the purpose is to highlight the similarities and the differences in the drivers of urban population as compared to ‘‘economic’’ growth and in doing so, reveal some interesting features of spatial adjustment processes in Europe and – briefly – how these compare to those in the USA. We start with a brief analysis of population growth in the major city regions of the EU of 12 over the period 1980–2000. These ‘‘city regions’’ are represented as Functional Urban Regions or FURs – as briefly explained in Sect. 16.2. Since we include the rate of population growth in the area of each country outside its major FURs as a control variable, we are, in effect, analyzing the pattern of net migration change over the two decades in each FUR. The conclusion is that interregional migration is orders of magnitude less in the EU than in the USA and that while internal migration flows do respond to the most obvious quality of life differences they do so only as quality of life varies within countries. We also find that national boundaries continue to be substantial barriers to spatial adjustment processes in Europe. The conclusion is, therefore, that in a European context one does not observe spatial equilibrium in a single ‘‘urban system’’; in other words there are P. Cheshire (*) London School of Economics e-mail: [email protected] 1

The authors have benefited from many discussions with colleagues as this work has developed but remain responsible for any remaining deficiencies or errors.

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_16, # Springer‐Verlag Berlin Heidelberg 2009

291

292

P. Cheshire and S. Magrini

people who could improve their welfare by moving to another city region in another country but constraints on mobility prevent them from doing so. This suggests that differences in measured income between places – even when adjusted for price differences – indicate differences in real welfare. The only practical data relating to the concept of ‘‘real income’’ is, of course, GDP per capita, so this implies that it is appropriate to analyze differential rates of growth of real GDP per capita not just if one is interested in productivity growth differentials but also if one is interested in investigating differences in welfare changes across Europe’s cities. City regions within the EU seem to behave like city-states, not as simply the spatial units from which a continental economy is constructed. The central assumption of perfectly mobile factors and the equalization of real marginal returns across cities, explicit in models of compensating differentials (the Quality of Life approach developed on the basis of Roback’s 1982 contribution), cannot reasonably be maintained in the European context. Understanding the differential growth rates of GDP percent across city regions in an EU context therefore has a double significance and it is to the analysis of this that the paper then turns. In this statistical analysis, we pay particular attention to testing significant hypotheses. The first is the hypothesis that the process of European integration has systematic spatial effects favoring ‘‘core’’ regions. Empirical interest in this goes back at least to Clark et al. (1969) and it is interesting to use as an explanatory variable the measure actually derived by Clark and his associates before the impact of European integration was significantly felt. Interest in these factors has been given a significant boost as a result of the theoretical developments of New Economic Geography as summarized, e.g., in Fujita et al. (1999) emphasizing the role of agglomeration economies in the face of reduced trade costs. Secondly, we explore the role of R&D and highly skilled human capital in explaining FUR growth rate differentials. Here we focus on testing a spatialized adaptation of endogenous growth theory (see Cheshire and Carbonaro 1996 or, for a more rigorous development, Magrini 1998). The third idea we are interested in investigating is the relationship between systems of city government and city growth performance. Here we test one of the basic propositions of fiscal federalism: that ‘‘the existence and magnitude of spillover effects clearly depends on the geographical extent of the relevant jurisdiction’’ (Oates 1999). Specifically we test that there is a positive relationship between the degree of co-incidence of governmental boundaries with those of the functionally defined city-region and the growth performance of the city-region. We find evidence supporting all three of our hypotheses about the impact of these variables on rates of growth of GDP percent but these results provide an interesting contrast with those from the population growth models and in our conclusion we try to draw out some of the implications of this contrast. In both these families of models we pay particular attention to issues of spatial dependence. Spatial econometrics tends to exist as a distinct field in which a finding of spatial dependence is often an end in itself – sometimes to be ‘‘corrected’’ by introducing spatial lags or other appropriate econometric devices. Our views are somewhat different. It seems important to test for spatial dependence since, if it is present, and the analysis does not properly take it into account, parameter estimates can be biased just as they can be in time series analysis if there are problems of

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 293

serial autocorrelation which are not offset for. However, it seems to us that the discovery of spatial dependence should trigger a further but economically inspired investigation. An indicated problem of spatial dependence suggests there is a specification problem. Something which explains the pattern of spatial dependence has been omitted and, if the model is specified better, then the problem should be resolved. This is particularly relevant in investigating spatial economic processes since theory suggests that there are important spatial adjustment mechanisms and other spatially determined features of economies. For example, labor markets and housing markets are likely to adjust to price and real wage differences in ways conditioned on some measure of distance. Theoretical and empirical investigations of agglomeration economies, human capital and innovation suggest there are important spatial aspects of these features of economies. These are possible sources of spatial interaction between cities’ economies which, if not represented in the model, would plausibly show up as spatial dependence. As the results reported below suggest, there seems to be some validity to this viewpoint. When we estimate growth models in which no spatial adjustment processes are explicitly included, tests show that there are problems of spatial dependence. However, deliberately including measures of spatial economic adjustment processes, which are a function of the distance between cities, eliminates spatial dependence and specification problems. In addition, the way in which the sensitivity of the models to measures of spatial dependence varies with the particular distance weights used to calculate ‘‘proximity’’ (the spatial weights matrix) provides, in our interpretation, insight into economic processes. In processes of both population and GDP growth, problems of spatial dependence only reveal themselves if an additional distance penalty to adjustment is included for national borders: this, we judge, tells one about the extent to which urban systems in Western Europe still interact as a set of national urban systems rather than as a unified EU urban system.

16.2

Data and Variables

All the analysis is performed on a data set built up over a 25 year period relating to Functional Urban Regions (FURs) defined2 so far as possible according to common criteria across the EU of 12. Such FURs correspond to the economic spheres of 2

For a detailed discussion of the definition of the FURs used throughout this paper see Cheshire and Hay (1989). They are defined on the basis of core cities identified by concentrations of employment and hinterlands from which more commuters flow to the employment core than to any other, subject to a minimum cut off. They were defined on the basis of data for 1971. They are broadly similar in concept to the (Standard) Metropolitan Statistical Areas used in the USA. As has been argued elsewhere (Cheshire and Hay 1989) the great variability in the relationship between administrative boundaries and the economic reality of European cities and regions introduces serious error and a strong likelihood of bias into data reported for administratively defined cities. The FUR/city and region of Bremen provide an extreme but not wholly unrepresentative example. Because of population relative to employment decentralization over the

294

P. Cheshire and S. Magrini

influence of significant employment concentrations and are relatively selfcontained in economic terms. The analysis is conducted only for FURs with a population of more than one third of a million and a core city which exceeded 200,000 at some date between 1951 and 1981. Cities of the former eastern La¨nder of Germany and Berlin have to be excluded because of lack of data. The new basis on which Eurostat estimated regional GDP from 1995 onwards means that the analysis stops then.3 The variables used are defined in Table 16.1. The table also provides a brief description of the sources used. More detail can be found in Cheshire and Magrini (2006b). All data are defined to common statistical concepts either weighting data available from the Eurostat REGIO database to estimate values for FURs or collected directly from national statistical offices or common data providers and adjusted where necessary to common definitions. There is necessarily some imperfection and imprecision in such data but they have the merit of not only allowing analysis of specifically European city regions but also of allowing the investigation of questions which, because of lack of variation, simply could not be investigated in the context of the US urban system. The analysis employs OLS but in Cheshire and Magrini (2006a, b) we provide substantial testing to see whether the results are subject to econometric problems. Since the observations represent the population of West European city-regions, the force of the standard objections to the use of cross sectional OLS for inference seem to be substantially mitigated. Compared to cross country ‘‘growth regressions’’ our observations represent a relatively homogeneous population and data are more comparable. We also try to minimize the impact of the standard problems associated with the use of regression to investigate causal processes by using spatial units which minimize nuisance noise in the data and formulating variables in ways which reflect views of causal mechanisms and minimize problems of endogeneity. As with all applied econometrics, however, in the end the credibility of results is not a categorical issue but depends on judgment. We are convinced multicollinearity is not a significant problem for our models and we judge that this is also true with respect to endogeneity. But proving endogeneity is neither present nor seriously distorting results and their interpretation is ultimately proving a negative and that is difficult. We necessarily make compromises but believe that the departure from the ideal conditions is not so great that the results are spurious for purposes of inference. relevant period, the growth of GDP percent is overstated by some 40% if the published Eurostat data for the administrative region is relied on. Even looking only at population growth if we rely on the NUTS data then apparently population of Bremen shrank by 1.8% during the 1980s while the data for the FUR show growth of 2.3%. In fact, the main feature of population change in Bremen during the 1980s was population decentralization. This, of course, contributes to the anomalous measure of GDP percent growth if the NUTS data is used. 3

We have made serious efforts to try to reconcile Eurostat regional GDP data estimated on the ESA79 and ESA95 methods but concluding it simply is not possible. Reluctantly we concur with the advice of Eurostat that: ‘‘Concepts and definitions between the two systems ESA95 and ESA79 are very different. In addition, ESA79 data is of very limited comparability between Member States. Therefore it would not be correct to create long series by linking data from the two systems’’ (Eurostat website – answers to Frequently Asked Questions).

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 295

Table 16.1 Variable definitions No Variable name Description 1 Ln population Natural log of population in 1979 2 Population density Density of population in FUR in 1979 3 Industrial emp. 1975 % of labor force in industry in surrounding NUTS 2 region 1975 4 Coalfield: core A dummy=1 if the core of the FUR is located within a coalfield 5 Coalfield: hinterland A dummy=1 if the hinterland of the FUR is located in a coalfield 6 Port size 1969 Volume of port trade in 1969 in tons 7 Agric emp.1975 % of labor force in agriculture in surrounding NUTS 2 region 1975 8 Unemployment Mean FUR unemployment rate 1977–1981 1977/1981 9 Nat Ex-FUR GDP growth Annualized rate of growth of GDP percent in the territory of each 1979–1993 country outside major FURs between 1978/1980 and 1992/ 1994 10 Nat Ex-FUR pop grow Annualized rate of growth of population in territory of country 1980–2000 outside major FURs between 1980 and 2000: so net natural population growth plus net international migration 11 Policy incentive Ratio of population of the largest governmental unit associated with the FUR (1981) to that of the FUR as a whole: see Cheshire and Magrini (2006b) for details. 12 University students emp. Ratio of university students 1977–1978 to total FUR employment ratio 1977/1978/1979 1979 13 R&D facilities per million R&D laboratories of Fortune top 500 companies per million population population 1980 14 South within country Distance south of centre of FUR from national capital city (Amsterdam taken as capital of Netherlands; Bonn of Germany) 15 West within country Distance west of centre of FUR from national capital city (Amsterdam taken as capital of Netherlands; Bonn of Germany) 16 South within EU Distance south of centre of FUR from Bruxelles/Brussel 17 West within EU Distance west of centre of FUR from Bruxelles/Brussel 18 Frost frequency Ratio of frequency of days with frost between FUR and national average (1970s and 1980s) 19 Wet days Ratio of wet day frequency between FUR and national average (1970s and 1980s) 20 Maximum temperature Ratio of maximum temperature between FUR and national average (1970s and 1980s) 21 Integration gain Change in economic potential for FUR resulting from pre-Treaty of Rome EEC to post enlargement EU with reduced transport costs 22 Peripherality dummy Dummy=1 if FUR more than 10 h time-distance from Brussels 23 University student density Sum of university students per 1,000 employees in all FURs employment within 150 min travel time discounted by distance with 600 time penalty added for national borders 24 R&D facilities density Sum of R&D Facilities per million population in all FURs within population 150 min travel time discounted by distance with 600 time penalty for national borders 25 Unemployment 1977/ Sum of differences between the unemployment rate (average 1981 density between 1977 and 1981) of a FUR and the rates in neighboring FURs up to 60 min away discounted by timedistance with a 600 min time-distance border penalty. 26 Interaction 1979–1991 Sum of the differences in the growth rate of employment in the FUR and in all FURs within 100 min travelling time discounted by distance over the period 1979–1991

296

P. Cheshire and S. Magrini

FUR growth in real GDP percent is estimated from common PPS values of GDP percent for Eurostat Level 3 regions. Estimates of GDP percent for FURs are derived by using the distribution of FUR population between Level 3 regions at the closest Census dates as weights and then applying those weights to the relevant Level 3 GDP percent data.4 Because of measurement error and short run fluctuations in Eurostat data, we take the start point of the series as the mean for 1978–1980 and the end point as the mean for 1992–1994. We are thus analyzing a period too short to correspond to a conceptual long run. Even if the system did tend to equalization of returns to factors on the margin, new shocks and disturbances will occur long before such a position is reached. We need, therefore, to model a system in which real incomes can permanently (in the sense of any period we can observe) vary between cities. The data used are derived mainly from Eurostat. Regional GDP data have been published for most Level 1, 2 and 3 regions since 1978 although for some they are available from 1977. There are, however gaps – data for Greek and Portuguese regions, e.g., only became available later. In both cases, Eurostat data have been supplemented with national data. For some countries, such as Italy, Eurostat data for earlier years were only published for Level 2 regions. In this instance, national sources for value added have been used to disaggregate from Level 2 to Level 3 values. The climate data are taken from the Climate Research Unit (University of East Anglia) database and for each city relate to the 30 km square which contains the geographical centroid of the FUR. In the case of Portsmouth and Southampton, the FURs fall within the same square but there is considerable climatic variation within most countries. Even within the Randstat cities of the Netherlands there is a 10% variation on most climate measures. A relatively standard set of control variables is used. These control for industrial structure where we find that detailed measures relating to old, resource-based industries tend to work better than broader measures of specialization in industry. In models in which GDP percent is the dependent variable, the unemployment rate at the start of the period is a useful additional control for structure. We do not use national dummies to control for country-specific differences in growth rates is but, as first employed in Cheshire (1995), the rate of growth of GDP percent in the area of each country outside its major FURs. This is an exogenous but appropriate control for national institutional, policy and other factors and also has what we see as the merit of being a continuous variable. Moreover, since some countries have only one or two major FURs, it would be necessary to make arbitrary decisions about country groupings if country dummies were used. The variable should also effectively control for national differences in the incidence of the economic cycle. In the models in which population growth is the dependent variable, a comparable 4

The EU institutions deal in so-called Nomenclature des Unite´s Territoriales Statistiques (N.U.T. S.) regions. This is a nesting set of regions based on national territorial divisions. The largest are Level 1 regions; the smallest for which a reasonable range of data is available are Level 3. These correspond to Counties in the UK (until 1996), De´partements in France; Provincies in Italy or Kreise in Germany. Because of cross border commuting flows there is inevitably built-in spatial nuisance dependence with this series. The use of self-contained FURs minimizes this problem.

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 297

variable – the rate of natural increase of population in the territory of the country outside its major FURs – is used. This has the advantage of converting our models of FUR net population growth into a proxy of an analysis of FUR net migration since we are offsetting for appropriate, but exogenously measured, natural change. Although national dummies have been the most frequently used method to offset for nation specific effects, it seems more elegant and powerful to use the continuous variable employed here. It is also consistent with our belief that our observations – all the large city-regions of the EU of 12 – represent in a statistical sense, a homogeneous population. Moreover, as was first shown in Cheshire (1995) and, in a more complete way in Cheshire and Magrini (2006b), the non-FUR growth variable performs very much better econometrically than national dummies. A further point of interest is that it eliminates the significance of any measure of the initial level of GDP percent Previous work has shown that both the significance and even sign of this commonly used variable were highly dependent on model specification (Cheshire 1995) and this confirms that result. It suggests that there is more variance in FUR growth rates across countries than within them and that the initial level of GDP percent acts in large measure as a national dummy. This finding is one factor underlying our skepticism with respect to the many estimates of so-called b-convergence following Barro (1990) and Barro and Sala-i-Martin (1991, 1992, 1995). In our data set all the results for GDP percent growth models which included the initial level of per capita GDP were unsatisfactory, with highly unstable co-efficient estimates associated with the variable, lack of statistical significance and problems of collinearity.

16.3

The Families of Models

As explained in the introduction, the purpose of this chapter is to compare the results of alternative families of models of urban growth processes: one family in which the rate of population growth is the dependent variable and another which analyzes differences in urban rates of growth of real GDP percent. We do not attempt to completely justify the specifications used nor explain the full set of tests applied. These are available in Cheshire and Magrini (2006a,b). Here we summarize the main findings and some of the test results relevant to the focus of this chapter in that they highlight differences (and sometimes similarities) in the models’ results.

16.3.1 Results of Modeling Urban Growth Rates: Population Growth We start with a brief summary of the results from modeling population growth in the FURs of the EU of 12 between 1980 and 2000 as reported in Cheshire and Magrini (2006a). As was noted above because the natural rate of increase of population for the area of the country outside its major FURs is included as an independent variable, we are in effect estimating a quasi-net migration model.

298

P. Cheshire and S. Magrini

Table 16.2 shows the results for a base model with no climate variables in column 1 and two of the best performing models in columns 2 and 3. In all models, a quadratic form for the climate variables performs best. As well as the standard controls, two other variables were included. The first is an ‘‘interaction’’ variable designed to measure localized employment opportunity differentials in the early part of the period. This is formulated on the basis that changes in commuting patterns are a potential source of spatial adjustment where there are densely packed FURs. Changes in commuting patterns induced by local differential employment opportunities at the start of the period are assumed to be likely to be at least in part converted to actual migration gain later. The variable is measured as the sum of all the changes in employment in a FUR and all surrounding FURs within 100 min travel time between 1979 and 1991 discounted by travel time-distance. This implies that, for a given change in employment, the value of the variable decreases as the distance between FURs increases and falls to zero if the distance exceeds 100 min. We do not use the simple mean of this variable but its total value on the grounds that the more FURs there are close to any FURi the more opportunities will be equalized over the interacting set of proximate FURs and the more total mobility will be induced into FURi. The second is the ‘‘Integration Gain’’ variable designed to measure the systematic spatial incidence of economic gains from EU integration. The variable measures the change in predicted economic potential for each FUR resulting from the integration of the EU of 12 (including lower transport costs). The values of the variable are calculated from Clark et al (1969) supplemented with the calculations of Keeble et al. (1988). The rationalization for including the variable is that changes in ‘‘economic potential’’ – which is a measure of the change in total distance costs discounted GDP available at each location (where ‘‘distance costs’’ include both transport an tariffs) – associated with EU integration will be a measure highly correlated with the economic gains from integration available at any location in Europe. Climate variables were formulated in two ways. The first was for each FUR relative to the mean value for the EU of 12 as a whole: the second was relative to the mean for the country. Variables relative to EU values proved wholly non-significant but when formulated relative to country means were strongly significant and also had a substantial impact in absolute terms: they were both significant and important. The linear estimates for each independent climate variable provided a simple guide to the overall impact of that aspect of climate on population growth. They showed that more cloud cover and wetness had a negative impact on growth and a warmer and drier climate had a positive impact. Model 1 is the ‘‘base’’ model: Models 2 and 3 include combinations of climate variables: frost frequency, maximum temperature and wet day frequency – all calculated as ratios of the country values. It will be seen that these models appear to perform well and provide striking evidence that climatic differences were strongly and significantly associated with differential rates of urban population growth. It was found, e.g., that these climate variables performed in a similar way, but statistically more significantly to, simple geographic variables measuring how far south or west FURs were relative to national datum points. Since the climate variables (and indeed the geographic ones) are only significant when measured as differences within countries there is no evidence to suggest that

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 299

Table 16.2 Dependent variable: FUR population growth rate 1980–2000: selected models Model ‘‘Base’’ model 1 Model 2 Model 3 R-squared 0.5180 0.6326 0.6405 Constant plus: Agric emp.1975 0.0004102 0.0004266 0.0004079 Std. err. 0.0000974 0.0000987 0.0000923 T 4.21 4.32 4.42 0.0000094 0.00000826 0.00000753 Agric emp.19752 Std. err. 0.0000026 0.00000249 0.00000246 T 3.61 3.31 3.06 Industrial emp.1975 0.0001693 0.0001457 0.0001213 Std. err. 0.0000416 0.0000393 0.0000341 T 4.07 3.71 3.55 Coalfield: core 0.0021143 0.001655 0.001812 Std. err. 0.0008684 0.0007881 0.000748 T 2.43 2.10 2.42 Coalfield: hint’land 0.0020548 0.001682 0.0018028 Std. err. 0.0008282 0.0007934 0.0007607 T 2.48 2.12 2.37 Port size 1969 0.0007278 0.0006274 0.0006521 Std. err. 0.0002844 0.0002422 0.0002469 T 2.56 2.59 2.64 0.0000366 0.0000294 0.0000315 Port size 19692 Std. err. 0.0000146 0.0000123 0.0000124 T 2.51 2.39 2.55 Nat Ex-FUR pop grow 1980–2000 0.4417852 0.5536141 0.4710524 Std. err. 0.1117606 0.1127851 0.1075922 T 3.95 4.91 4.38 0.0011278 0.0020954 0.0020679 Integration gain2 Std. err. 0.0004542 0.0004612 0.0004593 T 2.48 4.54 4.50 Interaction 1979–1991 0.0440806 0.0532723 0.0519908 Std. err. 0.0209222 0.0197226 0.0190658 T 2.11 2.70 2.73 Frost frequency ratio: country 0.0039281 Std. err. 0.001571 T 2.50 0.0020628 Frost frequency ratio2: country std. err. 0.0006133 T 3.36 Maximum temperature ratio : country 0.0752656 Std. err. 0.0322676 T 2.33 0.0379645 Maximum temperature ratio2: country Std. err. 0.0151008 T 2.51 Wet day frequency ratio: country 0.0247 0.0202854 Std. err. 0.0065655 0.0056615 T 3.76 3.58 0.008621 0.0069708 Wet day frequency ratio2: country Std. err. 0.0030658 0.0029409 T 2.81 2.37 All parameter estimates significant at 5% or better: N=121

300

P. Cheshire and S. Magrini

differences in climate across the EU as a whole were relevant: rather the results suggest that in the last decades of the twentieth century, people in all countries in the EU of 12 were attracted by, and able on the margin to choose to live in, places in their countries which had more agreeable climates. This is not inconsistent with a degree of international population mobility associated with climatic differences. But it suggests that in so far as people did make such moves, they selected the country first and then, in choosing locations within countries, chose cities with better weather. Table 16.3 reports the critical results of a series of diagnostics tests for specification and spatial dependence for the same three models. Full results are available from the authors but these show the important results. As is well known the major problem in testing for problems of spatial dependence is the choice of measures of ‘‘proximity’’. Following our non-conventional procedure, we searched for measures of ‘‘proximity’’ which related well to intuition, but generated the greatest sensitivity (as indicated by the appropriate test statistics) to spatial dependence. Despite not being the recommended method – which is to make a choice and stick by it on the grounds that enough trials will generate some result which is apparently significant – we prefer our method. We have not tested random or implausible measure of ‘‘proximity’’: only intuitively plausible ones recognizing that all applied econometrics is some combination of science and judgment. The approach we adopt provides a test for spatial dependence which is more difficult to pass when there are actual problems of spatial dependence and, moreover, allows economic rather than just statistical logic to be tested. In the present case, the most sensitive measure of distance when analyzing growth differences between European FURs was the inverse of time distance between pairs of FURs5 with an added time-distance penalty for all FURs separated by a national border. ‘‘Time’’ penalty effects tested for national borders varied from zero to 120 min. We found that the greatest sensitivity in the tests for spatial dependence was achieved if the time cost of a national border was set at 120 min. A valid criticism to this approach is that we should have continued testing for higher time-distance penalties until there was a reduction in indicated spatial dependence. This is the procedure we follow in the second family of models in which the rate of growth of GDP percent is the dependent variable. Where this is done, the most sensitive border penalty seems to be 600 min. However, for the population models we tested no larger penalty than 120 min partly because of the research effort involved but also because we found an apparently diminishing degree of sensitivity (as with the GDP percent models). Thus, the use of 120 min as the border penalty may not be the best that could be done but certainly makes little absolute difference to the results. In our defense we could claim that so far as we are aware no researchers have previously investigated any time-distance penalty for national borders and, in the present case that would mean that no problems of spatial dependence would have appeared. So we could

5

Measured as transit time by road including any ferry crossings and using the standard commercial software for road freight with origins as a historic central point, such as Charing Cross for London, in each core city.

0.0726 1 1

For weight matrix 120 mins borders+inverse time-distance squared 0.1248 3.2825 0.0010 0.0797 2.5592 0.0105 1 5.7734 0.0163 1 2.3529 0.1250 1 8.8033 0.0030 1 4.1270 0.0422

Moran’s I (error) Lagrange multiplier (error) Lagrange multiplier (lag)

DF 14

2

MI/DF 0.0124 1 1

Prob 0.3521

Prob 0.2996

Prob 0.0031 0.3866 0.1028

Value 15.3892

Value 2.4107

487.77

Model 3 568.063 13.4905 0.0000

For weight matrix 120 mins borders+inverse time-distance MI/DF Value Prob MI/DF Value 0.0245 3.1722 0.0015 0.0175 2.9603 1 1.4695 0.2254 1 0.7497 1 3.1892 0.0741 1 2.6616

DF 14

Value 9.4059

DF 10

Prob 0.4941

DF 2

Value 4.4466

DF 2

Prob 0.1083

143.0190

19.7911

Model 2 566.7440 13.0361 0.0000

Test Moran’s I (error) Lagrange multiplier (error) Lagrange multiplier (lag)

Diagnostics For Spatial Dependence

Regression diagnostics Multicollinearity condition number Test on normality of errors Test Jarque–Bera Diagnostics For Heteroskedasticity Random coefficients Test Breusch–Pagan Test

Table 16.3 Diagnostics for population growth – models 1, 2 and 3 ‘‘Base’’ model 1 LogLikelihood 550.3160 F-test 11.8200 F-test (prob) 0.0000

2.3999 1.9511 2.8366

Value 2.5297 0.3764 1.6872

Value 15.7706

1.3645

0.0164 0.1625 0.0921

0.0114 0.5395 0.1940

Prob 0.3276

0.5055

16 Growing Urban GDP or Attracting People? Different Causes, Different Consequences 301

302

P. Cheshire and S. Magrini

easily have reported results with quite plausible measures of ‘‘proximity’’ which showed no problems of spatial dependence. The diagnostic tests suggest that there are no problems of either heteroscedasticity or non-normality of errors. The value of the multicollinearity condition number is relatively high in the models in which climate variables are included in quadratic form but since the parameter estimates are stable and the functional form (effectively suggesting that it is asymptotic to an upper value) seems sensible, we are not concerned with this. The highest value for the multicollinearity condition was found for Model 3 but this may be because the functional form over the range considered is very close to linear. Of more concern are the results for the tests for spatial dependence. In the models in which climate variables (or ‘‘south within country’’) were included the LM error test – the most reliable and appropriate – suggests no problems of autocorrelation in errors but the results of the LM lag tests (again the most appropriate and reliable) suggest there could be some bias because of the omission of a spatial lag variable (or other specification problem). This seems likely to be a minor problem, however, only showing up as significant at all when distance is represented in the most sensitive form, as the inverse of time-distance squared including the 120 min national border effect: and even then, in Model 3, it is close to the 10 % margin of significance. The robust LM tests show no significant problems of spatial dependence for models 2 or 3. Fitting a spatial lag model using maximum likelihood estimation (Cheshire and Magrini 2006a) produces very similar results to those reported here. As suggested by the tests for spatial dependence, the spatially lagged value of population growth is significant. However, all signs remain appropriate and – except for the spatial Integration Gain variable in the ‘‘base’’ model – all variables are significant at least 10%. A few variables however, cease to be significant at 5%. All other variables are significant at 5% or better, however, and the diagnostics remain reassuring. Perhaps most reassuring of all, and again consistent with the conclusion that problems of spatial dependence are for practical purposes very minor, the coefficient estimates for equivalent models hardly change numerically in the spatially lagged estimate compared to the robust standard errors, OLS estimates reported in Table 16.2. Apart from these variables which are significantly associated with population growth it is worth noting some which are not. Neither the local concentration of university students per employee at the start of the period nor the concentration of R&D facilities of major companies was significant. Indeed, the sign with respect to the R&D variable was consistently negative and close to significant.

16.3.2 Results of Modeling Urban Growth Rates: GDP Percent Growth One conclusion from this analysis of urban population growth is that it is unreasonable to apply a full compensating differentials model to the major city regions of the EU. Although the influence on population mobility of some measures of differential

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 303

economic opportunities seems to be EU-wide, the effect of climatic differences – the most widely used measure of quality of life differences – are not. Moreover, strong national border effects are found when we test for spatial dependence. Not only is the EU orders of magnitude less than it is in the US, it still seems to be significantly confined within national borders. This means that the argument that population movement is the best measure of relative spatial welfare differences (see, e.g., Glaeser et al. 1995) is difficult to sustain in a European context. In turn, this suggests that (differences in) the growth of real incomes (or at least as proxied by real GDP percent) is a significant indicator of (changes in) welfare levels across the FURs of the EU. This lends additional importance to understanding sources of differential growth in real urban incomes. There is no data available on household disposable incomes across the EU so we have to use real GDP percent. It is perhaps more appropriate to think of these results as investigating sources of productivity rather than income growth, although the two are likely to be highly correlated. The results of three models are reported in Table 16.4. The first, Model 4, includes only non-spatial control variables. As the tests for spatial dependence (Table 16.5) show, this model is subject to significant spatial lag problems and so is open to concerns that it will yield inconsistent parameter estimates for the variables included to test hypotheses about the causal factors in urban economic growth. Indications of spatial dependence only appear when a time-distance penalty is added for national borders (so FURs separated by a national border are systematically less proximate to each other than FURs separated by an equal time-distance but within the same country) and the greatest sensitivity is demonstrated when the time-distance penalty for borders is set at 600 min. For this set of models, border time-distance penalties from zero to infinity were tested. Rather than attempt to fix these problems of indicted spatial dependence by simply introducing a spatial lag, we interpret this result as an example of a wider class of problem: that of omitted variables. We attempt to address the problem, therefore, by extending the logic applied to the population growth models and looking for variables which plausibly measure underlying spatial adjustment processes. Model 5 is constructed in this spirit, using what one might think of as ‘‘artisanal’’ methods – that is including only control variables and variables specifically chosen either to test hypotheses or to account for spatial adjustment processes. The presumption, therefore, is that if parameter estimates are significant then we can infer something about causal processes. Model 6 is a specific model emerging from the automated model selection algorithms employed in PcGets (see Hendry and Krolzig 2001). The set of variables available to the selection procedure included all variables available, including those relating to climate. All variables are significant and all those in Model 5 have the expected sign.6 The automated model selection procedures interestingly include one of the climatic variables (not included in the ‘‘artisanal’’ modeling process since there did not seem to be an obvious theoretical basis for expecting economic growth

6

Models were estimated in Stata using robust standard errors.

304

P. Cheshire and S. Magrini

Table 16.4 Dependent variable annualized rate of growth of GDP percent mean 1978/1980 to mean 1992/4: 4=base model without spatial variables, 5=artisanal’ & 6=PcGets best models Model 4 Model 5 Model 6 0.6785 0.7555 0.7719 R2 0.6372 0.7095 0.7235 Adjusted R2 AIC 10.8686 11.0440 10.9797 LIK 671.552 688.488 692.681 Observations 121 121 121 Constant 0.03200 0.0262573 0.03772 s.e. 0.00937 0.009193 0.01004 Nat Ex-FUR GDP growth 1979–1993 0.94416 0.902537 0.85222 s.e. 0.10238 0.097571 0.09720 Coalfield – core 0.00621 0.005213 0.00524 s.e. 0.00120 0.001287 0.00128 Coalfield – hinterland 0.00418 0.003176 0.00327 s.e. 0.00160 0.001526 0.00150 Port size 0.00147 0.000922 0.00096 s.e. 0.00040 0.000379 0.00037 Port size squared 0.00008 0.000045* 0.000047 s.e. 0.00003 0.000024 0.000023 Agricultural employment 0.00051 0.000484 0.00034 s.e. 0.00016 0.000159 0.00016 Agricultural employment squared 0.000013 0.000012 0.000010 s.e. 0.000004 0.000004 0.000004 Unemployment rate 0.00031 0.00035 s.e. 0.000136 0.00014 Population size 0.002118 0.001611 0.001496 s.e. 0.000600 0.000557 0.00055 Population density 0.0000015 0.0000013 0.0000013 s.e. 0.0000007 0.0000006 0.0000006 University students 0.0000309 0.000031 0.0000259 s.e. 0.0000116 0.000011 0.0000104 R&D facilities 0.000808 0.000845 0.00079 s.e. 0.000285 0.000275 0.00027 0.00770a Policy incentive 0.007500 0.008562a s.e. 0.00335 0.003455 0.00253 0.00253a Policy incentive squared 0.002089 0.002647*,a s.e. 0.001580 0.001554 0.00153 Wet days 0.03834 s.e. 0.01450 Wet days squared 0.01928 s.e. 0.00725 Integration gain 0.005162 0.00435 s.e. 0.001430 0.00149 R&D facilities density 0.262331 0.25088 s.e. 0.094307 0.09388 Peripherality dummy 0.005411 0.00632 s.e. 0.001318 0.00133 University students density 0.010527 0.01097 s.e. 0.003797 0.00371 Unemployment rate density 0.134403* 0.12129* s.e. 0.069318 0.06806

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 305

rates to be related to weather – but see below) as well as all those variables included in the artisanal model. As was noted in Sect. 16.2, the rate of growth of that part of each country outside its major FURs was used as a control – rather than national dummies – for basic differences in policy, the incidence of the cycle and other factors. Controls for industrial structure are employed in the population growth models, although the unemployment rate at the start of the period was also included. Other controls were designed as far as possible to reflect underlying (urban) economic theory and evidence. The log of population size is included with the expectation that larger cities will have grown faster in terms of GDP percent because of productivity gains in larger urban areas. Dynamic agglomeration economies is one of the possible reasons for expecting faster growth in GDP percent in larger cities although the ‘‘power couples’’ effect identified by Costa and Kahn (2000) is another.7 Initial population density was included since, allowing for agglomeration economies, cities with higher density will have higher costs of space and greater congestion and – other things equal – pollution. A negative relationship is expected. In our judgment, initial population density is likely mainly to reflect differences between FURs in the constraint on urban land supply produced by land use regulation. Higher density, other things equal, signals a tighter constraint imposed on development. Topography and the inertia of inheritance embodied in the built environment no doubt contribute to differences in densities but probably less than land use policy which varies substantially both across countries and between cities in Europe. Model 4, while it includes no purely ‘‘spatial variables’’ – see below – does include variables designed to test significant hypotheses about processes of urban economic growth in Europe. The first pair is straightforward. They are measures of specific, highly skilled, human capital and of relative concentrations of R&D activity. These are represented as the number of university students per employee over the period 1977–1979 in the FUR and as the number of R&D establishments of Fortune top 500 companies per million population in 1980. Thus, both are measured right at the start of the period to minimize possible problems of endogeneity. The last independent variable included in the non-spatial, base model, is designed to test whether there is a relationship between the boundaries of its governments and a FUR’s growth rate. With EU integration over the past 20 years there has been an associated development of territorial competition or competition between regional authorities, or representative agencies, to promote local growth. To the extent that there is an ‘‘output’’ from such activities, it is local economic growth. Suspending our disbelief in the possible efficacy of local growth

7 Costa and Kahn (2000) provide strong evidence for at least one important source of such productivity gains in larger cities – the increasing human capital of women. As they show, ‘‘power couples’’ – where both partners have high human capital – have an advantage of locating in larger cities since despite extra costs the benefits from richer labor market opportunities and better job matching will, on reasonable assumptions, lead to overall net gains. They further show that, empirically, such a concentration of high human capital couples has been occurring recently in US cities and seems to be related to increased educational qualifications of women.

306

P. Cheshire and S. Magrini

policies,8 any provision of additional local economic growth would be, in effect, the production of a pure local public good. Extra local growth is non-excludable in the sense that if a region’s economy grows as a result of local policy, those who did not contribute to the (costs of) the policy cannot be excluded from enjoying its benefits; and it would have a zero opportunity cost in consumption in the sense that if, say, one agent’s employment prospects or rents are improved, there would be no reduction in the employment prospects or rents of others. There are, therefore, the usual problems associated with the provision of (local) public goods, including a classic problem of spatial spillovers. Whether or not growth promotion policies are engaged in will be conditioned primarily on the structure of the incentives faced by the economic actors who may attempt to form a public/private consortium or ‘‘growth promotion club’’. It is reasonable to think of any FUR as being made up of one or more administrative units and that a ‘‘club’’ of administrative units (whether including private sector actors or not) will have to be formed to provide growth promotion policies. It is also reasonable to assume that the largest unit within the FUR – the central unit – will always be a part of such a club, either alone or together with other administrative units, so the territory of a FUR is made up of two potential sets of governmental units: the policy club members and the group of non-participating units. The expected gross payoff will be a direct function of the additional growth that a given club expects it can generate. Since FURs are defined to be economically self-contained, it is reasonable to assume that the territory their boundaries identify will contain any benefits that might be generated by local growth promotion policies. For a given potential growth gain for a FUR as a whole (which is, as just noted, the spatial unit containing the benefits of the growth) the expected payoff for any growth club will fall as the size of the territory it controls or represents falls in relation to that of the FUR within the boundaries of which the ‘‘club’’ is located. This is because the spillover losses to areas of the FUR not represented in the club will increase. Equally, assuming other factors are constant, the expected net payoff would fall as the transactions costs necessarily incurred to form the club increase. Transactions costs will be positively related to the number of relevant potential members and the institutional dominance of the lead actor (which we can assume will be a governmental unit). Thus expected net benefits will increase and transactions costs fall as the size of the largest governmental unit increases relative to the size of the FUR. Arguments such as these led Cheshire and Gordon (1996, p.389) to conclude that growth promotion policies would be more likely to appear and be more energetically pursued where ‘‘there are a smaller number of public agencies

8

As we stress in Cheshire and Magrini (2006b) we take a very broad view of ‘growth promotion policies’. We emphatically do not confine our definition to attempts to lure mobile investors with location incentives. Such policies probably have a very doubtful potential net benefit. Successful policies might mainly take the form of efficient local public administration, which is business friendly, the efficient co-ordination of infrastructure and economic development and effective education and training policies. Since none of these necessarily cost more than their ineffective counterparts, their strength cannot be measured by local expenditures.

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 307

representing the functional economic region, with the boundaries of the highest tier authority approximating to those of the region. . .’’. Applying this analysis it is possible to specify a variable closely reflecting this feature of FURs: the ratio of the total population of the largest (relevant) unit of government representing the FUR to the population of the FUR as a whole. We are implicitly assuming this ‘‘relevant’’ unit of government will be the unit with the largest population, usually representing the central administrative unit of the FUR, but this is qualified by ‘‘relevant’’: by which we mean that the governmental unit concerned must have significant powers of action. Even though it might be the largest N.U.T.S. region with a territory overlapping that of the London FUR, e.g., one could not define the South East Region as a ‘‘relevant’’ governmental unit for the London FUR because it had essentially no powers.9 The rules by which such ‘‘relevant’’ local government units were identified were established before any models including the variable were estimated so that the variable could be defined blind of the data. These rules are set out in Cheshire and Magrini (2006b). We call this the policy incentive variable because it is designed to measure the incentive and perhaps the capacity to prosecute policies promoting growth at the FUR level. In identifying the largest ‘‘relevant’’ unit of government, ‘‘relevant’’ is defined as a sub-national unit of government with an administrative area encompassing or corresponding to some proportion of the territory of a FUR and which has significant administrative and decision-making powers. Since the largest ‘‘relevant’’ unit was selected, it was also, in all cases, the highest tier of sub-national government relating to the territory of the FUR. Since one criterion was that the unit of government selected should have significant administrative and decision making powers, the Level 1 regions were potentially available for selection in European countries with an appropriate regional level of government. In practice, this means that the value of the variable ranged from only about 0.125 to over 2. We might further hypothesize that if the value of the variable were very high, so that the size of the ‘‘relevant’’ unit of government considerably exceeded the size of the FUR, then the capacity to generate local growth promoting policies would begin to weaken. This is because the interests of the FUR would begin to be lost in those of the larger unit which might pursue policies favoring rural areas or smaller centers. If this were the case then we would expect to observe a quadratic functional form with a maximum positive impact where the value of the policy incentive variable was above 1. Turning now to the results reported in Tables 16.4 and 16.5 we see that all the variables in the core model are significant and have the expected sign except that the co-efficient on the squared term associated with the policy incentive variable is non-significant. Thus, the variable itself is significant but the evidence that the appropriate functional form is quadratic as hypothesized is weak. The adjusted R2, with 121 observations in a cross sectional analysis, is 0.64. Table 16.5, reporting the

9

During the period analyzed there was a South East Regional Planning Council (SERPLAN) but this was effectively no more than a forum for discussion.

Table 16.5 Regression diagnostics for: 4=base model without spatial variables, 5=artisanal’ & 6=PcGets best models Regression diagnostics (SpaceStat) Model 4 Model 5 Model 6 Multicollinearity condition number 80.62 100.87 170.52 Test on normality of errors DF Value Prob DF Value Prob DF Test=Jarque–Bera 2 3.3273 0.1894 2 1.2374 0.5386 2 Diagnostics for heteroskedasticity Random coefficients DF Value Prob DF Value Prob DF Test=Breusch–Pagan 13 19.3825 0.1117 19 20.8169 0.3470 21 Diagnostics for spatial dependence For weight matrix (row-standardized) Inverse of time-distance with infinite national border effect Test MI/DF Value Prob MI/DF Value Prob MI/DF Moran’s I (error) 0.04344 1.8729 0.0611 0.04784 0.1457 0.8842 0.05938 Lagrange multiplier (error) 1 0.9212 0.3372 1 1.1171 0.2905 1 Lagrange multiplier (lag) 1 6.6183 0.0101 1 1.4510 0.2284 1 For weight matrix (row-standardized) Inverse of time-distance squared with infinite national border effect Moran’s I (error) 0.05593 1.4068 0.1595 0.06140 0.1916 0.8480 0.08193 Lagrange multiplier (error) 1 0.6996 0.4029 1 0.8432 0.3585 1 Lagrange multiplier (lag) 1 7.1177 0.0076 1 1.9795 0.1594 1 For weight matrix (row-standardized) Inverse of time-distance with 600 min national border effect Moran’s I (error) 0.0303 2.8693 0.0041 0.01504 0.6940 0.4877 0.02051 Lagrange multiplier (error) 1 1.5984 0.2061 1 0.3938 0.5303 1 Lagrange multiplier (lag) 1 5.8394 0.0157 1 0.9660 0.3257 1 For weight matrix (row-standardized) Inverse of time-distance squared with 600 min national border effect Moran’s I (error) 0.06620 1.7888 0.0736 0.03589 0.1484 0.8820 0.05842 Lagrange multiplier (error) 1 1.3233 0.2500 1 0.3888 0.5329 1 Lagrange multiplier (lag) 1 7.1366 0.0076 1 1.5291 0.2162 1 For weight matrix (row-standardized) Inverse of time-distance with zero national border effect Moran’s I (error) 0.0143 2.3972 0.0165 0.01538 0.4386 0.6610 0.0159 Lagrange multiplier (error) 1 0.5553 0.4561 1 0.6440 0.4223 1 Lagrange multiplier (lag) 1 2.4908 0.1145 1 0.4333 0.5104 1 For weight matrix (row-standardized) Inverse of time-distance squared with zero national border effect Moran’s I (error) 0.0573 1.7963 0.0724 0.02911 0.1375 0.8906 0.03337 Lagrange multiplier (error) 1 1.3549 0.2444 1 0.3498 0.5542 1 Lagrange multiplier (lag) 1 2.8781 0.0898 1 0.1902 0.6627 1 Prob 0.2039 Prob 0.1239

Prob 0.7253 0.1896 0.2091 0.6799 0.2204 0.1942 0.6427 0.3922 0.2899 0.8604 0.3100 0.2587 0.6079 0.4068 0.4580 0.8836 0.4978 0.7116

Value 3.1805 Value 28.6001

Value 0.3514 1.7209 1.5775 0.4126 1.5016 1.6855 0.4639 0.7321 1.1201 0.1759 1.0305 1.2755 0.5131 0.6882 0.5508 0.1464 0.4596 0.1367

308 P. Cheshire and S. Magrini

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 309

results of the diagnostic tests, suggests that there are no problems of non-normality of errors or heteroscedasticity but that if a time-distance penalty is included in the distance weights matrix there are significant problems of spatial dependence, most obviously problems of a spatial lag. For the set of models of economic rather than population growth we experimented with a full set of national border time-distance penalties ranging from zero to infinity. Problems of spatial dependence tended to show up most strongly when the penalty imposed for a national border was 600 min. As noted above, we interpret this result as signaling that there is, in effect, a problem of omitted variables. There are underlying mechanisms of spatial adjustment causing interaction in the growth rates of neighboring FURs as well as, perhaps, direct spatial effects, so we should not expect the growth behavior of a FUR to be independent of that of its neighbors. One such factor has already been identified in the context of the analysis of rates of population growth. The process of European interaction has long been argued to differentially favor ‘‘core’’ regions so the first spatial variable to include is the familiar measure of change in economic potential – or the Integration Gain variable, as used in the population models. A further obvious reason for expecting interaction in the growth rates of neighboring FURs can be found in the literature on labor market search behavior. If productivity, wages or job opportunities are relatively improving in one urban area then those who can access those opportunities at the lowest cost – who live nearest – will tend to do so. Migration is expensive but changes in commuting patterns respond to only small differences in opportunities (see, e.g., Gordon and Lamont 1982 or Morrison 2005). We should expect that if a FUR’s growth rate were negatively influenced by a concentration of unemployment in it at the start of the period then a concentration of unemployed in closely surrounding FURs would also have a negative impact. Given the possibility of job search in surrounding labor markets we would expect higher unemployment not to be just localized, moreover, but in densely urbanized regions, we would expect unemployment rates for workers of comparable skill levels to even out between neighboring FURs. This leads us to introduce as independent variables both the unemployment rate in the FUR itself at the start of the period and the weighted sum of unemployment rates in all surrounding FURs, discounted by time-distance. Since the search areas of low skilled workers, who are disproportionately overrepresented amongst the unemployed, are confined geographically, we should expect the impact of unemployment on the economic performance of neighboring FURs to decline rapidly with distance. Experiment confirmed this, showing that the best statistical results were achieved if the cut-off was set at 60 min. Similarly, the literature on the spatial pattern of innovation shows a distance decay effect, with patents tending to be applied more frequently nearer to the location of the patent and innovation rates declining with distance (see, e.g., Audretsch 1998). We should expect the impact of R&D with respect to innovation to be subject to a distance decay effect, therefore. This implies that we should expect R&D in one urban area to have a positive impact on innovation and growth in neighboring urban areas which would fall as the distance between them increased. Such mechanisms, leading to systematic spatial dependence in the

310

P. Cheshire and S. Magrini

growth rates of FURs, will depend on the costs of commuting and perhaps communication. It, therefore, seems not only most appropriate to formulate these ‘‘spatial’’ variables so that their impact declines with distance but also to include a specific time-distance penalty for national borders. We in fact experimented with alternative distance decay and national border factors but the best results were obtained using essentially the same formulae as employed to calculate the spatial weights matrix. The impact of unemployment and R&D on growth performance in neighboring FURs was assumed to decline with the inverse squared of time-distance and be subject to a 600 min national border time-distance penalty. As noted above, for unemployment, an upper cut-off of 60 min performed best but for R&D the best performing cut-off was found to be 150 min. The fourth ‘‘spatial’’ variable is the relative concentration of university students in neighboring FURs at the start of the period, again discounted by time-distance and with a national border penalty. Here we expect a negative impact on growth in a particular FUR of a stronger relative concentration of university students at the start of the period in neighboring FURs; and we also expect the distance over which such an effect would be measured to be longer than with unemployment. While a higher stock of unemployed within a tightly clustered set of urban areas should be expected to contribute to lower growth in all of them because of the way in which local labor markets work to equalize unemployment rates for workers of given skill levels between areas open to commuting,10 the same is not true of a higher relative stock of university students in surrounding FURs at the start of the period. Here, there is no tendency for their distribution to be evened out by the operation of local labor markets: rather a higher stock within a given FUR at the start of the period represented a resource for future growth. A concentration of workers embodying greater human capital is associated with faster growth over the subsequent period in the FUR in which they are found. Not only should this be expected to increase the growth performance of the FUR (captured in our direct University Student variable) but also the additional growth will increase relative job opportunities and tend to suck in complementary labor, including high human capital labor, from surrounding FURs over the period. Since the commuting range of higher skilled workers is greater, we should expect this effect to be measurable over a longer distance than was the case with unemployment. The best results were obtained if the cut-off was set at 150 min to which was again added a 600 min national border time-distance penalty. The final ‘‘spatial’’ variable was a dummy for peripherality. There has been much discussion in the literature of the impact of peripherality. We have already accounted for the impact of European integration via our Integration Gain variable but regions deemed peripheral may have common features (such as lower factor costs, e.g.) and also have tended to be recipients of regional aid from the EU.

10

Although FURs are defined to be as self contained in commuting terms as possible where they are tightly packed (e.g. in the Ruhr region of Germany) it is virtually zero cost for a worker living on the edge of any FUR to change to commute to the neighboring FUR(s).

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 311

Although the impact of such aid has been questioned (see, e.g., Midelfart and Overman 2002; or Rodriguez-Pose and Fratesi 2004) still it is unlikely to have been systematically negative. To avoid subjective judgments about what regions are – or are not peripheral this variable is formulated simply in terms of timedistance from Brussels; any FUR 600 min or more from Brussels – ignoring national borders – is classified as peripheral. When these variables are added, the model performance improves without significantly changing the parameter estimates associated with the main explanatory variables, while the problems of spatial dependence (see Table 16.5) are eliminated. This would seem to be a highly satisfactory result replacing a technical solution (which in this case would have been introducing a spatial lag) with one based on economic mechanisms. Two points about the results reported for Models 5 and particularly 6 should be noted. The first is that although the squared term on the policy incentive variable is still not significant at conventional levels, an F-test shows that neither term should be eliminated: both together perform significantly better than either alone. The second point is that when automated methods of model selection are used (see Hendry and Krolzig 2004) the same set of independent variables is selected plus a quadratic form associated with climate. All else taken into account, there was a statistically significant association between faster economic growth and a FUR having a wetter climate relative to its national average. One should add that the climate variables are highly correlated and if the wetness variable is excluded from the set of variables available for selection then the maximum temperature relative to the country is selected and is significant with a negative sign. Climate variable were not included in the ‘‘artisanal’’ models because theory does not obviously suggest that climate should be causally associated with economic growth. However, drawing on the literature deriving from Roback (1982) and reviewed in Gyourko et al. (1999) there is, in fact, a reasonable argument as to why one might expect a statistically significant association. A better climate will be capitalized into land prices and traded off by individuals against higher wages. As discussed in Sect. 16.3.1, there is strong evidence that national climatic differences are very significant in explaining patterns of population growth and mobility between FURs within countries. This is consistent with a process of sorting between FURs, with concentrations of human capital and R&D facilities being negatively but not significantly associated with population growth, while a drier and warmer climate, relative to a country’s mean, is strongly and significantly associated with population growth. This suggests that there was some selection process going on with people more motivated by quality of life and with lower skills tending to be differentially attracted to live in places with a better relative climate. This implies, other things equal, that more skilled people and those more work oriented – together with activities employing such labor – would find costs lower and welfare levels higher in FURs with relatively worse climates. Since this was a dynamic process – the dependent variable was a proxy for net migration over the 20 year period 1980– 2000 – it would imply a faster rate of productivity and wage growth in FURs with climates worse than their countries’ means. In essence, this is no more than the

312

P. Cheshire and S. Magrini

application of the insight that people who think they are likely to be unemployed anyway might as well live somewhere nice if there is a national system of welfare support.

16.4

The Contrasts and Similarities: Conclusions

The contrasts and similarities in these two sets of models not only reveal some interesting differences in the drivers of population compared to real GDP per capita growth but also suggest some insights into the underlying patterns of, and constraints on, urban change in the EU of 12. Cities (as FURs) with greater attraction to population and faster growth of GDP percent have some important structural characteristics in common. They share a common, minimal, inheritance of old resource based industries such as coal mining and port activity: these were underrepresented in the fast growing cities. Similarly, the faster growing cities were in wider regions with significant but not excessive agricultural employment (the very high proportions of agricultural employment in 1975 were found only in a few regions which still had a substantial, undercapitalized peasant population). Moreover, they had two related structural factors in common also: at the start of the period, there was a lower representation of industrial activity (favoring population growth) or lower relative unemployment (favoring economic growth). Finally there was one more EU-wide economic influence the more dynamic cities shared: they tended to be the systematic beneficiaries of the effects of European integration as measured by the change in economic potential associated with the formation and enlargement of the EU and falling transport costs. These were all common influences on growth of both population and GDP percent for larger European city-regions. However, of these only really the impact of European integration can be thought of as a European-wide factor. The other factors are common but could all be working within a national context. Having a coal mining inheritance, e.g., was a negative for both economic and population growth but that is consistent with it simply being that in the last two decades of the twentieth century, coal mining was declining in all of the old established areas and left behind a set of skills and an environment unattractive to migrants and new economic activity everywhere. A set of equally significant factors differs. We find that a relatively better climate within, but only within, countries was statistically the single most significant factor associated with differential population growth. Climatic differences have been found to be the most important indicator of quality of life differences between places but in the EU of 12 only quality of life differences between cities within the same country influenced population mobility. On the other hand, stronger economic growth was significantly associated with a city having a worse climate relative to the rest of its country, once all other factors had been allowed for. As argued above, this finding is consistent with the quality of life model and a process of sorting of population between locations (within countries) meaning that less work oriented/less

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 313

highly skilled people choose a better climate (driving up property prices) relative to employment opportunities. One may even be able to see the impact of this in the different role of a city’s share of industrial employment at the start of the period compared to unemployment. A higher level of unemployment was found to be associated with slower economic growth; and higher unemployment is associated with worse employment prospects and a less skilled labor force on average. However, although an initial relative specialization in industry is correlated positively with higher unemployment, the variable appearing as significant in the population growth models is industrial specialization. One should put this finding together with findings from the housing market hedonic literature (e.g., Cheshire and Sheppard 1995) or the literature on population and employment decentralization (e.g., Thurston and Yezer 1994) which concludes that the presence of industry is an environmental ‘‘bad’’ which people pay a premium to have less of in their neighborhoods or move away from. Looked at in this light then one can see that less industry in a city might have attracted mobile population while more unemployment would be less significant. At the same time, more unemployment would be a negative factor in terms of growth of per capita incomes or productivity while more industry was neutral. We also find that a concentration of potentially highly productive workers (university students) was favorable to growth in GDP percent but not significant in the context of population growth; and a concentration of R&D activity was also significantly associated with GDP percent growth but, if anything, negatively associated with population growth. This latter finding is again consistent with a process of sorting of more highly skilled and work oriented people concentrating in cities with faster economic growth while less skilled or work orientated people concentrated in cheaper and cities that were ‘‘nicer’’ to live in. We also find that having a government structure more favorable to promoting local economic growth helped a city’s growth performance in economic terms but had no impact on its population growth. When we compare patterns of spatial interaction and spatial dependence we find revealing features in common. The models do not work in identical ways – the details of the mechanisms of spatial interaction differ – but the fundamental patterns are similar. We can identify economic mechanisms, chiefly relating to search patterns in local labor markets and to differences in the costs of changing commuting patterns compared to migration in order to exploit spatial labor market opportunities, which produce systematic interaction in growth of both population and GDP percent. In the case of growth of GDP percent, we can also identify interaction mechanisms resulting from the tendency to apply innovations locally. Europe seems to be composed of city-states but these are not isolated city-states: where they are densely packed, they locally interact. They still exist largely within national urban systems, however, so even discounting for the low incidence of population mobility in Europe compared to the USA, we should not expect to observe a full spatial equilibrium across the whole set of city-regions in the EU. Not only do national borders still represent substantial barriers to spatial interaction – apparently the equivalent of a day’s travel time – but quality of life differences, although

314

P. Cheshire and S. Magrini

important determinants of the attractiveness of a city to mobile population, still seem be confined in their influence to their own national territories.

References Audretsch DB (1998) Agglomeration and the location of innovative activity. Oxford Rev Econ Policy 14(2):18–29 Barro RJ (1990) Government spending in a simple model of endogenous growth. J Polit Econ 98: S103–S125 Barro RJ, Sala-i-Martin X (1991) Convergence across states and regions. Brooking Pap Econ Activ 1:107–182 Barro RJ, Sala-i-Martin X (1992) Convergence. J Polit Econ 100(2):223–251 Barro RJ, Sala-i-Martin X (1995) Economic growth. McGraw-Hill, New York Cheshire PC (1995) Convergence/divergence in regional growth rates: an empty black box. In: Armstrong HW, Vickerman RW (eds) Convergence and divergence among European regions. Pion, London, pp 89–111 Cheshire PC, Carbonaro G (1996) European urban economic growth: testing theory and policy prescriptions. Urban Stud 33(7):1111–1128 Cheshire PC, Gordon IR (1996) Territorial competition and the logic of collective (in)action. Int J Urban Reg Res 20:383–399 Cheshire PC, Hay DG (1989) Urban problems in Western Europe: an economic analysis. Unwin Hyman, London Cheshire PC, Magrini S (2006a) Population growth in European cities: weather matters – but only nationally. Reg Stud 40(1):23–37 Cheshire PC, Magrini S (2006b) European Urban growth: now for some problems of spaceless and weightless econometrics. Paper given to 46th Congress of the European Regional Science Association, Volos, Greece Cheshire P, Sheppard S (1995) On the price of land and the value of amenities. Economica 62(246):247–267 Clark C, Wilson F, Bradley J (1969) Industrial location and economic potential in Western Europe. Reg Stud 3:197–212 Costa DL, Kahn ME (2000) Power couples: changes in the locational choice of the college educated, 1940–1990. Q J Econ 115(4):1287–1315 Fujita M, Krugman P, Venables A (1999) The spatial economy. MIT, Cambridge MA Glaeser EL, Scheinkman JA, Shleifer A (1995) Economic growth in a cross-section of cities. J Monet Econ 36:117–143 Gordon I, Lamont D (1982) A model of labor-market interdependencies in the London region. Environ Plann A 14:238–264 Gyourko J, Kahn M, Tracy J (1999) Quality of life and environmental comparisons. In: Cheshire PC, Mills E (eds) Handbook of regional and urban economics, 3: applied urban economics. North Holland, Amsterdam Hendry DF, Krolzig HM (2001) Automated econometric model selection using PcGets. Timberlake Consultants, London Hendry DF, Krolzig HM (2004) We ran one regression. Oxford Bull Econ Stat 66(5):799–810 Keeble D, Offord J, Walker S (1988) Peripheral regions in a community of twelve member states. Office of Official Publications, Luxembourg Magrini S (1998) Modelling regional economic growth: the role of human capital and innovation. Ph.D. Thesis, London School of Economics Midelfart KH, Overman HG (2002) Delocation and European integration: is European structural spending justified. Econ Policy 35:321–359

16

Growing Urban GDP or Attracting People? Different Causes, Different Consequences 315

Morrison PS (2005) Unemployment and urban labor markets. Urban Stud 42(12):2261–2288 Oates WE (1999) An essay on fiscal federalism. J Econ Lit 37(3):1120–1149 Roback J (1982) Wages, rents, and the quality of life. J Polit Econ 90:1257–1278 Rodriguez-Pose A, Fratesi U (2004) Between development and social policies: the impact of European structural funds in objective 1 regions. Reg Stud 38(1):97–113 Thurston L, Yezer AMJ (1994) Causality in the suburbanization of population and employment. J Urban Econ 35(1):105–118

Chapter 17

Urban–Rural Development in Sweden Johan Klaesson and Lars Pettersson

17.1

Introduction

Economic activities are confined to urban metropolitan areas in developed countries. The service sector oriented both towards consumers and producers comprises of a substantial part of the economy and, in general, these sectors are dependent on access to regional purchasing power. Jane Jacobs (1969) stressed the importance of urbanisation as an explanation for economic growth when she argued that the diverse economy in urban regions stimulate innovative activities. Thereby urban diversity is likely to cause productivity to increase and promote economic growth. In Jacobs (1984) seminal book ‘‘Cities and the Wealth of Nations’’ she argued that metropolitan regions and urban economies serve as the backbone and engine of the wealth of nations, not vice versa. The cause and effects of urbanization have received a great deal of attention within the endogenous growth theory. For e.g., Lucas (1988) recognizes the way metropolitan regions serve as engines for economic growth because of localized information and knowledge spillovers. In models of endogenous growth, accumulation of human capital and knowledge is understood to explain economic growth (Romer 1994). This means that the regional market size by itself, and hence, the process of urbanization influences the growth of the economy. Growth and urbanization are analyzed in models where workers’ and households’ are allowed to migrate between regions. The seminal core-periphery model attributed to Krugman (1991) represents a type of model that is widely referred to within this field of study that allows us to consider both rural and urbanized regions in one model. As pointed out by Fujita and Thisse (2002), since the ‘‘new’’ theories of growth and ‘‘new economic geography’’ rest on the same modeling framework of monopolistic

L. Pettersson (*) The Swedish Board of Agriculture, Jo¨nko¨ping e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_17, # Springer‐Verlag Berlin Heidelberg 2009

317

318

J. Klaesson and L. Pettersson

competition, there is a solid foundation of cross-fertilisation between the two fields of study. According to data from OECD (World Development Indicators) the growth rate of urban population is significantly higher than the growth rate of total population at the global level. This situation is present in around 175 countries of the world. This process has been stronger in less developed countries compared to developed countries during the last decades. This means that urban regions grow faster than rural regions, and urbanization is taking place in most countries in the world. In all EU15 countries the growth of urban population is higher than the growth of total population. Theoretical models in the field of new economic geography and urban economics explain the phenomenon of urbanization and geographical concentration based on different types of agglomeration economies. The market potential is stronger in dense markets with high population and purchasing power, and firms that face significant fixed cost in the production process have to produce large volumes in order to be price competitive. Since a substantial share of all employment, and the growth of employment, is found in service sectors oriented towards households, it is natural that economic activities cluster in large metropolitan regions. This also explains why it is relevant to characterize the process of development as ‘‘jobs follow people’’ and that urbanization drives economic growth, not the other way around. The development can be described as a process of cumulative causation that is self-reinforcing. There are several arguments as to why we should expect that urbanization is important for growth. But how are rural areas related to urban areas? According to the theory, ‘‘spillover effects’’ from urban areas can stimulate growth in rural areas. Geographical analyses that recognize the spatial structure show that there are large differences with respect to the location of population and jobs in different rural areas. This is also reflected in land rent gradients, where the land rent tends to increase near urban centers, and also has a propensity to increase with the city size. At another level of urban–rural interaction there is also a competition between these two types of areas with respect to jobs that serve external markets. Congestion effects in urban areas may also limit the scope from agglomeration economies that are present in the urban areas. In relation to this background it is relevant to contrast the development in rural areas from the urbanized areas. The role of urban centers is of particular interest for the rural areas, and also for different forms of policy efforts to make rural areas stronger. This is of relevance for a number of interesting questions. For example, do urban centers induce economic growth in nearby rural areas? – If this is the case, how do these links work and what is the mechanism? – Are there differences in growth opportunities in different types of rural areas depending on access to urban centers? The purpose of this paper is to analyze the influence of urban size on the development of neighboring rural population and employment. Doing this we also analyze whether or not jobs follow people or if it is the other way around. Studies in this field frequently utilize the Carlino–Mills (1987) type of models, which are well suited for this particular kind of analysis. The theoretical motivation originates from similar considerations that are stressed in Krugman’s core-periphery model. This is sometimes referred to as the growth centre theory. There is a significant number of

17

Urban–Rural Development in Sweden

319

studies in this field of analysis (for e.g., Hughes and Holland 1994; Henry et al. 1997; Capello and Faggiani 2002; Vermeulen and Van Ommeren 2004). During the recent years we also find a number of studies that focus more explicitly on the interrelationship between rural and urban development. In particular, suburbs and the fringe areas in metropolitan regions have received considerable attention. Voith (1998) found a positive relation between income growth in large city centers and suburbs, and Deitz (1998) also concludes that, according to a number of empirical studies, ‘‘jobs tend to follow people to the suburbs.’’ Henry et al. (1997) use a modified version of a Carlino–Mills type of model that originates from Boarnet (1994) when they analyze the presence of urban spread effects. Henry et al. (1999) and Schmitt and Henry (2000) also take spatial dimensions into consideration, using a spatial ‘‘linkage’’ matrix. A common feature in these studies is that they focus on ‘‘backwash-’’ and ‘‘spillover effects,’’ and also the process of suburbanization and the way urban core areas influence the development in their surroundings. In a study by Westlund (2002) it was shown that the Swedish countryside had a higher population increase than central municipalities during the period between 1990 and 1997. However, the increase was the strongest in municipalities surrounding regional centers. According to this view rural areas can be assumed to grow as a consequence of a kind of urban residual effect. Urbanization and agglomeration economies in densely populated metropolitan regions will be likely to generate some spin-off to adjacent rural regions. This means that we can expect rural areas with access to urbanized areas to have better growth opportunities compared to rural areas that are not able to benefit from these types of ‘‘spillover’’ effects. Adopting a functional economic region (FER) perspective, urban areas and their rural nearby surroundings will compete for jobs that are created in the FER. Rural employment can be viewed as a result of the characteristics of the rural population (working population), land and rural capital that may be associated with cost advantages compared to urban locations. Furthermore, congestion effects that increase with the size of cities may also stimulate a spread of employment to more peripheral locations.

17.2

The Swedish City Structure

In the following empirical analysis we focus on how the economy can be divided into rural and urbanized regions. In order to analyze how nearness to urbanized regions may influence rural areas we must make use of a definition of urban and rural areas respectively. In the literature, we find a number of different definitions that are used in different ways. Cities, and the boundaries of urban areas, are commonly determined with respect to density of houses and a lower boundary for population and/or number of houses. There is no common and widely used definition of ‘‘rural areas in Sweden.’’ Instead there are a number of definitions used by different scholars, official authorities, etc. This is in itself somewhat problematic. Statistics Sweden divides

320

J. Klaesson and L. Pettersson

the Swedish economy into areas that are considered as urbanized (‘‘ta¨torter’’) and non-urbanized (‘‘utanfo¨r ta¨torter’’). We have chosen to focus our study to this type of data published by Statistics Sweden. According to this definition an urbanized area has a population of at least 200 people and not more than 200 m between the housing units. Data is collected and published every fifth year by Statistics Sweden, following this definition time-series data is available from 1800 and onwards. If we consider non-urbanized areas as rural areas we can describe the long-term process of urbanization in Sweden as in Fig. 17.1. In an international perspective a population of 200 people does not seem very urban, in many countries a definition of 2,000 is more common. In any case, the empirical part of this study demonstrates that much larger units have to be used in order to observe expected population growth rates that are positive. The industrialization started in Sweden around 1870–1880. At the same time as the country was industrialized it was also urbanized, and the process of urbanization went on for approximately 100 years. Using this type of viewpoint it appears that the most significant migration from rural- to urban areas slowed down around 1970. From about 1970 this time and onwards around 85% lives in urbanized areas. However, there is still a process of interurban migration with a shift in population from small cities towards larger cities, which is not revealed in Fig. 17.1. The migration during the last decades was from cities with less than 10,000 people to larger urban regions. Table 17.1 shows the structure of cities and urbanized areas in the Swedish economy between 1950 and 2000. The number of small urban areas declines until 100 90 80 70 Percent

60

Urban Rural

50 40 30 20 10 0 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 Year

Fig.17.1 Urban and rural population in Sweden 1800–2000 Source: Statistics Sweden (2005)

17

Urban–Rural Development in Sweden

321

around 1990. During the 1990s there is an increase in small cities and urban areas near the largest cities in the country. We can also see a significant decline in the total number of cities and urban areas, which corresponds to around (–) 14% between 1950 and 1970. The number of cities and urbanized areas then increased by 10% during the last three decades of the nineteenth century. In 1950 more than half of all cities and urbanized areas had a population of less than 500 inhabitants. This share declined to around 40% from 1970 and onwards. Compared to the situation in 1950 another significant change which occurred is the increase in the number of cities and urbanized areas with 2,000 inhabitants and more, which is shown in Table 17.2. From Table 17.1 and 17.2 we can conclude that more people lived in larger cities in the 1990s and 2000s compared to 1950s, and 1970s. This means that the process of interurban migration prevails only after 1970 when the relative balance between population living in rural areas and urbanized areas is stabilized. The five largest metropolitan areas in Sweden host around 30% of the urban population in Sweden in the year 2000. Almost 60% of the urban population lives in cities with more than 20,000 inhabitants (see Table 17.3), which corresponds to almost half of the population in the country. The data describing the structure of cities indicates that the size of urban areas in itself can be a catalyst to growth in population. The development of rural areas can Table 17.1 Number of urban areas in different metropolitan size groups in Sweden Population size 1950 1970 1990 200–499 1,076 756 696 500–999 459 405 421 1,000–1,999 252 233 272 2,000–4,999 153 208 228 5,000–9,999 51 66 116 10,000–19,999 37 58 55 20,000–49,999 18 32 36 50,000–99,999 7 14 15 More than 100,000 3 3 4 Total 2,056 1,775 1,843 Source: Statistics Sweden (2005)

2000 780 440 272 225 111 52 36 15 5 1,936

Table 17.2 Share of urban areas in different; metropolitan size groups in Sweden Population size 1950 1970 1990 200–499 52.3 42.6 37.8 500–999 22.3 22.8 22.8 1,000–1,999 12.3 13.1 14.8 2,000–4,999 7.4 11.7 12.4 5,000–9,999 2.5 3.7 6.3 10,000–19,999 1.8 3.3 3.0 20,000–49,999 0.9 1.8 2.0 50,000–99,999 0.3 0.8 0.8 More than 100,000 0.1 0.2 0.2 Total 100.0 100.0 100.0 Source: Statistics Sweden (2005)

2000 40.3 22.7 14.0 11.6 5.7 2.7 1.9 0.8 0.3 100.0

322

J. Klaesson and L. Pettersson

Table 17.3 Population in urban areas of different size in Sweden 2000 Population size Number of residents 200–499 249,808 500–999 311,448 1,000–1,999 389,142 2,000–4,999 712,586 5,000–9,999 773,222 10,000–19,999 740,767 20,000–49,999 1,051,727 50,000–99,999 1,053,012 More than 100,000 2,183,149 Total 7,464,861 Source: Statistics Sweden (2005)

Percent 3.3 4.2 5.2 9.5 10.4 9.9 14.1 14.1 29.2 100.0

therefore also be assumed to be related to access to urban cores’ of a specified size. Empirical models of regional development often aim to capture the interdependence between household choice of residential locations and firm location decision. Studies in this field commonly address and describe the causality between household and firm location as a question if ‘‘jobs follow people’’ or ‘‘people follow jobs.’’ The seminal work by Carlino and Mills (1987) utilized a two-equation simultaneous model to analyze this type of problem. This type of model has subsequently been used in a number of studies focused on commuting zones (see for e.g., Boarnet 1994; Deitz 1998). In extended versions of this model Henry et al. (1997) and also Schmitt and Henry (2000) include more rural areas in the empirical analysis. The results from these studies indicate that small urban places and rural areas are dependent on the size of the neighboring larger urban core centers. These studies also show that households tend to be more sensitive towards the presence of amenities compared to businesses. Given the stylized facts for the Swedish cities structure and the development during the last decades, and experiences from similar types of studies in other countries, we expect that there is a ‘‘critical size’’ for cities, which have to be surpassed in order to establish a self-reinforced growth process. When we explore the Swedish data we find that the population in a city and urbanized area of around 25,000 inhabitants appears to be a level when the self-reinforcing mechanism works in a cumulative way and we can expect a positive growth of population. This relation is shown in Fig. 17.2, which pictures the relation between municipal population growth 1993–2003 (on the vertical axis) and the population size 1995 of the urbanized areas in the municipalities (on the horizontal axis). In order to further understand the spatial distribution of the population with respect to urban and rural living, we also explore the relation between urbanization and population size in municipalities in Sweden. In Fig. 17.3 this relationship is plotted for all 290 municipalities in Sweden for the year 2000. We use the logarithm of population size, which related to the share of people living in urbanized areas in the municipalities. According to the linear regression, and the plotted distribution, there is a positive relationship, which can explain more than 50% of the variation.

17

Urban–Rural Development in Sweden

323

40

Percentage population growt

y = 4.3288Ln(x) – 43,722, R2 = 0.3789 30 20 10 0 –10 –20 1000

10000 100000 Population in Urbanized Areas

1000000

Fig.17.2 Growth in municipal population 1993–2003 and population size in urbanized areas in Sweden 1995

Share of population living in urbanised areas

100% 90% 80%

y = 0.1047x – 0.2563 R2 = 0.5395

70% 60% 50% 40% 30% 20% 6

7

8

9

10 11 LN Population

12

13

14

Fig.17.3 Share of population living in urbanized areas and the logarithm of population in municipalities in Sweden 2000

One additional way to illustrate the relation between population growth and urbanization is to explore how the long-term growth of population varies with respect to the share of people living in urbanized areas. This relationship is pictured in Fig. 17.4 for municipalities in Sweden. Percentage change in population during 1993 and 2003 is measured on the vertical axis, and share of people living in urbanized areas in 1995 is measured on the horizontal axis. We find that the

J. Klaesson and L. Pettersson

Percentage population growth

324 40

y = 23.244x – 19.772, R2 = 0,2219 30 20 10 0 –10 –20 0.25

0.45 0.65 0.85 Share of population in Urbanised areas

1.05

Fig.17.4 Population growth 1993–2003 and share of population living in urbanized areas in Swedish municipalities 1993

estimated linear relationship intersect with the 0-change in population at the 85% level of urbanization (85% of the population in the municipalities lived in urbanized areas in 1993). It is also possible to show the spatial structure of the Swedish economy using maps (Fig. 17.5). The population density in Sweden is very low in the northern part of the country. Significant parts of northern Sweden have less than five persons per square kilometre. The highest density is found in the regions of the three largest cities (Stockholm, Go¨teborg, and Malmo¨). However, in an international comparison the overall density is relatively low in the country, in average 22 persons per square kilometre. We find that there are not more than approximately 40 cities that have at least 25,000 inhabitants (the critical size for self-reinforced growth revealed in Fig. 17.2). Furthermore, when we analyze the distribution of municipalities with respect to the share of people living in rural areas outside cities and urbanized areas, the picture is a bit mixed. We find for e.g., that, according to this type of measure, few people live in rural areas, in remote regions that have a very low population density. This means that in these regions people are likely to live in villages (‘‘urbanized’’) although the regions often are regarded as rural. From the explorative analysis presented above we find arguments’ supporting the fact that urbanization is still an on-going process in Sweden. Even if the division of population between rural areas outside cities and population living in urbanized areas reached a stable position around 1970, there is a movement from smaller places toward larger places. We should also expect a variable that reveals access to large places as a factor that can explain local development, also in rural areas. 25,000 inhabitants appears to be a critical size of cities in the Swedish economy, which we will be able to utilise in the following empirical analysis, which is also the critical size for a non-negative expectation of population development.

17

Urban–Rural Development in Sweden

Population density in municipalities

Municipalities having a Town with more than 25000 inhabitants

325

Four types of municipalities based on share of rural population

Fig.17.5 Three maps of Sweden showing density, large towns and share of rural population. The leftmost map shows quintiles of population densities – darker color signifies higher densities. The middle map identifies the municipalities having within its borders an urban centre of at least 25,000 inhabitants. The rightmost map shows the four groups of municipalities presented in Table 17.4, going from black to lighter colors. We have, Highly Urbanized, Urban, Rural and Highly Rural municipalities

17.3

Empirical Analysis

The hypothesis under study is that working age population and jobs develop together. An increase in municipal population gives an impulse for jobs to move there and vice versa. In addition, the initial working age population and the initial number of jobs are hypothesised to influence the subsequent development. As was shown earlier in this paper, municipalities that had a town of 25,000 inhabitants within its borders had an equal chance of growing or declining in population under the studied period. Municipalities that did not have a town of 25,000 inhabitants can be expected to decrease in population whereas a municipality that has can be

326

J. Klaesson and L. Pettersson

expected to grow in population. In order to capture this growth/decline impact an accessibility measure is constructed. This variable measures the accessibility in all municipalities to population in municipalities that have a town within its borders that have more than 25,000 inhabitants1. Following the tradition of Carlino–Mills type of models we can formulate the following simultaneous system of equations: Jtþt Ptþt ¼ a1 þ a2 ln þ a3 ln Jt þ a4 ln Acc Jt Pt

ð17:1Þ

Ptþt Jtþt ¼ b1 þ b2 ln þ b3 ln Pt þ b4 ln Acc Pt Jt

ð17:2Þ

ln

ln

where Jt and Jt+t measure the number of jobs in a municipality in the year 1993 and 2003 respectively. Pt and Pt+t measure the number of persons between the ages 20 and 65 living in the municipality for the same 2 years. Acc measures the accessibility to population in municipalities with a town of at least 25,000 inhabitants. We use the accessibility to cities with at least 25,000 inhabitants as a variable that takes the spatial structure into consideration2.Instead of using a spatial weight matrix we capture the spatial structure in this variable in the model, and consider this as one particularly important aspect of local attractivity. Before the regression analysis, the 289 Swedish municipalities are divided into four groups according to their degree of rurality. Five observations had to be dropped from the analysis since they did not exist as independent municipalities at the beginning of the period under study. As a proxy for rurality the share of the population living in cities and urbanized areas are used, people not living in these areas are regarded as living on the countryside in rural areas. In Fig. 17.1 it was shown that the share of population living in the rural areas in Sweden has been almost constant, around 15% since around 1970. Based on this information we, in the first step, consider municipalities with shares of population in the rural areas less than 15% as urbanized regions. The rest of the municipalities, with smaller share of people living in urban areas, are considered as rural. Using these criteria 67 municipalities are labeled highly urbanized and 222 as rural in the Swedish economy. The rural municipalities are in turn divided into three equally sized groups based on the share of population living in the rural areas of the municipalities. Table 17.4 gives information on the four different groups of municipalities.

1

Accessibility to population in municipalities with an urban centre larger than 25,000 inhabitants is P defined as: Accs ¼ Pt expðltsi Þ where Accs is the accessibility to population in municipality t

s, Pt is the population in municipality t and tst is the time–distance (travel time by car in minutes) between municipality s and t. l is an estimated distance-friction parameter. The sum goes over all municipalities with urban centers above 25,000 inhabitants. 2 The accessibility variable is assumed to reflect both size and diversity since these two variables are highly correlated with each other.

17

Urban–Rural Development in Sweden

327

Table 17.4 Share of population living in towns in the four groups of municipalities Municipality type Average share of population Cut-off points for the Number of living in urbanized areas intervals municipalities 1 (Highly urbanized) 0.923 0.85–1 67 2 (Urban) 0.801 0.76–0.849 73 3 (Rural) 0.704 0.64–0.759 74 4 (Highly rural) 0.541 0.3–0.639 72

Table 17.5 Description of data used in regression analysis Mun. Jobs Change %Working Change in type 1993 in jobs change age working in jobs population age 1993 population 1 2 3 4 Total

32,176 10,824 6,577 4,178 13,054

4,373 305 247 108 1,194

13,6 2,8 3,8 2,6 9,1

41,026 15,225 9,736 6,538 17,662

3,451 –22 –91 –185 733

%-Change in working age population 8,4 –0,1 –0,9 –2,8 4,2

Accessibility to municipalities with an urban centre larger than 25,000 inhabitants 146,989 59,039 44,926 20,924 66,396

Group 1 contains the urbanized regions with municipalities with a share of population living in cities and urban areas at least equal to 85%. The division of the rural groups 2, 3, and 4 of municipalities have increasing shares of population living in the rural areas as shown in Table 17.4. In Table 17.5 the data used is summarized for the four types of municipalities. The number of jobs and the size of the working age population in 1993 and the change to 2003 are displayed. The more rural the municipalities are the smaller they are and the weaker their growth. On an average all types of municipalities have had an increase in jobs. However, the increase has been much smaller in the rural types of municipalities (type 2, 3, and 4). The change in the size of the working age population are very different in the urban municipalities (type 1) compared to the rural ones (type 2, 3, and 4). The urban municipalities show quite a large growth whereas the rural municipalities of all three types decline. The decline in the rural municipalities increases as the degree of rurality gets larger both in absolute and relative terms. Turning to the accessibility variable it can be seen that the accessibility to population in municipalities with an urban population of at least 25,000 people gets smaller when going from urban to more rural municipalities. A 2-SLS regression is performed for the two (17.1) and (17.2) for each of the four types of municipalities. Table 17.6 provide the results of the regression analysis. The first striking feature of the information in the table is that the population equation seems to work much better than the job equation. In the population equation all parameter estimates are significant and have the expected signs. In the results for the job equation the coefficient for the population change

328

J. Klaesson and L. Pettersson

Table 17.6 Results from regression equation (17.1) and (17.2) Dependent variable DJobs Mun. type Const

DPop Jobs

Acc

R2-adj Const

DPop DJobs Pop

Acc

R2-adj

1

–0.039 0.64 –0.0014 0.012 0.48 –0.46 0.34 0.037 0.0073 0.61 (–0.34) (4.45) (0.16) (2.83) (–5.67) (3.98) (5.76) (2.21) 2 –0.081 0.74 0.0068 0.0046 0.39 –0.38 0.40 0.026 0.010 0.48 (–0.55) (5.6) (0.53) (0.72) (–3.51) (5.39) (2.64) (2.24) 3 0.11 1.19 0.0015 –0.0064 0.62 –0.35 0.42 0.018 0.015 0.74 (0.78) (8.51) (0.12) (–1.03) (–4.57) (8.28) (2.09) (4.62) 4 0.56 1.31 –0.053 –0.0041 0.48 –0.56 0.34 0.051 0.0088 0.73 (3.14) (7.17) (–2.91) (–0.78) (–8.98) (7.55) (7.12) (3.88) Note: t-statistics are presented in brackets. Bold letters indicate significance at the 0.05 level

variable is the only significant result with the exception of the accessibility variable for the municipalities of type one which is the most urban type of municipality. The working age population change parameter influences the job change in a positive way. This parameter is monotonically increasing when moving from the urban type of municipality to the more rural types. It rises from 0.64 to 1.31. The model explains between 40 and 75% of the variation in the independent variables. In general, the explanatory power (R2-adj) is higher for the population equation compared to the job equation for all four different types of municipalities. We also find that accessibility to larger urban areas is more important when we analyze the change in population compared to the change in jobs. The change in the number of jobs appears to be more tied to our measure of access to larger urban areas as we move our focus in the study from urban to more rural municipalities, except for the most peripheral regions. The change in population is the most significant explanatory variable in our model with respect to explaining the change in jobs and employment. Access to larger urban areas is only significant for municipalities in the largest regions in the economy. One interesting observation is that all the variables in our model are significant in all four types of municipalities when we focus on explaining the change in population. At the same time it is more or less only one variable – change in population – that results in significant estimates in the ‘‘job-equation.’’ Our interpretation is that our empirical analysis supports the hypothesis that ‘‘Jobs follow people,’’ and that there is a need for a larger array of variables when we analyze the change of population compared to the change of jobs and employment. There are also other observations that support this conclusion, for e.g., the fact that around half of all employment in Sweden is within the service sector oriented towards households. Since several parts of the service sector are dependent on access to local purchasing power it is logical to assume that the growth of jobs in the service sector is related to the growth in population to a substantial degree. Accordingly, the change in jobs appears to be dependent on local market size, also in terms of population (which represents purchasing power).

17

Urban–Rural Development in Sweden

17.4

329

Conclusions

Urbanization and access to urban areas of a ‘‘minimum-size’’ appears to be an important factor for development, both in urban and rural areas. There are theoretical explanations that support this consideration in the literature within the field of urban and regional economics, which builds on the endogenous growth theory. The so-called field of ‘‘new economic geography,’’ for e.g., the core-periphery model highlights the relation and the interdependence between rural and urban areas. According to these theories, the presence of agglomeration economies motivates why we can assume that there is a positive relation between urbanization and growth. However, congestion and economies of specialization are also important to acknowledge that can serve as motivation for why rural areas also may have a significant development. There are various forms of ‘‘backwash-,’’ ‘‘spillover effects’’ and also so-called ‘‘Christaller effects’’ etc. that indicate the phenomenon of interdependence between urban- and rural areas. In this paper we have explored and analyzed some features of the interdependence between urban and rural areas in Sweden. We find that a critical size of urban areas, in order to establish a self-reinforced growth of population, is around 25,000 inhabitants. In a way this may be noticed as a ‘‘minimum level’’ of city size when we can expect population growth to sustain and when the agglomeration of activities has reached its minimum size. In the empirical analysis we employ a Carlino–Mills type of model in order to analyze the interdependence between the change of jobs and the change of working age population at the municipality level during the years 1993–2003. According to the empirical results, the variable working age population is the most significant with respect to explaining the change in working age population. At the same time the change in working age population must be analyzed using a larger array of variables, and is also more dependent on access to urban areas of the minimum size (with at least 25,000 inhabitants). In municipalities that are considered to be more peripheral, access to urban areas of minimum size is more important compared to more urbanized areas. Our interpretation is that the empirical results support the hypothesis of ‘‘jobs follow people.’’

References Boarnet MG (1994) An empirical model of intrametropolitan population and employment growth. Pap Reg Sci 73:135–153 Capello R, Faggiani A (2002) An economic-ecological model of urban growth and urban externalities: empirical evidence from Italy. Ecol Econ 40:181–198 Carlino G, Mills ES (1987) The determinants of county growth. J Reg Sci 27:39–54 Deitz R (1998) A joint model of residential and employment location in urban areas. J Urban Econ 44:197–215 Fujita M, Thisse J-F (2002) Economics of agglomeration: cities, industrial location, and regional growth. Cambridge University Press, New York

330

J. Klaesson and L. Pettersson

Hughes DW, Holland DW (1994) Core-periphery economic linkage: a measure of spread and possible backwash effects for the Washington economy. Land Econ 70(3):364–377 Jacobs J (1969) The economy of cities. Random House, New York Jacobs J (1984) Cities and the wealth of nations. Random House, New York Krugman P (1991) Geography and trade. MIT, Cambridge, MA Lucas RE Jr (1988) On the mechanics of economic development. J Monet Econ 22(1):3–42 Romer PM (1994) The origins of endogenous growth. J Econ Perspect 8(1):3–22 Schmitt B, Henry MS (2000) Size and growth of urban centers in French labor market areas: consequences for rural population and employment. Reg Sci Urban Econ 30(1):1–21 Statistics Sweden (2005) Localities 2005, MI 38 SM 0601, Stockholm Vermeulen W, Van Ommeren J (2004) Interaction of regional population and employment: identifying short-run equilibrium adjustment effects. Tinbergen Institute Discussion Paper, TI 2004 – 083/3, Amsterdam Voith R (1998) Do cities need suburbs? J Reg Sci 38:445–464 Westlund H (2002) An unplanned green wave: settlement patterns in Sweden during the 1990s. Environ Plan A 34(8):1395–1410

Chapter 18

Patents, Patent Citations and the Geography of Knowledge Spillovers in Europe Manfred M Fischer, Thomas Scherngell, and Eva Jansenberger

18.1

Introduction

As interest in questions of the knowledge economy has grown, knowledge spillovers have received increased attention in recent years. For the purpose of this paper we use externalities and knowledge spillovers interchangeably to denote the non-pecuniary benefit of knowledge to firms, not responsible for the original investment in the creation of this knowledge. Such spillovers arise when some of the R&D activities have the characteristics of a non-rivalrous good and cannot be appropriated entirely. A fundamental question addressed by empirical research on knowledge spillovers is whether these spillovers are geographically bounded or not (see Karlsson and Manduchi 2001). Most of the studies on this issue thus far have concentrated on the spatial extent of local geographic effects that university research may have on the innovative capacity in a region, both directly and indirectly through its interaction with private sector R&D efforts. The studies vary somewhat in terms of research design, but they all find a strong and positive relationship between innovative activity and both industry R&D and university research at the state level in the USA. But the situation is different in terms of the significance of a local geographic spillover effect. Overall, the evidence is non-existent, weak or mixed, and only pertaining to a few individual sectors (Anselin et al. 1997). This may be due to the fact that knowledge spillovers are measured indirectly rather than directly. The only direct evidence we have for localized knowledge spillovers is based on Jaffe et al. (1993) pioneering analysis on patent citations. Our study lies in this research tradition1 and takes patent citations as a proxy for knowledge spillovers to M.M. Fischer (*) Institute for Economic Geography & GIScience, Vienna University of Economics and Business Administration, Austria e-mail: [email protected] 1 See Almeida (1996); Almeida and Kogut (1999); Hicks et al. (2001); Maurseth and Verspagen (2002); Agrawal, Cockburn and McHale (2003); Singh (2003), and Verspagen and Schoenmakers (2004) for examples of this research tradition.

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_18, # Springer‐Verlag Berlin Heidelberg 2009

331

332

M.M. Fischer et al.

test for spillover localization in Europe. We are particularly interested in knowledge spillovers between high-technology firms. High-technology is defined in our context to include the ISIC-sectors (ISIC Rev. 2) pharmaceuticals (ISIC 3522), computers and office equipment (ISIC 3825), electronics-telecommunication (ISIC 3832), and aerospace (ISIC 3845). Though some firms may choose not to patent inventions, patenting in high-technology industries is commonly practiced and indeed a vital component of maintaining technological competitiveness. The European coverage of our study is given by patent applications at the European Patent Office (EPO) that are assigned to high-technology firms located in the EU-25 member states2, the two accession countries Bulgaria and Romania, and Norway, and Switzerland. Space is considered in a discrete representation of 188 regions.3 The rest of the paper is structured as follows. The next section explains the nature of patents in some more detail and illustrates those inventive activities in the high-technology sector in Europe to be geographically clustered. Knowledge flows are notoriously difficult to measure. Following Jaffe, Trajtenberg and co-authors (see Jaffe and Trajtenberg 2002 we use patent citation data as an indicator for a specific type of knowledge spillovers between inventors. Sect.18.3 considers more carefully how citations might be used to infer spillovers. The Section, moreover, illustrates how patent citations between high-technology firms are spread across Europe and tend to be geographically clustered. Section 18.4 follows the pioneering methodology developed in Jaffe et al. (1993) and compares the extent to which actual citations are disproportionately located in space, relative to a distribution of control patents that have the same temporal and technological characteristics. This comparison allows us to control for any technology-based clustering of inventive activity, which may otherwise confound any inference drawn from co-location of citations. Section 18.5 concludes with a summary of our main findings and some suggestions for future work.

18.2

Patents and Patent Data

Patents have long been recognized as an important and fruitful source of data for the study of innovation and technological change (see Griliches 1990 for a survey of the use of patent statistics). A patent is a property right awarded to inventions for the commercial use of a newly invented device. An invention to be patented has to satisfy three patentability criteria. It has to be novel and non-trivial in the sense that it would not appear obvious to a skilled practitioner of the relevant technology, and it has to be useful, in the sense that it has potential commercial value. If a patent is granted, an extensive public document is created. The document contains detailed information about the technology of the invention, the inventor, the assignee that owns the patent rights, and the technological antecedents of the invention. 2

Except Cyprus and Malta. For the definition of the regions see Annex.

3

18

Patents, Patent Citations and the Geography of Knowledge Spillovers in Europe

333

Because patent documents record the residence of the inventors they are an important resource for analyzing the spatial extent of knowledge spillovers, as captured by patent citations. Patent related data have, however, two important limitations. First, the range of patentable inventions constitutes only a subset of all R&D outcomes, and second, patenting is a strategic decision and, thus, not all patentable inventions are actually patented. As to the first limitation, purely scientific advances devoid of immediate applicability as well as incremental technological improvements which are too trite to pass for discrete, codifiable inventions are not patentable. The second limitation is rooted in the fact that it may be optimal for inventors not to apply for patents even though their inventions would satisfy the criteria for patentability (Trajtenberg 2001). Inventors balance the time and expense of the patent process, and the possible loss of secrecy which results from patent publication, against the protection that a patent potentially provides to the inventor (Jaffe 2000). Therefore, patentability requirements and incentives to refrain from patenting limit the scope of our analysis based on patent data. Patents from different national patent offices are not comparable to each other because of difference in patent breadth, patenting costs, approval requirements, citation practices and enforcement rules across Europe. This makes patent data from the (EPO) rather than national patent offices a natural choice for our study.4 Our data source is the (EPO) database. The data on patent applications that we use in this study were drawn from the universe of European patents. By European patents, we mean patents assigned to corporations located in Europe, regardless of the nationality of the inventors. Our sample of patents is limited to those that are related to inventions in the high-technology industries or in other words to those patents assigned to patent classes which match the high-technology sector, at the four-digit level of the International Standard Industrial Classification, ISIC Rev. 2. We used MERIT’s concordance table (see Verspagen et al. 1994 between the fourdigit ISIC-sectors and the 628 patent subclasses5 of the International Patent Code (IPC) classification to identify the high-technology patents from the universe of European patent applications. Our database contains all the high-tech patents applied at the EPO from 1985 to 2002, totalling 177,424 patents. Each patent application produces a highly structured public document containing detailed information on the invention itself, the technological area to which it belongs, the inventor and her/his address, and the organization to which the inventor assigns the patent property right. By nature of the research question, we are interested in the geographical location of the inventor 4

At present national systems and the European system function in parallel though inventors tend to be making increasing use of the European system. This is especially true for inventors in smaller European countries looking for wider geographical protection for their inventions. But nevertheless it should be noted that patent data from the EPO cover only a subsample of patents applied for in Europe. 5 The IPC system is an internationally agreed, clear-cut non-overlapping hierarchical classification system that consists of five hierarchical levels. At the third level 628 subclasses are distinguished.

334

M.M. Fischer et al.

rather than the applicant and, thus, use the postal code of the inventor address for tracing inventive activities back to the region of knowledge production. For representing geographic space we use – as already mentioned in the introduction – 188 regions that cover the EU-25 countries (except Cyprus and Malta), Bulgaria, Romania, Norway, and Switzerland. Their definition is based on the Nomenclature des Unites Territoriales Statistiques [NUTS]. The regions are essentially in line with the NUTS-2 level of the regional classification in the case of Austria, Belgium, Germany, Finland, France, Italy, The Netherlands, Portugal, Spain, Sweden and UK, and in line with the NUTS-0 level in all other cases. See Table 18.6 (in the Annex) for the exact definition of the regions. The patent subclasses associated with the four digits ISIC sectors are outlined in Table 18.7. Figure 18.1 shows that inventive activities of high-technology industries, as measured in terms of EPO-patent activities (1985–2002), are unevenly distributed across Europe. High patenting activity is located in the Iˆle-de-France (9.21% of European patenting), followed by Oberbayern (6.76%), Switzerland (4.49%), Noord-Brabant (4.46%) and Darmstadt (3.52%). The Top-25 regions (see Table 18.1)

Fig. 18.1 Geographic distribution of high-technology EPO-patents across European regions (1985–2002) (measured in terms of shares in European high-technology patenting). Source: European Patent Office, Macon AG (Geodata)

18

Patents, Patent Citations and the Geography of Knowledge Spillovers in Europe

335

Table 18.1 Top-25 European regions in high-technology patenting (1985–2002) NUTSCode Region Share in European patenting FR10 Iˆle-de-France 9.21 DE21 Oberbayern 6.76 CH00 Switzerland 4.49 NL41 Noord–Brabant 4.46 DE71 Darmstadt 3.52 DEA1 Du¨sseldorf 2.87 IT20 Lombardia 2.83 DEA2 Ko¨ln 2.72 DEB3 Rheinhessen–Pfalz 2.41 SE01 Stockholm 2.28 FR71 Rhoˆne–Alpes 2.18 DK00 Denmark 1.93 DE11 Stuttgart 1.91 FI16 Uusimaa 1.83 DE12 Karlsruhe 1.78 UKH1 East Anglia 1.76 DE30 Berlin 1.65 UKI2 London region 1.52 DE13 Freiburg 1.45 DE25 Mittelfranken 1.31 UKH2 Bedfordshire and Hertfordshire 1.31 UKJ2 Surrey 1.27 DEF0 Schleswig–Holstein/Hamburg 1.22 FR82 Provence-Coˆte d’Azur 1.03 UKH3 Essex 1.02 Sum Data Source: European Patent Office

64.72

account for about two thirds of the total number of high-technology patents, which indicate a high geographic concentration in only a few European regions. It is notable that Eastern European regions and Southern European regions (except Northern Italy) display very little patent activity.

18.3

Knowledge Spillovers, Patent Citations and Data

Patent documents include references or citations to patents. These citations open up the possibility of tracing multiple linkages between inventions, inventors, firms, and locations. In particular, patent citations enable us to analyze the geographical extent of spillovers. There are, however, also some serious limitations to the use of patent citation data. Patent citations capture only those spillovers which occur between patented inventions and, thus, underestimate the actual extent of knowledge spillovers. Other channels of knowledge transfer – for e.g., transfer of knowledge embodied in skilled labor, knowledge transfer between customers and suppliers, knowledge exchange at conferences and trade fairs – are not captured by patent citations. Patent citations do not always represent what we typically think of as

336

M.M. Fischer et al.

knowledge spillovers. Some citations may represent only indirect knowledge spillovers since the patent examiner added them. This noise creates a bias against finding spillovers. Fortunately, bias in this direction is a problem of power which can be overcome with a sufficiently large sample size (Thompson 2003). In constructing the patent citation data set that forms the basis of our study we begin with the full set of issued patents that have their application year between 1985 and 2002. There are 177,424 high-technology patents. We then discard all patents that have not received any citations, since our study is using citations as a proxy for knowledge spillovers. We don’t believe that this elimination results in a selection bias since we are interested in comparing the fraction of citations that are from the same location as the original patent, a measure that is conditional on there being citations. Consequently, 42.9% of the patents are discarded, leaving 101,247 patents which generate 210,667 citations. The observation of citations is evidently subject to a truncation bias because we observe citations for only a portion of the life of an invention, with the duration of that portion varying across patent cohorts. This means that patents of different ages are subject to different degrees of truncation. To overcome this problem at least in part we have identified all the pairs of cited and citing patents where citations to a patent are counted for a window of 5 years following its issuance6. The analysis is, thus, confined to 1985–1997 in case of cited patents while citing patents appearing in 1990–2002 are taken into account. This process reduces the number of patents to 69,814 that generate 155,462 citations. Next, we discard 36.8% of those citations for which the citing patent is a self-citation7, because self-citations do not represent knowledge spillovers in the sense of externalities. This leaves us with 98,191 citations or observations that link a citing patent to a cited patent. The original data come in form of citations made (that is, each patent lists references to previous patents) while for identifying the knowledge flows one needs a list of cited and citing patent applications. To obtain the citations received by any one patent issued in year t, one needs to search the references made by all patents applied after year t. This requires in fact fast access to all citation data in a way that permits efficient research and extraction of citations not by the patent number of the citing patent, but by the patent number of the cited patent. The unit of analysis is the dyad “cited patent-citing patent.” A single originating patent, for e.g., that has two inventors and is cited by three subsequent patents will generate six unique observations. Each patent is assigned to one of the 188 regions based on the home address of the inventors as reported in the patent document. The 98,191 observations are illustrated in Fig. 18.2. The nodes represent the 188 European regions, their size is relative to their spillover generating power measured in terms of citations received. 6

The mean citation lag of all 210,667 citations is 4.6 years, with some sectoral differences: pharmaceuticals (4.4 years), computers and office equipment (4.4 years), electronics-telecommunication (4.7 years) and aerospace (5.4 years). 7 We consider assignee matches as self-cites. This is in agreement with most citation-based empirical research.

18

Patents, Patent Citations and the Geography of Knowledge Spillovers in Europe

337

Fig. 18.2 Knowledge flows between European regions, as captured by interfirm patent citations in the high-technology sector (1985–2002) Data Source: European Patent Office; visualization tool: Borgatti et al. (1999)

Figure 18.3 complements this picture illustrating the geography of knowledge spillovers across Europe. This figure classifies the 188 regions according to their spillover generating power (measured in terms of citations received; see Fig. 18.3) and their spillover absorbing power (measured in terms of citations made; see Fig. 18.3b). Both figures pinpoint to a centre-periphery pattern that is in close line with the pattern of patenting activity across Europe as observed in Fig. 18.1.

18.4

Testing for Geographic Localization

Patents linked by citations not only share a technology, but they are also often developed by inventors working in a common industry. Patents linked by a citation are, therefore, much more likely to share a geographic location than a pair of patents drawn at random from the entire pool of patents. To control for the tendency of inventive activities to be geographically clustered – as observed in Sect. 18.2 – we follow the case-control matching approach pioneered by Jaffe et al. (1993).

338

M.M. Fischer et al.

Fig.18.3 Knowledge spillovers between high-technology firms (1985–2002): (a) spillover generating regions and (b) Spillover absorbing regions Data source: European Patent Office, Macon AG (Geodata)

The essence of this case-control approach is to compare citing patents with control patents in terms of the frequency with which each is located in the same region as the originating patent. A finding of a disproportionate number of colocated citations relative to co-located control patents is interpreted as evidence of

18

Patents, Patent Citations and the Geography of Knowledge Spillovers in Europe

339

localized knowledge spillovers. The reason for utilizing controls is that patent citations will tend to be co-located with the original inventions even in the absence of knowledge spillovers when inventive activity in particular industries is clustered geographically Agrawal et al. (2003). Therefore the spillover effect is identified in our study as the extent of co-location which exists over and above what we would expect given the geographic concentration of inventive activity by the high-technology sector. More formally, let P (citation) be the probability that the originating patent and the citing patent are geographically matched, and P (control) be the corresponding probability for the originating patent-control patent match. Assuming binomial distributions, we test the null hypothesis H0: P(citation)=P(control) vs. the alternative hypothesis Ha: P(citation)>P(control) using the test statistic PðcitationÞ  PðcontrolÞ t ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 nfPðcitationÞ½1  PðcitationÞ þ PðcontrolÞ½1  PðcontrolÞg where PðcitationÞ and PðcontrolÞ are the sample proportion estimates of P(citation) and P(control). This statistic tests for the difference between two independently drawn binomial proportions. A positive significant value of Student’s t indicates support of the proposition that knowledge flows, proxied by patents cited by the originating patents, are geographically more located than expected. We use the following procedure to construct the set of control patents. A control patent is selected for each originating patent that matches the citing patent on the following two dimensions: application year and technology classification. Having generated the set of patents with the same application year and the same original three-digit IPC classification code as the citing patent, we identify the patent in the set that has the closest application date to the citing patent. Next, we confirm that the patent does not cite the original patent. If it does, the patent is removed from the set of potential control patents and the next best control patent is selected. Finally, if there are no patents that match the citing patent in at least the application year and the original IPC-classification without citing the original patent, the observation (originating patent) is removed from the data set. We consider two cohorts of originating patents, and corresponding sets of citing patents and control patents to test for spillover-localization. One consists of 1990 patent applications and the other of 1995 applications drawn from our patent database described in Sect. 18.2. Table 18.2 briefly describes these two samples. The 1990 cohort of originating patents contains 2,118 patents that have received a total of 2,362 citations (including self-citations) and 1,410 citations excluding selfcitations by the end of 1995. The 1995 cohort of originating patents contains 1,814 patents that have received a total of 2,387 citations (including self-citations) and 1,366 citations excluding self-citations by the end of 2000.

340

M.M. Fischer et al.

Table 18.2 Descriptive statistics Samepatent Meancitations Average Sample Patents Total Selfclassb (%) receivedc (number) citations citesa citation (%) lagd 1990 Cohort of 2,118 2,362 31.75 76.54 1.94 4.45 originating patents 1995 Cohort of 1,814 2,387 31.84 77.13 1.95 4.57 originating patents a A self-citation is defined as a citing patent assigned by its inventor to the same party as the originating patent b Comparison is at the three-digit level of the IPC classification c For those patents receiving any citations d Application year of the citing patent minus application year of the originating patent Table 18.3 Geographic matching fractions 1990 originating cohort 1995 originating cohort Number of citations Incl. self-cites 2,362 2,387 Excl. self-cites 1,410 1,366 Matching by country Overall citation matching (%) 60.1 61.2 Matching excl. self-cites (%) 36.6 35.9 Control matching (%) 21.9 25.4 t-statistic(excl. self-cites) 8.68 (p= 0.00) 6.01 (p= 0.00) Matching by region Overall citation matching (%) 36.7 37.0 Matching excl. self-cites (%) 13.7 14.8 Control matching (%) 5.2 5.4 t-Statistic(excl. self-cites) 7.91 (p= 0.00) 8.27 (p= 0.00) Note: The t-statistic tests equality of the citation proportion excluding self-citations and the control proportion. See text for details

The results of the case-control tests are provided in Table 18.3 for both cohorts of originating patents. Localization effects are reported at two spatial levels: the regional and the country level of analysis. “Number of citations” corresponds to the number of citations cited by the originating cohort of patents. “Overall Citation Matching,” “Citation Matching excluding Self-Cites” and “Control Matching” are the percentages of cited patents (with and without self-citations) and controls that belong to the same geographic location as the originating patent. The t-statistic tests the equality of the control proportions and the citation proportions excluding selfcitations. Let us focus first on the 1990 results as displayed in the second column of Table 18.3. Starting with the country match, we find that citations including selfcitations are intranational about 38% points more often than the controls. Excluding self-citations cuts this difference roughly in half. The remaining difference between the citations excluding self-citations and the controls is strongly significant statistically. Looking at the 1990 results for regions, we find that citations of patents come from the same region about 37% of the time. Excluding self-citations, however,

18

Patents, Patent Citations and the Geography of Knowledge Spillovers in Europe

341

makes a big difference. The proportions are cut to 13.7%. The matching frequency excluding self-citations is significantly greater than the matching control proportion. The results for patent citations of 1995 patents are similar (see the third column in Table 18.3). For both cohorts of originating patents and for both geographical levels, the patent citations are quantitatively and statistically more significantly localized than the controls. The citation matching percentages slightly rise at the regional level from 13.7% in 1990 to 14.8% in 1995, but slightly decrease at the country level from 36.6 to 35.9%. It is impossible, however, to tell from this comparison whether this represents a real change, or whether it is the result of differences in average citation lags.8 The results on the extent of localization can be summarized as follows. For citations observed by 1,410 of the 1990 originating cohort of patents, there is a clear pattern of localization at the regional and country levels. Citations are about seven times more likely to come from the same region than control patents, 2.6 times more likely excluding self-citations. They are 2.7 times more likely to come from the same country as the originating patents, and 1.7 times excluding self-citations. For citations of 1995 originating patents, the same pattern emerges. All these differences are statistically significant at a level much less than 1%. However, localization of knowledge spillovers is not a universal phenomenon. European regions reveal different patterns in the local diffusion of knowledge externalities. Table 18.4 presents the results for the Top-8 European regions in high-technology patenting plus Bedfordshire. For the samples, there are significantly higher proportions of citation matches than control matches (except Noord–Brabant in 1995) indicating localization effects. Results that are significant at the 0.05 level or better are given in bold. These results indicate quite strongly that knowledge is localized at the regional level. In 1995 Iˆle-de-France shows by far the strongest Table 18.4 Regional variations in localization: tests in selected regions Number Citation Control matching t-Statistica ofcitations matching (%) (excl. self-cites) (%) 1990 1995 1990 1995 1990 1995 1990 1995 ˆIle-de-France 130 197 27.9 28.4 13.9 8.6 3.30(0.000) 6.05(0.000) Oberbayern 82 88 12.1 10.2 2.4 2.4 2.22(0.009) 1.51(0.037) Switzerland 73 81 17.8 28.3 9.5 6.1 1.51(0.046) 3.81(0.000) Lombardia 68 43 26.4 16.2 7.3 11.6 3.38(0.000) 0.70(0.242) Noord–Brabant 65 14 24.6 7.1 13.8 7.1 1.72(0.044) 0.00(0.500) Darmstadt 53 76 11.3 28.9 0.2 3.9 1.93(0.029) 3.95(0.000) Ko¨ln 38 47 10.5 8.5 2.6 0.0 1.35(0.091) 2.06(0.041) Bedfordshire 36 13 46.1 23.0 5.5 0.0 3.21(0.001) 1.89(0.042) Du¨sseldorf 28 33 21.4 18.1 3.5 9.0 2.42(0.011) 1.78(0.022) a Results significant at the 5% level of significance are in bold

8

The average citation lag for the 1990 (1995) cohort of originating patents is 4.45 (4.57) compared to 4.14 (4.51) for the corresponding control patents.

342

M.M. Fischer et al.

Table 18.5 Test for regional variations in localisation: results of t-test ˆIle-de-France Oberbayern Switzerland 1990 1995 1990 1995 1990 1995 ˆIle-de-France – – –4.51 –5.55 –3.44 –4.17 Oberbayern 4.51 5.55 – – –2.21 –2.53 Switzerland 3.44 4.17 2.21 2.53 – – Note: All figures are t-statistics for differences in regional localisation and significant at the 0.05% level

localization effect. The results for the German regions (Darmstadt, Du¨sseldorf and Oberbayern), Switzerland and Bedfordshire are also significant in 1990 and 1995.9 In Table 18.5 we test whether the degree of knowledge localization is significantly different across regions. In order to compare, we use the Top-3 regions in high-technology patenting, Iˆle-de-France, Oberbayern and Switzerland. The results show that knowledge spillovers are significantly more localized in the Iˆle-de-France than in any other region though the other two regions also show considerable localization.

18.5

Summary and Conclusions

Localization of knowledge spillovers is implicit in most theories of new economic growth, but rarely studied empirically. In this paper we have analysed patent citation data pertaining to high-technology firms in Europe to test the extent of localization of knowledge spillovers. As described in the previous Section, we compared the probability that citing patents are from the same location as the originating patent with the probability that control patents selected to match the citing patents in terms of timing and technology classification are from the same location as the originating patent. The results strongly support the hypothesis that spillovers are geographically localized. The proportion of citing patents that match the location of their originating patents is significantly greater than that of control patent location matches at both spatial levels: the country and the region level. The t-statistics, which tests the equality of the proportion of citing-original versus control-original location matches, are large, with p=0.000. It is also interesting to note that spillover localization is specific to certain regions and that the degree of localization is significantly different across regions. Overall the results support the conclusion that regional and national systems of innovation matter (Fischer 2001). This is a conclusion that has important policy implications. European regional cohesion appears to be at stake, especially – but not exclusively – because of the localized nature of knowledge flows. 9

An examination of the citation data with self-cites reveals that localisation may often be driven by self-citations.

18

Patents, Patent Citations and the Geography of Knowledge Spillovers in Europe

343

Annex See Tables 18.6 and 18.7

Table 18.6 List of regions used in the study Country Nuts Region code Austria AT11 Burgenland AT12 Niedero¨sterreich/ Wien AT 1 Ka¨rnten AT22 Steiermark AT31 Obero¨sterreich AT32 Salzburg AT33 Tirol AT34 Vorarlberg Belgium BE10 Re´gion BruxellesCapital BE21 Antwer pen BE22 Limburg (B) BE23 Oost-Vlaanderen BE24 Vlaams Brabant BE25 West-Vlaanderen BE31 Brabant Wallon BE32 Hainaut BE33 Lie`ge BE34 Luxembourg (B) BE35 Namur Bulgaria BG00 Bulgaria Czech Republic FR21 Champagne-Ardenne Germany DE11 Stuttgart DE12 Karlsruhe DE13 Freiburg DE14 Tu¨bingen DE21 Oberbayern DE22 Niederbayern DE23 Oberpfalz DE24 Oberfranken DE25 Mittelfranken DE26 Unterfranken DE27 Schwaben DE30 Berlin DE40 Brandenburg DE71 Darmstadt DE72 Gießen DE73 Kassel DE80 MecklenburgVorpommern

Country

Nuts code DEB1 DEB2

Koblenz Trier

DEB3 Dec-00 DED1 DED2 DED3 DEE1 DEE2

Rheinhessen -Pfalz Saarland Chemnitz Dresden Leipzig Dessau Halle

DEE3 DEF0 DEG0 Denmark DK00 Estland EE00 Finland FI13 FI14 FI15 FI16 FI17 France FR10 CZ00 Czech FR22 FR23 FR24 FR25 FR26 FR30 FR41 FR42 FR43 FR51 FR52 FR53 FR61 FR62 FR63 FR71 FR72

Region

Magdeburg Schleswig-Holst ./Hamburg Thu¨ringen Denmark Estland Ita¨-Suomi Va¨li-Suomi Pohjois-Suomi Uusimaa Etel-Suomi ˆIle-de-France Republic Picardie Haute-Normandie Centre Basse-Normandie Bourgogne Nord-Pas-de-Calais Lorraine Alsa ce Franche-Comte´ Pays de la Loire Bretagne Poitou-Charentes Aquitaine Midi-Pyre´ne´es Limousin Rhoˆne-Alpes Auvergne (continued)

344 Table 18.6 (continued) Country Nuts Region code DE91 Braunschweig DE92 Hannover DE93 Lu¨neburg/Bremen DE94 Weser-Ems DEA1 Du¨sseldorf DEA2 Ko¨l DEA3 Mu¨nster DEA4 Detmold DEA5 Arnsberg

M.M. Fischer et al.

Country

Nuts code FR81 FR82

Greece Hungary HU00 Ireland IE00 Ital IT11 IT12 IT13 IT20

Region Languedoc-Roussillon Provence-Coˆte d’Azur Greece Hungary Ireland Piemonte Valle d’Aosta Liguria Lombardia

Table 18.7 Assignment of patent classes to the high-technology sector at the four-digit ISIC-level ISIC Industry sector IPC patent category category 3522 Pharmaceuticals A61J, A61K, C07B, C07C, C07D, C07F, C07G, C07H, C07J, C07K, C12N, C12P, C12S 3825 Computers and office B41J, B41L, G06C, G06E, G06F, G06G, G06J, G06K, G06M equipment G11B, G11C 3832 Electronics– telecommunications G08C, G09B, H01C, H01L, H01P, H01Q, H03B, H03C, H03D, H03F, H03G, H03H, H03J, H03K, H03L, H04A, H04B, H04G, H04H, H04J, H04K, H04L, H04M, H04N, H04Q, H04R, H04S, H05K 3845 Aerospace B64B, B64C, B64D, B64F, B64G

References Agrawal A, Cockburn IM, McHale J (2003) Gone but not forgotten: labor flows, knowledge spillovers and enduring social capital. NBER Working Paper 9950, Cambridge MA Almeida P (1996) Knowledge sourcing by foreign multinationals: patent citation analysis in the U.S. semiconductor industry. Strategic Manage J 17:155–65 Almeida P, Kogut B (1999) Localisation of knowledge and the mobility of engineers in regional networks. Manage Sci 45(7):905–17 Anselin L, Varga A, Acs Z (1997) Local geographic spillovers between university research and high technology innovations. J Urban Econ 42(3):422–48 Borgatti SP, Everett MG, Freeman LC (1999) Ucinet for Windows: software for social network analysis, User’s Guide. Analytical Technologies, Harvard

18

Patents, Patent Citations and the Geography of Knowledge Spillovers in Europe

345

Fischer MM (2001) Innovation, knowledge creation and systems of innovation. Ann Reg Sci 35(2): 199–216 Griliches Z (1990) Patent statistics as economic indicators: a survey. J Econ Liter 28(4):1661–707 Hicks D, Breitzman T, Olivastro D, Hamilton K (2001) The changing composition of innovative activity in the USA – a portrait based on patent analysis. Res Pol 30(4):681–703 Jaffe AB (2000) The US patent system in transition policy innovation and the innovation process. Res Pol 29(5):531–57 Jaffe AB, Trajtenberg M (eds) (2002) Patents, citations & innovations. A window on the knowledge economy. MIT, Cambridge, MA Jaffe AB, Trajtenberg M, Henderson R (1993) Geographic localization of knowledge spillovers as evidenced by patent citations. Quart J Econ 108(3):577–98 Karlsson C, Manduchi A (2001) Knowledge spillovers in a spatial context – a critical review and assessment. In: Fischer MM, Fro¨hlich J (eds) Knowledge, Complexity and Innovation Systems. Springer, Heidelberg, pp 101–123 Maurseth PB, Verspagen B (2002) Knowledge spillovers in Europe: a patent citation analysis. Scand J Econ 104(4):531–45 Singh J (2003) Multinational firms and international knowledge diffusion: evidence using patent citation data. Working Paper, Harvard Business School Thompson P (2003) Patent citations and the geography of knowledge spillovers: what do patents examiners know? Manuscript, Florida International University Trajtenberg M (2001) Innovation in Israel 1968–1997: a comparative analysis using patent data. Res Pol 30(3):363–89 Verspagen B, Van Moergastel T, Slabbers M (1994) MERIT concordance table: IPC–ISIC, rev. 2nd edn. Maastricht Economic Research Institute on Innovation and Technology, University of Limburg, Maastricht Verspagen B, Schoenmakers W (2004) The spatial dimension of patenting by multinational firms in Europe. J Econ Geogr 4(1):23–42

Chapter 19

Co-authorship Networks in Development of Solar Cell Technology: International and Regional Knowledge Interaction Katarina Larsen

19.1

Introduction

This paper examines the development of new science-based technology in the research area of nanostructured solar cells development – a science-based technology with potential for advancing renewable energy technology. As for other research areas, the production of new scientific knowledge in this particular field is not evenly spread across all geographic regions. Rather, scientific knowledge production and science-based innovation activities take place in regional nodes that often are located in metropolitan areas with a strong academic research ability and competitive private research and development. Another character of scientific knowledge production undertaken at universities is that one node (or location) of knowledge production within a certain field is connected to other locations through joint research initiatives, collaboration on technical development and mobility of researchers. In areas where advances in science-based technology are published in scientific journals, this interaction and exchange of knowledge can be analyzed through studies of the researchers’ joint publications. These two aspects (concentration of scientific knowledge production and knowledge networks between locations) motivate a regional dimension in studies of science-based technology and innovation. The focus on the regional context also incorporates the notion of crossregional knowledge networks and mechanisms of knowledge transfer. The regional dimension is also in the core of studies in the area of geography of innovation, following the early work on geographically mediated knowledge spillovers (Jaffe 1989; Acs et al. 1991). Studies of knowledge networks have also examined the effects of knowledge spillovers in science-based technology fields (Owen-Smith and Powell 2004). This paper focuses on examining the mechanisms by which science-based technical knowledge is transferred and applies a regional lens

K. Larsen KTH - The Royal Institute of Technology, Sweden e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_19, # Springer‐Verlag Berlin Heidelberg 2009

347

348

K. Larsen

to measures of scientific output, impact and structure. This leads to the following three components of the introduction. First, discussing some central aspects in previous work in studies of knowledge spillovers, then drawing on experiences from studies of science-based knowledge networks, and finally, outlining the scope of the study.

19.1.1 Knowledge Spillovers Studies of knowledge spillovers have been undertaken at the state level in the US (Acs et al. 1994; Audretsch and Feldman 1996) providing some insights to the major locations of inventive activity. Studies of knowledge spillovers and innovation activity acknowledge the importance of the ‘‘locational context’’ for economic activity (Feldman 1999, p. 6). This locational context is described as ‘‘the geographic unit over which interaction and communication is facilitated’’ (ibid). When analysis is made at the state level, it should be noted that it has been questioned whether the state ‘‘qualifies as a functional unit’’ (Karlsson and Manduchi 2001, p. 114). These studies of knowledge spillovers and innovation activity also analyze sectoral distinctive features, industry life cycle and firm size (Anselin et al. 2000; Karlsson and Manduchi 2001). For overview of earlier developments of the field, also see (Feldman and Audretsch 1999; Acs et al. 1994). For the current study with a focus on scientific knowledge spillovers and knowledge networks (not on economic activity per se or analysis of patents as a measure of innovation activity) it is necessary to also acknowledge previous studies in the area of economics of innovation that have to do with mechanisms by which science-based technical knowledge is transferred. This question is at the heart of studies of knowledge spillovers. A review of empirical work on spillovers and agglomeration distinguishes among four categories of studies, including studies of geographic innovation production functions, studies of linkages between patent citations (paper trails of knowledge), mobility of skilled labor (ideas of people), and knowledge spillovers embodied in traded goods (Feldman 1999). There are examples of studies examining agglomeration effects using the knowledge production function framework (Varga 2000) concerned with why knowledge (or economic activity) does not spread evenly across geographic space. For getting a better understanding of mechanisms for knowledge transfer, the two categories of studies of paper trails of knowledge and studies of ideas in people is discussed in the context of the current study. A previous study using patent citation data for a study of knowledge spillovers (Jaffe et al. 1993) finds that knowledge spillovers are geographically localized, but also that the localization fades over time. The current study, uses publication data (and not patent data) to study a field where sciencebased innovation is mainly located at academic institutions and resulting in scientific publications. However, patenting activity occurs in the field since the research inherently is targeting the development of new devices and application of the new discoveries. This provides scope for future analysis of both corporate activity and patenting in the field, but currently is beyond the scope of this study. Instead this study uses co-authorship data to study knowledge transfer and the mechanisms

19

Co-authorship Networks in Development of Solar Cell Technology

349

behind them. One e.g. of related work in the area of knowledge spillovers is concerned with mechanisms including complementarities in skills and competence of industries and sectors (Feldman and Audretsch 1999, p. 411) and the fact that knowledge travels with people (Feldman 1999). For the current study, this includes aspects such as researcher mobility triggered by, among other things, international research funding and international scientific collaboration.

19.1.2 Science-Based Knowledge Networks For the area of nanostructured solar cells technology, the scientific papers reporting on advances in development of this solar cell technology are predominantly found in natural sciences including chemistry, material sciences etc. It has been recognized that while pair wise co-authorship is frequent in areas such as economics (Beckman 1994), it is common with more co-authors per paper in fields of natural sciences. For e.g., studies of biomedical research show an average of four authors per paper, to be exceeded by areas such as high-energy physics with an average of nine authors (Newman 2001, p. 405). These examples of networking activities of scientists (Andersson and Persson 1993) reveal variation between fields. This gives rise to the question of why researchers collaborate. What are the benefits and costs? To discuss this, it is not only important to first recognize that the extent of collaboration (how many you collaborate with) varies between different research fields, but also to identify the level of development or commercialization process you are located at (when you collaborate). This is dealt with in both studies of science, technology and innovation (STI) and in studies of economics of science and technology. Review of factors encouraging scientific collaboration reveal a broad range of attributes including both institutional factors (such as changing patterns of funding), visibility and recognition, and the desire to increase crossfertilization of ideas and techniques, as well factors of geographic proximity (Katz 1994, p. 31). The US experience of Research Joint Ventures provides some insight into benefits and costs of collaboration1 in the stage closer to commercialization which is at the end of the technical development process. The benefits include knowledge spillovers captured from other members, reduced research costs due to a reduction in duplicative research, faster commercialization since the fundamental research stage is shortened, and finally, the opportunity to form (in some cases) industry-wide competitive vision (Audretsch et al. 2002, p. 180). The costs are related to a lack of appropriability since the research results are shared among the participants and managerial tension (in some cases) as participants learn to trust each other and work together (Audretsch et al. 2002, p. 181). At a European level, the science policy system also constitutes a driver for increased research collaboration at regional and international level in the framework of the European Research Area (EC 2000). In the EU-context, different measures of distance can change the European map considerably, depending on whether it is measured by time of air travel (Beckman 1993) or collaboration distance measured by degrees of 1

For overview see Hagedoorn et al. 2000.

350

K. Larsen

connection or separation between co-authoring scientists (Newman 2001). This triggers the question of the extent to which the collaboration takes place within regions, countries or across national borders. The answer, for the area of nanostructured solar cells, is that collaboration takes place both within regions, countries and internationally. A more interesting question in the context of this study is what the patterns of inter-regional collaboration look like and what different types of collaboration that data on co-authorship describes. The analysis of these knowledge networks is another way of studying drivers and reasons for why researchers collaborate by interpretative analysis of existing co-authorship links.

19.1.3 Scope of This Study The emergence of new scientific fields with a perceived potential for innovation and associated commercialization has spurred researchers to study novel technologies. One e.g. of studies of the bioscience research field is focusing on the emergence and development of a radical new technology ‘‘with a strong science base and a great commercial potential’’ (Feldman 2001, p. 346). The area of nanoscience and nanotechnology has also received increasing attention both in science and technology studies in academia (Darby and Zucker 2003; Meyer 2000a) and in an international science policy context (OECD 2003; EC 2003). The development of nanotechnology is considered to be important for national and European competitiveness and also to have benefits to society at large. It is an integrated part of the strategy for the European Research Area (EU 2005) and US science policy (NNI 2005). The wide range of areas of application of nanotechnology includes innovations in ICT, biomedical, and environmental related science (Royal Society 2003). Areas of environmental application of nanotechnology are found in development of sustainable energy (fuel cells and solar energy) as well as in end-of pipe technologies (soil remediation and nano-sensors for pollution monitoring). Applications for cleaner technologies are also suggested (EC 2003, p. 9). In other words, there are high expectations on nanoscience and nanotechnology to be responding to environmental pressures facing the industry. Having said this, there are arguments for an analysis of nano-science and technology (S &T) applications at a less aggregated level. For e.g., studies of nanoscience and technology suggest limitations of ‘nano-multidisciplinarity’ and claim a stronger bias towards ‘‘monodisciplinary fields which are rather unrelated to each other’’ (Schummer 2004, p. 425). This motivates a careful selection of a field of nano S&T application that enables a meaningful platform technology for a special area of application. Further studies of the use of nanotechnology for environmental applications are suggested (Larsen 2003) and the current study is concerned with the development of solar energy technology utilizing nanoscience and technology. The aim of the paper is to increase the understanding of science output, structure and impact within the area of nanostructured dye-sensitized solar cells. Central to this specialized area of solar cell technology is a nanoparticle film

19

Co-authorship Networks in Development of Solar Cell Technology

351

or nanowire set that ‘‘provides a large surface area for adsorption of light harvesting molecules’’ (Law et al. 2005, p. 455). For overview of this type of solar cell technology, see Hagfeldt and Gratzel (1994). The field is characterized by developing new solar cell technology and applied devices, building on advances in complementary knowledge bases in areas of physical chemistry, electrochemistry and material sciences. The importance of science for advances in this application-oriented field makes it particularly suitable for analysis of knowledge transfer using bibliometrics. The scope of the study is to analyze knowledge networks between locations with a strong science base in the subfield of dye-sensitized solar cells utilizing nanoscience and technology, hereafter described as nanostructured solar cells. The locations are the home-regions of authors as described in the Science Citation Index by the affiliation address of the authors. The research questions are not only concerned with structural properties of science based knowledge networks within this field measured by co-authored papers, but also in the different types of research collaboration and interaction that gives rise to co-authored papers. The two research questions are: l

l

What characterizes the science-based knowledge networks, in terms of network structure and centrality, in development of novel solar cell technology based on nanoscience and technology? In this specialized field, what types of interaction between individual nodes does the analysis of co-authorship links describe?

The study applies bibliometric methods and social network analysis (SNA) to examine the structure of scientific collaboration. Bibliometrics is described as ‘‘a generic term for quantitative analyses of relevant characteristics of the contents of scientific and technological texts, mostly across a set of research publications’’ (Tijssen 1991, p. 27). The publication data used in the analysis is retrieved from Science Citation Index (SCI). It enables analysis of research output measured by a number of publications and research impact measured by citations, and analysis of relational data about co-authorships. Social network analysis (SNA) is based on ‘‘an assumption of the importance of relationships among interacting units’’ (Wasserman and Faust 1994, p. 4). Scientific collaboration, knowledge transfer and exchange of ideas are important for advances in science and technology and innovation processes. These relationships are in the paper measured by the prevalence and structure of co-authored papers. It is argued that co-authored papers indicate substantive research relationships through which tacit knowledge can be shared (Hicks and Katz 1997). All collaboration does not lead to co-authored papers and there are other outputs of collaboration than co-authored papers (Melin and Persson 1996, p. 365). Bearing this in mind, co-authored papers are considered as a crude indicator of interaction giving rise to sharing knowledge and information. Earlier work on knowledge networks in science based technology areas (such as biotechnology) stress that advantages of a central position can give access to critical information and knowledge bases of other actors but

352

K. Larsen

also that a high centrality ensures both sustaining old and initiating new R&D alliances (Powell et al. 1999). In this paper, quantitative analyses of co-authorship links of the eight most prolific locations, in terms of research output within this specialized field, was further analyzed in terms of research output and structure. These include Beijing (China), Lausanne (Switzerland), Uppsala (Sweden), Baltimore (US), London (UK), Tokyo (Japan), Ibaraki (Japan), and Osaka (Japan). The first four locations are further analyzed using citations as a measure of research impact. In addition to quantitative analyses of individual locations, provided by the data from SCI, qualitatively analyses of the Swedish location in Uppsala was carried out to get a better understanding of what types of interaction between individual network nodes the co-authorship links describe. The paper is organized as follows. Section 19.2 provides a background to nanoscience and technology and definitions applied in the area. Section 19.3 outlines the research approach and data used in the paper. Section 19.4 presents the results and analyses the character of the knowledge networks and national characteristics of collaboration. The paper ends with conclusions of the study and discussion in Sect. 19.5.

19.2

Nanoscience and Technology in Solar Cell Development

The transformation of the energy system towards new renewable sources of energy, such as solar energy, is a central feature of national energy policy in several countries (Jacobsson et al. 2001; Silveira 2001). Energy policies that target this type of transformation also raise questions about the adoption and diffusion of new technology such as the application of nanoscience in development of solar cell technology. The potential of niche markets and characteristics of different types of solar cells, including nanostructured solar cells, has also been examined (STEM 2004). Nanotechnology, as the study of the very small (OECD 2003), has received increasing attention and research budgets in many industrialized countries. The body of literature analyzing the field of nanoscience and nanotechnology can be divided into three types of publications in addition to those reporting advances and new findings in natural sciences and engineering (material science, chemistry and physics journals). One branch of publications is concerned with the phenomenon of the emerging new science and technology area compared to the growth of biotechnology (Darby and Zucker 2003; Braun et al. 1997). The second branch is concerned with relations between science and technology in emerging nanocommunities (Meyer 2000a; Meyer and Persson 1998) and builds on work on patent citations and bibliometrics (Pavitt 1998; Meyer 2000b). The third branch is that of policy studies of nanoscience and technology raising questions about the impact of new applications of technology (Royal Society 2003). It reveals a multitude of applications and expectations on potential applications for nano-scale science and technology. Science-policy seminars are arranged to reach a wider audience (IVA 2004). Policy studies reviewing social and economic impact of nanotechnology

19

Co-authorship Networks in Development of Solar Cell Technology

353

focus on defining the conception of nanotechnology, and also outline the areas of application (ESRC 2003; Arnall 2003). The US national nano initiative (NNI 2005) considers something to be nanotechnology if it involves research and technology development at the atomic, molecular or macromolecular levels, in the length scale of approximately 1–100 nm. It implies creating and using structures, devices and systems that have novel properties and functions because of their small or intermediate size and an ability to control or manipulate on the atomic scale. A similar approach is adopted in science studies by Darby and Zucker (2003, p. 11) describing nanoscience and technology as ‘‘research and technology development at the atom, molecular and macromolecular levels, in the length scale of approximately 1–100 nm range, to provide a fundamental understanding of phenomena and materials at the nanoscale and to create and use structures, devices and systems that have novel properties and functions because of their small or intermediate size’’. The definitions applied generally fall into two categories. The one applied by the US nano initiative above, is rather technical and narrows it down to encompassing research concerned with ‘‘the novel properties and functions because of their small or intermediate size’’. The second category is wider in scope and states that nanotechnology ‘‘refers to a spectrum of new technologies that seek to manipulate atoms and molecules to create new products or all research activities undertaken at the nanometric scale’’ thus including much of research in the field of biotechnology and macromolecular chemistry undertaken during the last two decades (OECD 2003). Nanotechnology can also be characterized by distinguishing between the fabrication processes of top-down and bottom-up (Arnall 2003). Top-down is based on miniaturization while bottom-up also called molecular nanotechnology (MNT), that applies to the creation of organic and inorganic structures. The growth of the term ‘nano’ in the title of scientific publications has had an exponential doubling time of 1.6 years in the period 1986–1995 (Braun et al. 1997). The growth of the field of nanostructured solar cells is quite recent, due to some important findings during the 1990s. For this reason it is of interest to find out more about the development of nanostructured solar cells that can be traced in scientific publications in more recent years. To get an overview of the important countries in nanostructured solar cells in comparison to other nano-related2 publications during a more recent period, publication data was retrieved from the Science Citation Index (SCI) for the years 1990–2003. The elaboration of SCI-data of publications 2

Due to different set of definitions of the scope of nanoscience and technology (nano S&T), there are different approaches of defining publication output of nano-related science. One approach is to use bibliometric methods based on a set of keywords for limiting and defining the scope of the nano S&T field. For e.g., a study of nano communities used keywords for excluding papers (such as papers on nanometer and nanosecond) and other keywords for including relevant research areas (such as scanning probe microscopy) for the field of nano S&T (Calero et al. 2005). In Table 19.1 the main purpose is to get an overview of countries’ contribution to publication output in nanorelated science. Other ways to examine the nano S&T field are based on describing co-authorship networks at the level of universities and research institutes in the field of nano S&T (Heinze 2006) or studying budget allocations targeting nano-research (OECD 2003).

354

K. Larsen

Table 19.1 National shares of global publication output of nano-publications and nano-solar publications, total count of English scientific papers (1990–2003) identified in the Science Citation Index Nano-publications (1990–2003) Nano-solar publications (1990–2003) No. of papers Rank Percent No. of papers Rank Percent USA 13,944 [1] 27.7 49 [2] 15.4 Japan 7,146 [2] 14.1 80 [1] 25.1 China 6,858 [3] 13.1 13 [7] 4.1 Germany 5,496 [4] 10.5 22 [6] 6.9 France 3,912 [5] 7.7 9 [8] 2.8 England 2,329 [6] 4.7 35 [4] 11 Italy 1,682 [7] 3.3 8 [9] 2.5 Switzerland 1,053 [8] 2.1 43 [3] 13.5 Canada 1,033 [9] 2.1 6 [10] 1.9 The Netherlands 775 [10] 1.6 13 [7] 4.1 Sweden 727 [11] 1.3 33 [5] 10.3 Australia 591 [12] 1.2 4 [11] 1.3 Rest of the world 4,935 10.6 4 1.1 Total 50,481 100 319 100 Note: Ranking of countries in parenthesis is based on their share of publications in the defined field

provides an overview of the national publication stocks, shown in Table 19.1. The national share of papers with the string ‘nano’ in the title3 is showing that USA accounts for about 28% of the publications, followed by Japan and China. The national share of nano-solar papers identified specifying the sub-field of dyesensitized solar cells4 shows Japan’s leading position, with 25% of the publications, followed by USA and Switzerland. The countries examined in Table 19.1 included those where the concentration of nanoscale science and engineering articles is especially high (Australia, England, France, Germany, Japan, Switzerland, and USA) according to Darby and Zucker (2003). The search was also made for Italy, Canada, the Netherlands, Sweden and China to include countries of potential interest for examining the nano sub-field of development of dye-sensitized solar cells. The keywords used to identify the set of publications for nano-solar publications were selected in a dialogue with experts in the field.

3

In Table 19.1, the left column Nano-publications, shows results from selecting papers that include the string ‘nano*’ in the title. The ranking of the countries was checked by introducing keywords for excluding papers outside the core of the field. This resulted in a reduction from 50481 to 41611 papers, but the ranking of the countries was not altered. The keywords used to exclude papers outside the core of the field were: Title=(nano* NOT (nano-met* OR nano2 OR nano3 OR nano4 OR nano5 OR nano-secon* OR nano secon* OR nano-met* scale* OR nano-meterscale* OR nano meter length OR nanometer-scale OR nanogram* OR nano-molar* OR nano-plankton* OR nanoliter*)). 4 Search in English scientific papers, topic search (1990–2003) using key words: solar (OR photovoltaic), cell*, nano*, dye-sensitized (OR dye-sensitized). Search resulted in 320 publications where one was excluded since it was published in 2004.

19

Co-authorship Networks in Development of Solar Cell Technology

355

The stock of nano-papers comprised of 50,481 publications and 319 nano-solar publications in total, identified by a set of key words. The use of key words is further discussed in the section on data and methods. The ranking is based on total counting and the percent is calculated by dividing country’s research output based on number of publications within a field with the total number of publications identified for the field. For e.g., Sweden has 33 publications in the field of nanosolar publications. This gives 10.3% of total publication output within the special field (33/319=0.103) and a rank of five. The results show that some countries, such as USA and Japan, are highly ranked for both nano-publications and nano-solar publications. Other countries are higher ranked in the sub-field of nano-solar cells. This is the case for Switzerland, England and Sweden. The ranking are based on a total count of publications. The national share of publications, shown in Table 19.1, has been retrieved from the Science Citation Index. It reports on national share of publications where the country is involved. In other words, the use of total counting results in double counting in the case of internationally co-authored papers. The problem has been recognized also for international reporting of publication output (Gauffriau and Olesen Larsen 2005) as well as for models of knowledge networks (Beckman 1994). Therefore, it is important to recognize that the reporting of publications from ‘‘rest of world’’ is a proxy for the number of publications that did not include any of the listed countries. A limit of this proxy relates to the total counting method by a certain double counting of internationally co-authored papers. Studies examining the impact of different counting methods stress that differences in ranking may occur due to use of counting methods of total counting and fractional counting where fractional counting (at a country level) divides the publication by the number of country addresses contributing to the publication, according to the addresses in the publication (Gauffriau and Olesen Larsen 2005). This also applies to citation count as a measure of research impact. Of the 319 papers, one out of five papers approximately was internationally co-authored. The international character of this research field is recognized for the current study, and further analysis of citations will also take into account different outcomes in citations due to total or fractional counting. The use of total count in Table 19.1 fulfils the purpose of providing an orientation of the countries that are highly rated in the specialized area of nanostructured solar cells. The five highest rated countries are Japan, USA, Switzerland, England and Sweden. These countries are all included in the further analysis of regional nodes of knowledge production in development of nanostructured solar cells.

19.3

Data and Research Methods

The data used for the study was retrieved from the Science Citation Index (SCI). A combined approach based on a set of key words and identification of a seminal (and highly cited) paper in the field was used to select publications for further analysis.

356

K. Larsen

Number of papers 400 300 200 100 0 Papers citing

1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2

12

41

50

62

85

135

138

137

193

220

252

283

359

Fig. 19.1 Number of papers that cite the highly cited paper (O’Regan and Gratzel 1991) from 1991 to 2004

The key words used to identify a highly cited paper were the same as those used to identify the highest ranked countries in the area of nanostructured solar cells, shown earlier in Table 19.1. This resulted in a dataset of 1,930 publications with linked addresses in the Science Citation Index. This means that all these publications cite the seminal paper from the year 1991. The validity of this choice of approach was also confirmed by researchers in the specialized field of nanostructured solar cells who considered it to be a reasonable way to identify publications in this field. This special type of solar cell technology is also referred to as Gra¨tzel (or Gratzel) cells (Jake´lius 2001, p. 286; STEM 2004) because of advances by this scientist and collaborators in research and in technical applications during the early 1990s. The publications analyzed range between years 1991 and 2005 and a steady increase of citations to this seminal paper is exhibited, as shown in Fig. 19.1. The analysis of the research impact of the highly cited paper shows a steady growth of citations made by other papers even 12 years after the paper is published. Studies of highly cited papers’ citation life cycles characterize the highly cited papers according to how large the share is of the citations that are received during the first six years relative to the share of citations received during the last six years of a 12-year citation window (Aksnes 2003, p. 165). According to this, the O’Regan and Gratzel (1991) paper would be characterized as a medium rise-no decline paper. In addition to this, Fig. 19.1 shows that the largest number of citations were received during 2004, which was the most recent year included. This gives an indication of that there is actually still a growth in citations. But does this indicate a growing importance of this particular field or does it mean that the research field is growing with more active research groups in the research area? The interviews with researchers5 in the field suggest the latter explanation to the growth in citations.

5

Pilot interview in 2003/04, followed by interview with Anders Hagfeldt and Gerrit Boschloo 20 June 2005.

19

Co-authorship Networks in Development of Solar Cell Technology

357

The key word approach has been applied in other studies in the area of nanoscience and technology, including studies of nanoscience and technology. In the current study, publications within the area studied were initially selected by key words limiting the sample to a set of publications relevant to the area. The key words used6 resulted in selection of 319 publications that were all considered as relevant for the research field. This is consistent with recent bibliometric studies of nanoscience and technology which suggest limitations of ‘nano-multidisciplinarity’ and claim a stronger bias towards ‘‘monodisciplinary fields which are rather unrelated to each other’’ (Schummer 2004, p. 425). So, rather than studying several monodisciplinary nano-fields as one unit of analysis, the approach of this paper is to study application of nanoscience and technology in development of solar cells, thereby looking at a rather specific application of nanotechnology. The identification of papers using key words has been employed in other studies (Hicks et al. 2004) and is also applied in bibliometric studies in the energy area exploring fuel cell technologies (Godø et al. 2003). Other bibliometric studies in the energy area state that there is a blend of applied and basic research in development of solar cells (Tijssen 1991) which makes it a relevant area of work for analyzing science-based knowledge networks using co-authorship data. For the current study, the key word approach generated a set of publications within the nano-solar science field (319 publications) applying a set of keywords that eliminated non-relevant publications in areas, which deal with other aspects of solar radiation and other types of cells. The approach chosen for the analysis was to select a set of publications that cite a seminal paper within the specialized field. The key-word approach generated a starting-point with a highly cited paper7 in 1991 that in turn generated the set of publications for further analysis of coauthorship links. The validity of selecting this particular paper to identify scientific papers was discussed with researchers in the field and also confirmed in the literature. The O’Regan and Gratzel (1991) paper initiated a decade of research into the electrical transport physics of nanoparticle anodes (Law et al. 2005, p. 455) and the approach was therefore considered as a reasonable way to select papers in this field. Co-authorship indicates links within the scientific community (Hicks et al. 2004, p. 83) and is a useful relational measure that can be used to characterize publication patterns. The use of co-authorship in other bibliometric studies show that a high share (about 40%) of Swedish papers is internationally co-authored, thus indicating that co-authorship involves interaction over relatively long distances (Danell and

6

Key words used: solar (or photovoltaic), cell-, nano-, dye-sensitized (or dye-sensitized). O’Regan and Gratzel (1991)‘‘A low-cost, high efficiency solar-cell based on dye-sensitized colloidal TiO2 films’’, Nature, 353(6346), pp 737-740 (Oct 24, 1991). When this paper was retrieved from the Science Citation Index (27 April 2005), it had been cited by 2,057 papers.

7

358

K. Larsen

Persson 2003, p. 214). Earlier studies of co-authored papers show that international co-authorship, in general, result in publications with higher citation rates than purely domestic papers but without correlation between strength of co-authorship and relative citation eminence (Gla¨nzel and Schubert 2001, p. 199). Another aspect of effect of publications is the impact on patenting activity. It is suggested that in order to make more reliable statements about a connection between patenting and publishing activities of countries ‘‘one needs to go to the patents themselves and look at the extent to which they cite domestic scientific papers’’ (Meyer and Persson 1998). Although it could be of value to reveal national and international citation links between publications and patents, the study focuses on co-authorship links so patent-paper citation links are beyond the scope of this study. However, it can be noted that several of the research groups in the field are active in both publishing papers and patenting and that avenues for future of research could include an examination of citation networks between papers and patents. In the paper, co-authorships in 1,930 publications are used to analyze characteristics of science-based networks. For this purpose, two co-authorship matrices were created. The first matrix describes the co-authorship relations between the eight nodes. The second matrix also includes the relations between the eights nodes and their co-authors resulting in an extended matrix. The co-authorship matrix lists the locations in the first column and also repeats the list of countries (in the same order) in the first row of the matrix. The matrix is used for relational data and each position in the matrix describes if there is a connection (here by a co-authored paper) or not. To illustrate this, the matrix including only the relations between the eight nodes is shown in Fig. 19.2. The extended matrix (also including the relations between the eights nodes and their co-authors) results in a matrix of 140–140 cells (including names of locations in first row and first column). The matrix is used for generating a

London

Baltimore

Tokyo

Ibaraki

Uppsala

Uppsala Lausanne 1

1

0

0

Beijing Osaka 1

0

0

0

Lausanne

1

1

1

1

0

1

0

1

London

0

1

1

0

0

0

0

0

Baltimore

0

1

0

1

0

0

0

0

Beijing

1

0

0

0

1

0

1

0

Osaka

0

1

0

0

0

1

1

1

Tokyo

0

0

0

0

1

1

1

1

Ibaraki

0

1

0

0

0

1

1

1

Total

3

6

2

2

3

4

4

4

Fig. 19.2 Co-authorship matrix for the relations between the eight nodes Uppsala, Lausanne, London, Baltimore, Beijing, Osaka, Tokyo, and Ibaraki. A relation between two locations (established via a co-authored paper) is indicated with one (1) in the matrix. If no co-authorship is identified in the dataset, this is indicated with a zero (0) in the matrix. For example, in this data, a co-authorship relation exists between London and Lausanne, but not between London and Baltimore

19

Co-authorship Networks in Development of Solar Cell Technology

359

network of co-authorship relations in the area of nanostructured solar cells8 and characterization of knowledge networks using social network analysis (SNA). A co-authorship matrix at a country level has earlier been applied to the area of nanotechnology to identify patterns of cooperation among 23 different countries (Meyer and Persson 1998). The use of the home-regions at a sub-national level results in an expanded matrix and allows for identifying both national and international collaboration. The method of analysis measures links between two (or more) author addresses that are listed in the Science Citation Index for each publication. The co-author networks are considered to be relevant as a measure of science-based knowledge networks since the development of this type of solar cell technology is highly dependent on advances in science. Consequently, scientific publication is an important activity for advances within the field. The approach used is to classify the publications based on the address listed in the SCI. The use of addresses requires some careful thought about what it is that these addresses are to be used for. The interest for the current study is to use the home regions of the authors. The choice of home regions of the authors reflects the ambition to analyze networks of locations for research production. In this context, the fact that some authors list several addresses and that international research mobility also may lead to change in address for the same author is considered to occur within the science based knowledge network. Although this implies some international differences, also with regard to the way in which addresses and postal regional codes are organized, it serves the purpose of the study to increase the understanding of the character of science-based knowledge networks within the area of nanostructured solar cell technology.

19.4

Results

19.4.4 Network Structure The results section of the paper draws on data on publications and co-authorships. In contrast to other nano-related studies (Schummer 2004; Meyer 2000c; Meyer and Persson 1998), the principal contribution of this paper is to analyze the character of science based knowledge networks within a specialized sub-discipline with nanoscience applications, rather than for nanoscience as a more general field of research and technology. This targets an improved understanding of network characteristics and international and national science based knowledge networks based on co-authorship.

8

Using UCINET on data downloaded from SCI (27 April 2005) of papers citing the seminal paper (O’Regan and Gratzel 1991) resulted in 2,057 citing papers. In this dataset, author affiliation addresses were identified for 1930 publications.

360

K. Larsen

IBARAKI

OSAKA TOKYO

LONDON BEIJING BALTIMORE UPPSALA

LAUSANNE

Fig. 19.3 Co-authorship network figure created using UCINET. Network hubs: Ibaraki, Tokyo, Beijing, Uppsala, Lausanne, Baltimore, London, and Osaka (black squares), multiple-linked nodes (grey diamonds), double-linked nodes (triangles), and single-linked nodes (black circle) Source: Larsen (2008)

The science based co-authorship network is shown in Fig. 19.3, and includes locations active in publishing (co-authored) papers. The co-authorship network includes different types of nodes. Co-authorship links of the eight most prolific locations, in terms of research output within this specialized field, were further analyzed in terms of research output and structure. Social network analysis (SNA) was used to explore structural measures, such as network degree centrality (Wasserman and Faust 1994, p. 180). These are Beijing (China), Lausanne (Switzerland), Uppsala (Sweden), Baltimore (US), London (UK), and three locations in Japan including Tokyo, Ibaraki and Osaka. These eight locations, share many co-authorship links among themselves, but also with other locations. These critical nodes are here referred to as networks hubs. Hence, the eight network hubs are considered to be important to the connectivity of the network (Wasserman and Faust 1994, p. 112) also influencing the knowledge transfer within the network. The removal of these eight nodes would reduce the connectivity of the network as whole. Nodeconnectivity and line-connectivity measure other aspects of connectivity of the network. These measures are used to describe the minimum number of nodes and lines that need to be removed to make the network disconnected (Wasserman and Faust 1994, p. 115). The eight network hubs with high research output within the field studied are positioned in a circle in Fig. 19.3. The links between these and other nodes in the network illustrate co-authorship links used as a measure of research collaboration.

19

Co-authorship Networks in Development of Solar Cell Technology

361

The nodes (i.e. locations based on addresses in SCI) are illustrated as points that are connected to other points in the network. The lines between the nodes are coauthorship links, showing if there are co-authored papers. In addition to the network hubs, there are three other types of nodes in the co-authorship network. These nodes are not identified based on their high level of publication output as the eight network hubs were. Instead they are identified using measures of structural similarity of the individual nodes. Structural similarity is a less strict version of structural equivalence, enabling a characterization of nodes based on how many and what types of links a node is connected to. In earlier work on structural equivalence, two actors are described to be structurally equivalent if they have identical ties to and from all other actors in the network (Wasserman and Faust 1994, p. 356–357). Since the network, in Fig. 19.3, includes the eight network hubs’ co-author links, but excludes the nodes that are not co-authoring with either of the eight network hubs, nodes are considered to be structurally similar based on whether they co-author with one network hub or with several of the network hubs. A related approach is that of ‘‘high quality’’ links applied by Powell et al. (2005) to characterize the nodes that have few links but these links are towards well-connected groups. Measures of publication output define the eight network hubs, while the concept of structural similarity is used to characterize three more types of nodes in the coauthorship network. These are, firstly, the single-linked nodes that are locations only co-authoring with one of the eight network hubs, outlined as small black points in the periphery of the network in Fig. 19.3. The second type of node is doublelinked, illustrated by blue nodes with links to two of the network hubs. The third type, illustrated by grey diamond shaped nodes in the centre of the network, is multiple-linked with more than two links to the network hubs. The approach used focuses on the existence of links, not the intensity of collaboration. Therefore the single-linked nodes, as defined here, can have several links to the same node. Another limitation is that only the co-authorship links involving one of the network hubs are included in the analysis. The focus of attention is thereby on the structural properties of the network to do with knowledge diffusion involving central locations as a complement to other studies focusing on the intensity in collaboration for the distribution of knowledge. The different nodes can also be grouped together, or clustered, based on their structural similarity. Figure 19.4 shows this clustering of four of the eight network hubs. The figure should be read from left to right showing that Lausanne shares many nodes jointly with Baltimore. These two (Lausanne and Baltimore) in turn are structurally similar to London and Uppsala. Thereafter Petten in the Netherlands is linked to the group of nodes. In the network in Fig. 19.3, the node representing Petten is the triple-linked node (with links to Lausanne, Uppsala and London) located closest to Lausanne (grey diamond shape) that has co-author links to Lausanne, London and Uppsala. This tree structure will (if extended to the right) include all nodes in the network based on structural similarity in the relational data provided in the co-authorship analysis. Moreover, the clustering based on structural similarity will reveal some

362

K. Larsen

Uppsala

London

Baltimore

Lausanne

Petten

Fig. 19.4 Clusters of locations based on measures of structural similarity using co-authorship data

Table 19.2 Share of co-author links and publication output for eight network hubs in the area of nanostructured solar cells (percent) Lausanne (Switzerland) Uppsala (Sweden) Osaka (Japan) London (UK) Beijing (China) Ibaraki (Japan) Baltimore (USA) Tokyo (Japan)

Co-author links (percent) 14.3 6.9 6.1 5.8 5.3 5 4.8 4.5

Publication Output (percent) 4.1 4.1 3.5 1.7 5.9 2.2 1.5 2.7

different branches of the tree structure. In the current study, three branches are apparent. In addition to the European-USA branch, including locations outlined in Fig. 19.4, the network hub locations in Japan (Tokyo, Osaka and Ibaraki) constitute one distinct branch, as well as the more isolated Beijing branch. These patterns can also be distinguished in the network structure in Fig. 19.3, showing the complementarities between different ways to visualizing the co-authorship network. A set of measures of science output and structure was used in order to illustrate the position of the network hubs with respect to research output and centrality. These are measures of research output (share of publications within the selected field) and science structure (share of co-author links). These measures are calculated for the eight network hubs (Lausanne, Uppsala, Osaka, London, Ibaraki, Beijing, Tokyo and Baltimore), as shown in Table 19.2. The results in Table 19.2 show that Lausanne has a higher share of co-author links than the other network hubs. Lausanne accounted for about 14% of all co-authorships

19

Co-authorship Networks in Development of Solar Cell Technology

363

Share of publication output 7% BEIJING

6% 5%

LAUSANNE

4% 3% 2% 1% 0% 0%

2%

4%

6% 8% 10% Share of co-author links

12%

14%

16%

Fig. 19.5 Measures of research output (share of publications) and research centrality (share of co-author links) within the field of nanostructured solar cell development

links. The other measure, publication output, is calculated by relating the number of publications where, for e.g., Beijing is among the author addresses, to the total list of publications output from the full list of author addresses. This measure shows in Table 19.2 and Fig. 19.5 that Beijing exhibits a high level of publication output (6%) but not as high share of co-author links compared to Lausanne. To conclude, the two concepts of network hubs and structural similarity are introduced to address research questions about properties of networks and characteristics of collaboration patterns of individual actors in the networks. Analysis of knowledge networks gives rise to many different questions. Are all actors connected in one network or are there several isolated networks? Are there differences between how individual nodes (regions) interact with other nodes in the co-author network? The network characteristics examined by calculating the network degree centrality for the networks indicate a lower value (network degree centrality 0.38) for the network including all nodes, shown in Fig. 19.3, compared to only including the interactions between the eight network hubs (network degree centrality 0.48). The network degree centrality can range between zero and one.9 The network degree centrality reaches its highest value 1 when one node is connected to all other nodes, and the other nodes only interact with this one (Wasserman and Faust 1994, p. 180). The higher centrality value indicates that the most central actor (Lausanne) has a relatively more central position in the eight-node network compared to the larger network including all other nodes as well. An interpretation of this is that there are a considerable number of the co-authored papers between nodes not necessarily involving the most central node in the network. The analysis of the co-authorship network also revealed four categories of collaboration patterns identified based on 9

The network degree centrality quantifies the range of variability of the individual actor degree centrality indices. Calculated using UCINET, for further information on equation, see Wasserman and Faust 1994, p.180.

364

K. Larsen

measures of centrality, science output and structural similarity. The categories are labeled as Network hubs (the eight most prolific locations, in terms of research publications within the field studied), Multiple linked nodes (structurally similar nodes that co-author with two or more network hubs), Double linked nodes, and Single linked nodes (structurally similar nodes that co-author with only one of the network hubs). The following section examines these categories by focusing on one network hub (Uppsala) and its links to other network hubs as well as to other nodes in the network.

19.4.5 Co-authorship and Research Collaboration The research question on what types of interaction between individual nodes the analysis of co-authorship links describe, was analyzed using bibliometric data of co-authorship in combination with interviews with researchers in the field with the home region of Uppsala in Sweden. The case of Uppsala provides an insight into the collaboration patterns of a location showing early collaboration with strong nodes of competence in the field, such as Lausanne, and at the same time forges links with new sites of knowledge production in nanostructured solar cells. The combination of the two types of interaction is of particular interest to get a better understanding of the reasons for scientists to collaborate as well as facilitate mechanisms for exchange of skills and knowledge. The area studied can be characterized as science based technology development, where measurement techniques for determining the performance of solar cells include only one type of interaction that the analysis of co-authorship links describes. Researcher mobility supported by national or international funding is another type of interaction that gives rise to co-authorship. This was shown by the case of co-authorship links of Uppsala showing several co-authorship links with distant locations. The links between Uppsala and Peru, Ethiopia, Costa Rica and Kenya came about, according to the researchers interviewed from the Uppsala research group,10 as a result of the mobility programme namely International Science Program (ISP) funded by the Swedish Development Agency (SIDA) to promote mobility of researchers in developing countries. This can also be described as policy targeting convergence in the network as opposed to excellence-oriented policy only including central actors (David and Keely 2003). Also EU-programmes were mentioned as resulting in, among other things, co-authored papers. In an earlier description of the research on nanostructured solar cells it is stated that the group located in Uppsala has a leading international position and has its major partners in the EU (Jake´lius 2001, p. 286). The finding of the current study, based on co-authorship data, provides detailed knowledge about both international position as well as the different types of research collaboration within Sweden as well as outside the EU. 10

Interview with Anders Hagfeldt and Gerrit Boschloo 20 June 2005.

19

Co-authorship Networks in Development of Solar Cell Technology

365

Fig. 19.6 Journal co-citation network in nanostructured solar cell research Note: Colors indicate when co-citation most recently occurred (blue, 1991–1994; green, 1995– 1999; yellow, 2000–2001; orange, 2002–2005). Network visualization with CiteSpace shows abbreviations of scientific journal names Source: Larsen (2008)

In addition to bibliometric data on co-authorship links, illustrations of cocitation networks11 were used to visualize the set of journals in which the researchers within this particular field publish. The method of co-citation (of documents) is defined as the frequency with which two documents are cited together (Small 1973). Co-citation can be used to analyze emerging trends in a knowledge domain and the use of software, such as CiteSpace, allows the user to take a time series of snapshots of a domain and then merge the different time periods (Chen, 2005). The usefulness of co-citation in the context of this paper is primarily for illustrative purposes complementing the information from the researchers about the relevance of areas such as materials science and chemistry for advances in the development of nanostructured solar cells. The co-citation network at journal level, shown in Fig. 19.6, confirms the emphasis on chemistry and also that the field comprises publication in nano-journals, such as the journal Nano Letters. However, the publication in nano-journals is limited to a few percent of the total number journals. Co-citation can also be made at author and publication levels and be used to identify authors with a strong influence of the research field and reveal individual publications relevant to the field. Naturally, the seminal paper that was used to select the publications analyzed is among these documents, but other important papers can also be identified. Co-citation networks have also been used to identify knowledge diffusion between patents and scientific papers (Chen and Hicks 2004). The journal cocitation links, in Fig. 19.6, describes that two journals are cited together by papers within the field and show a strong emphasis on chemistry science for the field.

11

Using CiteSpace, figure provided by Ulf Sandstro¨m, is based on publication data downloaded from the papers citing the seminal paper (O’Regan and Gratzel 1991).

366

K. Larsen 25 20 15 10 5 0 1990 1992 1994 1996 1998 2000 2002 2004 2006 Beijing

Dublin

Lausanne

Fig. 19.7 Publication output (number of papers) for Beijing, Dublin, and Lausanne over time, based on first author address in the Science Citation Index (SCI)

The co-citation network also show at what point in time the co-citation occurred, illustrated by different colors of the lines between the journals in Fig. 19.6. This is one type of dynamics of the field. Another dynamic over time is the changes in research output from different locations measured by number of publications. The researchers interviewed were asked to comment on the growth in publication output for some of the eight locations. The increase in publication output by Beijing stands out in particular, see Fig. 19.7. The research output measured by papers from different locations exhibit different trends, as shown in Fig. 19.7. These shifts in levels of output may be due to researcher mobility, changes in sources of funds where other areas are given priority, increased research productivity or a growth of the area with more research groups entering the field. In addition to research output, measured by publications, it is of interest to examine if a high citation level also accompanies an increase in publication output. The citation mean was calculated for the four locations of Baltimore, Beijing, Lausanne and Uppsala, shown in Table 19.3. The results of comparison between citation mean of the fractional and total counting show that the citation levels dropped with around 15–30% for the locations analyzed (using data 1991–2004) with the highest reduction for Beijing. For the shorter time period Lausanne showed a drop of 35% when going from total to fractional counting. This is due to eliminating early (and highly cited) publications before 1996 by authors from Lausanne. The fractional counting was based on different locations, not on the basis of countries, since the science field has a strong regional concentration. The ANOVA analysis was made for fractional count of citations to determine if there is a difference in mean research impact (measured by citations) between the four locations. The results of the analysis are found in Table 19.4 in Appendix and confirm the difference in mean level of citations received by papers from Beijing compared to levels of citations received by the other three locations. Further on, the analysis shows that there is a statistically significant difference in the mean level of citations between Uppsala and Lausanne, but not between Lausanne and Baltimore. Further analysis with a smaller citation window could provide more insight into the interpretation of the importance if the size of the citation window for the level of citations. Altogether, the results here indicate that the publication output of Beijing is increasing, but is less cited than

19

Co-authorship Networks in Development of Solar Cell Technology

367

Table 19.3 Citations made to publications with authors from Baltimore, Beijing, Lausanne, and Uppsala for two time periods Citation mean Citation mean Reduction by going total counting fractional counting from total counting to fractional counting (percent) 1991–2004 1996–2004 1991–2004 1996–2004 1991–2004 1996–2004 Baltimore 34 24 26 20 22 15 Beijing 6 6 4 4 30 30 Lausanne 68 42 55 27 19 35 Uppsala 21 17 18 14 14 16 Source: Larsen (2008)

papers from Lausanne, Uppsala and Baltimore. Studies with a special focus on China (Leydesdorff and Zhou 2005) as well as Japan in this area could provide further knowledge about importance of factors such as visibility, language, and character of the science system and co-authorship networks that are beyond the scope of the current study. Instead the further analysis focuses on examining the importance of size of the citation window and other factors, such as in which journal the paper is published and the number of authors of the scientific publication. To do this, the current study also included regression analyses in order explore the significance of factors such as citation window and journal impact factor (JIF) on the level of citations. For the individual nodes of Lausanne, Uppsala and Baltimore, results indicate the importance of both a large citation window and high JIF for citation eminence. The results are reported in Table 19.5. The findings show that for the individual nodes of Lausanne, Uppsala and Baltimore, both the factors Citation window and JIF are significant for the level of citations (using fractional count). Other factors examined (number of co-authors) showed no significance using fractional count of citations. The interpretation is that it does matter when the paper is published and in which journal it was published for the level of citations it receives, but not so much if you wrote it together with many coauthors when applying fractional count of citations. Based on the interviews, one explanation relates to the type of papers that have many co-authors. For e.g., publications with many co-authors can be the result of internationally funded projects. Some of these type of publications, resulting from for e.g. EU-projects, have a stronger focus on review of current knowledge or project experiences compared to other scientific papers focusing on novel findings and scientific discovery. Although the former type of contribution has a higher visibility (by the many co-authors) the character of the field (being small where researchers know the other research groups fairly well) provides a rather well connected research community that exchanges new findings through specialized conferences12 in addition to formalized research collaboration resulting in joint publications.

12

One example is the conference on Nanoengineering: fabrication, properties, optics, and devices, arranged by the International society for optical engineering, California, US, August 2006.

368

19.5

K. Larsen

Conclusions and Discussion

The study of co-authorship links and knowledge networks was designed to get an increased understanding of science output, impact and structure within the area of nanostructured solar cells. The study employs a combined quantitative and qualitative approach based on bibliometric analysis of publications from the Science Citation Index (SCI), review of policy documents, and interviews with researchers in the field discussing the findings of co-authorship links and research output from different locations active in this particular research field. The analysis of coauthorship links describes several different types of research collaboration. The first type of collaboration is technical measurement and development of solar cell devices. This explains, in the Swedish case of Uppsala, several of the domestic co-author links. The research and development activities in the area can be characterized as highly dependent on chemistry, as illustrated by co-citation at journal level. Measurement techniques for determining the performance of solar cells are important types of, often domestic interaction. As a contrast to this collaboration between geographically proximate locations within a country, co-authored papers were also a result of researcher mobility programmes targeting researchers in developing countries. The use of co-authorship data for analysis of research structure and centrality raises some important questions about the dynamics of the field and measures of publications and citations. The analysis of research output and research impact for locations based on co-authorship data also means that both total and fractional counting must be considered when studying publications and citations. Locations can achieve a relatively lower citation impact going from total to fractional counting, simply because they have a high share of co-authored or even multi-authored papers. This is contrasted with the approach of social network analysis where these co-authored papers with many contributors will increase the level of centrality measured by the share of co-author links in the network. This can also be translated to different views on the science system as a commodity or as an infrastructure. Research output and impact, measured by publications and citations, reflect the view of science as a production system or a commodity, while measures of research structure, such as co-authored papers, stress science as an infrastructure for science based knowledge transfer and interaction. These views also have bearing on approaches for research evaluation and impact assessment. The study shows, that it is possible to identify knowledge networks based on coauthorship measures for the research field concerned with the application of nanoscience for development of solar cells. From a science policy perspective, the study shows that increase in research output is not necessarily followed by higher research impact in a short-term perspective. Factors such as citation window and journal impact factor are expected to have an influence on citation levels, and this is also confirmed in this study. The study of the development of this sciencebased technology can provide knowledge about future shape of an emerging industry in terms of geography and collaboration patterns. Further analysis of

19

Co-authorship Networks in Development of Solar Cell Technology

369

knowledge networks in the research field would preferably focus on the emergence and evolution of networks over time and analysis of mechanisms behind forming strong, repetitive links between research groups and between regions.

Appendix Table 19.4 Anova results Location Sig. Uppsala 0.000 Lausanne 0.000 Baltimore 0.000 Uppsala (N=89) Beijing 0.000 Lausanne 0.000 Baltimore 0.356 Lausanne (N=91) Beijing 0.000 Uppsala 0.000 Baltimore 0.657 Baltimore (N=41) Beijing 0.000 Uppsala 0.356 Lausanne 0.657 Dependent variable ln(1+Citation fraction) used to ensure normal distribution and to not exclude cases that has zero citations when analyzing difference in citation levels between the nodes Beijing (N=135)

Table 19.5 Regression analysis results Baltimore Lausanne Uppsala Citation window (years from 1991) 0.770*** 0.592*** 0.665*** Journal high (JIF 3.5 or above) 0.278*** 0.319*** 0.213** Journal medium (JIF from 2 to <3.5) 0.427*** 0.147 0.172** Number of co-authors on paper 0.060 0.018 0.027 R square 0.736 0.453 0.471 Notes: Beta (Standardized coefficients) reported from regression analysis ln (1+Citation fraction) = CitWindow X1+JournalHigh X2+JournalMediumX3+CoA X4+e Dependent variable: ln(1+Citation-fraction) Explanatory variables: Citation window (years from 1991), Journal Impact Factor, Number of co-authors on paper ***Significant at 0.01 level; **Significant at 0.05 level; * Significant at 0.10 level In four cases JIF values not identified resulting in total N=217

References Acs ZA, Audretsch DB, Feldman MP (1994) R&D spillovers and innovative activity. Manage Decision Econ 15(2):131–138 Acs ZJ, Audretsch DB, Feldman MP (1991) Real effects of academic research: Comment. Am Econ Rev 82(1):363–367

370

K. Larsen

Aksnes DW (2003) Characteristics of highly cited papers. Res Eval 12(3):157–170 ˚ E, Persson O (1993) Networking Scientists. Annals Reg Sci 27(1):11–21 Andersson A Anselin L, Varga A, Acs ZJ (2000) Geographic and sectoral characteristics of academic knowledge externalities. Paper Reg Sci 79:435–443 Arnall HA (2003) Future technologies, today’s choices: Nanotechnology, artificial intelligence and robotics; a technical, political and institutional map of emerging technologies. Greenpeace Environmental Trust, London Audretsch DB, Bozeman B, Combs KL, Feldman MP, Link AN, Siegel DS, Stephan P, Tassey G, Wessner C (2002) The economics of science and technology. J Technol Transfer 27:155–203 Audretsch DB, Feldman MP (1996) R&D spillovers and the geography of innovation and production. Am Econ Rev 86(3):630–640 Beckman M (1993) The case of scientific interaction at a distance. Ann Reg Sci 27(1):5–9 Beckman M (1994) On knowledge networks in science: collaboration among equals. Ann Reg Sci 28:233–242 Braun T, Schubert A, Zsindely S (1997) Nanoscience and nanotechnology on the balance. Scientometrics 38(2):321–325 Calero C, Buter R, Cabello Valde´s C, Noyons E (2006) How to identify research groups using publication analysis: an example in the field of nanotechnology. Scientometrics 66(2):365–376 Chen C (2005) CiteSpace II: Detecting and visualizing emerging trends and transient patterns in scientific literature. J Am Soc Inform Sci Technol 57(3):359–377 Chen C, Hicks D (2004) Tracing knowledge diffusion. Scientometrics 59(2):199–211 Danell R, Persson O (2003) Regional R&D activities and interactions in the Swedish Triple Helix. Scientometrics 58(2):205–218 Darby MR, Zucker LG (2003) Grilichesian breakthroughs: Inventions of methods of inventing and firm entry in nanotechnology, NBER working paper 9825, National Bureau of Economic Research David PA, Keely LC (2003) The economics of scientific research coalitions: collaborative network formation in the presence of multiple funding agencies. In: Geuna A, Salter A, Steinmueller EW (eds) Science and innovation: Rethinking the rationales for funding and governance. Edward Elgar, Cheltenham-Northampton EC (2000) Towards a European research area. Brussels, European Commission EC (2003) COM (2003)131FINAL: Developing an action plan for environmental technology. European Commission ESRC (2003)The social and economic challenges of nanotechnology. Report by Wood S, Jones R, Geldart A, ISBN 086226-294-1. The Economic and Social Research Council (ESRC), UK EU (2005) Nanotechnology, http://www.cordis.lu/nanotechnology (Accessed 4 August 2005) Feldman MP (1999) The new economics of innovation, spillovers and agglomeration: A review of empirical studies. Econ Innov New Technol 8:5–25 Feldman MP (2001) Where science comes to life: university bioscience, commercial spin-offs, and regional development. J Comp Pol Anal: Res Practice 2:345–361 Feldman MP, Audretsch DP (1999) Innovation in cities: science-based diversity, specialization and localized competition. Eur Econ Rev 43:409–429 Gauffriau M, Olesen Larsen P (2005) Counting methods are decisive for rankings based on publication and citation studies. Scientometrics 64(1):85–93 Gla¨nzel W, Schubert A (2001) Double effort = Double impact? A critical view of international co-authorship in chemistry. Scientomentrics 50(2):199–214 Godø H, Nedrum L, Rapmund A, Nygaard S (2003) Innovations in fuel cells and related hydrogen technology in Norway – OECD case study in the energy sector, NIFU report 35/2003 Hagfeldt A, Gratzel M (1994) Light induced redox reactions in nanocrystalline systems. Chem Rev 95:49–68 Hagedoorn J, Link AN, Vonortas NS (2000) Research partnerships. Res Pol 29:567–586 Hicks D, Tomizawa H, Saitoh Y, Kobayashi S (2004) Bibliometric techniques in the evaluation of federally funded research in the United States. Res Eval 13(2):78–86

19

Co-authorship Networks in Development of Solar Cell Technology

371

Hicks D, Katz S (1997) The changing shape of British industrial research. STEEP Special Report No. 6, SPRU – Science and Technology Policy Research, University of Sussex Heinze T (2006) Emergence of nano S&T in Germany. Network formation and company performance. Fraunhofer Institute for Systems and Innovation Research, Discussion Paper No. 7/2006, ISSN 1612-1430 IVA (2004) Royal Swedish Academy of Engineering Sciences’ seminar on Swedish nanotechnology, 16 March 2004 Jacobsson S, Andersson BA, Ba˚ngens L (2001) Transforming the energy system – the evolution of the German technological system for solar cells. SPRU electronic Working Paper Series, No. 84, SPRU – Science and Technology Policy Research, University of Sussex Jaffe AB (1989) Real effects of academic research. Am Econ Rev 79:957–970 Jaffe AB, Trajtenberg M, Henderson R (1993) Geographic loclisation of knowledge spillovers as evidenced by patent citations. Q J Econ 108:577–598 Jake´lius S (2001) Clean technologies for sustainable development. In: Silveira S (ed) Building sustainable energy systems: Swedish Experiences. Svensk Byggtja¨nst and Swedish National Energy Administration, Stockholm, pp 267–300 Karlsson C, Manduchi A (2001) Knowledge spillovers in a spatial context – A critical review and assessment. In: Fischer MM, Fro¨hlich J (eds) Knowledge complexity and innovation systems. Springer, Berlin Katz JS (1994) Geographic proximity and scientific collaboration. Scientometrics 31:31–34 Larsen K (2008) Knowledge network hubs and measures of research impact, science structure, and publication output in nanostructured solar cell research. Scientometrics 74(1):123–142 Larsen K (2003) Research assessment criteria and university–industry links in the social contract of ‘applicable’ science – Experiences from funding of biomedical science and strategic environmental research in Sweden and the UK, presented at conference in honour of Keith Pavitt: What do we know about innovation? Organised by SPRU – Science and Technology Policy, University of Sussex, England, 13–15th November 2003 Law M, Greene LE, Johnson JC, Saykally R, Yang P (2005) Nanowire dye-sensitized solar cells. Nat Mater 4:455–459 Leydesdorff L, Zhou P (2005) Are the contributions of China and Korea upsetting the world system of science? Scientometrics 63(3):617–630 Melin G, Persson O (1996) Studying research collaboration using co-authorships. Scientometrics 36(3):363–377 Meyer M (2000a) Does science push technology? Patents citing scientific literature. Res Pol 29:409–434 Meyer M (2000) What is special about patent citations? Differences between scientific and patent citations. Scientometrics 49(1):93–123 Meyer M (2000) Patent citation analysis in a novel field of technology: an exploration of nanoscience and nano-technology. Scientometrics 51(1):163–183 Meyer M, Persson O (1998) Nanotechnology – interdisciplinarity, patterns of collaboration and differences in application. Scientometrics 42(2):195–205 Newman MEJ (2001) The structure of scientific collaboration networks. Proc Natl Acad Sci USA 98(2):404–409 NNI (2005) National Nano Initiative in the US. http://nano.gov (Accessed 4 August 2005) OECD (2003) STI Scoreboard 2003. OECD, Paris, pp 44–45 O’Regan B, Gratzel M (1991) A low-cost, high efficiency solar-cell based on dye-sensitized colloidal TiO2 films. Nature 353(6346):737–740 Owen-Smith J, Powell WW (2004) Knowledge networks as channels and conduits: the effects of spillovers in the Boston biotechnology community. Organ Sci 15(1):5–21 Pavitt K (1998) Do patents reflect the useful output of universities? Res Eval 7(2):105–111 Powell WW, Koput WK, Smith-Doerr L, Owen-Smith J (1999) Network position and firm performance: Organizational returns to collaboration in the biotechnology industry. Res Sociol Organ 16:129–159

372

K. Larsen

Powell WW, Koput WK, Smith-Doerr L, Owen-Smith J (2005) Network dynamics and field evolution: The growth of interorganizational collaboration in the life sciences. Am J Sociol 110 (4):1132–1205 Royal Society (2003) Nanoscience and Nanotechnology, http://www.royalsoc.ac.uk/nanotechnology (Accessed 4 August, 2005) Schummer J (2004) Multidisciplinarity, interdisciplinarity, and patterns of research collaboration in nanoscience and nanotechnology. Scientometrics 59(3):425–465 Silveira S (ed) (2001) Building sustainable energy systems: Swedish experiences. Svensk Byggtja¨nst and Swedish National Energy Administration, Stockholm Small H (1973) Co-citation in the scientific literature: a new measure of the relationship between two documents. J Am Soc Inform Sci 24(4):265–269 STEM (2004) El fra˚n solen – energi och industri i Sverige [Electricity from the sun–energy and industry in Sweden]. Swedish National Energy Administration Tijssen RJW (1991) A quantitative assessment of interdisciplinary structures in science and technology: co-classification analysis of energy research. Res Pol 21:27–44 Varga A (2000) Local academic knowledge transfers and the concentration of economic activity. J Reg Sci 40:289–309 Wasserman S, Faust K (1994) Social network analysis: Methods and applications. Cambridge University Press, Cambridge

Chapter 20

Off-shoring of Work and London’s Sustainability as an International Financial Centre Ian Gordon, Colin Haslam, Philip McCann, and Brian Scott-Quinn

20.1

Introduction

The emergence, from the 1960s on, of a new spatial division of labor – with the old task-based division of labor within a firm taking on a spatial dimension, and comparative advantage increasingly shaping patterns of specialization by function/ process as well as by sector/product – reflected both new possibilities opened up by developments in management, control and communications technologies and intensified competitive pressures within (generally) mature industrial sectors. On an international scale this primarily affected manufacturing activities, and was driven essentially by labor cost factors. Within developed economies, however, it also affected a number of (mostly) office-based service activities, where the crucial cost factor more typically involved premises rather than labor, since these tended to occupy expensive space in central locations offering the face-to-face communication potential required for some at least of their functions. In these cases the new spatial divisions occurred within much more restricted territories, both because there were tighter constraints (on the kinds of labor deemed suitable and on the dispensability of face-to-face contact) and because there was much more local variation in the relevant cost factor. Even so, there were US examples from the 1980s of telecommunications links being used to effect substantial savings in typing/data entry costs by exploiting cheaper pools of English-speaking labor in offshore locations such as Ireland or the West Indies (Warf 1989). Since the end of the 1990s, however, service activities in advanced economies have taken initiatives to shift a much wider range of information-related functions to offshore locations in pursuit of labor cost savings, as they in their turn come to face more intense price competition. In this context, more of the jobs involve high levels of human capital – for which the core economies had been

I. Gordon (*) Geography Department, London School of Economics e-mail: [email protected]

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_20, # Springer‐Verlag Berlin Heidelberg 2009

373

374

I. Gordon et al.

presumed to possess a comparative advantage – raising questions about how far the process could be extended, and whether a number of their advanced service activities, notably ‘‘wholesale’’ financial services, have also become vulnerable to ‘‘hollowing out’’. London, which is the empirical focus of this paper, is both the dominant financial service centre for UK business (including its overseas operations) and the leading centre in the world for genuinely trans-national financial operations (i.e. activities which are not simply supportive of other domestic businesses). The competitive advantage of this financial service complex rests in large part on an array of scale economies, in terms of the depth and breadth of the specialist skill pools, related and support services, institutionalized markets, established customer links and international connections. Though all of these have a crucial qualitative aspect (involving issues of trust and tacit knowledge), sheer scale is also extremely important in each case, reflecting a degree of embodiment of competitive assets in specific firms, groups of workers and linkages. The relevant quantitative dimensions obviously vary, but it is not unreasonable to believe that several are associated with the number of workers employed in (wholesale) financial service activities – or perhaps rather the number of highly skilled workers involved. This qualification would imply that conventional back-office decentralisation away from the City, of the kind observed over the last 40 years, was likely to be more or less irrelevant to its competitive advantage (at least in quality terms, though presumably advantageous in relation to price) – but that more recent, and prospective, transfers of graduate work offshore could start to erode the place’s competitive capacity. Another perspective on this issue is that of Michael Porter’s (1990) cluster theory, which links the competitive advantage of successful cityregions to concentrations of the home bases of individually successful firms in related activities. Interpreting ‘‘home bases’’ literally in relation to firms’ command centres actually raises other important questions about the long term sustainability of the City’s competitive position, since a wave of foreign takeovers in the 1990s means that a third or more of City activity is now undertaken in externally owned finance houses. Interpreting the concept more functionally, in terms of where the ‘‘core competences’’ of each business rest, presents a rather more reassuring picture, since it is clearly for these that foreign banks have invested in London, whether to establish parallel capacities to those in their real ‘‘home bases’’ (as appears to be the case for Wall Street firms), or as the main site for particular specialist activities (as with the Deutsche Bank). The long term stability of some of these arrangements is questionable (Augar 2000). But, in any case the character of these core competences, and of the skills they require, are currently being renegotiated, or ‘‘re-engineered’’, as part of the same strategic response to intensified competitive pressure as proposals to move large bodies of work offshore to sites (in India or elsewhere) with radically lower costs for IT-based business processes. From a rather broader perspective, this may be seen as part of a recurring dynamic through which elements of the value chain for maturing products, subject to increasing price competition, are routinized and (then) dispersed, away from the metropole/ core where maintenance of activity levels depends on continuing innovation and product differentiation.

20

Off-shoring of Work and London’s Sustainability as an International Financial Centre 375

In the IT sector itself the off shoring process1 goes back to the late 1990s, particular in relation to defensive work against the ‘‘Y2K bug’’.2 And UK retail banking has participated in the great shift of call centre work to India. But serious interest in off shoring core business processes from wholesale financial services only appears to go back 3 or 4 years. Hence estimates of the numbers and significance of jobs (actually or potentially) involved are quite speculative, and mostly emanate from groups with a stake in the process. One influential study from Forrester Research projects the number of off shored European white collar jobs as growing from an estimated 82,000 in 2004 to 1.17 million by 2015 (Parker 2004), with UK growth following the strong US trend, while other European countries lag behind, perhaps fatally for their competitiveness. The greatest density of moves is projected for financial services, with perhaps (on a proportional disaggregation) some 225,000 jobs going from UK financial services over this period, representing a shift of about 1.6% p.a. (Gordon et al. 2005). Two very large recent UK moves to India (by AVIVA and HSBC) are cited in support of this projection, though (perhaps significantly) both actually involve ‘‘retail’’ rather than ‘‘wholesale’’ financial products. More typically, the number of jobs involved in individual offshore moves has been relatively small, though the companies involved were generally very large (Bronfenbrenner and Luce 2004). Flows are by no means all in one direction, and in the rapidly growing field of IT and business service exports the UK has been one of the major beneficiaries (in both net and gross terms). But it is the entry of low labor cost nations such as India into this sophisticated market which represents the most radical change affecting the position of advanced service centres in the developed world. In this context there is a quite immediate issue about the scale of work which may be moved offshore from the wholesale side of UK finance houses – and about what part of that work might otherwise have been located in or around the City of London rather than in secondary locations ‘‘onshore’’. Beyond this, however, there is a pair of related questions about: l

l

1

Firms’ capacity to restructure their labor demands so as to make a much more substantial proportion of their City jobs suitable for off shoring About the potential for offshore centres to further upgrade their capacities so as to handle more distinctively ‘‘financial’’ operations, opening up an even greater proportion of City work to off shoring

Off-shoring is defined here to cover all transfers of work abroad by companies which remain domestically-based, whether to ‘‘captive’’ establishments of their own/affiliates or through outsourcing to independent companies (cf. Norwood et al. 2006). As of 2001, two thirds of service sector off-shoring is estimated to involve captive suppliers, while just 5% of its out-sourcing was off-shore (Farrell et al. 2005). 2 This ‘‘year 2000’’ problem, engendering an estimated 300 billion dollars of remedial work, stemmed from fears of the inability of older computer systems incorporating two digit versions of the year-date to handle transition to a new century/millennium.

376

I. Gordon et al.

If either of these developments goes far in the longer run, they would open up a set of endgame possibilities involving a ‘‘hollowing out’’ of the core businesses in global centres, along the model developed by global ‘‘manufacturers’’ and currently being pioneered in financial services by the hedge funds (Irving et al. 2003). In itself that might not alter the centrality of City (or Wall Street)-based banks, though they would become much less significant to their host cities, but it would be much more likely to erode the external economies currently protecting them from competition by more specialist centres, new or old. To start to address these questions, this paper first discusses the current activity and employment base of London’s financial centre in relation to the kinds of capacity developing in offshore centres (particularly in India), and then examines the approaches which City investment banks are currently adopting to these issues, drawing on a set of interviews with senior managers involved in this process during summer–autumn 2004 (Gordon et al. 2005).3 In the concluding section we return to broader long run issues in relation to questions about how far the strong contextual factors affecting the strategic choices of specific firms allow any general conclusions to be drawn about the scale and balance of outcomes likely at the urban/sectoral scale.

20.2

Financial Service Activity and Employment in the City

Advanced (or ‘‘wholesale’’) financial services, concentrated in and around the traditional ‘‘square mile’’ of the City and in the recent Canary Wharf development in Docklands, 3 miles to the east, are the single most distinctive element in the London economy – though (for this reason) their role tends to be rather overestimated. Apart from this physical concentration, much of their distinctiveness involves the presence of a large component (accounting for about 50% of jobs) which is predominantly oriented to overseas markets. Another related feature is that they represent the only large element within the London economy which professes to see significant advantage in being located close to related businesses – primarily for reasons of access to shared intelligence (especially in the case of the foreignowned finance houses) but also for access to the pool of specialist labor (Buck et al. 2002; Gordon and McCann 2000). Rather against standard assumptions about the relation between internal and external economies, businesses within the financial centre tend to be large, both absolutely and relative to other areas. Within the financial district some 200,000 people are engaged in financial intermediation. These represent about one sixth of all those employed in the sector nationally, but with a quite distinctive occupational mix – including much larger 3

Within the investment banking sector in the City of London/Canary Wharf, interviews were focused on the larger firms, both because of their substantive importance and since (understandably) they have been leading the way in off-shoring. In total the firms which we covered were estimated to employ some 63,000 staff, representing about half of the sector’s employment in this area.

20

Off-shoring of Work and London’s Sustainability as an International Financial Centre 377

proportions in two groups (directors, financial managers, accountants and brokers; and information, communications and software professionals) than found in areas outside London – reflecting both the emergence of a strong spatial division of labor within UK, the ‘‘wholesale’’ sector and the predominance of ‘‘retail’’ financial operations elsewhere. On a peak–peak basis over the last full (1987–2000) macroeconomic cycle the sector’s employment in this core area showed quite modest growth, of around 10% – much below the local growth in other kinds of business service. This reflects both growing efficiency in business processing and a continuing decentralization of jobs to other parts of the UK, at an average rate of around 7,000 jobs per year. During the upswing of that cycle (before off-shoring became an issue) employment growth in the core was concentrated entirely in managerial, associate professional and (especially) professional jobs, with actual job losses in clerical and related occupations – against a pattern of occupationally balanced growth outside London. It is not clear, though, how far actual dispersal of the more routine jobs is responsible for this shift, as distinct from their more rapid replacement by IT in the higher cost location. Salary levels in city financial services are high (double the national average) and have been growing more rapidly than elsewhere, though both features reflect in part their distinctive occupational mix, high proportion of graduates (43% in 2002) and strong work pressures. Within particular occupational groups earnings tend to be about 40% above the national average. The annual average was £40,000–50,000 in 2003 for both IT professionals and computer technicians (two key groups in relation to off-shoring). The differential has grown very considerably since the early 1980s, and continued to do so even through the period of downsizing after the Stock Market peaked in 2000 (Gordon et al. 2005). This long run trend reflects both simple cost inflation and the increasing specialization of the local employment base in the highest value functions. Current and expected forms of off-shoring may be partly a response to that cost inflation, but certainly represent a continuation of the process of specialization, though the jobs which are being considered for relocation now would include (for the first time) a large element of graduate work. This graduate element has been the distinctive feature of the current round of service sector off-shoring, particularly to India, though simple number counts for the financial service sector (as, e.g. transport or telecommunications) include much call centre work with less specialist requirements. The initial motive for the wave of moves starting in the mid-1990s had more to do with labor shortages in the UK than with labor cost savings, and this was even more true for the ‘‘millennium bug’’ work. The success of these early moves not only demonstrated the scale of ITqualified labor available in India (with a flow of some 150,000 graduates each year), but also provided the motive for a rapid development of institutional and business capacity in key centres such as Mumbai. Indicators of this development include: l

l

Radical liberalization of the real estate sector; the emergence of sophisticated businesses supporting inward investment A growth of large scale local IT service providers placing a strong emphasis on quality standards (with more certified at the highest CMMI level than in the USA)

378 l

I. Gordon et al.

A broadening of the range of specialized skills on offer, epitomized by the case of Office Tiger, a Chennai-based firm with 2,000 employees which has moved from data management into equities research for US clients

There are some suggestions of emergent labor shortages – specifically of labor suitable for work with multi-nationals4 – and there have certainly been recent salary hikes in these sectors. But they are clearly offering advanced services of increasing breadth, and a substantial majority of off-shoring businesses now report gains in quality of service as well as value for money (McCarthy et al. 2003). The question is how far (and how) this reservoir of skills can be deployed to meet the specific requirements of City-based financial services, which have traditionally drawn on a quite different skill and experience mix for most of their needs (beyond the 11% or so directly involved in IT type work). This involves a raft of questions about how City firms manage their business which are currently up for re-examination.

20.3

Decision-Making in the City About Restructuring and Off-shoring

Current interest in and planning for the off-shoring of blocks of work from City businesses to (principally) India represents part of a much broader and more systematic approach to the strategic management of business processes within the wholesale financial sector, where emphasis has traditionally been placed much more on the generation of revenue, subordinating operations and IT issues to the needs of the ‘‘front desk’’ (i.e. the trading side). In large part this shift reflects intensified competition in the sector which has put increasing pressure on cost recovery, profit and cash generated from operations. Technology and system improvements mean that high-margin products can now be turned into low-margin, high-volume utility products remarkably quickly, just as in many other industries. Whereas in the past regulation offered some protection, it is currently adding to the pressures from the market and inter-firm rivalry to shave product margins, fee structures and commission rates – with these pressures being intensified over recent years by the impact of a general malaise in capital markets on profit margins in the industry. Investment banks have responded in two ways. On the revenue side, while many business areas are becoming ‘‘commoditized’’, they have sought to achieve higher returns elsewhere through forms of proprietary trading resembling hedge fund activity, or dealing in credit derivatives. Such activities, however, increase both capital requirements and risk exposure for the banks, while their long-run profitability is uncertain, as new entrants can be expected to drive down arbitrage 4

From extensive interviews with HR managers, Farrell et al. (2005) conclude that only 13% of university graduates in low-wage countries are suitable for such work, with notably lower proportions among finance/accounting graduates or (especially) generalists than for engineers.

20

Off-shoring of Work and London’s Sustainability as an International Financial Centre 379

opportunities over time, as in other areas. Hence much attention has been directed towards the cost side, focusing on the scope for rationalization and more integrated design of the business process to achieve significant economies. Here demands from the capital market to restore shareholder value have reinforced a more general process of business restructuring, partly linked to the transformation of ownership patterns in the City during the 1990s (Augar 2000; Kynaston 2001). The logic of the approach followed on the cost side involves the identification of functional blocks of work and analyses of the potential for cost saving in these through a combination of standardization, automation, outsourcing and/or relocation, having regard to issues involving specific skill requirements; sources of risk and commercial/regulatory constraints; as well as simple cost differentials (cf. Baldwin 2003). Calculations of prospective cost savings start with salary levels, since labor accounts for two thirds of overall costs, and alternative locations for particular blocks of work are seen to offer 20–30% lower pay requirements in the case of near shore locations (e.g. within the British Isles) and 80–90% lower in offshore locations such as India. On a fully loaded basis, however – including infrastructure costs, training, supervision, personnel management and (possibly) maintenance of a shadow capacity in London as insurance – the cost differentials narrow substantially, but can still represent savings of 40–60% for jobs successfully relocated to India. And it is important to note that many firms reported quality gains in the work that had been offshored, both through access to technically better qualified workers in IT specialisms and through reductions in staff turnover rates. Though prospective gains from relocating work seem great, they need to be seen in relation to the substantial savings which firms have been able to achieve in situ (on a cherry-picking basis) by simply re-aligning processes and streamlining systems – and those which some expect from global concentration of particular core functions (in any location), and others from dramatic speeding-up recording processes through more efficient use of ICT. To some extent all of these represent alternatives to off-shoring work (whether through outsourcing to local contractors or developing dedicated facilities of their own), but they can also be necessary precursors to successful off-shoring. And shared use of offshore suppliers (notably some of the major Indian firms which have developed since the late 1990s) can be one effective route to securing the benefits of scale economies in provision of noncore services. The choices that firms are in the process of making among different combinations of these approaches to cost-saving are complex, and vary in ways that are related to differences in their business structure and culture, notably in relation to types of function and product, and the differing risks/variability associated with these. In particular, our interviews revealed two key dimensions along which the risks of transferring work offshore were seen to vary, relating to the need for physical proximity in a function and the simplicity or complexity of the product involved. At the low risk extreme were routine support functions, including accounts, pay, treasury operations and database modifications, for which remote control was

380

I. Gordon et al.

adequate and system breaks could be remedied locally. A second least risky category involved the processing of commoditized products (‘‘plain vanilla’’ or ‘‘closed book’’), where regulatory and compliance requirements are all quite routine. In both these cases off-shoring is seen as feasible because an established structuring of processes involves a clear division of labor between front and back offices. This was much less true for a third category of work, characterized as structured business where team efforts and negotiation with multiple parties are more commonly required to handle complex compliance, legal or tax issues. But the extreme in terms of riskiness was represented by differentiated, over-thecounter products and provision of services to high value clients with sophisticated requirements – both with an indispensable face-to-face element to their handling. The general assumption among our interviewees was that only older and simpler product processing would be considered for relocation (even within the UK),5 and then with the aim of freeing staff resources in the City to deliver more complex products. One in particular was clear that the logic of decentralization of functions, down a chain from the City office to offshore centres, was to enable the level and quality of work in each location to be progressively upgraded. As between firms the most systematic differences were related to size. The largest global companies were able to exploit scale economies and labor cost differences through in-house IT operations in locations such as India, without the risks associated with use of third party vendors. The smaller players were more strongly attracted to outsourcing, though the risks were often unacceptable. And since even the larger firms were working on individual off-shoring projects involving groups of just 25–50 workers, it can be seen that the planning and transaction costs of such moves could well be disproportionate for smaller businesses. Because of the nature of the industry a wide range of risk factors were recognized in off-shoring work, in relation to data confidentiality; regulatory compliance; interruption of operations; qualitative shortcomings threatening business reputation; future cost inflation or labor shortages; ceding power to local suppliers, and loss of the capacity to reverse the strategy if circumstances changed. Attitudes to these clearly varied within as well as between firms, but some general lessons were widely recognized in terms of adopting an incremental/ experimental approach and securing clear contracts or internal service level agreements to codify all performance requirements, costs, responsibility for risks and means of enforcement/redress (cf. Willcocks et al. 2006). On the other hand, interviewees were clear that there was a need to keep a wide range of capabilities in London, partly for compliance or security reasons (as a fallback) but particularly for staff development, and the potential to flexibly

5 Y/Zen Ltd (2005) similarly conclude from interviews with London, New York, Frankfurt and Paris-based firms that ‘‘financial centers may lose certain types of commoditized activities to low cost cities but the important parts of the industry’’.

20

Off-shoring of Work and London’s Sustainability as an International Financial Centre 381

reconfigure business systems for product development across the traditional ‘‘silos’’ to meet changing client expectations. In this context it was argued that there was an increasing need for firms to develop staff with a broad mix of skills and experience of many parts of the product value chain, making a case for keeping some or all of many parts of that chain within (or at least directly managed from) central London. Access to a pool of skills of this kind was also seen as the most crucial reason for businesses to retain a London base, apart from its convenience for client access. Our interviewees were somewhat sceptical about the importance of other kinds of faceto-face interaction, for example as sources of intelligence, given the shift to screenbased trading. But they were clear that the city had crucial advantages in terms of recruiting and retaining staff with the specialized understanding of financial markets, products, investment banking practices and customer requirements, only attainable through employment experience in one of the global financial centers. This is very important in relation to the conjecture that offshore centres might continue to upgrade their capacities beyond IT and business process work to take over more core financial functions. Some support for this idea had come from the reported off-shoring of some ‘‘research’’ functions – a function now under particular pressure to cut its costs. In addition to actual IT work, accounting and financial tasks, salaries and staff records, some of our interviewees did refer to transfers of company based analytical research and credit analysis. However, these primarily relate to IT operations, and we came across no examples of more judgemental research operations being considered for off-shoring. As Grote and Taube (2004) point out from a Frankfurt-based study, there are significant variations among these (for example between international equity country analyses) in the degree of dependence on tacit knowledge, but: ‘‘research activities are locally embedded in Western financial centers to an extent that such a development is not likely’’ (p.2). And indeed the tacitness of key elements of City workers’ knowledge base seemed to represent a crucial limit to the scope for off-shoring of their work to centers whose comparative advantage is in particular sorts of highly codified human capital.

20.4

Conclusions

Discussion of off-shoring in advanced service activities has shown a tendency to over-excitement, reflecting: l l

l

The newness of this phenomenon The unexpectedness of the command and control centers of globalization becoming directly subjected to its logic Quite reasonable questions about ‘‘where this will all end up’’ if LDC’s have become effective competitors for graduate level work in knowledge-related sectors

382

I. Gordon et al.

Even from a more distanced analytic perspective than those of the proponents and protectionists, the phenomenon is a very interesting one, both in further exemplifying some logics which ought to be familiar, and in drawing attention to some of the conditionalities affecting how these operate in practice. In terms of familiar economic logics, the case exemplifies how processes of internationalization (legal, organizational, cultural and technological) continue to erode constraints on the operation of comparative advantage, allowing increasingly fine divisions of labor to be translated into spatial terms. They also illustrate the continuing interplay between this (price-based) Ricardian principle of comparative advantage and the (qualities-based) Porterian principle of competitive advantage. One take on this is in terms of a double spatial division of labor in which cityregions compete: First, in quality terms to enlarge/maintain the market shares and profitability of firms for which they provide (embedded) home bases; and Second, in terms of pure comparative advantage for shares of more/less desirable employment associated with divisible (mobile) functions of these and other firms. In relation to the major metropolitan centres one may also see this in product cycle terms as involving a continual tension between: l

l

l

On the one hand, their capacity to reproduce monopolistically competitive advantages through product innovation, differentiation and external market linkages; and On the other, the drive to routinize and decentralize work as competitive pressures develop in maturing segments of the market. This is a process which in the long run involves a succession of locally dominant sectors within successful centers, where much of the cost pressure in the past has come precisely from competing demands for local space and inelastic types of labor. In the case of off-shoring financial service work, however, this relationship is less central, since the (potential) operating cost differential has emerged through entry into the frame of alternative locations with historically lower labor costs – against the background of radically widening wage differentials in the UK as a whole since the 1980s.

In terms of less well understood (and more open-ended) conditionalities, the case first provides a humbling reminder of how hard it is to work out in advance the practical effects of these logics on the spatial distribution of activity. This is a point made by Massey and Meegan (1982) in relation to the potential spatial ramifications within the UK of the intensified international competitive pressures (in price terms) then facing British manufacturing. In this case they identified three major kinds of strategic response – work intensification, rationalization and capital deepening – each with significantly different locational implications, but ones which were inseparable from choices about factors such as technology and approaches to labor relations, and unpredictable except in relation to sectorally specific analyses. At a sub-sectoral level within major city investment banks a similar kind of

20

Off-shoring of Work and London’s Sustainability as an International Financial Centre 383

(intelligible) variability was evident in relation to the firms’ plans with respect to alternative strategies for cost reduction. Three specific aspects of open-endedness apparent in this case involve questions about: l l

l

The availability of effective productive capacities in alternative locations The range of skill profiles capable of successfully delivering financial service products The future innovation potential of advanced financial services within an established centre

In relation to the first issue, the basic point is that the emergence of off-shoring as a serious alternative for work above the level of call-centres and data entry depended not simply upon the discovery of a large (existing and potential) supply of highly qualified IT professionals in select LDC locations, but also on the development there of a rather sophisticated internationally credible infrastructure of suppliers, property intermediaries, consultants, place marketeers and other support services. This could not have been taken for granted, going well beyond what was assumed to be involved in spatial divisions of manufacturing labor, and marks a substantial upgrading of the agglomeration potential of LDC cities. In the rapid re-definition of the competitive capacity of cities such as Mumbai from a pure ‘‘IT support’’ to a much broader ‘‘business process’’ role there may be parallels with the kind of upgrading pursued in the electronics sector by East Asian ‘‘tiger economies’’ such as Taiwan. But even with this extension, the second important issue is about how far their capacities are substitutable for the more specifically finance/market-related kind of expertise on which the City’s competitive advantage rests. The evidence from our interviews is that – over and above specific issues about regulatory compliance – London banks do not yet see this expertise as dispensable, or offshore workers as competing in this area. But the logic of an increasingly radical approach to business process re-engineering is to push further at the separation of generic business process from specialist financial expertise, and raise questions about the proportion of workers required with that more place-bound expertise. This shades in turn into the third source of uncertainty, about the continuing potential within advanced financial services for the kind of innovativeness and customization of products on which maintenance of high proportions of dedicated financial professionals and of employment concentration in the financial centre depends. We could speculate about this, but it is truly an open-ended issue. For the more foreseeable future, the City bankers whom we interviewed were confident about the prospects both for achieving substantial cost savings, mostly through increased efficiency, and for continued innovation. They were becoming more comfortable with off-shoring as one element in their strategies, but were not expecting this to have a significant net effect on City-based employment – though a return to sales growth was expected to be accommodated more through productivity gains than substantial employment growth.

384

I. Gordon et al.

References Augar P (2000) The death of gentlemanly capitalism. Penguin, London Baldwin M (2003) More than off-shoring: smart-sourcing. Capco Inst J Financ Transform 8(August):95–102 Bronfenbrenner K, Luce S (2004) The changing nature of corporate global restructuring: the impact of production shifts on jobs in the US, China, and around the globe. US–China Economic and Security Review Commission, Washington, DC Buck N, Gordon IR, Hall PG, Harloe M, Kleinman M (2002) Working capital: life and labour in contemporary London. Routledge, London Farrell D et al (2005) The emerging global labor market: part II the supply of offshore talent in services. McKinsey Global Institute, San Francisco. http://www.mckinsey.com/mgi/reports/ pdfs/emerginggloballabormarket/part2/MGI_supply_fullreport.pdf Gordon IR, McCann P (2000) Industrial clusters: complexes, agglomeration and/or social networks? Urban Stud 37:513–532 Gordon IR, Haslam C, McCann P, Scott-Quinn B (2005) Off-shoring and the City of London. Corporation of London, London Grote MH, Taube FA (2004) Offshoring the financial services industry: implications for the evolution of Indian IT clusters. Faculty of Economics and Business Administration discussion paper. J.W. Goethe University, Frankfurt am Main Irving B, Shojal S, Gupta S (2003) Discovering the endgame in the off-shore debate. Capco Inst J Financ Transform 8:103–112 Kynaston D (2001) The City of London: a club no more 1945–2000. Chatto and Windus, London Massey D, Meegan R (1982) The anatomy of job loss: the how, why and where of employment decline. Methuen, London McCarthy JC, Ross CF, Schwaber C (2003) Users’ off-shore evolution and its governance impact. Forrester Brief, Forrester Research Inc. http://www.forrester.com Norwood J et al (2006) Off-shoring: an elusive phenomenon. NAPA, Washington Parker A (2004) Two-speed Europe: why one million jobs will move off-shore. IT view research and business trends, Forrester Research. http://www.forrester.com Porter ME (1990) The competitive advantage of nations. Free Press, New York Warf B (1989) Telecommunications and the globalisation of financial services. Prof Geogr 31:257–271 Willcocks LP, Cullen S, Lacity M (2006) The CEO guide to selecting effective suppliers. The outsourcing enterprise, research paper 3. LogicaCMG, London. http://www.logicacmg.com/ uk/350236490 Y/Zen Ltd (2005) The competitive position of London as a global financial centre. Corporation of London, London

Chapter 21

The Genesis and Evolution of the Stockholm Music Cluster Pontus Braunerhjelm

21.1

Introduction

I think that it is more prestigious for Ricky Martin to be allowed to work with Swedish song-writers than the other way round. A connection with Swedish song-writers is almost a necessity for an artist to have an international success.1

Why do superstars like and Bon Jovi, Britney Spears, Maddona and Ricky Martin, to an increasing extent choose Swedish composers and producers in an industry characterized by extremely fierce international competition? Bergen, Copenhagen, Dublin, London, Los Angeles, Manchester, New York, Paris and Seattle are some of the more prominent competitors to the Swedish – particularly Stockholm – music clusters. What triggered this evolution and which dynamic forces have been decisive in the creation of the Stockholm cluster? Despite the impressive research presented on spatial issues in the last decade, we know surprisingly little about the forces initiating the creation of clusters. Economic geography models originating in the international trade theory literature, view agglomeration as a function of linkages (pecuniary and non-pecuniary), trade costs and scale economies (Krugman 1991; Venables 1996; Fuijita et al. 1999; Braunerhjelm et al. 2000a,b). Still, the focus is rather on the re-location of already existing economic activities than the emergence of new clusters. New constellations of existing clusters that appear as altered trade costs (e.g. due to an integration process) induce a re-shuffling of firms and factors of production, leading to a new pattern of spatial distribution of economic activities. Notwithstanding that these models constitute a true contribution to our understanding of the spatial dimensions P. Braunerhjelm Royal Institute of Technology, Stockholm, Sweden e-mail: [email protected] 1

Kai R. Lofthus, reporter in the international music magazine Billboard, interviewed by the Swedish newspaper Expressen, March 16, 2002 (Lofthus 2002).

C. Karlsson et al. (eds.), New Directions in Regional Economic Development, Advances in Spatial Science, DOI: 10.1007/978-3-642-01017-0_21, # Springer‐Verlag Berlin Heidelberg 2009

385

386

P. Braunerhjelm

of economic activities. Empirical observations also suggest that clusters emerge for quite different reasons, such as exogenous shocks attributed to technological breakthroughs, locational specific factors, historically random events, deregulation, market structure and down-sizing of the government, or the disinvestment of a dominating firm (David 1985; Scott and Storper 1986; Arthur et al. 1987; Arthur 1989; Krugman 1991; Feldman and Francis 2002; Scott 2003; Braunerhjelm and Feldman 2006). The development of the Stockholm music cluster serves well as an example of how local knowledge, to some extent concentrated but also latent and partly related to adjacent fields of knowledge, can develop into a major international competitor in a remarkably short period. To argue that the emergence of the Stockholm music cluster can be traced to changes in trade costs, or the presence of important linkages, makes little sense. Linkages are no doubt important for strengthening and expanding the cluster; however, it requires a minimum critical mass among which linkages can be established. As will be shown below, the emergence of the Stockholm cluster can be attributed a whole set of reinforcing factors. Hence, the objective of this paper is to shed new insights on the forces prompting the emergence of clusters, and how those forces interact with the more well-known mechanisms – referred to above – worked to enforce and sustain existing clusters. These processes will be illustrated through a careful examination of the Stockholm music cluster. The theoretical foundations is provided by the findings of the new economic geography literature referred to above, together with the strand of economics emphasizing the role of systems and connectivity in evolutionary processes (Porter 1990; Carlsson and Stankiewicz 1991; Nelson 1993). The novelty of this paper thus relates to the very early phase as economic activities begin to concentrate spatially, and the ensuing dynamics of the cluster as a critical mass has been attained. The approach in this chapter is exploratory which primarily is due to the difficulties in obtaining data for an industry that is not captured well in the official statistics, rather it spans over a number of service and producing sectors, as well as the inherit complexities in measuring the evolutionary process in the emerging phase of an industry. Still, some statistical evidence will be presented at the industry level. To further come to grips with cluster formation and dynamics, conventional statistic sources must however be complemented by interviews since available data does not comply with the cluster concept. Special attention will be devoted to the central elements of the creation of the dominant Stockholm music cluster – the ‘‘igniting spark’’ – within the Swedish music industry, and the forces having propelled the development and specialization within the cluster over time. The results derived from the analysis are likely to be applicable to other (particularly service) clusters, even though there are undoubtedly features specific to the music industry. Initially, the theoretical framework and the method applied in the present study will be described (Sect. 21.2). Section 21.3 provides a summary of the extent of the Swedish music industry and the factors underlying its success in the last few decades. Sections 21.4 and 21.5 analyze the factors explaining the development of Stockholm as an internationally competitive cluster within the Swedish music industry. Finally, the concluding section summarizes the main findings and discusses the policy implications.

21

The Genesis and Evolution of the Stockholm Music Cluster

21.2

387

Theoretical Framework and Method

Even though the definitions of the cluster concept to some extent differ across disciplines and authors, there seems to be agreement on certain fundamental cluster characteristics (Porter 1990; Enright 1996; Fuijita et al. 1999; Gordon and McCann 2000). Overall, clusters are defined as the production of similar or complementary goods and services in spatially well-defined areas (Braunerhjelm and Carlsson 1999). Some internationally well-known examples are Hollywood (movie-production), Silicon Valley (ICT), London (finances) and Sophia Antipolis, France (biotechnology). In Sweden, which is at focus here, Kista and Karlskrona/Ronneby in information and communication technology products, Stockholm/Uppsala ¨ resund region (Medicon Valley) (e.g. Karolinska institutet and Novum) and the O in biotechnology, constitute examples of clusters representing different industries. This basic definition – proximity in physical (geographical) and product spaces – has ramifications in different directions where certain cluster characteristics are more or less emphasized. The strength of clusters supposedly depends on the frequency and extent of pecuniary linkages (to customers and suppliers) and nonpecuniary linkages (knowledge spillovers), and the presence of a certain set of agents in the cluster. On the other hand, locational stability of clusters depends on the size of the cluster, the mobility of factors of production, trade costs (defined as trade barriers and transports of output) and the propensity to restructure in response to market signals. Altered trade costs, either due to an integration of markets or technological progress (e.g. Internet) influence agglomeration and the pattern of cluster formation across countries and regions.2 The dynamics of clusters is considered to depend on their maturity, the degree of knowledge intensity and the extent of the diffusion of knowledge.3 Once more, vertical and horizontal links, a close interface between customers and suppliers, as well as simultaneous existence of competition and co-operation across firms and institutions are assumed to drive the dynamics of the clusters and their tendency to change (Clark et al. 2000). The inflows and outflows of firms and agents, contributing to increased variety, seem to be a vital component in this respect (Saxenian 1999). Clusters can also be defined in terms of cognitive, institutional/organizational and economic dimensions, interacting in innovation or technological systems (Nelson 1993; Carlsson 2001). Since the aim is to show how competence, impulses and specialization (cognitive dimension) have been created and evolved in the cluster, how the contacts between agents and the network (institutional/organizational dimension) are organized and knowledge is transmitted, and what the links to the market look like (economic dimension), the analysis is organized according to

2 See Porter (1990, 2000), Fuijita et al. (1999); Fujita and Thisse (2002); Braunerhjelm and Feldman (2006). 3 Martin and Ottaviano (1999), 2001???); Fujita and Thisse (2002); Maggioni (2006).

388

P. Braunerhjelm

these dimensions. In the following sections, the cognitive, institutional/organizational and economic dimensions will be connected to the classification of the Swedish music industry as presented by Hallencreutz et al. (2004): l

l l

l

Specialized services – artists, lyric writers, musicians, producers, composers, etc. Production – music production, music performances, copyrights, etc. Related activities and supportive institutions – musical studies, record companies, recording studios, technicians, printers, agents, managers, etc. Customers – the final consumers of music.4

Using this as our starting point, semi-structured in-depth interviews have been carried out with three different groups of agents. All interviews have been made with individuals on a managerial or operational level and they have all been carried out in the period June to August 2002.5 The size of the firms interviewed varies from individual proprietor enterprises which are predominant among composers/ producers, to 20, 50 and 58 employees at the record companies interviewed. Only one company is listed on the Swedish stock exchange (MNW). Overall, firms within these segments of the music cluster are thus small. The selection of respondents is based on the author’s own knowledge, previous studies, as well as agents who have turned out to be central for the Stockholm music cluster as the interviews have proceeded. This resulted in a large number of names, firms and institutions. The interviewees were taken from the largest and most important agents in each segment. This means that the sample is a subjective rather than a random choice, but it is likely that the selection is representative of the segments we have chosen to study in the Stockholm music cluster. Altogether 20 interviews were undertaken for the current study (see Appendix).

21.3

The Swedish Music Industry: Structure and International Position

The Swedish ‘‘music miracle’’ has been observed in several reports in the last few years, but little is known about its precursors and the extent of its claimed success in an international comparison (Forss 1999, 2003). There are numerous difficulties in international comparisons of the music industry. Often individuals have a part time 4

Hallencreutz et al. (2002) also include a fifth category Equipment, machines and related services (producers and suppliers of studio equipment, instruments, video producers, etc.) which is disregarded in the present study. 5 In the last few years, interviews have increasingly come to be considered as an important complement to statistical/econometric methods in empirical studies. See, e.g., The NBER project on industrial technology and productivity (http://www.nber.org), Borenstein and Farrell (1998) and Scherer’s (1986) earlier criticism of the skepticism of economists’ to use interviews as a complementary method.

21

The Genesis and Evolution of the Stockholm Music Cluster

389

engagement in music activities, incomes may be registered in other countries and definitions vary across countries. We will present statistics from several sources that amount to at least circumstantial evidence of a Swedish success in the music industry. As a first measure of international success, consider Billboard’s ranking of top records. According to a record’s ranking and time on Billboard’s weekly listing of the 100 most popular songs, a record is given a certain number of points. Figure 21.1 illustrates the ranking of Swedish music groups from 1974 to 1995. Before 1974, there were no Swedish groups, but since the Swedish pop-group ABBA won the Eurovision Song Contest in 1974, there has been a positive trend. Another indicator is the high revenue of Swedish music houses in relative terms, compared to other countries (Table 21.1). The per capita revenue is much higher in Sweden than in, for instance, more well-known music nations like the UK and the USA. A more detailed picture of the Swedish music industry, and how it relates to other regions and countries of similar size, is presented in Table 21.2. The two US regions included refer to the Cleveland and Seattle metropolitan statistical areas, which complement the comparison with four European countries of about similar

10000 9000 8000

Index

7000 6000 5000 4000 3000 2000 1000 0 71

73

75

77

79

81

83

85

87

89

91

93

95

Fig. 21.1 The Swedish music miracle, ranking of Swedish pop-groups at Billboards top pop singles list 1971–1995 Source: Billboard top pop singles 1955–1996 (Forss 1999) Table 21.1 Global revenue for music houses, by country, 1994 Country Revenue music houses (million US$) USA 1,206 Japan 896 Germany 851 France 625 UK 489 Sweden 97 Rest of the world 1,605 Total 5,769 Source: UNESCO 1998, STIM Annual report 1987–1997, OECD

Revenue per capita 4.6 7.2 10.4 10.6 8.5 11.1 –

390

P. Braunerhjelm

size (even though all of them are smaller) as Sweden. Cleveland and Seattle are well known musical centers in the USA, however for quite different reasons: whereas Cleveland hosts one of the leading symphony orchestras in the USA and has a classical tradition; Seattle is the place from which the ‘‘grunge’’ style of modern music originates. The other Nordic countries have about the same musical culture and tradition as Sweden, rooted in folksongs, choirs and classical music. Some of them have also been quite successful in launching new pop-groups, particularly Norway. The only non-Nordic country included in the comparison – Ireland – is perhaps most similar to Sweden; a strong music culture, a cultural tradition of story telling and poetry, and more recently some extremely successful pop- and rock artists (van Morrison, U2, Sinead O’Connor, and others). The aim of this study is not to undertake a detailed comparison of these countries with respect to the music industry, rather to ‘‘benchmark’’ the Swedish music industry against these countries. Even though statistical difficulties are ample, it helps in assessing the size and composition of the Swedish music industry and the Stockholm cluster before we start to analyze the causes of its emergence. The comparison is primarily based on the numbers of firms, but the bottom row in Table 21.2 includes employment data. There are several missing values at the industry level, particularly for Ireland, but the aggregate figures should be reasonably complete. The Swedish music industry consists of about 15,000 enterprises (1999), employing about 15,000 people (Table 21.2). Most firms are small, approximately 83% are self-employed (i.e. have no employees), and only seven firms have more than 200 employees. The majority of firms can be found in two industries: performing artists, producers of artistic and literary work (sic 92,310), and publishers of sound recordings (sic 22,140). These two branches account for 78% of the firms in the Swedish music industry. A closer examination, however, shows that roughly 1,500 firms can be defined as belonging to the core of the Swedish music industry, employing about 10,000 people, meaning that there are a large numbers of firms having no employees and negligible revenues.6 As compared to the other countries and regions, Sweden – together with Finland – turns out to have a comparatively large number of firms in the music industries. The apparent dominance of Sweden is most notable when it comes to performing artists and publishers (composers) of music. In relative terms the share of firms and employees in the music industry is quite modest in all countries. Still, judging from official statistical sources, Sweden has by far the largest percentage share of firms in the music industry, close to 3%. Remaining countries/regions all display a share below 1%. Employment data reveal a somewhat different pattern. From these data Ireland turns out to have the relatively largest music industry (0.60% of employment), closely followed by Norway, Sweden and Seattle in the USA (all around a percentage

6

For a more detailed description of the Swedish music industry, see Hallencreutz (2002), Lundequist (2002) and Hallencreutz et al. (2004).

Table 21.2 Total number of firms in the music cluster distributed on different industries and regions/countries, and total employment in the music cluster Nace Industry U.S.1 (2001) U.S.2 (2001) Dk (1999) Fin (1999) Nor (1999) Swe (1999) Ire (2001) Music creation 92,310 Performing artists (music) 31 67 219 1,041 301 9,884 n.a. Publishers 22,140 Publishers of Music 4 5 244 313 57 1,171 20 22,150 Other sound publishers n.a. – 417 – 127 647 22,310 Reproduction of sound recording 19 37 39 47 16 57 34 24,650 Prepared unrecorded media n.a. – 7 5 3 42 10 Distribution 51,433 Wholesale (CD, tapes) 135 246 107 49 45 207 n.a. 52,453 Stores for records, videotapes 97 105 158 – 251 507 n.a. 52,454 Stores for music Instruments 44 9 196 256 89 412 n.a. 92,320 Theatres, concert halls 22 54 416 151 148 800 n.a Musical instrument manufacturing 36,300 Music instruments 13 11 115 56 7 155 7 Other n.a n.a n.a 225 132 719 n.a Total no. of firms, music cluster 365 616 1,918 2,143 1,177 14,610 291 % of total firms 0.48 0.60 0.83 0.97 0.49 2.83 0.20 Total employment, music cluster 4,502 5,948 5,429 3,108 9,125 15,264 8,101 % of total employment 0.33 0.38 0.32 0.16 0.44 0.42 0.60 Source: Carlsson (2004), Census of Industrial Production (2001), Hallencreutz et al. (2002), Music Board of Ireland (2002), OECD (2004), Statistics Sweden (2002) and own calculations U.S.1 Cleveland metropolitan area, U.S.2 Seattle metropolitan area, DK Denmark, Fin Finland, Nor Norway, Swe Sweden, Ire Ireland For Ireland the numbers of firms in industry 22,310 also includes 22,320 and 22,330, whereas 24,650 also includes data for industries 24,640 and 24,700. Moreover, total employment figure and the share of employment in the music industry refer to year 2000

21 The Genesis and Evolution of the Stockholm Music Cluster 391

392

P. Braunerhjelm

share of 0.40). These data could be expected to be more reliable than data on the number of firms. Yet, these figures may also be questioned, mainly because individuals active in the music industry often also hold another job and it is unclear as to how this is classified.7 Taken together – absolute numbers, shares of firms, and shares of employees – the picture that emerges from Table 21.2 suggests that Sweden has a relatively large music industry, albeit there is no unambiguous evidence of a comparative Swedish dominance in the music industry.8 The size of the music industry can also be put in perspective with other industries within a country. For instance, the Swedish biotechnological industry is estimated to employ between 10,000 and 20,000 people in about 300 firms (Allansdottir et al. 2002). Hence, employment-wise it is about the same size as the Swedish music industry. In an international comparison, Sweden has been claimed to be the third largest exporter (per capita) of music in the world, only surpassed by the USA and UK (Forss 1999). This observation in itself motivates a study of what factors have made it possible for Sweden to reach this position. In 2002 Swedish music exports (including services, royalties and goods) amounted to 750 million Euro, which can be compared to a traditional Swedish industry such as iron ore (exports 420 million Euro) or medical instruments (exports 740 million Euro). There has been a steady increase of approximately 15% annually in 1990–2000. However, export growth fell to about 5% in 2000 and 2001. For more recent periods (2002 and 2003), export data is more difficult to assess due to the re-classification of items that may give an impression of continued increase in exports.9 In reality, export seems to have fallen in the last few years. The structure of exports earnings has also changed. The major part of the music industry’s export revenue can still be attributed to goods but an increasing share is related to services and royalties. In 1997 the music industry’s share of exports of services and royalties was 33% which increased to 41 in 2001. An illustration of this is provided in Fig. 21.2, where the royalty revenues from abroad for the period 1996–2001 are shown to be steadily increasing. Most people agree on the fact that the international Swedish success in the music industry started with the internationally extremely successful pop-group ABBA, in turn followed by Roxette, Ace of Base, The Cardigans, etc.10 A new generation of

7

Note the difference between number of firms and employees in the Finnish case. This underlines the data problems and calls for a cautious interpretation. 8 If, as indicated above, a more realistic measure of the number of firms in the Swedish music industry is 1,500, then the share of firms in the music industry declines to 0.20. Assuming that similar measurement problems accrue to the other countries, we stick to the figure of 15,000 firms. 9 For instance, DVDs are now included in the export figures which however mainly reflect exports of films. 10 The presence of ‘‘superstars’’ has been shown to be decisive for other, more research intensive industries, such as biotechnology (Zucker and Darby 1996???).

21

The Genesis and Evolution of the Stockholm Music Cluster

393

300 250 200 K E 150 S M 100 50 0 1996

1997

1998

1999 Year

2000

2001

2002

Fig. 21.2 Swedish music-royalty revenues from abroad, 1996–2002, Million SEK

music groups now seem to be on its way to the international market, e.g. The Hives, Division of Laura Lee, Soundtrack of Our Lives and Sahara Hotnights.11 At the same time, ABBA’s initial success, which constituted a springboard for a number of other groups, has been followed by a shift in the specialization of the Swedish music industry. Swedish music groups and artists still enjoy a certain level of success, but the focus has, to a larger extent, shifted to Swedish composers and producers. Among these, the studio collective Cheiron and Murlyn Music Group stand out in particular.12 These agents have composed and produced music for international artists such as Britney Spears, The Backstreet Boys, N’Sync, Jennifer Lopez, Ce´line Dion, Ricky Martin, and Bon Jovi, to mention a few. In the following sections, we will attempt to identify the underlying factors that have contributed to develop Stockholm into a leading international music cluster, despite fierce world-wide competition.

21.4

Cluster Emergence: The Precursors

Broken down into regions, the specialization often varies between clusters, even if similarities and interdependencies prevail across clusters belonging to the same general type of production. In certain Swedish regions, the emphasis is on musical 11

For instance, Sahara Hotnights was ranked among the top ten best released records in 2002 by Washington Post. 12 The Cheiron collective was dissolved in August 2000 and the members preferred to continue on an individual basis. Its founding father, Dennis Pop, was instrumental in developing this line of the Swedish music industry.

394

P. Braunerhjelm

performance and the formation of music groups, while in other regions, the strength of the music cluster is found in other segments. The Stockholm region – on which we have chosen to focus our analysis – seems to be the predominant Swedish music cluster (Hallencreutz et al. 2004). The majority of internationally successful composers and producers in the last few years are based in Stockholm. In order to explain why Stockholm has become so strong in the international music market, we must, however, start by studying the more general conditions for music production in Sweden.

21.4.1 A Broad Basis of Knowledge Base: Schooling, Language and Technological Know-how Some basic conditions for the formation of a strong Swedish music industry emerge from the interview material. First, most respondents agree on the municipal school of music being an important explanatory factor of the relatively high music competence in Sweden. The municipal school of music offered pupils the opportunity to get in touch with music without cost or time-consuming transports. Moreover, musical instruments were supplied free of charge, which meant that all pupils could devote themselves to music, irrespective of their financial situation. The municipal school of music created a ‘‘receiver competence’’. The implications included not only a relatively high knowledge of music among the Swedish audience, but also good conditions for a large inflow of new agents (musicians, producers, composers, etc.) into the Swedish music market – the opportunity space was widened. Hence, a good basis for a broad, general basis of knowledge in music was built up which, according to the interviewees, is necessary to create an internationally competitive music industry. In essence, a kind of ‘‘competent suppliers and demanding clients’ relation’’ was created in the Swedish music industry. The municipal school of music has now become a broader ‘‘Culture school’’. The difference as compared to the previous system is that there is now a fee for the education (varying between municipalities) and also that it is not necessarily located in the children’s own schools. This means that the children’s musical studies now depend more on the financial situation of the family and also the pupil’s active choice. In other words, the students’ access to music training has deteriorated. Moreover, music training is now competing with other creative activities since the ‘‘Culture school’’ also teaches dancing, theater, art, etc. An active choice (i.e. fees) does not necessarily imply a negative effect on the Swedish music industry. On the contrary, it might mean that particularly motivated and interested pupils are attracted to the schools of music.13 However, several respondents claim that they do see the effects of this policy, mostly through a 13

According to the ‘‘Culture School’’, the costs have not increased, nor has the availability decreased. This is, however, contradicted by both the respondents and earlier studies (Forss 1999).

21

The Genesis and Evolution of the Stockholm Music Cluster

395

lower musical competence among applicants for specialized and higher musical studies. Two additional explanations considered to be particular in the Swedish context should be added to a broad musical knowledge basis. First, the respondents stress that the generally good level of knowledge of English in Sweden constitutes an important partial explanation to the Swedish success on the international music market. The combination of the English language being taught at school already at an early stage, television programs and films that are not subtitled and the fact that Sweden is a small country easily subject to international influences, is considered to be the underlying cause of Swedes’ skills in the English language. English is, in turn, an important ‘‘input’’ in an industry strongly influenced by the USA and UK. Second, Swedes are considered to have a great interest in technology and possess considerable technological skills. The music industry has, in many ways, undergone a technological revolution, both with respect to recording techniques and also in general through the swift development of information and communication technology (ICT). The combination of technical knowledge and falling prices on technology (throughout the 1990s) means that accessibility has increased; almost everyone can record their own CD at home without any major financial sacrifice. Note the similarities to Ireland; The Irish experienced a technical revolution, spearheaded by the information and communication sectors, and the English language. Technical competence also implies that the musical quality can be further emphasized, and that more advanced and refined recording technologies can be used in the music production. Naturally, this is an advantage when the final product reaches the consumer. Hence, according to the respondents, there is a connection between high technical knowledge and musical success. The drawback of the development of ICT is pirate copying and free downloading. The problem often emphasized in knowledge-intensive production (as stressed in the ‘‘new economy’’ literature) obviously exists in the music cluster. That is, the costs for developing and marketing a new product is substantial, while the marginal cost for copying the new product is virtually zero (Quah 1999; Alexander 2002).14 The typical example is a CD. Swedish record companies work actively against pirate copying together within the trade association IFPI (The International Federation of the Phonographic Industry) as well as through a dialogue with both Swedish and European legislators. The old copyright system was based on regulations of the copying of pictures. This system was not considered to be fully applicable to music and a new legislation was introduced in Sweden in 2002 and 2003.15

14

See Hui and Png (2003) and Varian (2000). The protection for originators is now regulated through the recent law on Elektronisk handel (Electronic Trade) (Government bill 2002) and through the EU framework directive On the Harmonisation of Certain Aspects of Copyright and Related Rights in the Information Society (EU 2001). 15

396

P. Braunerhjelm

21.4.2 Culture, ‘‘Path Dependence’’ and the Market A generic component – ‘‘path dependence’’ – is often mentioned in the development of clusters. Silicon Valley builds on an electronic knowledge which goes back to the 1930s, by which other industries’ origin and history can be traced (Klepper 2004). As regards the effect of the Swedish national cultural heritage on Swedish music, the answers point in somewhat different directions. Some consider the Swedish connection to folk music and the general European connection to classical music to have strongly contributed to the ability to write catching tunes. In contrast, some consider the culture to be weak, since Sweden is a small country easily influenced by international trends, implying that artistic inspiration mainly stems from west Europe. That is, the interview suggests that Sweden lacks a cultural identity of its own. This might, on the other hand, mean that trends are easily picked up and copied, but that the ability to create unique products is weaker. Overall, the interviews show that Sweden is strong in mainstream production, but does not participate in the leading edge development of music. Even if there is some variation in the opinions about the cultural heritage, there is substantial agreement on the positive influence of immigration on the music life. Previous studies on creative environments and clusters have also emphasized the importance of multiethnic environments (Saxenian 1999; Florida 2002). In this light, the Swedish liberal view on immigration can be considered to be positive and beneficial to the development of the Swedish music industry and the Stockholm music cluster. Furthermore, it emerges from the interviews that Sweden is considered a ‘‘pop country’’ rather than a ‘‘rock country’’. This is most likely partly due to the fact that Europe has traditionally received its musical inspiration from classical music, while the USA has been more influenced by jazz and blues, partly due to the absence of live stages. In the USA, bands can be on a ‘‘never ending’’ tour for several years in order to create an audience.16 In Sweden, there are only a few performances, which means that the resources are concentrated to composition and professional productions, rather than artistry. The studio thus becomes the creative forum for the artist. The somewhat less polished rock music is instead often compensated by more frequent and spectacular performances.17 A general view among the respondents is also that the size of the Swedish music market is so limited that exports become a necessity. Launching an artist is a very costly venture. Projects only targeting the Swedish market are difficult to motivate financially. Thus, it is necessary to also target other markets, which means exports and the adaptation of music to markets with a different cultural basis. The firm

16

As music production become less profitable, there has been a tendency for an increase in price of live concerts tickets. Krueger (2002) claims that the price has doubled between 1995 and 2001. A similar tendency can be observed in Sweden, judging from income data from concert organizers, which is estimated to have increased throughout the period 1996–2002 (Forss 2003). 17 This is also claimed to be the case for Ireland (Clancy and Twomey 1997).

21

The Genesis and Evolution of the Stockholm Music Cluster

397

Stockholm Records is a good example; their central business idea is to export Swedish music. About 20–25% of the employees in the firm only work with other countries and 90% of the firm’s sales take place outside the Swedish borders.

21.5

Stockholm: Concentration and Dynamics

As discussed above, a broad musical knowledge – together with a considerable knowledge of languages and technological skills – constitute the core competencies in the Swedish music industry. In addition, a combination of the musical cultural heritage, multiethnic impulses, and a limited domestic market, has spurred internationalization. With this breeding-ground, Stockholm appears to have evolved as the dominating cluster for the segments of the Swedish music industry we have studied. According to the respondents, Stockholm is the only real alternative for agents wanting to succeed in the Swedish music industry. Hence, even if leading individuals in the Stockholm music cluster come from outside Stockholm, there seems to be strong centripetal forces attracting talent as well as other important cluster agents.

21.5.1 Proximity and Linkages First of all, the so-called major record companies – Sony BMG (a joint venture between Japanese Sony and German Bertelsmann), Universal (owned by Vivendi, France), Warner Music (USA) and the EMI Group Plc. (Britain) – are all represented in Stockholm. These companies hold about 80% of the world market and about the same part of the Swedish market. Their representatives in Stockholm are thus particularly important, since they constitute important links to both the Swedish and the international market. Second, the Stockholm dominance in the music industry is related to the strong dependence on a close co-operation with the media. Most Swedish and international newspapers are represented in Stockholm, and the proximity to TV and radio is also important. The music television channels Z-TV and MT as well as the major radio stations, many of them nation-wide are based in Stockholm. Music videos are also important for launching artists and the central video producers/companies are located in Stockholm. In fact, Stockholm displays several essential characteristics for the emergence of a cluster. It is sufficiently small for a high degree of connectivity, ‘‘everybody knows everybody’’, which earlier studies have also shown to be a requirement for successful, more culturally oriented, clusters (Hirsch 1972; Scott 1999). At the same time, it is sufficiently large to gather the agents of importance for the existence and the development of the cluster. This means that important cluster agents know that they are likely to be working together on several occasions and they know the

398

P. Braunerhjelm

relevant agents within the cluster. Since the network is geographically limited and does not contain too large a number of agents, rumors spread quickly and reputational capital (Scherer 1986) is important in this industry. One reason why ‘‘everybody knows everybody’’ is that the respective agents often hold many and various roles and positions within the cluster, i.e. multitasking seems to be frequent within the cluster. For example, according to the interviewees, it is common for an individual to teach part time at a musical senior high school and/or at an academy of music, and then also be a professional musician. Furthermore, it is not unusual for someone to work at a record company in daytime and compose/produce music in his own studio in his spare time. There are no clear borders between the formal and the informal, between the private and the public/ professional or between a friend, colleague and competitor. This helps create an open atmosphere, where proximity fosters great trust between agents. Despite stiff competition between firms, they thus help and inform each other about ongoing projects. These are typical characteristics of dynamic clusters – competition and cooperation are said to be the driving forces of development (Saxenian 1994). Most of the important educational institutions are also found in Stockholm and there is a tendency for these to profile themselves by focusing on different niches. The two principal musical senior high schools, So¨dra Latin and Rytmus, have consciously created a certain profile. So¨dra Latin has mainly focused its musical program on jazz and classical music, even if there is pop and rock as well. Rytmus has created its profile by mainly devoting itself to modern music, rock, soul, etc. The Royal College of Music, Stockholm (KMH) has a tradition in classical music, but has also opened up to other trends such as jazz, Afro-American, etc., even if the pop and rock culture is generally weak at KMH.

21.5.2 Cluster Dynamics and Diffusion of Knowledge As compared to, e.g. the USA and the UK, the music cluster in Stockholm is said to be characterized by openness and a lack of prestige. International agents are often surprised that agents are so easily accessible. Getting in touch with a celebrity composer/producer is no problem.18 In the USA or UK, it is virtually impossible to have personal contact with such ‘‘superstars’’. Another example of lack of prestige is the informal situation claimed to exist in studio work. Stockholm is not characterized by such clear-cut hierarchies as the USA, where the work is often carried out in very large studios, each agent having a specific place in the production. In Stockholm, the situation is very informal, the individual with the best knowledge for certain tasks gets to do that particular job (a spontaneous best practice organization), even if she is not formally trained for that 18

An example is Max Martin (Martin Sandberg), one of the most successful Swedish agents. He has composed/produced music for artists such as Britney Spears and Backstreet Boys.

21

The Genesis and Evolution of the Stockholm Music Cluster

399

task or even connected to the project. There is, e.g. nothing to prevent a musician from helping out with the production or a studio technician from making a musical contribution when there is a recording. International agents are often surprised when they experience the informal atmosphere in the Stockholm 12-m2 studios, where world-wide hits are created.19 As has been mentioned, the diffusion of knowledge partly takes place through individual movements between different parts of the cluster. There is always an intensive interaction between composers/producers and record companies, a strong interdependency between these two groups of agents. Knowledge of what takes place in the Stockholm music industry does, e.g. spreads to educational institutions through teachers also working as professional musicians. Teachers also move between educational institutes and commercial activities. At the educational institutions, knowledge is, not only, transmitted through the interaction between teacher and pupil, but also from pupil to pupil.20 Learning is partly characterized by tacitness, and transmitted through the interaction among individuals. Teaching and the creation of knowledge are thus developed at a local level, even though the impulses might be international. In the last decade, the Stockholm cluster has also created an increasing number of contact areas with the international music industry. Obviously, it is very difficult to point to the exact contribution of the respective agent in the cluster, but it seems as if the music houses have been particularly successful. It is mainly the music houses that have conveyed Swedish music to the international market, and most international contact areas seem to be found here. One respondent goes so far as to claim that music houses constitute the Swedish ‘‘music miracle’’. An example of the impact of music houses is the so-called ‘‘co-writing’’ activities. Music houses match their own composers/producers with corresponding agents in other countries. These form ‘‘song-writing teams’’ with the task of composing and producing music for various artists. Another task for music houses is to ‘‘pitch’’ songs for international artists. The international artists’ record companies send out a query to some hundred music houses (naturally, the number varies depending on the artist, the record company, etc.) for material for an artist’s new CD. The music houses immediately put a number of composers/producers to this task. The results are then sent to the artist’s record company which – together with the artist – finally decides what songs will be included on the CD. This trend towards more international contacts has also contributed to a more professional Swedish music industry. This is mirrored by the ‘‘fragmentation’’, or vertical specialization, that has been a conspicuous feature of the changing organization of the industry in the last 10–15 years. Services which were earlier found within one large firm have resulted in spin-offs and outsourcing, which more generally

19

At the same time, Stockholm is naturally not exempt from rivalry and conflicts among agents. This is, however, considered to be less widespread than in many other places. 20 See, e.g. Maskell et al. (1998) and Maskell and Malmberg (1999) for a description of local mechanisms for the diffusion of knowledge.

400

P. Braunerhjelm

reflect how the international music industry has been restructured (Hesmondhalgh 1998; Scott 1999).21 But there has also been a change in the sense that more people with an academic degree are now hired by record companies and more attention is also paid to marketing. In the past, many records were released by the record companies more or less on speculation. The business was more characterized by intuition and chance. Today, this is combined with extensive marketing knowledge before a record is released.22 The interviewees also indicate that composers/producers have become better at combining artistic integrity and entrepreneurship. When the contradiction between these can be bridged, the result is often said to be particularly successful. It seems that several agents have succeeded in this, while at the same time created a niche for a certain style of popular music.23

21.5.3 Degree of Competition Above we mentioned the dichotomy between competition and co-operation as a function of the informal character of the cluster. Firms are exposed to competition at both the international and the national level. Competition is fierce when composing/ producing music for an international well-known artist. It is of no importance whether this is done in Sweden or in the UK, in France or USA; there is competition with each agent on the international market. As regards production, the majority of the respondents claim geographical distance to be of no major importance. The same musical production can be carried out in parallel in two places in the world or more, by sending the music as sound files on the Internet. In other words, even if learning is local and concentrated to clusters, once the production phase has been entered, it can be decentralized and spread to other regions. At the local level, competition varies between being intensive around new contacts and production to turning into co-operation when an agent has appeared as a winner in a major project. Other parts of the music industry work under different conditions. Music for advertisements is usually targeted at a specific market in a region, city or country. It is a common practice for the advertising agency to contact a local composer/ producer for the music to be used in the advertising film. In particular, this is due 21

Note that there is an interesting parallel to other industries, e.g. the pharmaceutical industry where ‘‘big pharma’’ (the large companies) co-exist and complement smaller – often ‘‘drug discovery’’ – firms. 22 Another sign is the increased use of artist and repertoire (A&R) staff by record companies, responsible for the image, activities and career development of the artist. A&R is claimed to correspond to R&D in other industries (Clancy and Twomey 1997). 23 In particular, this seems to be the case with Maratone, The Location and A-Side (ex-members of Cheiron) as well as composers and producers connected with, e.g. the music houses Murlyn Songs AB and Tom Bone Music.

21

The Genesis and Evolution of the Stockholm Music Cluster

401

to the fact that different markets appreciate different kinds of music, depending on the underlying culture. Competition is thus mainly local. Competition has also opened a window of opportunity for less established Swedish composers/producers at the national level. The major Swedish composers choose to sell their productions where they can maximize their revenue, i.e. on the international market, which has created a tendency to a ‘‘lack of songs’’ for Swedish artists. There is thus room for less established Swedish song-writers to enter and possibly establish a position on the Swedish market. To summarize this section, we have identified the presence of pecuniary linkages related to suppliers (backward linkages) and customers (forward linkages), as well as non-pecuniary linkages (knowledge spillovers), which have played a vital role in strengthening and expanding the Stockholm music cluster. This is also what is predicted by economic models of clusters and agglomeration. Compared to for instance Ireland, the support structure seems considerably stronger in the Swedish case (Clancy and Twomey 1997). Moreover, the story outlined above confirms the old saying of ‘‘success breeding success’’, attracting entrance of new agents to the cluster. The Swedish ‘‘superstars’’ serve as role models for a new and younger generation of entrants on the Swedish music market. The respondents testify to having been very inspired themselves by the international success of Swedish artists and composers/producers in the last few years. This success has also made it easier for Swedish record companies to sell Swedish artists abroad. Both record companies and composers/producers confirm that the success has created an increased international interest in Swedish music. Self-confidence characterizes the music industry, and will most likely further increase the success by the blossoming up of previously less established agents. This also supports a generic development, path-dependence, often characterizing dynamic clusters.

21.6

Future Prospects for Regional Music Clusters

The ‘‘new economy’’ proponents in the 1990s no doubt exaggerated the extent to which the breakthroughs in the information- and communication technologies (ICT) could be expected to reshape the ways economies works. Still, certain sectors and industries are likely to be more influenced by the ICT progress than other industries. In particular, information- or ‘‘weightless’’ goods, i.e. goods where Internet provides a new channel to market and sell goods (and even partly produce goods), belongs to these industries (Quah 1999; Varian 2000). Music goes a long way in fulfilling these requirements. As a consequence, repercussions could be expected that relate to market structure, industrial organization and location. Information (and entertainment) goods – such as music – are characterized by asymmetries implying that the buyer cannot tell whether a good corresponds to expectations (preferences) before the good is consumed. If the consumer has access to the good without paying for it, this may considerably reduce revenues for music

402

P. Braunerhjelm

selling firms. First, because free access in general reduces the willingness to pay for a good even though preferences for owning a record has been reported high Frostling-Henningsson and Jacobsson (2003). Second, the potential customer may after listening to the record decide that it did not live up to expectations and therefore abstain from buying the record. Irrespective of the welfare gain for the individual – which of course is significant – this will cause a problem for the music industry in terms of declining demand and low willingness to pay. Hence, there are some clearly discernible threats to the music industry. Even though some claim that piracy and free downloading of music may have a beneficial effect on total demand through ‘‘information’’ externalities, empirical evidence points in a different direction (Hui and Png 2003). In the case of Sweden a recent study involving a total of 1,150 student at a Swedish and a US university found that the overwhelming majority of students downloaded music regularly (Gallaway and Kinnear (2001); Hellmer (2003). About 25% of the students had more than 1,000 music files on their computers. More interesting is that only 4% of the Swedish students and 7% of the US students reported that their purchase of music had increased as a consequence of downloading. In addition, even though 50% and 60% of the students in Sweden and the USA, respectively, claimed that they were prepared to pay for downloading, but only at prices considerably lower then current market prices. For newly released music, the revealed prices were in the range of US $0.30–0.80, while older music (six months and older) was valued between US $0.15 and 0.30.24 Compared to prices in stores – normally in the range US $10–20 – the concerns of the record selling business is easily understood. The potential threat that the record companies foresee have also been realized in terms of staggering demand, shrinking sales and lower profits. As a response to downloading, a number of legal procedures have been initiated, so far with mixed success. In addition, a restructuring of the market can be discerned. The major record companies aim at integrating, or form strategic alliances, with novel agents at the music market such as downloading operators and cell phone operators. At the same time steps towards increased collaboration and mergers have been taken; the ‘‘major five’’ – which have dominated the market since the 1920s – are now reduced to a ‘‘major four’’ and may soon shrink to a ‘‘major three’’. What that implies for the market, or the independent record companies, is at present hard to assess. The increased concentration does not necessarily imply tougher entry barriers, or higher prices, since Internet tend to mitigate such tendencies. Prices on CDs have declined 24

Preferences seem roughly to be in line with the price iTune (Apple) – which has roughly 70% of the market charge for downloading music, i.e. US $0.99. Competition is however increasing; Loudeye (works with Microsoft), MusicMatch, Napster and Real Networks are other actors. Still, even though iTune is limited in geographical scope, only 14 million songs were downloaded in a year as compared to the free access website Kaaza from which 700 million songs where downloaded. Total on-line sales is estimated to be 1% of the market, but expected to increase to 20–25% in 4–5 years (Financial Times 2003b). Music industry is estimated to have lost about US $3.4 billion due to illegal copying in 2002. The major record producing companies are therefore involved in restructuring, vertical integration, introducing online sales through various means, etc. (Financial Times 2003a).

21

The Genesis and Evolution of the Stockholm Music Cluster

403

since at least 2000 and the revenue share stemming from CD sales is dwindling. Rather these moves are made to secure sufficient scale in the operations of the record companies. Simultaneously firms try to cut costs by shedding labor and new firms enter the market, offering differentiated distribution means and new types of customer contracts. There seems to be a search for a new, or at least altered, business model somewhat similar to what happened to the film industry when the videotape recorder was launched.

21.7

Conclusions

Stockholm has established itself as an internationally leading cluster for composing and producing music in a relatively short period of time. Despite the recent emergence of the cluster, the success of today builds on an accumulation of knowledge which can be traced several decades back. There is a strong music tradition comprising of folk music, choir-singing but more popular music in Sweden, which has, to some extent, been reinforced by the activities in the municipal school of music.25 Yet, other mechanisms have been instrumental in visualizing those opportunities to the economic agents – the entrepreneurs – that have led to a commercial exploitation in this line of business. Still, rather than pinpointing one or two specific factors, the results suggest that a number of different factors – where some have developed over a long period of time – interacted in time and space to create the Stockholm music cluster. The most prominent among these were a broad basis of knowledge, including interdependency with related technologies, language skills, international exposure and role models. Once a critical mass had been attained, well-known centripetal forces were set in motion to strengthen and expand the Stockholm music cluster. Key actors in music clusters started to co-locate in Stockholm. The major international record companies are all represented in Stockholm and constitute important channels to both the Swedish and the international music market. Also international and nation-wide media firms have gathered in Stockholm. Close contacts with these are a necessary, though not sufficient, condition for success in the music industry. Moreover, the most important video producers/video production companies have representatives in Stockholm. The music cluster in Stockholm thus shows many typical characteristics of dynamic clusters: the ability for renewal, a large inflow of new agents, market experiments, and strong linkages between demanding customers and advanced producers of music. Together with strong international and national competition, paired with co-operation, as well as close and informal contacts, this has served to shape the cognitive dimension of the cluster. The music industry (as well as other 25

For example, about 175,000 adult education music circles are formed each year.

404

P. Braunerhjelm

kinds of cultural business) seems to be good at integrating and using the competence and resources of immigrants, which creates new impulses and serves as a driving force for the development within the cluster. The findings also correspond to previous conclusions as regards the embeddedness of industrial production into a social and institutional context, trust being one important component (Lee and Willis 1997; Autio and Yli-Renko 1998; Barnes and Gertler 1999). But there is also a distinct economic dimension to it. ABBA’s success since the 1970s seems to constitute a kind of ‘‘igniting spark’’ in this process, followed by ‘‘superstars’’ as Roxette, Ace of Bace and Cardigans.26 This created a basis for the emergence of the Swedish music industry, artistically as well as economically. The possibility to transfer ownership and income to corporations registered outside Sweden has contributed to strong private economic incentives to invest in – and enter – the music industry. Considering that the 25 top producer and composer companies in Sweden made a profit of more then 60% in relation turnover, there are obvious economic incentives to enter into the music industry.27 Rewards of some magnitude are necessary for the exploitation of future opportunities and continued successful entrepreneurship in the music industry.28 But that refers more generally to economic policies conducted within a country and should not warrant any specific incentives for the music industry. In terms of the institutional and organizational dimension, firms and individuals began to co-locate as a critical mass could be observed and due to the localized nature of knowledge and learning. Thus, the Stockholm music cluster entered a phase of endogenous agglomeration and growth. At the same time, specialization increased, followed by a vertical disintegration in the music industry. Many agents (educational institutions, composers/song-writers and record companies) seem to have developed different niches in one way or another. The initial ability of the Stockholm music cluster to produce new and successful music groups shifted towards more technology intensive and competence demanding segments (production and composing). Still, there are a few factors threatening the future role of the Stockholm music cluster. First of all, the strong international exposure of the industry makes it possible for successful and established agents to relatively easily reallocate their activities to other countries if the Swedish conditions were to deteriorate. The intercountry/region comparison revealed that there are competing location sites for agents in the Swedish music industry. In particular, this applies to ‘‘superstars’’ that are important for the long-term survival of the cluster. They remain in Sweden partly because they have the possibility to redirect their income and taxes to other countries. This should also be seen in the light of the positive inter-industry

26

Stig Andersson, Abba’s former manager, played a vital role in the group’s commercial success. Later on he initiated the well-known Polar prize, also referred to as the ‘‘Nobel’’ prize in music. 27 The distribution is quite skewed within the group of 25 companies. One firm accounts for 26% of turnover and 39% of profits (Veckans affa¨rer 2003). 28 The failure rate is much higher in the music industry than in most other industries (Negus 1992).

21

The Genesis and Evolution of the Stockholm Music Cluster

405

externalities that may stem from cultural clusters and positively influence location of other industries (Durkan 1994). Second, the future supply of knowledge seems uncertain. The exact extent of this problem cannot be determined on basis of the material, but the majority of respondents have expressed some fear that competence is being weakened.

Acknowledgements Constructive comments from David Audretsch, Magnus Henrekson and participants at a seminar held at Indiana University, have helped to improve the manuscript. Financial support from Vinnova is gratefully acknowledged.

Appendix Interviews categorized by sectors and individuals: l

l

l

29

Educational institutions – Principal Gunilla von Bahr at the Royal College of Music, Stockholm (KMH), Deputy Rectors Marie Linde and Gunnar Andersson, in charge of musical studies at So¨dra Latins Musikgymnasium (public high ¨ sterling at Rytmus Musikgymnasium (private school), Principal Johanna O high school)29 and the head of the cultural schools (primary and secondary level), Hans Skoglund at Stockholms Stads Kulturskola. Composers/producers and music houses30 – Andreas Grill, Christian Sandqvist, Andreas Claeson, Peter Hallstro¨m, Martin von Schmalensee, Mattias Ha˚kansson, Ben Male´n and Pelle Lidell.31 Record companies – President Niklas Nyman at Music Network (MNW), Vice President Eric Hasselqvist at Stockholm Records and President Per Sundin at Sony Music Nordic.

So¨dra Latins Musikgymnasium and Rytmus Musikgymnasium are senior high schools in Stockholm, specializing in musical studies. 30 The difference between record companies and music houses is that the former own the rights to the recordings while the latter own the rights to the works. This means that record companies sell and market transmitters of sound (CD, vinyl, DVD, etc.), while music houses own and administer the original works/copyrights and license these to record companies, film companies, advertising agencies, etc. Music houses also take care of the originators’ copyrights for existing works that are sold on transmitters, when the work is recorded live, broadcasted on TV, played on the radio, filmed, when music is sold or when texts are printed, etc. 31 Pelle Lidell represents about 40 Swedish composers/producers through the music house Murlyn Songs AB. Ben Male´n (previously under the stage name Ben Marlene) represents about 50 Swedish composers/producers at the music house Tom Bone Music.

406

P. Braunerhjelm

References Alexander PJ (2002) Peer-to-peer file sharing: the case of the music recording industry. Rev Ind Organ 20:151–161 Allansdottir A, Bonaccorsi A, Gambardella A, Mariani M, Orsenigo L, Pamolli F , Riccaboni M (2002) Innovation and competitiveness in European biotechnology. Enterprise Papers 7, Enterprise Directorate General, European Commission Arthur B (1989) Competing technologies, increasing returns, and lock-in by historical events. Econ J 99:116–131 Arthur B, Ermoliev M, Kaniovski M (1987) Adaptive growth processes modeled by Urn schemes. Kibernetika 6:49–57 Autio E, Yli-Renko H (1998) The network embeddedness of new technology-based firms: developing a systemic evolutionary model. Small Bus Econ 11:253–269 Barnes TJ, Gertler MS (eds) (1999) The new industrial geography: regions, regulations and institutions. Routledge, London Borenstein S, Farrell J (eds) (1998) Inside the Pin Factory: Empirical Studies Augmented by Manager Interviews. J Ind Econ 46(2):123–124 Braunerhjelm P, Carlsson B (1999) Industry clusters in Ohio and Sweden, 1975 –1995. Small Bus Econ 12:279–293 Braunerhjelm P, Feldman M (2006) Cluster genesis. Oxford University Press, London Braunerhjelm P, Faini R, Norman V, Ruane F, Seabright P (2000a) Integration and the regions in Europe: how the right policies can prevent polarization? Monitoring European Integration 10. CEPR, London Braunerhjelm P, Carlsson B, Johansson D, Karaomerliouglu D (2000b) The new and the old: the evolution of biomedical and polymer clusters. J Evol Econ 10:471–488 Carlsson B (2004) The music industry clusters in Cleveland. Case Western Reserve University, Cleveland, OH (mimeo) Carlsson B (ed) (2001) Technological systems in the bioindustries. An international study. Kluwer, Boston Carlsson B, Stankiewicz R (1991) On the nature, function and composition of technological systems. J Evol Econ 1:93–118 Clancy P, Twomey M (1997) Clusters in Ireland. The Irish popular music industry: an application of Porter’s cluster analysis. NESC, Dublin Clark L, Feldman M, Gertler M (2000) The Oxford handbook of economic geography. Oxford University Press, New York David P (1985) Understanding the Economics of QWERTY. Am Econ Rev 75:332–337 Durkan J (1994) The economics of the art in Ireland. The Arts Council, Dublin Enright M (1996) Regional clusters and the economic development: a research agenda. In: Staber U, Schaefer N, Sharma B (eds) Business networks: prospects for regional development. Walter de Gruyter, New York EU (2001) 2001/29/EG, On the harmonisation of certain aspects of copyright and related rights in the information society. Accepted on May 22, 2001, and can be found at http://europa.eu.int/ information_society/topics/multi/digital_rights/doc/directive_copyright_en.pdf Feldman M, Francis J (2002) Entrepreneurs and the formation of industrial clusters. University of Toronto, Toronto (mimeo) Financial Times (2003a) Music beats the drum for online sales. 6 June 2003 Financial Times (2003b) Sony plans to merge music arm with BMG. 16 Oct 2003 Florida R (2002) The rise of the creative class. Plersus, New York Forss K (1999) Att ta sig ton – om svensk musikexport 1974–1999. Ds 1999:28, Expertgruppen fo¨r studier i offentlig ekonomi, Finansdepartementet (Ministry of Finance) Forss K (2003) The export of Swedish music industry 2002. Report from Music Sweden Industry, Stockholm

21

The Genesis and Evolution of the Stockholm Music Cluster

407

Frostling-Henningsson M, Jacobsson A (2003) Downloading of music from a consumer’s perspective. Markets in the Information Age. The Swedish Competition Authority, Stockholm Fuijita M, Krugman P, Venables A (1999) The spatial economy. Cities, regions and international trade. MIT, Cambridge, MA Fujita M, Thisse J-F (2002) Economics of agglomeration. Cities, industrial location and regional growth. Cambridge University Press, Cambridge Gallaway T, Kinnear D (2001) Unchained Melodt: a price-discrimination based policy proposal for addressing the MP3 revolution. Rev Ind Organ 35:279–287 Gordon I, McCann P (2000) Industrial clusters: complexes agglomeration and/or social networks. Urban Stud 37:513–532 Government bill (2002) Regeringens proposition 2001/02:150, Lag om elektronisk handel och andra informationssamha¨llets tja¨nster. Stockholm, 14 Mar 2002 (Government bill on Electronic Trade implemented on July 1, 2002), downloadable at http://naring.regeringen.se/propositioner_mm/ propositioner/pdf/p200102_150.pdf Hallencreutz D (2002) Popula¨rmusik, kluster och industriell konkurrenskraft, En ekonomiskgeografisk studie av svensk musikindustri. Doctoral thesis, Department of Social and Economic Geography, Uppsala University Hallencreutz D, Lundequist P, Malmberg A (2004) Popula¨rmusik fra˚n Svedala. Na¨ringspolitiska la¨rdomar av det svenska musikundrets framva¨xt. SNS, Stockholm Hellmer S (2003) Will music file sharing over the internet influence the commercial market for pre-recorded music? Markets in the Information Age. The Swedish Competition Authority, Stockholm Hesmondhalgh B (1998) The British dance music industry: a case study of independent cultural production. Br J Sociol 49:234–251 Hirsch PM (1972) Processing fads and fashions: an organization-set analysis of cultural industry systems. Am J Sociol 77:639–659 Hui KL, Png IPL (2003) Piracy and legitimate demand for recorded music. B E J Econ Anal Policy 2:1–30. http://www.comp.nus.edu.sg/lu Klepper S (2004) The geography of organizational knowledge. Carnegie Mellon University, Pittsburg (mimeo) Krueger A (2002) Music sales slump, concert ticket costs jump and rock fans pay the price. The New York Times, 17 Oct 2002 Krugman P (1991) Geography and trade. MIT, Cambridge, MA Lee R, Willis J (1997) Geographics of economies. Arnold, London Lofthus KR (2002) Cited in Expressen, 16 Mar 2002 Article by Lindwall JT and Nordstro¨m A Lundequist P (2002) Spatial clustering and industrial competitiveness – studies in economic geography. Dissertation No. 48, Department of Social and Economic Geography, Uppsala University, Uppsala Maggioni M (2006) Mors tua, vita mea? The rise and fall of innovative industrial clusters. In: Braunerhjelm F (ed) Cluster genesis. Oxford University Press, Oxford Martin P, Ottaviano G (1999) Growing locations: industry location in a model of endogenous growth. Eur Econ Rev 43:281–302 Martin P, Ottaviano G (2001) Growth and agglomeration. Int Econ Rev 42(4):947–968 Maskell P, Malmberg A (1999) Localized learning and industrial competitiveness. Cambridge J Econ 23:167–185 Maskell P, Eskelinen H, Hannibalsson I, Malmberg A, Vatne E (1998) Competitiveness, localized learning and regional development: specialization and prosperity in small open economies. Routledge, London Music Board of Ireland (2002) The economic significance of the Irish music industry. Music Board of Ireland, Dublin Negus K (1992) Producing pop: conflicts and culture in the popular music industry. Edward Arnold, London

408

P. Braunerhjelm

Nelson R (1993) National systems of innovations. A comparative analysis. Oxford University Press, Oxford OECD (2004) OECD statistical database via Internet Porter ME (1990) The competitive advantage of nations. MacMillan, London Porter ME (2000) Location, competition, and economic development: local clusters in a global economy. Econ Dev Q 14:15–34 Quah D (1999) The weightless economy in growth. Bus Econ 30:40–53 Saxenian A (1994) Regional networks: industrial adaptation in Silicon Valley and route 128. Harvard University Press, Cambridge, MA Saxenian A (1999) Silicon Valley’s new immigrant entrepreneur. Public Policy Institute of California, San Francisco Scherer F (1986) On the current state of knowledge in industrial organization. In: De Jong H, Shephard W (eds) Mainstreams in industrial organization. Martinus Nijhoff, Dordrecht Scott A (1999) The US recorded music industry: on the relations between organizations, location and creativity in the cultural economy. Environ Plann 31:1965–1984 Scott A (2003) Origins and growth of the Hollywood motion-picture industry: the first three decades. UCLA, Los Angeles (mimeo) Scott A, Storper M (1986) Production, work, territory: the geographical anatomy of industrial capitalism. Unwin Hyman, London Statistics Sweden (2002) Data from the central register of enterprises and establishments, 1975–2002. Statistics Sweden, Stockholm STIM Annual report, 1987–1997. Swedish Performing Rights Society, Stockholm UNESCO (1998) World culture report. UNESCO, Paris U.S. Bureau of the Census (2001) Census of Industrial Production, Washington, D.C., USA Varian H (2000) Buying, sharing and renting information goods. J Ind Econ 48:473–488 Veckans affa¨rer (2003) Sa˚ ska det la˚ta 19, pp 24–26 Venables A (1996) Equilibrium locations of vertically linked industries. Int Econ Rev 37:37–72 Zucker L, Darby M (1996) Star scientists and institutional transformation patterns of invention and innovation in the formation of the biotechnology industry. Proc Natl Acad Sci 93:12709–12716

Index

A Academy-internal social capital, 131 Accessibility, 10–12, 14, 16, 17, 19, 22, 23, 25, 41, 181, 268, 269, 271, 276–279, 284–286, 326–328, 395 Accessibility measures, 278, 284, 286, 326 Advanced market economies, 94 Aesthetic solutions, 83 Agglomeration, 11, 14, 19, 60–62, 77, 122– 124, 225, 274, 275, 292, 293, 305, 318, 319, 329, 348, 383, 385, 387, 401, 404 Agglomeration economies, 14, 19, 60, 61, 122, 124, 225, 292, 293, 305, 318, 319, 329 Agglomeration forces, 60–62 Allochthonous, 254 Angst, 84 Applied policy analysis, 187 Approximate Entropy, 143–158 Archival information, 101 ARFIMAX model, 200, 202, 203, 212, 213 ARIMA model, 200, 211, 213 Artisanal models, 303, 305, 311 Artistically creative personalities, 82–84 Asymmetric individuals, 53–57 Automated Highway Systems, 156 Auto-Regressive, Fractionally Integrated, Moving Averaged (ARFIMA), 200– 203, 208, 211–213

B Backwash, 319, 329 Benchmark equilibrium, 182 Bibliometric, 351, 352, 357, 364, 365, 368 Bibliometric analysis, 351, 357, 364, 368 Big Government, 103, 306, 307 Bliss Point, 69, 70, 73–76

Braess’s Paradox, 38 Brain power society, 27, 59–63, 78 Bridging function, 233, 235, 237, 240, 242 Brownfield, 117, 118

C Catalyst, 321 Celebrity rents, 94 Centrality, 351, 352, 360, 362–364, 368, 376 Central place system, 13, 23 Centre-periphery pattern, 337 Centrifugal, 60, 61 Centripetal, 60, 397, 403 Centripetal forces, 60, 397, 403 Chennai, 378 Choirs, 390, 403 Citation matching, 340, 341 City bankers, 29, 376, 382, 383 City-based employment, 376–378, 383 City regions, 12, 134, 291, 292, 294, 297, 302, 312, 313 Civil society, 111, 117, 129, 136 Classical music, 89, 91, 93, 390, 396, 398 Client access, 381 Climatic variables, 298, 302, 303, 311 Cluster analysis, 237, 239 Cluster development, 19, 386, 396, 397, 404 Clustering, 267, 273, 274, 332, 361 Clusters, 14, 18–20, 24, 62, 102, 109, 121, 123–125, 128, 129, 136–139, 215, 237, 239–242, 267, 273, 274, 310, 318, 332, 337, 339, 361, 362, 374, 385–405 Cluster theory, 374 Cobb-Douglas, 93, 178, 187–189, 203 Cobb-Douglas type production function, 203

409

410 Co-ethnic clients, 249 Collective learning, 125, 222, 226, 227, 229, 233, 240, 241 Commoditized products, 380 Commodity balance equation, 41 Community power, 117 Community Voice Mail, 118 Compelling mythology, 104 Competition, 2, 9, 24, 35, 42, 61, 91, 99, 107, 109, 122, 180, 189, 273, 305, 318, 373, 374, 376, 378, 385, 387, 393, 398, 400–401, 403 Competitive input factor market, 197 Complex problems, 104, 105, 109 Complex systems, 61, 98, 100, 102, 105, 125, 252 Computational (Kolmogorov) entropy, 145–147, 151–157 Consumption-market potential, 17 Continual improvements, 9 Control matching, 28, 337, 340 b-Convergence, 297 Convex production technologies, 1, 2 Cooperating actors, 137 Co-operation, 226, 227, 268, 387, 397, 400, 403 Core-periphery model, 60, 317, 318, 329 Corporate innovation capitalism, 126 Cost of exchanging knowledge, 270 Creation of clusters, 385 Creation of knowledge, 13, 59, 61–65, 77, 78, 221–243, 272, 331, 399 Creative enforcement, 116 Creativity, 27, 59, 79–94, 103, 227 Creativity-based production systems, 27, 94 Creator of region, 93 C-society, 59 Cultural heritage, 396, 397 Culture, 27, 92, 97–100, 102, 104, 105, 110, 119, 124, 130, 133, 138, 230, 251, 254, 259, 379, 390, 394, 396–398, 401 Culture school, 394 Cumulative process, 15, 18 Customization, 4, 383

D Decentralized, 144, 400 Decentralized decision making, 144 Deskilling, 113 Deterministic interregional I–O model, 164 Development potential, 111, 116 Development trajectory, 4

Index Differential knowledge, 62–66, 74, 75 Diffusion, 4, 5, 7–9, 23, 83, 112, 133, 221–224, 228–230, 341, 352, 361, 365, 387, 398–400 Diffusion of knowledge, 7, 83, 221, 222, 224, 228–230, 341, 361, 365, 387, 398–400 Dispersion forces, 60, 61 Distance-sensitive, 15, 17, 19, 122 Distance-sensitive demand, 17, 19 Distribution of knowledge, 361 DNA, 98 Downsizing, 377 Durable links, 12, 276, 278 Dynamic agglomeration economies, 19, 222, 305 Dynamic competitive advantage, 27, 107–120 Dynamic efficiencies, 117 Dynamic regions, 1–29, 174 Dynamics, 1–29, 59–78, 97–105, 121–139, 262, 366, 368, 386, 387, 397–401 Dynamics of the clusters, 386, 387, 398–401, 403 Dynamic urban competitiveness, 110

E Econometric devices, 292 Economic adjustment processes, 13 Economic knowledge, 5 Economic Theory of Democracy, 115 Educational level, 260, 261 E-linkages, 62 Embedded uncertainty, 84 Embryonic development, 101 Emergence of clusters, 29, 128, 385, 386, 397, 403 EMI Group Plc, 397 Employment decentralization, 293, 313 Employment effects, 178, 193 Endogenous regional specialization, 14, 16 Endogenous specialization, 14–15 Entrepreneurial behavior, 112, 115, 255 Entrepreneurial City, 107–109, 117, 119 Entrepreneurial decision, 110, 111 Entrepreneurial firms, 9 Entrepreneurial knowledge, 5, 6 Entropy, 27, 45, 143–158, 163, 167–170, 172, 173 Entry and exit, 9, 10, 22 Equilibrium, 1, 2, 18, 27, 36, 39, 40, 42, 43, 60, 67–77, 88, 98, 124, 164, 177–190, 197, 291, 313 Equilibrium configuration, 69, 70, 75

Index Equilibrium dynamics, 68–77 Equilibrium path, 67, 71–73, 76, 77 Estimates of the unemployment elasticity, 190–191 Ethnic entrepreneurs, 28, 247–250, 252–255, 258–262 Ethnic entrepreneurship, 248–255, 259, 262 Ethnic firms, 247, 249, 253, 257, 260 European Patent Office, 274, 334, 335, 337, 338 Eurostat REGIO database, 294 Excludability, 270, 271 External economies, 14, 15, 17, 24, 376 Externalities, 14, 15, 27, 61, 62, 77, 114, 122, 123, 125, 128, 225, 271, 272, 274, 275, 331, 336, 341, 402, 405 External linkages, 233, 237, 240, 241 External scale economies, 15, 17, 19 Extra-market externality(s), 27, 122, 128 Extreme durability of great art, 91

F Factor mobility, 27, 178, 180–182, 184–186 Filtering-down theory, 22–24 Financial service, 374–378, 382, 383 Financial service sector, 377 Folksongs, 390 Four dancers, 69, 70 Fragmentation, 399 Franchise organization, 28, 247, 248, 251–255, 257–262 Free downloading, 402 Functional economic region, 10, 307, 319 Functional region, 3, 10, 12–19, 21, 24–26, 271, 278, 279 Functional urban regions, 20–25, 291, 293 Function/process, 373 Fuzzy definitions, 271

G General equilibrium, 1, 2, 27, 60, 177–190 General equilibrium model, 27, 60, 177–190 General equilibrium modeling, 180 Genes, 27, 97–101, 105 Geographic innovation production functions, 348 Geographic spillover, 331 Geographic transaction costs, 19 Global success, 89 Growth regressions, 294

411 H Heckscher–Ohlin framework, 13 Hidden order, 27, 143–157 High-tech patents, 333 Hollowing out, 374, 376 Human capital, 3, 6, 11, 13, 17, 19, 20, 28, 110, 128, 135, 269, 271, 272, 292, 293, 305, 310, 311, 317, 373 Hypothesis, 23, 25, 28, 135, 204, 206, 207, 211, 284, 285, 292, 325, 328, 329, 339, 342

I Immigrant groups, 249 Increasing returns, 2, 14, 27, 60, 178 Incubator function, 223, 224, 231 Industrial districts, 61, 121, 123–125 Industrial economists, 225, 229 Industrialism, 27, 80, 123, 127, 132 Industrial R&D, 28, 85, 267–269, 272–278, 280–286 Influence agglomeration, 387 Information and communication technology (ICT), 240, 242, 350, 379, 387, 395, 401 Infrastructure, 10–13, 16–18, 22, 27, 28, 82, 107, 114, 117, 122, 129, 157, 177, 181, 184–186, 197–218, 227, 271, 275, 276, 278, 368, 379, 383 Infrastructure development, 186 Infrastructure productivity, 197–218 Innovation and product differentiation, 374, 382 Innovations, 1–29, 35, 45–46, 59–78, 80, 94, 109–111, 113, 115, 117–119, 121–133, 135–139, 182, 198, 199, 208, 213, 214, 218, 222, 223, 225–233, 235, 237, 240– 243, 262, 267, 270, 273–275, 286, 293, 309, 313, 332, 342, 347–351, 374, 382, 383, 387 Innovative actors, 225 Innovative capacity, 224, 229, 240, 331 Input flows, 162 Institutional embeddedness, 247–262 Integrated economic system, 10 Integration of markets, 387 Interaction, 4, 5, 10, 11, 14–16, 21, 44, 60, 64, 67, 76, 77, 80, 82, 86, 93, 94, 97–99, 123–125, 128–131, 137, 139, 226, 227, 235, 249, 268, 276, 278, 286, 293, 298, 309, 313, 318, 331, 347–369, 381, 399 Internal market potentials, 14, 15, 17

412 International artists, 393, 399 International music cluster, 393 International trade theory, 53, 385 Interorganizational coordination, 110, 119 Interregional exports, 17 Interregional model, 161, 162 Intraregional labor market, 139 Intuition, 73, 111, 300, 400 Inventions, 5–7, 13, 20, 21, 27, 45, 79–82, 85, 97, 270, 271, 275, 276, 332, 333, 335, 336, 339

J Joint publications, 347, 367

K K-linkages, 61, 62 Knowledge, 4–9, 12, 13, 16, 18, 20, 22–24, 26– 28, 59–66, 68, 69, 71–78, 80, 83, 85, 92, 100, 107, 108, 113, 117–119, 122–125, 127–130, 132–135, 139, 217, 221–243, 247, 260, 261, 267–279, 286, 317, 331– 342, 347–369, 374, 381, 386–388, 394– 401, 403–405 Knowledge creation, 62–65, 77, 78, 221–243 Knowledge flows, 5, 267–272, 274–279, 286, 332, 336, 337, 339, 342 Knowledge networks, 5, 113, 271, 347–351, 355, 357, 359, 363, 368, 369 Knowledge production function, 275 Knowledge resources, 23 Knowledge spillovers, 28, 61, 122, 125, 128, 222, 224, 225, 271, 317, 331–342, 347–349, 387, 401 Kolmogorov entropy, 27, 145–147, 151–157 Kolmogorov’s entropy concept, 146 Kuhn–Tucke, 40

L Labor supply curve, 180 Lagrange multiplier(s), 170, 171, 173 Lagrangian multiplier(s) (LM), 52, 55, 211, 302 Language effect, 89 Lead-lag models, 25–26 Learning, 4, 8, 61, 62, 83, 124, 125, 222, 224– 229, 231, 233, 237, 239–241, 399, 400, 404 Leisure, 136, 138, 183

Index Leisure time, 48, 133 Leontief–Strout, 161, 164, 174 Level of service (LOS), 143–145, 147–154, 156 Life cycle(s), 22, 23, 129, 348, 356 Linkage(s), 3, 4, 19, 21, 27, 59–62, 78, 103, 168, 170, 174, 226, 228, 232, 233, 236, 237, 240, 241, 319, 335, 348, 374, 382, 385–387, 397–398, 401, 403 Localization economies, 14, 15, 18, 24, 225 Localized industry, 35 Local production system, 226 Location advantage(s), 13, 14, 18, 23, 24 Location dynamic(s), 16, 22–25, 117 Lock-ins, 136 Logarithmic utility function, 44, 45

M Macroeconomic management, 107 Management economics, 199 Manufacturing-industrial society, 134 Marginal productivity, 206, 215–217 Marginal productivity of infrastructure, 206, 213–215, 217 Market failure, 114 Market potential(s), 14–18, 318 Market size, 317, 328 Materially interdependent, 119 Measurement(s), 135, 157, 178, 182–183, 197, 199, 208–217, 223, 296, 364, 368 Medical instruments, 392 Meme(s), 27, 97–105 Meme-space, 27, 97, 104 Memetic(s), 100, 102, 104 Micro-model of knowledge creation, 64, 77 Milieu behavior, 240 Milieux, 110, 119, 227 Minimal Darwinism, 98 Ministry of Industry and Trade (MITI), 124, 127 Mobility of skilled labor, 348 Mobility of urban residents, 111 Monetary power, 112 Monodisciplinary nano-fields, 350, 357 Moran’s I, 207, 210 Moroccan entrepreneurs, 249 Multicollinearity, 294, 302 Multiplier(s), 169–173 Multi-regional model(s), 161, 164 Multi-stage processes, 86 Mumbai, 377, 383 Municipal R&D, 277–279, 282–285

Index Music cluster, 29, 385–405 Music industry, 86, 386, 388–397, 399–404

N Nanoscience, 350–355, 357, 359, 368 Nanostructured solar cell(s), 347, 349, 351– 353, 355, 356, 359, 362–365, 368 Nanotechnology, 350, 352, 353, 357, 359 Natural selection, 98, 101 Net migration, 291, 297, 311 Network hub(s), 29, 360–364 Networking, 112, 119, 229–231, 235, 239– 242, 249, 349 Networking behavior, 240 Networking function(s), 233, 235, 237, 240, 242 Network learning, 229 Networks, 2, 36, 92, 108, 122, 144, 161, 177, 227, 249, 269, 347, 387 New economic geography, 27, 59–63, 292, 329 New spatial division of labor, 373 Node(s), 3–5, 10, 20, 21, 112, 117, 122, 125– 128, 131, 271, 278, 336, 347, 351, 352, 355, 357, 358, 360, 361, 363, 364, 367, 369 Novel agents, 402 Numerical experiment(s), 178, 186 Numeric patterns, 147

O Oceanic capacity, 84 Off-shoring, 373–383 OLS estimates, 302 Opportunistic behavior, 228 Organizational transformation, 273 Output capacity, 168, 169, 172–174 Over-the-counter products, 380

P Pandora’s box, 77 Pareto-optimality, 97 Passive learning behavior, 241 Patent explosion, 127 Patenting, 85, 126, 275, 332–335, 337, 341, 342, 348, 358 Path-dependent risks, 21 People power, 112 Peripherality, 310 Platform-building, 138

413 Policy entrepreneur(s), 104 Population size, 15, 16, 305, 322, 323 Porterian principle of competitive advantage, 382 Price discrimination, 42 Probabilistic model(s), 168, 174 Probabilities of success, 86, 88 Probability of failure, 89 Processes of internationalization, 382 Product cycle(s), 22–26 Product innovation(s), 7–9, 21, 233, 235, 237, 382 Production methods, 6, 7, 9 Production patterns, 21 Proportional disaggregation, 375 Proximity, 11, 15, 19, 118, 123, 124, 221–226, 268, 270, 271, 274–276, 286, 293, 300, 302, 349, 379, 387, 397–398 Psychology of creativity, 83 Public policy, 97, 98, 101, 102, 104, 115, 123, 136–139

R Radical liberalization, 377 Random mutation, 100 Randstat, 296 Rank size distribution, 87 R&D function(s), 229, 230, 268, 285 Reagan administration, 103 Receiver competence, 394 Regional final demand, 174 Regional innovation systems (RIS), 124, 125, 128, 138, 139 Regional policy, 221 Regional wage-fixing curve, 180 Regression analysis, 88, 90, 198, 326, 327, 367, 369 Relation-building, 138 Relative capacity utilization, 168 Renewal processes, 21 RNA, 98 Routine processing, 377 Routine support functions, 379

S Scale economies, 15, 17, 19, 87, 197, 374, 379, 380, 385 Schumpeter, 2, 8, 9, 109, 115, 119, 121, 137, 270 Science and Research Parks, 223 Science-based technology fields, 347

414 Science Citation Index (SCI), 351–353, 355, 359, 361, 366, 368 Science output, 28, 275, 362, 364, 368 Science parks, 28, 221–243 Science, technology and innovation (STI), 349 Scientific explanation, 102 Second-order condition, 73 Seedbed function, 223, 242 Self-reinforcing processes, 16, 17 Small and medium enterprises (SMEs), 108, 123, 125, 233, 247, 248, 252, 253 Smith-Ricardo argument, 51 Smith-Ricardo paradigm, 26, 47, 48 Social capital, 27, 112, 121–139, 217 Social efficiency, 116 Social entrepreneurs, 110–116, 118, 119 Socialized processes, 228–230 Social links, 128–131 Social network analysis (SNA), 351, 359, 360, 368 Social safety, 133 Societal conditions, 135 Socioeconomic coordination, 109 Solar cell, 28, 347–369 Sony BMG, 397 Spatial adjustment processes, 291, 303 Spatial dependence, 292, 293, 300, 302, 303, 309, 311, 313 Spatial diffusion, 23, 221 Spatial extent, 10, 11, 331, 333 Spatial markets, 26, 35–46 Spatial pattern of innovation, 309 Spatial preferences, 3 Spatial spillover effect, 198, 210, 217 Spatial transaction costs, 14–15, 24 Specialization clusters, 14, 19, 393 Spillover effects, 198, 210, 217, 292, 318, 319, 331, 339 Spillovers, 15, 18, 28, 61, 62, 122, 125, 128, 163, 174, 198, 209, 210, 217, 222, 224, 225, 271, 292, 306, 317–319, 329, 331– 344, 347–349, 369, 387, 401 STEM, 352, 356 Strategic alliances, 227, 402 Strategic vision, 110, 119 Structural competitiveness, 109, 119 Success breeding success, 401 Supply-siders, 104 Swedish biotechnological industry, 392 Swedish composers, 385, 393, 401 Swedish music industry, 386, 388–394, 396, 397, 399, 404 Swedish Road Administration, 279

Index T Technical change, 8, 10 Technological innovation, 28, 46, 198, 208, 213, 214, 218 Technology-based clustering, 332 Technology change, 121, 162, 174, 175 Territorial entity, 226 Tiger economies, 383 Total infrastructure productivity, 217 Trading costs, 179 Traffic flows, 27, 143–158 Transaction areas, 15 Transaction cost, 4, 14–15, 19, 24, 92, 117, 122, 136, 179, 180 Transaction information, 270 Transfer channels, 222, 242 Transfer of knowledge, 28, 61–63, 65, 66, 77, 78, 100, 117, 118, 125, 128, 129, 222, 223, 226–229, 235–242, 270–272, 274, 335, 347, 348, 351, 360, 368 Transformation, 7, 81, 83, 94, 107, 111, 112, 114, 117, 133, 139, 147, 162, 202, 273, 352, 379 Transparency in governance, 109 Transport costs, 60, 92, 163, 167, 171, 174, 175, 177, 179, 183, 189, 298, 312 Transport network and cost, 36, 37, 163, 167, 174, 183, 184, 269 Transshipment costs, 167 Triple helix, 124, 125, 127, 129, 138, 139 Turkish entrepreneurs, 247, 254–256, 258–260 Turkish immigrants, 254, 255 Typology, 223, 224, 232, 241

U Uncertainty, 27, 77, 84, 117, 133, 163, 226, 383 Unemployment cultures, 133 Unites Territoriales Statistiques (NUTS), 334 University R&D, 28, 267–286 Urban economic milieu, 12, 77, 110, 317 Urban planning, 114 Utilized capacity, 168, 169, 173

W Walrasian model, 180 Warner Music, 397 War on Terror, 102 Welfare evaluations, 182 Welfare maximization, 39, 40, 42 Work Sharing, 51 World Development Indicators, 318

Un

co ted

rre c oo f

Pr