Regional Disparities in Small Countries (Advances in Spatial Science) 3540243038, 9783540243038

Citation preview

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

Titles in the Series

H. Eskelinen and F. Snickars (Eds.) Competitive European Peripheries VIII, 271 pages. 1995. ISBN 3-540-60211-9

G. Atalik and M. M. Fischer (Eds.) Regional Development Reconsidered X, 220 pages. 2002. ISBN 3-540-43610-3

A. Nagurney and S. Siokos Financial Networks XVI, 492 pages. 1997. ISBN 3-540-63116-X

Z. J. Acs, H. L. F. de Groot and P. Nijkamp (Eds.) The Emergence of the Knowledge Economy VII, 388 pages. 2002. ISBN 3-540-43722-3

M. M. Fischer and A. Getis (Eds.) Recent Developments in Spatial Analysis X, 434 pages. 1997. ISBN 3-540-63180-1

R. J. Stimson, R. R. Stough and B. H. Roberts Regional Economic Development X, 397 pages. 2002. ISBN 3-540-43731-2

P. McCann The Economics of Industrial Location XII, 228 pages. 1998. ISBN 3-540-64586-1

S. Geertman and J. Stillwell (Eds.) Planning Support Systems in Practice XII, 578 pages. 2003. ISBN 3-540-43719-3

R. Capello, P. Nijkamp and G. Pepping (Eds.) Sustainable Cities and Energy Policies XI, 282 pages. 1999. ISBN 3-540-64805-4

B. Fingleton (Ed.) European Regional Growth VIII, 435 pages. 2003. ISBN 3-540-00366-5

M. M. Fischer, L. Suarez-Villa and M. Steiner (Eds.) Innovation, Networks and Localities XI, 336 pages. 1999. ISBN 3-540-65853-X

T. Puu Mathematical Location and Land Use Theory, 2nd Edition X, 362 pages. 2003. ISBN 3-540-00931-0

J. Stillwell, S. Geertman and S. Openshaw (Eds.) Geographical Information and Planning X, 454 pages. 1999. ISBN 3-540-65902-1 G. Clarke and M. Madden (Eds.) Regional Science in Business VIII, 363 pages. 2001. ISBN 3-540-41780-X M. M. Fischer and Y. Leung (Eds.) GeoComputational Modelling XII, 279 pages. 2001. ISBN 3-540-41968-3 M. M. Fischer and J. Fröhlich (Eds.) Knowledge, Complexity and Innovation Systems XII, 477 pages. 2001. ISBN 3-540-41969-1 M. M. Fischer, J. Revilla Diez and F. Snickars Metropolitan Innovation Systems VIII, 270 pages. 2001. ISBN 3-540-41967-5

J. Bröcker, D. Dohse and R. Soltwedel (Eds.) Innovation Clusters and Interregional Competition VIII, 409 pages. 2003. ISBN 3-540-00999-X D. A. Griffith Spatial Autocorrelation and Spatial Filtering XIV, 247 pages. 2003. ISBN 3-540-00932-9 J. R. Roy Spatial Interaction Modelling X, 239 pages. 2004. ISBN 3-540-20528-4 M. Beuthe, V. Himanen, A. Reggiani and L. Zamparini (Eds.) Transport Developments and Innovations in an Evolving World XIV, 346 pages. 2004. ISBN 3-540-00961-2

L. Lundqvist and L.-G. Mattsson (Eds.) National Transport Models VIII, 202 pages. 2002. ISBN 3-540-42426-1

Y. Okuyama and S. E. Chang (Eds.) Modeling Spatial and Economic Impacts of Disasters X, 323 pages. 2004. ISBN 3-540-21449-6

J. R. Cuadrado-Roura and M. Parellada (Eds.) Regional Convergence in the European Union VIII, 368 pages. 2002. ISBN 3-540-43242-6

L. Anselin, R.J.G.M. Florax and S. J. Rey Advances in Spatial Econometrics XXII, 513 pages. 2004. ISBN 3-540-43729-0

G. J. D. Hewings, M. Sonis and D. Boyce (Eds.) Trade, Networks and Hierarchies XI, 467 pages. 2002. ISBN 3-540-43087-3

R.J.G.M. Florax and D. A. Plane (Eds.) Fifty Years of Regional Science VIII, 400 pages. 2004. ISBN 3-540-22361-4

Daniel Felsenstein Boris A. Portnov Editors

Regional Disparities in Small Countries With 61 Figures and 54 Tables

123

Professor Dr. Daniel Felsenstein Department of Geography Hebrew University of Jerusalem Mount Scopus 91905 Israel E-mail: [email protected] Professor Dr. Boris A. Portnov Department of Natural Resources and Environmental Management University of Haifa Mount Carmel, Haifa 31905 Israel E-mail: [email protected]

Library of Congress Control Number: 2005921919

ISBN 3-540-24303-8 Springer Berlin Heidelberg New York 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-Verlag.Violations are liable for prosecution under the German Copyright Law. Springer is a part of Springer Science+Business Media springeronline.com © Springer-Verlag Berlin Heidelberg 2005 Printed in Germany 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: Erich Kirchner Production: Helmut Petri Printing: Strauss Offsetdruck SPIN 11375630

Printed on acid-free paper – 88/3153 – 5 4 3 2 1 0

Contents Introduction 1.

Introduction DANIEL FELSENSTEIN and BORIS A. PORTNOV

1

Part I: Concepts, Theory and Methods 2.

The Liability of Smallness: Can We Expect Less Regional Disparities in Small Countries? DANIEL FELSENSTEIN and BORIS A. PORTNOV

13

3.

Country Size in Regional Economics MICHAEL BEENSTOCK

25

4.

Measures of Regional Inequality for Small Countries BORIS A. PORTNOV and DANIEL FELSENSTEIN

47

5.

Investigating Spatial Patterns of Income Disparities Using Coordinate Transformations and GIS Mapping BORIS A. PORTNOV and RIMMA GLUHIH

63

Part II: Empirical Evidence 6.

7.

8.

9.

10.

Regional Employment Disparities in Belgium: Some Empirical Results OLIVIER MEUNIER and MICHEL MIGNOLET Regional Income Convergence and Inequality in Boom and Bust: Results from Micro Data in Finland 1971-2000 HEIKKI A. LOIKKANEN, MARJA RIIHELÄ and RISTO SULLSTRÖM Regional Disparities in Ireland: The Roles of Demography, Profit Outflows, Productivity, Structural Change and Regional Policy 1960-1996 EOIN O’LEARY

85

109

129

The Persistence of Regional Unemployment Disparities in the Netherlands OEDZGE ATZEMA and JOUKE VAN DIJK

147

The Dynamics of Regional Disparities in a Small Country: The Case of Slovenia PETER WOSTNER

169

vi

Contents

11.

Interregional Disparities in Israel: Patterns and Trends BORIS A. PORTNOV

12.

Does Decentralisation Matter to Regional Inequalities? The Case of Small Countries CARLOS GIL, PEDRO PASCUAL and MANUEL RAPÚN

211

Regional Inequalities in the EU Enlargement Countries: An Analysis of Small Versus Large New Member States GEORGE PETRAKOS, YIANNIS PSYCHARIS and DIMITRIS KALLIORAS

233

13.

187

Part III: Policy Issues 14.

Has the Financial Economy Increased Regional Disparities in Switzerland over the Last Three Decades? JOSÉ CORPATAUX and OLIVIER CREVOISIER

251

15.

Regional Policy Lessons from Finland HANNU TERVO

16.

The Globalisation of Austrian Regions: New Policy Challenges and Opportunities MICHAEL STEINER

283

Innovation Policy: An Effective Way of Reducing Spatial Disparities in Small Nations? STEPHEN ROPER

297

Figures

313

Tables

317

Author Index

319

Subject Index

325

Contributors

331

17.

267

1

Introduction

Daniel Felsenstein1 and Boris A. Portnov2 1 2

Department of Geography, Hebrew University of Jerusalem, Israel Department of Natural Resources and Environmental Management, University of Haifa, Israel

During the Candiot War of 1645-1669, the Ottoman Sultan Ibrahim I ordered his chief admiral to attack Malta. Fearing imminent defeat by the superior Venetian forces stationed on the island, the admiral decided to trick the sultan out of the idea. As the story goes, he placed a candle on his naval map, allowing the wax to drip on the tiny island until it was completely covered. Then he exclaimed in false surprise, “Malta Yok!” (There is no Malta!), and convinced the sultan to sail his fleet to the Island of Crete instead. Although Malta is not featured in this volume, most of the countries it covers are of “wax drip” size. Intuitively, it may be expected that everything in small countries is diminutive: distances, population, economies, and even regional inequalities. Thus, at a symposium on “The Challenge of Development” convened in Israel in 1957 to mark the inauguration of a new building for the Department of Economics at the Hebrew University of Jerusalem, the eminent US economist Simon Kuznets stated that “developed small states seem to have succeeded in spreading the fruits of economic growth more widely among their populations than the larger states at comparable levels of income per capita”. While noting that he did not really have any empirical evidence to bolster this claim, he continued that “it is my belief that income is distributed more equally among the populations in the Scandinavian countries and Switzerland than say in France, Germany or even the United States…. These smaller countries have no proportionately large regions like our South with a per capita income distinctly lower than the rest of the country” (Kuznets 1960, p. 30). Similar sentiments also appear in discussions of the impacts of country size on economic development. For example, Streeten (1993) has claimed that ‘large countries show, of course, larger inequalities by region than small countries (p. 199). Perkins and Syrquin (1989) state similarly that “if inequality between regions in a country is a major source of inequality between households, then one would expect large countries to have greater regional diversity and hence higher 1 levels of inequality” (p. 1694) . This book revisits these common conceptions. The motivation for the volume is to examine both conceptually and empirically the “belief” that small countries (which are often not much larger than regions in a large country), do not exhibit 1

This expectation is not supported in their subsequent empirical analysis.

2

Daniel Felsenstein and Boris A. Portnov

significant regional differences. While we are not sure whether the empirical data would have supported Kuznets’ contention when it was stated nearly 50 years ago, half a century later we are able to garner evidence to test whether his intuitive feeling has stood the test of time. Neo-classical growth theory, as developed in the context of international trade and applied to regions, and Schumpeter’s (1934) theory of economic expansion, assert that competitive forces and interregional migration of labour and capital equalize differences and factor prices across regions and lead to more even regional development (Hirschman 1958; Siebert 1969; Richardson 1977). In contrast, the so-called “new economic geography” asserts the opposite: the uneven concentration of production that manifests itself, inter alia, in a “core-periphery geography”, is sustained by circular production linkages and may become increasingly entrenched over time (Krugman 1991, Brakman et al. 2001, Fujita et al. 2001). However much of the evidence, in both directions, is based on large countries such as US states or areas within a supra-national economy such as the EU (Armstrong 1995, Le Gallo and Ertur 2003, Tsionas 2000). Do any of these theories hold for small countries generally characterized by small land area and small population size? These two determining attributes lead to a slew of implications with respect to regional disparities. If distances are shorter, access costs are lower, the number of regions (and therefore inter-regional variance) is smaller, government structures more centralized and population more homogenous, ostensibly, this should point to narrower disparities across regions in small countries. On the other hand it can be argued that certain unique features of small countries may mitigate any regional convergence. For example, even in small countries physical distance between central cities, which are main centres of employment, and hinterland regions may surpass those practicable for daily commuting. Therefore, any interregional income equalization in such countries or spillover effects cannot but be limited in scope. Furthermore, small countries are, most often, densely populated. This leads to the emergence of considerable diseconomies of agglomeration, not only in their central areas but also in their hinterlands. Whereas in large countries, such diseconomies may be concentrated at major metropolitan areas, in small countries, they may spread over the entire national territory, resulting in considerable gradients of transport outlays and general production costs. In addition, small countries are characterized by a dependence on external markets, international trade and the global economy (Poot 2004). These activities are invariably conducted from the major population centres, leaving peripheral areas at a distinct disadvantage and further entrenching any agglomerative tendencies. In other respects, the characteristics of small countries may give rise to regional outcomes very different to those in large countries. For example, the measurement of spatial disparities in small countries may lead to very different results to those obtained for large countries due to very different spatial scales of analysis. In large countries, such units are often restricted to regions, which are internally heterogeneous. Since either aggregates or averages are compared, the results may often be misleading. In contrast, inequalities among municipalities and even

Introduction

3

individual localities in small countries may be analysed, leading (presumably) to more realistic estimates. Internal migration in small countries and its equalizing effects on interregional disparities may also be distinctively different from those found elsewhere. Smaller land areas mean that long-distance commuting can often substitute for internal migration. In addition it can be claimed that in small countries, the efficacy of public policy in closing regional gaps may be higher, compared to that in large countries with diverse economic, environmental and governance structures. How does this deductive reasoning hold up empirically? This volume attempts to come to terms with the empirical questions and with the attendant issues of conceptualisation, theory, measurement and policy that they presuppose. Primarily this is a book about regional disparities. Small country size and the unique features that stem from this attribute, form the context. Despite the intuition, the book seeks to examine whether there is any a-priori case to expect more regional convergence in small countries than in large ones. Counter to the contemporary trend in edited volumes, the motivation for this book is a real-world regional issue looking for a set of papers and not a set of (invariably, conference) papers looking for an issue. As such, the editors commissioned all the papers in this volume from authors with a publishing interest in the topic areas of small countries and the process of regional convergence therein. The result is a focussed series of theoretical, methodological, empirical and policy-oriented essays grounded heavily in the traditional nexus of regional science and calling on the research competencies of applied economists, urban/ regional economists, economic geographers, business economists and regional planners. We adopt a broad approach to the definition of “regional disparities”. While the common indicators of regional income, product and value added are all examined here, they do not form the exclusive focus. Regional employment and productivity are equally legitimate yardsticks and are also addressed. In addition, we further widen the focus to incorporate regional differences in technological innovation and R&D and responsiveness to global challenges and opportunities as indirect indicators of regional economic welfare.

1.1

Scope of the Book

We are ecumenical as to what constitutes a “small country”. While there is a tendency to distinguish between small “economies” based on level of GDP and economic diversification and small “nations” based on territory sometimes combined with population, we opt for the more nebulous small “countries” terminology. While inherently intuitive, this term suggests some logical combination of national territory, population and wealth (Alesina and Spolaore 2003, Crowards 2002). In this volume, while we do not overtly favour one criterion over another, the common denominator that emerges for defining “small countries” is the spatial (land area) factor. The result is a reasonably coherent set

4

Daniel Felsenstein and Boris A. Portnov

of case study countries comprising developed market or transitional economies characterized by small land areas (with the exception of Finland), small populations and a relatively uniform standard of living (with the exception of Slovenia) (Table 1.1). We also include two chapters that take an aggregate look at regional disparities in small and/or transitional countries. In all, the empirical evidence presented here accounts for a cross-section of small, developed countries that have internal regional divisions but could equally have included the likes of Denmark and New Zealand. Indeed, other “small country” studies have been much more synoptic in their choice including countries such as Italy and Portugal (Robinson 1960), Hungary and Canada (Freeman and Lundvall 1988) and Singapore and New Zealand (Ein-Dor, Myers and Raman 1997). This volume also follows in the wake of a recent surge of interest in “micro” and “peripheral” economies that has tended to look at the way remoteness and smallness impact on their economic performance (Armstrong and Read 1995, 2003; Bertram 2004; Poot 2004). Much of this interest is focussed on tiny island or city-states, protectorates and autonomous or dependent territories with some measure of political sovereignty. We have consciously chosen to avoid these examples irrespective of their level of economic development (Hong Kong, Singapore, Luxembourg, Malta etc) due to their lack of significant internal 2 regional divisions . The rationale for this is that our focus on inter-regional disparities in small countries presupposes some role for geographic mediating forces such as distance, density, factor mobility and space in creating regional disparities. The tiny physical scale of micro or city-states implies that interregional differences are virtually meaningless. There is also a tradition of interest in small countries in the development economics literature (Briguglio 1995; Easterly and Kraay 2000; Selwyn 1975; Streeten 1993). But again, this has been more concerned with questions of volatility and vulnerability in their national economies than with inter-regional gaps. The current volume with its emphasis on regional disparities, is therefore tangential but complementary to these areas of interest. Table 1.1 shows the extent of regional differences in the seven countries that form the empirical studies of this volume. In terms of the relationship between number of regions and size of population, all the countries (bar Ireland and Slovenia) show a remarkably consistent ratio of roughly one region per million population. This, of course, says nothing about the real inter-regional population distribution. Here we find that the range of values between the most and least populated regions in each country can be quite considerable. On the one hand a country like Israel displays a size distribution where the most populous region is 2

Luxembourg for example has 3 administrative divisions but the whole state is considered a NUTS II region of Belgium. Malta has 3 miniscule ‘regions’ the largest of which is 170 sq km and with a population of 215 thousand residents. Hong Kong has 18 districts distributed across 3 sections (Kowloon, Hong Kong Island and the New Territories) but these can hardly be considered equivalent to NUTS II or NUTS III regions.

Introduction

5

nearly double the size of the least populous. On the other hand, in the other small countries this range is much larger with the largest region in population terms being five or more time the size of the smallest (Table 1.1). Much of this is due to the presence of large metropolitan centres that dominate their regions. Thus in Austria the smallest region is Burgenland (278,000) and the largest Vienna (1.56 million), in Belgium the Brabant wallon region (358,000) is pitched against the Antwerp metropolitan area (1.66 million), in Ireland the Midland area (213,000) is compared with the Dublin and Mid East counties (1.52 million) and in the Netherlands the Zeeland regions is juxtaposed with the Zuid Holland region containing the Randstad metropolitan agglomeration (3.38m). In addition, the difference across countries in the number and size of regions notwithstanding, the data show substantial gaps in regional GDPpc across poorest and richest regions. In those countries with the lowest level of regional disparities (Finland, the Netherlands and Slovenia, Switzerland and Ireland) the richest region is roughly 1.7 times better off than the poorest. In the intermediate group, Israel and Austria, the richest region is more than twice as rich as the poorest. In the most unequal country (Belgium) the richest region is over 3 times richer than the poorest. These gaps are quite substantial considering the corresponding figures for inter-regional GDPpc differences in larger, developed EU countries such as Spain (2.08), Italy (1.93) and Germany (3.92). Table 1.1. Key attributes of the small countries examined in this volume

Country

Land GDPpc Population No. of 2000 Area 1 (Millions) Regions2 ($, pps)1 (Sq km)1

Austria Belgium Finland Ireland Israel Netherlands Slovenia Switzerland

83,000 30,200 305,000 69,000 21,000 33,800 20,200 39,700

8.1 10.2 5.2 3.9 6.5 16.0 2.0 7.2

25,000 25,300 22,900 21,600 18,900 24,400 12,000 28,600

9 10 5 7 6 12 12 7

Ratio of Ratio of Population Difference Size: Largest Between Richest versus and Poorest Smallest Regions: Regions3 Regional GDPpc 2000 ($, pps) 4 5.6 2.18 4.65 3.26 3.26 1.73 7.1 1.66 1.9 2.08 9.1 1.77 12.4 1.71 5.3 1.67

1. CIA World Factbook, http://www.cia.gov. 2. NUTS II regions; Ireland 7 counties - NUTS III regions, Israel 6 statistical districts; Slovenia 12 NUTS III regions; Switzerland 7 regions- NUTS II equivalents. 3. Administrative Divisions of Countries, http://www.statoids.com/statoids.html 4. Eurostat Regions Statistical Yearbook 2003; Eurostat, Regional Gross Domestic Product in the European Union 2000, Statistics in Focus, Theme 1-1/2003; Israel-based on Multi-regional I-O model; Switzerland - Federal Statistical Office. 5. Excludes Luxembourg. 6. This ratio excludes the Aaland NUTS II region with a population of only 26,000. The ratio compares the South Finland region (1.8m pop, including metropolitan Helsinki) with the North Finland region (.557m pop).

6

1.2

Daniel Felsenstein and Boris A. Portnov

The Structure of the Book

This volume engages the following questions: x What are the unique conceptual, theoretical and methodological challenges in analysing regional disparities in small countries? x Are small countries characterized by significant interregional disparities? Do interregional differences tend to converge or diverge over time? x Which policy measures might help to close regional gaps in small countries and what is their effectiveness? The three parts of the book address these issues in turn. Part I begins with a review of the main size-related attributes of small countries (openness and dependence on an external trade economy, number, size and density of regions, social cohesion and governance structure etc) and examines their impact on regional disparities. The chapter then proceeds to frame the issue as one in which the attributes of small size (land area, population and magnitude of the economy) are mediated by a series of spatial and non-spatial factors. These include factors such as distance, density and factor mobility, natural resources, land supply, social cohesion and governance structure. This generates regional outcomes that are expressed in income disparities, industrial structure, commuting and migration patterns, agglomerative forces and the like. Given the existence of these mediators, the size of regional disparities in small countries is not as surprising as it may seem at first glance. The economic theory behind these issues is taken up Beenstock (Chapter 3). He questions whether small countries are analytically different to large and require separate economic and statistical treatment. His conclusion is that size is something of a misnomer serving to deflect attention from the real issue of regional heterogeneity. Small countries can be regionally heterogeneous by the same token that large countries can be regionally homogeneous. In his view, this is the true justification for a regional perspective and not size, per se. Measurement issues are addressed in two separate chapters. Portnov and Felsenstein (Chapter 4) look at the sensitivity of commonly used income inequality measures to changes in the size and number of regions. A bootstrapping experiment and sensitivity test are set up to determine whether inequality measures commonly used in regional analysis produce meaningful estimates when applied to small countries. To this end, hypothetical distributions of populations and incomes presumably characteristic of small countries are compared with a “reference” distribution, assumed to better represent countries of larger size. According to results of the tests, only the population weighted coefficient of variation (Williamson’s index) and population-weighted Gini coefficient may be considered as more or less reliable inequality measures, when applied to small countries. Gluhih and Portnov (Chapter 5) investigate visualization issues involved in representing inter-urban income disparities in a small country using GIS tools.

Introduction

7

Four approaches are discussed with respect to their representational clarity and their ability to visualize spatial dynamics. Part II is the empirical heart of the volume. Here a series of empirical studies of regional disparities in the small countries featured in Table 1.1 are presented. These use a range of analytic tools from Barro-type growth models (Wostner, Chapter 10; Petrakos, Psycharis and Kallioras, Chapter 13) and extended shiftshare analysis (Meunier and Mignolet in Chapter 6 and O’Leary in Chapter 8), to inequality indices (Loikannen, Riihela and Sullstrom, Chapter 7; Gil, Pasual and Rapun, Chapter 12) and factor analysis (Portnov, Chapter 11). While the majority of these are single-country studies focussing on Belgium, Finland, Ireland, Holland, Slovenia and Israel two chapters present integrative pieces comparing a series of small and large countries. The first (Gil et al., Chapter 12) looks at the impact of decentralization on regional inequalities across small and large (mainly EU) countries. A major finding is that fiscal decentralization rather than political decentralization leads to regional convergence and this influence is likely to be felt more acutely in the small than the large countries. The second chapter offering an aggregate perspective (Petrakos et al., Chapter 13) compares regional imbalances across EU Accession States over the course of the 1990’s again contrasting the small with the large. Regional inequality across all countries is shown to have increased over time with the small countries showing slightly higher levels across some indices and certainly higher levels of volatility. The public policies and regulatory instruments used to address regional disparities in small countries are examined in Part III. Here a heterogeneous series of issues are addressed from a small-country perspective despite the fact that these are equally potent topics with respect to large countries. In an era of supra-national units such as the EU and global capital flows, Corpataux and Crevoisier (Chapter 14) question the wisdom of a small country (Switzerland) specializing in financial services and its role in increasing regional inequality. Similar issues of globalisation and its effect on regions in a small country are taken up by Steiner with respect to Austria (Chapter 16). The evidence presented shows that ironically, the advent of EU economic integration has led to heightened regional awareness and the desire to nurture regional competitive advantage. Finland, one of the small countries with relatively moderate inter-regional contrasts, is of course, a country with a rich legacy of direct regional policy interventions and these are critically examined by Tervo (Chapter 15). He finds that indirect national policies with regional implications (tax policy, income transfers and human capital enhancement) have been more effective in reducing regional income differentials than direct regional incentives. Finally, a chapter by Roper (Chapter 17) questions whether harnessing industrial strategy (in this case innovation policy) in order to moderate regional imbalances is really an efficient form of intervention. Drawing on the experiences of three small countries with rather different technological trajectories (Ireland, Israel and Finland) he shows how the trade-off between greater regional equality and its cost in terms of productivity loss, will vary across national contexts. Remarkably, all the policy discussions seem to lead to a similar conclusion. The emergent message seems to be that regional disparities in small countries are likely to become increasingly

8

Daniel Felsenstein and Boris A. Portnov

susceptible to the forces of globalisation and competitive advantage and less fashioned by the inherent characteristics of size and all that it implies (short distances, dense population, demographic homogeneity etc). Together, these chapters show that regional disparities exist in no small measure in the countries examined. Small countries, despite common intuition, do not seem to have significantly smaller regional gaps than the large. Just as size seems to be something of a non-sequiter with respect to the economic growth of small countries (Armstrong and Read 1995; Easterly and Kraay 2000) and at best the evidence is mixed (Perkins and Syrquin 1989), similarly with respect to regional disparities. As Beenstock points out in Chapter 3, economic and statistical theory do not suggest that small countries should have a more equitable income distribution than large ones. One question this volume leaves unanswered relates to the way small countries can harness economic forces in order to manage the liability of smallness and its consequent regional imbalances. As will be illustrated in Chapter 2, the geographic mediating forces of distance, density and factor mobility can all be transformed through technological advance, which pulls down transport and communications costs and increases accessibility ostensibly closing regional gaps. By the same token however those same forces encourage increasing agglomeration and lessen the prospect for regional convergence. This is of course a line for future research while the current volume is more concerned with establishing the existence of regional disparities in small countries and the policy measures taken to close these gaps. As always, a book project that involves over 20 authors demands considerable organizational resources. While not subject to constraints of size or even distance, the preparation of this volume nevertheless posed various technical challenges. In this task we were ably abetted by Katharina Wetzel-Vandai and Irene BarriosKezic of Springer who supported this project from its inception, by Michal Stern who fashioned the various chapters into a common format and by Tamar Sofer of the Cartographic Laboratory and the Hebrew University of Jerusalem who assisted with the graphic material.

References Alesina A, Spolaore E (2003) The size of nations. MIT Press, Cambridge, MA Armstrong H (1995) Convergence among the regions of the European Union. Papers in Regional Science 74:143-152 Armstrong HW, Read R (1995) Western European micro-states and EU autonomous regions: the advantages of size and sovereignty. World Development 23(8):1229-1245 Armstrong HW, Read R (2003) Microstates and subnational regions; mutual industrial policy lessons. International Regional Science Review 26(1):117-141 Bertram G (2004) On the convergence of small island economies with their metropolitan patrons. World Development 32(2):343-364 Brakman S, Garretsen H, van Marrewijk C (2001) An introduction to geographical economics; trade, location and growth. Cambridge University Press, Cambridge

Introduction

9

Briguglio L (1995) Small island developing states and their economic vulnerabilities. World Development 23:1615-1632 Crowards T (2002) Defining the ‘small’ states category. Journal of International Development 14:143-179 Easterly W, Kraay A (2000) Small states, small problems? Income, growth and volatility in small states. World Development 28(11):2013- 2027 Ein-Dor P, Myers MD, Raman KS (1997) Information technology in three small developed countries. Journal of Management Information Systems 13(4):61-89 Freeman C, Lundvall BA (eds) (1988) Small countries facing the technological revolution, Pinter London and NY Fujita M, Krugman P, Venables AJ (2001) The spatial economy: cities, regions, and international trade. MIT Press, Cambridge, MA Hirschman AO (1958; 1966 reprint) The strategy of economic development. Yale University Press, New Haven Krugman P (1991) Increasing returns and economic geography. Journal of Political Economy 99:483-489 Kuznets S (1960) Economic growth of small nations. In: Robinson EAG (ed) Economic consequences of the size of nations. Macmillan, New York, pp 14-32 Le Gallo J, Ertur C (2003) Exploratory data analysis of the distribution of regional per capita GDP in Europe 1980-1995. Papers in Regional Science 82(2):175-202 Perkins DH, Syrquin M (1989) Large countries; the influence of size. In: Chenery H, Srinivasan TN (eds) Handbook of development economics. Volume II, Elsevier, Amsterdam, pp 1691-1753 Poot J (ed) (2004) On the edge of the global economy. Edward Elgar, Cheltenham, UK Richardson HW (1977) Regional growth theory. Macmillan, London Robinson EAG (1960) (ed) Economic consequences of the size of nations. Macmillan, New York Schumpeter JA (1934; 1961 English Edition) The theory of economic developments: an inquiry into profits, capital, credit, interest, and the business cycle. Harvard University Press, Cambridge, Mass Selwyn P (ed) (1975) Development in small countries. Croom Helm, London Siebert H (1969) Regional economic growth: theory and policy. International Textbook Company, Scranton PA Streeten P (1993) The special problems of small countries. World Development 12(2):197202 Tsionas EG (2000) Regional growth and convergence: evidence from the United States. Regional Studies 34(3):231-238

Part I: Concepts, Theory and Methods

2

The Liability of Smallness: Can We Expect Less Regional Disparities in Small Countries?

Daniel Felsenstein1 and Boris A. Portnov2 1 2

Department of Geography, Hebrew University of Jerusalem, Israel Department of Natural Resources and Environmental Management, University of Haifa, Israel

2.1

Introduction

Ostensibly, size would seem to be a concrete physical notion that is easily observed and measured. It is hardly an elusive concept that is differently perceived and experienced by different individuals or groups. Objectively, size may be measured by three different, although interdependent, parameters - land area, population and economy. For the purpose of this study, the latter criterion (economy) is a more or less a non-starter. By defining a country as small, based solely on economic performance, we find ourselves including land-endowed giants such as Ukraine and Byelorussia, as well most African, Middle East and Central Asian nations. The physical magnitude of a country (measured by either population size or land area) would seem to dictate a whole string of attributes in which cause and effect are clearly delimited. Thus small countries are likely to have smaller markets and be more open to external trade. Smaller populations may lead to less extreme variation in social or economic characteristics. Similarly, should the magnitude of a country’s economy decline with physical size, then the effect of “economic smallness” would be equally clear: a small market means a more volatile economy, less ability to achieve scale economies and so on. Size, however, can also be conceived as a relative or contextual notion. No single index or measure of size will satisfy all research needs or policy contexts. For instance, small land area does not necessarily mean small population and vice versa. North Europe, Asia and the Pacific provide us with numerous examples of land-abundant but sparsely populated countries (e.g., Norway, Finland, Iceland, Australia, and New Zealand). Furthermore, the effect of size on economic outcomes is not absolute. The constraints and opportunities offered by small size and limited natural resources can be mediated by technological innovation and human capital embellishments. As a result, the “small countries” club may include both economically advanced nations (such as those mentioned above) as well as economic backwaters such as Mongolia, Nepal and Bhutan.

14

Daniel Felsenstein and Boris A. Portnov

Even those size factors considered absolute and “fixed” such as land supply and human capital endowments can be changed over time. For example, land can be re-claimed and workers can be re-skilled. Economic performance of a country can also change in both directions. While Slovakia, Slovenia, and the Baltic states have shown rapid improvements in their economic conditions, the economies of other small countries (e.g., Azerbaijan, Georgia and Armenia) have deteriorated considerably. Context and spatial scale are also important here. At certain levels of analysis such as the supra-national, a country may be considered “small” with all the implications that accompany this categorization. At other levels of aggregation, such as the trade-bloc, the same physical unit of territory or magnitude of economy, may assume a different relative size. These issues of absolute or relative scale are further compounded when dealing with regions within countries. If a country is small, then its regions may also be sized in proportion. If regions are simply countries writ small, then all the attributes relating to small countries should equally apply (and sometimes with greater potency) to their regions. For example, if small countries are open economic systems then their regions can be considered particularly exposed. If small countries are assumed to be culturally homogenous and socially cohesive, then their regions are assumed to exhibit these attributes even more pronouncedly. But just what are these issues and attributes and are regional characteristics just reflections of small country attributes? This chapter attempts to set the groundwork for the following chapters, by framing some of these issues. The key issue of course, is the effect of small country size on regional disparities. If regions just reflect their national structures, then small countries with more equitable distributions of income and product at the national level should also have smaller regional disparities. The chapter starts by outlining some of the distinguishing features of small countries that are likely to inform any analysis of their regional inequalities. We highlight the expected impacts of these attributes (conditionally defined as either spatial or non-spatial factors) on inter-regional convergence or divergence in a small country. The chapter then proceeds to test research assumptions concerning the effects of different size-related factors on the magnitude of regional disparities, using statistical data available for a number of small countries. The concluding section defines the general pattern of relationship between country size and regional inequality.

2.2

The Attributes of Small Countries

Some economists tend to consider country size a non-issue in terms of economic theory (see Chapter 3). This stems from a viewpoint that relates to countries or regions as individuals rather than groups. An alternative view however is that a country represents a group of regions, each region is a group of municipalities and so on. As Hare (1962) points out, small groups are inherently distinguished from

The Liability of Smallness

15

their larger counterparts by a number of distinctive characteristics - greater ability for self-organization, stronger social cohesion and smaller differences in goals and values among individual group members. Many of these “small group” characteristics are largely applicable to small countries. The small countries literature abounds with descriptions of the defining attributes of small nations. As noted in Chapter 1, much of this is grounded in the development economics tradition and as such focuses on micro, island and citystates (Armstrong, de Kervenoael, Li and Read 2000; Bertram 2004; Read 2004). The size definition used is invariably based on land area, population size or GDPpc. Most studies outline an upper size limit on the basis of statistical techniques (Crowards 2002) or “natural” break points in the size distribution. These however remain arbitrary choices. Work by Armstrong and his colleagues suggest a 3m population cut-off (Armstrong and Read 1995; Armstrong et al. 1998), others opt for a 10-15m population break (Robinson 1960) or a land area of 65,000km2 (Jalan 1982) and so on. What does emerge however is that over time, the growing complexity and diversity of “small” economies, makes issues of size as measured by standard population or territorial indicators, increasingly difficult to defend (Alesina and Spolaore 2003). The archetypal profile of the small country as portrayed in the development economics literature is one primarily characterized by small local markets, dependence on exports and an inability to reach scale economies (Scitovsky 1960; Streeten 1993). This is a prime feature that distinguishes large countries from the small, in both quantitative and qualitative terms. It is also an attribute that is not directly dependent on land or population size. Conceivably, a country with a large land mass and small population or with a large but poor population, could both be considered “small” under these terms. On the supply side, a small country is characterized by resource constraints. A labour supply constraint is likely to exist. However, in developed small countries such as those featured in this volume, this can work to their advantage. Constraints on the domestic supply of labour invariably result in an emphasis on developing high-skill human capital for high value added production. Labour market equilibrium and low-level labour supply can be attained via in and out-migration, especially when small countries are part of a larger continent, as in Europe (Armstrong and Read 2002). In other small countries, labour supply constraints coupled with the competitiveness and vagaries of the world market in they are forced to compete, leave the small country in a vulnerable position (Briguglio 1995). If physical area defines the small country, the land supply constraint is likely to be a particularly acute issue. On the one hand, a small land area makes for a small agricultural sector. This is a source of advantage for a small, developed economy (Armstrong and Read, 2002). On the other hand, as shown in Chapter 3 of this volume, limited land supply in small countries makes for limited stocks of building land and these are generally not uniformly distributed. As land and housing services are obviously non-tradable goods, they are likely to reinforce regional differences in small countries to a greater extent than in large countries.

16

Daniel Felsenstein and Boris A. Portnov

With an open economy dependent on imports to meet local consumption demands, the small country invariably finds itself subject to exogenous forces that determine many of its macro-economic parameters such as exchange rates, domestic price levels etc. In such circumstances the small country may align with a supra-national body such as the EU in order to try and mediate some of the liabilities of smallness (Marcy 1960). This however results in limiting the countries ability to effect an independent macro-economic policy via the monetary and fiscal tools at its disposal. All this would seem to point to size as a key conditioning factor in the economic performance of small states and the sub-optimality associated with being small (limited, high cost local production, lower incomes etc). However the empirics do not seem to support his view. Studies by Armstrong and his collaborators have shown that micro-states perform as well and sometimes better than their adjacent regions (Armstrong and Read 1995; Armstrong et al. 1998) and their income levels tend to converge to those of their patron economies (Bertram 2004). In addition, the empirical findings coming out of the development economics literature and attempting to link size to economic performance are often ambiguous (Milner and Westaway 1993; Perkins and Syrquin 1989). The attributes of small size extend beyond its impact on economic performance. Size also impacts on social cohesion and distributional impacts. Both these issues receive surprisingly short shrift in the literature. Social cohesion and homogeneity of tastes and cultures are assumed to be greater in small countries although this issue is not generally tested empirically (Kuznets 1960). Accessibility to decision makers is arguably easier and this makes for greater social consensus and solidarity. This could also be mediated by the more centralized governance systems in small countries. Stronger central government and less regional governance is likely to lead to more focused policy goals and greater attempts at regulating social cohesion (Shankar and Shah 2003). Political centralization in a small country is therefore likely to spawn fiscal centralization and this concentration of political power and budgetary control is likely to be selfreinforcing. Economic activity will choose to be close to the seat of power and resources further aggravating regional disparities. When compared to the big issues of vulnerability and export-orientation, the question as to whether small countries have a more equitable income distribution across social groups or regions is perceived as of secondary importance in defining their economic character. In addition, it may seem self-understood that small size implies less variation which in turn, implies a more equitable distribution. But is this linear reasoning so obvious and is it backed by empirical evidence? Streeten (1993) claims that “in small developed countries there seems to be less inequality in income distribution than in large ones” but that “large countries show, of course, larger inequalities by regions than small countries” (p 199). While this claim is not backed by any estimates, other work from development economics has not been able to verify this statement. Perkins and Syrquin (1989) test for a relationship between the size distribution of income and country size. They assume that the regional income distribution is reflected in the size distribution of income as regional inequality is one source of inequality in the

The Liability of Smallness

17

distribution by size. Based on data for 48 countries, they find no evidence to back this claim and the coefficient for size is in fact negative but insignificant.

2.3

Regional Impacts of Smallness

A-priori deduction of the relationship between country size and regional disparities does not point unambiguously in one direction. Table 2.1 sketches out some of the main expected outcomes of this relationship. Size-related attributes are presented and their impacts in terms of either regional convergence or divergence are hypothesized. In the following subsections, these impacts are considered separately for spatial and non-spatial factors and discussed in turn. 2.3.1

Spatial Influences

According to Tobler’s first law of geography, “everything is related to everything else but nearby things are more related than distant things” (Tobler 1970, p 236). The impact of inter-regional spillovers on regional disparities clearly follows this logic. On the one hand, shorter distances in small countries imply more spillovers and regional convergence. There is much evidence to suggest that knowledgebased spillovers are regionally bounded (Acs 2002) and thus where distances are small spillovers are likely to promote convergence. On the other hand, small countries often have one dominant metropolitan centre that casts a shadow or “Upas Tree” effect on other regions and limits any significant inter-regional spillover effect. For example, this effect has been noted for Helsinki, Tel Aviv and Dublin in their respective regional contexts (Roper and Grimes 2005). In addition, the dominance of the metropolitan centre is further entrenched as, even in a small country, the distance between such a centre and the hinterland regions generally surpasses those practicable for daily commuting (Portnov and Erell 2001). The small size of individual regions in small countries is another attribute with ambiguous effects. Small regional size means less likelihood of within-region extreme values and consequently less intra-regional variance. This makes for more evenly developed regions. Also, the smaller size of regions makes for smaller units of analysis and smaller aggregates are likely to show more equality. Alternatively, the small size of regions means that transport costs are less an advantage to domestic suppliers. With this form of protection removed, the small country becomes more dependent on exports. This dependence on external forces implies less freedom in setting a local policy agenda that includes regional preferences. All this can make for greater regional divergence.

18

Daniel Felsenstein and Boris A. Portnov

Table 2.1. Size-related attributes and their expected impacts on regional disparities Size-Related Attribute

Convergence

Divergence

Limited natural resources

More even regional development due to the absence of initial advantage in regional resource endowment

Specialization on tertiary industries and services leading to greater concentration of regional development

Small variation of climatic conditions and agricultural productivity

As above

Agglomeration forces are unobstructed by “natural attractiveness” of hinterland regions

High population density

Long-distance commuting substitutes for interregional migration; scale diseconomies spread over most national territory

Severe diseconomies of scale, specifically in overpopulated core regions, leading to growth spillover

Openness to the global economy

Direct representation of regions in the international markets; direct international investment in regional economies; advantages of both core and border regions for international trade

Less independence in setting social and regional priorities; pronounced concentration of development in few “global cities” and around major transport hubs

Centralized governmental structure

Fewer constraints on the implementation of regional development policies and programs

Stronger unitary governance; less regional budgeting

Short distances

High level of social cohesion and development interdependency; low transportation costs for local suppliers and service providers; possibility of daily interregional commuting; greater factor mobility; more development spillover

Functional domination (“shadow effect”) of major population centres (e.g., via jobs and service provision) over most national territory

Small number of regions

Less interregional variance of development rates

Small size of regions

Less intra-regional variance (smaller aggregates)

Limited agglomeration economies; small regional markets; greater dependence on exports; vulnerability to exogenous shocks (e.g., hyperinflation, economic slowdown), specifically in peripheral regions

The Liability of Smallness

19

The supply of land in a small country or region is both a geographic and an economic attribute that expresses regional size and mediates its effects on income distribution and agglomeration impacts (Figure 2.1). Land is a unique feature in the creating of inter-regional inequalities as its supply varies across regions. In addition, land and the housing services it produces are non-tradable across regions and exogenously determined. Even when all other factors are mobile, the regional differences in land supply serve to ensure that regional inequalities persist (see Chapter 3). This is not just a neo-classical insight. New Economic Geography (NEG) -inspired models arrive at similar conclusions. Helpman's model of the forces promoting agglomeration takes the supply of housing land in a region as the main force promoting dispersal and arresting agglomerative growth (Helpman 1998). In contrast to the original Krugman (1991) formulation where declining transport costs and the erosion of distance as a spatial mediator makes for greater agglomeration, in the Helpman model, lower transport costs make for less agglomeration and make regions more similar. The main geographic mediator is the supply of land which determines the distribution of the housing stock and consequently, the size of regions (large or small populations). Other attributes expected to promote regional convergence include first, the small number of regions in a small country. Again, the law of small numbers implies less extreme values and therefore more inter-regional equality. Second, small countries as noted above are likely to have greater social cohesion. Where distances are shorter, we can expect to find a greater homogenisation of tastes and cultures, more openness to change, greater national solidarity and more focus in setting national priorities and executing policy. All these factors are expected to work in favour of regional convergence. Finally, in a small country, exogenous shocks such as mass immigration and regional policy are likely to have a greater impact on promoting convergence as regions are smaller and less populous. In certain instances the expected outcomes seem clear cut while in others they can go either way. For example, in small countries certain factors of production are expected to be more mobile because of shorter distances (labour, goods). This is expected to lead to inter-regional convergence. However, small country size also means greater discontinuities generated by national boundaries. This has a differential effect on limiting factor mobility. In developed economies it hardly affects capital and technology but can still curtail the movement of goods and labour. These boundaries are not just national. In some instances they also represent cultural, linguistic, educational and social discontinuities as well. By constraining factor mobility these discontinuities can indirectly hamper regional convergence. Finally, greater population densities which are more likely to be found in small countries than in their larger counterparts, may lead to more severe diseconomies of scale in the central core region of the small country, thereby further retarding any regional convergence.

20

Daniel Felsenstein and Boris A. Portnov

Indicators of Size

Spatial and Nonspatial Mediators

Regional Outcomes

x Income Disparities x Land Area

x Distance

x Industrial Structure

x Density

x Export/Import

x Factor Mobility x Population

x Natural Resources x Land Supply x Social Cohesion

x Economy

x Governance Structure

Dependency

x Long-Distance Commuting Versus Interregional Migration

x Agglomeration Forces x Metropolitan “Shadow” Effect

x Development Policies and Regional Budgeting Fig. 2.1. The role of mediating factors

2.3.2

Non-spatial Factors

Foremost amongst the non-spatial factors unambiguously expected to promote regional divergence is the openness of the economy of the small country. This leads to dependence on external economic forces (trade, sources of supply) and in general less independence in setting social and regional priorities. This is expected to promote regional divergence. The centralized governance structure characteristic of small countries is also expected to work against regional convergence. A strong unitary system of government is less likely to consider regional budgeting or other forms of decentralization likely to promote regional fiscal autonomy (Shankar and Shah 2003). Factor mobility can be taken as the obverse of immobile land supply. Capital, labor, goods and technology are all mobile in differing degrees. This mobility mediates the effect of land area as a size factor. Land area may be an issue affecting the mobility of labor or goods (inducing higher transport costs) but it is hardly a factor affecting capital or technology mobility. Population size is also mediated by factor mobility (Figure 2.1). Different sub-sectors of the population have different propensities to commute or migrate (labour mobility) and the level of tradability of certain goods especially services can often be related to population size.

The Liability of Smallness

21

Transactions costs also play a large role in determining factor mobility and in determining the geographic peripherality of regions (McCann 2004). The evidence on the role of transactions costs in creating inter-regional inequalities is however ambiguous. This ambiguity relates to both information/technology transaction costs and to goods/labor transactions costs. As the cost of transacting over space has both increased and decreased it is necessary to be more circumspect when examining this issue and to differentiate across different types of activities. Thus in primary industrial activities and standardized services, transaction costs have fallen sharply increasing factor mobility and decreasing regional imbalances. In those activities where access to specialized technology or information is of prime importance, transaction costs may have risen promoting factor immobility and emphasizing the regional divide between regions with access to information/ technology and those without. Factor mobility and transaction costs are thus intimately linked to regional disparities. Puga (1999) has shown that when trade costs are high, economic activity will spread across regions to meet consumer final demand mediating regional disparities. However, when trade costs fall, agglomeration will occur and regional inequalities will become entrenched. Again, this is contingent on labour mobility. While lower transaction costs may bring more inter-regional equality in economic activity, if labor is not correspondingly mobile then inter-regional income gaps will persist.

2.4

An Empirical Test

In order to provide some initial indication of whether our assumptions concerning the effect of smallness outlined in the previous section are justified, we undertake a simple test. We estimate the magnitude of regional economic disparities (as measured by the ratio of largest to smallest regional GDPpc) as a function of select measures of country size (i.e., population, land area, number of regions etc). The effects of these factors are controlled by national per capita GDP, to reflect differences in economic development. Some seventeen countries are covered in this analysis. In addition to the seven countries reported in Chapter 1 of this volume, ten additional countries of comparable population size and economic development are included (Switzerland, Italy, Portugal, Spain, Greece, Sweden, New Zealand, Denmark, Hungary, Czech Republic and Slovakia). Using a simple OLS regression we identify and measure the effect of individual predictors on the extent of regional income disparities. The model we obtain is as follows: Inc_dif=3.94 -0.307*Land(log)-2.17*GDP(log)-0.04*Pop_dif-0.003*Nreg+1.35*Pop(log) t

0.92 2

-1.01

R =0.571; F=3.46*; N=17

-2.14*

-1.58

-0.06

3.3**

22

Daniel Felsenstein and Boris A. Portnov

where Inc_dif = the ratio of difference in GDPpc between the richest and poorest region of a country; Land = land area in km2; GDP = GDPpc, $US in PPS terms; Pop_dif=the ratio of difference in the population sizes of the smallest and largest region of a country; Nreg=number of regions, and Pop=population size of a country, residents. [* indicates a two-tailed 0.05 significance level; ** indicates a two-tailed 0.01 significance level]. The estimated model suggests that within the small countries universe, the poorer and more populated countries tend to have wider regional gaps. In other words, although smaller and richer countries do have regional disparities, these disparities tend (ceteris paribus) to be smaller than elsewhere. The model coefficients also suggest the relative importance of spatial versus non-spatial determinants of regional disparities. For the countries covered in the sample, regional disparities seem to reflect mainly economic factors (that is, the number of local residents and overall economic performance as a country as a whole), while spatial determinants (e.g., land area and number of regions) appear to be far less significant.

2.5

Conclusions

The subtitle of this chapter asks: “Can we expect small countries to have smaller inter-regional disparities?” The answer to this question is, not necessarily. There are a number of competing forces at work in small nations such as social cohesion, availability of natural resources, population composition, agglomeration economies, openness to external trade, etc. The combination and intensity of these forces may lead in either direction: both towards regional divergence and convergence. For instance, the shortage of natural resources may lead to more even regional development due to the absence of initial advantage in regional resource endowment. Concurrently, specialization in tertiary industries and services, common to small and resource-poor nations, may lead to a greater concentration of regional growth in selected metropolitan centres and severe underdevelopment of peripheries. As another example, high population density may also have opposite effects on regional development. On the one hand it may lead to greater regional convergence because long-distance commuting may effectively substitute interregional migration. On the other hand, high densities may cause severe scale diseconomies to spread over most of a given country, impeding any growth spillover (see Table 2.1). While small country size suggests greater homogeneity with the corollary that regions will be more similar as well, this chapter suggests that this might not necessarily be the case. Much depends on how small country size translates into measurable metrics such as, distance, density, factor mobility and supply of land (see Figure 2.1). These are real geographical issues that ultimately determine whether regional income distribution is more equitable in small countries, whether their regions are more socially cohesive etc.

The Liability of Smallness

23

While smallness is a comparative notion, it does dictate a host of social, political and economic conditions that ultimately determine the vibrancy of small countries. In this respect, territorial extent and population size are marginal issues. In some respects, the development of supra-national entities such as the EU, have resurrected the small country as a tenable political and economic unit. Furthermore, small countries have become much more complex economies belieing the stereotypical profile of a small country highly dependent on external markets, specializing in niche markets and engaged in sub-optimal production. Today, small countries such as the Netherlands, Ireland, Israel and Denmark all add value to a wide range of products and engage in international trade on the basis of competitive not comparative, advantage. The upshot of all this for regional disparities, is that as developed small countries increasingly become very much like the large, the same applies with respect to their regional disparities. There is no strong a-priori case to expect small, developed countries to be any more equitable than larger countries. These reasons are less to do with the raw attributes of size per se and more to do with the way size is translated into metrics that imply small magnitude, such as density, land supply and so on. Unlike the seminal work in organization theory by Freeman, Carroll and Hannan (1983) who found a distinct “liability of newness” attached to organizational life cycles, we cannot conclude that a parallel “liability of smallness” characterizes regional disparities.

References Acs ZJ (2002) Innovation and the growth of cities. Edward Elgar, Cheltenham, UK Alesina A, Spolaore E (2003) The size of nations. MIT Press, Cambridge, MA Armstrong HW, Read R (1995) Western European micro-states and EU autonomous regions: the advantages of size and sovereignty. World Development 23(8):1229-1245 Armstrong HW, de Kervenoael RJ, Li X, Read R (1998) A comparison of economic performance of different micro-states and between micro-states and larger countries. World Development 26(4):639-656 Armstrong HW, Read R (2002) The phantom of liberty? Economic growth and vulnerability of small states. Journal of International Development 14:435-458 Bertram G (2004) On the convergence of small island economies with their metropolitan patrons. World Development 32(2):343-364 Briguglio L (1995) Small island developing states and their economic vulnerabilities. World Development 23:1615-1632 Crowards T (2002) Defining the category of “small” states. Journal of International Development 14:143-179 Freeman J, Carroll GR, Hannan MT (1983) The liability of newness: age dependence in organizational death rates. American Sociological Review 48:692-710 Hare AP (1962) Handbook of small group research. 2nd Edition, The Free Press, London Helpman E (1988) The size of regions. In: Pines D, Sadka E, Zilcha I (eds) Topics in public economics: theoretical and applied analysis. Cambridge University Press, Cambridge, pp 33-54

24

Daniel Felsenstein and Boris A. Portnov

Jalan B (1982) Classification of economies by size. In: Jalan B (ed) Problems and policies in small economies. Croom Helm, London, pp 14-29 Krugman P (1991) Increasing returns and economic geography. Journal of Political Economy 99:483-489 Kuznets S (1960) Economic growth of small nations. In: Robinson EAG (ed) Economic consequences of the size of nations. Macmillan, New York, pp 14-32 Marcy G (1960) How far can foreign trade and customs agreements confer upon small nations the advantages of large nations? In: Robinson EAG (ed) Economic consequences of the size of nations. Macmillan, New York, pp 254-281 McCann P (2004) Geography, transactions costs and economic performance. In: Poot J (ed) On the edge of the global economy. Edward Elgar, Cheltenham UK, pp 48-71 Milner C, Westaway T (1993) Country size and the medium term growth process: some country size evidence. World Development 21(2):203-212 Perkins DH, Syrquin M (1989) Large countries; the influence of size. In: Chenery H, Srinivasan TN (eds) Handbook of development economics. Volume II, Elsevier, Amsterdam, pp 1691-1753 Poot J (2004) (ed) On the edge of the global economy. Edward Elgar, Cheltenham UK Portnov BA, Erell E (2001) Urban clustering: the benefits and drawbacks of location. Ashgate, Aldershot Puga D (1999) The rise and fall of regional inequalities. European Economic Review 43(2):303-334 Read R (2004) The implications of increasing globalization and regionalism for the economic growth of small island states. World Development 32(2):365-378 Robinson EAG (1960) (ed) Economic consequences of the size of nations. Macmillan, New York Roper S, Grimes S (2005) Wireless Valley, Silicon Wadi and Digital Island - Helsinki, Tel Aviv and Dublin in the ICT boom. Geoforum (forthcoming) Scitovsky T (1960) International trade and economic integration as means of overcoming the disadvantage of a small nation. In: Robinson EAG (ed) Economic consequences of the size of nations. Macmillan, New York, pp 282-290 Shankar R, Shah A (2003) Bridging the economic divide within countries: a scorecard on the performance of regional policies in reducing regional income disparities. World Development 31(8):1421-1441 Streeten P (1993) The special problems of small countries. World Development 12(2):197202 Tobler W (1970) A computer movie simulating urban growth in the Detroit region. Economic Geography 46(2):234-240

3

Country Size in Regional Economics

Michael Beenstock Department of Economics, Hebrew University of Jerusalem, Israel

3.1

Introduction

Does economic theory suggest that small countries are inherently different to big countries in the determination of output and the distribution of income? Should economists relate analytically to small countries, such as those analysed in this volume, differently to large countries such as the UK and the US? Or is it the case that small countries simply happen to be miniatures of an ideal economic type, while big countries happen to be larger versions? There is no separate medical theory for short and tall people. Should the same apply to economic theory, and economic size? Surprisingly, these questions have not been previously addressed in any systematic fashion, and the answers to them are not clear. A possible reason for this omission is that size is a non-issue, which does not merit attention. But it obviously is an issue for some. Small country economists are frequently berated by their large country counterparts, “We can fit the whole of your pint-sized country into New Jersey, not to mention Texas, so what is the sense in applying regional economics there?” The answer may, of course, be that it is worth applying regional economics to New Jersey too instead of simply treating it as a homogeneous economic entity. On the other hand there may be some wisdom in the “New Jersey Critique”. My own view is that size in itself is not important. What matters is regional heterogeneity. It may or may not be sensible to treat a small country as a “New Jersey”. If it happens to be regionally homogeneous it will be sensible, but if its regions are heterogeneous in terms of demographic and economic structure, the opposite is true. The same applies to large countries. The case for regional disaggregation in large countries does not simply follow from their size. If regions are homogeneous there will be no case. Therefore, there may be more sense to regional economics in a small country, where there is regional diversity and little economic integration, than in a large country, which is well integrated and homogeneous. In this chapter I attempt to bring together various strands of economic theory, which suggest that size might matter for the determination of aggregate output, its volatility, and for economic inequality. In doing so I distinguish between two aspects of size, population and territory. For given territory, more populated countries are obviously larger. However, territorial size might matter too. If A and B are two countries with the same population size, but A has more territory than B, we might expect A and B to have different income generating processes for

26

Michael Beenstock

reasons that go beyond the obvious differences in the supply of land and the greater distances in A. The issues raised are treated separately, although in principal they are related. This is obviously done for the sake of simplicity. I begin by considering two microeconomic issues. The first concerns the effect of labour market size upon the level of income and its distribution. To motivate the discussion I consider the unification of two separate, small labour markets into one large one. After critically examining the predictions of the influential Roy Model, which turns out to be insufficiently general, we are nevertheless able to make theoretical predictions of the effects of labour market size upon the level of income and its distribution. In particular, we can bound any increase in economic inequality that results from enlargement. However, in many cases inequality may vary inversely with size. The second microeconomic issue concerns the relationship between size and social interaction. In this case I consider population density, which determines physical distance between individuals. For given populations a large country has more territory. If the forces of social interaction are weaker in more dispersed populations, the level of income will tend to be higher in countries with greater population density and economic inequality will be lower. Four macroeconomic issues are considered. The first is concerned with economies of scale, according to which total factor productivity should be greater in larger economies. In this case size is measured by the labour force, which is in turn related to population size. There are several separate issues that are involved with this claim, which, however, do not seem to carry much empirical support. Therefore, it does not appear to be the case that larger countries enjoy greater scale economies. The second macroeconomic issue is concerned with the relationship between size and the volatility of output growth. In this case size is measured by the number of business establishments. Since risks are more diversified in larger economies, we expect that the volatility of output should vary inversely with size. However, these size effects are rapidly exploited so that for all practical purposes there is unlikely to be less volatility in larger economies. In short, this issue turns out to be red herring. The third macroeconomic issue, which also turns out to be something of a red herring, concerns the effect of population size on the distribution of income. It is well known that if the income data are independent, asymptotic convergence upon the limiting distribution occurs rapidly. Indeed, it occurs much too rapidly to be of any practical relevance for present purposes. However, if the data are dependent, due to social and macroeconomic phenomena that induce correlation among individuals, matters may be different and asymptotic convergence may be slower. Since income data are likely to be dependent we ask whether dependence is sufficiently strong to be of any practical relevance for our present concerns. The answer seems to be that it is of no practical relevance. This catch of red herrings is not entirely valueless. It is important to know a red herring when you see one. This was not apparently the case in Chandra (2003),

Country Size in Regional Economics

27

who argues that diversification may be important in the analysis of regional economic volatility. The fourth macroeconomic issue is concerned with the “New Jersey Critique”, already mentioned, and focuses upon the question of regional disaggregation in small economies. When does it make sense to regionalize, and how does this depend upon size? It is argued that the answer is only indirectly related to size. What matters is factor mobility, especially labour, and the nature of regional housing markets. Since geographical labour mobility varies inversely with distance, regional labour markets are likely to be more segmented in economies with larger territories. Indeed, it may or may not make sense to regionalize in large as well as small countries. However, in small countries regional labour and housing markets are likely to be more integrated because the distances between them are naturally shorter.

3.2

Microeconomics

3.2.1

Labour Migration and Wage Inequality

Assume that there are two countries A and B, which are ex ante the same in the sense that the distribution of ability among workers in A is similar to its counterpart in B. Capital is perfectly mobile between A and B, so that rates of return to capital in A and B are equated. By contrast, labour is not geographically mobile. Initially, residents of A have no right to work in B and vice-versa. We consider the analytical implications of creating one large economy out of A and B. In practice this means that the unified economy covers more territory, and that workers face job choices, which are more distant from them. However, our concern here is with the greater job choice that unification offers to workers in A and B. Indeed, I abstract by assuming that the unified economy is frictionless so that workers do not have to take distance, language and cultural barriers etc into account when making job choices. In the unified economy there is more job choice for all workers. Residents of A might find that they are better off working in B and vice-versa. I also abstract from other economic and political implications of economic unification, and focus entirely upon labour market size and job choice. If A and B are formally independent, but their labour markets are fully integrated, they are economically unified in our terms. For example, accession to the European Union enables workers access to the EU labour market, even though EU members are formally independent. The opposite applies in Russia, where in this large country labour markets are not integrated because Russians are still not free to choose where they live and work. Generally, however, larger countries have larger labour markets and greater job choice.

28

Michael Beenstock

The main question considered is: What is the effect of labour market unification on the level and distribution of income? To initiate the discussion, I apply the influential Roy Model (Roy 1951) to the question in hand.

Roy’s Model Prior to their unification the standard deviations of the logarithm of earnings (w = logW) in countries A and B are VA and VB and the overall standard deviation is V. It may be shown that:

V2

nV A2  (1  n)V B2  n(1  n)( w A  wB ) 2

(3.1)

where n = NA/N denotes the proportion of the population employed in A, and NA + NB = N is the total population, which is assumed to be fixed. Equation (3.1) shows that overall inequality varies directly with local inequality, as measured by VA and VB, and global inequality, as measured by the mean wage gap between A and B. Next, A and B are unified into one “big” country, and workers from A are allowed to migrate to B and vice-versa. We begin by considering the implications of the influential self-selection model, originally proposed by Roy, for inequality after the unification. Post unification variables are asterisked, e.g. n* denotes the proportion working in A after the unification. The key question is: what does Roy’s model predict for V* and how does it compare with V? Let wAi = PA + UAi denote the log wage of individual i = 1, 2,…,N if he works in A, and wBi = PB + UBi if he works in B. The Ps capture the general level of wages in A and B, which may be influenced by specific phenomena (such as weather, natural resources, etc) in A and B. The Us capture the idiosyncratic component of earnings as determined by heterogeneity in productivity. Each individual knows his own UA and UB. If he has a comparative advantage from working in A, then UA > UB. We assume that workers from A and B are ex ante similar in terms of their heterogeneity. This implies that on average A workers are no better or worse than B workers. In this case we may assume that:

wj

Pj

j

A, B

Roy’s rational selection model assumes that workers choose to work wherever their earnings are the larger. Individual i chooses to work in A if wAi > wBi, otherwise he works in B. He makes this choice regardless of whether before unification he worked in A or B, because his only consideration is income, not where he lives or works. Even if the general level of wages is higher in B because PB > PA, he will choose A if his comparative advantage of working in A is sufficiently large, i.e. Di = UAi - UBi > PB - PA. Roy’s model assumes that UA and UB have a bivariate normal distribution with variances V2A and V2B, and covariance VAB. If VAB > 0, workers who expect to do better than average in A also expect to do better than average in B. If VAB = 0,

Country Size in Regional Economics

29

there is no correlation between i’s productivity in A and B. Finally, if VAB < 0, workers who do better than average in A expect to do worse than average in B. We define s = (V2A + V2B - 2VAB)½, c = (PA - PB)/s, U = corr(UAi Di) and O = I(c)/)(c), where I( ) and )( ) denote the standard normal density and its cumulative counterpart respectively. It may be shown (Heckman and Honoré 1990) that:

O 2 (V A  V AB ) s

w A*

wA 

V *A2

V A2 [1  U 2 (cO  O 2 )]

(3.2) (3.3)

Equation (3.2) states that unification raises the average wage in A provided V2A > VAB. Note that VAB = VBrAB, hence the variance in B would have to be large relative to A for this condition not to be fulfilled. In this “standard” case migrants are positively selected, and post unification earnings rise on average. However, if this condition is not fulfilled, migrants are negatively selected, and post unification earnings fall on average. Equation (3.3) states that the post unification variance in earnings must be smaller than its pre-unification counterpart, because U2 is bounded between zero and unity. This happens regardless of whether migrants are positively or negatively selected. Similar conclusions apply to B. Positive selection implies that unification increases average earnings in B and the variance of earnings in B declines. In short, unification reduces inequality in A and B, and it raises the overall level of earnings in Roy’s model. The welfare implications of unification are unambiguously beneficial, because workers have more choice, average earnings rise, and nobody has lower earnings. It has already been mentioned that unification reduces inequality within A and B. What, however, happens to inequality as a whole? If n = n*, i.e. there is no net migration, Equation (3.1) implies global inequality (V*) will be reduced. If the mean wage gap between A and B (d*) happens to narrow, Equation (3.1) indicates that this will enhance equality. Equation (3.2) implies that the change in the wage gap is equal to:

d *2

d2 

O2 s

2

(V A2  V B2 ) 2  2d

O 2 (V A  V B2 ) s

(3.4)

Equation (3.4) states that if the initial level of inequality in A and B was the same, then d* = d. If d > 0 and V2A < V2B then the square of the wage gap may fall. The same applies when d < 0 and V2A < V2B. In this special case (n = n* and d*2 < d2) global inequality is reduced by unification. More generally, we cannot say what happens to global inequality because Roy’s model does not predict n*. For example, if n* > n (net inward migration to A) and inequality in A after unification happens to be greater than in B, this will tend to increase overall inequality. Also, we cannot assume that d*2 < d2. In summary, Roy’s Model predicts that unification increases the level of earnings in both regions, and it reduces inequality (as measured by the variance) in

30

Michael Beenstock

both regions. However, it has nothing to say about interregional inequality and therefore the overall level of inequality. Roy’s model has been widely used to simulate the effects of self-selection on inequality (e.g. Gould 2002). However, Heckman and Honoré (1990) have demonstrated that the conclusion that self-selection reduces inequality within groups (in our case inequality in A and B) is the peculiar result of Roy’s parametric assumptions. For example, if instead of being normally distributed, UA and UB have a bivariate Pareto distribution (which like the normal is symmetric), then unification may increase inequality in A instead of reducing it as in Equation (3.3). In short, the predictions of Roy’s model regarding the implications of unification for inequality are not robust to its parametric assumptions. The predictions of the Roy Model are sensitive to the way the model is set up. We shall therefore have to look further afield to obtain a more general understanding of the relationship between unification and inequality. What does Gini analysis have to say about the matter?

Gini Analysis Prior to unification the Gini coefficients for earnings (W) in A and B are denoted by GA and GB respectively. The overall Gini (Yitzhaki 1994) is exactly equal to:

G DI AG A  (1  D )I B GB  Gb

(3.5)

where D denotes the share of A in total earnings, IA denotes the coefficient of overlapping between the distribution of earnings in A with respect to B, IB denotes the coefficient of overlapping for B with respect to A, and Gb denotes the inter-regional Gini. The coefficient of overlapping measures the degree of stratification between earnings in A and B. Earnings are completely stratified when, for example, the lowest earner in A earns more than the highest earner in B. In this case IA = IB = 0. If the range of earnings in A and B is the same, there is perfect overlap in which case IA = IB = 1. If the earnings range in A falls inside the range in B then 2 > IA > 1 and 0 < IB < 1. Equation (3.5) reveals the relationship between global and local inequality. It implies that global inequality (G) may move in the opposite direction to inequality in A and B. This will happen, for example, when inequality in A and B increases, but inter-regional inequality (Gb) falls. Equation (3.5) further indicates that global inequality is independent of local and inter-regional inequality because it depends upon D and the degree of stratification (I). Therefore G may change without any changes in GA, GB, and Gb. For example if there is more inequality in A (GA > GB), and the share of A in total earnings (D) increases, it is obvious that G must increase. Inter-regional Gini for J regions may be written as: J

Gb

2

J J 1

¦ D j ( R j  R )(W j  W ) 1

W

(3.6)

Country Size in Regional Economics

31

The term J/ (J-1) corrects for degrees of freedom and tends to unity as the number of units tends to infinity. Note also that Equation (3.6) uses the average rank out of N of individuals in region j rather than region j’s rank out of J, as in Pyatt (1976). Pyatt’s between-group Gini ranks regions rather than individuals, and does not unfortunately satisfy the decomposition theorem in Equation (3.5). Hence we do not use it here. In short, Equation (3.6) uses the regional mean of the individual ranks rather than the rank of the regional means. When J = 2, as in our case, Equation (3.6) becomes: Gb

4

D ( R AW A  R BW B )  R BW B  0 .5W W

(3.7)

Equation (3.7) implies complete interregional equality, i.e. Gb = 0, under two conditions: when average earnings are the same in both regions, and when the average ranks are the same. The latter happens when the average rank is a half. Otherwise inequality varies directly with the earnings gap between the two regions, and the gap between the average ranks. Suppose, as in the Roy model, that after unification workers only migrate if this raises their earnings. It is obvious that mean global earnings must increase. This will happen if at least one individual migrates to improve his earnings. Hence:

W*tW

(3.8)

Using a first order stochastic dominance argument Yitzhaki (1982) has shown that: W * (1  G *) t W (1  G )

(3.9)

must hold. Note that Equations (3.8) and (3.9) are, unlike Equation (3.3), independent of arbitrary parametric assumptions. Expressing the percentage increase in global earnings by x, Equations (3.8) and (3.9) may be combined into the following inequality:

'G 

x (1  G ) 1 x

(3.10)

Suppose, for example, that G = 0.4 and x = 0.05, Equation (3.10) bounds the increase in Gini to 0.0286. Unification may increase global inequality, but it cannot do so by more than 0.0286. On the other hand, Equation (3.10) implies that unification may also reduce global inequality. The upper bound on the increase in inequality varies inversely with x. If x = 0.01 then the upper bound is 0.0059 instead of 0.0286. Also, this upper bound varies inversely with the initial level of inequality. If G = 0.5 and x = 0.05 then the upper bound is 0.0238 instead of 0.0286. To summarize: unification increases the global mean of earnings because some are better off and nobody is worse off. Unification may either increase or decrease global inequality, but it cannot increase it by more than the upper bound given by Equation (3.10). This is all that can be said. The upper bound on the increase in Gini implies bounds upon the components of Equation (5). However, unlike the

32

Michael Beenstock

Roy model, the change in intra-regional inequality cannot be predicted, nor can the change in interregional inequality be predicted. 3.2.2

Social Interaction

Suppose that two countries A and B are identical except for the fact that A has more space than B. If the populations are the same, the inhabitants of B will be living in closer proximity to one another than in A. For example, Finland and Israel have similar populations, but the territory of Finland is much larger than the territory of Israel. Finns are less likely to live close to each other than Israelis. If it is reasonable to expect that social interaction varies directly with population density, we may expect more social interaction in B than in A. Keeping up with the Jones is stronger in B than in A because the Jones live closer to each other in B than in A. The following simple social interaction model is hypothesized:

Wij

D j  E j E (W j )  J j X ij  uij

j

A, B

(3.11)

Equation (3.11) states that individual earnings depend upon the general level of earnings, observed individual characteristics such as education (X), and unobserved heterogeneity (u) with u ~ N(0, V2u). If E = 0 there are no social interactions. We expect that EB > EA because B is more densely populated than A. Mean earnings in social equilibrium are equal to: Wj

D j J jX j

(3.12)

1 E j

In Equation (3.12) the term 1/(1 - Ej) > 1 is the social multiplier. It implies that social interaction enhances average earnings. In the absence of social interactions, when E = 0, average earnings are obviously smaller. If A and B are otherwise identical apart from their population density, i.e. DA = DB etc, the greater social interaction in B implies, according to Equation (3.12), that average earnings in B will exceed their counterpart in A. Note that the effect of the exogenous X variables on W depends upon the social interaction coefficient. Exogenous shocks have a greater effect on economies where the social multiplier is larger. Equation (3.12) implies, for example, that the marginal return to X (e.g. education) in B is greater than its counterpart in A, because J/(1 - EB) > J/(1 - EA). No doubt the greater population density in B will affect other phenomena apart from E. For example, D may be larger or smaller. Presumably rents will be higher in B because the supply of land is smaller, and congestion will be greater. These factors would tend to reduce D. On the other hand, social networks are likely to be more developed in more densely populated societies, which would tend to raise both D and J. Also, social interactions are likely to affect the general level of X. This implies, the residents of B are likely to be more educated than residents of A.

Country Size in Regional Economics

33

It may be shown that the variance of earnings implied by Equations (3.11) and (3.12) is equal to: 2 V Wj

2  V uj2 J 2j V Xj

(3.13)

in which case the coefficient of variation (CV) for earnings is equal to:

CV j

(1  E j )V Wj

(3.14)

Dj J jX j

Equation (3.14) states that earnings will be more equally distributed where there is greater social interaction, since CV varies inversely with E. This happens because in Equation (3.13) the variance in absolute earnings is independent of E, whereas the mean varies directly with E. In the limit, when E = 1, Equation (3.14) predicts that there will be complete equality since CV = 0. In this case we all lose our individual heterogeneity and become identical to our neighbours. If regions A and B are otherwise identical, we can expect more equality in B than in A, because there is more social interaction in B. Note that this result is not an artifact of data measurement. Had Equation (3.11) been specified in logarithms (i.e. the dependent variable is w instead of W, and instead of E(W) the independent variable is logE(W), but not E(w)) the resulting nonlinear equation implies a result similar to that in Equation (3.14).

3.3

Macroeconomics

3.3.1

Economies of Scale

If countries or regions benefit from scale economies then total factor productivity will be greater in larger countries or regions. Broadly speaking there are two types of scale economy, which may be illustrated using the Cobb-Douglas production function:

Q

AK D LE

(3.15)

where Q denotes GDP, A denotes total factor productivity, K denotes the capital stock and L denotes employment. Returns to scale on factors of production are constant when D + E = 1 and they are increasing when D + E > 1. If product markets are perfectly competitive it is difficult to make a good theoretical case in favour of increasing returns to scale. However, if product markets are imperfectly competitive the case for increasing returns to scale is easier to make. When competition is imperfect firms face downward sloping demand curves and equilibrium occurs at the point of production where average cost is equal to average revenue. At this point average cost must be falling because the slope of the demand curve is negative. This simple intuition has formed the theoretical

34

Michael Beenstock

basis for much of the New Economic Geography, which has attached importance to increasing returns (Krugman 1991). Larger economies, i.e. economies that have larger K and L, will have lower average costs, which imply that D + E > 1. A second source of increasing returns is related to A in Equation (3.15). There may be economies of scale in the use of social infrastructure and the provision of public goods. For example, there are fixed costs in defending a country against invasion. It would not cost more to defend Israel if it was larger, but the costs would be defrayed over a larger population. Hence there are scale economies in the provision of defence services (Beenstock 1993). In short, larger countries may be more able to defray the costs of public goods and services, in which case A will vary directly with the scale of production. According to the New Economic Geography scale economies and urban growth are inextricably interwoven. However, our concern here is with the size of entire economies, not cities. Brakman et al. (2001) suggest that external economies of scale may apply to entire economies too. However, there is no evidence to support this conjecture. What is the empirical evidence in favour economies of scale at the country or regional level? This question has been investigated by cross - section comparisons between countries over the same time period, and time series comparisons for the same country. In the former case larger economies are compared with smaller ones to determine whether they enjoy a scale advantage. In the latter case the same country is compared over time to determine whether it enjoyed scale advantages when it was larger. A good example of the latter is Israel, which due to immigration serves almost as a laboratory experiment for these purposes. The empirical evidence in favour of scale economies is weak (Beenstock 1997). Nor do cross section studies indicate scale advantages to larger economies. Davis and Weinstein (1999) do not find that the size of the home market induces economies of scale, and Hanson (1997) fails to find conclusive evidence of spatial wage effects. This evidence suggests that scale economies, which may apply to firms and cities, do not apparently apply to countries and regions.

Endogenous Growth Theory Thus far we have considered the effect of scale on the level of GDP. Another branch of the literature has considered the effect of scale on the rate of growth of GDP. The so-called New Growth Economics or Endogenous Growth Theory (e.g. Grossman and Helpman 1991) predicts that larger countries will grow faster, and that countries that become larger will grow faster. This theoretical size effect stems from the fact that it is more profitable to develop new technologies in larger economies, where there will be more customers for the new technology, than in smaller economies. Since endogenous innovation depends upon the economic incentive to innovate, there will be more innovation in larger economies, and larger economies will grow faster than smaller ones. According to the empirical literature this size effect does not exist (Jones 1995; Segerstrom 1998). There is no evidence that larger countries are more innovative than smaller ones. Indeed, international differences in innovativeness do not

Country Size in Regional Economics

35

depend upon their size. Moreover, in an age of globalisation what matters is the size of the world economy rather than the size of individual countries, because new technologies can be sold abroad and not just nationally. Hence, the size effect in endogenous growth models would only apply to closed economies, or to the entire world. The size of the world economy is growing over time. If the size effect existed globally, we should find that economic growth and innovation is greater today than it was in the past, because the world economy is much bigger today than in the past. Accordingly, the rate of growth of the world economy should be increasing. However, the world rate of economic growth has not been increasing, hence this size effect does not appear to exist either between countries of different sizes, or in the world economy as a whole over time. In summary, country size has no apparent effect on either the level of GDP or its rate of growth, both within countries and between countries. 3.3.2

Growth Volatility and Diversification

It is a well known empirical fact that world GDP is less volatile than the GDP of the individual countries that it comprises. It is also well understood that the reason for this is that business cycles are not perfectly synchronized across countries. If they were, then world GDP would be just as volatile as the GDP in its component countries. This phenomenon creates the statistical illusion that larger economic units are less volatile than smaller ones. The same applies to countries and their regions. GDP has a lower variance than GRP because regional business cycles are imperfectly synchronized. However, this does not necessarily imply that larger economies are less volatile than smaller ones. There are two quite separate issues here: segmentation and diversification. Segmentation occurs when region A and region B are not perfectly integrated. For example, there is segmentation when workers from A are not allowed to work in B and vice-versa, or A has different fiscal and monetary authorities to B. Diversification refers to scale and the law of large numbers. In a more diverse economy economic risks are spread more widely, and volatility is reduced. Larger economies are likely to be both more segmented and diversified. However, whereas larger economies are naturally more diversified they need not be more segmented. Segmentation is partly inherent or social and partly political. Segmentation within countries may be induced by language barriers (e.g. Canada, Belgium and Switzerland), class barriers (e.g. India) and culture (Jews and NonJews in Israel), or it may be politically motivated. Federal countries will tend to be more segmented than countries that have strong central governments.

Diversification The relationship between size and segmentation is much more difficult to understand than the relationship between size and diversity. Several studies show, not surprisingly, that labour markets in the US are less segmented than in Europe, despite the size of the US in terms of population and territory. US workers are

36

Michael Beenstock

geographically more mobile than their European counterparts perhaps because Europe is more culturally diverse than America. In what follows I therefore address the relationship between size and diversity because this is much easier to analyse. A country comprises n economic units such as workers or firms. Total output is the sum of all the individual outputs. If the units are of similar size the rate of growth of output (g) is equal to the mean growth rate. The rate of growth of unit i is assumed to have a deterministic component Ni and a stochastic component Hi, with H ~ N(0, V2i):

gi

N i  Hi

(3.16)

The overall rate of growth is equal to:

g

1 n ¦ gi ni 1

and its variance, or volatility, is equal to: var( g )

1 n

2

n

n

¦ ¦ V ij

i 1

j 1

1 n

2

n

2 ¦ Vi 

i 1

1 n

2

n

¦ ¦ V ij

(3.17)

i 1j 1 iz j

The first term in Equation (3.17) tends to zero with n. To show this, denote the unit with the highest variance by v. If all units had this variance then the first term would be equal to nv/n2 = v/n, which obviously tends to zero with n. The first term will tend to zero at a faster rate than this because it is necessarily smaller than v/n. The second term in Equation (3.17) is approximately equal to the average covariance, which is defined as:

V

n n 1 ¦ ¦ V ij n( n  1) i 1 j 1

(3.18)

j zi

hence var(g) asymptotically tends to the average covariance since n -1 tends to n. This demonstrates that what matters for the variance of growth is not the variances of the individual units, which get diversified away, but the average covariance. If the average covariance happens to be zero then all of the growth risk will be diversified away and volatility will be zero. Since economic risks tend to be positively correlated the asymptotic variance is greater in economies in which risks are more correlated. Equation (3.17) demonstrates that volatility varies inversely with size as measured by n. This diversification effect suggests that economic growth in larger economies will tend to be less volatile than in smaller economies. The crucial question here is how rapidly does this diversification effect occur? The empirical answer seems to be that it occurs so rapidly that for all practical purposes it does not matter. Fama (1976) showed that for as few as 20 units the benefits of diversification are fully exploited. Since the number of economic units in small

Country Size in Regional Economics

37

countries typically runs into the thousands, we may safely conclude that the benefits of diversification have already been fully exploited.

Segmentation This suggests that the reason why larger economic aggregates display less volatility than smaller ones is not because of their size per se, but because of segmentation. To emphasize this argument, we assume that A and B are fully diversified so that the variance equals the average covariance, and that due to segmentation the average covariance in A and B happen to differ. Let V2A be equal to the average covariance in A, V2B be equal to the average covariance in B, and let V2 be equal to the average covariance of A and B combined. The latter must be smaller than either of the former, because due to segmentation there is more scope for diversification when the units of A and B are combined. The units of A are more correlated among themselves because they belong to the same segment, than they are correlated with the units of B, which belong to a different segment. The same applies to the units of B; the intra-group correlation is greater than the intergroup correlation. In short, segmentation means greater intra-regional correlation than inter-regional correlation. This argument can be made more formally using the following definitions:

Hj

1 nj ¦ H ji nj i 1

V j2

var(H j ) V j

V2

(

j

A, B (3.19)

n A 2 2 nB 2 2 n n ) V A  ( ) V B  2 A 2 B rV AV B n n n

where r denotes the correlation coefficient between the means of H in A and B. It measures the degree of segmentation between risks in A and B. If the risks are perfectly positively correlated (r =1) there is no segmentation at all, in which case Equation (3.19) implies that:

V

nA n VA  B VB n n

i.e. the overall standard deviation is a weighted average of the average covariances in A and B. More generally, however, when r < 1, the overall standard deviation is less than the weighted average of the average covariances. Indeed, when the risks are uncorrelated, i.e. when there is complete segmentation, the final term in Equation (3.19) is zero, and the overall variance attains a minimum.

38

Michael Beenstock

3.3.3

Size and the Central Limit Theorem

The true distribution of earnings is generally unknown, although earnings are often assumed to have a lognormal distribution. The question we ask here is: does the empirical distribution depend upon size, as measured by the number of earners? If the answer is yes, then we might find arbitrary differences in the distribution of earnings between A and B simply because A happens to be more populated than B. Common sense rightly suggests that the chance of observing extreme values in either tails of the distribution is greater in larger countries. We are more likely to find extreme talent in the US than in Andorra simply because the US is so much larger. By the same token Andorra is likely to have fewer murderers and other extreme cases of perverse behaviour. Without knowledge of the true distribution it is difficult to answer the question that has been posed, because we cannot say how tail, and other, phenomena depend upon size.

Lindeberg Condition The celebrated central limit theorem (CLT) states that whatever the true distribution happens to be, the standardized mean in a sample of n tends to be normally distributed as n tends to infinity. This means that we do not have to know the nature of the true distribution for CLT to hold. The true distribution may have a wide variety of shapes, however, CLT requires certain conditions to be fulfilled. The most important of these is the Lindeberg condition, according to which the average contribution of the extreme tails to the variance of the sum be negligible in the limit. If CLT does not hold, i.e. the standardized sample mean does not tend to be normally distributed, it means that the true unknown distributions do not satisfy the Lindeberg, and other relevant conditions, and that the contribution of extreme tails is not negligible. If, on the other hand, CLT applies in practice, the opposite is true. In this case, we can ask how CLT depends upon n. The answer will be informative regarding the question posed above: how does the empirical distribution depend upon size? Therefore the question posed is not answered directly, because this would require knowledge of the true distribution, which is generally unknown. Instead, it is answered indirectly through the validity of CLT. CLT comes in a large variety of versions depending upon the nature of the data generating process (DGP). Hence, to answer the question posed, even indirectly, we need to specify the DGP for earnings (W). The DGP is assumed to be linear:

Wi

P i  ei

(3.20)

where the stochastic component of earnings (e) has an unknown distribution with E(e) = 0. The limiting distribution of W is F(W), which is unknown because F(e) is unknown. If Wi are iid observations, i.e. they are identically and independently distributed, then Pi = P, Vi = V < ũ and Ŭ 0, and Vik = 0. The mean and variance

Country Size in Regional Economics

39

of earnings is the same for everybody, and earnings are entirely independent of each other. In this simplest case the Lindeberg - Levy CLT states: a o n (Wn  P ) / V n  N (0,1)

(3.21)

Equation (3.21) states that standardized average earnings in a sample of n are asymptotically normally distributed, despite the fact that F(W) is unknown. If instead of imposing homogeneity upon the DGP, we do not restrict Vi = V and Pi = P then the Lindeberg - Feller CLT states: a o n (W n  P ) / V n  N (0,1)

(3.22)

d o n (W n  P )  N (0, V 2 )

(3.23)

V n2

1 n 2 a o V ¦V i  ni 1

2

1 a max (V i2 / V n2 )  0 o n

(3.24) (3.25)

Equation (3.22) is the same as Equation (3.21) but for the fact that the population average for P is specified. Equation (3.23) states that the limiting distribution of the sample mean is normal with mean equal to the population mean and variance equal to the mean variance in the population. Equation (3.24) states that sample variance is asymptotically equal to the mean variance in the population, provided Equation (3.25) holds. The latter is the Lindeberg condition already mentioned, which restricts the tails to have a negligible effect on the variance. If it does not hold, nor will Equation (3.22). It is well known that Equations (3.21) and (3.22) approach the normal very rapidly with n. For example, the student t - distribution approaches the normal when n is as small as 40. Therefore these versions of CLT apply so rapidly that they are irrelevant to earnings distributions where n typically runs into the millions. Matters are quite different, however, when observations are assumed to be dependent, i.e. when there is dependence between earnings and Vik z 0. White (2001) reports CLTs for this case where the data are covariance stationary. Asymptotic convergence rates are, not surprisingly, slowed down by the dependence in the DGP, because we learn less from each individual the more they are dependent. If they were completely dependent we would learn nothing. Therefore it is the dependent case, which makes CLT relevant to our discussion.

Spatial Dependence and the Asymptotic Distribution of Gini Zitikis and Gastwirth (2002) show that when the data are independent Gini is root n asymptotically normal, i.e. CLT applies to Gini. This is not surprising because 1 Gini is a weighted mean rather than a simple mean. In this section I show that the 1

I wish to thank Yosi Rinott for advice and Aviv Zohar for research assistance.

40

Michael Beenstock

same applies when the data are dependent. Because of macroeconomic and social interactions individual earnings are often highly dependent. This may also be true of other economic phenomena such as the profits of firms and their output, employment etc. We may refer to this phenomenon as spatial dependence. Spatial dependence may be close, as in the case of social interactions between next-door neighbours, or remote, as in the case of interactions induced by commuting. Or, social interactions may occur within cultural groups located in different parts of the country. Surprisingly, little is known about the effects of spatial dependence upon convergence rates in CLT. One type of spatial dependence is circular dependence in which interactions occurs within one’s immediate circle of neighbours. I have carried out a Monte Carlo investigation of CLT for Gini with first-order circular dependence among n individuals living in a square lattice. Each individual has four neighbours, one on each side. Immediate neighbours are assumed to influence each other through a social interaction coefficient 0 < E < 1, which gives rise to circular dependence. The specific data generating process (DGP) investigated is similar to Equation (3.11) except that the neighbourhood mean replaces the global mean:

Wi

P  EWi*  ei

(3.26)

where W* denotes the average value of W among i’s four next-door neighbours. Note that when E = 0 and Pi = P Equation (3.26) reverts to Equation (3.20). Equation (3.26) takes account of first-order spatial spillovers between neighbours. As in Equation (3.20) e continues to be iid, but W is obviously no longer iid. We carry out a Monte Carlo study of the Gini for W given Equation (3.26) as the DGP. We set a value for E between zero and unity, and n, and then sample e 2,500 times from a lognormal distribution with mean zero and unit variance, thereby obtaining 2,500 Gini coefficients. We ask three main questions. The first concerns whether or not mean Gini has some limiting value, as the population size tends to infinity. The second concerns the rate of convergence of mean Gini on its asymptotic value. How rapidly does convergence occur in terms of n? The third question concerns the nature of the distribution of Gini around the mean. In particular, is Gini normally distributed? And if so, how rapidly does this distribution converge upon the normal? The Jarque-Bera statistic (JB) is used to test for normality. It has a chi-square distribution with 2 degrees of freedom because it tests whether the 3rd moment is zero, and whether the 4th moment is equal to 3, as under the null hypothesis. Values of JB greater than 3.78 are statistically significant at p = 0.95. The Monte Carlo results show that when E = 0 Gini converges upon its asymptote when n = 200, and the distribution of Gini is normal, which confirms Zitikis and Gastwirth (2002). When E = 0.5 Gini converges upon its asymptote when n = 800. This number rises to about 1,300 when E = 0.9. These results indicate that the speed of asymptotic convergence varies inversely with E. However, Gini is normally distributed in these cases even when n is as small as 200. The main conclusion is that circular dependence in the income generating

Country Size in Regional Economics

41

process slows down the rate of convergence, but by not enough to be of any practical relevance. 3.3.4

Regional Disaggregation

In this section we ask whether it makes sense to regionally disaggregate small countries. In doing so we address the “New Jersey Critique” raised in Section 3.1. We assume that a country consists of regions, which share laws, central government, language and other phenomena that characterize countries. If capital and traded goods are perfectly mobile between regions, returns to capital and the prices of traded goods are equated over the entire country. Labour may or may not be perfectly mobile between regions, hence wages are not necessarily equalized. Two kinds of labour mobility are distinguished. In the first, if workers from region A wish to work in region B they have to live in B. In the second, workers from A can commute to B, and vice-versa. In the first case domicile and location of work cannot be separated, whereas in the second they can. It is obvious that travel costs and distance between A and B will be important here.

The Model Each region produces two types of output: traded goods, which are produced using a common technology, and housing services. The latter depend upon the stock of building land, which is assumed to be fixed. It is obvious that under such circumstances house prices will not generally be equalized. Because building land cannot be traded the housing markets cannot be regionally integrated. There are countries such as Belgium where more than one language is spoken, and there are countries, especially less developed countries, in which the capital market is fragmented rather integrated. Also, there are countries such as Russia in which residential restrictions inhibit or prevent labour mobility between its regions. There are numerous countries such as Australia that have natural differences in regional endowments, such as natural resources. There may also be cultural and ethnic differences between regions that inhibit integration and induce regional segmentation. In such countries regional heterogeneity will be greater than in the simple model described below. The neoclassical model of regional equilibrium (Siebert 1969) assumes varying combinations and degrees of mobility in capital, labour and goods. Here, we introduce into the model the stock of building land (H), of which each region has a different endowment. The cost of living in region i is equal to Pi = PaPHi1-a, where 0 < a < 1, P denotes the price of traded goods, which is assumed to be equalized by free trade between regions, and PH denotes the price of housing services. There is a common, constant-returns-to-scale Cobb-Douglas production technology given by Qi = AKibLi1-b, where K and L denote capital and labour respectively and 0 < b < 1. Q is tradable and is sold at common price P in all regions. Finally, the demand for housing services or space in region i is assumed to vary directly with

42

Michael Beenstock

the population in the region and regional earnings, and inversely with the price of housing services: HiD = Li(PHi/P)-c(Wi/Pi)d. The structural parameters, a, b, c, and d do not vary by region, hence the regional equilibrium described below is symmetric, except for the fact that Hi varies by region. In equilibrium the supply of housing services is equal to demand, hence HiD = Hi. In all other respects, such as amenity endowments, all regions are assumed to be identical. Table 3.1. Taxonomy of regional equilibria

K: IMMOBILE L: MOBILE

Wi Wj

Li Lj

PH i PH j

S1

§ Hi ¨ ¨H © j

· ¸ ¸ ¹

· ¸ ¸ ¹

S1

§ · 1  c  ad ¸¸ b 2 ¨¨ © b(1  c  d )  1  a ¹

S2

b(1  a ) 1  a  bc

1

1

EXOGENOUS

Hi Hj

S4

§ Ki ¨ ¨K © j

· ¸ ¸ ¹

S3

(1  bd )(1  a) b(1  c  d )  1  a

S4

1 a 1  a  bc

§ Hi ¨ ¨H © j

S5

K: MOBILE L: MOBILE

S 2

§ Ki ¨ ¨K © j

S3

K: MOBILE L: IMMOBILE

· ¸ ¸ ¹

S 6

§ Ki · § H i · ¨ ¸ ¨ ¸ ¨K ¸ ¨H ¸ j j © ¹ © ¹ S 3 (1  db)  db S5 c  d (1  a) 1  S 4 (1  db) S6 c  d (1  a )

§ Li H j · ¸ ¨ ¨H L ¸ © i j ¹ S7

S7

1 c  d (1  a)

1

Country Size in Regional Economics

43

General Regional Equilibrium Table 3.1 reports various regional equilibria under different assumptions about factor mobility. Capital, labour and goods markets are assumed to be competitive and firms are assumed to maximize profits. S2, S4, and S7 are ambiguously positive. So are the other S coefficients, provided the income elasticity of demand for housing (d) does not exceed unity by too much. If both factors are mobile the regional distribution of employment (population) depends entirely on the regional distribution of building land. In this case house prices are equalized across regions and so are earnings. If labour is immobile, the mobility of capital ensures that earnings are equalized, but house prices will be relatively expensive in regions where the relative demand for housing is higher, i.e. with greater L/H. This means that although earnings are equalized, real earnings (Wi/Pi) are lower in regions where relative housing demand is greater. Finally, if capital is immobile and labour is mobile real earnings (Wi/Pi) must be equalized, but earnings and house prices must vary between regions. House prices will be relatively high in regions well endowed with capital, because labour will be attracted there, and relatively low in regions well endowed with building land. The same applies to the regional distribution of earnings. The regional distribution of employment (population) varies directly with relative endowments of capital and building land. Note that if commuting occurs, Wi = Wj and PHi = PHj even if capital is immobile. Hence, commuting creates the same regional equilibrium as labour and capital mobility. In general, however, Table 3.1 establishes that real earnings will tend to differ across regions under Classical equilibrium, unless labour happens to be perfectly mobile. Capital mobility and trade are not sufficient conditions for real earnings equality in our model because the regional distribution of building land is not necessarily uniform.

The New Jersey Critique If capital and labour are perfectly mobile, earnings and the returns to capital are equated, in which case there is no point to regional disaggregation, since one region is identical to the next. The same applies if there is commuting. Since commuting is more feasible in small countries, the New Jersey Critique is more likely to apply. The New Jersey Critique is an equilibrium concept and should apply therefore in the longer term, if it applies at all. In the short term real wages and the returns to capital may differ from their equilibrium values. What evidence is there in favour of factor price equalization within countries, and how does this depend upon country size? Beenstock and Felsenstein (2004) show that even in a country as small as Israel, real wages are not equated between regions. Indeed, the differences are large and persistent, and cannot be explained away by compensating differentials. The New Jersey Critique does not apply to Israel, despite its smallness. One wonders whether it applies in New Jersey. This piece of empirical evidence suggests that even if capital happens to be mobile, labour is not sufficiently mobile

44

Michael Beenstock

to induce real wage equalization between regions. Indeed, this result has been found for the UK (Johnston et al. 1996), the US (Eberts and Schweitzer 1994) and Brazil (Azzoni and Servo 2002) among other countries.

3.4

Conclusion

Economic theory has little to say about the effects of country size on the determination of national income and its distribution. Statistical theory has little to say too, beyond the fact that the laws of large numbers has already been fully exploited in even the smallest of countries. Small is irrelevant. If size doesn’t matter the New Jersey Critique is false. Regions, like countries, seem to be heterogeneous regardless of their size. If size doesn’t matter, the study of small countries is not inherently different to the study of large countries. Small countries obviously tend to be geographically more compact by nature, but this does not make them special.

References Armstrong HW, Read R (2002) The phantom of liberty: economic growth and the vulnerability of small states. Journal of International Development 14:435-438 Azzoni CR, Servo LMS (2002) Education, cost of living and regional wage inequality in Brazil. Papers in Regional Science 81:157-175 Beenstock M (1993) International patterns in military spending. Economic Development and Cultural Change 41:633-650 Beenstock M (1997) Business sector production in the short and long runs in Israel. Journal of Productivity Analysis 8:53-70 Beenstock M, Felsenstein D (2004) Mobility and mean reversion in the dynamics of regional inequality. mimeo Brakman S, Garretsen H, van Marrewijk C (2001) An introduction to geographical economics. Cambridge University Press Chandra S (2003) Regional economic size and the growth - instability frontier. Journal of Regional Science 43:95-122 Davis DR, Weinstein DE (1999) Economic geography and regional production structure. European Economic Review 43:379-407 Eberts RW, Schweitzer ME (1994) Regional wage convergence and divergence: adjusting for cost of living differences. Economic Review (Federal Reserve Bank of Cleveland) 39:224-231 Gould ED (2002) Rising wage inequality, comparative advantage, and the growing importance of general skills in the United States. Journal of Labour Economics 20:105-147 Grossman G, Helpman E (1991) Innovation and growth in the global economy. MIT Press, Cambridge MA Fama EF (1976) Foundations of finance. Basic Books, New York

Country Size in Regional Economics

45

Johnston R, McKinney M, Stark T (1996) Regional price level variations and real household incomes in the United Kingdom 1979/80 - 1993. Regional Studies 30:567578 Jones CI (1995) Time series tests of endogenous growth models. Quarterly Journal of Economics 110:495-526 Hanson GH (1997) Increasing returns, trade and the regional structure of wages. Economic Journal 107:113-133 Heckman JJ, Honoré BE (1990) The empirical content of the Roy model. Econometrica 58:1121-1149 Krugman P (1991) Increasing returns and economic geography. Journal of Political Economy 99:950-959 Roy AD (1951) Some thoughts on the distribution of earnings. Oxford Economic Papers 3:135-146 Pyatt G (1976) On the interpretation and disaggregation of Gini coefficients. Economic Journal 86:243-254 Segerstrom PS (1998) Endogenous growth without scale effects. American Economic Review 88:1290-1310 Siebert H (1969) Regional economic growth: theory and policy. International Textbook Company, Scranton, PA White H (2001) Asymptotic theory for econometricians. Revised. Academic Press Yitzhaki S (1982) Stochastic dominance, mean variance and Gini’s mean difference. American Economic Review 72:178-185 Yitzhaki S (1994) Economic distance and overlapping distributions. Journal of Econometrics pp 147-159 Zitikis R, Gastwirth JL (2002) The asymptotic distribution of the S-Gini index. Australian and New Zealand Journal of Statistics, 44(4):439-446

4

Measures of Regional Inequality for Small Countries

Boris A. Portnov1 and Daniel Felsenstein2 1

2

Department of Natural Resources and Environmental Management, University of Haifa, Israel Department of Geography, Hebrew University of Jerusalem Mount Scopus, Jerusalem Israel

4.1

Introduction

In ancient Japan, personal income was measured in terms of a “Koku.” One Koku was the amount of rice required to feed one person for a year, about 140 kg or 13 ounces per day. The income of a “Daimyo” or “great land owner” exceeded 10,000 Koku per year. The great Tokugawa Ieyasu, the first Shôgun of the Edo period (1543-1616), earned annually over four million Koku (Wayland 2003). The Koku system was both a simple and an ingenious measure of income inequality. It was not subject to inflation (only to personal appetite and availability of other food supplements) making it very suitable for both for longitudinal and crosssectional studies. For instance, by comparing regional Koku in years A and A+1, one could estimate that an average person in the central Yamashiro (Kyoto) province, who earned 7 Koku in year A and 7.7 Koku in year A+1, was 10-percent better off than the year before and twice as more affluent as his fellow citizen in the peripheral Hizen, who had to get by with only 3.5 and 3.8 Koku per year. However, this system of inequality measurement, though simple and affective for a pair-wise comparison, becomes nearly useless when we need to measure the inequality across more than two units (e.g., between Yamashiro, Hizen and Shimozuke provinces). Fortunately, no one seemed to have been concerned with such comparisons in those days. The computational problems associated with multi-group comparison of income inequality were noticed (apparently for the first time) by the American economist Max Lorenz. In his seminal paper published in 1905 in the Publications of the American Statistical Association, Lorenz highlighted several drawbacks associated with the comparison of wealth concentration between fixed groups of individuals. In particular, he found that while an increase in the percentage of the middle class is supposed to show the diffusion of wealth, a simple comparison of percent shares of persons in each income group may often lead to the opposite conclusion. For instance, while the upper income group in a particular period may constitute a smaller proportion of the total population, the overall wealth of this group may be far larger compared to another time period under study (ibid. pp.

48

Boris A. Portnov and Daniel Felsenstein

210-211). The remedy he suggested was to represent the actual inter-group income distribution as a line, plotting “along one axis cumulated percents of the population from poorest to richest, and along the other the percent of the total wealth held by these percents of the populations” (ibid. p.217). As he notes, “With an unequal distribution, the curves will always begin and end in the same points as with an equal distribution, but they will be bent in the middle; and the rule of interpretation will be, as the bow is bent, concentration [of incomes] increases” (p. 217). The Italian statistician Corrado Gini moved Lorenz’s ideas a step further, suggesting a simple and easy comprehendible measure of inequality known as the Gini coefficient. Graphically, the calculation of this coefficient can be interpreted as follows (Atkinson 1983): Gini coefficient =

Area between Lorenz curve and the diagonal Total area under the diagonal

Mathematically, the Gini coefficient is calculated as the arithmetic average of the absolute value of differences between all pairs of incomes, divided by the average income (see Table 4.1).1 The coefficient takes on values between 0 and 1, with zero interpreted as perfect equality. In 1920, the British economist Edward Hugh Dalton (1920) suggested an alternative measure of income inequality (G), which he estimated as the ratio between logarithms of the arithmetic (xa) and geometric (xg) means of total incomes:

G

where x a

log x a , log x g n

¦x i 1

i

/ n and x g

(4.1) n n

–x

u

u 1

(xi = total income of group i, n = number of groups under comparison). However, even Dalton himself did not attempt to test the proposed measure empirically due to the fact that the calculation of geometric means was vary laborious, if not impracticable, specifically if the number of individual incomes was large (ibid. p. 351). More recent empirical studies proposed and used a variety of additional inequality measurements, such as the Williamson index, Theil index, Atkinson index, Hoover and Coulter coefficients (Williamson 1965; Sen 1973; Atkinson 1983; Coulter 1987; Yitzhaki and Lerman 1991; Sala-i-Martin 1996; Kluge 1999; WBG 1999). Mathematical formulae for these commonly used inequality measures are given in Table 4.1. These inequality measures basically fall into two classes: measures of dispersion (e.g., the coefficient of variation and Williamson index), and measures

1

The computation includes the cases where a given income level is compared with itself.

Measures of Regional Inequality for Small Countries

49

of entropy. The measures in the latter class are given to the following generic formula: GE (D )

1 2 D D

ª 1 n § y ·D º « ¦ ¨ i ¸  1», «¬ n i 1 ¨© y ¸¹ »¼

(4.2)

where n is the number of individuals (groups) in the sample, yi is the income of individual i; y is the arithmetic mean of individual incomes, and parameter D represents the weight given to differences between incomes at different parts of the income distribution [low values of this parameter make the inequality measure more sensitive to changes in the lower tail of the parameter distribution, while high values make it more sensitive to changes in its upper tail]. The values of GE range from 0 to ũ, with zero representing the absolutely even distribution of incomes (WBG 1999). Table 4.1. Commonly used measurements of regional inequality Coefficient of variation (CV) (unweighted)

1 ª1 n  yi  y 2 º» ¦ « y ¬n i 1 ¼

CV

1

2

Theil index (TE(0)) y 1 n TE (0) ¦ ln n u 1 yi

Population weighted coefficient of variation (Williamson index (WI))

WI

HC

Gini (U) (unweighted)

Gini

1 n n ¦¦ yi  y j 2n 2 y i 1 j 1

1/ 2

Atkinson index (AT)

AT

Hoover coefficient (HC)

A 1 n Ai y i  i ¦ 2 i 1 Atot y Atot

1ª n 2 Ai º « ¦ ( yi  y ) » y ¬i 1 Atot ¼

ª 1 n ª y º 1H º 1 « ¦ « i » » «¬ n i 1 ¬ y ¼ »¼

1

(1H )

Coulter coefficient (CC) CC

2 ª1 n § A y A · º « ¦ ¨¨ i i  i ¸¸ » «¬ 2 i 1 © Atot y Atot ¹ »¼

1/ 2

Gini (W) (population weighted)

Gini

1 n n Ai A j yi  y j ¦¦ 2 y i 1 j 1 Atot Atot

Note: Ai and Aj= number of individuals in regions i and j respectively (regional populations), Atot= the national population; yi and yj= per capita development parameters observed respectively in region i and region j (e.g., per capita income); y is the national average (e.g. per capita national income); n = overall number of regions; H is an inequality aversion parameter, 0< H 1:

xni

C1

xi   xr  xi

5.3.3

IRi  1 ˜ C1 Di

y ni

yi   y r  yi

IRi  1 ˜ C1

(5.5)

Di

“Concentric Circle” Transformation

According to the third transformation method we propose, localities with equal parameter values are positioned at the same distance from the reference point, with the maximum parameter value being located on the circle closest to such a point. According to this method, the new distance of i locality from the reference point, Dni in equations (5.1) is estimated proportionally to the natural logarithm of IRi:

Dni

C3 ˜ ln IRi  ln C2 ,

(5.6)

where C2 and C3 are constants that define radii of the smallest and largest circle, respectively. Constant C2 depends on where it is decided to place localities with the highest parameter value i.e. it defines the radius of the smallest circle (C2 > max (IRi)). Concurrently, the radius of the largest circle, C3 can be calculated in different ways, for example: C3

1

max Di ln min IRi  ln max IRi

C3

§ · Di ¸¸ average¨¨  1 abs IR  © i ¹

The C1 coefficient serves two basic purposes. First, it facilitates keeping the transformed coordinates of localities within the limits of the original map, i.e. relative to either the maximum distance between a reference point and the most remote locality, max(Di) or mean distance, mean(Di). Second, it prevents ‘jumps’ of localities, which are more developed than the reference city (IR>1), across the reference point into the opposite quadrant. The tanh(x) function is introduced to suppress this undesirable effect (see the following section for discussion of this effect in more detail).

70

Boris A. Portnov and Rimma Gluhih

By substituting (5.6) in (5.1) we get formulae for coordinate transformation according to the “concentric circle” method:

5.4

xni

xr 

( xi  xr ) ˜ C3 ˜ ln IRi  ln C 2 Di

y ni

yr 

( yi  y r ) ˜ C3 ˜ ln IRi  ln C2 Di

(5.7)

Preliminary Testing and Discussion

To demonstrate the performance of the above transformation techniques, we positioned six test localities (1…6) around the reference point, Ref. (Table 5.1, Figure 5.2). In order to simplify the comparison of the outcomes of alternative transformations, the test localities are defined in such a way that they would constitute pairs. Both points in a pair have equal parameter value Vi (1 and 2; 3 and 6; 4 and 5) and therefore the same ratio IRi, while the distance from one locality in a pair to the reference point is about as twice as large as that from the second point (i.e. ~5 and 10 distance units, respectively). Figure 5.2B reports the result of the actual distance transformation method. Since, according to this method, Dni is directly proportional to Di (actual distance), more remote localities move farther from the reference point than less distant localities with identical parameter values (Pairs 1-2, 3-6, 4-5). Moreover, Point 3 (IR1=1.8) is shifted considerably less than Point 1 with a similar ratio of parameter values (1/IR1=2.5). The transformation in question thus does not accurately reflect the actual difference in the parameter values between localities and the reference point. The proportional increment transformation (Figure 5.2C) appears to correct the drawbacks of the actual distance method. In particular, after this transformation, pairs with identical parameter values (Points 1 and 2, 4 and 5) get equal shifts, irrespectively of their distance to the reference point. It should be noted, however, that since for points with IRi>1, C1 is a variable dependent on tanh of distance, shifts of localities with high IR values are not always proportional to actual differences in parameter values. In particular, the shift of Point 6 is somewhat smaller than that of Point 3, though these two localities have identical values of IR (IR3=IR6=1.8; Table 5.1). However, this partial distortion is clearly necessary to prevent “jumps” of points located close to the reference point and having high parameter values across the reference point into the opposite quadrant. (See, for example, the move of Point 6 o 6’’ (without tanh transformation) and 6 o 6’ (with tanh transformation; Figure 5.2C).

Spatial Patterns of Income Disparities

71

Fig. 5.2. Alternative methods of coordinate transformation A - Original positions of localities; B - “Actual distance” transformation; C - “Proportional increment” transformation. D - “Concentric circle” transformation Note: Coordinate axes are marked by regular numbers; bold numbers denote localities (1 through 6); ref. denotes the reference point. Filled circles show the original positions of localities; empty circles indicate the positions of localities after transformation.

Table 5.1. Test localities

Locality

Ref 1 2 3 4 5 6

X Y Distance from coordinate coordinate the reference point 0 0 0 -6 -8 10.0 3 -4 5.0 10 -2 10.2 2 5 5.4 9 5 10.3 -5 2 5.4

Parameter value, Vi

IRi=Vi/Vr

3500 1400 1400 6300 2800 2800 6300

1.0 0.4 0.4 1.8 0.8 0.8 1.8

72

Boris A. Portnov and Rimma Gluhih

Lastly, Figure 5.2D illustrates the results of the “concentric circle” transformation. As Figure 5.3D shows, after this transformation, three pairs of points with identical parameter values (1 and 2; 4 and 5; 3 and 6) are positioned on different circles, corresponding to the actual IR values: 0.4, 0.8 and 1.8. Points 3 and 6 with the maximum parameter value (IR=1.8) are closest to the reference point, whereas Points 1 and 2 (IR=0.4) are most remote.

5.5

Testing Against Actual Data Mapping Inter-urban Income Disparities in Israel

In this section, the proposed methodology of coordinate transformation is used to illustrate the spatial patterns of inter-urban income disparities across 170 urban localities in Israel, with at least 2,000 residents. The data for the analysis are obtained from two major sources: 1. Israel Regional Database, maintained jointly by Israel Social Science Data Centre (ISDC) and Israel Central Bureau of Statistics (ICBS) - average incomes in localities in 1991 and 1999, respectively; 2. List of Localities, their Population and Codes (ICBS 2001a) - geographic coordinates of localities. Two separate transformation methods - “concentric circle” and “proportional increment” transformation - are used. The mapping of transformation results is performed using the ArcView 8.3 GIS software. The results of the analysis are given in Figs. 5.3-5.5. Figs. 5.3 and 5.4 illustrate inter-urban income disparities in 1999, using both the concentric circle and proportional increment methods, whereas Figure 5.5 displays changes in income disparities between 1991 and 1999, using only the proportional increment method that appears to be most informative. As Figure 5.3 shows, the concentric circle method makes it possible to determine the relative degree of either wealth or income deprivation of local residents: long arrows away from Tel Aviv indicate considerably lower average incomes of residents compared to that in the reference city.2 Concurrently, long arrows towards Tel-Aviv (left diagram) indicate that local incomes are relatively high.

2

There are four major population centres in Israel - Jerusalem, Tel Aviv, Haifa, and Be’er Sheva. Although each of these cities could have been considered as the reference point for this study, Tel Aviv is selected as the centre of the largest population concentration in Israel; together with its hinterland, the city concentrates nearly 40 per cent of the country’s population (ICBS 2001b).

Spatial Patterns of Income Disparities

73

Fig. 5.3. Classification of localities according to relative levels of average income in 1999 (“concentric circle” method) A - Localities with relatively high levels of income; B - localities with relatively low level of income. Note: The starting points of the arrows show the geographic positions of localities; the lengths of the arrows indicate the relative difference in average incomes between the localities and Tel Aviv, selected as the ref. point.

74

Boris A. Portnov and Rimma Gluhih

A

B

Fig. 5.4. Classification of localities according to relative levels of average income in 1999 (proportional increment method) A - localities with relatively high levels of income; B - localities with relatively low level of income. See note to Fig. 5.3.

Spatial Patterns of Income Disparities

A

75

B

Fig. 5.5. Localities with different rates of income change between 1991 and 1999 (proportional increment method) A - Localities with increased levels of income; B - localities with decreased levels of income. See note to Fig. 5.3

However, it should be noted the maps obtained with this transformation may be somewhat misleading. Thus, for instance, localities in the southern part of the country (around the city of Be’er Sheva) have, in general, relatively low average incomes. However, most of them appear on the left diagram, with arrows pointing towards Tel Aviv. The explanation of this phenomenon is rather simple: these

76

Boris A. Portnov and Rimma Gluhih

localities are relatively remote from the reference point and are located outside circles designated for their respective income levels (see the previous section for explanation of the concentric circle method). Therefore, after coordinate transformation, these localities “moved” towards the reference points instead of moving from it, as could be expected. Technically, this effect could easily be corrected by increasing C3, i.e. the radius of the largest circle corresponding to max(IR). However, if C3 increases, many localities may “move” well outside the boundaries of the original map, which is clearly undesirable. The outcome of the proportional increment transformation, according which “shifts” of localities are determined in proportion to the absolute difference in incomes (Figure 5.4), appears to be more informative. For instance, this method allows us to distinguish clearly between localities with relatively high incomes (Vi>Vr; left diagram) and localities with relatively low average incomes (Vi 3 years), 1993 by RBA region Regional differences in long term unemployment (> 3 years), 1999 by RBA region Map of Slovene NUTS III regions Standard deviations of log GVA per capita and log Income tax base per capita and Gini coefficients for GVA The average annual growth rate in GVA pc 1990-1999 estimated with OLS and GVA pc in 1990 Natural regions and administrative districts of Israel (as of 1995) 1995 Census: indicators of population distribution population density and housing density 1995 Census: indicators of wealth and housing - ownership of personal computers and homeownership level 1995 Census: indicators of employment and wages - monthly income per employee and participation in the labour force

116 117 118 119 120 124 125 144 149 150 151 152 156 157 158

160 163 164 172 175 176 191 195 197 198

Figures 315

Fig. 11.5.

Fig. 11.6. Fig. 12.1 Fig. 13.1. Fig. 13.2. Fig. 13.3. Fig. 13.4. Fig. 13.5. Fig. 13.6. Fig. 14.1. Fig. 14.2. Fig. 16.1.

1995 Census: indicators of education and ethnic makeup average years of schooling and percentage of residents who had immigrated from Asia and Africa - 1st generation Changes in selected indicators of interregional inequalities over the whole country (1961-1995 Census data) Sigma and financial autonomy for the large and small countries samples The dispersion of GDP per capita around the national average: countries ranked by size, NUTS III level, 2000 GDP per capita at the NUTS III level (country average=100), 2000 E-convergence for EU enlargement countries; NUTS III level, 1995 - 2000 The weighted coefficient of variation of countries ranked by size: NUTS III level, 1995 and 2000 The weighted coefficient of variation and the size of the countries, 1995-2000 The evolution of the average coefficient of variation of small and large countries over the period 1995 - 2000 Evolution of employment in the various regional production systems, 1975-1995 Evolution of exchange rate of Swiss franc, 1973-1995 Regional gross domestic product 2000 per capita in Euro

200 201 225 240 241 242 243 244 245 256 259 287

Tables Table 1.1. Table 2.1. Table 3.1. Table 4.1. Table 4.2. Table 4.3. Table 5.1. Table 6.1. Table 6.2. Table 6.3. Table 6.4. Table 6.5. Table 7.1. Table 7.2. Table 7.3. Table 7.4. Table 8.1. Table 8.2. Table 8.3. Table 8.4. Table 8.5. Table 8.6. Table 8.7. Table 8.8. Table 8.9. Table 8 10. Table 9.1. Table 9.2.

Key attributes of the small countries examined in this volume Size-related attributes and their expected impacts on regional disparities Taxonomy of regional equilibria Commonly used measurements of regional inequality The reference and test distributions Results of sensitivity tests Test localities Shift-share components, Belgian regions employment, 19952001 (national benchmark) percent Shift-share components, Belgian districts employment, 19952001 (national benchmark) percent Dynamic shift-share components, Belgian districts employment, 1995-2001 (national benchmark) percent Belgian spatial administrative divisions Definition of industries The rich population by region in 1971, 1990 and 2000 Within group and between group inequality in 1971-2000, in percent Sub-group decompositions of the changes in disposable income inequality Decomposition of the squared coefficient of variation (I2) by main region and income source in selected years Decomposition of living standards in Irish regions: 1960-79 (percent p.a.) Decomposition of living standards in Irish regions: 1979-96 (percent p.a.) Sectoral productivity in Irish regions: 1960-79 (percent p.a.) Sectoral productivity in Irish regions: 1979-96 (percent p.a.) Adjusting manufacturing and aggregate regional productivity for transfer pricing: 1979 to 1996 (percent p.a.) Sectoral employment shares in Irish regions: 1960, 1979 and 1996 (percent) Decomposing regional productivity, 1960-79 (percent p.a.) Decomposing regional productivity, 1979-96 (percent p.a.) Definition of regional authority areas Sector classification Indicators of unused supply of labour in the Netherlands, 1999 Regional component in unemployment as percent of the labour force

5 18 42 49 55 56 71 90 91 94 104 105 120 121 122 123 132 133 135 135 136 138 139 139 144 145 153 155

318 Tables

Table 9.3.

Table 9.4.

Table 10.1. Table 10.2.

Table 10.3. Table 10.4. Table 10.5. Table 11.1. Table 11.2. Table 11.3. Table 11.4. Table 12.1. Table 12.2. Table 12.3. Table 12.4. Table 12.5. Table 12.6. Table 12.7. Table 12.8.

Table 13.1. Table 13.2. Table 13.3. Table 14.1. Table 14.2. Table 14.3. Table 17.1.

Deviation of the development of the number of NWW persons with certain characteristics compared to the total number of NWW persons Regional extremes in the development of the number of nonworking job-seekers per RBA-region, March 1993 - October 1998 Basic indicators of the regions Comparison of the internal regional disparities on the NUTS III level; population weighted coefficient of variation of regional GDP The determinants of the regional GVA pc; Equation 10.1 the fixed effects model; dependent variable lnGVApc The determinants of the regional GVA pc; Equation 10.2 pooled regression model; dependent variable lnGVApc The factor loadings after varimax normalized rotation of the principal components analysis Districts and natural regions of Israel (as in 1995 Census) 1995 Census: factor analysis - explanation of total variance 1983 Census: factor analysis - explanation of total variance 1972 Census: factor analysis - explanation of total variance Different measures of regional inequalities in GDP per worker Correlation coefficients of the inequality indices Decentralisation measures Public sector size and party orientation variables Coefficient of correlation between the decentralisation index and regional inequality Correlation between public sector size, parties in government and inequality indices Regression analysis of regional inequality Regression analysis of regional inequality with decentralisation, public sector size and party in government for Small Countries Basic demographic and economic characteristics of EU Enlargement Countries ranked by size Basic regional characteristics of EU Enlargement Countries, 2000 Regional inequalities in the EU Enlargement Countries, NUTS III level: 1995 and 2000 Regional implications of stages in banking development Evolution of employment in the various RPSs*, 1975-1995 Annual increase at the canton level of per capita income (1975-1995), in Swiss francs R&D and patenting comparisons: Israel and Ireland in context

161

162 172

174 179 181 184 192 203 205 207 219 220 222 223 224 226 228

229 235 236 239 253 257 258 303

Author Index A Aaberge R, 120 Aalbu H, 270 Abe H, 207 Acs Z, 17, 297 Adametz C, 290 Adams TM, 65 Ahola E, 279 Aiginger K, 291 Alanen A, 267 Alesina A, 3, 15, 217 Allégret JP, 254 Andréosso-O’Callaghan B, 298, 306 Andrienko GL, 65-66 Andrienko NV, 65-66 Anson J, 188, 190 Archambault E, 303 Armstrong H, 2, 4, 8, 15-16, 196, 213, 215, 277 Atkinson AB, 48, 113, 116 Attwood EA, 130 Audretsch DB, 297-299 Azzoni CR, 44 B Badinger H, 288 Baggi M, 262 Balaz V, 233 Balchin PN, 207 Bar-El E, 307 Barff R, 92-93 Barro RJ, 113, 176, 211 Baumont C, 53 Beaud M, 87 Beenstock M, 34, 43 Béguelin JP, 259 Bertram G, 4, 15-16 Berzeg K, 96 Best M, 299 Beugelsdijck S, 300 Birnie JE, 136 Björklund A, 120 Blattner N, 258 Blien U, 96 Bodenhöfer HJ, 291 Boncoeur J, 252 Boyle GE, 130, 132

Braczyk H-J, 299 Bradley J, 133, 141 Brakman S, 2, 34 Breathnach P, 140 Brennan G, 216 Briguglio L, 4, 15 Broersma L, 150 Buchanan JM, 142, 216 Buck TW, 89 C Callanan B, 306 Câmara G, 65 Cambell J, 65 Cameron D, 216 Carbonaro G, 177, 182 Carrington A, 53 Carroll GR, 23 Carsjens GJ, 65 Casson M, 285 Castles F, 217, 222, 229 Ceccato V, 66 Champernowne DG, 51, 219 Chandra S, 26 Cheshire PC, 177, 182 Chopard R, 262 Clapp J, 65 Clapp JL, 65 Clipson A, 57 Coe N, 300 Colomer J, 222 Colpaert A, 65-66 Constantin D, 234 Cooke P, 285, 299, 308 Cornet M, 300 Corpataux J, 256, 257 Coulter P, 48-49, 60 Cowell FA, 51, 219 Crepon B, 297 Crevoisier O, 256-257, 260, 264 Crone M, 305-306 Crowards T, 3, 15 Cuadrado-Roura JR, 137 D Dalton H, 48, 50-51, 53, 218 Davelaar E, 300 Davies C, 308

320 Author Index Davis DR, 34 De Brabander GL, 89 De Jong CF, 199 de Kervenoael RJ, 15-16 De la Mothe J, 297 Deloitte Touche, 260 Dent BD, 65-67 Dicken P, 300 Dillinger W, 212 Dinc M, 104 Dorling D, 63, 65-66 Dormard S, 87 Dosi G, 299 Dow SC, 251-253 Doyle E, 136, 142 Drabkin-Darin H, 188, 190 Drudy PJ, 305 Duguet E, 297 Dunford M, 207 Dunn ES, 86 Dykes JA, 63, 66 E Easterly W, 4, 8 Eberts RW, 44 Economou D, 207, 234 Ein-Dor P, 4 Ekamper P, 166 Elazar DJ, 213 Elfring T, 150 Epstein R, 57 Erell E, 17, 208 Ersson S, 216 Ertur C, 2, 53 Esteban J, 137, 218 Estevão M, 85, 87, 104 Etzion Y, 190 F Fama EF, 36 Farnetti R, 254 Fawcett JT, 199 Fazekas K, 233 Feitelson E, 307 Feldman M, 297, 299 Felsenstein D, 43, 106, 199, 307 Fernandez E, 300 Fernández MM, 87 Fitschen A, 65 Fitzgerald J, 133 Freeman C, 4 Freeman J, 23

Freinkman L, 215 Frenkel A, 298-299, 302, 304, 307-309 Friedlander J, 190, 199 Fritz O, 290 Fujita M, 2, 298, 309 G Gahegan MN, 63, 65-66 Gallagher L, 142 Garcia Greciano B, 137 Garretsen H, 2, 34 Gastwirth JL, 39-40 Gatrell A, 63 Geary RC, 130 Geldner N, 289, 291 Gerhardter G, 290 Gorg H, 302 Gorzelak G, 234 Gould ED, 30 Grabher G, 286 Gradus Y, 188-189 Gratzl B, 258 Griliches Z, 297-298 Grime K, 234 Grimes S, 17, 301, 305, 308, 310 Grossman G, 34 Gruber M, 290, 291, 292 Guesnier B, 87 Gulic A, 170, 180 Guptill SC, 65, 67 H Haapanen M, 272 Hanell T, 270 Hannan MT, 23 Hanson GH, 34 Hare AD, 14 Hartmann C, 290 Haukka J, 277 Hausermann H, 233 Hayek, 284 Heckman JJ, 29-30 Heidenreich M, 299 Heil, 216 Helpman E, 19, 34 Hemerijck A, 147 Henderson J, 300 Henry E, 130 Heshmati A, 297 Hess M, 300 Hesterberg T, 57 Hewitt-Dundas N, 306

Author Index 321 Hibbs DA, 217 Hirschman AO, 2 Hitchens DMWN, 136 Hofer H, 137, 288 Honkapohja S, 112 Honohan P, 133 Honoré BE, 29-30 Huovari J, 267, 271 Hutschenreiter G, 291 I Ingham M, 234 Isralowitz R, 190, 199 J Jacobsen A, 285 Jaffe AB, 297 Jalan B, 15 James A, 286 Jäntti M, 120 Jayet H, 95-96 Jeglitsch H, 291 Jenkins SP, 120 Johnston R, 44 Johnston S, 170 Jones CI, 34 Jun MJ, 65 Junquera B, 300 K Kalela J, 109, 112 Kangasharju A, 267, 270-271 Kaufmann T, 258 Kauhanen M, 272 Kavas D, 170 Kearney I, 133 Kennedy K, 129 Kiander J, 109, 112 Kimerling AJ, 65, 67 Kipnis BA, 188, 190, 300, 308 Kivikuru U, 109, 112 Klitgaard R, 65 Kloosterman RC, 150 Kluge G, 48-49, 60-61 Knight PL, 92-93 Koga D, 65 Kornai J, 182 Kortelainen S, 279 Koskela E, 112 Kowalski J, 234 Kraak MJ, 65 Kraay A, 4, 8

Krakover S, 188-189 Krugman P, 2, 19, 34 Kukar S, 170, 180 Kuznets S, 1, 2, 16, 211 L Lambelet JC, 255 Lampard EE, 86 Lane JE, 216 Le Gallo J, 2, 53 Lerman RI, 48, 51-52, 122 Levine R, 211 Leyshon A, 254 Li X, 15, 16 Lijphart A, 222, 227 Lipshitz G, 53, 188, 190 Loikkanen HA, 109, 112, 269-271 Lööf H, 297 Lorentzen A, 233 Lorenz MO, 47-48 Lundvall BA, 4 Luukkonen T, 301 M Maceachren A, 65 Mairesse J, 297 Majcen B, 170 Malmberg A, 286 Marcy G, 16 Marimon R, 86, 96-98, 101, 103 Markusen A, 63, 207 Martin R, 63, 255 Martinez-Vazquez J, 214, 220 Maskell P, 286 Massey D, 283 Maung AC, 50 McCann P, 21 McCarthy TG, 130, 132 McCrone G, 273 McKinney M, 44 McNab R, 214, 220 Menédez AJL, 87 Mera K, 189 Metcalfe JS, 286 Meunier O, 89, 106 Meyler A, 141, 305 Michalet CA, 254 Mignolet M, 89, 106 Milner C, 16 Minassian G, 234 Minns R, 255 Mitchell R, 65

322 Author Index Molho I, 199 Monaghan S, 57 Monteiro AM, 65 Mookherjee D, 121 Moore DS, 57, 88 Morgan K, 285, 299 Morgenroth E, 141 Morrison JL, 65, 67 Muehrcke PC, 65, 67 Muilu T, 65, 66 Murray AT, 63 Murray J, 306 Musgrave R, 213 Muth RF, 86 Myers MD, 4 Myrdal G, 211, 251-252 N Naukkarinen A, 65-66 Neary P, 286 Nemes-Nagy J, 233 Neubauer, 270 Niittykangas H, 278 Nijkamp P, 300 North DC, 213 Nyberg P, 112 Nygård F, 122 O O’Farrell P, 147, 150, 305 O’Leary E, 129-131, 136-137, 140-142 Oates WE, 212-213, 216 ÖIR, 289 Oosterhaven J, 155 P Palme G, 291 Paquet G, 297 Parr JB, 207 Patterson MG, 96 Pearlmutter D, 188, 194 Pecar J, 170, 172, 175-176 Pedersen PJ, 120 Pehkonen J, 271 Pekkala S, 270-272, 274, 278 Perkins DH, 1, 8, 16 Perloff HS, 86 Persky JJ, 50 Persson LO, 66 Petrakos G, 233-234, 238 Pierson P, 216 Pigou AC, 50

Poot J, 2, 4 Porteous DJ, 254 Porter R, 65 Portnov BA, 17, 61, 106, 188, 190, 194 Prudhomme R, 213-215 Puga D, 21 Pyatt G, 31, 51-52 R Raagmaa G, 233 Rainwater L, 116 Raman KS, 4 Ramos FR, 65 Rantala A, 269-270 Ratti R, 262 Raymond JL, 137 Read R, 4, 8, 15-16 Renelt D, 211 Richardson HW, 2, 196 Riihelä M, 119, 270-271 Ritsilä J, 272, 277 Robinson AH, 65, 67 Robinson EAG, 4, 15 Rodriguez M, 65 Rodríguez-Pose A, 176, 211, 238 Roper S, 17, 53, 298, 300-302, 304-310 Rosenthal H, 217 Roth JP, 259 Rovolis A, 238 Roy AD, 26 Ruane F, 302 Rusanen J, 65-66 Russel JS, 65 S Sala-i-Martin X, 48, 63, 113, 176, 211 Sandström A, 122 Schmidt MG, 217, 222-223 Schremmer C, 291 Schumpeter J, 2, 211, 284 Schweitzer ME, 44 Scitovsky T, 15 Segerstrom PS, 34 Selwyn P, 4 Sen A, 48-49, 51 Servo LMS, 44 Shachar A, 188, 190, 199 Shah A, 16, 20 Shankar R, 16, 20 Shaw M, 65 Shefer D, 188, 190, 299, 304, 307-309 Shefer S, 308-309

Author Index 323 Shorrocks A, 121 Shyy TK, 63 Siebert H, 2, 41 Siegel S, 132 Silber J, 51-52, 61 Simpura J, 109, 112 Smeeding TM, 116 Smith N, 120 Soen D, 188 Sonis M, 188 Spolaore E, 3, 15 Sposati A, 65 Stark T, 44 Steiner J, 291 Steiner M, 283, 285-286, 288-291 Stevens BH, 88 Stilwell FJB, 88, 92 Stöhr W, 288-289 Streeten P, 1, 4, 15-16 Streißler E, 285 Strobl E, 141, 305 Sturn D, 288 Sui DZ, 65 Sullström R, 119, 269-271 Sunley P, 63 Suoniemi I, 120 Syrquin M, 1, 8, 16 Szopo P, 291 T Taipale M, 271 Tam M-YS, 50 Tanzi V, 213, 215 Tarrant MA, 65 Taylor J, 196, 213, 215, 277 Tervo H, 268, 271-273, 275, 277-278 Teubal M, 307 Theseira M, 65 Thierstein A, 256-257 Thisse JF, 298, 309 Thouément H, 252 Thrall G, 65 Thrift NJ, 254 Tichy G, 289 Tobler W, 17 Tödtling F, 289 Tomaney J, 213 Tong S, 215 Totev S, 233-234 Toulemonde E, 97-98, 106, 131 Tsionas EG, 2

Tsui K, 215, 218 Tunstall R, 213 Tyner J, 65 U Url T, 288 V Van der Knaap W, 65 Van Dijk J, 150, 155 Van Marrewijk C, 2, 34 Van Wissen L, 166 Vartiainen P, 274 Vazquez CJ, 300 Venables AJ, 2 Vihriälä V, 112 Virkkala S, 275 Visser J, 147 Vonderohe AP, 65 W Ward N, 213 Wayland D, 47 Webster CJ, 66 Weinstein DE, 34 Wennemo T, 120 Westaway T, 16 White H, 39, 179 Williamson JG, 48, 143, 189 Wilson R, 308 Wojan TR, 50 Wolf K, 96 Wong C, 207 Wood J, 63, 67 Wörgötter A, 137, 288 Wrynn J, 302 Wu FL, 66 X Xie YC, 65 Y Yearly S, 301-302 Yeung HYC, 300 Yitzhaki S, 30, 31, 48, 51-52, 122 Yossifov P, 215 Z Zhang T, 215 Zhao X, 215 Zilibotti F, 86, 96-98, 101, 103 Zitikis R, 39-40 Zou H, 215

Subject Index A absolute dispersion, 55 accessibility, 9 actual distance method, 75, 83 additionality, 296 agglomeration, 9, 21, 23, 306, 330, 333 agglomeration economies, 20, 24, 155, 195, 319, 331 Albania, 250 allocation of resources, 226 allocative efficiency, 305 Andorra, 41 Armenia, 15 asymptotic distribution (Gini), 43 Atkinson index, 52, 53, 56, 65, 232-233 Australia, 14, 45 Austria, 5, 183, 186, 303, 307-308, 310, 313 Austrian school, 304-305 Azerbaijan, 15 B B2B spillovers, 322 backwash effects, 202 banking sector dualisation, 270, 281 Barro-type growth models, 7 Belgium, 4-5, 8, 39, 45, 92, 163, 226, 235 Beta convergence, 122, 141, 144, 146, 149, 189, 254, 257, 262, 288 Bhutan, 15 bi-national R&D funds, 323, 329 bootstrapping, 7, 61 Brazil, 48 Bulgaria, 250-252, 254-255, 257 business cycles, 39 business incubators, 330 Byelorussia, 14 C Canada, 4, 39, 234 capital investment, 224 capital mobility, 47, 268 capital stock, 224 Central Limit Theorem, 41 centralisation, 229 chart maps, 73

China, 229 city-states, 4, 16 civilian R&D investment, 324 closed economies, 38 cluster maps, 73 cluster policy, 314 cluster strategy, 311 clusters, 304, 306, 314 Cobb-Douglas production function, 37, 45 coefficient of variation, 36, 53, 55, 65, 131-133, 167, 169, 186, 260-261 cohesion policies, 251 collaborative networks, 321 commuting, 43, 179 comparative advantage, 319 competence clusters, 311 competitive advantage, 9 concentric circle method, 76, 79, 83, 85 constitutional structure, 235, 236 convergence, 25, 44, 119, 140, 142, 201202, 221, 262, 289, 308 coordinate transformations, 68, 74 correlation coefficients, 173, 234, 239 Coulter coefficient, 52-53, 65 counterfactuality, 143, 145, 149, 296 creative destruction, 305 credit rationing, 270 Croatia, 183 cross-border cooperations, 314 cumulative causation, 269, 291, 306 Czech Republic, 24, 251, 257 D Dalton transfer principle, 233 Dalton’s coefficient, 56 decartelisation, 269, 281-282 decentralisation, 8, 224-227, 230, 235, 239-242, 244 decomposition of change, 131 de-industrialisation, 269 demand and reserve coefficient, 65 demand-side policies, 228 demography, 140, 142 Denmark, 4, 24, 25, 186, 324 density, 7, 22 dependency ratio, 291

326 Subject Index deregulation, 120, 271 designated areas, 328 development economics, 5, 16-17 development policies, 251 development towns, 203 development zone, 293-294 direct subsidies, 294 discontinuities, 21 diseconomies of agglomeration, 2 diseconomies of scale, 19 disposable income, 119, 125, 127 distributional impacts, 17 divergence, 18, 140, 202 diversification, 39 double income households, 180 dual funding system, 272 Dublin, 57 Dutch Disease, 159-160 dynamic shift-share, 93, 100, 103, 112 E economic integration, 8, 270 economic theory, 7, 28, 48 economies of scale, 16, 29, 37 economies of scope, 306 empirical democracy theory, 231 endogenous growth, 38, 226, 229, 296 enterprise promotion, 293 entrepreneurship, 305, 323 entropy, 53 Estonia, 249, 251, 255, 257, 260 EU Accession States, 8 EU collaborative programmes, 323 EU enlargement, 249, 251, 254 EU membership, 312 EU regional policy, 296 export/import dependency, 21 export-led growth strategy, 152 export-orientation, 18 external trade, 3, 14, 24 F factor analysis, 8, 200, 206, 215-216, 220 factor income, 119, 125, 127-128 factor mobility, 7, 21-23, 30 FDI, 323 federalism, 226, 236, 239 female labour market participation, 162 financial autonomy, 235, 239, 242 financial centres, 268

financial globalisation, 271 financial institutions, 269, 272 financial markets, 268 financial metropoles, 282 financial services, 8, 268-269, 271-272, 283 Finland, 5, 8, 14, 35, 57, 118-119, 126, 136, 140, 285, 296-297, 299, 319, 323-324, 332 fiscal centralisation, 18 fiscal decentralisation, 228-229, 235, 239, 242, 244-245 fishnet/mesh surfaces, 74 fixed effects panel data regression, 190 France, 57, 234, 272 G Gamma convergence, 141, 144, 146, 150-151 general regional equilibrium, 47 generalised entropy measures, 131, 137 geo-demographic analysis, 71 Georgia, 15 Germany, 6, 163, 232, 234 Gini coefficient, 7, 33, 35, 43, 52, 54-56, 61-66, 119, 123, 127-128, 130, 133, 135, 137, 188, 232, 241 GIS Mapping, 7, 68 global capital flows, 8, 323 global challenges, 3 global cities, 268 global economy, 3, 19 global inequality, 33 global production network, 322 globalisation, 9, 303-304, 314 globalisation trap, 314 governance structure, 7, 22 graduated colour, 73 graduated symbol, 73 Great Britain, 276 Great Migration, 287 Greece, 24, 186 greenfield manufacturing facilities, 327 gross income, 119, 125 H head office economy, 275 high tech products, 288 high-tech industry, 322, 328 high-tech production, 217, 221 high-tech start-ups, 320

Subject Index 327 Hirschman-Henfirdahl index, 191 Holland, 8 Hong Kong, 4 Hoover coefficient, 52-53, 65 housing construction, 219, 221 housing density, 207-208 housing services, 45 human capital, 297 Hungary, 4, 24, 184, 250-251, 255, 257, 260 hyperinflation, 219 I Iceland, 14 income elasticity of demand, 47 income inequality, 68, 119 income redistribution, 226 income transfers, 295 income-output ratio, 145 increasing returns, 37 India, 39 Industrial Development Authority, 152 industrial policy, 154 industrial regions, 274 industrial SME systems, 278 industrial strategy, 8 industrial structure, 21 industrialisation, 152, 319 industry mix effect, 94 inequality, 119, 122-123, 136 inequality aversion, 233 inequality indices, 66 inequality measurement, 51, 54, 57 inflation, 231 information asymmetry, 271 information society, 297 information technology industries, 287 innovation, 38, 306 innovation by invitation, 326, 329 innovation policy, 8, 318, 324, 329 innovation-led regional development, 320, 322 institutional competencies, 320 institutional constraints, 235, 237 institutional pluralism, 235, 237 internal economies, 332 internal migration, 3 international competition, 270 international financial activities, 282 international trade, 2, 3 internationalisation, 249, 272, 282, 303

interregional convergence, 15, 21 interregional inequality, 33, 35, 200, 213, 219, 221 interregional migration, 2, 19, 21, 290 interregional spillovers, 18 intersectoral growth, 148 inter-urban income disparities, 7 intraregional inequality, 35 intrasectoral growth, 149 inward investment, 324, 326-327, 329 inward technology transfer, 326-327 Ireland, 5, 8, 25, 57, 140, 319, 320, 323325, 331-332 islands, 4, 16 isolines, 74 Israel, 5, 8, 25, 35, 57, 200, 318-320, 323, 324, 329 IT industry, 122, 136 Italy, 4, 6, 24, 163, 183, 186, 234 J Japan, 57, 234 job machine model, 162 K knowledge spillovers, 319, 321-322 Koku, 51 Kullback-Leibler redundancy index, 65 KwaZulu-Natal province, 72 L labour costs, 161 labour force participation, 209 labour market policy, 178, 180, 310 labour migration, 179 labour mobility, 23, 45 labour productivity, 244 labour supply, 16 land supply, 7, 20, 22-23 land-use mapping, 71 Latvia, 251, 255, 257, 260, 263 law of large numbers, 39 liability of newness, 25 liability of smallness, 14, 17, 25 Liberalisation, 270 Lijphart index, 235-236 Lindeberg condition, 42 liquidity preference, 269 Lithuania, 251, 255, 257 living standards, 141-142, 154 Ljubljana, 57

328 Subject Index local inequality, 33 local networks, 333 local supply-chains, 321 local technology transfer, 333 local venture capital markets, 320 location patterns, 287, 291 location quotient, 202 locational advantages, 320 locational patterns, 286 long-distance commuting, 3, 21, 25 long-term unemployed, 173 Luxembourg, 4, 283 M macroeconomic stability, 227 Malta, 4 mass immigration, 21, 58 max/min ratio, 254, 262 mean logarithmic deviation, 131 measures of decentralisation, 235 mergers/acquisitions, 272 metropolitan “shadow” effect, 21 metropolitan dominance, 257 micro data, 119, 122 migration, 232, 291 Mongolia, 15 monopolistic competition, 306 Monte Carlo simulation, 44 multinational companies, 272 N national business cycle, 105 national innovation systems, 324 national spatial strategy, 155 natural resources, 7, 19, 22, 24, 71 neoclassical convergence theory, 269 neoclassical growth theory, 2, 224-225 neoclassical model of regional equilibrium, 45 neoinstitutionalist economists, 226 Nepal, 15 network-building, 304 new economic geography, 2, 20, 37, 155, 182, 306 new growth theory, 155, 305 New Jersey Critique, 28, 30, 45, 47-48 New Zealand, 4, 14, 24 Norway, 14, 229, 232

O one-region-economies, 251 open systems architecture, 321 optimal regional convergence, 54 optimal spatial distribution, 332 output growth, 29 P panel data regression, 190 partisan influence, 231 patent applications, 326 path dependency, 196 pension funds, 271-272 performance effect, 94 Poland, 250-251, 255, 257, 260 political decentralisation, 235 pooled regression model, 190, 194-195 population density, 19, 21, 29, 207-208 population weighted coefficient of variation, 53, 206 Portugal, 4, 24, 234 principal components, 215 principle of transfers, 55 productivity, 140, 142, 145 productivity convergence, 154 productivity growth, 149, 224 profit outflows, 140, 142, 147 proportional increment, 79 proportional increment method, 75, 8485 public choice theory, 227 public finance, 280 public goods, 37 public sector, 225, 230 R R&D, 318-319, 321-322, 324, 327 R&D spillovers, 318 rank concordance measure, 144 regional authority areas, 140 regional budgeting, 22 regional competence, 299 regional competitiveness, 299 regional convergence, 2-3, 9, 18-19, 2122, 118, 123, 136, 148, 155, 286, 289 regional development, 286, 289-290, 305, 309 regional disaggregation, 45 regional disparities, 122, 227, 230, 240, 303 regional divergence, 25

Subject Index 329 regional effect, 95 regional employment disparities, 92 regional equality, 292 regional fiscal autonomy, 22 regional funding circuits, 268, 281-282 regional GVA, 186, 188, 192, 194, 196 regional heterogeneity, 7, 28 regional housing markets, 30 regional identities, 303 regional innovation and technology policy, 298, 310, 331 regional innovation premium, 311 regional innovation strategies, 320, 333 regional innovation systems, 320, 328 regional networks, 306 regional policy, 8, 21, 140, 152, 160, 182, 196, 228, 285-286, 292-295, 297, 299, 307, 310, 312-313 regional price convergence, 54 regional production systems, 273, 307, 314 regional productivity, 230 regional share, 94 regional spillovers, 227, 308 regional transport subsidies, 294 regional unemployment disparities, 159, 160, 179 regional welfare differences, 286 regionalisation, 295 relative dispersion, 55 residual effect, 95 residual growth, 148 resource constraints, 16 resource utilization, 292 restricted least squares method, 105 Romania, 250-251, 255 Roy Model, 29, 31, 33-35 Russia, 30, 45, 228 S sectoral employment shifts, 150 sectoral productivity growth, 145 securitisation, 270 segmentation, 39, 40 self-selection model, 31 shift-share analysis, 7, 93, 112, 148 Sigma convergence, 122, 141, 143-144, 146-147, 151, 187, 232, 239, 241242, 253 Sigma divergence, 145, 148 Singapore, 4

Slovak Republic, 186 Slovakia, 15, 24, 249, 251-252, 255, 257 Slovenia, 5, 8, 15, 57, 182, 195-196, 251-252, 255, 257, 263 small business advice centres, 330 SME, 282, 304 social cohesion, 7, 17, 20-22, 24 social infrastructure, 37 social interaction, 29, 35-36, 43 social multiplier, 36 social returns, 322 South Africa, 72 South-eastern Europe, 250 Spain, 6, 24, 163, 226, 234 spatial dependence, 43 spatial filtering technique, 308 spatial scales of analysis, 3 spatially referenced data, 72 special economic zones, 229 spurious correlation, 236 statistical theory, 48 stratification, 33 structural change, 94-95, 101, 122, 140, 145, 148, 249 structural funds, 296 structural unemployment, 172, 179, 290 subsidiarity, 296 Sweden, 24, 126, 234, 245, 318 Swiss franc, 268, 273, 276, 278, 280 Switzerland, 5, 24, 39, 229, 234, 245, 268-269, 273, 275-276, 278 T tacit knowledge, 306 targeted industrial policy, 153-154 tax reductions, 294 tax reform, 121 technological knowledge transmission, 305 technological linkages, 306 technology park, 298, 309 technology transfer, 321, 329 territorial networks, 311 tertiary industries, 25 the Netherlands, 5, 25, 159, 164-166, 234 Theil index, 52-53, 56, 65, 232, 241 thematic mapping, 72 tourist regions, 274 tourist sector, 273, 279 traded goods, 45

330 Subject Index transaction costs, 23, 306, 308 transitional economies, 4, 250 transport costs, 9, 19-21, 23 transportation mapping, 71 U UK, 28, 48, 226, 325 Ukraine, 14 unemployment, 120, 161, 164-168, 172174, 178-179, 231, 289, 296 unification, 35 untraded interdependencies, 321 Upas Tree effect, 19 urban diseconomies, 155

urbanisation, 333 USA, 28, 41, 48, 231-232, 276, 323, 325 V virtual employment, 105, 107-109, 111 visualization techniques, 70-72, 83 W weighted coefficient of variation, 253254, 262 welfare differences, 288 welfare policy, 295 Williamson index, 7, 52, 53, 65, 202, 212

Contributors OEDZGE ATZEMA, Department of Human Geography and Planning, Faculty of Geosciences, Utrecht University, P.O. Box 80.115, NL-3508 TC Utrecht, the Netherlands. Email: [email protected] MICHAEL BEENSTOCK, Department of Economics, Hebrew University of Jerusalem, Mount Scopus, Jerusalem 91905, Israel. Email: [email protected] JOSÉ CORPATAUX, Institute for Regional and Economic Research (IRER), University of Neuchâtel, Pierre-à-Mazel 7, 2000 Neuchâtel, Switzerland. Email: [email protected] OLIVIER CREVOISIER, Institute for Regional and Economic Research (IRER), University of Neuchâtel, Pierre-à-Mazel 7, 2000 Neuchâtel, Switzerland. Email: [email protected] DANIEL FELSENSTEIN, Department of Geography, Hebrew University of Jerusalem Mount Scopus, Jerusalem 91905, Israel. Email:[email protected] CARLOS GIL, Department of Economics, Universidad Pública de Navarra, Campus de Arrosadia 31006 Pamplona (Navarra), Spain. Email: [email protected] RIMMA GLUHIH, Jacob Blaustein Institute for Desert Research, Ben-Gurion University of the Negev, Sede Boqer Campus 84990 Israel. Email: [email protected] DIMITRIS KALLIORAS, Department of Planning and Regional Development, University of Thessaly, Pedion Areos 38 334, Volos, Greece. Email: [email protected] HEIKKI A. LOIKKANEN, Department of Geography, University of Helsinki, Arkadiankatu 7, FIN-00014 Helsinki, Finland. Email: [email protected] OLIVIER MEUNIER, Centre de Recherches sur l'Economie Wallonne (CREW), University of Namur, Rempart de la Vierge 8, 5000 Namur, Belgium. Email: [email protected]

332 Contributors

MICHEL MIGNOLET, Department of Economics and Centre de Recherches sur l'Economie Wallonne (CREW), University of Namur, Rempart de la Vierge 8, 5000 Namur, Belgium. Email: [email protected] EOIN O’LEARY, Department of Economics, University College Cork, Western Road, Cork, Republic of Ireland. Email: [email protected] PEDRO PASCUAL, Department of Economics, Universidad Pública de Navarra, Campus de Arrosadia 31006 Pamplona (Navarra) Spain. Email: [email protected] GEORGE PETRAKOS, Department of Planning and Regional Development, University of Thessaly, Pedion Areos 38 334, Volos, Greece. Email: [email protected] BORIS A. PORTNOV, Department of Natural Resources & Environmental Management, Faculty of Social Sciences, University of Haifa, Mount Carmel, Haifa 31905, Israel. Email: [email protected] YIANNIS PSYCHARIS, Department of Planning and Regional Development, University of Thessaly, Pedion Areos 38 334, Volos, Greece. Email: [email protected] MANUEL RAPÚN, Department of Economics, Universidad Pública de Navarra, Campus de Arrosadia 31006 Pamplona (Navarra), Spain. Email: [email protected] MARJA RIIHELÄ, Government Institute for Economic Research (VATT), Arkadiankatu 7, FIN-00100 Helsinki, Finland. Email: [email protected] STEPHEN ROPER, Aston Business School, Aston Triangle, Birmingham B4 7ET, United Kingdom. Email: [email protected] MICHAEL STEINER, Department of Economics, Karl-Franzens University, Universitätsstraße 15/F4, A-8010 Graz, Austria. Email: [email protected] RISTO SULLSTRÖM, Government Institute for Economic Research (VATT), Arkadiankatu 7, FIN-00100 Helsinki, Finland. Email: [email protected]

Contributors 333

HANNU TERVO, School of Business and Economics, University of Jyväskylä, P.O. Box 35, FIN-40014 University of Jyväskylä, Finland. Email: [email protected] JOUKE VAN DIJK, Urban and Regional Studies Institute (URSI) and Department of Economic Geography, Faculty of Spatial Sciences, University of Groningen, P.O. Box 800, NL-9700 AV Groningen, the Netherlands. Email: [email protected] PETER WOSTNER, The Republic of Slovenia Government Office for Structural Policies and Regional Development, Kotnikova 28, 1000 Ljubljana, Slovenia. Email: [email protected]