Regional Disparities in Small Countries (Advances in Spatial Science)

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 <ũ [the higher the value of H, the more society is concerned about inequality). In the literature on inequality measurements, formulas for inequality indexes often differ by a factor of 2 or ½. We scale all the indexes between 0 and 1, to facilitate the interpretation of results. Compiled from: Sen (1973); Coulter (1987); Kluge (1999); WBG (2001).

50

Boris A. Portnov and Daniel Felsenstein

Since the inequality indices in Table 4.1 are abstract mathematical formulae, one can assume that they can be applied to both large and small countries alike. Is this assumption correct? It is well known that the use of different measurement indices in regional analyses gives rise to highly variable results. For example, the notion of optimal regional convergence (i.e. that point where regional convergence also reduces overall nation-level inequality) has been shown to be highly dependent on type of inequality index used (Persky and Tam 1985) as is the measurement of regional price convergence (Wojan and Maung 1998). But does the number, size and rank of regions, also play a part? In this chapter, we shall attempt to answer these questions, using a number of empirical tests. The aim of these tests is to determine whether commonly used inequality measures produce meaningful estimates when applied to a small country. The chapter is organized as follows. It begins with a brief outline of general principles, which should govern, in our view, the selection of robust inequality measures, followed by an analysis of characteristic features of small countries, which may influence the choice of inequality indices. Then we move to testing the compliance of different commonly used inequality measures against the set of criteria that should characterize, in our view, a robust inequality measure. The tests are run in two phases. First, we use a number of pre-designed distributions, to verify whether a particular inequality measure meets our intuitive expectations concerning inequality estimates. Then, in the second stage of the analysis, we run more formal permutation tests to verify whether different inequality measurements respond sensibly to changes in the population distribution across the space.

4.2

General Requirements for a Robust Inequality Measure

As Dalton (1920) noted, many inequality measures, though having intuitive or mathematical appeal, react to changes in income distribution in an unexpected fashion. For instance, if all incomes are simply doubled, the variance quadruples the estimates of income inequality. Dalton’s second observation was that some inequality measures do not comply with a basic principle of population welfare set forward by Arthur Pigou in 1912. This is commonly referred to as the principle of transfers and is formulated by Dalton as follows: “if there is only two incomereceivers, and a transfer of income takes place from the richer to the poorer, inequality is diminished” (ibid. p. 351). After applying this principle to various inequality measures, Dalton found that most measures of deviation (e.g., the mean standard deviation from the arithmetic mean, and the coefficient of variation) are perfectly sensitive to transfers and pass the “test with distinction” (ibid. p. 352). The Gini index was found by Dalton sufficiently sensitive to income transfers. He also found that the standard deviation is sensitive to transfers among the rich, while the standard deviation of logarithms is less sensitive to transfers among the rich than to transfers among the poor but still changes when a transfer among the rich takes place.

Measures of Regional Inequality for Small Countries

51

Two other fundamental requirements for a “robust measure” of inequality, proposed by Dalton, are the principle of proportional addition to incomes, and the principle of proportional increase in population. According to the former, a proportional rise in all incomes diminishes inequality, while a proportional drop in all incomes increases it. According to the latter principle, termed by Dalton the “principle of proportional additions to persons,” a robust inequality measure should be invariant to proportional increases in the population sizes of individual income groups. Dalton’s calculations showed that most commonly used measures of inequality comply with these basic principles. Only the most “simple” measures, such as absolute mean deviation, absolute standard deviations and absolute mean difference, fail to indicate any change, when proportional additions to the numbers of persons in individual income groups are applied (ibid. pp.355357, see also Champernowne and Cowell 1998, pp. 87-112).2 Sen (1973, p.2) distinguished between “measures that try to catch the extent of inequality in some objective sense …and …indices that try to measure inequality in terms of some normative notion of social welfare.” He also undertook a systematic attempt to test the sensitivity of different inequality measures to changes in income distribution. The indices he tested included: the range; the relative mean deviation; the variance; the coefficient of variation; the standard deviation of logarithms; the Gini coefficient; Theil’s entropy measure; Dalton’s coefficient, and Atkinson’s index. Most of the tested measures appeared to exhibit substantial flaws. For instance, the range was found to ignore the distribution among the extremes (i.e., upper and lower incomes), whereas the sensitivity of the Gini index was found to depend critically “not on the size of the income levels but on... the rank-order position of the person in the ranking by income level” (ibid, p. 32). He also found that while the coefficient of variation appears to be sensitive to transfers across all income groups, the standard deviation and the standard deviation of logarithms appeared to be sensitive primarily to transfers in the lower income brackets, and insensitive to transfers among the rich. In a more recent paper, Yitzhaki and Lerman (1991) note another deficiency inherent to most inequality measures, viz. insensitivity to the position which a specific population subgroup occupies within an overall distribution. Their Gini decomposition technique takes group-specific positions into account. They suggest weighting subgroups by the average rank of their members in the distribution. This is in contrast to the weighting system used more conventionally in which between group inequality is weighted by the rank of the average (Pyatt 1976; Silber 1989). This latter system results in a large residual when inequality is decomposed into within and between groups. In contrast, the Yitzhaki approach results in a more accurate decomposition with no residual (Yitzhaki 1994). The question of weighting is, of course, intricately connected to the issue of country and regional size. The standard Barro-type growth regressions that look at 2

Dalton (1920, p. 352) distinguishes between measures of relative dispersion and measures of absolute dispersion. Whereas the former measures are dimensionless, the measures of absolute dispersion are estimated in units of income. The latter measures are easily transformed in the former by normalization.

52

Boris A. Portnov and Daniel Felsenstein

regional disparities over time, do not weight for regional or country size. This is because regions and countries are treated as individuals and not as groups. No compensation is given for small size just as large people and small people are not given any compensation when looking at a population income distribution (The case could be made that large people “need” more income due to their size and therefore deserve to be compensated). However, the growth regression approach assumes that compensating for population size is tantamount to obscuring the unique identity of places, one of which is their size difference. Whether regions are individuals or groups is of course a moot (philosophical) point. In contrast to the neo-classical growth regressions approach, the “inequality indices” (Gini) approach seems to accept the fact that countries and regions are groups. The question thus revolves around a suitable weighting system in which the conventional approach (Pyatt 1976; Silber 1989) is pitched against the alternative approach (Yitzhaki and Lerman 1991; Yitzhaki 1994).

4.3

Characteristic Features of a Small Country That May Affect Inequality Estimates

The aim of the present inquiry is not to test the conformity of commonly used inequality measures with basic inequality criteria (e.g., principles of transfer, proportional addition to incomes, and proportional addition to population). This task has been accomplished par excellence in previous studies, whose findings we have no reason to doubt. Instead, we shall focus our attention on the features which a robust inequality measure should possess in order to make it fully applicable to a small country, which is the main focus of this volume. First, however, let us outline some essential characteristics of such a country that may affect inequality estimates, at least in theory. Since most of these features were discussed in earlier (frame-setting) chapters (Chapters 1 and 2), we shall outline them only briefly, focusing mainly on the empirical aspects of their measurement. First, as previously noted, a small country is likely to have a smaller number of regions than a large and more populous nation. Thus, for instance, Japan with its 130-million strong population has 47 regional subdivisions (prefectures), while Israel (6.5 million residents) is split into only six administrative districts (mahozot, in Hebrew). Similarly, Finland (5.2 million residents) is composed by only six provinces (laanit, in Finnish), whereas France (60 million residents) is divided into 22 regions, which are further subdivided into 96 departments (CIA 2003). Although districts and provinces of a small country may further be subdivided into sub-districts and counties, the overall number of such administrative subdivisions in a small country is naturally smaller than the overall number of administrative subdivisions of comparable size in a more populous nation. The second feature of a small country, which may be important for our analysis, is the varying population sizes of the regions. Although a large country may also have regions of different population sizes, such variation is especially characteristic for a small country, which can be highly mono-centric with a clearly

Measures of Regional Inequality for Small Countries

53

emphasized urban core. Due to the geographic concentration of its population, the population size of the core region in a small country may greatly surpass the population of its sparsely populated peripheral regions. For example in Slovenia, the Central Slovenia region containing Ljubljana has over 26 percent of the country’s population and the smallest region (Zasavska) has a population one twelfth its size. Similarly in Ireland, the Dublin and Mid East region contains nearly 40 percent of the Irish population and has over seven times the population of the Midland Area. In Finland, the Helsinki metropolitan area dominates the Finnish regional population distribution accounting for nearly 20 percent of national population. Lastly, regions in a small country may be a subject to rapid change. For instance, economic growth may spread rapidly across neighbouring regions in a small country, reflecting “growth spillover” (Baumont et al. 2000; Carrington 2003). In contrast, in a large and polycentric country, regional growth may be more localized and slow-acting. For instance, we may recall the rapid regional growth attributed to the development of computer-related industries in Ireland in the late 1980s (Roper 2001). The long-term impact of mass immigration to Israel in 1989-1991 is another example of a rapid regional change in a small country. During this period, nearly 600,000 new immigrants arrived, increasing the existing population of the country by some 15 percent. Eventually many newcomers settled in the country’s peripheral areas, the Northern and Southern districts, whose populations nearly doubled within a short period of some 3-4 years, boosting the emergence of new major population centres (e.g. Be’er Sheva and Ashdod) and causing considerable changes in the existing urban hierarchy (Lipshitz 1998). Taking account of these peculiarities, we can introduce the following three basic requirements to a robust inequality measure which should make it applicable to a small country - the subdivision principle; tolerance to size difference, and rank-order insensitivity. These requirements are now outlined: x Subdivision principle: No matter into how many regions (subdivisions) a country is split, inequality estimates should not change, unless the parameter distribution alters. This requirement is basically in line with Dalton’s principle of population, according to which neither replication of population nor merging identical distributions should alter inequality. x Tolerance to size differences: A robust inequality measure should produce identical estimates for both geographically even and geographically skewed population distributions, providing that the parameter distribution (e.g., distribution of incomes) remains unchanged. For instance, most residents of a country may be concentrated in a single region or they may be dispersed evenly across 10 districts into which the country is split. As long as the income distribution stays the same, regional inequality should not alter. x Rank-order insensitivity. The inequality estimate should not alter as a result of a change in the sequence in which regions are introduced into the calculation, e.g. ranked either by population size or in alphabetical order. Since regions in a small country may be a subject to rapid changes, both in terms of their population sizes and parameter distributions, compliance with this principle

54

Boris A. Portnov and Daniel Felsenstein

will ensure that inequality estimates do not alter simply as a result of changing the position of a region in the rank-order hierarchy. In order to verify the compliance of commonly used measures of regional inequality with the above requirements, the analysis will be carried out in two stages: pre-designed sensitivity tests (Section 4.4) and random permutation tests (Section 4.5).

4.4

Pre-designed Sensitivity Tests

The following specific questions need to be answered: 1. Is an inequality measure sensitive to the overall number of intra-country divisions (regions) covered by analysis? 2. Is an inequality measure sensitive to differences in the population sizes of regions? 3. Does a particular inequality measure respond to changes in the rank-order in which individual regions are introduced into the calculation? Eight commonly used inequality measures (see Table 4.1) are tested here. The tests are designed as follows. First, we introduce the “reference” distribution (Table 4.2: “Reference distribution”). As Table 4.2 shows, this distribution has 16 internal divisions (regions). The average per capita income in its four central regions doubles that in the 12 peripheral regions - 20,000 and 10,000 Income Units (IUs), respectively. Let us call the former group of regions “H[igh-income]regions,” while 12 other regions will conditionally be termed “L[ow-income]regions.” As the table shows, in the reference distribution, the population is distributed evenly: there are 10,000 residents in each regional cell (see Table 4.2). The total population of the reference system is 160,000 residents and the average income is 12,500 IUs per capita. 4.4.1

Test 1 - Small Number of Regions

This test checks whether the overall number of regions matters. To this end, we reduce the overall number of regions to eight, from sixteen in the reference distribution. Total population for this distribution is 80,000 residents, while the average income remains the same and equals 12,500 IUs. Since there are no cardinal changes in income or population distribution, robust inequality indices should indicate the same level of inequality for both the reference and Test 1 distributions (see Table 4.2).

Measures of Regional Inequality for Small Countries

4.4.2

55

Test 2 - Uneven Population Distribution

This test is designed to trace the response of different inequality measures to regional distribution of population: evenly spread population in the reference distribution vs. unevenly spread population in the Test 2 distribution. Compared to the reference distribution, there are no changes in per capita incomes; only the pattern of population distribution is altered. In particular, the populations of the four central (H-regions) increased to 100,000 (4u25,000) residents, while the populations of surrounding L-regions shrunk to 60,000 (5,000u12) residents (see Table 4.2). The total population in this distribution is 160,000 residents and the average income is 16,250 IUs. Since the percent share of population concentrated in four H-regions increased to 62.5 percent [100,000u100/160,000 (total population)=62.5%] from 25 percent in the reference distribution [40,000u100/ 160,000 = 25%; see Table 4.2], the regional inequality of per capita incomes should expectedly decline. Table 4.2. The reference and test distributions Reference distribution 10,000 10,000 10,000 10,000

Average income 10,000 10,000 20,000 20,000 20,000 20,000 10,000 10,000

Test 1 (Number of regions) 10,000 10,000 10,000 10,000

10,000 10,000 10,000 10,000

Population size

Average income 10,000 20,000 20,000 10,000 Population size

10,000 10,000 10,000

10,000 10,000 10,000

10,000 10,000 10,000

10,000 10,000 10,000

10,000 10,000 10,000

10,000 10,000 10,000

10,000

10,000

10,000

10,000

10,000

10,000

Test 2 (Population distribution) Average income 10,000 10,000 10,000 10,000 10,000 20,000 20,000 10,000 10,000 20,000 20,000 10,000 10,000 10,000 10,000 10,000 5,000 5,000 5,000 5,000

Population size 5,000 5,000 25,000 25,000 25,000 25,000 5,000 5,000

5,000 5,000 5,000 5,000

20,000 10,000 10,000 20,000

Test 3 (District ranking) Average income 10,000 10,000 10,000 10,000 10,000 10,000 10,000 10,000

20,000 10,000 10,000 20,000

10,000 10,000 10,000 10,000

Population size 10,000 10,000 10,000 10,000 10,000 10,000 10,000 10,000

10,000 10,000 10,000 10,000

56

Boris A. Portnov and Daniel Felsenstein

4.4.3

Test 3 - Rank-Order Change

Our last test is designed to verify whether the sequence in which regions are introduced in the calculation matters. Compared to the reference distribution, there is no change in either the total number of residents (160,000) or in the average per capita income (12,500 IUs). The only change is the location of H-regions: if in the reference distribution these regions are located in the centre of the grid (6, 7, 10 and 11 sequence numbers), in the Test 3 distribution, they are moved to the corners of the grid (1, 4, 13 and 16 sequence numbers - see Table 4.2). Since the percent share of population concentrated in the H-regions has not changed [40,000u100/160,000=25%], no change in inequality should occur. Table 4.3. Results of sensitivity tests Inequality index

Reference distribution

CV WI TE AT HC CC Gini (U) Gini (W)

0.346 0.346 0.022 0.026 0.150 0.061 0.150 0.150

4.4.4

Test 1 (Number of regions) 0.346 0.346 0.022 0.026 0.150 0.087 0.150 0.150

Test 2 (Population distribution) 0.353 0.298 0.136 0.251 0.144 0.059 0.115 0.144

Test 3 (District ranking) 0.346 0.346 0.022 0.026 0.150 0.061 0.150 0.150

Sensitivity Test Results

The results of the tests are reported in Table 4.3 and discussed below. Test 1: Somewhat surprisingly, despite the unchanged distributions of incomes and populations, CC indicates a rise in inequality! The use of this index for small countries, with a small number of internal divisions (regions), may thus be misleading, specifically when a comparison with countries of larger sizes is planned. Test 2: While the five indices (WI, CC, HC, Gini (U) and Gini (W)) indeed indicate a drop in regional inequality compared to the reference distribution, three other measures (CV, TE and AT) indicate an increase (!) in income disparity. Characteristically, Gini (W) indicates only a marginal drop in inequality (from 0.150 in the ref. distribution to 0.144 in the Test 2 distribution) despite a considerable increase in the population share of H-regions. The use of CV, TE, AT, and Gini (W) for small countries (which are often characterized by extremely uneven regional distributions of population) may thus lead to erroneous results. Test 3: The test indicates no performance problems with any of the indices tested. Numerically, the results of the test appear to be identical to those obtained for the ref. distribution (see Table 4.3).

Measures of Regional Inequality for Small Countries

4.5

57

Permutation Tests

For more formal sensitivity testing of inequality measures, we used the statistical technique known as bootstrapping (Hesterberg et al. 2002). Traditional methods of calculating parameters for a given statistic (e.g., a certain measure of inequality) are based upon the assumption that the statistic is asymptotically normally distributed and use known transformations for parameter calculation. However, re-sampling techniques, such as bootstrapping, provide estimates of the standard error, confidence intervals, and distributions for any statistic by testing it directly against a large number of randomly drawn re-samples. 1000 re-samples are considered as a minimal number recommended for estimating parameters of a statistic, whereas larger numbers of re-runs increase the accuracy of estimates. In particular, we ran two separate tests, as described below: x Test 1 (Unrestricted test): The distribution of income was set identical to the reference distribution (see Table 4.2) and the average income was kept constant (12,500 IUs). Concurrently, the population was distributed across 16 regional cells at random and was allowed to vary slightly around the average population total, which was not restricted a-priori. x Test 2 (Restricted test): The income distribution, the average income, and the total population of the system were kept constant and identical to the reference distribution (see Table 4.2). In order to comply with these restrictions, the population was redistributed within the H-regions and L-regions, without allowing population exchanges between these two groups of regions. For each test, 1000 permutations (re-samples) were run. For the sake of clarity and brevity and to avoid overloading the reader with unnecessary technical details, we discuss below only those results that appear to exhibit most characteristic trends. 4.5.1

Unrestricted Test

The results of the re-sampling for five inequality indices - CV, Gini (U), AT, TE(0), and WI are reported in Figure 4.1. While CV, Gini (U), AT and TE(0) appear to exhibit the response pattern shown in Figure 4.1A, the rest of the indices tested (that is, WI, CC, HC and Gini (W)) exhibit the response pattern diagrammed in Figure 4.1B. The conclusion is thus straightforward: the former group of indices is not sensitive to the variation in population distribution across regional cells. They may thus lead to spurious results when used for small countries, which are often characterized by rapid changes in population patterns, due to (inter alia) the impact of immigration.

58

Boris A. Portnov and Daniel Felsenstein

Value (log)

1.00

0.10

0.01 1

101

201

301

401

501

601

701

801

901

Permutation #

CV

A

GINI (U)

AT

TE(0)

0.45

0.40

WI

0.35

0.30

0.25

0.20 1

101

B

201

301

401

501

601

701

801

901

Permutation #

Fig. 4.1. Results of permutation tests (Test 1: unrestricted test) for selected inequality measures - CV, Gini (U), AT and TE(0) (A) and WI (B) Note: see text for explanations.

4.5.2

Restricted Test

When population movements are restricted (i.e., the population is allowed to circulate only within the H-regions and within the L-regions, without direct population exchanges between the two), only the CC index appears to respond to population re-sampling, exhibiting the oscillation response pattern (see Figure

Measures of Regional Inequality for Small Countries

59

0.50

0.40

Value

0.30

0.20

0.10

0.00 1

101

201

301

401

501

601

701

801

901

Permutation #

WI

A

GINI (U)

0.400

0.398

CC

0.396

0.394

0.392

0.390 1

101

B

201

301

401

501

601

701

801

901

Permutation #

Fig. 4.2. Results of permutation tests (Test 2: restricted test) for selected inequality measures - WI and Gini (U) (A) and CC (B) Note: see text for explanations.

4.2B), whereas all other indices tested (i.e., CV, WI, HC, Gini (U), Gini (W), AT and TE(0)) fail to respond to changes in the population distribution across the regional cells (see Figure 4.2A). However, such a situation (in which population movements are geographically restricted) may be considered rather unlikely

60

Boris A. Portnov and Daniel Felsenstein

(specifically for open economies) and thus a failure of an inequality measure to pass this test may be considered only as a minor performance flaw.

4.6

Conclusions

Though individual studies of regional disparity may deal with separate development measures - population growth, wages, welfare, regional productivity, etc. the use of an integrated indicator is often essential, particularly if a comparative (cross-country) analysis is required. In order to measure the extent of disparities, various indices of inequality are commonly used. These indices may be classified into two separate groups (Kluge 1999): x Measures of deprivation (Atkinson index, Theil redundancy index, Demand and Reserve coefficient, Kullback-Leibler redundancy index, Hoover and Coulter coefficients, and the Gini index); x Measures of variation, such as the coefficient of variation and Williamson’s index. In this chapter, we did not attempt to assess whether these measurements reflect either the “true meaning” or “underlying causes” of regional inequality. Neither did we try to establish whether geographic inequality is a positive socio-economic phenomenon or a negative one. We shall leave these philosophical questions for other studies. Our task was simple: we attempted to determine whether commonly used inequality measures produce meaningful estimates when applied to small countries, thus making it possible to compare the results of analysis obtained for such countries with those obtained elsewhere. As we argue, a small country may differ from a country of larger size in three fundamental features. First, it is likely to have a relatively small number of regional divisions. Second, its regional divisions are likely to vary considerably in their population sizes. Lastly, regions of a small country may rapidly change rankorder positions in the country-wide hierarchy, by changing their attributes (e.g., population and incomes). In contrast, in a large country such rank-order changes may be both less pronounced and slower-acting. In order to formalize these distinctions, we designed a number of simple empirical tests, in which income and population distributions, presumably characteristic for small countries, were compared with a “reference” distribution, assumed to more accurately represent a country of a larger size. In the latter (reference) distribution, the population was distributed evenly across regional divisions and assumed to be static. In the first test, we checked whether the overall number of regions matters. In the second, we tested whether different inequality indices respond to differences in the regional distribution of population, viz., evenly spread population in the reference distribution vs. unevenly spread population in the test distribution. Finally, in the third test, we verified whether different inequality indices were sensitive to the sequence in which regions are introduced into the calculation.

Measures of Regional Inequality for Small Countries

61

Somewhat surprisingly, none of the indices we tested appeared to pass all the tests, meaning that they may produce (at least theoretically) misleading estimates if used for small countries. However, two indices - WI and Gini (W) - appeared to exhibit only minor flaws and may thus be considered as more or less reliable regional inequality measures. Although further studies on the performance of different inequality indices may be needed to verify the generality of our observations, the present analysis clearly cautions against indiscriminate use of inequality indices for regional analysis and comparison.

Acknowledgement Our gratitude is due to Jacques Silber for helpful comments on an earlier draft. We are also indebted to Alex Portnov for designing an algorithm for the permutation tests and to Goetz Kluge for verification of the calculation results.

References Atkinson AB (1983) The economics of inequality (2nd Edition). Clarendon Press, Oxford Baumont C, Ertur C, Le Gallo J (2000) Geographic spillover and growth: a spatial econometric analysis for European regions. Paper presented at the 6th RSAI World Congress 2000 “Regional Science in a Small World”, Lugano, Switzerland, May 1620, 2000 Carrington A (2003) A divided Europe? Regional convergence and neighbourhood spillover effects. Kyklos 56:381-393 Champernowne DG, Cowell FA (1998) Economic inequality and income distribution. Cambridge University Press, Cambridge, UK CIA (2003) 2002 World factbook. Central Intelligence Agency, Washington, D.C. (Internet edition) Coulter P (1987) Measuring unintended distributional effects of bureaucratic decision rules. In: Busson T, Coulter P (eds) Policy evaluation for local government. Greenwood Press, New York Dalton H (1920) The measurement of the inequality of incomes. The Economic Journal 30(199):348-361 Hesterberg T, Monaghan S, Moore DS, Clipson A, Epstein R (2002) Bootstrap methods and permutation tests, Ch18. In: Moore DS, McCabe GP, Duckworth WM, Sclove SL (eds), The practice of business statistics. WH Freeman and Co., NY 18:4-25 Kluge G (1999) Wealth and people: inequality measures (Internet edition) Lipshitz G (1998) Country on the move: migration to and within Israel, 1948-1995. Kluwer Academic Publishers, Dordrecht Lorenz MO (1905) Methods of measuring the concentration of wealth, Publications of the American Statistical Association 9(70):209-219 Persky JJ, Tam M-YS (1985) The optimal convergence of regional incomes. Journal of Regional Science 25(3):337-351 Pigou AC (1920; 1938 reprint) The economics of welfare. MacMillan and Co, London

62

Boris A. Portnov and Daniel Felsenstein

Pyatt G (1976) On the interpretation of disaggregation of the Gini coefficient. Economic Journal 86:243-255 Roper S (2001) Innovation policy in Ireland, Israel and the UK: evolution and Success. In: Felsenstein D, McQuaid R, McCann P, Shefer D (eds) Public investment and regional economic development. Edward Elgar, Cheltenham, UK, 75-91 Sala-i-Martin X (1996) Regional cohesion: evidence and theories of regional growth and convergence. European Economic Review 40:1325-1352 Sen A (1973) On economic inequality (The Radcliffe lectures series). Clarendon Press, Oxford Silber J (1989) Factor components, population subgroups and the computation of the Gini index of Inequality. The Review of Economics and Statistics 71(2),107-115 Wayland D (2003) Measure it in Japanese (http://www.thefightschool.demon.co.uk) WBG (2001) Inequality measurement. World Bank Group, Washington, D.C. (Internet edition) Williamson JG (1965; 1975 reprint) Regional inequalities and the process of national development: a description of the patterns. In: Friedmann J, Alonso W (eds), Regional policy. The MIT Press, Cambridge, Massachusetts 158-200 Wojan TR, Maung AC (1998) The debate over state-level inequality: transparent method, rules of evidence and empirical power. The Review of Regional Studies 28(1),63-80 Yitzhaki S (1994) Economic distance and overlapping distributions. Journal of Econometrics 61,147-159 Yitzhaki S, Lerman RI (1991) Income stratification and income inequality. Review of Income and Wealth 37(3):313-329

5

Investigating Spatial Patterns of Income Disparities Using Coordinate Transformations and GIS Mapping

Boris A. Portnov1 and Rimma Gluhih2 1

2

Department of Natural Resources and Environmental Management, University of Haifa, Israel Jacob Blaustein Institute for Desert Research, Ben-Gurion University of the Negev, Israel

5.1

Introduction

In urban and regional studies, spatial disparities (in average incomes, local productivity, and employment) are commonly investigated for large geographic units, such as states, counties, and administrative regions (Markusen 1996; Sala-iMartin 1996; Martin and Sunley 1998). However, such spatial units are often internally heterogeneous. For instance, a county or region may contain both developed and economically lagging localities. Since in comparative analysis either aggregates or averages are investigated, resulting assessments may thus be misleading. In contrast, if development levels of individual localities are compared, the analysis may yield more accurate estimates. There are various mapping techniques that help to visualize and compare development levels exhibited by small geographic units, such as individual localities, wards, and small census areas. These techniques include contour maps, density diagrams, graduated symbols, shaded colours, and bar charts (Gatrell 1994; Dorling 1994; Wood 1994; Gahegan 1998; Dykes 1997; Murray and Shyy 2000). However, none of these techniques is sufficiently informative. Thus, contour maps may assign high development levels (e.g. high average incomes) to places that lack any development, such as undeveloped land between localities. Graduated symbols and bar charts do reflect development levels in the places in which development actually occurs (i.e. in individual towns, villages, and neighbourhoods). However, if the overall number of observations increases, individual symbols on such maps tend to overlap, making the actual patterns of inter-area income disparities hardly unidentifiable. The problem may become especially acute when income disparities in small countries are analysed: in such countries, urban settlement is both dense and concentrated, and symbol overprints are thus especially likely (Figure 5.1). Zooming to clusters of individual localities on the computer screen may help to visualize inequality patterns more clearly. However, such zooming and other

64

Boris A. Portnov and Rimma Gluhih

interactive manipulations are of little help when the general picture of income disparity for the whole country is needed.

A

B

C

D

Fig. 5.1. Alternative mapping techniques of inter-urban income disparities. A - Graduated circles; B - Bar charts; C - Shaded symbols; D - Contours

Spatial Patterns of Income Disparities

65

The present chapter attempts to introduce an alternative way of visualization of inter-urban income disparities, which may assist in their exploration and analysis. The proposed technique is based on the transformation of distances between individual localities in proportion to actual differences in the average income levels. The chapter starts with a brief discussion of visualization techniques commonly used in urban and regional analysis. Previous studies of development inequalities in Israel are then discussed in brief. In concluding sections, the proposed visualization technique is discussed, and its performance is tested using income data for urban localities in Israel in 1990 and 1999.

5.2

Visualization Techniques of Spatial Disparities

Choosing the way of representing data on the map is an important step in geographic analysis (Cambell 1991; Tyner 1992; Dent 1993; Robinson et al. 1995). The quantitative characteristics of individual features (such as counts, ratios and ranks) can be represented on maps in different ways. Among traditional methods are symbols of varying size and colour; clusters of data; contours and charts. In recent years, the use of Geographic Information Systems (GIS) for mapping and analysing spatial data has become increasingly popular (Clapp et al. 1997; Kraak and Maceachren 1999; Gahegan 1998; Rusanen et al. 2001; Andrienko and Andrienko 1999). Thus, Clapp et al. (1997) discuss potential contributions of GIS to urban economic research and they open debate on how GIS can be applied to this field. The use of maps is perceived as a major tool for enabling new ways of analysing urban issues, including inter-urban disparities in development. The areas of GIS application in urban and regional planning include land-use mapping; transportation mapping and analysis of efficient transport routes for delivery and emergency response; geo-demographic analysis for facilities location, mapping of infrastructure utilities, and multiple applications in natural resource management (Carsjens and Van der Knaap 2002; Adams et al. 1992; Câmara et al. 2002; Jun 1999; Xie 1995). Sui (1998) reviews the practices, problems, and prospects of GIS-based urban modelling, contending that the integration of urban analysis with GIS must proceed with the incorporation of multi-dimensional concepts of space and time. In separate studies, Theseira (2002) and Mitchell et al. (2002) use GIS-mapping for a study of health, socio-economic and environmental inequalities in the U.K. Considerable attention has been devoted to the use of spatial analytical techniques in conjunction with socio-economic measures. Thus, Porter and Tarrant (2001) study socio-economic and racial inequalities to determine relationship between income and occupation and location of a number of federal tourism sites. In another study, Klitgaard and Fitschen (1997) investigate income inequality and poverty across rural areas of KwaZulu-Natal province of South Africa, using GIS for geo-statistical analysis.

66

Boris A. Portnov and Rimma Gluhih

Ceccato and Persson (2002) and Rusanen et al. (2001) use thematic mapping to identify the degree of agglomeration of economic activity through clustering patterns related to employment. Wu and Webster (2000) combine neo-classical urban economic theory with systems methods with explicit treatment of spatial relationships and time-series data; GIS in their model is used to explore microeconomic trends in urban areas, particularly those relating to space-dependent urban processes. Dykes (1997), Andrienko and Andrienko (1999), and Gahegan (1998) develop exploration tools for the analysis of spatial data using dynamic links between spatio-temporal data and GIS-graphics. Thus, Dykes (1997) uses such a link for querying the number of tourists at different locations of a city throughout the day. Descartes, the software package discussed in Andrienko and Andrienko (1999), provides a range of tools for visual exploration of spatially referenced data, including automated presentation of data on maps and their interactive manipulation. This software package includes advanced tools for zooming, handling geographic layers, and user-friendly controls that make it possible to adjust various visual properties of the data presentation. In the following subsections, visualization techniques that can be used for mapping geographic disparities are discussed in some detail. 5.2.1

Graduated Colour / Graduated Symbol Maps

A simple way to represent the different parameter values in localities is either to change the colour of the symbol (graduated colour) or to vary the symbol’s size according to the values of a particular attribute (graduated symbol). For instance, Figure 5.1A uses larger symbols to show localities with higher average incomes. However, graduated colour/graduated symbol maps may not always be sufficiently informative, specifically if individual localities (e.g. towns) are too numerous and located close to each other. The largest symbol needs to be small enough to prevent neighbouring localities from completely covering each other. At the same time, the difference in the size of symbols needs to be large enough for each class to remain distinct (Dent 1993). When urban settlement is dense, the compliance with these requirements may not always be possible. Dorling (1994) proposed an interesting method of visualization of development disparities between small geographic areas using cartograms, composed by nonoverlapping graduated symbols. According to his approach, aerial units are repelled from each other in proportion to their development characteristics (e.g. population size), in order to give more room for places with larger populations. Simultaneously, the “forces of attraction” are applied to the direction of a locality’s original neighbours. Though this method helps to avoid overlaps of individual symbols, it does not allow one to retain geographic features (such as coastlines, borders, etc.) and spatial relationships between individual places, specifically in densely populated regions.

Spatial Patterns of Income Disparities

5.2.2

67

Chart Maps

Chart maps display bar and pie charts over features, making it possible to symbolize multiple attributes (Robinson et al. 1995). Pie charts show relationships between parts and the whole and are particularly useful for showing proportions and ratios. Bar charts compare amounts of related values and are well suited to showing trends over time. However, the chart method has disadvantages similar to those of graduated symbol maps. In particular, when urban settlement is dense, graphs can be displayed only with leaders, i.e. not on the exact positions of localities (see Figure 5.1B). 5.2.3

Cluster Maps

This method of data representation is based on classifying the features with similar values into discrete classes and displaying them with a different symbol for each class (Figure 5.1C). This method may be useful when data values are too numerous to map them individually. The clustering can be based on several factors, which is an advantage of this method. However, like any grouping, clustering cannot reveal hidden patterns within classes. Moreover, it is sensitive to data classification. How the class ranges and breaks are defined (i.e. the high and low values for each class) will determine which features fall into each class and thus the way the map looks: by changing the class breaks, one can create completely different distribution maps (Dent 1993). 5.2.4

Isolines and Surfaces

Isolines and fishnet/mesh surfaces are commonly used to display general trends of parameter distribution (Wood 1994; Robinson et al. 1995; Dent 1993). Since in a study of inter-urban disparities, observed values are valid for particular localities, not for their hinterlands, neither meshes nor isolines reflect accurately actual distributions of parameter values. The use of this method for the visualization of intra-urban disparities may thus be misleading.

5.3

Alternative Visualization Technique

An alternative way to represent inter-urban disparities on the map is to use coordinate transformations. The general principle of the coordinate transformation we propose is to bring closer to a reference point localities with high values of a parameter while moving away the places in which the parameter takes low values. Subsequently, the original positions of the localities and their new positions can be connected by arrows, creating an effect of localities moving to and from a fixed point in space.

68

Boris A. Portnov and Rimma Gluhih

The general formulae for such a transformation can be derived using the properties of similar triangles (see Appendix):

xni

x r   xi  x r

Dni Di

y ni

y r   yi  y r

Dni Di

(5.1)

where: xi and yi - actual x and y coordinates of locality i; xni and yni - x and y coordinates of locality i after transformation; xr and yr - x and y coordinates of the reference city (e.g. a major population centre of a country); Di - actual aerial distance from locality i to the reference city; Dni - distance from locality i to the reference city after transformation. If Dni is defined as a function of the magnitude of income disparities between locality i and the reference city, then three different methods can be used to calculate Dni: “actual distance”, “proportional increment”, and “concentric circle” transformation. 5.3.1

“Actual Distance” Transformation

The simplest way to determine Dni is to estimate it as a function of the actual distance (Di) and the difference in development between locality i and the reference locality (IRi):

Dni where: IR i

Di , IRi

(5.2)

Vi . Vr

(Vi and Vr are parameter values in locality i, and the reference city, respectively. For instance, Vi and Vr may stand for average incomes in the two localities). By substituting (5.2) in (5.1), we obtain:

xni

5.3.2

xr 

xi  xr IRi

y ni

yr 

 yi  y r

(5.3)

IRi

“Proportional Increment” Method

Another simple way to determine Dni is to shift a subject locality by the distance proportional to the difference of parameter value in locality i and the reference city, IRi. Since Dni = Di + ţDi (Appendix) and the distance increment ţDi is assumed to be proportional to the difference in development between locality i and the reference locality (IRi), the new distance Dni in equation (5.1) can be calculated as follows:

Spatial Patterns of Income Disparities

Dni

Di  IRi  1 ˜ C1

69

(5.4)

where (IRi - 1) is a component which allows to preserve the original position of the reference locality (IRr = 1), and C1 is a scaling coefficient which may be determined as follows1: mean Di max Di or: if IRi < 1: C1 C0 C1 C 0 max IRi max IRi § Di · § · Di or: ¸¸ ¸¸ C 0 ˜ tanh ¨¨ C1 C 0 ˜ tanh ¨¨ © max Di ¹ © mean Di ¹ By substituting (5.4) in (5.1), we obtain the formulae for coordinates transformation according to the “proportional increment” method:

if IRi > 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
5.6

Summary and Conclusions

The present paper discusses different ways of visualization of inter-urban income disparities on thematic maps. These disparities reflect differences from a reference level, observed in a chosen (reference) locality. The approach we propose is based on transformations of distances between individual towns and a reference city (e.g. a major population centre of a country) in proportion to actual difference in their income levels. Three alternative methods of such a transformation are proposed and tested: x According to the actual distance method, the spatial “shift” of a locality is proportional to the difference in development between the locality and the reference city, and the aerial distance between them; x According to the concentric circle transformation, localities with identical levels of development are positioned at a certain distance from the reference city, forming concentric circles around it; x Lastly, according to the proportional increment transformation, the distance between a locality and the reference city is adjusted by a parameter whose values are proportional to development disparities between them.

Spatial Patterns of Income Disparities

A

77

B

Fig. 5.6. Changes in average incomes between 1991 and 1999 across urban localities in Israel, per cent A - localities with increased levels of average income; B - localities with decreased levels of income

The advantages and disadvantages of these transformation techniques are analysed, using both a hypothetical distribution of test observations and a case study of inter-urban income disparities in Israel in 1991 and 1999. As the study indicates, the “actual distance” and “concentric circle” transformations have a number of drawbacks, associated with their implementation and interpretation of results. For instance, in the actual distance method, more remote localities always move farther from the reference point than less remote localities with identical levels of development.

78

Boris A. Portnov and Rimma Gluhih

On the other hand, in the concentric circle method, remote localities with low average incomes may move closer to the reference point, instead of moving from it, as could be expected for localities with low income levels. In contrast, the proportional increment method makes it possible to illustrate both the existing patterns of inter-urban disparities and their dynamics over time, and may thus become a useful tool for the visualization and analysis of spatial disparities in urban and regional studies. Compared to graduated and coloured symbols maps (both unclassed and with predetermined class intervals and palettes), which are commonly used in urban and regional analysis, the proposed visualization method has a number of advantages: x First, the proposed method makes it possible to represent the spatial patterns of income disparity more clearly and accurately, since arrows are less likely to overlap than graduated circles etc. and the arrow lengths stay always proportional to actual differences in parameter values between individual localities; x Second, the proposed “arrow” maps can be printed both in black and white and in colour without loosing any essential visual information; in contrast, graduated colour maps may loose much of their visual information, if printed in greyscale; x Third, the proposed “arrow” method is especially informative for visualization of the spatial patterns of change in the level of socio-economic variables, such as population growth, income change, etc. In contrast to graduated symbol/graduated colour maps that look visually static, the arrows on the coordinate transformation maps create a dynamic appearance. A number of additional advantages of the proposed visualization method also deserve a note: Differenced between positive and negative values of a parameter can be shown clearly by using arrows leading into the opposite directions. Irrespective of the map scale, parameter value, and aerial proximity of localities, the beginning of the arrow shows the exact location of the locality (which cannot be done, for instance, using bar charts). It is also important that different attributes can be shown simultaneously on the map. For example, for the visualization of the second parameter, one can use different arrow types, such as solid, dashed, dotted etc. A possible way to incorporate the proposed visualization method into GIS software is to develop a dynamically linked two-display solution, with a possibility of choosing a reference object, and with “in” and “out” arrows for more developed and less developed localities shown on the second display in different colours.

Spatial Patterns of Income Disparities

79

Appendix Estimation of general formulae for coordinate transformations

yni

F

yi E

yr

Assume that Points C (xi, yi) and D (xni, yni) represent respectively the actual position of locality i and its position after transformation. Point O (xr, yr) describes the position of the reference centre. Using the property of similar triangles, ǻOAC and ǻOBD, and ǻOEC and ǻODF, we obtain:

ni D

i

C

O

A

xr

xi

B xni

OA

OB

OE

OF

OC

OD

OC

OD

Rewriting these proportions in terms of coordinates and distances, we get:

xi  xr Di

xni  xr Dni

yi  y r Di

y ni  y r Dni

After simple algebraic transformations, general formulas for coordinate transformations are obtained:

xni

x r   xi  x r

Dni Di

y ni

y r   yi  y r

Dni Di

References Adams TM, Vonderohe AP, Russel JS, Clapp JL (1992) Integrating facility delivery through spatial information. Journal of Urban Planning and Development - ASCE 118(1):13-23 Andrienko GL, Andrienko NV (1999) Interactive maps for visual data exploration. International Journal of Geographical Information Science 13(4):355-374 Câmara G, Monteiro AM, Ramos FR, Sposati A, Koga D (2002). Mapping social exclusion/inclusion in developing countries: social dynamics of São Paulo in the 1990s. Draft version Cambell J (1991) Introductory cartography. Wm C. Brown Publishers, Dubuque, etc. Carsjens GJ, Van der Knaap W (2002). Strategic land-use allocation: dealing with spatial relationships and fragmentation of agriculture. Landscape and Urban Planning 58(2-4):171-179

80

Boris A. Portnov and Rimma Gluhih

Ceccato V, Persson LO (2002) Dynamics of rural areas: an assessment of clusters of employment in Sweden. Journal of Rural Studies 18(1):49-63 Clapp JM, Rodriguez M, Thrall G (1997) How GIS can put urban economic analysis on the map. Journal of Housing Economics 6(4):368-386 Dent BD (1993) Cartography: thematic map design. Wm. C. Brown Publishers, Dubuque, etc 3rd Edition Dorling D (1994) Cartograms for human geography. In: Hearnshaw HM, Unwin DJ (eds) Visualization in geographical information systems. Wiley, Chichester, England, pp 85102 Dykes JA (1997) Exploring spatial data representation with dynamic graphics. Computers and Geosciences 23(4):345-370 Gahegan MN (1998) Scatterplots and scenes: visualization techniques for exploratory spatial analysis. Computers, Environment and Urban Systems 21(1):43-56 Gatrell A (1994) Density estimation and the visualization of point patterns. In: Hearnshaw HM, Unwin DJ (eds) Visualization in geographical information systems. Wiley, Chichester, England, pp 65-75 ICBS (2001a) List of localities, their population and codes. Israel Central Bureau of Statistics, Jerusalem ICBS (2001b) Statistical abstract of Israel. Israel Central Bureau of Statistics, Jerusalem Jun MJ (1999) An integrated metropolitan model incorporating demographic-economic, land-use and transport models. Urban Studies 36(8):1399-1408 Klitgaard R, Fitschen A (1997). Exploring income variations across traditional authorities in KwaZulu-Natal, South Africa. Development Southern Africa 14(3):363-376 Kraak MJ, Maceachren A (1999) Visualization for exploration of spatial data. International Journal of Geographical Information Science 13(4):285-287 Markusen A (1996) Interaction between regional and industrial policies: evidence from four countries. International Regional Science Review 19(1):49-77 Martin R, Sunley P (1998) Slow convergence? The new endogenous growth theory and regional development. Economic Geography 74(3):201-227 Mitchell R, Dorling D, Shaw M (2002) Population production and modeling mortality - an application of geographic information systems in health inequalities research. Health and Place 8(1):15-24 Murray AT, Shyy TK (2000) Integrating attribute and space characteristics in choropleth display and spatial data mining. International Journal of Geographical Information Science 14(7):649-667 Porter R, Tarrant MA (2001) A case study of environmental justice and federal tourism sites in Southern Appalachia: a GIS application. Journal of Travel Research 40(1):27-40 Robinson AH, Morrison JL, Muehrcke PC, Kimerling AJ, Guptill SC (1995) Elements of cartography. John Willey and Sons, Inc, NY, 6th Edition Rusanen J, Muilu T, Colpaert A, Naukkarinen A. (2001). Finnish socio-economic grid data, GIS and the hidden geography of unemployment. Tijdschrift Voor Economische en Sociale Geografie 92(2):139-147 Sala-i-Martin X (1996) Regional cohesion: evidence and theories of regional growth and convergence. European Economic Review 40:1325-52 Sui DZ (1998) GIS-based urban modelling: practices, problems, and prospects. International Journal of Geographical Information Science 12(7):651-671 Theseira M (2002). Using internet GIS technology for sharing health and health related data for the West Midlands Region. Health and Place 8(1):37-46 Tyner J (1992) Introduction to thematic cartography. Prentice Hall, Englewood Cliffs, NJ

Spatial Patterns of Income Disparities

81

Wood J (1994) Vizualizing contour interpolation accuracy in digital elevation models. In: Hearnshaw HM, Unwin DJ (eds) Visualization in geographical information systems. Wiley, Chichester, England, pp 168-180 Wu FL, Webster CJ (2000) Simulating artificial cities in a GIS environment: urban growth under alternative regulation regimes. International Journal of Geographical Information Science 14(7):625-648 Xie YC (1995) The overlaid algorithms for aerial interpolation problem. Computers, Environment and Urban Systems 19(4):287-306

Part II: Empirical Evidence

6

Regional Employment Disparities in Belgium: Some Empirical Results

Olivier Meunier and Michel Mignolet Centre de Recherche sur l’Economie Wallonne, University of Namur, Belgium

6.1

Introduction

Employment and unemployment are unequally distributed over space and Belgium is no exception to the rule, its regions recording strongly contrasting performances. The Flemish region displays high levels of participation rate in the labour market and low rates of unemployment. By contrast, Wallonia and Brussels are characterized by poor results for both employment and unemployment. Moreover, the interregional divergence in the performance of the labour market shows no significant tendency to decrease. See Appendix 1 for a breakdown of Belgian spatial administrative divisions. Over the period extending from 1973 and 2001, Flanders recorded an increase of 33 percent in the number of employees, while Brussels experienced a decline of 1 percent and Wallonia had a slow increase of 6 percent. This employment pattern feeds the political debate in Belgium, putting a strain on the national consensus between the Flemish-speaking Northern region and the French-speaking Walloon region. Within the regions, a similar performance disparity is observable between districts.1 For example, Bastogne and Charleroi are districts of the same region, Wallonia. Between 1997 and 2001, the number of jobs in Bastogne increased by 98 percent, while Charleroi lost 19 percent of its labour force. Figure 6.1 illustrates, for the different districts, the regional and inter-regional employment disparities in employment rates in 2001.2 To explain the interregional discrepancy, it has long been argued that poor employment results recorded in Wallonia relative to Flanders are primarily ascribable to an initially adverse industrial mix. However, recent work has questioned this hypothesis, stressing instead the mediocre sectoral growth performance in Wallonia (see Estevão 2003). This chapter sets new light on this issue, notably by using a highly disagregated employment series. More precisely, we intend to examine regional employment disparities in Belgium and their evolution in the recent past by developing a shift-share analysis and an econometric approach and applying this to employment statistics. 1

2

We are referring to the territorial entities between regions and municipalities, which in Belgium are called arrondissements. There are 43 Belgian districts (see Appendix 1). The employment rate is defined for each district as the ratio of the number of employed and self-employed workers to the total population.

86

Olivier Meunier and Michel Mignolet

Fig. 6.1. Regional and inter-regional disparities in Belgium: employment rate at the district level in 2001 (employed and self-employed workers as a percent of the total district population)

The layout of the chapter is as follows. The traditional shift-share identity is explained in Section 6.2, while Section 6.3 provides a short overview of the first empirical findings. To overcome some shortcomings of the traditional analysis, Section 6.4 and 6.5 turn to more advanced approaches. They deal respectively with the dynamic shift-share technique and the econometric model developed by Marimon and Zilibotti (1998). In both sections, we briefly outline the method before giving the main results. Section 6.6 briefly concludes.

6.2

The Traditional Shift-Share Method

The shift-share approach sets out to analyse the extent to which the difference in employment growth between each region and the national average is due to the regional industrial structure, or to a residual element that can be interpreted as indicating the locational (dis)advantages of each regional economy.3 First, the 3

The origin of the shift-share technique goes back to the seminal works of Perloff et al. (1960). This method has been extensively used as well as imaginatively applied in the field of regional economics.

Regional Employment Disparities in Belgium

87

growth disparity of regional employment can be attributed to the regional productive structure. The underlying intuition is that regional employment is growing if the region specializes in fast-growing industries and under-specializes in low-growth industries. Conversely, an over-representation of declining industries and an under-representation of high-growth industries in a region explain the relative decay of regional employment. This component is called the structural effect or the industry mix effect. The second explanation considers all other factors that are not related to the economic structure. Regional employment disparity is then ascribable not to the structure of employment, but rather to an inadequacy of job creation or to an excess of job destruction in some regions. In other words, a high unemployment rate reveals that the performance of the region’s industries in generating jobs is weaker than their performance at the national level. This second component is called either the regional share, or the residual or performance effect. The terminology is of little importance if it is understood that the residual component incorporates all factors not related to the activity structure. Some of these are region-specific factors, such as a lack of entrepreneurship on behalf of private agents, the inefficiency of public policies as regards unemployment and misallocation of public expenditures. Others are exogenous factors, such as shocks affecting all European economies, or the ability of the region to attract labourintensive multinational companies. The traditional shift-share model is defined as follows (see Equation 6.1).4 It splits into two component parts the growth rate differential of employment in a particular region and in a benchmark entity, usually the whole country.5 Both effects may be observed simultaneously:

gr - gn

6 iI 1win ( g ir - g in )  6 iI 1 g in ( wir - win )

(6.1)

i=1,…,I where: x gir and gin denote the rate of employment change in industry i, at the regional and national level, respectively; t 1

gr

4

5

t

¦i Eir  ¦i Eir 6 i Eirt

gn

6 i Eint 1  6 i Eint 6 i Eint

This formulation, sometimes called ‘two-factor’ shift-share analysis, was initially developed by Beaud (1966) and is now commonly adopted in a growing number of applications. See for instance Guesnier (1998), Fernández and Menédez (2002) or Estevão (2003). The conventional shift-share expression commonly distinguishes between three components: the national share, the industrial mix (or proportional shift), and the regional share (or differential mix). Its formulation is appropriate for examining employment change in a particular region but the two-factor shift-share approach seems more suitable when one region is compared to another (see Dormard 1999).

88

Olivier Meunier and Michel Mignolet

t x E irt and E in stand for sector i’s employment at the beginning of a time interval extending from t to (t+1), in the region and at national level, respectively;6 x i represents the industry; and x wir and win denote industry i’s employment share in the region and the nation, respectively.

wir

Eirt ¦i Eirt

win

Eint 6 i Eint

The second term on the right-hand side of Equation 6.1 measures the structural effect. It expresses the difference of regional employment growth that would have been observed if each regional industry had recorded its respective national employment growth rate. Regional disparities are thus only ascribable to the initial regional specialization in growing or declining industries. Accordingly, a positive (negative) result is indicative of a favourable (unfavourable) regional industry mix. The origin of a positive structural effect is twofold: either the region benefits initially from an activity portfolio consisting mostly of activities that have recorded a positive national employment growth or the opposite occurs and the region is characterized by a portfolio including the relatively few industries that have suffered from severe jobs losses at the national level. The first term on the right hand-side of the equation (Equation 6.1) is the residual or regional effect. It compares the regional employment growth rate in each industry with its respective growth rate at the national level. A negative residual effect is therefore indicative of a region-specific lack of growth performance. Shift-share analysis provides an identity that attributes the difference between aggregate regional and national employment growth rates to an industrial structure effect and a residual effect. The model does not allow for isolating the determinants of the residual effect. It does not explain why an industry performed better in a region or why, on the contrary, it was more affected by exogenous shocks. Shift-share analysis is just a descriptive tool whose main purpose is to explain past regional growth in terms of structural shift. Consequently, as Stilwell (1970) underlines, it is risky to implement economic policies exclusively grounded on shift-share results - that is, on a description of the past growth of employment.7 Most of the criticisms directed towards shift-share analysis seem to have arisen from an inadequate utilization of the model. However, even the simple use of the technique as a descriptive tool must be discussed. First, the quality of shift-share results depends greatly on the activities’ homogeneity in the sectoral breakdown. If an industry includes a fast-growing sector and a sector recording a zero growth, the aggregation of the two branches will give moderate sectoral growth. Now, if one region is specialized in the first sector and another in the second, the regional shift will capture some structural 6 7

Alternatively n stands, in a more general way, for any other reference region. See also Stevens and Moore (1980).

Regional Employment Disparities in Belgium

89

effects. Accordingly, the more thorough the sectoral decomposition, the more significant are the chances of correctly identifying the structural effects.8 However, a detailed sectoral and geographical breakdown, if it is desirable, leads to a computational difficulty related to infinite growth rates. This situation arises when a company locates in a region where its industrial sector was absent at the beginning of the period. Economic literature proposes several methods to meet this issue. These include the aggregation of sectors (so that the industry’s initial employment is no longer equal to zero), the replacement of every zero by one, or the replacement of all infinite growth rates by zero (De Brabander 1975).

6.3

Regional Employment Disparities in Belgium: Initial Results

We used employment data published by the National Social Security Organization (ONSS) between 1995 and 2001.9 Sectoral data for regional employment are provided for a highly detailed breakdown based on the general classification of economic activities, NACE-BEL, to five digits. As mentioned above, the use of very disaggregated sectoral statistics increases the accuracy of the shift-share analysis. The choice of the correct regional breakdown is equally significant. In particular, if geographic areas are too narrow, takeovers of companies and industrial mergers, as well as any corrections made to the industry classification, are likely to confuse the interpretation of our results. Accordingly, we have decided to use data published at the regional NUTS 3 level, which in Belgium correspond to the district level.10 However, the major difficulty in studying large series of data is to interpret the results. The shift and share effects that are observed stem from the different performances of employment in a great number of activities, each of them hardly affecting the final results. In addition, the role of shift-share analysis is precisely to provide the keys for the interpretation of aggregate statistics reflecting the evolution of employment. It is not relevant to undertake a shift-share study for observing, for example, the peculiar employment evolution at the industry level.11 The introduction of the new nomenclature of economic activities, NACE-BEL, in 1993, prevents us from comparing the employment statistics for a detailed sectoral breakdown before and after the classification change. Moreover, the ONSS made many corrections to its industry classifications during 1993 and 1994. 8

9

10 11

The issue of interaction between industries is linked to this first limitation. For example, a firm producing windscreens may have been classified into the glass industry, although it is probably highly dependent on the growth of the car industry (see Buck 1970). We consider data on employees for both the private and public sectors. We do not use data on civil servants within Public Administration (NACE 75 code) and Education (NACE 80 code), nor on self-employed workers. The reader may find results concerning self-employed workers in Meunier and Mignolet (2003). NUTS stands for Nomenclature of Territorial Units for Statistics. We have replaced all infinite employment growth rates by zero. See Section 6.1.

90

Olivier Meunier and Michel Mignolet

These corrections are likely to introduce a bias into the results of the shift-share analyses. Therefore, we have decided to restrict the investigation to the period extending from 1995 to 2001, which offers the most homogeneous series of data. Table 6.1 shows the results of the traditional shift-share analysis undertaken on employment in the Belgian regions between 1995 and 2001. The benchmark entity is the country as a whole. Table 6.1. Shift-share components, Belgian regions employment, 1995-2001 (national benchmark) percent

Regions Flemish region Walloon region Brussels

Residual share 2.019 0.424 -8.220

Industry mix -0.401 -1.044 2.921

Total shift 1.618 -0.620 -5.299

Over the period 1995-2001, employment in Flanders recorded a stronger progression than in the whole country. The positive growth difference between Flanders and the Kingdom amounts to 1.6 percent. This deviation is explained by a strong regional shift (2 percent) that outdoes the unfavourable effects of the productive structure in the north of the country (-0.4 percent). By contrast, the Walloon region displays a weaker growth rate of employment than the national average. The total growth difference over the period is -0.6 percent. This result is explained by an unfavourable structural effect (-1 percent) that dominates a positive residual effect (0.4 percent). Finally, if Brussels benefits from the favourable composition of its activity portfolio (2.9 percent), it experiences a very significant deficit of performance (-8.2 percent). This lack of “regional dynamism”, i.e. the residual effect of the firms located in Brussels, explains the negative growth differential of employment (-5.3 percent) in the capital as compared with the Kingdom. The observations on a regional scale are the result of the employment performances recorded by the districts that make up the regions. Table 6.2 provides the results of the shift-share analysis on the Belgian districts between 1995 and 2001. The Flemish region recorded faster employment growth than the country as a whole. This difference is not explained by a favourable activities portfolio, but rather by a strong positive regional effect. However, at the district level, one observes that the growth differentials of employment between Flanders and the country are negative in 17 districts out of 22. For these districts, the structural effect dominates even the residual effect in nine cases out of 17. Thus, the strong performance of Flanders as a whole is due to five districts out of 22 - in particular, to Hal-Vilvoorde, which is located close to Brussels. For these districts, both structural and residual effects are positive, except for Maaseik that records an unfavourable industry mix.

Regional Employment Disparities in Belgium

91

Table 6.2. Shift-share components, Belgian districts employment, 1995-2001 (national benchmark) percent Districts Flanders Eeklo Aalst Oostende Oudenaarde Diksmuide Kortrijk Dendermonde Roeselars Veurne Brugge Antwerpen Ieper Mechelen Hasselt Tielt Gent Tongeren Leuven Sint-niklaas Maaseik Turnhout Hal-Vilvoorde Wallonia Virton Mons Ath Arlon Bastogne Soignies Charleroi Thuin Dinant Huy Marche-en-Famenne Tournai Waremme Verviers Liege Philippeville Neufchateau Namur Nivelles Mouscron The results are ranked in ascending order.

Residual share

Industry mix

Total shift

-13.450 -3.001 -5.685 -2.638 -4.304 -5.929 0.395 -1.800 -6.951 2.115 -0.352 -2.433 -2.349 0.695 0.922 -1.744 2.388 0.990 3.006 5.760 4.923 3.124

-1.999 -6.705 -2.917 -5.531 -2.909 -0.946 -6.139 -3.543 2.785 -5.806 -3.289 0.102 0.282 -1.723 -1.678 1.322 -2.727 2.379 0.717 -1.062 1.030 11.012

-15.449 -9.706 -8.602 -8.169 -7.213 -6.875 -5.745 -5.343 -4.166 -3.691 -3.641 -2.330 -2.066 -1.028 -0.756 -0.422 -0.339 3.370 3.723 4.698 5.952 14.137

-18.726 -9.141 -7.255 -1.766 -2.158 -7.713 -5.047 0.023 -6.317 -5.270 -0.369 -1.436 -5.688 -0.126 -2.765 -2.684 2.814 2.236 0.231 13.118

-10.044 -2.645 -1.564 -6.794 -5.627 0.029 -1.047 -5.111 1.576 0.723 -3.135 -1.931 2.566 -2.861 -0.124 1.439 -2.120 -0.859 3.235 -7.067

-28.769 -11.786 -8.818 -8.560 -7.784 -7.684 -6.094 -5.087 -4.741 -4.547 -3.505 -3.367 -3.122 -2.987 -2.890 -1.246 0.694 1.377 3.466 6.051

92

Olivier Meunier and Michel Mignolet

Within Wallonia, the only district recording positive results for both shift and share effects is Nivelles, located close to Brussels. Namur, Neufchâteau and Mouscron are also characterized by higher growth rates of employment than the national average. For these districts, the growth differential is explained by a positive residual effect (2.2 percent for Namur, 2.8 percent for Neufchâteau and 13.1 percent for Mouscron) associated with a negative structural effect (-0.8 percent for Namur, -2.1 percent for Neufchâteau and -7 percent for Mouscron). Employment in the other Walloon districts experiences a lower growth rate than that of the Kingdom. This differential seems primarily ascribable to a lack of performance. In ten out of the remaining sixteen districts, the regional effect (when compared with the structural effect) appears to exert a dominant influence in terms of explaining the growth differentials. For Soignies, Dinant, Huy, Waremme and Philippeville, the lack of performance cancels out even the benefit of a relatively favourable activity structure. The effect of the productive structure dominates the districts of Arlon, Bastogne, Thuin, Verviers, Tournai and Marcheen-Famenne. For the district of Thuin, the residual effect is positive. These results illustrate for the first time the disparate district performances in terms of employment, and the dispersed contributions of the industry mix and the residual effects. Are these results due to the method and its simplicity? The economic literature stresses that traditional shift-share analysis has indeed an obvious shortcoming. It allows only a study of a time interval between two dates. It is well known that the composition of economic activity may be modified over time. Accordingly, Barff and Knight (1988) developed a dynamic shift-share analysis, which re-actualises the structure of activity at each period. The dynamic shift-share approach then amounts to adding up the results obtained by applying the traditional expression to annual data.

6.4

The Dynamic Shift-Share Model

Dynamic shift-share analysis has three major advantages: first, it is independent of the reference period (often the initial period, or, as proposed by Stilwell (1970), the last year). Second, it takes into account the compound effects, i.e. the automatic change of regional disparities at each period when the structure is not actualised. Finally, the dynamic shift-share model makes it possible to visualize easily, with a simple graph, the evolution of both components over time. The dynamic shift-share analysis that is defined as the yearly computation of both structural and regional effects, can accordingly be formulated as follows (see Equation 6.2): T 1

¦ ¦iI 1 (

t 1

T 1 Eirt 1 Eint 1 Eirt Eirt E int E int 1 I  ) ( ) ¦ ¦ 1 i t t t t 1 Eirt Eint ¦i Eirt ¦i E r ¦i Ein Ein

t = 1,…,T

(6.2)

Regional Employment Disparities in Belgium

93

Table 6.3 reports the results of the dynamic shift-share analysis on employment by districts compared with the Kingdom between 1995 and 2001. The results of the dynamic shift-share analysis broadly confirm the conclusions of the traditional analysis. The ranking of the districts remains basically unchanged. The residual effect of the Walloon districts is still a dominant cause of the growth differentials (in 14 districts out of 20), whereas in the Flemish region, the structural effect is dominant in ten districts out of 22. Table 6.3 shows that employment increases more in Flanders and less in the Walloon region and in Brussels compared with the whole country. The growth rates of employment attributable to the districts and regions are somewhat different from those shown in Tables 6.1 and 6.2. This is due to the dynamic process implemented by Barff and Knight. The sum of yearly components no longer equals the total employment growth differential. Accordingly, the dynamic approach breaks slightly with the traditional shift-share identity. A remarkable difference in comparison with the results of the traditional shiftshare approach should be highlighted. The residual effect appears to be negative for the Walloon region. The differential of growth rates between Wallonia and Belgium results from two negative effects: industry mix and regional shift. Figures 6.2 to 6.4 display the annual evolution of shift and share effects from 1995 to 2001 for Flanders, Wallonia and Brussels, respectively. 0.8% 0.6% 0.4% 0.2% 0.0% -0.2% -0.4% -0.6% 95-96

96-97

97-98

Regional shift

98-99

99-00

00-01

Industy mix

Fig. 6.2. Dynamic shift-share components, Flemish region, 1995-2001 (national benchmark)

In Flanders (Figure 6.2), the effect of performance remains positive over the whole period, except in 1999. The structural effect is sometimes positive, but more frequently negative. Figure 6.3 shows that the growth differential of employment between the Walloon region and the country results not only from an unfavourable structure of economic activity, but also from a relative lack of performance between 1995 and 1998 and after 2000. Regional shift becomes positive in 1999 and 2000.

94

Olivier Meunier and Michel Mignolet

Table 6.3. Dynamic shift-share components, Belgian districts employment, 1995-2001 (national benchmark) percent Districts Brussels

Regional share -6.753

Industry mix 1.791

Total shift -4.963

Eeklo Aalst Oostende Kortrijk Oudenaarde Dendermonde Antwerpen Diksmuide Roeselars Brugge Veurne Mechelen Hasselt Gent Ieper Tongeren Tielt Sint-niklaas Leuven Maaseik Turnhout Hal-Vilvoorde Flemish region

-12.768 -4.370 -7.289 -5.698 -2.326 1.501 0.379 -2.869 -2.143 0.983 -5.744 -2.014 0.297 -1.908 -3.003 1.461 1.759 2.245 0.526 4.711 4.878 2.705 1.816

-1.454 -4.655 -1.184 -0.805 -5.338 -6.368 -4.083 -3.959 -2.799 -4.182 2.142 -0.437 -1.132 1.543 0.491 -1.649 -2.184 0.652 2.455 -0.318 0.588 9.777 -0.367

-12.548 -8.116 -7.744 -6.393 -5.789 -4.075 -3.858 -3.847 -3.807 -3.041 -2.004 -1.891 -0.510 -0.395 -0.196 0.599 1.648 2.783 2.871 5.035 5.075 12.405 1.448

Virton Mons Soignies Ath Bastogne Charleroi Arlon Liege Marche-en-Famenne Huy Tournai Verviers Dinant Thuin Waremme Philippeville Namur Neufchateau Nivelles Mouscron Walloon region

-22.291 -9.797 -6.874 -6.673 -3.770 -5.421 -1.188 -4.231 -0.432 -3.400 -3.734 -0.413 -8.331 0.066 -4.977 -3.527 3.212 4.397 -2.500 13.067 -0.259

-9.949 -2.358 -0.648 -1.598 -3.988 -0.171 -5.883 0.619 -3.041 -1.020 0.301 -2.256 4.153 -4.119 2.098 0.500 -1.893 -3.863 6.557 -6.369 -0.322

-29.918 -10.923 -6.989 -6.038 -5.756 -5.594 -5.232 -3.820 -2.908 -2.873 -2.601 -2.159 -1.690 -1.658 -0.720 0.559 1.805 2.513 3.682 7.807 -0.581

The results are ranked in ascending order.

Regional Employment Disparities in Belgium

95

0.5% 0.4% 0.3% 0.2% 0.1% 0.0% -0.1% -0.2% -0.3% -0.4% 95-96

96-97

97-98

Regional shift

98-99

99-00

00-01

Industy mix

Fig. 6.3. Dynamic shift-share components, Walloon region, 1995-2001 (national benchmark) 2.0% 1.0% 0.0% -1.0% -2.0% -3.0% 95-96

96-97

97-98

Regional shift

98-99

99-00

00-01

Industy mix

Fig. 6.4. Dynamic shift-share components, Brussels, 1995-2001 (national benchmark)

Brussels (Figure 6.4) displays a negative regional effect each year except 1999. This lack of performance, which is at the origin of the differentiated growth of employment in Brussels and in the Kingdom, is the most significant effect, although the activity structure of Brussels generally shows a favourable sectoral composition. The dynamic shift-share approach reveals that the contributions of the residual shift and of the industry mix effect are variable in time in order to explain the disparities in regional employment performances. An important deficiency of shift-share methods is that the effects cannot be statistically tested. As Jayet (1993) emphasizes, without statistical validation, the interpretation of the results is likely to be founded on disparities without significance. Accordingly, numerous extensions have been developed in the economic literature aiming at integrating the shift-share technique into a statistical framework.

96

Olivier Meunier and Michel Mignolet

6.5

An Econometric Approach: Marimon and Zilibotti’s (1998) Model

Berzeg (1978, 1984), Jayet (1981), and, more recently, Patterson (1991) have contributed to converting the shift-share identity into an assessable stochastic formulation. Berzeg used an analysis of variance, while Patterson proposed to substitute the shift-share identity with an analogue regression model.12 More recent approaches in the economic literature tend to lay emphasis on the processing of the time dimension. Amongst these contributions, the statistical model proposed by Marimon and Zilibotti (1998) seems to offer an elegant new way of disentangling region-specific effects from other effects (sectoral and temporal) in the evolution of regional employment. In this way, their model is useful for assessing the broad performance of each region over time, with performance defined here, as before, as the ability of the region to retain a positive (net) employment growth rate. The model is specified as follows (see Equation 6.3):

e(i, n, t ) = h(i ) + m(i, n) + b(t ) + f (i, t ) + g (n, t ) + u (i, n, t )

(6.3)

i = 1,…, I; n = 1, …, N; t = 1, …, T where x e(i, n, t) represents the growth rate of total employment in industry i in region n at time t; x h(i) is a time invariant sectoral trend component that is common to all regions; x m(i, n) is a time invariant effect that is specific to industry i and region n; x b(t) is a pure time effect; x f(i, t) is the interaction between a fixed industry and a time effect; x g(n, t) captures the interaction effect between a fixed region and a time effect; and x u(i, n, t) is the idiosyncratic disturbance that is orthogonal to all other effects. To identify the model, Marimon and Zilibotti (1998) impose restrictions on the coefficients’ values. These restrictions take the sample average as reference points:

12

¦nN 1 mi, n 0 , i = 1, …, I

(R1)

¦iI 1 f i, t 0 , t = 1, …, T

(R2)

¦Tt 1 f i, t 0 , i = 1, …, I

(R3)

¦Tt 1 g n, t 0 , n = 1, …, N

(R4)

See Blien and Wolf (2002).

Regional Employment Disparities in Belgium

¦ nN 1 g n, t 0 , t = 1, …, T

(R5)

¦Tt 1 bt 0

(R6)

97

There are 2T + 2I + N + 1 restrictions, of which all but two are independent. Marimon and Zilibotti (2001) give the following interpretation of coefficients and constraints13: x The summing over regions of the specific effect of industry i and region n gives zero (restriction R1). Thus, the coefficients m(i, n) measure the difference between region-specific employment trends in industry i and the national average rate for the same industry. x The coefficients b(t) provide national business cycle effects, which homogeneously affect employment growth rates across both industries and countries. It is assumed that the business cycle effects average zero over time. x The coefficients f(i, t) represent industry-specific effects which cause a temporary deviation from the employment trend in industry i during t. Industry-specific effects at time t are assumed to average zero over industries (restriction R2). For industry i, these effects are also assumed to average zero over time (restriction R3). x The coefficients g(n, t) express regional transitory deviations of employment growth with respect to the business cycle. The model is estimated applying the restricted least squares method. The residuals of the regression are the estimates of u(i, n, t). This allows us to construct a “virtual” employment series, that is, the employment that would have been observed in the absence of any region-specific components. As the difference between actual employment and virtual employment can only originate from regional performance disparities, the comparison of both series offers another measure of the residual effect provided by the shift-share analysis. Setting m(i, n) = g(i, t) = u(i, n, t) = 0 for all i and t in Equation 6.3, gives the virtual employment growth rates (see Equation 6.4):

evirt = h(i ) + b(t ) + f (i, t )

(6.4)

To find virtual employment in industry i at time t in region n, it is sufficient to apply the virtual employment growth rate of industry i between time t-1 and time t to virtual employment in industry i at time t-1 in region n. Summing virtual employment of all industries located in region n at time t gives virtual employment in that region at time t.14 Considering the ratio of actual employment to virtual employment, Toulemonde (2001) defines a performance indicator: if the ratio is larger (less

13 14

See also Toulemonde (2001). Note that by construction, the virtual employment estimated for the whole country does not equal the level of national employment actually observed.

98

Olivier Meunier and Michel Mignolet

than) than one, industry i of region n (or region n as a whole) is performing better (worse) than the same industry in the country (or in the country as a whole).15 The choice of the employment data set, in particular the degree of sectoral and geographical breakdown, is dictated by a double constraint. First, the estimation of Marimon and Zilibotti’s (1998) specification inappropriately put up with an excessively thorough sectoral and geographical breakdown.16 Second, a highly detailed sectoral breakdown leads to the observation of a larger number of infinite rates of employment growth. Including this issue brings only a limited bias to the traditional shift-share analysis. By contrast, Marimon and Zilibotti’s specification is founded exclusively on employment growth rates that are not weighted by the level of employment. This leads to a number of questions: how is the “appearance” of one or two workers taken into account correctly?; how should the case of a newly created company that takes on 50 workers in the first year of activity be handled? and is it necessary to treat the recruitment of a worker in an identical way when employment increases from zero to one as when it increases from 100 to 101 units? Whatever method is adopted, the correction of the infinite growth rates ends in the addition of false information, which is certainly advisable to limit. Conversely, reducing the sectoral and geographical breakdown seems to provide a more appropriate way to resolve this infinite growth rate issue. Accordingly, we restricted our sample to the employment of the ten Belgian provinces (of which five are Flemish and five are Walloon) and Brussels, for a breakdown of 42 industries.17 The period of investigation extends from 1995 to 2001. The estimate of Marimon and Zilibotti’s (1998) specification, applying restricted least squares, gives the formal statistic of the regression R² = 0.427. In other words, the model seems to explain more than 42 percent of the employment growth rate variance in the ten Belgian provinces and in Brussels during the period 1995-2001. Our R² is definitely lower than the result of Marimon and Zilibotti (1998), who obtained an R² equal to 0.647 for a panel of 15 sectors observed in ten European countries. As Toulemonde (2001) emphasizes, our result indicates that, “at the local level, many shocks are idiosyncratic”, i.e. recorded by the error term u(i, n, t), which tends to reduce the R² value. Marimon and Zilibotti’s (1998) approach enables us to construct a virtual employment series by filtering out all the province-specific components from the actual employment series. The ratio of actual employment to virtual employment 15

16

17

A ratio larger than one does not necessarily mean that the employment level of industry i in region n is increasing; it may be decreasing more slowly than in the whole country (Toulemonde 2001). When the geographical areas are too narrow and/or when the number of sectors is too high, industrial mergers and takeovers or sectoral reclassification are considered exogenous effects, which do not contribute to the ‘spontaneous’ explanation of the employment evolution as estimated by the specification of Marimon and Zilibotti (1998). The sectoral breakdown is based on the classification of economic activities NACE-BEL to two digits, published by the ONSS. We have regrouped some closely related activities for computational ease. The listing of sectoral activities is provided in Appendix 2.

Regional Employment Disparities in Belgium

99

provides a performance indicator. A negative (positive) residual regional effect disclosed by shift-share analysis should be measured by a performance indicator lower (larger) than one. Indeed, the deviation between virtual and actual employment can only originate from factors that are not related to the provincial productive structure but are linked instead to some undefined regional impact. Summing up the virtual employment of all provinces in both regions gives the virtual employment of Flanders and Wallonia. Figures 6.5 to 6.7 show the evolution of actual and virtual employment, as well as the ratio of actual to virtual employment for the Flemish region, the Walloon region and Brussels, respectively, during the period from 1995 to 2001. Between 1995 and 1999, Flemish actual employment records a higher increase than Flemish virtual employment (see Figure 6.5). This relative growth surplus results from the greater Flemish dynamism during the first two years of the period. Until 1997, the growth rate of Flemish virtual employment is inferior to the growth rate of observed employment. This reveals the competitive advantage of the Flemish region, which is a positive regional effect. In 1997, the performance index is equal to 1.008. In other words, if the Flemish region had recorded employment growth for each industry identical to that observed in the country as a whole, employment in Flanders would have been 0.8 percent lower than the employment rate actually observed in 1997. Between 1998 and 2000, the locational advantage of Flanders fades: the virtual employment growth rate becomes superior to the actual employment growth rate. In 2000, virtual employment slightly exceeds real employment. The end of the period marks the return of a positive performance effect in the Flemish region. In 2001, the observed employment is very close to its virtual level, indicating that the good Flemish performance in 1996-1997 has been offset by the poor results for employment creation recorded at the end of the decade. 1700000 1650000 1600000 1550000 1500000 1450000 1400000 1350000

1.01 1.005 1 0.995 0.99 1995

1996

actual employment

1997

1998

1999

virtual employment

2000

2001

actual empl./virtual empl.

Fig. 6.5. Virtual and actual employment and the ratio of actual to virtual employment, Flemish region, 1995-2001

In the Walloon region, the observed employment remains lower than virtual employment over the whole period (see Figure 6.6). Moreover, the region never benefits from a competitive regional advantage. The real employment growth rate never exceeds the virtual employment growth rate. The years 1998 to 2000 are

100 Olivier Meunier and Michel Mignolet

characterized by a stabilization of the ratio between actual and virtual employment growth rates, but at the end of the period, the performance effect becomes unfavourable again. In 2001, the performance index is 0.975. Accordingly, the employment in Wallonia in 2001 might have reached a level 2.5 percent higher if employment growth had not been affected by factors related to the location in the South of the country. 1 0,995 0,99 0,985 0,98 0,975 0,97 0,965 0,96

700000 680000 660000 640000 620000 600000 580000 560000 1995

1996

actual employment

1997

1998

1999

virtual employment

2000

2001

actual empl./virtual empl.

Fig. 6.6. Virtual and actual employment and the ratio of actual to virtual employment, Walloon region, 1995-2001

Brussels records a lack of performance similar to that of the Walloon region (see Figure 6.7). The virtual employment growth rate remains higher than the actual growth rate over the whole period, except for 2000. In 2001, the performance ratio shows a relative growth deficit of employment up to 4.8 percent. 1,01 1 0,99 0,98 0,97 0,96 0,95 0,94 0,93 0,92

500000 480000 460000 440000 420000 400000 380000 1995

1996

actual employment

1997

1998

virtual employment

1999

2000

2001

actual empl./virtual empl.

Fig. 6.7. Virtual and actual employment and the ratio of actual to virtual employment, Brussels, 1995-2001

However, the regional results should be qualified, as a consideration of the evolution of virtual employment at a provincial level discloses significant intraregional disparities. Figures 6.8 and 6.9 show the ratio of actual to virtual employment in the ten Belgian provinces. The Flemish Brabant is the best

Regional Employment Disparities in Belgium 101

performing Flemish province, displaying a ratio higher than one over the whole period. The performance of Antwerp is not very impressive. The Walloon Brabant recorded an actual employment larger than virtual employment from 1999. The provinces of Hainaut as well as of Liège performed quite poorly over the period. 1,1 1,05 1 0,95 1995

1996

1997

1998

1999

Antwerp

Flemish Brabant

East Flanders

Limburg

2000

2001

West Flanders

Fig. 6.8. Ratio of actual to virtual employment, Flemish provinces, 1995-2001 1,08 1,06 1,04 1,02 1 0,98 0,96 0,94 1995

1996 Brabant 1997 Walloon

1998 Hainaut

Luxembourg

Namur

1999

2000 Liège

2001

Fig. 6.9. Ratio of actual to virtual employment, Walloon provinces, 1995-2001

A more circumstantial analysis shows that for most industries, regional performance remains close to the national average, i.e. actual employment is close to its virtual level for a majority of sectors. Figures 6.10 to 6.12 display the average of the ratios of actual employment to virtual employment, by industry, for the Flemish region, the Walloon region and Brussels, respectively.18 The air transport industry (26) performed remarkably in the three regions, but this result actually highlights a limitation of the model. The employment statistics in the air transport industry show a brutal decrease around 1997 in all provinces except Brussels, the Flemish Brabant and Liège (where employment has actually soared). Because of the various constraints imposed by Marimon and Zilibotti’s (1998) specification, the residual employment growth rate does not integrate the atypical employment evolution in these three provinces. Accordingly, residual 18

The industries are listed in Appendix 2.

102 Olivier Meunier and Michel Mignolet

growth tends to reflect the downward trend of employment recorded by the other provinces. 3

26 (3,1)

2,5 2 1,5 1

6

1 2 4 5

0,5

8 7

34 35 38 42 11 12 14 16 18 20 22 24 28 30 32 39 33 36 29 31 9 15 17 19 21 23 25 37 40 41 27 10 13

3

0

Industries

Fig. 6.10. Ratio of actual to virtual employment, by industry, Flemish region, average 1995-2001 3 26 (7,09)

2,5 2 1,5

27 17 37 40 42 19 21 23 29 3132 34 36 4 5 6 8 9 1112 14 1 39 41 13 15 16 18 20 22 24 25 28 30 33 35 38 3 7 10 2

1 0,5 0

Industries

Fig. 6.11. Ratio of actual to virtual employment, by industry, Walloon region, average 1995-2001 3 26

2,5 2 1,5 4

1 0,5

1

16 18 6 89 12 15 20 22 24 17 19 21 23 5 7 10 11 13 14

42 38 30 32 35 37 40 27 29 31 33 34 36 39 41 25 28

0 Industries

Fig. 6.12. Ratio of actual to virtual employment, by industry, Brussels, average 1995-2001

In the Flemish region, other activities that performed well are the computer industry (34), the renting of machinery and equipment (33) and insurance and pension funding (30). Walloon’s dynamic industries are the supporting and auxiliary transport activities (27), the manufacturing of office, shop and other

Regional Employment Disparities in Belgium 103

furniture and the recycling of metal waste (17). In Brussels, the sectors that performed well are post and telecommunications (28), research and development (35), sanitation and similar activities (38) and extra-territorial organizations and bodies (42). In Flanders, the activities that show a deficit in performance are mainly the extraction of coal, lignite and peat (3), mining of metal ores (10) and water distribution (25). In Wallonia, the industries that employ significantly fewer workers than their corresponding virtual employment levels are, like Flanders, the extraction of coal, lignite and peat and the mining of metal ores as well the production of leather accessories (7). In Brussels, the performance indicator is significantly lower than one for the extraction of coal, lignite and peat, the distribution of water and the supporting and auxiliary transport activities (27).

6.6

Conclusion

This chapter has examined regional and intra-regional performances of employment in Belgium in the recent past. Different methods were implemented to investigate this question: the traditional shift-share approach, the dynamic shiftshare version or the econometric model of Marimon and Zilibotti. The results are converging, whether expressed in terms of the industry mix effect and regional shift or in terms of virtual and actual employment. They show that employment dynamics are disparate in the three Belgian regions and are driven by different forces linked to the sectoral specialization (or the lack of specialization) of the regions. Neither Flanders nor the Walloon region has inherited a favourable industrial mix. In the Flemish region, the weight of the infavourable structure is overcompensated by better dynamic performances. In a way, Flanders holds some wining cards. The region gives firms locational advantages due either to natural endowments (easy access to the sea, for example) or to non-traded infrastructures (a dense highways network, well equipped harbours and airports, etc.). Firms also benefit from the agglomeration economies attributable to pecuniary externalities (the proximities to a large market) and to production externalities (knowledge spillovers, specialized workforce, opportunities for efficient subcontracting resulting from the closeness to firms engaged in similar activities, or cross fertilization, easy access to complementary services and inter-industry information exchanges due to urbanisation economies). These locational advantages stimulate regional economic activity and accordingly employment creation. On the contrary, the Walloon region’s ability to create jobs has remained too weak to overcome its structural handicap. Wallonia, which took the lead of the industrialization process in the 19th century, has remained too involved in declining industries (such as steel) and, with reference to the economic base theory, in non-basic activities (such as social and public services). Conversely, the region suffers from a lack of specialization in the service sector, notably in financial intermediation and business to business activities. In addition, Wallonia is not as well-endowed as Flanders. Accordingly, it faces a slight productivity

104 Olivier Meunier and Michel Mignolet

handicap that slows down the employment dynamics. This is particularly ascribable to a common labour market that insufficiently differentiates the wage rates between regions. As Estevão (2003) suggests, other likely reasons are “poor job matching” and “low labour mobility”. Let us observe that performance disparities appear to be even stronger at an intra-regional level, in the provinces and districts. Over the whole period considered, no convergence of employment performances was observed. In particular, Brussels, which is both an administrative region and a district, has benefited from a highly favourable industrial mix but has recorded poor residual employment performances. Basing public policies exclusively grounded on shift-share results is a risky process. The shift-share technique proves useful however “for providing guidance for industrial targeting” and contributes “to understanding and selection of key leading industries in a region” (Dinc 2002). Broadly speaking, our analysis stresses the need for political measures aimed at increasing the relative share of “basic” sector activities in the Walloon industrial mix and the need for raising productivity levels.

Appendix 1 Table 6.4. Belgian spatial administrative divisions FLANDERS Provinces (a) 1.Antwerp

Districts (b) 11.Antwerp 12.Mechelen 13.Turnhout

2.Flemish Brabant

23.Hal-Vilvorde 24.Leuven

3.West Flanders

31.Brugge 32.Diksmuide 33.Ieper 34.Kortrijk 35.Oostende 36.Roeselars 37.Tielt 38.Veurne

4.East Flanders

41.Aalst 42.Dendermonde 43.Eeklo 44.Gent 45.Oudenaarde 46.Sint-Niklaas

7.Limbourg

71.Hasselt 72.Maaseik 73.Tongeren

(a) NUTS 2; and (b) NUTS 3.

Provinces 2.Walloon Brabant

WALLONIA Districts 25.Nivelles

5.Hainaut

51.Ath 52.Charleroi 53.Mons 54.Mouscron 55.Soignies 56.Thuin 57.Tournai

6.Liège

61.Huy 62.Liège 63.Verviers 64.Waremme

8.Luxembourg

81.Arlon 82.Bastogne 83.Marche-en-Famenne 84.Neufchateau 85.Virton

9.Namur

91.Dinant 92.Namur 93.Philippeville

Regional Employment Disparities in Belgium 105

Fig. 6.13. Administrative regions and districts of Belgium

Appendix 2 Table 6.5. Definition of industries 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22

Agriculture and Forestry Fishing Mining of coal and lignite and extraction of crude petroleum and natural gas Mining of uranium, of metal ores and other mining Beverages, food and tobacco products Textiles Leather products Manufacture of wood Paper products and publishing, printing, and reproduction of recorded media Coke, refined petroleum products and nuclear fuel Chemical, rubber and plastic products Other non-metallic mineral products Basic metals and manufactured of fabricated metal products Machinery and equipment and office machinery and computers Electrical apparatus and radio, television, communication and medical optical instruments, watches, clocks Transport equipment Office, shop and other furniture Electricity, gas and water supply Construction Motor and cycle vehicles; retail sale of automotive fuel Wholesale and commission trade Retail trade

23 24 25

Hotels Land transport and via pipelines Water transport

26 27 28 29 30 31

Air transport Other transport activities Post and telecommunications Financial intermediation Insurance and pension funding, except compulsory social security Activities auxiliary to financial intermediation

32 33 34 35

Real estate activities Renting of machinery and equipment Computer Research and development

36

Other business activities

37

Human health and hospital activities

38 39

Sanitation and similar activities Activities of membership organisation and recreational, cultural and sporting and other service activities Domestic service Undefined activities Diplomatic representation, international organisations and allied forces [Administration of the State] [Education]

40 41 42

Regrouping by the authors on the basis of the NACE nomenclature.

106 Olivier Meunier and Michel Mignolet

Acknowledgements We would like to thank the members of the Service des Etudes et de la Statistique of the Ministry of the Walloon Region for their valuable comments. This research was carried out thanks to the financial support from the Ministry of the Wallon Region. Further details of this research may be found in Meunier and Mignolet (2003). The authors are also indebted to Eric Toulemonde, Daniel Felsenstein and Boris Portnov for their useful comments and to Eric Pynnaert for his literature survey. Of course, any errors are ours.

References Barff R, Knight PL (1988) Dynamic shift-share analysis. Growth and Change 19(2):1-10 Beaud M (1966) Analyse régionale-structurale et planification régionale. Revue Economique 1:1-10 Berzeg K (1978) The empirical content of shift-share analysis. Journal of Regional Science 18(3):463-469 Berzeg K (1984) A note to statistical approaches to shift-share analysis. Journal of Regional Science 24(2):277-285 Blien U, Wolf K (2002) Regional development of employment in Eastern Germany: an analysis with an econometric analogue to shift-share technique. Papers in Regional Science 81:391-414 Buck TW (1970) Shift and share analysis - a guide to regional policy?. Regional Studies 23(1):43-48 De Brabander GL (1975) The traditional and the Esteban-Marquillas shift-share models, a comparison with an application to Belgium (1846-1896). Workshop on Quantitative Economic History Discussion Paper, Katholieke Universiteit Leven Dinc M (2002) Regional and local analysis tools. Mimeo, World Bank Institute, Washington Dormard S (1999) Evaluation de l’emploi, gain de productivité et spécialisation régionale: développements et application de la méthode structurelle-résiduelle. Paper presented at the 35th ASRDLF Congress, Hyères, September 1st-3rd Estevão M (2003) Regional labor market disparities in Belgium. Reflets et Perspectives 42(1):95-114 Fernández MM, Menéndez AJL (2002), The evolution of the employment in the European Union. A stochastic shift and share approach. Paper presented at the European Regional Science Association Congress, Dortmund, August 27th-31st Guesnier B (1998) La dynamique spatiale du système productif. In: Louinger G, Némery JC (eds) Recomposition et développement des territoires. L’Harmattan, Paris Jayet H (1981) L’analyse de variance et l’extension de la méthode shift-share. Revue d’Economie Régionale et Urbaine 4:505-515 Jayet H (1993) Analyse spatiale quantitative. Une introduction. Economica, Paris Marimon R, Zilibotti F (1998) ‘Actual’ versus ‘virtual’ employment in Europe - is Spain different?. European Economic Review 42:123-153 Meunier O, Mignolet M (2003) L’emploi en Belgique et ses disparités régionales. Discussion Paper, Observatoire de l’emploi, to be published

Regional Employment Disparities in Belgium 107 Patterson MG (1991) A note on the formulation of the full-analogue regression model of the shift-share method. Journal of Regional Science 31(2):211-216 Perloff HS, Dunn ES, Lampard EE, Muth RF (1960) Regions, resources and economic growth. John Hopkins Press, Baltimore Stevens BH, Moore CL (1980) A critical review on the literature on shift-share as a forecasting technique. Journal of Regional Science 20(4):419-437 Stilwell FJB (1970) Further thoughts on the shift and share approach. Regional Studies 4:451-458 Toulemonde E (2001) ‘Actual’ versus ‘Virtual’ employment in Belgium. Regional Studies 35(6):513-518

7

Regional Income Convergence and Inequality in Boom and Bust: Results from Micro Data in Finland 1971-2000

Heikki A. Loikkanen1, Marja Riihelä2 and Risto Sullström2 1 2

Department of Geography, University of Helsinki, Finland Government Institute for Economic Research (VATT), Helsinki, Finland

7.1

Introduction

The purpose of this chapter is to analyse what happened to income differences between and within four main Finnish regions: The Helsinki Region, Southern, Middle and Northern Finland. We have results from 1971 until the turn of the millennium, but our main focus is on developments since the mid 1980s. At the beginning of this period the old institutional framework was still in operation including fixed exchange rate policy, capital import controls, interest rate regulation and a narrow-based tax system with quite a few deductions. Regional policies at large aimed at even development throughout the country. Urban growth problems were not emphasized although Finland was internationally lagging behind in the rate of urbanization relative to GDP per capita. Increase in mobility to urban centres when economic growth accelerated always seemed to come as a surprise. Immigration from abroad to Finland was marginal. This institutional framework in Finland changed in the late 1980s with financial liberalization, collapse of fixed exchange rate policies, EU and EMU membership. There were also tax and grant (to municipalities) reforms and deregulation of some markets. Besides institutional changes, the period since the mid 1980s is also unique in the economic history of Finland due to its volatility. A boom was followed by an economic disaster in the early 1990s, when GDP declined cumulatively by more than 10 percent in three years. In the mid 1990s a new, IT industry based growth phase started reshaping the economy and its regional structure. This chapter first briefly describes how Finland in the late 1980s experienced a boom and then came “Down from the heavens and up from the ashes” (see Kalela et al. 2001 for an extensive multidisciplinary analysis). Thereafter we shall concentrate on our main topic, namely what happened to regional income disparities (convergence or divergence) and inequality. In the economics literature analyses of income convergence and inequality are mainly separate topics. The former typically use national or regional macro aggregates (per capita GDP or the like) whereas the latter are based on micro data (household income, consumption or the like). These topics are related in studies,

110 Heikki A. Loikkanen, Marja Riihelä and Risto Sullström

which explain economic growth, among other things, by measures of income inequality. In our project we utilize micro data to study both regional convergence and inequality. We are especially interested in households’ situation. Household Survey (HS) data, available with about five year intervals, are used to give a long run view of income differences between and within regions in Finland during 19711985. As a second data source, annual income distribution statistics (IDS) are used to get a more detailed picture of the years 1990-2000. We apply three income concepts: factor income (i.e. wage, entrepreneur and capital income), gross income (i.e. before direct taxes) and disposable income (i.e. after direct taxes). Thus, we can get an idea how the mechanisms of the welfare state affect regional disparities and inequality. Our paper is organized as follows. Section 7.2 gives a summary of macroeconomic developments and policies since mid 1980s. With this background, Section 7.3 concentrates on what happened to income differences between and within regions since early 1970s. As for regional disparities, we consider how per capita incomes have evolved regionally relative to the respective national averages since early 1970s. Then, Gini coefficients are used to study income inequality nationally and within regions. We also present some results on the number of poor and rich. To find out what factors have affected changes in inequality, we decompose aggregate national and regional changes into components in order to examine the role of changes in demography, different forms of market income, and the mechanisms of the welfare state (taxation and transfers). A summary in Section 7.4 concludes the paper.

7.2

Economic Developments and Policies Since the Late 1980s

Since World War II Finland has been catching-up the average per capita income level in West-European countries. GDP per capita has doubled during the last thirty years and reached the level of Sweden for the first time at the turn of the millennium. Although economic growth has been fast on average, it has by no means been steady. On the contrary, Finland has been a rather volatile country from a West-European perspective, mainly due to being driven by the world market demand for forest sector products. In this respect, the 1980s were special, because during the first half of the decade it looked like the business cycle had been tamed. The growth rate of GDP was between 2 and 4 percent. Thereafter, macroeconomic stability problems became especially acute again. After the mid 1980s unexpectedly favourable international economic developments, improving terms of trade, and especially the most important phases of financial deregulation led to a boom with high GDP growth (Figure 7.1). Abolishment of interest rate regulation and capital import controls in 1986 ended extensive credit rationing and led to a credit expansion. It was predominantly based on an inflow of foreign capital to the economy. After deregulation, interest

Regional Income Convergence and Inequality in Boom and Bust 111

rates remained for a while at earlier levels and did not constrain borrowing in the first phase. Later, hikes widened the interest rate differential to the German Mark increasing importation of capital on the assumption that Finland would stick to the fixed exchange rate. As a matter of fact the Finnish Markka was revalued in 1988. Due to credit expansion asset prices increased substantially. The stock market boomed and especially prices of owner-occupied housing sky rocketed relative to normal times. 20 15

GDP UE PSS

%

10 5 0 -5 -10 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

Fig. 7.1. GDP growth rate, unemployment rate (UE) and public sector surplus (or deficit) as a percent of GDP (PSS) in Finland, 1988-2001 Source: Statistics Finland

Also the turn from the boom to a bust in early 1990s was unprecedented in Finland, more severe than the “great” depression in the 1930s. The declared policy of sticking to a fixed exchange rate, when current account deficits accumulated fast, led to a loss of confidence when export demand in Western markets decreased and barter trade with the former Soviet Union collapsed. The consequent currency and bank crises, together with pursued economic policies led to a cumulative decline of the real GDP of more than 10 percent in 1991-1993. The unemployment rate, which had been 3.2 percent in 1990, reached 16.6 percent in 1994. The bust increased public expenditure at the same time as tax revenues decreased. Public sector deficits led public debt to increase from about 15 per of GDP in 1990 to more than 60 percent level in mid 1990s. The share of public expenditure in GDP reached 62 percent in mid 1990s. When the boom turned to a depression in 1991 the fixed exchange rate policy lost credibility. After an unsuccessful defence with high interest rates, the Finnish Markka was devalued by 12 percent in November 1991, and in September 1992 the Markka was floated. In 1995 Finland became an EU member, in October 1996 Finland joined the ERM, and in 1999 the EMU.

112 Heikki A. Loikkanen, Marja Riihelä and Risto Sullström

The credit expansion of late 1980s, financed to a great extent by foreign borrowing, lead to a banking crisis when interest rates increased. The Markka was devalued and asset prices collapsed. Prices of owner-occupied dwellings decreased almost to half of their top level. The economic crisis resulted in banks with non-performing loans and missing collateral values and construction companies with unsold dwellings. A record high number of households suffered from unemployment and related income problems. Quite a few of them were unable to manage with their housing expenses, especially if their last dwelling was bought at a high price with credit in the boom, which soon collapsed. As for banks, some went bankrupt and others merged with domestic or foreign banks. Heavy state subsidies were used to save the banking sector. Bankruptcies and mergers also occurred in the construction sector. No major bank or construction company survived the turmoil of 1990s without some form of restructuring. To save the banking sector, it has been estimated that the government gave bank subsidies worth in all 7.5 percent of 1992 GDP (Nyberg and Vihriälä 1994). It has become customary to explain and summarize the developments which led to the boom of late 1980s and the economic crisis that followed referring to three ‘bads’: bad banking, bad luck and bad policies. The first refers to poorly designed financial deregulation without sufficient reforms in the banking sector and taxation. Bad luck refers to business cycle factors in Western markets and the sudden fall of East trade following the collapse of the Soviet Union. The third is related to the combination and timing of financial liberalization, exchange rate policies and fiscal policies. They first added fuel to an overheating economy at the end of 1980s, and later, when the bust started, had a neutral or in some years even a contractive effect on aggregate demand. A more thorough economic analysis of these developments can be found in Honkapohja and Koskela (1999) and in Kalela et al. (2001). The latter contains articles considering this period from a multitude of viewpoints. In the midst of the crisis, the government made quite a few reforms in 1993. There was a tax reform, which included adoption of a dual households’ income tax system where labour income is taxed on a progressive scale and capital income at a flat rate. Tax base was broadened by eliminating or limiting deductions. A modest local property tax was introduced and at the same time taxation of imputed income from owner-occupied housing was abolished. Also corporate income taxation was renewed such that statutory tax rates declined but tax base increased. Furthermore, two-stage taxation of capital income from corporate sector was eliminated. These reforms took place at a time when the government had difficulties in financing public expenditure. Declining tax revenues were supplemented by fast increasing public debt. Furthermore, the system of state grants to municipalities was reformed from a matching grants system to a lumpsum type system. Rent controls in the private rental sector were abolished, first from new contracts and by 1995 from all contracts. From 1995, economic growth was exceptionally fast until the last downturn of the economy in 2001. During the last growth phase the structure of the economy changed. IT industries, led by the success of NOKIA, became the fastest growing

Regional Income Convergence and Inequality in Boom and Bust 113

sectors and the stock prices of related firms boomed until the recent downturn. This period generated wealth and an exceptional flow of capital income to those who managed to benefit from the boom. Alongside those enjoying the success of the new economy, there now exists a huge amount of unemployed people. Unemployment is still around 9 percent, about three times higher than in the late 1980s. It has remained at high level (Figure 7.1) not only in areas with continuing economic problems (negative employment growth and net migration), but also in urban growth areas, which are undergoing structural change. After the above description of Finland's economic growth and especially the booms and bust after late 1980s, we shall concentrate on studying what happened to regional disparities and inequality. Were there major changes in the relative standard of living of people in the North relative to people in the Helsinki Region or in Southern and Middle Finland? Especially, did the bust of early 1990s lead to increasing disparities? On the other hand, what happened in the boom of late 1990s when economic growth was regionally less evenly distributed than earlier? During this phase the engine of growth was the IT sector rather than traditional (paper, pulp and metal) industries. Only half a dozen urban areas (Helsinki Region, Tampere, Turku and Salo Regions in Southern Finland, Jyväskylä Region in Middle Finland, and Oulu Region in Northern Finland; see Figure 7.2) attracted new investment and gained from net migration.

7.3

Studying Regional Income Differences and Inequality with Micro Data

Most analyses of convergence which have appeared since Barro and Sala-i-Martin (1992) typically use national or regional aggregates (per capita GDP or the like). Using econometric techniques they test whether there has been Beta or Sigma convergence. On the other hand, there is a vast literature on the distribution of income - thousands of entries in the EconLit database (Atkinson 1997). These studies on inequality are based on micro data. We shall use micro data in studying both regional convergence and inequality as our interest is in people's disposable incomes. Instead of using econometric methods, with only four regions, we simply present graphs of regional income developments in our convergence analysis. In our analysis of inequality, we use Gini coefficients and some other measures. Also, results on the number of poor and rich are presented. Furthermore, we decompose aggregate national and regional changes in inequality into components in order to find out the role of changes in demography, income types and the mechanisms of the welfare state (taxation and transfers).

114 Heikki A. Loikkanen, Marja Riihelä and Risto Sullström

1. 2. 3. 4.

Helsinki Region Southern Finland Middle Finland Northern Finland

Share of Number of Population Municipalities (in 2000) 18.4 4 44.1 216 25.2 158 12.3 74

Fig. 7.2. Main regions in Finland1

1

The number of inhabitants in Finland was 5.1 millions in 2000 and its total area 338145 km2. Geographically the extreme (mainland) point in north is Nuorgam 70o5'30'' and in south Hanko 59o48'30''. Their distance is 1157 km. Land boundary with Sweden (west) is 586 km, with Norway (north) 727 km and with Russia (east) 1269 km.

Regional Income Convergence and Inequality in Boom and Bust 115

7.3.1

Data Description and Definitions

Our long-run view of income differences between and within the main regions in Finland during 1971-1990 is based on time series data of Household Surveys (HS) from the years 1971, 1976, 1981 and 1985. A corresponding, but more detailed analysis of 1990s is done with annual Income Distribution Statistics (IDS) data for the period 1990-2000. In both data sources, information on different types of income makes it possible to calculate factor incomes for each household, then add transfers and get gross income, and subtract direct taxes to get disposable income. Thus, we can study regional convergence and inequality with all three income concepts. Our data are samples, which do not allow very detailed regional classifications. Here, we divide Finland into four main regions: the Helsinki (Capital City) region, Southern Finland, Middle Finland and Northern Finland. The sample size (households) in the time series data of the Household Survey is about 3000 in 1971 and 1976 and about 8000 in 1981 and 1985. The samples of annual IDS are large - ranging from about 9000 to 12000. Both have enough observations for our regional analyses. In studying regional income differences and inequality we apply the income per consumption unit metric by using the OECD equivalence scale. Despite this choice, we shall refer to per capita incomes in the text. Note that we have not deflated incomes by regional price indices, which deviate mainly because of housing price and rent differences. 7.3.2

On Income Differences Between Regions

At the national level and irrespective of income concept used, real income levels increased substantially until the year 1990. Thereafter, the depth of the depression of early 1990s is very clear. For the first time during the post World War II period in Finland real incomes per capita decreased. Figure 7.3 illustrates that in all main regions real disposable income per capita increased steadily during 1971-1990. When the economic crisis led to declining GDP during 1991-1993, real incomes did not decrease immediately, except in the Helsinki Region where the decline started already in 1991. Also, the drop in income level lasted longer in the Capital City Region than elsewhere but when the turn up finally took place in 1997, income growth in the Helsinki Region was faster than elsewhere. Next, we shall consider the evolution of regional per capita incomes relative to the national average (= 100) applying two income concepts: factor income and disposable income. Figure 7.4 indicates that relative income differences based on disposable income are somewhat smaller than those based on factor income. Thus, the mechanisms of the welfare state (direct taxes and transfers) decrease regional disparities. Furthermore, for both income concepts, there has been substantial regional convergence of relative income levels over time. Convergence was especially clear from 1971 to 1980. In the Helsinki Region relative income level declined, whereas the opposite tendency was prevalent in

116 Heikki A. Loikkanen, Marja Riihelä and Risto Sullström

Middle and Northern Finland. The relative position of Southern Finland has remained much the same. During 1980s relative income differences were much the same or even slightly increased on the basis of three observations from 1980, 1985 and 1990. 15000 13500 12000

Helsinki Region Southern Finland Middle Finland Northern Finland

Euro

10500 9000 7500 6000 4500 3000 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 Years

Fig. 7.3. Real per capita disposable income in main regions in 1971-2000

During 1990-2000 annual data indicate that relative income differences between main regions remained relatively stable. In the three year period of falling GDP, only in the final year (1993) did regional disparities in terms of relative factor and disposable incomes increase temporarily. After slight convergence in 1994-1996, the last four years indicate some divergence, mainly because of the improved position of the Helsinki Region. In the year 2000 relative regional disparities were somewhat greater than they were in 1990. Our findings thus suggest that there has been a long-run tendency for factor and disposable incomes to converge, especially before the 1980s. During the booms and bust of 1990s relative regional disparities did not change remarkably. To the extend that there were changes, the growth of relative disparities occurred in booms rather than in busts, and this pattern was mainly related to the position of the Helsinki Region. 7.3.3

Distribution of Income and Poverty Within Regions

International comparisons of income inequality before the 1990s indicate that Finland did not differ very much from other countries when the distribution of factor income is considered. However, together with Sweden, Finland had the most even distribution of disposable income (see e.g. Atkinson et al. 1995). Here, in addition to describing national developments, we also have results on inequality for main regions in Finland. We measure inequality by Gini coefficients based on per capita incomes (OECD equivalence scale).

Regional Income Convergence and Inequality in Boom and Bust 117 Factor income 180

Index (Finland=100)

160 140 120 100 80 60 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000

Disposable income

180

Index (Finland=100)

160 140 120 100 80 60 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 Helsinki Region Northern Finland

Southern Finland Finland (=100)

Middle Finland

Fig. 7.4. Per capita factor income and disposable income in main regions in 1971-2000, index (Finland = 100)

Figure 7.5 describes national developments of Gini coefficients. Our first observation is that direct taxes and transfers decrease inequality as Gini coefficients decline, when we move from factor income to gross income, and to disposable income. As for developments over time, after a slight decline, inequality based on factor income increased since mid-1970s. During the 1990s the sharpest increase took place during the deepest recession in 1992-1993, after which the growth of the Gini coefficient has continued at a slower rate.

118 Heikki A. Loikkanen, Marja Riihelä and Risto Sullström

50

Gini coefficient (%)

45

Factor income Gross income Disposable income

40 35 30 25 20 15 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000

Fig. 7.5. Gini coefficients based on three income concepts, 1971-2000

Gini coefficients based on gross income and disposable income both decreased from 1971 to 1976, and thereafter there were no major changes in their level. This is surprisingly also true for the deepest recession years 1991-1993 suggesting that the economic crisis left income distribution relatively unaffected. However, as economic growth began to speed up after 1994, inequality also began to increase. The general level and development of regional Gini coefficients is remarkably similar to the national Gini coefficients during 1971-1990 (Figure 7.6). The most noticeable regional differences are related to factor income Gini coefficients, whereas e.g. in 1985 and 1990 Gini coefficients for disposable income are almost the same in all regions. During the 1990s these trends change. Disparities in regional inequality based on disposable and gross income per capita increase since 1992. Especially, Gini coefficients in the Helsinki Region began to grow faster than those of the other regions. Somewhat surprisingly, regional disparities in factor income, after increasing in early 1990s, have remained roughly at the same level during the latter half of 1990s. Consideration of poverty supplements our analysis of inequality. The number of poor people (disposable income per capita below half of respective national median income) in the Helsinki Region was low during 1971-1990 (Figure 7.7) whereas in all other regions it was initially much higher and decreased over time.2 During the crisis years of early 1990s only in the Helsinki Region did the number of poor increase and this trend continued until 1998. In other regions, the number of poor only began to increase in 1994 when the economy began to grow fast.3 2

3

Note that we have not taken into account regional price differences, especially high housing costs in the Helsinki Region, which partly explains why the number of poor is low in this area. We have also calculated the number of poor by using the 1990 real income level of the national poverty line in considering later years. Using this criterion, the number of poor

Regional Income Convergence and Inequality in Boom and Bust 119 Factor income

Gini coefficient (%)

50 45 40 35 30 25 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 Gross income 38 Gini coefficient (%)

36 34 32 30 28 26 24 22 20 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000

Gini coefficient (%)

Disposable income 32 30 28 26 24 22 20 18 16 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 Helsinki Region Northern Finland

Southern Finland Finland

Middle Finland

Fig. 7.6. Gini coefficients by income variables and main regions in 1971-2000

increases during 1992-1996, whereas in 1997-1998 it decreases, but still in 1998, there are more poor than in 1990 (see Riihelä and Sullström 2001 p. 63).

120 Heikki A. Loikkanen, Marja Riihelä and Risto Sullström

200000

Helsinki Region Southern Finland Middle Finland Northern Finland

180000

Number of poor peopl

160000 140000 120000 100000 80000 60000 40000 20000

0 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000

Fig. 7.7. The poor population by region 1971-2000

When similarly observing the number of rich people by region (with disposable income per capita twice the national median income), the results of Table 7.1 indicate that their number and population share decreased from 1971 to 1990. Relative to the 1990 figures, in the year 2000 the number and share of rich people roughly doubled at the national level and in Southern and Middle Finland, whereas in the Helsinki Region and Northern Finland the increase was, especially in relative terms, slightly smaller. Table 7.1. The rich population by region in 1971, 1990 and 2000 Year/ Region 1971 1990 2000

7.3.4

Total % Total % Total %

Helsinki Region 133,000 21.1 65,000 7.8 116,000 12.3

Southern Finland 103,000 5.4 56,000 2.6 113,000 5.0

Middle Finland 51,000 3.7 23,000 1.8 46,000 3.6

Northern Finland 25,000 4.4 13,000 2.0 20,000 3.2

Finland 312,000 7.0 157,000 3.2 295,000 5.8

Decomposition of Inequality

In the previous section, we presented Gini coefficients for each region and the whole country. To find out which factors have caused change in total inequality of disposable income, we decompose the Gini coefficients into the contributions of demographic changes and changes in types of income. Methodologically our analysis is related to studies by Jenkins (1995), Aaberge et al. (2000) and Suoniemi (1999), who decompose either changes in Gini coefficients or generalized entropy measures. From the entropy family measures we use mean logarithmic deviation (I0) and squared coefficient of variation (I2) to

Regional Income Convergence and Inequality in Boom and Bust 121

decompose inequality to between and within region components, and I0 to decompose changes of inequality over time to the contributions of regional populations and relative incomes. To conclude this section, we summarize results based on decomposing changes in regional Gini coefficient by type of income. We use four income sources and direct taxes, which add up to disposable income. First, in Table 7.2 we consider the contributions of inequality within and between regions to aggregate inequality with our measures I0 and I2. Both measures indicate that total inequality was high in 1971, clearly lower in 1981 and 1990, grew slightly between 1990 and 1993, and increased considerably between 1993 and 2000. Most variation of disposable income is within main regions. The share of between regions variation was 12-13 percent in 1971, decreased to 5-8 percent in 1981, 1990 and 1993, and still more in 2000. Since 1981 all the changes of aggregate inequality are due to the within regions components. Table 7.2. Within group and between group inequality in 1971-2000, in percent

Year 1971 1981 1990 1993 1996 2000

Mean log deviation Within

Between

Total

11.1 7.1 6.4 6.9 7.9 11.8

1.6 0.5 0.5 0.5 0.3 0.8

12.7 7.6 6.9 7.5 8.2 12.5

Squared coefficient of variation

Between/ Within Total

12.5 6.3 7.6 7.3 3.5 6.1

12.2 6.9 7.4 10.4 10.7 70.9

Between

Total

Between/ Total

1.8 0.5 0.6 0.6 0.3 0.8

14.1 7.5 8.0 11.0 11.0 71.8

13.0 7.0 7.3 5.4 2.7 1.2

Gini 27.0 20.7 20.4 20.9 22.2 26.6

The above decomposition does not tell to what extent changes in inequality are due to changes in regional population shares, income levels and internal inequality. Following Mookherjee and Shorrocks (1982), a change in inequality measured by ' I0 can be approximately decomposed to four additive terms, where A is a measure of changes in inequality within each group (pure change in inequality), whereas terms B and C indicate the impact on the within groups and between groups components of inequality which result from changes in regional populations, and D gives the effect of changes in relative incomes of regions. In Table 7.3 results of this decomposition of change in inequality (in percent) are reported for three periods: (a) 1971-1981 (clear decrease in inequality), (b) 1981-1990 (slight decrease in inequality), (c) 1990-2000 (clear increase in inequality). They show that the pure effect A dominates and all other effects are small except term D during 1971-1981, when regional (relative) income changes decreased inequality. Changes in total and regional population are slow and do not help much in understanding changes in inequality especially in the 1990s. Changes in sources of income and taxation have changed much more following booms and busts of the economy, changes in unemployment, and reforms in tax and transfer systems. To study their role we first note that disposable income can be defined as a sum of four income sources minus transfers paid. The income items consist of wages, entrepreneurial income, capital income, transfers received, including separately

122 Heikki A. Loikkanen, Marja Riihelä and Risto Sullström

unemployment benefits. During 1971-2000 the share of wage and entrepreneurial income has had a decreasing trend. Capital income’s share surprisingly almost doubled from 1990 (6.6 percent) to 1993 (12.3 percent) and also increased thereafter. The share of unemployment benefits in the early 1990s increased to a still higher level. Table 7.3. Sub-group decompositions of the changes in disposable income inequality Period

' I0

% change in ' I0 accounted for by changes in Withingroup inequalities

(a) 1971-1981 (b) 1981-1990 (c) 1990-2000 (d) 1990-1993 (e) 1993-1996 (f) 1996-2000 (g) 1971-2000

-39.9 -8.8 80.1 7.7 18.3 52.2 -1.3

Population shares

Sub-group mean incomes

(term A)

Within groups (term B)

Between groups (term C)

(term D)

-31.0 -9.5 75.6 7.5 21.6 45.7 4.0

-0.2 0.0 1.2 0.1 0.2 0.7 1.2

0.6 0.1 0.5 0.2 0.1 0.2 1.4

-9.2 0.5 2.8 0.1 -3.6 5.7 -7.7

In Table 7.4 we present results of a decomposition of squared coefficient of variation by main regions and income sources (see Nygård and Sandström 1981). They indicate that the roles of income sources for inequality have changed over time. The positive contribution of wage income to aggregate (disposable) income variation has decreased since 1981. The contribution of entrepreneurial income increased until 1990, but has since decreased. The most noticeable change concerns capital income. Its contribution has increased considerably since 1990, and most of the change comes from the Helsinki Region. Direct taxes have decreased variation of disposable income and this effect peaked in 1990. Thereafter, the negative contribution has decreased to some extent. To further investigate the role of different sources of income for inequality, we have decomposed regional Gini coefficients by income type and calculated elasticities, which indicate how much a percentage change in each type of income affects total inequality during 1990-2000 (Lerman and Yitzhaki 1985). The results in Figure 7.8 indicate that marginal wage increases (evaluated at mean value) increase total inequality, but this effect has decreased over time. The opposite is true for capital income: the contribution of its change has increased, and it is clearly highest in the Helsinki Region. Increases in unemployment benefits and transfers decrease inequality and these effects were greatest at the end of the bust in 1994. An increase in taxes (transfers paid) decrease inequality and this effect is very similar across regions and years during 1990s.

Regional Income Convergence and Inequality in Boom and Bust 123

Transfers received

Transfers paid

Regional contribution

56.8

2.4

5.1

0.0

9.5

-18.9

54.9

Southern Finland

36.5

7.1

2.5

-0.1

1.9

-14.6

33.4

Middle Finland

11.3

1.2

1.6

-0.1

-1.8

-4.3

8.0

6.8

-0.4

0.1

-0.0

-0.7

-2.1

3.7

111.4

10.3

9.3

-0.1

8.9

-39.9 100.0

61.3

3.2

2.1

0.0

6.9

-24.5

49.0

49.0

7.5

2.4

-0.3

-2.4

-19.7

36.8

1971 Helsinki Region

Northern Finland Income source contribution

1981 Helsinki Region Southern Finland Middle Finland Northern Finland Income source contribution

1990 Helsinki Region Southern Finland Middle Finland Northern Finland Income source contribution

1993 Helsinki Region Southern Finland Middle Finland Northern Finland Income source contribution

1996 Helsinki Region Southern Finland Middle Finland Northern Finland Income source contribution

2000 Helsinki Region

Capital income

Unemployment benefits

Region

Wage income

Year

Entrepreneurial income

Table 7.4. Decomposition of the squared coefficient of variation (I2) by main region and income source in selected years

15.4

6.5

1.2

-0.3

-3.2

-10.2

9.7

8.7

1.7

1.1

-0.2

-2.0

-5.1

4.4

134.4

18.9

6.8

-0.9

-0.5

64.3

5.5

9.5

0.1

8.1

-32.9

54.4

40.9

11.8

7.7

-0.3

-1.2

-24.0

35.2

10.9

6.1

1.1

-0.4

-4.6

-7.3

6.2

7.3

2.6

0.9

-0.3

-1.9

-4.7

4.2

123.4

26.0

19.0

-0.9

0.5

40.8

3.7

19.9

0.0

13.4

-26.9

50.8

37.2

4.7

22.5

-1.6

0.0

-24.5

39.9

6.8

2.8

3.2

-1.8

-2.7

-5.5

4.6

-3.2

4.7

-59.6 100.0

-68.9 100.0

3.7

5.1

0.8

-0.8

-1.8

88.5

16.3

46.4

-4.3

8.8

-60.1 100.0

37.3

2.7

11.2

-0.4

7.6

-23.8

35.0

41.1

10.1

30.8

-1.8

2.9

-31.7

53.1

9.3

3.5

5.5

-1.5

-3.0

-7.2

8.1

-3.6

3.8

5.5

1.2

2.1

-1.0

-1.3

93.2

17.4

49.6

-4.7

6.1

-66.3 100.0

19.5

0.6

93.0

0.0

1.7

-42.4

72.4

Southern Finland

5.7

2.8

27.1

-0.2

-0.7

-11.6

23.3

Middle Finland

1.5

1.7

3.4

-0.2

-0.5

-2.2

3.9

0.6

0.2

0.5

-0.1

-0.3

-0.5

0.5

5.3 124.0

-0.5

0.3

Northern Finland Income source contribution

27.2

-56.7 100.0

124 Heikki A. Loikkanen, Marja Riihelä and Risto Sullström Wage income

Entrepreneur income

0,70 0,25 0,60 0,20

0,50 0,40

0,15

0,30

0,10

0,20 0,05

0,10 0,00 1990

1992

1994

1996

1998

2000

0,00 1990

1992

Capital income

1994

1996

1998

2000

1998

2000

1998

2000

Transfers received

0,45

1990 0,00

0,40

1992

1994

1996

-0,05

0,35

-0,10

0,30

-0,15

0,25

-0,20 0,20

-0,25

0,15

-0,30

0,10

-0,35

0,05

-0,40

0,00 1990

-0,45 1992

1994

1996

1998

2000

-0,50

Transfers paid

Unemployment benefits 1990 0,00

1992

1994

1996

1998

2000

1990 0,00

1992

1994

1996

-0,05 -0,05

-0,10 -0,15

-0,10

-0,20 -0,25

-0,15

-0,30 -0,20

-0,25

-0,35 Helsinki Region Southern Finland Middle Finland Northern Finland Finland

-0,40

Fig. 7.8. Gini elasticities of income components by main regions in 1990s

The most visible indicator of economic crisis in 1990s was the increase in unemployment, which took place during 1990-1994 in all regions. Surprisingly, during this phase inequality hardly increased, except in the Helsinki Region where the subsequent increase in Gini coefficient began in 1993 (Figure 7.9). Other regions followed with a lag.

Regional Income Convergence and Inequality in Boom and Bust 125

25

1996

1994

1999

Unemployment rate (%)

2000

20

1998 1997

15 2000

10

2000

1992

1991

2000

1998 1999

2000

1991

5

1990

0 15

16

17

18

19

20

21

22

23

24

25

26

27

28

29

30

31

32

33

Gini coefficient (%)

Helsinki Region Northern Finland

Southern Finland Finland

Middle Finland

Fig. 7.9. Trade off between unemployment rate and Gini coefficient by main regions

7.4

Concluding Comments

In this study, we have used micro (Household Survey and Income Distribution Statistics) data to study income differences both between and within main regions in Finland during 1971-2000. The former is related to the question of regional income convergence versus divergence and the latter to inequality at national level and within regions. Regionally, we are interested in differences between the Helsinki Region, Southern, Middle and Northern Finland. Besides long-term developments, we are especially interested in what happened during the boombust-boom period that began in the late 1980s. It included besides economic turmoil, also institutional changes ranging from financial liberalization, tax reforms to EU and EMU membership. During the last years considered, the structure of the economy also changed with the rise of IT industry. As for regional disparities, regional GDP as well as micro data based (factor, gross and disposable) income measures indicate that there has been convergence among Finnish regions over time with more occurring early in the period and less when we come closer to the end of the 1980s. At this level of regional aggregation, regional disparities remained more or less the same in the early 1990s, and increased only slightly thereafter. As for inequality, the greatest decline in (disposable) income differences since 1971 took place before mid-1980s, thereafter inequality remained pretty much at the same level. Somewhat surprisingly, this was also the case in the early 1990s, when output dropped and unemployment increased dramatically. When the recovery began in the mid-1990s inequality began to increase rapidly. These

126 Heikki A. Loikkanen, Marja Riihelä and Risto Sullström

developments were surprisingly similar both at the national level and at the level of main regions thinking of both time profiles and levels of most typical inequality indicators (Gini coefficients, generalised entropy measures). Only in the recovery period can we detect clear regional divergence with inequality which increased earlier and to greater extent in the Helsinki Region than elsewhere. This was very much due to capital income which increased both as a source of income and as a contributor to inequality especially in the Helsinki Region, a fact which is related to the rise of IT-industry since the mid-1990s. Our results indicate that the mechanisms of the Welfare State, namely transfers and taxation, decrease regional disparities and inequality. It is surprising that the joint effect of economic crisis and these mechanisms had almost no visible effect on our (relative) measures of regional income differences and inequality. Due to tax and transfer reforms of 1993, their re-distributive role has somewhat diminished and this partly explains the changes in disparities that have increased in the boom of the late 1990s. We also considered poverty to supplement our analysis of inequality. The number of poor people in the Helsinki Region was low during 1971-1990 whereas in all other regions it was initially much higher and decreased over time. During the crisis years of early 1990s the number of poor increased only in the Helsinki Region, and this trend continued until 1998. In other regions, the number of poor began to increase only in 1994 when the economy began to grow fast. As for the number of rich, their number also grew during 1994-1998, especially in Helsinki Region and in Southern Finland.

References Aaberge R, Björklund A, Jäntti M, Pedersen PJ, Smith N, Wennemo T (2000) Unemployment shocks and income distribution: how did the Nordic countries fare during their crises? Scandinavian Journal of Economics 120(1):77-99 Atkinson AB (1997) Bringing income distribution in from the gold. The Economic Journal 107:297-321 Atkinson AB, Rainwater L, Smeeding TM (1995) Income distribution in OECD countries. Evidence from the Luxembourg income study. OECD, Paris Barro RJ, Sala-i-Martin X (1992) Convergence. Journal of Political Economy 100(2):223251 Honkapohja S, Koskela E (1999) The economic crisis of the 1990s in Finland. Economic Policy 14:399-436 Jenkins SP (1995) Account for inequality trends: decomposition analyses for the UK, 197186. Economica 62:29-63 Kalela J, Kiander J, Kivikuru U, Loikkanen HA, Simpura J (2001) Down from the heavens, up from the ashes. The Finnish economic crisis of the 1990s in the light of economic and social research. VATT Publications 27:6. Government Institute for Economic Research. Helsinki Lerman RI, Yitzhaki S (1985) Income inequality effects by income source: a new approach and application to the United States. The Review of Economics and Statistics 67:151156

Regional Income Convergence and Inequality in Boom and Bust 127 Mookherjee D, Shorrocks A (1982) A decomposition analysis of the trend in UK income inequality. The Economic Journal 92:886-902 Nygård F, Sandström A (1981) Measuring income inequality. Almqvist Wicksell International, Stockholm Nyberg P, Vihriälä V (1994) The Finnish banking crisis and its handling. Bank of Finland Discussion Paper, 4/94 Riihelä M, Sullström R (2001) Income differences and inequality in major regions in the long run 1971-1998 and especially during 1990s (in Finnish). Government Institute for Economic Research, VATT-Research Reports 80. Suoniemi I (1999) Development of the Finnish income distribution and underlying factors affecting its evolution, 1971-1996 (in Finnish). Labour Institute for Economic Research, Research report 76. Helsinki

8

Regional Disparities in Ireland: The Roles of Demography, Profit Outflows, Productivity, Structural Change and Regional Policy 1960-1996

Eoin O’Leary Department of Economics, University College, Cork, Republic of Ireland

8.1

Introduction

The Republic of Ireland experienced rapid growth and convergence during the 1990s after a prolonged period of disappointing growth and failure to catch-up on other EU member states (Kennedy 2000/01). Compared to other small EU countries, the level of regional disparity within Ireland in the late 1990s is relatively low. In 1997 the ratio of maximum to minimum regional gross value added (GVA) per capita was 1.5, compared to 2.3 for Austria and Belgium, 1.9 for Greece and 1.7 for the Netherlands and Finland (European Commission 2001).1 At 57 persons per square kilometre, population density in Ireland is also very low, being the second lowest after Finland of all the small countries considered in this volume. This chapter presents evidence on the sources of living standards disparities in 7 Regional Authority Areas of Ireland from 1960 to 1996.2 Recent research has shown strong living standards convergence among Irish regions between 1960 and 1979 being replaced by weak divergence between 1979 and 1996 (O’Leary 2001a). The re-emergence of divergence in the 1980s and 1990s was also a feature of other small countries. This chapter investigates the sources of regional performance in Ireland. In doing so it sheds light on the causes of the convergent/divergent processes in small countries and the appropriate policy responses. It begins by outlining data and measurement issues. The roles played by profit outflows, labour productivity and demography in explaining living standards disparities are then investigated. The chapter then proceeds to decompose aggregate productivity into the 3 broad sectors and the contribution from structural change. Attention is then turned to considering the role of regional policy in explaining the changing performance. The final section considers Ireland’s future regional policy dilemma.

1 2

This refers to NUTS 2 regions for EU purposes. See Appendix, Figure 8.1 for map and Table 8.9 for definitions of regions.

130 Eoin O’Leary

8.2

Data and Measurement Issues

Until recently, lack of data has severely constrained analysis of Irish regions. It was only in the 1990s, that official regional GVA data were produced by the Central Statistics Office (CSO). The results presented in this chapter are based on a data set constructed from three sources, Attwood and Geary (1963) for 1960, Henry (1997) for 1979 and the CSO (1998) for 1996. This data set consists of GVA at constant factor cost,3 employment and population levels for NUTS 3 regions. Although there are 8 such regions, the standard practice is to amalgamate Dublin and the Mid-East due to the substantial amount of inter-regional commuting between these two areas (CSO 1998). Data are available for the 3 broad sectors, namely Agriculture, Manufacturing and Services.4 The construction methods used are described in O’Leary (2001a). The complete data set is available from the author. The chief measure of convergence used is the V convergence measure equal to the coefficient of variation, defined as:

V i,t = [Standard Deviation (Xi,j,t) / Mean (Xi,j,t)]

(8.1)

where i,j,t represent the aggregate (or sectors), regions and time respectively and X is living standards or labour productivity. If Vi,t+T < (>) Vi,t, then V convergence (divergence) is present between the t and t+T. The rate of V convergence is the average annual percentage change between Vi,t and Vi,t+T. This measure is a sufficient but not a necessary condition for E convergence. To circumvent this, the chapter follows Boyle and McCarthy (1997), who propose Kendall’s rank concordance measure, known as J convergence, as a supplement to V convergence. This measure tracks the degree of intra-distributional mobility or leap-frogging. This measure is calculated as:

J i,t = [Variance (AR (Xi,j,t+T) + AR (Xi,j,t))] / [Variance (2 u AR (Xi,j,t))]

(8.2)

where AR is the actual rank of X. The closer Ji,t is to zero (one) the greater (lesser) the degree of J convergence. In the absence of V convergence, E convergence may be tested for using this measure.

8.3

Decomposing Living Standards: Profit Outflows, Productivity and Demography

Growth in living standards, usually defined as regional “GVA” or output per capita, may be decomposed into two standard components. These are the growth in productivity, or GVA per worker, and the employment/population ratio. Due to the existence of large outflows from Irish regions, mostly attributable to the 3 4

At 1990 prices. National sectoral deflators are used as no regional deflators exist. See Appendix 1 Table 2 for definitions of sectors.

Regional Disparities in Ireland 131

emergence of profit repatriations by foreign owned firms since the mid-1980s, GVA per capita is widely regarded as a biased measure of living standards. Accordingly, O’Leary (1999) has suggested using regional “GNP” or income per capita, as a better measure of regional living standards. This involves adjusting regional GVA for net factor outflows. For profit outflows, the dominant component, the national profit outflow estimate is distributed using each regions share of foreign owned manufacturing profit, defined as the remainder of net output of manufacturing establishments, data for which are taken from the 1996 industry census (O’Leary 2003). The resulting regional “GNP” estimate gives an approximate estimate of regional income. The growth of this measure is decomposed, following O’Leary (2001a) as: Growth (I/N) = Growth (I/O) + Growth (O/L) + Growth (L/N)

(8.3)

where, I is regional “GNP”, N regional population, O regional GVA and L regional employment. The first term on the right hand side, which is the growth in the ratio of regional income to output, accounts for the changing gap between income and output, mostly due to profit outflows. The second term refers to regional labour productivity growth, while the final term is the growth in the employment/population ratio. This term is influenced by demographic factors including changes in the unemployment rate, the labour force participation rate and the age-dependency rate (inverse). Tables 8.1 and 8.2 present the decompositions for the two periods. Following O’Leary (2001a), counterfactuals are also presented in order to investigate the effects of each of the components in Equation (8.3) on the rate of living standards convergence (divergence) over the period. In order to examine the effect of differences in, for example productivity, on the rate of living standards convergence (divergence), the coefficient of variation of living standards levels is recomputed for each year after removing regional differences in labour productivity. If the resultant rate of convergence (divergence) increases compared to the observed rate, then it can be concluded that productivity has a divergent (convergent) effect. If the resultant rate decreases, then it has a convergent (divergent) effect. The size of the convergent (divergent) effect may then be calculated as the difference between the rate of V convergence (divergence) and the rate of V convergence (divergence) that results after differences in productivity have been removed.5

5

The use of counterfactual experiments in this way is indicative of the proximate causes of the degree of living standards convergence (divergence) observed. Isolation of the effect on the overall degree of convergence (divergence) of regional differences in each of the components of an accounting identity provides only an approximation of the nature and importance of effects. Underlying causal mechanisms are not uncovered by this method. However, the method does have the potential of indicating avenues for future research into ultimate causes. The method is a variant of shift-share analysis which may be used to address similar questions (see for example Toulemonde 2001).

132 Eoin O’Leary Table 8.1. Decomposition of living standards in Irish regions: 1960-79 (percent p.a.)

Border Dublin/Mid-East Midlands Mid-West South-East South-West West

+3.1 +1.3 +2.4 +3.1 +3.1 +2.8 +3.5

Income/ Output Ratio -0.5 -0.5 -0.5 -0.5 -0.5 -0.5 -0.5

State

+2.4

-0.5

Rate of V Convergence

-3.5

0.0

Living Standards

Productivity +4.5 +2.3 +3.7 +4.1 +4.2 +3.7 +4.7

Employment/ Population Ratio -0.9 -0.5 -0.8 -0.5 -0.6 -0.4 -0.7

+3.5 Counterfactuals -3.8

-0.6 +0.6

Table 8.1 shows that for the state as a whole, productivity growth was 1.1 percent p.a. greater than living standards growth. Reductions in both the income/output and employment/population ratios occurred. It is clear that the poorer regions fared better than the rich regions. The most prosperous region, Dublin/Mid-East, with a 1960 income level that was 146 percent of the national average, exhibited relatively low growth in living standards, mainly due to low growth in labour productivity. Similarly, the West, which was initially the poorest region with 65 percent of average income, registered the highest growth in living standards, mainly due to faster productivity growth. Although demographic factors have a negative effect on regional growth, this effect is weak. Changes in the income/output ratio in the regions are also negative and small. These flows are identical across regions, since the absence of significant profit outflows during this period results in net factor flows being distributed in proportion to regional GVA. Table 8.1 shows evidence of strong living standards convergence with the rate of V convergence being -3.5 percent per annum between 1960 and 1979. The presence of V convergence also implies E convergence, although there was an absence of J convergence, as evidenced by little change in the rank concordance measure.6 This suggests that E convergence occurred without leap-frogging. It therefore is clear from the counterfactuals that convergence was driven by labour productivity with the other factors having minimal effects. It appears as if there was a vibrant convergence process underway during the 1960s and 70s with poorer regions catching up mainly due to relative productivity improvements. Table 8.2 presents the decomposition for 1979 to 1996. For Ireland as a whole, productivity again dominates living standards growth. The growth of the 6

Rank concordance for living standards decreased very marginally from 1.00 to 0.94 over the period. It is not possible to present significance tests for these J convergence results as the test chi-squared statistic used by Boyle and McCarthy (1997) from Siegel (1956) is only valid for more than 2 years and more than 7 regions.

Regional Disparities in Ireland 133

income/output ratio is now exerting a stronger negative effect. Demographic factors have turned around and are now exerting a positive effect on living standards growth. This is due to the emergence of the “demographic dividend” in the 1990s and is consistent with the findings of Bradley, Fitzgerald, Honohan and Kearney (1997). Table 8.2. Decomposition of living standards in Irish regions: 1979-96 (percent p.a.)

Border Dublin/Mid-East Midlands Mid-West South-East South-West West

+2.7 +3.4 +3.2 +3.1 +3.2 +3.3 +3.1

Income/ Output Ratio -1.3 -0.6 -0.1 -0.8 -0.6 -1.4 -0.6

State

+3.3

-0.8

Rate of V Convergence

+0.9

-0.4

Living Standards

Productivity +3.9 +3.4 +2.7 +3.5 +3.4 +4.5 +3.6

Employment/ Population Ratio +0.1 +0.6 +0.6 +0.4 +0.4 +0.2 +0.1

+3.7 Counterfactuals -0.4

+0.4 +0.6

Dublin/Mid-East, where income was 120 percent of the average in 1979, subsequently grew faster in terms of living standards than any other region. On the other hand the Border, with an initial income of 81 percent of the average, was the slowest growing region over the period. Productivity growth played an ambiguous role in that two of the poorest regions, the Border and the West actually experienced faster productivity growth than Dublin/Mid-East, while the SouthWest had the fastest productivity growth. During the 1980s and 1990s, there were significant differences in the effect of the growing income/output ratio on regional performance. The Border and the South-West experienced larger than average outflows, while outflows from the Midlands and Dublin/Mid-East were lower. As regards demographic factors, Dublin/Mid-East and the Midlands benefited the most with Border and the West benefiting very little. Table 8.2 reveals evidence of weak living standards divergence, with the rate of V divergence being +0.9 percent per annum over the period. Based on the counterfactuals, divergence was driven by the employment-population ratio. Productivity, which was a strong convergent force in the previous period, was only a weak convergent force. The income-output ratio was also weakly convergent in its effect on living standards. In addition, the evidence on J convergence suggests that overall E convergence was not present.7

7

The J convergence measure decreased very marginally from 1.00 to 0.94 over the period.

134 Eoin O’Leary

It is clear from the foregoing that changing productivity growth was one of the main causes of the reversal in the regional convergence record. The next section proceeds to investigate aggregate productivity growth in greater detail.

8.4

Decomposing Aggregate Productivity: Sectoral Productivity Growth and Structural Change

The level of aggregate labour productivity, or GVA per worker, in a particular year, may be defined as the weighted sum of sectoral productivity levels, where the weights are base year sectoral employment shares. Thus, the growth in aggregate productivity consists of two components, productivity growth in each of the 3 broad sectors and the effect on aggregate productivity of shifting sectoral employment shares due to structural change over time. The affect of each of these components on both the degree of aggregate growth and convergence are now analysed in turn. 8.4.1

Sectoral Productivity Growth

Table 8.3 presents the aggregate and sectoral growth and V convergence rates for the period 1960 to 1979. Overall, Manufacturing exhibited marginally the fastest productivity growth, with Services being the slowest growing sector. However, overall strong V productivity convergence of 3.9 percent per annum is clearly driven by converging Services, with both Agriculture and Manufacturing exhibiting very weak convergence. The reason is that in Services the fastest (slowest) growing regions were those with the lowest (highest) productivity levels in 1960. Thus, Dublin/Mid-East, the clear productivity leader in Services in 1960, grew at 1.6 percent per annum subsequently, compared to approximately double this rate in the South-East and the Border. This strong convergence process was not present in the other sectors. For example, in Manufacturing the productivity leader, Dublin/Mid-East, grew faster than the Border. The presence of V convergence also signifies E convergence, although this again occurred without leap-frogging.8 Table 8.4 presents the results for the period 1979 to 1996. Overall, Manufacturing was the fastest growing sector, nearly doubling its performance compared to the previous period. On the other hand, productivity growth in Services declined significantly. In terms of the convergence performance, it now seems as if the overall divergence of 1.2 percent per annum was driven by both Manufacturing and Services. In Manufacturing the strongest growth performers were Dublin/Mid-East and the South-West, the two highest productivity regions to begin with. These regions also witnessed the emergence of concentrations of 8

The J convergence measure decreased very marginally from 1.00 to 0.93 for each of the sectors and from 1.00 to 0.96 for the aggregate.

Regional Disparities in Ireland 135

successful foreign-owned multi-national companies in the high-technology computer, electronics and pharmaceutical sectors. A similar pattern to that in Manufacturing also emerged in Services, with the most prosperous regions growing moderately, and productivity in the poorer regions declining slightly. The evidence on J convergence suggests that E convergence was not present to any great extent.9 Table 8.3. Sectoral productivity in Irish regions: 1960-79 (percent p.a.)

Aggregate

Agriculture

Manufacturing Services

Border Dublin/Mid-East Midlands Mid-West South-East South-West West

+4.5 +2.3 +3.7 +4.1 +4.2 +3.7 +4.7

+4.9 +2.0 +2.3 +3.5 +4.3 +3.5 +3.3

+3.3 +3.5 +4.2 +4.3 +4.1 +4.0 +4.2

+3.1 +1.6 +3.0 +2.9 +3.4 +2.9 +2.9

State

+3.5

+3.6

+3.7

+2.3

Rate of V Convergence

-3.9

-0.2

-0.8

-8.7

Table 8.4. Sectoral productivity in Irish regions: 1979-96 (percent p.a.)

Aggregate

Agriculture

Manufacturing Services

Border Dublin/Mid-East Midlands Mid-West South-East South-West West

+3.9 +3.4 +2.7 +3.5 +3.4 +4.5 +3.6

+7.9 +4.8 +7.2 +6.1 +3.8 +5.2 +6.3

+6.2 +7.9 +3.6 +5.9 +7.0 +8.6 +6.5

+0.0 +1.0 -0.6 +0.2 -0.2 +0.4 -0.2

State

+3.7

+5.9

+7.2

+0.5

Rate of V Convergence

+1.2

-2.8

+7.8

+9.1

It has been widely documented that manufacturing and aggregate GVA estimates may be distorted during this period, due to transfer pricing by foreign owned companies based in Ireland. In the present context this calls for adjustments to these data for 1996. Given the difficulties in measuring transfer 9

The J convergence measure decreased from 0.93 to 0.66 in Agriculture; 0.93 to 0.89 in Manufacturing; 0.93 to 0.86 in Services, while remaining unchanged at 0.96 for the Aggregate. Thus it appears the E convergence did not occur in any sector except, perhaps Agriculture.

136 Eoin O’Leary

pricing, it is not surprising to find substantial variations in approach. For example, Birnie and Hitchens (1998) adjust manufacturing GVA by the full amount of profit outflows, which are assumed to be generated completely by transfer pricing. This provides an upper estimate of the extent of transfer pricing, since it implies that no outflows occur for genuine profit repatriation reasons. In order to address this issue, O’Leary (1997) adjusts manufacturing GVA by 50 percent of profit outflows. This approach provides an average estimate of the degree of transfer pricing, as it lies between the two extremes provided by Birnie and Hitchens’s (1998) upper estimate and the lower estimate, where no adjustment is made. O’Leary’s (1997) method is adopted in this chapter. In order to operationalize it at a regional level, it is necessary to adjust regional GVA estimates in 1996 for 50 percent of profit outflows. Following O’Leary (1999), the national estimate of profit outflows is distributed regionally based on each regions share of foreign owned manufacturing profit. This is measured as the percentage of the remainder of net output in manufacturing establishments accounted for by foreign owned firms. Table 8.5 confirms that the regions with concentrations of foreign owned enterprises are most affected. These are Dublin/Mid-East, the Border and the South-West. The effects on the overall rates of V divergence are quite significant, with the rates for the Aggregate and Manufacturing both decreasing. The effect on the J convergence measure is negligible. Table 8.5. Adjusting manufacturing and aggregate regional productivity for transfer pricing: 1979 to 1996 (percent p.a.)

Border Dublin/Mid-East Midlands Mid-West South-East South-West West

Aggregate +3.0 (+3.9) +2.9 (+3.4) +2.5 (+2.7) +2.9 (+3.5) +2.9 (+3.4) +3.7 (+4.5) +3.1 (+3.6)

Manufacturing +4.2 (+6.2) +6.6 (+7.9) +3.1 (+3.6) +4.6 (+5.9) +6.0 (+7.0) +6.8 (+8.6) +5.2 (+6.5)

State

+3.1 (+3.7)

+5.8 (+7.2)

Rate of V Convergence

+0.6 (+1.2)

+7.3 (+7.8)

Note: The numbers in brackets are the unadjusted growth rates from Table 8.4.

8.4.2

Structural Change

Turning now to the second component of aggregate productivity growth, namely structural change. The approach used is based on Doyle and O’Leary (1999). It uses a modified shift-share technique to measure the effect of structural change on the degree of regional convergence (divergence) observed. Aggregate productivity

Regional Disparities in Ireland 137

growth is first decomposed into intersectoral, intrasectoral and residual component as follows: Intersectoral Growth = (6Pi,j,t.Si,j,t+T)/(6Pi,j,t.Si,j,t)

(8.4)

Intrasectoral Growth = (6Pi,j,t+T.Si,j,t)/(6Pi,j,t.Si,j,t)

(8.5)

Residual Growth = (6(Pi,j,t+T- Pi,j,t).(Si,j,t+T- Si,j,t))/(6Pi,j,t.Si,j,t)

(8.6)

where S refers to sectoral employment shares, P is labour productivity and the subscripts i, j and t refer to sectors, regions and time respectively. Intersectoral growth measures aggregate productivity growth arising from shifts in sectoral employment shares due to structural change. Intrasectoral growth measures aggregate growth due to sectoral productivity growth. The residual or interaction component represents the joint effect of changes in both sectoral employment shares and productivity growth. This effect, which is usually small, is here attributed to structural change. Intersectoral and intrasectoral productivity levels are the numerators in their respective equations. The effect of structural change on the rate of convergence (divergence) of aggregate productivity is then analysed using counterfactual experiments. The counterfactual rate is calculated for intrasectoral productivity levels, as this excludes the effects of structural change. If the resulting rate of convergence (divergence) increases relative to the observed rate, then structural change may be taken as having a divergent (convergent) effect. If the resultant rate decreases then it has a convergent (divergent) effect. The size of the convergent (divergent) effect may be calculated as the difference between the two rates. By measuring the affect of structural change on the rate of convergence (divergence), this technique, which has been applied to regions by O’Leary (2003), facilitates testing of the hypothesis that structural change has a convergent effect. Esteban (2000) employs a similar technique to analyse regional differences from the European average at a point in time. The approach differs from that of Cuadrado-Roura et al (1999), who in their study of Spanish regions, formulate the same hypothesis but use a different method, which involves investigating the trend of indices of inequality of productive structure. The method also differs from the standard conditional E convergence method, where the initial agricultural employment shares is included as a right hand side variable alongside the initial per capita income level, in explaining growth (see, for example, Hofer and Wörgötter 1997). Table 8.6 presents the evolution of the sectoral employment shares over the period. Changing employment shares may contribute to aggregate productivity growth if the shift is from low productivity sectors, like Agriculture, to high productivity sectors, like Manufacturing. This might facilitate convergence, if poorer regions have larger employment shares in low productivity sectors (O’Leary 2003). The Dublin/Mid-East region is clearly at a more advanced stage of development, with the services employment share increasing to 74 percent in 1996, a declining share in Manufacturing since 1960 and a very small share in Agriculture. The other regions are in the earlier stages of industrialization, with

138 Eoin O’Leary

scope for further declines in Agriculture, shares in Manufacturing only beginning to decline in the last twenty years and shares in services rising to only 55 percent in 1996. Table 8.6. Sectoral employment shares in Irish regions: 1960, 1979 and 1996 (percent)

Agriculture

Manufacturing

Services

1960 1979 1996 1960 1979 1996 1960 1979 1996 Border Dublin/Mid-East Midlands Mid-West South-East South-West West

52 9 49 48 44 40 65

29 4 33 26 24 25 39

16 3 15 13 17 15 22

19 36 22 18 20 24 11

34 34 34 32 34 31 24

34 24 31 33 31 29 23

29 54 30 34 35 37 24

37 62 33 42 42 44 38

50 74 53 54 53 57 55

State

36

19

11

25

32

27

39

49

62

Table 8.7 shows the effect of sectoral employment shifts on aggregate productivity performance between 1960 and 1979. Overall, structural change contributed +0.7 percent per annum or 20 percent of aggregate productivity growth. As a result of the strong industrialization underway in this period, the poorer regions benefited the most from this effect, with the West standing out with a 36 percent contribution. On the other hand, Dublin/Mid-East, which limited scope to benefit from decreasing agricultural employment shares, only had a 13 percent contribution from structural change. Thus, structural change had a convergent effect on the degree of aggregate convergence, with -2.0 percent per annum or over 50 percent of the overall rate of convergence being attributed to this effect. It had a negligible effect on J convergence.10 Table 8.8 displays the same decomposition for the later period.11 Overall, the beneficial effect of sectoral employment shifts had all but disappeared during the 1980s and 1990s. Only the West benefited from this effect. In each of the other regions, its effect was either neutral or slightly negative, indicating that sectoral employment shares changed to the detriment of aggregate productivity growth. Dublin/Mid-East and the South-East were especially notable in this regard.12 However, despite the smallness of the magnitudes, the difference between regions in the contributions from structural change had a strong convergent effect on the degree of aggregate divergence. In other words, divergence would have been 10

11

12

The J convergence measure decreased very marginally over the period from 1.00 to 0.98 for the intrasectoral level and from 1.00 to 0.96 for the aggregate level. Based on the adjusted aggregate and manufacturing GVA estimates presented in Table 8.6. This is due to the combination of decreasing productivity growth and increasing employment shares. This is part of the residual component, which is included here as part of structural change.

Regional Disparities in Ireland 139

stronger, at 1.5 percent per annum, only for the convergent effect of structural change.13 Table 8.7. Decomposing regional productivity, 1960-79 (percent p.a.)

Aggregate Productivity Growth

Contributions from: Sectors

Structural Change

Border Dublin/Mid-East Midlands Mid-West South-East South-West West

+4.5 +2.3 +3.7 +4.1 +4.2 +3.7 +4.7

+3.5 +2.0 +3.0 +3.3 +3.8 +3.2 +3.0

+1.0 +0.3 +0.7 +0.8 +0.4 +0.5 +1.7

State

+3.5

+2.8

+0.7

Rate of V Convergence

-3.9

-1.9

-2.0

Table 8.8. Decomposing regional productivity, 1979-96 (percent p.a.)

Aggregate Productivity Growth

Contributions from: Sectors

Structural Change

Border Dublin/Mid-East Midlands Mid-West South-East South-West West

+3.0 +2.9 +2.5 +2.9 +2.9 +3.7 +3.1

+3.2 +3.3 +2.5 +2.9 +3.2 +3.7 +2.8

-0.2 -0.4 0.0 0.0 -0.3 0.0 +0.3

State

+3.1

+3.2

-0.1

Rate of V Convergence

+0.6

+1.5

-0.9

The foregoing has identified the factors that were important in driving the productivity convergence performance in the regions. In the 1960 and 70s, overall strong convergence was driven by convergence in Services and by the convergent effects of structural change. The emergence of weak productivity divergence in the 1980s and 1990s was driven by strong divergence in both Manufacturing and 13

Once again, the J convergence measure exhibited very little change as the level for 1996 was 0.96 for both the intrasectoral and the aggregate levels.

140 Eoin O’Leary

Services. Divergence would have been stronger only for the continuing convergent effect exerted by structural change. Like many small countries, these trends may have been influenced by regional industrial policy. Accordingly, the next section proceeds by investigating the role of policy.

8.5

The Role of Regional Policy

In small countries government policy often plays an important role. Accordingly, this section analyses the extent to which the stance of Irish regional and industrial policy and the convergence performance of the regions were related. Balanced regional development has long been an objective of the Irish government and the EU. During the last forty years Irish regional policy has undergone shifts in line with the changing focus of industrial policy. In the early 1960s the government embarked on an export-led growth strategy. During the next two decades a regional policy of dispersal, implemented by the Industrial Development Authority (IDA), aimed to attract, using job targets and advance factories, foreign owned manufacturing firms to 177 towns and cities spread throughout the country (Breathnach 1982; O’Leary 2001b). However, given that the policy instrument was targeted at manufacturing, it is perhaps surprising that hardly any regional convergence of manufacturing productivity levels occurred during this period (see Table 8.3). A closer look at this sector indicates there was a very high degree of homogeneity of overall manufacturing productivity in Irish regions, with the coefficient of variation being very low at 9.3 percent in 1960 and 8.0 percent in 1979. Thus, there was very little scope for manufacturing convergence to occur. Indeed, it is surprising that weak manufacturing convergence had in fact emerged by 1979, as industrialization of regions outside Dublin/Mid-East (see Table 8.6) might have resulted in regions specializing in different manufacturing industries and consequently growing at different rates. This process of industrialization may have had the additional effect of increasing personal incomes and therefore the demand for services in the regions outside Dublin/Mid-East. This in turn may have lead to growth opportunities for the services sectors, which were largely sheltered from international trade at the time. These opportunities may have been realised, as evidenced by the significant growth in productivity levels there, relative to their counterparts in Dublin/MidEast. Thus, although the policy of regional dispersal had very little effect on the degree of convergence in Manufacturing, it may still have had a role to play in the strong convergence among Services. In addition, since regional policy was tied to industrialization, it may have contributed to the convergent contribution from structural change. Thus, the regional policy of dispersal may have contributed strongly to aggregate living standards convergence, through its effect on aggregate productivity convergence. This hypothesis warrants further study.

Regional Disparities in Ireland 141

During the 1980s, the widely documented poor performance of the Irish economy coincided with substantial job losses. The regional policy of dispersing foreign industry was criticized due to the low degree of linkage with indigenous industry. Accordingly the IDA effectively suspended regional policy and replaced it with targeted industrial policy, the aim being to develop sustainable numbers of high value-added companies in high growth sectors. Although, during the 1980s and 1990s substantial EU Structural Funding was made available, Ireland was treated as one region for EU purposes, with the result that little attention needed to be devoted to the development of coherent regional policies. The shift of emphasis by the IDA from an explicit regional focus to a targeted industry approach had a clear role to play in the emergence of strong divergence in Manufacturing during the period. The largely homogenous manufacturing base that was present in 1979 had disappeared by 1996, by which time concentrations of successful foreign-owned multi-national companies had emerged in the hightechnology computer, electronics and pharmaceutical sectors located in and around the main urban centres (Bradley and Morgenroth 1999; Meyler and Strobl 2000; O’Leary 2001a). Manufacturing divergence may have lead to service sector divergence, especially in the sheltered sectors. Thus, concentrations of high-technology manufacturing industry resulted in strong relative growth of incomes and therefore increased demand for services there. This can be seen in the regions where manufacturing sectors prospered, namely Dublin/Mid-East and the South-West, where service sectors increased in relative terms. Similarly, the opposite process seems to have been under way in regions like the Midlands, where diminishing relative manufacturing and service productivity coincided. However, it may also be that the strong tourism growth in the 1990s centred in Dublin/Mid-East and the South-West, and the emergence of internationally traded financial services also centred in Dublin/Mid-East, may also explain the strong service sector divergence that emerged.14

8.6

The Future Regional Policy Dilemma

This detailed investigation of the sources of regional disparities in Ireland since 1960 has confirmed the important role that has been played by regional and industrial policy. The re-emergence at the end of the 1990s of the problem of regional imbalance points to a renewed role for policymakers. The policy dilemma is how to address this problem without compromising future national performance. Other small countries have much to learn from watching Ireland’s future performance. By treating the problem of regional imbalance using redistributional measures, the danger is that regional and therefore national competitiveness will be undermined. Policies specifically targeted at improving 14

It is not possible to test this hypothesis in greater detail due to the limited sectoral disaggregation available.

142 Eoin O’Leary

regional productivity growth may avoid this danger. Unfortunately, the signs are that Irish policymakers may already be repeating past mistakes. By and large Irish policy has focused on manufacturing industry, although internationally traded services have also been targeted in more recent years. Policy not only affected the convergence performance of these sectors directly, but may also have influenced sheltered services. Thus, policy seems to have affected the degree of productivity convergence (divergence), which was found to be one of the main drivers of the degree of living standards convergence (divergence). The stance of policy underwent a major transformation over the period. In the 1960s and 70s, there was a regional dimension to industrial policy as a policy of dispersal was pursued. By spreading foreign direct investment throughout the country, this went against Buchanan’s advice to concentrate on a limited number of growth centres (1969). Although the policy did deliver balanced regional development, it coincided with very disappointing national growth rates and a lack of national convergence with the EU. During the last two decades of the twentieth century, where the chief concern of policy makers was reducing unemployment, regional policy was suspended and replaced by targeted industrial policy. Although this had the ultimate effect of helping to generate highly impressive national growth rates, it coincided with the emergence of regional divergence. At the beginning of the new millennium increasing regional disparity has put regional issues on the Irish policy agenda, with the inclusion for the first time of balanced regional development as a key objective in the National Development Plan 2000-06 (2000). The dilemma facing policymakers is how to bring about regional balance without compromising future national performance. There is widespread consensus that over the next twenty years, the chief determinant of Irish growth and competitiveness will be productivity growth of internationally competitive industry in both the manufacturing and service sectors (Gallagher et al. 2002). Based on the results presented here, the productivity performance of these industries will also play an important role in determining the degree of regional balance or convergence. Other factors, most notably structural change, which had a strong but declining convergent affect over the periods analysed, may be expected to continue to decline in future years as the pool of low productivity agricultural workers lessens further. The future issues involved may be considered under two scenarios. The first, the convergence scenario, might involve the emergence over the next number of decades of urban diseconomies in the form of congestion and higher factor costs in the major cities such as Dublin and to a lesser extent in Cork, Limerick and Galway. If these inhibit growth in the more prosperous regions, and if the less prosperous regions are able to exploit their catch-up potential, then regional convergence may result. In this scenario, a tendency towards balanced regional development would be the outcome. A second scenario might involve regional divergence. This would result if agglomeration economies in urban centres strengthen and outweigh emerging diseconomies and if poorer regions fail to exploit their catch-up potential.

Regional Disparities in Ireland 143

There are a number of theories in the literature that may be invoked. For example, neoclassical growth theory predicts regional convergence, while at a regional level, the Williamson (1965) hypothesis suggests that as countries catchup regional disparities may initially increase, due to the emergence of growth poles, but may subsequently lessen as urban diseconomies emerge. On the other hand, new growth theory and new economic geography, which have been the subject of a considerable research effort since the mid-1980s, generally predict divergence. Unfortunately, the absence of empirical economic research on these competing hypotheses in the Irish case, which is partly explained by lack of data, severely inhibits our capability to assess these or other scenarios at this time. However, there is a distinct danger that political expediency may result in a repetition of past mistakes. The recent announcement of the National Spatial Strategy (2002), which is an integral part of the National Development Plan, providing the blueprint for Irish regional development, confirms that policy will not be focused on a limited number of cities. Instead, the government is to concentrate on spreading growth through a greater number of towns and cities. This may bring more balanced regional development, but at the cost of diminished national growth. After decades of neglect, Irish local democracy is currently in no fit state to play a role in shaping the long-term economic performance of Irish regions. Irish economic policy is administered by central government in Dublin mainly through the Department of Finance. Regional policy is given low priority, being currently run chiefly by the Department of the Environment. By failing to build a coherent strategy in partnership with internationally competitive industries, national and regional policymakers may seriously jeopardize future prosperity. Such a strategy primarily involves forming economic, not political consensus. It necessitates trade-offs, by deciding for example, to target certain urban centres for industrial development and not others, or by selecting smaller towns as commuter towns for major urban centres and not as centres for industrial development themselves. Unfortunately, policymakers are currently showing no sign that they are prepared to face these choices. The only way for Irish regional policy to be taken seriously is if there is a cost of not doing so. By spreading funding around the country for short-term political purposes, the Irish government may be risking the heavy cost of diminished national growth rates over the next decade.

144 Eoin O’Leary

Appendix

Fig. 8.1. Regional authority areas Table 8.9. Definition of regional authority areas

Regions

Counties

Border:

Cavan, Donegal, Leitrim, Louth, Monaghan and Sligo

Dublin/Mid-East:

Dublin, Kildare, Meath and Wicklow

Midlands:

Laois, Longford, Offaly and Westmeath

Mid-West:

Limerick, Clare and Tipperary North Riding

South-East:

Waterford, Carlow, Kilkenny, Wexford and Tipperary South Riding

South-West:

Cork and Kerry

West:

Galway, Mayo and Roscommon

Regional Disparities in Ireland 145 Table 8.10. Sector classification

Sectors

Definition

Agriculture: Manufacturing: Distribution:

Agriculture, Forestry and Fishing Manufacturing and Building Retail and Wholesale Distribution, Transport and Communication Services Market Services, Other Non-Market Services (including Public Administration, Defence, Education and Health) and Rent of Dwellings

Other Services:

References Attwood EA, Geary RC (1963) Irish county incomes in 1960. ESRI General Research Series (16), September Birnie JE, Hitchens DMWN (1998) Productivity and income per capita in a peripheral European economy: the Irish experience. Regional Studies 32(3):223-234 Boyle GE, McCarthy TG (1997) A simple measure of E-convergence. Oxford Bulletin of Economics and Statistics 59:257-264 Bradley J, Morgenroth E (1999) Celtic cubs? Regional manufacturing in Ireland. In: Duffy D, Fitzgerald J, Kearney I, Smyth D (eds) Medium term review: 1999-2005. Economic and Social Research Institute, October 157-174, Dublin, Ireland Bradley J, Fitzgerald J, Honohan P, Kearney I (1997) Interpreting the recent Irish growth experience. In: Duffy D, Fitzgerald J, Kearney I, Shortall F (eds) Medium Term Review: 1997-2003. Economic and Social Research Institute, April, 35-66, Dublin, Ireland Breathnach P (1982) The demise of growth-centre policy: the case of the Republic of Ireland. In: Hudson R, Lewis J (eds) Regional Planning in Europe. Pion Ltd, London Buchanan and Partners (1969) Regional studies in Ireland. An Foras Forbatha, Dublin Central Statistics Office (1998) Regional accounts: GDP by region: 1996. CSO Statistical Release, Government Stationery Office, Dublin, November Cuadrado-Roura JR, Garcia Greciano B, Raymond JL (1999) Regional convergence in productivity and productive structure: the Spanish case. International Regional Science Review 22:35-53 Doyle E, O’Leary E (1999) The role of structural change in labour productivity convergence among European Union countries: 1979-1990. Journal of Economic Studies 26(2):106-120 Esteban J (2000) Regional convergence in Europe and the industry mix: a shift-share analysis. Regional Science and Urban Economics 30(3):353-364 European Commission (2001) European Union regional policy: working for the regions. European Commission, Brussels Gallagher L, Doyle E, O’Leary E (2002) Creating the Celtic Tiger and sustaining economic growth: a business perspective. Quarterly Economic Commentary, Economic and Social Research Institute, Dublin, Spring 63-81 Henry E (1997) GDP in Republic of Ireland for seven new NUTS3 planning regions. Mimeo, Economic and Social Research Institute, Dublin, April

146 Eoin O’Leary Hofer H, Wörgötter A (1997) Regional per capita income convergence in Austria. Regional Studies 31(1):1-12 Kennedy K (2000/01) Symposium on economic growth in Ireland: where has it come, where is it going? Reflections on the process of Irish economic growth. Journal of the Statistical and Social Inquiry Society of Ireland XXX:123-139 Meyler A, Strobl E (2000) Job generation and regional industrial policy in Ireland. Economic and Social Review 31(2):111-128 National Development Plan 2000-2006 (2000) Stationery Office, Dublin National Spatial Strategy for Ireland 2002-2020: People, Places and Potential (2002) Stationery Office, Dublin O’Leary E (1997) The convergence performance of Ireland among EU countries: 1960-90. Journal of Economic Studies 24(1/2):43-58 O’Leary E (1999) Regional income estimates for Ireland: 1995. Regional Studies 33(9):805-814 O’Leary E (2001a) Convergence of living standards among Irish regions: the roles of productivity, profit outflows and demography: 1960-96. Regional Studies 35(3):197205 O’Leary E (2001b) Regional divergence in the Celtic Tiger: the policy dilemma. Irish Banking Review, Spring: 2-15 O’Leary E (2003) Aggregate and sectoral convergence among Irish regions: the role of structural change: 1960-96. International Regional Science Review 26(4):483-501 Siegel S (1956) Nonparametric statistics for the behavioural sciences. McGraw Hill, Tokyo Toulemonde E (2001) Actual versus ‘virtual’ employment in Belgium. Regional Studies 35(6):513-518 Williamson JG (1965) Regional inequality and the process of national development: a description of the patterns. Economic Development and Cultural Change 13:3-45

9

The Persistence of Regional Unemployment Disparities in the Netherlands

Oedzge Atzema1 and Jouke van Dijk2 1

2

Department of Human Geography and Planning, Faculty of Geosciences, Utrecht University, the Netherlands Urban and Regional Studies Institute (URSI) and Department of Economic Geography, Faculty of Spatial Sciences, University of Groningen, the Netherlands

9.1

Introduction

The Dutch labour market has undergone profound changes since World War II. In the 1950s and 1960s labour markets in various sectors were tight and as a consequence, unemployment remained low. Relatively short periods of recession led to only minor unemployment increases and these small shocks dissolved quite rapidly. In the subsequent decades, namely the 1970s and 1980s, the situation changed quite dramatically. Industrial restructuring, oil price shocks and subsequent recessions caused severe increases in unemployment, without adjustment to previous levels when the economy picked up afterwards. This period was characterised by high and persistent unemployment, “sticky” wages, a rigid labour market, and within political and academic circles was customarily referred to as the “Dutch Disease”. As Lubbers declared in 1990 the Netherlands had become a “sick” country. In addition, the fast growth of unemployment in the 1980s was caused by developments on the supply side of the Dutch labour market, especially, the entry of vast cohorts of so-called “baby-boomers” born after World War II and the rapid rise of labour market participation of Dutch women. During the 1990s the labour market situation again underwent quite dramatic change and was dubbed the “Dutch miracle” (Visser and Hemerijck 1994). Unemployment fell from 7.5 percent in 1994 to a mere 2 percent in 2001. At the same time employment increased by 1.4 million jobs between 1994 and 2001. In fact, the growth of employment rates in the Netherlands has been more in line with the “American” rate than with the unemployment rates of the rest of Europe. The most important explanations for this “miracle” include the process of regulatory reform and deregulation, which created a more flexible labour market, on the one hand, and sustained wage moderation which led to a fall in labour costs relative to competitive countries on the other. This peculiar historical development from “disease” towards “miracle” raises two important questions:

148 Oedzge Atzema and Jouke van Dijk

1. To what extent does the time pattern of unemployment growth vary at the regional level? 2. Do regional disparities converge or diverge during the period of Dutch recovery? Answers to these questions might be determined by the fact that the Netherlands is a small country with an extremely open economy. Theoretically speaking, convergence of regional variation in unemployment might be expected because people in a small country like the Netherlands can change their places of work quite easily without having to change their place of residence. In other words, spatial “flexibility” in the labour market might be grounded in the small size of the country. Alternatively, small countries like the Netherlands might not be small enough to cause differences in regional unemployment to equalize. In that case the regional unemployment differences might be persistent despite several stages of economic upturn. As we shall illustrate for the Dutch case the latter scenario seems to be the case because most of the people active in the labour force in the Netherlands have a limited commuting tolerance at a maximum of half an hour driving time by car. Besides that most people do not want to take up employment in an area far away from the region where they live. Moreover, the low level of spatial flexibility in the Dutch labour market has been stimulated by the regional policy of the Dutch government. Right from the outset of regional policy in the 1950s, the main objective was to bring work to the people instead of providing incentives for people to move to the work. To substantiate our argument we will analyse the persistence of regional differences in unemployment in the Netherlands. For the first period (from a healthy economy to the “Dutch Disease”) we use data from a limited number of Dutch regions. For the 1990s, however, we use more detailed spatial data. Analysis of the spatial distribution of total unemployment categorised by socioeconomic groups is also possible with this data. The framework of this chapter is as follows. Section 9.2 outlines the development of unemployment in the Netherlands from an European perspective. The development of unemployment at the national level is described in Section 9.3. This section also pays attention to the impact on the analysis when alternative definitions of unemployment are used. The regional differences in unemployment since World War II are analysed in Section 9.4 and this section will also focus on the level of spatial disaggregation. Section 9.5 analyses the differences in regional unemployment for various socio-economic groups. Some concluding remarks are made in the final section of the chapter.

Persistence of Regional Unemployment in the Netherlands 149

9.2

The Dutch Labour Market in a European Perspective

In 2001 about 13 million people were unemployed in the European Union. This figure is about the same as it was at the beginning of the 1990s, while in the middle of the decade the number of unemployed reached a peak of 18 million. Still, the unemployment rate in the European Union at 8 percent was twice as high as the 4 percent in the United States (EU 2001). The development of unemployment is, in general, the combined result of economic growth and demographic changes. In the 1990s it was mainly the high economic growth in Europe that led to a reduction in unemployment. The contrary development is also true as the effect of the recent economic recession led to a rise in unemployment. According to Eurostat, unemployment in the European Union was over 9 percent in the summer of 2003. This relation between economic growth and unemployment is also pertinent to individual countries. However, Figure 9.1 shows that there are substantial differences in the level and development of unemployment in the last decade within the European Union.

Fig. 9.1. Level of unemployment in some European countries: 1988-2000

The low rate of unemployment in the Netherlands in the second half of the 1990s has to do with the growth of the economy and the fact that employment was above the European average. Figure 9.2 shows that the growth of employment in this period was even higher than in the United States. The relatively low level of unemployment in the Netherlands seems to be sustainable, because in the summer of 2003 the Dutch unemployment rate of 4.2 percent was still the lowest in the European Union, with the exception of the smallest EU country - Luxemburg (3.7 percent) (Eurostat 2003). The unique position of the Netherlands in Europe can be attributed to comparatively advantageous developments in labour costs. In accordance with

150 Oedzge Atzema and Jouke van Dijk

certain policy developments, a far-reaching reconstruction scheme of government finance was launched in the 1980s. This policy led to opportunities for readjustment of taxes and to the decrease in the burden on the taxpayer in the 1990s. The marginal part of labour costs was scaled down and the burden on taxpayer wages was reduced. Besides, the upward pressure from the labour market on wages was limited because of the substantial increase in labour supply (higher female participation rates). Wage restraint was mainly put into effect in the public sector, which led to lower average wages among public sector workers than in the private sector. This wage restraint was achieved through a general 3 percent reduction in the salaries of civil servants. Besides that, employment in civil services was reduced considerably, especially in higher earnings jobs. High paid civil servants were partly replaced by low paid employees according to the “job machine model” frequently used in the US labour market (Elfring and Kloosterman 1989). As a consequence, people whose income was linked to civil service scales functioned as a kind of cheap labour pool for entrepreneurs in the private sector. Furthermore, as a result of the rise of female labour market participation rates the increase of the double income household group became relevant. As a consequence, effective demand rose in spite of the moderate increase in wages. The labour-intensive service sector benefited especially from these changes. Wage restraint also played an important role for international firms since it partly compensated the appreciation of the then-Dutch guilder by the Dutch Central Bank. Firms also profited from a relatively stable financial and economic climate and low interest rates. Broersma and Van Dijk (2002) show that the speed of adjustment in regional labour markets in the Netherlands is similar to that of the United States and much faster than in other European countries. The way these adjustments take place through changes in the participation rates, however, are rather more “European” in nature. In contrast, migration is the main adjustment mechanism in the United States. 130 Netherlands

120 USA

110 EU-15

100

90 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

Fig. 9.2. Employment growth in the United States, the Netherlands and European Union (1990=100) Source: Netherlands Bureau of Economic Policy Analysis (CPB 2000), Centraal Economisch Plan, Appendix B3, pp 218-219

Persistence of Regional Unemployment in the Netherlands 151

While unemployment in most European countries decreased across the 1990s, the regional differences in unemployment continued to exist. Figure 9.3 presents the member states in ascending order of national unemployment rates and shows that the differences in unemployment within member states are larger than between member states (this accounts for over 200 European Nuts II regions which are equivalent in the Netherlands to the so-called 12 provinces). The number of unemployed people in some regions within a country is often more than four times higher than in other regions. The differences can be even higher at the sub-regional level, especially in larger countries (such as Spain, Italy, Germany). In a small country like the Netherlands this degree of regional disparity is less likely to be prevalent. Nonetheless, there must be more to this phenomenon than just the size of the country since the regional differences in Belgium are much greater than in the Netherlands. In the end, it is the regional (in)equality of the economic development that counts and this can vary from country to country for a whole host of reasons. For example, the differences between (parts of) the Walloon provinces of Belgium and Flanders are larger than between the Dutch regions. 30,0

% unemployment

25,0

20,0

15,0

10,0

5,0

Spain

Finland

Italy

Greece

France

ny Germa

Belgium

n Swede

m Kingdo United

Austria

al Portug

Netherl

ands

0,0

highe st

lowe st

av e rage

Fig. 9.3. Regional differences in unemployment within select European countries, 1998 Source: Eurostat (1999), Statistics in Focus, Eurostat, 5/1999

9.3

Decrease of Different Types of Unemployment in the Netherlands

The very favourable Dutch labour market situation in the 1990s was preceded by a very long period of high rates of unemployment, as shown in Figure 9.4. In the

152 Oedzge Atzema and Jouke van Dijk

1950s and 1960s unemployment was low, certainly compared to the pre-war crisis situation, but employment levels started to increase in the 1970s. Around 1980 in the aftermath of the second oil shock widespread cuts, closures, redevelopment, reorganisation and labour saving measurements caused an employment reduction of more than 100,000 jobs per year. As a consequence, between 1981 and 1983 registered unemployment rose from 3 percent to over 10 percent. It took about twenty years before unemployment levels settled down to levels comparable with the 1950s and 1960s. 12 .0

% unemployment

10 .0 8.0 6.0 4.0 2.0

96

92

88

84

80

76

72

68

64

60

56

00 20

19

19

19

19

19

19

19

19

19

19

19

19

52

0.0

Fig. 9.4. Registered unemployment 1952-2001 (percent of the labour force in the Netherlands) Source: Statistics Netherlands

In the second half of the 1990s annual employment growth was almost 100,000 labour-years per year and between 1994 and 2001 registered unemployment dropped from almost 8 percent to 2 percent. In spite of the enormous growth of the number of employed persons, we must bear in mind that the numbers of full-time jobs remained almost constant in the Netherlands. A lot of the new jobs are parttime or temporary contracts. Measured in full-time equivalents, the increase of employment in part-time jobs accounts for approximately two-thirds of the growth of the number of employed people. Partly due to these part-time jobs, the growth of the number of jobs during the period 1988-2000 exceeds almost four times the increase of the potential labour force. This growth clearly leads to an upward pressure on the Dutch labour market. Although the growth of employment started in 1993, it is striking that only after 1995 did this growth lead to a substantial decrease in unemployment. This timelag probably has to do with the labour market position of the unemployed people. New entrants (among others recent graduates) and those who re-entered the labour market after a period of inactivity (often women returning after child rearing) account for part of the job growth, such that people who were unemployed before benefit less from the upturn. Under these conditions, if the demand for labour

Persistence of Regional Unemployment in the Netherlands 153

intensifies, employers can appeal to unemployed people to try and fill their vacancies. However, this procedure differs for the different categories of unemployed person. Later on in this chapter, we will show that losing unemployed status is a persistent problem for the elderly and long-termed unemployed categories. Registered unemployment is just one of the possible indicators for unused supply of labour in circulation. Registered unemployment (in Dutch abbreviated GWL) is defined as those persons aged 16 to 64 years who are registered at the Public Employment Service and who do not work, or work less than 12 hours a week, and who are available within 2 weeks for a job of more than 12 hours a week. The information is based on data from the Labour Force Survey (in Dutch abbreviated EBB). The GWL is the official unemployment source in the Netherlands and is published monthly by Statistics Netherlands (in Dutch abbreviated CBS). This figure comes fairly close to the International Labour Organization (ILO) definition, which is generally accepted in international comparisons. Another indicator is the number of non-working job-seekers (in Dutch abbreviated NWW) as counted by the Public Employment Service. This indicator concerns persons registered at the Public Employment Service who do not work more than 12 hours a week. The advantage of the NWW figure is that detailed itemisation by region is possible by age, gender, education, and duration of unemployment. At first sight the difference in definition between the GWL and NWW seems minor (Table 9.1), but the NWW figure is approximately two and a half times higher than the GWL figure. Consequently, the availability for a job of more than 12 hours a week leads to a substantial difference of the number of unemployed. Nevertheless, the development over time for both figures is rather similar, implying that both measures are suitable for historical studies and changes over time. Table 9.1. Indicators of unused supply of labour in the Netherlands, 1999

Indicator Registered unemployment (GWL) Non-working job-seekers (NWW) Seeking paid work>12 hours a week ABW, WW or WAO-benefits*

Absolute number 221,000 548,000 805,000 1,500,000

Source: Statistics Netherlands * ABW = minimum subsistence benefits WW = unemployment Benefits WAO = disability benefits

Another delineation of the unemployed labour force concerns the number of people who would like to work for at least 12 hours a week, irrespective of whether they search for work, whether they are immediately available, or whether they are registered at the Employment Exchange. In 1999 this group consists of over 800,000 people which was 250,000 persons higher than the NWW figure.

154 Oedzge Atzema and Jouke van Dijk

Through the course of time, it turns out that this indicator does not follow the decrease of GWL as fast as the one for NWW. An even broader indication of unemployment is the number of people receiving social security payments (including disability allowances) or unemployment benefit. This group of 1.5 million people includes those who are not able or unwilling to work. It is, however, striking that the number of people with one of the above mentioned benefits has risen from 1 to 1.5 million since 1980, while the number of registered unemployed is still at the same level in 2000 as it was in 1980 (200,000 people).

9.4

Regional Differences in Unemployment Rates

As shown in Figure 9.3, the regional differences in unemployment in the Netherlands are rather small at the spatial scale of the 12 provinces, especially when compared to other European countries at the Nuts II level. In this section we analyse regional development of unemployment differences over time. Besides an analysis at the provincial level for the 11 regions we also analyse the regional unemployment evolution for a spatial subdivision in 40 COROP-regions. In the post-war period up to around 1975 the national unemployment rate was rather low in the Netherlands and then started to increase rapidly (see Figure 9.4). However, the regional components (regional unemployment rate minus the national unemployment rate) presented in Table 9.2 it is clear that there are substantial differences in unemployment across the 11 Dutch provinces even in the period of low unemployment. We are not able to trace the continuation of trends after 1989 because changes in definition in registration of unemployment. In Section 9.5 we give more detailed information about regional differences after 1989. In Table 9.2 the regional distribution of unemployment among all 11 provinces is expressed by standard deviations and coefficients of variation. It is striking that the standard deviations show only minor variation around 2, although these increase slightly when unemployment rises at the national level. Conversely, the coefficient of variation decreases sharply at the end of the period when national unemployment reached very high levels. The contradiction between both measures is a statistical phenomenon. Although the measures lead to opposing conclusions we prefer the coefficient of variation because it corrects the standard deviation for the national development. So, on the basis of these figures we can suggest that for provinces the national level of unemployment correlates positively with the regional variation of unemployment. In other words, if national unemployment is high the regional differences in unemployment are high too. This observation counts for both the total figures and the figures by gender. The statistics can be interpreted for specific regions. In the low unemployment period of the 1950s, 60s and 70s, unemployment rates were the lowest in the socalled Randstad area, the economic centre and most urbanised part of the Netherlands consisting of the provinces North-Holland, South-Holland and Utrecht. In most of the other provinces unemployment was higher. The provinces

Persistence of Regional Unemployment in the Netherlands 155

in the eastern part (Overijssel and Gelderland) and the southern parts (Zeeland, North-Brabant and Limburg) are close to the national average. The course of unemployment development differs between the provinces. The economic structure in the provinces Overijssel, North-Brabant and Limburg is traditionally specialised in manufacturing and mining production. In the 1970s and at the beginning of the 1980s a process of industrial restructuring resulted in growing unemployment figures in those provinces. Nevertheless, the highest unemployment rates are found in the three northern provinces (Groningen, Friesland and Drenthe). Over the whole period of forty years Groningen and Friesland show high unemployment rates. A notable change took place in the rural peat province of Drenthe. Unemployment rates were very high in the 1950s and 60s, but moved close to the national average at the end of the 1980s. Zeeland is another rural province, located in the south-western part of the Netherlands. The development of unemployment in this province was contrary to that of Drenthe: at the national average in the 1950s, 60s and 70s, but high in the 1980s. Table 9.2. Regional component in unemployment as percent of the labour force Province Groningen Friesland Drenthe Overijssel Gelderland Utrecht North-Holland South-Holland Zeeland North-Brabant Limburg The Netherlands (level) Standard deviation Coefficient of variation

1952 +3.7 +4.0 +8.5 -0.7 +1.0 -0.2 -0.5 -0.8 -0.5 +1.7 -1.8

1958 +3.3 +3.9 +8.1 +0.5 0.0 -0.8 -0.7 -1.1 +1.8 +0.3 -1.4

1967 +2.1 +1.2 +4.1 +1.0 +0.0 -1.1 -1.2 -1.2 0.0 +0.8 +1.9

1975 +1.5 +2.2 +4.3 +1.4 +0.5 -2.4 -1.6 -1.8 -0.7 +2.1 +3.4

1984 +5.8 +3.3 +1.8 +2.3 +1.5 -2.6 -2.3 -2.4 -3.2 +1.8 +3.1

1989 +7.7 +4.7 -0.4 0.0 -0.1 -1.7 +0.6 -0.6 -3.5 -1.0 +0.2

4.8

3.0

2.3

5.0

17.3

13.0

2.1 44%

2.0 66%

1.3 58%

2.0 40%

2.7 16%

1.9 14%

Source: Van Dijk and Oosterhaven (1992)

As regional differences are often more pronounced when a more detailed spatial disaggregation is used, we analyse the spatial unemployment differences for a subdivision of the Netherlands into 43 COROP regions. For this division, a consistent set of data on registered unemployment (GWL) for the period 19881998 is available. With the help of this data we address the question whether the regional unemployment differences decreased or increased over the economic upswing of the 1990s. Several answers can be found to this question since various measures give different results. A first indication gives the range, i.e. the difference between the highest and the lowest observed unemployment rate. This decreases from 10 (14-4

156 Oedzge Atzema and Jouke van Dijk

percent) in 1988 to 8 (10-2 percent) in 1998. This decrease in range also applies for men and women separately. The decrease was caused by the fact that regions with exceptionally high unemployment rates have been, in comparative terms, faring quite well.

3.5 3 2.5 2 1.5 1 0.5 0 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 Stand.deviation Stand.dev. male Stand.dev. females

Coef.variation Coef.var. male Coef.var. female

Fig. 9.5. The regional dispersion of registered unemployment per COROP region

When interpreting Figure 9.5, it is important to know that the calculations are based on figures expressed in rounded percentages (no digits) as published by the CBS for the 43 COROP regions. That is why coincidental fluctuations can easily occur. Furthermore, data by gender are related to only about 30 COROP regions since these figures have not always been published for the smaller COROP regions. In spite of the forced use of rounded percentages, the fluctuations in the standard deviation and variation coefficient are minor. The standard deviation increases slightly when national unemployment rises, while the coefficient of variation decreases slightly, especially at the end of the period. This observation relates to both total figures and the figures by gender. The results for the standard deviation show that regional differences fall when national employment decreases. The slight increase of the coefficient of variation does, however, indicate that not all regions can take advantage of this decline to the same degree. It is therefore interesting to examine the change in the order of individual regions on the basis of unemployment rates (Figures 9.6-9.7). In 1988, the regional pattern of unemployment presented a bi-polar view: urban, nationally central located regions, as well as rural, peripheral located regions, showed high unemployment rates. The city regions of Amsterdam and

Persistence of Regional Unemployment in the Netherlands 157

Rotterdam revealed the highest unemployment rates, as well as the three regions in the northern province of Groningen. Two of the three regions in the northern province of Friesland, as well as the city regions of Arnhem/Nijmgen (in the eastern part of the Netherlands) and South-Limburg (in the southern part of the Netherlands) showed rather high unemployment rates at 10 percent. On the other hand, the peripheral urban COROP-regions in the Randstad, outside the big cities, had the lowest unemployment rates.

Fig. 9.6. Registered unemployment per COROP region, 1988

158 Oedzge Atzema and Jouke van Dijk

Fig. 9.7. Registered unemployment per COROP region, 1998

In 1998, the unemployment rate of the city regions of Amsterdam and Rotterdam were halved (7 percent), yet the figures were still high for the Netherlands. Nevertheless, in 1998 the region Delfzijl in the northern periphery of the Netherlands led with 10 percent, which was just a little less than the 13 percent in 1988. This region took hardly any advantages of the national trend of falling registered employment. The other two regions in the province of Groningen illustrated a “normal” decrease of unemployment similar to the city regions of Amsterdam and Rotterdam. The four peripheral urban regions in the Green Heart of the Randstad had a very low unemployment rate in 1988 (4 percent), and also a decade later in 1998

Persistence of Regional Unemployment in the Netherlands 159

(2 percent). Alongside these regions Southwest-Gelderland at the eastern green edge of the Randstad also registered a very low unemployment rate in 1998 because of a 70 percent drop in unemployment. All these regions can be characterised as suburban residential areas in close proximity to the main urban economic centres of the country. Even though total unemployment was halved in the Netherlands, this reduction did not occur uniformly across regions. The relatively strong reduction of unemployment in the regions belonging to the so-called “Zandstad” region (the city region Arnhem/Nijmegen, Southwestern-Overijssel, North-Overijssel and the province of North-Brabant with the exception of Western-Brabant) merits attention. Unemployment decreased mostly in regions that were part of an urban ring outside the Randstad. Outside this area, there were also a few regions on the southern edge of the provinces of Friesland and Drenthe that showed a strong reduction in unemployment.

9.5

Regional Differences in the Characteristics of the Unemployed

The unemployment figures by GWL used in the previous section only permit limited breakdown by gender. There are no data available on the composition of the unemployed by education, age, and duration of unemployment. This itemisation does exist, however, in the NWW figures of the Public Employment Service. These data are only available in a consistent series for the period March 1993 to October 1999 at the spatial level of 18 so-called RBA-regions. To describe the developments per category and region, index numbers have been calculated with the NWW number in March 1993 set to 100 (see Figure 9.8). The course of the total NWW unemployment roughly matches the GWL number, aside from certain seasonal fluctuations. Besides that, we used absolute numbers of unemployed, while the GWL uses the number of unemployed in the total labour force in terms of percentages. When the absolute number of unemployed remains the same but the size of the labour force increases, the unemployment rate will decrease and the index does not change. When the labour force and the number of unemployed increase in the same way, the index will rise as well. We do, however, pay less attention to the total change, but focus more on the change over time of unemployment for certain socio-economic groups. This specification is relevant because, in spite of the decrease of registered unemployment, there is still a large share of structural unemployment in total labour supply that apparently does not fit the demand profile of the employers.

160 Oedzge Atzema and Jouke van Dijk

175 Age 50+ 150 DUR>3yr 125

Female

100

Low Educ TOTAL 99

99

98

98

97

97

97

96

96

95

95

94

94

94

93

93

75

NWW from March 1993 til November1999 per month (March 1993 = 100)

Fig. 9.8. Development of the number of non-working job-seekers in the Netherlands by socio-economic group, March 1993-October 1999

We focus on a few subgroups who are commonly believed to possess a competitive position on the labour market. It is expected that this group will have to cope with a fast increase of unemployment in times of recession and with a relatively slow decrease in times of cyclical upturn. This seems to be true because the decrease of unemployment of groups like women, elderly persons (50+) and long-term unemployed (longer than 3 years) stays behind national change for all unemployed (see Figure 9.8). Initially low educated unemployed (lower vocational education/lower general secondary education) possess a better position, even though they exhibit stronger seasonal fluctuations as a result of seasonal work such as catering industry and agriculture. This group covers about 60 percent of the total number of unemployed according to the NWW definition. The four distinct groups show considerable differences in their development over time for different regions. These differences have been measured using correlation coefficients (Table 9.3) between the index for a particular subgroup and the index for total unemployment for each of the 18 regions. The closer the value reaches 1, the more the change in the subgroup resembles the change in total unemployment. Table 9.3 presents the average value for all regions as well as the highest and lowest correlation coefficients found for an individual region. The development of the unemployment for low educated and women is very much in line with the course of the total unemployment. After 1997, the unemployment among these groups shows the same decline as the total. The share of women in unemployment does, however, rise in the period 1993-1999 from 43 percent to 52 percent, but this rise is in line with the overall increase in participation of women. Both the other groups show a completely different pattern. They are the “losers”, particularly in the case of the elderly. Their share in the NWW unemployment increases from 11 percent (1993) to 20 percent (1999). The related percentages for the long-term unemployed are 21 percent and 28 percent respectively. In spite of the favourable development of employment the

Persistence of Regional Unemployment in the Netherlands 161

labour market position of these groups hardly improves. Table 9.3 also shows that the change across the regions varies since the regions with the highest and lowest unemployment show substantial differences in correlation coefficients. This section further analyses the regional differences for the four socio-economic groups for individual regions. Table 9.3. Deviation of the development of the number of NWW persons with certain characteristics compared to the total number of NWW persons (calculated as correlation coefficients) Women Elderly (50+) Low educated (max. LBO/MAVO) Long-term unemployed (>3 years)

Average 0.89 -0.07 0.97 0.57

Highest 0.99 0.39 0.99 0.95

Lowest 0.74 -0.43 0.89 -0.16

Source: RBA/NWW

Table 9.4 presents the region or a cluster of regions with the highest and the lowest index number per socio-economic group at the end of the period (October 1999). The region of Flevoland shows high index scores for the problem groups, implying a large increase for these groups. The index has been calculated, however, on the basis of the absolute numbers over time and this may have led to misleading results for the rather particular situation in the province. Flevoland is built on new reclaimed land and shows exceptional growth rates labour force size and economic development. The high score in unemployment among problem groups is most likely related to the strong regional growth of the labour force in absolute numbers and not to a worsening of the labour market situation for particular groups. When this growth is taken into account, Flevoland’s score is not particularly exceptional for the elderly and long-term unemployed. The southern part of the province North-Holland belongs, in almost every case, to the worst performing regions. In contrast to Flevoland, the labour force hardly increases in this region. The three northern provinces, often seen as problem areas, take a reasonable position with regard to the development of unemployment among the problem groups. However, they still lag behind, but these regions closely follow the rate of national unemployment decline. Furthermore, the decrease among long-term unemployed is rather limited in Drenthe and Groningen compared to the rest of the Netherlands, while Friesland shows the opposite. The regions in the province of Brabant and the Green Heart region of Leiden form the leading regions with regard to the reduction of unemployment among problem groups. The situation in the Leiden region is less favourable for the long-term unemployed. Just behind the leading regions we find the Midden-Ijssel/Veluwe region followed by the Haaglanden region.

162 Oedzge Atzema and Jouke van Dijk Table 9.4. Regional extremes in the development of the number of non-working jobseekers per RBA-region, March 1993 - October 1998 (highest and lowest index numbers; 1993 = 100) Index figures October 1999 (March 1993 = 100) The Netherlands

Total

Women

Age 50+

Lower education

79

96

152

81

Leiden region / Brabant regions

61

Flevoland / South N-Holland

102

Leiden region / Brabant regions

72

Flevoland / South N-Holland

122

Leiden region / Brabant regions

110

Flevoland / South N-Holland

225

Duration of unemployment >3 years 106

Brabant regions

69

Flevoland / North N-Holland

305/155

Leiden region / Brabant regions

61

Flevoland / South N-Holland

101

So, it is clear that even in a small country like the Netherlands regional differences do exist with regard to the development over time for various socioeconomic groups. Figures 9.9-9.10 shows the relative position of the 18 RBAregions for 1993 and 1999 with regard to long-term unemployment. The province of Groningen shows a high score in both years. The position of the province of Friesland and the Rotterdam city region slightly improved but is still high. In the Amsterdam city region the situation even worsened over this period. In most regions the developments show similar patterns but the rate varies by region and socio-economic group. As a result, over space and time the share of the various subgroups in percentage of the total number of unemployed may show substantial changes.

Persistence of Regional Unemployment in the Netherlands 163

Fig. 9.9. Regional differences in long term unemployment (> 3 years), 1993 by RBA region (the Netherlands = 100)

164 Oedzge Atzema and Jouke van Dijk

Fig. 9.10. Regional differences in long term unemployment (> 3 years), 1999 by RBA region (the Netherlands = 100)

When we analyse the share of the various groups in total unemployment over time these figures indicate that the labour market of the province of Groningen appears in a less favourable light. Even though the unemployment stock in Groningen counted relatively few women and elderly, the average percentage of long-term unemployed still rose above the national average (23 percent) to 29 percent. By the end of 1999 39 percent of the unemployed in Groningen turned out to be unemployed for over 3 years. Thus, Groningen has to cope, more than other regions, with a sizeable hard core of unemployed who have immense

Persistence of Regional Unemployment in the Netherlands 165

difficulties finding work even, in the present situation of overall employment growth. The position of the other northern province of Friesland compares favourably with the province of Groningen. The shares for the various groups in Friesland are very close to the national average. There are a few regions with a strongly abnormal composition of the unemployed according to subgroups. Flevoland and the northern part of North-Holland, for example, show a relatively large share of women in total unemployment. A high share of unemployed elderly can be found in the Leiden region and in the province of North-Brabant. These are also the regions where total unemployment declines fastest. This concerns a relative issue: the reduction among the youngsters takes place at a fast pace, which reduces their share and increases the share of the elderly. With regard to the low educated, regional differences are rather indiscriminate. Regions in the central part of the Netherlands and the Leiden region show the lowest share of low educated unemployed. The highest shares can be found in the regions of Rotterdam, the Hague and in the province of Drenthe. The share of low educated unemployed might possibly be more related to the composition of the labour force than to the scarcity in the labour market. From the analysis by socio-economic group and region we can conclude that there is substantial regional variation in the development of unemployment over time. This conclusion implies that the share of the various types of problem groups that are most relevant for labour market policy also shows large regional variation over space and time, which in turn implies that the design and implementation of labour market policy requires a clear regional dimension.

9.6

Conclusions

In general, there is a very limited relationship between the development of national unemployment and regional variation in unemployment in the Netherlands. In periods of rising national unemployment regional variation of unemployment only partially increases. Over the period 1952 to 1989 most provinces moved closer to the regional average, with the exception of Friesland and Groningen diverged from the national average and together accounted for more than half of the average employment difference. On the other hand, reduction of national unemployment also involves decreasing absolute regional differences in unemployment rates. The analysis of the COROP regions for the exceptional decade of the 1990s shows that even though the course of unemployment percentages was not the same everywhere, the regions with high and low unemployment have roughly been the same over in the last ten years. However, on average, regional differences decreased in the same period. The regions close to the Randstad are in a relatively good position with the exception of North-Holland. All in all we can conclude that regional differences in the Netherlands are on average decreasing, but there are still regions that show a very persistent and large deviation from the national average.

166 Oedzge Atzema and Jouke van Dijk

In the year 2000 the Netherlands had a very low official unemployment rate. However, the number of persons entitled to benefits (the so-called unused supply of labour) is substantially higher and indicates that the unemployment problem is certainly not completely solved. Although regional differences in unemployment are rather small in the Netherlands, in comparison with other small European countries, the bi-polar pattern of regional unemployment (a lot of unemployment in the larger cities and in the periphery) still remains. Nevertheless, average regional unemployment disparities in the Netherlands are declining. There is a tendency for a weaker reduction in the rural areas of the north and in some other regions spread over the country. The reduction in the unemployment rate is especially strong in the south-eastern area of the Randstad. For some socioeconomic problem groups, moreover, the existing disparities between regions are more severe than for the total unemployment figures. The unemployment issue is increasingly concentrated among elderly and long-term unemployed. In the Amsterdam city region and in the provinces of Groningen and Drenthe the share of these problem groups in total unemployment is substantially higher. A considerable part of this stock of unemployed can be labelled as structural unemployment, i.e. labour supply that apparently does not fit the demand profile of employers. The extent in which this unused labour supply can be utilised to solve the problem of unfilled vacancies increasingly becomes a bottleneck for further economic growth. Even when the economy slows down, the aging of the population in the long run will lead to labour shortages. It is conceivable that within the geographical context of a small country like the Netherlands, regional variation in unemployment will fade away through the spatial flexibility of the labour force. This idea brings us to the role of commuting and labour migration with respect to the reduction of regional unemployment. In the 1990s commuting increased very rapidly, while migration, especially work related migration, remained constant and even declined. Ekamper and Van Wissen (2000) show that between 1987 and 1998 the number of commuters between COROP regions doubled to almost 1.2 billion people per day. It is quite possible that commuting disturbs the picture of regional disparities of unemployment. Nevertheless, this effect is spatially limited and occurs only if the regions are small because of the limited commuting tolerance of the working population in general and of the unemployed population in particular. Besides that, when a commuter looses his job he becomes unemployed in his region of residence, while the job remains or gets lost in the work region. In this way, unemployment in a region may rise due to economic decline in neighbouring regions. This effect has not been taken into account in this study. Looking at the decline in labour migration it is clear that this kind of labour market spatial flexibility is not well developed in the Netherlands. This fact might be a culturally inspired phenomena which has been enforced by government regional policy via subsidizing firms to bring work to people living in regions with relatively few jobs. The growing labour participation of women and the growth of double income households in the Netherlands, moreover, might have a positive effect on declining labour migration, because evaluation of the place of residence of the household has to be linked to the labour market position of at least both adults in the family. At the

Persistence of Regional Unemployment in the Netherlands 167

macro level this implies a reduction of the interregional labour migration which occurs irrespective of the size of the country. From the analysis in this chapter we conclude that even in a small country like the Netherlands regional disparities in unemployment are small but persist substantially over time. Although there is a tendency for the average difference to decline, there are still some regions that continue to lag behind. As spatial differences also vary by socio-economic groups regional and labour market policy aiming to reduce these disparities must take into account the characteristics of both the lagging regions and the specific characteristics of the unemployed people in those regions.

References Broersma L, Van Dijk J (2002) Regional labour market dynamics in the Netherlands. Papers in Regional Science 81(3):343-364 Van Dijk J, Oosterhaven J (1992) Het regionaal economische beleid: een nachtkaars voor North-Nederland? Beleid en Maatschappij 19(4):184-196 Ekamper P, Van Wissen L (2000) Regionale arbeidsmarkten, migratie en woonwerkverkeer (Regional labour markets, migration and travel-to-work). NIDI: Netherlands Interdisciplinary Demographic Institute, The Hague Elfring T, Kloosterman RC (1989) De Nederlandse “Job Machine”: de snelle expansie van laagbetaald werk in de dienstensector 1979-1986. Amsterdam EconomischGeografisch Instituut, Universiteit van Amsterdan (EGI-paper 38) EU (2001) Employment in Europe 2001. European Commission, Luxembourg Eurostat (1999) Statistics in Focus: EU unemployment still marked by regional variations. Eurostat, 5/1999 Netherlands Bureau of Economic Policy Analysis (CPB) (2000) Centraal economisch plan, Appendix B3, pp 218-219 Visser J, Hemerijck A (1997) A Dutch miracle: job growth, welfare reform and corporatism in the Netherlands. Amsterdam University Press, Amsterdam

10

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

Peter Wostner The Republic of Slovenia Government Office for Structural Policies and Regional Development, Ljubljana, Slovenia

10.1 Introduction Regional disparities received considerable attention over the last few decades both in academic as well as political spheres. Analysis and understanding of these trends has seen remarkable improvements during the 1990s especially with the development of new sub-areas like the New Economic Geography. On the other hand, regional disparities have always been important for politics, not least due to the importance of national cohesiveness and solidarity. Slovenia is no exception and has articulated its regional policy (as one of the Yugoslav republics) since the 1970s. This chapter analyses the extent of regional disparities in Slovenia and more importantly, attempts to identify the underlying causes of the dynamics in regional disparities. The Slovene case is interesting because Slovenia is a small country (2 million people; 20526 km2), a fact that might be instrumental in reducing the importance of location or suggesting that the regional analysis at such a scale does not make sense. On the other hand shorter distances could enhance the power of mutual interdependencies and have consequences for optimal regional development strategies. This study tries to capture the importance of initial conditions as these can have significant impact on the regions’ future development due to the path dependency effect. The existence of this effect would be important for policy making, since it would represent an important argument for regional policy and would also stress the middle to longer term effect of present policies on the growth potential of the regions. Regional disparities in Slovenia have been the focus of a number of studies. The 1967 IER study (IER 1967) set the analytical foundations for regional policy implemented during the 1970s. The study argued that without intervention there would be continued migration of population from less developed areas to industrial centres. The main problem of less developed areas was considered to be the structure of the economy, which called for a complex approach to regional policy involving the state, banks, firms and so on. For regional policy implementation, the national assembly requested an annual analysis of the implementation of the Social agreement of balanced regional development promotion. These studies concluded that during 1970s the less developed areas

170 Peter Wostner

grew faster then the national average and even more significantly, there were more jobs created in these areas than the Slovene average. This was primarily due to increased investment, which also contributed towards improved living standards also in peripheral areas. In this way a polycentric development pattern was successfully promoted. The economic crisis of the 1980s reduced the growth performance of the less developed areas, although in terms of employment these areas were still successful. The general index of development level put forward by Kukar et al. (1985), Kukar et al. (1989) and Kukar (1996) confirms that there was convergence at the regional level during 1970s, but that 1980s saw a worsening of the relative position of the less developed regions. After independence in 1991, regional policy was not considered a top national priority. Nevertheless, studies by Gulic and Kukar (1992; 1993) and IER 1993 made the case for the state intervention in this field due to the expected asymmetric burden of transition for the less developed, especially industrial, regions. In the beginning of the 1990s regional disparities actually diminished but then started to increase again (see Pecar various years and Kavas et al. 1998). Increasing disparities led to the passing of the new law in 1999 as well as a Regional Development Strategy in 2000, which were supported by two important analytical documents. The first was the White Paper on regional development in Slovenia (Johnston et al. 1999) and the second was the Analytical Basis for the Preparation of the Regional Development Strategy of the Republic of Slovenia (Majcen et al. 2000). These analyses concluded that regional disparities were still increasing and called for intensified regional policy, which would build on the endogenous development potential of the regions, combined with the attraction of domestic and foreign investors. The present study tries to use the econometric tools now available for regional analysis and is structured as follows. In the next section, the development context of the country is presented and the relative significance of regional disparities is considered. Then, trends in terms of convergence/divergence as well as agglomeration will be addressed. In the subsequent section the underlying growth factors are identified along with those that are disparity-enhancing and those that are disparity-reducing. Finally, conclusions and policy implications are presented.

10.2 Slovenia’s Development Context Slovenia is a small, open economy with a very short history as an independent state. With the disintegration of Yugoslavia in 1991, 2 million people founded a new sovereign country. Its neighbours are Italy and Austria in the west and north, while on the southern and eastern side, Slovenia has common border with Croatia and Hungary. Slovenia has a favourable geostrategic position both in terms of economic potential (proximity to some of the Europe’s most dynamic regions like Veneto, Trentino-Alto Adige and Steiermark), as well as in terms of its importance for transport (bridge between east and west). Slovenia is a very diverse country, which is reflected for example in different climate zones (mediterranean,

Dynamics of Regional Disparities in Slovenia 171

alpine and continental), topography, social patterns in different areas and similar. It is characterized by a large number of small settlements (approximately 6000), over 90 percent of which, do not have more than 500 inhabitants. There are only two cities with more than 100,000 inhabitants: Ljubljana, the capital (276,000) and Maribor (135,000). Furthermore, 68 percent of the populated areas have a construction density of only 1-2 buildings per ha, which causes specific problems and environmental burdens. On the other hand, 50 percent of the population live in the urban areas. Internal migration mobility is significantly lower than in the EU, even though it slightly increased in the second half of the 1990s (Draft Single Programming Document 2002:51). In 1999 5 percent of all employees were working outside the region of their permanent residence. Slovenia is the second most developed candidate country for accession to the EU after Cyprus, with GDP pc of 15,600 EUR in 2000 or 69 percent of the EU15 average. The average yearly real growth rate of GDP in the period between 1995 and 2000 was 4.3 percent, while the average level of unemployment in the same period amounted to 7.4 percent. The importance of agriculture is comparable to the EU average, since it represents only 3.3 percent of value added in the year 2000 - a fall from 5.9 percent in 1992. The share of services in value added, on the other hand, increased over the same period from 55 percent to 60.2 percent, putting Slovenia in the category of a modern, service-driven economy. According to the European Nomenclature of Territorial Units (NUTS), Slovenia as a whole is defined on the level 0 and 1. Intensive negotiations between Slovenia and Eurostat/EC have been conducted on the division on the second level, where Slovenia wanted to define 3 NUTS II regions. The EU insisted that for this programming period the country should not be divided. The issue, however, is to be re-opened before 2006. On the NUTS III level Slovenia has defined 12 so called statistical regions, which represent the basis for the implementation of national regional policy. They do not have any official political or administrative function and were defined for planning purposes back in the 1970s. The map (Figure 10.1) and basic indicators are presented in Table 10.1. Even though statistical regions have been defined on the principle of functionality, they also reflect historical and cultural backgrounds of different parts of Slovenia. Statistical NUTS III regions will represent the basis of our analysis. Note though, that prior to the year 2000 there was a slight difference in the definition of the border between Central and Southeastern region (Official Journal of Rep. of Slovenia No. 28/2000). The latter was smaller than today’s classification and was termed as Dolenjska region (106,626 inhab.), while Central region accounted for 516,997 inhabitants - both in 1999. Since this chapter looks at developments in regional disparities across the 1990s, preceding classification is used here.

172 Peter Wostner

NG HU Pomurska

Y AR

AUSTRIA

Murska Sobota

Koroska Maribor

Slovenj Gradec

Podravska Gorenjska

Savinjska

ITALY

Kranj

Celje

Zasavska

Central Slo.

Goriska

LJUBLJANA

Lower Posavska

Nova Gorica

Krško

Postojna

Novo mesto

CROATIA

Southeastern Slo.

Notranjskokraska Littoral-Karst Koper

0

10

20

30

40

50

km

Sc a le : 1 : 1 .3 0 0 .0 0 0

Fig. 10.1. Map of Slovene NUTS III regions Table 10.1. Basic indicators of the regions Region

Area (km2)

Population 1999

Southeastern Slo. Gorenjska Goriska Koroska Notranjsko-kraska Littoral-Karst Central Slovenia Podravska Pomurska Savinjska Lower Posavska Zasavska Slovenia

2675 2137 2325 1041 1456 1044 2555 2169 1338 2384 885 264 20273

137,925 196,436 119,998 74,012 50,470 103,298 485,698 319,468 125,037 256,562 70,100 46,553 1,985,557

Pop. density 1999 (pop./km2) 51.6 91.9 51.6 71.1 34.7 98.9 190.1 147.3 93.4 107.6 79.2 176.3 97.9

GDP pc PPS 1997-99; EU-15=100 61.8 62.0 67.6 58.0 57.5 69.9 88.9 55.2 51.7 61.6 56.6 54.3 68.1

Avg. yearly growth in GVA pc 1999/1990 3.5* 2.5 3.4 3.7 -1.0 2.1 2.7* 0.8 -0.3 2.2 -2.7 1.0 2.2

* Refers to the old territorial classification. Source: Statistical Office of the Republic of Slovenia (SORS hereafter), Statistical Yearbooks, Pecar, various years, author’s own calculations

Dynamics of Regional Disparities in Slovenia 173

10.3 Regional Disparities and Agglomeration Regional disparities within Slovenia are small compared to the other EU and candidate countries. In fact, when taking into account the population weighted coefficient of variation1 of regional GDP on the NUTS III level, we can see that only Sweden had smaller internal disparities than Slovenia (Table 10.2). While other candidate countries could still be found on both ends of the spectrum of the regional disparities in 1995, they have predominantly seen a move towards greater regional disparities by 1999. Slovenia was also among the countries with increasing regional disparities, yet the increase was fairly modest. These conclusions are somewhat in contradiction to the findings of the studies mentioned above due to newly available data. Nevertheless, despite the small regional disparities in absolute terms, they are perceived as problematic among politicians, the general public and in the research community. There seem to be two main arguments behind this attitude: Slovenia has been pursuing the concept of polycentric development since the 1970s, which on one hand strengthened the perception of regional disparities, but has also showed the (relative) development potential of non-central regions. This makes it clear to the average observer, that at present, regional endogenous development potential of the regions is not fully utilized. Second, regional disparities have increased more or less consistently across the 1990s, in contrast to the Slovene development strategy expressed in a number of strategic documents. On the other hand, the same trend of increasing regional disparities observed in more than half of the EU countries, implies that increasing internal disparities are likely to have at least some common denominators (for example the single European market) and will probably get still more attention in the future. The only countries where disparities diminished between 1995 and 1999 were Austria, Greece, Italy, Slovak Republic and Denmark. Since it would be preferable to analyse trends in regional disparities over the whole decade, we recalculated the “unofficial” gross value added (GVA) indicator, which is based on the balance sheets, collected by the Agency for Payment Transfers. The correlation between regional GVA pc and GDP pc for the available years varies between 0.93 and 0.94, which enables us to treat the GVA indicator as an acceptable, albeit imperfect, substitute.

1

The coefficient of variation was calculated as follows: §ª y

¦ ¨¨ « x i

c

©¬

i

i

2 yº x ·  » * i ¸ x ¼ ¦ xi ¸ ¹

¦y ¦x

i

i

where y and x stand for GDP in PPS and population, while i refers to region.

174 Peter Wostner Table 10.2. Comparison of the internal regional disparities on the NUTS III level; population weighted coefficient of variation of regional GDP 1995

1999

Change - %

Latvia

Country

29.0

53.5

84.5

2.

Poland

39.3

50.9

29.5

3.

United Kingdom

49.3

50.1

1.6

4.

Hungary

40.5

46.0

13.6

5.

Estonia

39.3

45.2

15.0

6.

Denmark

42.3

44.0

4.0

7.

France

41.2

41.3

0.2

8.

Czeck Rep.

31.0

40.7

31.3

9.

Belgium

38.4

39.5

2.9

1.

10.

Portugal

36.6

37.3

1.9

11.

Slovak Rep.

37.9

37.3

-1.6

12.

Bulgaria

33.2

33.3

0.3

13.

Italy

29.7

29.2

-1.7

14.

Denmark

27.8

27.6

-0.7

15.

Austria

29.3

27.4

-6.5

16.

Finland

21.6

25.7

19.0

17.

Romania

20.5

24.7

20.5

18.

Lithuania

13.0

23.8

83.1

19.

Spain

22.2

23.7

14.5

20.

Ireland

20.7

23.7

6.8

21.

The Netherlands

20.4

22.4

9.8

22.

Greece

23.3

22.0

-5.6

23.

Slovenia

19.4

20.4

5.2

24.

Sweden

13.0

16.5

26.9

Source: Regions: Statistical Yearbook 2002, European Communities 2002

Measuring V -convergence by unweighted cross-sectional standard deviation of the log GVA pc, we can undoubtedly say, that the regional differences in terms of economic activity per capita have increased during the 1990s - see Figure 10.2. Interestingly, this does not really hold for the first two years, which were still characterized by sharply falling aggregate economic activity. With the revival of the economy, regional differences started to increase. This effect, however, seemed to have run out of steam by the end of the period, in spite of the continuing aggregate growth. This observation however, should be complemented with available regional GDP pc data for the 1995-1999 period. According to this superior indicator, the trend was consistent with the GVA pc in the years

Dynamics of Regional Disparities in Slovenia 175

1995/1996, but then shows continuing increases in regional disparities over 19971999 period. 0.585 STD_GVA (L) STD_ITAX (L) GINI_GVA (R)

0.30

0.575

0.565 0.26 0.555

0.22

0.545

0.535 0.18

Gini coefficient for GVA

Stdev. of log GVA pc & log Inc. tax base pc

0.34

0.525 0.14 0.515

0.10

0.505 1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

Fig. 10.2. Standard deviations of log GVA per capita (STD_GVA - left scale) and log Income tax base per capita (STD_ITAX - left scale) and Gini coefficients for GVA (GIN_GVA- right scale) Source: Author’s own calculations using Pecar, various years, SORS, Statistical Yearbooks, various years

It is worth noting that while the regional disparities in terms of economic activity have been increasing, we cannot say the same for the personal income, measured with the personal income tax base per capita. While we can observe small fluctuations, the overall level of dispersion has not changed. This might reflect the high priority given to social stability through different “state insurance policies”. Apart from the above-mentioned regional disparities, one would expect the economic activity to increasingly agglomerate as well - especially in the central, metropolitan region. From the Gini coefficient for regional GVA we see that agglomeration dynamics have been broadly following disparities in GVA pc. The fall in agglomeration until 1992 was much steeper and resulted in only marginally increased agglomeration by 1997 compared to 1990. After that year, the trend actually shows falling agglomeration. At the end of the 1990s the trend is quite surprising and could be the effect of some sectors not being included in the aggregate data (like banking and insurance). Nevertheless, the level of agglomeration does not seem to increase to the extent that one would expect. Finally, has the focus of economic activity agglomeration been oriented towards

176 Peter Wostner

the centre? The greatest absolute increase in the region’s share of total Slovene GVA has indeed been in the Central region. In relative terms however, the Central region has not been the one, that benefited most. In fact the Dolenjska, Goriska and Koroska regions were the main beneficiaries, which at the first sight do not have anything particular in common - not in terms of their geographical position (position relative to Ljubljana or common border with the EU), nor in terms of economic structure. We will try to identify the underlying determinants of their success in the next section. Finally, a popular indicator of the regional growth is the E convergence estimator. The estimation of the convergence equation as proposed by Barro and Sala-i-Martin (1991)2, points to a totally insignificant (p=0.52) E estimator of 0.012. According to the test, there is no empirical support for the hypothesis, that the poorer regions in 1990 were growing faster than the richer ones over the 1990 - 1999 period, which can also be seen from the Figure 10.3.

Average annual growth rate in GVA pc 1999/1990

5 4

Koroska

Goriska Dolenjska

3

Central Slo. Gorenjska

SLO

Savinjska

Littoral-K.

2 Podravska

1 0

Zasavska

Pomurska Notranjsko-kr.

-1 -2 Lower Posavska

-3 -4 40

60

80

100

120

140

160

GVA pc 1990 (SLO=100)

Fig. 10.3. The average annual growth rate in GVA pc 1990-1999 estimated with OLS and GVA pc in 1990 (SLO=100) Source: Author’s own calculations using SORS, Statistical Yearbooks, various years, Pecar, various years

In terms of growth performance we can again see the strong performance of the two industrial regions, Zasavska and Koroska regions, which is not consistent with the EU experience, where industrial regions have been strongly under-performing (Rodríguez-Pose 1998). This is even more surprising when taking into account 2

The ‘shock variables’ that were originally included as the controlling variables and form part of the residual were not taken into account.

Dynamics of Regional Disparities in Slovenia 177

that Koroska is also a peripheral region, albeit with a common border with Austrian region of Koroska. The second best performer, the Dolenjska region, is on the other hand furthest away from the EU border, but borders the Central region, which makes the possibility of industrial decentralization as observed in Europe more likely (Cheshire and Carbonaro 1996). The same option should be available to Notranjsko-Kraska region however, which was the second worst performing region.

10.4 What Determined the Growth Performance of the Regions? In considering the causes for increasing regional disparities, we assume that the same factors are at play, as in large countries. Two further issues need, however, to be addressed. The first relates to whether geography can be verified as an important force on such a small scale as the Slovene NUTS III regions. From this perspective it is important to mention that Slovenia and its regions are infrastructurally well connected, which at least in principle, should reduce the importance of location. Second, special attention should be given to the role of initial conditions. As already mentioned, Slovenia conducted a complex regional policy in the period between 1971 and 1990, which resulted in more scattered industry location across its regions. Regional policy was implemented as a coordinated quasi-contract based effort with a whole range of partner-institutions, ranging from central and local governments, trade unions, Chambers of Commerce, banks and other so called “self-management interest groups” in charge of road infrastructure, agriculture, education, culture, research, health. This could have had important consequences for the development potential of the regions due to a path dependency effect. We use two econometric analyses for looking at the determining factors for regional GVA pc presented as (1) a panel data regression and (2) a pooled regression model, both for the 1990-1996 period. The fixed effects panel data regression is defined as follows (see Equation 10.1): lnGVAit = Di + E1 UNEMit + E2 SPINDit + E3 SPAGRit + E4 AIRDTit + E5 HIRSCHit + Ot + Hit

(10.1)

where lnGVAit stands for the logarithm of regional GVA per capita in region i in time t, expressed in million SIT (1992 constant prices). UNEMit is the regional unemployment rate. On the one hand this is expected to reduce the GVA pc due to lower employment ratio. On the other hand, increased unemployment can also control for the speed of restructuring that should bring about an increase in the GVA. SPINDit and SPAGRit stand for relative specialization in industry and agriculture, respectively. They are calculated as the ratio between the share of industry (agriculture) employment in the region relative to the same share for

178 Peter Wostner

Slovenia as a whole3. We would expect both variables to have negative impact on GVA pc. HIRSCHit is a measure of aggregate (and absolute) specialization expressed by the inverse Hirschman-Henfirdahl index4. An increase in this measure refers to greater diversification, which means that we would expect positive sign of the regression coefficient. Ot are the year dummies, while AIRDTit is the average inter-regional daily traffic of vehicles (cars, vans, buses, trucks, etc.). This means that AIRDT captures: (1) the movement of people in the sense of movement of purchasing power (shopping motive) and in the sense of mobile production factor (commuting for work); (2) the movement of goods, both end products and intermediate goods, which reflect business cooperation of the regions and finally (3) due to transit traffic, the AIRDT is a measure of accessibility and centrality, since more central and accessible regions, ceteris paribus, always reflect significantly greater volume of transit traffic. In this sense, the AIRDT is an instrument variable for economic potential as well as cost and demand effects. This reflects the extent of mutual cooperation of the regions. We would expect AIRDT to have a positive impact on GVA pc. Before going to the model results, very high and significant auto correlation between variables should be noted. Services seem to have very strong tendency to locate in the central regions (correlation between services specialization and AIRDT is 0.73, p=0.000) and these regions seem to be highly specialized (correlation between services specialization and HIRSCH is -0.71, p=0.000). Noncentral regions, on the other hand, show an industrial or agricultural orientation, usually connected with greater degree of diversification. In order to avoid multicollinearity as well as heteroskedasticity problem among these variables5, the relative specialization variables were taken as the basic variables. The average inter-regional daily traffic (AIRDT) was subsequently included as a residual of the unexplained variance by SPIND and SPAGR, while the aggregate specialization variable (HIRSCH) only captures the variation, that is not explained by SPIND, 3

The ratio is calculated as follows: Eij

6 j E ij 6 i Eij

sij

6 i 6 j E ij

4

where s refers to the relative specialisation index, E stands for employment, while i and j stand for region and sector respectively. It should also be mentioned that services, industry and agriculture relative specialization were all analysed and considered for inclusion in the regression. Since the last of the three relative specialisation variables is the residual of the first two, we only include relative industry (SPIND) and relative agriculture (SPAGR) specialization in the model due to their lowest auto correlation. The inverse Hirschman-Henfirdahl index is calculated as 1/

¦s

2

ij,

j

5

where s refers to the proportion of employment in region i in sector j. In particular among the following variables: SPIND, SPAGR, AIRDT and HIRSCH.

Dynamics of Regional Disparities in Slovenia 179

SPAGR and residual AIRDT variables. The regression results are presented in Table 10.3. It turns out that relative industry specialization has a positive and significant effect (E=0.005; p=0.027), while relative agriculture specialization has a negative but insignificant influence on GVA pc. Taking into account that services seem to have strong tendency to locate in the central regions, the best development strategy for non-central regions seems to be industry specialization. This is especially beneficial strategy for the regions with below average industry specialization, since residual variation of the aggregate specialization variable (HIRSCH) shows, that increasing diversification has an additional positive impact on GVA pc (E=0.644; p=0.078). A combination of these factors seems to explain an important part of the surprising performance of the two industrial regions mentioned in the previous section. Looking at the sectoral structure of GVA pc, however, we can see that primarily wholesale, retail, real estate, renting and other business services, education, public administration and business intermediation have the strongest concentration tendency. Even though these services all have very high GVA, the increased (aggregate) specialization towards the services segment unfortunately seems to be available only to the central region(s). Geography in this way seems to predetermine the choice of regions’ development strategy. Finally, the average inter-regional daily traffic variable shows that on top of sectoral specialization, the extent of cooperation with the other regions and improved accessibility is of great importance. Table 10.3. The determinants of the regional GVA pc; Equation 10.1 - the fixed effects model; dependent variable lnGVApc Variable

Coefficient

Std. Error

UNEM

0.0030

0.0077

SPIND

0.0055**

0.0024

SPAGR

-0.0010

0.0007

AIRDT

7.80e-06***

2.47e-06

0.6443*

0.3607

HIRSCH 2

R

within

0.4132 F(11,61) = 3.90

No. of obs.

Prob > F = 0.0003

84

Note: ***,**,* statistically significant values at 1 percent, 5 percent and 10 percent on a two tail test, respectively. F test (H0: all Ui=0): F(11,61)=10.83*** White Heteroskedasticity Consistent Standard Errors and Covariance

The panel data regression is able to explain 41 percent of the within variation. This is not entirely satisfactory, as is the inability to include other individual period-invariable factors, which would also capture initial conditions. For these

180 Peter Wostner

reasons the following variables were considered in the pooled regression framework which captures other socio-economic characteristics of the regions as well elements particular to Slovene case6: x OBS90i - index of the share of obsolete industrial sectors in total employment in 1990, as defined by Gulic and Kukar (1992); SLO=100; (-); x EXP90i - index of the region’s share in total turnover realized on foreign, nonYugoslav markets in 1990; SLO=100; (+) x EXPYUi - the dependency upon the demand from the other Yugoslav republics7 in the period, when the conflict escalated - 1991/1992; (-) x COMPSi - the number of small enterprises per 1000 inhabitants in 1990; (+) x COMPMBi - the number of medium and large sized enterprises per 1000 inhabitants in 1990; (+) x EDUPOPi - the share of highly educated population among all inhabitants aged over 15 years in 1991; (+) x RDEMBi - the R&D personnel in the private sector per 1000 inhabitants averaged over 1995-1997 period; (+) x RDEMPi - the R&D personnel in the public sector per 1000 inhabitants averaged over 1995-1997 period; (+) x FDIi - the number of employees per 1000 inhabitants working in the firms with more than 10 percent foreign ownership in 19988; (+) x DENSi - the average population density per square kilometre in the 1990-1996 period9; (+). Due to multicollinearity among the above variables as well multicollinearity of these variables with the ones used in the fixed effects model principal components analysis on the standardized variables was initially performed. This enabled us to define multiple orthogonal factors, which capture almost 97 percent of the variance of the original variables - for the factor loadings matrix see Appendix. The pooled regression model was then defined as follows (see Equation 10.2): lnGVAi = D + E1 F-PROD-ORIENTi + E2 F-PATHDEPi + E3 F-OPENNESSi + E4 F-RDEMBi + E5 F-AGGLOMi + E6 F-YUDEPi + E7 F-UNEMi + E8 F-RDEMPi + E9 TRANS93 + Hi 6

7

8

9

(10.2)

These variables were considered additional to the ones already included in the fixed effects model. At the end of each line is the expected impact of the variable on GVA pc. The data on the share of turnover realised on foreign markets allows the estimation of the regional dependency on Yugoslav markets, since these markets were not counted as foreign before 1992. The EXPYU variable is expressed as the regional change in export turnover relative to aggregate change between 1991/1992 - SLO=100. Due to limited data availability I had to assume that the last three stated independent variables are time invariant, which is admittedly a strong assumption and represents serious danger of endogeneity. This is why additional model versions were subsequently tested in order to evaluate this potential problem. Population density was not defined as panel variable because the inter-regional mobility of the population was very low.

Dynamics of Regional Disparities in Slovenia 181

where TRANS93 stands for a transition dummy variable (1990-1993 = 1, 0 otherwise). As can be seen from the Table 10.4, the overall fit of the pooled regression model is greatly improved and explains over 80 percent of the variance in the GVA pc. The strongest impact on the GVA pc comes from factor 1, which captures the general regional production orientation - specialized in services with small business enterprise culture and high human capital requirements vs. more industrial production type. Obviously, the regions from the former group are the better-off regions. These regions are more specialized and have high average interregional daily traffic indicators - they are in other words central and highly accessible. These results are compatible with the hypothesis from the fixed effects regression, that there is a trade-off in terms of production strategies of the regions, which among others, is also determined by Slovenia’s geography. The second most important determining factor is the one capturing R&D employment in the private sector (factor 4) and to a lesser extent R&D employment in the public sector (factor 8) (Table 10.4). Table 10.4. The determinants of the regional GVA pc; Equation 10.2 - pooled regression model; dependent variable lnGVApc Factor

Variable

Coefficient

Std. Error

D

-0.8163***

0.0205

Factor 1

F-PROD-ORIENT

0.1101***

0.0119

Factor 2

F-PATHDEP

0.0631***

0.0120

Factor 3

F-OPENNESS

0.0900***

0.0190

Factor 4

F-RDEMB

0.1031***

0.0120

Factor 5

F-AGGLOM

0.0742***

0.0120

Factor 6

F-YUDEP

-0.0041

0.0120

Factor 7

F-UNEM

0.0683***

0.0144

Factor 8

F-RDEMP

0.0690***

0.0119

Trans93

-0.0981***

0.0293

R2

0.8246 2

adj. R

0.8033

Prob > F

0.0000

No. of obs.

F(9,74) = 38.66

84

Note: ***,**,* statistically significant values at 1 percent, 5 percent and 10 percent on a two tail test, respectively.

Openness of the region (factor 3) also has a very significant positive impact on the GVA pc. This is expressed as openness: on one hand to FDI’s and on the

182 Peter Wostner

other, the regional market orientation towards foreign, predominantly western, markets. This is captured by the proportion of turnover realized on the nonYugoslav markets in 199010. The positive impact of the FDI’s comes as no surprise, while the latter confirmed, that the stronger presence of the firms in nonYugoslav markets requires greater competitiveness and gives a head-start with further market penetration, due to the established knowledge and distribution networks in these markets. The pooled regression model also confirms the positive impact of agglomeration economies measured by the average population density per square kilometre in the 1990-1996 period (factor 5), and the importance of the path dependency effect (Table 10.4). The latter is reflected via the location of medium and large sized enterprises per 1000 inhabitants in the 1990 (factor 2), which were fairly scattered due to extensive regional policy. This result confirms that existing firms positively contributed to GVA pc by retaining financial and especially human capital in the regions, while at the same time they mitigated the development of entrepreneurial culture and social capital. Ready access to information and know-how empowered these localities to respond to the transition process and the changed socio-economic environment with much more focus, leadership and flexibility. The only insignificant factor was the one, capturing the dependency on the Yugoslav markets during the conflict escalation (factor 6). This can be attributed to Slovenia’s very successful market re-orientation to the EU markets. The effects of the Yugoslav market collapse were probably averaged out. Finally, the pooled regression also included a dummy variable for the years 19901993, which were characterised by falling aggregate economic activity. This variable was meant to test the existence of the so called transformational recession, a concept put forward by Kornai (1995). The regression coefficient confirmed (Trans 93), that “when the post-socialist economy transfers from ... a sellers’ market, to the ... buyers’ market, it tips over too far, instead of arriving at an ideal state of equilibrium” (Kornai 1995:174), where aggregate supply equals aggregate demand.

10.5 Conclusion and Policy Implications According to Cheshire and Carbonaro (1996), convergence or divergence in a particular time period is the outcome of the net balance of the forces, which produce either of the two. Comparing the dynamics of individual variables with regions’ GVA pc from the beginning of the analysed period - 1990, we are able to 10

It should be noted that models with excluded business R&D employment as well as FDI variables were tested due to a potential endogeneity problem. The factors obtained were naturally somewhat changed. But the sign and significance of the initial variables remained robust. The only exceptions were (1) unemployment, which becomes insignificant and (2) population density, which in the model with both excluded variables gets connected with export orientation and therefore seems to have a negative influence on GVA pc.

Dynamics of Regional Disparities in Slovenia 183

identify, which variables fall into which group. Among the most obvious disparity-enhancing factors are the share of highly educated population (EDUPOP), the number of small enterprises (COMPS) and R&D personnel in the public sector (RDEMP). Somewhat less straightforward, but still with a clear central orientation are FDIs and R&D personnel in the private sector (RDEMB). It is of particular interest and importance that these two factors are not as concentrated as one might expect - in other words, they are not strictly selective on the basis of centrality or proximity to the EU border. This can be seen in particular in the intermediate regions of Dolenjska (whose successful performance was mainly based on the FDIs and private R&D) and Gorenjska (private R&D and human capital). The trends in the future will show whether this is only a temporary “anomaly” due to the privatisation process and path dependency effect. Our analysis nevertheless identified one variable that actually contributes to convergence - relative specialization in industry, whose effect was particularly strong when combined with the aggregate diversification of the region. Average growth of both variables actually tends to be higher in the worse-off and the more peripheral regions. Furthermore, according to the data geography does seem to play a very positive and significant role in determining regional GVA pc even in a small country like Slovenia. So, what can we learn from the Slovene case and the above analysis? First, location of the region should be considered as one of the crucial elements in the designation of a region’s development strategy. This is due to the importance that location combined with the quality of infrastructure has on the regions’ accessibility. Even more importantly, there seems to be very strong interdependence between a region’s location and the productivity of different economic sectors. Since certain high value added services have a very strong tendency to locate in central region(s), the non-central regions seem to be better off, if they opt for an industrial specialization option. The specialization in services is simply not an available option to these regions, due to their location. A second policy implication is that in order to mitigate increasing regional disparities, particular focus should be given to the development of industry, predominantly manufacturing, in less-developed regions, which would in fact also improve the aggregate growth potential of Slovenia. Finally, the analysis shows that initial conditions have strong influence on the development performance of the region. This would imply good grounds for continuing or strengthening direct regional policy.

184 Peter Wostner

Appendix Table 10.5. The factor loadings after varimax normalized rotation of the principal components analysis

UNEM SPAGR SPIND SPSER HIRSCH AIRDT OBS90 EXPYU EXP90 COMPS COMPMB EDUPOP RDEMB RDEMP FDI DENS Expl. Var. Prp.Totl

Factor 1

Factor 2

Factor 3

Factor 4

Factor 5

Factor 6

Factor 7

Factor 8

F-PRODORIENT

FPATHDEP

F-OPENNESS

FRDEMB

FAGGLOM

FYUDEP

FUNEM

FRDEMP

0.10

-0.14

-0.08

-0.07

0.16

0.09

-0.96

-0.01

-0.18

-0.54

-0.35

-0.42

-0.56

0.09

-0.05

0.12

-0.96

0.12

-0.10

0.13

-0.02

0.10

0.04

-0.09

0.96

-0.03

0.15

-0.06

0.09

-0.11

-0.07

0.08

-0.66

-0.46

-0.28

0.02

-0.42

0.19

0.08

0.08

0.75

0.03

0.01

0.44

0.02

-0.05

-0.12

0.40

-0.57

0.63

-0.21

0.14

0.29

0.20

0.01

-0.14

-0.26

-0.17

0.05

-0.10

0.03

0.91

-0.08

-0.03

-0.03

-0.11

0.61

0.23

-0.49

-0.49

0.11

-0.15

0.59

0.40

0.30

0.39

0.02

0.02

0.13

0.42

0.08

0.92

0.04

0.08

0.02

-0.20

0.16

0.19

0.81

0.30

0.12

0.25

0.11

-0.13

0.04

0.35

-0.03

0.12

0.19

0.94

0.02

-0.12

0.08

0.13

0.49

0.11

0.01

0.16

0.22

0.01

0.01

0.82

0.36

0.08

0.88

0.19

-0.04

0.12

0.07

0.09

0.09

0.07

-0.15

0.02

0.94

0.07

-0.19

0.18

4.67

2.10

1.59

1.65

1.79

1.28

1.06

1.30

0.29

0.13

0.10

0.10

0.11

0.08

0.07

0.08

References Barro RJ, Sala-i-Martin X (1991) Convergence across states and regions. Brookings Papers on Economic Activity 1:107-182 Cheshire PC, Carbonaro G (1996) Convergence-divergence in regional growth rates: an empty black box? in Armstrong HW, Vickerman RW (eds) Convergence and divergence among European regions. Pion Limited, London Gulic A, Kukar S (1992) Regional development and regionalisation of Slovenia: analysis of the situation with suggestions for change. 2nd phase (in Slovene). The Urban Planning Institute of the Republic of Slovenia and Institute for Economic Research, Ljubljana Gulic A, Kukar S (1993) Regional development and regionalisation of Slovenia: analysis of the situation with suggestions for change. 3rd phase (in Slovene). The Urban Planning Institute of the Republic of Slovenia and Institute for Economic Research, Ljubljana

Dynamics of Regional Disparities in Slovenia 185 IER (1967) Socio-economic background, problems and basic development orientations of the long term regional development of SR Slovenia (in Slovene). Institute for Economic Research, Ljubljana IER (1993) State’s role in balanced regional development promotion (in Slovene). Institute for Economic Research, Ljubljana Johnston S et al. (1999) White paper on regional development in Slovenia. Institute for Economic Research, Ljubljana Kavas D et al. (1998) Theses for the regional development strategy of the Republic of Slovenia (in Slovene). Institute for Economic Research, Ljubljana Kornai J (1995) Highways and byways; studies on reform and post-communist transition. The MIT Press, Cambridge Kukar et al. (1985) Regional aspects of development of SR Slovenia (in Slovene). Institute for Economic Research, Ljubljana Kukar et al. (1989) Balanced regional development promotion in Slovenia - results, problems and future orientations (in Slovene). Institute for Economic Research, Ljubljana Kukar S (1996) Regional policy in Slovenia. Results, problems and alternatives. Eastern European Economics, July-August: 41-103 Majcen B et al. (2000) Analytical basis for the preparation of the regional development strategy of the Republic of Slovenia (in Slovene). Institute for Economic Research, Ljubljana Pecar J (various years) Regional perspectives of Slovene development with the emphasis on firms’ financial results (in Slovene). Working paper, Institute for Macroeconomic Analysis and Development, Ljubljana Rodríguez-Pose A (1998) Social conditions and economic performance: the bond between social structure and regional growth in Western Europe. International Journal of Urban and Regional Research 22:443-459

11

Interregional Disparities in Israel: Patterns and Trends

Boris A. Portnov Department of Natural Resources and Environmental Management, University of Haifa, Israel

11.1 Introduction At the end of the 1960s, Israel joined the exclusive club of the world’s most developed countries. On the average, the country’s per capita GDP grew by some US$ 350 annually, reaching US$ 19,700 in 2003 (ICBS 2002; CIA 2003). When development occurs, it is almost unavoidably localized geographically: major cities, the main loci of capital and human resources, pick up the development pace first, whereas smaller localities and peripheral areas often lag behind. Spatial inequality is thus a side effect of any development. Viewing these two trends together (rapid economic growth and the spatial concentration of development), it is unsurprising that regional development in Israel is considerably uneven. The present study attempts to answer two main questions: What are the general patterns of interregional inequalities in Israel? Do they change over time? The underlying assumption of this chapter is that notwithstanding the presence of regional disparities resulting from historical reasons, culture, government policies, ecology etc., the effects of location may explain much of the difference in development rates among different regions. The chapter starts with a brief overview of previous studies of interregional inequalities in Israel, followed by an in-depth analysis of temporal changes observed in them. The aim of the latter analysis is to determine whether interregional inequalities in Israel have tended to converge or to diverge over time. Throughout, it is assumed that different measures of inequality, such as education, income, ethnic composition, etc. may result in different evaluations of spatiotemporal changes in development patterns. To test this assumption, we employ factor analysis to determine general (or underlying) patterns of interregional inequalities. In the concluding section of the paper, the results of the analysis are summarized in brief, and development strategies designed to reduce the extent of interregional inequalities are proposed.

188

Boris A. Portnov

11.2 Interregional Inequalities in Israel: Background Studies Since the foundation of the State of Israel in 1948, numerous attempts have been made to evaluate changes in interregional inequalities, and the effects of development policies on various aspects of the country’s development (DrabkinDarin 1957; Gradus and Krakover 1977; Soen 1977; Shachar and Lipshitz 1981; Sonis 1988; Shefer 1990; Kipnis 1996; Lipshitz 1996; Krakover 1998; Anson 1993; Portnov 1999; Portnov and Pearlmutter 1999). These studies have been focused on two distinctive aspects of interregional disparities: a) population distribution and b) regional economic development. Drabkin-Darin (1957) analysed changes in the geographic distribution of the country’s population between 1948 and 1955. He concluded that the country’s population appeared to shift towards the periphery, particularly to the south. The population of this region increased between 1948-55 by 1,130 percent (though from a very low base of some 21,000 residents). During the same period, the population of the core districts of the country (Haifa, Tel Aviv and Jerusalem) increased by less than twofold. Based on these data, he concluded that substantial convergence in the population distribution had been achieved. Soen (1977) reached a similar conclusion in another study carried out 20 years later, which tracked the location of the centre of the gravity of the Jewish population. He concluded that between 1948 and 1967, this centre of gravity had shifted some 11 km southwards. He attributed this change to the establishment of new towns and villages in the periphery, and interpreted it as a clear sign of ongoing population dispersal. Shefer (1990), however, questioned whether any substantial redistribution of population and development had actually been achieved in Israel. In particular, he argued that though the percentage of those residing in the core Tel Aviv district fell sharply from 1948 to 1983 (from 37.7 to 24.8 percent), the combined percentage of those in Tel Aviv and in the adjacent Central district decreased only marginally, from 49 percent in 1948 to 45.4 percent in 1983. He also pointed out that since Greater Tel Aviv was gradually expanding in area, even this marginal decrease of population within the static boundaries of the Tel Aviv district might have been misleading. Shachar and Lipshitz (1981) analysed regional inequalities in Israel during 1962-1976, using two measures of welfare: motorization rates and personal income levels. This analysis indicated that regional inequalities tended to increase over time, a trend that the authors attributed to the backwash effects that occurred in the country’s major metropolitan areas. In a later study, Lipshitz (1996), using the same indicators, came, somewhat surprisingly, to the opposite conclusion. According to his analysis of data for 1983-1995, inequalities in the rate of motorization and income per employee tended to diminish over time. The values of these two indicators were plotted using a spatial cross-section of sub-regions, from the Hula Basin in the north to the Southern Arava in the south. This representation of economic development in

Interregional Disparities in Israel: Patterns and Trends 189

Israel resembled an inverted V-function, often observed in empirical social studies: The highest level of development was observed in the Tel Aviv region, declining with distance to the north, south, and east. Gradus and Krakover’s (1977) study of the effect of government policy on the spatial structure of manufacturing and employment indicated the opposite: changes in the location quotient of the peripheral districts of Israel in the 1960s early 1970s clearly showed a more even distribution of manufacturing employment across the country’s various geographic areas. The substantial differences among these findings may have two possible explanations: x First, convergence and divergence in interregional inequalities may have alternated over time in Israel, as in various other developed countries (see inter alia Williamson 1965; Mera 1995). However, this does not resolve the apparent contradictions among studies dealing with similar time-periods; x Second, the discrepancies among findings may be attributed to the use of different indicators for assessing interregional disparities (population distribution, unemployment, wealth, etc.), and to the diversity of quantitative indices employed - absolute change, percent share, Williamson coefficients, etc. Thus different indicators may highlight different trends, even when the same time periods are considered. For instance, the concentration of population may be accompanied by a decrease in interregional disparity in personal income. In the following sections, these suppositions will be tested using the data available in five subsequent censuses of population and housing (1948, 1961, 1972, 1983, and 1995).

11.3 Inequality Variables and Data Sources 11.3.1 Data Selection Twelve indicators were selected for the analysis. These indicators cover most major aspects of socio-economic development, ranging from population distribution to employment, demographic composition and welfare: x x x x x x x x x

POPDENS: Population density [persons per km2 of land area]; HDENSITY: Housing density [persons per room]; HOWNER: Homeownership level [percent of households]; CAROWN: Car ownership level [private cars per 1,000 residents]; PCOWN: Ownership of personal computers [percent of households]; WAGES: Average household wages and wages per employee [US$]; LFORCE: Participation in labour force [percent of population]; HIGHED: Higher education [percent of population with academic degrees]; UNSKILLED: Unskilled workers [percent of population].

190

Boris A. Portnov

x YSTUDY: Average years of schooling; x MAKEUP: Proportion of Asia and Africa born - 1st generation [percent of population]; x HSIZE: Average household size [persons]. The proportion of Asia and Africa born (MAKEUP) is a measure that is specific to Israel, and deserves some explanation. After the establishment of the State of Israel in 1948, the country absorbed a large number of Jewish immigrants from countries in North Africa and the Middle East, more than doubling the total population in less than a decade. Most of these immigrants found it difficult to become accustomed to life in the Western society established by the founders of modern Israel. Consequently, they comprised a disproportionate part of the underprivileged socio-economic strata (Drabkin-Darin 1957; Isralowitz and Friedlander 1999). Geographically, the distribution of this category of immigrants was, from the outset, uneven. In the late 1940s - 1950s, many new immigrants were directed en masse towards peripheral areas of Israel, to implement a national policy of population dispersal (Portnov and Etzion 2000). Since the early 1970s, government policy evolved to include various incentives designed to encourage the growth of the so called “development towns” indirectly. These included government loan guarantees, tax exemptions, and the provision of public housing (Lipshitz 1996; Isralowitz and Friedlander 1999). The subsequent changes in the ethnic composition of various geographic areas of the country may thus become a source of valuable information concerning the trends of spatial redistribution of these groups of immigrants and their integration into Israeli society. Some of the above indicators are used in the analysis as they appear in the census database, viz. homeownership, years of schooling, taxable income, etc., while others were derived by calculation. For instance, car ownership [per 1,000 residents] was calculated using two inputs - the average number of cars per family (C), and family size (F): (C/F)*1000. Population density [persons per km2] was derived from the overall population of a region and its total land area. 11.3.2 Spatial Units for the Analysis Most previous studies of interregional inequalities in Israel have been restricted to administrative districts and sub-districts of the country (Drabkin-Darin 1957; Shachar and Lipshitz 1981; Shefer 1990; Kipnis 1996; Anson 1993; Portnov 1999). In part, such a high level of spatial aggregation is attributed to the availability of the results of the population counts in the official publications of the Israel Central Bureau of Statistics: Statistical Abstract of Israel; Census of Population and Housing, etc. In these publications, many essential data are provided only at the levels of administrative districts and sub-districts. In contrast, the present analysis deals with a finer spatial grain - so-called “natural regions”.1 1

Natural regions (NRs) is the term used by the Israel Central Bureau of Statistics to describe the smallest statistical divisions of the country for which comparable inter-

Interregional Disparities in Israel: Patterns and Trends 191

As of 1995, there are 51 such regions (Figure 11.1; Table 11.1), which are aggregated in 17 sub-districts and 8 administrative districts of the country. 291 211 244

Haifa District

212

245 243 246 242

213

241 222 237 221 321 235 322 233 234 236 232 323 231 324 311

292

Northern District 293

294

411 412 511 421 711

Tel Aviv District

521422 531

Central District 442

431

441

613

112

611 614

Gaza Area

Judea and Samaria

111 Jerusalem District

612

621 811 622

623 1 624 Southern District

626 0

10

30

50 km

Fig. 11.1. Natural regions and administrative districts of Israel (as of 1995) 627 625

For the list of natural regions, see Table 11.1. Each number on the diagram corresponds to the code of a natural region in the Census registry. The dotted line marks the cross-section of regions, used for the study of geographic disparities of average income.

census data are available. NRs do not necessarily imply geographic cohesion or welldefined topographic borders.

192

Boris A. Portnov

Table 11.1. Districts and natural regions of Israel (as in 1995 Census) Code 111 112 211 212 213 221 222 231 232 233 234 235 236 237 241 242 243 244 245 246 291 292 293 311 321 322

District Jerusalem " Northern " " " " " " " " " " " " " " " " " " " " Haifa " "

Natural region Judean Mountains Judean Foothills Hula Basin Eastern Upper Galilee Hazor Region Kinerot Eastern Lower Galilee Bet She’an Basin Harod Valley Kokhav Plateau Yizre’el Basin Yoqne’am Region Menashe Plateau Nazareth-Tir’an Mountains Shefar’am Region Karmi’el Region Yehi’am Region Elon Region Nahariyya Region Akko Region Hermon Region Northern Golan Middle Golan Haifa Region Karmel Coast Zikhron Ya’akov Region

Code 323 324 411 412 421 422 431 441 442 511 521 531 611 612 613 614 621 622 623 624 625 626 627 711 811

District Haifa " Central " " " " " " Tel Aviv " " Southern " " " " " " " " " " Judea and Samaria Gaza

Natural region Alexander Mountains Hadera Region Western Sharon Eastern Sharon Southern Sharon Petah Tiqwa Region Lod Region Rehovot Region Rishon LeZiyyon Region Tel Aviv Region Ramat Gan Region Holon Region Mal’akhi Region Lakhish Region Ashdod Region Ashqelon Region Gerar Region Besor Region Be’er Sheva Region Dead Sea Region Arava Region Northern Negev Mountain Southern Negev Mountains Judea and Samaria Gaza Area

11.3.3 Time-Related Changes of Inequality Patterns Temporal changes in the patterns of interregional inequalities were studied using data from different censuses. In particular, the data from the 1995 Census were compared with corresponding indices of interregional inequalities calculated from data provided in the previous population counts: 1948, 1961, 1972, and 1983. However, the inter-censual comparison could not be fully inclusive since the pre-1983 censuses do not report all of the socio-economic indicators found in later counts at the spatial level of natural regions. For example, in subsequent population censuses the number of the reported socio-economic measures increased, few indicators were reported in all of the population counts held since 1961. These include average household size, average years of schooling, overall population density, and participation in the labour force. Due to this restriction on

Interregional Disparities in Israel: Patterns and Trends 193

the availability of data, only a partial analysis of the respective time-related trends was possible. Finally, two indicators - ownership of personal computers and ownership of air conditioners - were reported only in the 1995 Census of Population and Housing. These indicators may thus be used only to illustrate existing patterns. Data from forthcoming population counts will be needed to analyse trends in their spatial distribution. 11.3.4 Indicator of Spatial Inequality To compare interregional disparities across different time periods, the population weighted coefficient of variation (the Williamson index-WI) of geographic inequality is used. The WI’s lower limit is zero, which corresponds to absolute equality; while its upper limit is not restricted and may increase (theoretically) indefinitely as the variation of the observations around the mean increases. Accounting for varying population size of regions is a major advantage of this index. This advantage justifies its use in the present study, in which regions of very different populations are compared. Furthermore, the WI is a dimensionless measure, since in its calculation the actual variation of a variable is normalized by its mean. It is thus possible to compare the extent of disparity for variables with distinctively different ranges of values, such as income, population density, etc. 11.3.5 Underlying Patterns of Interregional Disparities An analysis of the spatial distribution of different variables carried out in isolation, may often provide little useful information, or may even be misleading. For instance, if four indicators analysed separately show convergence among regions, while two others indicate divergence, one may be tempted to conclude that the differences among regions are decreasing. However, if the first four indicators are closely interrelated, while the latter two measures have little in common, the initial conclusion about convergence of inequalities may be misleading. Simple use of descriptive statistics is insufficient to clarify the underlying patterns of interregional inequalities. More sophisticated statistical techniques of data reduction such as, factor analysis, may be required. This multivariate technique makes it possible to reduce and summarize data and to analyse interrelationships among variables, explaining them in terms of their underlying dimensions (i.e. factors). In the present analysis, this task was performed using inequality variables reported in three subsequent censuses for which a sufficient number of comparable indicators were available - 1972, 1983 and 1995.

194

Boris A. Portnov

11.4 Recent Patterns of Interregional Disparities The spatial patterns of selected inequality measures, generated in ArcGIS software using the data drawn from the 1995 Census of Population and Housing, are given in Figures 11.2-11.5. In the following sections, each of four groups of inequality measures included in the analysis is considered separately - population distribution; housing and wealth; employment and wages; and education and ethnic makeup. 11.4.1 Population Distribution Two separate measures of population distribution - population density (POPDENS) and housing density (HDENSITY) - have been analysed. The spatial pattern of density distribution is relatively simple: it has three distinct peaks - around Tel Aviv, Haifa and Jerusalem. Outside these population concentrations, densities decrease gradually towards peripheral areas, towards both the north and south (Figure 11.2; left map). Between these urban centres, average densities also drop, reaching their lowest values in mainly agricultural regions on the Mediterranean plain - the Carmel Coast, Zikhron Ya’akov and Judean Foothills NR. At first glance, the pattern of housing density is clearly distinct from that of population density (Figure 11.2). In part, this may be attributed to the fact that housing density is related to two factors: family size and household welfare (i.e. the ability to purchase appropriate accommodation). However, a closer look at the spatial pattern of housing density shows that it is clearly related to that of population density. In fact, these two distributions are nearly inverse patterns of each other. Indeed, with the exception of the Jerusalem district, geographic areas with high population densities (Tel Aviv, Haifa and the Central districts) appear to have relatively low housing densities (see Figure 11.2). In contrast, most peripheral regions, which have low overall population densities, appear to have relatively high housing densities. For instance, the average housing density in the Tel Aviv District is only 0.56 persons per room, while in the peripheral Northern district it is 0.71 persons per room - 25 per cent higher. This relationship may be attributed partly to the predominant concentration of new housing construction (particularly private construction) in the more densely populated areas of the country, in which overall demand is greater and the average purchasing power of the local population tends to be higher (Portnov and Pearlmutter 1999). The “population - housing density” relation may also be the result of disparities in the ethnic makeup among different regions of the country. Israel has geographic regions with a disproportionably high concentration of ethnic minorities. In many cases, these minority groups have much larger families and a substantially lower income than the national average. For instance, average housing densities are considerably higher in mostly Arab-populated NRs, such as Shefar’am and Eastern Lower Galilee (see Figures 11.1 and 11.2; Table 11.1). In

Interregional Disparities in Israel: Patterns and Trends 195

these NRs, housing densities reach 0.94 and 0.91 persons per room, respectively, compared to the national average of 0.64 persons per room.

Haifa

#

Tel Aviv

#

#

#

Jerusalem

Be'er Sheva

Haifa

#

persons per sq.km 0 - 100 100 - 400 400 - 1000 1000 - 2100 2100 - 8700 no data

Tel Aviv

#

#

#

Jerusalem

Be'er Sheva

persons per room 0.49 - 0.56 0.56 - 0.64 0.64 - 0.71 0.71 - 0.8 0.8 - 0.94 no data

Fig. 11.2. 1995 Census: indicators of population distribution - population density (left) and housing density (right)

196

Boris A. Portnov

11.4.2 Housing and Wealth Interregional differences in housing and wealth may be illustrated by two different indices: homeownership (HOWNER) and ownership of personal computers (PCOWN) - see Figure 11.3. The spatial pattern of homeownership (Figure 11.3; right map) is similar to that of housing density (Figure 11.2; right map). This distribution has two extremes: Relatively low levels of homeownership are observed in the most centrally located NRs (Judean Mountains, Tel Aviv, Haifa and Carmel Coast) and in predominantly Jewish-populated peripheral areas (Ashqelon, Gerar, Besor, Northern Negev Mountains NRs etc.). Concurrently, high levels of homeownership are observed in mostly Arab-populated NRs (Alexander Mountain, Shefar’am region and Eastern Lower Galilee). The high level of homeownership in the latter group of NRs is hardly surprising since public housing, which is often available to Jewish families and, particularly to new immigrants, is generally unavailable to Israeli Arabs. The ownership of personal computers is especially high in Judea, Samaria and the Gaza Area and adjacent to major metropolitan centres of the country - Rishon LeZiyyon and Southern Sharon NRs near Tel Aviv, and Yoqne’am and Zikhron Ya’akov NRs near Haifa (see Figures 11.1 and 11.3, and Table 11.1). The relatively high levels of PC ownership in these regions reflect clearly the generally high level of education of the local residents - 12-13 years of schooling, as opposed to some 11 years on average for the country as a whole. 11.4.3 Employment and Wages Excluding Judea and Samaria, many of whose residents earn their income outside of their communities, the highest levels of income per employee are observed around the two major population centres of the country - Tel Aviv and Haifa (see Figure 11.4). In particular, the average wages are especially high in the suburban fringes of these population centres - the Southern Sharon NR, north-east of Tel Aviv (1,750 $US), and in the Zikhron Ya’akov NR, south of Haifa (1,600 $US). After peaking in these suburban areas, the average wage levels decline steadily towards more remote peripheral regions. The spatial pattern of labour force participation (Figure 11.4; right map) resembles that of average income (Figure 11.4; left map) relatively closely. This implies that the concentration of high-paying jobs and the availability of employment are spatially interlinked. By the same token, regions of scarce employment are generally characterized by relatively low wages, a well-known phenomenon reported in many studies (Richardson 1977; Armstrong and Taylor 1993). Relatively low participation in the labour force is also found in most peripheral regions of the country (see Figure 11.4; right map). For instance, in the Eastern Lower Galilee NR and in the Besor NR, average participation in the labour force is only some 39-42 per cent, i.e. 12-15 per cent below the national average. Such

Interregional Disparities in Israel: Patterns and Trends 197

low participation may be explained both by a large average family size and by a general lack of employment opportunities in these regions.

Haifa

#

Tel Aviv

# Jerusalem

# per cent 8 - 15 15 - 20 20 - 25 25 - 31 31 - 40

#

Be'er Sheva

no data

per cent less then 40 40 - 60 60 - 70 70 - 80 80 - 100 no data

Fig. 11.3. 1995 Census: indicators of wealth and housing - ownership of personal computers (left) and homeownership level (right)

198

Boris A. Portnov

Haifa

#

Tel Aviv

#

#

Jerusalem

Haifa

#

#

Be'er Sheva

Tel Aviv

#

$US per employee less then 1100 1100 - 1240 1240 - 1450 1460 - 1760 no data

#

#

Jerusalem

Be'er Sheva

per cent 38 - 45 45 - 51 51 - 57 57 - 63 63 - 76 no data

Fig. 11.4. 1995 Census: indicators of employment and wages - monthly income per employee (left) and participation in the labour force (right)

Interregional Disparities in Israel: Patterns and Trends 199

11.4.4 Education and Population Makeup The average number of years of schooling is not only a general indicator of the current state of regional development - it is also an important measure of a region’s growth potential (Figure 11.5; left map). In a modern, high-tech oriented economy, the presence of a highly educated and skilled labour force is essential for sustained regional growth (Shachar and Felsenstein 1992). The spatial pattern of the indicator in question (Figure 11.5) resembles the patterns of income per employee and participation in the labour force (see Figure 11.4): a large proportion of highly educated people is concentrated around Tel Aviv and Haifa, in Eilat (the Arava NR), and in the Judea, Samaria and Gaza area. Concurrently, a relatively low level of education is found outside the commuting rings of the aforementioned population centres (the Lod, Eastern Sharon, and Hadera NRs) and in mostly Arab-populated NRs of the Galilee and Golan - the Akko, Hermon, Shefar’am and Eastern Lower Galilee NRs (see Figures 11.1 and 11.5; Table 11.1). Notably, in the southern periphery of the country (the Be’er Sheva, Besor and Northern Negev Mountains NRs), the average level of education is relatively high (Figure 11.5). This is attributed to the fact that these regions absorbed many highly skilled new immigrants who arrived to the country in the wake of mass immigration from the former Soviet Union in 1989-1991 (Isralowitz and Friedlander 1999). Some 45 years after the bulk of new immigrants from the Middle East and Northern Africa were directed to the transit camps in the Negev and in the north (see Section 11.3.1), the effect of this policy of population dispersal is still reflected in the spatial patterns of ethnic makeup. As Figure 11.5 shows, in the peripheral areas of the country, the proportion of this group of immigrants (Asia and Africa born) is substantially higher than elsewhere. In part, this may be attributed to the “selectivity of migration” phenomenon, whereby the young and educated have a tendency to migrate, while the least educated and elderly tend to be less mobile (De Jong and Fawcett 1981; Molho 1995).

11.5 Time-Related Changes of Interregional Disparities In order to clarify temporal changes in the patterns of interregional inequalities, the population weighted means and the values of Williamson’s inequality coefficients were calculated using data available in different censuses, starting with 1961. The results of the comparison for selected inequality measures are shown in Figure 11.6, which illustrates the most characteristic temporal trends. Two important measures of interregional inequalities - participation in the labour force, ethnic makeup, and average income (LFORCE, MAKEUP and INCOME) - indicate that interregional disparities in Israel tended to grow (see Figure 11.6). Thus, for instance, the value of WI for the wages variable increased

200

Boris A. Portnov

from 0.139 in 1972 to 0.144 in 1983 and to 0.214 in 1995, implying that the differences in average wages among individual NRs diverged over time.

Haifa

#

Tel Aviv

# Jerusalem

#

Haifa

#

#

Be'er Sheva

Tel Aviv

#

#

Jerusalem

years 6 and less 6 - 10 10 - 11 11 - 12 12 - 13 no data

#

Be'er Sheva

per cent 0-3 3 - 10 10 - 14 14 - 19 19 - 26 no data

Fig. 11.5. 1995 Census: indicators of education and ethnic makeup - average years of schooling (left) and percentage of residents who had immigrated from Asia and Africa - 1st generation (right)

Interregional Disparities in Israel: Patterns and Trends 201

Differences among individual NRs in participation in the labour force (LFORCE) also diverged since the 1972 Census count. The value of the WI for this variable decreased between the 1961 and 1972 Censuses from 0.097 to 0.085, indicating a trend of convergence. However, since 1972, WI for this variable grew continuously: to 0.092 by 1983 and to 0.136 in 1995. This implies that participation in the labour force increased in some NRs and decreased in others. A similar tendency is observed in the ethnic makeup (MAKEUP), estimated by the percentage of Asia and Asia born in the total population. The values of the WI for this variable increased by nearly 16 per cent since the early 1960s, from 0.393 in 1961 to 0.454 in 1995 (Figure 11.6). 0.600 0.498

Williamson index

0.500

0.470 0.414

0.400 0.300

0.348 0.274 0.236 0.191

0.200

0.150

0.139

0.097

0.144

0.100

0.214 0.191 0.146 0.136

0.092

0.085

0.000 1961

1972

1983

1995

Census year INCOME LFORCE

YSTUDY HIGHED

HDENSITY CAROWN

Fig. 11.6. Changes in selected indicators of interregional inequalities over the whole country (1961-1995 Census data)

However, not all measures of inequality included in the analysis suggest an increase in regional disparities. The following four indicators show a trend of reduced inter-regional inequality: the average number of years of study (YSTUDY), housing density (HDENS), higher education (HIGHED), and motorization level (CAROWN). For instance, the WIs for the latter variable (CAROWN) decreased from 0.414 in 1972 to 0.146 in 1995, indicating that over time, motorization levels converged considerably across the country (Figure 11.6). The apparently contradictory trends displayed by different indicators warrant some discussion. The mean values of nearly all of the indicators that show that regional inequalities tend to decrease have a common property: They seem to approach certain quantitative thresholds beyond which further increase is highly unlikely. For instance, the mean value of the YSTUD variable reached 11.08 years in 1995. Theoretically, this value may grow to 16-17 years (12 years of primary and high school plus 4-5 years of college education). However, it is unlikely that the entire population of the country will opt for such a long period of study.

202

Boris A. Portnov

Likewise, the mean value of the housing density variable (HDENS) stood in 1995 at 0.64 persons per room. This means that about one and a half rooms are available for an “average resident” of the country. Though the average housing density may decrease even further as standards of living improve, it is likely to reach some quantitative limit eventually. On the other hand, many measures showing increasing inequality between regions do not have a natural limit: For instance, personal income may grow almost indefinitely as long as the overall wealth of the country increases, regardless of polarization between rich and poor.

11.6 Underlying Patters of Interregional Inequalities A total of eleven variables were considered in the factor analysis of the 1995 Census data. These variables covered all of the inequality measures introduced in Section 11.3.1, excluding two: ownership of personal computers (PCOWN) and ownership of air conditioners (AIROWN). The latter measures were omitted from the analysis in order to enhance the comparability of results with those obtained from the analysis of previous population counts covered by the study - 1983 and 1972, in which these variables were not reported. The same number of variables (11) was used for the analysis of the 1983 Census data. However, in the analysis of the 1972 Census, only ten variables were considered. One variable, homeownership level (HOWNER), was omitted from the analysis because data were not available at the required level of spatial aggregation (see Tables 11.211.4). 11.6.1 1995 Census As Table 11.2 shows, the model describes the variation of the observed (input) variables relatively well. This is indicated by the statistical significance of the results of two tests: Kaiser-Meyer-Olkin (KMO) test of sampling adequacy (KMO=0.725), and Barlett’s test of sphericity (P<0.0001). Two principal components are identified (see Table 11.2). Together, these components account for about 72 per cent of the total variance of the input variables, with 41 per cent attributed to the first factor (Comp. 1) and another 31 per cent explained by the second component (Comp. 2). Analysis of the rotated component matrix makes it possible to define each of these factors (components). In assigning such definitions, the following properties of the principle components are especially important: x Component 1 has a strong positive correlation with four input variables: average wages (r=0.750), years of study (0.848), participation in the labour force (0.824), and the proportion of residents with higher education (0.866). This component also correlates negatively with the proportion of unskilled workers (r=-0.809; see Table 11.2). The definition of the component in question is thus more or less clear: It reflects interregional differentials in

Interregional Disparities in Israel: Patterns and Trends 203

education and income, and can thus be defined as the “earnings-education” component of interregional disparities; Table 11.2. 1995 Census: factor analysis - explanation of total variance Component(1)

Initial eigenvalues Total

Percentage Cumulative of variance percentage 56.499 56.499 17.356 73.855 7.862 81.717 6.955 88.672 3.739 92.411 2.929 95.340 1.541 96.881 1.388 98.270 1.084 99.353 0.452 99.806 0.194 100.000

Rotated sums of squared loadings Total Percentage Cumulative of variance percentage 6.215 56.499 56.499 1.909 17.356 73.855

1 6.215 2 1.909 3 0.865 4 0.765 5 0.411 6 0.322 7 0.170 8 0.153 9 0.119 10 0.050 11 0.021 No. of observations: 50 Kaiser-Meyer-Olkin measure of sampling adequacy: 0.725 Bartlett test of sphericity - approximate Chi-Square: 459.444;. indicates a 0.001 significance level Principal Component Matrix(1) Indicator(2) Extraction Component matrix Rotated component matrix Component Component Component Component 1(3) 2(4) 1(3) 2(4) POPDENSI 0.392 0.376 0.500 0.453 -0.044 MAKEUP 0.852 0.483 -0.786 -0.037 -0.868 HIGHED 0.822 0.844 0.331 0.866 -0.287 UNSKILL 0.756 -0.796 -0.349 -0.809 0.237 WAGES 0.820 0.739 0.524 0.750 0.417 HDENSI 0.868 -0.908 0.209 -0.646 0.661 HSIZE 0.874 -0.921 0.160 -0.575 0.766 HOWN 0.835 -0.641 0.650 -0.113 0.906 CAROWN 0.387 -0.600 0.165 -0.373 0.555 LFORCE 0.625 0.790 -0.018 0.824 -0.293 YSTUDY 0.894 0.938 0.123 0.848 -0.429 (1) Extraction method: principal component analysis; rotation method: Varimax with Kaiser normalization. (2) See breakdown of indicators in Sect. 11.3. (3) Education and income. (4) Housing and ethnic composition.

x Component 2 is strongly correlated with household size (0.766), homeownership (0.906), and housing density (0.661). Concurrently, it has a strong negative correlation with ethnic makeup (-0.868), which is estimated as the percentage of Jewish immigrants of Asian and African descent in the total

204

Boris A. Portnov

population. We shall thus describe this component as the “housing conditions and ethnic makeup of the area.” To summarize, the underlying patterns of interregional inequalities in Israel reflected in the 1995 Census data are of two basic types: a) inequalities in education and income (Comp. 1), and b) inequalities in the ethnic makeup of the population and in housing conditions in a region (Comp. 2). 11.6.2 1983 Census Although factor analysis of the 1983 Census data also results in two principle components of interregional inequalities (Tables 11.2-11.4), the definitions of these components are quite different from those of the 1995 Census: x Component 1 has a strong positive correlation with the level of homeownership (r=0.927), housing density (0.810), and household size (0.683); it is inversely related to ethnic make up (-0.897), participation in the labour force (0.732) and average years of study (-0.677; see Table 11.3 - Rotated Component Matrix). In total, this component accounts for 43 per cent of the variation of the observed (input) variables; x Component 2 correlates positively with the proportion of residents with a higher education (0.858); average years of study (0.695), and average earnings (0.643), but is negatively related to the level of car ownership (-0.905) and to household size (-0.604; see Table 11.3 - Rotated Component Matrix). Altogether, the factor in question accounts for about 37 per cent of the variance of the observed (input) variables (Table 11.3). Therefore, in the early 1980s, the underlying patterns of interregional inequalities in Israel were substantially different from those indicated by the analysis of the 1995 Census. In particular, the main component of interregional inequalities in the early 1980s was related to housing conditions (housing density, household size and ethnic composition). The factor influenced by education and income occupies only the second position in the factors’ hierarchy. The same factors appear to be in the reverse order in the analysis of the 1995 Census data, in which inequalities in education and earnings came first, while inequalities in housing and ethnic makeup were ranked second. This reversal of the relative positions of the principle components between 1983 and 1995 may be attributed, at least in part, to dramatic improvements in housing conditions in Israel during this period. The average housing density dropped during this period from 1.19 persons per room in 1983 to 0.64 in 1995 (see Table 11.2). This considerable change may have contributed to the declining importance of housing as an indicator of interregional disparities. On the other hand, the transition of Israel’s economy from traditional industries (textile, chemicals, etc.) to high-tech production (electronics, optics, medical equipment, and internet technology), most of which is concentrated in the central part of the

Interregional Disparities in Israel: Patterns and Trends 205

country and near Haifa, is probably the reason for the increased importance of the “education-earnings” component in the factor hierarchy. Table 11.3. 1983 Census: factor analysis - explanation of total variance Component(1) Total

Initial eigenvalues Percentage Cumulative of variance percentage 59.289 59.289 19.534 78.823 7.851 86.673 5.975 92.649 2.604 95.253 2.274 97.527 1.245 98.772 0.567 99.339 0.339 99.677 0.230 99.908 0.092 100.000

Rotated sums of squared loadings Total Percentage Cumulative of variance percentage 6.522 59.289 59.289 2.149 19.534 78.823

1 6.522 2 2.149 3 0.864 4 0.657 5 0.286 6 0.250 7 0.137 8 0.064 9 0.037 10 0.025 11 0.010 No. of observations: 40. Kaiser-Meyer-Olkin measure of sampling adequacy: 0.761. Bartlett test of sphericity - approximate Chi-Square: 565.513; indicates a 0.001 significance level. Principal Component Matrix(1) (2) Indicator Extraction Component matrix Rotated component matrix Component Component Component Component 1(3) 2(4) 1(3) 2(4) POPDENSI 0.482 0.322 0.615 0.171 0.673 MAKEUP 0.820 0.585 -0.691 -0.897 -0.124 HIGHED 0.764 0.698 0.526 -0.167 0.858 UNSKILL 0.755 -0.864 0.089 0.702 -0.512 WAGES 0.636 0.781 0.163 -0.472 0.643 HDENSI 0.915 -0.943 0.162 0.810 -0.510 HSIZE 0.831 -0.911 0.001 0.683 -0.604 HOWN 0.899 -0.556 0.768 0.927 0.200 CAROWN 0.835 -0.698 -0.590 0.125 -0.905 LFORCE 0.792 0.883 -0.112 -0.732 0.506 YSTUDY 0.942 0.968 0.064 -0.677 0.695 See comment to Table. 11.2.

11.6.3 1972 Census The results of the analysis of the 1972 Census data appear to be quite similar to those obtained from the analysis of the 1995 Census data. Two principle components of interregional inequalities emerged from the factor analysis of the 1972 Census data (Table 11.4):

206

Boris A. Portnov

x Component 1 explains about 43 per cent of the variation of the input variables and correlates strongly with average wages (0.881) and car ownership (0.846). Concurrently, it has a significant negative correlation with household size (-0.974) and household density (-0.901). In other words, the highest values of this component are observed in regions with an affluent population, small families and good housing conditions. The factor in question may thus be defined as a general “indicator of welfare and housing.” x Component 2 explains about 28 per cent of variance of the observed (input) variables and correlates positively with the following two variables: the average years of study (0.905), and participation in the labour force (0.787). It correlates negatively with the proportion of unskilled workers in a region (r=-0.760). This factor thus reflects mainly educational and employment disparities among geographic areas. Since the former factor (welfare and housing) has a considerably higher explanatory power (about 43 per cent of the total variance vs. 28 percent for the second component), it implies that the patterns of interregional inequalities in 1972 were mainly attributed to interregional differentials in income and welfare (Comp. 1), while differentials in education (Comp. 2) played only a secondary role. The differences in the nature of interregional inequalities between 1972 and 1983 are thus fairly clear: inequality in 1972 was mostly due to differences in welfare, but by 1983 the difference between regions was mostly due to housing conditions. At least in part, this difference may be attributed to the structural changes in the Israeli economy during the period in question: Between the late 1960s and early 1980s, the country’s economy underwent a sharp transition from high annual rates of growth following Israel’s success in the 1967 War to a general economic slowdown in the late 1970s-mid-1980s. This transition was characterized by hyperinflation and low rates of housing construction, especially in the public sector (ICBS 2002). Reduced rates of housing construction were directly responsible for a rapid growth of interregional differentials in housing conditions and thus might have caused the emergence of the “housing-related” factor as the prime component of interregional inequalities in 1983. The reversal of the relative importance of the principle components between the 1983 and 1995 Censuses may be attributed, at least in part, to a roll-back of the “welfare state” in the middle 1980s. The economic crisis of 1983-85 and subsequent economic reforms, which included privatisation of many publicly held assets and a dramatic reduction of the state’s subsidies, caused further economic polarization of Israeli society and growing disparities of income in the 1990s. This reversal is illustrated by a decrease in the extent of interregional disparities for most measures from 1972 to 1983, followed by an increase in inequality from 1983 to 1995 (see Figure 11.6).

Interregional Disparities in Israel: Patterns and Trends 207 Table 11.4. 1972 Census: factor analysis - explanation of total variance Component(1) Total

Initial eigenvalues Percentage Cumulative of variance percentage 49.208 49.208 22.128 71.336 11.371 82.707 5.759 88.466 4.740 93.206 2.721 95.927 1.921 97.848 1.046 98.894 0.820 99.714 0.286 100.000 49.208 49.208

Rotated sums of squared loadings Total Percentage Cumulative of variance percentage 3.974 39.741 39.741 2.478 24.781 64.522

1 4.921 2 2.213 3 1.137 4 0.576 5 0.474 6 0.272 7 0.192 8 0.105 9 0.082 10 0.029 11 4.921 No. of observations: 39 Kaiser-Meyer-Olkin measure of sampling adequacy: 0.732 Bartlett test of sphericity - approximate Chi-Square: 339.925;. indicates a 0.001 significance level. Principal Component Matrix(1) (2) Indicator Extraction Component Matrix Rotated Component Matrix Component Component Component Component 1(3) 2(4) 1(3) 2(4) POPDENSI 0.209 0.399 0.223 0.456 -0.024 MAKEUP 0.248 -0.211 0.451 0.010 -0.498 HIGHED 0.432 0.657 0.024 0.600 0.268 UNSKILL 0.588 -0.430 0.635 -0.105 -0.760 WAGES 0.779 0.765 0.441 0.881 -0.058 HDENSI 0.900 -0.940 -0.130 -0.901 -0.298 HSIZE 0.970 -0.940 -0.295 -0.974 -0.150 CAROWN 0.744 0.834 0.222 0.846 0.168 LFORCE 0.624 0.409 -0.676 0.069 0.787 YSTUDY 0.958 0.734 -0.648 0.373 0.905 See comment to Table. 11.2.

11.7 Conclusions Interregional inequality is a multifaceted phenomenon, which has a variety of manifestations. Once created, these inequalities may become persistent and selfperpetuating, and may even lead to serious social divisions (Balchin 1990; Economou 1993; Dunford 1995; Wong 1995; Abe 1996; Markusen 1996; Parr 1999). The present analysis of interregional inequality in Israel and its temporal dynamics makes it possible to gain certain insights into this complex phenomenon. First, the extent of interregional disparities in Israel (as indicated by the most recent Census of Population and Housing, in 1995) appears to differ when

208

Boris A. Portnov

different measures of inequality are considered. Thus, population density, ownership of personal computers, ethnic makeup and the percentage of unskilled workers in the population indicate the highest degree of interregional disparities. Concurrently, there are smaller spatial variations when indicators of education, and motorization levels are considered. Second, temporal trends are complex. There appears to be no universal, nationwide tendency for socio-economic disparities among regions to either converge or diverge. Population density, average earnings, participation in the labour force, and ethnic makeup tend to diverge, reflecting greater interregional inequality. Other measures, mostly related to education and housing - average length of study, housing density, and the proportion with higher education - reflect a general tendency to converge. The mean values of nearly all of the factors which indicate convergence between regions seem to approach certain quantitative limits beyond which no considerable increase may be expected (housing density, average years of schooling, etc.). On the other hand, most measures indicating an increase in interregional disparities, such as population density and average income, are unlikely to reach any quantitative limit in the near future. For instance, population density may grow almost indefinitely as long as the overall population of the country increases and land resources remain constant. Third, the underlying patterns of interregional inequalities in Israel tend to change over time. Differentials in income and welfare were the main indicators of interregional disparities in the early 1970s; in the early 1980s differences in housing were the predominant characteristic; by 1995, education and income emerged as the predominant factors of interregional disparities. These changes are probably related to macro-economic processes in the national economy in the past decades. These include a drastic reduction of public involvement in housing construction in the late 1970s-early 1980, and the transition of the national economy from traditional low-tech industries to high-tech production in the late 1980s-early 1990s. While the former process resulted in a substantial divergence of housing conditions, the latter trend resulted in the strengthening of the educational component as the major feature of interregional disparities.

Acknowledgement This chapter summarizes the results of a joint research project entitled Interregional inequalities in Israel, 1948-1995: divergence or convergence? carried out with Evyatar Erell of Ben-Gurion University of the Negev and funded by the Ford Foundation.

Interregional Disparities in Israel: Patterns and Trends 209

References Abe H (1996) New directions for regional development planning in Japan. In: Aden J, Boland P (eds) Regional development strategies: an European perspective. Jessica Kingsley Publishers, London Bristol, pp 273-295 Anson J (1993) Geographical regions as ethnic groups? Social geography as the sociology of space. In: Anson J, Todorova E, Kressel G et al. (eds) Ethnicity and politics in Bulgaria and Israel, Avebury, Aldershot, pp 13-28 Armstrong H, Taylor J (1993) Regional economics and policy. Harvester, NY Balchin PN (1990) Regional policy in Britain: the north-south divide. Paul Chapman Publishing, London CIA (2004) 2003 World Fact Book (Internet Edition) De Jong CF, Fawcett JT (1981) Motivation for migration: an assessment and a valueexpectancy research model. In: De Jong CF, Gardner RW (eds) Migration decisionmaking. Multidisciplinary approaches to microlevel studies in developed and developing countries. Pergamon Press, NY, pp 13-53 Drabkin-Darin H (1957) Housing in Israel: economic and sociological aspects. Gadish Books, Tel Aviv Dunford M (1995) Metropolitan polarization, the north-south divide and socio-spatial inequality in Britain: a long-tern perspective. European Urban and Regional Studies, 2(2):145-170 Economou D (1993) New forms of geographical inequalities and spatial problems in Greece. Environment and Planning C 11(5):583-598 Gradus Y, Krakover S (1977). The effect of government policy on the spatial structure of manufacturing in Israel. Journal of Developing Areas, 1977 11:393-409 ICBS (2002) Statistical abstract of Israel (annual). Israel Central Bureau of Statistics, Jerusalem Isralowitz R, Friedlander J (eds) (1999) Transitions: Russians, Ethiopians and Bedouins in Israel’s Negev desert. Ashgate, Aldershot Kipnis BA (1996) From dispersal to concentration: alternative strategies in Israel. In: Gradus Y, Lipshitz G (eds) Mosaic of Israeli geography. Ben-Gurion University of the Negev Press, Be’er Sheva, pp 29-36 Krakover S (1998) Testing the turning-point hypothesis in city-size distribution: the Israeli situation re-examined. Urban Studies 35(12):2183-2196 Lipshitz G (1996) Spatial concentration, and deconcentration of population: Israel as a case study. Geoforum 27(1):87-96 Markusen A (1996) Interaction between regional and industrial policies: evidence from four countries. International Regional Science Review 19(1):49-77 Mera K (1995) Polarization and politico-economic change. Papers in Regional Science. 1995 74(1):175-185 Molho I (1995) Migrant inertia, accessibility and local unemployment. Economica 62(245):123-132 Parr JB (1999) Growth-pole strategies in regional economic planning: a retrospective view: part 2. Implementation and outcome. Urban Studies 36(8):1247-1268 Portnov BA, Erell E (1998) Long-term development peculiarities of peripheral desert settlements: the case of Israel. International Journal of Urban and Regional Research 22(2):216-232 Portnov BA, Etzion Y (2000) Investigating the effects of public policy on the interregional patterns of population change: the case of Israel. Socio-economic Planning Sciences 34(4):239-269

210

Boris A. Portnov

Portnov BA, Pearlmutter D (1999) Sustainable urban growth in peripheral areas. Progress in Planning Monograph Series 52(4):239-308 Portnov BA (1999) The effect of regional inequalities on migration: a comparative analysis of Israel and Japan. International Migration 37(3):587-615 Richardson HW (1977) Regional growth theory. Macmillan, London Shachar A, Felsenstein D (1992) Urban economic development and high technology industry. Urban Studies 29(6):839-855 Shachar A, Lipshitz G (1981) Regional inequalities in Israel. Environment and Planning A 13(4):463-473 Shefer D (1990) Innovation, technical change and metropolitan development: an Israeli example. In: Nijkamp P (ed) Sustainability of urban system: a cross-national analysis of urban innovation. Avebury, Aldershot, pp 167-182 Soen D (1977) Israel’s population dispersal plans and their implementation, 1948-1974: failure or success? GeoJournal 1(5):21-26 Sonis M (1988) Interregional migration in individual countries: Israel. In: Weidlich W, Haag G (eds) Interregional migration: dynamic theory and comparative analysis. Springer, NY Williamson JG (1965) Regional inequalities and the process of national development: a description of the patterns. In: Friedmann J, Alonso W (eds) Regional policy. The MIT Press, Cambridge, MS, pp 158-200 Wong C (1995) Developing quantitative indicators for urban and regional policy analysis. In: Hambleton R, Huw T (eds) Urban policy evaluation: challenge and change. Paul Chapman Publishing Ltd, Cardiff, pp 111-122

12

Does Decentralisation Matter to Regional Inequalities? The Case of Small Countries

Carlos Gil, Pedro Pascual and Manuel Rapún Department of Economics, Universidad Pública de Navarra, Pamplona, Spain

12.1 Introduction: Politics and Economic Performance Although most economic growth theories have concentrated on capital and technology, there is a rich tradition in the analysis of non-economic factors of growth. Schumpeter (1954), for example, draws special attention to the role of entrepreneurship in generating wealth with his notion of creative destruction. Kuznets (1966) studies the influence of factors such as social and political structure. Development economists like Myrdal (1957) have suggested the relevance of political conditions in the study of growth. It is in the last two decades, however, that there has emerged a widespread and increasing interest in the relationship between politics and political conditions on the one hand and economic performance on the other. Aspects such as political stability, the size of the public sector or fiscal policy measures are becoming common in growth regressions. Barro (1990; 1991) is one of the social scientists that have paid more attention to the relationship between political conditions and economic growth. However, in most models, which are embedded in neoclassical economic theory, social and political factors play only a secondary role, capital accumulation being the main driving force for growth. In this context, few political and social indicators are robustly correlated with growth (Levine and Renelt 1992). But the fragility of the results concerning social, institutional and political indicators may well be due to this supporting role (Rodríguez-Pose 1998). The presence of capital investment or capital stock variables may well distract from the relevance of social and political indicators. The close correlation between productivity or productivity growth (the endogenous variables), and capital stock or capital investment (the main exogenous variables), which are directly related and are probably, both the cause and the result of economic growth, may distract from the relevance of those variables that create the conditions needed to fuel a virtuous circle of growth and investment. Thus, Rodriguez-Pose (1998) argues that there is a need to consider social, political and even cultural factors as principal actors in the process of economic growth. National economic performance has usually been identified with national growth. But most countries pursue other relevant economic objectives, such as reducing personal and regional inequalities in income or productivity. In the last decade, especially since the publication of the article by Barro and Sala-i-Martin

212 Carlos Gil, Pedro Pascual and Manuel Rapún

(1991), and in an attempt to test the neoclassical growth theory, there has been a marked increase in the number of studies analysing regional disparities. The main aim in most of these works was not to explain their level or their evolution, but to prove or reject the hypothesis of productivity convergence predicted by the neoclassical model. Institutional, political or social variables were usually omitted in this kind of study, where it was argued that regions within a country share common political and social institutions. In this chapter we will try to analyse the influence of some potentially important political variables in the degree of regional inequality among OECD countries. We will focus mainly on decentralisation, both political and fiscal but we will also include two other potentially relevant groups of variables. The first refers to the size of the public sector or its economic impact, the second to the nature of its governing political parties. A second objective is to ascertain whether main conclusions drawn for the full sample are also true for small countries, defined in terms of population and land surface. We think that there are some quantitative, and perhaps qualitative, differences (related to the size of the country) that may generate diverse patterns in the object of our study related to the size of the country. This chapter continues by examining the debate over the different types of outcome that decentralisation could help generate in economic performance in general, and regional inequalities in particular. We also discuss arguments that could help explain a potentially different impact of decentralisation in small countries. Section 12.3 deals with the possible influence of other political variables, namely public sector size and parties in government, on regional disparities. This is followed in Section 12.4 by an explanation of the different measures of regional inequality, decentralisation and the remaining political variables that we use in the analysis. Empirical results are given in Section 12.5. This is followed by a brief concluding section.

12.2 Decentralisation and Economic Performance: Theoretical and Empirical Background 12.2.1 The Increasing Interest in Decentralisation For more than two decades, there has been an increasing interest in decentralisation all around the world. Many developing countries have embarked, or intend to embark, on some form of transfer of political power to local government (see, for example, Dillinger 1994). Furthermore, decentralisation has become a central issue in the political agenda of developed countries with more consolidated political systems. Belgium became a federal state in 1993 (Oates 1999). In the UK and Spain decentralisation is an ongoing process, though not without some measure of controversy. In the EU the regions are increasingly

Does Decentralisation Matter to Regional Inequalities? 213

perceived to be the relevant units for implementing political decisions (Tomaney and Ward 2000). There are numerous reasons that might explain this interest. Some of them were identified by Armstrong and Taylor (2000). First, the belief in decentralisation as an effective tool for increasing the efficiency of public expenditure. Second, the reaction against large centralised bureaucracies in areas such as the EU. The third has to do with the influences of changes that have taken place over the last decades in the way private corporations are managed. The fourth argues that policies designed to stimulate endogenous growth (through the encouragement of small firms, for example) are very difficult to run from the centre. Fifth, and last is the demand for a closer democracy, which could promote public participation in social policy and administration (Tunstall 2001). Within the main line of comparative political research with an interest in the consequences of federalism and decentralisation there is agreement over the importance of decentralisation issues. Some authors have claimed federalism to be superior to other democratic systems, because it provides a better safeguard for the democratic rights of citizens in general and minorities in particular (Elazar 1995). Neo-institutionalist economists have also made the case that certain institutional arrangements encourage individuals to engage in some economic activities more than in others, thereby giving rise to more successful economies (North 1990). It is widely accepted that the three main objectives of the public sector are those initially stated by Musgrave (1959): efficiency in the allocation of resources, income redistribution and macroeconomic stability. Traditionally, most public economists have agreed that while the first of these functions can be assigned to lower levels of government, the latter two should more appropriately be assigned to the national level. Decentralisation may generate more efficiency, but may also reduce economic stability and aggravate regional inequality. Thus, in recent years these assumptions have provoked considerable controversy. Emanating from public choice theory, with roots in the neo-classical school of thought, there is a suggestion that decentralisation could improve efficiency in the allocation of resources by better satisfying the needs and preferences of local citizens, through better knowledge of these preferences (Oates 1972). These gains in efficiency would be enhanced with mobility of citizens who could choose to live in the jurisdiction that best matched their preferences. Regions would also have incentives to compete with one another by attracting migrants, making more efficient use of their resources and increasing economic welfare. However, some authors (Prudhomme 1995; Tanzi 1996) think that preferences among individuals living in a country are quite similar, and that lack of coordination among regional governments could reduce efficiency in the provision of some public services. The existence of regional “spillovers” in the provision of some public goods could also generate an inadequate level of provision. Keynesian thought supports these arguments. Decentralisation reduces the capacity of central government to use demand policies to alleviate the effects of fluctuations in production and employment. Federal and highly decentralised states would therefore perform worse. Greater centralisation also permits more

214 Carlos Gil, Pedro Pascual and Manuel Rapún

efficient determination of macro-economic objectives, less diffusion in the utilisation of policy instruments, and a higher degree of co-ordination. 12.2.2 Regional Inequalities and Decentralisation The second main drawback traditionally attributed to decentralisation is an unbalanced distribution of resources across regions that might generate increasing economic differences among them. There are several issues that might influence the final outcome. The first is whether or not decentralisation results in more unequal distribution of public resources. Prudhomme (1995) argues that centralised public sectors will attempt to produce a more balanced distribution by channelling resources from richer areas to poorer ones. Conversely, centralised systems may create unequal distributions of public resources by favouring politically important jurisdictions. The second issue relates to whether centralisation could lead to a higher concentration of private investment. Investors seeking closer ties with politicians and the administration might tend to choose capital regions. The third point is that decentralisation can provide sub-national officials with the power to actively pursue economic development policies. This will include not only policies better suited to local needs or capabilities, but also several forms of competition among regional and local governments, which may include granting tax privileges and offering other forms of assistance to businesses willing to locate in a particular jurisdiction (Martinez-Vazquez and McNab 2003). It is difficult to assess whether or not they will contribute to reducing regional disparities. Both lines of argument linking political and fiscal decentralisation and economic performance could also be applied to the level of regional disparities. On the basis of public choice theory, we might expect less regional disparity in decentralised states. In the first place, the power to control most of the public budget locally could generate more competitiveness among regions, forcing regional governments to deliver services at minimum cost, thus enhancing efficiency. Besides, local governments could be removed if they failed to achieve standards of wealth and economic growth similar to those in the rest of the country. The power to design regional policies tailored to local needs, in an effort to promote employment and productivity, would give local officials the capacity to achieve economic goals. Furthermore, as central government would be reduced in size, the concentration of political and economic power around the capital region would also be less intense. From a Keynesian approach, however, a weaker central state would play a less crucial role in redistributing income among regions, and could not use demandside policies, such as public investment, to promote economic growth in the poorer regions. More diffusion in policies such as education or health could also lead to an increase in disparities in decentralised countries. Related to this is the fact that the benefits of regional policies spill over into other areas. For example, the creation of extra jobs in an assisted region will reduce the amount of unemployment transfers and raise tax revenue, to the benefit of the inhabitants of

Does Decentralisation Matter to Regional Inequalities? 215

non-assisted regions (Armstrong and Taylor 2000). Other less tangible benefits, such as those of a social or environmental nature, could also spill over regional boundaries. Since there are effects that spread beyond regional borders, totally isolated development policies are likely to produce inefficient levels of regional policy and equalisation among regions. The two opposing arguments about the impact of decentralisation on regional inequalities shed no light on the issue that provides the focus of this study: are regional inequalities in decentralised countries greater than, smaller than, or more or less the same as in centralised countries? We should seek the answer to this question in empirical studies. There are hardly any that address this question directly, however. Some national studies (Freinkman and Yossifov 1999, for Russia) in countries involved in asymmetric devolution processes find better economic performance in regions with more autonomy. However, it has already been mentioned that another study (Zhang and Zou 1998) yields the opposite result for China. Tsui (1996) directly addresses the issue of regional inequalities and decentralisation. He finds fiscal decentralisation in China to be related to the rise in disparities in the 1980s. Again though, very special circumstances prevailed during this period of analysis, such as the high level of foreign direct investment in the Special Economic Zones, which are to be found in the richest areas1. Also, the devolution process in China is asymmetric, thus leaving some regions with more political and fiscal autonomy than others. 12.2.3 Why Should Small Countries Be Different? The second objective of this chapter is to see whether the different indicators of decentralisation, and the two other political variables that will be presented in the next section, have a distinct impact in small countries. There are several reasons that could justify a divergent result to that obtained with larger countries. First of all, decentralisation could be less relevant when the political and administrative centre is close to all regions of the country. Politicians in central government may be more aware of regional peculiarities and problems when there are fewer regions (which is the case in small countries) than when there are many. Furthermore, the hypothesis that preferences among individuals living in a country are fairly similar (Prudhomme 1995; Tanzi 1996) could be particularly true when referred to countries with less diversity2. In large countries, meanwhile, a single region may be large enough to create the appropriate conditions to attract investment or to generate endogenous growth. In small economies, if local markets are too narrow in terms of consumption and employment, peripheral regions could face serious difficulties in trying to 1

2

Zhao and Tong (2000) argue that the “get rich first policy” and “coastal development strategy” has contributed largely to the increase in spatial disparities. Although it is not immediate that small countries, characterised in terms of population and/or surface, are less diverse in terms of economy, culture, natural resources, etc. (Switzerland or Norway could be good counter-examples) we can expect a smaller average degree of diversity.

216 Carlos Gil, Pedro Pascual and Manuel Rapún

compete. Investment would tend to concentrate in central regions, those most densely populated and those that house the country’s administrative centres. Centralisation would cause economic activity to concentrate close to where political power was located, thus causing even greater aggravation of inequalities than in larger countries. Finally, the answers to the questions about the influence of decentralisation on regional disparity and about the differing outcomes of decentralisation by size of the country, can only be answered through empirical work.

12.3 Public Sector Size and Parties: Their Influence on Regional Disparities Although the main purpose of this study is to examine the influence of different indicators of decentralisation on regional disparities, we also deal with other political variables, namely those relating to the size of the public sector and the political orientation of the parties in power. The main reason for this is that the omission of potentially relevant variables, especially those that could be correlated with decentralisation, could bias the results. Also, as we are about to explain, there are reasons to suspect that both the size of the public sector and partisan influence could be correlated with decentralisation. 12.3.1 Decentralisation, Public Sector, and Regional Disparities The “hypothesis of decentralisation”, proposed by Brennan and Buchanan (1980), suggests that decentralisation increases competitiveness among local governments and restrains the growth of the public sector. This hypothesis has been tested by Cameron (1978) Oates (1985), Heil (1991), Pierson (1995) and Lane and Ersson (2000) among others. Most of them have found a positive relationship between decentralisation and public sector size when the analysis includes only developed countries. If both variables are related, the size of the public sector should be included in the analysis in order to avoid the problem of omitted explicative variables. There are two opposite lines of arguments that could explain the role of the public sector in regional disparities. The first of these would support the hypothesis that a larger public sector could help to reduce inequalities. A public sector with more capacity to fund regional policies could activate lagging regional economies through subsidies, infrastructure investment, or financing the investment in human capital. This type of support would have a long-term impact on regional productivity and employment. Larger public sectors also have larger social security systems. Public pensions and unemployment benefits reduce disposable income disparities, and may indirectly generate less divergence in terms of employment and productivity in the non-tradable service sector of the poorer regions. Public service salaries are usually similar in all regions, thus high

Does Decentralisation Matter to Regional Inequalities? 217

numbers of civil service employees, for example in a public health system or in education, could also help to reduce disparities. The second argument predicts an increase in disparities, both in productivity and income per capita. Extensive public support for entrepreneurship and the unemployed in the form of subsidies generates fewer incentives to increase the competitiveness of existing firms, to increase human capital or even to accept lowpaid jobs. As the lagging regions are the more seriously affected by unemployment and lack of competitiveness (usually in decaying industries), net transfers from the more developed regions could delay the inevitable restructuring of a region. 12.3.2 Parties in Government, Public Sector and Regional Disparities According to the hypothesis of partisan influence on public policy, the existence of a left/right alternative leads to the application of different policies, with different political and economical outcomes. This hypothesis, which is based on the theory of a democratic political market, has been developed mainly in contributions to empirical democracy theory and research on partisan effects, like those of Castles (1982), Hibbs (1987a), Alesina and Rosenthal (1995) or Schmidt (1996). Although cross section national studies have not yielded conclusive results when studying the linkages between the party composition of government and macroeconomic indicators3, the analysis of country case studies (for example, Hibbs 1987b, for USA) and comparative research (Castles 1989) supports the hypothesis of partisan influence. According to Hibbs (1987a; 1987b), economic policy and macroeconomic outcomes such as unemployment and inflation are largely linked to the left or right wing tendency of governments, with the left more intent on pursuing lower unemployment levels, and the right putting more effort into controlling inflation. The increase in public spending in the post-1960 period is also clearly related to the participation of left-wing parties in government (Schmidt 1996). The correlation between public sector size and growth, and the support found in the literature for the hypothesis that partisan influence on public policy could influence economic outcomes in advanced democratic states, suggest the need to introduce measures of the party composition of government in our analysis. There are arguments linking regional inequalities with public sector size that could also be adapted to relate regional inequalities to partisan influence. First, the strength of conservative parties is associated with weaker social policy. This could lead to an increase in regional inequality resulting from the reduction of transfers to the unemployed (if the lagging regions are the ones with higher unemployment), or more differences in education and health systems, with impact both in the long run (larger disparities in human capital) and in the short run (reduction of the equalising effect of civil service salaries). Implications of the 3

See Schmidt (1996) for a review of this type of studies.

218 Carlos Gil, Pedro Pascual and Manuel Rapún

opposite nature, could be related to the presence in government of parties with a more market-oriented ideology (conservative and liberal) which could reduce the incentives to remain unemployed in the less developed regions, thereby increasing migration (reducing unemployment and increasing per capita capitalisation) and encouraging the mobilisation of endogenous factors. The weight of the public sector in the economy and the party in power may also have a differential impact, depending on the size of the country. For example, unemployment benefits are related to both larger public sector size and the presence of left-wing parties in government. This income could generate greater distortions in small countries, where barriers to the mobility of the work force may be less relevant, than in large ones.

12.4 Measures of Regional Inequality, Decentralisation, Public Sector Size and Parties in Power 12.4.1 Measures of Inequality We will use measures of regional disparities in GDP per worker (GDPpw). Other measures of affluence, such as GDP per capita (GDPpc), are less appropriate for this type of study, because the existence of commuters produces great distortion in some regions. Clear examples of this are Hamburg and Bremen in Germany, the District of Columbia in the USA, and Oslo in Norway. Another disadvantage of GDP per capita is that it is influenced by the age structure of the population and activity rates. In some countries, mainly federal states, such as Germany or USA, the relevant level of regional aggregation is quite clear. In others, such as France or the UK, this could pose a problem. In these countries, the use of NUTS1 or NUTS2 levels, giving the same weight to all regions, could lead to widely differing results. We will use inequality indices weighted by employment, so that the level of aggregation does not heavily influence the results (Esteban 1994)4. It is well known that different indices of inequality are based on alternative ethical assessments. By calculating alternative indices and using them in the analysis, it is possible to ensure that differences between countries are real, and the results robust. In this chapter we will use “sigma”, the standard deviation of the natural logarithm of the GDP per worker, a measure that has become widespread in the analysis of regional disparities in the convergence literature, and also three other measures: Gini, Theil and Atkinson indices. These are more common in the analysis of interpersonal income disparities, but they have been employed in studies dealing with regional inequalities (Tsui 1996, and Esteban 1994). All of them satisfy the Dalton transfer principle, i.e., a transfer from a 4

When the relevant unit is controversial, as in the UK, we have done all calculations on the basis of NUTS 1 and NUTS2 with little significant differences in results.

Does Decentralisation Matter to Regional Inequalities? 219

richer to a poorer region reduces inequality, but they have different degrees of sensitivity. Some of the indices have a high degree of risk or inequality aversion, thus an increase in the value of a very poor region would have a very strong impact in the index. Others have less inequality aversion, thus an increase in the value for a very poor region would have almost the same impact in the index as that of a region slightly below average5. The Atkinson index, A(H), is really an infinite set of indices, which ranks from “none” to “very high” inequality adversity. A low value of the parameter H indicates low inequality aversion (with H=0, A(0)=0). We will calculate Atkinson indices with three different levels of aversion: A(1) with low aversion to inequality, A(3) with medium aversion, and A(21) with high. Table 12.1. Different measures of regional inequalities in GDP per worker Sigma

Gini

Theil

A(1)

A(3)

A(21)

Value rank Value rank Value rank Value rank Value rank Value rank France

0.152

18

0.132

18

0.0126

18

0.0121

18

0.033

18

0.106

Germany

0.057

1

0.040

1

0.0017

1

0.0017

1

0.001

1

0.007

11 1

Italy

0.138

16

0.113

16

0.0089

15

0.0092

16

0.029

16

0.158

17

Spain

0.131

15

0.107

15

0.0080

14

0.0082

15

0.026

15

0.179

19

UK

0.100

7

0.087

9

0.0050

7

0.0050

7

0.015

7

0.086

10

Canada

0.107

10

0.089

10

0.0055

8

0.0056

10

0.017

10

0.144

16

US

0.129

14

0.105

14

0.0114

17

0.0057

11

0.022

14

0.121

13

Japan

0.141

17

0.123

17

0.0095

16

0.0098

17

0.030

17

0.161

18

Small Countries (SC) Austria

0.111

12

0.093

13

0.0065

12

0.0063

13

0.018

11

0.076

9

Belgium

0.114

13

0.087

8

0.0069

13

0.0066

14

0.019

12

0.073

7

Denmark

0.095

6

0.092

12

0.0047

6

0.0046

6

0.013

6

0.055

5

Finland

0.072

5

0.064

5

0.0027

5

0.0026

5

0.008

4

0.039

3

Greece

0.105

9

0.074

6

0.0055

9

0.0054

8

0.016

9

0.121

15

Ireland

14

0.111

11

0.091

11

0.0057

11

0.0059

12

0.019

13

0.121

Netherlands 0.065

3

0.054

3

0.0021

3

0.0021

3

0.006

3

0.040

4

Norway

0.105

8

0.086

7

0.0056

10

0.0055

9

0.016

8

0.076

8

Portugal

0.160

19

0.154

19

0.0131

19

0.0129

19

0.036

19

0.120

12

Sweden

0.058

2

0.049

2

0.0018

2

0.0017

2

0.005

2

0.023

2

Switzerland 0.072

4

0.063

4

0.0025

4

0.0025

4

0.008

5

0.059

6

Average

0.107

0.090

0.006

0.006

0.018

0.093

Average SC 0.097

0.082

0.005

0.005

0.015

0.073

Sources: Cambridge Econometrics, Statistisches Lamdesamt Waden-Wurttemberg (Germany), Statistics Norway, Canadian Statistics, Japan Statistical Yearbook (several years; Statistics Bureau, Management and Co-ordination Agency, Government of Japan), Bureau of Economic Analysis (USA). 5

For a detailed analysis of the properties of the various measures of inequality, see, for example, Champernowne and Cowell (1998).

220 Carlos Gil, Pedro Pascual and Manuel Rapún

We present the results for two different samples. The first includes all the OECD countries for which we have been able to gather data on regional GDP and employment, and indicators of decentralisation, public sector size and partisan influence. This means that we have data for nineteen countries. The second is a sub-sample of eleven small countries. The distinction could pose problems, but in our case it is quite evident that there is a clear cut difference between Spain and Canada, which has the smallest population and GDP among the large countries, or the Netherlands, which scores highest of all the small countries on these two indicators. With this distinction we aim to test the hypothesis of the relevance of size when analysing the influence of decentralisation. In order to avoid distortion in GDPpw generated by the economic cycle, we calculate the indices for the average of the period 1990-1997. Table 12.1 shows the results of these indices. The countries that show the greatest regional inequality on most of the indices are Portugal, France, Japan, Italy and Spain. A(21), however, produces quite a different ranking. On this index, Japan, Italy and Spain continues to register the highest regional inequality, but France and Portugal improve considerably. The countries with the least regional inequality are Germany, Sweden, the Netherlands and Switzerland. These indices suggest a relationship between federalism and regional inequalities that we will attempt to confirm in Section 12.5. On the whole, the ranking is quite similar on all the indices, with the exception of A(21). This idea is further confirmed by the correlation coefficients for these indices, which are shown in Table 12.2. All are very high, except those that include A(21), an index which is highly sensitive to small incomes, regardless of the size of the region. Table 12.2. Correlation coefficients of the inequality indices

Sigma

Sigma

Gini

Theil

A(1)

A(3)

A(21)

-

0.970

0.947

0.927

0.981

0.812 0.722

Gini

0.970

-

0.964

0.962

0.947

Theil

0.947

0.964

-

0.965

0.975

0.694

A(1)

0.927

0.962

0.965

-

0.989

0.730

A(3)

0.981

0.947

0.975

0.989

-

0.812

A(21)

0.812

0.722

0.694

0.730

0.812

-

Average

0.927

0.913

0.909

0.915

0.941

0.754

12.4.2 Measures of Decentralisation Next we present a group of variables that can be used to measure decentralisation. As Martinez-Vazquez and McNab (2003) have stated, this is a problematic issue, because there is no single or best measure of decentralisation. A country may allocate a large fraction of the public budget at the regional level but regions may

Does Decentralisation Matter to Regional Inequalities? 221

lack sufficient autonomy to make decisions on expenditure. It is therefore important to test several alternative measures of decentralisation. Nine measures of decentralisation are featured in Table 12.3. Sources and notes to this table provide precise details of each variable. Since the first five variables focus on political issues, we consider them to be indices of political decentralisation. The last four concentrate on revenues or expenditures, thus we call them indices of fiscal decentralisation. The first measure, Federalism, is a dummy variable that takes the value 1 when the country has a federal constitution. Only four countries have been federations for a relevant period of time. Belgium became federal in 1993, but this is too late for it to have had any impact on regional disparities. The next four variables (Constitutional Structure, Lijphart index, Institutional Pluralism, and Institutional Constraints) measure levels of political restraint to central government intervention. The Lijphart indices are standardised arithmetic means of z-transformed indicators of the federalism-unitarism dimension. The other three variables are additive indices. Fiscal Difficulty measures the capacity of central government to influence economic performance. Fiscal Decentralisation measures the share of regional and local taxes in total revenue. Fiscal Centralisation measures the share of central government revenue in total revenue, excluding supranational and social security taxes, so it is not the mirror image of Fiscal Decentralisation. The reason for excluding these taxes is that central government experiences more difficulty in manipulating these revenues than other taxes under its direct control. The last variable, Financial Autonomy, measures the proportion of local and regional government final consumption in relation to general government final consumption. This variable focuses on expenditures, whilst the previous two focus on the revenue side. 12.4.3 Measures of Public Sector Size and Parties in Power Mention has already been made of previous studies that have found a positive correlation between public sector size and centralisation (at least for developed countries). The detected relationship between regional inequalities and decentralisation may be a spurious correlation, the significant one being the relationship of productivity inequalities either to public sector size, or to parties in government. Table 12.4 includes three variables that measure public sector size, and three others that measure government profiles.

222 Carlos Gil, Pedro Pascual and Manuel Rapún Table 12.3. Decentralisation measures Federal

1

CS

2

3

Lijphart

IP

4

IC

5

6

7

FiscDif

FiscDec

FiscCen 48.9

8

9

FinAut

France

0

2

0.36

3

1

4.7

8.5

30.00

Germany

1

4

-1.79

4

5

6.7

30.8

33.4

77.25

Italy

0

1

0.01

4

3

4.4

2.6

60.8

41.25

Spain

0

1

-0.23

3

2

6.2

8.6

50.2

34.25

UK

0

2

1.4

1

1

3.4

8.8

73.9

38.75

Canada

1

4

-1.22

5

3

5.8

44.7

43.3

69.69

US

1

7

-1.62

6

5

7.9

28.8

41

50.57

Japan

0

2

-1.11

3

2

7.3

25

46.6

72.53

Austria

1

2

-0.37

4.2

21.6

51.8

60.25

Small Countries

1

3

2

Belgium

0

1

0.19

3

3

3.3

4.8

62.2

26.50

Denmark

0

0

0.49

2

2

2.8

29.9

66.9

63.00

Finland

0

1

0.46

3

2

3.8

24.1

59.5

63.25

Greece

0

2

0.64

0

1

4.5

4.3

65.7

25.25

Ireland

0

0

0.76

2

2

3.1

4.4

82.2

48.75

Netherlands

0

1

0.33

2

1

3.4

10

56.4

53.50

Norway

0

1

-0.08

1

2

3.5

20.4

54.4

56.00

Portugal

0

0

0.61

2

1

4.3

4.4

70.1

14.50

Sweden

0

0

-0.06

1

0

3.6

32

49.2

65.25

Switzerland

1

6

-1.53

6

5

11.3

39.9

27

74.75

Average

0.26

1.95

-0.15

2.84

2.26

4.96

18.61

54.92

50.80

Average SC

0.18

1.27

0.13

2.27

1.91

4.35

17.80

58.67

50.09

Federalism is coded: 0=no, 1=yes, from Castles (1999). Constitutional Structure, CS, from Schmidt (1996). Additive index composed of 5 indicators: federalism, parliamentary or presidential; representation, bicameralism, frequency of referendums. High values indicate decentralisation 3 Lijphart index: scale of federalism as developed by Lijphart (1984) and taken from Schmidt (1996). Negative values indicate strong decentralisation. 4 Institutional Pluralism, IP. Additive index of constitutional safeguards for sub national governance and modes of representations, based on Colomer (1995), and taken from Schmidt (1996). High values indicate higher barriers against national dominance. 5 Institutional Constraints, IC. Additive index of federal structures, taken from Schmidt (1996). Measures constraints due to policy harmonisation in the EU, difficulty of amending constitutions, bicameralism, central bank autonomy and referendum. Higher values indicate decentralisation 6 Fiscal Difficulty, FiscDif, is the reduction in central government revenue share that would be required to secure 1 percent of GDP increase in demand, as calculated by Castles (1999) 7 Fiscal Decentralisation, FiscDec. Share of regional and local taxes in total revenue. Averages from 1973, 1983 and 1992, taken from Castles (1999) and calculated from OECD Revenue Statistics. 8 Fiscal Centralisation, FiscCen, is central government revenue as a share of total revenue, averages from 1973, 1983 and 1992, taken from Castles (1999) from OECD Revenue Statistics. 9 Financial Autonomy, FinAut. Share of local and regional government final consumption in relation to general government final consumption. Average for years 1960, 1970, 1980 and 1990, from OECD National Accounts. 2

Does Decentralisation Matter to Regional Inequalities? 223 Table 12.4. Public sector size and party orientation variables Public Sector Size

Party orientation

France

46.8

18.3

28.7

Left percent4 35.1

111.3

2

Germany

43.5

18.8

24.5

77.8

198.0

2

Italy

42.9

16.3

29.0

90.5

201.1

2

Spain

37.2

15.1

22.8

63.6

190.7

3

UK

39.8

21.1

19.2

28.1

84.4

2

Canada

43.0

21.7

21.9

65.2

130.4

3

US

32.5

16.8

16.8

37.8

75.6

1

Japan

28.8

9.5

16.9

0.9

2.2

2

PSS

1

GFE/GDP

2

OPE/GDP

3

Left index5

L/R scale6

Small countries Austria

45.9

18.5

26.4

97.2

256.6

3

Belgium

50.5

15.4

36.6

81.8

208.3

2

Denmark

55.9

26.0

30.3

56.0

190.1

3

Finland

33.0

20.4

12.7

73.5

194.5

3

Greece

37.9

13.8

27.0

100.0

243.3

3

Ireland

37.4

16.3

21.6

30.0

69.7

2

Netherlands

51.6

15.3

36.7

72.1

179.1

2

Norway

46.9

19.8

23.5

85.5

247.3

4

Portugal

36.3

15.7

21.6

27.2

128.8

2

Sweden

59.8

27.6

31.8

87.2

257.3

4

Switzerland

32.4

13.7

17.5

54.4

164.6

2

Average

42.2

17.9

24.5

61.3

164.9

2.5

Average SC

44.3

18.4

26.0

69.5

194.5

2.7

Sources and notes: 1 Public Sector Size: average of overall government revenue and expenditure for 1980, 1990 and 1996, precent of GDP. Source: OECD, National Accounts, several years. 2 Government Final Expenditure for 1980, 1990 and 1996 as percent of GDP. Source: OECD, National Accounts, several years 3 Other Public Expenditure for 1980, 1990 and 1996 as percent of GDP. Source: OECD, National Accounts, several years. 4 Left percent: Percentage of years in power of Left and Left-Centre parties. Source: Schmidt (1996). 5 Left index: additive index of orientation of party in government: 3(percent left) + 2(percent leftcentre) + percent centre. Source: Schmidt (1996) 6 L/R scale: left-right scale of government. A higher number indicates governments more oriented towards the left. Source: Schmidt (1996)

12.5 The Results We begin our research by analysing the correlation between the different measures of decentralisation and the alternative indices of regional inequality. As can be

224 Carlos Gil, Pedro Pascual and Manuel Rapún

seen in Table 12.5, there is a negative correlation between decentralisation and regional inequality. The exceptions are “fiscal difficulties” and IP, which present contradictory signs for some of the inequality indices, although at insignificant levels. Fiscal difficulties is a variable with no clear relation to fiscal decentralisation, and its construction is so clearly open to criticism, that this result is not surprising. Note that with the Lijphart, Fiscal Difficulties and Fiscal Centralisation indices, a positive coefficient supports the hypothesis of a positive relationship between decentralisation and lower regional disparities, although in most cases this is not significant. Table 12.5. Coefficient of correlation between the decentralisation index and regional inequality Political Decentralisation Federal

CS

Lijphart

IP

Fiscal Decentralisation FiscDec

FiscCen

Fin Aut

Sigma

-0.222

-0.095

0.139

0.067 -0.121 -0.007

IC

FiscDif

-0.471**

0.254

-0.611***

Gini

-0.243

-0.150

0.178

0.056 -0.176 -0.024

-0.399*

0.274

-0.558**

Theil

-0.135

0.024

0.050

-0.402*

0.142

-0.597***

A(1)

-0.304

-0.219

0.200

-0.018 -0.232 -0.055

0.148 -0.056

0.061

-0.504**

0.253

-0.625***

A(3)

-0.286

-0.163

0.171

0.036 -0.192 -0.012

-0.488**

0.260

-0.612***

A(21)

-0.142

0.020

0.001

0.139 -0.036

0.163

-0.326

0.165

-0.403*

Average -0.222

-0.097

0.123

0.071 -0.136

0.021

-0.432

0.225

-0.568

Small Countries Sigma

-0.093

-0.278

0.364

-0.175 -0.053 -0.211

-0.614**

0.546*

-0.761***

Gini

-0.078

-0.308

0.324

-0.080 -0.062 -0.171

-0.451

0.486

-0.638**

Theil

-0.106

-0.293

0.339

-0.157 -0.113 -0.188

-0.574*

0.482

-0.774***

A(1)

-0.109

-0.294

0.348

-0.160 -0.111 -0.186

-0.584*

0.497

-0.774***

A(3)

-0.116

-0.294

0.367

-0.170 -0.107 -0.184

-0.607**

0.531*

-0.773***

A(21)

-0.076

-0.074

0.367

-0.250 -0.007 -0.046

-0.686**

0.569*

-0.704**

Av. SC

-0.096

-0.257

0.352

-0.165 -0.075 -0.164

-0.586

0.518

-0.737

*: significant at 10 percent, **: significant at 5 percent; ***: significant at 1 percent

The only highly significant indices are Fiscal Decentralisation and specially Financial Autonomy (percentage of local and regional government consumption in relation to general government consumption). These results may indicate that federalism and decentralisation matter, but only if they lead to greater regional fiscal powers. We will try to confirm this first result by introducing new variables into the analysis6. 6

We have also checked correlation with other variables that could influence the results, as total GDP, GDPpc, population and average population per region, surface and average surface per region, and the number of regions. The correlation of the relative variables (GDPpc, average population and surface of the regions) with the inequality indices are very close to 0. The variables related to the size of the country (total GDP, population, surface and the number of regions) are slightly correlated with inequality. Thus, there

Does Decentralisation Matter to Regional Inequalities? 225

The results for the small countries are similar to those obtained for the full sample. Correlation coefficients are even higher and present the expected signs, although the reduction in the sample means that significance levels do not improve. Again, Fiscal Decentralisation and Financial Autonomy are highly significant. Figure 12.1 shows the relationship between the most significant measure of decentralisation (Financial Autonomy) and sigma for both the small and the large countries. As can be seen, the relationship is very similar, but it is more significant for the small countries sample. 0.18

Small countries Large countries

0.16

Regional inequality

0.14 0.12 R2 = 0.3091 0.10 0.08 0.06 R2 = 0.5789

0.04 0.02 0.00 0

20

40

60

80

100

Financial Autonom y

Fig. 12.1 Sigma and financial autonomy for the large and small countries samples

Table 12.6 shows the correlation of the public sector and party in government variables with the inequality indices. Coefficients between variables related to public sector size and those that measure regional inequality are lower than those reported between decentralisation indices and regional inequality, but still favour the hypothesis that a larger public sector could help to reduce disparities. The are signs that small regions could be different. In the regressions that follow we have also included the number of regions in the country (highly correlated with the inequality indices), but the results are not reported, because this variable is not significant and the main results are not altered.

226 Carlos Gil, Pedro Pascual and Manuel Rapún

presence of leftist parties in government is also negatively correlated with degree of inequality, with significant values for Left percent and Left index. The values for the coefficients of the reduced sample are in this case lower, although they also suggest a positive influence of left wing parties in reducing disparities. The preliminary conclusion of this survey is that the relationship between decentralisation and regional inequality does not appear to be a spurious correlation resulting from the omission of variables controlling for the size of the public sector or political orientation of parties in government. This is because the correlation is stronger with decentralisation than with the other two groups of political indices. Table 12.6. Correlation between public sector size, parties in government and inequality indices PSS

GFE/GDP

Sigma

-0.342

-0.418*

OPE/GDP Left percent -0.122

-0.449*

Left index -0.443*

L/R scale -0.326

Gini

-0.318

-0.326

-0.168

-0.544**

-0.482**

-0.306

Theil

-0.326

-0.375

-0.129

-0.476**

-0.458*

-0.417

A(1)

-0.264

-0.396*

-0.045

-0.437*

-0.386

-0.266

A(3)

-0.332

-0.426*

-0.106

-0.460**

-0.438*

-0.295

A(21)

-0.497**

-0.498**

-0.254

-0.353

-0.454*

-0.211

Average

-0.347

-0.406

-0.137

-0.453

-0.444

-0.304

-0.295

Small countries Sigma

-0.279

-0.346

-0.086

-0.376

-0.305

Gini

-0.256

-0.204

-0.164

-0.530*

-0.380

-0.295

Theil

-0.248

-0.306

-0.071

-0.394

-0.282

-0.293

A(1)

-0.264

-0.316

-0.084

-0.414

-0.307

-0.303

A(3)

-0.297

-0.340

-0.109

-0.445

-0.356

-0.322

A(21)

-0.510

-0.573*

-0.203

-0.339

-0.419

-0.346

Average

-0.309

-0.347

-0.120

-0.417

-0.341

-0.309

*: significant at 10 percent, **: significant at 5 percent; ***: significant at 1 percent

The multivariate analysis of this relationship is restricted because of the small size of the sample. Nevertheless, it is worth testing to see if the above relationships are still valid when more than one variable is included at a time. We have regressed the different measures of regional inequality as dependent variables, and the three types of independent variables. We have regressed all possible combinations of the nine variables measuring decentralisation, the three measuring public sector, and the three partisan variables. The results of some of these regressions are displayed in Table 12.7. In all cases a measure of decentralisation and either an index of the size of the public sector or (alternatively) an index of parties in power7 is included. We do not present the 7

Results of the regressions when the endogenous variables are Gini and A(3) are very similar to those obtained with Sigma, thus they are not included here. Results with the

Does Decentralisation Matter to Regional Inequalities? 227

results for the three types of variables in the same regression because the variables that measure public sector size and parties in government are significantly correlated8. The results are quite encouraging. The signs obtained for all the variables support the positive influence of decentralisation in helping to reduce regional inequalities. Larger public sector size and the presence of left-centre parties in government also seem to favour minor disparities. When A(1) (small risk adversity) and Sigma (medium adversity) are introduced as endogenous variables, Fiscal Decentralisation and Financial Autonomy are significant at levels even below 1 percent. Left percent is significant in all the regressions, whilst results for public sector variables are better when including GFE (as percentage of GDP) than when including PSS. Financial Autonomy is the variable that produces the highest adjusted R2, indicating the relevance of regional spending in the process of keeping regional disparities low. If A(21) is introduced as an endogenous variable, results change slightly. Decentralisation variables and Left percent are less significant, and PSS is more relevant. We must keep in mind that this index is mainly determined by the distance between the poorest region and the average, with minor relevance of regional size and the rest of the units. In this situation, the existence of a large welfare state could be important in alleviating the economic situation of small regions that are lagging behind. We have also tested for differences between large countries and small countries, by performing the LM test of structural change (Table 12.7). There is no evidence to suggest that the results are not valid for the sub-sample of small countries. Table 12.8 shows that the results for the sample of Small Countries. We reproduce only those with fiscal variables, as they have proved the most relevant. These results should be interpreted with great care because of the obvious problem of the lack of degrees of freedom. Again, we obtain some remarkable results. We could expect a decrease in the significant levels of the decentralisation because of the rise in the standard deviations of the errors, but both are significant.

8

Theil index are somewhere in between those obtained with A(1) and Sigma. Among the decentralisation variables, we have selected Federal (which is not comparable to any other variable), and Lijphart and IC (because they are conceptually similar to CS and IP, and are slightly more significant in most regressions). We also present the results for PSS and GFE/GDP, excluding OPE/GDP, because the results for this variable are similar to those obtained with PSS but slightly less significant. Among the parties in government variable, Left% produces similar results to those obtained when the Left index is included in the regressions, and it is clearly more significant than L/R scale. For example, the correlation coefficient between Left% and PSS is 0.489, between Left% and Fiscal Decentralization is 0.02, and between PSS and Fiscal Decentralization is 0.019. We have also estimated regressions with the three types of variables, and tested the results for multi-collinearity (using the Farrar-Glauber test). The results fail to show any problem of multi-collinearity between public sector size and parties in government.

228 Carlos Gil, Pedro Pascual and Manuel Rapún Table 12.7. Regression analysis of regional inequality Depen. varp. A(1)

Structural Party in government or Heteros2 Ad. R2 change1 public sector variables Federal -0.0019 (.253) LEFT -0.48E-4 (.078) 0.163 (.093) 0.581 0.782 -0.0027 (.122) PSS -0.13E-3 (.157) 0.102 (.164) 0.561 0.653 -0.0022 (.18) GFE -0.3E-3 (.086) 0.155 (.101) 0.605 0.723 Lijphart 0.72E-3 (.378) LEFT -0.51E-4 (.063) 0.134 (.122) 0.855 0.453 0.001 (.244) PSS -0.13E-3 (.174) 0.041 (.278) 0.705 0.326 0.0011 (.208) GFE -0.35E-3 (.057) 0.144 (.112) 0.675 0.365 IC -0.5E-3 (.307) LEFT -0.51E-4 (.062) 0.149 (.107) 0.675 0.515 -0.9E-3 (.132) PSS -0.16E-3 (.112) 0.095 (.175) 0.638 0.397 -0.8E-3 (.129) GFE -0.37E-3 (.004) 0.182 (.078) 0.626 0.205 FiscDec -0.12E-3 (.017) LEFT -0.5E-4 (.036) 0.366 (.01) 0.912 0.366 -0.12E-3 (.027) PSS -0.1E-3 (.234) 0.234 (.046) 0.794 0.667 -0.1E-3 (.083) GFE -0.18E-3 (.297) 0.218 (.054) 0.819 0.535 FinAut -0.11E-3 (.002) LEFT -0.48E-4 (.026) 0.502 (.001) 0.824 0.813 -0.11E-3 (.004) PSS -0.1E-3 (.194) 0.385 (.008) 0.701 0.47 -0.99E-4 (.009) GFE -0.2E-3 (.195) 0.384 (.008) 0.569 0.228 Sigma Federal -0.0119 (.445) LEFT -0.47E-3 (.07) 0.134 (.122) 0.716 0.797 -0.0207 (.202) PSS -0.0014 (.095) 0.106 (.159) 0.672 0.907 -0.151 (.329) GFE -0.003 (.075) 0.127 (.131) 0.61 0.82 Lijphart 0.0048 (.542) LEFT 0.5E-3 (.059) 0.123 (.136) 0.824 0.865 0.0084 (.313) PSS -0.0015 (.103) 0.07 (.218) 0.66 0.678 -0.008 (.314) GFE -0.0034 (.054) 0.131 (.127) 0.545 0.75 IC -0.0025 (.6) LEFT -0.5E-3 (.06) 0.117 (.143) 0.818 0.743 -0.0062 (.252) PSS -0.0017 (.082) 0.087 (.188) 0.729 0.969 -0.0051 (.306) GFE -0.0035 (.051) 0.133 (.125) 0.613 0.87 FiscDec -0.0011 (.028) LEFT -0.49E-3 (.035) 0.341 (.013) 0.804 0.977 -0.0011 (.037) PSS -0.0012 (.121) 0.249 (.039) 0.717 0.907 -0.85E-3 (.126) GFE -0.002 (.231) 0.201 (.064) 0.671 0.811 FinAut -0.98E-3 (.002) LEFT -0.47E-3 (.023) 0.494 (.001) 0.69 0.85 -0.001 (.004) PSS -0.0012 (.086) 0.417 (.005) 0.613 0.534 -0.9E-3 (.012) GFE -0.0021 (.155) 0.381 (.008) 0.479 0.365 A(21) Federal -0.0111 (.667) LEFT -0.6E-3 (.164) 0.027 (.313) 0.372 0.439 -0.0273 (.253) PSS -0.0032 (.02) 0.222 (.052) 0.452 0.411 -0.0154 (.518) GFE -0.0057 (.033) 0.176 (.082) 0.291 0.338 Lijphart -0.8E-5 (.999) LEFT -0.62E-3 (.15) 0.015 (.344) 0.542 0.313 0.0077 (.534) PSS -0.0031 (.028) 0.174 (.084) 0.465 0.285 0.0058 (.636) GFE -0.006 (.031) 0.166 (.091) 0.275 0.218 IC -0.0011 (.886) LEFT -0.62E-3 (.15) 0.017 (.34) 0.458 0.376 -0.0087 (.271) PSS -0.0035 (.017) 0.217 (.055) 0.46 0.462 -0.0057 (.45) GFE -0.0062 (.025) 0.184 (.076) 0.254 0.48 FiscDec -0.0012 (.166) LEFT -0.61E-3 (.134) 0.13 (.128) 0.395 0.643 -0.0012 (.136) PSS -0.0029 (.027) 0.266 (.032) 0.467 0.12 -0.6E-3 (.487) GFE -0.005 (.075) 0.18 (.08) 0.382 0.362 FinAut -0.001 (.087) LEFT -0.59E-3 (.135) 0.184 (.076) 0.409 0.943 -0.001 (.059) PSS -0.0029 (.022) 0.327 (.016) 0.489 0.174 -0.79E-3 (.175) GFE -0.0049 (.061) 0.248 (.039) 0.361 0.443 In parenthesis: significant level (derived from t and F statistics). In bold: significant at 10 percent level. 1 Probability of the LM test for Structural Change (small countries/large countries). Large values indicate that we cannot reject the null hypothesis 2 Probability of the Breush-Pagan LM test for heteroscedasticity. Large values indicate that we cannot reject the null hypothesis (homoscedasticity). Decentralisation variable

Does Decentralisation Matter to Regional Inequalities? 229

Estimated values are higher than in the full sample, suggesting that the role of decentralisation in reducing regional disparities is even more important in the small countries. However, the other two groups of variables (public sector size and partisan variables) are less significant, although still with the expected sign. Table 12.8. Regression analysis of regional inequality with decentralisation, public sector size and party in government (Left percent) for Small Countries Depen. varp. A(1)

Party in government or Heteros(1) Adjusted R2 public sector variable FiscDec -0.13E-3 (.075) LEFT -0.43E-4 (.23) 0.32 (.087) 0.208 -0.13E-3 (.085) PSS -0.59E-4 (.545) 0.216 (.155) 0.276 -0.14E-3 (.125) GFE -0.27E-5 (.99) 0.176 (.188) 0.251 FinAut -0.12E-3 (.006) LEFT -0.39E-4 (.152) 0.618 (.008) 0.622 -0.12E-3 (.009) PSS -0.46E-4 (.547) 0.521 (.021) 0.36 -0.13E-3 (.013) GFE 0.43E-4 (.805) 0.502 (.025) 0.384 Sigma FiscDec -0.13E-2 (.059) LEFT -0.35E-3 (.284) 0.331 (.082) 0.68 -0.13E-2 (.065) PSS -0.58E-3 (.508) 0.265 (0.119) 0.664 -0.14E-2 (.106) GFE -0.15E-3 (.944) 0.222 (.150) 0.539 FinAut -0.11E-2 (.008) LEFT -0.32E-3 (.225) 0.567 (.014) 0.313 -0.11E-2 (.011) PSS -0.48E-3 (.507) 0.504 (.024) 0.932 -0.12E-2 (.018) GFE 0.11E-3 (.946) 0.474 (.031) 0.944 A(21) FiscDec -0.17E-2 (.028) LEFT -0.35E-3 (.331) 0.416 (.047) 0.424 -0.17E-2 (.02) PSS -0.15E-2 (.089) 0.548 (.017) 0.898 -0.14E-2 (.101) GFE -0.21E-2 (.345) 0.411 (.0449) 0.403 FinAut -0.12E-2 (.022) LEFT -0.34E-3 (.33) 0.444 (.039) 0.885 -0.11E-2 (.016) PSS -0.15E-2 (.091) 0.568 (.014) 0.553 -0.1E-2 (.066) GFE -0.23E-2 (.274) 0.462 (.343) 0.726 In parenthesis: significant level. In bold: significant at 10 percent level. (1) Probability of the Breush-Pagan LM test for heteroscedasticity. Large values indicate that we cannot reject the null hypothesis (homoscedasticity). Decentralisation variable

12.6 Conclusion Our main finding is that decentralisation, especially fiscal decentralisation, appears to be linked with lower regional disparities in labour productivity. The relationship between decentralisation and regional equality does not weaken when other explanatory variables related to public sector size and parties in government are included. Also remarkable is the difference in results that we obtain from different measures of decentralisation. The most significant, which also gives the highest adjusted R2, is the one most closely related to fiscal decentralisation. Castles (1999), in his analysis of the link between decentralisation and economic performance in a sample similar to ours, also found evidence to support the hypothesis that it is fiscal decentralisation rather than political structure that matters. The consequences of political and fiscal decentralisation are not merely a question of academic concern. Reduction of regional inequalities is one of the most important issues in the regional policy of the EU. If changes in the administrative level at which certain political and budgetary decisions are taken

230 Carlos Gil, Pedro Pascual and Manuel Rapún

can help to reduce inequalities, it would matter, not merely in the sense that we would know more, but in the sense that we could do more. Our results also suggest a larger public sector to relate positively to left and centre parties in power and to a lower degree of regional inequalities. A second objective of this chapter was to ascertain whether the influence of political factors, particularly decentralisation, on regional disparities might be different in small countries. Though some theorists might question the relevance of the geographical distribution of power in these states, our results suggest that we can expect decentralisation to have the same, or even a greater degree of positive influence in small countries than in large countries. Although federal governance in small countries is unlikely, it is not impossible (Switzerland is a federal state), and fiscal decentralisation can be achieved without federalisation, as illustrated by the case of Sweden. Perhaps devolution to local rather than regional administrations is the most feasible way of bringing power closer to the citizens and, eventually, of reducing regional disparities in small countries.

References Alesina A, Rosenthal H (1995) Partisan politics, divided government, and the economy. Cambridge University Press, Cambridge Armstrong H, Taylor J (2000) Regional economics and policy. Blackwell Publishers, Oxford Barro RJ (1990) Government spending in a simple model of economic growth. Journal of Political Economy 98-2:103-125 Barro RJ (1991) Economic growth in a cross-section of countries. Quarterly Journal of Economics 1062, pp 407-443 Barro RJ, Sala-i-Martín X (1991) Convergence across states and regions. Brookings papers on Economic Activity 1:107-182 Brennan G, Buchanan JM (1980) The power to tax. Analytical foundations of a fiscal constitution. Cambridge University Press, Cambridge Cameron D (1978) The expansion of the public economy: a comparative analysis. American Political Science Review 72-4:1243-1261 Castles F (1982) The impact of parties. Politics and policies in democratic capitalist states. Sage Publications, London Castles F (1989) The comparative history of public policy. Polity Press, Cambridge Castles F (1999) Decentralisation and the post-war political economy. European Journal of Political Research 36-1:1-26 Champernowne DG, Cowell FA (1998) Economic inequality and income distribution. Cambridge University Press, Cambridge Colomer J (1995) Introducción. In: JM Colomer (ed) La política en Europa. Ariel, Barcelona, pp 7-25 Dillinger W (1994) Decentralisation and its implication for urban service delivery. Urban Management Programme Discussion Paper 16. World Bank, Washington Elazar DJ (1995) Federalism. In: Lipset SM (ed) The encyclopedia of democracy. Routledge, London, pp 472-482 Esteban J (1994) La desigualdad interrregional en Europa y en España: descripción y análisis. In: Esteban J, Vives X (eds) Crecimiento y convergencia regional en España y

Does Decentralisation Matter to Regional Inequalities? 231 Europa. Instituto de Análisis Económico y Fundación de Economía Analítica, Barcelona, pp 13-84 Freinkman L, Yossifov P (1999) Decentralisation in regional fiscal systems in Russia: trends and links to economic performance. World Bank Working Paper 2100, Washington Heil (1991) The search for leviathan revisited. Public Finance Quarterly 19:334-346 Hibbs DA (1987a) The political economy of industrial democracies. Harvard University Press, Cambridge Hibbs DA (1987b) The American political economy. Harvard University Press, Cambridge Kuznets S (1966) Modern economic growth: rate, structure and spread. Yale University Press, New Haven Lane JE, Ersson S (2000) The new institutional politics. Performance and outcomes. Routledge, London, New York Levine R, Renelt D (1992) A sensitivity analysis of cross-country growth regressions. American Economic Review 82-4:942-963 Lijphart A (1984) Democracies. Yale University Press, New Haven, London Martinez-Vazquez J, Mcnab R (2003) Fiscal decentralisation and economic growth. World Development 31:1597-1616 Musgrave R (1959) The theory of public finance. McGraw-Hill, New York Myrdal G (1957) Economic theory and underdeveloped regions. Duckworth, London North DC (1990) Institutions, institutional change, and economic performance. Cambridge University Press, New York Oates WE (1972) Fiscal federalism. Harcourt Brace Jovanovich, New York Oates WE (1985) Searching for leviathan: an empirical study. American Economic Review 75:748-757 Oates WE (1999) An essay on fiscal federalism. Journal of Economic Literature 37:11201149 Pierson P (1995) Fragmented welfare states: federal institutions and the development of social policy. Governance 8:449-478 Prudhomme R (1995) On the dangers of decentralisation. Policy Research Working Paper 1252, World Bank, Washington Rodriguez-Pose A (1998) The dynamics of regional growth in Europe: social and political factors. Oxford University Press, Oxford Schmidt MG (1996) When parties matter: a review of the possibilities and limits of partisan influence on public policy. European Journal of Political Research 30:155-183 Schumpeter J (1954) Capitalism, socialism, and democracy. 4th edition. Allen and Unwin, London Tanzi V (1996) Fiscal federalism and decentralisation: a review of some efficiency and macroeconomic aspects. In: Bruno M, Pleskovic B (eds) Annual World Bank conference on development economics. World Bank, Washington, pp 295-316 Tomaney J, Ward N (2000) England and the “new regionalism”. Regional Studies 345:471-478 Tunstall R (2001) Devolution and user participation in public services: how they work and what they do. Urban Studies 38-13:2495-2514 Tsui K (1996) Economic reform and inter-provincial inequalities in China. Journal of Development Economics 50:353-368 Zhang T, Zou H (1998) Fiscal decentralisation, public spending, and economic growth in China. Journal of Public Economics 67-2:221-240 Zhao X, Tong S (2000) Unequal economic development in China: spatial disparities and regional policy reconsideration, 1985-1995. Regional Studies 34-6:549-561

13

Regional Inequalities in the EU Enlargement Countries: An Analysis of Small Versus Large New Member States

George Petrakos, Yiannis Psycharis and Dimitris Kallioras Department of Planning and Regional Development, University of Thessaly, Volos, Greece

13.1 Introduction The problem of unequal spatial distribution of income, economic opportunities and activities at a national and international level represents an important theoretical and practical issue, with major economic, social and political impact. A growing literature is now concerned with the regional aspects of the transition process and the type and evolution of regional disparities in Central and Eastern Europe. A number of earlier studies argue that the process of transition in Central and Eastern Europe is associated with increasing regional disparities. Petrakos (1996a, 2000) and Petrakos and Totev (2000) have attempted a theoretical inquiry concerning the interaction of the various processes of transition over space. They claim that transition will have a serious impact on the regional structure of Central and East European countries because the processes of internationalisation and structural change tend to favour metropolitan and western regions, as well as regions with a strong industrial base. In addition they claim that at the macrogeographical level the process of transition will increase disparities at the European level, by favouring countries near the East-West frontier. At the same time, a number of empirical papers have appeared suggesting disparities at the national level. Evidence from Estonia shows that core-periphery differences have increased, with Talin and Western coastal regions benefiting the most from the new orientation of the country (Raagmaa 1996). Evidence from East Germany already indicates that development is highly selective and depends on the behaviour of foreign capital. Berlin emerges as a development pole strongly linked to the West German and the international economies but with weak local linkages and low spread effects (Hausermann 1993). Similar trends have been detected in Slovakia, where Bratislava, with 9 percent of the national population, generates 30 percent of the country’s GDP (Balaz 1996). In Hungary, disparities increased during the early years of transition (Fazekas 1996), although regional unemployment patterns have remained stable (Fazekas 2000). FDI and domestic capital prefer metropolitan and western regions (Lorentzen 1996, 1999), turning the unbalanced pre-1989 situation of the regions into a serious core-periphery and east-west disparity (Nemes-Nagy 2000). Additional evidence comes from Poland

234 George Petrakos, Yiannis Psycharis and Dimitris Kallioras

(Gorzelak 2000), indicating that different regions adjust in a different way to the new economic environment. Another study (Ingham et al. 1996) has shown that the regional pattern of unemployment was relatively stable in the 1990-1994 period, indicating that initial best performing regions are also final best performing regions and initial losers are also final losers. This basic picture is also supported by reports for Albania (Petrakos 1996b, 1997), Bulgaria (Minassian and Totev 1996; Petrakos 1996b, 1997) and Romania (Ramboll 1996; Constantin 1997). A comparative regional analysis of Poland, Hungary, Romania and Bulgaria by Petrakos (2001) has suggested that the level of disparities is affected by national characteristics (such as institutional factors), economic factors (such as the level of development), the success of restructuring and catching up, as well as by size and the geographic co-ordinates of each country in the European spaceeconomy. A comparative analysis of the spatial structure of South-eastern Europe by Petrakos and Economou (2002) has found increasing regional disparities in all countries, an increasingly superior performance of the metropolitan regions, serious discontinuities at the borders which have, generated over-time border regions with below average performance and finally, an urban system with serious deficiencies in medium sized cities. Although the process of spatial adjustment to the forces of transition is very complex, the available evidence seems to reveal some general patterns. It shows that in countries sharing common borders with the EU and at a short distance from the European core, spatial adjustments have been favoured metropolitan and western regions. However, disparities have increased at various rates and degrees in all transition countries to levels that are higher than those in most of the EU countries. Given the findings of the literature in the early years of transition, two important questions arise: The first is concerned with the evolution of disparities in the more recent years. Have disparities continued to increase in the second half of the 1990s in all or most countries? This is an important question, for a number of reasons. If disparities increase, transition countries will be forced sooner or later, to shift the focus of their public policy and design more effective regional policies. Also, a large number of transition countries have already been accepted as the new members of the EU. It is already known that the new EU-27 will be characterized by a higher level of inter-state disparities, as the new members are characterized by a lower or significantly lower GDP per capita compared to the EU-15 average. Will the new members also be characterized by higher intracountry disparities? The answer to this question is crucial for the design of effective structural and cohesion policies on behalf of the EU. The second question is concerned with one of the driving forces of inequality. Does “size” mater? Or in other words, have small countries a tendency to generate more equal or more unequal distributions of production activities and income over space than the large ones? Can we talk about a “size” effect in regional disparities in the new EU member States? This question is important for the design of development policies. Traditionally, small countries were considered to be almost “dimensionless” or “one-region-economies”, where spatial variation of income is

Regional Inequalities in the EU Enlargement Countries 235

insignificant. This, in turn, affects the mix of development policies, which is disproportionately in favor of sectoral policies. In this chapter we analyse the level and evolution of regional inequality in the transition countries of EU enlargement over the period 1995-2000. Our goal is to examine whether inequalities have continued to increase in the second half of the 1990s, verifying the projections of recent literature, and whether they exhibit a “size” effect. In the next section we briefly present the basic demographic, economic and regional characteristics of the countries under examination. In Section 13.3, we explain the methodological issues related to the measurement of regional inequalities, while in Section 13.4, we present our findings. Section 13.5 concludes the chapter.

13.2 Basic Economic, Demographic and Regional Characteristics Our analysis includes the ten transition countries accepted to join the EU in 2004 or 2007. In Table 13.1, we divide these countries in two groups according to their population size. The first group of very small or small countries includes Estonia, Slovenia, Latvia, Lithuania and Slovakia. These five countries have a size ranging from 1.3 m (in the case of Estonia, which is the smallest EU enlargement country) to 5.4 m people (in the case of Slovakia). The second group of medium and large countries includes Bulgaria, Hungary, the Czech Republic, Romania and Poland. These countries have a size ranging from 8.1 m (in the case of Bulgaria) to 38.6 m people (in the case of Poland, which is the largest EU enlargement country).

Medium / Large

Very Small / Small

Table 13.1. Basic demographic and economic characteristics of EU Enlargement Countries ranked by size Country

Size (Sq. Km)

Estonia Slovenia Latvia Lithuania Slovakia Bulgaria Hungary Czech Rep. Romania Poland

45,228 20,273 64,589 65,300 49,035 110,910 93,029 78,860 238,391 312,685

Sources: Eurostat REGIO database.

Population Population GDP (‘000 Density ( bn) people) (Inhabitants / Sq. Km) (2000) (2000) (2001) 1,370 30.3 6,172 1,990 98.2 21,750 2,373 36.7 8,593 3,696 56.6 13,259 5,401 110.1 22,847 8,170 73.7 15,203 10,024 107.8 57,773 10,273 130.3 63,849 22,443 94.1 44,383 38,646 123.6 204,053

Per Capita GDP ( ) (2001) 4,520 10,920 3,650 3,810 4,250 1,910 5,670 6,220 1,980 5,280

236 George Petrakos, Yiannis Psycharis and Dimitris Kallioras

Although the difference in size between the two groups is more or less evident, the differences in terms of the other characteristics are not that obvious. The small countries group includes more low-density countries than the large countries group, although the exceptions of Slovakia (in the first group) and Bulgaria (in the second) blur the picture. In terms of GDP, the small countries’ group tends to include countries with lower GDP than the large countries’ group. This rule, however, does not hold in the cases of Slovenia and Slovakia that have higher GDP figures than Bulgaria. In terms of GDP per capita, there is serious variation within each group. The small countries’ group includes figures ranging from 3.6 to 10.9 thousand, while the large countries’ group includes figures ranging from 1.9 to 6.2 thousand. On the average, the small countries’ group fares better, as it includes Slovenia, the country with the highest figure, while the large countries’ group fares worse, as it includes Bulgaria and Romania, the countries with the lowest GDP per capita figures. Information about the regional structure of the countries is given in Table 13.2. The general rule is that small countries have a smaller number of NUTS III regions than large countries. Moreover, the average population size of the regions in small countries is smaller than that of the regions in large countries. Of course, there are exceptions. For example, the average regional size in Slovakia is larger than that of Bulgaria, Hungary and Romania, while the average regional size of Bulgaria is smaller than that of Latvia, Lithuania and Slovakia. This is not, however, an unusual situation. Large countries with small regions or small countries with large regions can be found also in the EU-15, as regional boundaries have been shaped by historical or political factors and do not follow any common administrative rule. In summary, we can claim that, although significant exceptions exist, small countries in our case tend to be characterized by a lower population density, a lower GDP, a higher GDP per capita and less and smaller NUTS III regions. Table 13.2. Basic regional characteristics of EU Enlargement Countries, 2000

Medium / Large

Very Small / Small

Country Estonia Slovenia Latvia Lithuania Slovakia Bulgaria Hungary Czech Rep. Romania Poland

Sources: Eurostat REGIO database.

Number of NUTS III regions 5 12 5 10 8 28 20 14 42 44

Average regional population (000 people) 274 166 475 370 675 292 501 734 534 878

Regional Inequalities in the EU Enlargement Countries 237

13.3 Measures of Regional Inequality Regional inequality can be measured employing a variety of variables, the most common of which is probably GDP per capita, by using either diagrammatic or statistical methods. Diagrammatic analysis is usually employed in order to examine the spread of regional values around the national average, as well as the evolution of the spread of regional observations over time. Statistical analysis aims to construct indices that can depict the level and the evolution of inequalities. The most commonly used index of inequality is the coefficient of regional variation (V/ x ) - or V-convergence coefficient1 - defined as the population weighted standard deviation of a variable divided by its mean value. The weighted coefficient of variation is a dimension-less index that allows cross-country and over time comparisons of the level of regional disparities. The value of the coefficient is basically determined by the value of the standard deviation of a variable and as a result it is affected by all observations. In principle, the greater its value, the greater is the level of regional disparities. Another frequently used index is the max/min ratio, which is also a dimension-less index of disparity, but its value is affected only by the two extreme observations of the variable under consideration. In principle again, the greater its value, the greater is the spread of the observations and the greater the level of disparities. Finally, the literature has extensively used the E-convergence coefficient,2 which estimates over time trends of convergence or divergence among regions. Positive values of E imply that regions with higher initial value of GDP per capita tend to experience higher growth. Negative values of E imply that regions with 1

The weighted coefficient of variation is estimated from the formula: 2

CVW = [ ¦ ( X i - x ) u ( Pi / P )]

1/ 2

/ x

t

2

where Xi is the price of per capita GDP in regional level, x is the price of GDP per capita in average country value, Pi is the population in regional level and P is the national population. It is the weighted towards population square standard error divided to the mean value of per capita GDP (V/ x ). The basic neoclassical E-convergence model for the evaluation of convergence or divergence trends across countries or regions adopts the form

1 1 - e Et Y ln( i,t ) = D + ln Y i,t -T ( ) + H i,t -T T T Y i,t -T where Yi,t represents GDP per capita of the country or region i; T is the period of analysis; E is the coefficient and H is the error term. In the present article, the E-convergence coefficient is estimated from the linear version of this model given by the regression:

Y2000/Y1995 = D + EY1995 + H, where Y is the per capita GDP value, D is the constant term, E is the convergence coefficient and H is the disturbance term. The Y2000/Y1995 ratio indicates the growth of regional GDP per capita in the period 1995-2000.

238 George Petrakos, Yiannis Psycharis and Dimitris Kallioras

lower initial value of GDP per capita tend to experience a faster growth. This indicates that positive values of E-convergence coefficient are associated with regional divergence, while negative values are associated with regional convergence. In order to study the level and evolution of regional disparities for the countries under consideration on a comparative basis, we use the Gross Domestic Product (GDP) per capita at the regional (NUTS III) level for the period 1995-2000. The results are presented in Table 13.3 and Figures 13.1 and 13.3-13.6.

13.4 Regional Inequality in the EU Enlargement Countries In Table 13.3 we present the weighted coefficient of variation and the max/min ratio for all countries of our sample and for the years 1995 and 2000. We also present the estimated value of the E-convergence coefficient in the period 19952000. On the basis of this information we can make a number of observations. First, in all countries, with the exception of Bulgaria, the coefficient of variation and the max/min ratio have increased between 1995 and 2000. The 2000 figures are in several cases higher than the figures of EU-15 member states of similar size (Petrakos et al. 2003), indicating that the process of transition in the 1990s has been associated with rapidly increasing regional disparities and that the EU-27 will have to deal with significant inter-national and also intra-national disparities. Second, we observe that the country with the highest coefficient of variation in 2000 is Latvia, followed by Hungary, Estonia and Poland. The country with the lowest coefficient of variation is Slovenia, followed by Lithuania and Bulgaria. This means that the small countries group includes the members with the highest and the lowest level of inequalities, as measured by the coefficient of variation. Note that the picture in 2000 is significantly different from that in 1995. For example, Latvia jumped from the 6th to the 1st place, while Bulgaria slipped from the 4th to the 8th place and Romania has jumped from the 8th to the 5th place. Third, we observe that in 2000 the regional max/min ratio was highest in Poland, followed by Latvia and Romania and lowest in Slovenia followed by Lithuania. In 1995, the 1st place was also taken by Poland, while the 2nd by Slovakia. The last two places were again taken by Slovenia and Lithuania. Fourth, we observe that the value of the E-convergence coefficient is positive and statistically significant for almost all EU enlargement countries (with the exception of Bulgaria, which has negative and statistically significant E-convergence coefficient). This indicates that over the period 1995-2000, the more advanced regions in each country grew at a faster rate than the less advanced regions. This trend has led to further divergence between the rich and the poor regions in the end of the period. Although the use of the E-convergence coefficient for the evaluation of regional inequalities has been questioned in the literature (Petrakos et al. 2003), the fact that almost all countries present the same trend and the fact that other indices of inequality provide similar results allow us to conclude

Regional Inequalities in the EU Enlargement Countries 239

that EU enlargement countries have been facing significant and increasing inequalities over the last period. Judging from the size of the coefficients, the phenomenon seems to be more intense in small countries (with the exceptions of Slovenia and Slovakia), probably attributable to the dominant role of the metropolitan centres in countries of such size. Table 13.3. Regional inequalities in the EU Enlargement Countries, NUTS III level: 1995 and 2000

Country Estonia

Very Small / Small

Slovenia

Latvia

Lithuania

Slovakia

Bulgaria

Medium / Large

Hungary

Czech Rep.

Romania

Poland

Measure of Inequality CVW max / min E - convergence CVW max / min E - convergence CVW max / min E - convergence CVW max / min E - convergence CVW max / min E - convergence CVW max / min E - convergence CVW max / min E - convergence CVW max / min E - convergence CVW max / min E - convergence CVW max / min E - convergence

1995 0.463 2.164

Year Ranking 2000 Ranking 2 0.562 3 6 2.718 7 2.802 (2.153)

0.207 1.681

9 9

0.236 10 1.780 10 0.273 (1.078)

0.341 2.041

6 8

0.747 1 4.327 2 18.454 (2.876)

0.156 1.574

10 10

0.314 9 2.432 9 20.744 (4.122)

0.372 3.080

5 2

0.414 7 3.486 5 0.428 (1.043)

0.394 2.878

4 4

0.391 8 2.617 8 -2.462 (-0.970)

0.483 3.054

1 3

0.583 2 3.597 4 1.444 (2.302)

0.328 2.359

7 5

0.448 6 2.765 6 0.934 (2.620)

0.211 2.140

8 7

0.478 5 4.316 3 6.115 (2.874)

0.415 4.213

3 1

0.527 4 5.188 1 0.896 (2.136)

Source: Author’s estimation from Eurostat REGIO database.

PC GDP 2000 (% country average)

240 George Petrakos, Yiannis Psycharis and Dimitris Kallioras

350 300 250 200 150 100 50 0 EST

SLN

LAT

LIT

SLK

BUL

HUN

CZE

ROM

POL

COUNTRIES (ranked by population size)

Fig. 13.1. The dispersion of GDP per capita around the national average: countries ranked by size, NUTS III level, 2000 Source: Author’s estimation from Eurostat REGIO database.

Figure 13.1 presents the distribution of regional GDP per capita around the national average of each country in 2000. Note that the countries under examination have been placed in the Figure with an increasing population order, i.e. the smallest country (Estonia) is placed first, while the largest country (Poland) is placed last. The first observation in this Figure is that nearly all countries have a metropolitan or core-periphery structure. This can be asserted in the Figure by the large distance between the first observation (which is the metropolitan region) and the rest of the observations. Most of the countries have developed an intense metropolitan structure and some of them, such as Estonia, the Czech Republic, Hungary, Poland, Latvia and Slovakia, have metropolitan centres with GDP per capita figures which are at least twice the figure of the national average. Based on Figure 13.1, we could say that the countries with the smaller regional spread of GDP per capita (or a more balanced regional allocation of per capita GDP) are Slovenia, Lithuania and Bulgaria. The results of Figure 13.1 are also verified by Figure 13.2, which shows that a pattern of metropolitan dominance in a number of countries. Figure 13.2 also shows that in the Enlargement countries bordering EU-15 members, an east-west pattern of development is also present, as western border regions in general fare better than eastern ones. In Figure 13.3 we plot GDP per capita growth rates on initial GDP per capita for the two size groups, using NUTS III data. The slope of the fitted lines (or the regression line) is the estimated E-convergence coefficient of the two groups. We observe that both country groups are characterized by regional divergence, as the E coefficients in both regression lines are positive. However, the coefficient of the large countries’ group is larger (with a steeper slope) and statistically significant,

Regional Inequalities in the EU Enlargement Countries 241

while the coefficient of the small countries group is smaller (with a flatter slope) and statistically insignificant.

Fig. 13.2. GDP per capita at the NUTS III level (country average=100), 2000 Source: Eurostat REGIO database.

242 George Petrakos, Yiannis Psycharis and Dimitris Kallioras

ALL COUNTRIES GDPGROWTH95-00 = 157.857 + 0.278uPCGDP95 (5.506) (0.926) VERY SMALL / SMALL COUNTRIES GDPGROWTH95-00 = 258.735 + 0.068uPCGDP95 (3.643) (0.888) MEDIUM / LARGE COUNTRIES GDPGROWTH95-00 = 98.779 + 0.670uPCGDP95 (3.842) (2.517) Fig. 13.3. E-convergence for EU enlargement countries; NUTS III level, 1995 - 2000 Source: Author’s elaboration from Eurostat REGIO database.

Regional Inequalities in the EU Enlargement Countries 243

In the next three figures we perform a number of different exercises, in order to further examine the relation between the level of regional inequality and country size. In Figure 13.4 we present the coefficient of variation of the EU enlargement countries for 1995 and 2000 in such a way that countries are placed on the horizontal axis in increasing population order. The relationship of regional inequality and country size does not seem to be an obvious one in this Figure. Small countries, like Estonia and Latvia, have indices of inequality as high or even higher than large countries like Hungary and Poland. 0,8 CV (NUTS III level)

0,7 0,6 0,5 0,4 0,3 0,2 0,1 0 EST

SLN

LAT

LIT

SLK

BUL

HUN

CZE

ROM

POL

COUNTRIES (ranked by population size) CV 1995

CV 2000

Fig. 13.4. The weighted coefficient of variation of countries ranked by size: NUTS III level, 1995 and 2000 Source: Author’s estimation from Eurostat REGIO database.

In Figure 13.5, we look at the same issue from a different angle. We plot the coefficient of variation on the vertical axis and the population size of the countries on the horizontal axis for period 1995-2000. This figure gives a somewhat different view, as inequality and size seem to be related in a positive way. The fitted regression line has a positive slope, indicating that larger countries, on the average will be more likely to be associated with higher regional disparities. However, a careful examination of the figure indicates that this relationship is not particularly strong for at least two reasons. Firstly, the data exhibits high volatility in small-sized countries, where we have very high and very low values of the coefficient of variation. Moreover, the slope of the regression line is very small and statistically insignificant.

244 George Petrakos, Yiannis Psycharis and Dimitris Kallioras

CV95-00 = 0.378 + 1.373u10-6uPOP95-00 (16.026) (0.889) Fig. 13.5. The weighted coefficient of variation and the size of the countries, 1995-2000 Source: Author’s estimation from Eurostat REGIO database.

Finally, in Figure 13.6 we undertake a somewhat different exercise, which is based on group average figures. On the basis of the coefficient of variation for each country in the period 1995-2000, we construct an average figure of inequality for the small / very small countries and an average figure for the medium / large countries. These two average figures are weighted by the national population of each country with the intent of measuring the level of regional inequality of the representative “small” and the representative “large” country. The average figures are shown by the two lines in Figure 13.6. We observe that on average, the larger countries in our sample tend to have a slightly higher index of regional inequality than the smaller ones. The gap however between the two lines tends to become smaller over time. We also observe that the small countries average is the

Regional Inequalities in the EU Enlargement Countries 245

weighted sum of figures that have relatively greater spread than that of the larger countries, especially over the last two years. This finding of convergence between the two groups seems to run in parallel with the findings derived from Table 13.3 (increasing intra-country variations) as the two groups of countries have experienced increasing inequalities over time. 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1995

1996

1997

1998

1999

2000

VERY SMALL / SMALL COUNTRIES' CVw MEDIUM / LARGE COUNTRIES' CVw Fig. 13.6. The evolution of the average coefficient of variation of small and large countries over the period 1995 - 2000 Source: Author’s estimation from Eurostat REGIO database.

13.5 Conclusions This chapter has used regional GDP per capita data for 10 EU Enlargement countries and for the period 1995-2000 in order to investigate whether regional disparities continue to increase in the late 1990s and whether the level of inequality is affected by a national “size” effect. For that purpose, we estimated the weighted coefficient of variation, the max/min ratio and the E-convergence index at the NUTS III level for each country, also using diagrammatic analysis. Our findings indicate that regional inequalities have continued to increase in transition countries in the late 1990s, perhaps with a greater speed than initially expected and certainly to levels that are comparable or higher than respective EU15 members. The results also indicate that the small or very small EU enlargement countries have similarly high levels of regional inequality with medium or large countries in the late 1990s. There is some weak evidence that larger countries tend to have relatively higher levels of inequality, but the differences between the two groups

246 George Petrakos, Yiannis Psycharis and Dimitris Kallioras

tend to diminish over time. On the other hand, the inequality index of the small group has a much greater spread than that of the large group, since it is affected by the highest (Latvia) and the lowest (Slovenia) values in our sample. This means that, although in average the two groups have almost similar inequality indices, the probability to experience an extreme value is higher in the small countries group. Overall, our findings indicate that small EU Enlargement countries may be characterized by relatively high levels of regional inequality that may not differ significantly from those of the large countries. As a result, domestic and EU policy makers need to realize that their development policies would be well served by incorporating a clear regional dimension.

References Balaz V (1996) ‘The wild East’? Capital markets in the V4 countries. European Urban and Regional Studies 3(3):251-266 Constantin D (1997) Institutions and regional development strategies and policies in the transition period: the case of Romania. Chapter presented at the 37th European Congress of The European Regional Science Association, Rome 26-29 August 1997 Fazekas K (1996) Types of microregions, dispersion of unemployment and local employment development in Hungary. Eastern European Economics 34(3):3-48 Fazekas K (2000) Regional labor market differentials during transition in Hungary. In: Petrakos G, Maier G, Gorzelak G (eds) Integration and transition in Europe: the economic geography of interaction. Routledge Publications, pp 150-169 Gorzelak G (2000) The dilemmas of regional policy in the transition countries and the territorial organization of the state. In: Petrakos G, Maier G, Gorzelak G (eds) Integration and transition in Europe: the economic geography of Interaction. Routledge Publications, pp 131-149 Hausermann H (1993) Regional perspectives of East Germany after the unification of the two Germanies. URDP, Special Edition, Topos, Athens Ingham M, Grime K, Kowalski J (1996) A geography of recent Polish unemployment. European Urban and Regional Studies 3(4):353-363 Lorentzen A (1996) Regional development and institutions in Hungary: past, present and future development. European Planning Studies 7(4):463-482 Lorentzen A (1999) Industrial development, technology change, and regional disparity in Hungary. European Planning Studies 4(3):259-277 Minassian G, Totev S (1996) The Bulgarian economy in transition: the regional aftereffect. Eastern European Economics 34(3):49-92 Nemes-Nagy J (2000) The new regional structure in Hungary. In: Petrakos G, Maier G, Gorzelak G (eds) Integration and transition in Europe: the economic geography of Interaction. Routledge Publications, p 170-186 Petrakos G (1996a) The regional dimension of transition in Eastern and Central European countries: an assessment. Eastern European Economics 34(5):5-38 Petrakos G (1996b) The new geography of the Balkans: cross-border co-operation between Albania, Bulgaria and Greece. Series of Transition in the Balkans, Vol. 1, University of Thessaly Press, Volos

Regional Inequalities in the EU Enlargement Countries 247 Petrakos G (1997) The regional structure of Albania, Bulgaria and Greece: implications for cross-border co-operation and development. European Urban and Regional Studies 4(3):193-208 Petrakos G (2000) The spatial impact of East-West integration. In: Petrakos G, Maier G, Gorzelak G (eds) Integration and transition in Europe: the economic geography of interaction. Routledge, London, pp 38-68 Petrakos G (2001) Patterns of regional inequality in transition economies. European Planning Studies 9(3):359-383 Petrakos G, Economou D (2002) The spatial aspects of development in Southeastern Europe. Spatium (8):1-13 Petrakos G, Totev S (2000) Economic structure and change in the Balkan region: implications for integration, transition and economic co-operation. International Journal of Urban and Regional Research 24(1):95-113 Petrakos G, Rodríguez-Pose A, Rovolis A (2003) Growth, integration and regional inequalities in Europe. Discussion Paper Series, Department of Planning and Regional Development, University of Thessaly Raagmaa G (1996) Shifts in regional development in Estonia during the transition. European Planning Studies 4(6):683-703 Ramboll (1996) Regional disparities in Romania. PHARE Program, Regional Policy Report, Bucharest: Ramboll- Consultants Group

Part III: Policy Issues

14

Has the Financial Economy Increased Regional Disparities in Switzerland over the Last Three Decades?

José Corpataux and Olivier Crevoisier Institute for Regional and Economic Research (IRER), University of Neuchâtel, Switzerland

14.1 Introduction Two processes can be identified in Switzerland over the last three decades. The first relates to the transformation of the financial sector which resulted in activity concentrated in one or two “metropoles” or “global cities”. The second process relates to the decline of these traditionally industrial regions, and more recently the tourist regions. This chapter examines the linkage between these two processes. The first part of the chapter deals with theories. Early theories of regional convergence/divergence (Myrdal 1957) emphasized the role of capital movements in the process of correcting disparities. But these approaches were very general and boiled down to the issue of the imperfection of financial markets. Few studies sought to examine those imperfections and their repercussions on an economy’s productive and spatial organisation. An exception is the work of Dow (1999). She describes a staged pattern of development in which a financial system initially deals in pure intermediation at the local level between lenders and borrowers. After that comes credit creation and then financial disintermediation. One of the consequences of this disintermediation is the disappearance of regional funding circuits and the emergence of specialized financial centres. The chapter continues with an application of this representation to Switzerland’s economic and spatial evolution over the period 1975 and 2000. A clear distinction is made between two phases in the relationship between the financial sector and the rest of the economy. The first phase, between 1975 and 1995, was characterised by the international development of the Swiss financial centre(s). This was basically concentrated into a few cities and led to an appreciation of the Swiss franc. In this context, the country’s large enterprises took advantage of a strong currency to internationalise. The appreciation of the Swiss franc nevertheless had some negative effects on exporting regions. The second phase, which started in 1990, resulted from a major recession in the domestic economy and a process of decartelisation. These two factors coalesced and created problems for the domestic banking sector. This resulted in a drying-up of regional business funding circuits and a strengthening of the control of the

252

José Corpataux and Olivier Crevoisier

domestic economy by the financial markets. The conclusions of the chapter assess the role of the financial sector in the de-industrialisation of Switzerland and in the increase in regional disparities.

14.2 The Financial Sector and Increasing Regional Disparities 14.2.1 Finance and Regional Development: Theoretical Links Neoclassical convergence theory postulates efficient capital markets, free operation of financial markets and the absence of any barriers to capital mobility. The central idea is that capital movements respond adequately to inter-regional variation in rates of return on capital. This mobility leads to the reduction or elimination of disparities. If demand for goods produced by a region falls, then the price of the region’s assets also falls. These assets then reattract investors, thereby stimulating the flow of capital into the region. In such a world, financial institutions are distributed efficiently and capital movements are the vital factor in resolving regional disparities. The financial sector, by enabling adequate capital mobility, is the key factor in resolving disparities. The corollary is that “imperfections” in the financial system reinforce regional disparities. Cumulative causality theory (Myrdal 1957) takes an opposing view. It suggests that the tendency for factors to converge towards economic centres creates imbalances. Where capital is concerned, the concentration of the banking system results in savings deposits filtering towards the central regions, depriving noncentral regions of the capital they need to develop. Little research however has been undertaken linking financial circuits and spatial economic development. From a post-Keynesian perspective, Sheila Dow (1990, 1992, 1993) is undoubtedly the author who has pushed theoretical thinking furthest relating the spatial structure of the financial sector with the increase in regional disparities. According to her, one result of liquidity preference1 is that, in non-central regions, economic agents tend to progressively abandon local funding circuits and their poor liquidity, preferring financial products offered by the centres (Dow 1992). Credit rationing in non-central regions adds to the polarisation of financial activities and investments in a country’s main financial centres, as economic agents progressively abandon local funding circuits. Liberalisation of capital movements and economic integration inevitably lead to the centralisation of the financial system and to regional divergence. A similar idea is expressed by Dow (1999) relating to the evolution of the financial sector. She distinguishes six stages of development and their correspond1

From a Keynesian perspective the individual displays a preference for liquidity. In the choice between two assets (financial or tangible), the individual will always prefer the one which is the more liquid (Boncoeur and Thouément 1993).

The Financial Economy and Regional Disparities in Switzerland 253

ing spatial structures (Table 14.1). At the early stages, banks play a single role of pure financial intermediation taking in and lending out local savings deposits. By stage 5 this intermediation is carried out by competitors. At stage 6, securitisation emerges with banks facilitating borrowing by issuing securities rather than by direct lending. Up to and including Stage 5, this system is regional and national. At Stage 5, competition emerges between banks and other non-banking financial institutions, and the system becomes more and more autonomous. Various financial agents seek to capture market shares through a policy of actively canvassing sales. More money is created and the economy is generously irrigated. In Stage 6, deregulation opens up international competition. Financial authorities seek new ways to contain monetary creation and this allows for the development of national financial centres. Spatial concentration takes place on the international scale too. The intermediation function becomes secondary and capital is procured via national and international financial markets. Banks no longer lend, but instead, supply services that enable their customers to get funding on the markets, while at the same time they remain powerful players in those markets. Table 14.1. Regional implications of stages in banking development

Stage 1:

Stage 2:

Stage 3: Stage 4:

Banks and regional development

Credit and regional development

Serving local communities Wealth-based, providing foundation for future financial centres Market dependent on extent of confidence held in banker

Intermediation only

Banking system develops at national level Central bank oversees national system, but limited power to constrain credit

Stage 5:

Banks compete at national level with non-bank financial institutions

Stage 6:

Deregulation opens up international competition, eventually causing concentration in financial centres

Credit creation focused on local community because total credit constrained by redeposit ratio Redeposit constraint relaxed somewhat, so can lend wider afield Banks freer to respond to credit demand as reserves constraint not binding and they can determine volume and distribution of credit within national economy Credit creation determined by struggle over market share and opportunities in speculative markets. Total credit uncontrolled Shift to liquidity with emphasis put on services rather than credit; credit decisions concentrated in financial centres; total credit determined by availability of capital, i.e. by central capital markets

Source: adapted from Dow (1999).

Thus Dow’s work suggests that the credit creation force in each region is the key to understanding today’s spatial imbalances. Innovation in the financial sector (deregulation, concentration, securitisation, growth of pension funds, etc.) and the

254

José Corpataux and Olivier Crevoisier

capacity of financial centres to transform companies’ tangible assets into liquid assets negotiable on the stock market both brings about an estrangement from sites of investment and confers the central location with a capacity for spatial trade-off (Leyshon and Thrift 1997) between regional, national and international scales. 14.2.2 The Drying Up of Traditional Funding Circuits: Consequences for Regions and SMEs The disappearance of the traditional credit system seems to challenge the unity of national economies. With financial institutions no longer managing to play a regulatory role dual development becomes reinforced. For Michalet (1998), the financial globalisation process thus brings with it a dualism within national space just as it does between nations. How can this process be accounted for? Traditionally, an economy based on bank credit implies a close relationship between bankers and firms. Loans to small enterprises (SMEs) play in part the role of equity capital. Proximity therefore plays an important role and can partly offset the repercussions of information asymmetry between lender and borrower (Porteous 1995). The reduction in the total amount of loans and their replacement by securities has dramatically reduced the need for proximity, giving way to standardised relationships in which distance is less of a factor. However, these relationships mean that banks are no longer involved to the same extent in the economy. This transformation also obliges the large financial institutions to make their investments more liquid. For example, financial reforms undertaken in France triggered the formation of a dual funding system (Allégret 2001). The financial changes of the 1980s did not facilitate SME access to external sources of finance, but had the effect of increasing investors requirement for liquidity. Thus, the growing sophistication of financial systems has, paradoxically, not led to easier funding for SMEs. In this context, large firms have a considerably wider range of funding instruments at their disposal - principally financial markets, securitised debt negotiable on the money market and bank loans. SMEs on the other hand have remained dependent on bank funding and cannot take advantage of the modernised financial systems. Farnetti (1995, 1996) highlights the parallel between, on the one hand, the good performance of Britain’s largest companies, particularly their internationalisation and their stock market performance, and the decline in the rest of the country’s economy. He dissects the sources of finance for the main mergers/acquisitions in recent years and defends the thesis that the competitive advantage of British firms stems primarily from having privileged access to financial markets that are particularly well-developed and liquid, via heavy recourse to the Euromarkets as well as the discretionary use of pension fund assets. This privileged access has favoured the move towards internationalisation of British multinationals via mergers/acquisitions, to the detriment of areas such as investment in research and development. To sum up, although British multinationals remain connected to the City of London and to the economy of the south-east, because they have their head

The Financial Economy and Regional Disparities in Switzerland 255

offices there, they are no longer truly linked to the rest of the British economy because they hardly invest there. The breakdown of regional financial channels and the concentration of financial activities in a country’s main financial centre results in the growing power of institutional investors who make a considerable contribution to increasing regional imbalances. Martin and Minns (1995) succeed in showing that the increasing power of British pension funds at the end of the 1980s had the effect of strongly magnifying regional imbalances. Savings deposits were collected in a homogenous manner across the whole country, but they were funnelled off into financial institutions that were mainly in the south-east of the country. Next, these funds are invested mainly on the London stock market and only listed companies - basically the large enterprises - benefit from them. In practice, almost nothing is reinvested in the other regions of the country. Thus financial markets become privileged and exclusive sources of funding for economic activities. As a result, the transition to a financial economy has favoured the economic agents who comprise stock markets (financial institutions, major banks, multinational companies, financial services, etc.) for over 20 years. By giving some economic agents privileged access to capital and to money creation, the development of financial markets reinforces regional disparities and a dual system emerges. This is characterized by the organizations associated with the global city, such as large companies, financial institutions, major banks, financial services on the one hand and SMEs, regional and local banks, and tourist regions on the other.

14.3 The Increase in Regional Disparities in Switzerland Traditionally, the Swiss economy has been divided into a successful export sector on the one hand and a protected sector orientated towards the domestic markets and much less successful on the other hand (Lambelet 1993). Today, this division is no longer valid. Instead, the dualised economy is comprised of a financial sector, with large industrialised firms and internationalised services and the rest of the Swiss economy, basically comprised of exporting industrial SMEs and a declining tourist sector. Spatially, the country is divided between financial metropoles and the rest. This raises the question of the linkage between the financial economy and the rest of the economy and between the financial metropoles and the other regions. Empirically, the objective of this chapter is to reveal the new economic and spatial structures in Switzerland. Two phases can be identified in the relationship between the financial sector and the rest of the economy. The first phase, between 1975 and 1995, is characterised by the international development of the Swiss financial centre. This growth in financial activities was accompanied by an almost continuous appreciation of the Swiss franc on foreign exchange markets. In this context, large enterprises took advantage of a strong currency to establish their strategy of

256

José Corpataux and Olivier Crevoisier

expansion and relocation while export sectors, made up basically of SMEs, experienced problems. With the recession at the start of the 1990s and the decartelisation of the domestic economy, a second phase began. It led to an opening up of traditionally protected sectors, as well as a drying-up of regional business funding circuits. A number of regional production systems (RPSs)2 (Crevoisier, Corpataux and Thierstein 2001) can be identified in Switzerland and these evolved in particularly contrasting ways. In terms of specialisation, the variations between them are extremely marked (Figure 14.1, Table 14.2). While the finance-structured metropolitan systems of Genève and Zürich experienced increasing prosperity between 1975 and 1995, the other Swiss regional production systems evolved in a contrasting manner. The industrial regions experienced a drastic reduction in job numbers from the 1970s. The tourist regions resisted exchange rate pressures until the funding circuits were ruptured at the start of the 1990s. The Ticino area and above all the Basel region were only partially successful in their metropolisation. Lastly, the regions that survived primarily on controlling domestic flows underwent major growth during that period until they, too, were hit by the crisis in public finance and the decartelisation process. The metropolitan industrial system In Basel

The metropolitan system in Zürich

The industrial system in the Swiss Jura

The industrial system in Eastern Switzerland

The Bernese administrative system

The urban and tourist systems in the Lausanne region

The tourist system in Graubünden

The metropolitan system in Genève The tourist, tertiary and industrial systems in Ticino

The tourist and industrial systems in Valais The tourist system in the Bernese Oberland

Evolution of employment in the various RPSs, 1975-1995 X < 15% 15% < X < 30% X > 30%

Fig. 14.1. Evolution of employment in the various regional production systems, 1975-1995 2

A regional production system (RPS) is defined as a geographical zone of productive specialisation(s) including a number of regional agents (companies - small or large institutions, authorities). They interact with each other on the basis of technical complementarity, competition and/or cooperation. An RPS receives and generates specific resources (in particular know-how) and this forms the basis of its competitiveness. Finally, a RPS possesses a certain degree of autonomy in its evolution.

The Financial Economy and Regional Disparities in Switzerland 257 Table 14.2. Evolution of employment in the various RPSs*, 1975-1995 RPSs identified in Switzerland

Number of jobs in 1975

Number of jobs in 1995

Variation in jobs

Metropolitan system in Genève

53,584

81,781

53%

Urban system in Lausanne conurbation Tourist system in the Côte, Riviera and Vaudois Alps Systems in Lausanne region

40,444 6,480

64,261 7,978

59% 23%

46,924

72,239

54%

Industrial system in Swiss Jura

100,453

70,206

-30%

Tourist system in Valais Industrial system in Valais Systems in Valais

35,436 11,922 47,358

44,650 11,834 56,484

26% -1% 19%

Bernese administrative system

54,029

85,822

59%

Tourist system in Bernese Oberland

16,125

18,878

17%

Metropolitan industrial system in Basel Tourist system in Ticino Tertiary system in Ticino Industrial system in Ticino Systems in Ticino

67,636 32,108 6,466 8,449 4,023

76,539 34,229 9,073 4,762 48,064

13% 7% 40% -44% 2%

Metropolitan system in Zürich

189,561

263,948

39%

Industrial system in Eastern Switzerland

106,889

88,958

-17%

Tourist system in Graubünden 36,105 41,849 16% * Total jobs does not correspond to total jobs in the region. It corresponds to total jobs in those specialized sectors that comprise the RPS. Source: IRER.

14.3.1 First Phase (1975-1995): From Export-Based Economy to Internationalisation of Large Enterprises and Decline of SME Systems During this phase, Switzerland evolved towards a “head office economy” specialising in the management and organisation of multinational production and the control of international financial flows. This was centred on the financial metropoles of Zürich and Genève, and grounded in the internationalisation of large enterprises. The result was negative repercussions via the appreciation of the Swiss franc, for the systems made up of SMEs (Corpataux, Crevoisier and Thierstein, 2002). The spatial scale is most relevant to this study is the RPS. These are based on MS regions.3 In Switzerland, statistical data is generally collected and calculated at canton level. There is therefore a discrepancy between the RPS and the statistical division of the country into politico-administrative units. Effort has been made to verify results obtained for RPSs against available statistics prepared at canton level. Regional per capita income at the canton level is a primary indicator of canton performance (Table 14.3). 3

MS regions are spatial entities based on commuters movements. There are 106 MS regions in Switzerland. By grouping some of them together on the basis of economic specialisations, it is possible to establish coherent RPSs.

258

José Corpataux and Olivier Crevoisier

Development of International Financial Centre and the Financial Metropoles Since the 1970s, the Swiss banking economy expanded its international dimension and specialization in international financial activities and services. Asset management, particularly for private customers, demonstrates the international role of the Swiss financial centre as an exporter of services (Blattner et al. 1996). According to estimates by Chase Manhattan Bank,4 Switzerland manages 35 percent of the world’s transborder asset management involving private customers, while Great Britain in comparison has only 21 percent and the USA a mere 12 percent. The place of the banks in the Swiss economy grew considerably until the start of the 1990s. In 1995 the banking sector in Switzerland accounted for a very high share of the total gross value added created (9 percent), with a corresponding figure of 7.2 percent in Great Britain. 4.6 percent in the US and 4.1 percent in Germany (BAK 1998; ASB 1999). Table 14.3. Annual increase at the canton level of per capita income (1975-1995), in Swiss francs Cantons

Zug Basel-Stadt Zürich Glaris Genève Vaud Schaffhausen Basel-Landschaft Nidwalden Aargau Schwyz St. Gallen Uri

Annual average increase of per capita income, 1975-1995 2431 1797 1708 1525 1427 1352 1340 1338 1287 1271 1258 1217 1214

Cantons

Ticino Luzern Solothurn Fribourg Graubünden Thurgau Bern Neuchâtel Obwalden Appenzell Ausserrhoden Appenzell Innerrhoden Jura Valais

Annual average increase of per capital income, 1975-1995 1193 1186 1173 1151 1130 1117 1111 1088 1027 979 970 919 830

Note: 1 Euro is more or less 1.55 SF Source: Federal Statistical Office (OFS).

The traditional advantages of Swiss economy and financial system such as banking secrecy, political stability and neutrality of the country and a qualified workforce coincided with the monetary policy conducted by the Swiss National Bank. In the first half of the 1970s the central bank decided to let the franc float on external markets and to pursue a monetary policy aimed at stable growth in the 4

Cited by the ASB (1996).

The Financial Economy and Regional Disparities in Switzerland 259

domestic money supply. The outcome was an almost continuous appreciation of the Swiss Franc on foreign exchange markets between 1973 and 1995. Figure 14.2 illustrates this appreciation compared to a basket of currencies that includes Switzerland’s 15 main trading partners. Index of real exchange rate of Swiss franc November 1977 = 100

140

100

USA Germany Total (15 countries)

60 1/1/73 1/1/76 1/1/79 1/1/82 1/1/85 1/1/88 1/1/91 1/1/94

Year

Fig. 14.2. Evolution of exchange rate of Swiss franc, 1973-1995 Source: Swiss National Bank (BNS).

This increase in financial activities was not concentrated on a limited number of metropoles. Of the total value added created by the Swiss banking sector in 1995, almost 60 percent came from the regions of Zürich and Genève. Zürich was clearly dominant, with a figure of 37.4 percent. Thus, the importance of the Zürich financial centre within Switzerland is relatively comparable to that of London (41 percent) within Great Britain (BAK 1998; ASB 1999). The metropolitan systems of Zürich and Genève, whose activities are basically centred around finance, have as a result experienced very strong growth. Zürich combines the functions of being a refinancing centre for Swiss large enterprises and a centre for institutional and private asset management. Genève specialises in private asset management and international trade finance (Roth and Béguelin 1992). This financial centre is first and foremost a service exporter and has very few links with the rest of the Swiss economy. However, it displays characteristics that are strongly complementary with the main sectors of the Genève economy especially in the areas of international trade/negotiation and international organisations. It should also be noted that the development of financial and banking activities, in both Genève and Zürich, has resulted in a very pronounced decrease in industrial activities. These observations are confirmed by an analysis of the evolution of per capita income (1975 to 1995) in the cantons concerned (Zug and Zürich for the Züricharea RPS and Genève for the Genève-area RPS). These three cantons rank in the top five places (Table 14.3). Zug turns in by far the best performance, with an annual average increase of 2431 Swiss francs. The canton of Zürich is third

260

José Corpataux and Olivier Crevoisier

(+1708) and the canton of Genève fifth (+1427). The gap between the results obtained by Zürich and Zug can be explained, in part, by the latter’s very competitive fiscal policy.

Internationalisation of Large Enterprises Over the period of appreciation of the Swiss franc, the large enterprises opted for a relocation of production and for multinationalisation via takeovers of foreign companies. They also invested massively abroad. Statistics published by the Swiss National Bank (BNS 1996) and Deloitte Touche (1997), concerning Switzerland’s outward foreign investment and inward foreign investment confirm this unilateral opening up to the exterior. Out-bound investment totalled over 42 billion Swiss francs over the period 1993-1995, while inward investment latter struggled to reach 7 billion Swiss francs. Of course, the strength of the Swiss franc, which is at once an incentive to invest abroad and a disincentive to produce in Switzerland, does not explain everything. But the strength of the franc undoubtedly favoured the external growth of large Swiss firms. Foreign companies who wanted to set up in Switzerland, were forced to “pay the price”. In the Swiss chemicals sector, for example, total outward direct foreign investment was in excess of 11 billion Swiss francs over the period 1993-1995, while inward direct foreign investment for the sector was 731 million Swiss francs!

Reaction of Industrial SME Systems Over the same period, regional SME systems specialising in industries such as watchmaking, machine tools, textiles were penalised on the export front and made increasingly vulnerable by the appreciation of the Swiss franc. Without the financial capacity nor the organisation to expand successfully abroad through external growth, SMEs reacted by specializing in high value added products and developing competitive advantage other than price. This behaviour characterizes the watch-making industry (Crevoisier 1993) for example. The huge machines sector also restructured itself and specialised in top-of-the-range machines, highly specialised and often bespoke. These “successes” however resulted in considerable job losses across the Swiss territory. Between 1970 and 1982, watchmaking lost two-thirds of its jobs. Other activities, such as textiles, practically disappeared. Thus, this process of specialisation in high-value-added market segments had serious limitations. Those RPSs that specialised in industrial activities saw their job numbers evolve more slowly than the national average. The industry of the Jura Arc for example, was strongly hit by the appreciation of the Swiss franc. All its industrial areas were orientated towards niche markets and high-value-added markets, involving limited production runs. The large firms restructured themselves (rationalised production, relocation, etc.) and many SMEs disappeared. Jobs in jewellery-watchmaking, for example, went from over 48,000 in 1975 to around 23,000 in 1995. The remaining companies nevertheless profited from an environment that was conducive to the development of a microtechnology industry, using

The Financial Economy and Regional Disparities in Switzerland 261

a combination of the region’s technologies and skills. The cantons of Jura, Neuchâtel and Solothurn (which make up, entirely or in part, the industrial system of the Jura Arc) registered weak growth in their per capita income, compared to the rest of the country. The Jura ranked 25th and next-to-last with an annual average increase of 919 Swiss francs. Neuchâtel ranked 21st (+1088) and Solothurn 16th (+1173) (Table 14.3). In eastern Switzerland, traditional specialisation in textiles and clothing was very exposed and took refuge in producing for niche markets, experiencing considerable job losses. The machine sector, in particular textiles machines, largely relocated its production abroad while maintaining jobs at home. The cantons belonging to the eastern Switzerland industrial system had diverse fortunes. Schaffhausen (7th with an annual average increase of 1340 francs) and St. Gallen (12th with an annual average increase of 1217 francs) emerged relatively well. But the other cantons were less fortunate: Thurgau ranked 19th (+1117) and the two half-cantons of Appenzell are 23rd and 24th (+979 and +970) (Table 14.3).

Growth and Decline of the Tourist RPSs The tourist RPSs (mainly the regions of Valais, the Bernese Oberland and the Graubünden), did not respond like industries to the appreciation of the Swiss franc because it was not possible to increase productivity to the same degree as in industry. By its very nature, tourist activity has little scope for automation and innovation. Competitiveness was mainly attained by squeezing costs and this was done in two ways. First, by huge recourse to poorly-qualified, poorly-paid foreign labour, explaining the relatively strong performance of these regions on employment. Second the tourist industry reduced costs by moving from hotelbased tourism to non-hotel based accommodation and services. The result was the development of the construction sector and the property business. However, this evolution came to an abrupt halt at the start of the 1990s when the Swiss franc appreciated once again. At this time, the structural limits of these systems were reached. It was no longer possible to avoid crisis by squeezing costs. Investment was needed. But it was also at this point that bank loans, the traditional source of funding for SMEs in this sector, were drastically rationed. This jolt was one too many and the tourist RPSs went into crisis, with particularly serious repercussions in the construction sector. The per capita income statistics for the Valais canton increased, on average, by a mere 830 Swiss francs per annum. This was the lowest growth in the country. The canton of the Graubünden did a little better and ranked 18th in per capita income change (Table 14.3).

Partially-Successful Metropolisation of Composite Systems The Basel metropolitan industrial system experienced lower growth than the country’s other large urban regions. Employment in chemicals, which structured the whole regional economy, fell by 12 percent between 1975 and 1995. The rising costs of research and manufacturing in Switzerland coincided with the attraction of relatively low investment costs abroad: the strength of the Swiss franc

262

José Corpataux and Olivier Crevoisier

was used as a lever and helped the internationalisation of Basel’s large industrial groups. From the end of the 1970s, acquisitions and direct investments multiplied to a point where jobs abroad today represent two-thirds of the jobs in these companies. The concentration of decision-making power on the international scale did not however lead on to a real metropolisation of the Basel region. The tertiary activities, and in particular the financial centre, did not follow. Per capita income of the half-canton of Basel-Stadt is ranked 2nd best, with an annual increase of 1797 Swiss francs (Table 14.3). This result is due to the presence of the big chemical companies’ head offices. The Ticino financial centre, with its specialisation in private asset management, grew strongly between 1975 and 1995. Nevertheless, its foothold at local level remained very embryonic and some commentators regard it as a large outpost of Zürich, owing its existence purely to its geographical, linguistic and cultural proximity with Italy (Chopard 1991). Other activities experienced problems. Shippers, totally dependent on the evolution of the Swiss franc, underwent “a process of decline” (Ratti and Baggi 1993). The tourism, clothing, textiles and leather sectors all collapsed. In other sectors, however, Ticino industry developed strongly such as industrial machines, electrical and electronic manufacturing and chemicals. Per capita income increase 1979-1995 ranked this canton 14th in Switzerland (Table 14.3).

Increase in Domestic Flows and Growth of Central Places The RPSs specialising in a domestic activities have little sensitivity to currency fluctuations. They are basically regional markets or domestic centres in activities such as services (health, government, higher education, the media, etc.). The RPS of the greater Lausanne area for example, profited from its central position, at the heart of French-speaking Switzerland, through the development of public services (health, education, etc.) and in the provision of private services (French-speaking head offices of banks and insurance companies, etc.). The canton of Vaud, which includes this RPS, ranked 6th in income growth in the country (+1352). Bern also experienced considerable growth in its activities. The funds that flowed in came, of course, from the whole of Switzerland, the public sector (government, health, etc.), the Federal State-owned companies (post office, national railway, etc.) as well as from the service providers that these entail. These centralised markets profited from both the increased productivity of the industrial systems and from the increase in financial systems by attracting the flow of funds generated by exports. However, these systems experienced more serious problems from 1992 onwards as a result of the crisis in public finance and of the revitalisation/liberalisation programme.

The Financial Economy and Regional Disparities in Switzerland 263

14.3.2 Second Phase: Decartelisation and Drying Up of Regional Funding Circuits At the start of the 1990s, high growth came to an abrupt halt. The domestic market was particularly affected by a series of institutional reforms such as decartelisation and the banking sector experienced a process of dualisation with large internationalised institutions pitted against regional and local banks. The outcome of this was a drying-up of regional funding circuits and, more globally, a pronounced disjunction between the large enterprise/sector and the domestic economy and SMEs.

Banking Sector Dualisation Two internal factors strongly influenced the restructuring of the banking sector. The first was the distribution of banking conventions in the early 1990s. This encouraged lively competition but penalised small establishments with regional markets. At the same time, the turn-around in the economic situation and the ensuing period of high interest rates (1991-1993) depressed the mortgage loan market, one of the main markets for regional and local banks. In this new context, the major banks simply accelerated their internationalisation, the private banks confirmed the trend, the major cantonal banks sought to follow suit and foreign banks continued to set up on Swiss soil. The regional banks, the local savings banks and the small and medium-sized cantonal banks found themselves in a more difficult position. The process penalised the banks with a regional orientation that supplied the regional funding circuits, and the SMEs. In 1985 the number of regional banks and savings banks stood at 216. This number fell to 204 in 1990, and then plummeted to 127 in 1995. By way of comparison, foreign banks, orientated mainly towards asset management, went from 120 in 1985 to 142 in 1990, reaching a total of 155 in 1995.5

Drying Up of Regional Funding Circuits Traditionally, regional business funding circuits were based on a number of mechanisms. First, ever since WWII the Swiss economy had become accustomed to having heavy recourse to bank loans. Until the start of the 1990s, the latter were plentiful and cheap. Where equity capital was concerned, the local stock exchanges (which existed in most Swiss cities) were all shut down and their operations concentrated in the Swiss Stock Exchange in Zürich. Unsuccessful activity in the property sector and a reversal in the economic situation led to a restrictive lending policy on the part of the banks. The total amount of bank loans plummeted: between April 1993 and March 2000, loans taken up by industry went from a little under 57 billion Swiss francs to around 32.5 billion Swiss francs. In this context, the SMEs were left to their own devices and it was at this time that 5

BNS, annual publication, various years.

264

José Corpataux and Olivier Crevoisier

diverse experiments were tried with regional risk-capital funds, or involving proximity capital. However, these were always very limited and never did provide SMEs with any satisfactory answer as to how to wean themselves off bank loans. Today, a limited number of companies have succeeded in entering the stock exchange or self-funding through effecting an employee takeover/buy-out operation. In the final analysis, regional funding circuits have practically disappeared. Consequently SMEs can expand in one of two ways. One option is through being traded on the stock exchange, with all the technical problems that poses for a small organisation and all the dangers that it poses to firm independence. Alternatively, a firm can sell out to one of the big international groups (Crevoisier, 1997). For example, the decade of the 1990s saw all the brand names in the watchmaking sector sell-out to the big groups.

14.4 Conclusions What lessons can be drawn from the Swiss case with respect to regional disparities in other small countries? What are the regional consequences of a strong specialisation in international financial activities rather than industrial activity? As has been shown, the Swiss economy was characterised at the outset by the development of financial metropoles, the internationalisation of large firms and the decline of SME systems specialised in industry or tourism. At a later stage, a process of decartelisation reinforced these spatial disparities by (a) targeting those regions orientated towards controlling domestic flows and (b) dismantling regional funding circuits. Therefore the dominant financial sector in Switzerland has tended to reinforce regional economic disparities. To what extent has the appreciation of the Swiss franc reflected the strong performance of the Swiss economy and the need for adaptation in its less competitive branches, such as industry and tourism? While the competitiveness and dynamism of the financial sector led to the strong Swiss franc, the strength of the Swiss franc was conversely the cause of the competitiveness of the financial market. Of course, there are other reasons for the development of the Swiss financial metropoles. Some centres such as Zürich, Genève and to a lesser extent Lugano have been able to combine Switzerland’s traditional advantages with their own specific resources. However the Basel region has not succeed in promoting its financial market, despite the presence of important national and international organisations (Association of Swiss Bankers, Bank for International Settlements, etc.). The industrial and tourist RPSs have had to adapt in order to survive an unfavourable macro-economic situation. The evolution of the Swiss franc eliminated all mass production in Switzerland. In areas of industry that were very competitive on the international scale, such as chemicals and watchmaking, there were massive job losses and major relocations. The large firms got out in good time, often to the detriment of their native regions. Thus only those regions that

The Financial Economy and Regional Disparities in Switzerland 265

had sizeable capacities in the financial sector at the outset, succeeded in transforming themselves. It can be said that Switzerland has evolved towards a metropolitan system of controlling European - even global - financial flows and that this is an enviable position. However it still remains to be seen whether this success is based on a sustainable competitiveness in these economic areas or if it is nothing but a reflection of specific, possibly transitory, conditions. What will happen when the European Union prevents Switzerland from acting as a refuge for capital or from making good use of its fiscal legislation? Moreover, how many people can survive in an economy that specialises in these activities? The financial services sector is undoubtedly a valid option for a micro-state such as Luxembourg. But is it equally valid for a small country of 7 million inhabitants?

Acknowledgements Frédéric Quiquerez is gratefully acknowledged for his helpful cantonal per capita income statistics.

References Allégret JP (2001) Introduction. In : Chanel-Reynaud G, Bloy E (eds) La banque et le risque PME. Presses universitaires de Lyon, Lyon ASB (1996) Le secteur bancaire suisse: évolution, structure et position internationale. ASB, Basel ASB (1999) Le secteur bancaire suisse: évolution, structure et position internationale. ASB, Basel BAK (1998) Internationaler Benchmark Report: Branchen und Regionen im internationalen Vergleich. BAK Konjonkturforschung Basel AG, Basel Blattner N, Gratzl B, Kaufmann T (1996) Das vermögensverwaltungsgeschäft der banken in der Schweiz. P. Haupt cop, Bern, Stuttgart Boncoeur J, Thouément H (1993) Histoire des idées économiques, de Walras aux contemporains. Volume 2. Nathan, Paris BNS (1996) L’évolution des investissements directs en 1995. Bulletin trimestriel, April. Chopard R (1991) Culture et langage bancaires : l’exemple de la place financière tessinoise. In: Hommage à un Européen. Fondation Jean Monnet pour l’Europe et Ecole des hautes études commerciales de l’Université de Lausanne, Lausanne Corpataux J, Crevoisier O, Thierstein A (2002) Exchange rate and regional divergences : the Swiss case. Regional Studies 36(6):611-626 Crevoisier O (1993) Industrie et région : les milieux innovateurs de l’Arc jurassien. EDES, Neuchâtel Crevoisier O (1997) Financing regional endogenous development: the role of proximity capital in the age of globalization. European Planning Studies 5(3):407-415 Crevoisier O, Corpataux J, Thierstein A (2001) Intégration monétaire et régions, Des gagnants et des perdants. L’Harmattan, Paris Deloitte Touche (1997) Foreign direct investment trends in Europe. IMR, February.

266

José Corpataux and Olivier Crevoisier

Dow SC (1990) Financial markets and regional economic development, the Canadian experience. Avebury, Aldershot Dow SC (1992) The regional financial sector : a Scottish case study. Regional Studies 26(7):619-631 Dow SC (1993) Money and the economic process. Edward Elgar, Cambridge Dow SC (1999) The stages of banking development and the spatial evolution of financial systems. In: Martin R (ed) Money and the space economy. John Wiley and Sons, New York Farnetti R (1995) Le royaume désuni. L’économie britannique et les multinationales. Syros, Paris Farnetti R (1996) Le marché financier et les multinationales britanniques : Royaume uni ou désuni? Economies et Sociétés R(9):153-191 Lambelet JC (1993) L’économie suisse. Economica, Paris Leyshon A, Thrift NJ (1997) Money/space: geographies of monetary transformation. Routledge, London Martin R, Minns R (1995) Undermining the financial basis of regions : the spatial structure and implications of the UK pension fund system. Regional Studies 29(2):125-144. Michalet CA (1998) Le capitalisme mondial. Quadrige, PUF, Paris Myrdal G (1957) Economic theory and underdeveloped regions. Duckworth, London OFS (various years) La prévoyance professionnelle en Suisse, Statistiques des caisses de pensions. OFS, Neuchâtel Porteous DJ (1995) The geography of finance. Avebury, Aldershot Ratti R, Baggi M (1993) Essai d’analyse dynamique et spatiale d’un réseau innovant dans le secteur des services : le cas des expéditionnaires à la frontière italo-suisse de Chiasso. In: Maillat D, Quévit M, Senn L (eds) Réseaux d’innovation et milieux innovateurs : un pari pour le développement régional. EDES, Neuchâtel Roth JP, Béguelin JP (1992) Rôle national et international de la place financière genevoise. In: Stepczynski M (ed) Genève et la Suisse, un mariage d’amour et de raison. Bourse de Genève, Genève

15

Regional Policy Lessons from Finland

Hannu Tervo School of Business and Economics, University of Jyväskylä, Finland

15.1 Introduction The Finnish economy and society has long been dominated by primary production. Post-war economic development was rapid and welfare gaps between the much more developed economies and Finland narrowed and were even partly reversed. Rapid economic expansion together with structural change has had the effect of centralizing both economic activity and population. The trend has been towards the southern and central regions where the metropolitan area of Helsinki and most of the other larger towns and urban centres are located. Together with vigorous technological progress in agriculture and forestry the rapid urbanization and industrialization of the 1960s and 1970s - the fastest in Europe - altered the status of the primary industries, creating enormous challenges for regional policymaking. Early regional policy based on building up infrastructure, decentralization of manufacturing industries and the creation of a welfare state was fairly successful, although it could not prevent huge out-migration from rural areas in the 1960s and early 1970s. While regional development was fairly even in the 1970s and 1980s, the 1990s were a time of great economic flux and drastic structural change. Finland was hit by a severe recession in 1991-93 and both production and employment fell sharply. Rapid economic recovery was based on export and knowledge-based industries. The investment in know-how turned out to be successful: several parallel analyses of international competitiveness show that Finland was one of the most dynamic and competitive economies in the world in the late 1990s and early 2000s.1 Although Finland as a whole flourished in the late 1990s, this was not true across all its regions. The positive aggregate development experienced by Finland was based on uneven regional growth, especially after the severe recession of the early 1990s. Regional competitiveness varied greatly, the most competitive regions being those containing an urban centre and especially those with a university (Huovari et al. 2001). Migration into the major centres of population accelerated in the late 1990s, and many of the smaller urban areas even saw a net loss in population. In recent 1

The World Economic Forum’s (WEF) Global Competitiveness Report ranked Finland 1st in 2001 and 5th in 2000, while Finland surged from 7th position in 1997 to 3rd in 2001 in the competitiveness study of the International Institute for Management Development (IMD).

268 Hannu Tervo

years Helsinki has in fact been the fastest growing metro region in Europe. The developments of the 1990s also seem to have brought regional convergence to a halt. Earlier, regional differences in Finnish per capita income levels tended to even out along with the process of regional concentration. New ideas and measures were desperately needed in regional policy, since it seemed that the old approaches were inadequate in the present era of knowledgeoriented growth. Finland has now taken the first steps in implementing a new regional policy. Technology- and knowledge- based support became more and more important. Special programmes based on know-how and the development of smaller urban centres have played an important role. Regional development has since become more even, but this is perhaps due more to renewed stagnation in the economy than a consequence of policy. This chapter examines the lessons to be learned from Finnish regional policy, the focus being the developments of the last ten years, although lessons from preceding eras of Finnish economic development are also highlighted. The chapter is organized as follows. The next section briefly examines lines of and gaps in regional development in Finland, by analysing regional development from the viewpoints of both location patterns and regional differences. Section 15.3 investigates regional policy and its effectiveness in different eras of development. In this connection, a difference between broad and narrow regional policies is made. Finally, Section 15.4 provides some discussion of the lessons to be learned from regional policy in Finland. The main messages are that regional policy should not work against market forces; regional policy cannot succeed if other public policies work forcefully against it and to be effective each era needs a regional policy of its own.

15.2 Regional Development in Finland Regional development can be analysed from at least two distinct standpoints, either from that of locational patterns or that of regional (welfare) differences. Levels and changes in regional production, employment or population reveal location patterns in the country. They indicate the extent to which regional development is centralizing or decentralizing. Another approach to regional analysis, however, is to identify regional differences in living conditions, such as in per capita income or in the unemployment rate. This reveals the magnitude of regional differences and whether these differences tend to converge or diverge. While the first approach takes the region as its starting point, the second one emphasizes the individual in a region. Broadly, two mega-trends can be seen in Finnish regional development: clear regional convergence in per capita income levels has taken place, but at the same time production and population has concentrated in the Helsinki region and other main centres of the country (Tervo 2000a). Convergence came to a halt in the 1990s, however, while the concentration process accelerated further. I will deal with both these processes in more detail in the following.

Regional Policy Lessons from Finland 269

15.2.1 Location Patterns A population shift from rural to urban areas has been underway in Finland for many decades now. This development is closely connected with the economic growth and structural change mentioned above. Over the period 1950-95, the population in the central and urban areas increased from around 2 million to 3.5 million (Loikkanen et al. 1998). At present about 2 million Finns, 40 percent of the total population, live in the six biggest centres. In addition to the regional and provincial centres, southern Finland in aggregate has benefited from the ongoing development, while other parts of Finland have lost throughout. For example, the population of Uusimaa has more than doubled since the 1950s, while in many eastern and northern provinces the population has declined. The late 1960s and early 1970s were characterized by the depopulation of rural Finland. This period has come to be known as the era of the “Great Migration”. People moved from rural areas to the cities and from the north to the south. A considerable part of this migration was directed at Sweden: Finland’s net loss to Sweden was 143 000 persons between 1961 and 1970. After this flow peaked at the beginning of the 1970s, regional agglomeration levelled out, and during the late 1970s and 1980s migration slowed down. The 1990s were a time of very rapid structural change. The recession years 1991-93 deeply affected every region in Finland. Employment declined rapidly throughout the country - almost every region lost one fifth of its jobs. While the recession treated regions “democratically”,2 the new growth based on the export and information technology industries took place highly unevenly across the country. Especially in the early phase of growth, new jobs were created only in a few big centres, which affected the direction of the rapidly increasing migration flows. For example, two thirds of the new jobs created during the period 19931996 were created in the three largest urban centres (Helsinki, Tampere, Turku) and only 9 percent of these jobs were found outside the ten most populated regions. As a result, in 2000 only six regions out of 82 had numbers of jobs at prerecession levels. In the wake of the recession, the significance of high technology industry for the Finnish economy has risen rapidly. High tech products accounted for 12.4 percent of exports in 1994, and 20.4 percent in 1999. The fastest growing sector is the manufacture of electrotechnical products, especially telecommunications equipment (Economic Council 2001). Nokia has had a strong effect on this. The rapid growth in the high technology industry has also affected regional development in Finland, since this growth has mainly taken place in big centres such as Helsinki, Oulu, Tampere, Turku (Salo) and Jyväskylä. Concentration substantially accelerated after the severe recession of the early 1990s. A new “Great Migration” period began, as people started increasingly to move from one region to another. For example, in 1998 five per cent of the population changed their municipality of residence. Net levels of interregional 2

This is exceptional, since the Helsinki region, for example, has normally had much better success in a relative sense than other regions during recessions.

270 Hannu Tervo

migration equalled those of the early 1970s. The greatest flows were from rural and smaller urban areas to a few big centres, whose growth is based on export and high technology industries. Consequently, the smaller urban areas now began to experience net losses, whereas they had experienced considerable net migration from their surrounding rural areas up to the mid 1970s. Using municipal boundaries as a rough proxy for real physical space, the migration figures show that over 90 percent of Finnish territory constitutes an out-migration area (Hanell et al. 2002). The danger for many rural regions is a depopulation process which cumulatively tends to retard development in these areas, while at the same time benefiting a few big centres. A scattered pattern of settlement, characteristic of sparsely populated countries, continues to strengthen in Finland. 15.2.2 Development in Regional Differences In spite of the strong trends towards regional concentration, welfare differences between regional economies are not very big in Finland and have tended to converge in the long run. Indeed, many Finnish studies show a convergence in per capita differences. Kangasharju (1998) approximated regional income levels by taxed income per capita and found clear convergence within the 88 Finnish subregions between 1934 and 1993. FinnishE convergence has been slightly above 2 percent annually and thus rather similar to that in many other countries. Kangasharju (1998) showed that in the 1930s per capita income (measured in terms of payment of per capita taxes levied locally on incomes) in the poorest subregion in Finland was only one tenth of that in the richest subregion, while in 1993 it was almost one half. Pekkala (2000) used regional GDP/capita data in her convergence analysis of 12 Finnish provinces and found that convergence was strong between 1960 and 1980, but weak thereafter. The per capita GDP figures for the year 2000 show that there are still relatively large differences between regions. The capital region Uusimaa is the most prosperous region with per capita GDP exceeding the EU average by 45 percent. On the other hand, regions in northern and eastern Finland such as Kainuu and Etelä-Savo lie some 32 percent and 29 percent, respectively, below the EU average. Accordingly, per capita GDP in the richest region is more than twice that in the poorest region in Finland. Loikkanen et al. (1998, 1999, this volume) examined income differences within the major regions by using the Statistics Finland Household Budget Survey, which makes it possible to examine different definitions of income and, more especially, to consider the role of the welfare state in regional income differences and their convergence. Their results showed that regional income differences decreased, particularly from 1966 to 1976, since which there has been either no further or only minor convergence, depending on the definition of income used. Convergence is clearest when taxation and income transfer systems as well as the effect of public services are taken into account, i.e. when disposable income or final income is considered. Accordingly, welfare state intervention through taxation, income transfers and the provision of social, health and other services

Regional Policy Lessons from Finland 271

has a strong levelling effect not only between individuals but also between regions. The long-run trend towards convergence seems to have ceased during the last few years. Loikkanen et al. (1999), for example, showed that when the recovery started in the 1990s, both average income and inequality of income increased in all regions. Kangasharju and Pekkala (2001) also show how the 1990s recession marked the end of a long period of regional economic convergence in Finland (see also Taipale 2002). Their results show that the evolution of both labour productivity and employment rate accounted for regional divergence in per capita GDP during 1990-1995. The productivity component, however, dominated the changes in regional disparities. Of the factors responsible, primary production contributed the most to the divergence in per capita GDP during the recession of the 1990s, and private services during the subsequent recovery (Kangasharju and Pekkala 2003). Diminishing convergence, of course, is a serious concern for regional development. There are also other indicators, such as the unemployment rate, which similarly show growth in regional disparities in the 1990s. Unemployment has traditionally been higher in the northern and eastern regions of Finland. Differences in unemployment have been fairly persistent, especially between greater regions (Pehkonen and Tervo 1998). The deep recession magnified the unemployment problem in Finland - the rate of unemployment jumped within a few years from about 3 percent to almost 18 percent. The increase in unemployment was much the same in each region, but during the recovery regional differences have grown (Tervo 1998; Huovari 1999). Both absolute and proportional differences in regional unemployment rates have increased, although there has been a successive decrease in all regions since the peak year of 1994. For example, in that year the unemployment rate was 13.9 percent in Uusimaa and 20.7 percent in Kainuu, while at the beginning of 2002 the rates were 5.8 percent and 18.9 percent, respectively. Accordingly, unemployment has fallen much more rapidly in Uusimaa than in Kainuu. The big migration flows of the late 1990s were not effective enough to stop this differentiating trend. It is noteworthy that the growth centres of the late 1990s are also suffering from structural unemployment. This follows from the sudden fall in employment in the early 1990s as well as from the rapid structural change in favour of the knowledge-oriented industries, which naturally effects the demand for labour. There is an increasing lack of fit between the demand and supply of labour in this area. As with unemployment rates, there are also big differences between regions in employment rates (the proportion of the working-age population who are employed). For example, in 2000 the employment rate varied from 58 percent to 74 percent between regions, the national rate being 66 percent. In their analysis of regional labour market dynamics, Pekkala and Kangasharju (2002) show that labour market adjustment to region-specific shocks initially occurs via labour force participation and, to some extent, unemployment. Only after a few years does interregional migration become an important adjustment mechanism. These results suggest that migration works as an equalizing mechanism only in a single-country context where the barriers to labour mobility

272 Hannu Tervo

are minimal. Interestingly, Pekkala and Kangasharju (2002) also found that the adjustment process was different in the case of total shock, in which case unemployment is the main adjustment mechanism and migration never attains a significant role. The effects of interregional migration on regional development are controversial. Even though interregional migration equalizes regional differences, it may still have negative effects on the losing regions in the long run due to the process of cumulative causation. The negative effects related to losses of the working population are strengthened by the selective nature of migration: the youngest and best educated typically leave their rural or smaller urban home regions and move to one of the bigger centres, in most cases to the Helsinki region. The process of the concentration of human capital is reinforced by interregional migration (Kauhanen and Tervo 2002; Ritsilä 2001; Haapanen 2003). Some research results show that interregional migration has not been a particularly efficient equilibrium adjustment mechanism in Finland, since the adjustment mainly takes place by means of regional unemployment, and not by personal unemployment (Tervo 2000b). Weak prospects “push” educated and young workers away from the poor regions, because the unemployed have decreased chances of local re-employment, and the employed are at an increased risk of losing their jobs. Further results indicate that it is the “quality” of migrants (e.g. age, education, human capital and unobserved ability), rather than the act of moving itself, that causes an improvement in employability (Pekkala and Tervo 2002; Tervo 2000c). These findings indicate that encouraging migration in general may not improve the geographical matching of jobs and the unemployed. Migration alone is not a very effective mechanism for alleviating individual unemployment, although the number of migrants increased substantially in the 1990s. 15.2.3 New Challenges to Regional Policy If we examine one of the individual “mega-trends” described above, we obtain a one-sided picture of regional development. A positive feature is the long-run convergence in regional differences in per capita income, although that now seems to have ceased. The other side of the coin is the strong process of concentration simultaneously taking place and which shows no signs of reversing. Income maintenance does not stop out-migration. From the point of view of regional economies and regional structure, location patterns are perhaps more important than regional differences. Centralizing trends and vicious circles, which support processes of cumulative causation devitalise and even depopulate regions with low population, such as those in Finland. A balanced regional structure is at risk even if the welfare of those inhabitants who are left behind remains unchanged or even improves when measured by differences in per capita income. The latest observations from the 1990s, which reveal an increase in regional per capita income differences, however, are alarming on that score.

Regional Policy Lessons from Finland 273

It is also to be expected that the living conditions of the people remaining in a region will worsen when depopulation processes have gone far enough. This will have a strong negative effect on the development of regional differences. The currently large differences in regional dependency ratios in Finland are one sign of this. A dependency ratio shows the number of persons outside the labour force and the number of unemployed persons for each employed person. This ratio is close to one in the advantaged regions, but more than two in the most disfavoured regions against a national average 1.3 in 2000. In certain municipalities in Lapland and Kainuu the ratio is even as high as 2.5. Even when one shifts from a local to a regional perspective the differences are still big. For example, at provincial level we find 1.9 non-working persons for every worker in Kainuu, while this ratio is 1:1 in Uusimaa. The major problems related to the regional concentration process in Finland will become apparent only in the long run. If regional policies do not succeed in countering this trend, the sparsely populated parts of Finland will have only a few growth regions, while most of the others will wither away. This presents enormous challenges for regional policy.

15.3 Regional Policy in Finland 15.3.1 Efficiency Versus Equity A major issue in the history of Finnish regional policy has been the question of the relationship between efficiency and equity. The traditional economic view is that regional policies conflict with macroeconomic efficiency so that a trade-off exists between the objectives of efficiency and equity; regional intervention has a cost in terms of real income and output arising from reduced aggregate efficiency (Tervo 1991). An economic case for regional policy, however, exists, and various economic arguments in favour of regional policy have been advanced. According to McCrone’s (1969) classic study, economic motives are related to factors such as the existence of externalities, the prevention of resource under-utilization and excessive inflation. Despite the potential conflict between the two overall objectives, the broad aims of Finnish regional policy have sought to involve both of these aspects. Priority has always depended on the current national economic condition (Tervo 1985; European Policies Research Centre 1996). It seems that when economic conditions are good, measures intended to reduce regional disparities in income, for example, are given more weight than in a time of general economic problems. On the other hand, the efficiency objectives are emphasized in periods of depression and high unemployment. As in many other countries, the goals of regional policy in Finland have been both ambitious and ambiguous (Tervo 1991). By and large, the principal aims of regional policy have been as follows: to reduce income disparities, to increase

274 Hannu Tervo

employment, to halt migration, to diversify regional economies and to improve infrastructure. Policy has for the most part aimed at greater regional equality. However, the efficiency argument has also been considered, even though this line of argument has been disputed. The early economic justification for regional policy was based on the resource utilization argument, whereas later the adverse consequences of regional imbalances in labour markets was also advanced (Tervo 1991). Since the 1990s, the priority in Finnish regional policy has increasingly been in macroeconomic development and efficiency. Regional development and policies have been more and more subordinate to the competitiveness of the country as a whole. 15.3.2 Developments in Finnish Regional Policy There are many ways of categorizing the different phases of Finnish regional policy. Vartiainen (1998) identified three main stages: policy with an industrial focus for development areas (up to the mid 1970s), regional policy planning (mid 1970s to late 1980s) - from 1975, at least in legislation, there was a marked shift away from a focus on development areas to a comprehensive nation-wide regional policy (European Policies Research Centre 1996) - and programme-based regional development (late 1980s onwards). Some form of regional policy has been practised, however, for as long as there has been a national policy. Indeed, it is obvious that all public policies have a spatial dimension. In its wider sense, regional policy constitutes the spatial dimension of a Nordic welfare state. The impact of policies from this standpoint on regional development is often labelled broad regional policy, while the more particular efforts directed at the development of weaker regions is termed narrow regional policy. Evidently, policies of the first kind have a much larger impact on regions than those more narrowly conceived. The amount of money spent on broad regional policies is many times higher than the money spent on regional policies in the narrow sense. While a broad regional policy brings with it welfare and redistribution in favour of the weaker regions, regional policy proper explicitly aims at influencing growth and development in the weaker regions. Regional policy is defined in the periodically renewed Regional Policy Act, which specifies objectives and measures, methods of regional policy coordination, tools of regional development policy and the criteria for selecting the areas eligible for regional aid. The first development zone laws took effect in 1966, and subsequently, these laws have been amended six times. Present legislation dates from 2003, and the laws preceding legislation from 1994. Both the aims and means of regional policy have undergone considerable modification and extension since the early days of the 1960s, in response to both changing regional problems, and lately, EU membership (Pekkala 2000). The early laws delineated development zones, in which firms received subsidized rates of interest on loans and tax reductions, but since then both the variety of policy means and the number of regions receiving support has expanded. Finland’s present regional policy also involves much more than enterprise promotion in backward areas. Regional

Regional Policy Lessons from Finland 275

problems have become more pronounced, and this has led to a shift in policy emphasis stressing the individual needs of particular areas. 15.3.3 Was Early Regional Policy Successful in Finland? The answer to this question is affirmative, especially if regional policy is considered over a longer period and in its broad sense. Four important policy orientations can be distinguished in Finnish regional policy. First, the infrastructure policy of the 1950s and 1960s laid the foundations for the development of the sparsely populated parts of Finland. This policy included for example, a major road-building programme. Although in its nature this was a counter-cyclical employment policy, it improved the operational preconditions for the manufacturing and service industries and eventually the competitiveness of more remote regions increased. Second, early regional policy, which aimed at decentralizing the manufacturing industries and dispersing them to the development areas, was successful. This policy had its beginning in a famous book “Do we have the patience to become prosperous?” published in 1952 by the then Prime Minister (later President) Urho Kekkonen in which he advocated systematic action for the development of northern Finland. The first phase in this policy took place in the 1950s and 1960s, when big state-owned companies were established in disadvantaged regions. The setting up of state-owned manufacturing firms and energy companies such as Rautaruukki and Kemijoki altered the entire economy of Finnish Lapland and turned the southern part of this vast province into an area characterized by big industry (Sisäasiainministeriö 1991). This alone indicates the strong role played by the state in the process of accumulation in post-war Finland (Virkkala 1994). The second phase came with the era of regional policy legislation, which started in 1966. This policy tended to encourage mainly small and medium-sized manufacturing firms to invest in the development zones by offering subsidized rates of interest on loans, tax reductions, regional transport subsidies and direct subsidies relating to investment, start-up and training etc. This policy turned out to be successful and a great number of new manufacturing plants were set up in the target areas. Evaluation studies have shown that the effects of this policy were significant (Tervo 1985). The results suggested that in the absence of this regional policy the level of employment would have been 12-17 percent lower, the level of output 8-15 percent lower and the number of plants 15-23 percent smaller in the manufacturing industries of the assisted areas in 1981. Many structural effects were also shaped by the goals of regional policy. For example, production became less one-sided. However, the positive results of the policy were at least partly based on a favourable phase for the development of manufacturing in the assisted areas. Industrial growth had moved to the less developed areas, which were able to provide good conditions for industries using standardised technologies requiring a large input of labour. In contrast, development outside the assisted areas concentrated on promoting the use of know-how. For these reasons, it might be said that deliberate regional policy began swimming aided by a favourable current.

276 Hannu Tervo

The third successful form of regional policy has been welfare policy and its spatial effects. The construction of comprehensive regional public service networks and expansion of the public sector promoted welfare and created jobs outside the central areas. These new jobs compensated for the loss of jobs in the agricultural and forestry industries. The most important reforms in the public services occurred in the 1960s and 1970s when many new jobs were created in northern and eastern Finland thereby levelling out Finnish regional development. In addition, many of these reforms, including the introduction of the comprehensive school and communal health centres, were regionally staggered, being timed to start in the north and move gradually south. This had a levelling effect on regional development in the late 1970s (Sisäasiainministeriö 1991). This “broad regional policy” continues to have large inter-regional effect, as we saw when looking at convergence in Finland. Welfare state intervention through taxation, income transfers and the provision of public services plays an important role in equalizing regional income differences. The fourth successful form of regional policy is the regionalisation of university education to cover practically the whole country. This action is also an example of a policy orientation which was not due to any official regional policy, but fashioned by regional interests (Sisäasiainministeriö 1991). In the early 1950s Finland had only two university towns, Helsinki and Turku. By the early 1980s this number had increased to 20 universities in 10 localities including areas such as Tampere, Jyväskylä, Oulu, Joensuu and Kuopio. This has turned out to be a very effective means of levelling out regional development. Those regions which succeeded in getting a university of their own have flourished. The long-run effects on economic development in these regions have been greater than perhaps were ever imagined at the time these decisions were taken. Many of these universities have become hosts to important technological and research and development initiatives which have increased training opportunities and jobs in the regions concerned. The best example of this is the University of Oulu, which has had an invaluable effect on the growth of this north-western region (Economic Council 2001). 15.3.4 Regional Policy Since the 1990s The environment in which regional policy has been pursued changed dramatically in the 1990s. This development was not only characterized by huge economic fluctuations, but also by deregulation and the opening-up of the economy (Economic Council 2001). A big and still unsettled question is whether regional policy as presently implemented can solve the distortion in regional development that has arisen in the present era of technology- and knowledge-oriented growth. Regional distortion is expressed in very uneven growth and widening income differentials between regions.

Regional Policy Lessons from Finland 277

EU Regional Policy Present regional policy can be divided into national and EU regional policy. The EU integration process has had a pervasive influence on Finnish society, including regional policy. Structural Fund programmes are important tools and they dominate regional policy in Finland. In addition to funding from the Structural Funds, EU membership also brought a profound change in the principles governing regional policy (Tervo 1996). EU regional policy is characterized by concentration of assistance, co-ordination, partnership, subsidiarity, programming and additionality. Although a controversial issue, it seems that the much-vaunted principle of additionality underpinning EU regional policy has not been realized in Finland. Like many other member states, Finland has seen the advent of the Structural Funds as an opportunity to cut back its own domestic regional policy commitments (Armstrong and Taylor 2000). The present period 2000-2006 in the EU regional policy programme includes two objectives. Objective 1 is to promote the development and structural adjustment of regions lagging behind in development, and Objective 2 is to support the economic and social conversion of areas facing structural difficulties. The largest EU funding per capita is allocated to Objective 1. The level of funding for Structural Fund programmes in Finland is quite high compared to that of the other EU Member States. In 1995-1999, the areas in Finland covered by EU Structural Fund programmes received altogether FIM 10.9 (1.8 ) billion of public development funding. Objective 6 areas in eastern and northern Finland accounted for half of this, the remainder being divided evenly among industrial areas in decline (Objective 2) and lagging rural areas (Objective 5b). The simplest way of assessing the effects of EU programmes is to compare economic development in programme regions with that in other areas. This gives quite a dismal picture of the effectiveness of the policy: for example, during the period 1995-1999 development in unemployment rate, net migration, per capita GDP and even in employment was worse in the programme regions than in other regions. This approach to evaluation meets with the problem of counterfactuality: we do not know what the outcome would have been without the policy. The outcome of EU programmes has been analysed in more detail in the evaluations required by EU law. These micro-level evaluations typically show more positive results for the EU regional policy and the projects undertaken within it than macro approaches. The real effects of policy are very difficult to disentangle here as well, due once again to the counterfactuality-problem. Potentially, the most important effects are those which start endogenous growth processes in the region: policy should act as a spark plug rather than fuel in regional development. Know-how, innovations and learning have become key issues in the new regional development logic. Ritsilä and Haukka (2003) analysed the role of structural funds in developing learning regions. The purpose was to find out how the European Social Fund interventions were integrated into the different processes involved in developing a learning region. The analysis indicates a good measure of integration achieved through developing and

278 Hannu Tervo

enhancing human capital, promoting the diffusion of know-how and transferring human capital into practise.

National Regional Policy Present Finnish regional policy laws conform to EU regional policy. National regional policy is closely related to the Structural Fund approach. Regions themselves have been given a greater role, one that accounts for regional strengths and weaknesses in policy implementation (Niittykangas and Tervo 1995; Pekkala 2000). Regional competitiveness is emphasized. In its national policy, Finland has also adopted a programme-based regional policy and a system of graduated regional support for business and industry according to a geographical breakdown of areas targeted for aid. Under the new law, each Regional Council draws up a fixed-term regional strategic programme. This has to include development targets based on the region’s potential and needs, the most important projects in terms of regional development, and other measures deemed essential to achieving the targets and financing the planned programme. As the Finnish economy worked its way out of the deep recession of the 1990s towards a competitive and knowledge-based economy, traditional regional policy instruments were displaced in great part by technology- and knowledge-based mechanisms. Suddenly a dividing line arose between the “old” regional policy with its industrial focus and the knowledge-based “new” regional policy. The former became underrated while the latter was seen as a solution to the regional problems of the information society. Actually, this “new” regional policy of the 1990s dates back to the 1980s when the Regional Policy Committee listed a number of measures required to strengthen regions. The fear was that the information society would polarize regional development. The measures included harnessing universities and training establishments to the economic development of their areas, expanding the regional activities of the Technical Research Centre of Finland (VTT) and of the Technology Development Centre (TEKES), and improving telecommunications (Sisäasiainministeriö 1991). At the beginning of 1988, the development of a specific regional technology policy was started when provinces began to prepare their own technology programmes. Compared with many other EU Member States, Finland spends rather little on regional and other industrial subsidies. National regional policy subsidies granted in Finland in 1996-1998 were the fifth lowest in the Union (Economic Council 2001). However, even if national regional policy has lost its financial importance, it continues to exert great importance as a channel through which attempts are made to launch new ideas and measures in regional policy. This policy does not tie up much public money but nonetheless steers activity and development in the regions, and, at best, may have a profound effect on regional development. The most important programmes are the Centre for Expertise Programme (started in 1994) and the Regional Centre Programme (started in 2000). Both of these include measures for strengthening provincial cities by helping them to develop specialities of national significance. The aim is to spread growth more evenly

Regional Policy Lessons from Finland 279

across the country through the smaller towns by utilizing the economies of concentration, specialization and know-how. The Centre of Expertise Programme supports regional specialization and cooperation between different Centres around the country. Finland is thus aiming to create a strong, effective network of Centres of Expertise to meet the challenges of the information society in the 21st century. The Centre of Expertise Programme for 1999-2006 is being carried out by 14 regional Centres of Expertise and two nationwide networks. The aim of the Centre of Expertise Programme is to promote collaboration between research and industry in order to utilize the top skills of both urban and rural areas. Technology centres are responsible for the practical implementation of the programme, and also provide a framework for actual projects. This programme is generally considered to have been successful, although the government’s contribution to its basic funding was low. The programme is committed to promoting a regional innovation policy. The projects implemented have promoted the creation of new structures stimulating investment in the development of growth sectors of real regional importance. In the period 19941998, the volume of projects launched with basic funding was already 13 times that set up by seed-stage funding. Overall some 3700 enterprises have been involved in projects. An evaluation of the first phase of the programme, has found that the programme has introduced three kinds of added value into the development of skills: funding, new contacts, and image benefits (Ahola and Kortelainen 1996; Economic Council 2001). The other programme, the Regional Centre Programme, is newer. Its aim is to spread growth, which in recent years has only taken place in a few main centres, across a network of smaller urban areas so that these towns might develop into strong cooperative centres for their regions. The number of designated centres is 34 and they are located in every part of the country. These regional centres comprise functionally homogeneous urban regions in terms of commuting, housing, service provision and demand. The programme is grounded in regional cooperation and the development of competence and know-how based on local strengths. It is too early however to draw conclusions as to its effectiveness.

15.4 Lessons from Finland Finland is a country small in population, but large in area. Finland’s regional problems follow from its main geographical and demographic features and are thus related to long distances and a scattered population. Nonetheless Finland is a relatively homogeneous country with no huge contrasts between regions or population groups. This is an advantage which usually distinguishes small countries from bigger ones, and which also creates a good starting point for regional policies. Typically, other factors have a larger impact on regional development than those relating to explicit regional policy alone. This is also a lesson learned from

280 Hannu Tervo

the experience of Finland. Intervention through taxation, income transfers and the provision of public services is of paramount importance in reducing regional income differentials and maintaining the level of welfare in regions. Broad regional policy is also important for other reasons. In the Finnish experience the early infrastructure-improvement policy had long-lasting effects on the development of the least developed regions in the north and east. The regionalisation of university education has turned out to be a very efficient means of spreading development across the country. An extensive network of higher educational institutions ensures that highly skilled human capital is available across the nation. This in turn acts as an incentive for firms and enhances the cultural as well as the economic attractiveness of otherwise weaker regions. The deliberate regional policy of the 1970s and 1980s aimed at decentralizing the manufacturing industries and locating them in the development areas. This policy was successful in its day, but gradually lost its importance as the country moved towards becoming an IT- and service-intensive economy. The trend towards the concentration of population and production accelerated in the 1990s in tandem with the halting of the trend to convergence in regional income differences. Market forces are strong and seem to be increasingly steering development towards centralized structures. This creates pressures on regional policy. Swimming against the current is hard and may not even be a good bet. Given that the driving force is now know-how and specialization, most recent regional policy is also based on these fields of activity. An attempt is being made to develop regional competence for the benefit of regional competitiveness. Support is directed at regional centres. Of course, naming specific localities as growth centres is a questionable practice, because growth can never be generated purely by public means. The aim, however, is to prevent growth from becoming wholly centralized in a couple of big cities, specifically the Helsinki region. Consequently, regional concentration is seen as unavoidable, but there are nonetheless hopes that concentration may take place at lower levels of intensity than hitherto experienced. This might be seen as only a delaying tactic, but it may also succeed, since a regional policy which follows this strategy is at all events trying to float with the tide. Whether the new policy will be successful is a question that remains to be answered. A turnaround has not been attained in very many regions so far. While the results do not yet seem very promising, it might be the case that they will only appear in the long run. In any case, Finland is obeying its earlier lesson that the ship of regional policy should move with the current, not drifting, but steering a deliberate course. To summarize, the general lessons to be learned from Finnish regional policy are: x Regional policy should not battle market forces. On the contrary, market forces should be harnessed for the benefit of more equal regional development. x Regional policy cannot succeed if other public policies work forcefully against it, since other policies may have a much larger impact on regions than explicit regional policy.

Regional Policy Lessons from Finland 281

x Regional policy should be constantly amended in line with economic development. To be effective each era needs a regional policy of its own.

Acknowledgement This article is based on a research project (number 200856) financed by the Academy of Finland.

References Ahola E, Kortelainen S (1996) An evaluation of the Centre of Expertise Programme (in Finnish). Ministry of the Interior, Department for Development of Regions and Public Administration Publications 17, Helsinki Armstrong H, Taylor J (2000) Regional economics and policy (third edition). Blackwell Publishers, Oxford Economic Council (2001) Regional development and regional policy in Finland - summary of the working group report. Prime Minister’s Office Publications 2001/2, Helsinki European Policies Research Centre (1996) Regional policies in EFTA and the EU: a comparative assessment. University of Strathclyde, Glasgow Haapanen M (2003) Studies on the determinants of migration and the spatial concentration of labour. University of Jyväskylä, School of Business and Economics. Jyväskylä Studies in Business and Economics 27, Jyväskylä Hanell T, Aalbu H, Neubauer (2002) Regional development in the Nordic countries 2002. Nordregio Report 2002:2, Stockholm Huovari J (1999) Regional unemployment and employment disparities (in Finnish). Pellervo Economic Research Institute Working Papers 27, Helsinki Huovari J, Kangasharju A, Alanen A (2001) Constructing an index for regional competitiveness. Pellervo Economic Research Institute Publications 44, Helsinki Kangasharju A (1998) Regional economic differences in Finland: variations in income growth and firm formation. Pellervo Economic Research Institute Publications 17, Helsinki Kangasharju A, Pekkala S (2001) Regional economic repercussions of an economic crisis: a sectoral analysis. Government Institute for Economic Research, Discussion Papers 248, Helsinki Kangasharju A, Pekkala S (2003) Why did regional disparities start increasing in the 1990s? Unpublished working paper Kauhanen M, Tervo H (2002) Who moves to depressed regions? - An analysis of migration streams in Finland in the 1990s. International Regional Science Review 25:200-218 Loikkanen HA, Rantala A, Sullström R (1998) Regional income differences in Finland, 1966-96. Government Institute for Economic Research, Discussion Papers 181, Helsinki Loikkanen HA, Riihelä M, Sullström R (1999) Income and consumption differences between and within urban, semi urban and rural municipalities (in Finnish). VATT, Government Institute for Economic Research, Discussion Papers 213, Helsinki Loikkanen HA, Riihelä M, Sullström R (this volume) Regional income convergence and inequality in boom and bust. Results from micro data in Finland 1971-2000

282 Hannu Tervo McCrone G (1969) Regional policy in Britain. George Allen Unwin, London Niittykangas H, Tervo H (1995) European integration, regional development and regional policy (in Finnish). Administrative Studies 14:323-334 Pehkonen J, Tervo H (1998) Persistence and turnover in regional unemployment disparities. Regional Studies 32:445-459 Pekkala S (2000) Regional convergence and migration in Finland, 1960-95. University of Jyväskylä, School of Business and Economics. Jyväskylä Studies in Business and Economics 4, Jyväskylä Pekkala S, Kangasharju A (2002) Regional labor markets in Finland: Adjustment to total versus region-specific shocks. Papers in Regional Science 81:329-342 Pekkala S, Tervo H (2002) Unemployment and migration: will moving help in finding a job? Scandinavian Journal of Economics 104:621-639 Ritsilä J (2001) Studies on the spatial concentration of human capital. University of Jyväskylä, School of Business and Economics. Jyväskylä Studies in Business and Economics 7, Jyväskylä Ritsilä J, Haukka J (2003) The role of structural funds in developing learning regions. Paper presented at the 43rd Congress of the European Regional Science Association, Jyväskylä, Finland 27-30 August 2003 Sisäasiainministeriö (1991) The international evaluation conference of Finnish regional policy. Ministry of the Interior, Department for Development of Regions and Public Administration Publications 14, Helsinki Taipale M (2002) Convergence of production and incomes between Finnish subregions (in Finnish). Pellervo Economic Research Institute Working Papers 58, Helsinki Tervo H (1985) The effects of regional policy on growth and development of manufacturing industries (in Finnish). University of Jyväskylä, Centre for Economic Research in Central Finland Publications 65, Jyväskylä Tervo H (1991) Studies on the economic case for and effects of regional policy. University of Jyväskylä, Jyväskylä Studies in Computer Science, Economics and Statistics 17, Jyväskylä Tervo H (1996) European integration and development of Finnish regions. In: Alden J, Boland P (eds) Regional development strategies: a European perspective. Jessica Kingsley, London Tervo H (1998) The development of regional unemployment differentials in the 1990s. Finnish Economic Papers 11:37-49 Tervo H (2000a) Regional structure and the factors affecting it (in Finnish). The Finnish Economic Journal 96:398-415 Tervo H (2000b) Migration and labour market adjustment: empirical evidence from Finland 1985-90. International Review of Applied Economics 14:343-360 Tervo H (2000c) Post-migratory employment prospects: empirical evidence from Finland. Labour - Review of Labour Economics and Industrial Relations 14:331-350 Vartiainen P (1998) Development phases in the Finnish regional policy (in Finnish). Ministry of the Interior, Department for Development of Regions and Public Administration Publications 6, Helsinki Virkkala S (1994) Economic restructuring and regional development in Finland. In: Doling J, Koskiaho B, Virkkala S (eds) Restructuring in old industrial towns in Finland. University of Tampere, Department of Social Policy and Social Work, Research reports A:6, Tampere

16

The Globalisation of Austrian Regions: New Policy Challenges and Opportunities

Michael Steiner Department of Economics, Karl-Franzens University, Graz, Austria

16.1 Changing Perspectives on Regional Inequality Austria and its regions have undergone profound changes both in economic situation and in policy orientation. These changes mirror a general transformation in the importance of regions and in the perspective on how regional inequality is understood and dealt with. As a paradoxical result of an enforced internationalisation of the (European) economy and of “globalisation”, the 1980s and 1990s saw the reemergence of regions as economic entities of their own and as promoters of economic development. For a long period regions were considered to be an irrelevant entity, just better or worse off parts of an eminent economy. As such regions were regarded as “passive” subdivisions of their respective national economies. This went hand in hand with changing interpretations on what aspect regions differ and according to which criteria they can be regarded as “unequal”. For a long time regions have been divided into “good” and “bad”, “centres” and “peripheries”, and “privileged” and “problem” regions. Often this division has been the basis for financial subsidies and other forms of economic support. This division thus both perpetuated, and partly answered, the old question as to the sense in which we can speak of regional inequalities. Massey (1979) emphasised in her seminal paper “In what sense a regional problem” that regional inequality has two aspects: areas have different degrees of attractiveness to the dominant form of economic activity and reveal differences in various indicators of social well-being. There are of course different ways of approaching these aspects both theoretically and empirically and of defining indicators for measuring regional disparities in regard to these aspects. Faced with growing internationalisation of the economy, different forms of adjustment enabling economic survival have developed. As a consequence there are several “types of region”, and there may also be different kinds of “problem region” that go beyond the simple dichotomies of “good” and “bad” regions, “centres” and “peripheries” etc. (Steiner 1990). New regional identities have led to new challenges and responsibilities: regions no longer can rely exclusively on national policies, economic competition becomes increasingly competition between regions.

284 Michael Steiner

The Europeanisation and globalisation of the economy especially has resulted in new reflections on what can be achieved on an international, national and regional/local level. This recognition of different levels of economic policy making implies that each level commands a restricted set of instruments which cannot automatically be used by other levels: the supranational level has goals and instruments different to those of the national or the regional levels. As a consequence, regional policy has become more aware of its limitations. It is not able to create employment by means of fiscal and monetary instruments as multipliers are much too low and spill-over effects too strong. Typical tasks for national economic policies include (de)regulation, tax systems, the construction of a legal framework and income/wage policy. Tasks more suited to a regional/local level include activities that promote clustering and network-building of innovation- and infrastructure-oriented instruments. These activities are centred on SMEs, have a medium-term perspective and as a goal an improvement of location and regional competitiveness. This chapter proceeds in the following manner. The “Austrian” perspective on regional inequality, grounded in the Schumpeter and Hayek tradition that stresses the region as the new locus of economic activity, is outlined in Section 16.2. Austrian regional policy and its reaction to changes in the dominant form of national economic activity will be analysed in Section 16.3. The paper concludes with some reflections on Austrian regional policy in a globalising economy.

16.2 New Emphasis on Regions and Regional Competitiveness an Extended “Austrian” Perspective A standard question in economic science, regardless of the level of aggregation, (national, regional or local) is why economies differ and in what respect. At a regional level this question has been posed with different emphases: are the differences in the development and in the wealth of regions a consequence of a different demand for their products, different production structures, different endowment of resources, different locational qualities, or different knowledge and/or use of technology? Furthermore, are they the result of geographical location (north/south, centre/periphery), or of dominant sectors (agriculture, industry, service)? From an “Austrian” perspective the answer cannot be found among these alternatives: If we regard regions as economic entities sui generis within national economies, we get a different perspective on regional inequality. One of the theoretical roots of territorial competitiveness can be found in the emphasis of the Austrian school on the creative function of the market. The market is an instrument which transfers incentives for economic change. Hayek, especially, underlined the explorative potential of the market with its capabilities not so much in allocation but in its flexibility to adjust to new situations. In the same sense Schumpeter always regarded the market as a process and not as a state.

The Globalisation of Austrian Regions 285

Economic competition is not price-taking behaviour under conditions of perfect competition, but a process with winners and losers. The process based character of the market brings about innovation via creative destruction. These innovations are not only of a technical nature, but also have organisational character thus opening new markets. Compared to these dynamic aspects the pure static, allocative advantages of the market are loosing importance (Streißler 1980). At the centre of this approach lies the creative function of the entrepreneur who has to create markets and to master production under conditions of uncertainty (Casson 1982). This personal element is a dominant feature of the market economy. The entrepreneur explores hitherto unknown opportunities and creates innovations connected with entrepreneurial risk and hazard. Extended to questions of regional development and of locational behaviour of entrepreneurs, this creative dimension leads to forms of behaviour surpassing allocative determinism. Entrepreneurs are “indecisive” with regard to location, they are “market-creating”, they can themselves induce change and are not restrained by locational costs and static allocative efficiency. Therefore approaches relying on static principles of regional distribution of economic activity have to be replaced by evaluations of regional potentials in the framework of the creative function of the market (Steiner 1990b, 2002). This process-oriented perspective gives special importance to the exploratory and experimental role of small and young firms within regions. Differential behaviour and variety are essential elements of the dynamic efficiency of markets and for the development of regions. In addition to this “Austrian-evolutionary” perspective several recent theoretical approaches have put new emphasis on the regional and spatial character of economic development: x New growth theory emphasises the importance of knowledge and human capital as an endogenous factor of growth: knowledge is not only created exogenously by means of research and development but is generated endogenously by processes of learning - knowledge turns into a by-product of co-operation and exchange of goods. As a consequence, the simple exchange of goods works as a mechanism for technological knowledge transmission and initiates a converging process between unequal trading partners. Trade and knowledge exchange turn into an engine of growth for less advanced regions as well (Jacobsen 1999). x Knowledge and learning appear to be an essential precondition for innovation. Innovation is increasingly based on interactions and knowledge flows. New insights into different forms of knowledge have received attention: tacit knowledge as non-ubiquified knowledge is regarded as a key determinant of the geography of innovative activities. This kind of knowledge defies conscious articulation and relies on close proximity (Cooke and Morgan 1998). x The nature of learning has also received growing attention. Originally understood as an individual effort and an adaptive response to a change in the environment, learning is now also considered as a social process of ongoing development embedded in a socio-cultural and regional context. Instead of an

286 Michael Steiner

“information processing” concept and of a cognitive performance, learning becomes a communicative process requiring new thinking about the nature and the forms of the transmission and dissemination of knowledge within a social and organisational context, such as the firm or the cluster (Grabher 1993; Maskell and Malmberg 1999). x The “new economic geography” approach puts the emphasis on economies of scope and spatial concentration under conditions of monopolistic competition and decreasing transaction costs. It gives a behavioural underpinning to theories of agglomeration and cumulative causation and explains agglomeration in a theoretical framework that is tractable and has solid microeconomic foundations. It is “new” also in the sense that its’ models result in lower costs of agglomeration and in the compensation for costs of concentration through technological linkages (Neary 2001). x A renaissance of interest in institutions and how they affect economic performance represents an attempt to identify factors, besides the rules of rational economic decision making, that influence human behaviour with regards to economic action. In this context, regional networks can be regarded as institutions for knowledge generation and diffusion substantially affecting the economic performance of regions (Metcalfe and James 2000; Steiner 2003). All these approaches emphasize the spatial and regional dimensions of economic activity. Territorial space is now an actor of development, knowledge an essential precondition for innovation demanding close proximity and organisational learning is one manifestation of clusters and regional networks. None of these topics is fundamentally new, some of them having a long tradition in economic thinking on technical progress and regional development. Yet their combination and recent reinterpretation have changed the perspective for interpreting regional inequality concepts.

16.3 Austria’s Regional Experience - Past and Present The main challenges of these new theoretical approaches for policy making can be summarised in three hypotheses: x Regional policy has to be understood as an activity that manages change. Its task is no longer to allocate and distribute given resources but to create and accommodate a process of dynamic evolution. x Regional policy needs to concentrate on the improvement of locational quality. x Locational quality is systemic in character. It is more than the sum of single elements (such as firms, development agencies, educational and research institutions, political bodies) but another tight network of interrelated agents. How has Austria’s regional policy reacted to these challenges? The change in perspectives was influenced by and reacted to a change in the dominant form of

The Globalisation of Austrian Regions 287

economic production taking place from the late 1960s to early 1980s. This change called for new entrepreneurial initiative and an inclination to take risks and to adapt to changing consumer demands. For this, highly qualified labour was needed and technological knowledge resided no longer in physical but in human capital. This implied changing locational requirements. Competitiveness was not embedded in cheap land, labour and inputs but in the existence of research and educational institutions, qualified infrastructure for fast communication and a critical mass of upper income consumers who signalled change in preferences. This called for a quality-oriented specialisation of regional production systems where the co-operation of firms, research institutions, business-oriented services, supporting public and semi-public institutions was connected via network relations. 16.3.1 Regional Differentiation in Austria - a Short Overview Regions in Austria are usually understood as the nine provinces (Länder) each with different sizes, population, and of course different economic potentials. They sub-divide into 99 districts which provide a more disaggregated level of analysis. The latest available data on differences in regional gross domestic product (RGDP) per capita (for 2000) reveals a marked differentiation between the Eastern and the Western part of Austria (with the obvious exception of Vienna (Figure 16.1). Burgenland, Austria’s only “Objective 1” region within the Structural Fund Programme and bordering Slovakia and Hungary, has the lowest RGDP. The western provinces (Vorarlberg, Tirol, Salzburg bordering Switzerland, Germany and Italy) are the ones with the highest. Carinthia and Styria (bordering Slovenia) and Lower Austria (bordering the Czech Republic and Slovakia) are roughly similar. Upper Austria which only then started to industrialize in the latter part of the first half of the 20th century, is somewhat in the middle.

Fig. 16.1. Regional gross domestic product 2000 per capita in Euro

288 Michael Steiner

Studies on regional differences and convergence/divergence in Austria confirm that there are significant differences among Austrian regions. Using data from the 1991 Austrian Population Census Badinger and Url (2002) show that substantial variation in unemployment rates (between 2.1 percent and 16.8 percent) at the district level. With the aid of a spatial filtering technique they find significant relations between regional unemployment rates and relative regional wages and transaction costs (such as the share of public housing, job vacancies, share of manufacturing sector, social assistance). A subset of the key factors explains about 50 percent of the regional variation in local unemployment rates. When they add the spatial component the explanatory power rises to 67 percent. They interpret these findings as indicating the considerable importance of transaction costs, regional amenities and other usually unmeasurable regional spill-overs. Hofer and Wörgötter’s (1997) analysis on regional per capita income convergence in Austria - from 1961 to 1989 - finds only weak support for convergence. The rate of convergence is only about 1 percent a year indicating slow progress in narrowing of regional differences. These results change with respect to spatial unit of analysis (region, district). Hofer and Wörgötter also show that controlling for socio-economic structure and vicinity of markets increase the pace of convergence to 2 percent per year. They conclude therefore that location specific factors play a role and that regional aid programmes should take region specific endowments into account. The question thus arises whether regional policy in practice meets this challenge. 16.3.2 Past Changes Up to the 1980s Austria’s economic situation and its regional policy were marked by some specific elements: x A large nationalised industry comprising almost a third of industrial employment with strong spatial concentration in the eastern parts of the country. Large sectors of economic activity were also heavily regulated. x A restricted and strongly fragmented set of policy instruments. The main government organizations with explicit regional objectives at the national level were the Ministry of Economic Affairs and more recently the Ministry of Science and Research. Policy instruments consisted mainly of capital incentives for underdeveloped regions and support for material infrastructure (Stöhr 1981). The “Länder” as the second level of Austria’s rather weak federal structure did not consciously pursue any consistent regional policy. x A regional situation strongly influenced and handicapped by geography. Austria’s long border with the formerly centrally planned economies marked the dead end of Europe’s market economies. For Austria’s eastern and southern regions close to this border this meant a strong reduction of their development potential (Steiner and Sturn 1993). The context for the various new regional initiatives of the 1980s was conditioned by specific factors. First, the aging of former centres of industrial

The Globalisation of Austrian Regions 289

activity. This brought about a change in the interpretation of regional development and regional strategies. It became evident that there was no linear logic to regional growth and that both rise and decline were possible. Second, the appearance of “bottom up” approaches to regional development formulated by Stöhr and Tödtling (1979) and Stöhr (1981) promoted a self-reliant form of development with special emphasis on local and regional internal feedback mechanisms and the mobilization of endogenous resources. These ideas were soon adapted as a key element in a strategy of revival for old industrial areas proposed by a team of experts (Tichy et al. 1982) to the Austrian and the Styrian (regional) government. Third, in 1986 the first technology park was opened in Graz, Styria setting a signal for new content and orientation in regional policy. For the first time a different kind of infrastructure - beyond roads, energy, availability of land - was considered important. Finally, two geo-political events also impacted regional approaches. The fall of the Iron Curtain in 1989 was of course a milestone for regional development especially in the Eastern parts of Austria. Yet the relief of not having to live in a dead end zone was soon followed by fears (Steiner 1998a). In addition, membership of the European Union since 1995 was probably the single most important event to influence also the strategic orientation and the perception of regional policy in Austria. It brought structural fund programmes and community initiatives to large parts of Austria and promoted new rules of the game for economic promotion and subsidies. 16.3.3 Present State The present state is a result of these changes. The main features of current regional thinking and initiatives can be summarised as follows: First, there is an awareness of the rising importance of the regional dimensions of economic policy. Many sectoral policies have a stronger and often explicit spatial aspect. This is especially true for research and technology policy. It represents part of a general tendency of increased targeting of technology policies and a greater attempt aimed at strategic direction expressed in the selection of key areas of technology adapted to specific regional characteristics. At the federal level the main instrument has been regional innovation and technology policy with a focus on improving local and regional infrastructure and stimulating innovation (ÖIR, JR, WIFO 1999). There has also been increasing regionalisation of labour market policies. Austria’s main labour market institution, has been reformed with a strong regional bias initiating territorial employment plans. Agricultural policy has adopted many features of endogenous regional development working on topics like renewable energy, direct marketing of local agrarian products and the upgrading of these products. Within Objective 5b programmes of the European Union, strong cooperation between regional and agricultural administrations have taken place. Second, the “Länder” have developed a stronger sense of economic selfawareness. This is especially true for those regions that are eligible for European Structural Fund Programmes. Going through the routines of programme design,

290 Michael Steiner

project selection and implementation has been an important step towards a rationalisation and reflection of the regions’ needs, strengths and weaknesses. Third, several regions have undergone a strategic re-orientation based on technology and innovation. Styria, for example, developed its own “Technology Concept” in 1995. The “Strategic Programme for Upper Austria 2000+” in 1998 has a key focus on “technology”. Carinthia made two consecutive attempts at formulating an “Economic Concept” in 1991 and a “Development Perspective for the Future of Carinthia” in 1998. Salzburg elaborated an “Economic Perspective” in 1997 focussing on technology and innovation oriented strategies and research. Fourth, the administration of regional policy has professionalised and has been handed over to semi-autonomous institutions characterised by arm’s-length autonomy with respect to their activities. In addition, regional policy has increased its range of instruments. Besides financial support services, infrastructure such as technology parks and incubator centres is provided, information and advisory services are offered and trade fairs and meetings for certain target groups are convened. In line with the increased stress on professionalism, the intuitive appeal of “alternative” approaches still exists but has lost influence. In the early 80s the “endogenous” approach was seen as a radical alternative to traditional regional policy borne by local people. As far as these ideas and such activities still exist they have become more professionalised and more influenced by economic reasoning (Gerhardter and Gruber 1999). At the same time many of these ideas live on in other concepts, such as the ideas of networking, intraregional cooperation and cluster formation. Regions have embraced “territorial networks” and “competence clusters” as a tool to improve locational quality. Styria was the first region to deliberately pursue such a cluster strategy, attempting to nurture an automobile cluster in the southern part of the region in addition to the old declining concentration of steel manufacturing in the northern part of the country. A network was formed between previously unconnected firms, partly situated in the old industrial area. The existing firms expanded, cooperation and specialisation increased and foreign investment was attracted. This approach was extended to other interlinked cores of economic activity (Adametz et al. 2000). This philosophy relies on the “discrete charm of clusters”. This implies that the well-founded assumption that regional specialisation in interlinked activities of complementary firms (in production and service sectors) and their cooperation with public, semipublic and private research and development institutions creates synergies, increases productivity, and leads to economic advantages (for an evaluation see Steiner 1998b). Cluster activities are a strong but not singular indicator of the increasingly “communicative” character of regional policy in Austria. This approach manifests itself in an elaborate process of consultation and discussion leading to the formulation of strategies and instruments. Communication in the form of consulting, delivering services and moderating has become a new task for regional institutions. This involves building and supervising trust relations, initiating learning processes and monitoring the outcomes. Finally, evaluation came to the forefront of policy activities in the 1990s. The regional special support actions of the Federal Government and the “Länder”,

The Globalisation of Austrian Regions 291

which accounted for the highest share of regional funding, were some of the first regional activities to receive ex-post evaluation in the late 1980s (Aiginger et al. 1989). These activities were followed by the evaluation of RIP (regional innovation premium) resulting in a report in which the new challenges of regional innovation policy were clearly taken into consideration (Hutschenreiter 1993). This exercise was repeated in 1999 (ÖIR, JR, WIFO) and elements of evaluation were also contained in critical assessments of regional development institutions (e.g. Geldner 1995 for Styria, Bodenhöfer and Steiner 1997 for Carinthia). The real catalyst for systematic evaluation of regional policy instruments however was the structural funded ex ante, interim and ex post evaluations. These were the basis for extensive evaluation efforts in which an institutional basis was to be installed consisting of accompanying councils, yearly reports, data systems for monitoring and guidelines (lists of indicators, criteria for evaluation etc.). 16.3.4 The Influence of EU-Membership The effect of Austria’s membership of the EU on regional policy has resulted in an adjustment and intensification of regional policy and at the same time, a measure of disillusion promoted by over-optimistic expectations (Gruber 2001; Schremmer and Steiner 2003). With the general rules of the competition policy leading to stronger territorial concentration of funds, expectations rose concerning the effects of the integration of Austria into the regime of EU-mechanisms: x Regional policy was anticipated to receive more attention and gain importance. x It was expected to lead to a reduction of regional disparities and accelerated structural change. x It was considered as a means to achieve a higher degree of rationality with new forms of policy implementation, clear responsibilities and innovative methods and instruments. On a general macroeconomic level, the effects of EU-membership between 1995 and 1999 amounted to a 0.5 percent growth in Gross Domestic Product (GDP). The welfare effects over the period summed up to 2 percent of GDP. Concentrating on the effects of the integration of Austria into EU-competition policy and the Structural Fund Programme and simulating a scenario without such participation, results in a rise of gross investment of 0.5 percent, an additional growth effect on GDP of 0.1 percent and an additional employment of 2000 (Breuss 2000). Altogether these macroeconomic effects appear to be rather modest. An expert-panel identified additional qualitative effects confirming some of the expectations and refuting others (Schremmer and Steiner 2003): x Regional policy in general gained recognition and importance. Regional policy funding within total economic subsidies increased in 1995 to 31.5 percent and more than doubled compared to previous years. The total sum of structural funds between 1995 to 1999 amounted to 1.623 Bn Euro (at 1995 prices). In

292 Michael Steiner

addition, there are clear indications of the high additionality of the structural funds. x Doubts increased as to the achievements of regional policy. There is no compelling evidence of a reduction in regional disparities. The turnarounds of specific parts of Objective 2 regions e.g. in the “old industrial area” of northern Styria were due more to a policy of endogenous renewal started in the 1980s than to the activities of the Structural Fund Programme (Gruber 2001). x There was little programmatic and methodological innovation. The main thrust of policy relied on the traditional tools of business aid and investment support. Also, programme design relied on existing infrastructure and used predominantly existing instruments. Only the smaller initiatives and the community initiatives (such as LEADER, URBAN) enabled new forms and content of activities. x Evaluation methods of traditional policy also remained. Methodology relied on simple straightforward quantification with an emphasis on final outcomes rather than processes, net-effects were often not separated from gross effects and adequate data was often not available or too costly (Gruber 2001). These caveats not withstanding, the main conclusions and general evaluative results did give a positive impression of the over-all added value of the integration of Austria’s regional policy into the framework of EU-policy: x The “profile” of Austria’s regional policy clearly improved. x An “enabling effect” brought about an increased awareness reducing barriers for the efficient use of the instruments and enforcing the systematic and control mechanisms of regional policy. x This also led to “institutional added value” implying programme orientation with the need for strategic considerations and longer time frames for policy evaluation. New institutions were generated, funds were redirected and even enlarged and overall this resulted in a more transparent public support system.

16.4 Regional Policy in a Complex World: Some Conclusions Regions in Austria certainly have considerably more of a global dimension now than they had twenty years ago: a rather closed system has been opened successively, their access to a wider world has been made easier, their tools have become more differentiated. While Austria’s membership of the EU did not generate this process, it did accelerate it. What gave Austria’s regions a more global dimension? x Regions in Austria have increasingly “clustered” as a form of regional production systems. They realised that locational advantage has turned from comparative to competitive advantage. This locational specification is founded on the nurturing of a specific regional profile. Without recognizable compe-

The Globalisation of Austrian Regions 293

tences based on highly qualified lines of cooperation regions get caught in the so-called “globalisation trap”. Clusters have become the local response to globalisation and a region specific tool for global competition. x Austria’s regions have integrated in a larger Europe. This has meant the acceptance of additional rules both of competition and of the regional policy. It has also meant specific responses to development challenges. The “Austrian” response has been more process and systemic based strategy with an emphasis on how markets emerge rather than how they allocate. Yet these processes still take time. The turn-around in some of the problem areas, such as the old industrial areas of Upper Styria, has taken more than twenty years. An active cluster policy then is not an additional policy of business support in the traditional sense of financial subsidies but rather an instrument aimed at empowering entrepreneurs. x The increased opening of borders to the East and South has also helped globalise the regions. The usual indicators of expansion and integration such as foreign direct investment, export-imports to the new members of the EU, cross-border cooperations, reveal that Austria’s regions (especially in the East and South) have positively taken up this challenge. These border regions have changed from distant, peripheral zones to functional spaces where different regions, cultures and systems touch each other and create a sphere for the exchange of cultures, ideas and goods. These former border regions may in the future turn out to be privileged regions exploiting new opportunities at their doorstep. Finally, changes have been influenced by specific features of Austrian society and its economy. Austria’s dominant political philosophy of dialogue, compromise and partnership has coincided with new challenges of regional change relying on slow processes of changes in mentality. EU regional policy however tends to emphasize quantitative goals (such as creation of jobs, support for SMEs, inducement of additional investment etc.). The strong time constraint implied in this leads to political pressure to make use of all available funds within the programme period and it favours projects with a predictable outcome. This creates a tension with Austria’s dominant philosophy grounded in a systemic line of thinking relying on slow processes of changes in mentality through endogenous development and active participation. This philosophy has its roots in the past where the lower profile and the lack of formal structures allowed for a more flexible approach. Austria’s small size may have afforded an opportunity for policy changes and meeting the regional challenges of globalisation. Small country size means lower costs of errors and better opportunities for trial. In this respect, the experience of small countries may be a harbinger of things to come in the large.

294 Michael Steiner

References Adametz C, Fritz O, Hartmann C (2000) Cluster in der Steiermark: Lieferverflechtungen, Kooperationsbeziehungen und Entwicklungsdynamik. Joanneum Research, Graz Aiginger K, Hutschenreiter G, Geldner N, Jeglitsch H, Palme G, Szopo P (1989) Die gemeinsamen regionalen Sonderförderungsaktionen des Bundes und der Länder. Österreichisches Institut für Wirtschaftsforschung, Wien Badinger H, Url T (2002) Determinants of regional unemployment: some evidence from Austria. Regional Studies 36(9):977-988 Bodenhöfer HJ, Steiner M (1997) Evaluierung des Kärntner Wirtschaftsförderungsfonds. Klagenfurt, Graz Breuss F (2000) An evaluation of the economic effects of Austria’s EU membership. Austrian Economic Quarterly 4/2000, Vienna Casson M (1982) The entrepreneur: an economic theory. Robertson, Oxford Cooke P, Morgan K (1998) The associational economy: Firms, regions and innovation. Oxford Economic Press, Oxford Geldner N (1995) Eine Einschätzung der Aktivitäten der steirischen Wirtschaftsförderung. Graz, Wien (mimeo) Gerhardter G, Gruber M (1999) Evaluation der “Förderung eigenständiger Regionalentwicklung (FER)” in Österreich. Joanneum Research, Institut für Technologie- und Regionalpolitik, Graz Grabher G (ed) (1993) The embedded firm: on the socio-economics of industrial networks. Routledge, London Gruber M (2001) Bewertungen von Programmen der EU-Regionalpolitik - Zwischen Anspruch und Wirklichkeit. Plattform Technologie Evaluierung,12:14-21 Hofer H, Wörgötter A (1997) Regional per capita income convergence in Austria. Regional Studies 31(1):1-12 Hutschenreiter G (1993) Regionalförderung unter neuen Rahmenbedingungen. WIFOMonatsberichte 11, Wien, pp 578-585 Jacobsen A (1999) Zentralosteuropäische Wachstumsperspektiven im Handel mit der Europäischen Gemeinschaft. Shaker, Aachen Maskell P, Malmberg A (1999) Localised learning and industrial competitiveness. Cambridge Journal of Economics 23:167-185 Massey D (1979) In what sense a regional problem. Regional Studies 13:233-234 Metcalfe JS, James A (2000) Knowledge and capabilities: a new view of the firm. In: Foss N, Robertson P (eds) Resources, strategy and technology. Routledge, London, pp 3152 Neary P (2001) Of hype and hyperbolas: introducing the new economic geography. Journal of Economic Literature 39(2):536-561 ÖIR (Österreichisches Institut für Raumplanung), JR (Joanneum Research), WIFO (Österreichisches Institut für Wirtschaftsforschung) (1999), Regionale Innovationspolitik 2000, Wien ÖROK (1973) Regionalpolitik in Österreich, Bericht des Bundeskanzleramtes (Büro für Raumplanung) an die OECD, Arbeitsgruppe Nr. 6 des Industriekomitees. Österreichische Raumordnungskonferenz, Schriftenreihe Nr. 3, Wien ÖROK (1978) Zweiter Raumordnungsbericht, Österreichische Raumordnungskonferenz, Schriftenreihe Nr. 14, Wien Schremmer C, Steiner J (2003) Regionale Auswirkungen des EU-Beitritts Österreichs: Bisherige Erfahrungen. Austrian Institute for Spatial Planning (commissioned by the Federal Chancellery), Vienna

The Globalisation of Austrian Regions 295 Steiner M (1990a) ‘Good’ and ‘bad’ regions? Criteria to evaluate regional performance in face of an enforced internationalisation of the European economy. Built Environment 16(1):52-68 Steiner M (1990b) How different are regions? An evolutionary approach to regional inequality. In: Peschel K (ed) (1990) Infrastructure and the space economy. Essays in Honour of Rolf Funck, Springer, Berlin Steiner M (1998a) Changing borders, extending markets and the feeling of “angst”. In: Kicker R, Marko J, Steiner M (eds) Changing borders: legal and economic aspects of European enlargement. Lang, Frankfurt, pp 165-183 Steiner M (1998b) The discreet charm of clusters: an introduction. In: Steiner M (ed) Clusters and regional specialisation. Pion, London, pp 1-17 Steiner M (2002) Clusters and networks - institutional setting and strategic perspectives. In: McCann P (ed) Industrial location economics. Edward Elgar, Cheltenham Steiner M (2003) Regional knowledge networks as evolving social technologies. International Journal of Technology Management 26(2/3/4):326-245 Steiner M, Sturn D (1993) From coexistence to cooperation: the changing character of Austria’s South-Eastern border. In: Ratti R, Reichmann S (eds) Theory and strategy of border areas development. Helbig und Lichtenhahn, Basel, pp 347-376 Stöhr W (1981) Regionale Wirtschaftspolitik. In: Abele H, Nowotny E, Schleicher S, Winckler G (eds) Handbuch der österreichischen Wirtschaftspolitik. Manz, Wien, pp 325-334 Stöhr W, Tödtling F (1979) Spatial equity - some antithesis to current regional development strategy. In: Folmer H et al. (eds) Spatial inequalities and regional development. Leiden, pp 133-160 Streißler E (1980) Kritik des neoklassischen Gleichgewichtsansatzes als Rechtfertigung marktwirtschaftlicher Ordnungen. In: Streißler E, Watrin S (eds) Zur Theorie marktwirtschaftlicher Ordnungen. Mohr & Siebeck, Tübingen Tichy G, Geldner N, ÖIR (1982) Regionalstudie Obersteiermark. BKA, Land Steiermark.

17

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

Stephen Roper Aston Business School, Birmingham, UK

17.1 Introduction There is now substantial evidence that investments in R&D and innovation increase firms’ capacity to achieve high levels of productivity and wealth creation (Crepon et al. 1998; Lööf and Heshmati 2000; 2001). There is also evidence from a number of different types of studies that, despite potential crowding-out and efficiency effects, public support for R&D and innovation has positive effects on the extent of R&D and innovation activity both within the assisted firm and, through spillovers, in co-located and/or related companies. Griliches, for example, summarises evidence from a range of studies “all pointing in the same direction: R&D spillovers are present, their magnitude may be quite large, and social rates of return remain significantly above private rates” (Griliches 1995, p. 72). Evidence from a number of US empirical studies, following Jaffe (1989), has also suggested that knowledge generated at universities spills over into the local industrial sector, leading to higher innovative outputs than would otherwise be the case. The existence of such positive spillovers from R&D investments creates the opportunity for technology-led regeneration strategies and emphasizes the potential importance for economic development of innovation policy1. The importance of the region as a unit of analysis for innovation, and as an important level at which strategic innovation policy and support is appropriate, is suggested by the social and interactive (i.e. uncodified/unwritten) nature of the innovation process itself (Feldman 1994; Audretsch and Feldman 1996; Acs 2000; De la Mothe and Paquet 1998). It can be argued, for example, that while 1

A key point, discussed in more detail later, however is the ability of the region to appropriate these positive spillovers. In a national context, Griliches (1995, p.83) notes: “All of the recent studies of R&D returns at the industry or national levels continue to find significant social returns ... A caveat should be entered here, however. For a relatively small country … such as Sweden or Israel, the same spillover arguments may not apply. Most of the technological spill-out is to the rest of the world and it is not clear why it should be supported by the citizens of the originating country. The social argument for R&D support in this case depends on externalities in education and training and on the development of a ‘knowledge-from-abroad-absorbing system’ which as a social good may not be provided adequately by the private market”.

298 Stephen Roper

information can readily be transmitted over long distances, transmitting knowledge, which is often tacit and sticky, necessitates face-to-face interaction and frequent and repeated contacts. For Audretsch (1998), this means that: “Knowledge spillovers tend to be spatially restricted... The increased importance of innovative activity in the leading developed countries has triggered a resurgence in the importance of local regions as a key source of comparative advantage.” The locus of economic policy has hence increasingly shifted to the regional level, and “a new policy approach is emerging, focusing on enabling the creation and commercialisation of knowledge... encouraging R&D, venture capital and new-firm start ups” (Audretsch 1998, p. 26). The question considered here is whether the positive and localized benefits of R&D and innovation, envisaged by Griliches and Audretsch, can be harnessed as a means of addressing regional disparities in small countries. Or, whether the centralising tendencies of agglomeration economies - discussed extensively in Fujita and Thisse (2002) - inevitably lead to disproportionate concentration of high-tech industry in the main urban metropolis. In countries the size of Finland, Israel or Ireland, for example, with populations of 5-6m, nearly half of the population is located in the main metropolitan area. Does this concentration mean that policy attempts to promote innovation activity, and hence economic development, in other regions are inevitably doomed to failure? Or, is the geographical scale of these countries sufficiently small to allow firms in any area to take advantage of the agglomeration advantages generated within the main metropolis? More specifically, we consider the evidence on the regional impact of innovation policy and support measures in two small countries, Israel and Ireland. The comparison is interesting because of the very different development paths of high-tech industry in the two countries - epitomised by a policy of R&D-led development in Israel and “industrialisation by invitation” in Ireland (Roper and Frenkel 2000). It is also interesting because in Israel local governance structures remain relatively weak, and policy has therefore been predominantly top-down with few specific regional initiatives. Instead, spatial disparities in innovation policy have largely reflected centrally determined differences in levels of public grant support for R&D and high-tech start-ups. In Ireland - arguably as a consequence of the availability of EU support for individual regions - innovation policy has had both top-down and bottom-up elements. In the first category are for example - national measures designed to expand and spatially disperse higher education and stimulate local venture capital markets. In the latter category are the type of regional innovation strategies developed by the Shannon region, on Ireland’s Atlantic coast designed to draw together the key social partners and knowledge-based institutions in a positive regional coalition (Shannon Development 1999; Andréosso-O’Callaghan 2001). The remainder of this paper is organized as follows. Section 17.2 sets the scene for the analysis by providing an overview of the conceptual and empirical issues involved in any assessment of the potential for innovation-led regional policy in small countries. We begin by noting the inevitable openness of small countries and their national dependence on the wider global economy. We then move on to

Innovation Policy and Spatial Disparities 299

consider more standard issues relating to spillovers, locational advantages and alternative types of policy interventions within the bounded geography of a small country. The approach adopted is evolutionary, emphasizing the social and interactive nature of the innovation process, and the important contribution of private and public sector institutional competencies and their degree of association (Cooke and Morgan 1998). Sections 17.3, 17.4 and 17.5 then introduce the case study countries before focussing on Ireland and Israel in turn, concentrating on the implementation and effectiveness of innovation policy and its regional dimension. Section 17.6 summarises the implications of the cases considered and provides our final assessment of the scope for innovation-led regional policy in small countries.

17.2 The Potential for Innovation-Led Regional Development in Small Countries Before considering the specifics of innovation-led development in small countries, it is perhaps worth pausing to consider briefly the nature of the innovation process itself. In our view, innovation is perhaps best understood as an evolutionary process in which knowledge and information are combined to generate new product and service offerings. This has both local and global, or at least international, dimensions. At the local level the systemic approach offered by the regional innovation systems literature (e.g. Braczyk et al. 1998) is useful, highlighting the fundamental importance of R&D and technology transfer as well as the importance of the capabilities of, and linkages between, local organisations (Braczyk et al. 1998; EU 1998), “untraded interdependencies” (Dosi 1988), knowledge “spillovers” (Audretsch and Feldman 1996), and knowledge integration through “open systems architecture” (Best 2000). An operationally useful distinction can be between two main elements - or subsystems - of a regional innovation system: x The knowledge generation and diffusion sub-system comprising those organizations within the region whose corporate objectives relate either to knowledge creation (i.e. researching organizations), knowledge sourcing, knowledge or technology transfer or regional or national economic development (e.g. universities, third-level colleges, government and industry research organisations, and technology transfer and technology mediating institutions). x The knowledge application and exploitation sub-system consisting primarily of firms linked through (vertical) local supply-chains and trading relations and (horizontal) collaborative networks. The final and crucial element of any RIS is the pattern of local linkages between the knowledge generation and knowledge exploitation sub-systems. Even within small countries, however, marked differences are likely to exist between the strength and characteristics of such local regional innovation systems. For example, Shefer and Frenkel (1998) emphasise the locational benefits for

300 Stephen Roper

innovation of firms located in the central region of Haifa, Israel in comparison to those in adjacent sub-urban or peripheral areas. Kipnis (1998) also points to differences in business culture between urban centres - even those as close as Haifa and Tel Aviv - and argues that this may be influential in terms of business location. Such observations are not universal, however. Roper (2001), for example, in an econometric examination of the innovation capability of manufacturing plants throughout Ireland finds no evidence of any positive metropolitan effect. Instead, the below average proportion of product innovating plants and the relatively high proportion of process innovators in urban areas of Ireland is more suggestive of the results of Davelaar and Nijkamp (1989; 1992) who found a lower incidence of innovation in Amsterdam compared to less urbanised areas of the Netherlands. Similar, and more recent evidence, comes from Beugelsdijck and Cornet (2001) who in their analysis of Dutch manufacturing firms find no evidence of stronger local B2B spillovers but do find some evidence of positive spillovers from local technological universities. Beugelsdijck and Cornet (2001) conclude: “This study thus suggests that the Netherlands is too small a country to have proximity play the leading role in facilitating knowledge spillovers. This conclusion might a fortiori hold for other regions of similar size” (Beugelsdijck and Cornet 2001, p. 17). The suggestion is that for regions within small countries - perhaps with the exception of the main metropolis - it may be difficult to appropriate much of any knowledge spillovers generated by local R&D and innovation activity. Such difficulties are likely to be particularly significant in less developed regions where the economy is likely to be dominated by smaller firms concentrated in less hightech sectors and where the innovation supporting public and private infrastructure is likely to be weaker. For example, Fernandez et al. (1996) have argued that the dominance of the Spanish economy by small and medium-sized firms, limits its capacity to appropriate locally the full benefits of publicly supported basic research activity, and suggest that for less developed regions - or those with an intermediate technological and industrial base - the locally captured social returns might be greater from strategic or applied rather than basic research. Aside from these local issues perhaps the other major influences on the effectiveness of innovation-led regional policies in small countries relate to the dependency of small nations on global economic and social conditions. Henderson et al. (2002) offer a useful framework for analysis here in the form of the global production network (GPN). This is defined as: “the global network of firms, institutions and other economic agents which shapes, and is shaped by the fundamental processes of knowledge and wealth creation, enhancement and exploitation; corporate, collective and institutional elements of organisational power; and, spatial and network embeddedness.”

Within each GPN Henderson et al. (2002) identify three key characteristics: the process of value creation, upgrading and distribution; the role (and power) of firms, organisations etc; and, the degree of territorial and network embeddedness. In each case, the unified notion of the GPN suggests an analysis in which the strengths of the small country - or its constituent regions - are positioned, for

Innovation Policy and Spatial Disparities 301

example, within global production and distribution systems. Roper and Grimes (2005), examine the contrasting positions of Helsinki, Dublin and Tel Aviv within the ICT or “new economy” GPN of the 1990s. In Dublin, and more largely in Ireland, they argue the process of value creation in high-tech industry has been supported primarily by inward technology transfer and FDI as US companies have established a European foothold. In Tel Aviv (although not in Israel more generally), they argue that high tech growth during the 1990s was based largely on locally generated knowledge and entrepreneurship supported by “knowledge seeking” capital inflows, primarily from the US. In Helsinki, and to some extent other parts of Finland, locally generated knowledge and “rent seeking” capital inflows were also important although the central driver of growth was not entrepreneurship (as in Israel) but rather the dramatic growth of Nokia and its suppliers. Perhaps three key points emerge from this view of the global positioning of small countries. First, it is hard to generalise about the nature of the technological linkages between small countries and the global economy. Put crudely, Finland and Israel, for example, probably had net technology outflows during the 1990s, while Ireland had a net inflow. From a regional perspective, the key question is which of these types of technology trade balance is most conducive to reducing spatial disparities? Secondly, it is perhaps easier to generalise about the importance of global capital flows in shaping development patterns in small countries: technology-seeking capital inflows are probably less likely to flow to underdeveloped regions than those either rent or factor seeking capital inflows. Thirdly, in addition to these “market” effects, supra-national organisations and partnerships may also play a particularly important role in small nation’s development. In the case of Israel, bi-national R&D agreements such as BIRD have been important for many companies, while EU collaborative programmes have been more significant for Ireland and Finland (e.g. Yearly 1995; Luukkonen 2000). In regional terms, R&D agreements such as BIRD - and EU programmes such as the Framework programmes - which involve existing companies seem likely to reinforce rather than alter spatial disparities (and may do little to reverse the pattern of international technology flows). On the other hand EU programmes - particularly the Structural Funds - which have an avowedly redistributive objective may work more positively to the benefit of less developed regions. Even here, however, potential issues arise if the EU agenda is not consistent with the development needs of particular regions (e.g. Yearly 1995).

17.3 Ireland and Israel: Some Innovation Benchmarks Before looking at the regional dimension of innovation policy in Ireland and Israel it is probably useful to look briefly at some overall benchmarks of national innovation investment and performance (Table 17.1). For example, historically investment in R&D in Ireland has been notably lower than that in Israel and other small countries. Indeed, in 2000, levels of civilian R&D investment in Ireland

302 Stephen Roper

were only a third those in Israel, with levels of per capita investment 40 percent of that in the US. In the same period, Israeli civilian R&D spending was 20 percent higher than that in the US, and notably above that in Finland and Denmark. These comparisons reflect both the historical priority given to R&D investment in Israel - and its low priority in Ireland - as well as the dependence of the Irish economy on inward technology transfer (Yearly 1995). They also make clear that levels of R&D investment in Israel, at least by 2000, were high by global standards while those in Ireland were relatively low. The case study countries therefore illustrate two very different contexts in which the effectiveness of regional innovation policy can be considered. The composition of R&D expenditure between countries is also interesting in this context. In particular, it is interesting to note that despite the difference in the scale of investment the composition of R&D spending in Ireland and Israel is broadly similar (Table 17.1). In other words, Irish R&D spending is consistently lower than that in Israel across all sectors - government, higher education and business. In both Finland and Denmark notably higher levels of government R&D activity are found. As indicated earlier the marked contrasts between levels of R&D investment in Ireland and Israel in 2000 reflect longer term differences in the technological trajectory of the two countries. This has also led to profound differences in the degree of “association” within the two national innovation systems. In particular, in Israel a relatively long tradition exists of positive interaction between the university and industrial sectors while such links remain relatively weak in Ireland. The historical lack of R&D investment in Ireland also contributed to the development of a dual economy: a technologically advanced externally-owned sector based largely on R&D conducted elsewhere and a technologically weaker indigenously-owned sector (Wrynn 1997). As a result, neither the externallyowned sector nor the weaker indigenously-owned manufacturing sector demanded or contributed much to the development of the Irish NSI (OECD 1974). Ireland’s technological dependency is not, however, limited to technology policy. Large scale inward investment since the late 1950s has meant that 44.1 percent of manufacturing employment, 68.4 percent of net output and 87.7 percent of manufacturing exports are now accounted for by foreign-owned enterprises (Ruane and Gorg 1997). Moreover, only two Irish-owned firms appear in the list of Ireland’s top 20 electronics companies (Roper and Frenkel 2000). Comparing innovation performance internationally is always difficult, particularly where as in Israel there is no national innovation survey. Roper (2002) provides an overview of the historical evidence in terms of patents, innovation among high-tech companies - for which some survey evidence does exist, and market shares in research driven sectors. These comparisons generally reflect the technological strength of Israeli firms and the relative weakness of many Irish companies, particularly those which are indigenously owned. In terms of patents, for example, Roper cites evidence from the USPTO which “points to Israel’s disproportionately strong patent performance. Not only is this evident in terms of a greater degree of technological self-sufficiency in the Israeli market but also in terms of the increasingly strong Israeli presence in US patent applications … In

Innovation Policy and Spatial Disparities 303

1994, the number of patent applications made by Israelis in the US was 27.3 percent of the number of applications registered domestically (by Israelis); in the UK and Ireland corresponding figures were 13.1 and 6.5 percent respectively”. Table 17.1. R&D and patenting comparisons: Israel and Ireland in context

Israel

Ireland

Finland

4.2 118.9

1.4 39.5

3.2 96.3

2.1 74.4

Composition of R&D (%, 2000) Private Non-profit Higher Education Government Business

3.7 17.2 6.4 72.7

0.7 19.2 7.0 73.1

0.7 19.7 11.4 68.2

1.2 20.3 15.2 63.3

Patent Statistics (USTPO) 1. Success Rates Patent Applications Patents Granted Success Rate (percentage)

16670 7333 44

2703 1178 44

13684 6957 51

11472 6346 55

1777 5308 25.1

260 906 22.3

853 5305 13.9

706 3959 15.1

6.1 22.3 18.3

13.1 37.9 32.3

6.1 8.0 7.7

8.2 17.7 16.3

R&D Expenditure (2000) 1. Percent of GDP 2. Per Capita (US =100)

2. Institutional and Institutional Patents Independent Patents Granted Institutional Patents Granted Independent (percentage) 3. International Collaboration by Type: Independent Patents Granted (%) Institutional Patents Granted (%) Overall %

Denmark

Sources: R&D Expenditure and Composition: Central Bureau of Statistics, Israel; Patent Statistics, Archambault (2001).

More recent evidence from Archambault (2001), also based on USTPO patent applications and patents data, largely reinforces the same points. In terms of both patent applications and patents granted, Israel has a markedly stronger performance than Ireland with nearly 8 times as many patent applications and seven times as many patents granted (Table 17.1). Success rates (the proportion of successful patent applications), however, are broadly similar for the two countries. Other aspects of the patent data outlined by Archambault (2001) reflect the pattern of R&D investments in the two countries. First, despite the level difference, the organisational composition of patents granted is broadly similar in the two

304 Stephen Roper

countries, reflecting the similarity in the composition of R&D spending across government, higher education etc. Secondly, in terms of independent and institutional patents a higher proportion of those registered in Ireland involved international collaboration than in Israel (Table 17.1), reflecting the technological openness of the Irish economy and potentially its dependence on external technology. The context for our discussion of the regional dimensions of innovation policy in Ireland and Israel is therefore very different. In Ireland, a relatively weak indigenous technology base is reinforced by inward technology transfer, while in Israel strong indigenous technological capability is reinforced by positive institutional linkages between knowledge generating and knowledge applying sectors.

17.4 Innovation by Invitation - Regional Dimensions of Innovation Policy in Ireland Since the late 1950s the key characteristic of industrial policy in Ireland has been its willingness to embrace mobile international capital. Successive waves of inward investment have resulted; during the 1970s inward investment was dominated by computer and component manufacturing; during the 1980s this was broadened to include software manufacturing and reproduction; and, in the 1990s further inward investments in both computer hardware and software manufacturing were accompanied by significant investments in chemicals and pharmaceutical manufacture. By 1998, figures published by the Irish Industrial Development Authority, suggested that 61 percent of electronics plants in Ireland were US-owned, and that these plants accounted for 82 percent of employment in the Irish electronics sector2. In export market terms, this inward investment also meant that by the mid-1990s, Ireland also had a 4.8-4.9 percent share of the world market for computer equipment and components and that the Irish electronics sector was exporting more than 90 percent of its production (Roper and Frenkel 2000). The key point here is that throughout almost the entire post-war period Ireland has benefited from a consistent flow of factor-seeking inward investment from outside, largely in the form of investments in greenfield manufacturing facilities. The knowledge-base for this investment came from outside Ireland, as did any impetus for changing or upgrading products and processes. This inward investment, and inward technology transfer, was accompanied by relatively low levels of domestic investment in R&D both in the public and private sectors. Innovation, the introduction of new products and processes, in the Irish economy

2

See Roper and Frenkel (2000), for a comparison of the growth of the Irish and Israeli electronics sector and Shefer, Frenkel and Roper (2003) for a comparison of innovation activity by firms in the two industries.

Innovation Policy and Spatial Disparities 305

was therefore primarily the domain of the inward investor, with locally-owned Irish firms making only a modest contribution. At a national level, Ireland’s achievement through to the late 1990s - and it is a substantial one - was to maintain the flow of inward investment, reinvestment and inward technology transfer. From a regional point of view, however, two questions arise. First, has it been possible to distribute this flow of activity to “… provide for the maximum spread of development, through all regions, giving an increased and wider range of economic and social opportunities and so minimising population dislocation through internal migration”?3. And, second, has inward investment and its associated inward technology transfer stimulated innovation in indigenous companies and generated viable regional innovation systems in Ireland’s less developed regions? Over the years a number of measures have been adopted in Ireland to encourage the dispersion of inward investment away from Dublin towards Ireland’s South and West coasts. Key policy instruments were: 1. From the early-1950s grant support was made available for capital investment and training, aimed to encourage locally-owned businesses to relocate or create industrial employment in certain underdeveloped or Designated Areas (DAs)4. This policy was maintained subsequently with more generous grant rates and assistance packages being offered to firms locating in less developed areas. 2. Institutional developments were also made with the creation of specific regional development bodies for the Gaeltacht (now Udaras na Gaeltachta) and the Shannon region (now Shannon Development). 3. Beginning in the mid-1960s, advance factories were built in geographically dispersed industrial estates to attract new FDI to areas outside Dublin (Drudy 1991). Evidence of early the success of this Irish policy of dispersal comes from O’Farrell (1980) who reports that from 1960-1973, 58.9 percent of new inward investment plants located in the DAs compared to 48.9 percent of new Irish firms. Meyler and Strobl (1997) also note that from 1972-1979 employment in the Designated Areas rose by 45.9 percent compared to an increase of only 6.9 percent in the non-Designated Areas (NDAs). A more contemporary perspective would suggest a similar picture with substantial clusters of externally-owned plants in Galway, Limerick and Cork and a significant representation in most Irish towns of any size. Entrepreneurial activity in high-tech sectors - notably software however, remains strongly concentrated in Dublin, however, suggesting the continued importance of agglomeration and informational economies in this sector in particular (e.g. Crone 2002; Roper and Grimes 2005). Top-down policies of dispersal have been effective in Ireland, and have brought high-tech industry to previously under-developed parts of the country, alongside developments in education and infrastructure. This has undoubtedly raised income 3 4

Source Review of Regional Policy (1972), quoted in Meyler and Strobl (1997). The Designated Areas included the counties of Sligo, Leitrim, Roscommon, Mayo, Galway, Clare, Donegal, Kerry and West Cork.

306 Stephen Roper

levels in previously less developed regions and contributed to the regeneration of cities such as Galway, Limerick and Cork. In many cases, however, inward investment plants are only weakly embedded in the local economy, with few local suppliers, and even fewer with whom they work on a collaborative basis (HewittDundas et al. 2002). It is less clear, therefore, that these developments have contributed to the creation of a well functioning regional innovation systems - a point implicitly recognised in the perceived need for a bottom-up type initiatives such as the Shannon regional innovation strategy. Starting in 1997, the Shannon Regional Innovation Strategy was Ireland’s only participation in the EU Regional Innovation Strategy programme. In Shannon, a period of analysis and strategy formulation during 1997-1998, was followed by the execution of an action plan covering the 1999-2002 period. This action plan had a number of key features. First, it was led strongly by staff from Shannon Development, the regional development agency for the region. This provided the initiative with a strong strategic focus and ensured the consistency of the measures being adopted with other regional development initiatives. Second, a key objective of the Shannon RIS was to strengthen collaboration between elements of the regional innovation system - particularly actors in the knowledge generation and knowledge application sub-systems (Callanan 2001). Third, the initiative involved physical developments designed to create foci for technological development (e.g. incubators, technology parks) in smaller towns linked wherever possible to thirdlevel educational establishments. In the Shannon region of Ireland at least what has emerged therefore is a largely complementary combination of top-down and bottom-up policy initiatives aimed to boost the region’s innovation capability. The national contribution - the top-down - has been to ensure that the Shannon region gets a reasonable share of inward investment, and hence technology transfer, and to support developments in educational infrastructure. The bottom-up initiative has been to work with local firms and public organisations to strengthen local co-operation and try to encourage technology sharing and diffusion. Ireland’s national industrial policy of innovation by invitation - coupled with the type of dispersal initiatives described earlier - has undoubtedly benefited the Shannon region and other areas in the South and West of Ireland over the last two decades. The existence of Shannon Development has also helped considerably in trying to develop positive measures to embed inward investment companies within the regional economy. Much less clear is whether - despite these advantages - the Shannon region has yet developed any very organic regional innovation system. This could just be a matter of time, but it may also suggest the difficulty of trying to develop a regional innovation system whose leading companies depend strongly on technology developed elsewhere and only weakly on the local knowledge base.

Innovation Policy and Spatial Disparities 307

17.5 Innovation from Invention - Regional Dimensions of Innovation Policy in Israel The key drivers of innovation in Israel are very different to those in Ireland. In Israel, the national search for technological and military independence dates right back to the foundation of the State, a factor reflected in large-scale government investment in both military and civilian R&D and higher level education. As early as the 1960s, for example, the Israeli government was supporting the development of Science Parks at the universities (e.g. the Kiryat Weizmann Science Park in 1967 at Rehovot; Felsenstein 1997) and giving R&D grants to individual firms (Teubal 1993). During the 1970s, as the Israeli economy became more open to trade, bi-national R&D funds were established and, albeit relatively unsuccessful, attempts were made to attract inward investment, particularly to more peripheral areas (Shefer and Bar-El 1993). The 1980s in Israel were marked by the continued development of the civilian electronics sector, macro-economic crisis and radical changes in the political complexion of the State. Previous “state regulated capitalism” had an implicit “bias associated with deep antagonism, or even hostility towards small business-owners’ and entrepreneurship. With changes in the political scene and a shift towards more free-market economic policies, a gradual change in attitude in favour of the small business sector occurred” (Feitelson 2001). Geopolitical changes in the 1980s and early-1990s reinforced this liberalising effect and released much of the human capital on which Israeli entrepreneurship of more recent years was based. Two factors were particularly important. First, the cancellation in 1987 of the Lavi fighter project, the end of the Cold War and the easing of the geo-political situation in the Middle East reduced both export and domestic demand for military hardware and released substantial amounts of highly skilled labour into the Israeli labour market. Second, post-1989, mass immigration to Israel from the FSU added nearly a million to the Israeli population and vastly increased the nation’s endowment of human capital. Throughout this period (i.e. until the early 1990s) the regional dimension of policies which might have influenced innovation capability in Israel were relatively limited. Higher rates of investment support were available in less developed areas and inward investors locating in more peripheral areas were offered a range of grants and tax incentives. In addition, while R&D support was available nationally, small grant premiums were available for companies locating R&D in more peripheral regions (Roper and Frenkel 2000). This R&D grant support was essentially spatially “neutral” and played a key role in the development of research-based start-up companies. A typical development path, for example, was that a company might spend two years in one of the networks of business incubators established in the early-1990s before receiving an R&D grant to further develop its products. The spatial premiums in investment and R&D grants in Israel during the 1990s, together with the development of national networks of business incubators and small business advice centres, were clearly intended to stimulate high-tech

308 Stephen Roper

development in Israel’s less developed regions. The predominantly free-market dynamic of innovation in Israel during the 1990s, however, was undoubtedly working against this policy aspiration with a strong tendency towards concentration in the central Tel Aviv conurbation. This tension between the free market tendency towards centralisation in Israel, and the policy aspiration of distributed development, was a consequence both of the technological dynamic of Israeli high-tech industry as well as the organisational and market context in which development took place. In technological terms, the tension arose because of the strongly knowledge intensive markets in which Israeli high-tech companies are concentrated (Roper and Frenkel 2000). For these companies, the advantages of agglomeration are likely to be stronger than that for firms engaged in less knowledge-intensive markets and the competitive penalties imposed by more peripheral locations are likely to be higher. Shefer et al. (2003), for example, demonstrate the positive impact of metropolitan locations on innovation in Israeli high tech companies and the lack of any such effect in Ireland. Another element of the strong centralising tendency of innovative industry in Israel derives from the dynamic of the development process itself, i.e. local entrepreneurship capitalising on technologies emerging from the military complex and the universities and inward flows of knowledge seeking capital (Cooke et al. 2001). First, there is a strong concentration of university and other public and military research centres in the Tel Aviv conurbation. This contributes to a knowledge-rich environment for high-tech development and ensures a supply of research trained staff. Second, Tel Aviv is the centre of the Israeli banking, finance and venture capital industries, and has strongly developed links to external financial centres and resources, particularly in the US. Thirdly, firms in Tel Aviv probably enjoy “cluster” based advantages due to a high concentration of other high-tech firms which might act as customers, suppliers, partners or sources of information or skilled manpower (Roper and Grimes 2005). The consequence of these strong centralising tendencies in Israel, which have far outweighed the advantages of regional grant premiums, has been a marked concentration of innovative high-tech industry in the Tel Aviv area so that by 2000, the greater Tel Aviv area was said to contain 86 percent of high-tech firms in Israel (see the discussion in Kipnis 2001).

17.6 Discussion Consideration of the cases of Israel and Ireland provides some insight into the potential costs and benefits of regional innovation policy in small countries. In particular, the two examples clearly illustrate the tension between the centralising tendency which results from agglomeration advantages and social and political aspirations towards more spatially distributed development. This raises fundamental questions about the desirability of using innovation policy to eliminate spatial disparities in the presence of significant agglomeration economies.

Innovation Policy and Spatial Disparities 309

In Ireland, for example, the innovation dynamic has been dominated largely by US inward investment in production facilities in the “mass market” phase of the high-tech value chain (Roper and Frenkel 2000). The evidence suggests that innovation in these firms derives no particular agglomeration benefits from a central location, and therefore suffers no competitive penalty from Ireland’s policy of dispersal (Roper 2000). In Israel, on the other hand, high-tech production is concentrated in more knowledge intensive elements of the global high-tech value chain (Roper and Frenkel 2000) and agglomeration advantages are significantly more important (Shefer et al. 2003). The implication is that, given the type of high-tech activity in each country, a policy of dispersal would be likely to be more costly in terms of national productivity in Israel than in Ireland. This distinction is also strongly suggestive of the trade-offs between national productivity and regional disparities discussed in Fujita and Thisse (2001) in the context of the economic and budgetary relationships between core and peripheral regions. In their terms, where agglomeration advantages are weak, as in Ireland, national productivity will be maximised by a policy of dispersal which minimises logistic and environmental costs. This will also have the social and political advantage of tending to equalise productivity between “core” and “peripheral” regions. For Israel, where agglomeration advantages are stronger and bounded within the conurbation (e.g. Shefer and Frenkel 1998), and therefore the productivity penalties for spatially distributing economic activity are higher, a wholly different scenario is suggested. Here, national productivity will be maximised by a high level of concentration which maximises the benefits of agglomeration albeit at the cost of higher logistic and environmental costs. Low productivity in peripheral regions would then be compensated by fiscal transfers; an economically efficient but socially and politically more difficult case to argue. In purely economic terms then the policy implications for Ireland and Israel are straightforward. In Ireland, agglomeration economies are weak and dispersion has no productivity cost and may have advantages in reducing logistic and congestion costs; for Israel any policy of dispersal is likely to reduce national levels of innovation and productivity albeit with potential savings in logistic and environmental costs. More generally, the Irish and Israeli cases emphasise that the optimal spatial distribution of innovation activity even in a small country will depend strongly on the nature of its industrial activity. The more research or knowledge-intensive the industry, the more likely it is that dispersion will have costs in terms of innovation and productivity with direct implications for the costs and benefits of regional innovation policy. Other issues arise for small countries in particular in terms of developing effective innovation strategy. For example, while large countries are able to support an R&D and innovation infrastructure across a range of industries and market segments, small countries are often forced to specialise. Finland is perhaps the clearest example here with its strong focus on mobile telephony, a sector with Nokia at its heart. Ireland has - per force - adopted a similar sectoral focus with a strong emphasis on the manufacture of computers, computer components and more recently software. Potentially, this type of specialisation strategy creates internal (Marshallian) economies but is clearly vulnerable to external market

310 Stephen Roper

trends, a particularly serious threat if a nation’s dominant urban centre is dependent on a specific sector. Other approaches are possible, however, as the case of Israel demonstrates, with the creation of a diverse high-tech cluster concentrated in the R&D phase of the global high-tech value chain (Roper and Grimes 2005). This creates the potential for cross-sectoral (Jacobs-type) economies and potentially reduces vulnerability to external market trends. This type of strategy, however, with a strong R&D focus, is likely to reinforce rather than offset naturally occurring agglomeration forces and contribute to increasing urbanisation. In either scenario, the importance of national industrial strategy as the context within which regional innovation initiatives are developed is clear. This is not, of course, to say that regional innovation initiatives are of no value. As the Shannon RIS demonstrates it is possible to shape bottom-up regional initiatives to complement whatever national policy is appropriate, and the evidence does suggest that strengthening local networks and promoting local technology transfer can still make a valuable contribution to embedding firms and increasing levels of innovative activity. The bottom line, however, is that in a world in which knowledge based competition is increasingly important, making a success of this type of local or regional strategy is going to become ever more difficult.

References Acs ZJ, Audretsch DB, Feldman MP (1994) R&D spillovers and recipient firm size. Review of Economics and Statistics 76:336-340 Acs Z (ed) (2000) Regional innovation, knowledge and global change. London, Pinter Archambault E (2001) Methods for using patents in cross-country comparisons. OST Observatoire des Sciences and Technologies, Montreal Andréosso-O’Callaghan B (2001) Territory, research and technology linkages - is the Shannon region a propitious local system of innovation? Entrepreneurship and Regional Development 12:69-87 Audretsch DB (1998) Agglomeration and the location of innovative activity. Oxford Review of Economic Policy 14(2):18-29 Audretsch DB, Feldman MP (1996) R&D spillovers and the geography of innovation and production. American Economic Review 86:630-640 Best M (2000) The capabilities and innovation perspective: the way ahead in Northern Ireland. Research Monograph 8, December. Northern Ireland Economic Council, Belfast Beugelsdijck S, Cornet M (2001) How far do they reach? The localisation of industrial and academic spillovers in the Netherlands. Centre discussion paper 2001-47 Braczyk H-J, Cooke P, Heidenreich M (1998) Regional innovation systems. UCL Press, London Callanan B (2001) Innovation strategy in the Shannon region. Paper presented at 31st European Small Business Seminar, Dublin, Ireland Cooke PN, Morgan K (1998) The associational economy: firms, regions, and innovation. Oxford University Press Cooke P, Davies C, Wilson R (2001) Innovation advantages of cities: from knowledge to equity in five basic steps. European Planning Studies 10(2):234-250

Innovation Policy and Spatial Disparities 311 Crepon B, Duguet E, Mairesse J (1998) Research, innovation and productivity: an econometric analysis at the firm level. Economics of Innovation and New Technology 7:115-158 Crone M (2002) Regional competitiveness - regional co-operation and cluster development - the Irish software industry. Paper presented at the European Regional Science Association Conference, Dortmund, August Davelaar E, Nijkamp P (1989) Spatial dispersion of technological innovation: a case study for the Netherlands by means of partial least squares. Journal of Regional Science 29(3):325-346. Davelaar E, Nijkamp P (1992) Operational models on industrial innovation and spatial development: a case study for the Netherlands. Journal of Scientific and Industrial Research 51: 253-284 De la Mothe J, Paquet G (1998) Local and Regional Systems of Innovation. Kluwer, Dordrecht Dosi G (1988) Sources, procedures and microeconomic effects of innovation. Journal of Economic Literature, XXVI(3):1120-1171 Drudy PJ (1991) The regional impact of overseas industry. In: Foley A, McAleese D, Gill, McMillan, Overseas industry in Ireland, Dublin EU (1998) Regional innovation systems: designing for the future - REGIS. Final report of the REGIS project, Targeted Socio-Economic Research (TSER) Programme (coordinator: Cooke P) European Commission DG XII Feitelson E (2001) Malicious siting or unrecognised processes? A spatio-temporal analysis of environmental conflicts in Tel-Aviv. Urban Studies 38(7):1143-1160 Feldman M (1994) The geography of innovation. Kluwer Academic Publisher, Dordrecht, the Netherlands. Felsenstein D (1997) The making of a high technology node: foreign-owned companies in Israeli high technology. Regional Studies 31(4):367-380 Fernandez E, Junquera B, Vazquez CJ (1996) Government support for R&D: the Spanish case. Technovation 16(2):59-66 Fujita M, Thisse JF (2002) Economics of agglomeration: Cities, industrial location, and regional growth. Cambridge University Press Griliches Z (1995) R&D and productivity: econometric results and measurement issues. In: Stoneman P (ed) Handbook of the economics of innovation and technological change, Blackwell, Oxford Henderson J, Dicken P, Hess M, Coe N, Yeung HYC (2002) Global production networks and the analysis of economic development. Review of International Political Economy 9(3):436-464 Hewitt-Dundas N, Andréosso-O’Callaghan B, Crone M, Murray J, Roper S (2002) Learning from the best: knowledge transfers from multi-national plants in Ireland: a north-south comparison. NIERC/Euro-Asia Centre, University of Limerick Jaffe AB (1989) Real effects of academic research. American Economic Review 79:957870 Kipnis BA (1998) Technology and industrial policy for a metropolis at the threshold of the global economy: the case of Haifa, Israel. Urban Studies 35(4):649-662 Kipnis BA (2001) Tel Aviv, Israel - a world city in evolution: urban development at a dead end of the global economy. Globalisation and World Cities Study Group, Research Bulletin 57, http://www.lboro.ac.uk/departments/gy/gawc/rb/rb57.html Lööf H, Heshmati A (2000) Knowledge capital and performance heterogeneity: a firm level innovation study. SSE/EFI Working Paper No 387, Stockholm School of Economics Lööf H, Heshmati A (2001) On the relationship between innovation and performance: a sensitivity analysis. SSE/EFI Working Paper No 446, Stockholm School of Economics

312 Stephen Roper Luukkonen T (2000) Additionality of EU framework programmes. Research Policy 29:711724 Meyler A, Strobl E (1997) Regional employment performance of Irish industry 1972-1996: job flow analysis. Trinity Economic Papers, Technical Paper No 9, June, Dublin O’Farrell P (1980) Multinational companies and regional development: the Irish experience. Regional Studies 14(2) Roper S (2001) Innovation, networks and plant location; evidence for Ireland. Regional Studies 35(3):215-228 Roper S, Frenkel A (2000) Different paths to success - the growth of the electronics sector in Ireland and Israel. Environment and Planning C 18(6):651-666 Roper S, Grimes S (2005) Wireless Valley, Silicon Wadi and Digital Island - Helsinki, Tel Aviv and Dublin in the ICT boom. Geoforum (forthcoming) Ruane F, Gorg H (1997) Reflections on Irish industrial policy towards foreign direct investment. Trinity Economic Papers Series, 97/3, Dublin Shannon Development (1999) Regional innovation strategy for the Shannon region. Limerick Shefer D, Bar-El E (1993) High technology industries as a vehicle for growth in Israel’s peripheral regions. Environment and Planning C 11:245-261 Shefer D, Frenkel A (1998) Local milieu and innovations: some empirical results. Annals of Regional Science 32:185-200 Shefer D, Frenkel A, Shefer S, Roper S (2003) Public policy, locational choice and the innovation capability of high tech firms: a comparison between Israel and Ireland. Papers in Regional Science 82(2):203-221 Teubal M (1993) The innovation system of Israel: description performance and outstanding issues. In Nelson RR (ed) National innovation systems: a comparative analysis. Oxford University Press, 1993 Wrynn J (1997) Foreign direct investment to a peripheral country - the case of Ireland. In: Fynes B, Ennis S (eds) Competing from the periphery, Oaktree Press, Dublin Yearly S (1995) From one dependency to another: the political economy of science policy in the Irish Republic in the second half of the twentieth century. Science, Technology and Human Values 20(2):171-196

Figures Fig. 2.1. Fig. 4.1. Fig. 4.2. Fig. 5.1. Fig. 5.2. Fig. 5.3. Fig. 5.4. Fig. 5.5. Fig. 5.6. Fig. 6.1. Fig. 6.2. Fig. 6.3. Fig. 6.4. Fig. 6.5. Fig. 6.6. Fig. 6.7. Fig. 6.8. Fig. 6.9. Fig. 6.10. Fig. 6.11. Fig. 6.12. Fig. 6.13. Fig. 7.1.

Fig. 7.2.

The role of mediating factors Results of permutation tests for selected inequality measures - CV, Gini (U), AT and TE(0) (A) and WI (B) Results of permutation tests for selected inequality measures - WI and Gini (U) (A) and CC (B) Alternative mapping techniques of inter-urban income disparities Alternative methods of coordinate transformation Classification of localities according to relative levels of average income in 1999 (“concentric circle” method) Classification of localities according to relative levels of average income in 1999 (proportional increment method) Localities with different rates of income change between 1991 and 1999 (proportional increment method) Changes in average incomes between 1991 and 1999 across urban localities in Israel, per cent Regional and inter-regional disparities in Belgium: employment rate at the district level in 2001 Dynamic shift-share components, Flemish region, 1995-2001 Dynamic shift-share components, Walloon region, 1995-2001 Dynamic shift-share components, Brussels, 1995-2001 Virtual and actual employment and the ratio of actual to virtual employment, Flemish region, 1995-2001 Virtual and actual employment and the ratio of actual to virtual employment, Walloon region, 1995-2001 Virtual and actual employment and the ratio of actual to virtual employment, Brussels, 1995-2001 Ratio of actual to virtual employment, Flemish provinces, 1995-2001 Ratio of actual to virtual employment, Walloon provinces, 1995-2001 Ratio of actual to virtual employment, by industry, Flemish region, average 1995-2001 Ratio of actual to virtual employment, by industry, Walloon region, average 1995-2001 Ratio of actual to virtual employment, by industry, Brussels, average 1995-2001 Administrative regions and districts of Belgium GDP growth rate, unemployment rate (UE) and public sector surplus (or deficit) as a percent of GDP (PSS) in Finland, 1988-2001 Main regions in Finland

20 58 59 64 71 73 74 75 77 86 93 95 95 99 100 100 101 101 102 102 102 105

111 114

314 Figures

Fig. 7.3. Fig. 7.4. Fig. 7.5. Fig. 7.6. Fig. 7.7. Fig. 7.8. Fig. 7.9. Fig. 8.1. Fig. 9.1. Fig. 9.2. Fig. 9.3. Fig. 9.4. Fig. 9.5. Fig. 9.6. Fig. 9.7. Fig. 9.8.

Fig. 9.9. Fig. 9.10. Fig. 10.1. Fig. 10.2. Fig. 10.3. Fig. 11.1. Fig. 11.2. Fig. 11.3. Fig. 11.4.

Real per capita disposable income in main regions in 19712000 Per capita factor income and disposable income in main regions in 1971-2000, index Gini coefficients based on three income concepts, 1971-2000 Gini coefficients by income variables and main regions in 1971-2000 The poor population by region 1971-2000 Gini elasticities of income components by main regions in 1990s Trade off between unemployment rate and Gini coefficient by main regions Regional authority areas Level of unemployment in some European countries: 19882000 Employment growth in the United States, the Netherlands and European Union Regional differences in unemployment within select European countries, 1998 Registered unemployment 1952-2001 (percent of the labour force in the Netherlands) The regional dispersion of registered unemployment per COROP region Registered unemployment per COROP region, 1988 Registered unemployment per COROP region, 1998 Development of the number of non-working job-seekers in the Netherlands by socio-economic group, March 1993October 1999 Regional differences in long term unemployment (> 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]