With the publication of Geography and Trade (MIT Press, 1991), Paul Krugman refocused attention on the role of geography in economics. While geography is now seen as important in subjects ranging from growth and macroeconomics to international trade, European integration, labour markets and technological change. The economics of geography formed the main focus of a joint conference with the Consorcio de la Zona Franca de Vigo, `The Location of Economic Activity: New Theories and Evidence', held in Vigo, Spain, on 17/19 December. The conference was organized by Paul Krugman, Professor of International Economics at the Massachusetts Institute of Technology and Research Fellow in CEPR's International Macroeconomics and International Trade programmes, and Anthony Venables, Professor of International Economics at the London School of Economics and Political Science and Research Fellow in CEPR's International Trade programme. Financial support and assistance with organization from the Consorcio de la Zona Franca de Vigo are gratefully acknowledged. A volume of papers presented at the conference is now available from CEPR.*

In his opening presentation, `Complexity and Emergent Structure in the International Economy', Paul Krugman focused on the possible contribution to the modelling of economic geography and international trade of `complex' systems, in which there are multiple equilibria and `bifurcations' in the behaviour of the system. Researchers in the physical sciences have developed non-linear, dynamic, numerical simulation techniques to model phenomena for which theory cannot provide analytical solutions. Their application to economics may prove useful in modelling the `emergence' of particular locational distributions of cities after urbanization or industries after national markets' economic integration. Krugman outlined a model of international trade in which complexity increases as trade barriers fall, in the sense that the structure of equilibria becomes richer, although theory cannot predict which equilibrium will be attained. Finally, he noted that welfare need not rise with the complexity of an economic system: for example, while Mexico remained a closed economy, industry concentrated around Mexico City to an extent that clearly reduced economic welfare, but trade liberalization could lead to a dramatic decentralization together with a reduction in complexity.

Both Richard Baldwin (Graduate Institute of International Studies, Geneva, and CEPR) and Jürgen von Hagen (Universität Mannheim and CEPR) suggested that empirical and, in particular, econometric applications of complexity will pose major challenges.

In `Expectations-Driven Spatial Fluctuation', Jordi Galí (Columbia University and CEPR) argued that characterizations of business cycles that have relied on the observation and analysis of national aggregate data may conceal important differences in cyclical performance across regions and cities. To fill this void, Galí developed a dynamic spatial model with monopolistic competition, increasing returns and labour mobility. He used this to show that rational expectations equilibria are characterized by fluctuations in the spatial allocation of resources, even when technology and preference shocks are absent. Such fluctuations result from the interaction of forward-looking decisions with agglomeration/congestion economies derived from the assumptions about market structure and preferences. Welfare losses result from unnecessary randomness of equilibrium allocations.

In `The Role of Market Size in Club Formation', Alessandra Casella (Columbia University and CEPR) noted that there is now a large literature examining how institutions shape markets, and she argued that the reverse may also be true: the size of the market may help shape institutions. She developed a model to investigate the links between market size and the optimal number of jurisdictions (or clubs).

Jürgen von Hagen pointed out that members incur the costs of club formation with every change in market size. In European terms, this suggests that a one-time expansion of the Community would be achieved at lower cost than a series of smaller incremental increases over a prolonged period.

In `Interactions Between International Migration and International Trade: Positive and Normative Aspects', with Efraim Sadka, Assaf Razin (Tel Aviv University and CEPR) used a standard two-country trade model to show that differences in their relative endowments of capital and labour will induce commodity trade and labour (or capital) mobility. These are perfect substitutes if technologies are identical, but they may be complements when the countries' technologies differ. East-West trade need not equalize wages and may even widen the wage gap, so policy should promote infrastructural and direct investment to raise East European productivity and alleviate migration pressures. Razin also distinguished labour mobility, which allows `guest-workers' simply to export their labour services, from migration, when migrants also change their national communities, which has major normative implications for assessments of overall social welfare in both host and sending countries.

Riccardo Faini (Università degli Studi di Brescia, Innocenzo Gasparini Institute for Economic Research, Milan, and CEPR) presented evidence that migrants to the US, France and Germany tend to be employed in import-competing sectors, which is consistent with their working in sectors in which the developed countries are at a comparative disadvantage; trade liberalization may thus be expected to reduce migration incentives.

In `Industrial Localization: An Empirical Test of the Marshall/Krugman Hypothesis', Jürgen von Hagen and George Hammond (Indiana University) tested the `localization hypothesis' introduced by Alfred Marshall and developed more recently by Paul Krugman: that localization economies lead to geographical clustering and hence that changes in employment in localized industries should be more closely correlated within than across regions. They used employment data for three two-digit industrial sectors and four US cities to show that asset-sharing produces significant localization economies, while the economies from labour pooling tend to play a more important role in rapidly growing labour markets.

Hideki Yamawaki (Université Catholique de Louvain) suggested extending their empirical work to take account of differences within industries; suppressing such differences through the use of broadly defined industries was inconsistent with the authors' theoretical model. James Markusen (University of Colorado, Boulder) suggested separating purely regional from purely sectoral effects.

In `Trade Unions, Fiscal Policy and Regional Development', Riccardo Faini used a two-region model to focus on the role of unions and show that inter-regional labour mobility may provide unions with an incentive to set wages in the poorer region at a higher level than otherwise. Unions' concern about the implications of a high-wage policy for employment in the less-developed region will also reduce if they anticipate that a central government will provide compensation through public transfers. Faini noted the particular relevance of his model to Italy and Germany.

In `Equilibrium Locations of Vertically Linked Industries', Anthony Venables considered locational choices of firms in imperfectly competitive upstream and downstream industries in which firms face increasing returns. The level of transport costs determined whether there is a single equilibrium (with production diversified between locations), or multiple equilibria (some of which involve agglomeration at a single location). Venables showed that the forces for agglomeration were greatest at intermediate levels of transport costs. Reducing such costs leads initially to agglomeration and consequent divergence of economic structures, but further reductions then cause the industries to operate in both locations, which leads to convergence of both structure and income.

Richard Baldwin pointed out that Venables's theory is essentially a European version of Krugman's theory of location developed for the US, in which labour moves to take advantage of the available employment opportunities. In Europe, capital is the relatively mobile factor and firms therefore move to regions endowed with the types of labour they require.

In contrast with the substantial recent literature on economic integration, trade and the location of production, there has been little work on geography of innovation. In `The Geography of Innovation and Production', David Audretsch (Wissenschaftszentrum Berlin für Sozialforschung and CEPR) and Maryann Feldman (Goucher College) first examined and controlled for the geographic concentration of production across industries using a data base which also identified innovative activity in individual states and specific industries in the US. They then used this to show that the propensity for innovative activity to cluster spatially is also greater in industries in which creation of knowledge spillovers is important. Production tends to be concentrated geographically in industries for which university research provides important inputs of such knowledge, but most clustering of innovative activity is nevertheless due to knowledge spillovers rather than concentration of production.

Paul Krugman noted that information may be transmitted almost costlessly over space, while the transmission of knowledge and especially tacit knowledge over long distances is much more difficult: proximity therefore plays an important role in such knowledge spillovers. James Markusen stressed the policy implications of these findings for public support for university research to encourage innovative activity.

In `Deindustrialisation as a Consequence of Market Integration', Massimo Motta (Universitat Pompeu Fabra, Barcelona, and CEPR) and Jacques-François Thisse (Ecole Nationale des Ponts et Chaussées, Paris, and CEPR) used a standard vertical product differentiation model to study the impact of trade liberalization on two countries with very different income levels. Under the `autarky equilibrium', which also represents the initial conditions of the ensuing trade game, the qualities of goods offered in each country increase with domestic income. Motta and Thisse then studied trade and welfare changes both in the short run (when firms can offer only goods of the qualities produced under autarky) and also in the long run (when they may also modify these qualities but this incurs adjustment costs). They identified conditions under which competition from higher-quality firms in the rich country causes exit by firms in the poor country.

Guillermo de la Dehesa (Banco Pastor and CEPR) suggested extending the model to consider the effects of market segmentation that allows firms to offer goods of varying qualities as well as prices at the same time and also the welfare implications of introducing a non-traded sector.

While many theoretical models have predicted the importance of agglomeration economies, their empirical identification has proved elusive. In `Increasing Returns and the Regional Structure of Wages', Gordon H Hanson (University of Texas, Austin) investigated the existence of such increasing returns by inference from regional differences in the structure of wages and employment for Mexico. The theory of agglomeration and increasing returns predicts that employment will concentrate in industrial centres while relative wages fall as the distance from them increases. Hanson found that Mexico's employment and wage patterns in the 1980s provided considerable support for the `increasing returns' hypothesis. While industry concentrated overwhelmingly in Mexico City under the closed economy, it has relocated to the US border region since liberalization. Mobile factor wages also tend to fall as distance from Mexico City or the US increases, although trade liberalization has tended to compress such regional differentials.

John Grahl (Queen Mary and Westfield College, London) remarked that Hanson's disaggregation by industry would add little precision unless there were in fact substantial intra-regional wage differentials by industry. He also called for a more detailed analysis of labour force participation.

In `Anti-Competitive and Rent-Shifting Aspects of Domestic-Content Provisions in Regional Trade Blocks', joint with Florencio Lopez-de-Silanes and Thomas Rutherford, James Markusen used an applied general equilibrium model to show that the discriminatory nature of domestic-content provisions generates both anti-competitive and rent-shifting effects in oligopolistic industries in which foreign multinationals operating within a regional trade bloc make greater use of imported intermediate goods than domestic firms. The anti-competitive effects of these rules tend to reduce welfare within the bloc, while rent-shifting tends to improve its welfare, provided this is not dissipated by new entry. The anti-competitive effect may be strong enough to raise total industry profits, and the equilibrium demand for domestic inputs may also fall if the scale effect of reduced output outweighs the substitution effect; this is more likely when foreign multinationals can switch from producing within a bloc to exporting to it from outside.

John Grahl pointed out that Japanese firms have responded to European domestic-content restrictions by substituting foreign direct investment for exports to Europe, so new Japanese-owned firms produce the local components such rules require. Hideki Yamawaki reported a similar response by Japanese multinationals to domestic-content provisions in the US. Carlos M Asilis (IMF) suggested extending the analysis to consider a longer-run, dynamic framework which would allow changes in the capital stock. Relaxing the assumption of full employment would also provide a more realistic characterization of the effects of trade liberalization on individual sectors.

In `Geography, Trade Patterns and Economic Policy', Carlos Asilis and Luis A Rivera-Batiz (Universitat Pompeu Fabra, Barcelona) developed a model of inter-regional trade in which firms, consumers and perfectly mobile workers view location as endogenous. Space is important on account of transport costs, market access and distance from pollution-generating industrial centres. They used their model to develop a compensating-differential theory of regional unevenness, which proved to be a generalization of the `gravity' model of trade patterns. They also investigated the geographic basis for industrial and environmental policies and the interactions between location and technological change in transport and other sectors.

Anthony Venables suggested applying the model to investigate the equilibrium number and locations of cities and the parameters that will change this equilibrium configuration.

In `Endogenous Growth in the European Regions', John Grahl and Jonathan Simms (University of Manchester Institute of Science and Technology) noted that studies of variation in economic growth across countries in two very distinct literatures endogenous growth and economic geography have both emphasized the importance of external economies of scale. They considered the determinants of growth rates identified by both approaches for EC regions during 1960-90 to show that both growth processes were closely associated and heavily concentrated in the first half of the sample period. Their results favoured explanations based on `density effects' whereby regions with larger shares of existing economic activity can expect faster growth rather than `institutional endowments'; local strategies to promote growth can only achieve their full potential under favourable macroeconomic conditions.

Massimo Motta disputed the authors' interpretation of a significant education variable as evidence against the approach of the `new growth' literature, to which the modelling of human capital formation is central; he also questioned the significance they attached to unemployment in determining the importance of institutional factors. Jürgen von Hagen suggested modelling institutions and policy endogenously, especially in the long run: the regional development programmes that are now prevalent throughout Europe provide clear examples of endogenous responses to historical experience of growth.

In `Growth, Industrial Mix and Structural Change in U.S. Regions: The Shift to a Service-Based Economy', joint with Therese McGuire, Teresa Garcia-Milà (Universitat Pompeu Fabra, Barcelona) examined linkages between regions' industrial compositions and relative growth rates to investigate whether the US economy's recent structural shift from manufacturing to services has benefited regional employment growth and, in particular, whether it induced knowledge spillovers from services that have supported growth in manufacturing. They found support for this view in productivity and earnings trends, estimations of industry-specific growth equations and input-output analysis. Nevertheless, their econometric analysis indicated that the employment share of business services had no effect on regional growth rates of manufacturing.

George Hammond suggested that knowledge spillovers from services to manufacturing may promote growth to a greater extent than the authors inferred, if the positive impact of services on regional growth reflects the impact of localized knowledge spillovers from education and research.

In `Regional Adjustment and Wage Flexibility in the EC', Filip Abraham (Katholieke Universiteit Leuven) developed a stylized theoretical model to integrate inter-regional migration, spillover effects from national to regional wage determination, and regional agglomeration effects. He used this to evaluate the role of wage flexibility in EC regional labour market adjustment to region-specific and aggregate shocks; his results indicated that regional wages are largely determined by the level of national wages, while region-specific shocks generate only small wage adjustments. This is not a promising picture for lagging regions that face specific structural adjustment problems: if rising regional unemployment does not induce a wage fall in response to adverse shocks, regions risk running permanent high levels of unemployment.
Georges Siotis (Université Libre de Bruxelles) suggested extending the analysis to apply Blanchard and Katz's finding for the US that regional employment shocks can have permanent effects in a European context. With lower labour mobility, `insufficient' wage flexibility and hysteresis in employment losses, public infrastructural investment and regional/industrial policies may help to promote the mobility of firms that is needed to reduce structural unemployment.

In `Externalities and Industrial Development', Vernon Henderson (Brown University) examined the impact of dynamic externalities on the employment growth of industrial sectors for US counties. He found that diversity tends to raise productivity and hence employment in a city's particular concentration of production and export activity. Cities, metropolitan areas and counties are all highly specialized in their manufacturing activities, which presumably reflects the benefits from concentration. Diversification also tends to have more persistent effects on employment growth than increased own-industry specialization, which suggests that a diverse environment can both attract emerging industries and also help to retain established industries within the locality.

Maryann Feldman suggested that variations in the lag structure between increased concentration and subsequent rises in employment across industrial sectors may in part reflect differences in the role of sunk costs across sectors.

In `Heterogeneity, Stratification, and Growth,' Roland Bénabou (MIT and CEPR) examined the effects of economic stratification on inter-regional inequality and growth over time. He found that the geographic segregation of agents of different income and social classes tends to minimize the losses from a given amount of heterogeneity, while their closer integration causes such heterogeneity to disappear more rapidly. Society therefore faces an intertemporal trade-off: social mixing reduces growth in the short run but increases growth in the long run.

Jacques-François Thisse suggested incorporating club theory within Bénabou's analysis to focus on the role of coalitions.

In `Competing Exchanges', Thomas Gehrig (Universität Basel), Konrad Stahl (Universität Mannheim) and Xavier Vives (Institut d'Anàlisi Econòmica, CSIC, Barcelona, and CEPR) investigated the implications of location for financial markets to determine whether large stock markets drive out small ones. If so, European integration would inevitably lead stock market trading to concentrate in London, which currently has Europe's greatest stock exchange turnover, and globalization of the economy would lead to an ultimate concentration of trading activity in New York. Information about firms and the potential returns on their equity is not ubiquitous, however, but concentrated rather in the locations where such firms operate, so trade in fragmented stock exchanges may represent an equilibrium outcome. The numbers of new firm births and the proportions of them surviving beyond their first five years have recently risen dramatically both in the US and in Europe, which should enhance the comparative advantage of local investors as some of these firms grow large enough to require equity financing.

Paul Krugman suggested that transport costs may play a greater role in decentralizing financial markets than the authors' model implicitly assumed.

One of the more striking phenomena experienced throughout developed and developing countries over the last half century has been a rapid degree of urbanization. In `A Monopolistic Competition Model of Hierarchical Urban Systems and Trade', Masahisa Fujita (University of Pennsylvania) and Paul Krugman pointed out that more than 70% of the world population now lives in cities, which are becoming increasingly important both for national economies and for the world economy and trade. They proposed a microeconomic model of endogenous urban agglomeration, based on Chamberlain's monopolistic competition model of firms providing a continuum of differentiated products. They examined this model's ability to yield a variety of urban systems, and investigated its usefulness in the design of effective regional development strategies.

* To order your copy of The Location of Economic Activity: New Theories and Evidence, please send remittance for £20.00/$30.00 to: CEPR Discussion Papers, 90-98 Goswell Road, London EC1V 7RR.