A joint workshop with the Centre for Economic Performance (CEP) on ‘The Spatial Economy: New Developments in Economic Geography’ was held at the London School of Economics (LSE) on 9/10 July 1996. The workshop was organized by Anthony Venables (LSE and CEPR) as part of the Human Capital and Mobility Network on ‘The New Economy of Europe: Market Integration, Regional Convergence, and the Location of Economic Activity’. The main objective of the workshop was to explore recent developments in the theory of spatial agglomeration.

In ‘The Rise and Fall of Economic Agglomerations’, Diego Puga (London School of Economics) starts by noticing that economic activity is geographically less concentrated in the European Union (EU) than in the United States (US), but that income disparities across EU members are much wider than across US states. He develops a stylised theoretical model that can account for these facts. The model highlights the importance of labour mobility (or the lack of it) in determining the strength of agglomeration versus dispersion forces and the associated spatial distribution of income. The author also explores the relationship between industrial agglomeration and trade. Unlike what is usually assumed, he shows that if labour is not mobile the relationship between trade barriers and industrial agglomeration need not be monotonic. Starting from high trading costs, agglomeration may increase and then decrease as trading costs decrease. Thus, a core-periphery pattern of economic activity may appear and then collapse as a result of increased mobility. Jacques Thisse (CORE, Université Catholique de Louvain, CERAS-ENPC and CEPR) praised the author for presenting an integrated framework encompassing many existing models of economic geography and nonetheless offering new results.

A model encompassing many previous analyses was presented by Yossi Hadar (Tel Aviv University) in ‘Homogeneous Products, Transportation Costs and their Effects’. Along with the traditional industrial and agricultural goods, he also considers housing to be a key determinant of location. In particular, it is shown that the equilibrium with two equal cities remains locally stable even when the transportation cost of the differentiated products is low. Furthermore when the transportation cost of the homogenous products is small and of the same magnitude as the transportation cost of the differentiated ones, and the individuals do not value variety much, then the equilibrium with two equal cities is unique. Otherwise, multiple equilibria are present. Johan Torstensson (Lund University, University of Nottingham and CEPR) questioned the robustness of the results since they rely mainly on simulations. According to Anthony Venables some of the results are corroborated by other papers.

In a joint work with Gianni Carbonaro (European Investment Bank), ‘The Sensitivity of Regional Growth Convergence to Spatial Factors’, Paul Cheshire (LSE) argues that conventional tests for growth rate convergence constitute neither adequate tests of growth theory nor of convergence. The paper tries to account for growth differentials as completely as is feasible. Given the available data, a growth model for the European regions is tested, explicitly introducing spatial factors in regional growth processes. The specification allows for agglomeration economies in R&D. Also included in the model and found highly significant are an adjustment mechanism when localized growth differentials emerge, a measure of the regional impact of integration and measures of localized agglomeration economies. All these are found to contribute significantly to explaining the growth performance of European regions during the 1980s. Barry McCormick (Southampton University) agreed with the necessity for careful use of the data and the emphasis on R&D, but he called for a more intensive use of educational variables.

Paul Krugman (Stanford University and CEPR) in ‘Confronting the Mystery of Urban Hierarchy’ recalled that the distribution of city size in a given country was very well described by a Pareto distribution of exponent minus one. He highlights that for the United States the relationship is very strong over more than one century. Unfortunately, existing theories in Urban Economics and Economic Geography do not give a satisfactory answer to this problem. It seems that the distribution of city size is best predicted by a random growth model. The paper then tries to give some micro-foundations to the random growth model, building on uneven transport costs in a network of cities. Finally the author suggested that European integration could lead to a distribution of city size similar to that in the United States. Simon Anderson (University of Virginia) wondered if there was any parallel between the problem of city size distribution and that of wealth distribution in a population.

Jacques Thisse addressed the issue of Marshallian districts, which he defined as locales that accommodate a large number of small firms producing similar goods to be exported and benefit from the accumulation of know-how associated with workers residing there. The paper ‘Learning-by-Doing and the Formation of Marshallian Industrial Districts’ is joint with Carlo Carraro (GRETA, Università di Venice and CEPR) and Antoine Soubeyran (GREQAM, Université de la Méditerranée). Assuming learning-by-doing at the district scale, the authors study the growth of Marshallian districts. This analysis of skill accumulation in terms of social process gives an important role to history which strongly affects the dynamics of the economy. In particular, when new firms set up in a locale, they tighten competition on the corresponding labour market, thus leading to a wage rise that reduces the incentives for firms to locate in the most efficient district.

Paul David (Stanford University and All Souls College, Oxford) presented ‘Marshallian Externalities and the Emergence and Spatial Stability of Technological Enclaves’, a joint paper with Dominique Foray (Université de Paris-Dauphine) and Jean-Michel Dalle (Ecole Normale Supérieure de Cachan). They analyse the long-run dynamic behaviour of a system in which firms’ choices among alternative production methods are influenced by both firm-specific random shocks and ‘Marshallian industrial neighbourhood’ effects. For instance, local factor market externalities tend to lower the relative marginal costs of those inputs that are used most extensively in the immediate locale. When two technologies are available, there is no ubiquitous diffusion of only one technology. On the contrary, it can be show, using the applied theory of Markov random fields, that the system exhibits a spatially localized form of ‘technological dualism’ in which at least two technological enclaves emerge and undergo path-dependant evolution. Richard Freeman (LSE) stressed that the model had some strong empirical implications.

‘Labour Specialization and City Formation’ was presented by Yves Zenou (CORE, Université Catholique de Louvain), written jointly with Marcus Berliant (Washington University). The authors explore the formation of cities through labour specialization, gains to trade, transportation costs and commuting costs of consumers. They use a very general setting which allows for multi-dimensional location space and firms using different types of labour to produce different outputs. In the models, all locations and prices are endogenous. First firms play a Nash location game anticipating the location of consumers. The equilibrium may or may not exist depending on the relative number of products and inputs. They also show that both welfare theorems fail in the framework. Alasdair Smith (University of Sussex and CEPR) wondered about the search for equilibria in urban economics.

A theory of the determination of city size over time was proposed by Gilles Duranton (LSE) in his paper ‘Early Urbanization and After’. In his analysis, increasing returns generated by the specialization of labour favour urbanization, whereas the transport cost of agricultural goods slowed city size expansion before the industrial revolution. Considering the existence of some urban institutions (in particular guilds), the equilibrium size of cities is derived and found to be efficient. Within this framework, the author can study the effects of urban domination, the emergence of primate cities as well as the formation of regional central place hierarchies. The transition between early and modern urbanization can also be analysed. Paul David argued that the historical reality was much richer and questioned some of the assumptions. Masahisa Fujita (Kyoto University) highlighted that the paper belongs to the New Economic Geography but also to the older literature of Urban Economics.

In a joint work with Philippe Martin (GIIS, Geneva, and CEPR) ‘Growing Locations: Industry Location in Models of Endogenous Growth’, Gianmarco Ottaviano (Università di Bologna, Università Bocconi and CEPR) integrates the new economic geography framework with the endogenous growth literature. He shows that with global spillovers in R&D, a high growth rate and a high level of transaction costs are associated with relocation of the newly created firms to the South (the location with a low initial human capital). With local spillovers in R&D, this activity will be agglomerated in the North and the rate of innovation will increase with the concentration of firms in the North. This, in turn, implies that a decrease of transaction costs, through for example trade integration, will increase the growth rate because it leads to a higher industrial concentration of firms where the R&D is located. The authors also show that industrial concentration improves welfare only for low enough transaction costs and high enough spillovers. Anthony Venables agreed with the need of precise modelling and specifications for spillovers.