Do physical location and geographical spillovers matter in the dynamics of regional income? Is interaction across regions important for economic performance? Is geographical proximity needed for interaction? The recent empirical literature on regional convergence/divergence finds evidence of increasing inequality, polarization and formation of convergence clubs. The purpose of a CEPR workshop on ‘Location and Regional Divergence/Convergence’ was to advance both theoretical and empirical understanding of the forces driving these dynamics. The workshop, organized by Jacques Thisse (CORE, Université Catholique de Louvain-La-Neuve, CERAS-ENPC, Paris, and CEPR) and Ana Lamo (Universidad di Alicante) was held at CORE, Université Catholique de Louvain-La-Neuve, on 25/26 October 1996. It was supported by a grant from the European Commission under its Human Capital and Mobility Programme.

Diego Puga (Centre for Economic Performance, London School of Economics) analysed the impact of regional integration on the spatial agglomeration of economic activity in his paper entitled ‘The Rise and Fall of Economic Agglomerations’. His framework allows for intermediate linkages in the industrial sector, and for both intersectoral and interregional mobility of the labour force. Earlier works by Krugman and by Krugman and Venables are special cases of Puga’s more general framework. High transport costs lead to an agglomerated economic structure. The higher the degree of interregional mobility of workers, the faster do decreasing transport costs lead to a concentrated economic structure. This result offers a possible explanation for the higher regional concentration of industry observed in the United States than in Europe. Moreover, the results of the paper suggest that the ability of peripheral regions to catch up depends on the degree of economic integration and on the flexibility of wages in response to changes in industrial employment. Konrad Stahl (Universität Mannheim and CEPR) pointed out that the labour market is a key element in the emergence of a geographical economic structure and that the framework presented by Puga was a first step in linking the functioning of the labour market to a spatial model of economic activity. He also suggested that population density might be important for agglomeration and should be included in further research.

‘Growth and Agglomeration’, presented by Philippe Martin (Institut Universitaire des Hautes Études Internationales, Geneva, and CEPR) and written jointly with Gianmarco Ottaviano (Università di Bologna, Università Bocconi, Milano, and CEPR), developed a model to explain the observed positive correlation between growth and urban agglomeration across countries. Cities play the crucial role for both the geography of innovation and the geography of production. The authors model growth and agglomeration as mutually reinforcing processes. Industrial agglomeration in one location spurs growth because it reduces the cost of innovation in that location through pecuniary externalities due to transportation costs. As the sector at the origin of innovation expands, new firms tend to locate close to this sector and growth fosters agglomeration. The model can be interpreted as illustrating one mechanism behind the emergence of cities seen as centres for production and innovation. The approach integrates the main features of two areas of research: it embodies circular cumulative causation – the distinctive feature of models of the new economic geography – and endogenous innovation as a source of growth. For Konrad Stahl the existence of a continuum of equilibria between the two extremes of total agglomeration and no concentration was a highly interesting and realistic feature of the model.

In his paper on ‘Regional Convergence and Transport Costs’, Tobias Adrian (London School of Economics) analysed the impact of transport costs in two sectors on the location of economic activity. He introduced agricultural transport costs in the Core-Periphery model proposed by Krugman (1991). Agricultural transport costs alter the relative impact of backward and forward linkages with the general result that the regional agglomeration of industry becomes less likely. For a given ratio of agricultural to industrial transport costs, regional concentration of the industrial sector can be excluded for a wide range of parameter values. For a given level of agricultural transport costs, a high degree of economic integration – with strongly decreasing industrial transport costs – always leads to instability of the agglomerated industrial structure. Hence, the regional economic structure crucially depends on the degree of economic integration. The range of parameters in which concentration occurs is decreasing in the size of agricultural transport costs. Furthermore, in the case of a large industrial sector and high scale-economies, an agglomerated economic structure is stable for high levels of transport costs but not for low ones. This result appears similar to Helpman’s 1996 result, in which the agricultural sector is replaced by a (non-tradable) housing sector.

In ‘Transport Investment, Infrastructure and Regional Convergence’, Roger Vickerman (University of Kent) examined the microeconomics of infrastructure. Aggregate models tend to find unreasonably high rates of return to infrastructure. In surveys conducted at the company level, however, infrastructure appears to be a heterogeneous phenomenon. It turns out that small projects – such as a new junction linking two streets – can yield extremely high returns, whereas costly projects can be inefficient means of fostering regional convergence. For example, the Eurostar rail project is expected to yield high returns at an aggregate level, but it could have adverse effects on regions – such as the Wallonie – which are bypassed by the train. Vickerman argued that, to achieve long-run regional convergence, decision-makers should look at the sectoral needs of specific regions instead of emphasising short-run aggregate demand effects. It is not sufficient to base policy towards infrastructure networks just on network measures; the role which the networks play in the development of peripheral regions and the process of cohesion is also relevant. Policies aimed solely at the reduction of infrastructural disparities may not necessarily have the desired effects. Jacques Thisse noted that it is not the mean, but the variance, of transport costs that is important. Furthermore, different kinds of infrastructure have very different effects on regional economic convergence. Konrad Stahl underlined the problems of measuring infrastructural capital.

In ‘Externalities and Interdependent Growth’, Antonio Ciccone (University of California and Universitat Pompeu Fabra) estimated the importance of cross-country externalities for growth. Whereas externalities in human and physical capital play a central role in the literature on endogenous growth, recent empirical tests still tend to rely on the Solow growth model without externalities. Ciccone constructed a model which allows for estimation of externalities within a country, as well as technological interdependencies across countries. International externalities operate through three channels: first, international technological interdependencies increase steady-state average labour productivity growth rates for any exogenous growth rate of total factor productivity; second, international interdependencies increase the effect of investment on steady-state labour productivity; and third, endogenous propagation becomes possible despite decreasing aggregate returns to scale for capital at the country level. Empirically, Ciccone found that external returns to scale for physical capital within countries were 8%, that a 10% increase in total factor productivity in a country’s neighbour raised its own total factor productivity by 6%, and that a 2% annual growth rate of labour productivity could be explained as an endogenous response to an exogenous 0.2% annual growth rate of total factor productivity in the steady state. Furthermore, internal returns to scale for physical capital were estimated at 28%, which was almost exactly the average share of capital in national income in the United States over the sample period.

Angel de la Fuente (Institut d’Anàlisi Econòmica, CSIC, Barcelona, and CEPR) presented his research on ‘The Sources and Growth of Convergence: A Close Look at the Spanish Regions’. Allowing for technological diffusion across regions, growth rate effects from human capital and fixed regional effects, he goes beyond the standard neo-classical growth model. His estimates, based on an extensive data set (1964–91), lead to the conclusion that some of the widely accepted stylised facts about regional convergence require qualification. Whereas many recent studies suggest that convergence occurs at a rate of 2% per year, de la Fuente’s results indicate that the stationary distribution of income is unequal and that convergence towards this unequal distribution is very fast. The regional level of education contributes most significantly to explaining regional growth rates. The estimates indicate omission of important variables, possibly relating to regional differences in sectoral structures. Initial estimates including sectoral data yield very promising results. Antonio Ciccone remarked that estimating the determinants of such fixed regional effects boils down to explaining region-specific Solow residuals – an investigation which is urgently needed.

In ‘Regional Dynamics’, Mario Fomi (Universita de Modena) and Lucrezia Reichlin (ECARE, Université Libre de Bruxelles, and CEPR) proposed a new empirical methodology for issues of economic geography. They outlined a dynamic factor model of the cross-sectional distribution of economic activity. The methodology allows common shocks to have heterogeneous impacts at the region and/or country level. The investigation of the regional structure of factors leads to the discovery of regional clusters. In initial empirical results with US data, it appears that two factors alone capture 98% of regional variance of output caused by macro- economic shocks. Local shocks have only short-run effects. The analysis of the autoregressive structure of the model furthermore allows the evaluation of geographical interaction. In Europe, regional and European factors appear to be more important than national ones. Jacques Thisse pointed out that the proposed methodology was initially atheoretical, but that – similarly to CAPM factor analysis – an economic interpretation of the identified factors can be done in a second step.

Yannis Ionnides (Tufts University) analysed the evolution of metropolitan areas in the US from 1900 to 1990. His paper, ‘The Evolution of City Size Distribution in the United States’, was written with Linda H Dobkins (Emroy Henry College). Their empirical work was based on a theoretical model which derives a ‘city production function’. The model emphasizes human capital accumulation and increasing returns to scale in the production of differentiated varieties. Demand for national output drives the evolution of cities in the economy. In the empirical analysis, based on a newly constructed data set, it turns out that the main characteristics of the dynamics of city size distribution in the twentieth century in the US are the emergence of new cities and the tendency towards increasing inequality in the sizes of cities. This ‘divergence’ result is based on regressions using the Pareto distribution. Nevertheless, a non-parametric model shows a tendency towards convergence of city size distribution for the last decade. It turns out that the coefficient of city population on its own lagged value is slightly smaller than one, which indicates a high degree of persistence in the city size distribution.

In ‘Illegal Immigration, Border Enforcement, and Relative Wages: Evidence from Apprehensions at the US-Mexico Border’, Gordon Hanson (University of Texas) and Antonio Spilimbergo (Inter-American Development Bank) showed that the evolution of real wages was the main factor determining the flow of illegal immigration from Mexico to the United States between 1976 and 1995. To overcome the lack of data on illegal immigration, an apprehension function was developed theoretically. Observed apprehensions were modelled as the probability of being caught at the border multiplied by the total number of Mexican residents attempting to cross the border. The model allows for dynamic specifications of illegal immigrants and considers both expected future border enforcement and expected future relative wages. The authors found – as a lower bound – that a 10% decrease in the Mexican real wage induces approximately a 8% increase in illegal immigration. Furthermore, economic instability – such as the Mexican debt crisis at the beginning of the 1980s – tended to cause increased immigration. This finding leads to the conjecture that an increase in real wages in Mexico should reduce illegal immigration flows. These results strongly imply that, if policy initiatives – such as the formation of NAFTA, or US loan packages to Mexico – favour Mexico’s economic convergence (in both levels and variability), they will also lower illegal immigration. On the efficacy of US border patrols, the authors found that an additional hour of policing at the border would yield 0.25 to 0.33 additional apprehensions.