Empirical analysis of spatial location of activity: a proposal of using spatial association statistics

Based on the initial paper of Krugman (1991), several works have analysed the consequences of the integration process on the activity location. In this sense, Puga (1996) and Puga and Venables (1996) among others, have derived an inverse U-shape relationship between integration and agglomeration. In order to check this hypothesis, it is habitual to compute some of the traditional concentration indexes. However, this kind of indexes considers each region as an isolated economy, discarding links with others. Taking into account these ideas, this paper proposes the use of some spatial association tests in order to shed light on the consequences of an integration process, overcoming the limitations shown by other indexes of concentration. First, we discuss, from a theoretical point of view, the utility of these spatial econometric tools to get the goal of the paper. So, we analyse the behaviour of the spatial correlograms (global spatial autocorrelation tests for different orders of contiguity) and the local spatial association tests for some theoretical spatial distributions of the product (even distribution, core- periphery scheme, monocentral distribution with and without hierarchical distribution around the centre, etc). Then, we derive different temporal evolutions both of these correlograms and the local statistics related to possible scenarios after the integration (more concentration, more random distribution, etc). Second, we apply this analysis to the European Union case. In this sense, we compare the evolution shown by the GDP pc among the European regions during the period 1975-1992 with the theoretical distributions defined in the first part of the paper in order to know the trend of the spatial distribution of the activity after the deepening in the integration process. Keywords: Spatial association test, spatial econometrics, integration process