Patterns of Localisation in the French Manufacturing and Service Industries: a Distance-Based Approach

We explore the patterns of localisation in the French manufacturing and service industries using the distance-based approach developed by Duranton and Overman (2005). The idea of this methodology is to consider the distribution of distances between pairs of plants in an industry and to compare it with that of a hypothetical industry with the same number of plants which are randomly distributed conditional on the distribution of aggregate activity. Relying on the same industrial classification as the one used by these authors, we show that their main conclusions remain valid in the case of French manufacturing industries. First, fewer manufacturing industries are found to be globally localised than usually found with discrete indices (Ellison-Glaeser, Maurel-Sédillot). Second, localisation takes place at small distance (< 50 km). Third, the degree of localisation varies drastically across industries. Furthermore, we find that service industries deviate more often from randomness than manufacturing industries, and that a higher proportion of them show global localisation. This localisation occurs at shorter distances than for manufacturing industries. Finally, these patterns of localisation remain fairly stable over the period 1996-2005. Within industries, we find that exiters are acting to reinforce localisation while entrants are acting in the opposite direction.