Concentration, spatial clustering and the size of plants : disentangling the sources of co-location externalities

Following the model-based approach of Ellison and Glaeser (1997), we develop a framework to test for the link between concentration, spatial clustering and the size of plants. Concentration is an a-spatial concept of variability that can be measured with the standard locational Gini or the more sophisticated Ellison and Glaeser index. By contrast, spatial clustering is directly concerned with distances. Therefore we also use a two-dimensional measure (the Moran index) to identify some specific distance-based patterns. We argue that, in a world where the size of establishments is independent of both concentration and spatial agglomeration, as the standard Dixit-Stiglitz (1977)-Krugman (1980) framework, all the variability in these measures should be accounted for by the variation in the number of plants. Using the Italian 1996 census year data on manufacturing industries, we therefore compare the values and significance of both the EG and Moran indexes computed on an employment and number of plants basis. Our results indicate that, for the majority of Italian manufacturing industries, big plants are much more concentrated than small ones, with size and concentration simultaneously influencing each other. On the other hand, small units are shown to be more spatially correlated, suggesting that different externalities may drive (or may be driven by) concentration and agglomeration patterns according to a size-related basis. These results therefore cast some doubt on the relevance of standard monopolistic frameworks to structurally account for the role of the so-called "pecuniary" externalities compared to more "localized" ones, such as Marshallian, Jacobian or factor endowments based externalities.