Industrial Location, Agglomeration, Markov Chains, Dynamic Increasing Returns.

C1, L6, R1

This work explores the spatial structure of location of production activities. We try to disentangle location- from sector-specific drivers in the dynamic process of spatial agglomeration. We argue that the former typically apply "horizontally" (i.e. across all industrial sectors), while the latter unfold in the form of non-decreasing dynamic returns to the current stock of installed business units. A stochastic model of location is developed and three different specifications are tested against Italian data on the location of manufacturing firms. Our results suggest that different locations exert different structural influences on the distribution of production activities. Moreover, a widespread horizontal power of "urbanization", which makes particular locations more attractive irrespectively of the sector, does emerge. However, after controlling for the latter, one is still left with sector-specific forms of dynamic increasing returns to agglomeration, which vary a lot across different manufacturing activities.