This paper analyzes empirically the effect of spatial agglomeration of activities on plant-level productivity,
using French firm and plant-level data from 1996 to 2004. We exploit short-run variations of variables
by making use of GMM estimation. This allows us to control for endogeneity biases that the estimation of agglomeration economies typically encounters. This means that our paper focuses on a subset of agglomeration economies, the short-run ones. Our results show that French plants benefit from localization economies, but we find very little – if any – evidence of urbanization economies. We also show that those localization benefits are relatively well internalized by firms in their location choice: we find very little difference between the geography that would maximize productivity gains in the short-run and the geography actually observed.