Location patterns of service industries in France: A distance-based approach

We study the location patterns of business-oriented service and manufacturing industries in France. We develop a new test of localization considering space as continuous. Our test relies on a measure of divergence in the space of density distributions that allows us to assess whether or not the density distribution of bilateral distances between all pairs of plants within an industry significantly departs from randomness. We improve the test proposed by Duranton and Overman (2005), which proves to be biased with respect to the number of plants in the industry under scrutiny. Our test does not suffer from such a bias. This property is crucial for the French case where industrial concentrations of service and manufacturing industries drastically differ. With this distance-based method, we highlight some distinctive locational features of service industries that have not been mentioned in the literature so far. We show that: 1/service industries diverge more often from randomness than manufacturing industries, 2/a majority of diverging service industries are localized at very short distances (less than 4 km) whereas a majority of manufacturing industries are localized at longer distances or even dispersed, 3/within a majority of service industries, the largest plants appear localized at shorter distances than other plants, 4/within most service industries, incoming plants reduce localization whereas exiting plants reinforce it over the period 1996–2005.

► We study the location patterns of service and manufacturing industries in France.
► We develop a new test of localization considering space as continuous.
► We improve the test proposed by Duranton and Overman (2005).
► Service industries diverge more often from randomness than manufacturing industries.
► A majority of diverging service industries are localized at very short distances.