Identifying the Geographical Agglomerations of Manufacturing Industries

Abstract

The presence of localized industrial aggregation, also from the viewpoint of regional and country growth theories, can be regarded as a potential competitive advantage for regions or nations. While there is an abundance of qualitative analysis of the clustering of single industries and the way in which spatial concentration changes into economies of agglomeration, for a long time it was difficult to undertake quantitative work on the same subjects because of the lack of any econometrically operational definition of industrial agglomeration.

Only recently, thanks to the work of Ellison and Glaeser (1997), it has become possible to measure the geographical concentration of production in a statistically correct way, excluding from the agglomeration phenomena cases of industrial concentration due to a random distribution of a small number of large plants.

Following these intuition many other measures of agglomeration have been proposed. The empirical evidence provided by this literature confirms that most industries are, in a large number of countries, more highly clustered than a model of random spatial distribution of firms might predict.

In these paper, starting from the work of Ellison and Glaeser we derive an algorithm to select and map industrial agglomeration. We propose, specifically, a test of the industrial specialization of an area based on the null hypothesis of absence of agglomeration advantages: in this way we can select the areas that pass the test and build around them a set of neighbouring areas that represent the space of probable diffusion of proximity advantages.

Then, we apply the algorithm to the geographical distribution of manufacturing activities in Italy and give a description of the main features of agglomerated areas in the country. We are able, among other things, to verify the high variability that subsists among Italian agglomerations in connection with their dimension, characteristics of specialization and presence of large firms. The considerable dissimilarities found in the structural features of agglomerations confirm that, in econometric research into the advantages of producing in an industrial agglomeration, a strictly dichotomous approach can be strongly misleading.

Finally, we discuss the conceptual differences between the “industrial agglomeration” obtained with our algorithm and the Marshallian notion of “industrial district” that represent a key analytical category in the literature on local development models in Italy.