Cluster Analysis using Microgreographic data

In this paper we try to identify manufacturing and service clusters in Spain, using data from Mercantile Registers of 2006. The proposed methodology partially follows contributions of Duranton and Overman (2005), Brenner (2003 and 2004) and Ellison and Glaser (1997), but departing from them we improve such approaches by several ways. In order to sum up, we can detail our approach and divide it into five stages. Firstly, we divide space into homogeneous cells. Secondly, we create industry specific maps departing from firms' georeferenced data. Thirdly, we create multiple random industry specific maps under two conditions: i) total number of firms at each industry remains constant and ii) total number of firms at each cell remains constant. Fourthly, we compare the observed spatial distribution of firms with random simulations of such distribution and we check if there is some kind of concentration compared to the random distribution. Fifthly, for each industry we map the areas where the concentration of firms is significantly higher than expected. Previous scheme allows us to identify real clusters (of different shapes and sizes) for all range of manufacturing and service activities and to use this information to design public policies related to such industries. Keywords: cluster analysis, geographic data, microeconomics, regional economics.