Geostatistics Approach for Mapping Potential Industry Clusters: Developing Isotropic and Anisotropic Scenarios at Firm Level

Abstract

Agglomeration economies are manifested in geographical concentration of firms, as stated by Agglomeration Theory. The objective of this paper was to examine the proximity and concentration of firms in continuous space and in specific directions for the mapping of potential industrial clusters, through an unprecedented approach. The methods used were the indicator semivariogram, anisotropy modeling and kriging interpolation. The geostatistical approach was initially applied in a simulated data with patterns known a priori: random and clusters. Then it was validated with real data of roasted coffee industry in Minas Gerais, Brazil. Each dataset was tested in isotropic and anisotropic scenarios. Geostatistics was sufficient to detect patterns of randomness and spatial dependence, measure the extent of proximity between firms, identify regions with a high level of industrial concentration on the map and estimate an index of concentration at the firm level. The results also advanced the specialized literature by indicating that the anisotropic scenarios, which adjusts the directional bias, better described the reality of the phenomena compared to the isotropic ones. Error measures in anisotropic scenarios were between 7 and 30% more accurate. It was concluded that the direction can be a determinant of the theory of agglomerations and that geostatistics can be useful for mapping potential industry clusters.