Regional Concentration and Confidence Regions

Abstract

Industries necessarily differ with respect to their type of geographical concentration. When some industries are overrepresented in urban areas (urban concentration), then some other industries must be overrepresented in rural areas (rural concentration). Unfortunately, the existing measures of concentration cannot distinguish between urban and rural concentration. They simply ignore the problem and rank industries with respect to their degree of concentration, even though these industries may exhibit completely different types of concentration. In the present paper we develop a new approach that avoids such misleading comparisons. Our approach distinguishes not only between urban and rural concentration but between seven different geographical patterns. The statistical identification of each industry's geographical pattern is based on two Goodman-Kruskal rank correlation coefficients and their bivariate confidence region. Using German employment data on 613 different industries, the power of our approach is demonstrated.