50th Congress of the European Regional Science Association: "Sustainable Regional Growth and Development in the Creative Knowledge Economy", 19-23 August 2010, Jönköping, Sweden
The description of geographical concentration and the search for its causes is at the focus of many studies. However, by not considering developments over an extended period of time, the majority of the studies mainly examine static relationships. This paper aims at filling this gap. We measure concentration in Western Germany for the time period from 1986 to 2006 with the Ellison-Glaeser index (EGI). In order to account for concentration processes at various levels, we investigate the long-run development of the EGI for NUTS3-regions and labour-market regions as well as for 43 two-digit-industries and 191 three-digit-industries. The Establishment History Panel provides detailed information about employment at the level of individual firms. The major part of the industries exhibits a larger degree of geographical concentration than one may expect when location decisions of firms are purely stochastic. This result holds true throughout our observation period irrespective of the applied spatial and industrial levels. Thus, regional concentration matters for about two thirds of all industries in Western Germany. However, despite concentration being significant, the value of the EGI is rather low for most of these industries. Since the mid 1980s, the shares of strongly localized and non-localized industries have been declining and increasing, respectively. Hence, dynamic changes of the industry-specific regional structure of economic activities have taken place at the lower and upper tails of the EGI distribution. A more detailed analysis of EGI rankings and changes is carried out for ten industry groups. All in all, sectoral shifts affect changes in industry-specific regional patterns. This holds true for old shrinking and stagnating industries and as well as for dynamic high-tech-manufacturing industries and business-related human-capital service industries. To confirm the descriptive findings in a multivariate setting, an econometric analysis is carried out. Using a multinomial logit model, we examine which forces lead to an increase, decrease or unchanged state of geographical concentration. Again, we find evidence that agglomeration externalities in old industries seem to have lost their importance. At the same time, the high-technology industries have not localized strongly enough to compensate for this effect.