Sector structure, Evolution and Agglomeration Externalities: Testing the Impact of Related Variety, Unrelated Variety and Specialization in a Cross-Section of European Regions

Abstract

The relation between regional sector structures and their resulting agglomeration externalities is one of the most thoroughly studied research topics in Economic Geography and Urban and Regional Economics. Recently however this research area has faced conflicting empirical findings, while insufficiently detailed theoretical models make it hard to find the cause of these disagreements. This study builds on the model of agglomeration externalities as proposed by Frenken et al. (2007), which inspired by Evolutionary Economic Geography distinguishes related variety and unrelated variety besides specialization and urbanization. A cross-section of 234 NUTS2 regions in 19 European countries is used to put the model to the test. Since this database is based on firm-level data, two types of sector structure variables are calculated which are rare in previous research: one based on the number of firms per sector, and the other weighted by the operating revenue of firms per sector. While several country-level tests have previously confirmed predictions based on this model, the empirical analysis in this research cannot find clear evidence in favor of the model. When variables based on absolute numbers of firms are used limited evidence for the employment enhancing effect of related variety can be found, but the positive relation between specialization and labor productivity is too weak to be confirmed. The expected protection against unemployment growth in regions with a structure of unrelated variety could not be found, and when variables are weighted by operating revenue none of the sector structure variables show a clear effect. Controlling for the number of hours worked (for employment growth and labor productivity growth) and for specialization calculated at different levels of sector detail does not yield significantly different effects. The paucity of confirmatory evidence may be caused by interference of the modifiable areal unit problem, besides the possibility that the expected effects of the model are only visible at different geographical or temporal scales. Further research is called for to disentangle these possible explanations of the research findings.