Identifying spatial nonstationarity in German regional firm start-up data

Background: We investigate the relationship between the rate at which new firms are established and regional characteristics and whether this relationship is constant over space or not. The characteristics reflect (i) agglomerations that in turn are related to increasing returns to production and (ii) measures which are influenceable immediately by regional decision makers. Method: In order to account for spatial nonstationarity, we use the geographically weighted regression method for German start-up data on the geographical scale NUTS3. Moreover, we discuss significance test for locally varying regression coefficients. Results: We are able to verify the global positive relationship between production convexities—measured by population density and growth amongst others—and the start-up rate on a regional level (Kreise). Furthermore, we find the share of industrial real estate to have positive influence on the regional start-up rate. Finally, we find strong empirically evidence that there is spatial nonstationarity in the data and hence the assumed relationship varies locally. Conclusion: The results give evidence that spatial nonstationarity could not be neglected in the analysis of start-up rates. However, we suggest to develop global models that account—at least partially—for the underlying spatial nonstationarity by exogenous variables. Copyright Springer-Verlag 2012