Abstract
Regional determinants of new firm formation are of interest to researchers and policymakers. In the analysis of new firm formation, most studies use econometric approaches that mask intra-unit variations, not recognizing counterbalancing and dilution effects as a result. Recent advances in spatial statistics such as Geographically Weighted Regression (GWR) take local variations into account. However, these approaches operate only on a bivariate level, making it impossible to detect the homogenous parts of the area under examination with regard to a number of relationships between new firm formation and its determinants. Based on a sample of 412 German regions, we apply GWR and subsequent graph-partitioning clustering to identify multi-relationally homogeneous sub-areas. Being that the results suggest a four-cluster solution, ‘one size fits all’ policies and premature unit zoning can be called into question.