Quasi-experimental methods in empirical regional science and policy analysis : is there a scope for application?

Applied econometrics has recently emphasized the identification of causal parameters for policy analysis. This revolution has yet to fully propagate to the field of regional science. We examine the scope for application of the matching approach – part of the modern applied econometrics toolkit – in regional science and highlight special features of regional data that make such an application difficult. In particular, our analysis of the effect of regional subsidies on labor-productivity growth in Germany indicates that such policies are effective, but only up to a certain maximum treatment intensity. Although the matching approach is very appealing due to its methodological rigor and didactical clarity, we faced difficulties in balancing the set of covariates for our regional data given that the regions differ strongly with respect to the underlying structural characteristics. Thus, results have to be interpreted with some caution. The matching approach nevertheless can be of great value for regional policy analysis and should be the subject of future research efforts in the field of empirical regional science. -- Generalized propensity score ; nearest neighbor matching ; labor productivity growth ; regional policy