This empirical paper wants to analyze determinants for regional differences in German unemployment rates using a spatial panel model. The analysis of regional differences is of particular interest in the German case due to its specific history. Twenty-two years after German reunification, there are still structural differences between both parts affecting economic activity. We identify the driving factors in the whole of Germany as well as in East and West Germany separately. To our best knowledge, this study is the first contribution investigating regional unemployment in the reunified Germany. We provide evidence of spatial dependence in regional unemployment data. Taking into account theoretical contributions to this literature, we exploit 24 possible explanatory variables on 412 German districts for the period from 1999 until 2007. To select the relevant variables for our model, we apply a two-step model selection procedure. Firstly, we divide our variables into three groups according to theoretical importance. Secondly, we regress regional unemployment rates on different combinations of variables where the essential variables are always contained. As unemployment data also exhibit temporal dependence, we specify both a static and a dynamic spatial panel model. The spatial econometric literature provides various types of spatial models. To decide which type of spatial model is appropriate in our context, we apply the specification test by Debarsy and Ertur (2010, Reg. Sci. Urban Econ.). The static model specification is estimated using the quasi-maximum likelihood estimator proposed in Lee and Yu (2010, J. Econometrics) and the dynamic model is estimated using the quasi-maximum likelihood estimator proposed by Lee and Yu (2010, Econometric Theory). To incorporate the spatial information into the model, we construct different spatial weights matrices. On the one hand, we use the binary contiguity matrix and, on the other, we apply a distance-based spatial weights matrix as well as the matrix proposed in Molho (1995, J. Reg. Sci.). We extend the existing literature by the following two aspects: Firstly, we apply both a static and a dynamic spatial panel model. Hence, we exploit the panel dimension of our data and, in addition to that, we account for both spatial and temporal dependence in the data. Our results show that the spatial dynamic panel model fits the data in the best way. Secondly, we provide evidence for regional unemployment in Germany being of disequilibrium nature. This finding justifies political interventions on regional labor markets.
