Abstract

Unemployment rates at the regional level are known to show persistent differentials. Building upon theoretically based hypotheses various factors can be identified. However, the conclusions with regard to the importance and the direction of the impact are sometimes contradictory. E.g., in recent empirical studies both a positive and a negative impact of the regional unemployment on the wages are found. Negligence or inadequate treatment of the spatial dimension or, in econometric terms, omitted-variable bias and error-in-variables bias may be one cause for these contradictory results.

Because the differentials in unemployment as well as many explanatory factors do not have a random geographical distribution, but generate observable spatial patterns, they should be analysed using methods of regional science: Within these, the approach of spatial filtering a spatial-econometric equivalent to the Principal-Component analysis seems very promising for a systematic detection of the spatial patterns, illuminating and capturing the relevant structures in the data,. Of particular interest is the interrelation between the spatial filters and the estimators of explanatory variables as well as a comparison of the estimations with other spatial econometric methods. A correct assessment in economic relationships between unemployment and explanatory factors and the proper inclusion in econometric analysis are necessary to receive consistent results.

Since unemployment, and particularly the high regional unemployment rates in some parts of the country, is one of the central and most challenging problems of economic politics in Germany, explaining these spatial matters has particular relevance even for policymaking reasons. In that, the spatial pattern of unemployment is of direct interest. Furthermore, it will be possible to conclude on the degree of estimation bias by analysing how the estimated partial correlations of other variables shift due to controlling for the significant spatial patterns.