The Shift-Share Regression: An Application to Regional Employ-ment Development

This paper presents an outline of the so-called Shift-Share Regression and an application of this method to the analysis of employment development. The method used is not a deterministic decomposition such as the classical Shift-Share-Analysis, but a powerful, yet simple and flexible econometric tool to test theory-related hypotheses, which can be applied as a ?work-horse? in studies of many kinds. Classical deterministic Shift-Share-Analysis has often been criticised, since it does not permit a model-based analysis. The detection of causal effects is at least problematic and the inclusion of additional explanatory variables is possible only in special cases. A major problem is the nature of the method as a deterministic procedure which excludes significance tests. Shift-Share Regression is able to overcome all these critical points. In a basic version it was introduced by Patterson (1991) as a method for analysing and testing regional industrial developments. In contrary to the deterministic Shift-Share-Analysis employment development was examined in a linear model. In Patterson?s analysis the industrial sector structure was used as the sole determining factor alongside the location effects and the national trend. We extend this basic structure further: In our case, the effects of sectoral structures, establishment size, qualification structures and locational determinants on employment growth are studied. The regional units used are districts of Western Germany (?Landkreise? and ?kreisfreie Städte?), especially in the present context the districts of the federal State of Bavaria. The analysis is motivated by theoretical considerations of different sources. The most important one refers to theoretical analyses of structural change. According to a specific theorem, the employment effect of technological progress depends on the elasticity of product demand. If demand is inelastic the direct labour saving effect of technological progress is dominating and the effect is negative. Then it is profitable for a firm to reduce its labour force. If, however, demand is elastic a compensating effect dominates. In this case price decreases following higher productivity lead to an extension of product demand which (over-)com¬pensates the direct labour saving effect. Then, it is profitable for a firm to increase the size of its labour force. It can be assumed that in different industries of an economy different demand elasticities are dominating. Therefore, an empirical analysis of employment effects should focus on the industries of an economy. In the paper the sources of different regional employment development in Bavaria are presented, analysed by Shift-Share Regression.