Using worker flows in the analysis of establishment turnover : evidence from German administrative data

"Economists have long been interested in the determinants and components of job creation and destruction. In many countries administrative datasets provide an excellent source for detailed analysis on a fine and disaggregate level. However, administrative datasets are not without problems: restructuring and relabeling of firms is often poorly measured and can potentially create large biases. We provide evidence of the extent of this bias and provide a new solution to deal with it using the German Establishment History Panel (BHP). While previous research has relied on the first and last appearance of the establishment identifier (EID) to identify openings and closings, we improve on this approach using a new dataset containing all worker flows between establishments in Germany. This allows us to credibly identify establishment births and deaths from 1975 to 2004. We show that the misclassification bias of using only the EID is very severe: Only about 35 to 40 percent of new and disappearing EIDs with more than 3 employees correspond unambiguously to real establishment entries and exits. Among larger establishments misclassification is even more common. We show that many new establishment IDs appear to be 'Spin-Offs' and these have become increasingly more common over time. We then demonstrate that using only EID entries and exits may dramatically overstate, by as much as 100 percent, the role of establishment turnover for job creation and destruction. Furthermore correcting job creation and destruction measures for spurious EID entries and exits reduces these measures and aligns them closer with the business cycle." (Author's abstract, IAB-Doku) ((en))