The ins and outs of German unemployment

This paper analyzes the relative flow contributions to different age cohort unemployment dynamics using administrative panel data for Germany. I consider a three-state model and allow for flows from/to inactivity (out of the labor force) to/from unemployment and employ both a steady-state and a non-steady-state approach. I use monthly data with exact information on the stocks of employment and unemployment as well as on flows between these two stocks and between unemployment and inactivity (three-state model). I use administrative data provided by the German Federal Employment Agency for the period January 2007 to December 2014 and consider two different aggregation levels, national and regional (402 counties). For both data, using variance analysis, I examine all unemployed and the age cohorts 15-24 years, 25-34 years, 35-44 years, 45-54 years, and 55-64 years old, to point out age related differences in unemployment dynamics.

I argue that it is not evident which implications the increasing relative appearance of older job seekers and job candidates may have relative to job-worker matching in the labor market and, ultimately, for unemployment dynamics. This is the first contribution that analyzes age related flow contributions to unemployment fluctuations and this is the first study in the literature on relative flow contributions to unemployment dynamics that considers panel data instead of time series only. The overall inflow and outflow contributions to the unemployment variance vary from 40%:60% to 55%:45%, depending on the approach. The dynamics that arise from the inactivity and activity flows at least account for about 20% (and up to 88% for the youth) of unemployment dynamics, while the contributions of separation and job finding are roughly 35% and 40% respectively. Across five age cohorts I find remarkable differences in flow contributions. The results provide strong evidence for decreasing labor market dynamics when the share of older workers increases. I also point out that the consideration of regional data instead of aggregated data provides more powerful results and is, in the case of Germany, more important than considering the time aggregation bias.