This project investigates the employment effect of heterogeneous training programs in Germany, with a special emphasis on multiple program participation. Most studies either lump program from a rather wide range together or they evaluate programs in isolation from others. As a result only average effects over different programs are calculated or interactions over different programs are neglected. In neither case it is possible to identify programs that work. This project aims at identifying successful programs by concentrating on two dimensions of program heterogeneity. Multiple program participation and different program contents. Current research on this topic suffers from two drawbacks:
The lack of data and methodological difficulties. In this project I plan to use unique recent data and to develop novel empirical methodologies. I have access to three unique recent data sets from the German Federal Services. I plan to follow two approaches for estimation of programs' effects:
First, I will extend the conditional difference-in-differences approach, to take potential endogeneity of multiple program participation into account. This part will be based on my previous work (Bergemann, Fitzenberger, Schultz, Speckesser, 2000).
Second, I will apply bivariate duration models that I have to extend to for complementary effects of multiple program participation. To work on the project I have to acquire further knowledge in theoretical statistics and on duration models. This will be done with the help of UCL workshops and courses. Richard Blundell and HideIchimura and several other members of the faculty will be a great help when solving methodological problems, as for example the modeling of the dynamic selection effects. Furthermore, I plan to discus problems connected with the use of large administrative data sets with members of the Cemmap and the Microeconomic Analysis of Fiscal Policy. With my research I will add to the existing research emphasis at UCL.