This study evaluates and compares the effectiveness of two alternative training schemes for the unemployed: short, job-search oriented training and long, human capital oriented training. We investigate the impact dynamics of these programs considering both unemployment and employment duration. Our analysis uses a rich administrative longitudinal data set for Germany where both programs were implemented at the same time. Our estimation strategy flexibly accounts for dynamic selection on observables and unobservables as well as heterogeneous treatment effects. The results indicate that short-term training schemes can be an effective tool to reduce unemployment in the long run, in particular if participation occurs early during unemployment. Long-term training schemes increase the average duration of employment spells more strongly than short-term training. However, they increase the expected unemployment duration if they are started early during unemployment. This negative short-run impact of long-term training is an important driver of its overall effectiveness.