Heterogeneous employment effects of job search programmes : a machine learning approach

The third chapter estimates negative average effects of a Swiss job search programme for unemployed persons. Those who are send to the job search programme are significantly less likely to find a job quickly than comparable unemployed persons who do not participate in such a programme. Recent causal machine learning tools are applied to investigate whether this is true for all groups of unemployed persons or if some groups actually benefit. The results show that only a tiny fraction of all unemployed persons seems to benefit from this specific programme and that caseworkers who assign them do not successfully identify this group.