"We re-analyze the effects of a Danish active labor market programme social experiment, which included a range of sub-treatments, including meetings with caseworkers, job search assistance courses, and activation programmes. We use newly developed non-parametric methods to examine how the effects of the experimental treatment vary during the unemployment spell. Non-parametric techniques are important from a methodological point of view, since parametric/distributional assumptions are in conflict with the concept of experimental evidence. We find that the effects of the experiment vary substantially during the unemployment spell."

Support: Paper