Does welfare to work work? Lessons from a pilot project

We are reporting from an evaluation of a municipal workfare pilot project. The program was intended to integrate young welfare recipients into employment. According to the workfare principle, participation was mandatory for those being assigned to the program. Welfare claims were denied for those who refused to participate. Since program capacity was limited, the assignment process could be viewed as a quasi- random process, which allowed for an impact evaluation by comparing program participants with recipients not assigned to the program due to capacity limitations. In retrospect, however, it turned out that the assignment process was still subject to uncontrolled selection, which strongly impedes comparability between treated and alleged controls. By means of econometric methods we have tried to correct for selectivity. Impact measurement is based on administrative data of the welfare authority which were merged with administrative data from the Federal Labor Agency in order to get information about employment histories before and after welfare claiming. The results are pointing to a strong positive employment effect of workfare, although not statistically significant. We conclude that slackness of project design may be compensated by econometric methods, but only at the price of a strongly increased number of observations in order to achieve statistically significant results. If policy makers need to draw conclusions from typically small scale pilot projects, there is no alternative to a project design, which allows for a controlled experiment.