We evaluate a pilot workfare project of a Berlin city community to integrate young people in social assistance into the labor market. Reference data are generated in collaboration with the German Employment Office. The participation effect is found to be positive and of relevant size, but not statistically significant. This indicates that selection biases detected in the allocation process of individuals to the program can be adjusted by econometric techniques, but only on the prize of reduced significance levels. Given the small sample sizes of pilot projects this calls for social experiments in controlled project designs.