This paper provides micro-econometric evidence on the effectiveness of Active Labor Market Policies (ALMP) in Poland. We sketch the theoretical framework of matching estimators as a substitute for randomization in labor market programs. Using retrospective data from the 18th wave of the Polish Labor Force Survey we implement a conditional difference-in-differences matching estimator of treatment effects. Treatment and control groups are matched over individual observable characteristics and pre-treatment labor market histories to minimize bias from unobserved heterogeneity. We also require that observations on controls are from the same regional labor market and from an identical phase of the transition cycle. Considering as the outcome a multinomial variable of labor market status, our first important finding suggests that training of men and women has a positive effect on the employment probability. For men public works and intervention works have negative treatment effects, while participation in intervention works does not affect women’s employment probabilities. We attribute the negative treatment effects for men to benefit churning rather than to stigmatization of intervention and public works participants.