The Albanian Ndihma Ekonomike is one of the first poverty reduction programs launched in transitional economies. Its record has been judged positively during the recession period of the 1990s and negatively during the more recent growth phase. This paper reconsiders the program using a regression-adjusted matching estimator first suggested by Heckman et al. (1997, 1998) and exploiting discontinuities in program design and targeting failures. We find the program to have a weak targeting capacity and a negative and significant impact on welfare. We also find that recent changes introduced to the program have not improved its performance. An analysis of the distributional impact of treatment based on stochastic dominance theory suggests that our results are robust.