The implementation of focalized education policy on the worst performing schools is controversial. Its positive discrimination nature has lead to both adherents and adversaries and the lack of serious empirical evaluations has left the debate in a highly theoretical level. This paper contributes to the debate by evaluating one of such policies implemented in Chile: The P900 Program. The paper´s findings suggest that given the compensatory nature of selection on unobservables, the traditional estimates (OLS and fixed effects) are downward biased. Bias free estimates find that the program has been increasingly effective in reducing the achievement gaps.