Evaluation of the P900 Program: A Targeted Education Program for Underperforming Schools

Education policies targeted at the worst performing schools are controversial. Its positive discrimination nature has been debated, while the lack of serious empirical evaluation has frozen the discussion at a highly theoretical level. The biggest problem has been to identify unbiased effects, given non-random participation even within underperforming schools. This paper contributes to the debate by evaluating the P900 program - that provides material support to low-achieving Chilean schools- estimating effects that are free of bias due to unmeasured fixed school-specific effects, which are correlated with participation. It explicitly considers the changing nature of both the program and the selection process, computing yearly effects and biases. The simplifying assumptions usually found in the literature (e.g. constant program effects, bias, and school effects) are tested and rejected, thus shedding doubts on previous findings. The paper finds that schools were selected for the P900 in an unobserved compensatory manner, thus downward biasing uncontrolled estimates. Moreover, the bias and the estimated effects of the Program are increasing in time. The 1992 program is not significantly different from zero, while the 1994 and 1996 program´s are significantly positive. The latter is significantly higher than the rest, thus it would be effectively helping the schools that participate in it.