Evaluating the impact of Mexico's quality schools program : the pitfalls of using nonexperimental data

The authors evaluate whether increasing school resources and decentralizing management decisions at the school level improves learning in a developing country. Mexico's Quality Schools Program (PEC), following many other countries and U.S. states, offers US$15,000 grants for public schools to implement five-year improvement plans that the school's staff and community design. Using a three-year panel of 74,700 schools, the authors estimate the impact of the PEC on dropout, repetition, and failure using two common nonexperimental methods-regression analysis and propensity score matching. The methods provide similar but nonidentical results. The preferred estimator, difference-in-differences with matching, reveals that participation in the PEC decreases dropout by 0.24 percentage points, failure by 0.24 percentage points, and repetition by 0.31 percentage points-an economically small but statistically significant impact. The PEC lacks measurable impact on outcomes in indigenous schools. The results suggest that a combination of increased resources and local management can produce small improvements in school outcomes, though perhaps not in the most troubled school systems.