Using data on primary schools in 10 Latin-American countries, we estimate the impact of decentralized school decision-making on student performance. We develop a model that shows that local autonomous effort will be jointly determined with student academic performance. The model predicts that least squares estimates are biased toward finding a positive impact of school autonomy on student performance. Empirical tests confirm these predictions. Least squares estimates show a strong positive effect of decentralized decision-making on test scores, but these results are reversed after correcting for the endogeneity of school autonomy. However, results support the role of parental participation in the schools as a positive influence on student achievement.