Students in private schools routinely outperform those in public schools both in the United States and around the world. But do private schools make students better or do they simply cream skim better students? In this article I take advantage of the remarkable fact that in many districts in India a majority of students attend private schools. As the private share of school enrollment increases, cream skimming becomes less plausible as the explanation for a higher rate of achievement in private schools. Evidence for cream skimming is found when the private share of schooling is low, in the range of 0–15%, and thus private schools have a large public pool from which to skim. But the private effect on achievement does not appear to diminish greatly even in districts where more than 70% of students are in private schools. Most importantly, mean scores taken over the entire population of students, private and public, increase with the share of private schooling. These findings support a significant productivity effect of private schools. (JEL I25, I2, L33)