Performance pay for teachers is frequently suggested as a way of improving education outcomes in schools, but the theoretical predictions regarding its effectiveness are ambiguous and the empirical evidence to date is limited and mixed. We present results from a randomized evaluation of a teacher incentive program implemented across a large representative sample of government-run rural primary schools in the Indian state of Andhra Pradesh. The program provided bonus payments to teachers based on the average improvement of their students' test scores in independently administered learning assessments (with a mean bonus of 30% of monthly pay). At the end of two years of the program, students in incentive schools performed significantly better than those in control schools by 0.28 and 0.16 standard deviations in math and language tests respectively. They scored significantly higher on "conceptual" as well as "mechanical" components of the tests, suggesting that the gains in test scores represented an actual increase in learning outcomes. Incentive schools also performed better on subjects for which there were no incentives, suggesting positive spillovers. Group and individual incentive schools performed equally well in the first year of the program, but the individual incentive schools outperformed in the second year. Incentive schools performed significantly better than other randomly-chosen schools that received additional schooling inputs of a similar value.