Abstract: Growing evidence suggests that teachers in developing countries often have weak or misaligned incentives for improving student outcomes. In response, policymakers and researchers have proposed performance pay as a way to improve student outcomes by tying concrete measures like achievement scores to teacher pay. While evidence from randomized experiments generally indicates that performance pay programs are effective at improving student achievement in developing countries, there has been considerable variation in how much these programs affect student achievement. The goals of this study are to: (1) examine the impacts of different teacher performance pay designs on student achievement, both for the average student and for students across the baseline achievement distribution; and (2) examine the mechanisms through which different teacher performance pay designs affect student achievement (for the average student and for students across the baseline achievement distribution). The sample includes a total of 8,892 students and their grade 6 mathematics teachers from 216 schools from 16 nationally-designated "poverty" counties in Yulin Prefecture (Shaanxi Province) and Tianshi Prefecture (Gansu Province) in rural, northwest China. To test the impacts of the different teacher performance pay designs, researchers designed a cluster-randomized controlled trial. In this trial, schools were randomly allocated to 4 different treatment arms: (1) control--no teacher incentive pay; (2) levels incentive--performance pay contract stipulating rewards based on student achievement levels on endline tests; (3) gains incentive--performance pay contract based on student achievement gains from baseline and endline tests; and (4) pay-for-percentile incentive--performance pay contract stipulating rewards based on student growth percentiles. Surveys were used to collect information from the students, teachers, and school administrators. Findings reveal that: (1) Only "pay-for-percentile" incentives had a positive, statistically significant impact on average student achievement; (2) Teacher incentives based on "levels" or "gains" were ineffective; (3) "Gains" incentives led teachers to only focus on certain types of students, which led to negligible learning (on average) across all students; and (4) Pay-for-percentile incentives led to score gains across all students (on average). The results of this study may have important implications for how Teacher Performance Pay Policy can be implemented in China and in other developing countries.