This paper presents evidence from the first three years of a randomized controlled trial of a government-administered pilot teacher performance pay program in Punjab, Pakistan. The program offers yearly cash bonuses to teachers in a sample of public primary schools with the lowest mean student exam scores in the province. Bonuses are linked to three school-level indicators: the gain in student exam scores, the gain in school enrollment, and the level of student exam participation. Bonus receipt and size are also randomly assigned across schools according to whether or not the teacher is the school's head. On average, the program increases school enrollment by 4.1 percent and student exam participation rates by 3.4 percentage points, both in the third year. The analysis does not find that the program increases student exam scores in any year. Mean impacts are similar across program variants. The positive mean impact on school enrollment is mainly seen in urban schools and the positive mean impact on student exam participation rates is only seen in rural schools.