We exploit these two margins using a regression discontinuity setup and find that the reform reduced the student-teacher ratio by around 10% (from a base level of 22 students per teacher), with no crowding out of other school resources or parental support. However, the reform did not improve test scores and longer-term academic outcomes, and we can reject even small positive effects. We do find that the reform improved the school environment from the students’ perspective, but with the largest impact on aspects most weakly associated with better academic outcomes.