Whether allocating more resources improves learning outcomes for students in low-performing public schools remains an open debate. We focus on the effect of increased instructional time, which is theoretically ambiguous due to possible compensating changes in effort by students, teachers or parents. Using a regression discontinuity approach, we find that a reform extending the school day increases math test scores, with a large effect size relative to other interventions. It also improved reading, technical skills and socio-emotional competencies. Our results are partly explained by reductions in home production by students, specialization by teachers and investments in pedagogical assistance to teachers.