We evaluate the causal effects of a program that constructed high quality "girl-friendly" primary schools in Burkina Faso, using a regression discontinuity design 2.5 years after the program started. We find that the program increased enrollment of all children between the ages of 5 and 12 by 20 percentage points and increased their test scores by 0.45 standard deviations. The change in test scores for those children caused to attend school by the program is 2.2 standard deviations. We also find that the program was particularly effective for girls, increasing their enrollment rate by 5 percentage points more than boys', although this did not translate into a differential effect on test scores. Disentangling the effects of school access from the unique characteristics of the new schools, we find that the unique characteristics were responsible for a 13 percentage point increase in enrollment and 0.35 standard deviations in test scores, while simply providing a school increased enrollment by 26.5 percentage points and test scores by 0.323 standard deviations. The unique characteristics of the school account for the entire difference in the treatment effect by gender.