Secondary School as a Contraceptive: Quasi-Experimental Evidence from Burundi

It is well-known that more educated women have their first child at later age compared to less educated women. The causality of this relationship and the mechanisms behind it however are another matter. In this paper we use a regression discontinuity design to infer the causal effect of prolonged schooling on the timing of the first child as well as the drivers of the effect. We tracked and interviewed 375 young Burundian women who took part in their countries ‘Concours Nationale’, a nationwide test taken at age 15.5 on average that decides whether or not someone can continue their education. Failure or success in this test strongly affects age at first child. As a lot of girls manage to circumvent the test score cut-off point set by the Ministry of Education, we employ a fuzzy method whereby the assignment into treatment serves as an instrumental variable for effective treatment. We find an ITT of -13 and a LATE of -54 percentage points on the probability to have had a pregnancy four years after the test and of -27 before age 20. An additional year of secondary school reduces the probability to become pregnant by 8 percentage points. We also find evidence for several channels raised in the literature to explain the finding (incarceration, knowledge and modernization). The results are robust to alternative specifications.