Article dans une revue: The empirical evidence that enrollment in higher education is constrained by access to credit is limited and usually indirect. We use a regression discontinuity design based on the fact that student loans are granted according to a score threshold at a South African credit institution (Eduloan) providing short-term loans at market conditions: we find that the credit constraint is substantial, as it reduces enrollment by more than 40 percentage points in a population of mostly middle-class applicants. However, this effect is entirely concentrated on women, and women granted a loan catch up with men's enrollment levels. This heterogeneity is not explained by lower incomes in the sample of women. It implies that women have lower access to credit, or that their options for managing without a credit are more limited than men's.