We consider the effect of legal access to alcohol on student achievement. We first estimate the effect using an RD design but argue that this approach is not well suited to the research question in our setting. Our preferred approach instead exploits the longitudinal nature of the data, identifying the effect by measuring the extent to which a student's performance changes after he gains legal access to alcohol, controlling flexibly for the expected evolution of grades as students make progress towards their degrees. We find that students' grades fall below their expected levels upon being able to drink legally, but by less than previously documented. We also show that there are effects on women and that the effects are persistent.