We use unique survey data linked to nearly a decade of administrative welfare data to examine the relationship between early marijuana use (at age 14 or younger) and young people's educational outcomes. We find evidence that early marijuana use is related to educational penalties that are compounded by high-intensity use and are larger for young people living in families with a history of welfare receipt. The relationships between marijuana use and both high school completion and achieving a university entrance score appear to stem from selectivity into the use of marijuana. In contrast, early marijuana use is associated with significantly lower university entrance score for those who obtain one and we provide evidence that this effect is unlikely to be driven by selection. Collectively, these findings point to a more nuanced view of the relationship between adolescent marijuana use and educational outcomes than is suggested by the existing literature.