Simple OLS estimates of the effect of school-imposed penalties for drug use on a student's consumption of marijuana are biased if both are determined by unobservable school or individual attributes. The potential reverse causality is also a challenge to retrieving estimates of the causal...