This paper estimates the effect of exposure to terrorist violence on education. Since terrorists may choose targets endogenously, we construct a set of novel instruments. To that end, we leverage exogenous variation from a local terrorist group's revenues and its affiliation with al-Qaeda. Across several Kenyan datasets we find that attacks suppress school enrolment more than predicted by difference-in-differences-type estimators. This indicates that terrorists target areas experiencing unobserved, positive shocks. Evidence suggests fears and concerns as mechanisms of impact, rather than educational supply.