This paper models decisions to apply to and attend charter schools in Boston using a generalized Roy selection framework linking preferences to the achievement gains generated by charter attendance. The model is estimated with instruments based on randomized admission lotteries and distance to charter schools. Charter schools generate larger gains for disadvantaged students, but demand for charters is stronger among more advantaged students. Similarly, gains are inversely related to unobserved preferences for charters. As a result, counterfactual simulations indicate that charter expansion is likely to be most effective when accompanied by efforts to target students who are unlikely to apply.