Why do parents invest so much in their children’s college education? I propose that college serves as a commitment device for altruistic parents facing a Samaritan’s dilemma. In a dynastic model where parents and adult children interact non-cooperatively, children who anticipate parental support under-save and over-consume. College mitigates this problem from the parent’s perspective: by permanently raising the child’s income and making shocks less persistent, it shrinks the set of states where parents optimally transfer, reducing the lifetime costs of moral hazard. However, the net effect on college attendance is theoretically ambiguous: anticipated transfers subsidize enrollment, raising attendance, but also cushion non-college children against income risk, reducing their incentive to invest in human capital. I document four facts from matched parent-child data (PSID and HRS) consistent with this mechanism: parents whose children attend college consume 8% more, accumulate wealth differently, face a substantially weaker precautionary saving motive, yet transfer only modestly less—suggesting that the option value of forgone transfers, rather than realized transfer flows, drives parental behavior. The structural model, estimated by the simulated method of moments, quantifies these opposing forces and shows that the sign of the moral hazard effect on college varies across the ability–wealth distribution.