We consider a college admissions problem with uncertainty. We realistically assume that (i) students' college application choices are nontrivial because applications are costly, (ii) college rankings of students are noisy and thus uncertain at the time of application, and (iii) matching between colleges and students takes place in a decentralized setting. We analyze a general equilibrium model where two ranked colleges set admissions standards for student quality signals, and students, knowing their types, decide where to apply to. We show that the optimal student application portfolio need not be monotone in types, and we construct a robust example to show that this can lead to a failure of assortative matching in equilibrium. More importantly, we prove that a unique equilibrium with assortive matching exists provided application costs are small and the lower-ranked college has sufficiently high capacity. We also provide equilibrium comparative static results with respect to college capacities and application costs. We apply the model to the question of race-based admissions policies
