Child marriage remains common even where female schooling and employment opportunities have grown. We experimentally evaluate a financial incentive to delay marriage alongside a girls' empowerment program in Bangladesh. While girls eligible for two years of incentive are 19 percent less likely to marry underage, the empowerment program failed to decrease adolescent marriage. We show that these results are consistent with a signaling model in which bride type is imperfectly observed but preferred types (socially conservative girls) have lower returns to delaying marriage. Consistent with our theoretical prediction, we observe substantial spillovers of the incentive on untreated nonpreferred types.