Abstract We ask whether the correlation between mortgage leverage and default is due to moral hazard (the causal effect of leverage) or adverse selection (ex ante risky borrowers choosing larger loans). We separate these information asymmetries using a natural experiment resulting from the contract structure of option adjustable-rate mortgages and unexpected 2008 divergence of indexes that determine rate adjustments. Our point estimates suggest that moral hazard is responsible for 40$\%$ of the correlation in our sample, while adverse selection explains 60$\%$. We calibrate a simple model to show that leverage regulation must weigh default prevention against distortions due to adverse selection.