AbstractThe transaction price of identical housing units can vary widely due to heterogeneity in buyer and seller preferences, matching, and search costs, generating what we term “markups” above or below the average market price. We measure markups for 3.4 million purchase‐money mortgages and show that they can predict mortgage defaults and credit losses conditional on default even after accounting for collateral coverage (loan‐to‐value ratio) and a comprehensive set of other covariates. The findings suggest that standard collateral coverage estimation may be inaccurate, with implications for both individual and portfolio‐level credit risk assessment.