This paper offers a possible explanation for the conflicting results in the literature concerning the empirical relation between collateral and loan risk. We posit that differences in collateral characteristics, such as liquidity, may be associated with the empirical dominance of different risk-collateral relations implied by economic theory. Using credit registry data and a novel identification strategy to control for borrower and lender selection effects allows us to differentiate between the ex ante and ex post theories of collateral. We find that collateral overall is associated with lower risk premiums and higher default rates. The results indicate an important role for collateral in mitigating losses and reducing risk-taking incentives, consistent with ex post theories. Liquid collateral is associated with especially low risk premiums, and these loans perform better than those with illiquid collateral or no collateral. We also find that individual collateral types exhibit significant variation in terms of risk-collateral relations, with some consistent with ex ante theories and others with ex post theories. Our results suggest that the conflicting results in the literature may occur because different samples may be dominated by different types of collateral with different economic characteristics.