We investigate whether collateral helps to solve adverse selection problems. Theory predicts a negative relationship between presence of collateral and risk premium, as collateral constitutes a signalling instrument for the borrower to be charged with a lower risk premium. However, bankers' view and most empirical evidence contradict this prediction in accordance with the observed-risk hypothesis. We provide new evidence with loan-level data and country-level data for a sample of 5843 bank loans from 43 countries. We test whether the degree information asymmetries affects the link between the presence of collateral and risk premium. We include five proxies for the degree of information asymmetries, measuring opacity of financial information, trust, and development. We find that a greater degree of information asymmetries reduces the positive relationship between the presence of collateral and the risk premium. This finding provides support for the adverse selection and observed-risk hypotheses, as both hypotheses may be empirically validated depending of the degree of information asymmetries in the country.