This paper provides a detailed examination of the cost imposed by thrift institutions resolved during the period 1980–1988. A simple model is presented to explain the cost of resolution. This model is tested empirically with a comprehensive data set that permits us to avoid some of the econometric problems present in earlier studies. The empirical evidence suggests that the model that explains resolution costs in the late 1980s is significantly different from the model for either the middle or early 1980s. This evidence is consistent with the changing nature of the thrift crisis and changes in the regulator's closure rule. Our econometric evidence, moreover, is consistent with the hypothesis that, for troubled institutions, tangible net worth systematically understates market-value net worth. In addition, the importance of including time effects as well as institution effects as determinants of the cost of resolution is revealed.