Bank Asset Risk: Evidence from Early-Warning Models

Using alternative measures of risk can increase the predictive accuracy of early-warning models of U.S. bank failures. However, the coefficients and the predictive accuracy of these revised models are not stable over time. Exploring the possibilities for accurately measuring risk using available accounting data involves examining the performance of a number of risk measures in early-warning models. An alternative measure improves model estimates for the United States in the mid-1980s, but the improvement deteriorates during the later stages of banking difficulties. The time specificity of the early-warning models affects their usefulness for bank policy and supervision. Copyright 1995 Western Economic Association International.