This study examines the usefulness of general price level information (GPL) and current cost information (CC) originally provided by SFAS No. 33 as compared to historical cost information (HC) in predicting bankruptcy. The study also examines the usefulness of GPL data versus CC data when each supplements HC data. In addition, this study tests the usefulness of the three types of information systems combined in one model (HC, GPL, and CC) versus HC data in predicting bankruptcy. The study focuses on the predictability of business failure using financial ratios as predictors. A comparison of these predictors is made in order to identify the accounting system that yields a better prediction of bankruptcy. Two multivariate statistical techniques, multiple discriminant analysis (MDA) and logistic regression analysis (LRA), are used to derive the ex—post classification and the ex-ante prediction results. Six functions are developed, based on ratios computed with HC, CC, GPL, the combined HC and GPL, the combined HC and CC, and the combined HC, GPL, and CC. The resulting functions are used to classify 40 firms as failed or nonfailed. The analysis is repeated for three time bases—one, two, and three years before bankruptcy. The main results of the various analyses indicate that the combined HC and CC model has more discriminant power than does the HC, the GPL, or the combined HC and GPL models in each of the three years before bankruptcy. The results also show that there are significant differences in the overall classification rate derived from the combined HC, GPL, and CC model and the HC model, the GPL model, or the combined HC and GPL model . The differences between the combined HC and CC and the combined HC, GPL, and CC models are not significant in each of the three years before bankruptcy. The results also indicate that the differences in the the performance of MDA and LRA are not significant except in the second year before bankruptcy when the combined HC and GPL model is used.