The financial crisis facing the United States savings and loan industry has been steadily escalating over the last decade. During this time, accounting treatments concerning various thrift institution transactions have also attracted a great deal of attention. The specialized accounting treatments used in the thrift industry, known as regulatory accounting practices (RAP) have been blamed as one of the culprits hindering the regulators' ability to detect serious financial problems within many institutions. Accordingly, RAP was phased out, and all federally insured savings and loan associations began preparing their financial statements in accordance with generally accepted accounting principles (GAAP) as of January 1, 1989. The purpose of this dissertation is to compare the relative predictive values of the two historical cost based accounting conventions (RAP and GAAP) available to the savings and loar? industry during the 1980's. For purposes of this dissertation, predictive value is defined as the usefulness in assessing future financial health and viability. The sample consisted of all the institutions reporting to the Federal Home Loan Bank of Dallas between 1984 and 1989. Year-end thrift financial report data, obtained from Sheshunoff Information Services, Inc. (Austin, Texas) was used to calculate several financial ratios. The Federal Home Loan Bank of Dallas provided a comprehensive listing of all institutions that failed between January 1, 1985 and March 31, 1989. The null hypothesis tested in this study was: no significant differences existed between the predictive values of RAP and GAAP financial statements. Using a dichotomous dependent variable (failed/not failed) and independent variables from prior research, several multinomial logistic models were developed to test the null hypothesis. All models developed failed to reject the null hypothesis.