Collection of Financial Statement DataFinancial-statement data of the failed firms were obtained fromMoody'sfor five years prior to failure. The "first year before failure"is defined as that year included in the most recent financial statementprior to the date that the firm failed. Another condition also had to befulfilled—the financial statements could not be more than six monthsold at the date of failure. The "second year before failure" is the fiscalyear preceding the first year. The third, fourtb, and fifth years are simi-larly defined. The financial statements of the nonfailed firms were ob-tained for the same fiscal years as those of their failed mates.The financial-statement data were then grouped according to yearbefore failure. For example, if two firms failed in 1964 and 1954, re-spectively, and their most recent financial statements were prepared onDecember 31, 1963 and 1953, respectively, the first year before failurewould include the 1963 statements of the former and the 1953 state-ments of the latter. The fifth year before failure would include the1959 and 1949 statements. Since the firms failed from 1954 to 1964,inclusive, the first year before failure includes financial statements froman eleven-year period, 1953 through 1963, while the fifth year includes1949 through 1959.The financial-statement data of the nonfailed firms were also strati-fied into years before failure, corresponding to the years that were as-signed to their failed mates. For example, the 1963 (1953) statements ofthe nonfailed mate to the first (second) firm mentioned above would beassigned to the first year before failure. Similarly, the 1959 (1949) state-ments would be assigned to the fifth year before failure.Because financial-statement data were not available for every yearbefore failure the number of observations decreases as the time periodbefore failure increases. The sample size is largest in the first year beforefailure—158 firms—and is smallest in the fifth year—117firms.^*Availa-bility of data over the five-year period was not made a criterion for in-clusion in the sample, since a possible bias might be introduced in this" The number of observations in each year before failure appears in Table A-3 inthe Appendix.