Corporate bankruptcy and prediction: An analysis of multi-discriminant, logit and survival models using the statement of cash flows

This study will compare and contrast three discriminating techniques, multidiscriminant analysis, logit and survival analysis, on their ability to predict corporate bankruptcy. It will evaluate these models over an 18-year period rather than the 6–12-years used in most previous studies. This time period includes an entire business cycle. The analysis will also investigate whether variables constructed from the statement of cash flows as well as macroeconomic variables add predictive ability to the models. The predictive models used are all combinations of models previously used in the literature with cash flow and macroeconomic variables added where appropriate. The results show the multiple discriminant technique is the most consistent and performs the best on the holdout sample. Macroeconomic variables add little to the predictive ability of the models. Variables constructed from the statement of cash flows perform quite well as stand-alone variables but add only marginally to models with accrual accounting variables