Predicting business failure using cash flow statement based measures

Purpose – Business failures during the economic recession of 2008-2010 years were unusually high in the USA. The purpose of this paper is to build a new model to predict business failure, using mostly cash flow statement based measures as predictor variables and discriminant analysis technique. Design/methodology/approach – The authors' data matrix consisted of 100 firms and seven predictor variables. A total of 50 “failed” firms were matched with 50 non-failed firms according to Standard Industrial Classification (SIC) code and size. Financial statement data for the year prior to failed year were pulled from COMPUSTAT database. Seven predictor variables were selected, namely Operating cash flow divided by current liabilities, Cash flow coverage of interest, Operating cash flow margin, Operating cash flow return on total assets, Earning quality, Quick ratio and Three-year sales growth. The SPSS-19 software was used to perform discriminant analysis (DA). Findings – The DA model classified 83.3 percent of original grouped cases correctly. The cross-validated approach (jackknife or leave-one-out method) correctly classified 79.5 percent of cases. The chi-square test of Wilks' lambda was significant at 0.000 level which means the model as a whole performed very well in predicting business failure. Originality/value – This study is unique in many respects. First, the sample companies are not industry specific. They come from more than 20 different two-digit SIC codes, which means the authors' model is very generic in nature. Second, the seven predictor variables (financial ratios) they selected are logically justified; these are not an outcome of step-wise procedure. Third, most of the predictor variables use operating cash flow information from the cash flow statement. Fourth, all the failed firms in the authors' test sample are from the most recent, 2008-2010, period.