A comparative examination of bankruptcy prediction: Altman MDA study versus Luther ANN study. A test of predictive strength between the two techniques.

摘要

As early as 16th century BC, business inability to pay their debt has been a factor that was considered as notable concerns. As a result, the bankruptcy laws that were enacted in 1800 in the United States were in response to business inability to pay debt and these laws set the threshold point for what was considered to be reasonable factors when a business is said to be bankrupt.; The painful economic depression with business failures that the country had undergone prompted many scholars to begin to develop tools that could be used to provide warning signal for businesses that were susceptible to failures. Many studies begin with financial ratios as the empirical study conducted by Euclid in the 300 BC did. His study like the ones that followed aimed at providing business the ability to predict the likelihood of failure before they occur.; In the 1930's, Paul J. Fitzpatrick conducted a sample analysis of industrial enterprises in 16 lines of business that failed during the period of 1920 to 1929. In his study, he tested several financial ratios in order to determine if they could be important tools to predict companies' failure.; Several other researches that followed did determine that, in addition to financial ratios, other factors must be weighed even though they might not be quantitative but they could provide clues to business failures. In 1967 and 1968, the works of Beaver and Altman set the stage that led to the opening of a floodgate of literature and the development of several statistical methodologies to help predict with accuracy when firms (companies) were susceptible to bankruptcy ex-ante.; Although these studies revealed that ratios alone could not predict ex-ante but they remained grounded in every study as predictor variables. The emergence of these researches gave businesses the ability which they did not have before, to predict failures. Edward I. Altman's development of a single Z score for predicting companies that were susceptible to bankruptcy to those that were not ex-ante was the pioneer study that used multivariate techniques.; Therein, many other studies have emerged using other statistical techniques in order to confirm that other statistical techniques were more robust and had more predictive strength when predicting business failures.; This study examined multiple discriminant (MDA) technique in Altman Z score theoretical study versus the technique known as artificial neural network (ANN) as studied by Luther in 1998 to determine whether or not there is a difference in the strength of prediction of bankruptcy and robustness using ANN model versus MDA.; A sample of 100 companies was drawn from the Compustat database from 1985 to 1995 with the set up of two hypotheses at an alpha of .05 or 95 percent confidence level. The set of financial ratios that was categorized in five groups in previous two studies mentioned earlier was extended to eight groups and these groups were used as the predictor variables. They were tested and trained under both neuro network and multiple discriminant models.; The results confirmed that the strength of prediction of bankruptcy and the robustness of ANN model did not differ from multiple discriminant analysis (MDA) technique, thus prompting this study to accept both null hypotheses.