A PROBABILISTIC MODEL OF FINANCIAL DISTRESS

The objectives of this study were specified in Chapter 1 as: (1) To develop a financial model to assess the vulnerability of firms to failure, and to determine the financial variables relevant for distinguishing failing from nonfailing firms. (2) To evaluate the model as a financial information system, by evaluating the information content of its predicted probabilities, using information theory. After assessing the economic impact of financial distress in Chapter 1, previous empirical and theoretical analyses of the problem were discussed in Chapters 2 and 3. Analysis of the empirical literature began with Beaver\u27s 1965 univariate study of the ability of financial ratios to predict business failure. Altman attempted to improve on Beaver\u27s results by incorporating a large number of ratios in a multiple discriminant analysis model. His classification error rate for the first year prior to failure is an improvement over Beaver\u27s, but error rates for earlier years are significantly worse. Although many later studies have attempted to improve on Altman\u27s technique, no improvements were found over Beaver\u27s error rate. Moreover, one cannot evaluate the significance of individual variables with this technique. Thus, the determination of relevant variables is impossible, and this is one of the primary purposes of exploratory empirical analysis. The traditional theoretical models have extended the Miller-Modigliani theory of capital structure to show that when imperfect arbitrage opportunities and nonhomogeneous expectations exist, nontrivial financial risk implies an optimal capital structure. The use of debt is expanded exactly to the point where benefits to leverage are offset by bankruptcy costs. The non-traditional approach emphasizes costs in addition to the small direct costs of bankruptcy. These include agency costs, such as the costs of monitoring to protect the interests of parties with claims to assets. If investors do not have perfect information concerning the true characteristics about the firm, financial variables may also be a signal to investors as to the magnitude and riskiness of the firm\u27s future income stream. Many ratios are potentially important signals. The importance of these variables has not been empirically determined, since the coefficients in the discriminant analysis models cannot be individually evaluated. This determination is necessary if this line of theoretical research can be effectively developed. Chapter 4 reported a model which assessed the vulnerability of firms to failure, conditional on the attribute vector of the firm. Maximum likelihood estimation of probit and logit models was performed for each of the five years prior to failure. Each model had a very high R(\u272) value and was also significant at greater than the .995 level by the likelihood ratio test. The sign of and significance of coefficients for the variables were largely consistent with a priori expectations. Information theory was used in Chapter 5 as a method for evaluation of the model\u27s predictions as signals from an information system. Using information theory, it is possible to objectively measure the quantity of information in a signal without reference to individual preferences. The quantity of information in a message is measured in terms of its ability to reduce the uncertainty over the occurrence of an event. The quantity of information provided by the logit model was found to be greater than that reported in the only existing comparable study. This technique was shown to be generalizable to the comparison of the information content of alternative information systems. This methodology will provide rankings of information systems consistent with rankings derived from individual preferences, if the cost of the information systems can be ignored