A Comparison of Corporate Bankruptcy Models in Australia: The Merton vs. Accounting-based Models

Actuaries have long employed logistic type regression models in their analysis of renewal rates for property and casualty insurance products. This paper introduces an application of such methodology to the prediction of corporate bankruptcy. This is an example of a widerfield area of endeavor where actuaries have the potential to add real value. The results presented in the paper have implications for levels of risk-based capital to be held by insurers and other financial organizations. This paper examines how effectively the default likelihood indicator (DLI) estimated from the Merton model can predict corporate bankruptcy in Australia during 1990-2003. In addition, the performance of the Merton model and the three most popular bankruptcy models, i.e., Altman (1968), Zmijewski (1984), and Shumway (2001) are also compared. Our findings suggest that the Merton model is the most informative model in explaining corporate bankruptcy, followed by the Shumway model. Among accounting and market-based variables in bankruptcy models, only two variables, namely the ratio of total liabilities to total assets (TL/TA) and the idiosyncratic standard deviation of stock returns are most significant in predicting corporate bankruptcy. Finally, the results from our comparative study of bankruptcy prediction models suggest developing a multivariable logistic regression model which includes both financial ratios and Merton's default likelihood indicator as predictors. We assess the predictive ability of this model by comparing 95% confidence intervals for the predicted probability of default for firms that default and firms that are not observed to default.