Quantifying Credit Risk I: Default Prediction

Abstract

Until the 1990s, corporate credit analysis was viewed as an art rather than a science because analysts lacked a way to adequately quantify absolute levels of default risk. In the past decade, however, a revolution in credit-risk measurement has taken place. The evidence from this research presents a compelling case that the conceptual approach pioneered by Fischer Black, Robert Merton, and Myron Scholes provides a powerful practical basis for measuring credit risk. “Quantifying Credit Risk II: Debt Valuation” shows that their approach provides superior explanations of secondary-market debt prices.

Until the 1990s, corporate credit analysis was viewed as an art rather than a science because of analysts' inability to adequately quantify absolute levels of default risk. More than 30 years ago, however, Fischer Black and Myron Scholes proposed that one could view the equity of a company as a call option. As subsequently elaborated by John Cox, Robert Merton, and others, this approach has come to be called “the Merton model.” In the 1990s, an empirical implementation of this approach developed by Oldrich Vasicek and the author (the KMV model) has enjoyed considerable success with practitioners. This article, the first of a two-part series, focuses on empirical tests of this approach for predicting default.

The article begins with a description of how the KMV model differs from the canonical Merton model in the academic literature. The main distinctions of the KMV model are its focus on the probability of default of the company as a whole, rather than valuation of the debt, and its use of an empirically determined default frequency distribution. It retains the main virtues of the Merton approach, however, in that it is a cause-and-effect model of default that transforms equity market prices into information on the credit quality of companies.

The article then turns to a description of the tests of the power of the KMV model versus agency debt ratings. Two kinds of tests and their results are described. First, power curve analysis shows that the outputs of the KMV model (“EDFs” or expected default frequencies) consistently outperform agency debt ratings in correctly predicting default—for any level of incorrect predictions.

Power curve analysis does not rule out, however, that in some cases, the agency debt ratings may do a better job of predicting default than the KMV analysis. Therefore, a second test, called “intracohort analysis,” is performed to evaluate the marginal information content of ratings relative to the KMV approach and vice versa. The intracohort analysis reveals that there is essentially no information in ratings that is not already captured in the EDFs but considerable information in EDFs that is not captured in ratings. “Noise” in equity prices does not cause overprediction of default risk.

The theoretical approach of statistical methods of default prediction as it relates to the performance of the KMV model is also discussed. Because correlation does not equal causation, statistical methods should be expected to add little power to the default–prediction power of the KMV model.

The findings discussed suggest that the Black–Scholes–Merton approach, appropriately executed, provides the long-sought quantification of credit risk. As an objective, cause-and-effect model, this approach also gives analytical insights into corporate behavior, thus creating the basis for a continuing rich research program into default risk. For instance, the link between default probability and equity values provides an understanding of the correlations within corporate debt portfolios as well as correlations between equity and corporate debt portfolios. Part II of “Quantifying Credit Risk” will examine application of the model to the valuation of corporate debt.