Bankruptcy Prediction Models and the Cost of Debt

Abstract

Financial institutions and academic researchers utilize bank - ruptcy prediction models to assess distress risk. However, predicting default can be problematic since 1) few firms actually experience default in any one year, 2) the lag between practical and actual default can vary significantly, 3) firms can strategically default, 4) firms can rework their obligations outside of bankruptcy, and 5) default frequency varies significantly over economic life cycles. Thus, relying on bankruptcy data alone to calibrate and validate these models can be problematic. In this article, the authors take an alternative approach by relying on the firm’s cost of debt as a market proxy for risk. They assess the validity of four widely used distress measures including two accounting-based models (Altman’s Z-Score and Ohlson’s O-Score), one reduced-form model (by Campbell, Hilscher, and Szilagyi, CHS), and one structural distance to default model (Merton-DD). The authors find dramatically different assessment of risk based on the models used. The CHS model has the most significant impact on the cost of debt followed by the Merton model. The accounting-based approaches of Altman’s Z-Score and Ohlson’s O-Score are highly ineffective. They caution researchers using Z- and O-Scores. The Merton distance to default model is superior to both accountingbased models but the authors recommend the use of the CHS model in research studies.

TOPICS: Factor-based models, credit risk management, statistical methods, analysis of individual factors/risk premia