Using Structural Models for Default Prediction

Abstract

I propose a new procedure for extracting probabilities of default from structural credit risk models based on virtual credit spreads (VCS) and implement this approach assuming a simple Merton (1974) model of capital structure. VCS are derived from the increase in the payout to debtholders necessary to offset the impact of an increase in asset variance on the option value of debt and equity. In contrast to real-world credit spreads, VCS do not contain risk premia for default timing and recovery uncertainty, thus yielding a purer estimate of physical default probabilities. Relative to the Merton distance to default (DD) measure, my measure (i) predicts higher credit risk for safe firms and lower credit risk for firms with high volatility and leverage (ii) requires fewer parameter assumptions (iii) clearly outperforms the DD measure when used to predict corporate default.