Dependence of defaults and recoveries in structural credit risk models

The current research on credit risk is primarily focused on modelling default probabilities. Recovery rates are often treated as an afterthought; they are modelled independently, in many cases they are even assumed to be constant. This despite their pronounced effect on the tail of the loss distribution. Here, we take a step back, historically, and start again from the Merton model, where defaults and recoveries are both determined by an underlying process. Hence, they are intrinsically connected. For the diffusion process, we can derive the functional relation between expected recovery rate and default probability. This relation depends on a single parameter only. In Monte Carlo simulations we find that the same functional dependence also holds for jump-diffusion and GARCH processes. We discuss how to incorporate this structural recovery rate into reduced-form models, in order to restore essential structural information which is usually neglected in the reduced-form approach.

► Functional relation between default and recovery rates in the Merton model ► This relation depends on a single parameter only. ► It is valid for correlated diffusion, jump-diffusion and GARCH. ► First empirical evidence: default and recovery data from Moody's annual study ► We suggest this relation as an alternative to reduced form recovery models.