Modeling Expected Loss with Unobservable Heterogeneity

Abstract

In this paper we model the expected loss over multi-year horizons. Investors usually have only incomplete information about the true state of a firm. To model this form of unobservable heterogeneity, we introduce a latent non-negative random variable into the model specification of the probability of default and the recovery rate. To estimate the expected loss over arbitrary horizons, we estimate the coefficients for the stochastic processes describing the covariates for the probability of default for each obligor and the loss given default. We provide an analysis of both in and out of sample performance of our estimates of the probability of default and the loss given default.

Keywords: Default, Bankruptcy, Recovery Rate, Frailty, Unobservable Heterogeneity

JEL Classification: G33, C41