A Flexible Approach to Modeling Ultimate Recoveries on Defaulted Loans and Bonds
by Edward Altman of New York University, and
May 10, 2010
Abstract: We present in this paper an intuitive Bayesian approach to modeling the distribution of discounted ultimate recoveries on defaulted debt using mixtures of distributions. We show that the technique is flexible enough to accommodate important idiosyncratic features of recovery distributions and how to adapt the results to target portfolios whose characteristics do not match that of the estimation sample. These applications do not require strong assumptions about the form of the relation between the characteristics of the borrower, the defaulted debt and recovery outcomes. Our empirical results provide insights to the attributes of debt and conditions at the time of default that are associated with economically important variation in the shape of recovery distributions. Expectations of industry level default conditions at the time of default, in conjunction with the Debt Cushion (an economic measure of debt subordination), prove to be of particular value in modeling portfolio-level recoveries. We benchmark the performance of our model against commonly used alternatives in estimating recoveries on portfolios of defaulted debt. Our simple "mixture-based estimates", adapted to reflect the Debt Cushion and industry-specific distress conditions at the time of default, substantially out-perform competing alternatives.
Keywords: Ultimate Recovery, Credit Portfolio Losses, Mixture of Distributions.