A nonparametric approach to incorporating incomplete workouts into loss given default estimates

Grazia Rapisarda and David Echeverry

Abstract

ABSTRACT

When estimating loss given default (LGD) parameters using a workout approach, ie, discounting cashflows over the workout period, the problem arises of how to take into account partial recoveries from incomplete workouts. The simplest approach would see LGD based on complete recovery profiles alone. While simple, this approach may lead to data selection bias, which may be the basis of regulatory guidance requiring the assessment of the relevance of incomplete workouts to LGD estimation. Despite its importance, few academic contributions have covered this topic. We enhance the literature by developing a nonparametric estimator that, under certain distributional assumptions on the recovery profiles, aggregates complete and incomplete workout data to produce unbiased and more efficient estimates of mean LGD than those obtained from the estimator based only on resolved cases. Our estimator is appropriate in LGD estimation for wholesale portfolios, where the exposure-weighted LGD estimators available in the literature would not be applicable under Basel II regulatory guidance.