Ensemble Predictions of Recovery Rates
- 372 Downloads
In many domains, the combined opinion of a committee of experts provides better decisions than the judgment of a single expert. This paper shows how to implement a successful ensemble strategy for predicting recovery rates on defaulted debts. Using data from Moody’s Ultimate Recovery Database, it is shown that committees of models derived from the same regression method present better forecasts of recovery rates than a single model. More accurate predictions are observed whether we forecast bond or loan recoveries, and across the entire range of actual recovery values.
KeywordsRecovery rate Loss given default Forecasting Ensemble learning Credit risk
JEL ClassificationsG17 G21
I would like to thank João Andrade e Silva, Telmo Pina e Moura, Joaquim Ramalho and an anonymous referee for their valuable comments and suggestions. All the remaining errors are mine. Financial support from FCT - Fundação para a Ciência e Tecnologia (grants PTDC/EGE-ECO/119148/2010 and UTA_CMU/MAT/0006/2009) is also gratefully acknowledged.
- Basel Committee on Banking Supervision (2006) International convergence of capital measurement and capital standards. Bank for International SettlementsGoogle Scholar
- Breiman L, Friedman JH, Olshen RA, Stone CJ (1984) Classification and regression trees. Wadworth International Group, Belmont, CaliforniaGoogle Scholar
- Breiman L (1996) Bagging predictors. Mach Learn 24:123–140Google Scholar
- Emery K, Cantor R, Keisman D, Ou S (2007) Moody’s Ultimate Recovery Database. Moody’s Investors ServiceGoogle Scholar
- Gupton GM, Stein RM (2005) LossCalc V2: dynamic prediction of LGD. Moody’s Investors ServiceGoogle Scholar
- Witten IH, Frank E (2005) Data mining: practical machine learning tools and techniques. Morgan Kaufmann PublishersGoogle Scholar