Loss functions for LGD models comparison - HAL-SHS - Sciences de l'Homme et de la Société Accéder directement au contenu
Pré-Publication, Document De Travail Année : 2017

Loss functions for LGD models comparison

Résumé

We propose a new approach for comparing loss given default (LGD) models which is based on loss functions de…ned in terms of regulatory capital charge. These loss functions penalize more the LGD forecasts errors made on credits associated to high exposure and long maturity than the other ones. We also introduce asymmetric loss functions that only penalize the LGD forecasts errors that lead to underestimate the regulatory capital. We show theoretically that the LGD models ranking determined by our approach may di¤er from the ranking obtained according to the traditional approach that consists in comparing the models according to their LGD forecasts errors. Using an original sample of credits and leasing provided by an international bank, we apply this new approach to compare the LGD forecasts issued from 6 competing models. The empirical results con…rm that the ranking based on a naive LGD loss function is generally di¤erent from the models ranking obtained with the capital charge symmetric (or asymmetric) loss.
Fichier principal
Vignette du fichier
LGD_28_avril_2017.pdf (1.08 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Loading...

Dates et versions

halshs-01516147 , version 1 (28-04-2017)
halshs-01516147 , version 2 (27-05-2017)
halshs-01516147 , version 3 (10-01-2018)

Licence

Domaine public

Identifiants

  • HAL Id : halshs-01516147 , version 1

Citer

Jérémy Leymarie, Christophe Hurlin, Antoine Patin. Loss functions for LGD models comparison. 2017. ⟨halshs-01516147v1⟩
823 Consultations
6030 Téléchargements

Partager

Gmail Mastodon Facebook X LinkedIn More