Title: Models of Corporate and Bank Default and Credit Migration
Authors: DIMOU PARASKEVI
Publication Year: 2007
JRC Publication N°: JRC42343
URI: http://publications.jrc.ec.europa.eu/repository/handle/JRC42343
Type: PhD Theses
Abstract: This thesis presents three studies on credit risk modelling. The first study compares the real default probabilities produced by three main structural models of default, Merton model, Longstaff and Schwartz model and Leland and Toft model, to the observed real default probabilities reported by Moody¿s for the BBB, BB and B rated bonds. We find that none of the models can accurately predict the default probabilities in all these cases. Merton as well as Leland and Toft models underpredict default probabilities. Longstaff and Schwartz model although it produces in some cases Expected Default Frequencies (EDFs) that are close to the observed ones, it tends to overestimate the default probabilities of riskier bonds as well as the default probabilities of bonds with the same rating but higher equity volatility. We also find that structural models tend to underestimate the default probabilities in early years. The second study examines whether information from equity markets, as summarized in the distance to default measure derived from a Merton-Moody¿s KMV (MKMV) model, provides useful additional information over accounting variables for predicting changes in bank credit ratings. Using a dataset of 98 equity listed banks from 1997 to 2004, we find that distance to default measure has additional explanatory power for modeling current ratings, or predicting credit rating changes over a 6-month or 12-month horizon, but only for the smaller sized banks. We find no evidence that changes in distance-to-default have additional explanatory power for predicting rating categories, regardless of the size of the bank. The third study compares two proprietary models, Moody¿s KMV (MKMV) and BARRA models that use information from the equity and debt market respectively for the estimation of market implied ratings that can be updated continuously. We compare the empirical performance of these models in terms of their ability to predict in a timely fashion changes in credit quality by employing a sample of 4594 bonds issued by 447 firms from US for a period of 3 years. We find that neither model provides a close mapping to observed ratings. Both however are useful for prediction of credit transitions.
JRC Institute:Institute for the Protection and Security of the Citizen

Files in This Item:
There are no files associated with this item.


Items in repository are protected by copyright, with all rights reserved, unless otherwise indicated.