Journal of Risk Model Validation

Risk.net

Breaking correlation breakdowns: non-parametric estimation of downturn correlations and their application in credit risk models

Oleg Burd

ABSTRACT

In this paper we provide an approach to integrating downturn correlations into the modeling of portfolio credit risk. The model is set in the wellknown asymptotic single-risk factor framework and is extended by random factor loadings. We show how the downturn correlation can be nonparametrically estimated by bootstrap, and then assess the impact of model misspecification and calibration errors on portfolio risk.