Systematic Credit Risk: CDX Index Correlation and Extreme Dependence

Abstract

Dependence is an important issue in credit risk portfolio modeling and pricing. We discuss a straightforward common factor model of credit risk dependence, which is motivated by intensity models such as Duffie and Singleton (1998), among others. In the empirical analysis, we study dependence under the risk-neutral measure using credit default swap (CDS) spread data of liquid large-cap U.S. obligors. The proxy for the commonfactor is the DJ CDX.NA.IG index. We document that (i) the CDX factor is significant but has low explanatory power, (ii) factor sensitivities show distinct time-varying nature and that (iii) systematic credit risk shows asymmetric extreme factor dependence, where extreme dependence is present for upward CDX movements only. This finding from an EVT-copula approach is what is predicted by various intensity models of joint defaults.