PT - JOURNAL ARTICLE
AU - van der Voort, Martijn
TI - Factor Copulas
AID - 10.3905/jod.2007.681816
DP - 2007 Feb 28
TA - The Journal of Derivatives
PG - 94--102
VI - 14
IP - 3
4099 - http://jod.pm-research.com/content/14/3/94.short
4100 - http://jod.pm-research.com/content/14/3/94.full
AB - Currently the industry standard model for correlated defaults is the single factor Gaussian copula, in which an issuer's default intensity comes from exposure to a common factor, which gives rise to correlation across issuers, plus an idiosyncratic component. In a rationally Bayesian world, occurrence of a default will raise the posterior expected value of the unobservable common factor, which in turn increases estimates of default correlation among the remaining credits. But this is inappropriate for defaults like those of Enron and Parmalat, that were clearly due to reasons unrelated to any common factor. In this article, van der Voort proposes modifying the Gaussian copula by adding another idiosyncratic term to deal with such cases. He shows that the model extended in this way is distinctly better able to capture the correlation skew in synthetic CDO tranche prices.