Loan Default Correlation Using an Archimedean Copula Approach : A Case for Recalibration

Appropriate modelling of loan default correlation is a prerequisite for effective credit risk management. In this study we provide an insight into the use of the copula functions, particularly addressing the key question of why the Gaussian copula approach failed to detect the high loan default correlation during 2007-08. We discuss other types of copulas which more efficiently capture the estimation of tails with respect to overall credit loss distribution. Our results show that by assuming linear dependence within loan default structures, the correlation modelling of marginal distributions may be mis-specified. We also demonstrate that an Archimedean copula, particularly Gumbel is a more efficient method for capturing tail-dependencies, illustrating the impact of non-linear correlation between fat-tails