Collateral damage : creating a credit loss model incorporating a dependency between PD and LGD

We apply a credit loss modelling method that incorporates a dependency between the default and loss processes. We use banking data to estimate parameters for the model and perform capital simulations. To this end, we describe and test several different estimation methods. We find that losses upon default are likely influenced by systemic factors, but we find no evidence that these factors correlate simultaneously with the the default process. We do find evidence that there is a 'lagged' correlation between defaults and losses. The systemic effect on loss may raise needed credit capital levels by about 15%.