CDO rating methodology: Some thoughts on model risk and its implications
BIS Working Papers No 163
Rating collateralised debt obligations (CDOs), which are based on tranched pools of credit risk exposures, does not only require attributing a probability of default to each obligor within the portfolio. It also involves assumptions concerning recovery rates and correlated defaults of pool assets, thus combining credit risk assessments of individual collateral assets with estimates about default correlations and other modelling assumptions. In this paper, we explain one of the most well-known models for rating CDOs, the so-called binomial expansion technique (BET). Comparing this approach with an alternative methodology based on Monte Carlo simulation, we then highlight the potential importance of correlation assumptions for the ratings of senior CDO tranches and explore what differences in methodologies across rating agencies may mean for senior tranche rating outcomes. The remainder of the paper talks about potential implications of certain model assumptions for ratings accuracy, that is the "model risk" taken by investors when acquiring CDO tranches, and whether and under what conditions methodological differences may generate incentives for issuers to strategically select rating agencies to get particular CDO structures rated.
Keywords: Collateralised debt obligations, credit risk modelling, rating agencies.
JEL classification numbers: C15, G11, G15, G20.