DRAFT 1.1 Parameter Uncertainty and the Credit Risk of Collateralized Debt Obligations

1The views expressed here are those of the author and do not reflect the opinions of the Federal Reserve Board or its staff. This paper is still very much a work in progress. If you would like to cite or distribute it, please contact the author beforehand, as a revised draft will be available shortly. This paper examines the empirical difficulties inherent in assessing the credit quality of collateralized debt obligations (CDOs). Because of the way CDO liabilities are structured, CDO note payouts are sensitive to tail collateral loss events. As a result, in order to assess the likelihood and severity of a CDO note’s losses, one needs to know the distribution of losses for each collateral asset, as well as the dependence of losses across collateral assets. In practice, CDO collateral losses are most commonly modeled using normal copulas. I show that for more senior CDO tranches, standard credit risk metrics such as probability of default, expected loss, and conditional expected loss are highly sensitive to model parameters which must be estimated or specified judgmentally by a credit analyst. Given assumptions about the historical data available to a credit analyst, I compute bounds on the accuracy of normal copula parameter estimates and show that in applied settings data constraints are likely to impose severe limitations on an analyst’s ability to accurately evaluate CDO tranches. Thus, CDO note credit ratings should generally be viewed as more preliminary and less informative than comparable corporate bond ratings.