CREDIT RISK

The Two Worlds of Credit Risk - Will They Converge?

Robert Merton's insight into the role of the implicit put on a corporation's assets that the debt holders have written in favor of the equity holders represents another important advance. This theoretical insight was given practical effect by the work of Oldrich Vasicek and his associates at KMV (now part of Moody's.) All this work was essentially micro-focused, concentrating on the potential for default of individual entities. Starting in the 1990s, the credit default swap market created an instrument that provided an estimate of expected losses (PD x LGD) due to default for credit risky assets relative to riskless substitutes. These too offer insights into the default likelihood of individual entities.

The introduction of basket derivatives, CDOs and CLOs presented a quite different challenge. The value of such structures is heavily dependent on co-variation of credit quality across obligors in addition to individual default probabilities. The Gaussian copula model provided a descriptive framework for summarizing the co-variability implicit in the prices of tranched CDOs, but it offered little or no structural explanation of this co-variation. Indeed, structurally inconsistent assumptions about a single underlying parameter are required to match observed CDO prices.

Traditional credit ratings also have played a questionable role in these markets. Intended to calibrate the probability of 100% payment in full of all obligations of an instrument, in this case individual tranches of bespoke CDOs, this calculation is heavily dependent on the behavior of highly uncertain co-variability. Even with unquestioned objectivity in their evaluation, the robustness of such probability estimates is far inferior to that of the default likelihood for a traditional corporate bond. Over-reliance on such estimates by investors has contributed to the still unfolding market turmoil.

Greater transparency driven by more detailed information on the underlying collateral and more intense modeling of structural drivers is a key to improved efficiency in this market. Accomplishing this, however, presents a major challenge in terms of consistent information management and detailed simulation capabilities. How quickly reduced form macro pricing models are supplemented with far more structural micro-analysis of the underlying collateral remains an open question.

Some issues worth discussion:

• Can factor-based models serve to reduce the dimensionality of micro analysis to manageable proportions?
• Should banks limit the volume of any instrument with no second means of valuation?
• Should a robustness estimate accompany standard credit ratings?