While credit risk is the obvious and necessary concern of The Journal of Credit Risk, the market turmoil of the last month or so has reinforced in all our minds how perceived credit risk and psychological factors can lead to a liquidity crisis. Credit risk in securitization has been especially the focus of discussion these last few weeks. The securitization market has taken a beating for many reasons. Two of the more important ones are: (i) investors' lack of understanding of credit risk in usual securitization tranches and in CDOs of ABS - investors seem to have been surprised to see widening of spreads, downgrading and principal losses in investor grade mezzanine tranches - they have also unwittingly extrapolated risk in mezzanine tranches to senior tranches; (ii) behavior of status-quo management in the face of potential drop in stock prices - the very institutions that depend heavily on securitization are also ones that have been growing offbalance sheet income rapidly, far in excess of required rate of return on equity - inability to access the securitization market is thus a double whammy for the stock valuation of such institutions - they behave in a manner leading the institution further to the brink of bankruptcy, irrespective of portfolio quality. The first one has everything to do with credit risk in securitization and credit rating of structured products, the second one has nothing to do with them, except via contagion.
With this backdrop in mind, it is certainly topical for The Journal of Credit Risk to solicit articles that address: credit risk in securitization tranches, lack of access to securitization market leading to bankruptcy (a credit event), credit concerns and market contagion leading to liquidity crisis and perhaps a host of related topics. Modeling credit risk in a complex structured credit product is inherently complex. Yet the average investor necessarily looks for a simple characteristic like rating. Perhaps it is meaningful to devise two or three measures or characteristics that an investor can take into account in making a common sense assessment of credit risk in a structured credit, fixed income instrument.
In this issue we present three full-length research papers and one technical report. In the first paper, "Generalized maximum expected utility models for default risk: a comparison of models with different dependence structures," Höcht and Zagst compare the success of alternative default probability prediction methods by choosing the model measure from a one-parameter family of paretooptimal measures defined in terms of consistency with the data and a prior measure. They introduce an alternative model derived from a combination of a wealth maximization strategy and maximum likelihood estimates: Maximum Expected Utility (MEU). The success of default prediction by different methods is compared using a Loans database from Fitch.
The second paper, "Markovian credit transition probabilities under inequality constraints: the US portfolio 1984-2004", by Christodoulakis presents a Bayesian perspective on credit transition probabilities. In the process the author is able to impose non-negativity restrictions on probabilities in the traditional stationary Markovian credit transition probability model. The posterior probability density is computed using Monte Carlo Integration. The methodology developed is then applied to empirically estimate the transition probabilities for the US aggregate data on non-performing loans from 1984 to 2004.
In the third paper, "Default intensity and expected recovery of Japanese banks and the government: new evidence from the CDS market", Ueno and Baba distinguish between default rates and recovery rates implied by CDS using the fractional recovery of face value technique: CDS written on the same reference entity but with different maturities enables one to separately identify default intensities and recovery rates expected by market participants. By assuming an affine process for the default hazard rate, and restricting the recovery rate assumption to a constant (but unknown) recovery of face value, they are able to obtain empirical estimates for hazard rate and recovery rate separately, and both in risk-neutral and in pseudo-actual probabilities.
In the technical report, "Computation of VaR and VaR contribution in the Vasicek portfolio credit loss model: a comparative study", Huang, Oosterlee and Mesters compare four approximate approaches to calculating portfolio credit VaR and marginal VaR contributions for precision, numerical ability, robustness and computational performance. The approximations they consider are: normal approximation, saddle-point approximation, simplified saddle-point approximation and importance sampling.