Journal of Credit Risk

Ashish Dev

Practice Leader, ERM and Structured Products Advisory, Promontory Financial, New York

In the recent turmoil in structured credit products markets and losses reported by many financial institutions and hedge funds, it appears that the effects of true credit risk and the effects of accounting procedures have got mixed up. Unambiguously, the credit risk in an equity tranche or the first loss piece is the highest, the credit risk in the mezzanine tranche(s) is next and the credit risk in the senior tranche is the least. Yet, investors in various tranches of a traditional securitization waterfall have reported losses that are not consistent with this risk ranking. This is particularly so in RMBS. Typically, the issuer institution holding the first loss piece does not mark it to market. They take impairment when losses exceed the assumptions in the original valuation. Since delinquencies take time to translate to losses, the typical equity piece holder has sustained little loss per unit capital at risk. Most mezzanine holders and senior tranche holders have marked the investment to market or to model or to a combination of mark-to-model and mark-to-market. Per unit capital at risk, the typical senior (AAA) holder has sustained a higher loss than the typical mezzanine holder. This spectrum of actual loss per unit capital is just the opposite of the true credit risk spectrum. The possible interpretations of the recent fair value accounting rules make matters only worse.

In the long run, accounting rules should not matter, but in the meantime, there has been a lot of unwinding in the name of cleaning up, resulting in losses once and for all. The effect of accounting also seems to throw up in perspective another phenomenon: in the financial world, more money may be made or lost through randomness and psychology than from skills of individuals. In other words, ex post outcome is hardly a measure of knowledge and skill. Instead, understanding of randomness ex ante is the real source of profits in a complex financial market. That is where modeling comes in. Although there is talk about rating reflecting market risk, liquidity risk, etc, it is an open question whether accounting issues need to enter into credit risk modeling.

In this issue, we present four full-length research papers and one technical report. The first paper "Validation of credit rating systems using multi-rater information", by Hornik et al, provides a quantitative approach for comparing credit rating systems across financial institutions. Three aspects of ratings are distinguished: association measures the degree of agreement in relative orderings of the customers; agreement measures the relative frequency of equal rating assignments; and bias measures the average distance between the pairs of rating assignments. Additionally, graphical representations (multi-dimensional-scaling and minimum-spanning-tree techniques) of proximity are suggested. The theoretical part is accompanied by an empirical study that compares rating systems of Austrian banks with respect to the aforementioned measures.

The second paper "Modeling basket credit default swaps with default contagion" is by Haworth and Reisinger. This paper investigates CDS baskets with asset correlation and default contagion. Firm assets are modeled as correlated geometric Brownian motions with exponential default thresholds. The contagion mechanism in this study allows a default event to trigger jumps in the volatility of related firms, which is incorporated into the model to evaluate the spread impact of credit contagion for CDS baskets. Results for two-firm and three-firm baskets are illustrated. The approach is also extended to allow for contagion with decay in the two-firm case. While the interesting theoretical model has some limitations in practical application, it provides a nice way to incorporate the contagion mechanism for risk assessment and CDS basket pricing.

The third paper "Correlation between default events and loss given default and downturn loss given default in Basel II" is by Wan and Dev. Since I am one of the coauthors, the complete editorial and refereeing process has been conducted and decided upon by Arthur Berd, an associate editor of this journal, for which I am very thankful. The concept of downturn loss given default (LGD), introduced late in Basel II, essentially has its roots in the fact that there is a systematic effect on LGD, or alternatively, the LGD is correlated with the event of default. While downturn LGD is to be determined primarily from data during stress periods and will vary somewhat across jurisdictions, it is helpful to have theoretical results based on a sound credit risk model in the background. In this paper, the effect of various levels of PD&-LGD correlation on credit VaR and economic capital is derived. It is shown that the linear formula relating downturn LGD to expected LGD introduced by the US regulators is a surprisingly good approximation except in cases of very low expected LGD.

The fourth paper "Firm value, diversified capital assets, and credit risk: towards a theory of default correlation" by Gru¨ne et al addresses the relationship between probabilistic measures of default and the diversification present within a firm's capital asset portfolio. The underlying concept is the necessary link between a firm's ability to reallocate productive resources from one sector to another with its credit rating and debt profile. The authors employ a dynamic decision approach, which is somewhat distinct from most of the conventional credit default quantification measures. They further show that the same causal explanation so successfully employed in traditional Markowitz-type portfolio diversification strategies can also be expected to generate changes in credit default spreads. The ability of a given firm to maintain productive capacity through well-diversified ownership and, thus, flexible productive capital, is examined and linked to default risk premium.

The last paper in this issue is a technical report. A technical report describes a particular practical technique and enumerates situations in which it works well and others in which it does not. Such reports provide extremely useful information to practitioners in terms of saved time and duplication of efforts. The contents of technical reports compliment rigorous conceptual and model developments presented in the research papers and provide a lot of value to practitioners.

The technical report "Multi-year dynamics for forecasting economic and regulatory capital in banking" is by Rösch and Scheule. This study develops a statistical framework that allows banks to model the credit risk of their risk segments and products point in time for multiple years. This is motivated by the fact that maturities of most credit risky products exceed the usual horizon of one year. The authors investigate the impact of multi-year forecasts of credit risk parameters upon the distribution of future losses to a credit portfolio, and the implications are demonstrated for collateralized debt obligations, using data from the S&P default study. They also investigate how different modeling methodologies, such as the point-in-time and through-the-cycle approach, affect correlation estimates for multiple years.