Journal of Credit Risk

Ashish Dev

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

The possible signs of a problem or crisis in a financial sector that could cause systemic effects may be actually easier to fathom than many other non-systemic, stochastic phenomena. It may be as simple as watching three things in conjunction: (i) the speed of growth of all non-niche financial sectors, (ii) the number of major private players in those sectors, and (iii) the extent of leverage used in those sectors.

By 2003, capital markets’ credit products were far from being a niche financial sector and therefore worthy of attention. The total notional amount of credit default swap (CDS) contracts, which was barely US$3 trillion prior to 2003, reached more than US$50 trillion by 2007. Of these, the synthetic collateralized debt obligations (CDOs) (CDS contracts with a super-senior CDO tranche as a reference entity) grew the fastest. During the same period, securitizations in general and CDOs in particular also grew phenomenally. All such products are traded over-the-counter, and the market-makers for a vast majority of deals are just a handful of dealers. CDS has inherent leverage embedded in it because it allows the assumption of a greater amount of credit risk per dollar invested than a cash bond. The synthetic CDOs and in particular their bespoke nature come with an overindulgence of complexity and lack of transparency. In small amounts this is fine and can be lauded as scope for innovation and a paradise for modelers. But in reality, the number of existing claims of CDS far exceeds the number of bonds.

Looking at all these in retrospect, it appears that some blinkers should have gone off well before 2007 as a possible warning of a problem of systemic nature, whether the events of 2007–2008 would have actually occurred or not. If the US sub-prime foreclosures had not spiked up, something else would have catalyzed a systemic problem of similar magnitude. All the supplementary fuel itwould further require is a significant lack of investor confidence. Those of us responsible for risk management in a bank rejoiced at the stellar growth of the CDS market and the structured credit products market in the period 2003–2007 for (i) sourcing of credit risk (much) more broadly beyond banks and (ii) for opportunities for deriving relationship income without creating inordinate “tall trees” of credit exposures through the use of active credit portfolio management. The regulators, however, should have spotted the stellar growth with some suspicion as their first concern is about systemic issues. Of course, looking at total notional amounts of derivative contracts is not an accurate way to assess the true risk. Nevertheless, the high growth rate per year shows up whichever measure is used, including market (or replacement) value. It is also argued that the important dealers in the CDS market have nearly net-flat CDS exposures; that is, the total size of CDS contracts in which protection is sold is just as large as the total size of CDS contracts in which protection is bought. However, both sides involve dealer-to-dealer transactions with associated transaction costs, differentiation across desks and accounting. These often result in compensations that create perverse incentives.

Each of the fewlarge dealers is subject to relatively small market value loss or gain because of the net-flat CDS exposures. But if one of those few dealers disappears, many of the dealerto- dealer CDS contracts of another dealer will be exposed to the loss of protection on one side. The risk exposure will increase immensely until all the CDS contracts are replaced. This is the direct consequence of a few large dealers with exceedingly large notional amounts in both directions with a relatively small net market value.

During its rapid period of growth, CDScontracts have often been highly bespoke in nature, including as many unique characteristics as the two parties entering into a trade come up with. But conceptually there is nothing inherent in the very nature of CDS contracts such that they have to be bespoke. It appears that, after all, the rapid growth of synthetic CDOs or even CDS in general owes its existence more to yield enhancement (it seems, with little questioning as to why the coupon or yield is higher) than anything else. The less-than-adequate understanding of credit risk and arbitrage when dealing with structured tranches as reference entities in CDS contracts was on the part of all concerned – investors, traders, structurers, modelers, issuers, rating agencies and valuation consultants.

From what has been said above, one can argue for the setting up of an exchange for the trading of CDS contracts in order to minimize events with systemic effect. By its very nature, exchange-traded instruments will entail more standardization, but surely life need not have to be so complex. It will certainly create more transparency. Could that exchange slowly move into all structured credit products as well? That is a tall order! Along with changes in the ratings of structured credit products, the setting up of an exchange may be just the trick to bring back the securitization market.

In this issue, two full-length research papers and two technical reports are presented. The first paper “On recovery and intensity’s correlation: a new class of credit risk models” by Gaspar and Slinko develops a reduced-form model of portfolio credit risk where the fundamental determinants of probability of default (PD) and loss given default (LGD) are functions of macroeconomic conditions. There has been substantial empirical documentation of a link between PD and LGD, but theoretical modeling of this phenomenon is relatively recent. In volume 3, issue 4, we presented a paper with a structural form model of portfolio credit risk that explicitly incorporates the correlation between PD and LGD. In this paper, the authors present a class of credit risk models for the term structure of credit spreads of a firm, using the mathematical framework of a doubly stochastic marked point process (DSMPP). This incorporates the realistic phenomenon that when firms default, they are restructured and continue to operate with bondholders accepting a fraction of the original face value as a loss. Within their class of models, the authors show that different assumptions, concerning default intensities, the distribution of LGD and specifically PD–LGD correlation, have a significant impact on the shape of credit spread term structures and, consequently, on the pricing of credit products.

The second paper “Asset correlations and credit portfolio risk: an empirical analysis” is by Duellmann et al. It is a detailed empirical analysis of the estimation of correlations. Correlation is the first major and the only fundamental distinction that separates portfolio credit risk management from traditional credit underwriting. At the same time it is the hardest parameter to estimate even in a relatively simple portfolio credit model. The different Basel II risk weight functions for loan portfolios are nothing but different correlation functions. Therefore, in the implementation of portfolio credit risk modeling and economic capital, asset value correlation measurements are critical. The paper does a thoughtful estimation of asset correlations within a particular data set (Moody’s KMV asset value data) and the resulting impact on creditVaR, particularly with its investigation of temporal and sector effects. This is a primarily empirical paper in which a few key empirical and conceptual points are explored in an efficient way.

This issue has two technical reports. 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 complement rigorous conceptual and model developments presented in the research papers and are of great value to practitioners. The first technical report “Anote on the survival probability in CreditGrades” by Kiesel and Veraart consists of enhancements on the estimation of default probabilities in the CreditGrades model from RiskMetrics, which is used by practitioners in many banks. The report first considers the approach to default probabilities in the CreditGrades in some detail. The authors provide both an approximation to the standard survival probability and the correct exact formula for survival probability in this model. The approximation approach is the same as that used in the CreditGrades model. The numerical properties of the approximated survival probability and the exact formula are then compared, documenting that when using the correct exact formula the two approaches have different numerical properties. This is in contrast to the findings presented in documentation for CreditGrades.

The second technical report “Correlation and asset correlation in the structural portfolio model” is by Frye. In this exposition, the author critiques a very common practice of using asset correlation value as the correlation parameter in a structural form portfolio credit model. Correlation in the structural model ties together events of default, while asset correlation is defined without reference to default. The paper shows that this difference in concept makes possible a meaningful difference in values. It provides an explanation for the fact that estimations of correlation from credit models and default data tend to yield lower values than estimations of correlation using data on firms’ asset values. The author concludes that for predicting the distribution of the default rate, historical default rates may provide a better guide than asset correlation and a chain of assumptions.

You need to sign in to use this feature. If you don’t have a account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an indvidual account here: