Journal of Credit Risk

Ashish Dev

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

Policymakers on both sides of the Atlantic, who deal with securitization, seem intent on loan originators retaining a percentage of the packaged securitized product for asset-backed securities, and probably for commercial mortgage-backed securities too. This concept of risk retention has come to be known as "skin-in-the-game" in the popular financial press. The logic behind requiring such risk retention is that issuers retaining a material risk of loss on securitized loans have a strong incentive to ensure that the loans are soundly underwritten at origination. In other words, it serves to better align the originator's risk with that of the investors purchasing the securitized assets. Underwriting has been a traditional credit risk management issue. But when it comes to risk retention in a securitized structure, what seems to matter just as much, if not more, are portfolio credit risk characteristics.

The extent of risk retention is proposed to be 5% in the European Union while US policy proposals put the figure at between 5% and 10%. In principle, the percentage can be a vertical slice across all the tranches or it can be a horizontal equity slice, which has two obvious problems. First, the one-size-fits-all percentage does not depend on the structure or on the diversification of the underlying collateral pool. Second, a 5% retention vertically across all the tranches represents very different skin-in-the-game than a 5% retention of a horizontal equity slice does.

Not too many years ago, the issuer would typically retain the first-loss-piece of any securitization. In terms of percentage, such first-loss-pieces would be about the same as the internal economic capital attributable to the portfolio of loans constituting the underlying collateral. Some banks would even compute economic capital for such off-balance-sheet exposure as if the portfolio were never securitized. In the case of credit card securitization, the issuer replaces receivables with new ones till almost the end date of the securitization, effectively retaining all the risk. Given the fact that the issuer will be compensated for the parts of the securitization it retains, either in the shape of coupons or in the shape of residual interest, not to speak of associated fee income, the risk-retention proposal is not necessarily bad for the issuers.

However, coupled with the skin-in-the-game proposals have come new accounting rules (FAS 166 and 167) in the US. These rules may indeed require that as well as retaining a portion of the risk, the issuer will not be able to treat the securitization as off balance sheet. Even after securitization, the underlying collateral pool will sit on the balance sheet, against which the issuer will have to hold capital for the full amount of the assets.

Capital relief to issuers is not the only benefit of securitization. It is not even one of the most fundamental efficiency benefits of the innovation of securitization. Those are: sourcing of credit risk beyond the banking sector; monetization of relatively illiquid financial assets, creating a source of funding; and enabling investors of various levels of risk aversion to chose from a spectrum of credit risk even though the underlying collateral pool may be relatively homogeneous in terms of credit risk. Overall, the existence of the securitization market should lower the cost of borrowing in general.

Therefore, with or without a skin-in-the-game provision and with or without capital relief to issuers, it is important that the securitization market is restored. More than two years into the credit crisis, hardly any non-agency residential mortgage securitization or commercial mortgage-backed securitization has taken place. The imperative of having a vibrant securitization market for the flow of credit, and more generally for economic growth, may be such that the constraints being proposed will have to be given up altogether.

Portfolio credit modeling does have a significant role to play in the risk-retention issue. The exact extent of risk retention that makes economic sense can only be estimated through modeling. The same modeling process that determines the rating of a tranche in securitization does not seem to be adequate for the purpose of determining what the appropriate risk retention should be.

In this issue we present three full-length research papers and one technical report. The first paper, "Credit-migration risk modeling", is by Andersson andVanini. The paper provides a new approach to the quantitative modeling of rating migrations that can be used to price credit derivatives based on credit transition matrices. The model constructs a point-in-time rating transition matrix. The authors show that traditional affine Markov chain models are not sufficient to generate meaningful point-in-time migration matrices in either an economic boom or a contraction. The authors introduce parameters reflecting rating direction and speed into their model, replacing the notion of rating drift, and then use a regime-shifting Markov mixture model to obtain suitable time-varying migration matrices that fit well with point-in-time data. The extended framework provides an analytical pricing formula for credit defaults swaps.

The second paper, "Extracting systematic factors in a continuous-time credit migration model", is by Thompson and Harris. The paper provides a methodology for estimating jump intensities in a continuous-time portfolio credit rating migration model. The authors construct an inhomogeneous Markov chain from a sequence of piecewise homogeneous chains and then extract the factors using an estimator for discretely observed, continuous Markov chains. From a risk management perspective, the fact that the intensities can be interpreted as drivers of systematic risk is important. They can be correlated with economic variables to perform portfolio stress testing. The authors address the estimation problem caused by missing data and by a setup inwhich the chain is continuous but is observed only at discrete times. Both a simulation experiment and a real-data application using corporates rated by S&P are provided in the paper.

The third paper, "From actual to risk-neutral default probabilities: Merton and beyond", is by Berg. The paper seeks to bridge a gap between theoretical structural models of portfolio credit risk and empirical testing of such model results. The author provides empirical assessment of risk-neutral probabilities of defaults as modeled by key structural models originating with the Merton framework. The paper also specifically evaluates the applicability of information uncertainty on current asset values as found in the Duffie and Lando hybrid model. Finally, the paper demonstrates that risk premiums constitute a significant portion of model-implied credit spreads.

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 effort. The contents of technical reports complement rigorous conceptual and model developments presented in the research papers and provide a lot of value to practitioners.

The technical report in this issue, "An empirical implementation of CreditGrades", is by Yeh. The author describes an empirical evaluation of the CreditGrades model by comparing the model's predicted spreads with actual credit default swap quotes over the period 2007-9. The paper focuses on the strengths and weaknesses of the chosen model by analyzing the key results via a number of statistical and qualitative tests.

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: