In our second example, we suppose that an investor is offered a differ- ent choice of selling credit protection via two different 5-year credit
8 The standard deviation we are referring to is of the results of simulations for each par- ticular rating cohort. It reflects the variability of portfolios formed at the same time.
c14-FrameworkEvalTrades Page 274 Monday, March 6, 2006 11:22 AM
derivatives—CDO equity and a basket swap. Both are first loss posi- tions, but one is based on exposure to an undivided portfolio while the other is based on tranched exposure to a higher-rated portfolio. The two choices are:
■ Sell credit protection for five years on a portfolio of 100 BB corporate names for 350 basis points via a basket CDS, “BB basket.”
■ Sell credit protection for five years on 100 A-rated corporates for 35%
upfront and 500 basis points per annum, being responsible for any default losses that occur on the 0% to 3% synthetic CDO tranche,
“CDO equity.”
We again assume $100 notional for both trades. We still make use of the yearly loss given default data displayed in Exhibit 14.7 and the 25% interyear standard deviation of loss given default. But now, we use marginal default rates for BB and A rating cohorts from 1970 through 2000. These data are shown graphically in Exhibits 14.11 and 14.12.
Note the difference in scale on the exhibit’s vertical axis.
Exhibit 14.13 shows the expected present values of the CDO equity and BB basket alternatives for each rating cohort from 1970 through 2000. Again, there is no contest between the two alternatives. The aver- age of the CDO equity series is $49.14 versus $10.08 for the BB basket, giving CDO equity a $39.06 advantage. The CDO begins with a pricing advantage, because—in the absence of any default loss payout—the
EXHIBIT 14.10 Present Value of Two BBB-Underlying Trades
276 SYNTHETIC CDOs
EXHIBIT 14.11 Yearly A Rating Marginal Default Rates
Source: Moody’s Investors Service.
EXHIBIT 14.12 Yearly BB Marginal Default Rates
Source: Moody’s Investors Service.
c14-FrameworkEvalTrades Page 276 Monday, March 6, 2006 11:22 AM
present values of the CDO and BB basket are $56.88 and $15.32, respectively. The slightly higher expected loss on CDO equity versus the BB basket, $7.74 versus $5.24, is not enough to overcome the advan- tage with which the CDO begins.
In Exhibit 14.14, we again equalize the expected present value of the two trades, this time by inflating the notional of the BB basket to
$488. Now, both trades have an expected present value of $49.14.
Besides having a higher theoretical worst-case potential loss because of its higher notional, the BB basket shows more volatility year-to-year.
The CDO ranges from $23.73 to $56.88 while the BB basket ranges from $9.62 to $69.14. Finally, in Exhibit 14.15, we show average present values minus one standard deviation. Again, the basket trade is more volatile than the CDO equity trade. Our conclusion is that the CDO equity trade is more attractive than the BB basket trade.
Methodological Caveats
Our analysis assumes that the credits we looked at are about as default- prone as any other credit of the same rating. In other words, we assume no adverse (or verse) selection bias in the choice of credits. We are assuming that the creators of these trades have not gone through Moody’s ratings and picked the worst (or best) credits out of each rating
EXHIBIT 14.13 $100 A Equity versus $100 BB Basket
278 SYNTHETIC CDOs
EXHIBIT 14.14 $100 A Equity versus $488 BB Basket
EXHIBIT 14.15 $100 A Equity versus $488 BB Basket
c14-FrameworkEvalTrades Page 278 Monday, March 6, 2006 11:22 AM
category. We also assume that the CDOs are about as diverse (e.g., by industry) as the entire rating cohort.
Finally, our analysis focuses on long-term buy-and-hold results. We do not address short-term mark to market volatility, especially price volatility that is not the result of defaults and recoveries. We focus solely on the present value of cash flows and note that ultimately market values must reflect actual cash flows. We also do not address the work- ing capital cost of collateralization or the cost of economic capital to support trades.
Caveats and Wishes
Our methodology yields immediate insights from history. By no means, however, do we believe that we provide the final answer to the challenge of analyzing tranched credit risk if for no other reason than that an accurate view of the past is unavailable. In this chapter, we relied upon default studies based on credit ratings knowing that there is a great deal of variability in the credit quality of corporates that carry the same rat- ing. In the past, this has lead to adverse selection in some synthetic CDO portfolios, where the worst credits of a particular rating are selected as reference entities.
We would prefer that default studies were performed on more homo- geneous groups of corporate credits. That way, variability in default rates over time would be less a function of the variability in name credit quality and misratings and more attributable to economic and under- writing factors.
There are numerous avenues that researchers could pursue to develop more credit-homogeneous cohorts and thus better default studies. For example, each quarter Moody’s shows how bond spreads rank the credit quality of corporates better than do ratings.9 A default study based on rel- ative bond spread, instead of on rating, would better stratify credits and help us to make better predictions about the future. Moody’s has already done a default study incorporating rating outlooks, rating watches, and prior rating changes, in addition to the credit’s current rating.10 This study shows that the predictive power of ratings is refined with this addi- tional information. A long-term default study using these variables would also give the market a better view of the past to apply to the future.
9 For example, Richard Cantor and Christopher Mann, The Performance of Moody’s Corporate Bond Ratings: September 2004 Quarterly Update, Moody’s In- vestors Service, October 2004.
10 David T. Hamilton and Richard Cantor, Rating Transitions and Defaults Condi- tional on Watchlist, Outlook and Rating History, Moody’s Investors Service, Febru- ary 2004.
280 SYNTHETIC CDOs
Finally, statistical default modeling based on the Merton model or structural model, financial statement variables, spreads, or combina- tions of these variables has the potential to form more homogeneous credit cohorts and thus eliminate misratings as a source of default rate fluctuation over time. It would also help our predictions if these models were applied more to the prediction of absolute default rate rather than to the relative default propensity of particular credits. All these meth- ods, and others we have not thought of, would give us a better credit view of a portfolio.
The second reason why our method is not the final answer to the problem of tranched credit analysis is that even with a perfect view of the past, one always has to determine the applicability of history to the current situation. We have more modest goals for our methodology than those who approach portfolio credit risk like a physics problem whose solution is calculable to an infinite number of decimal places. We simply want to know what has happened in the past before we try to predict what will happen in the future. The question that naturally follows from this is “What part of the past is most applicable to the future?” An informed opinion depends on an assessment of current economic condi- tions and underwriting standards.
CONCLUSION
We applied a historical perspective to the analysis of a single credit, a portfolio of credits, and a CDO tranche, as well as to credits with the same and different underlying ratings. We asked, and answered, the question: “What would have happened if we had done this trade in 1970, 1971, 1972 … 2000?” In the matchup between the BBB CDO tranche and the single-name BBB CDS, the CDO unambiguously would have performed better. In the matchup between the CDO equity tranche and the BB basket swap, the results were more ambiguous, but were favorable to the CDO equity.
In our analysis we used only actual data on marginal defaults and yearly loss given default. We think this brings a much needed dose of reality to the analysis of credit portfolios.
c14-FrameworkEvalTrades Page 280 Monday, March 6, 2006 11:22 AM
CHAPTER 15
281
Structured Finance Credit Default Swaps and Synthetic CDOs
ntil 2005, there was only a trickle of synthetic CDOs based on struc- tured finance underlyings. Synthetic CDOs were almost always based on corporate credit. That changed in 2005 when Wall Street dealers released a series of templates for transacting single-name credit default swaps on structured finance underlyings (SF CDS). This jump-started the market for SF CDS, which in turn breathed life into the market for synthetic SF CDOs.
Not that documentation issues are settled. A group of end users, headed by monoline insurers, have proposed their own, very different, template for trades. This has lead to different varieties of SF CDS and synthetic SF CDOs. In this chapter, we look at SF CDS documentation issues and how they affect synthetic SF CDOs.
With respect to SF CDS documentation, we cover:
■ The differences between corporate and structured finance credit.
■ The evolution of SF CDS documentation.
■ The competing dealer and end user templates.
■ The SF CDS terms which best replicate the economics of owning a cash SF bond.
With respect to synthetic SF CDOs, we cover:
■ Manager’s new found flexibility in accessing credit risk.
■ The creation of new SF CDO structures.
■ The effect on SF CDO credit quality.
U
282 SYNTHETIC CDOs