In my first blog about the FER paper I asked the rhetorical question, How can we talk about elevating the standards of analysis and raising the accountability of SROs when we do not enable our MBAs to understand and follow the paradigm shifts etc. A guest commenter asked if I were implying that the FER were not qualified to comment (since the theory of securitization is not taught in any of the universities with which they are affiliated). Another guest commenter asked me to clarify whether I agreed or disagreed with the FER recommendations.
I would say that in the domain of the blog, anyone with internet access is entitled to comment on anything (including non-economists specializing in structured finance, like me) but what is difficult to establish in the blogosphere is not the backgrounds or motivations of bloggers but the factuality of their assertions. In intellectual matters factuality counts too, not just credentials. Members of the FER having risen to positions of prominence and academic leadership in finance or economics are eminently well-qualified to comment on matters related to the capital markets. But my recognizing their qualification to comment does not obligate me to agree with everything they say.
Moreover, it cannot be denied that over the past twenty-five years, non-mainstream financial ideas have been filtered out of the university curriculum, including mainstream financial practices like securitization. How else to explain that credit derivatives are a curricular item, led by experts in market risk analysis, but securitization (the spot market) is scarcely taught? My first blog on this topic was simply an attempt to calibrate the FER recommendations to market realities and try to stimulate discussion. In this blog, my purpose is to point out comments about SROs which the FER has offered as factual that, regrettably, are not facts but dangerous myths. Dangerous, because their perpetuation perpetuates the corruption of the financial system.
First myth: the assertion that SROs stopped evaluating the collateral they securitized, and that this was a causal factor in the market meltdown.
The notion that SROs ever evaluated collateral is a myth. Even the SROs have made no secret about accepting data provided by sellers and financial arrangers at face value without investigation. From the beginning of structured finance ratings in 1983, this had always been true. The data was reviewed by the seller and additionally, sometimes, by accounting firms or special servicers. The special servicer is a relatively new market institution, one whose primary mission has been to ensure the integrity of the data collected on the collateral. Anecdotally, SROs did not want to formalize the role of special servicer in their rating process because they perceived the special servicer as a competitive threat. Whether or not this is true, the reason offered by SROs as to why they never second-guessed the quality of data provided to them was their belief in the integrity of the banking system and its processes. This is a preservation of reputational capital argument—the same rationale that was (and still is) offered in defense of the SRO business model. Alas, the flaw in the argument is only becoming clear now. When fraud and misrepresentation are believed to have no adverse consequences, the value of reputational capital is $0.
Second myth: the notion that SROs played a central role in creating and marketing tranches of graded claims. I call this assertion a myth, but perhaps a better word is “half-truth.” It is true that the target rating—without which the market cannot establish a price—determines tranche size. But what is missing is the explicit recognition that rating and structure are non-linearly interdependent. The size of a tranche determines its rating; the rating determines its cost of capital and hence the weighted average cost of capital in the transaction; the WA cost of capital in turn determines the quantity of residual cash flow (excess spread or XS) available to the support tranches; and XS is another source of credit enhancement that drives the rating. To ignore the nonlinear relationship between structure, rating and price or interest cost, is to build in a loophole in the system for one party, usually the seller, to exploit at the expense of the other, usually the buyer.
The sensitivity of value to default risk and structure, credit convexity, is an intermediate to advanced level problem in fixed income mathematics that, as far as we know, is not taught in any academic finance program other than ours (at the Hong Kong University of Science and Technology; the University of California at Irvine, KAIST program; and Baruch at CUNY). The dearth of learning opportunities appears to be related to the lack of exposure to issues of structural design within academic finance, and seemingly also a deep reluctance within the academy to cultivate a critical perspective on the core concepts of academic finance.
Because one picture is worth a thousand words, I present figures at the end of this blog that are taken from a simple cash flow model of a structured deal. It is the same cash flow model that my students typically build after about 20 hours of instruction. The pictures show graphically the impact of nonlinearity (and credit convexity) on value. Usually when my students (who typically have five to ten years of deal experience) make these diagrams, they gasp in astonishment at the clarity and stark simplicity of what they have never seen before.
The basis of the figures is a simple analysis of a two-class, senior-subordinated (A/B) tranche transaction with the following collateral assumptions:
· the loan being securitized is a fixed-rate, level pay instrument
· defaults are 10% cumulatively in a logistical pattern
· recoveries are a constant 50% of defaults
· prepayments are cumulatively 20%
A sensitivity analysis on credit quality in the A and B tranches is run with defaults rising from 10% to 80% and subordination decreasing from 20% to 1%. Repayment in this scenario is sequential, meaning that the Class A has 100% of principal allocated to it before the Class B is entitled to any return of principal. The output in this analysis is the “reductions of yield” on each class—that is to say, the yield to maturity (expressed in basis points) in each scenario, with the scenario jointly defined by default rate and subordination level. The average reduction of yield serves as a rating-equivalent in Moody’s structured analysis. So, to aid the viewer in discerning the credit meaning, I have superimposed the credit grades AAA/Aaa, BBB/Baa and CCC/Caa.
Inspection of the A tranche output scenarios via Figure 1 matches a linear intuition of structured credit quality. Each line represents the –IRR of a default isocurve (x-axis) on the sub-domain of subordination (12% – 1%) or, alternatively, leverage (6.5x – 99x). These outcomes form beautiful parallel lines. As you move away from the upper RH corner (low-default scenarios) the security holder’s absolute likelihood of loss rises, and as you move from left (less leverage/more subordination) to right (more leverage/less subordination) the security holder’s relative likelihood of loss rises, in well-spaced, parallel lines.
The analysis in Figure 1 is offered more from the seller’s perspective, who may have a better idea of the collateral risk and creates a structure around that information. Changing the axis in Figure 2, offers more the perspective of a buyer who knows the structure and watches the impact of unfolding collateral quality on his or her slice of risk over time. Here the –IRR of a subordination isocurve (domain: 1% – 12%) is mapped to default ranges (10%-80%). The picture is still very linear, but note how A tranches supported by thick B tranches in the upper RH corner enjoy slight positive convexity with respect to a small increases in default rates and highly levered A tranches high have slight negative convexity. “Convexity” here refers to credit convexity, defined above.
By contrast, inspection of the B tranche dynamics leads to a few surprises:
Figure 3 for Class B is analogous to Figure 1 for Class A. As mentioned above, the picture of the nonlinear relationship between risk and credit protection in the Class B inevitably astonishes people and still has that effect on me. It reveals the clear separation between securities that will repay in full and those that will default—in other words, the default boundary which, without doing a cash flow analysis and knowing how to scale the results, can only be discovered after it is too late. However, knowledge of structuring and rating permits those possibilities to be mapped out and their ex ante probabilities to be assigned.
This picture shows how it is possible for sellers to originate securities that they know to be under-collateralized, up to two or three years before the evidence of fraud becomes clear. They are able to exploit the chasm in analytics and financial knowledge through plausible deniability. “How could we know?! Everything seemed fine.”
Note that in this graph, the credit enhancement variable is referred to as “leverage” (because the Class B does not have subordination) and leverage does not change by 1% increments. It changes much faster because as the denominator goes down by 1%, the numerator goes up by 1%. To move from the left to the right (X axis, above) is to increase leverage. Risk in structured deals arises because credit enhancement is levered, which is to say, shared by all the classes in claims-paying priority. To move from the right to the left is to de-lever. A reasonably well structured deal, one in which the risk is contained in the envelope of credit enhancement, will naturally move from the right to the left as the senior classes pay off and credit enhancement becomes dedicated to the remaining classes.
Figure 4 for the Class B is analogous to Figure 2 for the Class A. By reversing the axes and looking at the problem the other way, leverage can be seen to actually help the Class B because it lowers the total financing cost of the structure. Call this the Alan Greenspan strategy.
Lowering the funding cost frees up additional excess spread (XS) for both classes, but the marginal utility of that incremental XS for the Class B will be much higher than for the Class A, and over some scenarios (up to 33% defaults) the highly levered Class B generates very high returns. But at some point the benefit from high XS and leverage reaches its maximum. After the maximum comes a cliff, beyond which value plummets as the risk continues to rise. That dynamic explains the first “twist” around the 33% default level in Figure 4. The second twist, above the 70% default level, comes about because the Class B receives interest before the Class A receives principal. Because the Class B receives interest ahead of Class A, and because the Class B rate of interest is higher, and because (in this case) sequential repayment of the Class A means that the Class B is not being repaid, so that the absolute amount of yield is higher—because of all these factors, the curves in the picture “twist” again.
These graphs show the dynamics of all basic structured finance transactions with a pay-through waterfall. All senior securities exhibit similar risk/return dynamics to those shown in Figures 1-2, and all subordinated securities exhibit similar risk/return dynamics to those shown in Figures 3-4. Transactions where the cash is not allocated in a pay-through manner (i.e., those with triggers) will have securities with a target pattern of returns that may deviate from these norms. Such securities will typically siphon cash away from the other tranches in stressed scenarios.
In fraudulently structured transactions – those in which the collateral is known to be worse than stated – the senior securities with high ratings ultimately turn out to be under-collateralized (illustrated in the ranges moving down and to the left on Figures 1-2) while the subordinated tranches, which typically bear low investment grade ratings, will be virtually wiped out with the passage of time. Most likely, they were not structured for immediate sale but rather held in portfolio by the arranging banks for repackaging in future CDOs (collateralized debt obligations), ABCP (asset-backed commercial paper) or SIVs (structured investment vehicles), where the inflated rating could be re-used to prop up the nominal value of the collateral.
An overhang of this low-value debt had existed on the books of the banks for several years when, in August of 2007, the most conservative segment of the U.S. capital markets suddenly awakened to the presence of subprime mortgages in ABCP collateral. Banks arranging RMBS transactions on behalf of themselves and their clients had believed they could hide behind the illusion of solvency for many years to come by selling the illusion of safety and soundness to a naïve and complacent public, a strategy that has become much more familiar after the Madoff revelations.
Our banks today still hope for forbearance. Thus far, the government seems to be letting them have their way. The accounting system is badly damaged. The financial system still clings to the illusion that ratings mean something. The truth of ratings and the structured finance market is still muddled in emotion, ignorance and confusion. So far, the solutions have all been political. But, any political solution is a zero sum game. If we do not want to go on being manipulated by the same institutions who had profited illicitly from a perversion of securitization and ought now to earn back the trust of the public or face resolution, our only choice is to begin to insist on factual analysis.
In my next blog, I consider two more myths about SROs raised in the FER statement:
· Building a reputation for accuracy is critical to the success of any SRO.
· Ratings firms prospered to the extent that their predictions of the probability of default proved reliable after the fact.