When AAA securities are downgraded to single C the excuse that our home price appreciation assumptions were wrong appears insufficient. Indeed the assumptions were faulty – but that might more appropriately have resulted into AAA securities being downgraded to AA, or maybe single A. When a AAA security becomes a single C security, despite all the triggers, excess interest trappings, and other forms of structural protections put in place to preserve the integrity of the AAA rating, the problem must be much bigger. It begins with transparency, or lack thereof.
(Buchanan 2006-1, one of Morgan Stanley’s now infamous “Dead Presidents” deals [CUSIP 118011AA3] provides an example of a security originally rated Aaa by Moody’s in November 2006 and which was ultimately downgraded to their lowest rating, single C, in April 2009.)
When industry experts like Joseph Mason and Arturo Cifuentes disagree with the rating agencies as to the level of transparency of their ratings, they’re typically arguing remarkably different concepts.
| Arturo Cifuentes, Ph.D. in How to avoid a ratings fiasco: just say No to black boxes:
“… what is a black box anyway? The term, used commonly in computer science, refers to a computer program that receives a specific input (data) and generates an output (in this case a rating) without offering much insight as to how it works. |
Vickie Tillman (S&P Executive Managing Director) in A good example of a highly transparent model:
“If Mr Cifuentes were to visit our public website, he could download the model for nothing and use it. It comes with a user manual, as well as disclosure of the assumptions (on probability of default of the assets, recovery potential, correlation, and so on) and of the formulas that drive the model. Scores of financial institutions and investors around the world use the model in order to help assess how changes in underlying collateral and assumptions might impact on our ratings.” S&P press release: “In contrast to the issuer-pay model, the subscriber-pay model tends to be a ‘black box’ with limited disclosure to ordinary investors.” |
| Joseph Mason, Ph.D. in congressional testimony:
“While the statistical techniques used by the NRSROs are transparent, the ratings criteria (the variables incorporated into the statistical techniques) are not disclosed up to a level of replicability. Without disclosure, even to a regulatory authority, NRSRO models are black boxes. Hence, it came as a surprise when Moody’s revealed that their ratings models lacked many key variables needed to properly evaluate non-prime loan products. In one of the more striking recent reports, Moody’s commented in that “the data fields essential for running the model were established when the model was first introduced in 2002. Since then, the mortgage market has evolved considerably, with the introduction of many new products and an expansion of risks associated with them.” |
Mark Adelson in congressional testimony:
“In my view, it is entirely clear that the rating methodologies are fully transparent. The evidence of the transparency of rating methodologies is in the voluminous reports that come from the agencies; the fact that they make the actual quantitative models available; the fact that analysts, hundreds of analysts, leave the rating agencies each year to take jobs with issuers, underwriters, etc.; and most important of all, the spirited debate about the pros and cons, the strengths and weaknesses of the methodologies that takes place in the open from individuals like me writing research reports. I have cited many of those reports in my written testimony, as you will see. So they are not black boxes, but they are also not totally simple. They are actually quite technical, and you have to have the right kind of technical background to grasp it. I mean, if my watch was broken, I couldn’t fix it to save my life. But if I went to watch repair school, eventually I would learn how to fix a self-winding mechanical watch, and then I could do it. So it is a technical area. It is not going to be graspable by everyone. But to folks in the business, it is perfectly graspable.” |
Let me address these disconnects. (The ramifications of a lack of ratings transparency, in an otherwise already opaque world of securitization, will however be explored in a subsequent piece.)
Complete ratings transparency, for me, means that if you wanted to establish the appropriate rating on a security at any point in time you could simply apply the rating agency’s assumptions to a model that is materially similar to the one they ought to have used – and out would pop the rating. In other words, barring some minor qualitative factors, one could essentially reproduce the rating. (I am not going to take a stance as to whether complete ratings transparency is ideal. Certainly, much is to be said for ratings analysts bringing a qualitative measure to their analyses.)
While some rating agencies might argue that the assumptions used in their models are completely transparent, this is certainly not always the case and certainly not the case for all rating agencies. In many situations, Nationally Recognized Statistical Rating Organization (“NRSROs”) do not even have defined methodologies published for products they have rated. But this again is a topic for another day.
Today we’ll explore the different levels of transparency.
Assumptions Transparency
Even if all assumptions applied are being fully disclosed – which they are not – there remain at least two aspects of transparency that beg attention.
Aspect 1 is the call for transparency not only in connection with the assumptions being applied, but pertaining to alternative reasonable assumptions that may otherwise have been used but were rejected. I would like transparency as to how the rating agencies came up with their egregiously poor correlation and recovery rate assumptions for RMBS tranches in ABS CDOs. I know the numbers they were (apparently) using. I have no idea why. I would like to know the extent of analysis performed that allowed them to come up with those crucial inputs. That, for me, is part 1 of transparency into assumptions.
Aspect 2 is crucial for it relates to the source of the data that are being entered into the models. Are these data inputs being provided by intenral ratings analysts, by independent third parties, or by incentivized parties to the transaction? To the extent they’re being provided by biased external parties, are they being verified internally? And do the ultimate clients have full disclosure as to the extent to which the ratings depend upon unverified data that is being provided by conflicted parties?
In my experience the answer to these questions will not leave you a happy investor: there are numerous examples of crucial inputs being provided by banks, collateral managers and the like which pass by the ratings analysts and into their models, unverified and untested. Many of these deficienct practices remain, to this day, undisclosed.
Model Transparency
If as an investor you’re forced to use an NRSRO’s model to verify their ratings, your capacity for doing your own due diligence is limited: you’re relying for one on the assumption that their models are accurate. We have seen that this is not always the case.
| Ratings Elements | Transparency Level | Notes and Examples |
| Quantitative Assumptions | Typically good | May be completely transparent for some NRSROs. For others, may be non-existent. |
| Qualitative Assumptions | Poor | Typically not disclosed. Example: a rating agency will typically not provide (publicly) its internal opinion on a CDO collateral manager’s management style – and its affect on their approach to rating deals managed by each manager. |
| Data Behind Assumptions | Average to Poor | Default history data, recovery data may be provided; data supporting correlation assumptions and choice of assumptions is seldom provided, if ever. |
| Model | Good to Non-existent | Certain models can be re-produced by a sophisticated investor; others are “black boxes.” On the corporate debt side, rating agencies would typically not disclose the total levels of off-balance sheet debt that banks were holding, of or credit default swaps that insurers like AIG or MBIA had sold protection on. Had this data and the model been available, investors would quickly have realized that their investments in MBIA or Ambac or ACA were no better than investments in an index of n senior tranches of risky CDOs, supported by low levels of capital. Like many CDO-squareds, the “monoline” insurers had a (disproportionately large) exposure to subprime ABS CDOs. |
| Ratings Scale and Metrics | Good to Poor | Application of ratings scale elements not always clearly defined or segmented. Example: NRSROs rating sovereign debt issues do not always bifurcate their rating into the two important, distinguishing elements of (1) ability to pay and (2) willingness to pay. Separately, it is also not always completely clear to what extent NRSROs choose the upper bound or the lower bound when their ratings analysis fits between two rating subcategories. |
Conclusion
The burden is on the investor to perform the due diligence required when investing in securities – particularly complex securities. The rating agencies have shown that you rely on them at your peril. Indeed, those who performed their due diligence (think Paulson) were able to find those instances where ratings provided were least predictive of the performance of the bonds rated. Investors who did this were able to profit from taking the other side of the trade against investors who chose to rely solely on ratings. We all know who won that bet.
Per Kurowski
May 14, 2010
One of the problems with credit ratings is that they are never sufficiently publicly debated, unless when it is too late, and when that happens then it is mostly the case of a small questioner against the mother of all father authorities in the markets.
Too often have I heard bankers ask me “Per, how on earth do you think I could convince my colleagues on the Board that the credit rating agencies were getting it so extraordinarily wrong that we should exit from what seemed to be an extraordinarily good business for us?”
In our efforts to solve the asymmetry in information we have increased the asymmetry of the credibility with respect to financial information, making it now almost impossible for divergent opinions to nudge the markets on the margin, and being only considered when the causes for the divergence become much too apparent, which is of course then much too late.
The first thing that should happen is that the credit rating agencies should be required to post, real time, all the questions and answers received with respect to every particular ratings, so to allow the market to express their viewpoints and to allow configure the necessary opinion majorities that could force the credit rating agencies to revise what they are doing.
If that Bank Director friend of mine could have referred to a place where those same suspicions were uttered by others, then he would stand a much better chance of being heard.