Tuesday, February 5, 2013

Smoking E-Mails and More From Standard and Poors

Journalists from the New York Times report today that Federal prosecutors have subpoenaed 20 million pages of emails from Standard and Poors in relation to Federal investigations about the company's ratings of structured finance vehicles.  That sounds like prosecutorial over-reaching, but that's the climate of our political environment, I guess.

Just as in every Wall Street crisis, there are smoking emails, and the Times reports these two:

“Rating agencies continue to create an even bigger monster — the C.D.O. market,” one S.& P. employee wrote in an internal e-mail in December 2006. “Let’s hope we are all wealthy and retired by the time this house of card falters.”
Another S.& P. employee wrote in an instant message the next April, reproduced in the complaint: “We rate every deal. It could be structured by cows and we would rate it.”
 The original 2010 Senate hearings on the credit ratings issue feature a high level cast of characters from the rating agencies.  A student or reader who wants a laugh, or a headache, can listen to the audio.  Like most hearings on the Hill, they are not enlightening.

One expert's testimony, I found today, was enlightening and educational for me.  He is Dr. Arturo Cifuentes, a Professor of Industrial Engineering who also earned his M.B.A. in Finance at NYU's Stern School of Business.  When Dr. Cifuentes testified in 2008 and again in 2010, he was Managing Director in the Structured Finance Department of R.W. Pressprich and Company in New York.  He also writes with a sharp sense of humor.

In relation to my post from yesterday, Cifuentes told the Senate Banking Committee in 2008,

 "A study should be conducted by an independent internationally-recognized statistical consulting organization (there are well-established mathematical methods to conduct this type of analysis) to see if the ratings have been “independent.”  Take, for example, all the CDO ratings given in a specific time period by Moody’s and S&P (to the same transactions) and compare them to see if they are “statistically different” or not.  This is a much needed exercise"
The point I was making yesterday is that this exercise is the first, objective "smell test" which would show if in fact the global financial system had three independent credit rating agencies or not.  Professor John Coffee  concluded the evidence shows a "race to the bottom" as the three agencies competed for market share and for large consulting fees.  They had nothing to gain by giving appropriately lower ratings to CDO's, since this would automatically exclude them from being considered for an investment bank's business.  Issuers needed AAA ratings from two agencies in order to go forward marketing to their institutional investors, who had to be rating driven by their charters.

If was very clear, as Cifuentes points out, that once these CDO's were trading in the secondary market, buyers and sellers looked right through the published ratings and priced the paper appropriately.  He cites the example of two CDO's issued in March-April 2007, both rated (BBB/Baa).  One was trading at LIBOR +1000 and the other at LIBOR+120.  Cifuentes says this situation is "unheard of."

The much more technical paper by Cifuentes and Katsaros convincingly demonstrates a problem facing rating agency analysts and investment bank analysts.  As we said before, the performance of the CDO depends on the credit risk behavior of the underlying pool of assets.  The analyst has to determine the probability of default, for which market proxies like CDS spreads are available.  In addition, there are ratings and fundamental analyses of assets which can provide guidance.  The big problem is how to estimate the default correlation, for which there should be relatively few events from which to make an estimate.  Then, as Cifuentes points out, the default correlations proved to be time-dependent, which is not a usual model assumption.

Rating agency analysts turned to a specific model to solve their problems, the One-Factor Gaussian Copula, which allowed a modeler to use asset correlations as proxies for the unknown default correlation.  Using this model derives implied default correlation values that are tranche-dependent, something that should not happen.  Just to give the punch line from their interesting paper,

"To sum up: the one-factor Gaussian copula method is a flawed technique to
model something that does not exist -- two very good reasons to move on
and leave all this correlation/copula nonsense behind.  Future efforts should
be focused on estimating default probabilities better.  Period.  End of story."
I believe that when one thinks about the flawed model and how it propagated itself through all the rating agencies, it explains something about the behavior of the investment banks.  Their analysts and quants are, like it or not, of a much higher caliber than those of the rating agencies.  I would guess that they knew relying on the Gaussian copula was fine for generating the AAA rating the banks needed to move the paper into the market, but they also knew that a rating based on this flawed model was unjustified.  Thus, it's no surprise that investment banks like Goldman shorted CDO's in the secondary market. Just a thought.

It really is a shame that five years after these problems were clearly identified, essentially nothing has happened to hold the culprits and their enablers accountable for a crisis whose after-effects still permeate our economy and our financial system. Politicians of both parties and our regulators are squarely to blame.

No comments: