The line of thinking that underlies an investigation by New York attorney general Andrew Cuomo is a challenge to conventional wisdom about the financial crisis. The prevailing view is that since credit ratings were one of the single biggest points of failure in the crisis, the ratings agencies were one of the biggest, if not the biggest, perp.
Now this crisis had many parents, with the cast of characters including Alan Greenspan, Bob Rubin, Phil Gramm, Gaussian copula models, negative basis trades, captured regulators, undercapitalized banks and dealers, and of course, ratings agencies.
And the ratings agencies DO deserve a lot of blame. They played a compromised role in the structured credit market, via assisting in the design of deals they ultimately rated. In addition, the traditional ratings system which is hard wired into a lot of investment processes does not translate very well into structured credit deals. As Joe Mason and Josh Rosner demonstrated in a 2006 paper, structured credit transactions inherently are exposed to the risk of more dramatic downgrades than a corporate bond. So an AAA rating for those deals was NEVER as solid as an AAA on a corporate or municipal bond, even before considering the impact of a housing bubble, lousy origination standards, and design of some deals to favor the interest of the short side.
But the rating agencies have managed to argue a First Amendment exemption, which so far has made them judgment-proof (although a recent decision challenged that view in a structured credit case). So Cuomo could be argued to be unfairly targeting the banks due to his inability to pursue the ratings agencies.
But the Cuomo investigation is honing in on a crucial issue: did the banks misrepresent the assets in the deals rated? That has the potential to be a Big Deal, since it could result in bad ratings and the resulting losses being attributed to bad information from banks, who could be sued, and conveniently also are deep pockets.
The issue of misrepresentation is a common thread in many current and potential lawsuits. It is at the heart of the SEC’s case against Goldman. It also forms the basis for Fannie and Freddie action against the four biggest mortgage originators in the US: Bank of America, Citigroup, JP Morgan, and Wells Fargo. These banks had represented contractually that the mortgages underlying the bonds they sold to the agencies met certain standards, when they didn’t. The losses to these banks resulting from these false claims is $5 billion for 2009 and higher numbers are expected in 2010.
A New York Times story on Cuomo’s probe covers a lot of ground that will be familiar to readers of the blog: that low and behold, the banks tried to game the ratings! Since banksters try to game everything, this is hardly news:
At Goldman, there was even a phrase for the way bankers put together mortgage securities. The practice was known as “ratings arbitrage,” according to former workers. The idea was to find ways to put the very worst bonds into a deal for a given rating. The cheaper the bonds, the greater the profit to the bank.
While this behavior is to be expected, and the rating agencies should have been on guard, this sort of thing is another matter:
A central concern of investors in these securities was the diversification of the deals’ loans. If a C.D.O. was based on mostly similar bonds — like those holding mortgages from one region — investors would view it as riskier than an instrument made up of more diversified assets. Mr. Cuomo’s office plans to investigate whether the bankers accurately portrayed the diversification of the mortgage loans to the rating agencies.
Yves here. On the one hand, correlation was a crucial input into structuring these deals, and one big reason so many supposedly AAA bonds failed was the underlying assets were highly correlated. A story earlier this week in Bloomberg on a Merrill CDO provides support to the Cuomo thesis:
Neo CDO Ltd. was a complex construction. More than 40 percent of its holdings were slices of collateralized debt obligations sold by Merrill, according to Moody’s Investors Service and data compiled by Bloomberg. Many of those were CDOs made up of other CDOs backed by bonds linked to home loans. About one-sixth of Neo was invested in junk-rated debt.
Yves here.This deal was managed by Harding, widely seen as captive to Merrill. While the assets Harding selected presumably adhered to the restrictions set forth in the offering memo (limits on the % by rating, on regional and servicer concentrations, etc). it seems astonishing that a deal would be 40% CDO squared, 1/6 junk, and still have a AAA tranche that was 75%+ of total par value (which I would imagine it did). Normal mezz CDOs (ones made mainly of BBB subprime bond tranches) usually had only 10% maximum from other CDOs; so-called high grade CDOs, which were made from “better” cuts of subprime (A, AA, and junior AAA) did permit more CDO squared, but even then, the usual limit was 30%. And to have so much from the same issuer was also unusual.
Moreover, banks themselves often did these deals off their correlation desks. That further raises the possibility that the banks were arbing the correlation risk against their investors (in other words, they could be narrowly truthful in insisting, as Goldman and Magnetar do, that they weren’t designing the deals to fail; they were designing the deals to have highly correlated exposures. But that would mean if they got in trouble, they WOULD fail, as oppose to merely be somewhat impaired).
On the other hand, we argued in ECONNED that the use of correlation models on mortgage bonds was misguided:
ABS CDOs were the financial equivalent of turning pigs’ ears into silk purses, and in the end, they worked about as well. How could anyone at the time have convinced himself that these junior exposures to low credit quality instruments could produce AAA-rated paper? The problem is that procedures that made some sense on first generation securitizations were dangerously misleading here. It’s easy to blame rising real estate prices and ratings agencies, but the real roots lie again in flawed economic models.
Recall the discussion of correlation risk from chapter 3. The theory, developed by Harry Markowitz and William Sharpe, was that investors could create an optimal portfolio that suited their appetite for risk. But to do that, they needed to find investments whose prices moved differently, and they needed to have precise information about how these prices would move in relationship to each other (“covary”) in the future. In other words, this was a clever idea that would seem to have little practical application, except that a whole industry of faux science was constructed on this flawed foundation.
The way this approach was applied to structuring collateralized debt obligations was particularly dubious. The ratings agencies, the monoline insurers, and many investors looked at the risk of default using correlation models. But correlation is a concept in financial economics used to estimate overall portfolio risk based on price movements of the instruments in that portfolio in relationship to each other. If the price of one holding increases 5% in a day, another could change in a whole range of ways: up even more, up but not as much, no change, or down a lot or a little.
But if one loan defaults, the next will either default or not default. Only simple binary outcomes are possible. Thus using Markowitz/Sharpe-type models to analyze defaults was fundamentally wrongheaded.
Yves here. The interesting bit is from a legal standpoint, the logical response for the investment banks would be to say the credit agency models were bunk, the way that correlation models that were developed in the corporate loan market were repurposed to the asset backed securities market was problematic. But the difficulty here is the banks were hawking correlation products and correlation trading strategies; they were even deeper into these approaches than the rating agencies. So Cuomo may indeed be able to land a very solid blow if his inquiry does establish that the investment banks misrepresented