Bear Stearns and the Vagaries of Models

We had wanted to write about the role of models and more important, model assumptions in the ongoing Bear Stearns hedge fund debacle, and Gretchen Morgenson of the New York Times, in her story, “When Models Misbehave,” provided some useful intelligence.

With all due respect to Morgenson, while she touches on some dimensions of the problem, she doesn’t begin to capture how woefully inadequate the risk management and risk modeling processes are that are apparently standard practice on Wall Street for collateralized debt obligations, which has been a rapidly growing market in recent years. And the worst is these shortcomings have been in plain view.

Let’s start with Morgenson:

First, marking illiquid securities to a model that makes certain assumptions about their future behavior is not the same thing as marking to an honest-to-goodness market of buyers and sellers…

In worst-case scenarios, such models may reflect the fantasy that a firm’s principals prefer, not the reality of a security’s likely value. And yet, investors and financial firms everywhere are relying heavily on these models and building their balance sheets accordingly — a very dangerous game, especially when it comes to complex pools of securities backed by assets like home loans.

What does this mean in cold, hard cash? On a conference call with clients on Thursday, a Credit Suisse analyst estimated that the markdowns would likely be in the billions of dollars.

That brings us to our second lesson, which is another blinding glimpse of the obvious emerging from this debacle: the rating agencies, which investors rely on to be prescient cops on the beat, are stunningly behind on downgrading mortgage-backed securities and the pools that own them. Do the math: Bear Stearns is paying $3.2 billion to shore up a fund that once had $10 billion in value, according to one investor. That’s 32 cents on the dollar.

THE portfolio wasn’t just made up of toxic stuff, either. While 60 percent of the fund was invested in residential mortgages, 40 percent was in commercial loans. Moreover, 90 percent of the fund consisted of securities with AA or AAA ratings, according to the investor.

Officials at ratings agencies have said in the past that their ratings reflect their estimates of future performance, not market pricing. So the agencies are also marking to model. And that keeps people playing the fantasy game about values, especially in hard-to-analyze collateralized debt obligations that are essentially pools of other asset-backed securities. Some $1 trillion of C.D.O.’s have been issued. (Yep, C.D.O.’s were in the troubled Bear funds.)

“The C.D.O. sector is still extremely rich versus where the underlying collateral is trading,” said Albert Sohn of Credit Suisse on the conference call. “Either subprime has to get richer or C.D.O.’s have to get cheaper.”

I hate to sound like I am picking on Morgenson, who is a fine journalist, but market arcana is not her beat. There is enough wrong with this piece so as to make it somewhat misleading, and almost all of it is in the direction of making the situation sound better than it is.

Morgenson is right that marking to model is problematic, but she only skims the surface of how detached from reality the CDO assumptions, developed by the ratings agencies, are. This post from Minyanville give a much clearer picture:

I asked a large broker firm to send over its smartest math person on Collateralized Debt Obligations (CDO) structuring. I wanted to know what I am missing: why is the market so sanguine in the face of deteriorating collateral values in the mortgage market? One of my firm’s theses has been that, as the mortgage market deteriorates, investors holding CDO as an investment would realize losses and this would feed into other risky asset classes. Why aren’t losses being seen when the market is clearly deteriorating?

The team that came over was headed by a very smart gentleman. He was very good at math and very straightforward. Working for a broker I was prepared for some sugar coating. I didn’t get any.

The answer is simple and scary: conflict of interest.

He explained that due to the many layers of today’s complicated credit products, the assumptions used to dictate the pricing and outcome of CDO are extremely subjective. The process is so subjective in fact that in order to make the market work an “impartial” pricing mechanism must exist that the entire market relies upon. Enter the credit agencies. They use their models, which are not sensitive to current or expected economic activity, but are based almost entirely on past and current default rates and cash flow to price the risk. This of course raises two issues.

First, it is questionable whether “recent” experienced losses over the last few years really represent the worst of the credit market (conservative). But even more importantly, it raises a huge conflict of interest: the credit agency’s customers are the very issuers of the tranches they rate. The credit agencies, therefore, need to compete for business based at least in part on the ratings they are willing to give these tranches. As a result, they will only downgrade when forced to by experienced losses; not rising default rates, not a worsening economy, but only actual, experienced losses. Even more disturbing, they will be most reluctant to downgrade the riskiest tranches (the equity tranches) since those continue to be owned by the issuers even after the deal is sold.

So even though the mortgage market has deteriorated substantially, mark-to-market losses by those holding the CDO paper have generally not been realized simply because the rating agencies have not changed their ratings for all the above reasons. Accounting rules only require holders of the paper to mark prices according to the accepted model, not actual prices. For example, below is a chart of the actual BBB minus tranche of the mortgage-backed securities pool from November ’06 to present. Actual prices where traders can really buy and sell is substantially lower than where investors are marking their positions.

The levels at which investors are carrying the paper is not reflecting underlying reality as the holders simply hold their collective breath and the rating agencies ignore a worsening environment.

I asked them what would force the rating agencies to change their ratings and the response was “it’s just a matter of time if the market continues to deteriorate, for the agencies at some point will be forced by the cumulative losses to acquiesce.” Because these losses have been compressed, any re-adjusting of ratings by these agencies are likely to result in a massive repricing of risk.

For those of you who don’t live in the debt world, the 25% decline in the chart above is a huge move.

Back to Morgenson. She got a very important issue wrong:

Officials at ratings agencies have said in the past that their ratings reflect their estimates of future performance, not market pricing. So the agencies are also marking to model.

Ratings agencies are not Bloomberg terminals. They provide ratings. Even Shedlock got this slightly wrong. The Financial Times provides perhaps the clearest explanation:

The agencies highlight that their ratings explicitly do not address market pricing or trading liquidity for the security in question, but rather focus on the likelihood of default. However, some investors may not appreciate that distinction and could be surprised by volatile prices. Mr Fuss also notes that many newer structured instruments have not been through a serious, broad-based market downturn, making their behaviour hard to predict.

Connecting the dots between the Shedlock post and the FT, market participants have chosen to rely upon a credit model for pricing of illiquid CDOs. This is clearly nuts, but as Shedlock made clear, it was certainly convenient and thus became the norm in the absence of any other convention. And Shedlock implies, and may well be correct, that the agencies have been complicit (ie, it suited their investment bank clients to fob off this approach on clients so they could find chumps to buy the high risk, least liquid tranches).

In fact, it appears that at least some hedge funds (one would hope most) were using their own models to value this paper, although they may merely have been tweaked versions of the rating agency models. Similarly, investment banks were using their own methods for valuing CDOs as collateral, to the point where hedge funds would shop among brokers to see which would lend on the most favorable terms.

Moggenson again:

Bear Stearns is paying $3.2 billion to shore up a fund that once had $10 billion in value, according to one investor. That’s 32 cents on the dollar.

No, Bear is not making an equity infusion. That is why its stock only fell 1.4% on Friday, in line with the market. Tanta at Calculated Risk quoted this note from Fitch:

The credit ratings of The Bear Stearns Companies Inc. (Bear Stearns) will not be affected by today’s announcement to provide up to $3.2 billion in secured financing to The Bear Stearns High-Grade Structured Credit Fund (High-Grade Fund), according to Fitch Ratings.

The High-Grade Fund is a hedge fund managed by Bear Stearns Asset Management (BSAM). The Bear Stearns facility is a collateralized repurchase agreement, which can be readily funded with existing internal cash sources. The provision of repo financing is a product offered in Bear Stearns’ usual commercial activity and does not constitute an equity investment.

And $3.2 billion vs. an initial value of $10 billion tells us less than it appears. Some assets have apparently been sold at close to par, so the total (for her purposes) is less than that. The credit facility Bear is providing is presumably to take out, presumably to a large degree, the remaining margin loans. I don’t know how collateral is treated in this market (as in what percentage of market value one can margin) and given that there are assets of varying credit quality, the margin percentages doubtless vary.

Back to Morgenson:

The portfolio wasn’t just made up of toxic stuff, either. While 60 percent of the fund was invested in residential mortgages, 40 percent was in commercial loans. Moreover, 90 percent of the fund consisted of securities with AA or AAA ratings, according to the investor

The commercial mortgage market has been engaged in the same type of practices you see in residential lending, so the particular credits in the Bear fund could well be dodgy. And the high ratings mean very little. A hedge fund buddy of mine who doesn’t normally play in CDOs was looking at some AAA rated paper in March and mused that it was trading at 270 basis points over Treasuries, a junk bond spread. He decided not to buy it because he figured it still had downside risk.

Enough on pesky but important detials. What does this mean?

1. We see widespread use of pricing models that make no, zero, zip, nada allowance for liquidity risk

2. We nevertheless see a lot of people, some of them very smart (Cioffi was no fool and Bear supposedly has among the best risk management for this sort of thing) making very large bets apparently using this sort of pricing approach

3. This market happens to be huge in aggregate. The structured securities market in the US alone is estimated at $9 trillion in market value, which is twice as large as the Treasury market. However, it isn’t clear what the split is between reasonably liquid and illiquid instruments

When did we last see this movie? LTCM. LTCM had the best and the brightest, it had enormous confidence in its models, it too the risky side of “shot vol” trades across many markets, assuming they were uncorrelated, and they all became correlated in the Russia default in 1998 when there was a flight to quality.

Now we are talking about only one market (unlike LTCM) but this is a very big market. And you have a lot of hedge funds that are specializing in this exotic debt paper, so and so by definition are not diversified.

As we have discussed earlier, if LTCM-like problems were to develop across multiple players, there is not enough legal, regulatory, and managerial bandwidth to handle multiple Bear-scale crises at once. I don’t know what the exact threshold is, but my sense is that six simultaneous meltdowns like this would cause the system to choke. And we aren’t alone in this view. Richard Bookstaber, a hedge fund manager himself and innovator of some of the risk management techniques used on Wall Street, warns that the industry, like a nuclear power plant, is subject to “tight coupling.” A large error jeopardizes the entire system.

Let’s hope the shock of the Bear fund failures leads other players to reduce their risk exposures in an orderly fashion…..