The Role of CDOs in Merrill’s Losses (Updated and Expanded Version)

We’ve been having too much fun on other fronts to spend much time on Merrill’s stunning announcement of $8.4 billion in losses for the quarter, which came from a combination of operating losses and writedowns.

The reason we’ve taken particular interest in this earnings announcement, aside from the magnitude of the red ink, is that write-downs of collateralized debt obligations were far and away the single biggest component of the loss.

CDOs are structured investments that contain a variety of assets, often tranches of residential mortgage securitizations (often the BBB or BBB- portion), commercial mortgages, pieces of collateralized loan obligations (LBO debt), and sometimes whole residential mortgages, There are also CDO squared (CDOs made from CDOs) and CDO cubed, and synthetic CDOs (CDOs constructed from the cashflows of credit default swaps written as credit enhancement for other CDOs).

In other words, CDOs are arcane and hairy paper. Ever since subprimes went wobbly, they looked primed for trouble. For more background, see here and here.

As we’ll discuss in greater detail, Merrill has taken the largest visible writedown on CDOs to date. This begs the question as to whether its haircuts will either encourage or force other firms to revalue their holdings.

Because CDOs don’t trade much and are very heterogeneous, there is a lot of legitimate uncertainty as to what they are worth. However, it is pretty certain that they are not as valuable as they were early in the year. It is widely believed that a number of organizations such as pension funds and others that can use “mark to model” may be keeping CDOs on their books at their old values if their credit rating remains the same.

Now to the Merrill results. Merrill is the only one of the largest investment banks to have reported a quarterly loss, which means a reduction of capital. Dana Cimiluca of the Wall Street Journal points out that the $7.9 billion writedown equals 19% of Merrill’s equity base fell by (although another way to frame it is that Merrill gave up a tad more than its record earnings of last year).

The unanswered question is what really lay behind Merrill’s losses. Were they so steep by virtue of being deeper into the wrong businesses at the wrong time, having gone further than its peers in writing down inventory, or having made some serious trading/business errors?

There is a sovereign irony in Merrill’s lousy results. The firm was expected to be less exposed to fixed income loss because it was better diversified. Lehman and Bear Stearns were far more exposed to those businesses. So this really may be a case of poor management or simply not cutting positions fast enough in a deteriorating market.

The reason we think that Merrill might have taken deeper writedowns is the fact that it sold a fairly large amount of CDOs last quarter as well as wrote them down. That may mean that they felt they had less latitude with respect to how they valued their remaining holdings.

Consider these statements, courtesy the Financial Times:

Merrill said it had cut its exposure to asset-backed securities in CDOs by 53 per cent to $15.2bn. Subprime mortgage exposure was down by 35 per cent to $5.7bn.

Total CDO exposure was written down by $6.8bn in the quarter, while subprime mortgage exposure was written down by $1.1bn.

So the great majority of the writedowns was due to CDOs. We have no way of knowing how much of the writedown was realized losses on CDOs sold versus marking down the value of the holdings that Merrill still has. However, if you simply take the beginning CDO holdings of $32.3 billion ($15.2 billion/.47), the $6.8 billion represents a 21% loss on that balance. Query whether any other Wall Street firm have taken that sort of reduction, and whether that is sufficient.

Another irony of this situation is that Merrill was the least aggressive user of “mark to model” accounting of major Wall Street firms, so this is not a case of having to walk away from phony-baloney accounting.

By way of background, firms can use something called Level 3 accounting for assets that trade infrequently, and CDOs certainly fit the bill. A not very charitable reading of this sort of treatment came in a recent Bloomberg story:

You can thank the Financial Accounting Standards Board for this. The board last September approved a new, three-level hierarchy for measuring “fair values” of assets and liabilities, under a pronouncement called FASB Statement No. 157, which Wells Fargo adopted in January.

Level 1 means the values come from quoted prices in active markets. The balance-sheet changes then pass through the income statement each quarter as gains or losses. Call this mark-to- market.

Level 2 values are measured using “observable inputs,” such as recent transaction prices for similar items, where market quotes aren’t available. Call this mark-to-model.

Then there’s Level 3. Under Statement 157, this means fair value is measured using “unobservable inputs.” While companies can’t actually see the changes in the fair values of their assets and liabilities, they’re allowed to book them through earnings anyway, based on their own subjective assumptions. Call this mark-to-make-believe.

A MarketWatch story, citing Bernstein research, was kinder, but also noted that Merrill was the most conservative in its Level 3 valuation methods:
Some assets are so esoteric and trade so infrequently that investment banks have to value them based on mathematical

models, rather than the market prices of similar or related securities. These are known as Level 3 assets.

This, in theory, gives firms lots of leeway in valuing these assets, which include things like derivatives, private-equity investments, residuals of CDOs and mortgage-servicing rights, Bove said.

Wall Street firms use mark-to-model techniques to value 9% of the Level 3 trading inventory on their balance sheets, estimates Bernstein’s Hintz. Goldman is top at 15%, while Merrill is bottom at 2%.

brokerage Percentage of Level 3 trading inventory valued using mark-to-model techniques
Goldman Sachs 15%
Morgan Stanley 13%
Lehman Brothers 8%
Bear Stearns 7%
Merrill Lynch 2%

Source: Bernstein Research

Consider this comment from the Financial Times’ Lex column:

A few weeks ago, investors were in masochistic mode. A string of hefty writedowns from investment banks were greeted with rising share prices, on the assumption that the worst was now out in the open. Merrill Lynch appeared to take a “kitchen sink” approach, with a massive $4.5bn writedown on collateralised debt obligations and subprime mortgages. Now it is clear that a large bathtub would have been more appropriate. Merrill on Wednesday increased the writedown to $7.9bn.

That is shocking. How could Merrill have got the scale of its exposure to losses so wrong? Was it not conservative enough the first time round, factoring in instead some sort of price recovery in these illiquid instruments? Or was it simply unaware of the true depth of its problem? Neither explanation will give any comfort to investors. Nor should either be enough to remove the spotlight that must now be shining on Stan O’Neal’s position as chief executive.

More broadly, investors might well reassess their confidence in the mark-to-market losses taken so far by other banks. One thing Merrill’s mess underlines is how subjective the marks appear to be for the more complex and illiquid instruments. No wonder the whole sector sold off alongside Merrill, on renewed fear that there might be more monsters lurking on investment banks’ balance sheets.

So if Merrill, which by all accounts seemed to be most conservative in its valuation of illiquid assets, made a large change in a very short period of time, how much confidence can we have in the valuations made by other firms?

One more tidbit suggests that most of the CDO holding were AAA paper. That profile would be consistent with the average haircut, since lower rated paper has plummeted in value. From the Wall Street Journal:

For much of the mortgage boom, Merrill was able to sell the bulk of the CDOs it underwrote to investors all over the world. But from late 2005 onwards, it became harder for the investment bank to find buyers for the growing volume of mortgage CDOs it was arranging. Many investors felt they had invested enough money in this asset class, and financial guaranty companies, which wrote credit insurance on many CDOs, were getting skittish about their growing exposures to mortgage securities in a slowing housing market.

For Merrill, the fees it earned from arranging deals were too lucrative to give up. Its profits averaged 1.25% of the deal volumes it did, or around $12.5 million for each $1 billion CDO. More than 70% of the securities issued by each CDO bore triple-A credit ratings. Traditionally these top-rated securities were insured by a financial guaranty company, which effectively bore the risk of losses. But by mid-2006, few bond insurers were willing to write protection on CDOs that were ultimately backed by subprime mortgages to people with poor credit histories.

According to people familiar with the matter, Merrill put large amounts of AAA-rated CDOs onto its own balance sheet, thinking they were low-risk assets because of their top credit ratings. Many of those assets dived in value this summer.

If this is indeed what happened, everyone in the chain of command that understood what was being done ought to be fired.

The explanation is a bit confused (the discussion of financial guarantors is a distraction), so let me translate. Merrill was making tons of money underwriting CDOs. For those unfamiliar with industry norms, the proper practice of underwriting involves earning fees for distributing securities. Despite the use of the term “underwriting,” Wall Street firms whenever possible pre-sell all of their deals. For highly liquid securities, they might enter into a “bought deal” where they purchase a new issue and then sell it as quickly as possible. But for normally-illiquid paper, novel securities, or IPOs, while technically the issuer again sells the deal to the underwriting syndicate, which in turn re-sells it to investors, the “ownership” is nominal since the investors have put in orders for the securities.

So what happened to Merrill? It was enamored of the CDO underwriting fees (managers also get to take “management” fees). But the market was backing up and Merrill was having trouble placing paper. (note despite the discussion of guarantors, Merrill did not step in to provide credit enhancement. It didn’t have the AAA rating to enable it to do that. The financial guarantor business was added either out of confusion or to illustrate that guarantors as well as investors were abandoning the paper). The fact that Merrill was holding AAA paper means it was already structured and rated, not inventory waiting to be securitized (although a related story suggests Merrill was stuck with some of that too).

So Merrill kept doing these deals so it could continue to book the new issue profits, but wound up keeping some of each deal on its balance sheet because it couldn’t sell it. That is a cardinal sin in the underwriting game. When that happens to an underwriter, you can be certain that they will take losses on their unsold positions, as Merrill finally did.