All the Devils Are Here [164]
By 2005, the hot new thing in the market was “correlation trading”: going long one tranche, maybe triple-A, while shorting another tranche, the triple-B, say. It was all driven by demand by investors for a particular slice of risk and models that purported to show what the spread, or the difference in price, between various tranches “should” be. It also allowed firms to make these correlation trades with immense leverage, because their models told them that the trades balanced out and therefore carried little risk. Remember 2004, when the SEC allowed Wall Street firms to use their own models to calculate the amount of capital they had to hold? Here was the consequence of that decision: because the models for correlation trading said they were near riskless trades, the firms didn’t have to put much capital against those trades. It was like Long-Term Capital Management on steroids.
In the spring of 2005, a warning shot was fired about the dangers of this brave new world when the credit rating agencies unexpectedly downgraded the debt of General Motors and Ford. All the models went haywire; press reports speculated that some hedge funds—and some Wall Street banks—had lost huge sums. That May, Investment Dealers’ Digest ran an article entitled
“The Synthetic CDO Shell Game: Could the Hottest Market in All of Fixed Income Be a Disaster in the Making?” It noted ominously that many players in the market weren’t capable of assessing the risks. Michael Gibson, the Federal Reserve’s chief of trading risk analysis, told the magazine, “What we are hearing from market participants is that there is a minority of CDO investors—perhaps 10 percent—who do not really understand what they are getting into.” Said an unnamed market participant: “I imagine the number is higher. It’s mind-boggling how much data you have to get a handle on to measure your exposure.” Risk expert Leslie Rahl explained, “[T]here could be a substantial difference between what a theoretical model tells you something is worth and where a buyer and seller are willing to transact.”
As usual, what should have been a wake-up call was ignored. Quickly, the synthetic CDO market in corporate bonds rebounded. And Wall Street took the next step: it began repackaging credit default swaps on mortgage-backed securities into synthetic CDOs. (Some of these synthetic CDOs only referenced mortgage-backed securities; others also included some actual mortgage-backed securities, or even other asset-backed securities, such as commercial mortgages, credit card debt, student loans, and so on.) Not only did this up the ante in complexity, but it also upped the amount of leverage at work. A corporate credit default swap had its share of leverage, but at least the underlying instrument was an obligation of a real company. More often than not, the underlying instrument being referenced in these new synthetic CDOs was a 100 percent loan-to-value mortgage made to a homeowner who probably couldn’t pay.
And yet the appeal was overpowering. There was so much demand for these securities that no matter how fast the originators made mortgages, it wasn’t fast enough. In 2005, Lars Norell, who worked with Chris Ricciardi at Merrill Lynch, told U.S. Credit, “In ABS [asset-backed securities], the availability of assets has been a sticking point. There’s a finite amount of them issued.” He pointed out that after subprime mortgages were crafted into residential mortgage-backed securities, there was only about $10 billion to $12 billion in the lowest-rated triple-B tranches. Since those were the tranches with the best yield, and therefore were the most attractive raw material for CDOs, it had the effect of putting a “natural cap,” as Norell put it, on CDO issuance. Plus, buying all the securities for a CDO could take up to six to nine months.
But with a synthetic CDO, it didn’t matter anymore if the originators could make new mortgages—or even if they went