Preventing structured product crises

The securitization market at the heart of the subprime fiasco may be dormant now, but it will reawaken, says one industry expert, and reforms are needed to tame the risks it exposed.

Loan standards in the U.S. were relaxed starting around the year 2000, with increasing recourse to subprime mortgages. Before then, subprime mortgages had been primarily the domain of second mortgages. What was crucial, however, was that the mortgage companies did not keep the loans on their books, says John Hull, the Maple Financial Group Professor of Derivatives at the University of Toronto’s Rotman School of Management.

Those loans were packaged with other loans such as asset-backed securities or collateralized debt obligations (CDOs). Such CDOs might contain other debt instruments, such as auto loans and credit card receivables, notes Hull, who was giving a talk this week on the credit crunch, sponsored by the Toronto CFA Society.

Hull gives the example of a $100-million ABS that is sliced into three tranches, though more typically there are six tranches. The first tranche, worth $75 million, would be rated AAA and come with a coup of 6%. The second or mezzanine tranche of $20 million would pay 10% and be rated triple B. The last, or equity tranche, of $5 million would promise 30% and be unrated.

Typically, the CDO originator would keep the equity tranche or sell it to a hedge fund. While the promised rate of return is high, that rate depends on whether the assets default or not. Like a waterfall, the interest on the loans first fills the senior tranche, then the mezzanine tranche, then finally the equity tranche.

In this example, the equity tranche would bear the first $5 million of losses, at which point it would be wiped out. The next $20 million would be borne by the mezzanine tranche.

What makes the picture more complicated is that, while the triple-A senior tranche is relatively easy to market, the mezzanine tranche is not. So very often, that mezzanine tranche was packaged with other mezzanine tranches to create a second-order CDO. Nor did it stop there. There were “third and fourth levels of securitizations getting more and more dodgy,” Hull says.

As evidence of the risk embedded in these types of CDOs, when Merrill Lynch sold $30 billion of the instruments to Lone Star last summer for 22 cents on the dollar, the implied loss was on the order of 20% to 25% on the underlying mortgages, Hull says.

Getting a handle on the risk in a CDO is complicated by the fact that only two things are known about the underlying mortgages: the credit or FICO score of the borrower and the loan-to-value ratio. “Everything else was lost in the securitization process,” Hull notes, including income. “That’s why mortgage originators didn’t check, because it was irrelevant.”

Beyond that, Hull cites a whole host of problems in assessing the riskiness of the CDOs. For one thing, “we need assumptions about correlations” among the underlying loans. For another, ratings tended to be negotiated, with CDO sponsors making changes to the senior tranche to earn the triple-A rating.

Another factor was that CDOs are not bonds. A mezzanine CDO has characteristics quite different from those of a triple-B bond. There is what Hull calls “cliff risk.” If it’s hit with defaults it is wiped out almost immediately. By contrast, investors in a defaulting bond might recover 50% of their investment.

Also problematic was the fact that the interests of the CDO originators and the investors were not aligned. “Quite often the originators kept the equity tranche, but once it was wiped out, they were no longer interested,” Hull notes. He suggests that originators keep 20% of each tranche they create.

Compensation played a role, too. The annual bonus for contributing to the bottom line, says Hull, “creates short-term horizons.”

But investors also misjudged the products. Institutional investors traded the triple-A credits “without having models, which is dangerous…you can’t do risk management.” But the kind of risk modelling also matters. “It is important to stress-test based on scenarios generated by the risk-management committee as well as historical data,” he explains. “You can’t do risk management just by mechanically applying value at risk.” Beyond that, senior management has to be aware of the importance of the stress-testing.

That said, however, “models in finance are not like models in the physical sciences,” Hull warns. “Models in the physical sciences tell you exactly what’s going to happen.”

A good case can also be made for more transparency. But it’s not through longer prospectuses. Words sometimes fail to convey the complexity of structure products. Hull suggests using software rather than words to describe the products.

In the end, if the credit crunch is not to be repeated, Hull argues that incentives must be realigned, risk models should not rely simply on value at risk, traders have to look beyond the ratings of the products, and stress-testing has to be taken more seriously.

(02/06/09)