The risk fallacy, page 2

Because such loans were new, there was little data an investor could use to predict their performance. Still, they were mostly made to prime borrowers, whose risk of defaulting was considered quantifiable, so it was possible to make a reasoned judgment of the default risk.

Of course, there were higher fees to be made by investment banks if they could apply the magic of securitization to dodgier assets. And if the dangers were gauged correctly, low-quality assets could be transformed into highly rated (i.e., easily sold) securities in much the same way a pastry chef could take a motley assortment of old fruit and turn it into a delicious pie.

These particular tarts were known as asset-backed securities, or ABS. They were created in the 1990s but really took off beginning in 2003. Here's how they worked: Subprime and other risky loan assets were packaged together. They were then divided into slices (called tranches) by likelihood of default. The least risky tranche was known as "senior," and the riskier ones were dubbed "subordinate," with multiple gradations along the way.

Then, to earn even more fees, the banks served up a more complex dish. They resuscitated an old vehicle from the junk bond era called collateralized debt obligations (CDOs) and used them to mix ABS, subprime loans, and, in some cases, credit derivatives (unregulated contracts that receive payments if a certain credit event occurs).

This process of repeatedly dividing the mortgages into securities, so the thinking went, diffused the risk of defaults among many thousands of investments. But each time banks concocted (or "structured") another level of increasingly complex assets and sold them globally, the chance that payments wouldn't come in as anticipated was harder to detect and easier to misjudge, and therefore posed more danger to investors.

In theory, rating agencies served as quality control, assigning a wide spectrum of grades to the various tranches. The highest rating was AAA, considered to be nearly risk-free. A BBB tranche, for example, had a higher chance of defaulting, but was designed to absorb losses and thus protect the pristine AAA tranches. At least, that's what the computer models said.

Oh, the models. Here we get to the nub of the problem. Wall Street based its actions on what its complex computer analysis concluded about the risks of securities based on subprime mortgages. But the models were fatally sabotaged by three basic fallacies.

NEXT: The risk fallacy, page 3

1 2 3 4