Good article from the WSJ (subscription) on how some CDO’s went bad along with why risk became concentrated. Includes a nice graphic that explains how CDOs work. From the article:
the system works only if the securities in the CDO are uncorrelated — that is, if they are unlikely to go bad all at once. Corporate bonds, for example, tend to have low correlation because the companies that issue them operate in different industries, which typically don’t get into trouble simultaneously.
Mortgage securities, by contrast, have turned out to be very similar to one another. They’re all linked to thousands of loans across the U.S. Anything big enough to trigger defaults on a large portion of those loans — like falling home prices across the country — is likely to affect the bonds in a CDO as well. That’s particularly true for the kinds of securities on which mezzanine CDOs made their bets. Triple-B-rated bonds would typically stand to suffer if losses to defaults on the underlying pools of loans reached about 10%.