Back in December, the New York Times ran a story about Goldman and some collateralized debt obligations they underwrote. Turns out the CDOs, which were constructed by pooling together other bonds, performed badly. Really, really badly. The thrust of the Times piece was that Goldman routinely created CDOs with the worst mortgage bonds it could find, and then bet against them. Goldman and a few hedge funds pocketed billions when the CDOs failed, while Goldman’s clients who bought the CDOs endured huge losses. Goldman, of course, has denied all of this.
Few think Blankfein & Co. are innocent, but a report that is getting a lot of attention lately because of Michael Lewis’ new book The Big Short offers some fodder for Goldman’s side of the story. Here’s why:
Lewis’ book is all about the people who saw the housing bubble and the financial crisis long before most of Wall Street realized that their profit machine was about to blow. You can read Barbara Kiviat’s review here. In the acknowledgments of the book, Lewis praises a report written by Anna Katherine Barnett-Hart as the best piece of research written on CDOs, ever. By now you may have seen the WSJ story about her. Turns out she is not some high-paid hedge fund analyst, but at the time of writting the report a college senior just trying to earn her degree. Her thesis looks at why so many CDOs performed worse than expected.
One thing Barnett-Hart examines is how the CDOs of different investment banks performed. Turns out Goldman wasn’t the worst CDO underwriter after all. Quite the opposite. Barnett-Hart looked at CDO deals underwritten by investment banks from 2002 to 2007, and found that out of about 700, Goldman’s CDOs performed better than every other major underwriter of the investment product on the street. Through the end of 2008, just 10% of the bonds that Goldman packed into its CDOs had gone bad. J.P. Morgan’s rate of default was about four times that, making it the worst US investment bank in the CDO game. But plenty of others had similarly bad numbers. Merrill and Bear came in at a default rate of about 35%, and Citigroup posted a similarly depressing 30%. Barnett-Hart goes on to praise Goldman’s CDO underwriting prowess. Here’s what she says:
Based on the rankings in A.1, we can say that the CDOs of Goldman Sachs consistently outperformed, and are associated with a decrease of 6% in Default after controlling for CDO asset and liability characteristics. Among the consistent underperformers are Morgan Stanley, Merrill Lynch, Deutsche Bank, and JP Morgan – JP Morgan’s CDOs are associated with a staggering 18% increase in Default on average, after controlling for CDO asset and liability characteristics.
If Goldman was trying to create the worst performing CDOs, they clearly failed.
I ran all this by Goldman critic Janet Tavakoli and her response was that the overall performance of Goldman’s CDOs hides the truth of what they were doing. Goldman was good at structuring CDOs, and to be able to short the market they didn’t need dozens of deals to go bad. They just needed one or two well place bombs, hidden among the rest of their very good deals. Once you had those clunkers in place, you could buy as much credit default swap insurance on any single deal as you wanted. And the insurers like AIG would probably give you a good rate, knowing that 95% of your CDOs were performing above average. They didn’t know you set up a few to fail.
The problem with Tavakoli’s scenario is that AIG and others would sure be soon to catch on. If Goldman asked for CDS contracts on only one or two particular deals, deals they had set up, even AIG would have eventually asked what the heck did Goldman knew that they weren’t telling. And we know from the Madden Lane facilities that Goldman was indeed buying insurance on a number of CDOs and bond deals, not just one or two.
Another more plausible explaination of the low default rate by Goldman is that they weren’t alone. J.P. Morgan has a large CDS business, so it is entirely possible that banks was doing the same thing. But that doesn’t explain Bear, Citi and Merrill’s poor underwriting record. There is no evidence those banks were looking to do anything but go long mortgage bonds and the resulting CDOs.
Does this mean we have to rethink the whole Vampire Squid thing? Perhaps. But a few caveats before I go. The analysis goes up to 2008. It is possible that these numbers changed dramatically last year, and that Goldman’s deals are now fairing much worse than the other investment banks, though I am not sure why that would be. Second (and this one I really don’t buy as a reason to not believe the report, but putting it out there because I know some Goldman conspiracy theorists will find it anyway) Anna Katherine Barnett-Hart just recently landed a job with, you know who, the Vampire Squid itself. She will be starting there shortly. But Barnett-Hart finished the her thesis nearly a year before she got an offer from Goldman. In fact, the bank she was planning on heading to at the time and work briefly for was Morgan Stanley, and that bank doesn’t seem to get any special treatment in the report. Lastly, I haven’t been able to get in touch with Barnett-Hart to verify that I am reading her charts correctly. Ann Rutledge at R&R Consulting believes the chart on defaults has to do with the underwriters of the collateral, not the underwriters of the CDOs. That would make a big difference. But I am 99% sure I am right and Rutledge, despite being very smart and thanked specifically in the Barnett-Hart’s thesis, is wrong. So take a look yourself and decide who you side with, so we can all finally solve once and for all if Goldman is actually the root of all evil, get on with our lives and enjoy a little spring weather.