Felix Salmon

Goldman’s outperforming mortgage CDOs

By Felix Salmon
March 25, 2010

I’m a bit late to this, but Stephen Gandel had an interesting post this weekend about the now-famous Harvard thesis upon which Michael Lewis based a lot of his latest book, and whether it partially exonerates Goldman Sachs from the charge of deliberately building CDOs which were designed to go bad, to the profit of Goldman Sachs.

Gandel gets comments from both Janet Tavakoli, who will never admit that Goldman is anything but pure evil, and from AK Barnett-Hart, the author of the thesis, who keeps a certain amount of scholarly caution while agreeing that Gandel basically read the thesis correctly.

Here’s what he found:

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 U.S. 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.

Of course, the likes of Merrill, Bear, and Citigroup never went short mortgages, so they can’t possibly be accused of deliberately constructing highly-toxic CDOs in the knowledge that those CDOs would end up defaulting. It’s hard to be evil when you’re fundamentally incompetent. But as the Lewis book shows, it was really Deutsche, not Goldman, which was most cognisant of the coming subprime collapse. And Goldman’s decision to go short came really rather late in the game.

Lewis is a storyteller, not a historian, so when it comes to finding evil intent at investment banks, absence of evidence is not evidence of absence. But Lewis also clearly has no love for Goldman, and the Goldman press office, for one, is upset with the contents of the book. Given the book’s subject matter and Lewis’s sourcing, I’m sure he would have loved to include a Goldman plot to rip off its own buy-side clients. But in fact it seems that Goldman’s clients were better served than the clients of other sell-side players in the CDO market.

13 comments so far | RSS Comments RSS

You are just not an experienced enough insider to understand Goldman.

Posted by randolfduke | Report as abusive

The evidence may point to the fact that Goldman issued higher quality CDOs than Citi or Merrill. But it seems patently ridiculous to claim that JP Morgan was “the worst US investment bank in the CDO game” without stating explicitly that JPM chose to exit the market in 2005 per Barnett Hart p. 94. (In other words, JPMs poor underwriting result is most likely conditional either directly or indirectly on a comparison to the performance of the earlier CDOs issued by Goldman et al.)

At 10%, Goldman’s CDO failures were more than 1.5 times JP Morgan’s total issuance of CDOs.

Posted by csissoko | Report as abusive

does anyone know which bank AK Barnett-Hart is now working at? I bet its GS…

Posted by davidbrentbrent | Report as abusive

Wait, isn’t the problem with Goldman that it was creating specific CDO’s that were bound to go bad? If that’s the case, the aggregate performance doesn’t tell you much — if anything, it would be an exacerbating factor. (ie Goldman understood what it was doing generally and produced higher quality CDOs, but for the purposes of dumping garbage on AIG went out of its way to produce poor ones.) Perhaps I don’t understand this.

Posted by jeff12345678 | Report as abusive

Brandon Adams says
“When I sent him my former student AK Barnett-Hart’s senior thesis, The Story of the CDO Market Meltdown: An Empirical Analysis, he read all 115 pages the night I sent it and pointed out an error that AK or her advisors didn’t catch! This thesis won the award for the best thesis in the Harvard economics department in 2009 and has since been read by hundreds of people. Apparently Michael caught the only error, and none of the material in the thesis was particularly new to him”.
quote taken from:
Debunking Michael Lewis’ Subprime Short Hagiography – 03/25/2010 – Yves Smith

Posted by velobabe | Report as abusive

AK Barnett-Hart was last seen working at Morgan Stanley.

Posted by Uncle_Billy | Report as abusive

AK Barnett-Hart was last seen working at Morgan Stanley.
Posted by Uncle_Billy

uhmm i know you. we are just spreading the LOVE⁄

Posted by velobabe | Report as abusive

Yves Smith rips AK Barnett-Hart and Lewis a new one in that article:
http://www.nakedcapitalism.com/2010/03/d ebunking-michael-lewis-subprime-short-ha giography.html

Felix, can we expect a response from you on this lively debate?

Posted by davidbrentbrent | Report as abusive

davidbrentbrent, my response is this: http://twitter.com/felixsalmon/status/11 075199623


This to me always seemed like a specious deduction. Sure, on page 33, you see that Goldman has had lowest rate of default specifically as of 2008. But on page 26, you see that Goldman drastically cut its CDO underwriting from 2007 to 2008 although it was one of the biggest underwriters in the overall period of 2002-2007 and the second and third largest respectively in 2005 and 2006 when sub-prime mortgages across the industry proliferated in the CDO makeup. Perhaps then, couldn’t Goldman’s comparatively superior performance as measured in 2008 be a result of a quick exit from the market in 2007?
A hypothesis: Maybe, JP just didn’t care about its relatively small exposure to the industry and the 40% default rate it incurred wouldn’t have made them much more money than if only 5% defaulted as Goldman’s did. But, Goldman’s exposure to the market meant it carefully evaulated its position and got out of bad positions just in time, which was reflected in its 5% default rate. Maybe, if the CDOs had defaulted just one or two years earlier, the story would be different.

Has anyone else actually read the thesis? Any comments?

Posted by anjalik | Report as abusive

P.S. It’s been about a week since I’ve read it, so perhaps I’m forgetting another important argument or piece of evidence raised. If so, please remind me.

Posted by anjalik | Report as abusive

She’s on MSNBC today:

http://www.msnbc.msn.com/id/21134540/vp/ 36088862#36088862

Posted by moneymonger | Report as abusive

Post Your Comment

We welcome comments that advance the story through relevant opinion, anecdotes, links and data. If you see a comment that you believe is irrelevant or inappropriate, you can flag it to our editors by using the report abuse links. Views expressed in the comments do not represent those of Reuters. For more information on our comment policy, see http://blogs.reuters.com/fulldisclosure/2010/09/27/toward-a-more-thoughtful-conversation-on-stories/