Monday, December 28, 2009

It’s impossible to price a CDO

Posted on Reuters by Felix Salmon:

The discussion about Goldman’s synthetic CDOs just happens to coincide with a separate thread about a working paper by four researchers at Princeton, who seem to have a solid mathematical demonstration that CDOs simply can’t be priced: the amount of computing power necessary to do so just doesn’t exist.

Robert Oak has a good English-language explanation of the implications of the article: it boils down to the idea that if Goldman was loading up its synthetic CDOs with nuclear waste, there’s a good chance that its clients couldn’t have worked that out. Not because they were insufficiently sophisticated, but just because there aren’t enough computers in the world to do so.

It’s possible to overstate the connection between the paper and the events that happened at the height of the credit bubble. There’s a hidden implication in the paper that maybe with better models or faster computers we might have avoided some of this mess — but the solution to model risk isn’t more complex models, its less reliance on models altogether. And anybody who applied a simple smell test to the mortgages underlying the CDOs in question — rather than deciding instead to trust various quants both in-house and at the ratings agencies — would have come to the right conclusion without any computing power at all.

That said, there’s a case to be made that Goldman — along with much of the rest of Wall Street — was indeed arbitraging the models that investors and ratings agencies were using to price CDOs in general and synthetic CDOs in particular. The latter were particularly toxic, because synthetic CDOs are zero-sum, meaning that Goldman stood to gain in precise proportion to the degree to which the buyers of the paper were underestimating the risk involved.

The result was that Goldman had every incentive to structure CDOs which would blow up spectacularly, leaving investors with massive losses and Goldman with equal and opposite gains. The investors, meanwhile, might not have trusted Goldman (after all, they understood the zero-sum nature of these instruments), but did trust — far too much — mathematical models which turned out to be deeply flawed. And which, now, seem to be far too complex to get a grip on in any case, given the limitations of today’s computers.

In the final analysis, much of the problem here was a function of banks waving the magic letters “AAA” in front of the eyes of lazy institutional investors who thought they were getting free money by buying risk-free debt at hefty spreads over Treasuries. Some banks, like Merrill Lynch, did that and still contrived to lose billions; Goldman, by contrast, did that and managed to make a profit on the deals.

The whole sorry episode I think was fairly summed up by Goldman CEO Lloyd Blankfein when he said that “we participated in things that were clearly wrong and have reason to regret”. At the same time, however, Goldman wasn’t particularly evil or cunning here. It was simply acting like everybody else in the finance world, trying to take as much advantage of the credit boom as it possibly could. The main difference was that it had a bit more control over its own balance sheet than most of its rivals.

No comments: