The New York Times Magazine recently ran a fascinating article by Joe Nocera on the quantitative "risk management" that prevailed on Wall Street in recent years leading up to the crash. But, the experts interviewed in the story seem to me to miss a very major point of failure of the Wall Street risk analysts in recent years.
First let me say that the story is an almost comical story of people who believe that our analog, chaotic world can be quantified. Too many people believe this. They reify IQ tests, body counts, heart rate monitor readings, and medical tests of all kinds, dollar flows, and economic measures of all kinds.
One problem is that the world of human social relations is simply far too complicated to ever be quantified. This is the core problem with the discipline of economics, which in this case appeared to have helped lead some people into the mistaken belief that history can be regularized. Books have been written on this subject, but the nub of the problem is that regularities never last very long in history (not to mention the fact that even where a local regularity is discovered, the market quickly incorporates that regularity into its understanding of the world and therefore destroys it by "jumping out of the system").
In the Times article the historicist view (which anyone with a sense of history and discontinuity and a healthy skepticism toward economics should feel in their bones) is represented by Black Swan author Nicholas Taleb. As I have mentioned, I sometimes think of the possibility of discontinuity in terms of the movie "Empire of the Sun," in which a boy's stable, steady world collapses around him. But there is a more specific way to explain where Wall Street went wrong in its overreliance on the "Value at Risk" (VaR) quantitative model for measuring risk.
The basic problem appears to be that the VaR model used on Wall St. is essentially linear. The key thing that the risk analysts needed to track, but did not, was where in the range of possible market behavior the boundaries lay where the effects of market fluctuations upon their assets would go nonlinear -- where the feedback effects would kick in. Many disasters occur when organizations allow things to be set up in ways that permit some negative fluctuations set off chain reactions that quickly become far bigger in their effects than the factors that set them off.
The implosion of AIG and its 116,000 employees, for example, resulted from the activities of one small 377-person unit in London, according to another fascinating piece of reporting by the New York Times (which will surely win a Pulitzer for its superb explorations of the crisis). That unit began messing with credit derivatives, through which AIG insured the holders of packages of debt known as "collateralized debt obligations" (CDOs) -- basically, as the Times explains, the equivalent of mortgage insurance, only instead of for homeowners, it was for the owners of toxic mortgage debt. The CDOs served as amplifiers, taking the misfortune of AIG's clients and funneling it with greatly intensified force at AIG itself.
A similar story seems to have taken place with Merrill Lynch, which invested heavily in subprime mortgages. As the subprime market weakened, it makes sense that Merrill's investments would go down -- but it seems surprising at first glance that such a large company could not withstand the blows. In a linear world, that is what would have happened. But again, the problem is that it did not just invest in the subprime market, it also tried to amplify its profits through the use of CDOs, and when the market went sour those derivatives had set Merrill up for a negative feedback spiral that killed the whole giant company. Similar feedback mechanisms killed Enron (which hid the problem for a time through fraudulent activities, but the fraud is a separate issue from the stupidity).
The lesson for everyone: keep a sharp eye on situations where feedback mechanisms are created.
Comments