"History is a wonderful thing, if only it was true"
-Tolstoy

Sunday, August 12, 2007

Rocket Science ... doesn't work


A Week When Risk Came Home to Roost - NY Times


Unfortunately the Times has made this a "pay per view" piece.

In short, computer models aka "black boxes" work great, until they don't

Once the exogenous variable comes on the scene, it is likely that the model will fail, catastrophically.

Extensive quote ... with some deletions for brevity:


Using what are known as market-neutral strategies designed by computer models, hedge fund traders have been blindsided by a correlation between bonds and stocks that they never expected would occur.

Portfolios of this stripe are often known as quantitative funds; some of their most common trades are called statistical arbitrage. These bets are suggested by brilliant mathematicians and academics, using computer models to scour the markets for interesting trading patterns that continue for long periods.

...

Seeing that such bets typically generated profits over long periods left traders believing that their stakes were conservative.

The only trouble is, financial markets do not always trade in a way that is typical or predictable. And when they deviate from the norm, all the wonderful and smart trades stop behaving according to plan.

ANALYSTS call it model misbehavior.

...

Compounding the problem, of course, is the borrowed money these funds use to enhance their performance. When things start to unravel, leverage aggravates an already painful fall.

Mr. Rothman also pointed out that so many fund managers had the same trades on their books that when they went to cash out of them, the ill effects were exaggerated.

The losses that investors are suffering this month, he wrote, are comparable only to those in the 1960s and during the bursting of the Internet bubble. “This appears to be an event with little precedent,” he wrote.

None of this would be a problem, of course, if fund managers were not relying so heavily on just that — precedent — to make their decisions. Computer models seem so perfect, so scientific, so flawless, and they are advertised to investors in precisely that fashion. Ingenious models lull investors into a dangerous complacency about the risks they are taking. It is almost as if the models eliminate risk entirely from the markets.

But risk is never gone, as investors are recognizing with a jolt. And that is so even if Wall Street assigns conservative-sounding labels to portfolio strategies that are, in fact, aggressive.

“They have their standard deviations, correlations, ‘stable value’ and ‘real return’ funds and nothing for what the normal human being would call risk at all,” said Frederick E. Rowe Jr., a money manager at Greenbrier Partners in Dallas. “They’ve taken the word ‘risk’ and hijacked it. The concept of risk — the permanent loss of capital — vanished in the minds of the people who speak the new language.”

Risk, and all that it should connote to investors, is back in the language now. Unfortunately, it has brought an awful lot of losses with it. "

No comments: