A good example is in the recent Wall Street Journal article “One Quant Sees Shakeout For the Ages – 10,000 years” by Kaja Whitehouse. The protagonist is Matthew Rothman, Global Head of Quantitative Equity Strategies for Lehman.

Here are two quotes from the article.

*“The trouble started Aug. 3, when stocks started moving not only in ways that commonly used models didn't predict, but in precisely the opposite direction from what was expected. Equally troubling, the moves were far more volatile than models based on decades of testing assumed were likely.”*

*"Wednesday is the type of day people will remember in quant-land for a very long time," said Mr. Rothman, a University of Chicago Ph.D. who ran a quantitative fund before joining Lehman Brothers. "Events that models only predicted would happen once in 10,000 years happened every day for three days."*

History suggests that hedge fund managers and quant managers won’t remember what happened for long. What long-term impact did the Long Term Capital Management collapse have?

Perhaps hedge fund managers and quant managers shouldn’t “remember” for long. Most of the time, they walk away with plenty of money in the bank. Their clients are the ones who lose big.

The primary problem with many hedge fund managers and quant managers is that they mistake the map for the territory, despite abundant evidence to the contrary.

Quant models are models, not reality. A quant model that describes reality reasonably accurately most of the time is valuable. However, it is inadvisable to use a quant model to compute the probability of a large loss from using the model aggressively. It is also inadvisable to compute the probability of a large loss from using the model aggressively from a simulated application of the model.

Suppose a quant model implied that so called “10 sigma” and “one in a 1,000 year” adverse events were fairly frequent? Would it be used? No. The only quant models that are used are the ones that are unrealistic enough to make such adverse events appear to be “10 sigma” and “one in 1,000 year” events. The same goes for long-term simulations.

The ability to hedge and diversify aggressive investment strategies effectively depends on strong assumptions about the market’s behavior, i.e., on the realized interactions among its securities. These are not known.

To illustrate, only an estimated covariance matrix is available and it is not the true covariance matrix. Moreover, the fund’s realized hedging effectiveness depends on the realized covariance matrix, which is different from the true covariance matrix. Worse yet, not even the realized covariance matrix captures a fund’s true risk.

The true probability distributions of returns and relative returns of aggressive hedge funds and quant funds have much more probability of devastating negative returns and negative relative returns than the normal distribution suggests.

A quant model may allow an aggressive fund’s risk to be known pretty well most of the time, but the other times occur frequently enough with devastating effect to make aggressiveness a loser’s game.

“10 sigma” and “one in 1,000 year” market events have occurred frequently enough over recorded history to convince any reasonable observer that they are nothing of the sort. They are quite common.

There is nothing wrong with the market. It’s the quant models and those who use them that are the problem.

## No comments:

Post a Comment