It's everywhere. Magazine covers herald: "Wall Street's rocket scientists thought they had a sure-fire way to beat the market. Boy, were they wrong!" Nobel laureates, former Ivy League professors and sundry Ph.D.s by the busload are weeping in their beer over yet another dreadful tale of mathematical investing gone wrong. Quantitative woes at Long-Term Capital Management LP, Bank of America and the Harvard University endowment fund get headlines, but these are not isolated incidents. Never has mean-reversion been so mean to so many.
Is quantitative investing dead? Have we finally seen the folly of attempting to apply this revolutionary technology in investing? Is it time to shut down all of the computer models and go back to the good old days of lunch with the chief executive officer and sampling the new flavors? What a wacky idea. It's hard to imagine a financial world 50 or 100 years from today with less reliance on information technology. Does anyone really believe computerized investment analysis will be gone and forgotten? No, computers are here to stay.
The recent episodes are a painful maturation process for Wall Street's rocket scientists. Few things are certain in any market, at any time. Price differences between truly equivalent securities are quickly eliminated by arbitrageurs. Beyond these pure structural arbitrages, other forecast-based investments rely on scientific and mathematical, but imprecise, educated "guesstimates."
Some scientific guesses are better than others. None is perfect, but in aggregate, diverse, risk-controlled portfolios of investments based on reasonable but imperfect forecasts have done well over their five- to 10-year histories. Many institutional quantitative managers consistently have delivered unleveraged returns in excess of passive benchmarks such as the Standard & Poor's 500 by five to 10 percentage points annually, with low volatility from year to year, because of a well-developed approach to risk controls.
Betting the farm, and the neighbors' farms, on any market forecast using fifty- or hundredfold leverage is pure hubris. But it does not invalidate the underlying notion that the past can provide information about the future. Foolish, overconfident actions can show the limits of a technology, but do not invalidate it.
Here is the main point: computerized "rocket science" strategies can analyze more securities, more often and more consistently than the most diligent humans, who also rely on the past in framing judgments. There are more than 50,000 investible stocks worldwide. Many are followed by zero analysts. Information is the engine of market efficiency; where information is thin, there is opportunity to profit from transient inefficiencies. Computerized methods might be the only way to wade through the myriad conflicting, error-prone information available to a global investor.
But quantitative hubris can cause some people to confuse the model with the object being modeled. Sometimes these two things are close enough for practical purposes. Nations with nuclear weapons are willing to give up actual testing because the computer models are so accurate they can determine everything they need to know without detonations. People in markets are much less well-behaved than plutonium nuclei. They get nervous about Boris and Bill and Monica. Instantaneous communication transforms local jitters into a global anxiety attack. Good news is seen as bad news and bad news is seen as apocalyptic.
Models that have worked well for years lately are seeing conditions wildly different from what they have seen before. Spreads in the fixed-income markets go into unexplored regions. Short-term volatility is in new territory as well. The concentration of equity returns in a few very large stocks makes the "Nifty Fifty" of the 1960s look like a broad market. Historically, value stocks have done slightly better than growth; last year the pattern reversed with a vengeance, with value lagging nearly 15%. Quantitative investors sometimes speak of three or four sigma events -- outcomes that fall three or four standard deviations from the historical average. Such events understandably are supposed to be rare, the four-leaf clovers of quantitative investing. Not last year.
A useful analogy is to compare quantitative investing to driving while looking through the rear-view mirror. This backward looking approach will work when the road changes direction smoothly. When the road turns sharply, it's time to slow down or risk going over the cliff. If the road keeps switching direction, it's time to ease up on the accelerator by reducing bets, tightening risk controls and eliminating leverage. No one has a clear windshield on the future of the markets, but today's quantitative models might be slower to adapt than intuitive humans. Econometricians would say this is a "regime shift." Country singers would say we need to know when to hold 'em and when to fold 'em.
Turmoil and unpredictability in the markets will not last forever. Henry Ford once described failure as "the opportunity to start again with better information." The current difficulties do not invalidate quantitative investing. They set the stage for its evolution into something better than it has been.
David J. Leinweber is managing director-global equity strategies at First Quadrant LP, Pasadena, Calif.
