Bringing Analysis to Bear and Bull

By Bill Alpert

1706 words

18 June 2007




(c) 2007 Dow Jones & Company, Inc.

A lot has changed in the 60 years since W. H. Auden told Harvard's Phi Bates, "Thou shalt not sit with statisticians." That was obvious last week, when Mohamed El-Erian and his colleagues from Harvard's endowment sat in on a symposium on quantitative finance, sponsored by the university's statistics department.

When El-Erian took charge of Harvard's $30 billion-plus endowment last year, he had to fill a gaping hole in the ranks of Harvard Management Company. His predecessor had just left with about 30 staffers to start a hedge fund.

El-Erian began hiring talent from Wall Street, but he also sought advice from Harvard's economics and statistics departments. "We realize that there's a tremendous amount of brainpower at the university whose incentives are totally aligned with those of the endowment," El-Erian told the statistics seminar. "By reaching out, we can be smarter investors."

But the world's biggest school endowment isn't the only investment shop where "quants" are gaining influence.

Ten years ago, the Ph.D.s on Wall Street were confined to the derivatives units of big investment banks, or a few hedge funds like D.E. Shaw and Renaissance Technologies. But last week's meeting featured prominent speakers from Merrill Lynch and Lehman Brothers. Everyone agreed that statistical methods have spread to the mainstream of trading and investing. Some of the investment techniques described at the conference came from far-flung domains, including predator-prey studies and Cold War radar technology.

Harvard Management chief El-Erian has to play smart as he tries to extend the remarkable performance record established by his predecessor, Jack Meyer. The giant endowment gained an average of 15.2% over the past 10 years, compared with 8.7% for the median large trust fund. In the fiscal year ended June 2006, Harvard's endowment returned 16.7%.

Traditional asse-allocation strategies are having trouble in today's world, El-Erian told the two-dozen seminar guests gathered in a building financed by the founders of Microsoft. Formerly reliable market indicators now give contradictory signals. So last year, Harvard Management invited faculty financial wizards to help analyze the changing economic and financial environment. That examination paid off.

El-Erian credits the quants with helping him discover changes in the correlations of the fund's diverse assets, which left Harvard overexposed to moves in the stock market. The endowment lightened its positions in time to blunt the impact of a 4% drop in the Standard & Poor's 500 on Feb. 27. It subsequently bought back at cheaper levels.

Harvard Management's returns are supplying $1 billion a year to the university -- about one-third of the school's budget. On about $30 billion in assets, each additional percent of return equals a typical year's fund-raising.

Even though the endowment can't pay its portfolio managers as handsomely as a hedge fund can, the endowment's noble mission is a big draw. Already, Harvard Management has rebuilt almost 90% of the talent that it lost with Meyer's departure. "I had an easier time hiring people here than I expected," said Stephen Blyth, a Wall Street veteran who is now a vice president at Harvard Management and a teacher in the statistics department.

Statistics offers sensible tools for investing, said stats-department Chairman Xiao-Li Meng. That's because the discipline looks for practical ways to deal with uncertainty. In most scientific investigations, uncertainty is a nuisance. But in the financial markets, said Meng, profit can be made on uncertainty -- which Wall Street types call "volatility."

The high volatility of growth stocks seems to be the reason for a peculiar phenomenon discussed at the seminar by Samuel Kou, a Harvard statistics professor.

It turns out that the stock-market capitalizations of companies in the Internet or biotech sectors tend to fall along a predictable size distribution. Rank all the Internet stocks by market-cap. Then plot the market-caps against their ranks, both on logarithmic scales. The companies fall along an almost perfect line. Likewise for biotech.

This peculiarity was first observed in the 1990s by analysts at Credit Suisse. It's reminiscent of regularities in size found in many parts of the world: the distribution of personal wealth; the size of businesses; the population of cities. But in stocks, the alignment of market-caps appears only in volatile growth industries.

Kou has modeled this market-cap alignment using an approach from population studies known as a birth-death process. Theoretically, an investor could use such a model to trade stocks that get out of line. Kou expressly warned that he was not promoting the approach as a trading tool. But he did say that at least one hedge fund has exploited the phenomenon profitably.

Another speaker was quite eager for Wall Street to make use of his research. That was Jan Vecer, a Columbia University professor who is also a well-known sports statistician with a system for predicting the outcomes of pro tennis matches. Vecer told the seminar of his work on "maximum drawdown," a measure of investment risk that's gained some popularity among hedge funds; Vecer thinks it should also be the basis for a new kind of derivative that would protect investors against stock- market crashes. Maximum drawdown is defined as the drop from an asset's high to its low price, within a given time frame. "You can think of it as the worst thing that can happen to you as an investor," said Vecer.

Today, investors can try to protect themselves by buying a put option on a stock or the S&P 500 Index. But if the market continues rising-as in a bubble-the option will lose its value. The market might later crash and still remain above the option's strike price.

By contrast, a contract based on maximum drawdown wouldn't require the buyer to time the market top. And it would be relatively less expensive than options.

Signal-processing techniques invented by Russian radar scientists can help an investor to decide it's time to protect her portfolio with a maximum drawdown contract, said Vecer. During the Cold War, the Russians developed thresholds for judging whether a radar blip was just noise or an American spy plane. It was costly to scramble planes or launch a missile, so the Soviet researchers wanted to avoid false alarms.

Biology supplied the financial insights reported by another speaker, MIT finance professor Andrew Lo. His laboratory has hooked up traders to medical monitors, tracking their physiological responses as they trade. It turns out that the traders with the best P&Ls aren't those who are hyper-rational. Some fear and greed actually seems to make for better decision making, said Lo, by focusing the trader's attention on the profitable trading opportunities. Lo hopes his research might yield tools that tell a trader when she is "in the zone" to make good trades.

The processes of evolutionary biology can help explain the trading strategies seen on Wall Street, said Lo. Traders keep mutating their strategy until they find something that works in the environment of a particular market. They'll then stick to that strategy -- even if better strategies exist -- until changes in the environment erode the strategy's success. "The hedge-fund industry is the Galapagos Islands of finance," he said.

The hedge-fund industry is home to diverse species of investment strategy, all competing for survival. The profitability of any particular approach waxes and wanes over a short number of years, like the populations of predators and prey in an ecosystem. Statistical arbitrage did really well as a strategy from 1995 to 2000, for example, but has done poorly since. By studying the flora and fauna of each securities market, the MIT professor hopes to predict the cycling of hedge-fund investment styles.

It is so easy to raise money now to start a hedge fund, says Lo, that there are way too many predators and not enough prey. "In a few months or years, there is going to be a major hedge-fund event," he predicted. "People are going to get burned."

Investors will then pull money out of the hedge-fund industry. The prey-that is, the investment opportunities-will repopulate. And the cycle will start again.

It's a good time to be a quant on Wall Street, agreed members of a Tuesday industry panel. Ten years ago, quants were pigeonholed into solving specific problems, such as derivatives pricing, said Andrew Morton, a celebrated interest-rate researcher who now heads the European fixed-income business of Lehman Brothers. Today, nearly every single unit of an investment bank has people analyzing that unit's data to make decisions on a more rational basis. Lehman has five times as many quants as it did a decade ago.

Derivatives aren't all that interesting to study any more, said Emanuel Derman, a famous veteran of Goldman Sachs who now teaches at Columbia and works with the fund-of-funds firm Prisma Capital Partners. One subject Derman is examining is hedge-fund replication. That's a technique that some folks say can build a portfolio that mimics the returns of a hedge fund without the expense of a hedge-fund manager who would keep between 2% and 20% for herself.

There's plenty of work to be done. Harvard's Blyth said that investment-banking firms are looking more deeply into the tremendous amount of transaction data that they have on their order books. But the brokers have to handle the data carefully, lest buy-side customers worry that they're being studied for the benefit of the brokers' proprietary trading desk.

Indeed, statisticians should see no end of opportunity on Wall Street, as the buy-side and sell-side study themselves and each other -- looking to profit from anomalies in each other's operations, and even from anomalies in each other's statistical models.