"Magic Formulas" => Not-so-Magic Results
"Investing by the Numbers Doesn't
Offer a Great Edge"
By JOSEPH KAHN -- July 21, 1998
NEW YORK -- Put some of the world's smartest math
scholars together with the fastest computers. Give
them
a wealth of historical data and a few billion dollars.
Do they have a chance of predicting swings in the stock
market?
David Shaw, a computer-science professor turned Wall
Street trader, believes that his chances are not much
better than the odds of picking red or black correctly
in roulette. Shaw has all the tools at his disposal --
the brains, the computers, the data and the money. But
each time his secret algorithms spot a buy or sell
opportunity in stocks or bonds, he figures his chances
of making money are only marginally better than if he
flips a coin.
Shaw is a quant, or a quantitative trader, who uses
algorithmic and statistical analysis of market data the
way other investors use research on individual
companies and instinct.
He is among the most successful of a generation of
scholars who have left their ivory towers in the last
20 years to try their skills at hedge funds and big
investment banks. A great deal of fanfare greeted their
arrival on Wall Street, with some analysts predicting
that mathematical models would revolutionize the way
money is made or lost in the financial markets.
It has not. Math has made a few multimillionaires,
including Shaw. But the quant business has become a
highly specialized and professional one -- small,
competitive and, only sometimes, a lucrative way to
manage money.
Quant funds now control perhaps 5 percent of all the
money invested in stocks, and a greater part of the
speculative capital invested in futures and options.
But only a few people make a good living playing the
stock-market numbers game. Fads that once gripped the
imaginations of quant traders, like chaos theory and
its hoped-for applicability to modeling the market,
have faded. And some of the most prominent names in
quantitative-oriented proprietary trading, like Salomon
Smith Barney, are getting out of the business
altogether.
"A few people have done well but the majority of people
have not and many have been forced out," Shaw said. "If
we had not gotten in early, it probably would not be
possible to break in at all."
That sober analysis does not match the popular image,
most recently embodied in the brilliant but disturbed
hero of the movie "Pi." Convinced that numbers lie at
the heart of everything in the universe and aided by an
advanced silicon chip, Max Cohen, the lead character in
the independently produced film that won a top prize at
the Sundance Film Festival, discovers a number that
unlocks the secrets of the Torah, the infinitely
variable game of Go and stock-price movements.
"There are patterns everywhere in nature: the cycle of
disease epidemics, caribou populations in the Arctic,
sunspot cycles, the rise and fall of the Nile," the
obsessive Max says in one of the movie's philosophical
monologues. "Within the stock market, there are
patterns as well, right in front of me. Always have
been."
That's fiction, of course, but not fantasy. A British
hydrologist, pondering the mysterious repetition of
droughts and floods on the Nile at the turn of the
century produced the theory that seemingly random
events, like the roll of the dice and the flip of a
coin, are not necessarily perfectly random.
He was H.E. Hurst, and his contribution to statistics
is the Hurst exponent, which measures the probability
that one event -- like a flood or a surge in the
Standard & Poors 500 stock index -- is likely to
closely follow a similar event.
This theory, with many mutations, lies at the heart of
quantitative active trading strategies today. No
serious quant trader aspires to predict tomorrow's
price of, say, Ford Motor shares. But most rely on
computers that take into account many variables, like
economic conditions, analysts' recommendations, sales
and profit growth, interest rates and the money supply,
to help identify anomalies in market prices that might
provide opportunities for profit.
Take a hypothetical example. By analyzing stock-market
history, a quant trader might conclude that Ford shares
tend to move in tandem with those of General Motors.
That is not a big surprise, given that both companies
make mass-market cars and trucks, selling a lot of them
in the United States. But for any number of reasons,
the shares may not always move together, with one
company's shares rising or falling faster than the
other's.
Computers can spot when such anomalies occur and
recommend ways -- known as arbitrage -- to exploit that
inefficiency. A trader might buy shares in the laggard
or sell short the forerunner, or try the reverse in a
falling market.
The universe of quantitative traders seeking to exploit
such loopholes is diverse. Barclays Global Investors,
based in San Francisco, has some $95 billion in
actively managed quantitative funds. Fidelity and
PanAgora Asset Management, both located in Boston, run
some mutual funds that use quantitative techniques.
Big investment banks, including Morgan Stanley Dean
Witter, Swiss Bank Corp., and Donaldson, Lufkin and
Jenrette, at least dabble in quant trading, as do at
least a dozen big private hedge funds.
The field has attracted some of the most formidable
minds in math and economics. Robert Merton and Myron
Scholes, who won the Nobel Memorial Prize in Economic
Sciences in 1997 for their work on options pricing, are
partners in a hedge fund based in Greenwich, Conn.:
Long Term Capital Management, which is run by John
Meriwether. Doyne Farmer, a physics scholar, runs the
hopefully named Prediction Co. in Santa Fe, N.M.,
backed by Swiss Bank.
Among them is Shaw, who left Columbia University to set
up his own company. He works from a self-consciously
un-buttoned-down office in midtown New York, where
young traders dress in beach clothes and sandals and
the walls are back-lighted in green neon.
Though Shaw employs some 1,050 people, none visits
companies or reads stock reports. Instead, Shaw deploys
some $1.7 billion to exploit inefficiencies identified
by the company's computer calculations, trading shares
in volumes that can account for more than 5 percent of
the daily turnover of the New York Stock Exchange.
Some people familiar with Shaw's operations said he had
consistently made solid returns in both up markets and
down markets -- approaching 25 percent annually in
recent years.
But even Shaw sees the limits of quantitative trading.
Predictably, he does not want to call attention to his
track record, lest he attract more imitators. But that
fear illustrates the point that his trading formulas,
however secret, are usually absorbed by the market over
time.
"The inefficiencies we identify are very small, and we
have a small edge that gives us a small amount of
predictive value some of the time," he says. Moreover,
he says that his company does not need more money,
because if it throws its financial weight around too
much in pursuit of its small-margin arbitrage
opportunities, they tend to disappear even more
quickly.
In fact, the fastest-growing part of D.E. Shaw is not
proprietary trading -- buying and selling for its own
account -- but arranging trades for big institutional
clients.
The market inconsistencies the company seeks to
capitalize on are so small that trading costs often
make the difference between profits and losses. So it
has become skilled at buying and selling large blocks
of securities at low prices and without disrupting the
market, a skill it now offers to many nonquant clients.
Shaw's experience is not unique. Secret
quantitative-trading strategies, when they work at all,
do not usually work for long. One limiting factor is
competition. If one trader seems to be making a great
deal of money buying Ford shares and shorting General
Motors, others will sooner or later figure out why, and
end the arbitrage opportunity.
Another reason is that historical trends are not
reliable predictors of future events, though they might
hold true for a while. Under the hypothetical Ford-GM
scenario, Ford might hire an innovative chief executive
who overhauls production methods and catapults the
company well beyond GM, earning the Ford shares a
market premium and justifying a disparity that history
dictated should not exist. Computers running algorithms
fail to anticipate such fundamental shifts.
Eric Sorensen, head of quantitative research for
Salomon Smith Barney, says that some kinds of such
statistical arbitrage work well in stable markets but
fall apart when the market changes course abruptly.
"It's like flying -- boring for nine or 10 hours --
then you suddenly hit a storm," Sorensen said. "In good
times you can work out a routine, make a little bit of
money reliably. But you lose lots of money when a big
event changes everything."
Salomon itself worries about such sudden shifts. Once
known as one of Wall Street's largest proprietary
traders of debt instruments, it has de-emphasized its
"black box" secretive computer-trading strategies.
Salomon earlier this month closed its high-flying
bond-arbitrage operation, citing losses. But it and
many other big investment banks still try to keep
current on quantitative-trading theories.
People who follow hedge-fund investments say many
quants that play in the mortgage-backed securities
market -- including Long Term Capital, which is
reportedly down about 15 percent this year after four
years of strong gains, have recently suffered sharp
losses. These people attributed the losses to
unexpected swings in the rate at which homeowners repay
their mortgages. Long Term Capital officials declined
to comment.
Despite such setbacks, some experts in the field still
contend that it is only a matter of time before
computers running mathematical models replace human
judgment as the dominant force in stock picking.
But others say the unexpectedly slow development of
quantitative trading supports a long-held but
hard-to-prove belief about stock markets, especially
the United States' extraordinarily deep ones: They are
formidably efficient, meaning that stock prices are
fair reflections of what the investing public will pay
for shares at any given moment.
Any opportunities to profit from a difference between a
share price and what the market will really pay are so
limited, and shrinking, that the quest is ultimately
doomed.
"The idea of some Holy Grail strategy that works a
majority of the time is nice, but it is just a dream,"
says Edgar Peters, a quantitative strategist with
Panagora. Good thing, too. Because if Peters has it
right, any market that allows a small number of smart
people to make a disproportionate share of the profits
cannot last long.
"If anyone could really predict with any degree of
reliability," he added, "the market would die."
Copyright New York Time 1998 |
