NEWS by Jennifer Ouellette

Physicists Graduate from Wall Street ver the past decade, the number of fall in value. methods for predicting the stock market. OOPh.D. physicists employed in the finan- Hence, “risk management is more techni- Then, in the 1960s and early 1970s, Benoit cial community has increased dramatically. cal than ever,” says Neil Chriss, a vice presi- Mandelbrot—now widely known as the Once considered something of an anomaly dent and portfolio manager at Goldman “father of fractals” and an IBM Fellow Emeri- on Wall Street and in banking, physicists— Sachs Asset Management, who heads a fledg- tus at IBM’s T. J. Watson Research Center— and their fellow Ph.D.’s in mathematics, ling master’s program in financial mathemat- proposed a model of price variations that computer science, and engineering—have ics at . “The need to eventually evolved into the concept of frac- become a critical element to successful control risk has become a computationally tional Brownian motion in multifractal time. investment strategies, gradually replacing intensive problem, involving the ability to Among other conclusions, Mandelbrot, who many employees who lack strong statistical price many different assets quickly.” worked at IBM from 1958 to 1993, demon- and analytical backgrounds. Today, quanti- Not surprisingly, the problem-solving strated that wealth acquired on the stock tative methods are commonplace on Wall skills of physicists are useful in this capacity, market is typically acquired during a small Street, despite concerns about their predic- as are their abilities to view a problem in a number of highly favorable periods—a find- tive accuracy, and the proliferation of Ph.D. broader context, separate small effects from ing markedly different from the Brownian physicists in financial activities has made larger ones, and translate intuition about model, which predicts small gains consis- competition for these lucrative positions how something works into formal models. tently over time. more intensive than ever. “Bond traders will try to persuade you that A major turning point occurred in 1973, “Investing is increasingly becoming domi- there’s an emotional aspect that must be when economists Fischer Black and Myron nated by physicists, mathematicians, electri- understood behind certain bonds, but that Scholes devised an equation to calculate the cal engineers, and programmers,” says Adri- really isn’t the case,” says Cooper. “A bond value of options in simple derivative deal- an Cooper, founder and president of Wall is a mathematical instrument which per- ings, best described as an option to buy a Street Analytics (Palo Alto, CA), where forms according to precise characteristics, stock in the future at a specified price. (The roughly one-third of the employees are Ph.D. and in order to analyze it properly, you need term derivative is used because the value of physicists. Peter Carr, who heads the Equity people capable of understanding the math the contract derives from the value of the Derivatives Research Group at Bank of Amer- behind those characteristics.” underlying stock.) The Black-Scholes ica Securities (New York, NY), recalls that all Although physicists have helped foster the approach was later extended and applied to of his interviewers for his first position at widespread use of quantitative methods in more complex derivatives, particularly inter- were physicists. the financial community, the revolution est rate derivatives. Today, more than $14 Physicists in finance generally fall into two actually began with fundamental develop- trillion is invested in derivative securities, categories: those attempting to predict the ments in the mathematics of finance, dating three times as much as is invested in the stock market to achieve superior return, back to 1900, when Louis Bachelier intro- ordinary stocks and bonds from which they and—more commonly—those who duced a Brownian motion, or “random are derived, and the quantitative analysts use quantitative methods to walk,” model of price variations. In trading these staggering sums include many assess and manage investment 1953, mathematician Harry Ph.D. physicists. risk, a group known as quantita- Markowitz introduced his “Without the problem-solving skills of tive analysts, or “quants.” Invest- Nobel Prize-winning physicists, there would be a great employ- ment banks are highly leveraged work on mean-vari- ment shortage on Wall Street,” says Steven institutions, with book assets that ance analysis, which Shreve, a professor of mathematics at often greatly exceed the value of the gave birth to the Carnegie-Mellon University, because finan- firm. Their goal is to maintain a neutral use of quanti- cial institutions now use quantitative meth- position—a balance between gainers and tative ods to hedge risk in losers—as various assets in a portfolio rise and

9 The Industrial Physicist DECEMBER 1999 © American Institute of Physics News

trading derivative securities and other finan- els is to detect the mispricing of an asset, cial instruments. “Physicists didn’t create and make a trade based on [that],” says Nor- that fact, but they helped build the human man Packard, one of PC’s founders. resource needs of the banks.” Along with fellow high-energy physicist The demand for financial-modeling sys- Doyne Farmer, Packard developed a com- tems has driven the formation of numerous puterized system for beating the roulette start-up companies, many founded by Ph.D. wheel in the 1970s based on the then- physicists drawn to the industry by the tech- emerging field of chaos theory. They subse- nical challenges and potential monetary quently sold it to other entrepreneurs for fur- rewards. (Base salaries on Wall Street can be ther development. The pair then concluded as much as three times that of traditional that financial markets offered another exam- physics positions.) Cooper earned his Ph.D. ple of a complex system that might be in theoretical physics from Stanford Univer- amenable to predictive technology. They sity, but found himself comparing the career founded PC in 1991, and within a year, the satisfaction and financial rewards of Ph.D.’s company had signed an exclusive agreement his age who had followed the traditional to provide predictive signals and automated career path with those who had gone into trading systems to O’Connor and Associates finance. “It was pretty clear which direction (now part of Swiss Bank), a highly successful was more appealing,” he says. Cooper went Chicago-based trading firm that had made on to found Wall Street Analytics, which millions in derivatives trading using the This fractal develops software for modeling financial sys- Black-Scholes equation. design, known as tems for mortgage-pool investments. The skeptics, however, still remain uncon- “Leather,” is wildly com- Nigel Goldenfeld, professor of physics at vinced. “If they could do it, they wouldn’t be the University of Illinois, Urbana-Cham- wasting their time with a company. They plicated, yet B. B. Mandelbrot paign (UIUC), earned his Ph.D. from the would just be sitting there buying and selling obtained it by iterating a very University of Cambridge in England and spe- IBM share options,” Cooper says. simple “dynamical” rule. Financial cializes in statistical, theoretical, and compu- Prevailing attitudes among academics tational physics. His first Ph.D. student at toward physicists working on Wall Street data are also wildly complicated, UIUC ended up working for Goldman have changed in the last decade. Emanuel yet many of their features are Sachs, which sparked Goldenfeld’s interest Derman, who earned his Ph.D. in physics reproduced by Mandelbrot’s in the physics of finance. Convinced he from Columbia University in the 1970s, is could improve on the calculation techniques now a managing director at multifractal model of price used, he founded NumeriX in 1996 with fel- and head of its Quantitative Strategies variation, which is also low physicists Alexander Sokol and Mitchell Group. He finds that the financial world is of surprising simplicity. Feigenbaum and entrepreneur Michael no longer viewed as a second-rate “alterna- Benoit Mandelbrot Goodkin. NumeriX is a New York-based ven- tive” career for physicists unable to obtain ture that markets fast numerical software positions in academia and industry. Instead, programming isn’t enough to land a job on products for derivative-risk management. finance now is a highly desirable first choice, Wall Street. New Ph.D. physicists also need a Physicists attempting to predict the stock as evidenced by the number of tenured pro- basic understanding of options, pricing theo- market look for patterns in the data of stock fessors who have left their academic posi- ry, interest rate theory, and other foundations prices and foreign-exchange markets. The tions for more lucrative careers in finance. of finance to be considered. “The stakes are Prediction Company (PC), based in Santa Ironically, physicists interested in pursu- higher than they’ve ever been in terms of the Fe, New Mexico, is perhaps the best-known ing careers in finance today may have become level of knowledge expected for entry-level company focusing on this sector of finance. victims of their predecessors’ success. “It [quant] positions,” says Chriss. The company develops advanced forecasting used to be a gravy train for physicists and This increased competitiveness has technologies for prediction and computer- other mathematically oriented people, but spurred the formation of numerous master’s ized trading of financial instruments, based now the job market has become saturated,” degree programs in computational finance on the assumption that the stock market is says Carr. Few financial houses are hiring (also called financial engineering or mathe- not completely random but has short-term additional quants, and today the ability to matics in finance) at institutions around the pockets of predictability. “Our task for mod- solve differential equations and perform basic country. Carnegie Mellon University pio-

10 The Industrial Physicist neered the concept in 1994 with a 12-month Although he supports the rationale for in physics, mathematics, and other sub- program combining coursework from four master’s programs in computational finance, jects,” Cooper says. “If the physics depart- separate academic departments: mathemati- Cooper argues the need to preserve the tradi- ments aren’t protecting the subject of cal sciences, statistics, computer science, tional focus of university physics programs. physics, who is going to do it?” But Packard and business. His primary concern is that the growing believes the threat of a “brain drain” is not Other schools have followed suit, includ- emphasis on finance as a career option for limited to finance. “Financial markets are the ing Purdue University, the Massachusetts physicists could undermine graduate educa- least of physics’ worries,” he says. “The field Institute of Technology, Columbia Universi- tion and turn Ph.D. candidates into mort- is facing an even more severe brain drain ty, Cornell University, the University of gage traders too early in their development. from a number of other areas, such as elec- Michigan, the , and Several universities are mulling the possibili- tronics and computer engineering.” New York University. At UIUC, students are ty of running physics and finance graduate As successful as today’s financial models able to complete an accelerated master’s in programs in tandem. “These are the universi- have been, there remains substantial room finance program in conjunction with their ties that have been entrusted to pass on the for improvement. Although pleased at the physics Ph.D.’s. learning that has taken generations to amass increased trust placed in quantitative meth-

11 The Industrial Physicist News

ods by the financial community, Thierry the heart of what goes on in financial mar- Kaufmann, a theoretical physicist who heads kets,” Mandelbrot concludes. Underestimat- Purdue’s computational finance master’s pro- ing the frequency of 10 sigma events, a tech- gram, admits that existing models are not nical term for enormous price fluctuations, 100% accurate, which poses serious potential can have serious global economic implica- consequences. “Sometimes excessive trust is tions, as evidenced by last year’s fears of placed in these quantitative results by people damage to financial markets worldwide who lack the proper background and make brought on by heavily leveraged trading by a very risky decisions based on them,” he says. called Long Term Capital Man- “They can lose a lot of money.” agement. This potential threat was narrowly Although the models used for equity averted by a bailout of the hedge fund paid derivatives have proven fairly robust, many for by major investment houses. surprises still take place in markets heavily Even academic physicists are beginning to dependent on bonds and interest-rate move- look more critically at some of the prevailing ments and in volatile foreign-exchange mar- modeling assumptions used by the financial kets such as Indonesia. Part of the problem, community. “It’s becoming its own disci- says Goldenfeld, is that many models were pline,” says Derman. In fact, financial markets created with an eye to being easily calcula- provide an excellent practical field of study for ble. Hence, these models are not faithful to those interested in the behavior of complex the complexities of real market dynamics. and nonequilibrium dynamical systems “It’s no good having people help you calcu- because there is a wealth of data available. late if you’re using the wrong model,” he “Previous attempts to look at complex sys- says. “You’re getting the wrong answer faster tems, in my view, have not been successful and more accurately. You want to be able to because people have operated at a level of get the right answer fast and accurately.” generalities rather than rolling up their Among the sharpest critics has been Man- sleeves and doing honest spadework,” says delbrot himself, now Abraham Robinson Pro- Goldenfeld. “What’s happening now is fessor of Mathematical Sciences at Yale Uni- exactly what I had hoped: people are digging versity, who believes that the current models in and trying to understand the financial seriously underestimate the frequency of markets from a physicist’s perspective rather large fluctuations in stock value. For exam- than that of a financial economist.” ple, Alcatel, a French telecommunications However, he remains a strong proponent equipment manufacturer, experienced severe of the value of firsthand experience with volatility in its stock prices last year, which financial markets when studying such sys- fell 40% in one day, fell another 3% over the tems. “Thermodynamics was invented by next three days, and then rebounded by 10% engineers who wanted to make steam on the fourth day. “The classical financial engines, not by people thinking about quan- models used for most of this century predict tum states and other abstract concepts,” that such ‘10 sigma’ precipitous events Goldenfeld says. should never happen,” Mandelbrot says, with estimated probabilities of a few millionths of For further reading: a millionth of a millionth of a millionth. In Kelly, K. “Cracking Wall Street,” Wired, reality, such spikes occur quite regularly—as June 1994. often as every month—with probabilities Mandelbrot, B. B. “A Multifractal Walk closer to a few hundredths. Far from varying Down Wall Street,” Scientific American, Feb- continuously, as such models tend to ruary 1999. assume, prices oscillate wildly, often discon- Mandelbrot, B. B. (Ed.). Fractals and Scal- tinuously, at all time scales ing in Finance: Discontinuity, Concentration, “Volatility, far from being a static entity to Risk, Springer-Verlag, New York, 1997; 456 be ignored or easily compensated for, is at pp., ISBN 0-387-98363-5. Ω

13 The Industrial Physicist