My Life as a Quant (20 page)

Even when my managers graciously allowed me to telecommute from home one day a week, I chafed at my shackles. I yearned for academic life and considered retracing my steps, exploring how to become a long-term physics postdoc again with Baqi Bég at Rockefeller or Norman Christ at Columbia.

Then, in late 1983, Wall Street began to beckon.

The first hints of my life-to-be began with occasional calls from head-hunters in New York City. Soon, we all knew their legendary names—Jory Marino, Rick Wastrom of Smith Hanley, Rita Raz of Analytics, Steve Markman of Pencom, to name just a few of those who preyed on Bell Labs. Many of them are still active now, still advertising in the Sunday
New York Times
or on the Internet. Out of the blue, someone you had never heard of would cold-call you at work and ask if you wanted a job that paid $150,000—a huge amount in those days for an ex-physicist making less than $50K—and then command you, urgently, to come to his or her office immediately, before it was too late. These recruiters might flatter you by saying they had heard of you; in fact some acquaintance at the Labs had simply given them your name, or they had found it in a Bell Labs internal directory wheedled from some disgruntled employee. A few times, at their behest, I would leave work early and drive back into Manhattan to meet one of them, putting on my rarely worn navy blazer and 1960s knitted tie in a poor effort to simulate business attire. I owned no suits. Headhunters could keep you waiting for hours, like God as the Puerto Rican janitor in Bruce Jay Friedman's
Steambath
I saw on PBS around that time. Many headhunters were imperious, high-handed and low-mannered, and none of us knew enough to doubt their implicit claim that they held the keys to the kingdom.

I had several mostly unpleasant interviews at Wall Street firms in late 1983. I knew nothing about options theory and regarded myself as a software person. In those days, information technology on Wall Street was a land of mainframers who were educated in COBOL or FORTRAN or MIS and you needed only one eye to be king. At the Labs I had learned to design and maintain programs with a sophistication beyond that of most of the renegade scientists and coders on the Street. On interviews, because of my experience with HEQS, I advertised myself as someone who could design a computer language to match the Street's modeling needs.

One of my first interviews was with Zach Cobrinik, now an ex-Goldman partner and then a very boyish member of the Goldman Sachs Quantitative Strategies group that I headed seven years later. At that time the group was run by David Weinberger, who shortly thereafter left for O'Connor in Chicago. Zach seemed most interested in finding a systems administrator for their VAX, a task of little interest to me, and so I declined to continue the process.

Some months later, another headhunter sent me to Salomon Brothers about a position whose nature was never explained. I spent no more than ten minutes naively informing my interviewer about the benefits I could bring them by solving financial models with simultaneous algebraic equations using HEQS. Then, in the middle of our conversation he was interrupted by a colleague; he suddenly stood up, apologized curtly, and left for something that he said had urgently arisen, promising to resume our discussion at some future time. Neither he nor the head-hunter ever called again. Despite repeated attempts, I could never get through to the woman in Personnel who arranged the interview; she was always unavailable. For a long time I thought they were simply very busy; finally I realized that this was their way of brushing me off. I never understood why it wouldn't have been easier to tell me that they simply weren't interested.

Then, in late 1983, when Eva was pregnant with our daughter Sonya, a headhunter I knew arranged an interview for me in Stan Diller's Financial Strategies Group (FSG) in the Fixed Income Division at Goldman Sachs. There I met Ravi Dattatreya, a former PhD engineer from Bell Labs who had himself migrated to the Street a short while earlier.

Stan's group addressed a new need for quantitative modeling at Goldman. By tradition a gentlemanly equity house that concentrated on IPOs and stock trading for large institutional clients, Goldman had recently begun to step onto Salomon's turf, the hurly-burly and plebeian world of bonds and mortgages. Stock trading was a simple, gutsy, risk-taking business that required little intellectual or technical capital. Bonds were more complex; they involved numbers, arithmetic, algebra, even calculus. As my trader friend said, there's no competitive edge to being smart in equities.

The trick in the stock market was to estimate the right price for a share of stock in one of several thousand different companies. The bond market comprised fewer securities, but each security was more complex, sometimes bewilderingly so. The 900-pound gorilla in the market was the American Government, the source of liquid, ever-flowing Treasury bills, notes, and bonds that sprang from the government's need to borrow. Government bonds were characterized by a wide choice of maturities and coupons and, since they were backed by the full faith and credit of the United States, they would never default. Foreign governments issued bonds, too, some of them more prone to default than others. Corporate bonds were even riskier; a company might run out of cash and then be unable to make their promised payments. Some bonds could be “called”—the corporation had the option to end its obligation to keep paying a high rate of interest on its loan after interest rates had dropped by prematurely repaying the amount they had borrowed. Mortgages sold by homeowners were among the most horrifically elaborate bonds; they too could be suddenly called (that is, prepaid) by homeowners who had either sold their houses and no longer needed the principal they had borrowed, or who had discovered that they could now get cheaper financing as interest rates dropped.

All this complexity made it difficult to decide where value lay, and thus opened the door to mathematically adept modelers on the Street. The Fixed Income division at Salomon relied for their mathematical analysis on Marty Liebowitz's renowned, large, and enviably experienced Bond Portfolio Analysis (BPA) group. Goldman was slowly realizing that they needed something similar. FSG and its leader, Stan Diller, a former Columbia University PhD in Economics and one of the earliest academics to cross over to the Street, was their solution.

The problem on everyone's mind was the sudden increase in the volatility of interest rates. Before the 1980s, investors used to apportion their investments between stocks and bonds in a fairly static sort of way. Traditionally, they viewed the bond portion as safe and the stock stake as risky. Then, in the late 1970s, about the time I was teaching physics at Boulder, American interest rates rocketed and gold and oil prices soared. Bonds, previously perceived as nonvolatile, became risky. While everyone knew that stock investors might have to stomach a 40 percent drop in a bear market, few investors had imagined something similar could happen to Treasury bonds. The trading desks at investment banks, which naturally carried large inventories of bonds with which to supply their customers, saw their portfolios abruptly decline in value. As the intrinsic riskiness of fixed-income securities became manifest, a new approach to managing interest-rate risk began to diffuse through the industry. Desks wanted to hedge their varied and complex bond portfolios with large offsetting trades in cheap liquid bond futures. Hedging and risk management became the new thing, crucial both for wholesalers like Goldman Sachs and for institutional investors, too.

In the universe of equities, arithmetic was enough to do business; you didn't have to worry about even the simplest algebra until you got involved with options. In fixed-income land, by contrast, investors measured a bond's value by its yield, the average percentage return you would earn over its lifetime if you bought it at its current market price and received all future payments of interest and principal. As soon as you began to contemplate the relation between a bond price and its yield, the storm clouds of algebra, sequences, series, and, finally, calculus loomed blackly in the appendices at the backs of textbooks. A stock is only a stock, but even the simplest bond is a derivative security whose value depends on interest rates.

Therefore, suddenly, bond traders in the early 1980s needed analytic and mathematical skills to understand a portfolio of hundreds or thousands of bonds and its characteristics. They also required computing power; paper and pencil, a book of yield tables, even a hand-held calculator, were too slow and inflexible to amalgamate all the angles. Only on a computer could you estimate the value, sensitivity, and risk of many different securities in real time.

You couldn't buy commercial risk-management systems in the late 1970s. Do-it-yourself tools like spreadsheets were not yet ubiquitous. Most of the programmers in information technology couldn't handle bond mathematics, and most of the traders on the desk were unable to program. A trading desk had to turn to some jack-of-all-trades who could build their risk management tools, from the low-level databases through the financial valuation models to the high-level user interface.

The jack-of-all-trades in that era was unlikely to be an MBA or a PhD in Finance—even if they knew enough quantitative finance, most of them looked down with disdain on programming and mathematics as cheap, geeky skills they could pay other people to do for them. Mathematicians, too, tended to avoid programming, preferring analysis to computation. Computer scientists, though they knew discrete mathematics and Boolean algebra, were often uncomfortable with continuous-time mathematics.

PhDs in physics or engineering fit the jack-of-all-trades bill pretty well. First, the mathematics of finance closely resembles the mathematics of physics. Furthermore, physicists don't grow up wearing white gloves; they have no scruples about tackling tasks beneath their so-called dignity. They do their own math and programming; the willingness to do so is an essential part of graduate student and postdoc subculture.

This was probably why Stan's Financial Strategies Group consisted mostly of former physicists, applied mathematicians, and engineers, many with PhDs. All of Stan's hires came from a culture in which you did your own dirty work—you developed your own theory, did your own mathematics, and then wrote your own programs. It was a hiring model I tended to repeat a few years later, when I staffed the groups I led, not out of principle but rather because of a natural affinity for that intellectual style.

When I first started interviewing on the Street, Stan was among the most famous practitioner-quants. An article in
Forbes
about him in the early 1980s was titled “Diller's Dillies,” a patronizingly flattering reference to the awkward, foreign quant-nerds the reporter met. Stan had a reputation for hiring foreigners; I had heard some people imply that he liked inarticulate technophiles so that he could manage their work and present their results, but I think that was unfair. Most quants, then and now, came from abroad because immigrants often see the quickest path to success in hands-on work. It's the next generation that prefers management and business school.

Stan wrote occasional lengthy research reports on his group's work, original and creative papers phrased in a distinctively unorthodox and inbred style.
1
They were insightful and intuitive, but hard to classify and always a touch off-center: too technical for early-1980s Wall Street, not quite rigorous enough for real finance academics—and insufficiently hard-sell for a sales piece. As a consequence Stan gained less influence than he deserved. But his work on the options embedded in mortgage portfolios was prescient, foreshadowing the more formal and rigorous work of the growing number of finance academics already entering the field.

When I met them, Diller and Co. were already pulling together a team skilled in finance, mathematics, and programming, and they wanted me to join them. I think it was my software skills that appealed to them as much as my talents in physics. Stan was a true pioneer in embedding financial models in portfolio trading systems, and a good decade ahead of his time in understanding the importance of professional software engineering in this endeavor. Later, after leaving Goldman Sachs in 1985, he built AutoBond, an early mortgage portfolio valuation system at Bear Stearns. Today he runs Polypaths, a company that produces fixed-income portfolio analysis software.

Nowadays, the cosmos of trading systems is very different. PCs are ubiquitous, spreadsheets are easy to use, and risk management software is increasingly available from tens of companies selling everything from building blocks to turnkey systems. Nevertheless, the largest banks still build their own software in order to book, value, and hedge the latest products as soon as they come to market. Even today, though, risk systems are balkanized, each one focusing on one or two product classes at most. There is still room out there for a system or language which can handle all the classes of securities—mortgages, swaptions, currencies, equities, metals, energy derivatives, and so on—that a large firm trades.

I returned to Goldman for another inspection with no idea what to look for. An article in the
New York Times
about switching jobs suggested asking your potential employer what you would be doing ten years in the future. At the end of my second visit there I sat down again with Stan. He explained to me that Wall Street was one of the few places you could eventually make $150,000 a year without running your own small business like an accountant or a doctor.

“If I come here,” I said, “What will I be doing ten years from now?” Stan became instantaneously irate.

“In ten years,” he proclaimed, “You'll be doing whatever the hell you'll be doing now, only making more money! I don't want to be jerked around by someone coming in here and thinking about doing something different!”

I couldn't comprehend then what infuriated him. Ten years later, I understood his frustration. After a few years on the job, quants in banking often grow envious of the better-paid traders and salespeople who sit in the driver's seat. He must have thought that I was applying to his Financial Strategies Group with the devious plan to insinuate myself into trading. The last thing Stan needed was a new employee who wanted to tunnel out of the quantitative group into the business area before even beginning his work.