Stat Arbitrageurs: Merchants of Volatility

Neil OHara

Statistical arbitrageurs have demonstrated the ability to make money during tough times. “Stat arb tries to remove market risk to a degree that you really don’t see in any other strategy,” says Alain Sunier, portfolio manager of Highbridge Capital.

Like many hedge fund firms, Highbridge Capital Management had an absolutely miserable year in 2008. Its flagship Highbridge Capital Corp. multistrategy fund, which previously had experienced only one down year since Highbridge co-founders Glenn Dubin and Henry Swieca launched it in 1992, fell by more than 27 percent. Its convertible arbitrage portfolios, long a strength of the New York–based multistrategy powerhouse, did even worse, dropping about 45 percent. Between the losses and investor redemptions, the firm saw its hedge fund assets under management plummet from $27.8 billion at the start of 2008 to $17.3 billion when this year began. 

Highbridge, which is majority-owned by banking giant JPMorgan Chase & Co., had one bright spot, however. Its statistical arbitrage hedge fund was up more than 22 percent in 2008, taking advantage of some of the same forces that sent its other funds tumbling. As Highbridge portfolio manager Alain Sunier explains, statistical arbitrage funds profited last year from equity price distortions caused by sudden drops in liquidity and the spikes in volatility that typically accompany them. 

“Stat arbs take the other side of a move,” Sunier says. “We provide liquidity. If it’s a good time to do that, we earn an economic rent as prices return to normal.” 

In a world as desperate for liquidity as most equity markets were in 2008, rents soared for anyone willing to supply it — stat arbs included. And with no sign that equity markets will settle down anytime soon, stat arb funds look poised to enjoy success through 2009 and perhaps beyond. 

Statistical arbitrageurs employ a quantitative rapid-fire trading technique that relies on computer models to detect anomalous price movements among equity securities that normally trade in tandem. Although stat arbs as a group didn’t come close to matching Highbridge’s performance — Morningstar’s statistical arbitrage index fell 6.21 percent in 2008 — they did well in a year when the Morningstar 1000 hedge fund index dropped 22.41 percent and the Standard & Poor’s 500 index declined by 38.5 percent. And though most hedge fund managers suffered their worst losses in the fourth quarter, stat arbs managed to deliver positive returns as the year came to a close: The Morningstar statistical arbitrage index was up 1.44 percent in the final three months of 2008, compared with a 10.62 percent loss for the broader Morningstar 1000. 

It’s a dramatic turnaround for a strategy that had been all but left for dead after the summer of 2007. During the first two weeks of August that year, statistical arbitrageurs — including major players like AQR Capital Management, D.E. Shaw & Co., Goldman Sachs Asset Management and Renaissance Technologies Corp. — suffered huge losses. Goldman’s once–$5 billion Global Equity Opportunities Fund, for example, was down 30 percent; the then–$1.7 billion Highbridge Statistical Opportunities Fund fell 18 percent. For managers like Goldman and Highbridge that were able to hold on, performance snapped back later that month, but those that were forced to liquidate missed the rebound. In retrospect the crippling losses were more the result of margin calls that originated in the credit markets than of any flaw in statistical arbitrage theory, according to Andrew Lo, a finance professor at the MIT Sloan School of Management. 

“There was some kind of unwinding, most likely due to a multistrategy fund that needed to raise cash to meet margin calls for other investments,” says Lo. 

Investors nonetheless were spooked and took flight. Judith Posnikoff, a managing director and co-founder of Pacific Alternative Asset Management Co., an Irvine, California–based firm that manages about $9 billion in funds of hedge funds, estimates that between one third and one half of the hedge fund capital dedicated to statistical arbitrage had fled the strategy by early last year. At the same time, proprietary trading desks at many of the big investment banks also got out of the game. The exodus set the stage for a rebound in 2008. Lo points out that stat arbs tend to be long volatility, which shot up to record levels in the fourth quarter of 2008. 

“Spreads widened, volatility increased, and as a result, the profitability of these strategies grew,” says Lo, who is also founder, chairman and chief scientific officer of Cambridge, Massachusetts–based AlphaSimplex Group, a $700 million quantitative hedge fund manager. “Those were ideal conditions for stat arb.” 

Stastical arbitrageurs behave like quasi–market makers — buying stock at the bid price or selling it at the offer, but not doing both at the same time. True market makers, of course, have an obligation to do both. That makes them vulnerable when stocks are moving quickly; they can end up short in a rising market or long in a falling market. To compensate for that risk — known as being “short gamma” in traders’ lingo — market makers widen bid-offer spreads in volatile markets. Although stat arbs are not required to make a two-way price, they still benefit from the wider spreads. 

“The strategy makes money by providing structured liquidity to the market,” explains Paul Simpson, who co-manages two statistical arbitrage funds for Old Mutual Asset Managers in London. “As volatility increases, so does the opportunity to profit from more extreme price action.” 

Managers credit their fancy computer models rather than market structure as the source of their alpha. Every stat arb fund has a different model — often more than one — but Simpson says the models typically track the relationship between either stocks that normally move in the same direction at the same time (correlation) or stocks that don’t always move in the same direction but fluctuate over time around an observable constant such as a dollar spread or a price ratio (co-integration). 

Simpson knows whereof he speaks. After starting his career in the late 1980s in risk management, he worked as a proprietary trader doing equity arbitrage at several major securities houses, including Deutsche Bank and UBS. He eventually ended up at hedge fund Millennium Capital Management in London, where he and colleague John Dow jointly managed the statistical arbitrage program. In January 2006, Old Mutual hired Simpson and Dow, who together run about $320 million in two statistical arbitrage funds using models they built about seven years ago and continue to refine. Their flagship fund returned 18 percent in 2008. 

Like almost all stat arb funds, the Old Mutual portfolio is market-neutral — meaning its returns owe nothing to movements of the market as a whole. Although some stat arbs may use fundamental information about equities, such as price-earnings ratios, in their models, most — including Simpson and Dow — do not. 

“Our model is entirely driven by relative price evolution,” Simpson notes. “Mean reversion is a well-observed phenomenon in equity markets that exists at every time interval from a few minutes out to several years.” Simpson and Dow’s flagship fund trades large liquid equities on a global basis; the average holding period for each position is typically measured in days, not weeks. 

Old Mutual’s average holding period, in fact, is shorter than that of many stat arbs (Paamco’s Posnikoff estimates the industry average to be two to three weeks). But even a week is practically an eternity for Marek Fludzinski, whose New York–based Thales Fund Management turns over 10 to 40 percent of its $1.2 billion portfolio every day. Flud­zinski, who has a Ph.D. in theoretical physics from Prince ton University, applies mathematical principles to extract tiny distortions from mountains of data and thereby predict market market moves that may last for a few hours at most. 

He has been doing statistical arbitrage since the early 1990s, first at D.E. Shaw and then at Swiss Bank Corp. In 1994 he started Thales Financial Group with seed money from Paloma Partners Management Co., the now–$2 billion Greenwich, Connecticut–based investment firm founded by Donald Sussman. In January 1999, Fludzinski launched the Thales Fund, which had a largely successful run until it hit the skids in August 2007. Thales finished that year down 8 percent and suffered significant withdrawals as a result. Instead of soldiering on with a model that was no longer delivering the goods (it looked for trends that would play out over three to ten days), Fludzinski elected to return investors’ money. He launched a new fund last May using what he says iss an improved model. He shortened the average duration of the trends he looks for and built in an optimization routine that chooses the best time scale for each trade. 

Thales wasn’t the only shop to experience redemptions after the 2007 carnage. In fact, even managers who did well suffered withdrawals, not only at the end of 2007 but all through 2008. Paamco’s Posnikoff says stat arbs offer investors better liquidity than most hedge fund strategies do — they can’t justify long lockups for portfolios that turn over every week or two — so investors who needed cash dipped into stat arb funds as if they were ATMs. 

The credit crunch exacerbated the pressure on stat arb managers, as those who stayed in the game had to cut back their leverage too. Before August 2007 it was not unusual for statistical arbitrage funds to use as much as eight times leverage, says MIT’s Lo. “Nowadays leverage has come down dramatically,” he notes. “You would be hard pressed to find a stat arb manager getting more than five to six times leverage, and the typical leverage being used is probably three to four times.” 

Highbridge’s Sunier says that as a result stat arb was better placed than many other strategies to weather the market storms in 2008. “The supply-demand balance has been quite favorable for the short-term forecasts that lie at the heart of stat arb profitability,” he explains. “I do feel that those conditions will remain with us for some time.” 

Sunier is not alone. Posnikoff projects that the CBOE volatility index, or VIX, a key measure of the U.S. equity market’s expected near-term volatility as expressed by the prices of S&P 500 index options, will remain high through the end of this year, trading between 40 and 60. To protect against the eventual decline in volatility, she favors high-frequency statistical arbitrage funds that turn over their portfolios every two or three days. “The short time frame adapts to changes in volatility more quickly,” Posnikoff explains. She expects these funds to thrive even when the markets calm down and volatility recedes to a more normal range — a VIX of 20 to 35. Her biggest challenge is to find capacity; managers who trade like whirling dervishes can handle only a limited amount of assets. 

Fludzinski, for his part, intends to cap Thales’s gross assets — its longs plus its shorts — at about $2.5 billion. He is almost halfway there, having raised some $300 million in capital for the new fund, which has a leverage target of four to one. “The limitation is liquidity, the need to trade in and out of stocks quickly,” says Fludzinski, whose fund was up about 13 percent last year in just eight months of operation. Thales’s positions tend to be in large- and midcap stocks that can accommodate the frenetic pace of trading. 

Although high volatility usually boosts returns for stat arbs, too much of a good thing can be harmful. In extreme market conditions indiscriminate forced selling overwhelms normal trading relationships and the models stat arbs use don’t work. Jaeson Dubrovay, senior consultant for hedge funds at Cambridge, Massachusetts–based investment consulting firm NEPC, sees the sweet spot for stat arbs in a VIX range of 20 to 40. “Below 20 nothing is moving, and above 40 everything is moving either up or down together,” he says. “Stat arbs need dispersion, where the best stocks outperform the worst stocks.” Dubrovay expects volatility to slide toward that sweet spot later this year and foresees an increase in dispersion too. Not surprisingly, he favors an allocation to statistical arbitrage at the moment as part of a diversified portfolio for clients who invest in hedge funds. 

At a time when investor confidence is low, liquidity is scarce and market volatility remains high — albeit lower than the record levels seen last fall — stat arbs are likely to enjoy a benign environment for some time to come. But in theory, statistical arbitrage is a strategy for all seasons because it depends on short-term relative price movements, not the intrinsic value of securities. The returns it generates are particularly attractive on a risk-adjusted basis even when market conditions are not so favorable. 

“Stat arb tries to remove market risk to a degree that you really don’t see in any other strategy,” says Highbridge’s Sunier.