By Eric Uhlfelder
With $2.1 billion in assets, John Burbank’s Passport Capital has long had a reputation for taking controversial positions that create volatile swings in returns. Investors have learned to brace themselves for the ride, believing they’d come out on top in the end. Even last year’s dramatic losses at Passport, which is based in San Francisco, didn’t change that. Though the flagship fund lost 51% for the year, its three-year annualized returns were still above 21% at the end of 2008.
What did change, however, was the way the California hedge fund looks at risk.
But, similar to many hedge fund managers, Passport has had to redefine how it looks at risk. Managers are placing more emphasis on liquidity, rethinking the value of traditional risk-modeling tools such as Value at Risk, or VaR, and taking account of other factors that can affect the markets, such as the concentrated positions connected with quant models.
Passport says its current emphasis on liquidity is crucial given that the firm runs concentrated portfolios. “We believe the return on illiquid securities is no longer enough to compensate for the risk,” says Garry. Now, Passport designs the portfolio so that it can sell more than half the fund’s assets within 10 days if necessary, he explains—and Garry would like to get that up to 75% by yearend. Passport hasn’t tossed the standard measure of risk—the controversial VaR—but it is tweaking it.
“One always must be mindful that VaR, and any version thereof, is helpful to show potential trouble,” says Garry, “but discretionary interpretation is always necessary to fully respond to risk.”
The firm is also taking the unusual step of running live quant models that simulate programs, based on factors such as earnings estimate revisions and price momentum, that Garry believes drive about one-third of all U.S. equity trades. “This is helpful in understanding a key market force,” he says, “and how it may impact Passport’s portfolios.”
Garry attributes the firm’s rebound this year (Passport Global Strategy was up 21% through July) to the re-emergence of liquidity, which helped long-term emerging market investments recover, along with a fairly stable yield curve. Not until there is another crisis will Garry know whether the changes the firm has made to its risk management are working.
Much of the current debate regarding the adequacy of risk management centers around VaR, a risk metric designed by derivatives experts at firms such as Bankers Trust and JPMorgan, that evolved in the aftermath of the 1987 stock market crash.
“It was invented to determine how much a manager could lose in one day with 95% to 99% accuracy,” explains Aaron Brown, chief risk manager at quant firm AQR Capital Management, which runs $7.8 billion in hedge funds, down 14% from a year ago. Since then, he says, AQR, of Greenwich, Conn., has improved its liquidity modeling and reduced its counterparty collateral exposure.
Brown says that VaR is a risk-warning device, telling managers when daily statistics and models are no longer sufficient to trust as guidance and when to shift focus to more rigorous stress testing to get a better read of near-term risk. “Not adequately interpreting the warning signals VaR was giving off as the current crisis unfolded was one of the key failures in risk management,” says Brown.
But VaR doesn’t tell you what to do when it indicates that you need to go beyond your models. For instance, when discrepancies appeared between VaR and stress tests, Brown says that a common response was to simply multiply VaR readings by three to presume the worst-case scenario. This practice was akin to how structural engineers would anticipate worst-possible structural stresses—taking their calculations, then multiplying them by a factor to believe they’ve accounted for the most intense forces. But that doesn’t work in finance when the degree of potential stress is unknown.
During each financial meltdown that has occurred since the creation of VaR, critics have pointed to its limitations. Take the issue of liquidity. VaR takes into account the liquidity difference between private equity assets and those of a publicly traded company. But it is unable to take into account when mark-to-market pricing no longer functions because liquidity freezes up, as it did for a protracted period late last year. VaR knows that a thinly traded stock can move more than a widely traded one. But it doesn’t know the difference between assets that take two days to sell versus ones that take two years to sell. And this problem gets exacerbated when a manager must sell his most liquid assets and is left with a portfolio that’s increasingly less liquid. The VaR may stay the same, but the risk may rise substantially.
Another point in question is leverage. VaR does incorporate some of the risks associated with leverage. A position levered two to one has twice the volatility of an unlevered one, and twice the VaR. Brown says, however, that VaR doesn’t differentiate between the duration of financing behind leverage: overnight versus long-term repo. Brown says, “A lot of folks ignored this basic risk associated with the type of leverage used, which required a basic plan to respond when refinancing is challenged.”
VaR knows an option is riskier than the underlying asset. But, says Brown, it is not designed to account for how leverage changes over periods longer than a day. If you buy an option and it moves against you, your leverage goes down, limiting your risk tomorrow. If you sell an option and it moves against you, your leverage goes up, increasing your risk tomorrow. VaR only measures your risk today.
Ken Tropin, chairman of $5.7 billion Graham Capital Management, agrees that VaR doesn’t inherently look at liquidity or reflect leverage, but says he would much rather have the tool than not. “VaR can be a good flag for changes in overall risk profile from one day to the next,” says Tropin, “but we think it’s critical to use additional analytical tools to address its shortcomings. That most frequently means stress testing.” He has sets of scenarios customized by portfolio that look at more extreme events. Because of its limitations, Tropin thinks VaR is more useful for global macro than relative value and option trading strategies, which involve intermarket relationships.
Like both Passport and AQR, Graham is also focusing on liquidity. Its 11 discretionary traders are surveyed weekly about how much it would cost to liquidate their entire positions over one, three, and five-day periods based on market liquidity and position size in both normal and distressed markets. The firm is concerned when projected liquidation expenses (related to expedited selling) exceed 3%, at which point it might take steps to ratchet back. Graham also keeps track of its exposure to crowded trades, determined qualitatively through discussions with counterparties and traders, as well as the assessment of bid-offer spreads.
Fortunately for Graham, it enhanced risk management in November 2007, well before the crisis exploded. Since then, the firm instituted daily rather than biweekly risk-management meetings with all executives. In these meetings, Tropin explains, “the group reviews external risks [liquidity, counterparty credit, bank safety where cash is held, and those associated with clearing and executing brokers] and internal risks [positions, performance and current drawdowns of each discretionary trader].” He thinks the daily analysis of trends and risk, which allows Graham to respond more quickly to unexpected changes in the market, is a key to confronting the truly unpredictable.
So far, so good. According to its investor letters, Graham’s Proprietary Matrix Portfolio, which reflects the firm’s systematic and discretionary trading strategies, was up 9.36% this year through August, and nearly 30% for the past 12 months.
But not everyone is keen on VaR. Last year, David Einhorn, the founder of $6 billion Greenlight Capital, made waves when he said that VaR was “relatively useless as a risk-management tool and potentially catastrophic when its use creates a false sense of security among senior managers and watchdogs. This is like an air bag that works all the time, except when you have a car accident.”
Others have been even harsher. Nassim Nicholas Taleb, the best-selling author of “The Black Swan,” calls VaR and other mathematically sophisticated models “a fraud.” He believes, in fact, that they’ve done more harm than good because they only measure events that have already occurred. And as history shows, the most severe hits are taken from events that are considered highly improbable. He believes such events can’t be measured.
VaR proponents, such as Ron Papanek, head of business strategy at RiskMetrics Group, an independent financial risk consultant, says the problem is not in models but in people. The drive for short-term profitability ended up trumping longer-term concerns, he argues. “Intensive competitiveness among hedge fund managers to attract investor capital,” says Papanek, “drove many to imprudently shift their focus away from systemic and specific risks. As a result, the industry rewarded managers who took on excessive leverage and exposure to riskier assets over those who were demonstrating greater discretion.”
He believes managers made faulty assumptions and used inappropriate distribution models as inputs into VaR. In large part, this was because they were following investors’ quest for returns at the expense of sufficiently questioning their risk assumptions. Today, as investors question managers more about risk, hedge funds are placing more focus on it.
Some managers who avoided last year’s calamity think a bit more common sense is largely what’s needed. Jim Melcher, founder of the $1.1 billion Balestra Capital in New York City, says the current crisis was brewing for years—as was the need for appropriate risk and investment adjustments.
“With the problems in residential mortgages becoming obvious in 2006, the subsequent deterioration in real estate values across the board, growing unemployment, increasingly constrained consumption, and enormously expanded credit risk, there were plenty of signs that mandated a move away from the bull-market group-think and extreme risk-taking of the previous four years,” says Melcher.
Balestra’s risk management is built into the way it makes and manages every individual investment—from idea generation, instrument selection, portfolio construction and monitoring, through realizing profits and losses. The fund uses some of the same risk controls that many hedge funds do, including portfolio stress testing and investment correlation analysis. Melcher also has stringent limits of 15% on individual theme exposure. And portfolio equity is not leveraged, except for the inherent leverage associated with some derivatives.
“But we also integrate analyses of market and systemic risk with common sense,” adds Melcher.
Active hedging and the avoidance of acrowded trades help Balestra maintain a flexible portfolio that can shift positions quickly. The fund limits individual positions to no more than 2.5% of the portfolio. It can liquidate 90% of the portfolio in an hour. And Melcher’s team has no problem moving into large cash positions, which hit 80% in 2008.
Another key component of risk management is the fund’s asymmetrical investment exposure, which increases profit opportunities while limiting downside. For example, when investing in credit default swaps, Balestra is only interested in buying protection. Therefore, it limits its risk to the premiums involved in buying such derivatives. “We would never sell a CDS and expose ourselves to potential losses equaling the size of the contract,” says Melcher. Common sense was required in that decision because VaR did not adequately measure the risk of writing premiums on CDS, which turned out to be lethal last year.
Balestra was one of the few funds that soared in 2008: Its $900 million core global macro fund, Balestra Capital Partners, gained nearly 46% percent in 2008, and it was winner of the Absolute Return Global Macro Fund of the Year award. It has registered more modest gains of 4.18% so far this year through August.
Another fund manager who did well last year, Roy Niederhoffer, believes the problem is that many hedge funds make money only when stocks are rising and economic conditions are tranquil—precisely when risk management may not be needed. “They often fail to provide a hedge during bear markets and periods of illiquidity,” Niederhoffer explains, “when many managers do not deliver value beyond that of equities.”
Niederhoffer, who runs $800 million in quant funds and whose Negative Correlation Fund won Absolute Return’s 2008 award for Managed Futures Fund of the Year, argues that his funds are less correlated than most hedge funds to the broader markets. Niederhoffer says his investment approach is based on short-term trades (lasting on average between one and two days) configured around historical investor behavior, which he believes inherently controls risk. No one particular trade will have a meaningful effect on total returns. He relies on generating small positive returns across many bets rather than hitting home runs.
Niederhoffer further controls risk by relying on diversified trading strategies that respond to a variety of signals. Each strategy is scaled and designed to exit positions at different levels, each influenced by volatility. “We intentionally seek to have a large number of individual strategies that have extremely diverse return profiles,” adds Niederhoffer, “so that when a few strategies are down, we want others to be up.” Because his trades are of such short duration, he uses a proprietary VaR risk measure that is based on more sensitive intraday trading data to determine both the movement of assets and correlations between markets, which advises each investment. When his particular version of VaR gets too high, he reduces his positions proportionately. As a last backstop, the firm’s risk-management committee will occasionally use its discretion to reduce risk, especially before big market events—such as moments preceding the release of unemployment figures—or during periods of high volatility.
This approach, and his investment selection process, helped Niederhoffer’s global macro diversified program soar more than 51% in 2008. This year through August, however, like a lot of CTAs his diversified program is off 3.6%. Niederhoffer explains this performance as a direct result of his strategy, which is designed to do best during volatile conditions, profiting from day to day or intraday swings. “The one-way 50%-plus rally in equities since March has not provided as many opportunities for us recently as it did the previous three years, during which our flagship diversified program rallied a total of 123%,” he says.
He doesn’t believe risk management is holding back performance. “When there is a problem,” Niederhoffer explains, “you can typically point to a single trade, a single market, a big bet that went awry. This year, both our losses and our profits have been quite small, and if anything we’re suffering from a lower percent of correct investments rather than a few big losing positions.”
As traders like Melcher and Niederhoffer know, when a bubble gets under way, many market players are less likely to pay attention to the type of stress testing, or plain common sense, needed to accompany VaR. Following each crisis since its inception, there have been calls for changes in risk management. Given the severity of last year’s financial crash, are they finally coming to fruition?
Some say yes. Passport’s Garry acknowledges that VaR doesn’t control against tail-risk loss, or the unlikely scenarios that fall outside of VaR’s 95% percent degree of certainty. So Passport complements VaR with a shorter-term focused metric called conditional VaR. Incorporating a Monte Carlo simulation of 5,000 possible market scenarios during the past day, the model aims to manage short-term volatility, especially when there is a breakdown in uncorrelated asset class performance. Fifty percent of the model’s weighting is focused on the past 10 days; the other half, on the prior 70 days.
“This helps provide us with an early warning sign when substantial value can be lost,” says Garry. He believes, with 95% certainty, that this may limit daily tail-risk loss to no more than 4%. And even if the market were to slide continuously as it did earlier this year, Garry believes management has better tools to reduce risk.
In September, Passport enhanced its risk management by installing SunGard’s Front Arena software, which provides live risk calculations for potential profit and loss in existing positions. This differs from other risk-management solutions that provide such information only after the last trading day.
Garry says various improvements—along with existing risk controls that limit positions to no more than 10% of the portfolio and with one-fifth of the portfolio invested in global macro trades (where losses are strictly limited)—should help to further reduce the blind spots inherent in risk-management calculations.
“The current crisis has produced fundamental changes in the way we define and manage risk,” he says. “We now explore risk like we explore any other thematic asset class, meaning we think we can generate alpha not only from security selection but also by better understanding risk.”
