When an investment bank that is supposed to know better loses billions of dollars betting on subprime mortgages, you have to wonder what happened to the concept of risk management. The people in control at UBS certainly did.
In April, after taking $37 billion in subprime-related write-downs and shutting down its Dillon Read Capital Management hedge fund division, the big Swiss bank released a soul-searching postmortem. UBS concluded that one of the problems was that its risk management systems faltered and a key reason was an overreliance on a model called Value at Risk, which contributed to low-ball assumptions of exposure.
UBS isn’t the only one looking warily at VaR. The venerable risk-modeling tool, widely used in the investment world, has come under increasing criticism lately, derided as a fancy placebo sold to suckers by risk systems vendors. In an op-ed commentary in the Wall Street Journal in April, Lazard chairman and CEO Bruce Wasserstein pronounced VaR a “dud.” David Einhorn, president of Greenlight Capital, a New York–based hedge fund firm with $6 billion under management, followed up the same month with a speech in which he declared that VaR had provided absolutely no help in protecting assets during the recent market mayhem.
But risk analysts and managers say VaR is being unduly singled out as a cause of the recent troubles, when the problem is its misuse. Although just about every risk system on the market contains a VaR calculator, it is just one of a number of tools typically packed into a risk management software suite. The trick is knowing when and how to use VaR and understanding its limitations.
“You can’t rely on VaR as your only metric,” says Leslie Rahl, president and founder of New York–based Capital Market Risk Advisors. “We recommend people use three to five different metrics. It’s like a doctor ordering an X ray, an MRI and a CAT scan — they all tell you slightly different things.”
A veteran of 35 years in the financial industry and a financial engineering pioneer, Rahl ran the derivatives business at Citibank in the 1980s before establishing her consulting firm in 1991. She preaches the importance of rigorous risk analysis and testing to cope with the impact of the types of investments she peddled in her earlier role.
Rahl recommends applying stress tests to see how a portfolio would react to sharp drops, market shifts, unusual situations or changes in underlying assumptions. Stress-testing models, which are included in risk systems, can reveal weaknesses that a simple VaR test misses. But Rahl says too many financial firms continue to rely mostly on VaR. Back in April 2000, Rahl’s firm conducted a survey of risk practices and found that 45 percent of financial firms, including hedge funds, were not using stress tests at all. Although she hasn’t updated the survey, she says she has noticed only a slight improvement since then.
As a concept, Value at Risk has been around at least since the 1950s. It got its first major boost in the 1980s, when broker-dealers began using it to help determine market exposure. VaR revolves around a formula to calculate the maximum potential loss in a portfolio at a high probability (usually set at 95 to 99 percent) over a fixed time period.
VaR’s big move to the buy side came in 1994 when J.P. Morgan & Co. built a system around a VaR calculator to measure portfolio risk for an end-of-day report to senior management. That was the genesis of the RiskMetrics system that J.P. Morgan began offering to institutional clients in 1996, then spun out in a stand-alone company in 1998. RiskMetrics Group, which went public in January of this year, remains a leading risk systems provider in what has become an increasingly crowded field of vendors selling to hedge funds, banks and other asset managers.
Today risk systems boast considerably more computing power and flexibility than in the early days. They include several kinds of VaR calculators, a Monte Carlo simulator for historical testing, stress-testing models and various other models for running projections. Most systems allow users to create customized tests and produce a wide range of reports, charts, graphs and spreadsheets. Vendors increasingly offer hosted systems, in which the servers and technology are owned and maintained by the provider, with users logging in to run their tests and create reports. The hosted approach eliminates the need for costly software installations and hardware purchases and upgrades to keep the system running.
By all accounts, these new-generation risk management systems can digest and analyze massive quantities of historical and market data faster and more accurately than ever before and are capable of churning out reports in a matter of minutes. Risk systems providers compete not only on speed and functionality but also on supplying historical market data that users can load into calculators to test against their own portfolios.
Yet with all that computing power and sophistication at hand, some risk managers still fall back on the simplest VaR calculators as their primary tools. Old habits are hard to break.
“We have been trying to get people to not do that for ten years,” says Gregg Berman, co-head of the risk business group at RiskMetrics. “Risk management is a lot more than VaR. It has to do with scenario analysis, stress testing, a whole suite of things.”
Berman joined RiskMetrics in 1998 after serving as the co-manager of several multiasset hedge funds run by British hedge fund manager ED&F Man Group (now Man Group). Like many in the risk management field, he has a background in science rather than business. Berman has a Ph.D. in physics from Princeton University.
Satyajit Das, a Sydney, Australia–based risk consultant to banks and fund managers and author of Traders, Guns and Money, believes the limits of VaR make it less and less useful for hedge funds that deal in increasingly complex, multiasset strategies. “If I am a hedge fund manager, I would rely entirely on stress tests,” Das says.
Berman says the idea that hedge fund risk managers should simply turn off the VaR calculators in their risk management software is the wrong response to recent market disruptions. Depending on the portfolio, some type of VaR analysis might provide useful information. The task of the risk manager is first to understand the particular nature of the portfolio to be analyzed, then to determine what sorts of questions to ask and finally to decide on the correct tools to do the job.
“The first question is, How do I put all this information together?” Berman says. “The second is, Am I measuring the right things?”
VaR is often misunderstood as a concept. It can be a simple calculation for evaluating listed securities. But it can also be molded into a more powerful and complex analytical tool by adding different data and modeling techniques, including stress tests.
Oleg Movchan, chief risk officer at Alexandra Investment Management, a New York–based hedge fund firm with
$1 billion under management, says knowing which tools to use and how to employ them is a big part of his job.
“This whole discussion about the usefulness of VaR makes very little sense,” Movchan says. “Is a screwdriver a good tool or not? That is a meaningless question. Is a screwdriver good for screwing? Yes. Is it good for hammering nails? No. The question that should be asked is whether VaR is the right tool for a specific problem.
“For some complex products, where assumptions about stability of volatility and correlations and perfect [ample] liquidity are likely to be drawn, VaR is completely inappropriate,” he adds. “But for a long-short equity analysis, it is a useful tool.”
Movchan, who has been running risk analysis at Alexandra for 12 years, says experience helps him decide how to use the tools in the risk software suite he gets from New York–based Imagine Software. Each day he produces a report that runs seven to nine pages and gives a broad snapshot of the firm’s risk exposure.
Similar risk analyses were being run at UBS and Dillon Read. But the conclusions were often wrong. In its dense, 50-page April report on its losses, complete with a three-page glossary filled with definitions of such instruments of modern finance as credit default swaps and collateralized debt obligations, UBS offered one of the more detailed looks at how a risk management system failed in the recent market meltdown.
One of the problems at UBS and Dillon Read was that too much of the number-crunching was being done with VaR calculators, which do not take into account liquidity issues. They are based on assumptions of “perfect” markets, or those in which there are always buyers and sellers. Instruments like mortgage-backed securities, which have inherent liquidity risk, are ill suited to basic VaR analysis.
That problem was compounded by reliance on overly rosy credit ratings that agencies like Standard & Poor’s placed on mortgage-backed securities. Even a stress test to determine how those securities would perform under adverse conditions missed the mark, because the data included an assumption that they would hold their value. That gave the green light to traders to leverage up on mortgage-backed securities. When defaults skyrocketed, the MBS market collapsed — and so did Dillon Read.
Steven Harrison, president and co-founder of Imagine Software, says some firms did a fine job of forecasting potential problems. The secret was not necessarily in their chosen software, but rather in doggedly raising questions about the underlying assets in mortgage-backed securities, Harrison says. A risk manager who bothered to look inside the baskets of securities being traded would have discovered stacks of default-prone subprime loans. With that information in hand, stress tests, which could be constructed on Imagine’s software, would have shown the true risk exposure.
“If you are making a billion-dollar bet,” Harrison says, “you are supposed to be analyzing what is really in this basket that somebody sold you.”
What leads a risk manager to ask the right questions, run the right tests and avoid the pitfalls of using the wrong tools for the job? Rahl says experience counts for a lot, especially with the complex nature of derivatives and the equally complex risk management systems used for analysis. But the rapid growth of hedge funds — there are now an estimated 10,000 worldwide — has made it tough to find talent. As a result, some firms have relatively inexperienced risk managers.
“There are an awful lot of very smart people in risk,” Rahl says. “But there are also a lot of kids who are still wet behind the ears.”
Andrew Aziz, executive vice president for risk solutions at Toronto-based risk vendor Algorithmics, believes that the recent focus on VaR is probably a good thing because it has caused firms to review their risk practices. Those that relied heavily on VaR to the exclusion of other risk models will now likely change their approaches.
“I think going forward there will be few organizations that base everything on a simple VaR calculation,” Aziz says. “VaR won’t be discarded. For normal markets it is still a very useful measure.”
As the subprime situation has shown, finding the right data to run tests can be a challenge in itself. To help him gather intelligence, Alexandra’s Movchan says he resorts to the decidedly low-tech technique of talking to people, both in his firm and elsewhere, in a constant search for any tidbit that might shed light on a trend or issue of importance.
“In risk management only about a third is quantitative,” Rahl says. “A third is still a big part of the puzzle, so it is quite valuable.”
The remaining two thirds of the puzzle is where good risk managers earn their money. Ultimately, an accurate forecast depends on knowledge, experience and chutzpah.
“It has nothing to do with the computer,” Rahl says. “It has to do with wisdom and experience.”
And perhaps a bit of luck.
