In part one of his two-part feature, Anthony Malakian examines the state of risk management at the portfolio/trade-book level among buy-side firms. Part two will look at the middle- and back-office functions of risk across the buy side.
In 2008, the capital markets witnessed the collapse of some of Wall Street’s most prestigious investment banks. The resulting typhoon laid waste to many a hedge fund, asset manager and pension fund. Risk systems proved to be impotent. Much was learned and improved upon, but as the markets become increasingly automated and fragmented, risk concerns weigh heavy on the minds of fund managers.
This examination by Waters of risk management on the buy side looks at the new technologies, strategies and models being deployed inside buy-side firms. Part one explores front-office risk management concerns at the portfolio/trade-book level, in addition to data and analytics, while part two will deal with the larger issues affecting operational and enterprise risk, the systems used, and greater regulatory themes.
The VaR Debate
Two hedge funds. Both managing less than $300 million. They have different strategies, and, as a result, have vastly different views of their Value at Risk (VaR). According to one of the managers—who parted ways with a large, New York-based hedge fund at the height of the financial crisis, but has since launched his commodities-focused fund—“a VaR-based approach is the way to go,” when it comes to managing his firm’s portfolio exposure. He acknowledges that VaR isn’t infallible, but he says that most of the problems that arise with these models are due to either incorrect or dishonest input into the model.
“When it comes to risk management, looking at what people are doing today as opposed to before, my intuition tells me that not much has changed, although the names that people use for things have changed,” he says. “Since the financial crisis, there’s been a whole industry created to come up with new names for VaR. A lot of this is window dressing for investors.”
An opposing argument comes from the head trader of a hedge fund specializing in debt. He was formerly a trader at Goldman Sachs and has more than a decade’s worth of experience in the industry. He says VaR-based models are useless on the buy side, as they are only appropriate for large investment banks. He points to the European debt crisis in 2011 as an example of VaR’s shortcomings, as it couldn’t properly quantify the market fluctuations, in his estimation.
“My personal sense is that VaR is just not used on the buy side—nobody looks at these numbers. They are not very useful for risk management purposes,” he says. “They were developed for large banking institutions so that they could communicate with their regulator. They’re not useful for the buy side. What people care about on the buy side are primarily sensitivities of the portfolio to short-term moves, and, a distant second, shock analysis to various scenarios across instruments.”
These different views demonstrate the lack of consensus on the buy side when it comes to VaR—a staple of risk management—and therefore risk management systems and processes, in general. Everyone can agree, however, that if the data being used is not clean and timely, then VaR─or any other risk measurement, for that matter─is essentially useless.
Since 2008, if they had not been doing so already, many buy-side firms have begun coupling stress-testing with their VaR calculations. Finding systems to generate these reports simultaneously is a challenge.
“VaR hasn’t changed that much, but definitely stress-testing has picked up dramatically,” says Stuart Breslow, managing director of Chicago-based risk management technology vendor, Citadel Technology. “VaR is more of an approach you do for everyday risk, but it has to be coupled with stresses.”
Citadel Technology, founded in 2009, is the licensing arm of $13 billion hedge fund Citadel. It was set up to sell the hedge fund’s internally built technology to other buy-side and sell-side firms. Citadel Technology provides front- and middle-office technologies that Citadel’s portfolio managers use. As such, Breslow has an intimate knowledge of the fund’s risk management systems. He notes that the focus on data quality and daily updates has taken on greater importance since 2008.
“There is a lot more attention to detail in terms of some of the things we do with our risk applications, and not just looking at a VaR number, but looking at what goes into that VaR number—the underlying risk factors, managing the quality of the risk factors, looking at the quality of the underlying data that feeds the risk process, making sure there is daily certification by the chief risk officer, and ensuring that was reviewed every day. In the past, it was more of a weekly process, and some places had it as a monthly process,” Breslow says.
Meru Capital is based in New York and manages $1.3 billion. It was launched in 2009 by a number of former founding partners of Citigroup’s now-defunct hedge fund, Old Lane Partners. At its inception, Meru turned to software provider Imagine Software to help manage its risk management, profit-and-loss (P&L) monitoring and position-keeping functions. It chose Imagine because the New York-based vendor could cover those three important segments, according to Jonathan Barton, principal at Meru.
“What was important for us [when we launched] was a multi-asset class delivery,” he says. “For us, it’s not just about being able to do it on US equities—it’s about being able to apply stresses on credit default swaps, interest-rate swaps, and swaptions, and the ability to stress, not just the equity market, but changes in volatility, changes in credit spreads, and changes in foreign-exchange rates.”
Another now-popular method for measuring risk is that of tail risk, a statistical model that does not include a normal distribution curve, such as VaR does, although it’s a measurement that has caught the eye of investors, and as such, fund managers. Some equate tail risk to stress-testing, although David Merrill, CEO of buy-side risk management technology vendor FinAnalytica, says that stress-testing is more of a substitute for those who don’t have the capabilities to measure tail risk.
“By doing a stress-test, they essentially force a tail event and feel that they then know what the tail risk is. Essentially, what you’re saying is: ‘If the S&P 500 index drops by 30 percent, what happens to my portfolio?’ In doing a stress-test, you’re assigning what happens based on some historical period to have all the results calculated," although these do not accurately model fatter tails, as historical VaR and Monte Carlo simulations account for thin tails.
Sandy Warrick, chief risk officer at equity market-neutral hedge fund QuantZ Capital Management, says his firm can provide tail risk and “black-swan protection” through its different sub-strategies, and, as a result, he says QuantZ does “asymmetrically well” during extreme market events.
“A key philosophical differentiator versus most quants and bottom-up stock pickers is our macro regime overlay as well as the use of ‘meta-models,’” Warrick explains. “Our world macro regime model classifies the world into a finite number of terminal states, which in turn maps to different alpha models. One can use what we call ‘meta-models’ to predict the likely efficacy of our sub-strategies, which in turn drives the dynamic leverage across them. Dynamic leverage combined with some meta-model foresight can allow one to truncate the left tail of the returns distribution and presumably accentuate the right skew at the same time. Extreme diversification across stocks, sectors, models and even horizons keeps the idiosyncratic risk contained, and usually results in portfolios where the realized VaR is substantially lower than that predicted by most vendor risk models.”
Tail risk and scenario analysis hasn’t changed much since 2008—it’s more the interpretation of those numbers that has changed, according to the head of risk management at a $10 billion New York-based hedge fund.
“It used to be just this report that you would look at and say, ‘OK, that’s the answer,’” he says. “Now, it’s become more of an evolving scenario where risk managers need to reinterpret this report so we can describe it to a front-end user or portfolio manager.”
He says that rather than leaving portfolio managers up to their own devices, as a risk manager he wants to be a bit more prescriptive. “In the past you’d say, ‘You have “X” amount of exposure to this concept that you’re not familiar with.’ You would then tell them to become familiar with it, hand them the report, and wish them luck. Now we’re saying: ‘Here is exactly how you should act on this information, and here’s a way to use it to shift your investment process.’”
Part two of this feature will explore the systems that deliver data for those reports, although it is important to note that developing a system that consolidates data from various silos and platforms—including order management systems, risk management platforms, portfolio management systems, and/or accounting systems—is a significant undertaking, according to Thomas Lee, partner at NorthPoint Solutions.
“In the past, risk management was often considered a separate function from portfolio management, trading, and operations—it was really where the risk manager was looking at stuff and was more of a downstream person who would take data and produce reports,” he says. “Nowadays, hedge funds want risk management tightly integrated with everything else. They want one consistent set of data for both portfolio and risk management.”
Once that data is brought together in a report, it must be interpreted. That is a function that has changed of late. In order to better interpret risk exposures, the head of risk management at the $10 billion fund says it is developing a reporting system that more dynamically disseminates those risk reports.
Instead of using a simple PDF that is read at the portfolio manager’s leisure, the firm is looking to create internal web pages with risk analytics that are updated regularly to reflect the portfolio’s real-time status. So when portfolio managers ask questions, rather than the risk manager having to hand them a report, he can direct them to a link that will be updated with near real-time, relevant analytics.
The general idea is that a model can no longer simply be taken at face value to say what a firm wants it to say—it’s more of an interpretation of an event and being able to react to it. The Flash Crash in May 2010 was a prime illustration of the fact that models cannot protect anyone from freak occurrences. It did reinforce that analytics must be clean and firms must believe in those numbers, which allows traders to trust their instincts and experience, says Mehmet Yanilmaz, partner at Illinois-based proprietary trading firm Myra Trading, which develops and implements its own trading and risk analytics software.
“We strive to improve our risk analytics for systemic risks that occur due to fat fingers or some other rare conditions that occur in electronic markets,” he says. “Our test and software acceptance processes for trading and risk algorithms—and their implementation—are exhaustive, and have not changed since the inception of the firm.”
While the ability to use true, real-time data is not a reality in most asset classes, the drive to gain access to clean, quick data has become the focus du jour at investment houses. Risk management has moved from an isolated function that generated reports that would get pushed out once a day, to more of a real-time interaction that can examine shifts in the book, says the risk manager.
“Before, people were more concerned with how their book looks right now, whether they will blow up tomorrow, and did I make money yesterday,” he says. “Now we’re more interested in where exactly did I make money, how I did it, was it consistent, is it replicate-able, and is the portfolio manager good or bad at that?
• Value at Risk (VaR) is still important, but many firms are now coupling it with functions like stress-tests and calculating tail risk.
• Portfolio managers are incorporating stress-tests and developing new methods for stress-tests, more today than pre-2008. • While some may look at tail-risk simulations as simply another form of stress-testing, potential investors are now demanding hedge funds prove they have the capability to run these models.
• Rather than handing over a risk report on a monthly, weekly or even daily basis, risk managers are looking to develop reporting tools that provide real-time risk analysis that portfolio managers have immediate access to.
Click Here to view as PDF