Value at risk, a statistical calculation of risk exposure designed to help investors protect their assets from volatile markets, may hold some risks of its own for institutional investors.
Some in the institutional investing community have suggested incorporating VAR limits into portfolio management can increase market volatility and reduce liquidity. They believe if investors are getting the same information at the same time from their VAR models -- say, that they are too heavily weighted in Asian currency -- then they may all make similar market moves at the same time - in this case, reducing currency exposure in Asia.
In their book, "Risk Management: Approaches for Fixed Income Markets," Bennett W. Golub and Leo M. Tilman of BlackRock Inc., New York, explain how putting hard VAR limits on a portfolio can lead to even larger losses. In a market crisis, they write, investors are forced to sell as VAR rises past the set limits. Soon, the selloff begins to feed on itself and losses become much larger than they would have been had investors been able to hold their positions through the crisis and not been forced to sell to stay within their VAR limits.
The theory is not unlike one put forth in analysis of the 1987 market crash in which so-called "portfolio insurance" resulted in greater selloffs and thus greater volatility. Messrs. Golub and Tilman point out in their book that portfolio insurance, which used put options and index futures to mitigate losses, actually led to more selling as the markets fell, further exacerbating the selloff.
The authors point to the 1998 market crisis a hypothetical example of how VAR could make a selloff worse. As the historical volatility of interest rates increased and the correlation between the three-month spot rate and the spot rate on longer maturities became negative, Messrs. Golub and Tilman suggest, the VAR of leveraged portfolios could have risen quickly. When it broke through the limits, portfolio managers would have had to sell to reduce their VAR. They would have found themselves trying to sell in the middle of a crisis, which likely would have made the selloff worse and led to greater losses.
But risk experts, while acknowledging such a VAR scenario is technically possible, said it's unlikely because for most investors VAR is only one risk management component in an overall investment strategy. Nobody, they said, would make portfolio decisions based solely on VAR.
Data issues
"I agree there's a possibility of that, although the problem is less the VAR model than the data used in the VAR model," said Andrew Aziz, executive director of professional services for Toronto-based Algorithmics Inc. "VAR is one tool in the risk-management process. It's an important tool, but not the only tool."
Also, as more investors adopt a historical simulation VAR model over the more simplistic variance/covariance matrix model, the likelihood of VAR setting off any "liquidate now" alarms is reduced, Mr. Aziz said.
Timothy Morris, vice president for risk management at GE Asset Management, Stamford, Conn., agrees few would base an entire portfolio strategy on VAR.
"My view is that VAR is just one piece of the risk-management process," he said. "People are looking at a variety of different items. They focus on exposure and concentrations in different countries and regions, and use other quantitative measures that also give us a view on the world besides what a VAR model may be saying."
Although he sides with Messrs. Aziz and Morris, Philippe Jorian can lay out the argument of those who insist VAR can indirectly increase volatility and contribute to liquidity losses like those that doomed Long-Term Capital Management. Mr. Jorian is vice dean and professor of finance at the University of California at Irvine Graduate School of Management and author of the book "Value at Risk: The New Benchmark for Measuring Financial Risk."
The argument goes this way: In August 1998, Russia's default on its short-term debt obligations caused a balloon in Russian VAR portfolio components. The overall VAR number, fueled by the Russian component, bumped up against, and in some cases broke through, hard VAR limits set by investors. Traders either had to secure more capital to cover the increased risk or liquidate riskier, more volatile holdings.
Broader markets
Risk managers at investment banks and hedge funds called for liquidation not just in Russian positions, but in broader swap and government bond markets. They had to, they reasoned, in order to raise money to cover their Russian losses.
Author Nicholas Dunbar uses this example in his book, "Inventing Money: The Story of Long-Term Capital Management and the Legends Behind It," to promote his theory that VAR indirectly contributed to the 1998 liquidity crisis.
One risk manager at a New York hedge fund, who asked not to be identified, said it's clear VAR can cause market volatility. Take investment banks, for example. A major event like the Russian default can cause them to re-examine and likely reduce certain portfolio positions. The risk manager said the banks would comb their VAR models to find the positions generating the most volatility and liquidate them. Many different banks selling the same security at the same time would drive prices down, increasing volatility.
"It's not generally known, but guys on the sell side recognize that this can occur," he said. "For once, we're actually thinking ahead of the curve."
On his VAR website, www.GloriaMundi.org, Barry Schachter, head of enterprise risk management at Caxton Corp., a large macro hedge fund, recounts Mr. Dunbar's logic. Mr. Schachter said he thinks VAR was a factor 1998, but not the only factor.
"I think that the Dunbar story is plausible, but I think that there are other factors that sort of pre-empted VAR from being the driving force," Mr. Schachter said. "It was a combination of the impact of VAR and a whole bunch of risk management factors not related to VAR."
Thomas K. Philips, chief investment officer for Paradigm Asset Management Co., New York, said he doesn't know many people who use VAR as their only risk measurement. Consequently, they would be unlikely to make portfolio decisions solely based on VAR.
"It can't predict risk," Mr. Jorian said. "VAR cannot predict all possible outcomes because it has to look back at the past, and the future is not necessarily a replay of the past."
Just one part
Todd E. Petzel, executive vice president and chief investment officer for Commonfund Asset Management Co., Wilton, Conn., said hypothetically, VAR could set off a warning signal that global bond exposure may be too large, "but it's only one part of the decision making process. We don't make it on VAR alone."
Mr. Jorian argued that VAR is less important for pension funds than for commercial banks because pension funds typically look at longer-term returns while commercial banks sometimes hold investments for only a day. That means pension funds are less likely to quickly sell securities based on a spike in VAR.
Also, saying that VAR models lead to cutting positions, and thus volatility, doesn't take into account the expected rate of return, Mr. Jorian said. "In periods where volatility is high, it's usually because markets are falling," he said. "This is precisely when expected returns are higher. You really have to look at the tradeoff between the two. If you have a hard VAR cap, you could lose out."
