The sub-prime explosion has shown managers that they cannot rely on VaR – and that VaR poses a risk in itself. By Nick Fitzpatrick
Not everyone has been a loser in the sub-prime blow up. “The managers in which we’ve invested have not lost money. They correctly forecasted the markets and were net short of sub-prime situations,” says Larry Jones, the chief investment officer of Nedgroup Investments, which is partly owned by Old Mutual. He adds: “Our managers were forward-looking and understood the market risks. They were looking at sub-prime as an accident waiting to happen.”
‘Forward-looking’ is the key phrase here. Forward-looking risk management is meant to provide a handrail in market turmoil, as Jones – who is a former chief risk officer at Attica Alternative Investments – indicates. “It is worth spending on risk systems,” Jones adds. “But risk management does not prevent institutions from losing money. You can still be surprised by big moves in the markets.”
Anybody relying solely on value at risk (VaR), the standard risk management tool that is widely employed at present, would almost certainly have been surprised by the recent jolt from the sub-prime sector. VaR accords a risk weighting to past events but has been adapted by users so that it gives relatively less weight to ‘50-year events’, or extreme market shocks. In short, VaR never saw the sub-prime explosion coming.
The VaR model is used as a risk modelling tool across the board by many banks, brokers and fund managers. In certain areas such as the Ucits funds regualtions, it is recommended by regulators as a default tool.
Robin Cresswell, managing principal at $50bn fund manager Payden & Rygel, says: “[Under VaR] the ‘50-year event’ is seen as exceptional, so less weight is given to that kind of historic data. Instead the investor gives more weight to what . happened more recently – like last week, last month and last year – with decreasing emphasis, or weight attributed to events the further back they go.”
Nedgroup’s Jones says: “VaR would have led some people to believe that their risks were lower than they really were. It works in normal markets, but in abnormal markets VaR is not enough. Additional stress testing is needed.”
Giovanni Beliossi, portfolio manager at FGS Capital, says: “VaR only works in normal times. We do not use it – we do not think it is relevant.”
The sub-prime blow-up has caused many investment managers to cast a worried glance at the robustness of their risk modeling. This happens periodically when any market shock occurs. But what, if anything, will be different this time?
“I think sub-prime will change everything,” says Phillipe Carrell, global head of business development at Reuters Trade & Risk Management. “It will make people look at correlations between asset classes – just like dotcom saw a return to fundamental analysis, or in the way that Enron increased the focus on corporate governance.”
Similarly Beliossi, at FGS Capital, says: “From the point of view of an equities manager, what’s happened recently has shed more light on the possible links between things that in a normal market environment do not seem to be linked. For example, the link between forms of debt and equities.”
These correlations have, at least in part, been created by hedge funds investing in fixed-income-based derivatives, such as sub-prime collateralised debt obligations (CDOs), who have had to sell some of their best quality equity holdings to realise cash when their sub-prime instruments began to shake and margins were called in. Normally, there should be no obvious correlation between equities and CDOs in this way and the problem with risk modelling is that there was scant data owing to the short life so far of the CDO market, says Beliossi. “The extent of cross-selling is something that I personally have not seen since 1998,” he notes.
Multi-factor stress testing
But there has also been another form of correlation. According to David Brierwood, chief operating officer at MSCI Barra, a provider of risk management tools and indices, recent weeks have also shown a correlation of investments styles. “We’ve seen some of the traditional sources of return – such as value – underperform and this is incredibly strange. All of the different styles – such as earnings yield, value, growth or earnings momentum – are typically uncorrelated. But now, if you use one of these styles, they are correlated, meaning everything is moving against you.”
The future of risk management could lie in multi-factor stress testing. “Multi-factor stress scenarios are the future, where you might use the 1987 stock market crash, for example, as a state-of-the-world involving all combined risk factors, not just equity prices and interest rates,” says Carrell, at Reuters. “Managers need to look at the details of all their positions by various factors, including geography, sector, counterparty risk and concentration, and then run scenarios, such as a burst in the real estate bubble.”
But how many scenarios can there be? Brierwood, at MSCI Barra, says: “The problem with creating scenarios is that there is a good probability that there are millons of scenarios. But we know how markets behave during times when certain stresses appear. There are patterns that emerge and this means we have been able to identify that the risk in equities is down to style factors, such as earnings momentum or earnings yield, for example.”
Beliossi, at FGS Capital, says he uses MSCI Barra products coupled with in-house tools. Payden & Rygel uses VaR, but just as one diagnostic tool. Important also is experience, says Payden’s Cresswell. “In our Ucits offshore funds we are required by the regulators to measure VaR on a daily basis. These VaR reports provide confirmation that within certain parameters we are not running excessive risk. We then broaden VaR as a modeling tool. VaR is simply one diagnostic tool. At the end of the day, we have to have the wisdom and experience to know when to use it and when not to rely on it too much.”
As well as its use under Ucits, VaR has also been backed by regulators as a tool for banks to use in order to stay within their Basle II minimum capital requirements. VaR’s use is in telling a risk manager how much money might be lost in a given scenario. But in prescribing the rules, the regulators have also unwittingly shown the pathway around them. The banks have been able to take loans off of their balance sheets and sell them on as derivatives, thus reducing their capital requirements.
According to Cresswell, the widespread adoption of VaR as a risk management tool poses a risk itself. “Problems start to occur when the regulator selects VaR as a standard way of measuring risk because this sends a signal to fund managers that one of several VaR models should be implemented. By default, a lot of risk managers are grouped together in one bucket, meaning their model will be as strong or as weak as each other’s. I think that introduces the threat of systemic risk when the markets start to behave differently.”
© fe September 2007