Stressed by VaR!

Value at risk (VaR) is but one of a number of measures available to the risk manager, which includes, amongst other mathematical calculations, standard deviation, concentration metrics, tracking error and stress testing. However, good old fashioned common sense, experi­ence and prudence need to be considered as paramount when formulating decisions relating to risk management.

VaR is generally used for guidance and because it is an industry standard accepted by regulators, amongst others, it is a useful statistical measure to track the trend in risk taking within the organisation. VaR’s strengths and weaknesses have been de­bated over the years and whilst there is no question of discarding VaR, other tools such as stress testing are featuring prominently within the risk manager’s tool kit.

It should be considered in the cases of both Bear Stearns and Lehman Brothers, VaR estimates quadrupled between 2004 and 2007, clearly indicating something amiss. During 2007/8 VaR trends pointed to the steady increase in risk in banks’ financial portfolios.

One of the hardest hit European banks, UBS, reported exceptions to its VaR esti­mates at a 99% confidence level, equating to approximately one trading day every week throughout 2008. The bank stated the exceptions “highlighted the limitations of VaR and illustrated the need for multiple views of risk exposure such as macro and more targeted stress scenarios’’.

It is interesting that for UBS, the previous period with a high number of reported VaR exceptions was in 1998. These related to the Asian debt crisis, Russian rouble crisis and the failure of the Connecticut-based hedge fund Long Term Capital Management. Simi­larities between these crises and the 07/08 financial crisis are worth noting!

The UK Financial Services Authority (FSA) stated the low number of exceptions re­ported before the start of the current fi­nancial crisis and the high numbers during 2007/2008 were symptomatic of poor mod­elling. They considered the following were causes for the exceptions:

• increased volatility and correlations
• beyond those predicted in the VaR model
• non-negligible risk factors not covered
• in the VaR model; and
• concentrated positions.

Stress tests were considered as being within the range of tools available to risk managers prior to and during the crisis, but criticism centred around the lack of creative thinking employed in their compilation. Stress tests were indeed conducted but they simply served to prove what was already known rather than tackle the unthinkable.

A number of organisations consider stress testing in terms of either running single or multiple shocks related to interest rate movements, for example imposed on a portfolio to determine the effect of such hy­pothetical movements. Another common form of stress testing involves imposing a historical period of extreme movements, such as the period around September 11th 2001 on the portfolio to determine the effect should the same event with the as­sociated market movements reoccur. But we all know the possibility of such an event re-occurring causing exactly the same set of market movements is highly unlikely, bring­ing into question the worth of such tests. Nonetheless, they are interesting and can provide an insight into the specific risks to which the portfolio is exposed.

A major strength of stress testing, newly considered, lies in the definition, plotting and tracking the effect of an evolving set of economic conditions rather like writing an unfurling story. For example, consider the events triggered by the effective bank­ruptcy of Thailand and the decision to float the Thai baht cutting its peg to the US dollar in July 1997. The start of the Asian crisis had wide reaching effects and a world­wide economic meltdown was feared. The crisis caused a heavy decline in commodity prices and Russia, heavily dependent on ex­ports, including oil and natural gas, severely suffered. At the height of the Russian rouble crisis, inflation peaked at 84%. The main reason for the amazing speed at which Russia recovered was due to the steep in­crease in world oil prices during 1999 and 2000…and so the story unfolds.

We have a wealth of information related to previous financial crises and the sheer ease of availability of economic information provides the fuel for the organisation’s im­agination to run wild. For example, consider the Norwegian kroner and its correlation with oil prices for a starter. Indeed, it would be easy to become carried away with the sheer range of possibilities and resources involved in such an operation would need to be tightly constrained!

Such a stress test would involve an organ­isation considering economic conditions and demands the collaboration of the C-levels , economists and risk managers but it is not difficult to visualise the benefit to the organisation if such a project is expertly im­plemented. Indeed with the move towards enterprise wide risk management, one could easily predict that the organisation’s ability to operate this type of stress test could be directly linked to its profitability.

Stress tests are generally becoming more widely used but, unlike VaR, it is difficult for regulators to determine an effective stand­ard, with organisations advised to conduct stress tests applicable to relevant risk expo­sures and lines of business.

In May 2009, the Basel Committee pub­lished a set of guidelines designed to in­tegrate the results of stress testing within bank’s governance and risk management culture so that they are considered within the strategic decision making process. However, if stress testing is to be regarded as a compliment to VaR there would appear to be a moral hazard. To produce creative stress tests under the shadow of having a penalty imposed in the form of higher regu­latory capital, for example, would undoubt­edly deter the most creative stress tester.