Weaknesses of VaR Approaches
It is worth concluding on the subject of VaR by reiterating that it is not a panacea and its results must be treated with caution. Ninety-nine percent confidence levels sound very impressive but for a one-day holding period mean that in one day in 100 we should expect greater losses than specified:
Most models assume normal distributions and no serial correlation and this is not always the case.
Correlation coefficients are calculated on the basis of historic data and there is no guarantee that such relationships will persist.
Standard holding periods may be insufficient to liquidate or hedge all positions.
These three concerns are all most likely to exhibit themselves at times of extremes when the level of potential losses is likely to be highest.
The use of VaR techniques does not eliminate the risk that losses will be greater than that expressed at the stated confidence level.
VaR models do not usually capture all material higher order price or realization risk factors.
Most VaR models do not capture intra-day positions and the actual risk profile may be much higher than that suggested by VAR calculated on the basis of open positions at the end of each day.
VaR is a reporting and analytical tool and most banks with active trading books overlay standard VaR approaches with sensitivity analyses and stress testing. None of these remove the need for banks to put other controls in place. Only offices or branches with sufficient derivatives expertise and adequate risk control systems should be allowed to deal in derivatives, for example. Management has to put in place control systems to set and enforce limits on exposures across a range of different dimensions including counterparty, location, product, currency, country, risk type, risk factor, position limits and VAR limits.
HA ·