Value-at-Risk

Both correlation calculations as distribution test highlight the importance of performing stress simulations as a supplement to VaR calculations.

The VaR-target has a weakness, which originates from the fact that both volatility and correlations are assumed constant over time. This means that you should carefully consider how these numbers are estimated, and secondly that VaR only can be used to estimate risk in a relatively short horizon. A VaR numbers calculated for a time horizon of one or more months will therefore contain no reliable meaning.

It is therefore recommended that, implied volatilities are to be used, and that the most recent data period are used when estimating based on historical data, and that on a regularly bases the data are assessed to whether there should be changed in the estimates.

Furthermore it is recommended to use a short horizon for the calculation of VaR, optimally one day, since the parameters in practice will change. It may well be justified to calculate the VaR of a week or 2 weeks if it is not possible to collect data and make the necessary calculations daily. The longer the period - the greater the uncertainty. To check these calculations, you obviously need to do stress tests.

A completely different thing is that VaR only manages market risk, whereas for example, credit risk is not included and must therefore be handled separately. Likewise, there are problems with instruments that provide a non-linear or asymmetric return eg. Options. VaR can, as designed today, only calculate risk on linear instruments. Therefore there must be supplemented with key figures on the given options.

Although that VaR has some weaknesses the measurement provides, however the best ever bid on a portfolio's risk, since both the individual position risk and the diversification effect is included in the calculation.