Value at risk is defined as:
The maximum loss for a given exposure over a given time horizon with x% confidence
VaR helps a user to define the maximum loss on an exposure for a given confidence level and has helped investors and corporations in managing their risk. VaR is on the face of it an excellent risk management tool, which can be used to measure a variety of risk types.
However, it should be noted that:
VaR does not define the worst case scenario
It may give the maximum loss for an exposure with 99% confidence using a 3000-iteration Monte Carlo simulation. The question remains however, what happens to the exposure for that 1% point of confidence? The frank truth is a VaR model is incapable of answering that question. Thus, a degree of both care and common sense is needed. The more sophisticated corporate Treasuries frequently seek to refine their VaR model to go beyond the natural confidence level limit to try and define the maximum loss with 100% confidence. A practical way of trying to achieve this is to impose operational limits (such as in terms of number of contracts, nominal amount, sensitivities or stop loss orders) in addition to VaR limits. That relates to the aspect of care. The common sense aspect relates to never trusting your risk to a computer model alone. If you cannot quantify it itself without use of the model, you have a problem.