Selecting Value-at-Risk methods according to their hidden characteristics

The foremost concern of a modern risk manager is to estimate Value-at-Risk (VaR), that is the maximum loss likely to occur over the next two weeks for a given confidence level. Although widely spread as a standard measure for quantifying market risk, most of the techniques involved in measuring VaR are based on unrealistic assumptions. The purpose of this paper is to highlight the deficiencies of these techniques and to illustrate the exact steps for the estimation of portfolio-VaR according to all basic methods, namely, the variance–covariance (VC), the historical simulation method (HS) and the Monte Carlo simulation method (MC). Furthermore, the paper examines the accuracy of both the HS and the MC methods according to the Basel II regulatory framework.