An Analysis of the Maximum Losses Expected Calculated by VaR (Value at Risk) in Moments of Systematic Crisis

One of the main ways of measuring risks is the calculation of the VaR (Value at Risk), where the risk is measured in value. One of the assumptions of VaR is that the distribution of the financial assets returns follows a normal distribution, but what has evidenced in recent years is a distribution with more extreme values, mainly in function of unexpected financial crises by normality. Therefore, financial disasters are not often enshrined in the maximum expected losses by the financial market. The present work had as its overall objective: To verify if the VaR manages to capture the maximum expected loss in moments of systemic crisis, as well as to analyze which of the three methods used for the calculation has the slightest error in their estimates. To achieve the general objective, the VaR was calculated daily, during the period from 1993 to 2010, with mobile windows, using three methodologies: normal linear, historical simulation and Monte Carlo simulation. It was found that in moments of systemic crisis the VaR is unable to predict accurately the expected maximum loss, not properly safeguarding the investor as the volatility of the asset. Among the three methodologies, the one that possesses the less error (excess loss) is the historical simulation, being that the other two, normal linear and Monte Carlo simulation, are technically equally in the errors, this means, there is no difference in the forecasts of maximum losses