Scrutinizing Parametric Value-at-Risk Measure under Real-World Assumptions

The thesis compares an industry-standard parametric Value-at-Risk estimate with alternative approaches. The intention of the thesis is to find out, whether, or to what extent can the inappropriate assumption of normally distributed returns influence the Value-at-Risk estimate. We used the exceedance rate as a back-testing procedure in order to test the accuracy of parametric Value-at-Risk estimate. We look whether the exceedance rate of the estimates approaches the given confidence level or not. We contrasted the parametric measure to its historical and Monte Carlo counterparts. The latter assumes Student's t-distribution as an example of a fat-tailed distribution, because the estimate of tails is crucial for the accuracy of Value-at-Risk estimate