| Abstract [eng] |
The thesis examines the variance-covariance approach to the estimation of portfolio Value-at-Risk using multivariate GARCH models. Two multivariate GARCH models, DCC and O-GARCH, are estimated using Baltic stock market data. Based on the results of these models, Value-at-Risk of randomly generated portfolios is calculated using the variance-covariance approach. This approach was improved by taking quantiles of generalized hyperbolic distributions instead of standard normal ones. The analysis suggests that the use of generalized hyperbolic distribution considerably improves the accuracy of Value-at-Risk estimates. Therefore, it is proposed to use the family of generalized hyperbolic distributions in practice. |
