In this study, the performances of GARCH models with different distribution assumptions in modeling Value-at-Risk are evaluated by the backtesting procedure for three equity indexes. Recent researches indicate that Extreme Value Theory (EVT) is good candidate to model rare extreme events and unpredictable losses. Due to return series have non-normal characteristics, standardized residuals of GARCH are modeled by EVT and leptokurtic distributions. Empirical findings show that EVT based GARCH model is outperformed according to the backtesting results modeling daily VaR for all equity indexes.
ISSN: 2241-0996 (Online)