The purpose of this study is to compare the performance of forecasting Value at Risk(VaR) through GARCH model and Stochastic Volatility(SV) model by using daily returns of KOSPI200 Index from Jan. 1993 to Sep. 2001. For estimating the parameters of the models, ML method and EMM were employed for GARCH and SV, respectively. Under Basel’s back-testing criteria, VaR were forecasted for 4 sample periods and compared with realized returns in out-of-sample. The results of comparing are as follows; First, though SV brought in exceptions more than GARCH, it was verified as a valid model with the exception of the 1st sample period. through the test of failure rate under 5% significance level. Second, SV outperformed GARCH in accuracy and required capital. Additionally, it can not be judged to adapt square-root rule for calculating long-term VaR or not. Judging from the results of this study, it was verified that SV model can be adapted as a method to forecast VaR of KOSPI200 Index.
