전통적 방법과 이분산성과 꼬리위험을 고려한 방법에 의한 VaR 예측치의 성과비교: 한？중？일 증시를 중심으로

Associate Professor, Cheju National University, Department of Business Administration, Korea.攀攀In this paper, I compare the performance of five models in estimating VaR of market indices. Two of them are conventional methods(HS model, RiskMetrics model), the rest are new methods considering heteroscedastic and tail risk(EVT-GARCH model, GARCH(t) model, GARCH(gpd)model)They are applied on KOSPI(Korea Stock Exchange Index), SSE(China Stock Exchange Index), NIKKEI(Japan Stock Exchange Index) with the intension to compare the performance of the various estimation techniques on markets with different capitalization and trading pratices. The main results can be summarized as follows.First, the distribution of stock return series(KOSPI, SSE, NIKKEI) is not normal but fat tail and heteroscedastic. When we calculate VaR by conventional methods(HS Model, RiskMetrics Mode) we can ignore the heavy tails of the innovations or the stochastic nature of the volatility. Second, EVT-GARCH model, GARCH(t) model, GARCH(gpd) model are vindicated by the very satisfying overall performance in various backtesting experiments.Third, at very high confidence levels the EVT-GARCH model, GARCH(t) model, GARCH(gpd) model produce the more accurate forecasts.