보험 섹터펀드의 VaR추정과 모형의 적정성 검증

This paper undertakes two types of fund in VaR model selection for KOSPI200 index and insurance sector index at the 95% and 99% confidence levels. Nine models are compared using daily returns from January 2000 to December 2009. The nine models are simple moving average model(SMA:50, SMA:125, and SMA:250), RiskMetrics model(RM:0.94, RM:0.96, and RM:0.99), and GARCH model using three different distribution assumptions. In order to choose appropriate models in types of fund among the various methods, three backtesting criteria(unconditional, independence, and conditional coverage) are used to test the statistical accuracy of models. Empirical results are as follows; At the case of 95% level, the performance of SMA models using historical data as 50, 125, 250 days, RM model using as 0.99, and AR(1)-GARCH(1.1) models with normal, , and GED distribution are better than that of other models in the KOSPI200 portfolio. The performance of SMA models using historical data as 50 days, RM model using as 0.96, and AR(1)-GARCH(1.1) models with -distribution are better than that of other models in the insurance stocks portfolio. At the case of 99% level, the performance of RM models using as 0.94, 0.95, and 0.96 in the KOSPI200 portfolio and SMA models using historical data as 50, RM models using as 0.96, AR(1)-GARCH(1.1) model with normal, , GED distribution are better than that of other models in the insurance stocks portfolio.