혼합 빈도 GARCH 모형을 이용한 주식시장 변동성 예측

We study the volatility forecasts in Korean stock market based on the GARCHMIDAS(GARCH with the Mixed Data Sampling) model consisting of low frequency information variables and high-frequency financial variables. Through this work, we analyze the effect of long-term components of macroeconomic variables on predictability of stock market volatility. The variance ratio statistic based on in-sample estimation of GARCH-MIDAS models shows that macroeconomic components including both growth rates and volatilities of macroeconomic series statistically significant contribution to movements of stock volatility. In addition to in-sample estimation, we conduct out-of-sample forecasts to investigate adequacy of volatility models. Our results show that the GARCH-MIDAS models generally yield superior predictive power than the widely-used benchmark model of GARCH(1,1) model. It is also found that not only domestic macroeconomic components but foreign macroeconomic variables including output, price level, oil price and exchange rates significantly help predict the stock price volatility. In particular, macroeconomic volatilities are shown to better perform than macroeconomic growth rates in terms of stock return predictability. Based on our analysis of out-of-sample forecasts as well as a set of empirical results, proposed GARCH-MIDAS models could provide useful tools to predict volatility of the stock market in various practical areas.