변동성 예측에서 실현 왜도와 첨도가 갖는 정보효과: 이질적 자기회귀모형의 개선을 중심으로

In financial markets, there are many studies for improving the predictability of future volatility by combining the heterogeneous autoregressive (HAR) model and realized volatility (RV) using intraday high-frequency data. We examine the predictability of future volatility based on expanded HAR-RV models including realized skewness and realized kurtosis proposed newly in this study using the intraday high-frequency data of KOSPI. In the study, we utilize a modified method that improves the problems from the previous studies, which may occur in the process of separating the continuos elements and the discontinuous jump elements from the realized volatility. The main results are as follows. First, the HAR-RV model shows the high explanatory power with respect to changes in future volatility, and application of the continuous element and discontinuous jump element of realization volatility into the model separately is useful for enhancing the explanatory power. Second, the HAR-RV model including realized skewness and kurtosis proposed in this study shows obviously improvement of both the predictability and explanatory power for the future volatility. Realized skewness and kurtosis can have additional information reflecting the volatility characteristics resulting from the action and reaction of various heterogeneous investors with different time horizons. And our results suggest that realized skewness and kurtosis are useful variables with additive information for predicting the future volatility in the HAR-RV model. The usefulness of these variables is well established regardless of controlling the volatility leverage effect and intra-day return seasonality effect in the model.