Title: LIHAR model for forecasting realized volatilities featuring long-memory and asymmetry
Authors: JiWon Shin - Ewha Womans University (Korea, South) [presenting]
DongWan Shin - Ewha Womans University (Korea, South)
Abstract: It is well known the fact that financial time series have asymmetric variances and a recent paper revealed that an integrated HAR model beats the HAR model. So, we add a leverage term to an integrated HAR model, called as LIHAR model. Comparisons of forecasting ability of several models show that the IHAR model with leverage (LIHAR) is superior to the HAR and IHAR model. The model is applied for 20 real data sets of RV for financial indices DJIA, S$\&$P 500, Russell 2000, KOSPI Composite, etc. The volatilities of the financial indices like stock price and foreign exchange are characterized by very persistent long memories and asymmetry. These features are so well-suited for the integrated heteroscedastic autoregressive model with leverage that the LIHAR model generally produces better out-of-sample forecasts than other models like HAR, IHAR and LHAR for the real data sets. This result supports that recent IHAR model is more suitable for forecasting RV for financial indices. It is good to consider long-memory, asymmetry, nonstationarity for forecasting realized volatilities.