The realized volatility (RV) calculated using intraday high-frequency returns, is used as an estimator of asset price volatility. The heterogeneous autoregressive (HAR) model, which specifies RV as a function of the previous daily, weekly and monthly RVs, is recognized as efficient. The realized GARCH and stochastic volatility (RSV) models, which augment the GARCH and stochastic volatility models with RV, have also attracted research attention. After conducting a survey of previous studies on models using RV, this paper compares the volatility predictive abilities of some major models using the daily returns and RV of Nikkei 225 stock index. Evidently, the HAR and realized exponential GARCH models perform better than other models in certain periods, including the first wave of the COVID-19 pandemic. The RSV model provides the best results for other periods.
