Impacts of asymmetry on forecasting realized volatility in Japanese stock markets

Abstract

Abstract This study investigates the most important asymmetric property for forecasting realized volatility in the Japanese futures and spot stock markets. We employ heterogeneous autoregressive (HAR) models allowing for three types of asymmetry: positive and negative realized semivariance (RSV), asymmetric jumps, and leverage effect. We find a clear difference among HAR models allowing for asymmetry. The HAR model with leverage effect performs best among asymmetric models. Additionally, the forecast performance of the HAR model with RSV is superior to that with asymmetric jumps. The asymmetric jump components do not produce better forecast performance compared with the standard HAR models. The empirical results indicate that asymmetric information, particularly leverage effect and RSV, yields better modeling and more accurate forecast performance for the realized volatility of Japanese stock markets.