Forecasting the realized volatility based on jump, jump intensity and regime switching in stock market

Abstract

This paper constructs a series of new volatility prediction models by introducing the jump and jump intensity to the heterogeneous autoregressive (HAR) model and together considering a Markov regime switching model with a fixed transition probability matrix. To evaluate the predictability of the above prediction models on Chinese stock market volatility, popular statistical tests such as the model confidence set (MCS), out-of-sample R2 test and direction-of-change test are used. The empirical results indicate that: 1) compared with the benchmark, the regressions incorporate the regime switching can achieve superior forecasting performance; 2) the forecasting model with jump component, jump intensity and regime switching can outperform peers; 3) the MS-HAR-TJI model still can successfully predict the stock realized volatility during the COVID-19 and high volatility periods.