结合日内跳跃识别方法和马尔可夫机制转换模型,对已实现波动率异质自回归模型(HAR-RV)进行拓展,以刻画连续波动、跳跃波动以及不同方向跳跃波动对未来波动影响的差异和波动的结构转换特征,并运用该模型对上证综指和深证成指高频数据进行实证分析。研究结果表明;在短期内,连续波动和跳跃波动对未来波动影响具有显著的差异;负向跳跃和正向跳跃往往同时发生且幅度相当,但负向跳跃波动对未来波动的影响更大;在不同波动状态下,历史波动对未来波动的影响存在较为明显的差异。MCS检验结果显示,区分跳跃波动方向和考虑波动的结构转换特征可以显著提升模型的样本内和样本外的预测能力。
For capturing the different effects of continuous volatility, jump volatility and jump volatility with different directions on future realized volatility and structure changes of realized volatility, this paper expands the HAR-RV model by combining intraday jump test method and Markov Regime Switching and then uses the expanded models to analyze Chinese stock market high-frequency data. The empirical results conclude: (1) in short term, continuous volatility and jump volatility have significant different effects on future realized volatility; (2) negative jumps and positive jumps often appears in the same trade day with same magnitudes, but negative jump volatility make stronger effects on future realized volatility; (3) under different volatility conditions, past realized volatility has distinct effects on future realized volatility. MCS test shows that differentiating the directions of jump volatility and considering the structure changes of realized volatility can improve the forecasting performance of models significantly.