摘要
根据投资者情绪对股市波动具有重要影响这一观点,引入投资者情绪的传统GARCH类波动率模型出现因不同频率数据建模而产生的效率损失问题。文章基于混频数据结构,分别从不同行业、不同情绪状态和不同经济阶段3个角度切入,引入自适应权重形式的广义自回归条件异方差混频数据抽样模型(GARCH-MIDAS-adapt),对中国股市日度波动率进行估计与预测比较。实证结果表明,自适应权重形式融合的混频数据结构可以更好地解释投资者情绪对股市产生的长期波动作用,不同行业表现出有显著的解释力和预测力。此外,在不同行业下,情绪低落时对股市的冲击更大。
According to the view that investor sentiment has an important impact on stock market volatility, the traditional GARCH-type volatility model with investor sentiment has the problem of efficiency loss caused by different frequency data modeling. Based on the mixed frequency data structure, a generalized autoregressive conditional heteroscedasticity mixed frequency data sampling model with an adaptive weight function(GARCH-MIDAS-adapt) is used to estimate and forecast the comparison of daily volatility of the Chinese stock market from three perspectives of different industries, different sentiment states and different economic stages. The results show that mixed data structure with adaptive weight can better explain the long-term volatility of investor sentiment on the stock market, and different industries show significant explanatory and predictive power. In addition, in different industries, the impact of depression on the stock market is larger.
作者
吴文诗
宋泽芳
张兴发
李元
WU Wen-shi;SONG Ze-fang;ZHANG Xing-fa;LI Yuan(School of Economics and Statistics,Guangzhou University,Guangzhou 510006,China;Center of Statistical Science of Lingnan,Guangzhou University,Guangzhou 510006,China)
出处
《广州大学学报（自然科学版）》
CAS
2022年第4期68-79,共12页
Journal of Guangzhou University:Natural Science Edition
基金
国家自然科学基金重点资助项目(11731015)
国家自然科学基金青年资助项目(11701116)
广州市基础与应用基础研究资助项目(202201010276)。
