摘要
文章基于日内动量效应的视角,增加日内收益率作为解释变量,并考虑日内交易的交互关系,推广得到结合混合频率的条件均值方程和条件方差方程,构建混频Realized GARCH模型(Mixed Frequency Realized GARCH模型),进一步考虑传统的正态分布假设不能够刻画金融时间序列的非对称性、非正态性、厚尾性等特征,将偏t分布引入混频Realized GARCH模型中,构建了基于偏t分布的混频Realized GARCH模型,推导其参数估计方法,并运用滚动时间窗预测技术和优于SPA检验的MCS检验判别扩展模型对我国黄金期货市场波动的预测结果。由样本内估计结果和MCS检验证实表明:考虑高频信息的Realized GARCH模型在样本内估计和样GARCH模型和Realized GARCH模型有着更高的拟合优度和预测精度,其中基于偏t分布的MF Realized GARCH模型在6种模型中是拥有最佳表现的波动率模型。
From the perspective of intraday momentum effect,this paper adds the intraday rate of return as an explanatory variable and considers intraday trading interaction to generate conditional mean equations and conditional variance equations with mixed frequency,and construct the mixed frequency Realized GARCH model.And then,the paper further considers that the traditional normal distribution hypothesis can not describe the characteristics of financial time series such as asymmetry,non-normality and fat tail,etc.,and introduces the skewed-t distribution into the mixed frequency Realized GARCH model to construct the mixed frequency Realized GARCH model based on skewed-t distribution,deducing the estimation methods of its parameters.Finally,the rolling time window prediction technique and the MCS test,which is superior to SPA test,are used to judge the prediction results of the extended model for China’s gold futures market fluctuations.The results confirmed by in-sample estimation and MCS test show that the Realized GARCH model considering high frequency information is superior to the GARCH model considering low frequency data in both in-sample estimation and out-of-sample prediction,and that the mixed frequency Realized GARCH model combining low frequency data with high frequency data enjoys higher goodness of fit and prediction precision than the GARCH model and Realized GARCH model,and the MF Realized GARCH model based on the skewed-t distribution is the best volatility model among the six models.
作者
蔡光辉
徐君
应雪海
Cai Guanghui;Xu Jun;Ying Xuehai(School of Statistics and Mathematics,Zhejiang Gongshang University,Hangzhou 310018,China)
出处
《统计与决策》
CSSCI
北大核心
2021年第2期130-135,共6页
Statistics & Decision
基金
教育部人文社会科学研究青年基金项目(16YJC910001)
浙江省哲学社会科学规划课题(18NDJC189YB)
浙江省一流学科A类(浙江工商大学统计学)资助项目(1020JYN4118004G-58
1020JYN4119004G-94)