摘要
基于长记忆性的视角,将混频数据抽样(mixed data sampling)引入Realized GARCH模型的条件方差方程中,扩展得到Realized MIDAS GARCH模型,结合金融波动的非对称杠杆效应(asymmetric leverage effect),构建得到Realized MIDAS EGARCH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Realized MIDAS EGARCH模型中,构建了基于偏t分布的Realized MIDAS EGARCH模型,推导其参数估计方法,并基于滚动时间窗技术预测和比SPA检验更具优势的模型置信集(model confidence set,MCS)检验评估各种波动率模型对我国期货黄金市场波动的预测能力.实证结果表明:Realized MIDAS GARCH族模型比起Realized GARCH族模型有更优的拟合能力,捕捉长记忆性的能力更强;考虑杠杆效应和非对称厚尾分布能够提升模型的解释能力和预测能力;基于偏t分布的Realized MIDAS EGARCH模型是本文所探讨的8种高频波动率模型里面拥有最高样本内拟合优度、最强长记忆性的捕捉能力和最优样本外预测精度的高频波动率模型.
Based on the perspective of long memory,the mixed data sampling architecture was introduced into the conditional variance equation of the Realized GARCH model,and the Realized MIDAS GARCH model was extended to further consider the asymmetric leverage effect of volatility.Realized MIDAS EGARCH model was constructed,further considering that the traditional normal distribution assumption cannot characterize the asymmetry,non-normality,and fat tail of financial time series.Skewed t distribution was introduced into the Realized MIDAS EGARCH model.Realized MIDAS EGARCH model based on skewed t distribution,derived its parameter estimation method,and based on rolling time window technology prediction and model confidence set(MCS)test,which has advantages over SPA test,evaluates the ability of various volatility models to predict the volatility of China’s futures gold market.The empirical results show that the Realized MIDAS GARCH family model has better fitting ability and better ability to capture long memory than the Realized GARCH family model.Considering the leverage effect and asymmetric thick tail distribution can improve the model’s interpretation and prediction capabilities;The Realized MIDAS EGARCH model based on skewed t distribution is the high-frequency volatility model of the eight high-frequency volatility models discussed in this article that has the highest in-sample goodness of fit,long-memory capture capability,and out-of-sample prediction accuracy.
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;Zhejiang Institute of Economics and Trade,School of Finance and Accounting,Hangzhou 310018)
出处
《系统科学与数学》
CSCD
北大核心
2021年第7期1985-2005,共21页
Journal of Systems Science and Mathematical Sciences
基金
国家社会科学基金项目(19BTJ013)
浙江省哲学社会科学规划课题(18NDJC189YB)
教育部人文社科项目(16YJC910001)
浙江省一流学科A(浙江工商大学统计学)(1020JYN4118004G-58,1020JYN4119004G-94)资助课题
