基于广义已实现测度的Realized GARCH模型改进及应用  

Forecasting VaR:Realized GARCH Incorporating Generalized Realized Measures

作　　者：蒋伟 顾研 Jiang Wei;Gu Yan(Fanhai International School of Fianace,Fudan University;School of Economics,Fudan University)

机构地区：[1]复旦大学泛海国际金融学院 [2]复旦大学经济学院

出　　处：《数量经济技术经济研究》2019年第7期156-172,F0003,共18页Journal of Quantitative & Technological Economics

基　　金：教育部人文社会科学重点研究基地项目“全球金融市场联动与中国经济增长”(16JJD790011)的资助

摘　　要：研究目标:从日内已实现信息视角获得更准确的市场波动率估计和VaR预测。研究方法:将广义已实现测度引入Hansen等(2012)提出的Realized GARCH(RGARCH)模型中,用于中国股票市场的波动率估计与VaR预测,并应用无条件覆盖检验、独立性检验、条件覆盖检验、损失函数和MCS检验方法,综合比较了RGARCH模型采用不同已实现测度的VaR预测效果。研究发现:无论是样本内的模型拟合结果还是样本外的模型预测效果,微观结构噪音稳健的RK和跳跃稳健RBV的引入并没有对RGARCH模型有实质性的改进效果,而当引入更为广义的已实现测度,尤其是日内已实现VaR时,模型样本内和样本外的表现都有了显著的提高。研究创新:引入广义已实现测度改进RGARCH模型的估计与预测效果。研究价值:提高了RGARCH模型在波动率估计和VaR预测中的表现,并对日内交易的已实现信息的应用提供了新的思路和方法。Research Objectives:In order to get more accurate volatility estimation and VaR forecasting based on the intraday information.Research Methods:We apply realized GARCH models by introducing generalized realized measures of intraday returns into the measurement equation,to forecast the daily value at risk(VaR)of Chinese stock mark index returns.Besides using the conventional realized measures,realized volatility,realized kernel,and realized bipower variation as our benchmarks,we also use generalized realized measures,realized absolute deviation,and two realized loss measures,realized value-at-risk and realized expected shortfall,and realized lower partial moment.The intraday high frequency data and daily data of SSCE and CSI 300 index are during 2005-2016 in this paper.Moreover,unconditional coverage test,independence test,conditional coverage test,loss function and MCS test are applied to compare the results of the estimation of volatility and the forecasting effect of VaR in this paper.Research Findings:The empirical results show that realized GARCH models using the generalized realized risk measures provide the best parameter estimation for the in-sample and substantial improvement in VaR forecasting for the out-of-sample.In particular,the realized VaR performs best for all of the alternative realized measures.Our empirical results reveal that future VaR may be more attributable to present losses(realized loss measures).The results are robust to different sample estimation windows.Research Innovation:We introduce the generalized realized measures to improve the estimation and forecasting effect of Realized GARCH model.Research Value:It improves the performance of Realized GARCH model in volatility estimation and VaR prediction,and provides new ideas and methods for the application of realized information in intraday trading.

关 键 词：Realized GARCH 波动率 VAR 已实现测度 

分 类 号：F222[经济管理—国民经济]