SOE
Chow Institute
User Center
中
EN
About WISE
People
Committee of Academic Consultants
Faculty Directory
Staff Directory
Research
Publications
Working Papers
Facilities&Centers
Education
Overview
Undergraduate Programs
Graduate Programs
Study-Abroad MA Programs
Exchange Programs
Executive Education
News & Events
News
Announcements
Conferences
Seminars & Conferences
Job Openings
SOE
Chow Institute
User Center
中
EN
About WISE
Introduction to WISE
Contact Us
Map and Direction
People
Committee of Academic Consultants
Faculty Directory
Staff Directory
Research
Publications
Working Papers
Facilities&Centers
Education
Overview
Undergraduate Programs
Graduate Programs
Study-Abroad MA Programs
Exchange Programs
Executive Education
News & Events
News
Announcements
Conferences
Seminars & Conferences
Job Openings
Research
Home
->
Research
->
Publications
->
Content
Research
Publications
Working Papers
Facilities&Centers
Finance & Economics Experimental Lab
MOE Key Lab in Econometrics
Fujian Provincial Key Lab in Statistics
Center for Econometrics Research
Center for Financial Research
Center for Research in Labor Economics
Center for Macroeconomics Research
Center for Statistics Research
Center for Information Technology
SAS Center for Excellence in Econometrics
High-Speed Computing Cluster
基于广义谱和MCS检验的VaR模型预测绩效评估
Id:2320
Date:20160223
Status:
ClickTimes:
作者
张玉鹏, 洪永淼
正文
条件VaR 模型的正确设定检验等价于检验均值化的“撞击序列”是否服从鞅差分序 列,然而通常的反馈检验方法只检验了该序列的部分性质。采用对该鞅差分性质进行直接检 验的广义谱检验方法,全面考察中国股票市场( 香港恒生指数、上证综合指数和台湾加权指 数) 上各参数、非参数和半参数共22 个VaR模型在采用滚动窗口预测机制时的样本外预测绩 效。鉴于条件VaR模型正确设定检验无法反映超过某VaR水平的尾部风险信息,为避免极端 损失的发生以及增加结果的稳健性,同时采用模型置信集检验方法。研究结果表明,采用通 常的反馈检验方法常会得出错误的结论; 在1%和5%置信水平,与历史模拟法、极值理论模型、 CAViaR模型和CARE模型相比,误差项为t分布的GARCH模型族在金融危机期间具有较好的样 本外预测绩效; 涨跌停板制度对于选取预测绩效最优的VaR模型具有重要影响。
JEL-Codes:
关键词:
VaR模型; 预测绩效; 涨跌停板制度; 广义谱检验; MCS 检验
TOP