基于高频数据的中国股市VaR风险研究.pdf

Research on VaR of Chinese Stock Market based on High Frequency Data A Thesis Submitted to Chongqing University in Partial Fulfillment of the Requirement for the Master’s Degree of Science By Wu Xili Supervised by Prof. Fu Qiang Specialty:AppliedMathematics College of Mathematics and Statisticsof Chongqing University, Chongqing, China April 2013 重庆大学硕士学位论文中文摘要 I 摘要金融市场的风险度量一直是学术界和风险监管当局关注的重点。传统的风险度量大多数都是基于低频日间数据建立的GARCH类模型或SV类模型。虽然这些模型本身能较好的度量时间序列的波动状况,但股市日内交易频繁,低频数据模型会损失大量的日内重要信息。现有研究表明,传统的GARCH类模型并不能直接用于估计高频波动率。建立有效的高频数据风险度量模型,为金融机构和监管当局的风险监控提供一种有效的理论方法参考和政策建议具有重大意义。本文结合前人对已实现类高频波动率的研究,对已实现波动率 RV、已实现双幂次波动率RBV和赋权已实现双幂次波动率WRBV进行比较,针对WRBV具有的长记忆性,建立了ARFIMA-WRBV-VaR模型对中国股市风险进行度量,并与采用低频日间收益率序列建立的GARCH类模型相比较。实证结果表明:ARFIMA-WRBV-VaR模型比EGARCH-VaR模型估计效果更好。而且,已实现类高频波动率出现了跳跃点、日内U型周期性日历效应和长记忆性特征,这些特征受市场微观结构中的信息不对称和投资者心理等因素影响。进而为风险监控提出了完善信息披露机制和增强投资者素质的政策建议。关键词:VaR,高频波动率,ARFIMA模型,GARCH模型,微观结构重庆大学硕士学位论文英文摘要 II ABSTRACT The academic and financial regulatory authorities alwaysfocus on the risk measurement of financial market. Most of traditional risk measurementsare using GARCH model or SV model based on low-frequency intraday data. Although these models can measure the volatility of timeseries well, because of the stock market transaction frequently ineveryday, low-frequency data model will loss a large number of daily important information. The current study shows that, the traditional GARCH models cannot be directly used to estimate high-frequency is significant to establish an efficiency risk measurement model based on high-frequency data. So that we can provide an effective theory method and policy mendations for risk control of financial institutions and regulatory authorities. Based on previous research on realized high-frequency volatility ,this paper compares realized volatility, realized bipower volatility and weighted realized volatility. Lighting of long memory of WRBV, establish A