HAR类已实现波动率模型及其在中国股票市场中的应用研究

发布时间：2018-01-01 02:17

本文关键词：HAR类已实现波动率模型及其在中国股票市场中的应用研究　出处：《长沙理工大学》2013年硕士论文　论文类型：学位论文

【摘要】：股票市场的波动率以及相关特征的研究，是国内外金融学者和业界金融从业者研究金融风险度量、金融资产定价、金融衍生品定价等金融实务问题的基础。定量分析这些实务问题，其前提是对资产波动率进行准确的度量和预测。因此，资产波动率的度量和预测一直是金融学者关注和研究的热点。近年来，随着通讯技术及计算机的使用和普及，在很大程度上降低了金融交易数据的记录和存储成本，从而使得金融高频交易数据日益成为研究金融资产波动率的重要手段。Anderson和Bollerslev（1998）运用日内高频数据计算的已实现波动率，相比传统的资产波动率度量模型，对股票市场波动率的度量精度更高。因此，本文从已实现波动率的角度，通过构建新的已实现波动率模型来研究中国股票市场的波动率。本文首先介绍已实现波动率的基本理论，以及分析中国股票市场已实现波动率的特征。然后，考虑隔夜收益方差，对现有的HAR-RV和HAR-CJ模型进行改进，构建新的HAR-ARV和HAR-CJ模型；并在新的HAR-CJ模型的基础上，加入投资者的行为偏差因素（动量效应），构建了HAR-CJ-M模型。接着，以中国股票市场沪深300指数的5分钟高频数据为研究样本，对HAR-ARV模型、HAR-CJ模型和HAR-CJ-M模型进行参数估计，再使用这三个模型的其他两种常用的表达形式、选择不同的样本长度和选择不同的参考价格（HAR-CJ-M模型的研究）对三个模型的参数估计结果进行稳健性检验。最后，运用损失函数法比较这三个模型对中国股票市场未来调整已实现波动率的预测能力。研究结果表明：中国股票市场已实现波动率存在明显的尖峰厚尾性、右偏性、跳跃性和长记忆性特征；HAR-RV模型、HAR-CJ模型和HAR-CJ-M模型的构建具有较强的理论支撑，且它们适用于中国股票市场的已实现波动率的研究。其中，，HAR-ARV模型的估计结果证实了中国股票投资者异质性的存在，与Müller等（1993）提出的“异质市场假说”理论相吻合。HAR-CJ模型和HAR-CJ-M模型的估计结果可以看出中国股票市场中历史的连续样本路径方差成分对未来波动率有一定的预测作用，而历史的离散跳跃方差成分的预测能力很弱，同时从HAR-CJ-M模型的估计结果中，发现不同期限（日、周和月）的动量效应因素（资本利得突出量）的系数大多都显著，说明中国股票市场中各类投资者的这一非理性行为对未来的波动率有一定的预测作用。另外，本文的研究还表明：HAR-RV模型、HAR-CJ模型和HAR-CJ-M模型对中国股票市场波动率有不错的预测能力，而加入动量效应因素的HAR-CJ-M模型预测中国股票市场波动率的能力明显强于其它两个模型，它更有利于金融风险度量、金融资产定价和金融衍生品定价等中国金融实务问题的研究。
[Abstract]:The research on volatility and related characteristics of stock market is a study of financial risk measurement and financial asset pricing by domestic and foreign financial scholars and industry financial practitioners. Financial derivatives pricing and other financial practical issues. Quantitative analysis of these practical issues, its premise is to accurately measure and predict the volatility of assets. The measurement and prediction of asset volatility has always been the focus of attention and research by financial scholars. In recent years, with the use and popularization of communication technology and computers. It greatly reduces the cost of recording and storing financial transaction data. As a result, financial high-frequency trading data is increasingly becoming an important means of studying volatility of financial assets. Anderson and Bollerslev 1998). Realized volatility calculated using intraday high frequency data. Compared with the traditional asset volatility measurement model, the measurement accuracy of the volatility of the stock market is higher. Therefore, this paper from the point of view of realized volatility. The volatility of Chinese stock market is studied by constructing a new realized volatility model. This paper first introduces the basic theory of realized volatility, and analyzes the characteristics of realized volatility in Chinese stock market. Then, the variance of overnight return is considered. The existing HAR-RV and HAR-CJ models are improved to construct new HAR-ARV and HAR-CJ models. On the basis of the new HAR-CJ model, the HAR-CJ-M model is constructed by adding the investor behavior deviation factor (momentum effect). Based on the 5-minute high frequency data of CSI 300 index in Chinese stock market, the parameters of HAR-ARV model, HAR-CJ model and HAR-CJ-M model are estimated. Then use the other two common expressions of these three models. Select different sample length and choose different reference price to study HAR-CJ-M model. The loss function method is used to compare the forecasting ability of the three models for the future adjustment of the realized volatility in the Chinese stock market. The results show that: the volatility of Chinese stock market has obvious characteristics of sharp tail, right deviation, jump and long memory; The construction of HAR-RV model HAR-CJ model and HAR-CJ-M model has strong theoretical support, and they are suitable for the research of realized volatility in Chinese stock market. The estimation results of HAR-ARV model confirm the existence of heterogeneity of Chinese stock investors. M 眉 ller et al. 1993). Heterogeneous Market hypothesis. The estimation results of HAR-CJ model and HAR-CJ-M model show that the component of path variance of continuous sample in Chinese stock market can predict the future volatility to some extent. The history of discrete jump variance components of the prediction ability is very weak, and from the HAR-CJ-M model estimates, we found that different periods (days). Most of the momentum factors (capital gain protruding) of the week and month are significant, indicating that the irrational behavior of various investors in the Chinese stock market has a certain predictive effect on the future volatility. The research of this paper also shows that the HAR-CJ model and HAR-CJ-M model have a good ability to predict the volatility of Chinese stock market. However, the HAR-CJ-M model with momentum effect is better than the other two models in predicting the volatility of Chinese stock market, which is more favorable to the measurement of financial risk. Financial asset pricing and financial derivatives pricing and other issues of financial practice in China.
【学位授予单位】：长沙理工大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.51;F224

【参考文献】