基于高频数据的金融资产共同跳跃建模研究

发布时间：2018-02-21 18:18

本文关键词： 高频数据共同跳跃 MHAR模型风险度量　出处：《福州大学》2014年硕士论文　论文类型：学位论文

【摘要】：在日益复杂的金融市场环境下,投资者在追求报酬最大化的同时,更加关注风险的评估和管理,因此风险管理成为金融学者们相当重要的研究课题之一。随着非参数估计广泛应用于估计波动率,人们更加关注日内丰富信息对风险管理的重要性,基于高频数据的跳跃行为研究在现实中可以很好地解释日内异常事件对价格和风险的影响。跳跃行为对准确预测方差有重要作用,同样地,共同跳跃行为对准确预测协方差也非常重要,进而对合理资产组合配置、优化资产管理都要重要的作用。本文在已有文献研究的基础上,基于高频数据,从以下几个方面对不同金融资产之间的共同跳跃行为进行有关建模方面的研究：1)系统阐述近几年国内外关于金融资产共同跳跃的研究现状,并从理论上分析了资产价格服从的跳跃-扩散过程,为后面的研究提供理论基础。2)系统阐述四种常用的日内跳跃检验统计量,并从实证角度对这些日内跳跃检验统计量进行比较,通过分别利用模拟数据和实证数据对统计量比较后发现,ABFN统计量在总体上在检测日内的跳跃行为方面具有一定的优越性。3)应用ABFN统计量剥离出跳跃方差序列和跳跃协方差序列后,分别建立预测方差和协方差的多资产MHAR-RCV-CJ模型,并且与MHAR-RCV模型比较预测精度,证明共同跳跃对于预测协方差的重要性,考虑共同跳跃能够更好地预测协方差。4)在用MHAR-RCV-CJ模型和MHAR-RCV模型预测出方差和协方差的基础上,对构造出的等权重资产组合分别进行VaR和ES风险度量分析并比较后发现：考虑跳跃和共同跳跃的MHAR-RCV-CJ模型构造出的资产组合风险度量效果比不考虑跳跃和共同跳跃的MHAR-RCV模型的效果好；相对于VaR度量方法,ES度量方法更加精确,是用于资产组合风险度量的更好方法。本论文是国家自然科学基金资助项目《基于已实现测量非参数方法的金融资产跳跃行为研究》(NO.71171056)的阶段性研究成果。
[Abstract]:In the increasingly complex financial market environment, investors pay more attention to risk assessment and management while pursuing the maximization of returns. Therefore, risk management has become one of the most important research topics for financial scholars. With the wide application of non-parametric estimation in estimating volatility, people pay more attention to the importance of rich information in the day for risk management. The study of jump behavior based on high frequency data can explain the effect of intraday abnormal events on price and risk in reality. Jump behavior plays an important role in accurate prediction of variance. The common jump behavior is also very important for accurate prediction of covariance, and it is also important for the rational allocation of asset portfolio and the optimization of asset management. In this paper, based on the existing literature, based on high-frequency data, From the following several aspects of the joint jump behavior between different financial assets related modeling research: 1) A systematic description of the domestic and foreign financial assets in recent years on the joint jump research status. And theoretically analyzes the leapfrogging and diffusion process of asset price clothing, and provides the theoretical basis for the later research. (2) systematically expounds four commonly used intraday jump test statistics. And from the perspective of empirical comparison of these intraday jump test statistics, By comparing the statistics with simulated data and empirical data, it is found that ABFN statistics have certain advantages in the detection of hopping behavior on the whole. 3) the ABFN statistics are used to peel off the jump variance sequence and jump. After the jump covariance sequence, The multi-asset MHAR-RCV-CJ models of predictive variance and covariance are established, and compared with the prediction accuracy of MHAR-RCV model, the importance of common jump in predicting covariance is proved. Considering that the common jump can better predict covariance. 4) based on the prediction of variance and covariance using MHAR-RCV-CJ model and MHAR-RCV model, The VaR and es risk measurements of the constructed equal-weight portfolio are analyzed and compared respectively. It is found that the risk measurement effect ratio of the portfolio constructed by the MHAR-RCV-CJ model with jump and joint jump is less than that of skip and joint jump. The effect of MHAR-RCV model is good. It is more accurate than the VaR measurement method. It is a better method for portfolio risk measurement. This paper is a periodic research result of the project supported by the National Natural Science Foundation of China, "Research on the jump behavior of financial assets based on the realized measurement of non-parametric methods" (no. 71171056).
【学位授予单位】：福州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F832

【相似文献】