基于贝叶斯厚尾DCC-MSV模型的股市波动溢出效应研究

发布时间：2018-01-10 00:18

本文关键词：基于贝叶斯厚尾DCC-MSV模型的股市波动溢出效应研究　出处：《湖南大学》2014年硕士论文　论文类型：学位论文

【摘要】：20世纪80年代以来，经济全球化与金融一体化在全球范围内不断推进，极大增强了世界各国金融市场之间的相互依存性，单个金融市场的波动不但受到其自身前期波动影响，还受到其他金融市场前期波动的影响，这就存在着波动溢出效应。而现有关于金融市场波动溢出效应的研究大多采用GARCH类模型，与GARCH类模型相比，随机波动模型被认为更加适合金融领域的实际研究。因此，本文构建了具有双向格兰杰检验的贝叶斯厚尾DCC-MSV模型，用于研究不同金融市场之间的波动溢出效应与动态相关关系。首先，对厚尾DCC-MSV模型进行贝叶斯分析，设计了模型参数估计的MCMC抽样算法，解决了多变量随机波动模型参数难以估计的难题。然后，介绍了随机波动模型参数收敛性诊断的方法、随机波动模型比较准则以及波动溢出效应显著性的检验方法。最后，本文选取沪深300指数、恒生指数、标准普尔500指数、英国富时100指数、日经225指数作为研究对象，分析2005年中国大陆股市股权分置改革后中国大陆股市、中国香港股市、美国股市、英国股市、日本股市的波动溢出效应与动态相关关系，同时，运用DIC准则对CCC-MSV模型、厚尾CCC-MSV模型、DCC-MSV模型、厚尾DCC-MSV模型进行比较分析。研究结果表明，在波动溢出效应方面，仅存在中国香港股市对中国大陆股市的单向波动溢出效应，美国股市、英国股市与日本股市对中国大陆股市不存在波动溢出效应，且中国大陆股市对其他四个股市均不存在波动溢出效应。而中国香港股市与英国股市、中国香港股市与日本股市、美国股市与英国股市之间均存在双向波动溢出效应，但美国股市对香港股市、美国股市对日本股市、英国股市对日本股市仅存在单向波动溢出效应。在动态相关性方面，各股市之间的相关关系具有时变特征，，且这种相关关系存在长记忆性，同时这种相关性在金融危机期间呈现出上升趋势；此外，中国大陆股市与中国香港股市联系最为紧密。在波动模型模拟效果方面，引入t分布的厚尾CCC-MSV模型与厚尾DCC-MSV模型要优于未引入t分布的CCC-MSV模型与DCC-MSV模型。
[Abstract]:Since 1980s, economic globalization and financial integration have been continuously promoted in the global scope, which has greatly enhanced the interdependence of financial markets around the world. The volatility of a single financial market is affected not only by its own pre-fluctuation, but also by other financial market's pre-volatility. There is volatility spillover effect. However, most of the existing researches on volatility spillover effect in financial markets are based on GARCH model, compared with GARCH model. Stochastic volatility model is considered to be more suitable for the practical research in the financial field. Therefore, a Bayesian thick-tailed DCC-MSV model with bidirectional Granger test is constructed in this paper. It is used to study the relationship between volatility spillover effect and dynamic correlation between different financial markets. Firstly, the Bayesian analysis of the thick tail DCC-MSV model is carried out, and the MCMC sampling algorithm of the model parameter estimation is designed, which solves the difficult problem of estimating the parameters of the multivariable stochastic wave model. This paper introduces the method of parameter convergence diagnosis of stochastic volatility model, the comparison criterion of stochastic volatility model and the test method of volatility spillover effect significance. Finally, this paper selects Shanghai and Shenzhen 300 index and Hang Seng index. The Standard & Poor's 500 Index, the FTSE 100 Index and the Nikkei 225 Index are used as research objects to analyze the Chinese mainland stock market and the Chinese Hong Kong stock market after the split share structure reform in mainland China in 2005. The volatility spillover effect of American stock market, British stock market and Japanese stock market is related to the dynamic correlation. At the same time, the CCC-MSV model and the thick-tailed CCC-MSV model are analyzed by using DIC criterion. DCC-MSV model and thick tail DCC-MSV model were compared and analyzed. The results show that, in terms of volatility spillover effect, there is only one-way volatility spillover effect of Hong Kong stock market on mainland China stock market, the United States stock market. There is no volatility spillover effect on the mainland stock market in the UK and Japan stock market, and there is no volatility spillover effect on the other four stock markets in the Chinese mainland stock market, while the Hong Kong stock market and the UK stock market in China have no volatility spillover effects. There is a two-way volatility spillover effect between Chinese Hong Kong stock market and Japanese stock market, US stock market and British stock market, but the US stock market has a two-way volatility spillover effect on the Hong Kong stock market and the US stock market against the Japanese stock market. There is only one-way volatility spillover effect in the British stock market to the Japanese stock market. In terms of dynamic correlation, the correlation between the stock markets has time-varying characteristics, and this correlation relationship has a long memory. At the same time, the correlation showed an upward trend during the financial crisis; In addition, the mainland stock market and Hong Kong stock market are most closely linked. The thick tail CCC-MSV model and the thick tail DCC-MSV model with t distribution are better than the CCC-MSV model and DCC-MSV model without t distribution.
【学位授予单位】：湖南大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：F224;F832.51

【参考文献】