基于滞后系数突变的结构突变AR（p）模型的上证指数变点研究

发布时间：2018-02-24 01:23

本文关键词： 伯努利分布贝叶斯推断 Gibbs抽样分层先验均值方差双重突变　出处：《华侨大学》2013年硕士论文　论文类型：学位论文

【摘要】：中国股市作为一个新兴市场，由于法律制度、市场交易和运行机制的不完善以及微观交易主体投资心理的不成熟等原因，导致我国股市常出现较大的波动，因此其发生结构突变的几率更大，然而国内关于我国股市结构突变问题，尚缺乏细致、系统的研究。因此研究我国股市结构突变问题，进一步探讨导致结构突变发生的主要原因，对我国的金融风险管理和政策的制定有着重要的理论和实践意义。本文首先对国内外关于结构突变检测方法和股价指数结构突变问题的研究现状进行总结。在此基础上，结合贝叶斯计量经济学，首次尝试将二项分布引入滞后项系数结构突变的AR(p)模型，综合考虑均值突变与方差突变，全面分析了时间趋势项、截距项、滞后项或方差可能发生的突变。基于贝叶斯推断的基本思想，通过构造似然函数，同时利用先验信息和样本信息，为模型参数引入分层先验分布，并采用Gibbs抽样和MH算法完成突变点位置与个数的判断。进一步地，为了分析我国股价指数的结构突变问题，本文选取上证指数作为研究对象。首先基于滞后系数结构突变的AR(p)模型对上证指数的结构突变问题进行了细致的研究，发现1992年5月至2012年4月上证指数共存在14个结构突变点，其中4个为均值-方差双重变点、1个为均值变点、9个为方差变点。在此基础上详细阐述了均值-方差双重变点的成因。得出的结论是：上证指数发生均值方差双重突变，更多的受国内宏观经济和国外经济形势的影响。接着，采用滞后系数恒定的结构突变AR(p)模型实证分析上证指数结构变点，，并与本文的研究进行比较，结论是：在上证指数结构突变问题上，滞后系数结构突变的AR(p)模型更有效。最后，给出本文的结论和政策建议：中国股市具有明显的周期性，可划分为高速下行、缓慢下行、上行三个阶段。长期而言，政府的多次救市操作大部分都未达到改变股价指数运行趋势的目标。为保持国内股票市场的健康、快速发展，国家应大力发展国内经济，保持经济平稳快速发展，并营造良好的经济环境，进一步完善国内股市的各项制度。
[Abstract]:As a new market, China's stock market often fluctuates because of the imperfection of legal system, market trading and operation mechanism, and the immaturity of investment psychology of micro trading subjects. Therefore, the probability of structural mutation is greater. However, there is still a lack of detailed and systematic research on the structural mutation in China's stock market. Therefore, the problem of structural mutation in China's stock market is studied. It is of great theoretical and practical significance for financial risk management and policy formulation to further explore the main causes of structural catastrophe. First of all, this paper summarizes the research status of structural mutation detection methods and stock price index structural mutation at home and abroad. On this basis, combined with Bayesian econometrics, For the first time, the binomial distribution is introduced into the ARP) model of the structural mutation of the lag term coefficient. Considering the mean and variance mutations, the time trend term and intercept term are comprehensively analyzed. Based on the basic idea of Bayesian inference, by constructing likelihood function and using prior information and sample information, a hierarchical prior distribution is introduced for model parameters. Gibbs sampling and MH algorithm are used to judge the location and number of catastrophe points. Furthermore, in order to analyze the structural catastrophe of the stock index in China, this paper selects the Shanghai Stock Exchange Index as the research object. Firstly, the structural catastrophe problem of the Shanghai Stock Exchange Index is studied in detail based on the ARGRP model of the structural mutation of the lag coefficient. From May 1992 to April 2012, 14 structural mutation points were found in Shanghai Stock Exchange Index. Four of them are mean variance double change points, one is mean variation point, and nine are variance change points. On this basis, the causes of mean variance double change point are expounded in detail. The conclusion is that the Shanghai Stock Exchange index has double mean variance mutation. It is more affected by the domestic and foreign macroeconomic situation. Then, the structural change point of Shanghai stock index is empirically analyzed by using the structural mutation ARP model with constant hysteresis coefficient, and compared with the research in this paper. The conclusion is that the ARP model is more effective for structural mutation of Shanghai stock index. Finally, the conclusions and policy suggestions are given: the Chinese stock market has obvious periodicity, which can be divided into three stages: high speed downward, slow downward and uplink. Most of the government rescue operations failed to achieve the goal of changing the operating trend of the stock index. In order to maintain the healthy and rapid development of the domestic stock market, the state should vigorously develop the domestic economy and keep the economy developing steadily and rapidly. And create a good economic environment, further improve the domestic stock market system.
【学位授予单位】：华侨大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：F832.51;F224

【相似文献】