股票间相关性测量方法的研究及应用

发布时间：2018-04-26 23:21

本文选题：股票相关性 + 金融时间序列　；参考：《哈尔滨工业大学》2017年硕士论文

【摘要】：2014年以来我国股市进入一轮显著的牛市行情,全国掀起一波火热的炒股的浪潮,技术分析在投资者投资决策中的作用和需要日渐重要。股票相关性是研究股价或者收益率间的关系和行业分类的技术工具,它对股票市场系统性风险与资产组合的有效性的衡量具有重要价值。所以,个人与机构投资者均把股票间的相关性作为一个重要标准,以此权衡股市风险的大小与组建的的投资组合有效性。通常情况,对于股票相关性的衡量,国内外学者用相关系数大小加以表示,股票相关系数越大,相关性就越强。然而在我国股票市场中显著的牛市或熊市行情会引起不同行业的股票间过度相似的的同涨同跌现象,说明单边上涨或下跌行情中引起了股票间的超额联动效应,这将导致测算的股票间相关系数偏离真实值,影响投资者的投资决策、行业分类及投资组合有效性。因此,研究出能尽可能规避股票间的超额联动效应,测算出反映股票收益率间真实相关性的相关系数对于行业股票相关性的研究有非常重大的应用价值。针对该问题,本文首先研究了OLS、资本资产定价模型(CAPM)及不同股市行情特征下的股票相关性分析方法,并以上证A股行业股票的对数收益率为实证研究对象,通过大数据程序算法验证了股票市场中超额联动效应存在的普遍性。其次,针对CAPM模型测算股票收益率存在的超额联动效应,本文将传统的CAPM公式进行相应的变形剔除引起超额联动效应的市场系统性风险,构造出改进的股票相关性的测算模型;应对金融时间序列下存在的超额联动效应,本文通过自动依照成交量截取、人工截取、依照滚动时间窗口的股票间相关系数截取等方法构造尽可能规避OLS及行情特征下的超额联动效应影响的新型行业股票间相关性测量方法。最后,文章以上证A股行业股票进行实证研究,将尽可能规避股票间超额联动效应的新型相关性测算方法计算的相关系数与传统测算方式进行横向与纵向对比分析,验证出新型相关性测算方法的合理有效性。同时,文章将尽可能规避超额联动效应的股票间相关性的测量方法测算的新型相关系数应用于复杂网络对股票间的相关性的研究中。规避超额联动效应的股票间相关性的新型测量方法更有利于提高投资者投资决策、投资组合选择的有效性以及行业分类的准确性。
[Abstract]:Since 2014, China's stock market has entered a remarkable bull market, and a wave of hot stock speculation has been launched all over the country. Technical analysis plays an increasingly important role and needs in investors' investment decisions. Stock correlation is a technical tool to study the relationship between stock price and yield and industry classification. It is of great value to measure the systematic risk and the efficiency of portfolio in stock market. Therefore, both individual and institutional investors take the correlation between stocks as an important criterion to weigh the stock market risk and the efficiency of the portfolio. In general, for the measurement of stock correlation, domestic and foreign scholars use the correlation coefficient to express it. The greater the stock correlation coefficient is, the stronger the correlation is. However, in China's stock market, a significant bull market or bear market will lead to excessive and similar simultaneous rise and fall among stocks in different industries, indicating that a unilateral rise or fall has caused an excess linkage effect between stocks. This will lead to the deviation of the correlation coefficient between the measured stocks from the real value, which will affect the investment decision, industry classification and portfolio efficiency of investors. Therefore, the study can avoid the excess linkage effect of stock as much as possible, and calculate the correlation coefficient which reflects the real correlation between stock returns. It is of great application value for the research of industry stock correlation. To solve this problem, this paper first studies OLS, capital asset pricing model (CAPMM) and stock correlation analysis methods under different market market characteristics, and takes the logarithmic return rate of A shares in Shanghai Stock Exchange as the empirical research object. The universality of excess linkage effect in stock market is verified by big data program algorithm. Secondly, in view of the excess linkage effect of CAPM model, the traditional CAPM formula is deformed to eliminate the market systemic risk of excess linkage effect, and an improved stock correlation calculation model is constructed. In order to deal with the excess linkage effect in financial time series, this paper, by automatically intercepting according to the transaction volume, manually intercepting, According to the method of interstock correlation coefficient interception of rolling time window, a new type of industry stock correlation measurement method is constructed to avoid the influence of excess linkage effect under OLS and market characteristics as much as possible. Finally, the article carries on the empirical research with the Shanghai Stock Exchange A share industry stock, carries on the horizontal and the longitudinal contrast analysis to the correlation coefficient and the traditional calculation method which avoids the excess linkage effect of the stock as far as possible. The validity of the new method is verified. At the same time, this paper applies the new correlation coefficient to the study of the correlation between stocks based on complex networks. A new measurement method to avoid the correlation between stocks with excess linkage effect is more helpful to improve the investment decision, the validity of portfolio selection and the accuracy of industry classification.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F832.51

【参考文献】