基于分位数回归方法的A股市场收益率影响因素的研究

发布时间：2018-01-03 19:04

本文关键词：基于分位数回归方法的A股市场收益率影响因素的研究　出处：《哈尔滨工业大学》2012年硕士论文　论文类型：学位论文

【摘要】：关于股票收益率的影响因素的理论和模型一直是各国学者研究的重点问题，从最初的资本资产定价模型和套利定价理论到三因素模型等各种因素模型，学者们在探索影响因素的过程中发现了许多模型和理论并不完善。尽管学者们不断的改变影响因素和完善模型，但所有的检验方法都是检测收益率均值处的效应，常常因模型和数据样本的不同往往会得出不同的实证结果，尤其是很多在外国得到论证的结论在中国市场上并不成立。本文立足中国股市并结合学者们的研究成果选取了26个可能影响收益率的影响因素。然后利用聚类分析方法从中筛选了12个代表性因素。最后以中国A股市场股票为研究对象，以最小二乘法和分位数回归方法对12个影响因素进行全部数据和分行业的分析。经过分位数和最小二乘法两种回归分析方法的对比发现：（1）同一个影响因素与股票收益率在最小二乘法下的相关性和显著性是唯一的，但在分位数回归方法下不同分位点处的结果却各不相同；另外，同一影响因素在不同行业中与收益率的关系在不同分位点处也是不同的，因此还要分行业利用分位数回归方法分析二者之间的关系。（2）全部数据分析结果显示在大多数分位点处分位数回归的拟合效果好于最小二乘法的拟合效果，，但分行业分析后，最小二乘法的拟合效果更好。（3）无论是全部数据还是分行业分析，大多数财务指标对股票收益率的解释力度远不如一些公司外部的特殊因素对收益率的解释，表明我国大多数投资者在进行价值和投资分析时更加重视那些反映公司市场表现的特殊指标对股票收益率的影响力，忽略了反映公司经营业绩的财务指标的影响力。通过两种回归分析方法的结果对比可以看出来，最小二乘法下回归系数、敏感性和拟合度都是唯一的；但在分位数回归方法下，不同分位点处的回归系数、敏感性和拟合度差异还是很明显的，能深刻的刻画这三个变量的变化及趋势。为以后更加深入的研究股票收益率影响因素提供了更好的分析方法。
[Abstract]:On the impact of stock returns and the theory model of factors has been the focus of researchers all over the world, the model from the capital asset pricing model and arbitrage pricing theory to the original three factor model and other factors, the scholars in the exploration of the influencing factors were found in many models and theories are not perfect. Although the factors to affect the scholars continuous and perfect model, but the test methods are all effect detection yields at mean, often because of the model and data samples often have different empirical results obtained different, especially in many foreign get the conclusion of the argument does not hold in China market.
Based on the China stock market and the research results of the scholars selected 26 factors that may affect the rate of return. Then use the clustering analysis method is selected from 12 representative factors. At last China A shares stock market as the research object, analysis of all the data and industry of the 12 factors, using the least square method and the quantile regression method.
After comparing the two quantile and least squares regression analysis found that: (1) the same factors and the correlation of stock returns in the least squares method and is the only significant, but different in the quantile quantile regression method under the results are different; in addition, the relationship between the same factor effect of rate and yield in different industries in different sites is different, so the relationship between the industry and using the method of quantile regression analysis of the two. (2) all the data analysis results showed that the fitting effect of the fitting effect in most quantile quantile regression is better than the punishment of the least squares method, but the industry analysis after the fitting effect is better than the least square method. (3) whether all data and industry analysis, explanation of most financial indicators on stock returns than outside the company for some special The explanation of yield shows that most investors in China pay more attention to the influence of special indicators that reflect the company's market performance on the stock returns, ignoring the influence of financial indicators that reflect the company's performance.
Through the comparative analysis of two kinds of regression results we can see that the least squares regression coefficient, sensitivity and fitting degree is unique; but in the method of quantile regression, the regression coefficients of different sites, and the fitting degree of sensitivity difference is very obvious, which can describe the changes and trends of the three variables deep. For a more in-depth study of stock returns influencing factors analysis method is provided for the better.

【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】：2012
【分类号】：F832.51;F224

【参考文献】