三因子模型在沪深A股市场的实证研究

发布时间：2018-01-06 17:21

本文关键词：三因子模型在沪深A股市场的实证研究　出处：《复旦大学》2012年硕士论文　论文类型：学位论文

【摘要】：目前,在国外特别是美国,Fama-French三因子模型已被大多数人所接受,并被广泛应用于收益率预测、风险管理、基金业绩评价等各个方面。由于中国证券市场的出现晚于西方国家上百年,中国证券市场存在着投机跟风严重、信息披露不透明、信息失真等不成熟市场所具有的特点。因此以西方成熟市场为基础的Fama-French三因子模型在中国证券市场的适用性受到了学者们的质疑。过去许多学者对Fama-French三因子模型在中国证券市场是否适用做过很多相关研究,但得到了一些不同的结论。本文在扩大样本的基础上,选取的数据包括沪深A股市场全部上市公司(除创业板),数据样本区间为1995年5月到2011年12月,检验了Fama-French三因子模型在中国证券市场上是适用的。另外,本文还对三因子模型的回归系数进行了分析。(1)利用肯德尔和谐系数(Kendall's Coefficient of Concordance)检验方法研究,发现SMB系数s和HML系数h具有多期排序稳定性；市场因子(RM-RF)的系数的多期排序稳定性弱于SMB系数s和HML系数h。(2)市场组合(RM-RF)的回归系数b几乎不可能小于0,因此我们很难在中国这个卖空不太容易的市场上,通过投资分散化来完全消除市场因子(RM-RF)的风险,但我们可以尽量做到消除风险因子SMB、HML的影响。(3)使用日收益率数据比月收益率数据更能使回归系数稳定。(4)总的来说,特别是对于风险因子SMB的回归系数s和风险因子HML的回归系数h,增加股票组合中所含股票的数目会使回归系数的预测结果更加稳健。(5)按照回归系数的大小排列形成的投资组合,其组合的回归系数存在着均值回复的特征,特别是对于当期回归系数最大的那个组合,下期的回归系数会下降很多,但是回归系数的排序几乎不发生变化。
[Abstract]:At present, especially in the US, Fama-French three factor model has been accepted by most people, and has been widely applied in various aspects, such as yield prediction, risk management, fund performance evaluation and so on.
Due to the emergence of Chinese stock market later than in western countries for hundreds of years, China securities market there is speculation to follow suit serious, transparent information disclosure, information distortion and immature market features. Therefore the applicability based on Western markets Fama-French three factor model in China stock market has been questioned by scholars. In the past many scholars on whether the Fama-French three factor model is done a lot of related research in the Chinese securities market, but there have been some different conclusions. In this paper, on the basis of the sample expansion, the selected data including Shanghai and Shenzhen A stock market listed companies (except the GEM), data sample interval from May 1995 to December 2011, inspection Fama-French the three factor model is applicable in China securities market.
In addition, the regression coefficient in the three factor model is analyzed. (1) by Kendall (Kendall's Coefficient of coefficient Concordance) test method, SMB coefficient s and HML coefficient h with multi period scheduling stability; market factor (RM-RF) coefficients of multi period sequence stability coefficient s and weaker than SMB HML coefficient h. (2) market portfolio (RM-RF) regression coefficient B can hardly be less than 0, so it is difficult for us in this is not easy Chinese short selling on the market, through the diversification of investment to completely eliminate market risk factor (RM-RF), but we can try to eliminate the risk factor of SMB, the influence of HML. (3) the use of daily return data than the monthly return data to make regression coefficient stability. (4) in general, especially for the regression coefficients of risk factors SMB regression coefficient s and risk factors of HML h, increase the stock portfolio contained in stock The number of votes will make the prediction of regression coefficients results more robust. (5) according to the size of the regression coefficient are arranged to form a portfolio, the regression coefficient combination exists characteristics of mean reversion, especially for the combination of the current regression coefficient is the largest, the number of regression lines next will drop a lot, but the sort of regression coefficient there is almost no change.

【学位授予单位】：复旦大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：F832.51;F224

【引证文献】