基于投资者情绪的多因素资产定价模型设计及实证分析

发布时间：2018-10-19 17:03

【摘要】：随着行为金融学的蓬勃发展,投资者情绪受到了越来越高的关注。本文主要研究了投资者情绪与情绪资产定价模型。我们提出了一套构建复合投资者情绪指标的方法,并说明这套方法可以建立出优良的情绪指标。进一步地,将该指标作为情绪因素,引入CAPM和FF三因子模型,可以对现有的资产定价模型起到明显的改善。首先,本文对情绪指标进行了界定,对原始的情绪指标进行了说明。情绪指标可以分为两类,一类是反映投资者情绪的指标,另一类是影响投资者情绪的指标。投资者情绪是一个灰箱,影响投资者情绪的指标是输入项,反映投资者情绪的指标是输出项。本文中着重于反映投资者情绪的指标,这类指标包括换手率、封闭式基金折价率等,它们是构建复合情绪指标的原始数据。同时,本文还建立了一套评价体系,来判断情绪指标的优劣。优秀的情绪指标需要与股票价格趋势有关联性,情绪指标的变化需要对股票收益率产生影响,而且还需具有稳健性。然后,我们建立了状态空间模型来计算复合投资者情绪指标。由于模型中的参数是未知的,需要把该模型转化为自适应系统识别问题,即非线性状态空间模型,之后再利用扩展的卡尔曼滤波来求解。虽然主成分分析法和TOPSIS法也可以构建复合情绪指标,但是状态空间模型法建立的指标与股票价格趋势的因果性更强,其变化与股票收益率的因果性更强,而且稳健性更好。所以,相比主成分分析法和TOPSIS法构建的情绪指标,状态空间模型法得到的复合投资者情绪指标是最优的。最后,本文在现有的资产定价模型中引入了复合投资者情绪指标。我们发现在CAPM或FF三因子模型中添加情绪因子,均会使AIC下降、调整后的可决系数上升、拟合优度提高,同时,引入情绪的资产定价模型有更强的预测能力。资产定价模型原有的单因子或三因子在解释收益率时,还存在无法解释的部分,情绪因子的引进,会对无法解释的因素做出一部分合理解释。这说明引入的情绪因子是有效的,它可以改善现有的资产定价模型。
[Abstract]:With the vigorous development of behavioral finance, investor sentiment has been paid more and more attention. This paper mainly studies investor sentiment and the pricing model of emotional assets. We put forward a set of methods to construct complex investor sentiment index, and show that this method can establish good emotion index. Furthermore, using this indicator as an emotional factor and introducing CAPM and FF three-factor model can improve the existing asset pricing model. First of all, this paper defines the emotional indicators and explains the original emotional indicators. Emotional indicators can be divided into two categories, one is an indicator of investor sentiment, the other is an index that affects investor sentiment. Investor sentiment is a gray box, the index that affects investor sentiment is input item, the index that reflects investor sentiment is output item. This paper focuses on the indicators that reflect investor sentiment, such as turnover rate, closed-end fund discount rate and so on, which are the original data of constructing compound emotion index. At the same time, this paper also established a set of evaluation system to judge the merits and demerits of emotional indicators. Excellent emotional indicators need to be related to the trend of stock prices, and the changes of emotional indicators need to have an impact on stock returns, and also need to be robust. Then, we establish a state space model to calculate the composite investor sentiment index. Because the parameters in the model are unknown, it is necessary to transform the model into an adaptive system identification problem, that is, the nonlinear state space model, and then use extended Kalman filter to solve the problem. Although principal component analysis (PCA) and TOPSIS method can also be used to construct composite emotional indicators, the state space model method has stronger causality to stock price trend, stronger causality to stock return and better robustness. Therefore, compared with the emotional indexes constructed by principal component analysis and TOPSIS method, the state space model method is the best one. Finally, this paper introduces the composite investor sentiment index into the existing asset pricing model. We find that adding emotional factors to the CAPM or FF three-factor model can decrease AIC, increase the resolution coefficient and improve the goodness of fit. At the same time, the asset pricing model with emotion has stronger predictive ability. When the original single or three factors of the asset pricing model explain the yield, there are still some parts that can not be explained. The introduction of the emotion factor will give some reasonable explanation to the unexplained factor. This shows that the introduced emotional factor is effective, it can improve the existing asset pricing model.
【学位授予单位】：吉林大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：C912.6;F832.51

【参考文献】