【作者】 刘跃；

【导师】 王志强；

【作者基本信息】 东北财经大学 ， 金融工程， 2021， 博士

【副题名】来自中国A股市场的证据

【摘要】 资本资产定价模型(CAPM)是线性单因子定价模型的研究基础,经过半个世纪的发展,从中演化而来的因子投资实践凭借其在资产配置和获取超额收益方面的卓越表现,使之成为当今实证资产定价领域重点的研究对象。在传统金融学的理论框架中,资产定价模型对于个股预期收益始终遵循风险补偿原则,认为定价因子面临风险暴露就需要获得高于交易价格的风险溢价,即更高的预期收益。但随着市场上出现更多不能被主流资产定价模型所解释的收益异象,风险补偿原则受到各方质疑,同时在理论分析和实证研究中也发现传统因子定价模型可能在构建逻辑上存在瑕疵。股票市场中系统性风险对应着价格溢价,多因子定价模型表现不够优秀是因为风险的代理变量并不能完全覆盖描述市场中存在的系统性风险。在构建定价因子时对于市场中投资者是理性人的假设也与中国A股市场情况不符,在沪深股市中遵从价值投资的投资者占比较少,个人投资者较机构投资者占比较多,同时股票的换手率和股价波动率普遍偏高,这些非理性的投资特征较为明显。传统定价因子构建方法一种是基于有效市场假说下的理性资产定价理论,另一种是应用理性资产定价理论中隐含的结构性约束去对资产的收益率或资产投资组合的排序特征进行统计分析来提取因子特征,按照构建原则可以分为公司基本面特征因子和经验技术类因子,这样的构建方法和划分方式缺少对投资者行为的关注,忽视了投资者在投资行为上的偏差所造成对股票错误定价的影响。基于以上逻辑,本文的创新点和研究内容如下:第一,系统的梳理了国内外传统定价因子、收益异象和主流多因子定价模型的理论文献和实证文献,通过对文献资料的分析,尝试论述在我国A股市场可以用行为金融学理论构建行为定价因子,力争能够对构建我国股票定价因子在研究思想脉络和方法论上实现创新。通过对国内外关于因子投资领域的理论文献和实证文献进行系统分析,发现我国投资者的行为偏差是造成股票被市场错误定价的原因之一,通过对构建系统性定价因子的方法研究和心理学原理进行交叉论述后,发现投资者过度自信和有限注意力的心理学原理能够作为本文实证研究的理论支撑,可以作为行为定价因子的代理变量。结合中国A股市场的特征,尝试提出从投资者行为角度去衡量股票被错误定价程度,在行为金融学理论框架内构建和优化系统性行为定价因子。第二,在Daniel等(2020)研究基础上,优化构建了长、短时间期限行为定价因子和行为因子定价模型,并使用1997年1月1日至2020年12月31日中国A股市场量价交易数据进行定价效率的检验,实现了在我国实证资产定价领域中因子投资工具和模型的使用创新。本文以行为金融学理论作为基础,在心理学领域内寻找能够衡量投资者行为偏差的代理变量:使用中国A股上市公司5年复合股权发行量作为排序变量,用来衡量由于投资者过度自信造成的对股票长期错误定价的偏差程度,优化构建了长期行为因子;用中国A股上市公司财报披露日前后4天的累计异常收益率作为排序变量,用来捕捉投资者有限注意力关注原因导致对股票市场新出现的盈利信息反应不足产生的短期错误定价,构建了短期行为因子。在与传统定价因子的统计性分析中发现长期行为因子构建的投资组合相比于其他定价因子构建的投资组合在平衡市场风险与收益率之间的关系具有比较优势。在与传统定价因子的相关性分析中,发现长期行为因子和短期行为因子之间、行为定价因子与传统定价因子之间在函数单调性和几何意义上均不存在显著的相关性,说明每个定价因子在捕捉错误定价能力上具有独立性。在构建行为因子时使用的数据与过往研究存在主要区别:(1)预测个股收益过程中使用行为因子载荷作为抑价定价的代理变量;(2)使用A股市场样本股票1个月内的超额收益进行日度的时间序列排序来替代常规60个月的股票量价数据回归计算行为因子的因子载荷。在行为因子预测个股收益率方面,发现长期行为因子的因子载荷对个股收益未来走势有积极显著的预测能力,稳定性检测中,在控制了 7个类别的24个常规异象变量后,预测能力依然积极稳健。同时发现,并不能明确短期行为因子的因子载荷同样具有预测个股收益的能力,原因之一是短期行为因子的因子载荷在估计时可能存在系统性偏误。在行为因子对收益异象投资组合中多头、空头收益率影响的实证研究中,发现长、短期行为因子的因子载荷能够更好地解释投资组合中被高估的空头收益率。进一步在对行为因子个股收益预测效率影响的实证研究中,发现中国股票市场上存在的交易摩擦,对长期行为因子个股收益预测效率有显著提升作用。同时认知偏差对长期行为因子的个股收益预测效率亦有影响,用分析师覆盖率和分析师分歧度作为投资者在投资行为上的认知偏差程度的代理变量,研究发现投资者认知偏差的存在使得长期行为因子的因子载荷对个股预期收益率的敏感性会显著提升。本文同时验证了机构投资者在投资过程中并没有发挥价格发现的作用,反而会做出泡沫骑乘的理性投机行为,实证结果显示随着机构持股比例的增大,长期行为因子投资组合的收益率显著提高,机构投资者股票价格泡沫骑乘行为能够显著提升长期行为因子个股预期收益率,长期行为因子能够解释机构投资者股票价格泡沫骑乘行为。在对A股市场存在收益异象的解释效率上,构建的行为因子模型表现优异,尤其在解释长期收益异象上具有比较优势。第三,检验证明我国A股市场存在长时间和短时间范围内的行为收益异象,实现了对A股市场收益异象划分类别完善性上的创新。本文分别从定价因子构建多空投资组合获得异常收益角度,从定价因子相关性角度,同时使用单变量组合价差法以及定价因子间相互解释能力四个方面系统对A股市场长、短期行为收益异象的存在性进行实证检验,并且在进一步控制住22个影响公司股价变动的基本面特征变量后,证明了中国A股市场在长时间(一般3-5年)和短时间(不满1年)的期限内稳定存在行为收益异象。第四,构建能够预测股价崩盘事前概率的三元Logit模型,用长期行为因子与构建股价崩盘事前概率模型联合使用预测股价崩盘风险,实现了两者在联合使用方法上的创新。该模型能够把股票未来实现的极端负收益率、极端正收益率和正常收益的概率区别开。并且证明了使用样本外预测方法计算股票价格崩盘概率是有效的。通过实证检验发现个股复合股权发行量越大,投资者行为偏差导致的股票价格被高估程度就越高,个股未来预期收益率会越小,同时股票价格崩盘风险概率就会显著增高。进一步研究发现,股灾期间出现股票流动性枯竭的极端市场风险不会影响长期行为因子通过股价崩盘事前概率模型预测股价崩盘风险的能力,而且长期行为因子预测股价崩盘能力具有时变性,表现为随着时间推移预测能力显著提升。在同时考虑了企业规模和价值的影响后,发现长期行为因子对预测股票价格崩盘风险的能力显著提升,其中企业规模作为控制变量提供较大增量贡献。在机构投资者行为对长期行为因子预测股价崩盘能力的实证研究中,发现对于同一组股票,复合股权发行量CSI值越大,具有高机构持股比例的股票价格崩盘风险就越高,机构投资者的存在可以解释长期行为因子对崩盘概率的预测能力。进一步地研究发现股票价格崩盘事前概率对长期行为因子在获取未来收益率方面没有显著影响,不具备内在相关性。总体来看,基本实现了股价崩盘概率模型和行为因子联合使用预测股价崩盘风险在方法论上的一次创新尝试。本文以行为金融学为理论依据,遵循美国证券市场成熟的实证资产定价研究标准范式,并借鉴了构建系统性定价因子的研究方法,使用中国A股市场股票量价交易数据构建了行为定价因子,并证明我国股市存在长、短两个时间期限范围内的行为收益异象,发现了行为因子对个股未来收益有稳健的预测能力,进一步研究发现股票市场上存在的交易摩擦、投资者认知偏差对行为因子个股收益预测效率存在影响,机构投资者的存在对行为因子个股未来收益率预测准确性有影响。证明了构建的股价崩盘事前概率模型和长期行为因子联合使用能够定量预测股价崩盘风险。在论证过程中总结的思想和方法丰富了我国已有实证资产定价的研究思路,实证结果亦对我国因子投资实践具有重要的参考价值。更多还原

【Abstract】 The Capital Asset Pricing Model(CAPM)is the research foundation of the linear single-factor pricing model.After half a century of development,the factor investment evolved from it has made it an empirical evidence today by virtue of its outstanding performance in asset allocation and obtaining excess returns.The key research object in the field of asset pricing.In the theoretical framework of traditional finance,the asset pricing model follows the principle of risk compensation for the expected return of individual stocks.It is believed that when the pricing factor faces risk exposure,it needs to obtain a risk premium higher than the transaction price,that is,a higher expected return.However,with the emergence of more income anomalies that cannot be explained by mainstream asset pricing models,the principle of risk compensation has been questioned by all parties.At the same time,theoretical analysis and empirical research have also found that traditional factor pricing models may have flaws in their construction logic.,The systemic risk in the stock market corresponds to the price premium,and the performance of the multi-factor pricing model is not good enough because the proxy variables of risk cannot fully cover the systemic risk that exists in the market.The assumption that investors are rational in the market when constructing pricing factors is also inconsistent with Chinese A-listed market.In the Shanghai and Shenzhen stock markets,there are relatively few investors who follow value investment,and individual investors account for more than institutional investors.At the same time,the turnover rate of stocks is generally high and the volatility is relatively high.These irrational investment characteristics are more obvious.One method of traditional pricing factor construction is based on the rational asset pricing theory under the efficient market hypothesis,and the other is to apply the structural constraints implicit in the rational asset pricing theory to make statistics on the return rate of assets or the ranking characteristics of asset portfolios.Analysis to extract factor characteristics,according to the construction principles can be divided into company fundamental characteristics factors and empirical technology factors,such construction methods and division methods lack of attention to investor behavior,ignoring the deviation caused by investors in investment behavior The impact on stock mispricing.Based on the above logic,the innovations and research contents of this article are as follows:Firstly,systematically combed the theoretical literature and empirical literature of traditional pricing factors,income anomalies,and mainstream multi-factor pricing models at home and abroad,and tried to argue that behavioral finance theory can be used to construct behavior in Chinese A-listed market through the analysis of literature data.Pricing factors,and strive to achieve innovation in the research ideology and methodology of the construction of Chinese stock pricing factors.Through systematic analysis of theoretical and empirical literature in the field of factor investment at home and abroad,it is found that the behavioral deviation of Chinese investors is one of the reasons for the mispricing of stocks by the market.Through the methodological research and psychology of constructing systematic pricing factors After cross-examination of the principles,it is found that the psychological principles of investor overconfidence and limited attention can be used as the theoretical support for the empirical research in this article,and can be used as proxy variables for behavioral pricing factors.Combining with the characteristics of Chinese A-listed market,this paper attempts to explain the mispricing of stocks from the perspective of investor behavior,and to construct and optimize systematic pricing factors.Secondly,based on the research of Daniel et al.(2020),optimized the construction of long-term and short-term behavioral pricing factors and positioning factor pricing models,and used the Chinese A-listed market from January 1,1997 to December 31,2020,The volume and price transaction data is used to test the pricing efficiency,which realizes the innovation in the use of factor investment tools and models in the field of empirical asset pricing in China.Based on the theory of behavioral finance,this article looks for proxy variables that can measure investor behavior deviations in the field of psychology:using Chinese A-listed companies’ 5-year review equity issuance as a ranking variable to measure the overconfidence caused by investors The degree of deviation of the long-term mispricing of stocks is optimized to construct a long-term behavior factor;the cumulative abnormal return rate of 4 days before and after the disclosure date of the Chinese A-listed company’s financial report is used as a sorting variable to capture investors’ limited attention The short-term mispricing caused by insufficient response to the newly emerging earnings information in the stock market constitutes a short-term behavioral factor.In the statistical analysis with traditional pricing factors,it is found that the investment portfolio constructed by long-term behavior factors has a comparative advantage compared with the investment portfolio constructed by other pricing factors in balancing the relationship between market risk and the rate of return.In the correlation analysis with traditional pricing factors,it is found that there is no significant correlation between long-term behavioral factors and short-term behavioral factors,and between behavioral pricing factors and traditional pricing factors.The pricing factor is independent in its ability to catch mispricing.There are differences between the data used in constructing behavioral factors and previous studies:(1)Using behavioral factor loads as underpricing as proxy variables in the process of predicting individual stock returns;(2)Using the excess returns of sample stocks in the A-listed market within 1 month for daily the degree of time series sorting replaces the conventional 60-month stock volume and price data regression to calculate the factor loading of the behavior factor.In terms of predicting the return rate of individual stocks by behavioral factors,it is found that the factor loading of long-term behavioral factors has a positive and significant predictive ability on the future trend of individual stocks’ returns.In stability testing,after controlling 24 conventional anomalous variables in 7 categories,the predictive ability still positive and stable.At the same time,it is found that the factor loadings of short-term behavior factors that cannot be clearly defined also have the ability to predict the return of individual stocks.One of the reasons is that the factor loadings of short-term behavior factors may be systematically biased in the estimation.In the empirical study on the influence of behavioral factors on long and short returns in the return anomaly portfolio,it is found that the factor loading of long-term and short-term behavior factors can better explain the overestimated short return in the portfolio.Further,in the empirical research on the impact of behavioral factors on the forecasting efficiency of individual stocks,it is found that the trading frictions in the Chinese stock market have a significant effect on the efficiency of long-term behavioral factor stocks forecasting.At the same time,cognitive biases also have an impact on the efficiency of long-term behavioral factors in the forecast of individual stock returns.Using analyst coverage and analyst divergence as proxy variables for investors’cognitive biases in investment behavior,the study found that investors’ cognitive biases The existence of the factor load of long-term behavior factors will significantly increase the sensitivity of individual stocks’ expected returns.This article also verified that institutional investors did not play the role of price discovery in the investment process,but instead made bubble-ridden investment behaviors.As the proportion of institutional lists increased,the return rate of long-term behavioral factor portfolios increased significantly.It is concluded that institutional investors’ riding behavior of stock price bubbles can significantly increase the expected return of long-term behavior factors of individual stocks,and that long-term behavior factors can explain institutional investors’ riding behavior of stock price bubbles.In terms of the efficiency of explaining the anomaly of return in the A-listed market,the constructed behavior factor model performs well,especially in explaining the long-term return anomaly.It has a comparative advantage.Thirdly,the empirical test proves that there are long-term and short-term behavioral income anomalies in Chinese A-listed market,which has achieved innovation in the perfection of the classification of the A-listed market anomaly.In this paper,from the perspective of pricing factors to construct long-short investment portfolios to obtain abnormal returns,from the perspective of pricing factor correlation,the univariate portfolio spread method and the ability to explain each other between pricing factors are used in four aspects to systematically affect the long-term and short-term behavioral returns of the A-listed market.The existence of the elephant is empirically tested,and after further controlling 22 fundamental characteristic variables that affect the company’s stock price changes,it proves that Chinese A-listed market has a long-term(usually 3-5 years)and short-term(less than 1 year)There is a stable behavioral income anomaly within the period of time.Fourthly,constructing a three-dimensional Logit model that can predict the pre-probability of stock price crashes,using long-term behavior factors and building pre-disruption probability models to predict the risk of stock price crashes,realizing the innovation of the combined use of them.This model can distinguish the probability of extreme negative returns,extreme positive returns and normal returns realized by stocks in the future.And it proves that it is effective to use the out-of-sample forecasting method to calculate the probability of stock price collapse.Through empirical testing,it is found that the greater the issuance of individual stocks for review equity,the higher the degree of overestimation of stock prices caused by investor behavior deviations,the smaller the expected future return of individual stocks,and the higher the probability of stock price collapse risks.Further research found that the extreme market risk of stock liquidity exhaustion during the stock market disaster will not affect the ability of long-term behavioral factors to predict stock price crash risks through pre-crash probability models,and the ability of long-term behavioral factors to predict stock price crashes is time-varying,which manifests itself as time-varying.The ability to predict progress has been significantly improved.After considering the impact of company size and value,it is found that long-term behavior factors have significantly improved the ability to predict the risk of stock price collapse,and the company size as a control variable provides a larger incremental contribution.In the empirical research on the ability of institutional investor behavior to predict stock price collapse by long-term behavioral factors,it is found that for the same group of stocks,the greater the CSI value of the rechecked equity issuance,the higher the risk of a stock price collapse with a high institutional listed holding ratio.Institutional investment the existence of the person can explain the long-term behavior factor’s ability to predict the probability of a crash.Further research found that the prior probability of stock price collapse has no significant impact on long-term behavioral factors in obtaining future returns,and has no inherent correlation.On the whole,it has basically realized an innovative attempt in the methodology of stock price collapse probability model and behavioral factors to predict the risk of stock price collapse.Based on the theoretical basis of behavioral finance,this paper follows the mature empirical asset pricing research standard paradigm of the American securities market,and draws on the research methods of constructing systematic pricing factors,using the stock volume and price trading data of Chinese A-listed market to construct behavioral pricing factors.It also proves that there are behavioral returns anomalies in the long and short time horizons of the Chinese stock market.It is found that behavioral factors have a robust predictive ability for the future returns of individual stocks.Further studies have found that there are trading frictions and investor cognitive biases in the market.It has an impact on the efficiency of predicting the returns of individual stocks of behavioral factors,and the existence of institutional investors has an impact on the accuracy of predicting future returns of individual stocks of behavioral factors.It is proved that the combined use of the constructed pre-probability model of stock price crash and long-term behavior factors can quantitatively predict the risk of stock price crash.The thoughts and methods summarized in the demonstration process have enriched our country’s existing empirical asset pricing research ideas,and the empirical results also have important reference value for factor investment practice.更多还原