中国股票市场价值反转投资策略有效性研究

【作者】 肖军；

【导师】 徐信忠；

【作者基本信息】 对外经济贸易大学 ， 国际贸易， 2003， 博士

【摘要】 股票市场有效性是一个在金融领域争论了近百年的问题，而如今，学者们依旧对它情有独钟。股票市场如果是有效的，那么任何投资者将不可能持续获得无风险的超额收益。价值反转投资策略在美国等发达国家股票市场上的显著有效对股票市场有效性提出了严峻的挑战，然而对这种投资策略超额利润的产生原因，学者们并未形成一致看法。 本文以中国深沪A股股票市场为考察对象，试图解决两个问题：(1)在中国股票市场上，价值反转投资策略是否有如同美国、日本股票市场那样显著的效果。(2)如果在中国股票市场上存在显著的价值投资策略效果，那么超额利润产生的原因是什么?是风险溢价还是投资者的系统性定价失误，或两者兼而有之。最后，作者对资产定价模型进行了改进。 作者采用目前较为前沿的金融计量等实证研究方法对上述内容展开研究，具体研究内容和方法如下： 在文章第二章，作者首先以买入并持有组合收益率(buy—hold return)和Fama(1998)所支持的组合月均收益率为标准考察了规模效应，结论是规模效应非常显著。接着作者先以买入并持有组合收益率(buy—hold return)为标准考察了持有期为三年、二年和一年的一维、二维等权和权重价值反转投资策略的效果，得出结论：在持有期为三年、二年的策略中，以B/M为指标构造的一维价值投资策略效果最显著，以B/M—GS、E/P—B/M、B/M—CF/P和GS为指标构造的价值投资策略效果次之；但在持有期为一年的策略中以B/M—GS为指标构造的一维价值投资策略效果最显著，以E/P—B/M、B/M、B/M—CF/P和GS为指标构造的价值投资策略效果次之。作者接着以Fama(1998)所支持的组合月均收益率为标准重新考察了上述价值反转投资策略的效果，得出了基本相同的结论，但与前者相比，后者的超额收益率显著性有所提高。这启发我们能否结合规模效应和B/M效应构造出一种新的二维价值投资策略，结果是成功的。作者对不同规模股票组合的价值投资策略效果进行了考察，发现在B/M、E/P—B/M价值投资策略中，小规模股票组合的投资效果高于大规模股票组合的投资效果，而在GS、BM—GS价值投资策略中，小规模股票组合的投资效果低于大规模股票组合的投资效果。这说明投资者对小规模公司的GS指标持谨慎态度。投资者对小公司的经营业绩不稳定性有一定的认识，反而会更容易对大公司的经营业绩产生过度外延反应。最后，作者对价值投资策略的一月和二月效应进行了考察，发现:价值投资策略的超额利润不是由一月效应和二月效应产生，价值投资策略在一月和二月的表现低于在其他月份的平均表现，而且在二月份超额利润更低。这与在美国等发达国家股票市场中价值投资策略超额利润在一月最为显著的现象有所不同。 在文章的第三章，作者利用传统的风险指标。，·刀，和全距以及夏普指数、特雷诺指数和詹森指数对上述持有期为一年的一维、二维等权和权重价值反转投资策略的价值投资组合和魅力投资组合的风险和投资业绩进行了计算，同样从买入并持有收益率和组合月均收益率两个角度入手。作者发现无论是利用买入并持有收益率或组合月均收益率，价值投资组合的业绩均高于魅力投资组合的业绩和市场组合业绩。但在风险度量上，利用买入并持有收益率或组合月均收益率计算的组合风险却有较大差异，具体表现为:在以买入并持有收益率为标准时，价值组合的风险小于魅力组合风险，这样的结果明显支持投资者的过度反应。但在以组合月均收益率为标准时，价值组合的风险大于魅力组合风险，这样风险溢价对价值投资策略的超额利润有一定的解释作用，这正是LSv(1994)和Fama(1995)对价值投资策略超额利润来源的分歧根源之一。最后，作者检验了在不同情况下价值投资策略的表现差异，发现价值组合的市场下降风险并未超过魅力组合。 在文章的第四章，作者利用LSV(1994)的思路，检验了过度外延模型，结论为:在价值投资策略中存在着明显的过度外延现象。而且作者发现C用和E护指标并不是划分价值投资组合和魅力投资组合的有效指标。而且二维指标对价值投资组合和魅力投资组合的区分能力高于一维指标，这与LSv(1 994)的结论相同。最后，价值组合和魅力组合在组合形成前后的收益率反转现象也支持了DeBondt&Thaler(1 985)有关股票市场存在过度反应的观点。 在第五章，作者引入了新的风险因子，利用多因子模型来解释价值投资策略的超额利润。新引入的风险因子包括B服因子、规模因子、协偏度(coskewness)和协峰度(cokurtosis)因子。结论为:CAPM无法解释价值反转投资策略的超额利润的产生原因;Fama一French三因素模型对价值反转投资策略的超额利润的解释能力最为显著;但对于有些投资组合，在Fama一French三因素加上协偏度后，解释能力超过原来的Falna-F~h三因素模型，而且在引入的新的风险因子，有关支持投资者过度反应的证据消失，这是Lsv(1 994)和Fama(1995)对价值投资策略超额利润产生原因的意见分歧的另一个产生根源。 在第五章的基础上，作者在第六章进一步研究了资产定价问题。作者从多因子模型对股票?更多还原

【Abstract】 The Efficiency Market Hypothesis (EMH) on Stock Market is evergreen discussing focus in financial field, which has been coming into academician’s eyes for nearly one hundred years. According to EMH, it is impossible for any investor to get risk-adjusted anomalous return for a certain long period. The Value Contrarian Investment Strategy (VCIS), which has been used by investors for over seventy years, and has been proved significantly in economic terms, take a great challenge to EMH. At the same time, there is not a consensus between academicians on the reason why VCIS can bring in anomalous return.The purpose of this dissertation is to address the two following research proposals: firstly, is there any significant anomalous return produced by VCIS in China stock market, which has been proved in US Stock Market; Secondly, what is true reason behind this significant anomalous return, the reason may be risk premium or investor ’s systematic error pricing, or both? Finally, the author make an attempt to improve existing asset pricing models.The author adopt several update financial econometrical methods to carry out above-mentioned research proposals, please see the details in the following.hi Chapter Two, firstly, the author find significant size effect. Secondly the author evaluate the anomalous return produced by three- year-holding, two- year-holding , one- year-holding VCIS ,including equal-weighted and value- weighted, in the most popular Buy-Hold Average Returns (BHARs) and Average Monthly Returns (AMRs)term advocated by Fama(1998). The conclusions drawn from two methods are same: VCIS is effective investment strategy and can produce significant anomalous return, but the statistic significance from the latter is larger than that of the former. Based on the finding, the author construct a very successful investment strategy, the indices of which is stock capitalization and B/M factor. In additional, the author evaluate the anomalous return produced by VCIS, including equal-weighted and value- weighted, after controlling stock capitalization. Finally the author evaluate the anomalous return produced by VCIS in January and February .In Chapter Three, the author adopt conventional risk indices including σp, Bp andFull Range, and such portfolios management evaluation ratios as Jenson’s Alpha , Treynor Ratio and Sharpe Ratio to evaluate risk-adjusted investment performance and relevant risk indices of Value stock portfolio and of Glamour stock portfolio in Buy-Hold Average Returns (BHARs) and Average Monthly Returns (AMRs) term. The conclusions drawn from two methods are same: VCIS is good and can produce significant final result, but the risk calculated from the latter method is larger than that of the former. This is one of source of divergence between LSV (1994) and Fama(1995) on the reason why VCIS can bring in anomalous return.In Chapter Four, the author adopt the ideas of LSV (1994) to test the naive Exploration Hypothesis. The conclusions drawn from it are that there is clear investor’s tendency of overdue Exploration of past performance. In addition , the author find that the two indices :C/P and E/P are not valid indices to distinguish Value stock from Glamour stock, and two-dimension indices have better ability to distinguish Value stock from Glamour stock than one -dimension indices ,which is same as LSV (1994) .Finally, the evidence of return mean-reverting from this chapter support the ideas of DeBondt& Thaler (1985) on stock overreaction.In Chapter Five,, the author adopt new risk indices to construct multi-factor models to explain the anomalous return produced by VCIS. The new risk indices include B/M factor size factor coskewness and cokurtosis. The conclusions drawn from it are that: CAPM has not any ability to explain the anomalous return produced by VCIS; Fama-French three-factor model has the most significant ability to do so. It is noted that ,after including coskewness, the four-factor model has a greater significant ability than that of Fama-French three-factor model. Af更多还原