我国证券市场呈现高度散户化的基本特征，且此情况将长期持续存在。这一尚不 成熟的市场发展状态意味着股市中有着大量不成熟、风险承受能力低的个体投资者， 他们的认知心理和行为偏差在市场上表现尤为突出。个体投资者进行交易决策时常常 受到专业人士、亲友、邻居、或同事等参考群体成员同期行为的影响，从参考群体成 员处获得的信息是形成个体投资决策的重要依据，盲从行为十分普遍。且个体投资者 羊群行为对市场稳定性的威胁和破坏比机构投资者更大。因此，确定我国个体投资者 羊群行为的内部驱动机理是一项十分迫切和重要的任务。 本研究首先采用问卷调查法，聚焦于参考群体如何与何时驱动个体投资者羊群行 为。构建两个嵌套的数据层通过跨层次回归模型分析参考群体与个体投资者羊群行为 的关联。在此基础上运用Mplus和Matlab软件，分析和验证不同外部决策情境下参考 群体驱动个体投资者羊群行为过程中，内部异质性（自信和特质焦虑）的跨层次调节 中介作用模型，延展了以往关于羊群行为的研究思路及相关研究工具的使用。然后， 为验证我国个体投资者并非只关注基本面而是倾向于跟随参考群体发生羊群行为，通 过某营业部账户日交易的每笔交易成交数据集来分析我国个体投资者羊群行为的决定 因素及其对股票价格的影响。市场指标测度与问卷调查方式测量个体投资者羊群行为 间高度正相关，且市场指标测度研究结果支持了问卷调查方式得出的结论。最后是全 文的总结和展望。主要研究结论如下： （1）基于参考群体的理论整合新视角，研究发现其他人（同龄人）、亲友和专家 群体导致的个体投资者羊群行为程度依次递增。参考群体是诱发羊群行为的重要外部 线索，个体投资者通过内化参考群体，与群体心理、态度和行为形成联结，发生羊群 行为。填补了微观层面上参考群体驱动个体投资者羊群行为内部机理这一研究空白。 （2）以内部特质和外部情境相整合为切入点，首次发现个体投资者自信和特质焦 虑在参考群体驱动羊群行为过程中具有跨层次调节中介作用，且他人评价对这一跨层 次调节中介模型产生影响。具体地，无他人评价效应时，个体投资者自信程度对参考 群体和羊群行为的关系起到正向跨层次调节作用，特质焦虑程度对参考群体和羊群行 为的关系具有负向跨层次调节作用，且特质性焦虑水平中介自信水平对参考群体类型 和羊群行为之间关系的跨层次调节作用。然而，他评效应缩小甚至模糊了不同自信和 特质焦虑水平个体的最终决策信心和投资意愿的差异，即他评效应导致自信和特质焦 虑的跨层次调节中介模型不成立。 （3）结合股票买入和卖出过程中股价预测一致与否这一未被理论界发掘的现实中 的决策情境，发现个体投资者买入和卖出羊群行为驱动机制既有相同之处亦存差异。 参考群体与个体投资者股价预测一致情境中，他评效应减弱或模糊不同参考群体类型 第II页 西南交通大学博士研究生学位论文 对个体投资者买入（卖出）羊群行为影响程度间的差异；他评效应导致自信和特质焦 虑不再影响参考群体类型与买入（卖出）羊群行为间的关系；他评效应导致参考群体 驱动买入（卖出）羊群行为过程中，自信和特质焦虑跨层次的调节中介效应不再显著。 参考群体与个体投资者股价预测不一致情境中，无他评效应时个体投资者自信和 特质焦虑对参考群体和买入（卖出）羊群行为间关系的跨层次调节中介作用不显著。 他评效应模糊（减弱）不同参考群体类型对个体投资者买入（卖出）羊群行为影响程 度间的差异。并且，他评效应亦缩小甚至模糊了不同自信和特质焦虑水平个体的最终 买入（卖出）决策信心和投资意愿的差异。此时，跨层次调节中介作用不显著。 参考群体驱动买入与卖出羊群行为过程的不同之处在于，个体投资者卖出股票时 更倾向于追随参考群体而羊群行为程度更强烈，且卖出股票时不同参考群体类型引致 羊群行为程度间的差异更大。 （4）采用真实交易记录通过压力期间羊群行为特征以及羊群行为后观察到的股价 动态这一确定羊群行为驱动源的更为实质和充分证据，验证了我国个体投资者确实倾 向于追随参考群体发生羊群行为。研究发现市场压力期间个体投资者羊群行为更明显， 且此期间集中买入和卖出股票的收益延续二到三周后出现逆转。由此，个体投资者羊 群行为会破坏证券市场稳定，该行为并非源于信息而是基于行为因素而表现为追随参 考群体，验证和支持了问卷调查的结论。本研究还发现，个体投资者在市场压力下跟 随参考群体卖出股票较买入时更普遍。此外，不同于台湾个体投资者，沪深市场个体 投资者偏好买入过去收益为正而卖出过去收益为负的股票。 关键词：个体投资者；羊群行为；自信；特质焦虑；参考群体；他评效应 西南交通大学博士研究生学位论文 第III页 Abstract Nowadays, as the Chinese stock market develops, more and more people are investing in stocks. The individual investor has become an important part of the stock market and constitutes a considerable proportion of the investor structure, and this situation will persist for a long time. This immature market development status means that there are a large number of immature individual investors with low risk tolerance in the stock market. The weakness of individual investors' familiarity with the stock market, information acquisition and processing ability, experience and other factors will inevitably deepen their irrationality in the decision-making process. When individuals make investment decisions, they are often influenced by the behavior of the reference group members such as professionals, relatives, friends, neighbors, or colleagues. The information (whether right or wrong) obtained from the reference group members is an important basis for forming individual investment decisions. It can be seen that the blind behavior of individual investors in China's securities market is very common. Moreover, individual investors' herding behavior is more threatening and destructive to market stability than institutional investors. It is a very urgent and important task to clarify the driving mechanism of individual investors' herding behavior. This dissertation first used a questionnaire survey to focus on how and when the reference groups drive the individual investors' herding behavior. The questionnaire survey method was used to construct two nested data layers to analyze the association between the reference groups and the individual investors' herding behavior through the cross-level regression model. The study found that individual investors tend to follow the reference groups. On this basis, this study used Mplus and Matlab software to analyze the multilevel mediated moderation effect of individual heterogeneity (self-esteem and trait anxiety) on the mechanism of reference group driving herding behavior in different external decision-making situations, which extends the previous research ideas of herding behavior and expands the use of relevant research tools. Then, this dissertation analyzed the determinants of individual investors' herding behavior and its impact on stock price return through real transaction records to verify whether the herding behavior is caused by behavioral factors and thus follows the reference groups. The herding behavior measured by market indicators is highly positively correlated with the herding behavior measured by the questionnaire survey, and the results of market indicators support the conclusions drawn from the questionnaire survey. The main findings are as follows. 第IV页 西南交通大学博士研究生学位论文 (1) Based on a new perspective of theoretical integration of the reference group, the research found that the herding behavior resulting from peers, friends and relatives, and experts presents a growth trend. The reference groups are the important external clue to induce herding behavior. Individual investors form a herding behavior by internalizing the reference groups and connecting with the cognitive psychological process and behavior of the group members. This research fills in the research gap that how the reference groups drive the individual investors' herding behavior at the micro level. (2) Taking the integration of internal traits and external contexts as the starting point, the dissertation found for the first time that the self-esteem and trait anxiety of individual investors have a multilevel mediated moderation effect in the process of reference group-driven herding behavior, and others‘ evaluations have an impact on this multilevel mediated moderation. Without others‘ evaluations, self-esteem positively moderates the relationship between reference group type and herding behavior, and trait anxiety negatively moderates the relationship between reference group type and herding behavior. Meanwhile, the multilevel moderated moderation is significant. When others‘ evaluations are present, self-esteem has an insignificant effect on the random slope of reference group type and