本章导读 本章梳理一些对解释金融市场异象可能有用的心理学概念。§10.1 认知心理学：10.1.1 偏好心理（前景理论 §2、模糊厌恶 Ellsberg 1961、时间不一致 Beta-Delta §3）；10.1.2 信念心理（锚定/代表性/可得性 Tversky & Kahneman 1974；过度自信——过度精确/过度定位/过度估计；小数定律 Rabin 2002 与赌徒谬误/热手谬误；信念固执/确认偏差；外推 Greenwood & Shleifer 2014 投资者预期与过去收益正相关、与模型隐含预期负相关，图 10.1；凯恩斯选美竞赛高阶思维 + Thaler 猜数字游戏图 10.2）；10.1.3 情绪影响偏好与信念（情感性 Johnson & Tversky 1983、认知性=熟悉度）。§10.2 社会心理学（行为 = 人格 + 环境，环境或更重要）：认知失调、从众（Asch 1951）、群体思维 Janis 1972、服从。图 10.1–10.2 已转述。

信念心理的概念 / Concepts in the psychology of beliefs 锚定、代表性与可得性偏差（或记忆偏差）：Tversky & Kahneman (1974) 提出的三大基本启发式与偏差，见 §1.2.4。过度自信 (Overconfidence)：其伞下有多个概念——过度精确 (over-precision)：人们总是对自己信念的准确性过于自信；过度定位 (over-placement)：人们倾向认为挫折更可能发生在别人而非自己身上；过度估计 (over-estimation)：人们对自身能力或成功机会总是过于乐观。小数定律 (Law of small numbers)：Rabin (2002) 提出该术语，指人们夸大小样本对总体的代表程度；这与 §6.3 讨论的赌徒谬误（脚注：赌徒谬误的例子——公平抛硬币中，主体在看到一连串反面后赋予正面更高的主观概率）(Tversky & Kahneman 1974) 与热手谬误 (Gilovich et al. 1985) 相关，因为两种情形下主体都相信小样本频率应反映总体均值。信念固执 / 确认偏差：人们一旦形成观点就总不愿改变，即便收到新信息、尤其是与其先验相悖的新信息。Anchoring, representativeness and availability bias (or memory bias): the three fundamental heuristics and biases proposed by Tversky & Kahneman (1974), see §1.2.4. Overconfidence: there are many concepts under its umbrella — over-precision: people are always too confident about the accuracy of their beliefs; over-placement: people tend to think setbacks are more likely to happen to others than to themselves; over-estimation: people are always too rosy (optimistic) about their own abilities or chances of success. Law of small numbers: Rabin (2002) proposed this term, saying people exaggerate the degree to which a small sample represents the population; this is related to the gambler's fallacy (footnote: an example for the gambler's fallacy — in a fair coin toss, an agent whose subjective probability of a head after seeing a series of consecutive tails has a higher subjective probability of a head) (Tversky & Kahneman 1974) and the hot-hand fallacy (Gilovich et al. 1985) discussed in §6.3, because in both cases agents believe small-sample frequencies should reflect the population mean. Belief perseverance / confirmation bias: once people have formed an opinion, they are always reluctant to change it even after receiving new information, especially new information against their prior beliefs.

外推与凯恩斯选美竞赛 / Extrapolation and Keynes' beauty contest 外推 (Extrapolation)：人们倾向基于对当前或过去数据的观察来做未来预测。Greenwood & Shleifer (2014)（脚注：该文在 He 2020a 第 27.1 节详述）用 1963–2011 年六个数据源关于投资者对未来股市收益的预期，发现投资者预期与过去股票收益及股市水平正相关、却与模型隐含的预期收益负相关（图 10.1）。这是反对"理性预期代表性投资者模型"的证据，体现投资者的外推行为。凯恩斯"选美竞赛"（higher order thinking）：重要的不是某物真实如何，而是别人认为它如何。凯恩斯以一个报纸竞赛作比喻：参与者从一百张照片中选出六张最有吸引力的脸，选中最受欢迎面孔者有奖；最重要的不是凭自己最佳判断挑最美的脸，而是三阶思维——找出"平均意见所预期的平均意见"。另一个例子：Thaler 在《金融时报》玩"猜数字"游戏，读者猜 \(0\) 到 \(100\) 的数，猜得最接近全体平均的 \(\tfrac23\) 者得两张伦敦到美国的商务舱机票；理性均衡本应是人人猜零，然而现实并非如此（图 10.2）。Extrapolation: people tend to make future projections based on their observation of current or past data. Greenwood & Shleifer (2014) (footnote: this paper is discussed in detail in §27.1 in He 2020a) use six data sources between 1963 and 2011 on investor expectations of future stock-market returns, and find investor expectations are positively correlated with past stock returns and the level of the stock market but negatively correlated with the model-implied expected returns (Figure 10.1). This is evidence against the rational-expectations representative-investor models and shows the extrapolation behaviors of investors. Keynes' "beauty contest" (higher-order thinking): the important thing is not what something really is, but what others think it is. Keynes describes an analogy about a newspaper contest: participants choose the six most attractive faces from a hundred photographs, and those who pick the most popular faces are eligible for a prize; the most important thing is not to pick the prettiest face to one's own best knowledge, but third-order thinking — to find what the average opinion expects the average opinion to be. Another example: Thaler played the "guess the numbers" game in the Financial Times, where readers guess a number between \(0\) and \(100\) and those whose guess is closest to \(\tfrac23\) of the average get two business-class tickets from London to the U.S.; the rational equilibrium should be everyone guessing zero, however this didn't happen (Figure 10.2).

情感性与认知性情绪 / Affective and cognitive feelings 情感性情绪 (Affective feelings)：由外生刺激带来的更愉快心情，会导致对无关的未来结果作出更正面的判断。Johnson & Tversky (1983) 用简短的报纸报道操纵被试的情绪 (affect)，结果表明：一则悲剧性报道使被试对无关风险与不良事件的频率估计大幅上升；相反，一则愉快事件的报道则降低其对风险频率的判断。认知性情绪 (Cognitive feelings)：人们倾向对那些更易于获取与处理（即更熟悉）的事物作出更正面的判断。Affective feelings: a happier mood due to an exogenous stimulus leads to more positive judgments about unrelated future outcomes. Johnson & Tversky (1983) conduct experiments in which the subject's affect (emotion) is manipulated by a brief newspaper report, and the results show that a tragic report leads to a substantial increase in the subject's estimates of the frequency of unrelated risks and undesirable events; on the contrary, a report of a happy event reduces their judgment of the frequency of risks. Cognitive feelings: people tend to have a more positive judgment on the things that are easier for them to access and process, i.e. more familiar things.

图 10.1、10.2（已转述 / Figures 10.1, 10.2, paraphrased） 图 10.1（Gallup %看多−%看空 与过去收益）：横轴为时间（约 1996–2010），虚线为 CRSP 价值加权股指的 12 个月滚动名义收益、实线为 Gallup 调查的"%看多−%看空"投资者预期；两线同向波动——投资者预期随过去收益与市场水平上升而上升（外推）。图 10.2（猜数字报纸游戏）：《金融时报》实验 1,468 名被试的猜数分布直方图，横轴为 \(0\)–\(100\) 的猜测值、纵轴相对频率；分布散布且在 \(22\)、\(33\) 等处有小峰，平均 18.91、获胜数字（平均的 \(\tfrac23\)）为 13——远非理性均衡的 0。Figure 10.1 (Gallup %Bullish−%Bearish and past returns): the horizontal axis is time (about 1996–2010), the dashed line the 12-month rolling nominal return on the CRSP Value-Weighted stock index, the solid line the "%Bullish−%Bearish" investor expectation from Gallup surveys; the two lines move together — investor expectations rise with past returns and the market level (extrapolation). Figure 10.2 (guess-the-numbers newspaper game): a histogram of the guesses of 1,468 subjects in the Financial Times experiment, the horizontal axis the guess value \(0\)–\(100\), the vertical axis relative frequency; the distribution is dispersed with small peaks at \(22\), \(33\) etc., with an average of 18.91 and a winning number (\(\tfrac23\) of the average) of 13 — far from the rational equilibrium of 0.

社会心理学概念 / Concepts in social psychology 认知失调 (Cognitive dissonance)：当人们采取与其（通常正面的）自我形象相冲突的行动时会感到不适，于是可能操纵自己的信念以避免这种不适。从众 (Conformity)：人们倾向顺从群体；Asch (1951) 的从众实验见 §4.6.1。群体思维 (Groupthink)：Janis (1972) 讨论该术语，指群体共识与凝聚力比事实本身更重要、从而压制异议与对替代方案的评估的思维方式；当下列情形时效应更强：群体被珍视且有吸引力、群体与外部意见隔绝、有强势领导者表明其观点、情境紧张、没有考虑其他意见的流程。服从 (Obedience)：人们倾向听从权威人物的命令或指示，即便任务看起来不当；即便日后永不再见该权威，仍不想让其失望。Cognitive dissonance: people feel discomfort when they take an action that conflicts with their (typically positive) self-image, so they might manipulate their beliefs to avoid the discomfort. Conformity: people tend to conform to the group; see §4.6.1 for the Asch (1951) conformity experiment. Groupthink: Janis (1972) discusses this term, referring to a thinking in which group consensus and cohesiveness are more important than the facts themselves, so that dissent and appraisal of alternatives are suppressed; the effect is stronger when: the group is valued and attractive, the group is isolated from other opinions, there is a strong leader who makes his view known, the situation is stressful, and there is no process for considering other opinions. Obedience: people tend to follow the orders or instructions from an authority figure even when the task seems inappropriate; even if they will never see the authority again in the future, they still don't want to disappoint him/her.