Individual investors' dispersion in beliefs and stock returns

• Published date: 01 September 2022
• Author: Junjun Ma,Xindan Li,Lei Lu,Weixing Wu,Xiong Xiong
• Date: 01 September 2022
• DOI: http://doi.org/10.1111/fima.12389

DOI: 10.1111/fima.12389

ORIGINAL ARTICLE

Individual investors’ dispersion in beliefs and stock

returns

Junjun Ma1Xindan Li2Lei Lu3Weixing Wu4Xiong Xiong5

1School of Economics and Management, Beijing University of Technology,Beijing, China

2School of Management and Engineering, Nanjing University,Nanjing, China

3Asper School of Business, University of Manitoba, Winnipeg, Canada

4School of Finance, University of International Business and Economics, Beijing, China

5College of Management and Economics, Tianjin University, Tianjin, China (Email:xxpeter@tju.edu.cn)

Correspondence

LeiLu, Asper School of Business, University of

Manitoba,Winnipeg, Canada.

Email:lulei2000@gmail.com

XiongXiong, College of Management and

Economics,Tianjin University, 92 Weijin Road,

Tianjin,300072, China,

Email:xxpeter@tju.edu.cn

Fundinginformation

NationalNatural Science Foundation of China,

Grant/AwardNumbers: 71733004, 72101009,

72141304,71720107001, U1811462; Social

Scienceand Humanity Research Council

InsightGrants, Grant/Award Number: 435-

2021-0041

Abstract

We construct a measure of dispersion in beliefs among indi-

vidual investors. We find that dispersion in beliefs nega-

tively predicts future cross-sectional stock returns, and it

is positively related to trading volume and stock volatility.

We also find that illiquidity does not affect the significance

of dispersion in beliefs in predicting future stock return,

and that the negative disagreement-return relation is signifi-

cant under high-sentiment periods but becomes insignificant

under low-sentiment periods. Moreover,investor character-

istics affect their dispersion in beliefs even when controlling

firm fundamentals. In particular, stocks with more wealthy,

younger, and male investors tend to havehigher dispersion

in beliefs, and stocks with more experienced investors have

lower dispersion in beliefs.

1INTRODUCTION

Investors’ beliefs are centralfor asset pricing. It is well known that investors have different beliefs about firms’ funda-

mentals, and this heterogeneity affects stock price and its dynamics (e.g., Carlin et al., 2014; Cen et al., 2017; Huang

et al., 2021; Miller, 1977; Scheinkman & Xiong, 2003). The existing literature usually uses the forecasts of financial

© 2022 Financial Management Association International.

Financial Management. 2022;51:929–953. wileyonlinelibrary.com/journal/fima 929

930 MA ET AL.

analysts as the proxy for investors’ opinions about firms’ fundamentals (e.g., Diether et al., 2002;Yu,2011). How-

ever,individual investors may have limited information or knowledge compared with institutional investors, and thus

analysts are not representativeof all investors, especially in the markets dominated by individual investors (i.e., China).

In addition, it has been documented that due to analyst’s conflicts of interest with firms or analysts’ herding behavior,

the information conveyedby financial analysts is subject to bias (e.g., Hirshleifer et al., 2021; Malmendier & Shanthiku-

mar,2014; O’Brien et al., 2005). In this paper, we construct a new measure of investors’ differences in beliefs using the

trading data of individual investors. Weexamine its effects on stock return, stock volatility, and trading volume when

market has frictions (i.e., illiquidity,sentiment, gambling, and short-sale constraints) or investors are not rational (e.g.,

gambling), and explore whether investors’ characteristics(i.e., age, education, gender, wealth, and investment experi-

ence) influence their difference in beliefs.

In recent years, with the development of financial innovations and technologies, it is possible for researchers to

measure investors’ beliefs using large sets of data (e.g., Goldstein, 2021). We use account-level daily trading data to

construct a measure of disagreement. There are at least two advantages in using daily account data. First, it avoidsthe

deviation between investors’ beliefs and actions. Giglio et al. (2021) argue that surveydata are often based on unrep-

resentative samples and may not reveal those beliefs on which agents actually base their actions. Also, respondents

and investing platform users may not be as serious as they would if they actually had to takeaction when expressing

their beliefs. Second, we can calculate investors’ disagreement at a weekly frequency.The literature usually estimates

investors’ disagreement monthly (e.g., dispersion in analyst earnings forecasts) or quarterly (e.g., dispersion based on

mutual funds) and then could not examine the effects of dispersion in beliefs on short-term stock returns.1

Our measure of dispersion in beliefs is strongly and positively related to turnoverand is not significantly correlated

with dispersion in analysts’ earnings forecasts. This result is consistent with the finding of Li and Li (2021), which sug-

gests that financial analysts represent a distinct group of economic agents. Using our disagreement measure, we test

its relation with future stock return and find that higher dispersion in beliefs among individual investorspredicts lower

stock returns, which is in line with the overvaluedhypothesis (e.g., Miller, 1977). Our results are qualitatively the same

when we use investors’stock holdings (not trading) to construct the disagreement measure or control for fixed effects.

Next, we examine how imperfections of financial markets or investors’ behaviors affect the disagreement-

return relation. First, this relation is not significantly affected by liquidity, supporting the argument of

Golez and Goyenko (2021). This finding partially reconciles the concern of Cookson and Niessner (2020)

that it is difficult to separate the effect of investors’ differences in opinion on asset prices from that

of liquidity even using individual investors’ trading data. Second, we test the effect of sentiment on the

disagreement-return relation and show that stocks with high disagreement are prone to overpricing dur-

ing high-sentiment periods. Third, this negative relation is significant for lottery-type stocks but weakly

significant or insignificant for nonlottery-type stocks. We also examinethe relation between disagreement and future

stock trading volume and stock volatility.Our results indicate that dispersion in beliefs among individual investors is

significantly and positivelyrelated to future trading volume (e.g., Ajinkya et al., 1991; Cookson & Niessner, 2020; Hong

& Stein, 2007) and stock volatility (e.g., Anderson et al., 2005; Banerjee, 2011; Hibbert et al., 2020).

The existing literature analyzes the relation between dispersion in beliefs and firms’ characteristics (e.g., Diether

et al., 2002) and analysts’ strategy behavior (e.g.,Liu & Natarajan, 2012). However, investors’ characteristics also mat-

ter for their beliefs and investment decisions (e.g., Barber & Odean, 2001; Giglio et al., 2021; Goetzmann & Massa,

2005). Our results show that dispersion in beliefs among individual investorsis significantly related to their character-

istics when controlling for firms’ fundamentals. In particular, stocks with more wealthy,younger, and male investors

tend to have higher dispersion in beliefs, and stocks with more experiencedinvestors have lower dispersion in beliefs.

Our paper contributes to three strands of literature. First, it is relevant for the literature on the measure of dis-

persion in beliefs. The existing literature uses various proxiesto measure dispersion in beliefs, including dispersion in

analysts’ earnings forecasts (e.g., Diether et al., 2002;Park,2005;Yu,2011),trading volume (e.g., Garfinkel & Sokobin,

1Oneexception is the paper of Cookson and Niessner (2020), which calculates daily disagreement.