Communication and Comovement: Evidence from Online Stock Forums

• Author: Baozhong Yang,Jinyu Liu,Lei Jiang
• Date: 01 September 2019
• DOI: http://doi.org/10.1111/fima.12245
• Published date: 01 September 2019

Communication and Comovement:

Evidence from Online Stock Forums

Lei Jiang, Jinyu Liu, and Baozhong Yang∗

We study investor communication and stock comovement using a novel data set from an active

online stock forum in China. We find substantial comovement among the returns of a stock and

its “related stocks,” which arefrequently discussed in the subforum dedicated to the given stock.

Comovement is greater when the discussion of related stocks is more intensive. Further, the

effect of communication on comovement is strongerfor stocks associated with higher information

uncertainty. Codiscussed stocks aremore actively traded and experience more correlated trading.

A trading strategy that exploits communication-driven comovement generates abnormal returns.

Our findings highlight the impact of investor communication on asset comovement.

One fundamental question in financial economics is how asset prices are determined. In the

rational expectations paradigm, price variations reflect changes in fundamental values. However,

the empirical literature documents that there can be comovement in stock prices that is difficult

to explain by fundamental values.1Understanding the source and extent of comovement can shed

light on the structure of asset prices and facilitate the design of portfolio management strategies.

In this paper, we study whether communication among investors can generate comovement of

stock returns. It is usually difficult for researchers to observe and measure the communication

among investors about specific pairs of stocks. Using a novel data set from online stock forums

in China, we can clearly identify communication about stock pairs and therefore directlymeasure

We especially would like to express appreciation to Bing Han (Editor) and one anonymous referee for their insightful

and detailed comments that substantially improved the paper. Weare grateful for helpful comments from Vikas Agarwal,

Haiqiang Chen, LaurenCohen, Zhi Da, Will Goetzman, Allen Guoming Huang, Lixin Huang, Fuwei Jiang, Raymond Kan,

WenjinKang, Omesh Kini, Laura Xiaolei Liu, Ruichang Lu, Xingguo Luo, Kevin Mullally, Lin Peng, Michael R. Powers,

Zhen Shi, Rob Stambaugh, Ajay Subramanian, YuehuaTang, Paula Tkac, Jun Tu,Larry Wall, Baolian Wang,Hao Wang,

Neng Wang, Jinqiang Yang, Chunyang Zhou, Guofu Zhou, Xiaoneng Zhu, and participants at the Central University

of Finance and Economics, Federal Reserve Bank of Atlanta, Georgia State University, Peking University, the 2014

China Finance Review International Conference,the 2014 Chinese Finance PhD Symposium at Xiamen, the 2015 China

International Conference on Finance, the 2016 FMA Meetings, the Second Annual Volatility Institute at NYU Shanghai

(VINS) Conference,the 25th European Financial Management Association, and the Shanghai Risk Forum.We appreciate

the financial support from the 2015 CICF Yihong Xia Best Paper Award. Lei Jiang gratefully acknowledges financial

support from the AXA researchfund. This research is also supported by Tsinghua University Initiative Scientific Research

Program (20151080398), the National Science Foundation of China (71572091), the National Science Foundation of

China (71803018), and Tsinghua National Laboratory for InformationScience and Technology. The authors thank Yaru

Huang and Jie Liu for their excellent researchassistance.

[Correction added on 19 July 2019, after first online publication: additional funding information, National Science

Foundation of China (71803018), has been inserted in the author footnote.]

∗Lei Jiang is an Assistant Professor of Finance in the School of Economics and Management at Tsinghua University

in Beijing, China. Jinyu Liu is an Assistant Professor of Finance in the School of Banking and Finance at University

of International Business and Economics in Beijing, China. Baozhong Yang is an Associate Professor of Finance in

the Robinson College of Business at Georgia State University in Atlanta, GA. Send correspondence to Jinyu Liu at

liujy.12@sem.tsinghua.edu.cn.

1See, for example, Lee, Shleifer, and Thaler (1991), Pindyck and Rotemberg (1993), and Froot and Dabora (1999).

Financial Management •Fall 2019 •pages 805 – 847

806 Financial Management rFall 2019

investor communication. We document substantial comovement among the returns of stocks that

are discussed together by investors on online forums. Such comovement cannot be explained

by previously known causes, such as market and industry factors, common risk factors, investor

attention, or media coverage. Codiscussed stocks are more actively traded and exhibit more

correlated trading, suggesting that communication-driven trading is a potential mechanism for

observed return comovement.

To motivate our empirical analyses, we develop a simple Grossman and Stiglitz (1980)–type

model with two assets in which investorscommunicate before trading. The model shows that asset

returns can exhibit comovement beyond what is implied by fundamental values when investors

exhibit persuasion bias (DeMarzo, Vayanos, and Zwiebel, 2003) or fail to account for possible

repetition of the information they receive. In the model, investors receive a sequence of signals

when communicating with one another and update their beliefs before trading. Persuasion bias

allowsrepeated communication to have a continuing, amplifying effect on the correlation structure

of investors’ beliefs. As a result, the model predicts that communication among investors can

generate excess comovement in asset prices.

The model also predicts that comovement in asset returns is positively related to the com-

munication frequency among investors before trading. Intuitively, more frequent communication

leads to a greater dependence of investor beliefs on common signals, resulting in greater co-

movement. Further, the model predicts that the effect of communication on comovement is more

pronounced when investors possess less accurate beliefs—that is, for stocks associated with

greater information uncertainty.

We test these predictions using a unique data set from the East Money Stock Forum, one

of the most active online stock forums in China. The Chinese stock market provides an ideal

environment to study investor behavior. Established in the 1990s, the modern Chinese stock

market has developed rapidly but still witnesses irrational behaviors of individual investors (e.g.,

Wang,Shi, and Fan, 2006; Xu, 2000). While the importance of institutional investors has increased

over time, individual investors still dominate trading. In 2015, retail investors account for 86.9%

of the trading volume in China.2Indeed, the predominance of individual trades has been used

to explain the high market volatility in 2015.3In the Chinese stock market, individual investors

frequently exchange information and ideas on online forums.4While such communication can

help to propagate and incorporate information into stock prices, it can also potentially lead to

distortions, such as excess comovement, in the market, when investors are not fully rational.

For any given stock, there is a subforum on the online forum devoted to exchanging ideas

about this stock among investors. We refer to the stock that the subforum focuses on as its

target stock. Investors are also free to discuss other stocks in the subforum. Based on our model,

we expect the returns of the stocks discussed in the same subforum to exhibit comovement.

To test this hypothesis, for any target stock, we consider the most frequently discussed stocks

(henceforth referred to as “most related stocks”) on the target stock’s subforum. We construct

arelated portfolio that consists of the top five discussed stocks for every target stock and

rebalance the portfolio each month. We then estimate time series regressions of target stock

returns on the returns of their related portfolios to examine their correlations. We find that the

average correlation between the returns of a stock and its related portfolio is positive and highly

significant, even after controlling for known factors that drive comovement, including market,

2The trading volume data are availablefrom the 2015 Shanghai Stock Exchange Statistics Annual.

3For example, see “FT Explainer: WhyAre China’s Stock Markets So Volatile?”Financial Times, July 2, 2015.

4For example, an Internet survey shows that 65.9% of individuals are willing to share information and ideas on online

forums (Sixth Survey of Chinese Internet Community Development in 2010 by iResearch).

Jiang, Liu, & Yang rCommunication and Comovement 807

industry, and macroeconomic variables. This comovement is also economically significant; for

example, a 1% increase in the related portfolio return is associated with a 0.11% increase in the

target stock return.

To address the concern that the correlation may be spuriously generated by a temporal trend

or comovement among industries, we conduct a falsification test. We first create for each target

stock a placebo portfolio that consists of placebo stocks in the same industries of the related

stocks. We then estimate the same regressions, replacing the returns of related portfolios with

those of the placebo portfolios. We find the coefficients on the returns of the placebo portfolios

to be insignificant, suggesting that the comovement we document is unlikely to be caused by

temporal or industry factors.

We next examine the prediction on the relation between the frequency of communication and

stock comovement. Foreach target stock, we proxy for communication intensity by the number of

messages about its top five most related stocks in its subforum. Wethen include this frequency and

its interaction with the related stock portfolio return as independent variables in the regressions

of the target stock returns. We find that, as the model predicts, more frequent communication

leads to higher comovement between the returns of the target stock and its related stocks.

We then investigate the prediction that the effect of communication on return comovement is

larger for stocks associated with greater information uncertainty. We use five proxy variables

for the information uncertainty of stocks: stock illiquidity, market capitalization, analyst forecast

dispersion, systematic volatility, and return volatility. We stratify our sample of stocks into five

quintile groups based on each of the information uncertainty proxies and conduct our regressions

separately for each group. Consistent with our model’s prediction, we find that for more illiquid,

smaller, volatile, and dispersedly forecasted stocks, the frequency of forum discussion has a

greater effect on stock comovement.

We address potential endogeneityconcer ns in two ways. First, we employ an exogenous varia-

tion in the extent of investorcommunication caused by an outage in the East Money Forum in June

2010 (henceforth, the “outage month”). Wef ind the number of posts during the outage month to be

substantially lower than that in adjacent months. We conduct our tests of comovement separately

for the outage month and for the adjacent months and we find that the comovement between target

and related stocks in the outage month is substantially lower than that in the adjacent months. We

also conduct placebo tests using the years 2009 and 2011 and find that the comovement between

target and related stocks in June of those years is statistically indistinguishable from that in the

adjacent months.

Second,we use the introduction of the mobile application (App) for the forum in November 2012

as an exogenous shock to online communication. Tothe extent that mobile phones and tablets have

vastly expanded online interactions, the introduction of the mobile App facilitates access to the

forum. This event is also exogenous to the stock marketas it is driven by technology advancement.

We find that comovement among related stocks and targetstocks increases signif icantly after the

launch of the mobile App. In contrast, placebo tests using pseudo-events in adjacent years do not

reveal significant changes in comovement.

A plausible mechanism through which communication affects comovement is correlated trad-

ing. Upon reading online messages, investors may generate trade ideas that they then execute.

Their trades can thus impound their beliefs into asset prices and influence comovement conse-

quently. Due to persuasion bias, investors do not fully incorporate the consequences of repeated

communication and thus their beliefs, trading, and stock returns should exhibit excess comove-

ment. Consistent with this channel, we find trading volumes of target and related stocks are

indeed higher when they are more intensivelydiscussed in the for ums. Further,there is a positive

correlation between the trading volumes of target and related stocks, which also increases with