The rate of communication

https://doi.org/10.1016/j.jfineco.2021.03.013Get rights and content

Abstract

We study the transmission of financial news and opinions through social interactions among retail investors in the United States. We identify a series of plausibly exogenous shocks, which cause “treated investors” to trade abnormally. We then trace the “contagion” of abnormal trading activity from the treated investors to their neighbors and their neighbors’ neighbors. Coupled with methodology drawn from epidemiology, our setting allows us to estimate the rate of communication and how it varies with the characteristics of the underlying investor population.

Introduction

People do not operate in a social vacuum; instead, they constantly interact with one another. Social interactions ensure that we are privy to the latest news, ideas, and opinions. They also expose us to the spread of noise and disinformation. While the theme of contagion and diffusion has been examined by many disciplines (e.g., Berger, 2014, 2016; Jackson, 2014, 2019), there is likely no field that has looked at this subject more extensively than the field of epidemiology (e.g., Keeling and Grenfell, 2000; Heesterbeek, 2002; Heffernan et al., 2005; Delamater et al., 2019). The goal of this study is to draw from the epidemiology literature and to estimate the “effective transmission rate” of financial news and opinion and how it varies with investor characteristics.

The ideal experiment with which to address our research question would be to randomly seed pieces of information among investors and then track their diffusion through the investors’ respective networks. Our empirical design draws inspiration from this ideal. In particular, we consider a series of cross-industry stock-financed mergers and acquisitions (M&As). At the completion of each cross-industry stock-financed M&A, investors in the target firm, residing in some industry x, receive shares of the acquirer firm, residing in some industry y. We conjecture that the endowment of shares from the acquirer industry leads at least some of the affected investors to form opinions about the acquirer industry and to start trading firms in the acquirer industry (aside from the acquirer firm itself). If such “target investors” communicate their newly gained industry perspectives to other investors in their neighborhood, we may observe abnormal trading activity in the acquirer industry not only by target investors but also by their neighbors and their neighbors’ neighbors. Tracing the contagion of abnormal trading activity in the acquirer industry thus enables us to estimate the degree to which financial information spreads through social interactions and the extent to which the “effective transmission rate” varies with the characteristics of the sender of the financial information and the receivers of it.

To implement our empirical tests, we combine detailed trading records of about 70,000 households in the United States from a discount brokerage for 1991 to 1996 with data on all cross-industry stock-financed M&As that occurred over the same time period.

To gauge the validity of our empirical design, we first conduct a simple difference-in-differences analysis to see how much more intensely target investors trade in the acquirer industry in the post-M&A period (excluding trading activity in the acquirer firm itself). We repeat the above difference-in-differences analysis for “target neighbors”; target neighbors are non-target retail investors who reside within three miles of a target investor.

Our results reveal that in the year following the completion of a cross-industry stock-financed M&A, target investors more than double the number of trades they execute in the corresponding acquirer industry as compared to other investors. This abnormal trading activity in the acquirer industry dies out within two years.

Consistent with the presence of contagion, we find that target neighbors also trade substantially more actively in the acquirer industry compared to investors who do not live within three miles of a target investor. Target investors and target neighbors tend to trade in the same direction; that is, if a target investor is buying in the acquirer industry, so are the investor's neighbors. Consistent with “word of mouth” playing a role in generating our results, our effect becomes statistically and economically weaker the further away an investor resides from a target investor.

In a placebo test to help rule out alternative interpretations, we consider cross-industry M&As that are 100% cash-financed. In cash-financed M&As, target investors receive cash as opposed to shares in the acquirer firm and, as such, are less incentivized to study the corresponding acquirer industry. We find that our effect disappears when we consider cash-financed M&As. Moreover, inconsistent with a simple local attention story, we observe little abnormal trading activity when a stock-financed M&A is first announced. Instead, abnormal trading activity accrues only after target investors receive shares of the acquirer firm.

Our main analysis builds on the above findings and utilizes methodology drawn from the epidemiology literature to estimate an analog of the reproduction number; the reproduction number is the average number of new infections generated by a single infective. We hereafter refer to this analog as the rate of communication, or, simply, the communication rate. We also estimate how much the communication rate varies with the characteristics of the underlying investor population, including age, income, gender, past investment performances, measures of lifestyle, and state of residence.

Our estimate of the overall communication rate is 0.32 with a 95% confidence interval of 0.17 to 0.46. In other words, one “infected” investor, on average, “infects” 0.32 neighbors.1 As we discuss in Section 2.1, an outbreak will fade if the reproduction number falls below one; a disease will continue to spread and grow if the number is above one. Our communication rate of 0.32 thus suggests that while the transmission of financial information through social interactions is significant, it eventually dissipates without intervention, at least in our setting. To put this number in perspective, Cao et al. (2020) estimate that the effective reproduction number of COVID-19 in China during the onset of its outbreak (December 2019–January 2020) was 4.08 with a 95% confidence interval of 3.37 to 4.77.

A key difference between the transmission of a pathogen and the transmission of an idea is that the latter occurs voluntarily. That is, for an idea to transmit, a mere interaction between two individuals is not sufficient. The sender of the information must be motivated to share the idea. In addition, the receiver must be willing to listen and consider the idea interesting and credible enough to absorb and act on the idea. This line of thinking forms the basis for our analysis of how much the communication rate varies with characteristics of the underlying investor population.

Our first set of determinants is motived by studies of homophily. These studies note that people prefer to interact with people of similar backgrounds. They are also more likely to trust information received from these individuals (Lazarsfeld and Merton, 1954; McPherson et al., 2001). Transmissions are thus substantially stronger between people with similar backgrounds (Jackson, 2019).

Consistent with this perspective, we find that the transmission of an investment idea is strongest when there are few differences in age, income, or gender between the sender of financial information and the receiver. While our estimate of the overall communication rate is 0.32, we find that the communication rate between investors of the same age, the same income category, and the same gender rises to 0.47 with a 95% confidence interval of 0.29 to 0.65. When comparing the relative importance of differences in age, income, and gender in slowing down transmission, we find that a ten-year age gap, a one-category difference in income, and being of an alternate gender lowers the communication rate by 9%, 3%, and 12%, respectively. That is, in the investment context, differences in age and gender represent higher barriers to information transmission than differences in income.2

A key strength of our setting is that we can pinpoint both the sender of financial information and the receivers. We use this feature to uncover notable asymmetries. In particular, our results suggest that while transmission is strongest among investors of similar age, gender, and income, relatively speaking, transmission from older, high-income, female investors to younger, low-income, male investors is stronger than transmission in the reverse direction. One possible explanation for these asymmetries is that investors perceive information conveyed by older, wealthier, female investors as more credible and, thus, are more likely to act on any views transmitted by such investors.

Our second set of determinants relate to investors’ past investment performances. The psychology literature finds that people are more likely to share a story and others are more likely to listen if the story helps receivers re-access positive emotional experiences. That is, people are more likely to converse about a story if it invokes pleasant memories (Lovett et al., 2013; Berger, 2014; 2016). Consistent with this view, Kaustia and Knüpfer (2012), Heimer and Simon (2015), and Escobar and Pedraza (2019) find evidence that investors more frequently share stories of investment success than stories of investment failure. Han et al., 2021 model the implications of agents’ preferences for sharing successes over failures.

In our setting, we conjecture that the communication rate is a function not only of the sender's past investment performance but also of that of the receiver. If receivers have suffered recent investment failures, they are unlikely to entertain a conversation about investment-related topics and, consequently, act on any ideas so transmitted.

In line with this view, we find that the communication rate is the highest, 0.44, when both the sender's and the receiver's recent portfolio performances are above the sample median. If the sender's recent portfolio performance is above the median, yet the receiver's performance is below the median, the communication rate drops by 16% to 0.37. The communication rate is the lowest, 0.29, when both the sender's and the receiver's portfolio performances are below the sample median. Comparing these figures with those based on differences in investors’ socioeconomic backgrounds, we can infer that recent investment performances are a stronger determinant of the rate of communication than differences in socioeconomic backgrounds.

Our third and final set of determinants captures similarities (or differences) in lifestyle and state of residence. In short, we find that the communication rate is highest when the sender and the receiver lead a similar lifestyle as approximated through common ownership of unique vehicles (truck, recreational vehicle (RV), motorcycle). Moreover, the communication rate is highest in states for which survey evidence indicates that people spend more time visiting friends (Putnam, 2000).

The rest of the paper is organized as follows. Section 2 describes and situates our paper in the corresponding literatures. Section 3 describes our data. Section 4 introduces our empirical setting and presents evidence that people turn to each other for financial advice and investment ideas. Section 5 estimates the rate of communication and how it varies with the characteristics of the underlying investor population. Section 6 examines whether, in our setting, investors transmit value-relevant information or merely spread noise. Section 7 concludes.

Section snippets

Literature review and contribution

Our paper builds on two streams of research: medical studies of how frequently various pathogens are transmitted and finance studies providing evidence of word-of-mouth effects in financial markets.

Data sources and descriptive statistics

We obtain detailed investor trading records for a subsample of US households for the 1991–1996 period from a discount brokerage firm. These are the same records used by Odean (1998) and Barber and Odean (2001), among others.3 The

Our empirical setting

A major challenge facing empirical, non-experimental research on diffusion and contagion is the presence of common shocks that affect everyone. To illustrate by example, prior studies generally infer the transmission of financial information through positive correlations in trading patterns between investors residing in the same locale. Yet, if two investors in the same locale exhibit correlated trading patterns, how can we be certain that they actually communicate with one another rather than

The rate of communication

To estimate the rate of communication for our particular type of financial information and to study how much the rate varies with the characteristics of the underlying investor population, we propose a “dynamic” estimation procedure drawn from studies that examine the contagion rate of diseases (Kermack and McKendrick, 1927, 1932).

Fig. 2 contrasts our previous “simple” difference-in-differences analysis in Section 4 with our new “dynamic” estimation procedure. Each dot represents an investor in

Dissemination of value-relevant information or merely spreading noise?

Do investors in our setting transmit unique and value-relevant news or simply spread noise? If any newly acquired views about the acquirer industry transmitted through social interactions represent unique and value-relevant information, the stocks bought by target investors and their neighbors in the acquirer industry (“long leg”) should subsequently outperform the stocks sold by target investors and their neighbors in the acquirer industry (“short leg”). On the other hand, if the views about

Conclusion

In this paper, we identify cross-industry stock-financed M&As as a series of plausibly exogenous shocks, which cause target investors to trade abnormally in the acquirer industry. We then trace the contagion of abnormal trading activity from the target investors to their neighbors and their neighbors’ neighbors. We use our setting to provide causal evidence that financial information is contagious. We also produce novel estimates of how contagious financial information is and how much such

References (43)

  • A. Banerjee et al.

    Using gossips to spread information: theory and evidence from two randomized controlled trials

    Rev. Econ. Stud.

    (2019)
  • B. Barber et al.

    Boys will be boys: gender, overconfidence, and common stock investment

    Q. J. Econ.

    (2001)
  • J. Berger

    Contagious: Why Things Catch On

    (2016)
  • J. Brown et al.

    Neighbors matter: causal community effects and stock market participation

    J. Finance

    (2008)
  • Z. Cao et al.

    Estimating the effective reproduction number of the 2019-nCoV in China

    MedRxiv

    (2020)
  • H. Chen et al.

    Listening in on Investors’ Thoughts and Conversations

    (2020)
  • P. Delamater et al.

    Complexity of the basic reproduction number (R0)

    Emerg. Infect. Dis.

    (2019)
  • M. Doucleff

    No, Seriously, How Contagious Is Ebola. Public Health, National Public Radio

    (2014)
  • E. Duflo et al.

    The role of information and social interactions in retirement plan decisions: evidence from a randomized experiment

    Q. J. Econ.

    (2003)
  • L. Escobar et al.

    Active Trading and (poor) Performance: the Social Transmission Channel

    (2019)
  • S. Flaxman et al.

    Report 13 - Estimating the Number of Infections and the Impact of Non-Pharmaceutical Interventions On COVID-19 in 11 European countries

    (2020)
  • Cited by (6)

    View full text