Investors’ information sharing and use in virtual communities

Authors


Abstract

Research demonstrates that information disseminated and circulated in online forums may have a significant impact on investors and on the securities market, so an understanding of that environment is critical. This article reports on an analysis of information sharing and use in three investment discussion forums. Threads containing 1,787 posts were coded using previously developed typologies for Internet-based discussion. Citations were studied in their context and sources were categorized into types. A high degree of collaborative information behavior was identified, but the study also reveals some areas of information use that may compromise investors’ decision making, including heavy reliance on personal sources of information and other sources that vary greatly in trustworthiness, including commercially sponsored information, blogs, and investor guru sites. These challenges are discussed and recommendations are made for improving services to investors. Questions for additional research are also identified.

Introduction

Investors throughout the world, including an estimated 80% of investors in China and India, are now trading online (FinWeek, 2011). Low transaction costs, easy access, and aggressive marketing by online brokerage companies are attracting investors in unprecedented numbers. Online brokers also recently have launched a dizzying array of services available through mobile technologies, including applications for the iPhone and BlackBerry, guaranteed to make online trading more accessible, convenient and attractive than ever.

The impact online trading has on investors is complex, and there is reason for concern. According to Barber and Odean (2002), investors earn less when they move to the online environment. They write, “Those who switch to online trading perform well prior to going online, beating the market by more than 2% annually. After going online, they trade more actively, more speculatively, and less profitably than before—lagging the market by more than 3% annually” (Barber & Odean, 2000, p. 455). Frith (2011) agrees that “the speed and volatility of such instant trading has made it riskier for small investors” (p. 52).

How investors find and use financial information is also transformed in the online environment (Barber & Odean, 2001). Online investors, who are more likely to be new to investing, avoid interacting with brokers (Barber & Odean, 2001). Forgoing the counsel of professional advisors places the burden for finding, evaluating and using information squarely on the shoulders of investors. Online investors are also more likely to restrict their information search to online sources (Williamson, 2008). Williamson and Smith (2010) conclude that online investors need help “dealing with information overload, learning to balance the need for speedy delivery of information with making considered investment decisions, undertaking systematic analysis using information, [and] using advice from interpersonal sources of information judiciously” (p. 72).

Although a burgeoning body of research on investors’ information behavior exists to provide such an understanding, less is known about it in virtual environments. Understanding information behavior in online discussion environments is important not only because of its impact on the success or failure of individual investors, but also because collectively their results affect the entire market (Barber & Odean, 2001). In fact, research demonstrates that information disseminated in online forums may have a direct and significant impact on stock price movement (Antweiller & Frank, 2004; Regnier, 1999). Furthermore, information or misinformation can be introduced and circulated in chat forums for the express purpose of manipulating stock prices, so it is critically important online investors know how to evaluate the quality of information they find there (Langevoort, 2002). This article reports on a study of investors’ information sharing and use in virtual discussion forums conducted by analyzing the content and context of citations to formal information sources. It describes the types of sources used, analyzes the collaborative information behavior exhibited, and demonstrates how Internet discussion groups function as investment information channels.

Literature Review

Investor Information Behavior

Much of what is known about investor information behavior is based on individual investors and often focuses exclusively on information use. This research nonetheless provides a helpful foundation for this study. Reporting on their study utilizing a telephone survey of 911 randomly sampled U.S. investors, Hira and Loibl (2006, 2009) demonstrate the types and constellations of sources investors use. They found that the investors in their study (65% of whom were male and 35% of whom were female) fell into five clusters of information use: high, online, moderate, workplace, and low. Investors in the high and online categories were self-reliant, information driven and practice multisource, diversified, high-information use strategies. Unfortunately, these investors constituted only 22% of investors studied. They were also highly likely to be male, well-educated, and possess the largest financial assets. The moderate group (24%) practiced a broad but less frequent information use strategy and was also highly likely to be male and well educated. The remaining 54% of investors, who constitute the low and work-place oriented information users, were primarily female with the lowest educational levels and financial assets. Their strategies were characterized by the use of few, if any, formal, mediated information sources and the nonadoption of technology for investment information.

Older studies, including McKay et al. (1996), Peter D. Hart Research Associates (1997), and Mezick (2001), suggested that investors prefer printed sources of information, such as newspapers, magazines, and annual reports. For example, Mezick found 77% of investors cited magazine and newspapers as the most frequently used sources of information for investors. Friends and relatives (69%), web pages (54%) and search engines (36%) followed. Financial advisors, television and stockbrokers were each cited by less than 6% of those surveyed as primary information sources. Hira & Loibl's (2006) findings were quite different: investors in their study cited financial advisors (28%), magazines and newspapers (22%), the Internet (21%), and the workplace (10%) as the most often used sources of financial information. Friends, television, radio, and classes all followed with less than a 5% response for each.

Even as early as 2001, however, Mezick found that the growing popularity of the Internet was having an impact on investors’ information use. Fifty-five percent of respondents in her study reported using the Internet daily or weekly, whereas only 6% used library sources with that frequency for investment information. Williamson (2008, 2010) and Williamson and Kingsford Smith (2010), who studied information use by 520 online investors (most of whom were male, outnumbering women five to one) found that 82% of their participants used the Internet for investment-related information: brokers’ websites were the most often used sources of information, followed by company investor relations websites, advice from brokers or analysts via email, general financial information portals, and financial data or charting services. Traditional media were used as well; 79.2% of online investors indicated they obtained information from newspapers and 55.5% from electronic media (radio and TV) at least “frequently.” The majority of participants in their study (62.8%) reported rarely receiving information or advice from family, friends and acquaintances. However, Williamson (2008) reported that “this result turned out to be questionable during the individual interviews. There was much more discussion with family and friends than people either wanted to admit, or thought of admitting” (p. 10).

Mezick's participants described convenience (38%), currency (14%) and ease of use (14%) as primary reasons for Internet adoption. Williamson (2008, 2010) and Williamson and Kingsford Smith (2010) also provide some insight into how online investors select information. In interviews with 26 investors selected from their larger pool of 520, they found that convenience/ease of access and content were important criteria to all their participants, whereas reliability/accuracy, currency/timeliness, and speed of access were important to the majority of them. It is important to note that adoption of technology for investing and investment information seeking varies markedly by sex. In one study, women were 20% less likely to use the Internet for investing information and three times less likely to trade on the Internet (Hira & Loibl, 2006). When asked why they do not use the Internet for investing and investment information, 85% of female investors surveyed indicated they preferred working with people, 49% worried about security, and 44% found financial websites confusing.

Barber and Odean (2001) discovered that the Internet may transform investors’ information behaviors. Because the Internet reduces the costs of some information, but not others, it may impact source selection. They write:

The Internet especially facilitates comparisons of real time data, and thus has changed investors’ focus by emphasizing the importance of speed and immediacy. While the serious individual investor of a decade ago may have checked stock positions once a day in the morning paper, casual investors now check theirs several times a day. Many more investors pay attention to short term—even intraday—returns than ever before. (p. 48).

Other effects of Internet adoption are clear. According to Barber and Odean (2001) new communication channels on the Internet and the popularity of online trading are closely related: the explosion in web-based investment information is “substituted for brokerage firm guidance, supporting (if not inflating) the sense of confidence for the retail investor” (Barber & Odean, 2001, p. 42). Their research indicates that “when people are given more information on which to base a forecast or assessment, the accuracy of their forecasts tends to improve much more slowly than their confidence in the forecasts. Although the improved accuracy of forecasts yields better decisions, additional information can lead to an illusion of knowledge and foster overconfidence, which leads to biased judgments” (Barber & Odean, 2001, p. 46). The trend to bypass professional investment advice is particularly troublesome because inexperienced nonprofessional investors earn lower returns as their use of unmediated information rises relative to their use of mediated-information (Elliott, Hodge, & Jackson, 2008). Furthermore, in lieu of professional advice, investors “turn to numerous sources of fundamental and technical market information, to chatroom gossip, to online journalists, and to sophisticated advice engines. However the quality of such cyber-resources varies greatly. If investors are unable to distinguish high quality advice from low, they are unlikely to pay more for quality. Indeed with so much information available for free on the Internet, many investors will be unlikely to pay anything for information alone” (Barber & Odean, 2001. p. 44). Thus the abundance and immediacy of Internet-based information strengthens the illusion of being informed (Barber & Odean, 2001). Because investors are heavily influenced by mass media, which lures them to purchase “attention grabbing” stock, greater exposure to more information may also alter the types of companies in which they choose to invest (Barber & Odean, 2008).

Williamson (2008, 2010) and Williamson and Kingsford Smith (2010) also identified several problematic effects of investors going online, including participants’ higher levels of trust in information sources than might be warranted, a predilection for speed in the delivery of information and the impact of information overload, which they speculate may be greater on information seekers less experienced than most of those who participated in their study. Overall, Williamson and Kingsford Smith conclude, however, that online investors are a “relatively engaged and knowledgeable group,” which, they add, is “by contrast to the wider populations of investors who research demonstrates are extremely difficult to engage, even in their own self-interest” (Williamson and Kingsford Smith, 2010, p. 69).

Online Information Sharing and Use

In addition to literature on investor's information use, research on information sharing and use in virtual communities provides a useful context for this study. Several authors have examined the exchange of health information in online groups (see, for example, Donnelle & Hoffman-Goetz, 2009; Burnett & Buerkle, 2004; Wikgren, 2001, 2003. Savolainen (2001) studied consumer information exchange in a Finish newsgroup. These studies generally find richer information environments than were expected. Strandberg (2008) and O'Connor and Rapchack (2012) examined online information use in political discussion forums and found that virtual environments are not always collaborative. Political discussion boards often included “negative comments, superficial topics, and unsubstantiated claims rather than true conversations” (Strandberg, 2008, p. 83). Clearly information behavior in online communities varies with the make-up of their membership and the nature of the interests that unite them.

A few studies have examined investors’ online forums. Antweiller and Frank (2004) examined 1.5 million postings to assess the impact of investors’ online discussions on the market. They found that contentious discussions induce trading, that forum discussions predict market volatility and that forum content reflects public information extremely rapidly. They discovered that discussions about news preceded and predicted treatment of that same topic in print newspapers by about 48 hours and concluded that internet investor boards do contain useful information.

However, some studies describe negative aspects of investor chat forums. Langevoort (2002) asserts that the “illusion of control provided by the Internet combined with an immediate audience for hype, fraud, or even ordinary opinion can make investors in chat groups more vulnerable” (Langevoort, 2002, p. 15). The Financial Industry Regulatory Authority (FINRA), an organization that regulates the behavior of professional securities brokers, does delineate acceptable behavior of professionals within social media. They prohibit professionals from posting content that is “unbalanced, overly positive or predicts an imminent price increase” (FINRA, 2011, p. 2). But the agency is still concerned about investor manipulation in chat forums and plans to issue more regulations in the near future (FINRA).

Park, Konana, Gu, Kumar, and Raghunathan (2010) analyzed 502 postings from the largest finance message board in South Korea. They found that investors exhibit confirmation bias, the tendency to seek out information that confirms what they already believe, when they select and use information from message boards. This well-documented tendency is exacerbated in virtual communities, because they enable people to interact with individuals who share their beliefs and opinions (Frick, 2011). Barber and Odean (2001) agree that “investors are more likely to visit chatrooms of like-minded investors and, if controversies ensue, they are likely to be convinced by those with whom they already agree. Investors who believe that additional information makes them better investors are unlikely to seek out or attend to evidence that indicates otherwise” (p. 47). Park et al. (2010) also demonstrate that investors with stronger confirmation bias also exhibit greater overconfidence. Consistent with the findings of other studies, overconfident investors in their study also had higher expectations about their performance, traded more frequently, and realized lower returns. They conclude that “these results suggest that participation in virtual communities increases investors’ propensity to commit investment mistakes and is likely to be detrimental to their investment performance” (Park, et al., 2010, p. 1).

Additional literature on group investing is also relevant to this study, although research in this area is slim and its findings are ambiguous and even contradictory. Barber and Odean (2000) demonstrate that overall, investment clubs do not perform well. During their 18-month study, 60% (n = 100) of participating clubs underperformed the market by an average of 4. Club returns were also consistently lower than individual returns by 2 pps (points per share) per year. Hens (2008) supported Barber and Odean's findings. However, Gort and Gerber (2008) compared the returns of individual investors to those of groups of investors and came to different conclusions. They note “large performance discrepancies across groups” and concluded that “the best groups significantly outperform individuals” (Gort & Gerber, 2008, p. 24). They found that a high level of information exchange (where members not only share information but also evaluate and weigh contradictory information) was the best predictor of strong market performance. “Only if the group members’ opinions are communicated and discussed, do groups outperform individuals” (Gort & Gerber, 2008, p. 24).

Though investors’ group information behaviors have been studied in face-to-face contexts, little is known about it in virtual communities. Existing studies about investors in online forums tend to focus on individual information behavior instead. This study will begin to address that gap. Burnett (2000) categorized the types of interactions in online communities and, with Buerkle, revised them in a 2004 study. He separated them into two broad categories; interactive and noninteractive behaviors. Although non-interactive behaviors, often called “lurking” are important, they are beyond the scope of this study. Burnett categorized interactive behaviors, which require active posting of messages, as fundamentally either hostile or collaborative. He furthermore divided collaborative behaviors into those that are explicitly information-oriented and those that are not. These typologies will be used for data analysis in this study and will be discussed at greater length in the Methods and Findings sections of this article.

Collaborative Information Behavior

Although both collaborative and noncollaborative interactions will be considered, this article will focus on analyzing collaborative, information-oriented behavior. Thus, a brief discussion of research on collaborative information behavior (CIB) provides a useful context for this work. Karunakaran, Spence, and Reddy (2010) define CIB as “the totality of behavior exhibited when people work together to identify an information need, retrieve, seek and share information, evaluate synthesize and make sense of the found information, and then utilize the found information” (p. 2). Reddy and Jansen (2008) and Reddy, Jansen, and Spence (2010) describe four triggers for CIB: (1) complexity of information need, (2) lack of immediately accessible information, (3) lack of domain expertise, and (4) fragmented information resources.

Collaborative information behavior has been identified as occurring in both organizational and nonorganizational contexts (Karunakaran et al., 2010). When CIB occurs in investor forums, where participation is voluntary, rather than in organizations, where it is mandatory, it occurs within a greater context of community building. Burnett (2000) describes these online exchanges of texts as virtual communities that “function as social spaces supporting textual ‘conversations’ through which participants can find both socio-emotional support and an active exchange of information” (p. 3). Burnett and Jaeger (2008) argue that these communities can be viewed as “computer-mediated small worlds” with the same types of normative attitudes and behaviors that shape information behavior found in the nonvirtual world (p. 10)1. These social norms “provide a shared understanding of propriety and correctness of those visible aspects of social activities within the world,” including information sharing and use (Burnett & Jaeger, 2008, p. 6). Burnett and Buerkle (2004) note that the variance between communities can provide an important means for understanding the small worlds that exist in the communities. This study will extend the literature of virtual CIB by describing a previously unknown social information environment and providing an additional point for comparison.

Methods

As previously stated, the purpose of this study is to describe investors’ behavior in online forums and analyze their information sharing and use. Specifically, it seeks to answer three research questions:

  • R1: What types of communication occur in online investing forums and to what degree can the forums be considered information environments?
  • R2: What motivates collaborative information behavior in online investing forums?
  • R3: What types of information are cited and valued by participants in online investing forums?

The first significant challenge in a study like this is selecting data sources from an overwhelming array of existing investing forums. Big Boards (search.big-boards.com) was used to identify top discussion groups by subject. Morningstar Discussion Boards, Market Thoughts, and Finance Forums were the most active boards that were exclusively investment related and moderated. Unmoderated forums were considered, but they were too spam heavy to yield good data. The organization of each of the boards also had to offer reasonable options for data collection. It was also important to balance these discussion groups, so effort was made to select forums that differed from one another in the apparent sophistication of their discussions. Although every effort was made to select three forums representative of the array of existing types of forums, the selection of just three sites should be understood as a limitation of this study. Because unmoderated sites were avoided, the data in this study may provide a more sophisticated picture of information use and sharing than would have otherwise resulted.

Because financial cycles can swing so much, and conceivably have a real impact on the discussions, data were collected across four years, from 2007 (prior to the recession) through 2010 (as the economy began to rebound). Sampling procedures were used to select a reasonable data set. Each thread in that time period was assigned a number from 1 to the last thread in the forum. A random number generator was used to select a stratified random sample 20% of the threads per year, which, as Table 1 illustrates, yielded a total 358 threads for all three forums. Four of the threads in Morningstar had no postings (presumably they had been deleted by the moderator), so a total of 354 threads containing 1,787 posts were analyzed. Postings were copied to a Word file for data analysis.

Table 1. Threads and postings across forums.
 2007200820092010

Total

sampled

Total threads

analyzed

Total postingsAverage postings per thread
Morningstar4657456321120711595.59
Market Thoughts4020181795953673.86
Finance Forums913151552522615.02
Total Threads9590789535835417875.05

Individual posts, which were the unit of analysis for this study, were coded according to Burnett's (2000) and Burnett and Buerkle's (2004) typologies and placed into one category. As described previously, these typologies are separated into two broad categories; interactive and noninteractive behaviors. This study focuses on interactive behaviors, which are defined as the active posting of messages. Interactive behaviors are coded as fundamentally either hostile, or collaborative. Hostile postings include flaming, which Burnett defines as “ad-hominem argumentation, aiming neither for logic nor for persuasion, but purely and bluntly at insult” (Burnett, 2000, p. 14); trolling, which is “posing a message or the purpose of eliciting an intemperate response” (Burnett, 2000, p. 15) and spamming, which is “the online equivalent of unsolicited junk mail” (Burnett, 2000, p. 15) and cyber-rape, which is “unsolicited, unwelcome (and violently assaultive) information that transgressed the behavioral norms and any shared sense of subject scope held by the community” (Burnett, 2000, p. 16). As previously stated, collaborative information behavior is broadly defined as the “totality of behavior” in identifying and meeting an information need (Karunakaran, Spence, & Reddy, 2010). Burnett defines collaborative behavior more generally and in contrast to hostile behavior as those positive behaviors that “reinforce the community” (Karunakaran, Spence, & Reddy, 2010; p. 17). Collaborative information behaviors are further classified as either explicitly information-oriented or non-information-oriented or neutral. According to Burnett and Buerkle, neutral behaviors include pleasantries and gossip, humorous behaviors, and empathetic behaviors For this study, collaborative information behavior was coded when the “exchange of information—in terms of seeking or offering of information—is an explicitly motivating factor in an interaction” (Burnett, 2000, p. 18).

Coding was conducted first by the author of this study. To assess consistency and reliability of coding, a graduate student was asked to code a data sample. Because Burnett's definitions are clear and well-defined, coding was highly consistent between coders. Other than the “neutral” classification, which was expanded as will be discussed in the findings section, Burnett's (2000) and Burnett and Buerkle's (2004) typologies accurately describe the variety of communication identified by this study. Examples from the data will help clarify coding distinctions; they are presented and discussed in the findings section, where they will simultaneously help illustrate investors’ information behavior.

Posts that were coded as collaborative and information-oriented were further analyzed using a content analysis method. Specifically, content was examined to learn whether or not one or more of Reddy, Jansen, and Spence's (2010) triggers could be identified as motivating collaborative information behavior. Possible triggers are lack of domain knowledge, complexity of information, lack of access to information and fragmentation of information (Reddy & Jansen, 2008; Reddy, Jansen, & Spence, 2010).

This study then used methods employed by Wikgren (2003) to examine health information citation behavior in Internet discussion groups. She sampled 30 discussion threads from English-language Internet groups based on two criteria; the threads contained at least one citation to a formal source and the discussion revolved around their topic, nutrition. Posts with citations were extracted, and cited sources were coded as a single source type. Source type codes (for example, books, journals, and websites) were informed by those used by Wikgren (2003), Kingsford-Smith and Williamson (2004), and O'Connor (2013), but because the investing environment is unique and the online environment is ever-changing, new information categories (such as stock and bond rating services) became necessary. These were developed iteratively as source types were coded and type codes were assessed for how well they functioned to describe identified sources.

For the purposes of this study, which occurs in an online environment where observation is one-dimensional (i.e., message postings provide all evidence and behavior cannot be observed separately from self-report), the presence of a citation in a posted message constitutes both information use (the poster clearly drew on the information themselves) and information exchange and sharing (the poster is providing the information to others), which are all used synonymously here. The one-dimensional nature of online information use and sharing should be understood as an inherent limitation of this study because it does not allow for the same distinctions among these behaviors as would be possible in an in-person study.

Findings

The Typologies of Postings

Analyzing discussion board postings by type allows researchers to “determine the degree to which that community can be thought of as an information (or purely social) environment for its participants” (Burnett & Buerkle, 2004, p. 2). Table 2 illustrates how posts were distributed across Burnett's (2000) and Burnett and Buerkle's typologies. According to Burnett (2000), “hostile interactive behaviors—flames, trolls, spam and cyber-rape—are those behaviors which do suggest interaction between members of the virtual community, but which emphasize overt aggression and conflict rather than congeniality or the social exchange of information” (p. 10). Because these forums were heavily moderated, the low occurrence of hostile behaviors across all three forums is not surprising. Spam postings, disguised as informative postings, did occasionally make it past moderators, particularly in the Market Thoughts board.

Table 2. Posting types across forums.
Message typeMorningstarMarket ThoughtsFinance Forums
Hostile Interactive Activities % % %
Flaming00.0000.0000.00
Trolling00.0000.0000.00
Spamming00.00143.8141.53
Cyber-rape00.0000.0000.00
Responses to Hostile Postings00.0020.5400.00
Collaborative Interactive Behaviors—NonInformational
Neutral786.73246.5483.07
Humorous110.9530.8283.07
Emotional40.3500.002.77
Responses to Neutral35130.283910.63215.36
Collaborative Interactive Behaviors—Informational
Announcements30.2620.5400.00
Queries11610.005214.173914.94
Responses to Queries59651.4223162.9417971.26
Total1159 367 261 

Non-information-oriented interactivity includes neutral behaviors, which Burnett (2000) defines as pleasantries and gossip, humorous behaviors, and empathetic behaviors. The latter two types functioned well, but the neutral category was not broad enough to be useful for this study. A common type of posting in these forums was a request for other participants’ opinion about the poster's finance –related activities. These postings often described a financial decision or strategy (or in some cases a summary of the poster's portfolio) and invited other members to comment on it. A typical example is:

Looking for some tips/advice on what to do as far as saving for grad school.

Currently I have no debt except for my vehicle, around 23k, 4.9% interest rate over 60 months. Have had the vehicle for almost a year …

I have a little over 14k in cash, and 3.5k in a ROTH IRA started a couple of years ago.

I plan to start a two-year program in the summer of 2011. By that time, I should have at least 30k saved up.

Tuition will be a little over 20k not counting additional fees or cost of living. Still, I feel pretty good about the amount I've saved up.

I'd like to contribute the max to my ROTH this year as eventually I won't qualify for it anymore. Additionally, I think any student loans can be deferred (hopefully) until after graduation meaning that I'll be using them mostly as a safety net and can hopefully wipe a large part of them out with saved cash after graduation.

Just looking for any tips or advice about what you'd be doing in my shoes. [Morningstar, 2009]

These types of postings were coded as neutral, since they were not specifically information-oriented, and a category was added for responses to these requests, since they were numerous. Although it could be argued that asking for such advice may be information-oriented, and occasionally posters shared information in response to the requests, they were generally closer to requests for support than for information. As Table 2 demonstrates, humor and emotional support were not substantial activities on these boards, but this type of “tell me how you think I'm doing” posts were.

Information-oriented interactive behaviors composed the overwhelming majority of the behaviors exhibited on these boards. In fact, across the three forums, an average of nearly 75% of the postings were either requests for information or responses to such requests. Requests for information varied from direct statements of information need:

I have been searching for a good source of M&A and company spinoff information, so that I can do some research on this aspect of the market. Any info is much appreciated.

Thanks … Colonel redfeather. [Market Thoughts, 2008]

to broader, more speculative questions, such as:

Assuming the bank doesn't become insolvent or taken over by the FDIC, what might its future be? How low can the stock fall as a practical matter? It does not seem to have many similarities to Bear Stearns, yet its future seems destined to end the same way. Is the financial condition of Wachovia as bad as the market seems to think, or is it a victim of FUD? [Morningstar, 2010]

Posters also frequently asked for instructions on how to use certain financial tools or calculate a variety of financial measures, as in the example below:

How can I calculate standard deviation of a portfolio of funds (historical measurements)? I want to see how I am doing against say S&P500 or Russell2000. [Morningstar, 2008]

Requests for information nearly always received multiple responses, and this kind of exchange constitutes a central activity type on these investor boards. Postings were coded as “responses” both because they were included as a “reply” in the thread and contained content that was apparently formulated in response to the original posts. Posts were coded distinctly as “responses to neutral” and “responses to queries” only because the original post to which they were responding was coded as either a neutral or query post and, thus, should not be understood as inherently distinct from each other in any other way. A typical response to a neutral posting is:

Look at Ohio 529 plan for yourself. You get some benefit on state taxes. With school starting just next year and mm not paying anything meaningful, just use laddered CDs for investment option.

M* shows a stable return option but I don't see anything other than MM and bank CDs at plan site. [Morningstar, 2010]

A typical response to a query is:

I think what you might be looking for is portfolio Beta, which measures both portfolio volatility and risk. Standard deviation is a broad measure of a security's price movement due to market fluctuation.

Investopedia has an excellent discussion of these two measures of portfolio volatility, along with R squared and Alpha, which you can see at http://www.investopedia.com/articles/mutualfund/03/072303.asp. BruceM. [Morningstar, 2008]

Both types of responses contain a variety of opinions, statistical data, and excerpts from and citations to formal information sources. They also contain other interactive behaviors such as humor and gossip, however because, for the sake of consistency with previous studies, postings were coded as a single type, these activities are not illuminated by these data. Further research should include a more finely grained analysis of online communication activity by coding messages for multiple types of behavior.

Content Analysis

After postings were coded by type, those coded as collaborative queries were further analyzed to ascertain what motivated the collaborative information behavior. Recall that the four possible triggers are lack of domain knowledge, complexity of information, lack of access to information, and fragmentation of information (Reddy & Jansen, 2008; Reddy, Jansen, & Spence, 2010).

Lack of domain knowledge was the most frequently cited trigger for collaborative information behavior. Statements about a lack of domain knowledge were often framed by personal experience. “I'm new to this,” “I just don't know much about bonds yet,” and “The problem of asset allocation seems so complex” are examples of the types of statements that were coded for domain knowledge. Nearly 36% of the queries fell into this category. Framing requests for information in terms of personal knowledge and experience not only served to identify the posters’ information gaps, but it also provided critical information to other posters about the level of information that would be accessible to the individual.

Complexity of information was the second most frequently identified reason for CBI. Statements such as “These charts are confusing,” and “Is there an easier to understand source for this information,” typified postings coded in this category. Approximately 28% of the postings made statements about complexity of information.

Only a few postings (6%) cited lack of access as a trigger for CBI. Lack of access was typically because of costs (for fee-based information) rather than to lack of physical access. Note that more multiple triggers may have occurred in a single posting, so there is some overlap in the percentages cited here. Statements about fragmentation of information were not noted in the postings.

Once the content of collaborative, information-oriented posts was analyzed, the sources cited by those posts were examined.

Analysis of Sources Cited

As Table 3 illustrates, 485 (27%) of the 1,787 posts analyzed cited at least one information source. A total of 768 sources were cited by investors in all three forums for an average 1.58 sources per posting containing a citation. Citation density (percentage of posts with citations) varied widely by forum with a high of 63% for Market Thoughts, 19% for Morningstar and 12% for Finance Forums.

Table 3. Citations to sources across forums.
 Total postingsNumber of postings with citation

Percentage of

postings with citation

Number of sources citedAverage number of sources citeda
  1. aIn posts with citations.
Morningstar1,15922119%4261.61
Market Thoughts36723263%2561.10
Finance Forums2613212%862.47
All Forums1,78748527.14%7681.58

Figure 1 illustrates the types of sources that were cited across the three forums. Specifically, it reports the number of postings that cited each source type. Discussion forums and posts (n = 113), books (n = 108), rating services sources (n = 91), and news sources (n = 87) were the most often cited types of sources. The discussion forum category contained citations to postings within the same board, as well as to postings in other financial boards and to other boards generally. Information, such as stock and bond ratings, historical price information, and articles, available through securities rating services, including Standard and Poor's, Mergent's, Valueline, and Morningstar, compose the rating services category. The books category contained both print and electronic books. The titles listed in Table 4 were cited repeatedly. News sources include traditional news media, such as the Wall Street Journal, news services, such as Reuters, and financial news aggregators.

Figure 1.

Sources used in three online discussion forums. (Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.)

Table 4. Books cited in forums four or more times.
Book title with author(s) and publicaton yearaTimes cited in postings
  1. aPublication dates are for the latest edition of these works. Dates were typically not provided in posters’ citations.
Graham, B. (2009). The Intelligent Investor: The Definitive Book on Value Investing.11
Bernstein, W. (2010). Four Pillars of Investing: Lessons for Building a Winning Portfolio.10
Bogle, J. (2007). The Little Book of Common Sense Investing: The Only Way to Guarantee your Fair Share of Stock Market Returns.9
Larimore, T., Lindaur, M., & LeBoeuf, M. (2007) The Bogleheads’ Guide to Investing.6
Schultheis, B. (2009). The Coffeehouse Investor: How to Build Wealth, Ignore Wall Street, and Get on with Your Life.6
Perrucc, D., & Miccolis, J.A. (2011). Asset Allocation for Dummies.6
Malkiel, B. (2007). The Random Walk Guide to Investing.5
Ferri, R.A. (2010). All About Asset Allocation.5
Bogle, J. & Swenson, D.F. (2009). Common Sense on Mutual Funds4

Comprehensive financial websites (n = 67), financial tools (n = 63), the websites and blogs of investment gurus (here defined as people who are or want to be famous for dispensing investment advice; analogous to spiritual gurus) (n = 59), commercially sponsored financial websites (n = 42), and magazines, journals, and newsletters (n = 40) form the middle range of sources cited. Comprehensive financial websites are sites that offer one-stop portals for investors’ information needs, such as Yahoo! Finance (http://finance.yahoo.com), CNN Money (http://money.cnn.com), and SeekingAlpha (http://seekingalpha.com). They include a wide variety of sources such as news, portfolio management tools, glossaries, and social networking spaces. The financial tools category includes sources such as interest calculators, portfolio management software, and charting programs not embedded in comprehensive financial sites. Guru sites and blogs are information sources that revolve around the advice of a finance guru (from the well known to the unknown). A few of the blogs were institutionally sponsored, but most were written by individual “experts.” Commercially sponsored financial information includes a variety of advisory and educational content on the websites of companies or individuals who have a financial interest in the services and products discussed. For example, Fidelity's page on selecting mutual funds would be coded for this category. Additional examples are provided in the discussion section of this article. Magazines, journals and newsletters include print and electronic sources. If a newsletter was written by a guru, it is coded in the guru and blogs category, thus the newsletters in this category are published by a variety of organizations and associations.

Company websites (n = 19), reference sources (n = 17), government information (n = 13), and broadcast media (n = 7) were the least cited types of sources. Company websites provided information on a particular company, typically as a potential stock investment or purveyor of products and services. Reference sources included stand-alone encyclopedias, dictionaries, and directories. Government sources included information published by federal and state governments and regulatory agencies, such as the Securities and Exchange Commission (SEC). They also included sources of demographic or economic statistical information. Television shows, such as Jim Kramer's Mad Money on the Fox network, and radio broadcasts such as Marketplace from American Public Media, are coded in the broadcast media category. Forty-two sources were coded either as having a nonfunctional link or as a miscellaneous source type.

Differences Across Forums

At first glance, Market Thoughts seems to be, by far, the richest information environment, with 63% of its postings citing an information source compared to 19% and 12% in Morningstar and Finance Forums respectively (see Table 2). However this is largely the result of an information sharing and use pattern that is unique to Market Thoughts among the three. Many of the postings were a sort of positive trolling, for lack of an existing term. Burnett (2000) defines trolling as “deliberately posting a message for the purpose of eliciting an intemperate response” (p. 15). Positive trolling posts were clearly meant to elicit a response, just not an intemperate one. Typically, the poster would post a source and ask for comment. The following post from Market Thoughts is a good example:

Wachovia to exit the wholesale mortgage lending business. About time, isn't it?! http://www.bizjournals.com/charlotte/stories/2008/07/21/daily11.html?jst=b_ln_hl [Market Thoughts, 2008]

Because maximum comparability was sought with Burnett and Burnett and Buerkle's (2004) findings and because trolling and positive trolling are far too dissimilar to code together, these messages were not coded as trolling but, rather, as queries as illustrated in Table 2. However, a re-coding of positive trolling posts accounted for as much as 23% of the citations in Market Thoughts. Seldom were there any responses to these types of posts. Thus, the average posting per thread is much lower for Market Thoughts at 3.86 compared to 5.59 and 5.05 (see Table 1) and the average source per citation per thread is also lower at 1.10 compared to 2.47 and 1.61 (see Table 2), which is also a result of this behavior.

Figure 2 illustrates the source types for which there were substantial citation differences between the two large forums, Morningstar and Market Thoughts. The number of postings in the Finance Forum board was simply too small to note substantial variations with the other two boards. Books, other discussion forums, rating services, and commercially sponsored finance websites were cited by participants in Morningstar far more often than in Market Thoughts. On the other hand, news sources, financial tools, and blogs were cited more frequently by participants in Market Thoughts than in Morningstar.

Figure 2.

Variation in source use across forums.

Despite the variations across forums, the data in this study provide a clear picture of how investors use and share information in online environments. The findings not only have clear implications for practice but also identify areas for which further inquiry is necessary. Both are included in the discussion that follows.

Discussion

This study's data, on the one hand, confirm some of what previous research suggests about information behavior across contexts, but, on the other hand, highlight what is unique about information behavior in virtual environments. They also demonstrate how the investing domain uniquely impacts online information behavior in relation to other domains. Specifically, this study demonstrates that online investors tend to form more collaborative and information dense (i.e., containing more information-oriented posts and higher ratio of citations to posts) environments than have been found in other types of communities; that online investors post substantially fewer messages that function primarily as social (humor and play, for example) and hostile (flaming and spamming, for example) behavior than do participants in other online communities; that, although they still draw heavily on personal sources of information, they are more likely to value and utilize traditional materials such as monographs and newspapers, and that the information sources themselves are rarely contested and are more widely accepted as authoritative. Each of these assertions will be discussed in turn.

Information Behavior

The findings of this study support Burnett's (2000) claim that virtual communities often provide a more collaborative and information dense than is expected, which, again, is defined by both the ratio of sources cited to the total number of postings and by the overall percentage of information-oriented posts. Clearly, as demonstrated by the high percentage of posts (68% ; n = 1,213) coded as queries and responses to queries, these online investment forums constitute an information environment (as opposed to merely a social one). The results of this study demonstrate that, in terms of collaborative behavior, investor forums are more similar to the health-oriented forums analyzed by Wikgren (2003) and Burnett and Buerkle (2004) than the consumer or political forums examined by Savolainen (2001), O'Connor and Rapchack (2012), and Strandberg (2008), respectively.

The citation patterns in this study also demonstrate a normative information culture in which sharing information is an important element of the investing community. Wikgren (2003), who used citation density (the ratio of postings that cite at least one source to the total number of postings analyzed) as a measure of collaborative information behavior, discovered 18% of the posts in her study contained citations, which is much lower than the average 33% identified for the three forums in this study. However, if the single “positive trolling” posts (posts that are intended to spark a reaction from other participants), which constituted 23% of the postings with citations, are removed, the average citation density for Market Thoughts is only 48%, which would reduce the average for the three forums to a little more than 26%, which provides a more representative picture overall. So these investing forums, together, are still a little more information rich than the health forums Wikgren studied, although Morningstar was very similar at 19% and Finance Forums is a little lower at 12%. The investors’ forums, however are much more information dense than the political forums studied by O'Connor (2012), who found that only 6.3% of the postings she analyzed cited an information source. Clearly, higher citation density is consistent with previously stated findings that lack of domain knowledge, complexity of information, and, to a lesser extent, lack of access to information sources motivate investors to participate in these virtual communities.

The low occurrence of hostile behaviors observed in this study may furthermore support the author's contention that a collaborative social norm governs these virtual investors’ communities. Even if this low occurrence is the result of vigilant moderating, it still provides evidence that these investors are less tolerant of noncollaborative behaviors, such as flaming and spam, than were members of boards analyzed by Burnett and Buerkle (2004), O'Connor (2012), and Strandberg (2008). One limitation of using Burnett's (2000) typologies is that they only measure disagreement when it takes the form of flaming and cyber-rape, thus it is impossible to tell the extent to which conflicting viewpoints were considered and weighed in a collaborative manner. Because Gort and Gerber (2008) connect high information exchange, which they define as time spent weighing a variety of viewpoints, to higher returns on investment, measuring the level of information exchange in these forums would be useful. Burnett's typologies would need to be expanded if they were used for such a study. Furthermore, because several authors (Barber & Odean, 2001; Frick 2011; Park et al., 2010) have noted that confirmation bias may be a more significant issue in online communities, it is important to establish whether the lack of hostile behaviors indicates a general a-priori agreement across investors participating in this discussion board. In other words, are these boards less hostile because they have simply drawn participants with compatible viewpoints on investing or are the investors in online communities more successful in weighing diverse viewpoints in a collaborative manner? This is an important question for further exploration.

Information Use

Although this study demonstrates that virtual communities can provide rich environments for collaborative information behavior, it also confirms some of the challenges of information seeking and use in the online environment, including those outlined by Williamson (2008, 2010) and Williamson and Kingsford Smith (2010), such as participants’ higher levels of trust in information sources than might be warranted and a predilection for speed in the delivery of information.

As has been found in many previous information seeking and use studies, people have a strong preference for interpersonal sources of information (Case, 2007). This is also true for investors. For example, although nearly 63% of respondents in Williamson and Kingsford Smith's (2010) study said they rarely or never received information or advice from family, friends, or acquaintances, follow-up interviews indicated otherwise. In an online discussion group context, assessing how much interpersonal information sharing and use occurs outside the forum is difficult; however reliance on personal sources is evident in the heavy citation of other discussion forums and posts (see Figure 1). The high level of social inter-communication found in this study is unprecedented, though predictable (because all the subjects of the study are by definition participants in discussion forums and thus are presumably more likely to cite them). In contrast, Wikgren (2003) found only 4% of citations were for other types of discussion groups and Williamson and Kingsford Smith indicate that only 10% of investors in their study used chat rooms regularly. However even though one would expect higher citation to discussion forums by discussion forum participants, the fact they were the most cited sources, particularly in the Morningstar forum, demonstrates how highly information obtained from other investors is valued. This supports Barber and Odean's (2001) view that investors may be substituting informal, interpersonal advice for professional, fee-based counsel.

Another concern raised by Wikgren (2003), Williamson (2008, 2010), and Williamson and Kingsford Smith (2010) is the extent to which individuals in online environments tend to rely on popular rather than scholarly or scientific sources of information. Nearly 40% of the sources cited by investors in this study are from popular literature. Wikgren (2001) also found that only 60% of her citations referred to what she categorized as scientific sources, which included Medline abstracts, scientific journals or books, educational cites, and researchers or physicians. Investors’ high citation rate for books, particularly in the Morningstar forum, was a unique finding of this study. Although the cited titles were almost entirely from the mainstream, popular presses, it would be interesting to examine their content to assess how much they do or do not rely on scholarly information.

Although not as pronounced as Wikgren's (2003) finding that nearly 25% of web citations and 9% of all citations were to pharmaceutical or other related companies, investors’ reliance on commercial sources of information is nonetheless another area of concern. The use of investment-related information provided by an entity that has a vested interest in selling related products and services is analogous to physicians and patients who obtain their information from pharmaceutical companies. At times, it was evident that a commercial sponsor was providing the information in a website. More troublesome, however, were the websites whose commercial sponsorship was not clear and could only be ascertained by reading the “about us” link or conducting additional research on them.

Finally, what was unique about information use in investors’ forums was that, overall, information was not contested. Members seemed to agree implicitly on what made sources trustworthy for investing. Very seldom were sources challenged. Data for this study were gathered before the political neutrality of Standard and Poor's was broadly questioned after downgrading U.S. Bonds while having failed to downgrade mortgage companies or banks prior to the recession. It would be interesting to assess whether or not the ensuing conversation about the political nature of stock and bond analysis created a more skeptical attitude toward these sources.

Conclusion

This study takes a unique approach to understanding investors’ information behavior in online communities by drawing on and combining the distinct approaches of Wikgren (2003), Burnett (2000), and Burnett and Buerkle (2004). It describes a virtual environment that is highly collaborative and information dense compared to similar communities in other domains, such as health and politics. However it also illustrates areas of concern, such as investors’ heavy reliance on inter-personal and popular sources of information, an uncritical stance toward investment-related information, particularly well-established sources such as stock and bond rating services, dependence (sometimes unwitting) on commercial sources of information and a tendency to solicit and use information without evaluating or contesting its quality. Further research is necessary to expand our understanding of these behaviors and determine methods for improving investors’ ability to identify, select, and use high-quality information for their investing decisions.

Footnotes

  1. 1

    Burnett and Jaeger (2008) provide an excellent discussion of Chatman's concept of small worlds as it relates to online communities.

Ancillary