Can Twitter Help Predict Firm Value?
Aggregating stock information from tweets might allow investors to tap into the wisdom of crowds.
New research shows that aggregating stock-related information from Twitter can enable investors to tap into the ‘wisdom of the crowd’.
on information intermediaries — INVESTORS HAVE LONG RELIED from financial analysts and advisors to the business press, credit rating agencies, short sellers and auditors — to acquire timely and value-relevant information regarding the prospects of stocks. However, the past decade has witnessed an explosion in new sources of information that are easily accessible to capital market participants.
With the rise of the Internet, individual investors are increasingly relying on each other as peer-to-peer sources of information. The early days of the Internet saw the rise of financial websites such as Yahoo! Finance, Motley Fool and Raging Bull. More recently, individual investors have shared information and opinion through portals such as Seekingalpha. However, by far, the biggest revolution in the dissemination of information on the Internet has been the advent of social media platforms such as Twitter, which allow users to instantaneously post their views about stocks to a wide audience.
While Twitter is undoubtedly an exciting and emerging new source of information for the capital markets, it is unclear whether it is actually useful to investors. As Twitter is an unregulated platform with potentially anonymous users, the information in tweets may be uninformative or even intentionally misleading. For example, in two days in January 2013, a series of damning but false tweets on two stocks — Audience Inc. (ticker symbol: ADNC) and Sarepta Therapeutics, Inc. (ticker symbol: SRPT) — sent their prices plunging by 28 per cent and 16 per cent, respectively.
Even though information from Twitter might be uninformative or worse, misleading, there are also many reasons why it might actually prove to be a treasure trove of timely and relevant information for the capital markets. The main driver for this is a concept many readers will be familiar with: The Wisdom of Crowds.
The Wisdom of Crowds
The Wisdom of Crowds concept goes back over a century and refers to the phenomenon that the aggregation of information provided by many individuals often results in predictions that
are better than those made by any single member of the group, or even experts.
One classic example from the turn of the 20th century is Sir Francis Galton’s surprising finding that the crowd at a county fair accurately predicted the weight of an ox when their individual guesses were averaged. The crowd’s average (or median) prediction was closer to the ox’s true weight than the estimates of most crowd members, and even closer than any of the estimates made by cattle experts. Similarly, trial by jury can be understood as a manifestation of the wisdom of crowds, especially when compared to trial by a single expert (i.e. a judge).
The principle of the wisdom of crowds has been validated by the ability of the Iowa Electronic Markets to predict election results with little or no bias. In fact, its predictions for highprofile elections outperform polls conducted by experts. Recent academic papers that build on the wisdom of crowds show that user-generated research reports and commentaries posted on the Seekingalpha portal help predict stock returns and that Twitter information can predict future returns around Federal Open Market Committee (FOMC) meetings.
Aggregating stock information from tweets might therefore allow one to tap into the wisdom of crowds. Twitter has two features that make it an attractive source and conduit of information: Its users come from diverse backgrounds and they are truly independent. Hence, in contrast to traditional information intermediaries and other social media platforms (e.g., blogs, investing portals), they are less likely to be subject to ‘herding’ behaviour. Finally, Twitter’s short format (140 characters, with a 280-character limit in testing phase) and ease of information search (due the use of ‘cashtags’ like $AAPL) make it an ideal medium for sharing opinions and information in a timely and succinct fashion.
Our Research
Previous research on the impact of Twitter on the capital markets has focused primarily on how companies are exploiting this new channel to communicate with investors. While some researchers have examined whether aggregate investor sentiment on Twitter predicts the overall stock market, my colleagues and I decided to focus our attention on a potentially more intriguing question: Does firm-specific information from Twitter help predict a firm’s future earnings and stock returns?
In a recent paper published in the Accounting Review, Eli Bartov of NYU’S Stern School of Business, Lucile Faurel of Arizona State University and I explored three questions:
1. Does the aggregate opinion in individual tweets regarding a stock’s prospects predict its quarterly earnings?
This was a test of the ‘relevance’ of the information provided by Twitter. We hypothesized that the aggregation of opinions provided in individual tweets may result in a more accurate estimate of forthcoming earnings than the opinions of analysts, because individual tweets reflect opinions of a large and diverse group of people making independent and timely assessments of a company’s future prospects. If the wisdom of crowds and the value of the diversity and independence of Twitter users outweigh any concerns about the credibility of information from Twitter, then the information should be relevant to capital markets.
2. Does the aggregate opinion in individual tweets regarding a stock’s prospects predict the stock price reaction to the firm’s earnings realizations?
While information from Twitter may be relevant for predicting earnings, this question examined whether that information is new to the markets and hence moves stock prices. If this
Two features make Twitter an attractive source and conduit of information: Its users come from diverse backgrounds, and they are truly independent.
information is indeed new, we hypothesized, then it should also have the power to move stock prices.
3. Is the ability of information from Twitter to predict future earnings stronger for firms in weaker information environments?
We posited that the incremental impact of additional information from Twitter may mean a lot more for a firm with a weaker analyst following and press coverage, as opposed to a widely followed and covered firm.
To address these three questions, we analyzed a broad sample of tweets spanning a four-year period, from January 1, 2009 to December 31, 2012. Our sample only covered tweets providing information related to a stock’s prospects and value, written by individuals in the nine-trading-day period leading up to the firm’s quarterly earnings announcement. Our analysis focused on earnings announcements because they are recurring, high impact corporate events that are scrutinized closely by capital market participants.
We obtained complete historical Twitter data from GNIP, the first authorized reseller of Twitter data. The data consisted of the full archive of tweets with ‘cashtags’—stock symbols preceded by the dollar sign, such as $AAPL for Apple Inc. and $PEP for Pepsico Inc. In an effort to increase confidence that the tweets related to the firm financial performance and value, we limited our sample to tweets with cashtags. Given our interest in the ability of tweets to predict earnings, we focused only on tweets made in the nine-trading-day period leading up to the earnings announcement. This resulted in a sample of 998,495 tweets from 3,662 unique firms.
An important research design choice was how to measure aggregate Twitter opinion. To enhance the robustness of our results, we used two different approaches. The first was a machine learning approach for analyzing entire tweet phrases and classifying them as being positive or negative in their tone. The second approach focused on vocabulary, classifying words as positive or negative and then counting the number of positive versus negative words. Note that this was done at the individual tweet level. Our next step was to aggregate this to the firm level across all tweets. To do this, we averaged the opinion across all tweets for a given firm, giving greater weight to Twitter users with more followers. Surely, we felt, the opinion of a frequent Twitter user with 100,000 followers should mean more than an infrequent user with a handful of followers.
Our Findings
Our first question pertained to the ability of social media to predict quarterly earnings, and collectively, our results strongly suggest that the aggregate opinion from individual tweets does help to predict earnings.
Additional analysis provided some interesting insights. When we decomposed our sample into tweets that provided new information (e.g. ‘$TPLM red from entry looking to time add spot as small oils out of favour... still love this co’) and those that disseminated existing information (e.g. ‘Triangle Petroleum Announces Fourth Quarter Fiscal Year 2012 Earnings and Conference Call $TPLM #oil #petrol #TFB #OIL http://t.co/gvjdsjth’), we found that both kinds of tweets have predictive value. This suggests that Twitter has a dual role, both as a source of new information coming from individual users and as a means of disseminating existing information.
In addition, we found that tweets that directly deal with stock fundamentals (e.g. ‘Seen the lines for the new Iphone X?
To manifest itself, the wisdom of crowds needs a non-trivial number of distinct users providing their insights.
I think $AAPL will kill revenue estimates’) are more likely to be informative than general tweets (e.g. ‘I added more $AAPL in the 90’s today... developing’).
Our second research question pertained to whether the information from Twitter predicts future stock returns. Here too, we found a strong association between aggregate Twitter opinion and the announcement period’s stock returns. This suggests that not only is aggregate Twitter opinion relevant, it is also incrementally informative.
Our third research question dealt with how the ability of Twitter to predict future earnings varies with the information environment. We looked at three dimensions of information environment — analyst following, institutional investment and press coverage — and found that the ability of aggregate Twitter opinion to predict future earnings is much stronger for firms in weaker information environments. In such firms, the dual role of Twitter as both a new source of information and a new channel of dissemination is particularly important.
Finally, one of the assumptions underlying the wisdom of crowds is that the ‘crowd’ has enough participants that the ‘noise’ in individual opinion is diversified away, allowing the ‘truth’ to emerge. Consistent with this, we found that to manifest itself, the wisdom of crowds needs a non-trivial number of distinct users providing their insights in order for the information to be useful to the capital markets..
Implications for Practitioners
The dramatic increase in the use of social media in recent years has had a significant impact on the capital markets. Firms now regularly use social media to communicate with their investor base and, increasingly, individual investors use social media to ‘crowd-source’ information and insights about the prospects of particular stocks.
While a healthy skepticism regarding the value of information on Twitter is warranted, our results suggest that information aggregated from a number of users regarding a stock’s prospects
does provide timely and relevant information. Hence, Twitter is another embodiment of the wisdom of crowds and has a dual role as a source of new information as well as a vehicle for the dissemination of existing information. As indicated, these roles become particularly salient for firms in weak information environments.
Our results have important implications for the role social media plays in the investing community. While investing may be viewed as a non-cooperative, zero-sum game, our results demonstrate that individuals are using social media to share information regarding companies’ future prospects for their mutual benefit.
Our findings are also important for regulators. Skeptics argue that individuals exploit social media by disseminating misleading and speculative information, and thus call for more regulation of it. However, we found that the wisdom of crowds and the value of diversity and independence trump any concerns about the lack of credibility of information on Twitter. Put simply, information in social media may actually help investors make good investment decisions.
Overall, our findings highlight the importance for capital market participants to consider information on stocks from Twitter when assessing a stock’s future prospects and value.
Partha Mohanram is the CPA Ontario Professor of Financial Accounting, Director of the India Innovation Institute and Professor of Accounting at the Rotman School of Management. This article summarizes his paper, “Can Twitter Help Predict Firm-level Earnings and Stock Returns?”, co-authored with Eli Bartov of the Stern School of Business at New York University and Lucile Faurel of the W.P. Carey School of Business at Arizona State University. This paper is forthcoming in The Accounting Review and is available online.