Dr. Stock Microblogging

The TUM School of Management has awarded one of the first phd titles worldwide for research on Twitter. Timm Sprenger's dissertation on "The Information Content of Microblogs and their Use as an Indicator of Real-World Events" investigates whether information extracted from Twitter can serve as an indicator of real-world events and explores the mechanism that explains the efficient aggregation of information. 

The results illustrate that the sentiment (i.e., bullishness) of stock-related tweets is correlated with abnormal stock returns. In addition, they provide empirical evidence supporting the idea that followership relationships and retweets represent the Twittersphere's "currency" for weighing information, since users providing above average investment advice are retweeted (i.e., quoted) more often and have more followers. Thus, this dissertation contributes to the understanding and use of social media content in social science research.

International Conference on Weblogs & Social Media

TweetTrader.net presented its website as a demo at the International Conference on Weblogs and Social Media (ICWSM 2011) in Barcelona, Spain. The associated demo paper is available online as part of the conference proceedings. On this occasion, we also met with a "Who is Who" of the Twitter research community including Johan Bollen from Indiana University (author of "Twitter mood predicts the stock market"). Despite an intense conference program, there was time for a little bit of sightseeing to find that the Sagrada Family has a lot in common with our platform - it's beautiful, but still under construction...

What Topics are Stock Microbloggers talking about?

We have investigated which topics are on the minds of stock microbloggers. We have classified thousands of tweets into the following news or event categories: Corporate Governance, Financial Issues, Operations, Restructuring Issues, Legal Issues, and Technical Trading. In addition, to allow for more granular analysis, we have defined subcategories for which we found a substantial number of messages among our data set (e.g., stock-related vs. market-related technical trading signals).

The table below provides an overview of the event categories and sample messages assigned to the respective class. It shows that technical trading signals are the most frequently mentioned event category with references in almost one third of all messages (34.0%), among which stock-related signals make up the vast majority (28.2%). Next are comments regarding company operations (20.3%), especially product development (9.4%), operational performance (4.2%) and marketing (4.2%). Financial issues come third as an event category (13.3%) with a majority of messages dedicated to the discussion of earnings results (6.9%). Restructuring issues (6.1%), Legal Issues (3.5%) and topics related to Corporate Governance (3.3%) are mentioned less frequently. The distribution of topics shows some interesting differences to professionally edited newspapers and confirms that stock microblogs provide information "off the beaten track".

You find the current share of topics for each stock in real-time on the TweetTrader Scoreboard:

 

TweetTrader research featured in the New York Times

The New York Times (online edition) Freakonomics blog has recently featured results from our research on stock microblogging.

The Winner is... - Ranking of the Best Investment Advisors on Twitter

In order to find our whether there are stock microbloggers who consistently provide better investment advice than others we have used our proprietary classification method to classify each tweet in a 6-month sample period as a recommendation to buy, hold, or sell a stock. We define the quality of a tweet as the accuracy of this recommendation relative to same-day returns of the stock in question (i.e., the tweet "$AAPL going up" gets a point if AAPL was up by the end of the day). The quality of a particular user is the average quality of all tweets posted by this individual. We find that even among users with hundreds of messages, we can identify some that seem to consistently provide higher quality investment advice than others. And the winner is...

Sentiment of Stock Microblogs

Our sentiment analysis of stock microblogs shows that users tend to be much more bullish than bearish. We manually classified 2,500 tweets as either buy, hold, or sell signals.  Roughly half of these messages were considered to be hold/neutral signals (49.6%). Among the remainder, buy signals were more than twice as likely (35.2%) as sell signals (15.2%). This indicates that stock microblogs appear to be more balanced in terms of bullishness than internet message boards where the ratio of buy vs. sell signals ranges from 7:1 (Dewally, 2003) to 5:1 (Antweiler & Frank, 2004).

The table below shows a few typical examples. Our analysis of the most common words per class draws a semantic profile of buy, hold and sell signals. Obviously, some features occur frequently in all classes (e.g., numbers and hyperlinks). However, beyond these universal features, the most common words reasonably reflect the linguistic bullishness of the three classes. Positive emotions, for example, are much more likely among buy signals. In addition, buy signals often contain bullish words with an origin in technical analysis (e.g., “moving average”, “resistance”, “up”, or “high”), operations (e.g., “acquire”), financials (e.g., “beat”, “earn”), or trading (e.g., “buy”, “long”, “call”). Sell signals contain many corresponding bearish words in the areas of technical analysis (e.g., “support” and “cross”), financials (e.g., “loss”) or trading (e.g., “short” and “put”). As a results of the frequent occurrence of negative adjectives (e.g., “weak”, “low”) and verbs (e.g., “decline”, “fall”), negative emotions are among the most common features in sell signals. Positive and negative emotions are much more equally balanced in hold messages, which also contain more neutral words such as product names (e.g., “ipad”, “iphone”) and make fewer references to specific price targets (i.e., dollar values).

Stock Microblogs Capture the Pulse of the Market

The business press has come to describe stock talk on Twitter as "the modern version of traders shouting in the pits" and calls related third-party applications, such as TweetTrader.net, “a Bloomberg for the average guy” (Business Week). In the first 6 months of this year, we have collected a large number of stock-related microblogging messages containing the dollar-tagged ticker symbol of an S&P 100 company. The figure shows the distribution of messages throughout the day. We observe a significant spike in message volume before the markets open. The majority of tweets are posted during the trading hours between 9:30 am and 4:00 pm. This provides further evidence that stock microblogs are truly used by financial professionals to exchange relevant trading ideas in real time and when it matters. They truly capture the market conversation.

 

Research on stock microblogs featured at ICWSM

TweetTrader.net and the related research got some attention at the International Conference on Weblogs and Social Media (ICWSM), which took place in Washington this May. At the time, the working title for the precursor to TweetTrader.net was stockTUitter.com, to pay hommage to the TU School of Management. Media consultant Koki Smith featured the interview and further detail regarding our research on her blog.

Social media as information markets

"Is all that talk just noise? Predicting the future through social media"

Ever since the launch of Twitter, the information stream on this platform has grown exponentially. The site is no longer used only for personal status updates, but represents the fastest way to distribute new and valuable information. In the context of the social semantic web, there are numerous attempts to aggregate this information in a meaningful fashion.

Last year, a group at the TUM School of Management has launched the research project TUitter. As part of this project, we investigate in how far the information content on Twitter can be used as an indicator of events in the offline world. In the context of the German federal elections, we have already shown that the information content of Twitter messages can serve to predict election results and reflect the political landscape surprisingly well (for details, read the paper or watch the presentation at the International Conference on Weblogs and Social Media, which took place in Washington in May). Initial results to use Twitter as an indicator of financial market activity are equally promising. As part of this analysis we are using machine learning techniques to classify message board content automatically and extract the sentiment contained in the postings.

Currently we are transferring these research results into an online application that harnesses this "wisdom of crowds" and may help aggregate stock-related financial information. It provides the online community with an innovative approach to generate fresh trading ideas and investment advice.