Using Twitter to Predict the Stock Market?

Next Story

Google Spends Millions Lobbying US Government

Traditionally, it’s been considered a 50-50 gamble to predict the stock market’s fluctuations with any accuracy.  Some traders have made a name for themselves by consistently getting slightly better odds than this, but few can claim an 87% or better prediction rate.  Except for Twitter.

Yep, Twitter.  Well, more accurately, the Google-Profile of Mood States (GPOMS) can.  That is, if some people at Indiana University are right – and their study of over 10 million tweets from 2008 shows that they might be.

GPOMS has six “mood states” it pulls from the live Twitter stream: happiness, kindness, alertness, sureness, vitality, and calmness.  What Johan Bollen and friends at Indiana University were interested in was the “calmness” stream.

Going through 9.7 million tweets from March to December of 2008, extracting the moods, and correlating that with the stock market, these researchers found that the calmness index matches a change in the market between 2 and 6 days later.  They were able to predict changes (up or down) in the Dow Jones Industrial Average with an 87.6% accuracy.

Let’s see Merrill Lynch pull that off.

There are some questions about the data used by the team, however.  First, Bollen’s team used global tweets instead of filtering down to U.S.-only tweets.  This may, however, mean that they could be even more accurate rather than less so; and, as most people should be aware, the stock market isn’t really local anyway.

The big one, though, is how this mechanism actually works.  What is the reasoning behind the apparent correlation between Twitter moods and the market?  There’s no hard-and-fast answer, but I have a suggestion.

Most people who trade stocks are, in all likelihood, also socially connected through sites like Twitter.  Since it’s been well-established by economists and even sociologists that the public mood often dictates the movement of the market more than any other factor, it would make sense that these moods, as translated to social media, would go along with market changes.

What I can’t understand is the long delay between the Twitter mood and the market itself.  2 to 6 days is an eternity in Wall Street, where the average stock is held for a whopping total of 11 seconds.

So who’s to know, really?


TEKTRONIX  MAINFRAME / CHASSIS TM503 (B1) picture
TEKTRONIX MAINFRAME / CHASSIS TM503 (B1)
Tektronix R7103 1GHz 3-slot Analog Oscilloscope mainframe, superb condition picture
Tektronix R7103 1GHz 3-slot Analog Oscilloscope mainframe, superb condition
NH Research 4700 NHR High Current DC Load Mainframe with all  modules  picture
NH Research 4700 NHR High Current DC Load Mainframe with all modules
National Instruments SCXI-1000 Mainframe SCXI-1102, 1180, 1161, 1124 Modules picture
National Instruments SCXI-1000 Mainframe SCXI-1102, 1180, 1161, 1124 Modules
Tektronix TLA 714 Logic Analyzer Color Portable Mainframe WIN2000 PRO picture
Tektronix TLA 714 Logic Analyzer Color Portable Mainframe WIN2000 PRO