New York – A study published a few days ago by has shown a correlation between the mood captured on a large-scale collection of Twitter feeds, and the Dow Jones Industrial Average. The study analyzed more than 9.8 million tweets from 2.7 million users during a 10 month period in 2008. Associate Professor at IU Bloomington’s School of Informatics and Computing Johan Bollen says that he found “an accuracy of 87.6 percent in predicting the daily up and down changes in the closing values of the Dow Jones Industrial Average.” The other researcher on the study was PH.D. candidate Huina Mao.
The researchers used two mood-tracking tools to analyze the different twitter feeds. The first was OpinionFinder and the second was Google-Profile of Mood States which measures mood in six categories: calm, alert, sure, vital, kind, and happy. Together the two tools gave the researchers seven daily data points which they could then correlate to the Dow Jones closing values.
The researchers then showed how even a basic prediction model such as a Self-Organizing Fuzzy Neural Network could have its accuracy improved by utilizing the public mood indicators they had been researching. Their hope was that more sophisticated market models would show an even better improvement in accuracy. Indeed, while likely not sufficient in itself for a trading strategy, it is possible that this data would be a good non-correlated input to quantitative trading models.
The most relevant leading factor they found was the level of Calm from the Google-profile of Mood states data. The odds that an accuracy rate of 87.6 percent would show up in their study by pure chance were only 3.4 percent.
Integrity currently has a list of research providers tracking social media, and last month we wrote about a new company called PeopleBrowsr whose approach was focusing on determining the attitude and emotion of various articles on the web including Twitter conversations. We wrote at the time that data such as this needs to be tested to a greater degree before investors can benefit from it in a meaningful way, and perhaps this study goes a little way in accomplishing that testing. Of course, if this study is to be fully believed, the next logical question would still be why aren’t these IU researchers’ millionaires yet?