Twitter may be able to foresee the ebbs and flows of the stock market better than any financial analyst.
The mood of the Twitter-verse can predict swings in the Dow Jones Industrial Average.
Calm or anxious tweets are the best predictors of market performance.
Although this research has only been applied to past market conditions, the study could help investors make a quick buck.
A mere 140 characters can help investors make a quick buck.
By analyzing nearly 10 million tweets over 10 months, Indiana University scientists found they could predict the rise and fall of the Dow Jones Industrial Average -- in some cases nearly a week in advance.
"These online environments are starting to drive real-life social phenomona," said Jonah Bollen, a scientist at IU who, along with graduate student Huina Mao, published a paper on the arXiv preprint server.
A growing body of research is mining information from social media sites like Twitter, Facebook and blogs. Recently scientists have used this information to measure the public's general state of happiness with a "hedonometer" and to predict box office sales of new movie releases by the amount of online chatter they generate.
A tweet here or a blog post there won't make much of a difference. It's the cumulative effect of millions of tweets, blogs and status updates that gives a broad view of the nation's mood.
The researchers intended to find out how anxious, happy, vital, calm, sure and alert people were feeling during a politically and economically volatile year. They were looking to see what kinds of correlations they would find by comparing the performance of the Dow Jones Industrial Average with the mood of the Twitter-verse. The team was able to detect a strong correlation, but only when the Twitter data was shifted back three or four days.
The IU researchers found that only calmness could predict market swings, as determined by two mood-tracking tools: OpinionFinger and Google-Profile of Mood States. Boiled down, these tools look for specific words and phrases that indicate how the Twitter user is feeling.
For example, if a person tweeted "I am happy," that would influence the happiness score. If a user tweeted "I am so sad right now," that would count as especially negative, and move the happiness score down.
The link isn't just correlation either, it's causation, said Bollen. The calmness or anxiety of the Twitter community actually predicts the fluxes in the market, and with nearly 90 percent accuracy.
The IU research is currently limited to events that have already happened. However, scientists will soon turn to future events, said Brian Uzzi, a professor at Northwestern University who also mines social media for information.
"The kind of predictions made by the Bollen group will be ever more finely tuned, and faster," said Uzzi.
One group of researchers is already doing almost real-time evaluation of Twitter, using a powerful Cray supercomputer. Scientists from Georgia Tech and the Pacific Northwest National Laboratory crunched a single day's worth of Twitter messages to find who the movers and shakers of the Twitter-verse are. In about an hour, the group found most tweets are about media or government, or small groups talking amongst themselves.
Couple that kind of computing power with the predictive abilities of Twitter, IM, Facebook and other social networks, and all kinds of possibilities arise.
"The electronic data that has recently become available to scientists gives us unprecedented new clues into the spread of disease, how people should evacuate a city or how the stock market behaves," said Uzzi.