A team of computer scientists has found that Twitter chatter can be used to predict stock prices, even though their study was not looking to do that. Researchers at the University of California, Riverside, first defined a variable they called connected components, meaning “the number of distinct conversational threads taking place around a company in any given day.” Three possible threads around Apple, for example, could be the iPad, iPhone and the Foxconn factory. The more connected components, “the higher the volume of trading for that stock in subsequent days.”
What’s the Big Idea?
The model also suggested that the presence of more connected components signaled a rise in stock prices, prompting researchers to developed a trading formula that took advantage of this pattern. Their formula did quite well, outperforming the Dow Jones Industrial Average over four months (a 2.2% loss for the Twitter model versus a 4.2% loss for the Dow). So why the connection? Lead researcher Vagelis Hristidis theorizes that when bad news strikes, everyone focuses on that one topic, reducing the number of connected connections. Good news, perhaps being less interesting, allows for more digressions in the conservation.
Here is the comment of astute conservative commentator Yuval Levin on the recent mega-gaffe by Romney’s communications director:If, say, yesterday, you had asked me what kind of statement by a Romney […]