Nate Silver made modeling elections famous, using poll data. Now it seems that the Twitterati are also pretty good predictors of electoral wins.
An Indiana University study analyzed 537 million tweets in the three months prior to elections for the U.S. House of Representatives in 2010 and 2012. The researchers found that in most cases, if a candidate got a bigger proportion of the tweets, they won. And it didn’t matter whether the tweets were “I hate that guy” or “She’s my favorite candidate!” the results could help gauge trends in future elections when polling data is incomplete.
The study’s lead author, Joseph DiGrazia, a doctoral student in sociology, presented the findings at the annual meeting of the American Sociological Association.
DiGrazia looked at 406 races in 2010 and 395 in 2012. In 72 percent of the races the candidate with a bigger tweet-share mentioning his or her name won. DiGrazia said that to get a good sample of people the program that monitored tweets looked for names mentioned in text, not as Twitter handles. People who use the handles to refer to candidates tend to be an “elite” group of Twitter users, he said, and aren’t always representative of the general population. The study also didn’t include third-party candidates, and races in which a candidate ran unopposed. “We didn’t want to add a complicating factor,” DiGrazia said.
In a smaller percentage of the races, more tweets didn’t predict the final outcome. During the presidential primaries, for example, Rick Santorum got up to 40,000 mentions per hour, but that didn’t help him win the Ohio Republican primary. And sometimes candidates were mentioned a lot because of a serious scandal that ultimately sunk their campaign.
But DiGrazia’s Twitter results could help predict Congressional races, which tend to be thinly polled, if at all, at least by the big public polling firms. What polls there are tend to be internal ones, used by the candidates’ campaigns. So there’s a real dearth of good data to build models with, and that’s part of what this study tries to address.
DiGrazia said a future project will be to watch the Twitter traffic and see if it actually does predict results with any accuracy.
Credit: Twitter Screenshot