Who needs polls? A study published Monday reports that campaigns could use Twitter to successfully predict the winner of most races, findings that might bolster the social media service's already robust political presence.
The key measure, researchers from Indiana University found, was a candidate's "tweet share," the percentage of total tweets about a race that mention them. The more often a candidate is mentioned on Twitter relative to their opponent, the study reported, the greater their chance for victory.
The findings were comprehensive: An analysis of tweets from the 2010 midterm elections found the data correctly predicted the winner in 404 of the 406 House races.
"We plotted it and thought, 'Holy moly, it was a very strong correlation,' " said Fabio Rojas, a sociology professor at Indiana and one of the study's coauthors. He added that preliminary analysis of last year's congressional elections show similar results.
The findings rest on two important points: The raw number of tweets about a candidate doesn't matter, and neither does whether the tweets are positive or negative. Rojas and his colleagues, who collected hundreds of thousands of tweets from the 2010 race, initially measured the total number of times the candidate was mentioned, but the findings failed to correlate with which candidate won. Well-known candidates, like Rep. Michele Bachmann, R-Minn., or candidates from bigger and wealthier districts would inherently receive more attention.
It was then, Rojas said, they realized that what mattered was the Twitter "horse race," or the number of tweets a candidate earns vis-à-vis his or her opponent. Just as the candidates would compete for a limited percentage of the vote, they would also compete for a limited percentage of the total Twitter traffic.
Perhaps most interesting, whether the tweet praised or criticized the recipient was irrelevant. When it comes to Twitter and politicians, apparently all publicity really is good publicity.
"Are you going to talk about the guy who loses or the guy who wins?" Rojas asked. "You're going to talk about the winner, even if you hate the winner."
He added that although campaigns could seemingly skew the results by paying social media directors to tweet or by asking volunteers to pitch in, such a problem has not yet arisen. Such efforts are usually canceled out by similar action taken by their rival, he said.
In Rojas's view, the findings should revolutionize how campaigns conduct themselves. Rather than spending hundreds of thousands, or even millions, of dollars on surveys, campaigns could simply gauge their status on Twitter. That should help campaigns with fewer resources compete with well-heeled incumbents, he said.
"The point is, it's cheap," he said. "Once you start up software for collecting tweets, it's very cheap. It took one of my Ph.D. students a couple of weeks to set it up."
Of course, professional polling isn't likely to disappear from politics any time soon. For one, it's used for more than just the horse race—campaigns test a variety of things with polls, including their message. Twitter doesn't offer help that way. And few politicians would be willing to switch off from a battle-tested pollster in favor of a technology unproven in the heat of a critical race.
But Twitter is also often derided as a hangout for political and media elites, producing a debate that bears little resemblance to the thoughts and opinions of most voters. This study suggests that the social media outlet does provide an accurate reflection of the electorate.
Guy Harrison, a former executive director of the National Republican Congressional Committee, said he'd have to see a Twitter analysis produce accurate results in a competitive race before putting his faith in it. But the service's importance to the political debate, he said, is beyond question at this point.
"Social media and digital media across the board is here to stay, and it's going to be here a long time," he said.
Play of the Day: The American Politician
This article appears in the August 13, 2013 edition of NJ Daily.