Facebook Knows How to Trick You Into Voting
Peer-pressure on the social network drove 340,000 people to the polls in 2010, a new study has found.
Contagion is a funny thing. Carly Rae Jepsen is catching (248 million YouTube views and counting), as are West Nile virus (1,993 cases as of last week) and iPhone 5 hysteria (larger screen! Built-in soft-serve dispenser!). Obamania proved infectious in 2008, but the 2012 strain is markedly less intense. Romneymania, for its part, has yet to sweep the land.
As it turns out, civic participation also exhibits viral qualities: In an article that appears this week in Nature, political scientists at the University of California, San Diego, demonstrate that a single Facebook message during the 2010 midterm elections drove some 340,000 voters to the polls -- simply because their friends had voted, too.
James Fowler and his colleagues partnered with the social-media site to develop what he describes as perhaps the largest randomized, controlled trial ever performed: on Election Day, 2010, nearly every American user over the age of 18 who logged onto Facebook -- some 60 million people -- saw a message at the top of their News Feed encouraging them to go vote. A link was provided to their polling location, as were an "I Voted" button and prominent photos of six friends who'd already done so. But the researchers had worked out two control groups with Facebook, each 600,000 people large. One control group saw no civic message at all. The second was encouraged to vote, but without any of the social pressure markers -- no photos of friends, no reminder that "Jaime Settle, Jason Jones, and 18 others" had done their democratic duty.
When the researchers later compared state voter rolls with Facebook users who'd been targeted (a subset of all voters, 217 million of whom were eligible to vote that year), they found that civic participation was infectious among friends. Not only was Sara on Facebook more likely to go vote if she'd received social pressure through the site to do so, but Sara's friends and friends of Sara's friends were more likely to go vote, too. The subtle encouragement, or guilt, rippled across two degrees of Facebook separation.
"There has been a lot of interest in how online behavior affects other online behaviors," Fowler says. "There has been a lot of interest in how real-world behaviors affect other real-world behaviors. What we have shown here is that those two worlds are not separate: the online world and the real world affect one another. And in this case, we find that this message that started online, that spread online, actually affected real-world behavior. It got a third of a million people to the polls."
He points out, too, that the message was only contagious between close friends. Pressure from casual acquaintances didn't induce people to vote, but pressure from intimates did. "What our research makes abundantly clear is that that's where you get most of your effect. For every one person who you get to the polls directly, you get four additional friends" from social contagion.
Strategists have long wondered just how effective Get Out the Vote appeals really are. The evidence is fluid, and a little dispiriting. A 1999 study, from Yale political scientists Alan Gerber and Donald Green, showed that door-to-door canvassing (by handsomely paid grad students) produced a six percent uptick in voting. A more recent study, from the University of Notre Dame's David Nickerson, asked, "Does Email Boost Turnout?" Thirteen field trials and 232,000 subjects later, the answer was a resounding, "Nope." Nickerson has done other work suggesting that Get Out the Vote appeals are contagious within households, between spouses, for instance, and further Yale research indicates that homeowners are induced to vote when told that their neighbors have done so -- probably because they fear being outed as bad civic participants.
Fowler's Facebook study indicates that online social messaging falls somewhere between email (cheap, useless) and door-to-door canvassing (pricey, powerful) in its impact, combining the sheer scale of the Web with the pressure of human relationships.
Fowler and his usual collaborator, Harvard's Nicholas Christakis, are famous (and in some circles, infamous) for their work on contagion and social networks. Using data from the Framingham Heart Study, a longitudinal study that's followed one Massachusetts town and its offspring since 1948, the pair have examined the "spread" of everything from obesity and happiness to smoking and depression among neighbors, families, co-workers, and even friends (the real-life kind). Their best-known work, published in the New England Journal of Medicine in 2007, concluded that a person's chance of becoming fat increased by more than fifty percent if they had fat friends, and to a slightly lesser extent a fat spouse or siblings, but they've also studied the way that adolescents induce one another to smoke pot and the contagiousness of cooperation among Tanzanian hunter-gatherers.
Fowler's research has its critics (untangling correlation and causality among social networks is messy and imprecise) but he maintains that the conclusions he and his colleagues draw in the Facebook study are, if anything, low-ball figures. Some users saw the News Feed message too late in the day to vote, and others went by different names on their profiles (Larry) than on the voter rolls (Lawrence), and so weren't counted by researchers. "I think if you added all these things up," he says, "you'd find that this was actually one of the stronger Get Out the Vote messages that we've seen in the literature."
The next step, for academics and campaigns alike, will be to isolate "super users": Democrats and Republicans with exceptional influence among their online friends. Fowler was only able to crunch data at the group level, owing to privacy concerns. But, he says, "I suspect, like other network effects, that what you would see is a skewed distribution with a fat tail, where most people didn't effect anyone, and there were a small number of people who had a very large effect on a large number of people."
Both the study authors and Facebook are quick to argue that, in the end, their Get Out the Vote appeal favored neither party. "It looks like the message got just as many Democrats to the polls as Republicans to the polls in this particular instance," Fowler says. That's a little hard to believe, as Facebook users are mostly young people, and young people tend to be more liberal. But fewer than two percent of Facebook users self-report their political ideology, so the data are weak and it's tough to extrapolate. And given that more than 90 million Americans cast ballots in the 2010 midterm, it seems improbable that 340,000 votes -- spread across the entire country -- did much to swing any races. Down the line, of course, that could change. Even if Democrats and Republicans are equally susceptible to social contagion, Democrats are fishing from a much larger pool on Facebook. A strategic Election Day blast by the Obama campaign would reach a lot of young, supportive eyeballs -- far more than a Romney one.
Fowler, for one, seems pleased to have nudged the course of democracy, however slightly. "A lot of times, people will criticize these online social network studies by saying, 'Well, you're just studying the behavior of playing Farmville.'"