Starting in 2008, Facebook started nudging people to vote by posting special reminders that say: “Today is Election Day.” These messages include a button that Facebook members can click to indicate they have voted—information that is in turn shared automatically with their friends.

New research suggests these messages have striking real-world power. On Election Day 2010, these reminders, when adorned with faces of those who’d clicked “I voted,” drove an additional 340,000 voters to polls nationwide, according to research conducted by Facebook and academics, and published today in Nature.

James Fowler, a political scientist at the University of California, San Diego, who led the research says “there were one-third of a million people who actually showed up at the polls who wouldn’t otherwise have if the message hadn’t been shown.” The reason, he says, was straightforward. “Seeing the faces of friends accounted for all this effect on voting. And it affected not only people who saw them, but friends, and friends of friends.”

Teasing out the effect of just one message is a challenge. The run-up to Election Day is always crammed with appeals to vote. What’s more, about one-third of voters don’t actually vote on Election Day, but do so earlier with absentee ballots. To isolate the effect of the Facebook message, Facebook’s data-science team (see “What Facebook Knows”) worked with the social scientists to run the experiment on all of the 61 million voting-age Americans who visited Facebook on November 2, 2010. 

Those 61 million people were divided into three categories. The first group, which had around 60 million people, got a message that said “Today is Election Day”; it featured a round red-white-and-blue symbol that said “Vote” and offered help finding a local polling place. Importantly, it also included the faces of up to six friends and the information that these friends had clicked an “I Voted” button. The second group, with 600,000 people, got the same message, but with no information from friends—just the nudge to vote and a way to find a polling place. The third group, the control, had 600,000 people who got no message at all.

Then Fowler and colleagues followed up with shoe-leather work to study available state-by-state polling records to determine who had actually shown up to cast a ballot. The researchers found that the first group voted at a slightly higher rate than members of the other two groups, enough to translate into 60,000 additional votes cast.

The rest of the increase was attributed to social contagion. People who had seen the messages with faces of their friends were also more likely to themselves click the “I Voted” button. This cue spread, in turn, to their friends. Those friends also voted at higher rates, translating into another 280,000 votes. 

The authors added that, if anything, they understated the effect of the Facebook messages, because they could include data only on people whose identities could be definitely matched to voting records.

A political strategist was struck by the result. “It certainly validates the idea that social media can have an enormous effect on voting,” says J.D. Schlough, a Democratic political operative who designs online advertising campaigns (see “Facebook: The Real Presidential Swing State”). “In politics, a face-to-face interaction between family or neighbors or friends is the Holy Grail of persuasion. The idea that people trust information from trusted sources, and act on it, is intuitive. But this study nails that to the wall using Facebook.”

Was the increased voting enough to throw someone out of office? Facebook’s voting messages are nonpartisan, and the authors says the increased voting effect was the same on a subset of self-reported conservatives as on self-reported liberals. But one day Facebook’s get-out-the-vote efforts could have a bigger effect. Nationally there are far more registered Democrats than Republicans—72 million to 55 million. (Another 42 million are registered as Independent.)  “Increased voter turnout is generally associated with favoring Democrats,” Schlough notes. And slim margins can have big consequences. The 2000 presidential race was decided in favor of George W. Bush by 537 votes in Florida. He sealed his 2004 reëlection battle against John Kerry by fewer than 120,000 votes in Ohio.

Cameron Marlow, a Facebook data scientist who participated in the study, wouldn’t say whether Facebook plans to again urge people to vote in the upcoming election, but he did not rule out a repeat effort. “In general, we are committed to being part of the democratic process,” he says.

The study adds to evidence that online messages in social media can strongly affect offline actions. Earlier this year, Facebook founder Mark Zuckerberg decided to use his company’s network to encourage more people to enroll as organ donors. A clickable box appeared on Timeline pages to let people indicate that they were registered donors, and such clicks showed up in notifications to friends.  This campaign was associated with a huge boost to donor enrollments, which increased by a factor of 23 in 44 states.