Read our COVID-19 research and news.

'Social Voting' Really Does Rock the Vote

Brace yourself for a tidal wave of Facebook campaigning before November's U.S. presidential election. A study of 61 million Facebook users finds that using online social networks to urge people to vote has a much stronger effect on their voting behavior than spamming them with information via television ads or phone calls.

The study comes hot on the heels of a Science paper originally published online on 21 June that tracked how people influence each other's online behavior through Facebook. A lingering question remained: Does that online social influence translate to real-world behavior when people step away from the computer? The challenge is to find online and real-world behaviors where cause and effect can be teased out with controlled experiments on a large enough scale.

In the spring of 2010, a golden opportunity fell into the lap of James Fowler, a social scientist at the University of California, San Diego. He had recently been introduced to Cameron Marlow, the director of a new "data science" team at Facebook's headquarters in Menlo Park, California. Fowler wondered if he could create an experiment using Facebook's giant social network. It just so happened that the U.S. congressional elections were coming up in November of that year, and Facebook was planning on rolling out a nonpartisan "Get Out the Vote" campaign, reminding people to vote by publishing a message on Facebook users' news feeds. With just a few tweaks to how those messages were published, the campaign could be converted into a massive controlled social science experiment: With less than 40% of eligible U.S. citizens normally voting in congressional races, would the Facebook campaign have an impact? Since no personal data of Facebook users would be released, Marlow agreed. Fowler's Ph.D. student Robert Bond led the research team.

On Election Day, about 60 million people received a message that encouraged them to vote. It included links to local polling stations, a clickable "I Voted" button, and photos of six of their randomly chosen friends who had already clicked the "I Voted" button. Two control groups, each containing about 600,000 people, either received a version of the message with voting information but no photos of their friends, or no message at all. Then, to track who actually voted in the election, the team matched people's names and birth dates with those in the official state election rolls. If the influence of Facebook friends extends beyond the Internet, then seeing the profile photos should translate to voting out in the real world.

The photos apparently worked: People who received messages alerting them that their friends had voted were 0.39% more likely to vote than those who received messages with no social information. That translates to an additional 282,000 votes cast, the team reports online today in Nature. The effect was four times stronger than just seeing the voting message without photos of friends, and most of that boost came from the people's closest friends (judging closeness by the frequency of interaction on Facebook).

The study is "both significant and convincing," says Dylan Walker, a social scientist at Boston University School of Management. The next step, he says, is to see what kinds of relationships matter most. "For example, I have different types of friendships with my online peers that go beyond the distinction of casual versus close. Some are work colleagues that I see on an everyday basis; others are old college friends; yet others are high school peers with whom I seldom engage offline but whose updates I read on a regular basis. Do they influence me in different ways? Absolutely."