How to Tell When 'Yes, I'm Voting' Means 'I'm Going to Flake Out'

Flakers, like bad poker players, have tells.

Saying you will vote doesn't earn you a sticker.

Photographer: MICHAEL B. THOMAS/AFP/Getty Images

Psychologists have finally started to shed some light on flaking out -- that exasperating tendency of people to say they’ll do something and then not do it. People flake out over all kinds of things, from calling you back to taking life-saving medications. Voting is particularly susceptible to this failure of follow-through, says Todd Rogers, a professor of public policy at Harvard’s Kennedy School.

But there, he has found something of a counter-strategy. Flakers, like bad poker players, have tells. And he’s found that pollsters can be surprisingly good at picking these up from subtle cues revealed in a short phone conversation. 

Being able to predict who will actually vote could make polls more informative. Fewer than half of people who say they are going to vote actually show up at the voting booth.

Rogers said he wondered whether this behavior could be predicted, given that human intuition has proven pretty good at picking up personality traits, mood, sexual orientation and tendency toward racial prejudice. To test the idea, he and colleagues teamed up with callers from the organization Get Out the Vote, and asked them to place percentage odds on whether people actually would vote, regardless of what they said they would do.

He said he was surprised by the acumen of the callers’ predictive ability. Of those respondents who said yes, they would vote, the ones the callers predicted most likely to vote proved twice as likely to actually vote as those deemed unlikely to do so. Only 26 percent of those who were judged very unlikely to vote followed through. The results were published in this week’s issue of the Proceedings of the National Academy of Sciences.

“It’s surprising to me because it’s such a short exchange for callers to be able to make useful inferences about whether respondents are actually going to do what they say,” said Rogers. There were some limitations, however. The Get Out the Vote callers systematically erred on the side of optimism. They underestimated the total flakiness of the people they called. Their predictive power was in judging people’s relative odds of flaking.

He also noted that a few of the people who said they would not vote turned around and voted. Callers were not able to do better than random chance at predicting which ones would switch in the direction from no to yes.

Katherine Milkman, an associate professor of operations, information and decisions at the University of Pennsylvania’s Wharton School, said she thought callers were probably picking up on lack of commitment in the people who said they would vote and didn’t. People may flake out because they forget to vote or they procrastinate. Or maybe they never intended to vote, but they felt reluctant to admit that to the callers.

She said she wouldn’t expect the callers to predict the behavior of people who said they weren’t voting and then did vote. That’s not flaking out; that is a change of mind. 

To get a sense of the cues that callers might have been picking up on to predict who would fail to follow through, Rogers and his colleagues set up a second experiment in Texas, where it’s legal to record phone conversations with permission from only one party. The researchers listened to the recordings and analyzed a list of possible nonverbal cues. They found that the people that the callers deemed less likely to vote showed more uncertainty, insecurity and nervousness, and hesitated more before giving an answer.

Since Rogers and his colleagues knew who actually voted, they went back to examine which speech patterns correlated with not voting. There, they found that the callers were correct in identifying uncertainty, insecurity and delayed response, but that nervousness wasn’t a good clue. The researchers also found that those who flaked used more "ums," "uhs" and other speech fillers. Rogers said this information might make people even better at predicting who will fail to vote -- or perhaps fail to fulfill a task like taking medication.

Rogers and Milkman both said that they expect this finding to apply to many other areas of human behavior beyond voting, and that plans were afoot to start broader testing. We’ll be watching to see if they follow through.

This column does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners.

    To contact the author of this story:
    Faye Flam at fflam1@bloomberg.net

    To contact the editor responsible for this story:
    Philip Gray at philipgray@bloomberg.net

    Before it's here, it's on the Bloomberg Terminal.