CoauthorMasa Aida, Greenberg Quinlan Rosner Research
When assessing the effects of negative advertising on political engagement, or studying how political campaigns influence the behavior of voters, or conducting political polls, understanding people’s voting intentions is crucial. But despite its importance to political science and practice, the accuracy with which we can measure voter intention to vote has been surprisingly understudied, write Todd Rogers, assistant professor of public policy, and Masa Aida, of Greenberg Quinlan Rosner Research, a polling and political strategy firm.
They found that past voting behavior was a better predictor of whether someone would vote than a person’s own predictions.
The researchers analyzed six pre-election surveys with post-election vote validation, five from the 2008 presidential election and one from the 2009 New Jersey general election. They were able to survey more than 29,000 people who had previously been asked by pollsters about their intention to vote.
In their analysis of 2008 general election respondents, Rogers and Aida found that 54.8 percent of those who said they would not vote actually did vote, and that 13.3 percent of those who said they would almost certainly vote did not vote. (The results were the same whether a voter was contacted closer to Election Day — and maybe more caught up in the excitement of an election — or earlier in the election season.)
The 2009 New Jersey general election saw less than half of voters turnout. Poor self-prediction among voters was striking: nearly 30 percent of those who said they would not vote did in fact vote, and 54.2 percent of those who said they would vote did not.
"Both forms of inaccurate self-prediction follow a similar pattern: people are more accurate when predicting they will behave consistently with their past behavior than when predicting they will behave inconsistently with their past behavior," the authors write.
Their findings have important implications. They argue that political scientists should explicitly address voting history in their analysis of voter intention.
Their research also suggests that because many voters who express no interest in voting actually end up voting, setting deadlines for voter registration ahead of an election, as many states do, may reduce turnout. "Registration in advance of an election requires that citizens anticipate their interest in casting a vote in an election," the authors write, "and the results we report show that people’s ability to do this is limited."
Political pollsters, too, should take note, since screening likely voters by asking them about their intentions is "grossly inadequate."
"The present data suggests that when calling from a list of registered voters, a hybrid approach incorporating both vote history and self-report could substantially increase predictive accuracy," Rogers and Aida write.
Why did so many of the voters studied inaccurately predict their voting behavior? The data discount two possible explanations: that voters are especially bad at predicting if they will vote when Election Day is far away, or that they tell pollsters they don’t plan to vote simply to end the phone call.
But the authors offer some other hypotheses. Respondents may have underestimated how much others’ excitement about the election would affect their own behavior. The influence of, for example, lots of friends voting, may have pushed them toward the polling station. The researcher’s results, which showed higher turnout in the 2008 presidential election than in the 2009 New Jersey election among that those who predicted that they would not vote, supports this idea.
Voters may also have said they wouldn’t vote as a way of venting their disaffection with the political process, rather than expressing a genuine lack of desire to vote. And voters may also have been genuinely bad at predicting their own behavior, even though they were fully aware of their past behavior.
"For reasons we do not understand, respondents seem to not correctly access or weight that information when predicting whether or not they will vote," the authors conclude.
— by Robert O'Neill