Polling caller guesses found to be useful in predicting which respondents will follow through on claims
(Phys.org)—A trio of researchers has, by conducting two studies, found evidence that suggests untrained polling callers are able to "guess" with almost 60 percent accuracy, which people they call will actually follow through on claims they have made regarding whether they will vote in an upcoming election. In their paper published in Proceedings of the National Academy of Sciences, Todd Rogers, with Harvard University and Leanne Brinke and Dana Carney with the University of California, describe their two studies and their assessment of possible signals given by respondents that potentially serve as cues to polling callers that give away their true intent.
Some researchers in the polling business have begun to question the accuracy of current polling methods—as new tools for testing accuracy and new technology changes the social landscape, some believe that the results of polls are not very reliable and that news organizations might be reporting inaccuracies that don't become evident until important events have passed. One such important case surrounds elections. More and more groups are seeking to discover which people will vote so that they can be targeted with ads. Oddly, despite the move to cell phones, most polls are still conducted by speaking to people on land lines, which obviously leaves very large gaps in polling. But, there is also the problem of people responding inaccurately, whether intentionally or not, when asked via an unsolicited phone call, whether they will be voting in an upcoming election. In this new effort, the researchers took a unique approach to increasing the reliability of such polls, by surveying the callers and simply asking them whether they thought a particular respondent would in fact vote regardless of which answer they had given.
The researchers conducted two studies, the first involved querying callers working on a 2009 campaign in New Jersey—they simply asked each to guess which voter that said yes they would vote, would actually do so. They then compared the responses to voter records, and in so doing, found the callers were right approximately 58.5 percent of the time. Meanwhile, they also found that only 47 percent of respondents who said they were going to vote, actually did so.
In the second experiment, the researchers listened to recorded calls made by pollsters speaking with respondents prior to an election in Texas in 2010—they found that cues given by respondents, such as pauses and changes in voice, were picked up by pollsters as signs that they likely would not vote despite saying they would. Interestingly, they also found that other traditional cues, such as a person sounding nervous or tense, were not a factor.
More information: Unacquainted callers can predict which citizens will vote over and above citizens' stated self-predictions Todd Rogers, PNAS, DOI: 10.1073/pnas.1525688113 , www.pnas.org/content/early/2016/05/18/1525688113
Abstract
People are regularly asked to report on their likelihoods of carrying out consequential future behaviors, including complying with medical advice, completing educational assignments, and voting in upcoming elections. Despite these stated self-predictions being notoriously unreliable, they are used to inform many strategic decisions. We report two studies examining stated self-prediction about whether citizens will vote. We find that most self-predicted voters do not actually vote despite saying they will, and that campaign callers can discern which self-predicted voters will not actually vote. In study 1 (n = 4,463), self-predicted voters rated by callers as "100% likely to vote" were 2 times more likely to actually vote than those rated unlikely to vote. Study 2 (n = 3,064) replicated this finding and further demonstrated that callers' prediction accuracy was mediated by citizens' nonverbal signals of uncertainty and deception. Strangers can use nonverbal signals to improve predictions of follow through on self-reported intentions—an insight of potential value for politics, medicine, and education.
Journal information: Proceedings of the National Academy of Sciences
© 2016 Phys.org