Research Article: Decisions among the Undecided: Implicit Attitudes Predict Future Voting Behavior of Undecided Voters

Date Published: January 29, 2014

Publisher: Public Library of Science

Author(s): Kristjen B. Lundberg, B. Keith Payne, Brock Bastian.


Implicit attitudes have been suggested as a key to unlock the hidden preferences of undecided voters. Past research, however, offered mixed support for this hypothesis. The present research used a large nationally representative sample and a longitudinal design to examine the predictive utility of implicit and explicit attitude measures in the 2008 U.S. presidential election. In our analyses, explicit attitudes toward candidates predicted voting better for decided than undecided voters, but implicit candidate attitudes were predictive of voting for both decided and undecided voters. Extending our examination to implicit and explicit racial attitudes, we found the same pattern. Taken together, these results provide convergent evidence that implicit attitudes predict voting about as well for undecided as for decided voters. We also assessed a novel explanation for these effects by evaluating whether implicit attitudes may predict the choices of undecided voters, in part, because they are neglected when people introspect about their confidence. Consistent with this idea, we found that the extremity of explicit but not implicit attitudes was associated with greater confidence. These analyses shed new light on the utility of implicit measures in predicting future behavior among individuals who feel undecided. Considering the prior studies together with this new evidence, the data seem to be consistent that implicit attitudes may be successful in predicting the behavior of undecided voters.

Partial Text

What does it mean when someone says they are undecided about how to vote just weeks before an important election? Do undecided voters truly have no clear preferences? Do they have preferences, but are still weighing all the information at hand? Or, do they simply lack conscious insight into what their preferences are? Regardless, undecided voters are hard to ignore. Less than two months before the 2012 U.S. presidential election, five percent of voters in swing states were undecided, and another 17% said they might change their minds [1]. These percentages are more than enough to decide most elections, making the prediction of undecided voters’ behavior an important problem for pollsters and social scientists. Is it possible to predict the behavior of individuals who cannot predict their own?

In this first analysis, we began our re-examination of the role of implicit attitudes among undecided voters by predicting voting behavior from implicit and explicit attitudes toward the presidential candidates.

The same ANES data used for the candidate attitudes were also used for the examination of racial attitudes. In the analyses that follow, measures of voting behavior and confidence in the decision were identical to the analyses of candidate preferences described previously. Measures of confidence were selected from the wave that corresponded with the date of AMP administration and, as before, a control variable was included to indicate this factor.

One reason that implicit attitudes should predict behavior among the undecided is that when people introspect about whether they have reached a decision, they attend primarily to consciously endorsed attitudes and neglect implicit attitudes. This does not imply that implicit attitudes are necessarily unconscious, but simply that people tend to consider them to be a less valid basis for judgments [5], [7]. To test the hypothesis more directly, we next examined the relationships between attitude extremity and confidence. We expected that more extreme explicit attitudes would be associated with greater confidence in one’s vote, but that this association would be weaker for implicit attitudes. Statistically, this hypothesis predicts a curvilinear relationship between explicit attitudes and confidence.

The idea that implicit attitude measures can predict the behavior of undecided voters has generated a great deal of interest, because it suggests that precursors of decisions can be detected before respondents feel they have made up their minds. This idea has also generated some disagreement over whether and when such effects might be expected, as some prior studies have found implicit attitudes to predict voting among undecided voters whereas others have not. We re-examined these ideas using a large nationally representative sample and a longitudinal design, with implicit measures of both candidate attitudes and racial attitudes. We found that when implicit and explicit attitudes were highly correlated, as in the case of attitudes toward political candidates, the role of implicit attitudes depended critically on whether explicit attitudes were statistically controlled. When explicit candidate attitudes were not controlled, implicit candidate attitudes were more predictive among decided than undecided voters. However, this finding appears to result from the shared variance between implicit and explicit candidate attitudes. When explicit candidate attitudes were controlled, implicit candidate attitudes were equally predictive for decided and undecided voters. The results for candidate attitudes and racial attitudes both provided evidence that implicit attitudes may predict the behaviors of voters who feel that they have not made up their minds.

Implicit attitudes have been suggested as a key to unlock the hidden preferences of undecided voters. Past research, however, offered mixed support for this hypothesis. The present research found that, when the influence of explicit attitudes was controlled, implicit attitudes predicted voting as well for the undecided as for the decided voters. Implicit tests may offer a useful tool in forecasting elections, especially at early stages when many voters have yet to make up their minds.