December 18, 2007
Study suggests polls may overestimate support for Obama, underestimate backing for Clinton among Democrats
A new national study of voters who say they might vote in Democratic primaries and caucuses shows a striking disconnect between their explicit and implicit preferences, according to University of Washington researchers.
When asked who they would vote for, Sen. Barack Obama held a 42 percent to 34 percent margin over Sen. Hillary Clinton. Former senator John Edwards was in third place with 12 percent. However, when the same people took an Implicit Association Test that measures their unconscious or automatic preferences, Clinton was the runaway winner, the favored candidate of 48 percent of the voters. Edwards was second with 27 percent and Obama had 25 percent.
Bethany Albertson, a UW assistant political science professor, and Anthony Greenwald, a UW psychology professor and inventor of the Implicit Association Test, emphasized that their participants were not a representative sample of Democrats but were self-selected volunteers who took an experimental test over the Web. The data came from 926 people age 18 and over who took the test between Oct. 16 and Nov. 5. Of that total, 687 people said they might vote in the Democratic primaries.
“In the past, poll numbers have often overestimated support for black candidates when compared to their actual vote percentages,” said Albertson. “Findings of this study suggest that this familiar pattern may be about to repeat itself in the 2008 Democratic presidential primaries.”
“What is new here is a pre-election indicator that this may happen,” added Greenwald. “We don’t know what will happen in the Iowa caucuses when people who say they favor Obama have to convince other participants. And we don’t know if some of those participants in the caucuses who say they are planning to vote for Obama will end up choosing Clinton.”
Albertson said implicit preferences may shape the way voters take in new information as the presidential campaign develops.
The Implicit Association Test was developed nearly a decade ago to measure the unconscious roots of people’s thinking and feeling. Since it was created, more than 6 million people have taken versions of the test that have measured unconscious attitudes about such topics as race, gender, sexuality and various ethnic groups. The test is widely used around the world by psychological researchers to probe people’s attitudes.
In the researchers’ new version participants took the test, which randomly administered a measure of implicit attitudes for either an Obama vs. Clinton race, Edwards vs. Clinton or Obama vs. Edwards. Then they were asked a number of explicit questions, including how warmly they felt toward the candidates, how likely they were to vote and whom they favored.
Albertson and Greenwald said the disconnect between implicit and explicit preferences for Obama held up for both white and black participants as well as for both men and women. Just under 70 percent of the participants in the study were female. Whites made up 72 percent of the sample while blacks numbered 10 percent.
“The result doesn’t disappear when we just look at men or women or when we look separately at black and white voters. This gives us greater confidence that these results have meaning,” said Greenwald.
Individuals can take a similar Implicit Association Test for either the leading Democratic or Republican candidates on the Web at https://implicit.harvard.edu/implicit/demo/featuredtask.html. These versions of the tests allow participants to examine their preferences for Democratic candidates Clinton, Obama, Edwards and Bill Richardson and Republican hopefuls Rudy Giuliani, Mike Huckabee, John McCain and Mitt Romney. As the primary season proceeds, the tests will be revised to follow the top four candidates for each party.
Funding to create the primary election Implicit Association Test was provided by the University of Washington and Project Implicit, a collaborative research effort of researchers at the UW, Harvard University and the University of Virginia.
###
For more information, contact Greenwald at (206) 543-7227 or agg@u.washington.edu, Albertson at (206) 295-8803 or balberts@u.washington.edu .
