No matter how much ground the economy has gained (or lost) since 2009, President Obama’s electoral fate will depend on how the economy performs during the final six months before the election. And voters’ implicit racial attitudes may prove to be a significant hurdle for Obama, perhaps shaving as much as 10 percentage points off his otherwise expected voting performance.
Those were two of the main messages from a University of Virginia conference on “The Political Unconscious and the 2012 Election,” held Friday in Garrett Hall’s Great Hall.
The keynote speakers were political scientist Larry Bartels of Vanderbilt University and psychology professor Tony Greenwald of the University of Washington, two “giants in their respective fields,” said Paul Freedman, associate professor of politics in the College of Arts & Sciences and a member of U.Va.’s Political Psychology Working Group, which organized the conference.
Speaking to an audience of about 50, Greenwald shared his findings on how voters’ implicit racial attitudes may impact the presidential election. His research was based on systematic review of the prediction errors in polls leading up the 2008 primary elections and general election, in combination with results of the Implicit Association Test that he co-developed with Mahzarin Banaji, a Harvard University psychology professor, and Brian Nosek, a U.Va. associate professor of psychology.
After the 2008 election, Greenwald explained, pollsters patted themselves on the back for accurately predicting Obama’s margin of victory, concluding that Americans no longer had any significant disinclination to vote for a black candidate that they might be reluctant to share with public opinion surveyors. The phenomenon of voter reluctance to share racist attitudes is known among political scientists as the Bradley/Wilder effect, named for African-American candidates in two elections where the effect was first observed.
The polls, looked at in aggregate, were indeed very accurate, Greenwald said. But when analyzed more closely, state-by-state, the polls were actually significantly less accurate than in other recent elections, with substantial errors in two separate directions that offset each other: blacks cast votes for Obama more frequently than polls predicted, while whites cast less votes for Obama than predicted.
In general, among the states that Obama carried in the 2008 election, polls over-predicted Obama’s margin of victory, with larger error rates than observed in similar polls ahead of the 2004 election, implying that something undisclosed to pollsters was involved, Greenwald said.
With a hunch that racial attitudes were the culprit, Greenwald examined results from the Implicit Association Test among 2,200 eligible voters who took the test in the few days before the 2008 election. The test asked the voters to pair "good" and "bad" words with images of black, white, Hispanic and Asian faces, and the speed at which one sorts and categorizes is an indication of how strongly those concepts are associated in the mind. The test has enabled observation of unconscious attitudes and has revamped understanding of stereotyping and prejudice.
Greenwald found that while candidate preferences were predicted strongly by one’s self-described liberalism or conservatism, they were predicted moderately by implicit racial attitudes.
Greenwald reviewed other published studies that suggest racial attitudes influence voters, including one that found that the volume of “racially charged” Internet searches (by state) predicted Obama’s underperformance relative to Hilary Clinton in the 2008 Democratic primaries.
Taking stock of all the studies, Greenwald concluded that “race may be the cause of at least a 10 percent handicap for Obama,” meaning that if polls predict the race is a dead heat, the actual vote can be expected to give Romney a 55 percent to 45 percent victory.
However, when questioned from the audience, Greenwald offered a caveat to that prediction, noting that there may be an as-yet-unidentified variable that can explain the observed correlation between candidate race and voting preference. "I’m not saying the Bradley effect exists," he said.
In separate remarks, Bartels, co-director of Vanderbilt’s Center for the Study of Democratic Institutions and May Werthan Shayne Chair of Public Policy and Social Science, made a case that voters do not behave as if they were fully informed, and voters are heavily swayed by their perception of the economy’s performance in the months leading up to the election.
Historically, Bartels explained, when voters perceive that the economy is improving or doing well as Election Day approaches, the incumbent party generally does well at the polls, and vice versa when the economy is perceived to be in a slide.
Election-year income growth is a much stronger election outcome predictor than cumulative economic growth over the previous four years, he said.
"Voters are myopically focused on current conditions," forgetting or ignoring the recent past – such as how bad the economy was when Obama took office, Bartels said.
Because of this voter focus on short-term income growth, it does not matter that from the late 1940s to the present, average annual income growth for middle-class and lower-income Americans has been overwhelmingly better under Democratic presidents than under Republicans. However, the economy’s shifts during election years – the only performance that matters to voters – have often advantaged Republican presidential candidates.
In conclusion, Bartels noted that election outcomes have very significant policy consequences, but those election outcomes are shaped much more by short-term income growth and other idiosyncratic factors than by any "romantic notions of how democracy actually works," such as voters voting on the basis of public policy positions or governmental performance.
The conference, the first organized by U.Va.’s Political Psychology Working Group, was sponsored by U.Va.’s Page-Barbour and Richard Lecture Series, the Frank Batten School of Leadership and Public Policy, the Woodrow Wilson Department of Politics, the College and Graduate School of Arts & Sciences and the Office of the Vice President for Research.