Democrat L. Douglas Wilder was supposed to cream Republican Marshall Coleman in the 1989 race for the governor's mansion. He was up significantly in the polls before the election and exit polls indicated a 10 percent lead.
By the end of election night, though, his margin had evaporated to just 0.4 percent, or fewer than 7,000 votes.
The phenomenon — where voters are more supportive of African-American candidates in polls than in voting booths — has become known in political circles as the "Wilder effect." Whether Democratic Sen. Barack Obama becomes the first African-American president — bolstered by last week's upset win over Hillary Clinton in the Iowa caucuses — his time in the public spotlight has generated considerable polling data and discussion, bringing the Wilder effect back in vogue and prompting academics and political pundits to debate the strength of its hold.
In 1989, nobody saw it coming save Paul Goldman, Wilder's longtime political Svengali, and Wilder — now Richmond's mayor — himself. Thanks in part to advice from campaign pollster Michael Donilon, who went on to advise John Kerry's 2004 bid for the White House, Goldman assumed that anything short of a definitive commitment of support from a white poll respondent couldn't be trusted and weighted his poll's statistical model accordingly. His poll indicated a virtual tossup, putting Wilder's chances at 50-50.
Polls from The Washington Post had him leading by 11 percent; the Richmond Times-Dispatch guessed he would win by 9 percent. Most promising of all, exit polls taken as voters left the precincts indicated Wilder held a huge 10 percent lead.
"This was a historic campaign," Goldman says. "Everybody was talking about it — race, race, race — so you give whatever answer's the socially acceptable one."
Still, the gap between the early polls and Wilder's slim margin of victory surprised many, including Scott Keeter, who was running Virginia Commonwealth University's Commonwealth Poll in the late 1980s. His research bore similar results, showing Wilder in the lead outside the margin of error. Since then, Keeter's become the director of survey research for the Pew Research Center.
"It's something that all of us in the polling business have wondered about, if [the Wilder effect] was still around, but we haven't had to confront it since," Keeter says, because of a lack of high-profile African-American candidates running for statewide and, obviously, national office. There were plenty of black-versus-white matchups in 2006 in which the Wilder effect didn't appear to play a role, and then there was Obama's Iowa victory.
"The pattern was that we didn't see the overstatement of the support for the black candidate," Keeter says.
In a Pew study released last year, Keeter notes that in the late 1980s and early 1990s, Wilder wasn't the only one who polled significantly higher than his final draw.
The same year Wilder squeaked by Coleman, David Dinkins was running for mayor of New York City against Rudy Giuliani. One New York City poll gave him an 18 percent lead; another, 14 percent, but he won by only 2 percentage points. Similar patterns emerged in the mayoral races in Los Angeles and Chicago in 1982 and 1983.
In 2006, Ohio, Pennsylvania and Massachusetts all had lopsided governor's races with African-American candidates. Those races had clear favorites, and the polls in the run-up to those elections were pretty much on target. The close race was in Tennessee for U.S. Senate.
"The Tennessee race was very interesting," Keeter says, where black Democrat Harold Ford Jr. lost to white Republican Bob Corker. "You had a moderate-conservative Democrat running against a Republican who was within the mainstream of the Tennessee Republican Party, and a lot of people thought it was a place where we'd maybe see the Wilder effect again."
It's not that race didn't play into the campaign. It did. Corker ran an attack ad against Ford with a ditzy blonde telling the camera, "I met Harold at the Playboy party," then inviting the candidate to call her. The ad was widely viewed as unfairly playing on old stereotypes of a black man unable to resist the temptations of white women. But in this case, the polls didn't appear to be tainted by the Wilder effect, because, well, Corker was predicted to win the election by a slim margin — 2 percentage points — and did.
"I just think that Americans are more comfortable with the idea of black elected officials," Keeter says. Or, in Ford's case, feeling unencumbered enough to admit they weren't voting for a candidate just because he or she was black.
Keeter says public attitudes about race have changed more than the polling techniques, wherein statistical modeling has remained largely the same.
"There's greater acceptance of racial diversity," he says. "We've been asking over the years if Americans believe in blacks and whites dating, and we've seen increasing approval among blacks and whites over time." Keeter says there's also a generational component, with younger people indicating they're more comfortable with racial diversity than their parents were.
So does that mean Obama is in the clear?
Polling is a notoriously slippery business, says Daniel J. Palazzolo, associate professor of political science at the University of Richmond. "We don't want to attribute too much of a historical change," he says. "Just because we didn't see too much of a Wilder effect in Tennessee, we still don't know. It depends on the size of the electorate, how race has been used as an issue in the campaign, how a candidate responds to it. [If] Obama holds out for that effect, the potential is there. We don't really know."
If there is historical change, he says, it could be among African-American respondents, too. "To what extent is it going to happen in reverse, with African-American voters saying they will [vote for a black candidate] and then not? I don't know," he says. "I wouldn't put it to bed." S