I haven’t been able to find one major pollster who thinks the so-called “Bradley effect,” (the notion that voters overstate their support for a black candidate to pollsters for fear of being perceived as racist) will be a factor Tuesday. In fact, some think this election could finally shatter what they see as the “myth” of the Bradley effect.
The name comes from former Los Angeles mayor Tom Bradley, who, after being ahead by double-digits in some polls right before Election Day in his 1982 run for California Governor, unexpectedly lost to his Republican challenger.
It happened again during Doug Wilder’s run for Virginia governor in 1989. Wilder squeaked out a much narrower victory than most polls predicted to become the nation’s first African American governor.
The unifying thread in both of those races is that the pollsters who worked them say their internal numbers showed the contests much closer than the public polls predicted, and express skepticism that race was a factor in the discrepancies.
“Some people say they lie to pollsters, but they really don’t,” says GOP pollster Jim McLaughlin. He did the polling for Wilder’s opponent in 1989, Republican Marshall Coleman. McLaughlin says his internals had Wilder up by just a point as Election Day neared.
“There was a bad Washington Post poll that came out close to Election Day that had Wilder up by 17,” McLaughlin says. “It was just a bad poll, and that bred the perception.”
McLaughlin also notes that RNC Chairman Lee Atwater was pouring money into the Virginia governor’s race, which he wouldn’t have done without internal numbers showing the election much closer than it appeared.
McLaughlin’s conclusion: the Bradley effect is “complete nonsense.”
What about the polls in this year’s Democratic primary in New Hampshire where Obama led by double-digits only to lose to Hillary Clinton? No Bradley effect there either. Most pollsters attribute that to a late Clinton surge, helped by her tearful moment in a New Hampshire diner.
If the Bradley effect really did exist, argues Gallup’s Jeff Jones—another skeptic—you might expect to pick it up already in the public polls currently out there. The entire phenomenon hinges on the notion that respondents fear that interviewers will perceive them as racist. If true, says Jones, that impact would be severely diminished with so-called “robo pollsters.”
And while he observes a slight difference, Jones says there isn’t a statistically significant drop in Obama’s support in automated surveys. It’s yet another reason to doubt the legitimacy of the Bradley effect, unless you think white voters fear that robotically generated voices may think them racist.
This is not to say that attitudes on race aren’t playing a role in this election. It just means there’s little evidence that white voters are systematically hiding their intentions from pollsters.
If this election does turn out to be closer than some polls predict, it will have much more to do with suspect likely voter models than it will the Bradley effect. The possibility that some pollsters may overestimate the turnout among young voters, black voters or Democrats in general seems a much stronger one.
Shane D’Aprile is web editor at Politics magazine. firstname.lastname@example.org