In the beginning, there was the Bradley Effect.The Bradley Effect describes an election event where the a candidate who is clearly leading in the polls ultimately loses the election. Certainly you’ve heard about it. Everyone has. It made the front page of the India Times. The effect is named for the 1982 California Governor’s race in which Bradley succumbed to Deukmejian despite having a substantial lead in the polls up to a few days before the election. The effect suggests that there is a gap between what people report to pollsters and what they did/will do in the voting booth. Election analysts suggested that race and social desirability were factors shaping voting behavior. To wit: when polled about their intentions (particularly by an African-American pollster) many white voters reported intent to vote for Bradley (or undecided). But in the privacy of the voting booth, they voted for Bradley’s white opponent. These voters, it is said, feared that if they reported their real preferences they would be labeled racist. Most people don’t want to be called racist. It’s generally not socially desirable.
The so-called Bradley Effect is a challenge to pollsters. It indicates that there is an margin of errors in the polls. And this year, preliminary election poll-to-outcome analysis by Albertson and Greenwald (2008) suggests a Reverse Bradley Effect emerged next to the Bradley Effect. A Reverse Bradley Effect occurs when polls in traditionally Republican areas underreport support for the African-American candidate. The same social desirability explanation applies. In a traditionally Republican area when a pollster calls, many voters simply reflect the local sentiment independent of their actual leanings and (interestingly) their race.
Sure, theoretically phone canvassing is random and anonymous. But there is something personal about talking to someone in voice. You are talking to another person. That person knows your phone number. Other people make judgments. Most people don’t like to self-identify as “outsiders” — even to complete strangers.
Taken more broadly, the Bradley and its Reverse Effect are the same thing. The visible difference between them is that the Bradley Effect was reported in Democratic California and the “Reverse” Bradley Effect in Republican strongholds: Montana, South Dakota, Oklahoma, Missouri, and Indiana. The local sentiment where they happen is flipped. It’s not about Race, it’s about social inclusion and social desirability. Both Bradley Effects describe people’s desire to blend in… to appear “normal,” where “normal” is defined by local sentiment. Whatever that means, wherever you are.
The real irony is that, in the event, the Bradley loss was not an example of the Bradley Effect. Elections are complicated. Campaign insiders suggest a careful look at the polls leading up to the election (rather than the weekend poll often cited) indicate that Bradley had lost measurable ground before the election because of an unpopular gun control initiative. Further, the GOP mounted an aggressive absentee ballot program that year. Hundreds of thousands of Republican votes that no pollster anticipated rolled in. Polling error, again.
The other irony is that despite all this discussion, more comprehensive analyses of voting behavior (Hopkins, 2008) indicate that social desirability driven gaps between poll reporting of, and actual, voting booth behavior is diminishing for race and (as an original finding) are negligible for gender split races. By the time we figure out what it really is, the Bradley Effect may no longer be a problem.
Article posted on HFI.
Blair, Levin (2008). What Bradley Effect? NYT Op-ed.
Greenwald, Anthony G. and Albertson, Bethany (2008) Tracking the Race Factor Pew Research Center Publications.
Hopkins, Daniel, J. (2008). No More Wilder Effect, Nevera Whitman Effect: When and Why Polls Mislead about Black and Female Candidates. Poster presented at the Annual Meeting of the Society for Political Methodology, University of Michigan, Michigan, July 11th, 2008.
Schwarz, Joel (2008). Polls may underestimate Obama’s support by 3 to 4 percent
Wikipedia: Bradley effect.