Bradley effect: The purported cause of a discrepancy between voter opinion polls and election results in a race between a white and a nonwhite candidate. The term is named for Tom Bradley (1917–1998), the first African-American mayor of Los Angeles, who led by a substantial margin in the polls but lost the 1982 California gubernatorial election to his white opponent, Republican George Deukmejian. The phenomenon is less commonly known as the Wilder effect, after Douglas Wilder of Virginia, the first African-American governor in the United States.
William Safire devoted his "On Language" column in yesterday's New York Times Magazine to the Bradley effect. About the unexpected outcome of the Bradley race, he writes: "Speculation ranged from inaccurate sampling, to last-minute mind-changes, to latent racism, to freely lying voters, to the reluctance of those being polled to admitting a preference that may be socially unacceptable — anti-black — in talking to interviewers."
The Bradley effect has surfaced in coverage of the current presidential campaign, too. Safire again:
The phrase burst out of the starting gate of this year’s presidential campaign in the coverage of Hillary Clinton’s comeback victory in the nation’s first primary election in January. Barack Obama, victor in the Iowa caucus, had been polling ahead of Clinton but lost by 2.6 points. “Did ‘The Bradley Effect’ Beat Obama in New Hampshire?” headlined the liberal Nation magazine over a posting by John Nichols. John McWorter [sic: it's McWhorter] of The New York Sun concluded the answer was no, under the headline “Bradley Effect, R.I.P.”
Read my post about other eponymous effects (Doppler, Droste, Forer, etc.) here.
¹ Safire—who once was a Nixon speechwriter and has proclaimed himself a John McCain supporter—provides a citation from "the liberal Nation magazine" but doesn't identify the slant of the Manhattan Institute. Not to belabor this, but that liberal article appeared in a Nation blog, not in the pages of the magazine itself.