The Bradley Effect

After correctly projecting winners of gubernatorial, U.S. Senate, and presidential elections in California in 76 out of 80 statewide contests since 1948, the Field Poll is widely considered the state’s most accurate pollster. Indeed, the last Field Poll survey of a general election often tracks the actual election margin within one or two percentage points, and the last survey to predict the wrong winner in a polled contest came in 1982. At the time, Los Angeles mayor Tom Bradley, the Democratic nominee for governor, was shown leading Republican state Attorney General George Deukmejian by a convincing margin of eight points. As potentially the first elected African-American governor in history, Bradley’s candidacy had generated considerable excitement. On Election Night, exit polls indicated a Bradley victory. But once the votes were tallied, Deukmejian won the election by about 100,000 votes and 1.2% of the total 7.5 million votes cast. Ever since his unexpected triumph pollsters have spoken of a supposed Bradley effect that plagues black candidates. This theory suggests that voters often tell pollsters they are either undecided or will support the black candidate in an election, but will actually stay home or vote for a white candidate on Election Day. Whether or not a Bradley effect exists—and a former Deukmejian pollster denies that it did even in that 1982 contest—its causes are hotly disputed.

Deukmejian won popularity during his governorship as a fiscal conservative following President Reagan’s mold of slashing allegedly wasteful domestic programs. He defeated Bradley once more in 1986, this time by a lopsided margin of 61-37. Ultimately, the first elected black governor in American history would arrive in 1989 when Democrat Doug Wilder won the governorship of Virginia. Again, observers noted a “Bradley effect” in that Wilder’s margin was close enough to warrant a statewide recount after polls had shown him leading Coleman by four to 15 points and even the Election Night exit poll had predicted a 10-point Wilder win. The same night, David Dinkins, also a Democrat, became New York City’s first African-American mayor, defeating Republican nominee Rudy Giuliani by some two percentage points after leading him by 18 and 14 points, respectively, in the two final polls of the race. A similar outcome had occurred in the Chicago mayoral election of 1983.

Wilder and Dinkins both left office after elections in 1993—Wilder was limited to one term in office and Dinkins lost another close race to Giuliani. The later 1990s and early 2000s proved a fairly poor era for black candidates, with no African-American governors elected and the only black U.S. Senator (Carol Moseley Braun) defeated for reelection in 1998. After the election of another black governor, Deval Patrick of Massachusetts, in 2006, and Sen. Barack Obama’s 2007 presidential announcement, the Bradley effect received considerable scrutiny regarding its continued applicability in an age of purportedly higher racial tolerance.

Throughout the 2008 primary campaign, pundits speculated whether Obama’s poll numbers would significantly overstate his on-the-ground support. However, with the single exception of a campaign-resuscitating come-from-behind win for Hillary Clinton in New Hampshire, Obama’s numbers largely matched or exceeded the polls when votes were tallied. In Southern states particularly, he was known to surpass his polled support. In South Carolina, the poll aggregator website RealClearPolitics showed an Obama lead of 12 points, but he won by 29. He similarly over-performed in Democratic primaries in Georgia, Virginia, Wisconsin, and North Carolina. During the general election campaign, with “Bradley effect” talk again in full swing as Obama maintained consistent polling leads over Republican Sen. John McCain, popular analyst Nate Silver wrote in Newsweek that “there is no reason to conclude that the polls are systematically overestimating Obama’s support.” Indeed, RealClearPolitics’ final poll average on November 3 showed an Obama popular vote lead of 7.6%, an almost perfect match for the eventual 7.3% outcome. RealClearPolitics also projected that Obama would win 338 electoral votes to McCain’s 200. The final result was 365-173.

Whether or not the Bradley effect existed in the 1980s, most agree that there is little evidence for it in closely watched, competitive elections today.