James Taranto: Is the "Wilder effect" the result of pro-black bias?
Barack Obama's presidential campaign has prompted a lot of talk about the "Bradley effect," a k a the "Wilder effect"--the supposed tendency of opinion polls to overestimate support for black candidates. The effect is named for Tom Bradley and Douglas Wilder, the Democratic nominees for governor of California in 1982 and Virginia in 1989, respectively. Both men were black; their Republican opponents, George Deukmejian and Marshall Coleman, were white. Both seemed to be doing much better in opinion polls than they actually did at the ballot box.
Surveys a month before the election showed Bradley leading Deukmejian by a margin of between 9% and 22%. By late October, Bradley's lead had closed to 6%. Deukmejian won the election by a bit over 1%. (In mid-October Deukmejian had fired his campaign manager, Bill Roberts, for predicting Bradley's erosion of support and ascribing it to racial prejudice.) Wilder defeated Coleman, but by less than 0.4% after having led in the polls by 4% to 15%.
One problem with the hypothesis of the Bradley effect is that it rests so heavily on these two examples, both a generation old. A new paper by Daniel Hopkins, a fellow at Harvard's Kennedy School of Government, addresses this objection. Hopkins looked at all races for governor and U.S. senator between 1989 and 2006 in which one party nominated a black candidate and the other a white one. He found that the Bradley effect does exist--or rather that it did until 1996.
"African Americans running for office before 1996 performed on average 2.7 percentage points worse than their polling numbers would indicate," Hopkins reports. "Yet this effect subsequently disappeared." As a control group of sorts, Hopkins also looked at elections that pitted a woman against a man. He found no similar effect.
If the Bradley effect is real, what accounts for it? The typical explanation is that voters lie to pollsters and say they support the black man, and then, alone with their prejudices in the voting booth, cast their ballots for the pol of pallor. Hopkins theorizes that the effect disappeared in 1996 because "racialized rhetoric about welfare and crime" had "receded from national prominence."
This could be true, but it is purely speculative. The Bradley-effect voter is not a creature whose behavior has been directly observed, merely a statistical artifact. There is convincing evidence that before 1996, a small but significant number of voters were inclined to tell pollsters they supported a black candidate but actually vote for a white one. As to why this would be the case, one can only guess....
Read entire article at WSJ
Surveys a month before the election showed Bradley leading Deukmejian by a margin of between 9% and 22%. By late October, Bradley's lead had closed to 6%. Deukmejian won the election by a bit over 1%. (In mid-October Deukmejian had fired his campaign manager, Bill Roberts, for predicting Bradley's erosion of support and ascribing it to racial prejudice.) Wilder defeated Coleman, but by less than 0.4% after having led in the polls by 4% to 15%.
One problem with the hypothesis of the Bradley effect is that it rests so heavily on these two examples, both a generation old. A new paper by Daniel Hopkins, a fellow at Harvard's Kennedy School of Government, addresses this objection. Hopkins looked at all races for governor and U.S. senator between 1989 and 2006 in which one party nominated a black candidate and the other a white one. He found that the Bradley effect does exist--or rather that it did until 1996.
"African Americans running for office before 1996 performed on average 2.7 percentage points worse than their polling numbers would indicate," Hopkins reports. "Yet this effect subsequently disappeared." As a control group of sorts, Hopkins also looked at elections that pitted a woman against a man. He found no similar effect.
If the Bradley effect is real, what accounts for it? The typical explanation is that voters lie to pollsters and say they support the black man, and then, alone with their prejudices in the voting booth, cast their ballots for the pol of pallor. Hopkins theorizes that the effect disappeared in 1996 because "racialized rhetoric about welfare and crime" had "receded from national prominence."
This could be true, but it is purely speculative. The Bradley-effect voter is not a creature whose behavior has been directly observed, merely a statistical artifact. There is convincing evidence that before 1996, a small but significant number of voters were inclined to tell pollsters they supported a black candidate but actually vote for a white one. As to why this would be the case, one can only guess....