New Research on the "Wilder effect"
Although many considered the poll an outlier when it was published nine days before the Virginia gubernatorial race in 1989, a Washington Post poll showing Lt. Gov. L. Douglas Wilder with a "commanding" lead was among those that spawned a generation of research on potential bias in public opinion surveys on contests between black and white candidates.
The poll showed Wilder, who is African American, with a 15-point advantage over Republican J. Marshall Coleman, a white candidate. On Election Day, Wilder squeaked by with a razor thin margin of under 1 percent to become the country's first black governor.
At the time of the Post poll, both campaigns put Wilder's final-stretch lead at about five percentage points, as reported alongside the Post's own poll.
Nonetheless, the "Wilder effect" - where whites overstate their support for black candidates - merged with the "Bradley effect" - where whites say they have no opinion when they really support a white candidate in match-ups between white and black candidates - in lore, casting doubts on the accuracy of polls in such contests.
Despite years of successes polling just such campaigns, the theory lingers. But a new analysis by Daniel Hopkins may yield a better understanding of the dynamics at play.
Hopkins shows how any "Wilder effect" that existed in the early 1990s has disappeared, countering the notion that there is a systematic bias in such polls. Instead, he argues that the effect that did exist resulted from the particularities of those earlier elections. Those dynamics could return in this year's election, but it is far too simplistic to assume a problem.
Hat top to Prof. Sides from the Monkey Cage.
Posted by: George | July 29, 2008 10:16 PM | Report abuse
Posted by: che | August 1, 2008 11:29 AM | Report abuse
The comments to this entry are closed.