Saturday, October 25, 2008
From the Newsroom: 'Wilder Effect' may bring surprises
From the Newsroom column
Dwayne Yancey, Senior editor for new channels
981-3113
- Previous column: Many variables in political polling (Oct. 5, 2008)
From the Newsroom blog
Recent posts
You'd think by the way some of my media colleagues are acting that the election was already over.
Newsweek recently had a cover headlined: "How a President Obama would govern a center-right country."
That's not quite as catchy as the infamous headline "Dewey defeats Truman" in the Chicago Tribune in 1948, but you get the idea.
Here's why I'm not rushing to inaugurate Barack Obama just yet: The Bradley Effect. Or, as we prefer to call it here in Virginia, the Wilder Effect.
That's the fancy catch phrase for the belief that some white voters will lie to pollsters when a black candidate is on the ballot.
The phenomenon got its name from Tom Bradley, the black mayor of Los Angeles who led the polls in his 1982 race for governor of California -- only to lose in an upset to Republican George Deukmejian.
The Virginia version came in 1989, when another black candidate for governor -- Doug Wilder -- led the polls by increasingly wide margins as Election Day approached, only to wind up winning by less that one-half of 1 percent in an election that precipitated a statewide recount.
I still remember being at the Democrats' headquarters in Richmond on that election night, waiting for the returns to confirm the big Wilder win that most of us had already written up, just waiting to fill in the actual numbers.
Except the numbers weren't coming in as expected. The early returns stubbornly wouldn't budge beyond an oh-so-narrow Wilder lead. I remember asking Bobby Scott -- now a U.S. representative, then a state legislator from Newport News -- what the matter was.
He breezily assured me that everything would be fine. Just wait until the 9th Congressional District-- Southwest Virginia -- comes in, he said. We're going to win big there, he said.
Uh, I hate to break this to you, I said, but the 9th has already come in, and Wilder didn't win it. I remember a look of horror passing across Scott's face. He hustled off to get more information and I never saw him again that night.
So, enough reminiscing, and back to the point at hand: Are the polls this year right? Is there still a Bradley Effect distorting the polls? And was there ever really such a thing?
Scholars and pollsters still debate its very existence.
Lance Tarrance, who was the pollster for Deukmejian in the 1982 campaign, recently wrote an article for the political Web site Real Clear Politics in which he said there was no Bradley Effect, just bad polling.
He says his internal polling always showed his candidate closing fast enough to overtake Bradley, but the public polls showing a Bradley lead were wrong because they didn't account for absentee ballots, where the Republicans had an advantage.
But what about Virginia in 1989, where polls showed Wilder with a big lead that never happened?
Did some white voters simply lie to pollsters that they were supporting a black candidate because it seemed the socially acceptable thing to do?
"Yes, it happened, no question," says Larry Sabato, the University of Virginia political analyst who studies Virginia election returns the way football coaches study game films. "The dead giveaway were the exit polls, which confirmed a 55 [percent]-56 percent Wilder lead (there were two good exit polls that year, separately conducted). And the exit poll that asked about lieutenant governor and attorney general hit those two races on the head. Remember, same voters responding. They only lied about governor."
The trend was most pronounced in rural areas, such as Southwest and Southside Virginia.
For the record, Paul Goldman, who ran Wilder's campaign that year, insists voters weren't lying, the pollsters just goofed. He says his campaign's private polls never showed Wilder winning big -- and that an unpublicized exit poll on Election Day actually showed Wilder losing.
So what about this year's polls? Are they right? Or should we automatically deduct a certain number of percentage points from Obama's totals, whatever they are?
Most pollsters say no -- that modern polling techniques account for racial bias, either by matching the race of the interviewer with the race of the interviewee, or by using push-button surveys to allow more anonymity, or by other statistical tricks of the trade.
Sabato is inclined to agree. "This was 19 years ago," he says. "The country has made great racial progress. Maybe there's a point or two of racial leakage this year, but I doubt more. And the large black turnout and youth vote for Obama may provide more than enough votes to make up for the loss, in Virginia and everywhere else where the contest is competitive."
Still, we should probably wait until the votes have actually been cast -- and counted -- before proclaiming any winners.
Just ask President Dewey.
