NEW YORK (AFP) — They stumbled in 2000, and fell on their faces in 2004 -- so can anyone believe what opinion pollsters say about Barack Obama's lead over John McCain today?
Scientific surveys aimed at predicting the winner of the November 4 election are proving harder than ever, polling experts admit.
"It's going to be a very difficult election to predict because we can't use normal statistical models," said Steffan Schmidt, a political science professor at Iowa State University.
In 2000, major media US broadcasters may have helped tip the balance of an extraordinarily tight race in favor of Republican George W. Bush by announcing that he had beaten Democrat rival Al Gore in Florida.
The call, based on calculations of preliminary data, was retracted two hours later, but the issue of who won Florida proved the key to overall victory, eventually awarded to Bush.
Even more embarrassing was the US media's reporting of exit polls in 2004 that showed Democrat John Kerry cruising to victory against Bush. When real votes were counted, Bush emerged the winner.
Now, four years later, pollsters face a new and unique set of problems.
The biggest imponderable is the significance of racism in an election where Democrat Obama is bidding to be the first black president in US history.
One theory -- the so-called "Bradley-Wilder effect" -- is that people tell pollsters they support Obama because they are afraid of sounding racist. Then in the privacy of the polling booth on November 4 they promptly opt for the white Republican candidate, McCain.
The latest national opinion polls show Onama maintaining an average lead of six points over McCain.
In other words, the slim, yet firm lead held by Obama in opinion polls may prove hollow on election day.
The theory took shape when a black former mayor of Los Angeles, Tom Bradley, lost the 1982 California governor's race to a white opponent, while opinion polls had pointed to victory.
A 1989 race in which the black candidate, Doug Wilder, only just squeaked through as Virginia governor, despite having been forecast to win by 10 percent, appeared to confirm the phenomenon.
Many other elections involving black candidates have not adhered to the pattern, but a presidential election puts unique pressures on voters. "We have no idea exactly how that's going to affect the polling," Schmidt said.
Carroll Doherty, associate director at Pew Research Center, a polling agency, told AFP, "we are entering the unknown."
"The Wilder effect is one of the great issues hanging over this election that we've never encountered before."
Doherty said there was little evidence of any Bradley-Wilder effect during the Democratic primary battles Obama fought against Hillary Clinton and other white contenders.
Yet in a tight election, as many predict on November 4, every element counts.
"We know that race is a factor in the consideration of some people. How big is not clear. But even a two- or four-percent change can make a difference."
Another hurdle in making accurate polls this time is the reliance on the landline telephone for reaching respondents -- automatically excluding almost the fifth of the population that uses mobiles or other communications.
"We are now down to only 79 percent of the population using landlines," Humphrey Taylor, chairman of The Harris Poll, said.
"Particularly among the young and college students, almost no one has a landline. Then you have Skype and there are... (those) with no phone at all."
Polling by Pew that does incorporate responses from mobile phone users results in a small but significant upturn for Obama, whose supporters tend to be younger and higher-tech than those of McCain, Doherty said.
"Perhaps Obama could get as much as two to four points more support if cell phones are included. That doesn't sound like a great deal, but then the last election was decided by three points."
What reputable polling organizations, such as Pew, Harris or Gallup, stress is that their own questioners bear primary responsibility for producing accurate results.
A one word change in a question can elicit a different response.
For example, in the aftermath of the Wall Street meltdown, the phrasing to describe Congress' multi-billion-dollar response was especially loaded. Was it a rescue, an intervention, or a bailout?
The National Council on Public Polls, an association of polling agencies, warns that the order of questions can also have a subtle impact.
"If people are asked what they think of the economy before they are asked their opinion of the president," the council says on its website, then "the presidential popularity rating will probably be lower than if you had reversed the order of the questions."
As Taylor, of the Harris Poll, said, there is no rest for the conscientious pollster.
"We are always running scared, running nervous," he said.
Copyright © 2009 AFP. All rights reserved. More »
