Who's Leading Presidential Vote? This Time, Polls Offer Little Help By: Staff
The race is dead even. No, it's Bush in a cakewalk.
Americans have been bombarded and bewildered by a blizzard of reputable polls showing very different results. What's going on?
IBD/TIPP's latest poll shows President Bush and Sen. John Kerry all tied at 45%, with Ralph Nader taking 2%. Head to head, Kerry actually leads, 46%-45%.
But several other surveys show Bush with a tidy lead.
Gallup's most recent poll for CNN and USA Today gives Bush an eight-point lead (52% vs. 44%), while ABC News and Time magazine show six-point advantages.
Why the difference between these other polls and IBD/TIPP?
Methodological differences — such as likely voter screens, weighting for party affiliation, timing of polls and allocation of undecided voters — are key factors that contribute to the differences.
With the election looming, most polls now focus on likely voters. Pollsters want to weed out people who won't vote, but don't want to screen out those who likely will.
Most observers think turnout will rise from 2000 because the race is close and so important. But no one knows how much.
"Obviously there is more intensity and engagement by voters this election," said Carroll Doherty, editor at the Pew Research Center. "The challenge to pollsters is to tease out that higher turnout in their surveys."
IBD/TIPP defines likely voters as adults who say they're very likely to vote in November, have a high level of interest in the election and have voted in every or nearly every presidential election or are eligible for the first time.
Pollsters also are having more trouble reaching people. Cell phones and answering machines are a big reason. Polling firms call back several times to reach people. IBD/TIPP calls four times.
Even when someone answers, fewer take part. So the overall response rate is just 30%.
So IBD/TIPP weights poll results to account for demographic variations based on census proportions for age, gender and race.
Most other surveys do this. But pollsters are divided over weighting for political party identification.
"There are two schools of thought. The first one says that party ID varies from poll to poll, while the second school believes it's relatively stable and changes over several years," said Raghavan Mayur, president of TIPP, a unit of TechnoMetrica Market Intelligence, IBD's pollster.
IBD/TIPP takes the latter view, weighting surveys to a benchmark of party composition based on a profile of voting-age Americans polled over five to six months. The latest IBD poll considers voting-age Americans to be 37% Democratic, 35% Republican and 26% independent.
John Zogby, whose election eve polls were very accurate in 1996 and 2000, uses a party weight of 39% Democrat, 35% GOP and 26% independent, based on 2000 exit polls.
Other pollsters say party weighting is a bad idea overall. "The problem is that party ID is an attitude," said Mark Blumenthal, a Democratic pollster who runs MysteryPollster.com. "The reality is a lot of people change their affiliations over a short period of time."
The latest Gallup poll, which doesn't weight for party, has 43% Republicans, 31% Democrats and 25% independents.
Roughly 90% of Republicans and Democrats vote for their party's candidate, so a poll's makeup has a huge impact on final results. Not surprisingly, IBD shows an even race; Gallup gives Bush a nine-point edge.
Scott Rasmussen, whose daily tracking poll has shown Bush with a one- to four-point lead over the past month, says weighting party ID on past results gives you a "plausible starting point."
"That may change a bit, but it's highly unlikely we'll see more (self-described) Republicans than Democrats," he said.
IBD's poll stands out right now because others showing a closer race haven't released fresh results lately.
Zogby's most recent poll, taken Sept. 17-19, showed Bush with a three-point lead. Over those same dates, the Wall Street Journal/NBC poll, which adjusts for party ID, showed a four-point Bush lead.
Also, there has been a downward shift in Bush's lead the past 10 days or so. IBD's poll has gone from Bush up three to even. Gallup says his lead has shrunk from +14 to +8. Time says it's fallen from +11 to +6.
However, Pew Research Center's newly released Sept. 22-26 poll shows Bush with a 48%-40% lead among registered voters. A Sept. 11- 14 poll had the race tied.
Pew's new "unweighted" poll included 33% Democrats, 32% Republicans and 30% independents.
In a recent paper, Pew compared how several major polls changed from early August to early September. In all, voters' GOP ties rose from one to 10 percentage points.
Weighting has problems, too.
Republicans fumed this summer when a Los Angeles Times poll showed a sudden shift in party ID, with 38% Democrats and 25% GOP. Not surprisingly, Kerry had a seven-point lead. Other surveys didn't confirm those results.
"I understand the confusion and strong reaction when a poll shows a 10-point GOP or Democrat advantage at a time when we are saying the nation is evenly divided," said Pew's Doherty.
Pollsters agree you shouldn't rely on a single poll. A rough and ready solution is to average out recent surveys. Real Clear Politics' average shows Bush with a 5.9-point lead over Kerry in a three-way race.