Divergent Opinion Polls Reflect
New Challenges to Tracking Vote
By JOHN HARWOOD
Staff Reporter of THE WALL STREET JOURNAL
September 20, 2004
WASHINGTON -- Widely divergent poll results in recent days underscore a paradox of the 2004 presidential race: Despite all the surveys, it may be the toughest election in memory for anyone to track.
Opinion polls themselves had been getting harder to conduct long before the matchup between President George W. Bush and his Democratic rival, Massachusetts Sen. John Kerry. The reasons range from growing reluctance to participate in surveys to increasing reliance on cellphones rather than the land lines pollsters have long used to ensure demographic and geographic balance in surveys.
But this year's bitter presidential contest has heaped on new challenges. They include an exceptionally close race and a polarized electorate that magnifies the consequence of different polling methods. In addition, unprecedented voter-mobilization drives by both parties make it especially tough for pollsters to say which voters probably will show up on Election Day.
"It makes it harder" to forecast the likely electorate, says Fred Steeper, a longtime pollster for Mr. Bush. In the six weeks to Election Day on Nov. 2, he adds, disparate polls may reflect sampling error and methodological differences more often than shifting opinion. "My advice to the consumer is ... the day-to-day reports of polling will exaggerate the changes in this race."
CAMPAIGN 2004
Read the latest news from the campaign trail at our Campaign 2004 page, including Zogby Interactive's latest poll of voters in 16 battleground states.
• Employment Picture Is Mixed in Election Battleground States
Media coverage of the campaign last week appeared to prove that point. On the same day last week, USA Today cited a new poll by the Gallup Organization in reporting that Mr. Bush "has surged to a 13-point lead" over Mr. Kerry, while other news organizations reported surveys by Pew Research Center and Harris Interactive showing the contest tightening to a dead heat.
Adding to the confusion is the way poll reports themselves become weapons in the campaign. The Bush campaign swiftly touts favorable surveys and seeks to discredit those showing Mr. Kerry drawing closer. The approach plays on the so-called bandwagon effects that energize supporters of a surging candidate and dispirit those of a lagging one.
Kerry advisers embrace dead-heat polls as a way to halt high-profile critiques of their campaign's inner workings and shift public dialogue to more fruitful ground such as violence in Iraq or domestic issues. Thus, even as Bush aide Matthew Dowd argued that Mr. Bush's lead was widening at week's end, Kerry spokesman Joe Lockhart told reporters, "The trends are going in our direction."
Underlying those conflicting arguments aren't just different political calculations but also differences in polling philosophy and techniques. Consider last week's Pew Research Center survey, which showed strikingly different research during two consecutive polling periods.
In the portion of the survey conducted Sept. 8-10, Mr. Bush led Mr. Kerry 52%-40% among registered voters. In a separate portion conducted Sept. 11-14, Messrs. Kerry and Bush were tied at 46%. But there was one other key difference, too: Among voters sampled in the first portion, self-described Republicans outnumbered Democrats by two percentage points; in the second, Democrats outnumbered Republicans by four percentage points.
Pew pollster Andrew Kohut described that difference as normal week-to-week drift -- because party allegiance is a fluctuating attitude -- that doesn't call his results into question. In fact, he says his surveys show the race is more volatile than other analysts have suggested. But the Bush campaign insists the partisan variation exaggerated the appearance of a trend toward Mr. Kerry.
Party allegiance "does not change in seven days" by that much, says Mr. Steeper, the Bush pollster. He says Mr. Kohut should have "weighted" his poll with a common assessment of partisanship for both samples; averaging the two would have shown the president with a steady lead of about six percentage points.
Bush advisers were more pleased by a CBS-New York Times survey late last week showing the president leading Mr. Kerry by nine percentage points, 50% to 41%, up from seven points the previous week.
Yet those CBS surveys were conducted the same way the Pew polls were -- without making any adjustment for the different number of Republicans and Democrats surveyed. And in the CBS polls, the number of Republicans surveyed rose sharply from the first week to the second.
Last week's CBS sample, in a mirror image of Pew's, contained four percentage points more Republicans than Democrats. Because this polarized contest has left roughly nine in 10 adherents of each party supporting its nominee, such variation in the number of Republicans and Democrats surveyed has an unusually large impact on polling outcomes.
In a close race, in fact, that can make the difference between an apparent dead heat and a solid lead for one candidate. If the CBS and Pew surveys are adjusted to reflect comparable numbers of Republicans and Democrats, their results would have been virtually identical.
Indeed that's precisely what liberal polling analyst Ruy Teixeira did on his Web log, called Emerging Democratic Majority. As the New York Times report of the poll carried the headline "Bush Opens Lead," Mr. Teixeira's blog declared, "CBS News/New York Times poll has it close to even."
(The Wall Street Journal and NBC News plan to release the latest of their surveys later this week. The Journal/NBC poll does adjust for variations in self-described party identification.)
Mr. Teixeira argues that the Democratic edge Mr. Kohut found is realistic, since exit polls from the 1996 and 2000 campaigns indicated that in both cases four percentage points more Democrats than Republicans showed up to vote. Slightly more self-described Democrats than Republicans voted in the 1984, 1988 and 1992 elections as well.
There's no assurance that will be the case this year, since both realignment of voter attitudes and party turnout drives can sharply affect that balance. Mr. Dowd says a roughly equal number of Democrats and Republicans will show up on Nov. 2.
Just who will turn out represents one of the biggest quandaries facing pollsters. About 105 million ballots were cast in 2000, and all sides agree more Americans will vote this time. Bush strategist Karl Rove predicts a total of around 110 million; Democrats estimate an even larger turnout, with some projections as high as 120 million.
Close to election time, pollsters like to report results among those considered most likely to vote on the theory that those results will align most closely with the final outcome. But weeks away from Election Day that's especially difficult to do, since many of the campaign's mobilization activities occur immediately before the election.
"I don't know how you factor that into your polling," Mr. Steeper says. Adds Democratic pollster Peter Hart, a veteran of presidential politics who helps conduct the Journal/NBC survey: "This is art. This isn't science. Nobody knows."
The Journal/NBC survey uses a single question to identify likely voters. It asks respondents to assess their interest in the election on a 10-point scale with 10 as the highest; those responding 9 or 10 are called likely voters.
The Gallup Poll, which provides surveys for CNN and USA Today, among others, assesses likelihood of voting in a different way that has raised the ire of the Kerry campaign. Gallup asks a series of questions first devised decades ago that assigns voting probability to each respondent; it then uses their answers and an overall estimate of voter turnout to identify the likely electorate.
Since mid-July, that method has yielded a likely electorate that is substantially more Republican-leaning than those of recent presidential contests. For instance, the likely-voter sample in last week's survey showing Mr. Bush ahead by 13 points contained seven percentage points more Republicans than Democrats. Given the current polarization by party, the survey would have showed a near-even race had the sample's partisan balance matched the 2000 exit polls or the registered-voter sample in the Pew poll.
As a result, Kerry pollster Mark Mellman has loudly accused the high-profile Gallup survey of using a likely-voter identification method that is "not very accurate," in part because the screening questions are outdated and because they can't properly measure voting intention so long before Election Day. The substantial variation between the likely-voter results and Gallup's registered-voter findings -- which showed an eight-percentage-point Bush lead -- is larger than what other likely-voter assessments usually record, Mr. Mellman says.
"We're open to any scientific evidence that would point to our modifying our likely-voter model," responds Frank Newport, editor-in-chief of the Gallup Poll. Mr. Newport says so far he hasn't seen any.
In 2000, Gallup's election-eve sample of likely voters showed Mr. Bush leading by two percentage points over Al Gore. Its registered-voter sample, showing Messrs. Bush and Gore neck and neck, was closer to the actual Election Day results. But Mr. Newport notes that in 1996 the likely-voter model more accurately forecast the size of Bill Clinton's victory over former Senate Republican Leader Bob Dole.
Mr. Bush narrowly lost the popular vote while winning enough electoral-vote battlegrounds to capture the presidency. And the same polling variations that can affect assessments of the race nationally are multiplied this time around by the intense focus on polls in battleground states. Democrats insist Mr. Kerry is more competitive with Mr. Bush in the race toward a 270-electoral vote majority than national polls would suggest. They argue Mr. Bush is piling up exceptionally large polling margins in states he's already expected to win, masking closer contests in battlegrounds such as Ohio and Florida.
Mr. Dowd argues, to the contrary, that Mr. Kerry's advantages in Democratic-friendly states such as California and New York offset Mr. Bush's edge in the South and Mountain West. As a result, he says, the president's advantage in battleground states matches his national lead.
The argument is especially hard to sift since different surveys of battleground states as a group show different results. There are multiple public polls of many individual battlegrounds, and the campaigns rarely publicize their private battleground-state surveys.
Write to John Harwood at john.harwood@wsj.com
