Swings may reflect voters' changing minds
By David Jackson / The Dallas Morning News
NEW YORK â€“ A film scout once checked out the busy back room at Blum & Weprin Associates, a drab, peach-painted box where workers busily phoned strangers to tap public opinion.
He had an inspiration: What a great set for a bookie joint.
Critics of polls such as the ones conducted by Blum & Weprin find the analogy apt, calling the surveys little more than dice rolls on the presidential race between George W. Bush and Al Gore.
But the principals of the New York firm and other pollsters say the candidates' shifting fortunes simply reflect a volatile electorate. The attacks on their credibility, pollsters said, are as cyclical as the elections themselves.
"People hate us," Julie Weprin said. "People who don't agree with the majority feel like they've been left out."
Added Micheline Blum: "People always think something's wrong with the polls if they don't agree with them."
Campaign aides and some political analysts have questioned polls that have bounced around like a roulette ball. Darrell M. West, a political scientist at Brown University, raised the specter of polling's biggest nightmare: Harry Truman's unexpected win over Thomas Dewey in 1948.
"The 2000 elections may discredit the polling industry in as dramatic a fashion as the 1948 presidential contest did," Mr. West said.
He pointed out that in September, a Newsweek poll gave Mr. Gore a 14-point lead, while the Voter.com "Battleground Poll" put Mr. Bush up 5.
Countdown: Seven days
Al Gore says financial good times returned under Democrats and that changing course could "drive our economy into the ditch."
George W. Bush says Mr. Gore would squander the nation.s economic riches and continue the "politics of division in Washington."
"There's no reason that two polls talking about the exact same race should have two entirely different results," Mr. West said.
But the pollsters of 1948 stopped weeks before Election Day. This year, surveys will be made every day through Nov. 7. And pollsters said these daily tracks show that voters change their mind more often than previously thought, especially in a close race.
"I think there are incorrect assumptions about the stability of the electorate," said Frank Newport, editor in chief of the Gallup Poll. "I think who you're going to vote for is a very loosely held attitude."
The daily tracking poll that Gallup conducts for CNN and USA Today has drawn the most criticism of late. In mid-October, it gave Mr. Bush double-digit leads on three consecutive days; less than a week later, Mr. Gore led by 1 point. Now Mr. Bush is back ahead.
Pollsters said the Gallup methodology reflects the fact that all polls are not alike. They said the tracking poll is designed solely to pick up movement within the presidential horse race.
A different purpose
Other surveys are designed to predict the outcome on Election Day, so pollsters weigh the results using turnout rates from past elections.
Some polls rely on registered voters, while others try to screen for those likely to cast ballots on Election Day. Even then, the definition of a "likely voter" may differ from poll to poll, and even within the same poll on different days.
"For several days, you'll find Republicans more enthusiastic and determined to vote," said William Schneider, who analyzes polls for CNN. "On other days, Democrats catch up. Enthusiasm and determination to vote are very volatile this year."
There can be other differences among surveys. Some polls wind up interviewing more Republicans than Democrats, often depending on which days phone calls are made and whether they are made at night or on the weekends.
Pollsters can apply a variety of methods, but many say their profession has a common challenge: Finding people to talk to isn't as easy as it used to be.
John Zogby, a pollster based in Utica, N.Y., said that when he began polling in 1984, his response rate averaged about 55 percent. These days, it's closer to 35 percent.
"But all that means is it takes us longer to do the poll," Mr. Zogby said. "It does not mean we are not getting a representative sample."
Amid the pallid lights and metal chairs at Blum & Weprin Associates, poll takers learned that even surveys about religious views and cable television viewing habits can draw suspicion.
"I'm sorry, I'm on the phone long distance," one potential respondent said archly, cutting short a poll taker.
"Can I call you back later?" the caller persisted.
"Oh, I don't care ... thank you."
While bigger firms use automatic dialing machines, phone operators at Blum & Weprin dial in the old fashioned way. The poll takers â€“ including retirees, college students and out-of-work actors â€“ nosh grapes or drink coffee, sitting at foldout tables shoved against the wall or in the middle of the rectangular room. The firm, which conducts polls for The Dallas Morning News and other media outlets, has about 50 phones.
Ms. Blum said her employees average two to five valid responses per hour, depending on how in-depth the survey is. She estimated the response rate at between one in five and one in 10.
Many of the computer-generated phone numbers have been disconnected or are business numbers. Often no adults are home, or they don't want to talk to strangers.
"We have to be very quick to say we're not trying to sell you anything," Ms. Blum said.
Ms. Blum said response rates have gone down everywhere, but not horribly so.
"If you tell people what you're calling for, they are cooperative â€“ as long as they believe it really is a poll," Ms. Blum said.
But critics such as Mr. West said lower response rates can create a "selective bias" within polls, which increasingly reflect extreme rather than moderate views.
Part of the problem, he added, is that there are too many polls.
"There are three, four, five surveys coming out every day," Mr. West said. "It reinforces public skepticism about polling."
Ms. Blum said that once you get people to talk, they will answer questions truthfully if not completely. Questions about income are the most frequently refused.
"People don't lie," Ms. Blum said. "They may not answer, but they don't lie."
Pollsters also said that, fluctuations aside, the polls have been in general agreement.
Mr. Bush led the race before the Democratic convention, then Mr. Gore surged ahead for most of September. The three debates in October seemed to give Mr. Bush new momentum.
Mr. Schneider of CNN said that the word "lead" can be a misleading this season because most of the advantages have been within the polls' margin of error. He also said many voters have expressed ambivalence about both candidates.
This year, analysts said, the polls are less unusual than the closeness of the presidential race.
"It's been many years since it was this close this late," Ms. Blum said.
And whether one believes in polls or not, they are always put to an acid test, Ms. Blum said: "If you don't think it works, wait until Election Day and see what happens."