Covering politics in North Carolina and beyond, VoterRadio.com is streaming 24 hours a day. Listen live or on-demand.
Are Pollsters Getting It Right this Year?
By J. Barlow Herget
Published: Oct. 4, 2004
RALEIGH - Along with falling leaves, this is the season for political opinion polls.
If you’re what pollsters call a “frequent voter” you probably have been called at least once. My household has been polled four times.
We read about the results almost daily. The cable news channels tout their own surveys. The national parties do likewise and North Carolina media have reported almost weekly on their own polls.
There’s growing chatter, however, among political observers and pollsters themselves that the political surveys this year may not be reporting the full story.
Phil Meyer, Knight Professor at the University of North Carolina School of Journalism and a veteran pollster, says that this might be the year that the profession “stumbles into a major trap.”
Gene Lyons, columnist and author of “The Hunting of the President,” reminded his readers recently that “back in 2000, the Gallup organization showed George W. Bush leading Al Gore by 13 percentage points two weeks before the election -- a veritable landslide. Its final count showed Bush up by 5. Gore won the popular vote by roughly one half of one percent nationwide. Did that many people change their minds? Or were the polls skewed by poor methodology and wishful thinking?”
This year, polls may be skewed by some big changes in the electorate involving technology and voter turnout. Says Scott Crosson, PhD, a political psychologist and opinion pollster, “There are some long-term and short-term trends that make you concerned about some of the polling numbers out there.”
Those changes include:
Cellular phones. More and more people use cell phones rather than traditional landline phones, especially young people. These people do not show up on the lists that pollsters need to obtain accurate random samples for their telephone calls.
Voter turnout. Election officials believe that Americans are passionately and equally divided about President George W. Bush and expect a larger than normal vote. In North Carolina, State Elections Board Executive Director Gary Bartlett predicts a 68 percent turnout among registered voters. Turnout in 2000 was 60 percent. This means a large number of new voters will show up, and they typically are not on “frequent voter” lists.
Young people. This demographic group is the hardest to reach for opinion surveys, but they’re engaged in this election far more than in past elections.
Non-response rate. In the early days of polling, George Gallup’s employees conducted personal interviews in respondents’ homes. The response rate was 80 percent. People today either don’t like pollsters as much or don’t want to share their opinions; only 30 percent respond. Minorities and poor people respond even less, and if they show up to vote, their numbers will throw off pre-election surveys.
Crosson illustrated the challenge in a recent poll he conducted over several evenings. His team made 1,400 calls one night of which 10 percent were answering machines and 74 percent were people who hung up or simply didn’t answer.
Raleigh Pollster Dean Debnam, president of Public Policy Polling, employs “robocalls” or computers that make automated calls in which respondents simply push number pads on their telephones to answer questions. He surveyed a House district last week and his machines dialed 19,000 numbers in about an hour.
“We got 8,017 answering machines; 597 busy signals; about 2,500 disconnected lines; and about 4,530 hung up; 1,190 didn’t answer. We got 1,082 legitimate responses,” says Debnam.
As the election nears, voters can expect to hear from “push polls,” too. They show up typically in the final week of the campaign and are not legitimate polls. Their real purpose is to smear a candidate by asking such questions as, “If you knew Harry Truman lied about his World War I experience, would you still vote for him?”
Tracking polls measure the horse race aspect of the election. They ask few questions, require 400 to 600 respondents, and are taken almost daily. They help candidates quickly identify swings in voter mood.
The final surveys of the season are “exit polls” taken on Election Day. They canvass voters who have just voted in selected precincts. Unless the election is a dead heat as in 2000, pollsters and their employers will know by 4 p.m. who has won the election. (If TV anchors begin to smirk when they report in their 6:30 p.m. broadcasts, you know that they know.)
Despite this year’s challenges, candidates and media are relying on polls more than ever because, says Meyer, “it’s better than guessing.”