Phone polls have lost ring of truth Cellphone users don't get the call Response rates concern pollsters
TIM HARPER WASHINGTON BUREAU TORONTO STAR Oct. 20, 2004 WASHINGTON - The last weeks of the U.S. election campaign may have doomed telephone polling, which became common three decades ago late in the Jimmy Carter-Gerald Ford contest. The reason? Cellphones. The traditional method of polling via random phone calls to American homes may be missing a significant and active part of the electorate, and an increasing number of analysts are suggesting a surprise may be in store Nov. 2. Some 8 million Americans, equivalent to almost a quarter of the population of Canada, have "cut the cord" to use cellphones, and no longer have a home telephone on a land line. Their numbers are growing among the 170 million wireless communications users in this country, according to the Cellular Communications and Internet Association. The effect these CPOs (cellphones only) may have as voters in this campaign is unknown — for the simple reason that no one has asked them. American pollsters do not call cellphone users, who must pay for minutes used on incoming calls. But it's not just cellphones. Call display, U.S. "Do Not Call" legislation and an electorate which doesn't have time to answer a series of questions have all radically reduced response rates for pollsters. They are all changing the proportion of the estimated 110-120 million Americans expected to vote who are actually being polled in this closely watched, highly polarized race between President George W. Bush and Democratic challenger John Kerry. Pollsters are concerned. If, as George Gallup once said, the odds of an American being polled are roughly equivalent to being hit by lightning, the odds lengthened appreciably in 2004. "I think telephone polling will be a thing of the past, certainly 10 years from now," says Karlyn Bowman of the American Enterprise Institute, who has studied U.S. public opinion research for more than two decades. She says many early polls should be viewed with suspicion and predicts the era of internet polling is coming quickly. "The polling industry in general, and Gallup in particular, is concerned about anything that could affect our data," says Frank Newport, editor-in-chief of Gallup in the United States. "It is an issue like call waiting and call display that we are watching, but so far we have no evidence that our data is being compromised." If the spectrum of polls seem to indicate the race is anywhere from a dead heat to Bush eight points ahead, which — if any — are accurate? No one seems certain. Conventional wisdom suggests CPOs are young and urban, more likely Kerry supporters — while those who are answering land lines are older and possibly rural, more likely Bush supporters. More than six in 10 voters over 65 use land lines exclusively. Minorities in the U.S. are also less likely to respond to pollsters and they are also more likely Democrats. The bottom line — Kerry's support may be being under-reported and Bush's support over-reported. But Newport says he has no data to suggest that. While he says younger voters do tilt toward the Democrats they do not do so in such overwhelming numbers as to skew polls. He and other pollsters say they ensure their samples include 18-to-29 year-olds, expected to vote in possibly record numbers this year. There are 40.7 million Americans eligible to vote in that age category, an estimated 6.8 per cent of whom are CPOs. The Do Not Call legislation which governs parts of the country blocks telemarketers but not pollsters, but is symptomatic of a country fed up with unsolicited calls. Gallup and other major pollsters display their names rather than unknown numbers so they can get past call display, but Pew Research has found that in a typical five-day survey period, they receive responses from 27 per cent of households, down from 36 per cent as recently as 1997. "The decline results from increased reluctance to participate in surveys and not from an inability by survey organizations to contact someone in a household," Pew said in its April study. "The growing use of answering machines, voice mail, caller ID, and call blocking is not preventing survey organizations from reaching an adult in most of the households sampled." Gallup has come in for particular criticism during this campaign from Democrats and their surrogates, and was the target of a full-page ad in The New York Times by MoveOn.org criticizing its methodology after it reported a 14-point lead for Bush. "We're used to that, it's normal politics," Newport said. "We believe we are giving an accurate picture." Still, he says internet polling is probably on the way and Gallup may begin polling cellphone users. Frank Graves, president of Ottawa-based Ekos Research, said in an interview he polls cellphone users in Canada, and call display is not the hindrance to polling in Canada first feared. Like American pollsters, he says there is a surprising consistency in sample results, even if his researchers have to fight through high non-response rates. The disconnect between Canadian polling numbers and election results in June were the result of a last-minute shift back to the Liberals, Graves said, a "victory of caution over censure" and not a result of polling methodology. The number of Canadians who use cellphones only is relatively small, he said. But American pollsters warn about other numbers as well. Beware of weekend polling, they say, because Fridays and Saturdays are particularly poor days for reaching anyone. Wednesdays and Fridays are also bad because they can be bowling nights and Democrats are out. And Sunday morning polling is suspect because more Republicans are at church. Sunday afternoon polling gets a disproportionate number of female responses during the football season. Polling done the night before the election is useless, as is anything attempted the night before Halloween, most say.
