Should the Internet replace phones as the polling technology of choice?
October 20, 1999
by Elizabeth Wasserman
(IDG) -- The landslide winner of a recent Time.com poll on the GOP candidates was a shocker. The respondents' choice for the presidential nomination was none other than... Orrin Hatch.
That's the same Orrin Hatch who's wallowing near the bottom of the Republican field by almost every measure – including most polls. Time.com's straw poll allowed anyone, not just registered voters, to cast a ballot and to do so as many times as they wished. The Utah senator's Web site linked supporters to the poll, and he pulled in 60 percent of the vote.
While this unscientific survey is the Web's version of the just-for-fun 900-number telephone polls, there's concern among researchers that the public has trouble distinguishing one poll from another on the Web. In fact, the news media has picked up on some of these pseudo-polls, and in at least one instance reported on the poll as if it were credible. An ABC.com straw poll on the Democratic presidential candidates found 25 percent of respondents favored Warren Beatty. Amazingly, the New York Times' Maureen Dowd cited the results as evidence that the political "nutty season" had begun.
The confusion surrounding online polls is adding fuel to the biggest debate now raging among pollsters: Is it time for the Internet to replace the telephone as the polling technology of choice?
Already, the Harris Poll Online has chronicled the upcoming presidential contest by assembling a pool of 5 million Internet users and surveying demographically adjusted samples of that pool every month. The company's efforts have attracted wide interest, but Harris' methods have prompted concern that Internet users are not representative of the national population: They're too white, too male, too young and too wealthy.
"Our opponents think you should not let it out of the garage yet," says Jonathan Siegel, director of Harris' Election 2000. "But Internet polling is going to happen."
Meanwhile, media outlets and well-established polling entities like the Gallup Organization say they are staying away from Internet polling – for now.
"I will be doing what I normally do: blocks of nationally representative telephone surveys," says Kathleen Frankovic, director of surveys for CBS (CBS) News. She admits she's intrigued by the flexibility of Internet polls and the potential for sharing visual materials with respondents online. But she notes that the Net still poses problems for pollsters. "You can't just randomly sample Internet users. You're not supposed to send unsolicited e-mail. And there is a truly nonrepresentative nature to the people actively engaged on the Internet." Most disturbing of all, she says, is the under-representation of the over-65 age group, which studies show votes in high numbers.
At the same time, there are new efforts under way to perfect the use of the Net as a polling technology. InterSurvey, a company recently started by two Stanford University professors, pulls a random sample the old-fashioned way – through random-digit phone dialing – and then equips panel members with Internet-access devices, such as WebTVs, so they can respond to Web-based surveys. The company, which is backed by venture capitalists and the university, has thus avoided much of the criticism leveled at Harris over its sampling; InterSurvey is also in talks with several campaigns and consultants about work on the 2000 election.
Doug Rivers, CEO of InterSurvey, says that he and his partner, Norman Nie, were inspired to start the company after controversy sprang up over Harris Poll Online's methodology. "We think the web is a great way to interview," Rivers says. "The content that can be included in a Web survey includes videos and multimedia presentations. A year ago, George W. Bush was leading all the polls in the race for president. They asked people why they would support Bush and a quarter of his supporters said it was because he made a good president. An image showing this was a different George Bush might have generated a different result."
Rivers admits that the Web allows pollsters to do things they couldn't do before. "The problem of the Web," he says, ''is that you can not get a representative sample."
When pollsters talk about a "representative sample" they mean a sampling that accurately reflects the population at large. The most widely used method is random-digit dialing, in which the first six digits of a telephone number are selected to allow for every region to be well represented, while the remaining four digits are dialed at random.
One problem with online polling is that pollsters can't e-mail people at random. E-mail addresses follow no standard format. There's no central phone book for all e-mail addresses. And some Internet service providers block spam to subscribers.
Technology has always influenced the way pollsters do their job. Starting in the 1960s, the phone replaced door-to-door polling as the preferred means of taking the nation's pulse. That was only after phone use became ubiquitous. But a 1998 Department of Commerce study found only 26 percent of households had Internet access, though more recent private studies estimate that share to be between 38 percent (Nielsen NetRatings) and 44 percent (Jupiter Communications).
Gallup, for one, is waiting until Net use becomes more widespread before embracing online polling. "We will continue to consider it," says Jack Ludwig, VP and research director at Gallup. He says the company might use the Internet "at some point, if the penetration gets up to where it is with the telephone or television."
A recent study by Mediamark Research bears out assertions that Net users aren't representative of the population at large. Some of the findings: 52 percent are male, 42 percent are between the ages of 18 and 34, 39 percent have yearly household incomes exceeding $75,000 and 88 percent are white. The comparable figures for the population at large: 48 percent are male, 34 percent are in the 18-to-34 age bracket, 21 percent have a household income above $75,000 and 84 percent are white.
Michael Traugott, president of the American Association of Public Opinion Research (OPI) , asserts that weighting the data – as Harris plans to do – won't compensate for the differences in any meaningful way. But Harris' Siegel notes that in 1998 the online Harris Poll accurately called 21 of 22 U.S. Senate and gubernatorial races, while traditional polls miscalled five of the 22 races.
Traugott argues that pollsters need to look deeper at the data from online polls. "The question is not whether you called the right candidate; it's the margin of victory between the two candidates. In the Web surveys, they got the bias you would expect. The results were pro-Republican. Web users are from a higher socioeconomic status; they're better educated, and they're more likely to identify with Republicans."
Will candidates embrace Internet polling? Alex Gage, president of Market Strategies Washington, the polling firm that's working with the George W. Bush campaign, says his opinion of the Net's potential has shifted. "I began the year thinking the Internet would be used in storing, retrieving and distributing information," he says. "More and more, I've come to realize that you can have selected panels of suburban women who you know have 90 percent Internet access, and you can talk to them every week about what they've seen." Gage says the potential lies in testing advertising, keeping tabs on opponents and getting instantaneous reactions from events like debates.
Democracy Online project to find out exactly how Web affects elections
RELATED IDG.net STORIES:
Study into Web's Impact on State and Local Elections Goes Public
|Back to the top||
© 2001 Cable News Network. All Rights Reserved.|
Terms under which this service is provided to you.
Read our privacy guidelines.