Net polls miss constituent groups, pollsters say

The Associated Press

05.22.00

Printer-friendly page

PORTLAND, Ore. — Some pollsters say the Internet allows them to collect public attitudes more quickly and cheaply than the telephone. But many of the public opinion researchers gathered here this weekend say surveys done strictly online don't measure up.

They say pure Internet polling fails to survey people who don't have computers — people who tend to have lower income and less education, people more likely to be minorities. And they say it ignores some basic principles of survey research, especially the concept of random sampling.

They raise the specter of 1936, when a famous survey miscalled the presidential election because it relied on lists of people who owned telephones and cars — at a time when those were luxuries.

These days, probability theory says a national sample of Americans, often about 1,000 or more drawn with a telephone technique called random-digit dialing, can mirror the attitudes of the entire population if the poll is done properly and then weighted for demographics.

As for Internet polling, Mike Traugott, president of the American Association for Public Opinion Research, said, "Clearly, the Internet is the wave of the future.''

The turmoil running through the industry is similar to the anxiety faced by pollsters when they began to make the switch from face-to-face to telephone questioning, he said.

But Traugott said the number of American adults who go online, roughly half, is not high enough yet to provide a cross-section of the population. And he said no one has figured out how to draw a random sample of computer users the way traditional pollsters draw a probability sample of the population.

Pollsters are working hard to figure out how to harness the speed, power and efficiency of the Internet. Two differing approaches were evident on May 19 at the annual meeting of the survey-research group.

InterSurvey of Menlo Park, Calif., is blending the methods of traditional research, starting by drawing a panel of respondents using a telephone poll. Anyone in the panel who doesn't have Internet access is given interactive television, at InterSurvey's expense, to file responses.

When InterSurvey wants to conduct a poll, it contacts the respondents by lighting the boxes on top of their televisions, a technique less disruptive than a dinnertime phone call.

"You don't have to abandon scientific sampling to poll on the Internet,'' said Doug Rivers, chief executive of the company. Traditional pollsters and major media outlets have used InterSurvey to get quick reaction on events such as the State of the Union address.

Another wave in Internet research is the collection of panels of potential respondents like those put together by Harris Interactive of Rochester, N.Y., Greenfield Online Inc. and other firms.

Harris Interactive has built up a panel of 6.2 million people, most of whom volunteer through Web sites, said George Terhanian, a company official.

"We get bashed by people who talk about 1936 and 1948,'' Terhanian said, referring to two of the most famous fiascos in polling history. The Literary Digest predicted Alf Landon would win the 1936 presidential election based on polls taken from lists of people who owned telephones and cars. In 1948, pollsters predicted Thomas Dewey would beat Truman, and later overhauled their methods to avoid such an embarrassing repeat.

"If we used unweighted data, our results would be as bad as the Literary Digest,'' said Humphrey Taylor, a veteran pollster who's also a Harris executive. He said Harris is able to weight its data according to demographics and the "propensity'' of people to be on the Internet. Some traditional pollsters don't accept his theories, but Harris, Greenfield and other firms have found a more willing audience in market surveys.

Harris Interactive says its political polls have been largely successful, but traditional pollsters say they're worried about offers of inexpensive research they feel isn't scientific.

"My view is that it's an enormous threat,'' said pollster Warren Mitofsky, an early proponent of telephone use in surveys. "There are too many unsophisticated people willing to pay and this kind of bad research is going to drive out some good research.''

In other discussions at the AAPOR meeting, public opinion researchers said the wealth of new political information available electronically probably won't increase voter participation in this year's presidential election — even though half of all potential voters now use the Internet.

Instead, pollsters and political scientists at the annual meeting of the American Association of Public Opinion Research suggested last week that the Internet's major impact on the 2000 elections will be as an organizing and fund-raising tool.

"It will make it easier for the people who are interested in politics to really be informed, to enrich their experience,'' said Andrew Kohut, director of the Pew Research Center for People and the Press. "But it's not for the people who are marginal in terms of their interest in elections and politics and the larger world.''

Surveys by Pew and other organizations estimate that the number of Internet users has doubled since the 1996 presidential election. The new users tend more to be female, less educated and less well-off than the group that was online in 1996.

But while campaigns and interest groups offer a vast array of political information to voters over the Internet, recent surveys have found that Internet users rank politics far down on their list of reasons for going online.

"As the users of the Internet become more like the real world, it becomes less political,'' said Michael Margolis, a political scientist at the University of Cincinnati. "Most people aren't interested in politics.''

For those who do turn to the Internet for political information, Margolis said the most-visited Web sites are those maintained by the traditional sources of political information, the major political parties and news organizations.

"Where we had hoped to see change possible in the Internet, the Net creating a new kind of politics, what in fact we have is more of a politics as usual,'' he said.

Kohut compared the Internet's impact on politics to that of C-SPAN, the cable television network that began providing gavel-to-gavel coverage of Congress nearly two decades ago. Despite that saturation coverage, he said, Americans probably know less about Congress now than they did before the broadcasts.

Still, Kohut said, the Internet already has proved an invaluable tool for candidates to organize their supporters, raise money quickly and efficiently from newly energized followers, and inexpensively target potential voters.

It also is becoming increasingly important as a source for campaign news, he said, with 24% of Americans saying they use it for some political news.

Among Internet users, one in five has sent or received an e-mail related to politics, 26% have taken part in an online public opinion poll and 20% have downloaded political information.

Newspapers and the television networks were by far the main source of campaign news for most voters four years ago, but Kohut said regular Pew surveys over the past four years have found that the Internet's increasing presence has fractured that dominance.

Asked to identify their main source of campaign news, Kohut said, 31% of Americans now cite cable television, 31% newspapers, 25% local television news, 24% network news and 6% the Internet.

"The audience is now dispersed all over the place,'' he said. "The Internet is clearly a player and the player with the most dramatic growth. All of the other sources are going in opposite directions in terms of audience size.''