In a move that has ignited a fiery debate among public-opinion researchers, the Harris Poll plans to canvass cyberspace to predict the outcomes of primaries and elections in 2000.
Researchers at Harris Black International Ltd., in Rochester, N.Y., say Internet polling is less expensive and faster and offers higher response rates than telephone surveys. "All research is going to migrate to the Internet," says Gordon Black, Harris Black's chairman. "If we can do what we want to do in 2000, it will all but eliminate telephone polling going forward."But most researchers condemn the Web polls as flawed. Critics such as Michael Traugott, president-elect of the American Association for Public Opinion Research, say Internet users are still too white, too rich, too educated and too male to use in drawing inferences about the general public.
Problem No. 2: The people polled in Harris Black's Internet surveys aren't randomly chosen, as telephone participants are; they are a "self-selected" group. In other words, they have volunteered to be part of the test sample, which could mean they are more comfortable with technology, more informed about news and events, and more politically engaged than Americans who aren't online.
"There's a fair amount of cachet attached to being the first company to try (large-scale Internet polling), and sooner or later this will be a viable method, but not right now," Mr. Traugott says.
David Iannelli, a senior consultant with Coldwater Corp., Ann Arbor, Mich., a research company that conducts opinion polling for The Wall Street Journal and NBC News, agrees with Mr. Traugott. "When you want to talk to shoppers, you go to the mall," Mr. Iannelli says. "When you want to talk to households, you call them on the phone. You only poll on the Web if you want to know what the technology types are thinking."
Though controversial, Harris Black's approach is methodical. Over the past year, with the help of the Web-portal company Excite Inc., Harris Black has amassed a database of three million Internet users for its experiment. By agreeing to take part in periodic Harris Poll Online surveys, poll subjects are offered the occasional chance to win prizes and cash awards.
For each survey, the company sends electronic messages to targeted participants inviting them to visit the Harris Poll Web site and answer questions. Participants are given passwords to ensure that they answer the survey only once. By next year, the company hopes to have five million potential respondents.
Mr. Black's confidence in the Internet is bolstered by a successful set of polls the company conducted in the run-up to last November's gubernatorial and Senate elections. The Harris Internet poll accurately projected the winners in 21 of the 22 races it tracked. "We won over a lot of people with those results," says George Terhanian, Harris Black's director of Internet research. "At least a lot of the doubters are now willing to listen."
Some of Harris Black's competitors -- the Gallup Organization, the Roper Center for Public Opinion Research and Yankelovich Partners -- have dabbled in various Internet-polling efforts, but none are as ambitious as the Harris Poll Online, perhaps for good reason.
Another source of controversy is that there are small but important political differences between wired Americans and those who remain unplugged. According to February data, a slight plurality of Internet users, 34 percent, claim to be independent voters, 31 percent say they are Democrats and 29 percent call themselves Republicans. But among all registered voters, Democrats have the edge: 35 percent claim to vote Democratic, compared with 29 percent who say they vote Republican and 28 percent who call themselves independent.
Some of the differences in online polls and their participants are evident in results of a poll America Online Inc. conducted of 118,000 of its subscribers last summer. The respondents in the AOL survey were less supportive of President Clinton than the general population in the months preceding his impeachment. According to the poll, 52 percent answered yes to the question "Should Clinton resign?"; 48 percent answered no. By contrast, three telephone polls conducted during the same period by Gallup, ABC News and CBS found that a majority of Americans (72 percent, 61 percent and 61 percent respectively) didn't wish the president to relinquish his office.
The AOL poll was unadjusted to account for the demographic disparity between Web surfers and the general population. However, Andrew Kohut, director of the Pew Research Center in Washington, D.C., says that adjustment of Internet polls isn't the solution to increasing online polls' reliability. "These differences are evident even when the online sample is statistically adjusted to account for the underrepresentation of some demographic groups," he says.
Mr. Black, however, is ready to accept Internet polling with open arms. He points out that about 45 percent of Americans now use the Internet, either at home, at school, at work or in public libraries, and says they are increasingly representative of the U.S. population at large. "The Internet looks more like America than it ever has," Mr. Black says. Taking issue with Mr. Kohut, he insists his researchers can adjust Internet polling data to compensate for imbalances that still exist online.
It is an adjustment, however, that the company still hasn't perfected. Last November, in the Georgia governor's race, Harris Black's pollsters projected Republican Guy Millner would defeat Democrat Roy Barnes. Instead, Mr. Barnes won. Observers agree that Democratic anger over the congressional GOP impeachment efforts spurred black voters to turn out in larger numbers than usual, helping to tip the scales in Mr. Barnes's favor. In making its projection, the Harris Poll overestimated the impact of third-party voters and failed to accurately tap the sentiments of black Democrats.
In Harris Black's defense, "unexpected surges in turnout by particular groups can confound pollsters," says Atlanta GOP pollster Whit Ayres. Still, Mr. Ayres's own telephone poll correctly picked Mr. Barnes as the winner.
For detractors, Harris Black's misjudgment in Georgia points to another subtle flaw of its Internet poll. African-Americans may cruise the Internet in numbers relative to their presence in the general population, says Mick Couper, an assistant research scientist at the University of Michigan's Survey Research Center. However, just like the rest of the online population, he says, "blacks who are on the Internet" also "are going to be richer, more educated and more informed politically."
Mr. Black is unfazed by the skeptics. "It's a funny thing about scientific revolutions," he says. "People who are defenders of the old paradigm generally don't change. They are just replaced by people who embrace new ideas."