As part of Time's ongoing three-year series on the 20th century, visitors to the magazine's Web site have been asked to cast their votes for the greatest men and women of the century. When the votes for the "Warriors and Statesmen" category were tallied every minute or so over the last year, Mustafa Kemal Ataturk was often at the top of the list. He was also tops in the "Entertainers and Artists" category, as well as in "Leaders and Revolutionaries."
Just who is M.K. Ataturk? He happens to be the father of modern Turkey, and quite a popular figure there. But the greatest entertainer of the 20th century? That's a stretch. (Time says it does not use the online poll to make its choices for the magazine.)
Similarly, many of the Internet polls this past summer and fall reported 70 percent or more of respondents supported impeachment and Clinton's removal. An interesting set of numbers, but they signify nothing-not even the views of Internet users as a group. All of the scientific surveys of American public opinion-from Gallup, CBS News, NBC News, and others-have found majorities of Americans are against impeachment.
In fact, most Internet polls -where you surf to a Web site and then cast your vote with a click of the button-are good for a few laughs, but little else. The latest in a long series of junk masquerading as polls and indicators of public opinion (joining the 900-number call-in polls and the mail-this-coupon-back-with-your-opinion surveys), these polls are popular in the broadcast, print, and online worlds because they give the illusion of a link between the reader/ viewer/user and the particular media: "Let's hear what you have to say" and "The People Speak" are common come-ons. (The Web sites have another reason: There are usually ads on the pages where ballots are cast and again on the pages displaying results. Thus, lots of people voting means more ad revenue for the site.)
But all of these efforts are devoid of meaning for one reason: Only when you can calculate the probability of any given person being sampled for a poll can you say a poll is representative of everyone's opinion. Since there is no such probability function when people choose themselves to vote in an online poll, the results reflect absolutely nothing.
And given that many of the Internet polls make only limited efforts to keep users from voting over and over again (the electronic equivalent of ballot-box stuffing), small groups of people with a point of view or an ax to grind can dominate the results. That is exactly what happened in the Time survey.
Several organizations are trying to figure out how to use the Internet as an accurate poll data collection tool, but no one has succeeded yet. First, defining how many people "use the Internet" is a slippery concept. Does that mean everyone with e-mail? Everyone who can surf the Web? Or those who use a computer at the office, but can connect only to their own company's Intranet site? Second, there is no list of everyone "on the Web" as there is, say, of people with telephones, nor any way to randomly contact a sample of that list.
In line with the hype about the huge number of Internet users and the great potential of the medium, the defenders of online junk polls are quick to say that 10,000, or 25,000, or even 100,000 people participate in their surveys, so they must mean something. But big numbers don't change the reality of the junk. After all, more than 10 million people responded to the Literary Digest poll in 1936. That big number didn't mean much when the poll predicted Alf Landon would defeat Franklin D. Roosevelt in the presidential elections.
(P.S. By mid-October of this year in the Time online survey, Ataturk had fallen behind some guy named Winston Churchill.) n