Web polls let you exercise instant sway -- over meaningless numbers
At BigVote, a site devoted to the kind of live "interactive polling" that's all over the Web these days, they urge you to "Make a Difference!"
So I did. The first question on BigVote's politics page is: "Would you vote for Dole?" As it happens, I wouldn't, so I clicked on the "no" button. The site then informed me that 45 percent of voters shared my dislike of Dole, but 55 percent would support him.
That just wouldn't do, so I clicked on "no" again. And again. I clicked away until I boosted that sucker to 46 percent.
It took maybe ten clicks for me to make a difference. If I cared enough, I guess I'd have clicked until I made even more of a difference, but life is short, and RSI is no fun.
Live polls have proliferated across the Web because they're easy for programmers to script and they offer static sites a quick fix of pseudo-interactivity. Trouble is, they don't mean anything; they are to real polling what your local paper's syndicated daily horoscope is to serious astronomy.
Professional polling isn't exactly a foolproof science, to be sure. When I worked a summer job two decades ago for Penn and Schoen Associates -- then a small startup Democratic polling operation and now a big firm whose clients have included the Clinton White House -- I learned the basics: get 1000 people to answer your questions and you can be pretty confident of a poll's accuracy.
If, that is, you phrase the questions effectively. And if your telephone inquisitors ask them clearly. And if you program your computers to analyze the data sensibly.
Much can go wrong twixt the query and the result. But by far the most important factor in a poll's success is the quality of the sample. Any old 1000 people won't do; they have to be carefully distributed geographically and demographically, and that takes work. It turns out, for instance, that it's much harder to collect usable responses from lower-income zip codes. If you can't get enough responses from a particular area or group, you can always "weight" the ones you do get to carry more influence in the final tabulation -- and hope that you aren't compromising your results too much.
Polling, in other words, is something of a rough art. But at least the aim is to come up with a result that mirrors reality. The kind of polling that's rampant on the World Wide Web today is a pathetic travesty -- in which the goal is simply to foster an illusion of involvement in what is inevitably a meaningless enterprise.
Most obviously, few Web polls are able to enforce any kind of "one man, one vote" rule. Some sites require you to register before you can vote, but since virtually none of them can charge real money and few actually bother to authenticate e-mail addresses, you can register over and over again almost as easily as you can vote over and over again. Many, like Rock Mall, are reduced to politely requesting, "Please cast only one vote per day" -- as if Bon Jovi's fans are going to restrain their ardor.
But even if you could keep the masses from voting early and voting often -- and even if you could rig your ballot so that you wound up with a good geographic and demographic distribution -- you're never going to be able to conduct a valid, trustworthy poll from a vote-tallying Web page for one simple reason: any Web-based sample is self-selecting. And the people who bother to vote will, most likely, be people with an extra passion for or against the question at hand.
This principle is well understood at places like Pat Buchanan's web site, where, next to a hotlist of Web poll sites, you'll find the exhortation: "OK Internet Brigade, your orders are to get out in full force... Go to these Web sites and show your support for Pat!" For that matter, you can visit the Pat Paulsen for President Florida web site and find that an astonishing 45 percent of respondents polled there support Paulsen's candidacy. Stop the presses!
These flaws in Web polls seem pretty obvious, yet polls continue to turn up as regular features at high-profile sites, like ESPNet Sportszone, Hotwired's Netizen and C|Net -- where at least they require you to use an authenticating password. (Yes, we had a brief fling with daily polling here at Salon, too, before deciding that there were better uses for our bits.)
Worse, some forms of Web-based polling are beginning to be taken seriously by old-line journalists. Earlier this week, the New York Times reported -- with the bemused air of slumming anthropologists -- on the voting preferences of a strange new "bloc" known as "netizens." These Net natives "turn out to be fiscally conservative social libertarians who would take money from veterans to support endangered species."
There are plenty of such characters floating around the net, no doubt about it, and plainly many of them chose to participate in Crossover Technologies' Reinventing America, the budget-simulation game from which the Times drew its data. But the Times story, although labeling the game's 500 devotees and thousands of casual participants a "less-than-scientific sample," nonetheless dubbed it "a good look at what the folks who surf the Net under screen names want to do politically."
Reinventing America is in many ways a smart, fun site -- but, like the National Budget Simulator, its benefits are in education through hands-on involvement more than in gauging group opinion. To treat its results as any kind of indication of political attitudes on the Internet, let alone in the nation at large, is to mistake the enthusiasm of a group of volunteers for the mood of their entire community.
Most Web-based polls come with boilerplate rhetoric about participation: The National Opinion Registry calls itself "a medium designed to give your opinions real clout...for those of you who really value your opinions and want to influence the social and political affairs that affect your life and the lives of those you love." But since the most recent question listed at the time of my writing was "NEW! - What did you think of the President's State of the Union Speech Tuesday 01/23/96?," there doesn't seem to be anyone awake at the other end of the line.
Very few Web polls -- with the possible exception of fan extravaganzas, like the Star Trek polls, and excursions into contemporary mores, like the Potty Poll -- are motivated by a desire to find out what people think. At best, they're attempts by Web developers to pay lip service to the new medium's "interactivity." Ironically, though, these polls transform participants right back into what the old media always treated them as -- statistics.
In more and more cases, the Web poll phenomenon is, even more perniciously, a Trojan horse: While these Web sites are pretending to listen to your opinions, they're really just compiling your demographics. eVote, for instance, announces that "eVote Polls are not simply popularity contests. All results are demographically analyzed and corrected to represent a true cross-section of the U.S. population, with standard margin of errors comparable to traditional national polls." That doesn't deal with the problem of self-selection, but it's a start. To register with eVote, you fill out a full questionnaire on your age, residency, education, income and more -- then you get to vote on the issues of the day.
The only trouble is, there's nowhere on the eVote site that reports those "demographically analyzed and corrected" poll results. Which, on today's hyper-competitive Web, leads one to wonder whether eVote cares what we think of Dole and Clinton -- or just wants to deliver our demographically sorted eyes to its sponsors.