The Problem with Web Polls

By Deane Barker on January 24, 2004

Gay Marriage Poll Gets Annulled: Here’s a example of why Web polls are silly. The American Family Association put a poll on their Web site asking people if they were opposed to or in favor of gay marriage with ultimate plans to take the results to Congress. Of course, they assumed that with their sympathetic visitor base, they’d have great results. Sadly:

Against the wishes of the AFA and its members, the poll leaked to the outside. And soon, people like Gabe Anderson began posting it to blogs, social-networking sites such as Friendster and sundry e-mail lists. When Anderson posted it to his blog on Dec. 18, 2003, the anti-gay-marriage position was leading, with 51.45 percent of respondents opposing gay marriage or civil unions.

But with his posting, in which he alerted his readers to the poll, Anderson — and many like him — began to unleash the democratic power that the Internet promises, and which organizations like the AFA must have forgotten: the ability to bring people together to fight for, or against, a cause.

You have to wonder why the AFA would do this in the first place. Web polls are not even remotely scientific, so what results did they plan to take to Congress? Incredibly biased ones?

This reminds me of something that happened with Microsoft two years ago. There was a poll by ZDNet about which platform was better: J2EE or .Net. .Net was trailing, until…

Only 21.5 percent said they planned to use Microsoft .Net—even less than the figure (23.5 percent) planning to use neither. But by the time the poll closed, on January 5, the results had dramatically changed, with three quarters of voters claiming to be implementing .Net.

Sadly, there was a clear trail of an email campaign.

Several of the voters evidently followed a link contained in an e-mail, the subject line of which ran: “PLEASE STOP AND VOTE FOR .NET!” ZDNet logs include the Web address from where the e-mails were sent and showed that the people who followed that link all had e-mail addresses in the domain.

It got worse. ZDNet claimed to have proof of bots voting multiple times, and people casting multiple votes.

These things are obviously bad, but is there any problem with the email campaigns in either case? Companies put these things on the Web to get input, and can they realy complain about word-of-mouth? It’s anyone’s right to email anyone about something they think they’d be interested in.

You see, a real, scientific poll chooses its audience — either a random sample or a specific demographic. With a Web poll, you can’t choose your audience’ instead, the audience can choose your poll and can choose only to vote into something where they think the vote will project a view they want.

Which leads me back to an important point: Web polls are of no use to anyone. Period.