PhotoDude.com

Mon. Oct 18, 2004

Pondering the Polls

Over the next two weeks, we’ll likely be inundated with polls from various media outlets and other entities. It’s likely no two of them will say the same thing. We’ve heard they’re neck and neck, Bush is up by 2, Kerry has the momentum, and down the stretch Secretariat rises from the dead to win by a half length.

Personally, in this most volatile election year, I don’t think they’re worth the effort expended to create them (other than the fact there’s now an entire polling industry to support).

During my days in radio, I had to learn far more than I ever cared to know about polling and statistics due to the fact we lived and died by the Arbitron ratings book. Sample sizes (and sample controls) are critical, and I tend to ignore polls with less than a thousand respondents (preferably 1,200), as the margin of error and lack of a statistically significant sample potentially overwhelms the accuracy of any outcome. If you want to convince me about the breakdown of 100,000,000+ voters, and you can’t ask 1,000+ people, I’m not listening.

But when you’re polling voters, another element is often added; the poll of “likely voters,” a far narrower subset than simply “registered voters.” How the heck do they determine who is a “likely voter”? Well, one common way is to ask a simple question: on a scale of 1 to 10, what is your interest level in the election? Only those that respond 9 or 10 get counted in a poll of “likely voters.” To me, you’ve immediately skewed your results by sampling only the most partisan of all those questioned, i.e., those who’d likely answer 10 or 9 to that question. Frankly, my honest answer to that question would be “8,” and I’ve written tens of thousands of words here about the election. But I know many other people far more interested and active in this election than I.

However, there’s an even bigger flaw in these polls that that. In the 2000 election, 105 million voted. I’ve seen various sources predict a range of between 112 million and 120 million may vote this year. A couple of weeks ago, I read in the local paper that the Georgia Secretary of State’s Office was receiving new registrations at a rate of 40-50,000 per day. Just for Georgia.

Nevermind cell phone users and others excluded from current samples, no one is polling these newly registered people. They comprise 10% or more of the coming electorate, and it seems they are a complete unknown.

It’s also good to keep in mind that some of the polls were saying at this point in 2000 that Bush led by 52-39. Those numbers turned out to be off by a country mile.

And this time around, there’s 7-15,000,000 people showing up who didn’t in 2000. Despite those who’d like to convince you (one way or the other) that’s it’s all over but the counting, right now, we know nothing. Nor are we likely to until November 3. If we’re lucky.