Wednesday, October 22, 2008

One last thing: The AP Poll

Once upon a time, the Associated Press was considered to be the most impartial of all news services. Since the wire service is where most newspapers get the bulk of their news content, it certainly would be a nice reality to live in.

Today, the AP released a national tracking poll of the presidential election that showed Obama leading McCain 44%-43% among "likely" voters. A statistical tie that Drudge--who's spent the last few weeks highlighting outlier polls (including a Nickelodeon online poll of children) showing McCain within five points--naturally jumped all over.

A statistical tie? you say. That's right, a statistical tie! But, you ask, don't most polls show Obama at or near 50%? Yes they do. So why does this one show him six points lower than that? That's exactly the question I asked the Internet, and somebody called AmericaBlog answered. Ready?

I'll summarize the biggest point: As with all polls of so-called "likely voters" (as opposed to registered voters) the AP is performing a little precognition to determine what the likely make-up of the electorate will be. There's all sorts of methodology, to figure out the general demographics, but "likely voters" generally reflect the make-up of the last election, making it, in my opinion and in this election, a bit of a problematic designation. Still, most polling agencies have some way of determining the what a likely voter is, and differences in methodology account for some differences in polling results.

Now, you say, this summary sure does seem to go on and on without saying much.

Keep your pantyhose on, I'm getting there. In this particular poll, 44% of the likely voter pool is comprised as self-identified evangelicals. If you think that seems like a big number, don't worry, history proves you right. As I said before, likely voter tabulations are determined, in part, by looking to the past (i.e. Exit polls from recent elections).

The problem with the 44% number is that it is nearly double the evangelical turnout from 2004, and nearly four times that of the 2000 election. In short, the AP is assuming evangelicals will make up nearly half of the over all electorate when the most they've ever amounted to is a little less than a quarter. If you consider that evangelicals voted Republican in 2004 by a ratio of 4:1, the AP, by doubling the evangelical proportional turnout from 2004, is essentially assuming that a group that gave Bush, a president they felt was one of their own, about 16% of the overall vote is going to give McCain, a candidate they generally mistrust, about 32% of the overall vote. Not 32% of his vote--one third of the entire electorate!

Anyhow, the point I'm trying to make, is that if you saw that AP Poll and felt a little worried. Forget about it. But still vote, okay? You know, just in case.

Update: Nate Silver, noted sports statistician and honcho at, discusses the discrepancy between likely voters and registered voters in a way that leaves little doubt as to whether or not some of these polls are a bit more suspect than they let on. Worth a read.

No comments: