The Associated Press has released a poll that shows a much closer race than any other poll out ther. Now, Liz "I know what kind of donuts John McCain really likes" Sidoti must have been gleeful that she was chosen to write John McCain's "comeback" piece. (The infamous AP/McCain donut video is below.)
But, the poll doesn't make sense on many levels. Here's one example: 45% of this poll's respondents are evangelicals or born-again Christians (this is on page 20 of the poll's crosstabs (it's a pdf):
The problem? In 2004, evangelicals/born-again Christians made up 23% of voters. But that same group makes up 44% of likely voters in AP's poll released today. That's almost double the number - it's totally implausible.
In 2000 and 2004, there was a very aggressive push for the evangelical vote. In 2000, when the question was asked "Do you consider yourself part of the conservative Christian political movement," 14% of voters said "yes" in exit polls. In 2004, when the question was changed to "Would you describe yourself as a born-again or evangelical Christian?" - the very category AP uses in its current poll - 23% of voters said yes in exit polls.
Did you get that? The percentage of evangelicals/born-agains voting in 2004 was 23%. The percentage of evangelicals/born-agains that AP included in their likely voters scenario is 44%. That's almost twice as many. Consider that 79% of evangelicals voted Republican in the 2004 presidential elections, and we can assume that anyone calling themselves "born-again" might be more prone to voter Republican. This means AP disproportionately skewed its polling towards the GOP base. So it's no surprise that the AP poll shows McCain doing better than in other polls.
With such an outlier, one wonders why the brain trust at AP decided to move ahead releasing this poll. But, the AP brain trust loves McCain. It's not just Sidoti. Remember, AP's Washington bureau chief almost went to work for McCain. He's in that donut video, too.
As Nate Silver so clearly explains, the likely voter models being used by some pollsters, including AP's partner, GfK, aren't making sense. In fact, Nate has issued a challenge to the pollsters who have a wide gap in their models:
I would like to issue a challenge to those pollsters like Franklin & Marshall and GfK which in spite of all the facts above, are showing a substantial shift toward the Republicans when they apply their likely voter models. E-mail me -- my contact information is at the top of the page -- and tell me why you think what you're doing is good science.I would like to issue a challenge to the Associated Press, too: Put aside your friendship with John McCain, please. At least for the next two weeks, be the objective new source you used to be.
