Attention: No matter what the AP says, Obama is not leading McCain by 7 points today, and there has not been a 12-point swing in Obama’s favor since the GOP convention. At least, such a thing can’t be determined by the most recent poll from AP.
I’ve said before on this blog that some things can be measured by polls, and some can’t. Of course, presidential preferences among voters can, but only if you do it right: among other things, you have to be careful to keep your sample constant, or at least credibly reflective of available public data. Newsbusters has uncovered the methodology of the last two presidential polls conducted for AP by GfK Roper, who has a history of this kind of sloppiness (and yes, I call it sloppiness…unlike Newsbusters, I doubt it was deliberate “cooking,” as they put it).
AP/Roper’s poll of 9/5-10 measured a party division ratio in that poll’s sample of 31Rep/33Dem/36Ind, which sounds about right coming out of a GOP convention (since not all states register voters by party, national polls ask respondents the party with which they affiliate or identify most closely, and that is a more subjective question than simple party registration).
The sample for their 9/27-30 poll, during which Nancy Pelosi was blowing the bailout vote with her ridiculous, spiteful, childish floor speech just before her bill went down, swung wildly in the Democrats’ direction to 29Rep/40Dem/31Ind. That measure is entirely counterintuitive, especially given that Barack Obama did nothing to distinguish himself during that negotiation and vote.
Democrats normally enjoy a national party registration advantage (they were ahead in raw numbers even before they lost both houses of Congress in 1994), and it’s hardly unusual for registration numbers nationwide to fluctuate from one year to the next, but a nine-point swing toward the Democrats in party identification, in three weeks? That just doesn’t happen, and for AP/Roper to try to sneak these poll results into the media stream while burying their party balance stats is inexcusable.
The AP/Roper poll after the GOP convention showed McCain ahead by 5 points. I would never argue that McCain hasn’t noticeably lost ground nationally since then, but this follow-up poll by the exact same pollster showing Obama now ahead by 7 doesn’t even pass the laugh test, when considered in light of the crazy swing in party affiliation. Unfortunately, it’s already been factored into the RealClearPolitics average, and become part of the conventional wisdom.