August 04, 2004

Saletan's Latest Misunderestimation

Will Saletan tries valiantly to spin the post-convention polls in Kerry's favor. His analysis is flawed.

Throughout, Saletan ignores likely voter figures; even where they're available, he uses registered voter data and links to pages where the likely voter data is missing. But as I mentioned in yesterday's AmSpec column, polls that measure likely voters are generally most accurate. (Elsewhere on Slate is a good explanation of how likely voter models work.)

The entire "What's changed" section is based on registered voter polls; to be fair, I've had trouble finding the internals he talks about here from the CNN/USA Today/Gallup poll broken down by likely voter, though Gallup does use likely voter measures-- see here for the first version of the poll Saletan refers to (USA Today extended polling for a day to double-check the negative bounce).

In the "Trial heats" section, Saletan writes:

Before the convention, Bush led Kerry 48-46 among registered voters in the ABC poll. After the convention, Kerry leads 50-44.
But that poll also measured likely voters, who came out with a far less impressive-- statistically insignificant, in fact-- 49-47 lead for Kerry. (It's typical for likely voter tallies to look better for the Republican; Republicans are usually more likely to vote.)

Saletan continues quoting registered voter tallies:

In the CBS poll, Kerry turned a 45-42 lead into a 48-43 lead. The CNN/USA poll goes the other way, boosting Bush from a 47-43 deficit to a 48-47 lead. That's counterintuitive, given the pro-Kerry media coverage around the convention. It doesn't square with the CBS or ABC polls. Nor does it square with an American Research Group poll, which bumps Kerry from a 47-44 lead to a 49-45 lead, or a Newsweek poll—taken on the last night of the convention and the night afterward—which bumps Kerry from 47-44 to 49-42. So my guess is that the CNN poll is off the mark.
That Newsweek poll measured adults, not registered voters, an even more inaccurate method. And of course the counterintuitive poll is closer to ABC's likely voter tally than the rest.

In the "Locking up support" section, Saletan argues that Kerry's base is more solid than Bush's. Once again, he's looking at registered voters; as I mentioned at the beginning of yesterday's column, the "may change mind" supporters on either side are about the same size among likely voters.

The one perfectly fair point that Saletan makes is the "Bush's flat line" section: Bush's approval ratings are consistently lower than those of presidents who've won re-election. (The Zogby poll he refers to even measures likely voters.) But Bush's ratings are also higher than those of presidents who've lost, so this proves less than Saletan implies.

In 2000 Saletan made the embarrassing-in-hindsight declaration that "Bush is toast," and he cautions that he's not going that far now. But declaring a race this tight "Kerry's race to lose" is going too far already.

UPDATE: Since Mickey Kaus seems to read this post as somewhat more bellicose than I meant it to sound ("John Tabin snipes at Saletan"), I should mention that I like Will Saletan-- he's a smart guy, and a class act. But on this one he's wrong.

Posted by John Tabin at August 4, 2004 07:06 AM

As a Bush supporter I have previously been discouraged by the drumbeat of the media. They consistently say undecideds are very few and will break for Kerry. (I know no undecideds.)

Posted by: lee tabin at August 4, 2004 09:08 AM

Dad: I wouldn't completely dismiss the CW that undecideds break to the challenger, but that doesn't tell the whole story; there are also the "approval gap voters," those who give Bush positive marks but still lean toward Kerry. I wouldn't be surprised to see Kerry shed enough support from that cohort to cancel out his advantage among undecideds.

Posted by: John Tabin at August 4, 2004 10:43 PM

What particularly bothered me is that Saletan wanted to throw out the CNN result - this suggest a profound misunderstanding of the polling process and statistics (or intelllectual dishonesty, but I won't go there). Each individual polling will likely change by an amount similar to the statistical error, even if nothing has changed. This is because different people are sampled each time. The CNN result is not at all inconsistent with the other polls. Saletan is simply engaging in wishful thinking.

Posted by: Lone Con at August 5, 2004 01:23 AM

Saletan can't imagine the circumstances under which he would change his vote, therefore he can't imagine the circumstances under which anyone in the entire United States would, and assumes that Kerry has a base floor and Bush a base ceiling and there's a narrow slice of true undecideds in between. I don't think there's any reason to believe that either of those phenomena actually exists; I can easily imagine scenarios such as a strong Bush performance and a weak Kerry one in the wake of a terrorist attack which could significantly move voters in the middle toward the Bush camp. Less ominously, I can imagine a general dissatisfaction with the Kerry alternative that erodes his floor as he proves to be weak on defense issues and excessively soft on old school Democratic issues-- as his convention speech was-- and he ends up being done in by the drip, drip, drip of deflating hopes in favor of the guy they already know. I can also imagine opposite scenarios in which Bush blows it in similar fashion, which I guess means I have more imagination than Will Saletan....

Posted by: Mike G at August 5, 2004 11:21 AM

I enjoy reading Will Saletan but I have to admit I use him like a reverse barometer when it comes to political pronostication. Aside from the infamous "Bush is toast" comment of 2000' he didn't do any better predicting a net republican pick up of seats during the 2002' congressional elections. If I was a Kerry follower I would take his predictions with a big grain of salt.

Posted by: gk at August 5, 2004 01:34 PM