Better Poll Analysis Needed

I'm hoping I've been reading the Gallup polls wrong, because if I'm not, then this Washington Post write-up of the "who will vote for McCain if their favorite Democrat isn't the nominee" polls makes a fundamental error in the third paragraph: "A slew of recent national polls report Democratic 'defection rates' ranging from 18 to 35 percent, depending on the candidate. If the numbers held in November, it would be a big turnabout from recent elections."

Let's walk through this slowly. Current Gallup polls show the race is basically a 50-50 contest. In addition, the pollsters recently started asking voters supporting Hillary Clinton who they'd vote for in an Obama-McCain contest, and vice versa. Of each candidate's supporters, between 18 and 35 percent say they might vote for McCain. But each candidate only has support from half the Democratic party! Indeed, Gallup spells this out [emphasis mine]: "The accompanying table displays the percentage of Clinton supporters who say they would vote for McCain if the general election matchup were McCain versus Obama". This means that the range of possible defection rates is roughly between 9% and 18%, well within historical norms.

What's more, as has been noted elsewhere, these polls are close to meaningless in March; early polls showed half of McCain's supporters would vote for Gore in a Bush-Gore matchup, but by November Bush won 60% of McCainiac. Now, an we stop freaking out about defections and get back to pointing out that John McCain's proposal to bear the white man's burden in Iraq for a hundred years doesn't include any way to prevent the day of reckoning when we do leave?

Comments

[Shock Mouse]

Um what, "Bush won 60% of the McCainiac?" That's a horrible percentage (for Bush)! Also I don't see how that's much of a contradiction from the preceding phrase "half of mcCain's supporters would vote Gore". So between March and November, 40% of McCain supporters, or 80% of those who said they'd vote Gore, ended up actually defecting in some way. No that number does not make me feel comfortable at all.

Now I too hope the number of Dem defectors will go down as the party unites (and these reflections on past polls comfort me that Bush had higher defectors than Clinton or Obama face). But that just emphasizes the need to get around to unifying. Bush had 8 months to unify, the Dems right now are looking at 3.

[Nicholas Beaudrot]

Yeah, what you really want is an exit poll question that says "did you vote for McCain in the primaries". But the point is that Bush outperformed the March polls.

I get the feeling that people want this done by the end of the primaries. They don't want Clinton to drag this out if Obama is clearly ahead. Already Obama's closed the gap in PA from the high teens to the low teens. I think if PA is close and Obama wins NC big, then it's over.

[Ari]

Oh Nick, you and your numberrs. Don't you know that it's all about narrative, silly. You'll never understand John McCain's mavericky awesomeness if all you do is sit inside with your sliderule.

[Shock Mouse]

Certainly a May 6th end date is the current operating theory. I don't see why necessarily Clinton would end then, even if the undecided supers do come out against her. At the moment she is hopeless unless Obama crashes or she threatens to destroy the party (over things like getting Michigan counted) until they give her the nomination. After a blow out on May 6th she will be... hopeless unless Obama crashes or she threatens to destroy the party.

The ironic thing now is that having lost the pledged delegates, her only hope is Supers. Which means supers can end it by saying "no". Unfortunately that means a loss NOT at the ballot box, so if HRC or her supporters want to be bitter, they say it was back-room politics and not popular elections. Until HRC tells them to calm down, it will be incredibly hard to unite the party, for they will feel it was "stolen" from them by supers.