Another wacky poll from Gallup, this time showing Bush with an eight-point lead among "likely voters", but a thirteen-point lead with registered voters. There is no info provided about the partisan breakdown of whom they chose to interview, but since this is almost a reverse of their previous post-convention polls, which showed Bush with a larger lead among LV's than RV's, it stands to reason that in order to now give Bush a significant lead, Gallup has to find a sampling base where Republicans outnumber Democrats by close to a dozen points.

Or to put it another way, it is as if Gallup chose to poll only in the state of Montana, and use those results to extrapolate data for the entire nation. Now, it might well be that the national political landscape, post-9/11, now looks like Montana did four years ago, with a massive, historical partisan shift poised to give Bush a landslide victory reminiscent of FDR over Alf Landon, or Reagan over Mondale. Since Gallup's state polls (and other state polls, collected here by Donkey Rising) actually show a race similar to 2000, with Kerry comfortably ahead in California and New York, and swing states like Ohio and Florida within the margin of error, and since Gallup's national pre-election polling in 2000 also sucked, I'm going to stick my neck out just a little and predict that is not the case, and Kerry will probably pick up more than a dozen electoral votes.

Now, it may be easy to laugh off Gallup's quadrennial folly, but this poll (and others like it) have very serious tactical consequences. One of the ways in which polling data is used right now is to demoralize the side that's losing. If the punditocracy can convince enough people that Kerry is toast, or that he needs a miracle, perhaps in one of the debates, to turn around the election, the public perception that he can't win will set in. Through the use of rigged polling data, enthusiasm for the trailing candidate is diminished, his crowds dwindle, and his supporters marginalized and effectively silenced. And that is why the partisan breakdown of the polling sample is so important to know, and why any poll that doesn't provide that information be looked at with extreme skepticism.