Lying Liars

Visiting Pandagon, I re-encounter a crazy story that I’d heard before but never got around to posting about. No, Pandagon is not the source of the story, they just link to it, and also point out that it’s crazy. Via Yahoo News:

Republican officials expect the Democratic ticket — Kerry and running mate John Edwards — to receive a bounce in polls of about 15 percentage points after the convention. Their goal is to try to cut into that likely post-convention advantage by ensuring that media coverage of the Democratic convention includes the GOP’s views.

Clearly, the purpose of this story is for Republicans to wait until there’s a blip of less than 15 points and then say, “See! Americans hate Kerry/Edwards!”, even if there’s a +5 to +7 point blip. It’s the converse of the primary Bush-promotion stratagem (set expectations ridiculously low, and then shout triumphantly from the hilltops when Bush surpasses them by, e.g., committing only a few malapropisms in a speech/debate/interview).

In any case, where did the 15-point convention blip myth come from? As far as I can tell, it is spun out of whole cloth and earwax somewhere deep within the bowels of RNC headquarters.

For example, in 2000, the Democratic Convention was held from 8/14-8/17 and the Republican Convention was held from 7/31 to 8/3. Were there blips in the polls? Did they come even close to 15 points? Here are the data from Gallup, to which I’ve added a blue vertical line for the Democratic convention and a red one for the Republican convention:

If you look closely, if anything, there appears to be a short-lived spike in the Democratic numbers and drop in the Republicans’ numbers, after the Republican convention. Then, after the Democratic convention, Gore’s numbers drop and Bush’s jump — the exact opposite of the alleged 15-point convention blipTM.(*)

So by the Republicans’ own logic, this time around, if Kerry/Edwards fail to drop after the convention, then it will have been a smashing success.


(*) I suspect that if you search enough polls, you could probably find one poll such that on or around 8/17-8/14, the Gore increase plus the |Bush decrease| equals 15. Failing that, there’s probably a poll out there such that Gore increase plus the |Bush decrease| + the margin of error equals 15. Failing that, there’s probably a poll out there such that Gore increase plus the |Bush decrease| + 2 X (the margin of error) = 15. This is irrelevant: if it’s an effect that we should “expect” this time around then it should be present in most polls from the prior election, not appear only in carefully chosen outliers.