Okay, let’s return to those late congressional generic polls. I think we now have all the polls likely to be released before the election. The total is seven. And one obvious pattern shows up. The two polls released on Saturday had big Democratic margins in line with recent polls (Time, Newsweek). The three released yesterday showed substantial or dramatic tightening of the race (Pew, WaPo, Gallup). And then the two released today showed the race back to the big Democratic leads of the past few weeks (CNN, Fox).
Time 2-3rd: D+15
Newsweek 2-3rd: D+16
Pew 1-4th: D+4
WaPo 1-4th: D+6
Gallup 2-5th: D+7
CNN 3-5th: D+20
Fox 4-5th: D+13
The pattern that jumps out is that the polls in the middle period — particularly those that included the fourth but not the fifth — show the big tightening. The ones before and after show the big Dem leads.
The first thing to point out is that there just isn’t enough data here — at least not in the toplines — to say whether this is really a blip in public opinion on the days in question or whether the different polling outfits are just using different likely voter methodologies and that difference happens coincidentally to show a illusory chronological pattern.
I’ll be eager to hear from more poll savvy folks, once they’ve dipped into the internals, about what to make of all this.
(ed.note: Notations about the dates of the polls were hastily pulled together from media reports. In some cases they were ambiguous. If anyone has corrections, please send them in. I was unable to find the dates of the polling for the Newsweek poll — as noted by the question mark after the date. So I’ve inferred polling dates of Nov. 1-3rd from the date of release, Nov. 4.)
Late Update: Pollingreport says the Newsweek poll was conducted on the 2nd and 3rd. So I’ve changed the above stats accordingly. Thanks to TPM Reader MB for the tip.
