Replies

And it's sheer nonsense. The ground game tells us this. Gallup can "observe" whatever he wants to, but the real message we get on the ground---and apparently OK STATE, you never bother to walk a precinct or do any "poll flushing" or make calls, so you might not know this---is that the GOP's ground game will be pretty close to what we had in 2004, and the Dems will not even be as good as 2004.

And let's please remember Gallup's organization was infiltrated by British MI-5 in WW II and produced bogus polls without even knowing it.

And lets not forget Gallup in 1996 was one of ALL the pollsters who was dramatically off to the left and was guilty of oversampling Dems.

It's the same story, over and over. These guys have badly flawed methods that never seem to show up until election eve, when they hustle to get in line with something close to the real electoral verdict.

"And lets not forget Gallup in 1996 was one of ALL the pollsters who was dramatically off to the left and was guilty of oversampling Dems. "

That's untrue. Gallup has Clinton winning by 11 -- he won by 8, well within the margin of error. The only poll that eleciton to be outside the margin of error was conducted by CBS News.

LS, you are right about the polls prior to just before the election. In all the recent elections, Republicans "came from behind" to close the gap (although not in a monotonic fashion, as sometimes the Democrat rebounded as Al Gore did in 2000). The fact that polling companies continue to survey through the campaign, enables them to claim they are "right" (meaning, within the "margin of error") a high percentage of the time (but not as frequently as sampling theory would imply).

Based on this history, we can suppose the following: 1. For whatever reason, Republicans tend to do better than early polling "predicts;" and, 2. Polling, thusfar, is a bit biased even among good companies (such as Gallup [nationwide] and Mason-Dixon [statewide]) (as distinct from Research2000 and - ohmyGod - the Minneapolis Star-Tribune).

With respect to 1996, the fact that more than a dozen reputable polls overpredicted Clinton's number means that their combined poll was well out of the margin of error for the implied sample (of about 10,000). The same thing occured with the Exit Polls of 2000 and 2004. The odds that these polls have overpredicted the Democrat vote by more than the margin of error in three consecutive Presidential races are too low to be dismissed as coincidence.

In fact, the consortium involved in the Exit Poll has itself concluded that their poll suffers some bias, which they are hoping to rectify by having their survey-takers "look like America," instead of being predominately liberal Democrats. However, I don't think the problem is isolated to the fact that their bottom people are predominately of one persuasion, but that all their people are predominately of that persuasion.