Polls to watch
Iowa
Mason-Dixon
Des Moines Register
Aggregate average xARG
Rasmussen
New Hampshire
Mason-Dixon
Aggregate average xARG
Rasmussen
That is all, for now.
ARG is just wrong and should be ignored. Other polls vary in quality but in total should be pretty good.
ARG’s latest silliness includes McCain at 20% in Iowa. They’ve been wrong over and over again this year (and all previous years) compared to reputable pollsters and actual results.
Iowa
Mason-Dixon 12/3 - 12/6
Huckabee 32
Romney 20
Thompson 11
McCain 7
Des Moines Register
Nothing in December
Aggregate x ARG
Huckabee 30.4
Romney 26.4
Thompson 10.6
McCain 8.8
Rasmussen
Huckabee 28
Romney 27
McCain 14
Thompson 8
The polls after Christmas should show a good bit of movement
It is interesting to see the MOE for each poll. For example, according to ARG, Fred is at 5% in Iowa (MOE 4) so be might be as high as 9%. Another poll (SurveyUSA) has him at 16% (MOE 4.5) so he could be 11.5%. This is not overlapping with ARG poll so clearly some poll - by definition - is incorrect.
I suppose it is very difficult to have a reliable polling in small states. Still, some trends can probably be spotted and average is typically quite good predictor. (”Very highly likely voters” has also been useful indicator - but not all polls use that).
From my experience in the past, Mason-Dixon is the best. However, Ras was pretty close in 2006, and overall, all of the polls were dead on.