Interestingly, not even Raz saw the level of discontent in 1994. No one saw that coming.
A bit of a lesson in polling.
The talk you hear of oversampling Ds or Rs or Is to some extent doesnt mean anything, and to another extent means everything. There are three flavors of turnout model.
1) The turnout mix of the last presidential election.
2) A new measurement of party affiliation done periodically and recently.
3) A semi random sample (non random only to the extent of being sure 80 of 100 telephone prefixes are not the same and providing a corrupted, accidentally non random mix) is taken. The respondents self identify party and that is what is reported.
The vast, vast majority of the polls showing big Obama leads are using (forcing) a partisan mix that fits 2008, which was 6 weeks after Lehman and very likely an aberration. If you used 2010s partisan mix, youd see a huge Romney lead.
As best I know, Gallup uses flavor 3. Totally random, respondents self identify and the samples partisan mix is reported as what the random measure got.
Not sure, someone will correct, but I think Rasmussen does a recent partisan mix measure of the population.
Regardless, no matter who you are as a pollster, no matter how honest and pure your measurement, if you use 2008s turnout as the turnout model for your poll, youre going to show an Obama lead.
Michael Barone ( and I did ) :-)... He called it to the numbers, I was off a few seats in each house :-)...
I think Frank Lutz was Advising Mr Newt, and told him to prepare to be Majority Leader, and I think even the Newster had trouble believing it...