It may very well be garbage in, garbage out tomorrow but it is very, very unlikely. These polling outfits know what they are doing, they are professional statisticians and social scientists. It would be an insane statistical anomaly for 40 or so polling outfits to all be getting similar results....we’re talking 1 in millions. The only alternative is that the thousands of employees of all the polling outfits are involved in a very large conspiracy to depress the conservative vote but keeping that quiet when the incentive to leak is so high is probably 1 in a quadrillion.
Sample sizes of Party ID do not get ridiculous to me unless people screw with them too much..aka...weighting. One thing I have tried over and over and over to get across to people is that Party ID is an attitude or a feeling, NOT a rigid identity like American or Canadian or Black or White or Hispanic. The mushy middle will call themselves Republican or Democrat VERY often based on how they vote. The party ID FOLLOWS the vote and not the other way around.
In other words, if Bush is up 5 points against Kerry in Oct 2004, you will see a Dem party ID advantage of only 3 points. If Obama is up 7 points in Oct 2008 you see a Dem party ID advantage of 6 or 7 points and if you see the Republicans up 6 points in the 2010 elections you see party ID tied.
The party ID is very fluid and not rigid at all and to screw with it too much, you risk becoming the outlier. This is exactly what Gallup and Rasmussen did in 2010:
http://www.realclearpolitics.com/epolls/other/2010_generic_congressional_vote-2171.html
As you can see, Gallup and Rasmussen were way off that year. They saw the enthusiasm gap in their data and over-corrected the party ID thinking there MUST be a republican advantage when the party ID was actually tied. But since dems usually outnumber republicans in registration by 3 to 10 points historically for over 30 years, a party ID tie was a HUGE advantage for Republicans already...there was NO NEED to weight the data.
Rasmussen and Gallup are doing the same thing this year and tomorrow we will see if they screwed up again or are doing something novel and genius. I honestly think they are screwing up, they will be off by at least 3 points (possibly more) based on what I’ve plugged in to SPSS many times today.
Your comparison of a congressional race and presidential race is invalid. They are two different animals.
Rasmussen and Gallup are two of the most accurate pollsters for presidential elections.
Are you aware Rasmussen was the MOST ACCURATE for the 2008 presidential election?
You’re giving the pollsters too much credit. Most of these pollsters are run by liberal newspapers. Party identification is VERY IMPORTANT. So is enthusiasm.