I've been reading dailykos all day and there's been a gradual collective conclusion that all the polls are inherently flawed by not giving Dean's supporters enough weight.
Basically, the polls use a flawed sample set. They call voters that have been registered long enough to show up in their lists, and some only call people that have participated in past caucuses/primaries.
The assumption is that all of the new voters that the polls don't catch would split among the candidates by the same ratio as their polled support. In other words, that every candidate has "x" percent of their support as new supporters, and that that x is identical to all candidates.
It's obviously a stupid assumption when there's reason to believe that there's a huge influx of new voters, and that these new voters are gravitating towards certain campaigns. Dean's claim that 65% of their 1's were new caucus attendees. The number itself is meaningless because we don't know what that percentage number is for other candidates. But we do know it made a lot of buzz, which implies that that number is higher than for other candidates.
If that's true, it means that the polls' methodologies are seriously flawed. I think the story after Iowa will be that the real loser of the Iowa caucuses will be the press and polling outfits.
Will the polling outfits change their methodology for future races? What could they do?