Howie P.S.: Markos downplays the DMR poll's recent accurate track record.
Polling the Iowa caucuses is notoriously difficult, and while none nailed the entrance poll results in 2004, the Des Moines Register's effort came closest by far:
Just hours ago, the Register released its final poll of the contest:
Des Moines Register. 12/27-30. Likely Democratic caucusgoers. (11/25-28 results)
Obama 32 (28)
Clinton 25 (25)
Edwards 24 (23)
Past performance is no guarantee of future results, and all that jazz. It's a poll, and one that doesn't seem to pick up the Edwards surge we've seen in other polls. The Edwards campaign is defensive about the results, which I guess is understandable but whatever. It's all spin at this point. This thing is tight enough that the second-choices can push anyone over the top. This poll has Richardson at 6% (much lower than in some other polls, which show him breaking into double digits), Biden at 4%, Dodd at 2%, and another 6% uncommitted. Throw in Kucinich's 1 percent, and we have 19 percent of caucus goers who can tip any of the top three candidates over the top. And that's assuming no deals are cut between candidates.
Despite the cheers of some candidates' supporters, and the wails of others, all this poll tells us is what all the other polls have told us -- any of the top three candidates can win, and anyone that tries to "predict" which one will do so has the same 1 in 3 chance of getting it right as anyone else.