The Des Moines Register poll adopts an unprecedented new turnout model for the caucuses, and its new poll is out of sync with the other polling done in the race.
When you look at Democrats who last time were 80% of the turnout, Hillary wins with that group by 6%, 33 to 27 for Obama and 25 for Edwards. And as David Yepsen points out, had their pollsters used the 2004 turnout model, Hillary would lead by 29 to 27, figures in line with the other polls.
The Des Moines Register Poll this time has 40% independent voters and 5% GOP voters in the poll when past independent participation has been 15% in 2000 and 19% in 2004, and the GOP has generally made up 1% of the vote. So they are depicting an unprecedented departure from historically established turnout patterns in the caucus. Under their model, only 55% of the caucus goers would be Democrats.
The other recent polls all show Hillary trending up and leading or within 1 point of the lead, and many show her moving up from a substantial deficit to tie and having the momentum in this race.
Insider Advantage, 12/28-12/19: Clinton 30, Edwards 29, Obama 22 Zogby, 12/27-12/30: Clinton 30, Edwards 26, Obama 26 Mason-Dixon, 12/26-12/18: Clinton 23, Edwards 24, Obama 22 Research 2000, 12/27-12/27: Clinton 28, Edwards 29, Obama 29 ARG, 12/26-12/28: Clinton 31, Edwards 24, Obama 24 Insider Advantage, 12/28-12/19: Clinton 30, Edwards 29, Obama 22
So we do not see this poll as accurately reflecting the trends we are seeing in other polls, on our nightly canvasses or in our own polls, and voters should understand this is a very close race, and that their participation on caucus night could make all of the difference.