The Loras result is generally in line to the latest polls IA from Suffolk, CNN, PPP, Gravis and Marist. It is possible that the DMR poll is superior and has a unique insight that the others do not. But it still is an outlier.
Selzer's always been uniquely astute at predicting the voter universe in Iowa. The 2014 midterms are an obvious example, and in 2008 she correctly predicted the margin of Obama's victory by assuming that the majority of caucus goers would be first-time caucus-goers and that a sizable number would be independents.
I assume she's weighting a higher number of independents and first-time caucus-goers than other pollsters again, which helps Sanders and hurts Clinton. That said it remains to be seen whether a) the majority of independents partake in the Democratic caucuses when it's the GOP field that is sucking up all the media oxygen (unlike 2008) and b) Sanders can build the sort of GOTV operation Obama had to get first time voters out on caucus night. I'm not sure we'll know till the winter.
