More on NH polling

Charles Franklin and Mark Blumenthal have some good posts at Pollster.com trying to explain what went wrong in NH for pollsters. Some takeaway points:

• The polls did not overestimate Obama's levels of support. Rather, they greatly underestimated Clinton's. (This would tend to negate the race effect I mentioned previously.)
• The polls predicted the GOP share of the vote pretty well, suggesting that there was nothing systematically wrong with the way polling firms conducted their surveys. For some reason, they just really screwed up the Clinton predictions.

Why did the polls miss Clinton's share of the vote by so much? They offer some theories, including a late surge for her among undecideds and even a reluctance among Clinton supporters to answer polls.

Update: More on this topic from John Sides.