One thing that everyone is overlooking in the New Hampshire polls is that they were largely correct--No really:
Pollster.com is run by two academic pollsters with Charles Franklin creating estimates of the polling based on aggregating the different polls.
Obama estimate 36.7% from Pollster.com
Obama actual vote: 36%
Edwards estimate: 18.4%
Edwards actual vote: 17%
There are several reasons to think racial slippage didn't happen in New Hampshire. That the polls nailed the two other major candidates should tell us that what happened wasn't strange in terms of the polling, but what happened after the polling was complete. We had a significant event after most of the polling was done with the Hillary moment that was covered on all the national media including the top three national newscasts on the networks.
There are rules of thumbs about how undecideds break and while those rules of thumbs were not accurate this time, there's a very good reason that may be the case--a flurry of coverage over Clinton that was unsympathetic, but at the same time created a sympathetic backlash.
On top of that, we are talking about a Northern State in a Democratic Primary where the effect is the least likely to be observed. And it is far from a universal effect with the opposite generally being observed in Illinois even during general elections. Obama and Moseley Braun have done better in primaries than the final polls demonstrate and I believe Jesse White has as well, but I don’t have the numbers handy.
Then again, the national media is still howling about the polling being so off by talking about the gap between Clinton and Obama when the support for Obama was dead on.
Can we get back to Tweety’s sexism please?
All this talk of demonstrating that the polling was accurate but didn’t account for how fluid the race really was is growing boring. 😉