Recriminations -- and speculation -- have already begun in the wake of Hillary's unexpected win in New Hampshire last night. The debate centers on why the polls were so wrong. David Kuo and Andrew Sullivan (among others) suggested that it might be the so-called Bradley Effect in which white support for black candidates is lower at the ballot box than in polling. However, as a friend pointed out to me last night (and as Matthew Yglesias also noted), the polls came close to predicting Obama's support. They were just way too low on Hillary. So where did the additional votes come from? There are some indications that it was a surge of previously undecided voters, which would be surprising because she is, in effect, the quasi-incumbent in the race, and undecideds usually break heavily against the incumbent.
PS: There's already a search on to find "moments" and "events" that explain the outcome (particularly the debate or Hillary choking up), but I think it's going to be hard to establish that those made the difference.
Update 1/9 9:53 AM: The Washington Post has a good rundown of the various explanations for what happened and the available evidence for them. The short answer: It's unclear.
Is political polling the only field of science where a failed prediction isn't evidence that they just got it wrong?
You're all arguing that Obama really had a surge and that Clinton rallied back at the final hour.
There's no real evidence for any of that.
Honestly, if climate scientists, biologists and physicists make a prediction and it fails, they don't start creating all sorts of ad-hoc explanations to save the original theory. They start over, modify the theory to match the observations, or simply admit that they don't understand the world as well as they thought they did.
Posted by: Jinchi | January 09, 2008 at 12:38 PM