Why Clinton victory doesn't compute

The polls had Hillary Clinton losing by 9 percentage points in New Hampshire. She won by 3. So why were the polls off by such a wide margin?

Talking heads have been asking this question all morning - and then not answering it. Instead, they point to demograpic slices of the electorate from the exit polls, such as the data showing that most women voted for Hillary. Or they speculate that the now famous Hillary-getting-all-weepy-in-the-diner incident suddenly changed minds.

The former explains who voted but not why the polls were wrong. The latter is just plain bunk. Much as the media would like to believe it's so, the election didn't turn on one 11th hour teary sound bite.

I can see two possible reasons why the polls were wrong:

1. The polls were flawed, or

2. There was a sudden shift in voter preferences at the last minute.

There is some data to support the theory that in the the very high voter turnout in New Hampshire included disproportionately larger or smaller representation of some groups (women, young people) than what was reflected in the survey samples.

But the polls were also changing very quickly. Witness Barak Obama's last minute surge in the the last two polls over the weekend. In an unpredictable race with so many undecided voters, it's possible that the ground simply shifted again.

Then again, the polls did very well in representing the final results for the other candidates, such as Edwards and McCain.

Conspiracy theorists might want to chime in here with e-voting fraud explanation. But then, much of New Hampshire still uses paper ballots.

So what is the truth?