Bernie Sanders has surged to a 14-point lead over Hillary Clinton in the Democratic primary in New Hampshire, according to a poll just out from Monmouth University. He leads with every demographic, including women, where he used to trail. He's got strong favorability ratings and appears to be well-positioned for the primary vote, now less than a month away.

That's a big improvement from a poll released by American Research Group earlier this week which had Sanders up only three points. Which was a big drop from a poll released last Friday from Fox News showing Sanders up 13. Which is vastly different from a poll released recently from NBC/Marist/Wall Street Journal which had Sanders up four.

So is the electorate in New Hampshire really that fickle? Sanders up big, then a close race, then up big again? Now the kicker: The ARG and Monmouth polls covered the same time period, as did the Fox and NBC/Marist ones.

This is what the polling looks like.


(We didn't include Martin O'Malley in this because the chart would have had to have been twice as big to include his sub-10-percent numbers.)

This is weird. There are 10-point swings in polls that are conducted over the same time period. So what's going on?

The answer is methodology, as our polling guru Scott Clement pointed out to us. The ARG and NBC/Marist pollsters called New Hampshirites at large, filtering down to those they identified as Democratic primary voters. Fox and Monmouth called from actual voter lists which were matched to telephone numbers where possible. Monmouth's group was identified as "registered Democrats and independent voters who participated in a primary election in the past two election cycles or voted in both the 2012 and 2014 general elections," as Clement pointed out to us.

We've color-coded the methodologies used in the polls below. You may notice a pattern.


The question is: Which methodology is more accurate? Does calling people who you know have voted in the past result in better predictions? Or does including a wider universe -- including people who might be inspired to head to the polls for the first time -- give a better sense of the electorate?

In 2014, as Pew Research reported last week, polling of people with demonstrated voter histories was more accurate. In 2012, though, there wasn't much difference.

Who you poll matters as much as when you poll. There's little question that Sanders has an advantage over Clinton in the state. It also seems to be the case that both Clinton and Sanders are doing good prep work to get voters to the polls. And it is also the case that polls will likely change dramatically as election day approaches.

If the election were held today, as the saying goes, Bernie Sanders would probably win New Hampshire by somewhere between 3 and 14 points. The election, though, is in February.