Pollsters in Iowa are searching for answers after they failed to predict Ted Cruz's victory over Donald Trump and Marco Rubio. It might even be worse next week in New Hampshire.

Pollsters and other observers surveyed by POLITICO this week pointed to a number of reasons why New Hampshire is such a uniquely difficult state to poll. Chief among them: The volatility in the electorate that persists right up until voting begins.

"People are moving around from candidate to candidate," said Dante Scala, an assistant professor at the University of New Hampshire and author of a book on the state's first-in-the-nation primary. "I think if you called the same person in two successive nights, you might get two different answers of who they like."

Recent polls indicate that Rubio, the first-term Florida senator who placed third in Iowa, is the early beneficiary of the post-caucuses momentum. Early tracking polls – a daily tracker launched this week by the University of Massachusetts-Lowell, and a small-sample, post-Iowa survey from the University of New Hampshire – show Rubio's numbers ticking up.

While the changes aren't huge at the moment, they are significant in this regard: Rubio is seeking to parlay his momentum to break away from a pack of candidates who had been jockeying for second place behind Trump, a group that includes Cruz, Jeb Bush, John Kasich and Chris Christie. (The early post-Iowa surveys show Cruz slightly ahead of Bush and Kasich, with Christie sagging lower.)

But before Rubio starts writing his victory speech Monday night, it's worth recalling the most famous example of the failure of pre-election surveys to predict the outcome of the New Hampshire primary.

On Jan. 3, 2008 – the day of the Iowa caucuses – Hillary Clinton led Barack Obama by 7 points in the RealClearPolitics average of New Hampshire polls. And then Obama won Iowa.

Despite just four days separating Iowa and New Hampshire, Obama quickly surged in the polls, opening up an 8.3-point lead in the final RCP average.

But something happened on the way to an Obama sweep of the first two states: Clinton stormed from behind in the race's final 24 hours, leading pollsters – including the industry's professional organization, the American Association for Public Opinion Research – to conduct thorough reviews of industry practices.

Their most basic conclusion: Polls that stop days before the primary miss important shifts in voter preference that occur right up until Election Day.

"Late deciders haunt everybody," Scala said this week, recalling his expectation in 2008 that Obama would run away with the primary. "I remember watching the actual returns come in and realizing half an hour in that wasn't going to happen."

It's a lesson the pollsters who survey New Hampshire are taking to heart. Late-deciding voters are the biggest reason why pollsters are lowering expectations that the pre-primary surveys will match the results next Tuesday.