WASHINGTON (AP) — Voters might want take all those presidential polls coming out of wintry New Hampshire with a grain of road salt.
The Iowa caucuses showed that survey-takers had a tough time pinning down the state's fickle electorate, and pollsters face some of the same challenges in next-up New Hampshire.
A polling primer:
Most poll-takers took it on the chin when Donald Trump's lead in pre-election polls ended with a Ted Cruz victory at the Iowa caucuses. Marco Rubio's strong third-place showing among Republicans seemed to sneak up on pollsters, too.
They did better at capturing a tight race on the Democratic side that favored Hillary Clinton. Ultimately, she squeaked out only the slimmest of victories over Bernie Sanders.
Those with good memories will recall that it's not a rare occurrence for pollsters to miss the mark in Iowa.
In general, election polls are trickier than most surveys, and polling on caucuses is tougher than in primary states.
On top of that, the whole polling business is facing big challenges caused by people's increasing reliance on cellphones, declining survey response rates and growing difficulty in identifying likely voters.
"Polling is much harder to do well than it was four and eight years ago," says Cliff Zukin, a Rutgers University professor and former president of the American Association for Public Opinion Research.
In December, Zukin wrote an election primer warning that "2016 will present election polls with a very challenging environment."
Voters in Iowa debated their options until the last minute — even switching sides mid-caucus. But most polling ended several days before the caucuses, missing late movement. Caucus-goers who said they decided late tended to break toward Rubio and Cruz. Among the 45 percent of caucus-goers who decided in the final week, 29 percent went for Rubio, 27 percent for Cruz and just 14 percent for Trump, according to polls conducted by Edison Research for The Associated Press and the television networks as Iowans entered the caucus precincts.
"Polls are very good at telling you the situation when you take them," says Peter Brown, assistant director of the Quinnipiac University Poll. "They're less accurate about telling you what the situation will be next Thursday. They're god-awful about two weeks from Tuesday."
For all the talk about record turnout in Iowa, only a sliver of the state's voters attend caucuses, making it difficult to nail down who will show up. "When you have low-turnout elections, a little volatility could have a major impact," says Patrick Murray, director of the Monmouth University Polling Institute. And the 2016 race has been nothing but volatile.
Also, conventional wisdom had it that high turnout Monday would signal lots of new caucus-goers showing up for Trump. It turned out that a big share of the increase in turnout came from sporadic caucus-goers who were motivated to come out as the Rubio and Cruz organizations worked off of newly developed lists of past caucus-goers going back decades, says Murray.
In races with lots of candidates and lots of information, it's harder to predict which direction undecided voters will ultimately tip. That's always the case in politics-obsessed Iowa and New Hampshire, but it was an even bigger factor with this year's crowded and cacophonous field. "The level of indecision is unprecedented in any election that I've ever polled," says Murray. He talked about running into the same voters in New Hampshire on multiple occasions, and "every time I talked to them they were leaning toward a different candidate."
Pollsters undershot on turnout among evangelicals in Iowa. A whopping 64 percent of GOP caucus-goers were evangelicals, and a third of them favored Cruz, compared with 22 percent for Trump.
One of the final major polls of the Iowa campaign, by The Des Moines Register and Bloomberg, had Trump leading Cruz by 5 percentage points, based on a pool of likely caucus-goers that included 47 percent evangelicals. The poll-takers did say that if evangelical turnout hit the same high levels as 2012, the race would be tighter, with Trump ahead of Cruz by just 1 point.
In the end Cruz got 28 percent of the Iowa vote, Trump, 24 percent and Rubio, 23 percent.
WHAT ABOUT NEW HAMPSHIRE?
Pollsters must contend with some of the same complicating factors as in Iowa: voters debating among lots of candidates, being bombarded with information and considering options until the last minute.
But at least turnout will be higher than in a caucus state. Plus, pollsters have lots of lessons from Iowa to ponder.
In the end, though, "you're still trying to predict what somebody's going to do X days out when they don't even know what they're going to do," says Michael Dimock, president of the Pew Research Center.