By Ed Murray From NJ Advance Media for NJ.com

TRENTON — Statistics guru Nate Silver successfully called all 50 states and the District of Columbia for President Barack Obama in 2012.

This time, Donald Trump defeated Hillary Clinton even though Silver gave her a 71 percent chance of winning.

Pollsters and statisticians were left wondering Wednesday why none of their models predicted a near-sweep of swing states and Trump’s election as the nation’s 45th president.

“The polls clearly got it wrong this time,” said the American Association of Public Opinion Research, which is looking what happened in 2016.

Final U.S. election predictions – see http://www.nj.com/politics/index.ssf/2016/11/what_the_pundits_and_the_punters_expect_to_happen.html The analysis is done and the betting windows are closed. How right will they be?

Among the possible problems: Fewer people willing to respond to pollsters, bad assumptions on who would vote, and voters’ last-minute changes of heart.

“The polls were not capturing something, maybe undecideds or maybe hidden Trump votes, but they were generally wrong,” said Julian Zelizer, a professor of history and public policy at Princeton University.

Even, however, Tuesday, most reputable survey results were within the polls’ margins of error. What that means is a survey with a 4-point margin of error putting Clinton ahead by 2 points could mean she’s leading by 6 — or trailing by 2.

“It’s not as if we’re talking about a chasm of difference that separates the outcome from what was predicted,” said Krista Jenkins, a political science professor at Fairleigh Dickinson University and director of its PublicMind survey. “When you have such a slim margin for victory, it really is tied up. We might be in for some surprises on Election Day.”

Indeed there were. Andrew Gelman, professor of statistics and political science at Columbia University, said the difference between a Trump win and a Clinton victory was just 2 percent.

“A 2 percent shift in public opinion is not so large and usually would not be considered shocking,” he wrote. “In this case, the race was close enough that 2 percent was consequential.”

Quinnipiac University’s final pre-election polls had Clinton tied with Trump in Florida and 3 points ahead in North Carolina with margins of error of 3.3 percentage points in each state.

The polling institute said both states were too close to call. In end, Trump won Florida by 1.3 points and North Carolina by 3.8 points. The Quinnipiac poll of Ohio gave Trump a 5-point lead with a 4-percentage-point margin of error; he won by almost 9 points.

“That’s why there is a margin of error, because polling is an inexact science,” said Peter A. Brown, assistant director of the Quinnipiac polling institute. “We think we did pretty well.”

Silver on Wednesday called Trump’s victory “the most shocking political development of my lifetime,” even as he said polling indicated “a fairly competitive race with critical weaknesses for Clinton in the Electoral College.”

After all, he wrote a week earlier, “Trump remains an underdog, but no longer really a longshot.”

That didn’t stop him from giving Clinton a better than 7 in 10 chance of becoming president.

Nor did it stop Nate Cohn of the New York Times Upshot from giving her a better than 8 in 10 chance, or the same odds of a National Football League kicker making a 37-yard field goal.

After it was all over on Wednesday, Cohn wrote: “Field-goal kickers frequently miss 37-yard field goals.”

IMAGE: Republican presidential nominee Donald Trump arrives on stage with his family during election night at the New York Hilton Midtown in New York on November 9, 2016. Trump won the US presidency. / AFP PHOTO / Timothy A. CLARYTIMOTHY A. CLARY/AFP/Getty Images

For more on this story go to: http://www.nj.com/politics/index.ssf/2016/11/why_were_trump-clinton_election_results_a_surprise.html