This is an archived article that was published on sltrib.com in 2016, and information in the article may be outdated. It is provided only for personal research purposes and may not be reprinted.

The 2016 election dealt a major blow to the science of polling, with prominent firms whiffing on not just the outcomes in Wisconsin, Michigan and Pennsylvania, but also missing the big picture in Donald Trump's win.

But in none of the 17 states deemed to be "in play" did pollsters miss by as much as they did in Utah, according to an analysis by the statistics site FiveThirtyEight.

On election night, pollsters predicted Trump would win the Beehive State by nearly 10 points. It turns out that Trump, as he did in many states, performed much better, clinching an 18-point margin of victory — an outside the margin-of-error miss that has led to some head-scratching.

FiveThirtyEight founder and editor Nate Silver posited that pollsters may not have reached enough white voters without a college education — "since that is a group that flocked to the Republican nominee" — and found pollsters were off the mark by larger margins in states where those voters are prevalent.

Utah may be a different story, however, and that hinges largely on the performance of independent anti-Trump candidate Evan McMullin.

Dan Jones, the pollster for The Salt Lake Tribune and Hinckley Institute of Politics, said that in the final days of the election, he detected voters who were torn between Trump and McMullin abandoning McMullin when it looked like the outcome in Utah could matter.

"Where I saw it turn is when Vice President-elect [Mike] Pence came and said, 'All you Republicans, come on back home.' … There were those who decided to do that," Jones said.

It is an explanation that is borne out by the tens of thousands of Republican voters who either voted on Election Day, swamping the polling places, or waited until the last minute to mail in their ballots. According to the Utah Colleges Exit Poll, 30 percent of Utahns who voted for Trump made up their mind in the last week of the campaign, and 12 percent of them decided in the last few days.

Contrast that with Hillary Clinton supporters, 58 percent of whom had made up their minds when she became the party's nominee at the Democratic National Convention and didn't waiver.

"What the story here is … is that Trump had a late surge and that's evident by the last day or two before the election," said David Magleby, a Brigham Young University political science professor who coordinated the exit poll. "The trend-line was very clear: People moved to Trump, if they moved to him, very late."

It was a trend that was perhaps more dramatic — but not isolated — to Utah, Magleby said. National exit polls showed that half those who made up their mind in the final week went to Trump, compared to 38 percent to Clinton. It is hard to know why that's the case, he said, although the Clinton campaign has blamed a letter by FBI Director James Comey that reignited a controversy over her use of a private email server when she was secretary of state.

Another indicator pointing to the fluidity of Utah voters in the presidential race: Several surveys that missed the margin in the presidential race were almost spot-on in Utah's race for governor.

Jones, for example, forecast Gov. Gary Herbert winning by 36 points and Utah polling firm Y2 Analytics projected a 38-point victory. Herbert won re-election by 38 points.

And in the Senate race, Jones saw Sen. Mike Lee beating Democrat Misty Snow by 43 points, while Y2 projected a 40-point win. Lee won by 41 points.

"Y2's polling was extremely accurate for Utah governor and Senate races," said Quin Monson, a partner at Y2 Analytics. "The major story here is the presence of McMullin. Viable third-party candidates can really mess with polling accuracy. … There was a lot of instability between McMullin and Trump."

While pollsters may have been surprised by the outcome, one prominent Trump supporter was not. A few days before the election, Don Peay, who was chairman of Trump's campaign in Utah, confidently predicted a Trump win, and by a healthy margin.

Peay said Friday that, while the Trump campaign didn't do its own polling in Utah, he and several members of the Mormon community hired their own pollster to take the temperature of the race. He said he was never in doubt.

Utah didn't suffer the worst polling breakdown — in Tennessee and West Virginia, for example, the polls were off the mark by more than 14 points, and in South Dakota they were nearly as bad. Missing the mark by 8 points made Utah the worst among the 17 states considered to be in play on Election Day.

"Those folks who were on the sideline, it was really at the end that they came on," said Spencer Kimball, polling adviser at Emerson College in Boston.

Of those polls conducted down the stretch, Emerson came closest to predicting the final result in Utah, showing Trump up by a dozen points a few days before the election. But Kimball said the election results overall have him reassessing how the craft is practiced.

Like most pollsters, Emerson conducted its surveys and then weighted the raw results based on the assumption that the 2016 electorate would look a lot like voter makeup in 2012. The raw results actually would have had Trump at 41 percent in Utah, Clinton at 27 percent, and McMullin at 23 — within a couple of points of the actual outcome, which saw Trump at 45, Clinton at 27 and McMullin at 21.

The results — along with big misses in the three Rust Belt states of Wisconsin, Michigan and Pennsylvania that helped Trump win the presidency — has Emerson reassessing how to weight future polls, or whether to do so at all.

"It was a lesson we should've learned in '48 with Truman and Dewey," Kimball said, referring to President Harry Truman's historic upset of Thomas Dewey. "But I guess every 60 or 70 years we need to be reminded, because you've got to poll up to the end."

Magleby said that points to one of the major problems in polling — figuring out who is actually going to show up and vote. In 2012, Mitt Romney's pollster had told Romney on Election Day that, based on his polling, Romney would be the next president.

But that projection assumed the voters that turned out in the 2010 midterms would come out in 2012, when, in fact, the electorate looked a lot more like it did in 2008, and President Barack Obama won re-election by a relatively comfortable margin (3.8 points).

gehrke@sltrib.com Twitter: @RobertGehrke —

Inside

Ex-Utah Gov. Jon Huntsman Jr. may be in play as Trump's pick for secretary of state. › A6

As Trump shrugs off criticism of his talk with Taiwan's leader, Huntsman says he sees signals of a new U.S.-China approach. › A8