What pollsters could have missed when trying to predict the next president

Posted at 7:58 PM, Nov 10, 2016
and last updated 2016-11-11 10:33:56-05

A number of national polls conducted the week before the election showed Hillary Clinton winning the presidency. Of the eleven polls posted on RealClearPolitics, only one put President-elect Donald Trump on top.

Over the next few weeks and months there's going to be a lot of analysis and digging into why polls showed a different outcome than reality, but there are already some early theories. It could be that people who said they were going to vote didn't, or there were a lot of “hidden Trump voters,” people who lied about voting for Hillary Clinton, or others who didn't take the survey because they didn't want to admit to voting for Trump.

Even though most polls got it wrong, it wasn't by much.

“In a lot of cases Hillary Clinton and Donald Trump were within the margin of error of each other but things just broke Trump's way,” said Dr. Mileah Kromer, the director of polling at the Sarah T. Hughes Field Politics Center at Goucher College.

So while they weren't dramatically off from hitting the mark, there was something keeping the numbers from adding up.

“There's a few things that could go potentially wrong. Pollsters rely on two fundamental things - one, individuals telling us the truth of their likelihood of turning out and two, them telling us the truth about their voter preference. And so right now, it really remains to be seen where the error lies,” said Dr. Kromer.

Jason Vasquez is a freshman student at Goucher College. He thinks people lying about their presidential preference affected the polls.

“A bunch of people I know would say they're voting for Hillary, but if you asked them behind closed doors they'd lean in and be like 'I really think I'm going to vote for Trump,' and I think there were a lot more people like that than originally planned,” said Vasquez.

Another reason could be people changing their minds about voting based on what they saw in the polls.

“They see, oh, Hillary's going to win, so we don't really need to do it,” said Annie Carter, a senior at Goucher College.

It'll take time to pinpoint what happened but Dr. Kromer said it could also have to do with the polling process. Most are done using a dual frame telephone sample where pollsters call household land lines as well as cell phones, but they're finding that people are less willing to participate when they’re called.

“What pollsters refer to as nonresponse bias. So, these are individuals that when pollsters called their households or cell phones they just opted not to answer the survey thus introducing a systematic bias of under calling or under sampling Trump voters. Not because the sampling or the math was wrong, but because individuals for some reason, perhaps a distrust of media, were just not interested in opting into the surveys,” said Dr. Kromer.

To get around that, she said, there could be a shift in how polling is done.

“We'll see the introduction of more non-probability based but representative online samples and that methodology is still very new and pollsters are really trying to work it out,” she said.

While the national polls were somewhat off, the Goucher College poll in Maryland was fairly close. According to the unofficial results from the Maryland State Board of Elections, Hillary Clinton won 60 percent of the vote while Donald Trump won 35 percent.

The Goucher poll from September showed Clinton winning 58 percent of the vote and Trump with 25 percent.

Dr. Kromer said they didn't do anything different when conducting their survey, there just wasn't a lot of campaigning in Maryland.

Download the ABC2 News app for the iPhone, Kindle and Android.