Why polls didn’t show Trump was going to win in 2016 election

FALLS VILLAGE — Douglas Schwartz, director of the Quinnipiac University Poll, described polls as “pieces of data woven together to tell stories about who we are as a country at any given time.”

Schwartz spoke at Housatonic Valley Regional High School on Friday, Oct. 18, as part of the Salisbury Forum series of talks.

Schwartz said the current political atmosphere is “unprecedented and complex.”

He said that lawmakers monitor public opinion closely, and noted that before her announcement of an impeachment inquiry on Sept. 24, House Speaker Nancy Pelosi (D-Calif.) spoke in terms of public sentiment and bipartisanship.

On Sept. 23, 37% polled supported impeaching President Donald Trump, and  57% did not.

One week later, there was a 20-point swing, so 47% in favor of impeachment and 47% opposed.

Asked a second question — do you support formal impeachment — 51% said yes.

These results tell of “the intense political polarization gripping the nation.”

The situation is much different from the impeachment (or threat of impeachment) of Presidents Bill Clinton and Richard Nixon, Schwartz said.

Quinnipiac was not polling nationally in 1998, but in the Gallup Poll, support for impeaching Clinton never topped 35%.

When Nixon fired special prosecutor Archibald Cox in 1973, Gallup showed support for impeachment at 33%. Schwartz said that number didn’t rise until the very end of the Nixon presidency, in 1974, when Nixon resigned rather than face impeachment and a trial in the Senate.

“So it took time for Nixon and it never happened for Clinton.”

How they do their polls

Schwartz provided a look at how Quinnipiac polls are conducted. He asked for a show of hands from people who had been called by pollsters. A lot of people raised their hands. “That’s more than I expected,” said Schwartz.

Unlike online polls, people cannot volunteer to take a Quinnipiac poll. Quinnipiac does 19 state polls and national polls, which is unusual in the polling business.

For a national poll, Quinnipiac interviews about 1000 people.

(Sometimes odd things happen. Schwartz told the story of conducting a poll about Pennsylvania Gov.  Ed Rendell. One of the people reached for an interview was … Ed Rendell.)

To be accurate, a poll must have a random sample. Quinnipiac uses “random digit dialing,” where lists of phone numbers are generated by computers using area codes and exchanges. The list includes land lines and cell phones. The latter is crucial to the process, Schwartz said, because more than half of Americans use only a cell phone.

If the number is a land line, the interviewer  asks to speak to the person in the household with the next birthday.

A poll usually runs for five days, with between 100 and 200 interviewers working the phones and recording answers into a computer.

The raw data  is weighted for demographics to match census data.

Schwartz said for people trying to make sense of a poll, he advised looking at the number of respondents and the margin of error.

A 40% approval rating with a 3% margin of error means approval is between 37 and 43%, he noted.

Quinnipiac also asks respondents for their political affiliation, gender, income and race.

During the question-and-answer portion of the program, when he asked about the polls that showed Hillary Clinton winning in 2016, Schwartz said one thing pollsters did not weigh for was education.

White voters without a college education turned out to be a significant — and overlooked — group of voters.

Unexpected outcomes

Schwartz said polls give the public a voice and can make a difference in policy. As an example, he said that in March 2017, a Republican healthcare bill was polling poorly, at 17% approval, and was ultimately withdrawn.

Sometimes polls and policy are at odds. When Virginia Gov. Ralph Northam found himself in hot water over some old yearbook photos and comments that could be construed as racist, several prominent Virginia politicians called for his resignation.

But polling showed that almost 50% of Virginia voters, and more than half of black voters, thought Northam should finish his term.

And sometimes polls and conventional wisdom clash. Schwartz said polling in Texas, a state popularly supposed to be a hotbed of pro-Second Amendment opinion, shows 53% favor stricter gun laws and 90% favor background checks.

Looking ahead to the 2020 presidential election, Schwartz said the sheer number of Democratic candidates add more pressure to “winnow the field.”

(The Democratic National Committee includes Quinnipiac in the group of polls used to determine eligibility for debates.)

Clinton v. Trump

A key thing to remember: “Polls are not forecasts.”

Asked why polls failed to predict Trump’s win in 2016, Schwartz said the polls were not as off as popularly thought.

He said Clinton won the popular vote by 2%, and most polls had her winning by 3 or 4%.

But Clinton lost critical swing states. “There was clearly a change in the final days.”

Schwartz said that exit polls showed that a significant number of voters made up their minds in the last few days of the 2016 campaign, and most of them voted for Trump.

One problem was how polls were reported. Quinnipiac’s final Florida poll had Clinton up by one point — too close to call.

“That’s not how it was reported.”

Some of the responsibility lies with the press, he said. “They didn’t accurately describe how close the race really was.”

How do the Democratic candidates fare in head-to-head match-ups with the president?

Schwartz said that Joe Biden does best at 11 over Trump, followed by Elizabeth Warren, Bernie Sanders, and Pete Buttigieg, who lead Trump by single digits. 

At the moment, that is. “Remember, we’re not predicting.”