What Went Wrong With the Polls?

Surveys may have missed big swaths of the white electorate

A stunned crowd, including the hotel staff, at the Nevada Democrats' election night watch party at the Aria Hotel & Casino in Las Vegas watch as Donald Trump delivers his victory speech after being elected the 45th President of the United States on Nov. 8. (Bill Clark/CQ Roll Call)
A stunned crowd, including the hotel staff, at the Nevada Democrats' election night watch party at the Aria Hotel & Casino in Las Vegas watch as Donald Trump delivers his victory speech after being elected the 45th President of the United States on Nov. 8. (Bill Clark/CQ Roll Call)
Posted November 9, 2016 at 3:46am

How, where and why did the polls go so wrong?

How could Donald Trump have won? The same Republican presidential nominee who on Tuesday morning had less than a 30-percent chance of winning.

How could the media, the pollsters, the pundits and the Democrats all be so, so wrong?

“This could put the voter projection industry out of business,” remarked CNN’s Jake Tapper on Tuesday.

Somehow, everyone was wrong. Everyone but Trump, who claimed all along he had the support of a silent majority of working-class Americans, racked with economic anxiety and feeling abandoned by the Washington elite.

White, rural voters turned out for Trump in droves, and pollsters vastly misjudged entire swaths of the white electorate. Trump won Florida, North Carolina, Ohio, Wisconsin and Pennsylvania; FiveThirtyEight.com says polling averages in each of those states were between two and seven points off.

So, what went wrong? For starters, most polls are still conducted by phone. And not only by phone, but by landline.

[Trump Pulls Off Stunning Upset]

The problem is that people just don’t answer their phones for pollsters anymore. Fewer people have landlines, there are new prohibitions on autodialing cellphones and people rarely pick up phone calls from numbers they don’t recognize; not to mention suspicion of solicitations from strangers.

Christopher Leavitt, who ran Republican Ed Gillespie’s 2014 Virginia Senate campaign, in which inaccurate polling led the GOP to push resources elsewhere, said campaigns must redouble their efforts to educate the voters and the media about the deficiencies of polling.

“It comes down to consistency and beating the message home from Day One: Ignore the polls and work toward the goal,” said Leavitt.

It’s also possible that many Trump supporters were never reached by pollsters. Or, as some have suggested, they misled pollsters about who they were supporting because they were embarrassed about the way the media has portrayed Trump’s base.

Some Republicans posited that their party’s voter-turnout operations, long maligned by the media, were far more efficient than anyone thought.

“Where the heck is the vaunted Democratic turnout machine?” wrote Robert Blizzard, a GOP pollster at Public Opinion Strategies, in a tweet. “The RNC crushed this.”

Something else to consider might be the residual effects of a Supreme Court decision that gutted the Voting Rights Act of 1965, leading to the closing of hundreds of polling places nationwide. In a blog post, FiveThirtyEight’s Carl Bialik pointed out that Georgia, North Carolina and Arizona were affected by that decision — Trump won the first two, and was leading in Arizona, as of press time. 

[GOP Aides Predict Trump Loss, Control of Both House and Senate]

It may take time to figure out exactly why Trump’s victory was such a surprise. The curious can at least consider the one poll — the USC/L.A. Times Daybreak tracking poll —   that consistently pointed to a Trump victory. That poll used a more complicated weighting process, meaning the way the poll adjusts to represent diversity, than other systems.

It was also conducted online.

“That doesn’t necessarily mean that a poll conducted online … necessarily will be more accurate than polls conducted by phone,” the Times wrote Tuesday evening. “But it is yet another indication that polling needs more, diverse ways to look at public opinion, not fewer.”

Shawn Zeller contributed to this report.