WESTPORT, Conn. — Though few expected Donald Trump to win the presidential election, “the polls were not as bad as is commonly assumed,” the director of the Quinnipiac University Poll told the Y’s Men of Westport/Weston.
The “quality” polls — those done by the major television networks, Pew Research and Quinnipiac — projected a 4 percentage point popular vote victory by Hillary Clinton, Q poll director Doug Schwartz said in a talk Thursday. Clinton won by 2 percentage points.
“If you’re off by 2 or 3 points on the margin, that’s not bad,” he said.
The biggest surprises came with Trump's victories in Michigan, Wisconsin and Pennsylvania — where the polls tightened at the end, and where the pollsters missed, Schwartz said.
Fewer than 80,000 votes — less than 1 percent — decided these states, he said. Had those gone the other way, and Clinton hard earned the 46 Electoral College votes from those three states, she would be president-elect today, he said.
In Pennsylvania, polls showed Clinton in the lead — but not by a wide margin, he said.
Quinnipiac’s final poll, released one week before the election, showed Clinton up by 5 percentage points, Schwartz said. She lost by less than 1, essentially a tie. Unfortunately for Quinnipiac, it was on the incorrect side, Schwartz said.
And in Florida, Quinnipiac’s final poll showed Clinton winning by 1 point, but Trump won by 1 point. Again, the Q Poll was on the wrong side. Even though Quinnipiac termed these results “too close to call,“ the media ran with the numbers, but not the caveat, Schwartz said.
So what happened? Schwartz said late deciders in the swing states — about 10 percent of all voters — went overwhelmingly for Trump. What was to blame?
“Polling data doesn’t answer the question,” Schwartz said.
Asked during the Q&A whether the Quinnipiac Poll missed a “certain kind of voter,” Schwartz said that the white working-class voter, a key to Trump’s victory, was correctly sampled, but voted for Trump in larger numbers than the polls showed.
Why? One reason was what Schwartz called the “social desirability bias.” That accounts for people who did not want to admit they planned to vote for Trump even though they had made up their minds, he said.
Pollsters also missed Hispanics, who voted overwhelmingly for Clinton, but still went for Trump more than they had for Mitt Romney in 2012, and African-Americans, who voted in smaller numbers than in the two previous presidential elections, where they voted heavily for Barack Obama.
Schwartz noted that poll consumers often expect “pinpoint accuracy.” But polls have a built in sampling error, a statistical reality that makes them less predictive the closer the projected outcome, he said.