Skip to main content

Bucknell’s Ellis Assesses Presidential Polling Problems

Amid the drama of another presidential election cliffhanger came this storyline — how could the political polls be even more inaccurate this year than they were in 2016? After all, most polls had former vice president Joe Biden leading by comfortable margins nationally and in battleground states going into Election Day. And yet, President Trump’s performance defied the polls again — calling into question their accuracy this year and in future elections.

Bucknell University Professor Chris Ellis, political science — co-director of the Bucknell Institute for Public Policy — saw three main areas of polling error.

1. The polls seemed to underestimate Trump’s strength among nonwhite voters. “It still wasn’t enough to win them, but it looks like Republicans won a higher share of the nonwhite vote than they have in 60 years,” says Ellis.

2. Election turnout was really difficult to model this year. “With the rapid shift to vote-by-mail, it was just hard to figure out exactly who was going to vote,” Ellis says. “This is difficult to do in a normal year, but in an unprecedented time, pollsters had to make a lot of guesses that they don’t usually have to make, and many of these turned out wrong.”

3. The pollsters tended to underestimate how many Republicans “came home” to vote for Trump, even if they had to “hold their nose to do it.” “In a time of polarization that we haven’t seen in more than 100 years, party identification is king, and we can expect elections to always be close, no matter who the candidates are, in what is essentially a 50-50 nation,” says Ellis, who wrote on the phenomenon with Texas A&M political scientist Joseph Daniel Ura in a paper titled Polarization and the Decline of Economic Voting in American National Elections, which was published in a recent issue of Social Science Quarterly.


CONTACTS: Ellis at 412-849-6500 (c), 570-577-1960 or; Mike Ferlazzo, 570-238-6266 (c),

Comments are closed.