Pollsters had good 2018 but have they made the right adjustments for 2020?

Many political pollsters re-calibrated following President Trump’s widely unexpected (including by him) 2016 electoral victory in the presidential election and the results seem to have paid off in 2018.

In the RealClearPolitics average of public polls, Democrats were projected to win the majority of votes cast for House candidates by 7.3 percent. The official national totals are still rolling in but according to the Cook Political Report’s tally, Democrats lead by 8.5 percent, well within the margin of error of most polls.

Even at the district level which is notoriously harder to survey, pollsters had a better year than normal, largely thanks to the 43 polls taken in hotly contest races conducted by Siena College for the New York Times which finished with an average error margin of just 3 percentage points, half the typical district error rate of 6 percent.

But, Republican pollster Ed Goeas said on Tuesday’s Hill.TV broadcast of “What America’s Thinking” that pollsters still need to change their way of predicting electoral outcomes.

He says that instead of trying to use past polling and educated guesses to assume what the final demographics of the electorate are going to be, pollsters ought to focus on survey respondents’ characterization of how likely they are to vote.

“I’m hoping what we’ve set aside, finally, is that coming out of 2008, all the public polls started polling the sample based on what they thought the electorate was going to be. As political pollsters, we don’t do that,” Goeas said.

He added that “We go fairly wide, we do take out people that are definitely not going to vote. I take age, education, intensity to vote, intensity for the candidates and then set up a matrix” where his clients can then decide for themselves based on likelihood to vote.

“It’s so hard to predict who’s a likely voter,” said Mallory Newall, research director at Ipsos Public Affairs polling.

— Matthew Sheffield