U.S. Pollsters Mark Worst Performance in 40 Years in 2020 Campaign
(Bloomberg) -- Donald Trump supporters who didn’t respond to pollsters were a likely cause of wildly wrong pre-election surveys in 2020 as political surveys suffered their worst performance in 40 years, a new industry report finds.
If that trend continues, those non-responses could pose an even greater crisis for political polling than the advent of mobile phones a generation ago.
A new report from a task force of the American Association for Public Opinion Research found that national polls overstated President Joe Biden’s lead over Trump by 3.9 percentage points. In state-level polls, it was 4.3 points.
In 2016, pollsters heavily favored Hillary Clinton to defeat Trump. But even though they correctly predicted Biden’s win in 2020, the error margin was much larger.
Biden’s predicted victory margin was the biggest miss by survey-takers since 1980. And while polling errors have favored Republicans as recently as 2012, two presidential elections with an increasing Democratic bias have pollsters concerned.
The report identified the two most likely reasons for the error as Trump supporters who refused to participate in polls, and an influx of new voters who were not accounted for in surveys. But the report said the data are “unsatisfyingly inconclusive” and arrived at those root causes by ruling out many of the factors blamed for similar misses by pre-election polls in 2016: Late deciding voters, too few non-college educated voters, and Trump supporters who would claim to be undecided in polls.
Unlike traditional “shy” voters, who would respond to polls but claim to be undecided despite supporting a candidate, these voters aren’t responding at all — making it harder for pollsters to adjust.
“Trump provided explicit cues to his supporters that polls were ‘fake’ and intended to suppress votes,” the report said. “These statements by Trump could have transformed survey participation into a political act whereby his strongest supporters chose not to respond to polls.”
It’s not clear whether the problem will persist in the 2022 midterm elections and beyond. Polls in the 2018 congressional elections were the most accurate in 20 years, suggesting that non-response is only a problem when Trump is on the ballot.
But Trump has hinted he’ll run again in 2024 and has continued to blast polls and pollsters.
“The polls were a joke,” he said in a May 15 statement. “Even the rigged final result was extremely close. It’s called SUPPRESSION POLLING and it should be illegal.”
But the report failed to find any major differences in polls based on who conducted them or how they contacted voters.
“It’s not clear that Republican pollsters did any better, so it’s an issue that is pretty pervasive,” said Joshua Clinton of Vanderbilt University, the chairman of the 19-member AAPOR task force. “All partisan pollsters have the same incentive to get this right, but Democratic pollsters and Republican pollsters were off by equal amounts.”
©2021 Bloomberg L.P.