2020 election polls suffered historic errors and underestimate Trump’s support in nearly every state
Polls predicting the outcome of the 2020 presidential race — as well as numerous state races — suffered the worst mistakes in four decades, according to a new survey.
But experts say the reason why is a mystery.
Public opinion polls significantly overestimate President Joe Biden’s margin of victory over Donald Trump, according to a study of 529 national presidential election polls and 1,572 state-level presidential polls conducted by the American Association for Public Opinion Research.
Days after Americans went to the polls, the final results showed Biden won more than 81 million votes, while Trump won just over 74 million.
Projections ahead of November’s election overestimate Biden’s margin to Trump by 3.9 percent in the popular vote and 4.3 percent in state polls.
A YouGov poll put Biden a whopping 11 points ahead of Trump a week before Election Day.
Biden defeated Trump in the November election with 4.4 percent of the vote.
Pollsters suggested Trump refers to polls as “fake news” and that politicking the results may have contributed to the errors and reduced reaction among Republican voters, but they wouldn’t say it was the main reason.
A survey of nearly 3,000 state and national polls found that most forecasts significantly underestimate Trump’s performance in the 2020 election
Your browser does not support iframes.
Your browser does not support iframes.
The polls conducted in the two weeks before Election Day were 5 percentage points behind the state level.
Support for Trump exceeded expectations in nearly every state by more than three percent on average.
By contrast, support for Biden was expected to be one point higher than he ultimately won.
The same polls were heavily criticized in the weeks following the election after Trump surpassed expectations.
A Morning Consult poll the day before the election showed Biden leading Trump in five battlefield states
AAPOR came to its conclusion after research: nearly 3,000 different studies at the state and national levels.
But the task force of 19 election and political science experts doesn’t know why so many polls underestimate Trump.
“With the available data, it is impossible to conclusively determine why polls overestimate the Democratic-Republican margin over the certified vote,” the report said.
Support for Trump has fallen since the start of the coronavirus pandemic in the national Morning Consult poll, but the AAPOR task force has been inconclusive as to why the final results were so different from the polls
What is ‘social desirability bias’ and why does it affect polls?
Social desirability bias is the name some pollsters have given to the effect that keeps Trump supporters quiet when asked who they vote for.
It simply means that people are more likely to give the answer they think the person asking them wants to hear, rather than what they actually think.
While this could be true for any question, in Trump’s case it is compounded by the abuse Trump supporters are often subjected to, polls say.
Robert Cahaly, a pollster for the Trafalgar Group, said the abuse has worsened between 2016 and this year.
“In 2016, the worst thing that was said about Trump voters was that they were ‘deplorable,'” he said.
Now ‘people’ [are] getting beat up for wearing the wrong hat, people being harassed for having a sticker on their car. People just don’t want to say anything.’
Fellow surveyor Arie Kapteyn said this effect can be measured by asking people who they vote for and then who they think other people vote for.
Asking people who they will vote for in 2020 gave Biden a solid 10-point lead, he said. But if you ask them who other people will vote for, you cut that in half.
This is because it leaves people free to speculate about how popular they think a candidate is without involving themselves.
The lack of answers makes it difficult to provide solutions to improve accuracy in the next election.
“It is unclear whether the problems facing polls in 2020 will persist in 2022 or 2024, and what happens in 2022 may not be informative to know if there are longer-term problems,” the report states.
Congressional and governor polls showed an even greater discrepancy with overestimating Democrats’ performance.
National-level polls even predicted Democrats would win in the House, but they lost 13 seats.
Josh Clinton, chair of the AAPOR task force and professor at Vanderbilt University, ruled out the possibility that some Trump voters are lying to pollsters as a reason for the discrepancy.
The report also failed to find a significant under-representation of groups, thus excluding this as a cause for errors.
That polls overestimate Biden’s support more in whiter, more rural, and less populous states is suggestive (but inconclusive) that the poll error was the result of too few Trump supporters responding to polls. “Greater poll error was found in states with more Trump supporters.”
Adding to the confusion of experts, Clinton said the polling problems discovered in 2016 — which correctly predicted Hillary Clinton would win the popular vote — appeared to improve in 2018.
Polls in 2016 did not take into account voters’ education levels, according to a similar survey on that election cycle. That issue was seemingly resolved in the ensuing poll, with the task force ruling out that as a possible cause for errors.
Clinton said it is uncertain whether the historic error levels are related to Trump or problems with polls in general.
“If the polls do well in 2022, then we don’t know if the problem is solved,” Clinton told the… Washington Post. “Or that it’s just a phenomenon unique to presidential elections with certain candidates calling on ‘don’t trust the news, don’t trust the polls,’ making polling a political act.”