Polls that predict the outcome of the 2020 presidential election (and many state elections as well) have suffered the worst errors in 40 years, according to a new survey.
But experts say the reason is a mystery.
According to a survey of 529 national presidential election polls and 1,572 state-level presidential polls conducted by the American Polling Association, polls significantly overestimate the margin of victory for President Joe Biden over Donald Trump. Was there.
A few days after the Americans went to polls, the final results showed that Biden received more than 81 million votes and Trump received more than 74 million votes.
Biden’s margin for Trump was exaggerated by 3.9% in general polls and 4.3% in state polls, leading up to the November elections.
In a YouGov poll, Biden was a whopping 11 points ahead of Trump just a week before the election day.
Biden beat Trump by 4.4% of the popularity poll in the November elections.
Pollsters have suggested that Trump called the poll “fake news” and politicized the results, which may have contributed to the reduction of Republican voters’ mistakes and responses. I didn’t say it was the reason.
Approximately 3,000 state and national polls found that most forecasts significantly underestimated Trump’s performance in the 2020 elections.
Your browser does not support iframes.
A poll conducted two weeks before the election date was off by 5 percentage points at the state level.
Trump’s support exceeded expectations by an average of over 3 percent in almost every state.
In contrast, Biden’s support was predicted to be one point higher than he would have won.
Joe Biden was predicted to lead Donald Trump by 10 points nationwide from 52% to 42%, according to recent poll averages, but in reality it leads 51% to 48%. The count is still going on. Voters predicted an average of 10 points lead for Biden (pictured) towards election day. This lead evaporated significantly as the results came out.
The same poll was widely criticized in the weeks following the election after Trump exceeded expectations.
As a result of the investigation, AAPOR came to a conclusion. Approximately 3,000 different surveys at the state and national levels..
But the task force of 19 election and political science experts is at a loss as to why so many polls underestimated Trump.
“It seems impossible with the available data to definitively identify why the vote is exaggerating the Democratic-Republican margin for a certified vote,” the report said. ..
What is “social desirability bias”? Why does it affect polls?
Social-desirability bias is the name given by some pollsters to the effect of keeping quiet when Trump supporters are asked who to vote for.
Simply put, it’s more likely that the person asking the question will give the answer they want to hear, rather than what they actually think.
This may apply to any question, but pollsters say it gets worse in the case of Trump because of the abuse often suffered by Trump supporters.
Robert Cahaly, a Trafalgar Group pollster, said the abuse worsened from 2016 to this year.
“The worst thing that was said about Trump voters in 2016 was that they were’sorrowful’,” he said.
Now’people [are] He was beaten for wearing the wrong hat and harassed for sticking a sticker on his car. People just don’t want to say anything. “
Pollster Ally Kaptein said the effect can be measured by asking who is voting and then who he thinks others are voting for.
When asked who to vote for in 2020, Biden returned a solid lead of 10 points, he said. But if you ask them who else will vote, you can cut that in half.
This is because you are free to guess how popular a candidate is without involving yourself.
It is difficult to provide a solution to improve accuracy in the next election because no answer can be found.
“It is unclear whether the problems faced in 2020 will continue in 2022 or 2024, and what happened in 2022 may not be useful in knowing if there are longer-term problems. “The report said.
Parliamentary and governor polls suffered even greater discrepancies by overestimating Democratic performance.
National polls even predicted that the Democratic Party would be profitable in the House of Representatives, but lost 13 seats.
The chair of the AAPOR Task Force and Professor Josh Clinton of Vanderbilt University have denied that Trump voters may be lying to voters as a reason for the disagreement.
The report also failed to find a significant underestimation of the group and ruled it out as the cause of the error.
“Public polls exaggerate Byden’s support in whiter, more rural, and less densely populated states is due to polling errors due to too few Trump supporters responding to polls. (But not definitive), “the report said. “A larger voting error was found in a state with more Trump supporters.”
In addition to the confusion of experts, Mr. Clinton said the voting problem detected in 2016 (which correctly predicted Hillary Clinton to win the popularity vote) seems to have improved in 2018.
A similar survey of the election cycle found that a 2016 poll could not explain the level of education of voters. This issue seems to have been resolved in a subsequent poll, and the Task Force has ruled it out as a possible cause of the error.
Clinton said it’s unclear whether historical-level errors are related to Trump or general voting issues.
“If the poll goes well in 2022, I don’t know if the problem has been resolved,” Clinton told The Washington Post. “Or, whether it’s a phenomenon peculiar to the presidential election by a particular candidate who complains,’I don’t trust the news, I don’t trust the poll,’ as a result, the poll becomes a political act. “
The 2020 election poll suffered from historical mistakes and underestimated Trump’s support in almost every state
Source link The 2020 election poll suffered from historical mistakes and underestimated Trump’s support in almost every state