What went wrong with the presidential polls

in Features by
FiveThirtyEight’s final forecast, the left diagram, classified the Rust Belt states as Democratic strongholds, something that did not hold up in the actual election where Biden and Trump were very close. The Congressional race predictions were even more off-base, with Democrats expected to win back the Senate and receive more seats in the House, both of which did not happen. (Diagrams/FiveThirtyEight and BBC)

Nate Silver, founder of famed election prediction website FiveThirtyEight, made a bold prediction Election Day, claiming presidential candidate Joe Biden had an 89% chance of winning and a 29% chance of a landslide victory. Swing states Florida and North Carolina were classified as likely Democrat in his forecast, while Wisconsin and Michigan were very far left, favoring Biden by 94%. While Silver was correct to declare Biden as the winner, there was a sharp contrast between what polling data suggested and the actual results.

Followers of the 2016 election likely remember polls all but guaranteeing Hillary Clinton’s win by a large margin, only for Donald Trump to take the White House. Going into the 2020 election, pollsters across the country promised they fixed their polling process, but this year’s polls were even more wrong than before. For example, the final polls for Wisconsin gave Biden a nearly nine point lead, whereas the actual results were less than one point apart. Even Republican-lean states, like Ohio, which was predicted to be a very close race with Trump taking the win by less than a percentage point, in actuality went to Trump by almost 10 points. In fact, every single state except for Colorado ended up more Republican than the polls predicted.

There are a few theories that can possibly explain such an unexpectedly high Republican lean. According to The Minnesota Post, the “shy Trump voter” is a common theory. This theory states that people are embarrassed to tell pollsters they voted for Trump due to his controversial reputation. However, many pollsters disagree with this theory since Trump supporters are stereotypically very open about who they support (think MAGA hats and Trump rallies). 

Instead, they have proposed another theory: the “submerged Trump voter.” For the four years he has been in office, Trump has claimed that news sources such as CNN, NBC, The New York Times and The Washington Post are fake and corrupt. Since most polls are run by these agencies, Trump supporters who believe in his claims likely will not want to participate in the allegedly fake polls.

The record-breaking turnout on Election Day also contributed to the Republican lean found in polls. Pollsters use a small sample size and then adjust for what they expect the turnout to be. However, since the turnout was unexpectedly high, the adjustment did not work as well as the pollsters hoped. 

“Though we are talking about just a one-point shift in the result, that still puts us in the right direction and it’s consistent with the actual turnout that we had,” pollster Charles Franklin said to The Minnesota Post. His poll was more accurate than others, and this is because he over adjusted his numbers to account for the large turnout.

If pollsters do not want a repeat of this year, they need to find better ways to survey Republican voters. Whether it be an anonymous poll, over adjusting numbers or increasing the scope of the poll to account for more people, something likely needs to be done to make the predictions as accurate as possible. The American Association for Public Opinion Research is beginning to analyze the polls in order to get a definitive answer of what went wrong, but for now, it is up to the pollsters and forecasters to take a look at their current system and make adjustments accordingly. With a brand new electoral map due to Census data, it will be an uphill battle for pollsters to make the most accurate predictions, but it is one that they can achieve if they revamp their polling procedure.

Senior Ellaheh Gohari is entering her fourth (and sadly final) year on staff and third year as co-EIC of the Patriot Post. She loves learning new things and can often be found going down Wikipedia rabbit holes in search of random knowledge. Outside of room 25310, she serves as co-president to both the Girls Excelling in Math and Science club and the Science National Honor Society. A science-lover, she enjoys exploring the subject through research projects with UMiami, volunteer tutoring with OTTER and fact-checks with MediaWise. She hopes you enjoy your time reading the Patriot Post.