Were Polls Wrong In the 2020 Election?

We have analyzed lots of polls prior to the 2020 election, and they have formed a large part of our pre-election predictions. Obviously, there is no “one-size-fits-all” type analysis for the polls, since polls were wildly incorrect in some states but almost spot-on in others. In this post, we are going to take a look at some of the presidential polls and how they were very wrong (or right) about the results of this election.

Make sure to first check out our post discussing the final results of the 2020 Presidential Election.

Let’s start with the swing states in which the polls were correct in. This means that the actual result was pretty similar to what the polls were predicting, and there were no wild swings between the expected result and the actual result.

The first state on this list is Georgia. President-Elect Biden won by a 0.3-point margin in this traditionally Republican-leaning state. Biden won 49.5 percent of the vote share and President Trump won 49.2 percent of the vote. This was very similar to the FiveThirtyEight final polling average in the state, where Biden was up by 1.2 percent and had 48.5 percent of the vote compared to Trump’s 47.4 percent. Most of the final reliable polls showed the race within a two-point margin.

Monmouth University (an A+ rated polling firm), Public Policy Polling (B rated polling firm), YouGov (B), and Data for Progress (B-) polls showed Biden up by two points. Siena College/The New York Times (A+), Opinion Insight (A/B), Morning Consult (B/C), and the University of Georgia (B/C) polls showed the race even. Emerson College (A-), Landmark Communications (B), InsiderAdvantage (B-), and AtlasIntel (B/C) polls showed Trump leading by small margins.

Therefore, we can conclude that Georgia was, in fact, a relatively correct call.

Nebraska’s Second District was also called quite well. Biden actually out-performed his margin here, winning the district by about seven points. The polling margin was 3.8 percent, but the Siena College/The New York Times poll was almost bang on.

Another state in which the polling was fairly accurate was in Minnesota. Though the polling average was 9.2 percent for Biden compared to the actual result of 6.9 percent for Biden, this small shift is within the margin of error. Highly-rated A polls from Suffolk University and SurveyUSA, for instance, actually underestimated Biden, putting him ahead by only five to six points. Likewise, Colorado, Virginia, and New Mexico were also polled pretty good. Biden led by 12.5, 11.8, and 11.7 percent there respectively in the polling average. Biden won these states by 13.5, 10.2, and 10.8 percent margins respectively. (In an unrelated note, this election may fully solidify these three states’ positions as blue states.)

Unfortunately, polling in the rest of the Sun Belt and Rust Belt were quite overwhelmingly off.

In Florida, the polls were shockingly off by six points. What was a 2.5 percent lead for Biden became a 3.3 percent win for Trump. Again, this is likely due to Hispanic and Latino voters flipping for Trump in the traditionally extremely liberal Miami-Dade County (and others). Aside from ABC News/The Washington Post (A+) polling showing Trump ahead by two points, just about every other reputable pollster, like Monmouth University (A+), Marist College (A+), Siena College/The New York Times (A+), and the University of North Florida (A/B), showed Biden leading. Suffolk University (A) showed an even race. The fact that so many reputable pollsters got Florida wrong is a cause for concern, and further analysis is required.

North Carolina was also called incorrectly. Despite polls showing Biden up by 1.8 percent, Trump ended up winning by 1.3 percent. This is still within the margin of error of polls, though, so it cannot be definitively be said to be “wrong.” That said, though, a number of highly-rated pollsters, including Marist College (A+), Siena College/The New York Times (A+), ABC News/The Washington Post (A+), and Monmouth University (A+) all expected Biden to win. SurveyUSA (A), Emerson College (A-), and the University of Massachusetts, Lowell (A/B) showed the race even. N.C. is quite weird in the fact that a large percentage of the undecided vote splits Republican, making the state always lean Republican.

Though the winners were correctly called in Arizona, Iowa, Michigan, Ohio, Pennsylvania, Texas, and Wisconsin, the margins were way off in these states, some by way more than others.

The less-disastrous states include Arizona, Pennsylvania, and Texas. Biden won the first two by margins of 0.31 and 1.2 percent respectively, while Trump won Texas by 5.7 percent. In these states, polling showed Biden up by 2.6 percent in Arizona, 4.7 percent in Pennsylvania, and Trump up by 1.1 percent in Texas.

In Arizona, though a close race was expected, many thought that Biden owuld win by more than one percent, simply because of the fact that most polls showed Biden having a two to three-point lead. Though this result is definitely still within the margin of error, Democrats were slightly overestimated.

Pennsylvania was expected to be the closest of the three Rust Belt states, but it actually voted to the left of Wisconsin (we’ll get to that state in a minute). Polls were only off by 3.5 percent, still in the margin of error of most polls. Most A+ polls did predict a larger vote share for Biden here, but ultimately, we see that the average was not too far off the actual result.

Texas was expected to be somewhat close, but no one really expected it to flip this year. The polling average was a bit off, but notably, all highly-rated pollsters called the race correctly for Trump, by varying margins. Siena College/The New York Times, for example, put the race at a four-point lead for Trump. Emerson College only gave Trump a one-point lead.

Michigan’s result was not quite as disastrous as Wisconsin, Iowa, and Ohio, but still wrong enough to be notable. Though Biden had a 7.9-point lead in polls, he only ended up carrying the state by 2.7 percent (though this may increase as votes haven’t bee fully finalized yet). That a larger error than in 2016, despite the polls calling the winner correctly this year. Most highly-rated pollsters expected the margin to be at least five percent.

The polling was quite a bit off from the actual result in Iowa and Ohio, though the winner was still called correctly. Iowa and Ohio were expected to be close races: Trump led by just 1.1 and 0.8 percent respectively. However, he ended up carrying these states by 8.3 and 7.9 percent respectively. That is quite a significant swing, and virtually no pollster predicted this result. The only exception was Iowa-based Selzer & Co. (A+), which showed Trump up by seven percent in Iowa. Selzer is considered the gold standard in the polling industry, so it’s no surprise that their outlier poll was almost bang on. This error may be limited to Trump only, though: Trump was able to convince a larger-than-expected number of unionized workers to vote for him in these two blue-collar dominant states, allowing him to carry a fairly safe victory. In any case, though, this result further cements the states as moving steadily to the right.

Wisconsin was, simply put, a major catastrophe for polling. It was one of the most-polled states, yet, the margin was wildly incorrect (despite still being called correctly). The polls showed Biden leading by 8.4 percent, even more than in Michigan. Yet, he was only able to carry the state by 0.62 percent, making this state see an over seven-point shift right compared to the polls, not unlike what was seen in 2016. There is something strange in Wisconin that caused the polls to swing so wildly, and further analysis will be required. Every reputable pollster had Biden leading at least by four to five percent, and some even double-digits.

Make sure to check out the 2020 election coverage page for more.