In the weeks leading up to the 2024 presidential election, Guardian polling averages showed Kamala Harris winning an incredibly close race. Still, Donald Trump won, showing that yet again his volume of support was underestimated.
Though Guardian US averages had Harris winning, the final result was within the margin of errors for high-quality polls conducted this election cycle.
With an estimated 99% of the vote counted, Guardian US national polling averages were off by three percentage points as of 20 November 2024, an analysis of the 2024 presidential election returns has found.
Though this may change as more votes are counted, preliminary results show that polls underestimated the level of support for Trump for the third election in a row. The Guardian US national polling average had Harris at 48% and Trump at 47%, with the final results being Trump 50%, Harris 48%, an error similar to the polling errors in past elections.
Guardian US national averages only took the average of high-quality polls, but swing state polling averages used every poll gathered by 538. There are fewer swing state polls conducted, but are important because national polling is not indicative of who is going to win the election, only the level of support nationally.
Guardian US averages had Trump winning four out of seven key swing states and there was a bigger spread in the accuracy for swing states.
“You don’t have a lot of high-quality polls in swing states in particular,” Andy Crosby, assistant professor of teaching at the University of California Riverside School of Public Policy, said.
Swing state polling averages were not off by a significant amount, and nationally, swing state polls were accurate, Crosby said.
“High-quality polls were in the margin of error for a lot of these critical swing states,” Crosby said.
In the key swing states the Guardian US followed, the difference between the polls and the results ranged from a 0 percentage point difference in Georgia to a 4 percentage point difference in Arizona.
The Guardian US began tracking national and swing state polling in August and in the weeks leading up to the presidential election, the rolling 10-day averages showed an extremely tight race.
Throughout October 2024, Harris polled one to two points above Trump, with the gap falling to a single percentage point on 31 October, the last day that Guardian US took an average of state and national polls.
The election result shows that the 2024 presidential election polling error is similar to past elections. A 538 analysis of past polling errors found that between 1996 and 2020, the polling errors for their averages ranged from .1 pp in 2008 to 4.3pp in 1996. The Guardian’s national polling error was 3 percentage points, which means the polling in 2024 was not a deviation from years past.
Though the polling averages were close to the final results, the averages all biased the Democrats, with Trump doing better in every swing state than the polling averages expected.
This is a recurring pattern in elections where Donald Trump ran for office. Polls underestimated Trump’s performance in 2016, when he defeated Hillary Clinton, and in 2020, even though he lost to Joe Biden.
Though polling in 2016 has been attacked for missing Trump support, national polling was accurate according to the American Association for Public Opinion Research.
“After the 2016 election, there was a big effort to kind of do a debrief on what,” Crosby said of the AAPOR report. “There was a lot of inward thinking, like, what are we going to do to address this? Largely again the conclusion was, look, the national polls were accurate.”
It was, again, issues in swing state polling that missed the mark.
“We see this kind of pattern repeating itself in the sense of, oh gosh, it’s going to come down to the swing states and we don’t have the data for the swing states,” Crosby said.
The polls in 2020 had the highest polling error in 40 years for the popular vote, even though Biden won in line with what the polls expected, according to the 2020 AAPOR report. Why the polls were off remains a mystery though.
“Identifying conclusively why polls overstated the Democratic-Republican margin relative to the certified vote appears to be impossible with the available data,” the report said.
There are several reasons the polls might have had errors in the same direction across three elections, according to experts like Crosby and polling pontificators like Nate Silver.
Random chance, the demographics of who responds to polls, the elections being extremely close, or even people reluctant to voice support for Trump to strangers on the phone could all make up part of the overall explanation. But Trump won’t be running in 2028, so it remains to be seen if there is some unique aspect of Trump voters that continues to elude pollsters ability to measure his support.
