Research · When the polls missed

Third time in a row: the smaller Trump miss of 2024

Abstract. The final averages had Kamala Harris 0.7 points ahead and Donald Trump won the popular vote by 1.5, a 2.2-point miss that was smaller than in 2020 but the third presidential election running in which polls understated Trump. Weighting on party and recalled 2020 vote tightened the polls, but they still missed Republicans in Republican areas, the swing of Hispanic voters to Trump and new voters.

Polling Forecast · 2 November 2026 · 7 min read · United States, 2024 presidential election, polling error, final polls, inquiry, pollster methods

Summary

On election eve the main national polling averages had Kamala Harris ahead of Donald Trump by 0.7 points. Trump won the popular vote by 1.5 points (49.8% to 48.3%), so the averages missed the lead by 2.2 points, the smallest miss of his three elections but the third in a row in the same direction. State polls had their best presidential year since 1944 by AAPOR's count, yet they still showed Harris narrowly ahead in Michigan, Wisconsin and Pennsylvania, and Trump won all seven swing states. The industry's post-2020 shift to weighting on party and past vote narrowed the error, but polls still undercounted Republicans in Republican areas, Hispanic Trump voters and people who had not voted in 2020.

What the polls said

National polls in the last two weeks clustered tightly around a tie, with most showing Harris ahead by 1 to 3 points. The final polls from the larger pollsters in our database:

Pollster	Fieldwork end	Harris	Trump	Lead
TIPP	4 Nov	48.3	48.6	Trump +0.3
Ipsos/Reuters	3 Nov	50	48	Harris +2
Research Co.	3 Nov	48	46	Harris +2
Marist	2 Nov	51	47	Harris +4
NBC News	2 Nov	49	49	Tie
Emerson	2 Nov	49	49	Tie
ABC News/Ipsos	1 Nov	49	46	Harris +3
Morning Consult	31 Oct	49	47	Harris +2
YouGov/Economist	29 Oct	49	47	Harris +2
Rasmussen	28 Oct	46	48	Trump +2
New York Times/Siena	23 Oct	48	48	Tie

The published averages on 5 November were Harris +1.2 at FiveThirtyEight (48.0 to 46.8) and 270toWin (48.4 to 47.2), +1.0 at Silver Bulletin, +0.9 at the Cook Political Report, +0.1 at RealClearPolitics (48.7 to 48.6) and a tie at Decision Desk HQ. Their mean was Harris 48.5, Trump 47.7, a lead of 0.7 points.

In the swing states, our final-poll averages had Harris ahead by 1.7 points in Michigan, 1.0 in Wisconsin and 0.3 in Pennsylvania, and Trump ahead by 1.0 to 1.7 points in Georgia, North Carolina, Nevada and Arizona. All seven were within the polls' margin of error.

What happened

The Federal Election Commission's official totals gave Trump 77,302,580 votes (49.80%) and Harris 75,017,613 (48.32%), out of 155.2 million cast. Trump won 312 electoral votes to Harris's 226.

As in 2020, the Democrat's share was close: 48.5 in the averages, 48.3 in the count. Trump's was understated by 2.1 points, 47.7 against 49.8. The lead was off by 2.2 points.

Trump won every swing state: Pennsylvania by 1.7 points, Michigan by 1.4, Wisconsin by 0.9, Georgia by 2.2, North Carolina by 3.2, Nevada by 3.1 and Arizona by 5.5. The polls had the wrong leader in the three northern states, but by margins small enough that a Trump sweep was well inside the expected range of error.

Why it happened: what the pollsters found

AAPOR's Task Force on 2024 Pre-Election Polling, chaired by Josh Pasek of the University of Michigan, released its report on 29 October 2025. Its findings:

Polls were more accurate than in 2016 and 2020. Across 611 polls in the final two weeks for all offices, the average absolute error on the margin was 3.3 points, down from 5.3 in 2020 and 5.2 in 2016. National presidential polls missed by 2.6 points and state presidential polls by 3.0, the most accurate state-level presidential polling since 1944.
But they leaned the same way for the third time. Polls overstated Democratic margins by 2.7 points across all offices (4.6 in 2020, 3.1 in 2016), and by 2.5 points in national presidential polls. The report calls three consecutive presidential misses toward one party something that "has few precedents."
Three groups were missed: Republican voters in Republican-leaning areas, who were under-represented relative to Democrats in those same areas; Hispanic voters, whose support for Harris the polls overstated; and 2024 voters who had not voted in 2020, whom surveys correctly saw leaning Republican but whose share of the electorate they underestimated.
Turnout shifted unevenly. Counties that backed Trump in 2020 had turnout surges, while Biden-leaning counties saw declines. Most surveys assumed an electorate shaped much like 2020's.
No evidence of herding. The polls' tight clustering came from more firms using the same political variables in sampling and weighting, not from firms copying each other. When the task force reweighted raw data from different firms to common targets, the spread was about the same as in the published polls.
No single method won. Firms weighting on party identification, using detailed likely-voter models, and Republican-affiliated pollsters, did slightly better, but the differences were small.

What our data shows

Our database holds 25 national pollsters' final polls from the last two weeks. Their simple average put Harris 1.6 points ahead, a 3.0-point miss on the lead. That is larger than the aggregators' 2.2, partly because the aggregators included Trump-leaning firms, such as AtlasIntel, that are not in our national set. Twenty-three of the 25 overstated Harris. Only five showed Trump ahead and four a tie.

The state errors were smaller than in 2016 and 2020 but still almost all one way. Taking each pollster's final poll in each state over the last three weeks, polls in 37 states overstated Harris's margin by 3.1 points on average, with an average absolute error of 3.4. Polls overstated her in 33 of the 37. The errors were smallest where polling was heaviest:

State	Final-poll average	Result	Error on lead
Iowa	Trump +5.3	Trump +13.2	8.0
Florida	Trump +6.8	Trump +13.1	6.3
Texas	Trump +7.6	Trump +13.7	6.1
Arizona	Trump +1.7	Trump +5.5	3.8
Michigan	Harris +1.7	Trump +1.4	3.1
Nevada	Trump +0.8	Trump +3.1	2.3
North Carolina	Trump +1.1	Trump +3.2	2.1
Pennsylvania	Harris +0.3	Trump +1.7	2.0
Wisconsin	Harris +1.0	Trump +0.9	1.9
Georgia	Trump +1.0	Trump +2.2	1.1

The seven swing states averaged a 2.3-point error; Florida, Texas and Iowa, all lightly polled, missed by 6 to 8 points. Iowa included the largest single miss of the year: Selzer & Co.'s final Iowa poll had Harris ahead by 3, and Trump won by 13. Arizona, with its large Hispanic electorate, was the only swing state where the error grew from 2020 (from 2.4 to 3.8 points), which fits AAPOR's finding on Hispanic voters.

Across the three Trump elections, the national error on the averages went 1.9, 3.4, 2.2, and the Midwestern state errors shrank: Wisconsin went from 7.2 and 7.3 points to 1.9, Pennsylvania from 5.0 and 4.0 to 2.0. The drop came as firms moved to weighting on party and past vote, though AAPOR found only small accuracy differences between firms that did so and firms that did not.

What the pollsters changed

The change between 2020 and 2024 was the spread of political weighting. In AAPOR's survey of 39 firms, 68% always used a political variable (party, past vote or turnout history) in their weights, 32% always weighted on recalled 2020 vote and another 42% sometimes did. Voter-file samples, which carry each respondent's turnout record, became more common, and 62% of firms with a likely-voter model used voter-file turnout history in it.

AAPOR found that past vote is powerful but hard to use. Reweighting 50 surveys showed that 2020 vote alone accounted for 50% to more than 80% of the gap between raw samples and 2024 results. But recall is imperfect, and weighting hard to the 2020 electorate "can effectively hard-code past electoral composition into a poll," which is exactly how polls missed the new and returning voters who helped Trump in 2024.

After the election, Pew Research Center added the 2024 presidential vote to the weights for its American Trends Panel and its reference survey from June 2025. Pew said its own polls had underestimated Trump in all three elections even after weighting on party, though the change moved its estimates by less than a point.

What it means for our forecast

Three misses in the same direction are hard to ignore, but AAPOR also notes that Democrats and Republicans have each been underestimated about equally often since the 1930s, and that midterm polls in 2018 and 2022 showed no such lean. Our correction rule handles this by design: we correct polls by half the average past miss, and keep that correction only if it would have improved past forecasts made for each election without knowing its result. After 2024 that means a modest shift toward Republicans, not a full one.

2024 also shows what a small miss can do. A 2-point error shared by every swing state turned seven toss-ups into a 312-vote Electoral College win. Our 5,000 simulated elections replay past state errors as whole patterns, so a clean sweep of close states by one side comes up often in a race this tight. The method is in our US simulation paper, and firms' 2024 records are part of our US pollster ratings.

Sources

How to cite

Polling Forecast (2026). Third time in a row: the smaller Trump miss of 2024. Retrieved 2 November 2026 from https://www.pollingforecast.com/research/missed-us-2024. Live figures change as new polls arrive; the date above is when the text was last revised.