Skip to main content
Everything PR News
Marketing

Polling Errors: What Went Wrong in 2016, 2020 and 2024

EPR Editorial Team4 min read
Share
Polling Errors That Change Headlines: 2016, 2020, 2024 — and the Brand-Research Implications

Polling errors are systematic gaps between survey predictions and real outcomes, not a single poll missing by a few points. The 2016 and 2020 U.S. presidential polls underestimated support for Donald Trump in key states, and 2024 polling recovered in part. The misses damaged trust in all online-panel research, including the brand and press surveys communications teams run.

Updated October 22, 2026. Part of the survey research guide.

What is a polling error?

A polling error is a systematic deviation across many polls and pollsters, not the gap between one poll and the result. Every poll carries a margin of error, usually reported at 95% confidence. A poll showing a candidate at 51% when the result is 48% sits inside the normal range for a 1,000-respondent survey of about ±3 points.

The consequential kind runs in one direction across the whole polling cycle. That is what happened in 2016 and 2020, when national and state polls repeatedly understated Trump's support in the Midwest and other swing states.

What went wrong in the 2016 polls?

The 2016 national polls were roughly right in direction, since Hillary Clinton won the popular vote by about 2.1 points. State polls in Pennsylvania, Michigan, and Wisconsin missed, and those states decided the Electoral College.

The American Association for Public Opinion Research (AAPOR) reviewed the cycle and pointed to several causes. The leading one was that many state polls did not weight by education, so respondents with college degrees, who leaned toward Clinton, were overrepresented. Late-deciding voters breaking toward Trump also contributed. The "shy Trump voter" explanation found less support in the evidence.

Election-night forecasts built on those inputs, such as the New York Times model and FiveThirtyEight, gave Trump low odds, and the narrative that polling had failed became part of the story.

Why was the 2020 miss bigger?

Pollsters rebuilt their weighting after 2016, and the 2020 error was larger. Final polls pointed to a Joe Biden popular-vote lead of roughly 8 to 10 points, while the actual margin was about 4.5 points. State errors were bigger than the national one.

AAPOR's 2021 review found the polls were off in a way that weighting by education did not fix. It could not pin down one cause, and it pointed to differential nonresponse, meaning Trump supporters were less likely to take surveys, as the most plausible explanation. The pandemic also distorted response patterns.

How did 2024 polling recover?

Polling improved in 2024. National averages showed a close race, and Trump won the popular vote by about 1.5 points while winning the Electoral College decisively. Several pollsters weighted samples to match how respondents said they voted in 2020, a technique meant to offset differential nonresponse.

The recovery is partial. The credibility damage from two consecutive misses has not fully healed, and skeptics now compare polling with prediction markets such as Polymarket and Kalshi.

What do polling errors mean for brand and press surveys?

Brand tracking, net promoter programs, and press surveys use online panels similar to those behind political polls, so they inherited part of the skepticism. Journalists now ask about sample source, weighting, and response rate before they cover a number, and executives ask the same questions about internal research.

Five lessons follow for communications teams:

  1. Disclose the method. State sample size, sample source, weighting, mode, and response rate in the release. The PR survey guide has a methodology note checklist.
  2. Size the sample to the claim. A finding about a narrow sub-group needs that sub-group's own defensible sample. See Survey Sample Size: 400, 1,000 or 30,000 Respondents?.
  3. Triangulate. A claim supported by several independent sources, such as Pew, Gallup, and your own data, holds up better than a single-source number.
  4. Remember what tracking surveys miss. Brand trackers that sample current customers tend to lag fast-moving reputation shifts, so pair them with monitoring of news and social conversation.
  5. Treat results as estimates. A survey is a probabilistic estimate, not a measurement of fact, and the wording of the release should say so.

Frequently Asked Questions

What is a polling error?

A polling error is a systematic deviation between survey predictions and actual outcomes that is larger than the normal margin of error and appears across many polls and pollsters. A single poll missing within its margin is statistical noise. A consistent miss in one direction across the industry points to a flaw in method.

What caused the 2016 polling error?

AAPOR's review pointed mainly to state polls that did not weight by education, which overrepresented college graduates who leaned toward Clinton. Late-deciding voters moving toward Trump also contributed, while the "shy Trump voter" idea found less empirical support.

Why was the 2020 polling error larger than 2016?

Weighting by education did not fully fix the problem, and pandemic conditions changed who answered surveys. AAPOR's 2021 review could not identify one cause and pointed to differential nonresponse as the most plausible explanation.

Did 2024 polling produce more accurate forecasts?

Yes, partly. Final polls were closer to the result than in 2016 or 2020, helped by methods such as weighting on recalled 2020 vote. The wider credibility damage has not fully recovered.

How do polling errors affect brand research?

Brand tracking and press surveys rely on online panels similar to political polls, so the misses raised scrutiny of their methods. Journalists and executives now ask for sample source, weighting, and response rate before trusting a survey number.

What should communications teams do about polling errors?

Disclose the full method, size samples to the claims, check findings against independent sources, and describe results as estimates. Publishing the question wording and data table alongside the headline number builds credibility.

EPR Editorial Team
Written by
EPR Editorial Team

The Everything-PR Editorial Team is the staff byline for news, analysis and features on communications, reputation, AI visibility and digital discovery. Everything-PR has published since 2009. Every article is reviewed by a human editor before publication. Coverage follows the Editorial Policy, and substantive corrections are noted on the article under the Corrections Policy.

Related reading

Other news

See all

Most brands are invisible inside AI search. Is yours?

EPR publishes the data every week.

Free. Weekly. Unsubscribe anytime.