Pre-election polls this year underestimated the strength of statewide Democratic candidates. In recent presidential elections, they underestimated the GOP share. A new report on the pre-election poll performance in 2024 and an interview with two seasoned veterans of the Election Day polling describe some of the problems the pollsters have had and what lies ahead. The debut of a new election polling operation last Tuesday is the latest effort in the evolution of the business.
The 2024 post mortem: Election pollsters had serious, well-documented problems in both 2016 and 2020, and after those elections, a distinguished group of pollsters and academics came together to examine what went wrong. On October 29, the American Association for Public Opinion Research (AAPOR) released its assessment of what went right and wrong in 2024, building on excellent work of the earlier task forces.
The nearly 80-page report is probably more than most people want to read about a past election, but the executive summary hits the high points. The AAPOR task force analyzed 116 general election polls from 2024 conducted between October 23 and November 5, 2024 concluding that “public polls painted an essentially accurate picture of an extraordinarily close contest.” State polls were the most accurate since 1944, they said, while national polls looked “about average.” There were twice as many state-level polls (mostly concentrated in swing states) as national polls in 2024, a development that will likely continue.
In 2024, polls again underestimated the Republican vote share, as they had in 2016 and 2020, but to a lesser extent. Polling firms used different methodologies and approaches with polls “from higher-volume firms, those weighting on partisan self-identification, those using detailed likely-voter models, and Republican-affiliated pollsters” being slightly more accurate. The task force said that but no single methodology guaranteed higher accuracy. The researchers also found no evidence of “herding,” that is, pollsters adjusting results so that they were in line with other polls. There were shifts in turnout within states, and the polls did not accurately anticipate it.
The Exit Polls: From Past to Present. In late October, the Roper Center’s Roper Roundtable hosted Murray Edelman, who was present at the creation of voter exit poll in 1967 and Joe Lenski whose firm ran the National Election Pool, the network exit poll consortium of ABC, CBS, NBC, and CNN from 2004 until this year. The two veteran pollsters are this year’s recipients of the Warren J. Mitofsky award for excellence in survey research.
They discussed the challenges that exit pollsters face including changes in the way people vote and declining response rates. According to Edelman, 8% of people voted early or absentee in 1992. In 2024, 59% did. So, in addition to talking to voters at the polls, exit pollsters have added pre-election surveys to capture pre-election or mail voters. Lenski noted that response rates were probably somewhere in the sixties when the first exit polls were conducted. In 2024, the response rate was 39%. Lenski discussed some of these problems in a June 2021 interview with AEI.
The Future? At the Roundtable, Lenski introduced the Voter Poll by SSRS which made its debut last Tuesday and will replace the AP/NORC/Fox News VoteCast pre-election surveys and the NEP’s consortium’s exit poll. Lenski believes it will take “the evolution of the election polling to the next step” by taking what his firm developed over the years in terms of in-person Election Day interviewing and what AP/Vote Cast did in their pre-election online surveys. The new Voter Poll has will use SSRS’s experience in building a panel and doing voter roll-based surveys for many news organizations. The new Voter Poll juggernaut will be a multimode survey which will incorporate in-person surveys with multiple samples and multiple modes of interviewing verified voters.
Is one exit poll worse than none? In the past when there was more than one poll, analysts gained insights through the different questions pollsters asked and the methodologies they used. But financial pressures and other changes have made the new collaboration necessary, and the variety of perspectives the expert teams involved provide assurance that we will have a good understanding of the messages voters are sending.
2025 and Beyond? The pollsters have taken past problems seriously. We will have to wait for a scholarly assessment of last Tuesday’s results and a verdict on the new Voter Poll, but the evolution is ongoing.
Source: https://www.forbes.com/sites/bowmanmarsico/2025/11/10/election-polls-their-past-present-and-future/
