Can we trust the polls?
Experts and AAPOR's president review the limits of the craft.

With the midterms less than three weeks away, Northeastern Global News brought together academics and Mary Losch, president of the American Association for Public Opinion Research (AAPOR), to revisit a recurring question: how far can polls be trusted? The short answer is that they are good at spotting competitive races but less reliable at separating candidates one or two points apart.
The diagnosis starts from known facts. Pew Research Center documents a steep drop in response rates over recent decades; those who do answer do not always reflect the electorate’s diversity, and some are less candid about their views. For professor Katherine Haenschen, the core problem is knowing whom to poll and who will actually vote: recent misses are mostly explained by failing to anticipate turnout.
History weighs on the field. In 2016, forecasts gave Hillary Clinton a 70% to 99% chance of winning, and she lost Pennsylvania and Wisconsin. In 2024, Ann Selzer’s final Des Moines Register poll had Kamala Harris three points ahead in Iowa; Trump won the state by about 13. Political scientist Nick Beauchamp notes that in 2016 polls actually pointed to a close race and missed by a couple of points.
According to Losch, AAPOR’s review found no consistently superior method, but some approaches were somewhat more accurate on average: higher-volume pollsters, those weighting by party identification, those using detailed likely-voter models and Republican-affiliated firms. Her warning: in close races, even small errors change who appears to be ahead.
Journalism professor Dan Kennedy adds another critique: media have focused too much on the horse race at the expense of issues and candidates. Aggregators and prediction markets such as Polymarket and Kalshi reinforce that trend, although Beauchamp points out that nearly all of those markets’ information comes from the polls themselves.
For consultants and campaign teams, the lesson is both methodological and about communication. The likely-voter screen is a poll’s most important decision and should be documented and stress-tested against turnout scenarios. And when presenting results, talk about ranges and margins, not winners. The November 3 midterm will be another public test for a trade that lives on its credibility.
The likely-voter screen is the most important methodological choice: document it and test turnout scenarios.
Summary written by the FOCUS editors based on the original story. Rights to the original content belong to its publisher.
FOCUS courtesy space


