Current midterm polls do not look like they are about to repeat the kind of broad miss that marked some recent presidential surveys, according to the person who ran the polling and analytics programs on Biden’s 2020 campaign and later on the Harris campaign in 2024. The argument is simple: polls are often good at showing what voters want, even when they are less reliable at calling the exact outcome of a close election.
That distinction matters now because readers are once again asking whether the polls are right. The answer, from this veteran campaign pollster, is that it depends on what “right” means. In a midterm, the president’s party almost always loses ground, and the current signs point in the same direction: Trump’s approval ratings are at all-time lows, Democratic voter registration and primary turnout trends are strong, and many pollsters are trying to do a better job with vote recall this cycle.
The case for trusting the numbers is not that polling has become perfect. It is that recent history does not support assuming the same miss every time. Most public polling overestimated Democratic support in 2016, again in 2020 and again in 2024, although the 2024 error was smaller than the earlier two. But in the past two midterms, polling was quite accurate and did not tilt too far toward either party. That is the part that gets lost when people treat every new survey like a rerun of the last presidential election.
The industry has also changed. Phone polling is harder to do, fewer people are willing to answer surveys, and many firms have moved to cheaper, faster opt-in online panels. Those panels can be hard to get right, which is why some polling remains vulnerable to nonresponse bias. But they are not the same as a blanket excuse for bad numbers, and the fact that recent off-year elections in 2023 and 2025 underestimated Democrats cuts against the idea that the current cycle must automatically repeat the old error.
That is the friction in the polling debate: the public remembers the misses, but the recent midterms were broadly on target, and the latest off-year results leaned the other way. Pollsters are still adjusting methods, including how they account for vote recall, and that makes the next round of surveys more than a memory test for 2016 or 2020. They will be judged on whether they match the electorate now, not on whether they can be blamed in advance for a mistake they have not yet made.
For campaigns, that means the practical question is not whether polls can be trusted in some absolute sense. It is whether they are tracking the race closely enough to show where support is moving. On that score, the evidence so far says they are usable, and the stronger claim that they are destined to overstate Democratic support this year is not supported by the recent record.
