Polls generally understated Democratic performance in a preliminary review of 27 Senate and governor races in the 2026 elections, according to Pew Research Center.
The analysis by Scott Keeter, published Friday, found that Democratic candidates outperformed their polling averages by at least 2 percentage points in 23 of the 27 contests. The average difference was about 4 points.
The findings use unofficial results available Nov. 4 and polling averages from RealClearPolitics, Silver Bulletin and Fifty Plus One. They are an early assessment that could change as ballots are counted and results become final.
The differences ranged from less than half a percentage point to about 9 points. Pew measured the difference between the actual Democratic margin and the polling average’s Democratic margin.
That measure tracks the direction of an error. It differs from an absolute-error calculation, which measures the size of a miss regardless of which party benefited.
Most races moved in the same direction
The preliminary pattern means Democratic candidates generally finished farther ahead, or less far behind, than their polling averages suggested. It does not mean Democrats won every race in the analysis.
Pennsylvania’s governor race was an exception. The polls overstated Democratic Gov. Josh Shapiro’s margin by less than 3 points; he won reelection by about 22 points in the results used for the analysis.
Pew also examined the national generic congressional ballot, which asks voters which party they would support for the U.S. House rather than naming individual candidates.
As of Nov. 5, the RealClearPolitics average put Democrats ahead by nearly 9 points on that question. The available national House vote margin was about 12 points, and The New York Times expected it to reach about 13 points as later-counted West Coast votes came in, according to Pew’s account.
Those national comparisons also remained provisional. The generic ballot, the national House vote and polling for individual statewide races are different measures and do not describe an identical group of contests.
A change from recent presidential polling
If the roughly 4-point average holds, Pew said the error would be larger than in the 2018 and 2022 midterms and comparable in size to errors in 2024 Senate and presidential polling.
Its direction would differ from the prominent polling misses in recent presidential elections. In 2020, for example, state presidential polls overstated Joe Biden’s margin by about 4 points, Pew noted.
A separate American Association for Public Opinion Research review of the 2024 election found that polling errors continued to favor Democrats on average, even though presidential polling improved from 2020. That history provides context for the reversal in the preliminary 2026 findings.
Why the polls missed remains unsettled
Pew cautioned that it was too early to establish why the errors occurred. Polling averages can reveal a broad pattern without showing whether it came from who responded, how the responses were adjusted or who ultimately voted.
Pollsters make different decisions about weighting respondents by party identification or recalled past vote, and about estimating the likely electorate. Those choices can affect results, but the preliminary comparison does not isolate their individual effects.
Keeter also discussed possible differences in voter motivation involving President Donald Trump and the economy. Those are potential explanations to investigate, rather than causes demonstrated by the early results.
Turnout and voter preference present separate problems for election surveys: a poll can misjudge which people will cast ballots even if it accurately records the preferences of those it interviews.
Pew’s review therefore offers an initial picture of polling performance in the selected races. A fuller assessment will require completed vote counts and more information about how the polls were conducted.


