How Accurate Were the 2026 Polls? A Race-by-Race Check
Dated editorial analysis; figures in this article are not recalculated automatically. Visit the linked race pages for current data.
Before Election Day this site published a polling average for every Senate and governor race that had at least one verified poll in the final 14 days. Now that most of those races have been called, we can put each average next to the result. This is a check of the polls and of our own averages; it is not a judgment about any candidate.
How we measured it
For each race we took the final polling average, expressed as the Democratic (or independent) share minus the Republican share, and the same margin in the unofficial result. The difference between the two is the miss. A positive miss means the result was more Democratic than the polls; a negative miss means it was more Republican. We counted only races that were called or at least 95% counted, which gave 25 races.
The headline numbers
- Average size of the miss: 4.6 points.
- Average direction: about 1.3 points toward Democrats.
- Leader matched: in 21 of the 25 races, the candidate ahead in the polling average was also ahead in the result. Two of the four exceptions were races where the average was exactly tied: Nebraska Senate and South Carolina Senate.
More polls, smaller misses
The clearest pattern is the number of polls behind each average. The three largest misses all came in races with one or two polls:
- South Dakota Senate: one poll, R +8; result R +21.
- Maine governor: two polls, D +6; result D +18.4 with 86% counted.
- South Dakota governor: one poll, R +18; result R +27.
Races with three or four verified polls were much closer. Ohio's special Senate election (three polls, D +5.2) finished at D +4.9. Kansas Senate (three polls, D +5.3) finished at D +4.9. Michigan Senate (four polls, D +7.9) finished at D +7.0. Maine Senate (four polls, D +5.5) finished at D +7.7.
A single poll carries the full weight of its own sampling error and its own method. An average of several polls from different firms tends to cancel some of those errors out. That is why this site labeled single-poll races clearly before the election, and why one-poll averages should be read as rough guides.
The Texas exception
Texas Senate had four polls and still missed by about nine points: the average was D +3.4 and the result D +12.3 with 96% counted. The four polls ranged from R +1.1 to D +7, so the polls themselves disagreed by eight points. Texas governor was the one called race where the polling leader lost: the two-poll average was R +1, and the result was D +7.4. We look at both Texas races in a separate article.
What this does and does not tell us
- Results are still unofficial. Late-counted ballots can change several margins by a point or more, which would change these numbers slightly.
- Twenty-five races is a small sample. A 1.3-point lean in one year does not show a lasting bias in either direction.
- Polls measure opinion during their field dates. Voters who decided late, or turnout that differed from pollsters' assumptions, show up as a miss even if a poll was carried out well.
The full table, including races still being counted, is on Polls vs results. For how our averages are built, see the methodology and how we verify polls.
Figures are unofficial. Vote shares, share of the expected vote counted and race calls come from the results snapshot this site published on November 4, 2026 (8:51 a.m. ET), drawn from the Fox News Decision Desk results pages; House calls are by The Associated Press. Polling averages are the mean of verified polls ending in the 14 days before Election Day, as shown on each race page. Certified results from state election offices may differ.
Relevant race pages: Kansas Senate · Maine Senate · Michigan Senate · Nebraska Senate · Ohio Senate · South Carolina Senate · South Dakota Senate · Texas Senate · Kansas Governor · Maine Governor