How far US polls miss, and how our method would have done
How close the US's final polls have come to the result at past elections, party by party and pollster by pollster, and what that means for today's average (Democrats 54.3, Republicans 44.8).
Final polls (each pollster's last poll ending in the month before polling day) of 863 Senate, governor and presidential races since 2000 missed the result by 4.7 points on the margin on average (Senate and governor races, 13 cycles). We replayed 2018, 2022 and 2024 with our method, fitting every parameter without the year being tested.
Every election, all ten countries: polls on the eve vs the result · which countries poll best · pollster leaderboard
The 2026 midterm polls against the results
Across 57 Senate and governor races with final polls, the polls overstated the Republicans by 3.1 points on the margin on average, and missed each race's margin by 5.3 points either way. Since 2000 the average statewide miss has been 4.7 points a race.
The final generic ballots (7 pollsters) averaged D+9.6 on the decided vote; the national House vote is heading for D+14.2, a miss of −4.6 points on the margin.
Provisional: where a count is not finished, the result is our live model's projection of the final count. The shares hardly move with the votes still to come, so the marks will barely change; they are redone automatically as each count is completed and when the states certify.
These results feed the pollster ratings and the 2028 forecast: the 2026 races join every past race the model learns from (how far polls miss, how much of the miss races share, each pollster's lean, and the Senate and governor fundamentals), and the election map gains an "if the polls miss like 2026" reading.
Race by race
| Race | Final polls | Poll average | Result | Miss (D-R) |
|---|---|---|---|---|
| Montana Senate * | 4 | R+22.5 | R+36.2 | +13.6 |
| Illinois Senate * | 2 | D+12.3 | D+23.7 | −11.4 |
| South Dakota governor * | 3 | R+17.2 | R+27.2 | +10.0 |
| South Dakota Senate * | 4 | R+12.0 | R+21.2 | +9.2 |
| Vermont governor | 2 | D+2.6 | R+6.6 | +9.2 |
| Texas governor * | 22 | R+1.9 | D+7.1 | −9.0 |
| Texas Senate * | 27 | D+3.2 | D+12.1 | −8.9 |
| Nevada governor * | 4 | D+1.6 | D+10.4 | −8.9 |
| Tennessee governor * | 3 | R+12.7 | R+4.0 | −8.7 |
| Georgia governor | 8 | R+0.1 | D+8.6 | −8.7 |
| Georgia Senate | 7 | D+9.4 | D+18.0 | −8.6 |
| North Carolina Senate | 12 | D+11.5 | D+19.7 | −8.2 |
| Rhode Island governor * | 3 | D+36.9 | D+44.7 | −7.8 |
| New Mexico Senate * | 2 | D+19.1 | D+26.3 | −7.2 |
| Alabama Senate * | 3 | R+16.6 | R+9.4 | −7.2 |
| New Mexico governor * | 2 | D+7.5 | D+14.6 | −7.0 |
| South Carolina governor | 6 | R+11.5 | R+4.5 | −7.0 |
| Colorado governor * | 3 | D+21.0 | D+27.9 | −6.9 |
| California governor * | 8 | D+24.5 | D+31.3 | −6.8 |
| Tennessee Senate * | 1 | R+20.6 | R+14.1 | −6.5 |
| Arizona governor * | 8 | D+6.6 | D+13.0 | −6.4 |
| New York governor * | 9 | D+12.8 | D+19.0 | −6.3 |
| Rhode Island Senate * | 3 | D+31.4 | D+37.5 | −6.1 |
| Maine governor * | 13 | D+10.6 | D+16.5 | −5.9 |
| New Hampshire Senate * | 7 | D+8.0 | D+13.7 | −5.8 |
| South Carolina Senate | 9 | R+1.2 | D+4.5 | −5.7 |
| Pennsylvania governor * | 5 | D+26.9 | D+21.2 | +5.7 |
| Alabama governor * | 3 | R+9.5 | R+4.0 | −5.6 |
| Arkansas governor | 1 | R+20.2 | R+15.2 | −5.0 |
| Oregon governor * | 1 | D+4.3 | D+9.1 | −4.8 |
| Colorado Senate * | 3 | D+20.9 | D+25.4 | −4.4 |
| Idaho governor * | 1 | R+20.4 | R+24.7 | +4.3 |
| Minnesota Senate | 12 | D+6.0 | D+10.2 | −4.2 |
| Ohio governor * | 20 | D+3.8 | D+7.9 | −4.2 |
| Iowa Senate * | 15 | D+0.1 | D+4.2 | −4.1 |
| Wisconsin governor * | 4 | D+4.3 | D+8.0 | −3.7 |
| Iowa governor * | 12 | D+7.5 | D+11.1 | −3.6 |
| Florida Senate (special) | 18 | R+6.4 | R+2.9 | −3.5 |
| Mississippi Senate * | 1 | R+3.4 | D+0.1 | −3.5 |
| Massachusetts Senate | 3 | D+28.6 | D+32.0 | −3.4 |
| Massachusetts governor | 3 | D+24.8 | D+28.0 | −3.2 |
| Nebraska governor | 7 | R+4.3 | R+7.5 | +3.2 |
| Illinois governor * | 2 | D+20.2 | D+23.2 | −3.0 |
| New Hampshire governor * | 7 | R+7.3 | R+4.3 | −3.0 |
| Florida governor | 20 | R+4.2 | R+1.3 | −2.9 |
| Oklahoma governor * | 2 | R+3.9 | R+6.7 | +2.8 |
| Michigan Senate * | 21 | D+4.7 | D+7.1 | −2.4 |
| Michigan governor * | 15 | D+8.5 | D+10.8 | −2.3 |
| Maine Senate * | 16 | D+3.8 | D+5.7 | −2.0 |
| Ohio Senate (special) * | 22 | D+2.9 | D+4.9 | −1.9 |
| Louisiana Senate | 1 | Even | R+1.8 | +1.8 |
| Minnesota governor * | 8 | D+11.6 | D+12.8 | −1.2 |
| Arkansas Senate | 2 | R+12.9 | R+11.7 | −1.2 |
| Kansas Senate * | 10 | D+3.9 | D+4.9 | −1.1 |
| Connecticut governor * | 5 | D+20.2 | D+19.1 | +1.0 |
| Alaska governor * | 4 | D+1.1 | D+1.6 | −0.5 |
| Kansas governor * | 8 | D+0.2 | D+0.6 | −0.4 |
Poll average: the mean of each pollster's last poll ending 4 October to 3 November 2026, Democratic (or the independent the polls tested) minus Republican, on the decided vote. Miss: poll average minus result; positive means the polls overstated the Democrats. * Result projected from the count so far.
Pollster by pollster
Average miss: how far each pollster's final polls were from the result on the margin, either way. Leaned: which side its misses favoured on average.
| Year | Office | Called right | Polls alone | Fundamentals alone | Our miss | Polls' miss | Seats D: ours / actual |
|---|---|---|---|---|---|---|---|
| 2018 | Governor | 33 of 36 | 30 of 34 (ours 31) | 30 of 36 | 3.8 | 3.9 | – |
| 2018 | House | 338 of 372 | – | 338 of 372 | 8.6 | – | 201 / 226 |
| 2018 | Senate | 30 of 32 | 26 of 29 (ours 27) | 27 of 32 | 3.4 | 3.6 | – |
| 2022 | Governor | 34 of 36 | 27 of 29 (ours 27) | 33 of 36 | 5.5 | 3.6 | – |
| 2022 | House | 312 of 336 | – | 312 of 336 | 6.4 | – | 180 / 184 |
| 2022 | Senate | 32 of 33 | 23 of 23 (ours 22) | 30 of 33 | 5.0 | 5.5 | – |
| 2024 | Governor | 11 of 11 | 8 of 8 (ours 8) | 10 of 11 | 8.2 | 4.7 | – |
| 2024 | House | 384 of 396 | – | 384 of 396 | 4.9 | – | 215 / 215 |
| 2024 | Senate | 30 of 32 | 23 of 24 (ours 22) | 29 of 32 | 4.4 | 4.5 | – |
Called right: the winner's party matched the result. Senate and governor races: our blend of the final polls (partisan polls adjusted) and the fundamentals, against the polls alone where there were polls and the fundamentals alone everywhere. Misses are on the D-minus-R margin, points. House: contested districts only in "called right"; the seat totals count every district in the test, uncontested ones included.
2018 senate: environment on the eve D+7.1 from the final generic ballot (5 polls); the House vote came in at D+8.6. Wrong: 2018 Florida Senate (we had D+3.4, result R+0.1), 2018 Indiana Senate (we had D+1.4, result R+5.9).
2018 governor: environment on the eve D+7.1 from the final generic ballot (5 polls); the House vote came in at D+8.6. Wrong: 2018 Florida governor (we had D+3.3, result R+0.4), 2018 Iowa governor (we had D+2.2, result R+2.7), 2018 Ohio governor (we had D+3.1, result R+3.7).
2018 house: environment on the eve D+7.1 from the final generic ballot (5 polls); districts from their previous House margin, incumbency taken out and put back (no presidential result by district is held for these lines); the House vote came in at D+8.6. Wrong: Utah 04 (we had R+4.2, result D+0.3), Texas 23 (we had D+6.9, result R+0.4), Minnesota 01 (we had D+6.7, result R+0.5), Maine 02 (we had D+0.9, result R+0.7), California 21 (we had R+5.2, result D+0.8), Georgia 06 (we had R+15.1, result D+1.0), New Jersey 03 (we had R+12.2, result D+1.3), South Carolina 01 (we had R+11.2, result D+1.4), Oklahoma 05 (we had R+12.0, result D+1.4), Florida 26 (we had R+3.5, result D+1.8), New Mexico 02 (we had R+14.9, result D+1.9), Virginia 07 (we had R+7.1, result D+1.9) and 22 more.
2022 senate: environment on the eve R+1.0 from the final generic ballot (6 polls); the House vote came in at R+2.6. Wrong: 2022 Georgia Senate (we had Even, result D+1.0).
2022 governor: environment on the eve R+1.0 from the final generic ballot (6 polls); the House vote came in at R+2.6. Wrong: 2022 Arizona governor (we had R+1.8, result D+0.7), 2022 Nevada governor (we had D+0.7, result R+1.5).
2022 house: environment on the eve R+1.0 from the final generic ballot (6 polls); districts from their 2020 presidential lean (states whose lines changed in 2024 left out); the House vote came in at R+2.6. Wrong: California 13 (we had D+7.1, result R+0.4), Washington 03 (we had R+8.3, result D+0.8), Arizona 01 (we had D+1.4, result R+0.9), Pennsylvania 07 (we had R+7.5, result D+2.0), Oregon 05 (we had D+4.5, result R+2.1), Pennsylvania 08 (we had R+7.0, result D+2.5), Nebraska 02 (we had D+2.2, result R+2.7), New Jersey 07 (we had D+3.9, result R+2.8), California 22 (we had D+12.8, result R+3.0), Maine 02 (we had R+10.3, result D+3.7), Nevada 03 (we had R+1.4, result D+4.0), Virginia 07 (we had R+1.0, result D+4.7) and 12 more.
2024 senate: environment on the eve R+0.8 from the final generic ballot (4 polls); the House vote came in at R+2.4. Wrong: 2024 Pennsylvania Senate (we had D+4.1, result R+0.2), 2024 Ohio Senate (we had D+0.1, result R+3.6).
2024 governor: environment on the eve R+0.8 from the final generic ballot (4 polls); the House vote came in at R+2.4. No race wrong.
2024 house: environment on the eve R+0.8 from the final generic ballot (4 polls); districts from their 2020 presidential lean; the House vote came in at R+2.4. Wrong: Ohio 09 (we had R+5.8, result D+0.6), Colorado 08 (we had D+1.6, result R+0.7), North Carolina 01 (we had R+1.3, result D+1.7), Nebraska 02 (we had D+3.2, result R+1.9), Ohio 13 (we had R+0.2, result D+2.2), Washington 03 (we had R+7.0, result D+3.9), New Jersey 07 (we had D+1.0, result R+5.4), New York 17 (we had D+7.0, result R+6.3), Michigan 08 (we had R+0.9, result D+6.6), California 22 (we had D+9.7, result R+6.8), Pennsylvania 01 (we had D+1.6, result R+12.8), Maine 02 (we had R+8.9, result D+51.1).
The miss each cycle
| Cycle | Races | Mean miss (D-R) | Average miss |
|---|---|---|---|
| 2026 | 57 | −3.1 | 5.3 |
| 2024 | 71 | +2.7 | 4.0 |
| 2022 | 54 | +1.1 | 4.2 |
| 2020 | 82 | +5.8 | 6.1 |
| 2018 | 66 | −0.1 | 3.7 |
| 2016 | 84 | +4.0 | 5.2 |
| 2014 | 71 | +3.3 | 5.0 |
| 2012 | 74 | −2.2 | 3.8 |
| 2010 | 62 | −1.7 | 4.7 |
| 2008 | 83 | +0.3 | 3.7 |
| 2006 | 53 | −0.8 | 3.9 |
| 2004 | 70 | +0.8 | 2.9 |
| 2002 | 30 | +3.5 | 4.6 |
| 2000 | 6 | −0.9 | 2.6 |
Poll average minus result on the Democratic-minus-Republican margin of the decided vote: positive means the polls overstated the Democrats. Senate, governor and presidential races by state (odd-year governors counted with the next even year).
Should we correct the polls?
Leave out each cycle, take half the other cycles' mean miss off every race's poll average, and see if it comes closer. Average miss per race: 4.65 as polled, 4.66 corrected, closer in 6 of 13 cycles. It does not help, so the state polls are taken as they are.
The same test with a correction that grows with how Republican the state is (the polls have overstated the Democrats most in states the Republican won for president): 4.62, closer in 8 of 13 cycles. It does better than one figure for every state, so we use it: 0.8 points off the Democratic margin in a state leaning 15 points Republican, 0.3 in an even state, -0.2 in one leaning 15 points Democratic.
Other things the record gives us: a race average's error is √(4.7² + 5.7²/k) points for k polls; the part every race shares in one cycle is about 2.9 points; partisan and campaign polls lean 2.5 points to their side against the independent polls of the same race (652 polls); polls taken 30-45 days out drift a further 3.2 points by polling day.
Final generic ballots against the House vote
Each election is predicted with the correction worked out from the other elections only, so it never sees the result it is tested on. Correcting the generic ballot was closer at 11 of 13 and further out at 2 (average miss per party 1.92 against 2.44 for the polls as published). So our projection uses the correction.
| Election | Polls' average miss | Corrected | Verdict |
|---|---|---|---|
| 2000 | 3.51 | 2.52 | Correction closer |
| 2002 | 5.34 | 4.43 | Correction closer |
| 2004 | 3.49 | 2.50 | Correction closer |
| 2006 | 1.54 | 0.52 | Correction closer |
| 2008 | 1.41 | 2.60 | Correction further out |
| 2010 | 1.91 | 0.84 | Correction closer |
| 2012 | 1.47 | 0.40 | Correction closer |
| 2014 | 4.89 | 3.96 | Correction closer |
| 2016 | 1.09 | 0.70 | Correction closer |
| 2018 | 0.77 | 1.93 | Correction further out |
| 2020 | 4.65 | 3.72 | Correction closer |
| 2022 | 0.80 | 0.58 | Correction closer |
| 2024 | 0.80 | 0.31 | Correction closer |
Final generic ballots held, each pollster's last poll before polling day: 2000, Gallup and Pew Research Center; 2002, Gallup; 2004, Gallup; 2006, Opinion Research Corporation, ABC News / Washington Post poll and Pew Research Center; 2008, Opinion Research Corporation; 2010, Fox News, Opinion Research Corporation and ABC News / Washington Post poll; 2012, Opinion Research Corporation; 2014, Fox News, Opinion Research Corporation, ABC News / Washington Post poll and Pew Research Center; 2016, Fox News, Opinion Research Corporation and ABC News / Washington Post poll; 2018, Fox News, CNN / SSRS, ABC News / Washington Post poll, Pew Research Center and Rasmussen Reports; 2020, CNN / SSRS; 2022, Fox News, CNN / SSRS, ABC News / Washington Post poll, Pew Research Center, Rasmussen Reports and Marist College; 2024, CNN / SSRS, Rasmussen Reports, Hart Research Associates / Public Opinion Strategies and Emerson College (35 polls from 10 pollsters). Fox News's final polls (2010, 2014, 2016, 2018, 2022) are read from the trend table in its April 2025 topline. Aggregators' averages are not used. Shares of the decided vote.
| Election | Final polls | Average D | Result D | Average R | Result R | Margin miss |
|---|---|---|---|---|---|---|
| 2024 | 34 | 48.8 | 48.3 | 47.5 | 49.8 | +2.8 |
| 2020 | 52 | 53.3 | 51.3 | 43.9 | 46.9 | +5.0 |
| 2016 | 28 | 46.6 | 48.1 | 41.7 | 45.9 | +2.7 |
| 2012 | 26 | 49.8 | 50.9 | 48.6 | 47.0 | −2.7 |
| 2008 | 22 | 53.6 | 52.8 | 45.7 | 45.4 | +0.5 |
| 2004 | 2 | 48.2 | 48.2 | 50.8 | 50.6 | −0.2 |
| 2000 | 2 | 45.5 | 48.2 | 49.7 | 47.7 | −4.7 |
National polls of the decided vote against the national popular vote (summed from the state results, MIT Election Data and Science Lab). These elections are the US rows in the cross-country leaderboard.
Figures as of . Published by Polling Forecast.




