Tracker | US elections

US midterms 2026

Who wins the Senate, the House and each race on 3 November, from our own model of the polls and past results. No other forecaster's numbers go in.

Forecast of 2026-10-16
18 days to the election

Senate: Democrats 60% Chance of a majority; Republicans 40% Senate seats 51D Most likely Democratic total (51 for a majority) House: Democrats 83% Chance of a majority; Republicans 17% House seats 226D Most likely Democratic total (218 for a majority)

Senate control

Democrats hold 34 of the seats not up this year, Republicans 31. Democrats need 17 of the 35 races to reach 51; at 50 the Republican vice president breaks ties. Point at a state to see its race; click it to jump to the details.

Safe RLikely RLean RTossupLean DLikely DSafe DNo race

Colours: Safe = 95% or more, Likely = 80-95%, Lean = 60-80%, Tossup = 40-60%.

How many seats

Each bar is the share of 20,000 simulated elections that end with that many Democratic seats (with King and Sanders).

Democratic majority Republican control

The market: Polymarket gives Democrats 64% (2026-10-16). It is shown as a benchmark; it is not used in the model.
The biggest risk: the polls overstated Democrats in each of the last four Senate elections. Our model, which also weighs past voting, leaned their way by 4.8 points on average in 2020, 2.7 in 2024 and 1.6 in 2022, and was on target in 2018. If 2026 repeats the average of those four, Democrats' chance falls to 26%. We tested correcting for it in advance and it made past forecasts worse, because the lean was large in the presidential years and small in the midterms (see Track record).
A judgement call: polls paid for by a party, or run by a partisan firm, are moved toward the other party, by an amount fitted on past elections. Without that adjustment, Democrats' chance is 52%.
Independents: Dan Osborn (Nebraska) has said he will caucus with neither party. Counting every independent winner with the Democrats changes their chance to 61%.

Every Senate race

The chance each side wins, our expected margin with its 80% range (the result should fall inside it four times in five), and how much the estimate leans on the state's polls rather than its past voting. Independents marked (I) are counted on the Democratic side where no Democrat is the main challenger.

Close

Iowa

Josh Turek (D)48% 52%Ashley Hinson (R)

Expected: Hinson +0.1 (range -5.5 to 5.2, D minus R). 34 polls, average D+2.2; polls carry 62% of the estimate.

Texas

James Talarico (D)54% 46%Ken Paxton (R)

Expected: Talarico +0.4 (range -4.9 to 5.7, D minus R). 56 polls, average D+3.1; polls carry 62% of the estimate.

Ohio (special)

Sherrod Brown (D)59% 41%Jon Husted (R)incumbent

Expected: Brown +0.8 (range -4.5 to 6.2, D minus R). 30 polls, average D+5.9; polls carry 61% of the estimate.

Alaska

Mary Peltola (D)38% 62%Dan Sullivan (R)incumbent

Expected: Sullivan +1.2 (range -6.6 to 4.3, D minus R). 21 polls, average D+4.1; polls carry 60% of the estimate.

Kansas

Adam Hamilton (D)18% 82%Roger Marshall (R)incumbent

Expected: Marshall +3.8 (range -9.3 to 1.7, D minus R). 14 polls, average D+1.5; polls carry 59% of the estimate.

Michigan

Abdul El-Sayed (D)89% 11%Mike Rogers (R)

Expected: El-Sayed +4.9 (range -0.4 to 10.2, D minus R). 54 polls, average D+4.1; polls carry 62% of the estimate.

South Carolina

Annie Andrews (D)11% 89%Darline Graham (R)incumbent

Expected: Graham +5.4 (range -11 to 0.2, D minus R). 12 polls, average R+0.1; polls carry 58% of the estimate.

Maine

Troy Jackson (D)90% 10%Susan Collins (R)incumbent

Expected: Jackson +5.3 (range 0 to 10.7, D minus R). 28 polls, average D+3.4; polls carry 61% of the estimate.

Leaning

Nebraska

Dan Osborn (I)7% 93%Pete Ricketts (R)incumbent

Expected: Ricketts +6.9 (range -12.5 to -1.2, D minus R). 12 polls, average R+0.7; polls carry 57% of the estimate.

Florida (special)

Angie Nixon (D)4% 96%Ashley Moody (R)incumbent

Expected: Moody +7.8 (range -13.3 to -2.3, D minus R). 17 polls, average R+7.3; polls carry 60% of the estimate.

Minnesota

Peggy Flanagan (D)96% 4%Michele Tafoya (R)

Expected: Flanagan +7.9 (range 2.5 to 13.4, D minus R). 16 polls, average D+5.8; polls carry 60% of the estimate.

New Hampshire

Chris Pappas (D)97% 3%John E. Sununu (R)

Expected: Pappas +8.6 (range 3.2 to 14.1, D minus R). 22 polls, average D+7.8; polls carry 60% of the estimate.

Louisiana

Jamie Davis (D)3% 97%Julia Letlow (R)

Expected: Letlow +10.6 (range -17.1 to -4.1, D minus R). 2 polls, average R+9.6; polls carry 43% of the estimate.

North Carolina

Roy Cooper (D)97% 3%Michael Whatley (R)

Expected: Cooper +9.1 (range 3.7 to 14.4, D minus R). 47 polls, average D+11.7; polls carry 62% of the estimate.

Safe

Mississippi

Scott Colom (D)2% 98%Cindy Hyde-Smith (R)incumbent

Expected: Hyde-Smith +12.6 (range -19 to -6.3, D minus R). 3 polls, average R+7.4; polls carry 46% of the estimate.

New Jersey

Cory Booker (D)incumbent98% 2%Justin Murphy (R)

Expected: Booker +17.5 (range 8.9 to 26.1, D minus R). No public polls: estimate from past voting only.

Georgia

Jon Ossoff (D)incumbent99% 1%Mike Collins (R)

Expected: Ossoff +11.5 (range 6 to 17, D minus R). 16 polls, average D+12.1; polls carry 60% of the estimate.

Illinois

Juliana Stratton (D)99% 1%Don Tracy (R)

Expected: Stratton +14.6 (range 7.6 to 21.6, D minus R). 1 poll, average D+10.3; polls carry 34% of the estimate.

South Dakota

Brian Bengs (I)1% 99%Mike Rounds (R)incumbent

Expected: Rounds +13.6 (range -19.4 to -7.9, D minus R). 9 polls, average R+6.5; polls carry 56% of the estimate.

Kentucky

Charles Booker (D)1% 99%Andy Barr (R)

Expected: Barr +17.4 (range -24.3 to -10.5, D minus R). 2 polls, average R+15.2; polls carry 36% of the estimate.

Colorado

John Hickenlooper (D)incumbent99% 1%Mark Baisley (R)

Expected: Hickenlooper +21.9 (range 13.3 to 30.5, D minus R). No public polls: estimate from past voting only.

Arkansas

Hallie Shoffner (D)1% 99%Tom Cotton (R)incumbent

Expected: Cotton +15.4 (range -21.4 to -9.4, D minus R). 5 polls, average R+7.5; polls carry 52% of the estimate.

New Mexico

Ben Ray Luján (D)incumbent>99% <1%Larry Marker (R)

Expected: Luján +17.9 (range 11.7 to 24.1, D minus R). 3 polls, average D+18.2; polls carry 48% of the estimate.

Oregon

Jeff Merkley (D)incumbent>99% <1%David Brock Smith (R)

Expected: Merkley +24.9 (range 16.3 to 33.5, D minus R). No public polls: estimate from past voting only.

Delaware

Chris Coons (D)incumbent>99% <1%Michael Katz (R)

Expected: Coons +25 (range 16.3 to 33.6, D minus R). No public polls: estimate from past voting only.

Alabama

Everett Wess (D)<1% >99%Barry Moore (R)

Expected: Moore +18.7 (range -25.3 to -12.1, D minus R). 2 polls, average R+19.1; polls carry 41% of the estimate.

Montana

Seth Bodnar (I)<1% >99%Kurt Alme (R)

Expected: Alme +16.2 (range -21.7 to -10.7, D minus R). 16 polls, average R+20.5; polls carry 60% of the estimate.

Tennessee

Marquita Bradshaw (D)<1% >99%Bill Hagerty (R)incumbent

Expected: Hagerty +23.4 (range -30.4 to -16.3, D minus R). 1 poll, average R+24.4; polls carry 34% of the estimate.

Oklahoma

N'Kiyla Jasmine Thomas (D)<1% >99%Kevin Hern (R)

Expected: Hern +25.5 (range -32.5 to -18.4, D minus R). 1 poll, average R+32.4; polls carry 34% of the estimate.

Virginia

Mark Warner (D)incumbent>99% <1%Bert Mizusawa (R)

Expected: Warner +21.6 (range 15.1 to 28.1, D minus R). 2 polls, average D+27; polls carry 44% of the estimate.

Idaho

Todd Achilles (I)<1% >99%Jim Risch (R)incumbent

Expected: Risch +26 (range -31.7 to -20.3, D minus R). 9 polls, average R+23.7; polls carry 57% of the estimate.

West Virginia

Rachel Fetty Anderson (D)<1% >99%Shelley Moore Capito (R)incumbent

Expected: Capito +33.4 (range -42 to -24.7, D minus R). No public polls: estimate from past voting only.

Wyoming

James W. Byrd (D)<1% >99%Harriet Hageman (R)

Expected: Hageman +32 (range -40.6 to -23.3, D minus R). No public polls: estimate from past voting only.

Rhode Island

Jack Reed (D)incumbent>99% <1%Raymond McKay (R)

Expected: Reed +27 (range 21 to 33.1, D minus R). 5 polls, average D+29.6; polls carry 51% of the estimate.

Massachusetts

Ed Markey (D)incumbent>99% <1%John Deaton (R)

Expected: Markey +30.5 (range 24.7 to 36.3, D minus R). 10 polls, average D+27.6; polls carry 55% of the estimate.

House control

Democrats win a majority (218 of 435 seats) in 83% of our simulations. The most likely result is 226 Democratic seats; four times in five it falls between 214 and 241. Each district's chance comes from its lean in past presidential votes, whether its member is running again, and the national House poll average (Democrats ahead by 6.2 points). There are too few district polls to use.

Democratic majority Republican majority. Democratic seats, share of 20,000 simulations.

Louisiana: its House primaries are on 3 November and its general election on 12 December. The forecast is for the final winner. Nine states drew new maps for 2026 (Alabama, California, Florida, Louisiana, North Carolina, Ohio, Tennessee, Texas and Utah); challenges to the Louisiana and Tennessee maps were still open on 2026-10-16.

Polymarket gives Democrats 90% to win the House (2026-10-16), shown as a benchmark only.

Every district

Each of the 435 districts on the 2026 maps, coloured by each side's chance. Point at or tap a district, zoom to a state, or find a district below. City districts are tiny on a map; the equal-squares view gives every district the same size (within a state, squares are not in their real positions).

Zoom to a state

Safe RLikely RLean RTossupLean DLikely DSafe D

Find a district

The 53 closest districts

Districts where neither side has better than a 90% chance, closest first.

North Carolina 1 Don Davis (D) · R+5 R 53%

Ohio 9 Marcy Kaptur (D) · R+5 R 53%

Michigan 10 John James (R), not running · R+3 D 53%

Iowa 3 Zach Nunn (R) · R+2 R 54%

Wisconsin 1 Bryan Steil (R) · R+2 R 54%

Pennsylvania 7 Ryan Mackenzie (R) · R+1 D 58%

Florida 22 open seat · R+4 R 60%

Iowa 2 Ashley Hinson (R), not running · R+4 R 60%

Maine 2 Jared Golden (D), not running · R+4 R 60%

Texas 35 open seat · R+4 R 60%

Florida 14 Kathy Castor (D) · R+4 D 60%

Michigan 4 Bill Huizenga (R) · R+3 R 67%

Pennsylvania 10 Scott Perry (R) · R+3 R 67%

Virginia 1 Rob Wittman (R) · R+3 R 67%

Wisconsin 3 Derrick Van Orden (R) · R+3 R 67%

Arizona 6 Juan Ciscomani (R) · Even D 70%

Colorado 8 Gabe Evans (R) · Even D 70%

Michigan 7 Tom Barrett (R) · Even D 70%

New Jersey 7 Thomas Kean Jr. (R) · Even D 70%

Virginia 2 Jen Kiggans (R) · Even D 70%

Florida 7 Cory Mills (R), not running · R+5 R 71%

Montana 1 Ryan Zinke (R), not running · R+5 R 71%

Florida 25 Jared Moskowitz (D) · R+3 D 72%

Texas 28 Henry Cuellar (D) · R+3 D 72%

Texas 34 Vicente Gonzalez (D) · R+3 D 72%

Alabama 2 Shomari Figures (D) · R+7 R 76%

Arizona 1 David Schweikert (R), not running · R+1 D 76%

Iowa 1 Mariannette Miller-Meeks (R) · R+4 R 77%

Missouri 2 Ann Wagner (R) · R+4 R 77%

New York 1 Nick LaLota (R) · R+4 R 77%

Ohio 10 Mike Turner (R) · R+4 R 77%

Pennsylvania 8 Rob Bresnahan (R) · R+4 R 77%

California 22 David Valadao (R) · D+1 D 80%

New York 17 Mike Lawler (R) · D+1 D 80%

Pennsylvania 1 Brian Fitzpatrick (R) · D+1 D 80%

Florida 16 Vern Buchanan (R), not running · R+6 R 81%

South Carolina 1 Nancy Mace (R), not running · R+6 R 81%

Washington 3 Marie Gluesenkamp Perez (D) · R+2 D 81%

Florida 9 Darren Soto (D) · R+8 R 84%

Colorado 3 Jeff Hurd (R) · R+5 R 85%

Colorado 5 Jeff Crank (R) · R+5 R 85%

Florida 4 Aaron Bean (R) · R+5 R 85%

New Jersey 2 Jeff Van Drew (R) · R+5 R 85%

North Carolina 11 Chuck Edwards (R) · R+5 R 85%

Ohio 7 Max Miller (R) · R+5 R 85%

Ohio 15 Mike Carey (R) · R+5 R 85%

Washington 5 Michael Baumgartner (R) · R+5 R 85%

Florida 11 Daniel Webster (R), not running · R+7 R 87%

Kentucky 6 Andy Barr (R), not running · R+7 R 87%

Nevada 2 Mark Amodei (R), not running · R+7 R 87%

Texas 23 open seat · R+7 R 87%

Michigan 8 Kristen McDonald Rivet (D) · R+1 D 87%

Ohio 1 Greg Landsman (D) · R+1 D 87%

The national House vote

Our average of 406 national polls puts Democrats ahead by 6.2 points (80% range 1.9 to 10.4). In three of the last four elections the same average, 18 days out, overstated Democrats by about four points:

Poll average 18 days before the election against the result (points, Democrats minus Republicans)
Year	Poll average	Result	Miss
2018	8.6	8.7	0.1
2020	7.4	3.1	-4.3
2022	1.8	-2.8	-4.6
2024	1.3	-2.6	-3.9

Track record

Before publishing, we ran both models on the last four elections as if it were 18 days before each, fitting them each time only on the other three. We set the tests in advance: a model's chances would be published only if it beat simple rules of thumb. Both did.

Senate

Brier score: the average squared gap between the chance given and what happened (0 is perfect, lower is better; 0.25 is a coin toss)
Election	Races	Our model	Poll average alone	Last presidential result	Incumbent party holds
2018	34	0.069	0.049	0.145	0.176
2020	33	0.045	0.066	0.062	0.061
2022	33	0.015	0.031	0.051	0
2024	35	0.034	0.051	0.044	0.086
All four	135	0.041	0.05	0.075	0.081

Our model called 128 of 135 races right (9 wrong for the poll average alone, against 7 for ours) and missed the margin by 4.6 points on average, against 6.
The weak spot: in midterm years alone (2018 and 2022), it only tied the poll average (0.042 against 0.041), and in 2018 it did worse.
Its 80% ranges caught 78% of results, so they are slightly too narrow.

House

Democratic seats: our forecast 18 days out against the result
Election	Forecast	80% range	Result	Chance of a Democratic majority
2018	230	213-248	234	82% (won)
2020	235	223-249	222	96% (won)
2022	217	204-230	213	47% (lost)
2024	212	200-224	215	28% (lost)

District by district (1597 contested races), the Brier score was 0.038, against 0.041 for a uniform national swing and 0.07 for "the incumbent's party holds"; it called 75 districts wrong, against 85 and 112.
Its worst year was 2020, when the national polls overstated Democrats and it gave them 13 seats too many.
After 3 November we will score these forecasts against the results, and against the market, here.

What we tried and did not use

Each was set out in writing before it was run, with the rule for using it, and tested the same way on the four past elections.

Correcting for the polls' past lean toward Democrats. Removing the average lean made Senate margins more accurate (an average miss of 4.1 points against 4.6) but winners less so (Brier 0.0425 against 0.0408; in midterm years 0.055 against 0.042). A first version that corrected the polls and the national mood separately counted the lean twice and was set aside.
Linking states by their population. States with more white adults without a degree, and states with more Hispanic voters, saw the model overstate Democrats more, the same way in every test. But building that link into the forecast made it worse on past elections (Brier 0.044), so it is not used.

How it works

Senate: three steps

Poll average. Every public poll of the race, weighted by its size and how recent it is. Where a poll asks several versions of the question, the one with the full list of candidates counts.
Past voting. What the state would do on its own record: its lean in the last presidential election, the national mood (the House poll average) and whether an incumbent is running.
Combining them. The two are weighted by how accurate each has been in past elections; more polls earn more weight. Then 20,000 elections are simulated, with a shared national error so that a polling miss in one state comes with misses in others.

House

The same past-voting step for every district, using its lean (the Cook Partisan Voter Index, a measure computed from presidential results) on the 2026 maps, then the same simulation. A national miss moves every district together.

What does not go in

No other forecaster's averages, ratings or probabilities, and no betting prices. Markets appear beside the model only as a benchmark.

Known weaknesses

Strong independents (Nebraska, Montana): the past-voting part judges them by the state's lean, which may understate them. In the two past cases it under-predicted both, by 3 and 5 points.
States without polls, and every House district, rest on past voting and the national polls alone: a local scandal or a strong candidate is not seen.
Georgia's run-off rule and Maine's ranked-choice voting are not modelled.

Data: election results 1976-2024 from the MIT Election Data and Science Lab (CC0); polls for 2018-2024 from FiveThirtyEight's poll files (CC BY 4.0); 2026 Senate polls, House districts and candidates from Wikipedia (CC BY-SA 4.0); 2026 national House polls from VoteHub; state and district boundaries from the Census Bureau (public domain) and, for the nine states that redrew for 2026, from each state's enacted plan: Alabama Legislature, California Statewide Database (AB 604), Florida Senate, Louisiana Legislature, North Carolina General Assembly, Ohio's state GIS office, Tennessee's TNMap, Texas Legislative Council (PLANC2333) and Utah's UGRC (CC BY 4.0). Want to go deeper? Contact us for the methods.