U.S. Senate

Why this race matters

This is the 13th closest Senate race in the forecast, 9.8 points, out of 35.

35 Senate races are on the 2026 ballot in this forecast (15 Democratic leads, 20 Republican leads). Seats that are not on the ballot are not assigned, so this is not a 100-seat chamber.

No incumbent is on the ballot.

No earlier forecast of this race is archived, so there is no previous projection, date, or reason to show.

History block in the model: D+12.9. That is the office blend, not one past election.

Last same-office result in the candidate file: 2022 D+9.1. A special on the 2026 ballot is not automatically that contest.

Polling block: D+9.1. The published result is still the county sum.

Counties

Every reporting unit in the race. Tap a county to open its precincts. The projected votes above are this table added up. They are not ballots cast.

Counties

County	Leader	Party	Margin	Vote share	Projected ↓
Hillsborough County	DChris Pappas, DChris Pappas	D	D+10.3	54.7%	173,407
Rockingham County	DChris Pappas, DChris Pappas	D	D+5.0	52.0%	150,147
Merrimack County	DChris Pappas, DChris Pappas	D	D+12.7	55.8%	68,754
Strafford County	DChris Pappas, DChris Pappas	D	D+19.5	59.3%	56,745
Grafton County	DChris Pappas, DChris Pappas	D	D+25.6	62.3%	29,972
Cheshire County	DChris Pappas, DChris Pappas	D	D+16.1	57.5%	24,364
Belknap County	RJohn E. Sununu, RJohn E. Sununu	R	R+6.8	52.9%	24,119
Carroll County	DChris Pappas, DChris Pappas	D	D+4.6	51.8%	20,520
Sullivan County	DChris Pappas, DChris Pappas	D	D+5.6	52.3%	14,924
Coos County	RJohn E. Sununu, RJohn E. Sununu	R	R+7.5	53.3%	10,248

Model

1. Raw data

   2024 presidential baseline

   National R−D 1.5

   County presidential file. Not the 2026 result.

2. Raw data

   Generic ballot

   D 51.7 R 43.1 R−D -8.6

   National Wikipedia aggregate, 2026-10-30. State generic ballots, where a public poll exists, are mixed with the national swing for a race that does not have its own poll: ME, MI, NH, NY, OH. A state with no generic-ballot poll keeps the national swing and is not assigned a neighbor's.

3. Raw data

   Party identification

   Pew D+5 and Gallup D+10, blend R−D -7.3

   Pew NPORS 2026, identifiers plus leaners: Republican 43, Democratic 48. Gallup party identification plus leaners, first two quarters of 2026: Democratic 49, Republican 39. Pew is 55% of this blend and Gallup is 45%.

4. Model input

   National block

   R−D -8.1

   60% generic ballot and 40% party identification. The swing versus the 2024 presidential margin is what gets transmitted.

5. Model input

   Swing transmitted

   0.40 × -9.5 = -3.8

   This race has its own poll. The swing in the history blocks, before they shrink toward that poll, is the national block minus the 2024 presidential margin. Elasticity is shrunk halfway toward 1.

6. Model input

   History block

   R−D -12.9

   Weights 2018 5 / 2024 35 / 2020 20 / 2022 40. If that office was on the ballot in 2022, that year takes the 40 and 2018 takes the 5. Otherwise 2018 keeps the 40. A missing year is filled with the 2020 result for the same office. Another office is not used. The year lines are Hillsborough County, the largest county, not the state total: 2018 from 2020 D+15.7 w5, 2024 from 2020 D+15.7 w35, 2020 D+15.7 w20, 2022 D+9.2 w40. The state number above is the vote-weighted sum. A 2024 county map is not copied forward unchanged.

7. Model input

   Electorate block

   R−D -12.9

   History reweighted by usual midterm turnout, plus half of the transmitted swing. Rural and presidential-only counties still move farther when the result is spread back to counties. That spread is ballot size, not the demographic file. Published ballots are the 2018 midterm canvass for this office, split by these modeled shares. They are not the 2024 presidential vote. The 2026 early-vote count is not inside this margin.

8. Model input

   Turnout block

   R−D -11.7

   Each county is weighted by its 2018 ballots divided by its 2024 presidential ballots. Small counties shrink halfway toward the state median. 2020 mail shares are not used.

9. Model input

   National environment

   R−D -14.1

   History plus the full transmitted swing. With no poll of this race, that swing is the state generic ballot mixed with the national block, or the national block alone.

10. Model input

    State environment

    R−D -12.8

    History plus 45 percent of the transmitted swing. This is a Senate and governor weight.

11. Model input

    State partisan lean

    R−D -2.8

    The 2024 presidential margin of this state, or of this district for a House race. It is heavier than the old state-environment block and is not shrunk toward the polling term.

12. Model input

    Trump popularity

    R−D -10.0

    New York Times approval average, October 29, 2026: 37.4% approve, 60.2% disapprove. Net approval -22.800000000000004. That is not a vote share. Four-tenths of the net is a margin, then scaled by this place's partisan lean (factor 1.09), so an unpopular president costs more in a Democratic state.

13. Model input

    Candidate favorability

    R−D 0.0

    Open seat or no incumbent. Favorability is zero.

14. Model input

    Demographics

    R−D -14.5

    One statewide block only. A county that is more college-educated or more nonwhite than the other counties in this race takes more of the transmitted swing, and a more rural county takes less. That changes this block, at its weight. It does not choose which counties move when the margin is spread back out. That spread uses ballot size. The shares are the 2012–2016 American Community Survey and the 2010 Census, from the 2018 county election-context file. Alaska is not in that file.

15. Model input

    Finance

    R−D -11.6

    No sourced spending edge. The block is history, or history shrunk toward the poll. It is not a finance model.

16. Model input

    Issues

    R−D -11.6

    No sourced issue score. Same treatment as finance. It is not a separate issue model.

17. Model input

    Polling

    R−D -9.4

    Individual polls -6.9 (80%) and the Wikipedia average -7 (20%). Named support R−D -6.9. Allocated support R−D -6.9. Named support does not move. Only the undecided line is split, 49.3% Republican and 50.7% Democratic. That is the 2024 two-party Republican share 48.6 moved halfway to 50 and capped at 60/40. A minor-party line is not published as its own candidate. Banned individual rows excluded: 0. One same-office miss, 2022: the poll was R−D -2.3 and the result was -9.1, miss -6.8. A single cycle is noisy, so 40 percent is kept. Capped at 2.5 points. It is added to the polling term only. It is not a separate weight. Applied to the polling term: -2.50. Public pollsters: 7. Outliers: 0. Recency half-life 16 days. Sample weight n/(n+350). A registered-voter poll is 0.90. An internal poll, or a poll sponsored by one party, is weighted 0. It is not in the average. An internal-only file is dropped. A poll that names both parties is not an internal. A poll more than 8 points from the median of the others is cut in half. Pollster quality nudges that weight by only a few percent. The same rule is used on a House poll. House district polls are likely-voter polls only.

18. Model input

    Generic ballot weight

    R−D -16.7

    The generic ballot stays in the blend next to the independent polls. It is not shrunk toward the poll and an internal does not replace it. Weight 9.5%.

19. Model input

    Top line

    Not a House term

    Senate and governor are the top line. They are not moved by themselves.

20. Model input

    Candidate

    R+0.00 added whole

    Open seat. No incumbency term. A negative popularity score is not added on top of the office baseline. No overperformance. An open seat, a scandal, or a missing incumbent gets none. These points are not watered down to the candidate weight. The candidate weight is a separate slice of the baseline. On a Senate or governor race with three public polls and no outlier, the points stop at that poll instead of moving past it.

21. Model input

    Scandal

    None

    No candidate on this ballot is scandal-defined.

22. Model input

    Endorsements

    R−D -0.32

    Trump endorsed John E. Sununu. The presidential lean here is R−D -2.8, so the endorsement costs 0.32. Ballotpedia, Endorsements by Donald Trump, 2026 tables, page updated August 5, 2026. A primary endorsement is counted when that person is the nominee. A lost primary, a candidate not on the ballot, and a withdrawn endorsement are not.

23. Model input

    Halfway floor

    Not applied

    Not applied. At least three public polls from different pollsters are not outliers. An outlier elsewhere in the file is already cut in half and does not put the floor back on.

24. Model input

    Pooling

    35% of the way from each history block toward the polling term

    Otherwise the shrink is 35%.

25. Model input

    Prediction market

    91% win probability

    Chris Pappas 91.0% is a win probability, not a vote share. Republican win probability 9.0 from Polymarket 9.0% Republican. A price of 50 is a tie. Each point above 50 is 0.35 margin points, capped at 12, so the raw margin is -12.0. That margin is shrunk 40% toward the pre-market margin -9.6, which leaves -11.0. It enters once, at 5.0%. The polling term is not nudged. Books under $10,000 and primary markets are ignored.

26. Model input

    State bend

    None

    Not applied.

27. Model input

    State environment adjustment

    R−D +0.2

    Saint Anselm College / AARP, early October 2026, New Hampshire. That state ballot, against the 2024 presidential margin, swings -9.2. The generic-ballot swing is -10.1. Twenty percent of the difference is kept, capped at 1.5, so this state is 0.2 versus the national environment. The ballot is already inside the swing for a race without its own poll. This is the extra.

28. Model input

    Early vote

    None

    No 2022 same-day party comparison, or the state does not report party on the ballots already cast. The raw Democratic mail lead is not used as a margin.

29. Model input

    Other ratings

    R−D -4.9 at 3.0%

    Cook, Sabato, Inside Elections, and the other columns on the same published tables (Decision Desk HQ, RealClearPolitics, Split Ticket, FiftyPlusOne, The Economist, Fox News, Silver Bulletin). 10 ratings, average R−D -4.9. A toss-up is 0, tilt 1.5, lean 3.5, likely 7, safe or solid 12. If this race is already further to that side, the rating does not pull it back. Mixed in after everything else, at 3.0 percent. It does not replace the poll. A seat nobody has rated is left alone.

30. Model input

    Weights

    polling 20.7% · generic 8.8% · stateLean 14.1% · trump 9.4% · favorability 8.7% · history 6.3% · candidate 4.7% · national 4.7% · electorate 3.9% · stateEnv 3.1% · turnout 2.3% · demographics 2.3% · finance 1.6% · issues 1.7% · market 4.9% · ratings 3.0%

    History is lighter. State partisan lean, Trump popularity, and candidate favorability are their own terms. A market, when it qualifies, scales every other weight and is not taken out of polling alone. Published ratings, when a race has them, are mixed in after that. They are not taken out of polling alone.

31. County-sum margin

    -9.8

    This is the sum of the county projections. Counties are not moved afterward to force this total.

32. Minor parties

    None

    Other is the 2024 presidential vote that was neither Democratic nor Republican, weighted by these counties. It is not the last forecast. A missing opponent is not Other. No Libertarian, Green, Constitution, or other minor line was on this ballot.

33. 80% interval

    -15.8 to -3.7

    10th to 90th percentile of 50,000 draws. Poll, national, state, turnout, and candidate error.

34. 95% interval

    -19.0 to -0.5

    2.5th to 97.5th percentile. Senate and governor state errors in the same state share a correlation of 0.30. The band shown is the marginal interval.