2026 Midterms Forecast
Control of both chambers, priced the same way we price rate decisions and credit events: market odds where a contract exists, poll aggregates where a survey exists, and a structural model that translates national margin into seats. The mechanics are on the page so you can disagree with a specific coefficient rather than the conclusion.
Will Democrats win control of the US House in the 2026 midterms?
Stored in your browser only. No stake, no payout, nothing is submitted.
Will Republicans hold the US Senate after the 2026 midterms?
Stored in your browser only. No stake, no payout, nothing is submitted.
House seat-swing simulator
Move the three inputs that historically explain most midterm House movement. The seat count recalculates from stated coefficients, so the projection is reproducible rather than a simulation of chance.
4.6 seats per point, the postwar midterm average
2.1 seats per point of relative turnout advantage
0.34 seats moved per open seat, incumbency advantage forfeited
Baseline of 213 seats reflects the chamber as currently seated. The model is a linear approximation of a non-linear system: district geometry compresses the seats-per-point coefficient in wave years and expands it in close ones.
Senate races that decide the chamber
| State | Rating | Poll margin | Presidential lean |
|---|---|---|---|
| Georgia | Toss-up | +1.2 | -2.1 |
| Michigan | Lean out-party | +3.4 | +1.4 |
| North Carolina | Toss-up | -0.8 | -3.2 |
| Arizona | Lean out-party | +2.6 | -0.6 |
| Ohio | Lean in-party | -4.1 | -8.0 |
| Nevada | Toss-up | +0.9 | -1.1 |
Margins are shown from the out-party's perspective in percentage points. Presidential lean is the state's deviation from the national margin in the most recent presidential cycle, the single strongest structural predictor now that split-ticket voting has declined.
Method and limits
How are the 2026 midterm probabilities calculated?
Each chamber question blends prediction-market implied odds, an aggregate of publicly released polls, and a structural model built from generic-ballot history, seat geometry and turnout differentials. The weights are published on the forecast board.
Why do market odds and poll aggregates disagree?
Polls measure stated preference at a point in time; markets price expected outcome including turnout and late movement. The gap between them is itself information, which is why both are shown separately rather than merged silently.
What historically predicts midterm House swings?
Three variables explain most of the variance: the shift in national generic-ballot margin, the relative turnout differential between the parties, and the number of open seats created by retirements. The simulator on this page uses all three with stated coefficients.
Does LUMINAIRE endorse candidates or parties?
No. Coverage is structural and quantitative. The board forecasts outcomes and explains the mechanics behind them; it does not advocate for any candidate, party or policy.
Full forecast boardTrack recordUS Debt Clock
Non-partisan structural analysis. No endorsement of any candidate or party is expressed or implied, and no stakes are accepted on any outcome.