2026 House.
435 districts.

A district needs a contested Democratic–Republican 2024 result on a comparable plan to receive a projection. Redrawn districts and races without two-party results stay unmodeled, even if local polls exist. We do not assign those seats a 50–50 prior or invent a baseline. National House-control odds are withheld.

The outcome chart includes only modeled districts. Its seat totals cannot be compared with the 218-seat majority threshold. District probabilities, poll blending, and shared-error assumptions have not been historically calibrated. Senate validation does not validate this separate House model.

The starting point is the 2024 district two-party margin, capped at ±60 points. We shift it by the change from the 2024 national House two-party margin to today's weighted generic ballot. The national 2024 anchor is affected by uncontested races. Local polls update this estimate with a maximum 80% weight.

Polls use a 90-day window and 21-day recency half-life. We prefer likely voters, take one question per poll, retain a single recent named matchup, and allow generic party questions. Sample weights are capped at 2,000 respondents; partisan polls receive half weight, and each pollster's total weight is capped. Local candidate names are polling observations, not a verified ballot roster.

10,000 simulations combine a shared 3.5-point national error with 5–10 points of district error. This is a provisional two-party model: incumbency changes, candidate quality, independents, vote transfers, and regional correlation need further work. Alaska's baseline uses first-round party totals, not ranked-choice transfers.

Geometry comes from the Census Bureau's 120th congressional district data and notices. Alabama, California, Florida, Louisiana, North Carolina, Ohio, Tennessee, Texas, and Utah are flagged for changed plans. Census warns that its Missouri 120th layer may not reflect the election plan; we use its 119th layer for Missouri. Legal changes may require further updates. Boundaries are simplified for display.

2024 results were extracted from the House Clerk's official election statistics. Major-party candidate totals include recognized fusion ballot lines. These are historical results, not a current incumbent or nominee list. Uncontested and same-party races do not supply comparable two-party margins.

Polling collected by The New York Times, reused under CC BY 4.0 and independently filtered and aggregated by DutyStation. The Times does not endorse this model. Polling refreshes every four hours, independently of the Senate update. A failed update retains the last good forecast and its timestamp. Source files, model version, hashes, and selected district polls are included in the download. Boundaries and historical baselines require a reviewed rebuild.

House forecast JSON · Update health