Data downloads

• Alabama: US House (partial), State Senate (partial), State House (partial)
• Arizona: US House (none), State Senate (none), State House (none)
• Georgia: US House (partial), State Senate (partial), State House (partial)
• Hawaii: US House (none), State Senate (none), State House (none)
• Kentucky: US House (partial), State Senate (partial), State House (partial)
• Louisiana: State Senate (not-up), State House (not-up)
• Mississippi: US House (partial)
• Nebraska: State House (complete)
• New Hampshire: US House (partial), State Senate (partial), State House (partial)
• New Jersey: State Senate (not-up), State House (not-up)
• Nevada: US House (partial)
• Oklahoma: State Senate (partial), State House (partial)
• South Carolina: State Senate (complete)
• Virginia: State Senate (partial), State House (partial)

Partial means the candidate list we hold may leave someone out. None means we hold no list. Not up means no election for that office this year.

• Only races present in the candidate file are counted. States whose official list omits districts with no candidates (for example MA, WV) cannot show those districts, so uncontested counts there cover only districts that have at least one candidate row.
• Some "complete" states carry status "filed" or "nominee" rather than "certified" (for example WV filing list); the status is the state source's own.
• Holders come from the officials files as of their build; a vacancy or mid-term change after that is not reflected.
• Open means no current holder is on this race's ballot. It does not mean the holder is retiring: holders_running_elsewhere lists holders on the 2026 ballot for another district or office in the same state (common after redistricting). Open seats are counted only where the holder is known; races with no holder in the officials file are listed under open_status and excluded from open totals.
• Staggered multi-member chambers (for example WV Senate) are checked against the holders of the district, so a holder not up this year who is not on the ballot is not counted open.
• States with coverage partial, none or not-up are listed in not_counted and contribute nothing to any total.

Each of the 35 seats on the 2026 ballot is compared across its last two regular (non-special) elections, which are 2014 and 2020 for Class 2 and 2016 and 2022 for the two Class 3 specials (FL, OH). Where a general went to a runoff (GA 2020, LA 2014), the runoff decides the seat, so its winner and margin are used and the first-round margin is kept under first_round. Special elections held in the same seat since the earlier regular election are listed under specials and are not part of the comparison.

Districts redrawn for 2026 (district_changes.py) get no comparison, because the 2022 to 2024 shift no longer describes that district. Boundaries come from Census TIGER/Line 118th (2022) and 119th (2024) district files. Overlap is intersection area divided by union area of the same-numbered district in an equal-area projection (EPSG:5070). A state counts as having new lines when any of its districts overlaps its 2022 version by less than 0.99. In those states a district is compared only if overlap is at least 0.9999; elsewhere a district is compared if overlap is at least 0.99. Everything else is marked lines_changed_no_comparison and gets no margin change or flip. Special elections are ignored.

• Margins are of the total vote, so a district with strong third parties or write-ins shows a smaller margin than a two-party calculation would.
• Swing is a subtraction of two results; it does not say why a margin moved, and turnout changes are not adjusted for.
• Boundary overlap is computed from Census files built at different dates. Small differences below 1% of a district's area appear even where no law changed (water edges, municipal boundary corrections); they are treated as the same lines, and districts with a visible difference carry a boundary_note.
• The Census files show whether lines changed, not why; we name no court or legislative action.
• Uncontested races have no margin. Their flips are still reported when the winning party changed.
• DC's House delegate has no 2022 row in our results file, so DC is not compared.
• Candidate names are the files' display names. A flip is a change in the winner's party code, so a seat won by an independent in either year counts as a flip when the other year was won by D or R.