Congress Swing TrackerLegislative intelligence Swing vote analysis

Methodology

The swing score (0–100) estimates how likely a lawmaker is to break from their party on a bill they haven’t committed to. Each input is compared within the member’s own party and chamber.

How the default model (swing-v3) works

Scores computed October 10, 2026

Ideology gapweight 0.25How far the member’s overall voting record (DW-NOMINATE) sits from the typical member of their own party in the same chamber.
Scorecard gapweight 0.25How far the member’s ratings from advocacy groups (LCV and Heritage Action) stray from their party colleagues’. Where a group has scored votes similar to this bill, agreement on those votes is used instead of the overall rating.
Cross-pressureweight 0.20How strongly the member’s district or state leans toward the other party, so voters back home may pull them across.
Similar votesweight 0.30How often the member voted against their own party on earlier roll calls about this bill or similar bills.
Committee gatekeeper bonus+0.10 bonus, added on topMembers of the committees handling the bill decide whether it moves, so they get a bonus (largest for the chair and ranking member) instead of a weighted input.
Cosponsor penalty−0.70 penaltyThe sponsor and active cosponsors have already committed to the bill publicly, so their score drops toward entrenched.

Putting it together. Each input is scored from 0 to 1 against the member’s own party in the same chamber. The ideology gap, scorecard gap, cross-pressure, similar votes are averaged using their weights, counting only the inputs that have data and apply to the bill. The committee bonus is then added, and the total is capped at 100. Sponsors and cosponsors lose 70 points, which keeps them at 30 or below.

Why do scores change between models?

The same member can score differently on the same bill depending on the model (v2 or v3) and on directional mode. Each view uses different evidence and weights, so a member’s rank can move even though nothing about them changed. Pick the view on any bill or lawmaker page; it stays in the link.

swing-v2: who the member is

Uses only member-level inputs, so a member’s score is nearly the same on every bill they haven’t committed to: how far their DW-NOMINATE voting record sits from their party’s median (weight 0.35), how far their published advocacy scorecard ratings stray from party colleagues (weight 0.35; LCV counts half off topic), and how hard their seat leans toward the other party (weight 0.30).

swing-v3: who the member is, on this bill

v3 adds three inputs that depend on the bill:

Similar votes (weight 0.30): how often the member voted against their own party’s majority on earlier contested roll calls about similar bills (Voteview, their chamber, this Congress and the one before). A vote on the same bill counts 3, a bill with the same policy area and overlapping subjects counts 1, and the same policy area alone counts 0.5. The rate is pulled toward the member’s usual rate, as if they had cast 10 more votes, so a handful of votes can’t swing it to 0% or 100%. If a bill has little similar history (total weight under 5), the input is left out for everyone; if a member cast fewer than 3 similar votes, it counts as missing data for them.
Topic-matched scorecards: for LCV and Heritage Action, which publish the roll calls they score, the member’s agreement with the group on its scored votes similar to this bill replaces the group’s overall rating, at full weight, when there are at least 3 such votes. Otherwise the published rating is used as in v2.
Committee gatekeeper (bonus up to +0.10): members of the committees the bill was referred to control whether it moves, so they get a bonus: full for the chair or ranking member, 85% for a referred subcommittee’s member and 70% for other committee members. It never counts toward completeness, and scores are capped at 100.

Bands and rules

Swing is 70 and above, entrenched is below 30, and everything between is middle.
A score needs at least 50% of the input weight to be banded or ranked. Below that it is shown as insufficient data, with the missing inputs listed.
Cross-pressure is 0 when the seat leans 15 or more points toward the member’s party, and reaches 1 at 10 points toward the other party.
Partisan lean is computed by Congress Swing Tracker from presidential results (pres:2020=0.25,2024=0.75); it is not the Cook PVI.
Topical scorecards count at 50% weight on bills outside their topic. Non-voting delegates are scored but never ranked or counted.

Sources

The DownballotPresidential results by congressional district (partisan lean).
Federal Election CommissionOfficial statewide presidential results (partisan lean).
VoteviewDW-NOMINATE ideology scores. Lewis, Jeffrey B., Keith Poole, Howard Rosenthal, Adam Boche, Aaron Rudkin, and Luke Sonnet (2026). Voteview: Congressional Roll-Call Votes Database.
Congress.govMembers, bills, status and cosponsors (Library of Congress API).
Heritage ActionConservative legislative scorecard.
League of Conservation VotersEnvironmental legislative scorecard (LCV).

Glossary

DW-NOMINATE: A score from −1 (most liberal) to +1 (most conservative) computed from every roll-call vote a member casts, published by Voteview at UCLA.
LCV rating: The League of Conservation Voters, an environmental advocacy group, scores members from 0 to 100 on how often they voted with its position on the environmental votes it selects.
Heritage Action rating: Heritage Action for America, a conservative advocacy group, scores members from 0 to 100 on how often they voted with its position on the votes it selects.
Partisan lean (D+4, R+5): How a district or state voted for president compared with the whole country, from the 2020 and 2024 results (weighted 25% and 75%). R+5 means 5 points more Republican than the nation, D+4 means 4 points more Democratic, and EVEN matches the nation. We compute it ourselves; it is not the Cook PVI.
Swing, middle and entrenched: Bands of the 0–100 swing score. Swing (70–100): most likely to break from their party on the bill. Middle (30–70). Entrenched (0–30): least likely, including every sponsor and cosponsor.
Cross-pressure: How strongly a member’s district or state leans toward the other party. It is 0 for a seat 15 or more points in their own party’s favor and rises to 1 for a seat 10 or more points toward the other party.
Similar votes: Earlier roll calls on the same bill or on bills with the same policy area and subjects. The swing-v3 score uses how often the member voted against their own party’s majority on them.