Skip to main content
PTP

Money-Vote Gap

Leaderboard

Two views into the gap between PAC dollars in and votes out. Members ranks (member, industry) cells by how far the member's Yes-rate diverges from the median for their own party in the same chamber. Bills ranks bills by revealed-preference subterfuge — industries lobbying to pass the bill while not appearing in its stated subject tags.Methodology →

MembersBills (subterfuge)

Members tab — Party Δ (primary) = member's Yes% − party median Yes% on the same industry's bills (same chamber, same cycle), with Wilson 95% confidence bounds. Chamber Δ (secondary, dim) = member's Yes% − chamber median Yes%. Chamber Δ is partisan-biased when the chamber has a partisan majority; shown for context but ranking uses Party Δ.

Sorting: rows are ranked by the lower bound of |Party Δ| at 95% confidence — i.e. the minimum deviation we can claim given the sample size. When the confidence interval on Party Δ crosses zero, the row is marked CI ∋ 0: the direction of the deviation is not confidently signed and the row ranks below all cells with a confidently-signed deviation. This kills the small-sample noise floor that point-estimate ranking lets through.

Confidence column: CI ∋ 0 means direction not confidently signed (Wilson CI on the deviation crosses zero). Preliminary means the direction IS signed but the cell does not pass Benjamini-Hochberg FDR correction across all leaderboard cells — could be a false positive from running ~1500 tests; treat as suggestive. A severity label (Slight / Moderate / High / Extreme) means the cell is BOTH confidently signed AND passes FDR — the rigorous-stats-claim tier.

Min 5 votes per cell, ≥$1 in industry PAC dollars. Min 5 same-party members in the pool for Party Δ; rows below that floor show "low-n party" and rank by Chamber Δ. See docs/audits/mvg-partisan-median-artifact-2026-05-14.md (party-conditional fix) and docs/audits/mvg-wilson-ci-2026-05-15.md (Wilson CI rigor) for the methodology audits.

Bills tab: subterfuge_score = Σ over unstated supporters of (1 + log₁₀(filings)). See methodology for the full audit.

Compare+2