Methodology
Version 1.1 · formula unchanged since June 2026 · changelog
Greederboard answers one question with public data: how much is a member of Congress turning public office into private money? It doesn't judge their politics. The number comes from a fixed formula, and every input is sourced.
The six components
| Component | Weight | What it measures |
|---|---|---|
| Stock self-trading | 25% | Personal and spouse/dependent stock trades disclosed under the STOCK Act. Weighted up for volume, trades in sectors the member's committees oversee, suspicious timing and late filing. |
| Net-worth enrichment in office | 15% | Growth in net worth beyond a congressional salary and broad-market returns over the member's tenure. Wealth that predates office isn't counted. |
| Money → action alignment | 20% | Concentration of campaign/PAC money from specific industries alongside votes and bills favoring them. Correlation, never proof of a quid-pro-quo. |
| Campaign-fund personal use | 15% | Campaign or leadership-PAC money spent on personal-benefit categories: travel, dining, family payroll, clubs. |
| Revolving door & family | 15% | Immediate family lobbying or contracting tied to the office, and the staff-to-lobbyist pipeline. |
| Skips the job to fundraise | 10% | Missed votes and low attendance set against heavy fundraising or paid speeches. |
How the score is computed
Each component gets a 0–100 sub-score from sourced public records. The Self-Dealing Score is their weighted average: score = Σ(sub-score × weight) / Σ(weight), over only the components that have data.
Who does the research
AI research agents gather and cite the evidence for each component; the formula assigns the number. The model is the researcher, never the judge. Every claim links to its source, and each source is labeled by type: official record, reference database, news reporting, or data aggregator. Scores built mostly on aggregator and news sources are the first to be re-checked against primary filings.
Stock trades
Trade-level data comes straight from the official filings: House periodic transaction reports (PDFs from the House Clerk) and Senate reports (eFD). E-filed reports are parsed automatically; paper filings are listed but not itemized. A trade counts as late when the report was filed more than 45 days after the transaction, the outer limit in the STOCK Act. Amounts are the reported ranges, never invented point values.
Principles
- Behavior, not ideology. Policy positions never touch the score. We test whether scores line up with party and ideology and publish the result, including when it isn't flattering.
- Receipts or it doesn't count. No source, no score.
- Correlation is labeled as correlation. "Money → action" shows that funding and votes line up. It never asserts a bribe.
- The household, disclosed. Spouse and dependent trades count because they benefit the household, and they're flagged.
- Wealth isn't self-dealing. A fortune earned before office isn't scored.
- Words matter. We don't call anyone corrupt or an insider trader. Where an official body (a court, the House or Senate Ethics Committee, the Office of Congressional Conduct) has made a finding, we quote it.
Limits
193 of 539 members are scored; the sample was chosen for party balance, not at random. Sub-scores are calibrated estimates from disclosed records, not audited figures, and disclosure ranges are wide. It's a tool for transparency and questions, not a verdict or financial advice.
Changelog
- 2026-10-08 · v1.1. Added trade-level parsing of official House and Senate filings, late-filing detection, ban-vote records, FEC and attendance context, source-type labels, and a published party/ideology test. Formula and weights unchanged. Three scored members who left Congress (Greene, Mullin, Graham) moved to a former-members list.
- 2026-06-22 · v1.0. Six-component formula and weights published; full Senate scored.