How we turn public records into evidence
GovGreed Research is the independent analysis arm of GovGreed. We take federal records that are already public — but scattered across millions of rows nobody aggregates — and turn them into reproducible working papers. This page documents what every paper in the series stands on: the data sources, the methods, the verification protocol, and the correction policy. Everything here is free to read, quote, and reproduce.
The working-paper series
Each paper carries a stable ID (GGR-WP-YYYY-NN), a version, a published date, an abstract, numbered tables and
figures, a limitations section, and an appendix with the exact queries behind every number. They are working
papers: independent analysis, released openly, not externally peer-reviewed — and verified instead through the
protocol described below.
441,165 federal contract awards (~$1.86T) mapped to 857 public companies. Extreme concentration — the top 10 firms take 88% — overwhelmingly defense, and the same contractors sit in the portfolios of dozens of members of Congress, both parties.
The Cabinet divested to take office (270+ sales); the President is exempt under 18 U.S.C. §208. His disclosure lists 588 companies — 243 federal contractors that collect ~90% of all mapped contract value between them.
STOCK Act compliance, 2012–2026: 12.5% of 100,875 disclosures filed past the 45-day deadline, 2,397 over a year late. The penalty is a flat, often-waived $200 fee and zero prosecutions — most members comply, the rule barely binds the rest.
An emerging instrument: the federal government taking grants, loans, and now equity in public companies. ~$41.5B across seven deals — and in 2025 the DoD became a major shareholder of MP Materials, the first time the state owned a critical-minerals firm outright.
Congress trades the AI stack in three layers — chips, platforms, and power. The tell is the power layer: ~91% of members' buying in Constellation, 68% in Vertiv, and 54% in Vistra came after 2024. The build-out's bottleneck became the portfolio's frontier.
Further papers are in preparation. Browse everything published so far in the newsroom.
Data sources
Every figure we publish traces to a primary federal record. We do not use scraped or paywalled intermediaries for headline claims. The core sources:
| Source | What it provides | Used for |
|---|---|---|
| USAspending.gov U.S. Treasury | Every federal contract award — recipient, obligated amount, awarding agency, product/service code, dates. | The contract leaderboard, sector/agency/PSC breakdowns, contractor identification. |
| STOCK Act PTRs House & Senate | Periodic transaction reports — every stock trade a member of Congress discloses, with filing dates. | Congressional ownership, disclosure-gap analysis, cross-referencing trades to bills and contracts. |
| OGE disclosures Office of Government Ethics | Executive-branch Form 278e (annual) and 278-T (periodic) financial disclosures — President, VP, Cabinet, senior staff. | Presidential and Cabinet holdings, divestiture tracking, executive conflict-of-interest analysis. |
| Congress.gov / GovTrack | Bills, sponsors, committee referrals, votes, and pipeline status. | Bill tracking, the Bill Pass Index, committee context. |
| SEC EDGAR | Corporate insider Form 4 transactions and 8-K material events; CIK ticker mapping. | Insider-trading signals, government-equity-stake detection. |
| FEC | Campaign contributions and PAC spending. | Donor and lobbying overlap analysis. |
How we build a finding
1 · Mapping awards to public companies
USAspending records a recipient name and a Unique Entity Identifier (UEI). We match these to a publicly traded ticker so that a second public record — congressional and executive disclosures — can be joined on. Matching is by name and UEI; subsidiaries are rolled up to the listed parent where identifiable. Coverage is the mapped subset (857 firms): awards to private firms, universities, and nonprofits are out of scope, so our totals are a floor, not the entire federal contract universe.
2 · Parsing disclosure documents
OGE 278e/278-T filings are PDFs. We parse them into structured line items — asset, value bracket, transaction type — and match assets to tickers. STOCK Act PTRs are ingested into per-trade rows keyed by the member's bioguide_id. Both are automated pipelines, and both are imperfect on some filings; where a parse is partial we say so, and we treat counts (how many holdings/contractors appear) as more robust than any single value, because disclosure values are reported only in wide brackets.
3 · Cross-referencing
The analytical core of most papers is a join between two public ledgers that are almost never placed side by side — for example, the contract recipient list and the congressional disclosure list, on ticker. We report the conjunction (who is paid and who in government holds the payee). We are explicit that conjunction is not causation: a disclosed holding is legal, often spouse- or fund-held, and is evidence of a conflict-of-interest pattern, never of a specific improper act.
Verification & corrections
Every headline figure is re-derived live from the production database on the day a paper is published or revised, and the queries that produce it are printed in that paper's Appendix A so any reader can reproduce it against the same primary source.
The protocol: (1) re-run headline totals against the live data; (2) cross-check leaderboards against record counts to flag single-record outliers; (3) re-count overlaps with COUNT(DISTINCT ...) rather than trusting cached statistics; (4) reconcile against the prior version and correct any figure that has moved.
Corrections are logged in the open. Each paper carries a revision history. When re-verification changed a fact, we say what changed and why. Recent examples: the contract total in GGR-WP-2026-01 moved from $2.13T to $1.86T on live re-derivation; and GGR-WP-2026-02 removed several widely-reported figures (Trump Media / World Liberty / specific crypto each >$50M) because they were not present in our parsed OGE data — we publish what the records contain, not what the press reports.
Standards & ethics
Conflict of interest & funding
GovGreed is a commercial congressional-trading-intelligence platform; GovGreed Research is its analysis function. Our papers receive no external funding, and no person or entity named in them is given prior review. The platform monetizes subscriptions, not the entities we analyze.
Models & technical methodology
Some of our work is model-driven rather than purely descriptive. The flagship model is the Bill Pass Index — a calibrated probability that any congressional bill becomes law (AUC 0.74, calibrated against 37,132 labeled bills). Its full technical methodology — inputs, formula, empirical priors, backtest, and known limitations — is documented separately:
Working with our data?
Every paper is free to cite with a link. Want a custom cut, the underlying dataset for a specific paper, or to verify a figure before you publish? Email govgreed@gmail.com — usually 24–48h, free with a link credit.
Browse the series
Not financial advice. All data from public federal disclosures.