Data Connections

Exact name matching means we miss cases where the same organization uses slightly different names across filings (e.g., "American Petroleum Institute" vs "API"). Fuzzy matching would increase coverage but also increase false positives. We chose precision over recall.

Currently based on ~429K comment details (about 4.4% of 9.76M total comments, up from 2% in March 2026). As full comment text collection continues against the Regulations.gov 1,000-req/hr rate limit, this overlap count grows daily. Organization names are self-reported and unstandardized, so matches depend on commenters using their organization's exact legal name.

100% of stock trades are now matched to a bioguide_id; an April 2026 rework of the name-and-chamber matcher closed the previous 14.5% gap. As of April 2026, trades also carry an SEC CIK identifier for the traded issuer on 67.7% of records, populated via a ticker-SIC crosswalk. Committee assignments reflect current membership only; historical rotations are not tracked.

The FEC-to-bioguide crosswalk covers 1,712 members. Contributions are PAC-to-candidate only (not individual donors — see the PII section of the methodology page for why individual-donor data is held locally but not deployed). Committee assignments are current session only — we don't track historical rotations, so a member's past committee work won't show here.

Temporal proximity does not imply causation. Members give many speeches and make many trades — some overlap is expected by chance alone. We chose 7 days as a meaningful window, but this is an editorial choice. The speech content is not automatically matched to the traded stock's sector; users must evaluate relevance manually.

SIC codes cover publicly traded companies only — mutual funds, ETFs, and ADRs don't have SIC classifications and are excluded. The committee-to-SIC mapping currently covers primary jurisdiction only; some committees (Appropriations, Budget, Foreign Affairs) are omitted because their jurisdiction is cross-cutting rather than sector-specific. The mapping reflects editorial judgment about which SIC ranges fall under each committee.

Not all lobbying filings reference specific bills — many describe issues generally. Congress number inference from filing year may be off for filings near session boundaries. The regex doesn't catch informal references like "the infrastructure bill" or bill names without numbers.

Name matching misses ~50–80 members who go by nicknames in lobbying filings (e.g., "Billy Tauzin" vs. "William Tauzin", "Dick Gephardt" vs. "Richard Gephardt"). We plan to add a manual nickname mapping table to capture these. The position text is free-form and inconsistent, so some staffers who list their boss's title may be incorrectly included or excluded.

GAO coverage is now dual-sourced: GovInfo for the historical archive through mid-2008, plus direct scraping from gao.gov for 2008–present. The combined gao_reports table holds 73,725 reports. Many GAO reports reference laws by USC section rather than Public Law number, which this matching doesn't capture. Approximately 4,990 reports have public law references, and 2,479 match to bills in our legislation table (which covers Congresses 93–119).

Exact name matching is conservative — organizations that testify under slightly different names (e.g., "Hogan Lovells" vs "Hogan Lovells US LLP") won't match. The 83 matches represent a lower bound. Fuzzy matching could increase coverage but risks false positives.