What we count, and what we miss.

An interest group here is a list of FEC committees tracked together, plus the donors and employers that give alongside them. 23 groups carry money. Iran Policy is tracked for roll-call votes only. No committee or donor is tagged into it, so it carries no money and appears on no money page.

Groups are rows in the database, seeded from a file in this repository. Adding one is a data change, not a change to how any of the matching works.

Tracked committees counts every committee tagged into the group. With any recorded flow is the subset that has produced at least one contribution or expenditure to a sitting member in cycles 2010–2026: 884 of 971 across all groups. A committee with no flow is still tracked. It filed with the FEC and it belongs to the group; it has not given to anyone currently in Congress.

The FEC’s own committee-type field is not a usable label here. Some tracked committees it labels “PAC” are not, in the ordinary sense: among them independent-expenditure-only filers, electioneering-communication filers and communication-cost filers. The site never prints that field, and the word “committee” above should not be read as “PAC.”

Money reaches a member four ways. The database stores five flow types across them, because an individual contribution can be tied to a group by two different routes.

Committee money

A payment from a tracked committee’s treasury to a member’s campaign committee. Only the FEC’s direct-contribution code counts, so refunds, transfers and expenditure codes filed in the same report drop out. Memo lines are excluded: a filing often carries an informational attribution line beside the real receipt when a PAC bundles earmarked contributions, and counting both would double the total.

What it misses: money from any committee not on the tracked list. There is no inference here. A committee is in a group because it was put there.

Leadership PAC money

A payment from a tracked committee’s treasury to a leadership PAC the member sponsors. The FEC records the sponsor on the committee itself, and that record is the only link used here; a leadership PAC that merely names a member somewhere in its filings does not count. The same contribution code and the same memo-line rule apply as for committee money, and only the giver’s record is read, since the leadership PAC files the same payment again as a receipt. This money joins every total on the site. It came to $364,316,810 across cycles 2010–2026 in the August 29, 2026 rebuild, and a measurement taken before that rebuild put it at 27% on top of the committee money above, with about 78% of what leadership PACs receive from committees coming from tracked ones.

What it misses: contributions from individual donors to a leadership PAC. Only committee money is counted on that side.

Individual contributions

A person’s contribution straight to a member gets tied to a group two ways. Either the donor gave $200 or more to one of the group’s committees in one of the cycles 2000–2026, which fingerprints them for that group; or the employer they reported on the filing matches one of the group’s employer patterns.

What it misses: anyone who gave under $200 to the tracked committee, anyone who never gave to one at all, and any filing whose employer field is blank or written a way no pattern catches.

Where it over-reaches: a donor is keyed on last name, first name and state, so a common name fingerprinted by one group’s PAC can pick up a different person’s contributions. About 1.5% of individual dollars leak across groups this way. A fingerprint also applies to a donor’s whole career in our data, so one 2026 contribution to a new super PAC pulls that donor’s earlier giving into the group.

Independent expenditures

Money a committee spends for or against a candidate without coordinating with the campaign. Filings mark each one support or oppose. Both are stored; only support counts toward any total on this site. An attack ad a group ran against a member must not read as money that member received.

What it misses: many rows are filed with a blank candidate ID. Those are matched by the candidate’s name and seat instead, and only when exactly one sitting member or candidate holds that seat; $29,954,405 was recovered this way in the August 29, 2026 rebuild. The rest, about $80,000,000 measured before that rebuild, names candidates who are not sitting members, mostly presidential and Senate candidates who lost, and is still skipped. Every FEC candidate ID a member has held is carried, so spending aimed at a House member’s Senate run counts.

Measured figures in this section come from the August 29, 2026 rebuild review.

211 employer patterns cover 18 of the 23 money groups. Pro-Israel Lobby, Organized Labor, Health Professionals, Environment and Gun Lobby have none: their money arrives as committee contributions, independent expenditures and fingerprinted donors, so an employer-name match would add nothing.

Patterns are anchored to word boundaries and tried in file order. The first match wins, and a contribution is attributed to one group at most. If a contribution matches both a donor fingerprint and an employer pattern, the fingerprint takes it.

Every pattern runs against a committed corpus of real-looking employer strings on every test run, and each entry has to match its own group’s pattern and no other. A few over-matches are kept rather than narrowed, each with a note in the corpus saying why. The pattern for Boeing also catches “BOEING EMPLOYEES CREDIT UNION,” and “AM-PM ARCO CHEVRON,” “BLACKSTONE CONSTRUCTION” and “MORGAN & MORGAN LAND PREPARATION” all match patterns written for their industries. Together they are worth under about $150,000, measured August 29, 2026. One pattern is case-sensitive on purpose and therefore matches nothing at all, since the FEC upper-cases every employer field.

Totals cover sitting members of Congress only: 537 of them in this build. Money to someone who has since left Congress is still in the filings. It is not in these numbers, and a member who leaves takes their history out of every total on the site.

Contributions and expenditures are downloaded for cycles 2010–2026. Donor fingerprints are built from a wider span, cycles 2000–2026, because a fingerprint identifies a person rather than dating a contribution.

The people running against sitting members are in the database too: 2,026 candidates for 2026 House and Senate seats, read from the FEC’s candidate master, limited to declared candidates with a principal campaign committee. Their money is collected the same way. They appear in one place only, the race table on the incumbent’s page, and never in a leaderboard or a group total. A former member running again is not among them. In a state with a regular and a special Senate election in the same cycle, both races’ candidates appear together.

A contribution can belong to more than one group. One committee tagged into two groups writes one check, and both groups count it. That is deliberate: a group’s total answers what that group sent, and the same dollar is genuinely part of both answers.

Any figure that adds groups together therefore adds those dollars more than once. The combined career figure on the home page and on the groups index, and a member’s career total across all groups, are all sums of per-group totals. $3.71B is the sum of the 23 group totals in the table above, not a count of distinct dollars. Read it as a ceiling.

A single group’s own total does not double count. A contribution matches a fingerprint or an employer pattern, never both, and an amended filing replaces the version it supersedes instead of adding to it.

A contribution that matches both a donor fingerprint and an employer pattern counts once, toward the fingerprint. Seeding a new group adds committees, which adds fingerprints. A donor newly fingerprinted under one of those committees who reports a defense-contractor employer now counts for the new group instead of for defense.

That happened on August 29, 2026: the defense career total fell from $122,509,210 to $122,379,756, a drop of $129,454, all of it in employer matches. The pro-Israel total has no employer patterns behind it, so it could not move this way and did not.

This follows from one match per contribution rather than being a bug, but it means a group’s employer-matched total only means something alongside the list of groups that were seeded when it was computed.

A group name is a label on a committee list, and some of the lists are wider than their names suggest. Shares below were measured August 29, 2026.

• Big Tech & Telecom follows OpenSecrets’ Communications/Electronics sector, so it holds broadcast, streaming and studio PACs alongside software, hardware, cable and wireless. The name understates that.
• Artificial Intelligence holds both industry-funded PACs and the AI-safety PACs formed in opposition to them, so one figure covers money spent on both sides of the same argument. About 35% of the group’s support-direction independent-expenditure dollars come from the opposition side.
• Pharmaceuticals & Health Products includes health insurers, about 27% of the group’s PAC total, and the AMA PAC, a physicians’ group rather than a drug maker, which is its largest single committee. Both follow the anchors in the group’s own definition.
• Gun Lobby and Real Estate career totals come mostly from the donor-fingerprint path through mass-membership PAC donor lists, not from PAC money. Gun Lobby is $60,346,357 of individual contributions out of $72,782,285, against $7,285,777 of gun-PAC contributions; Real Estate is $86,571,226 of $183,398,046. Neither compares to Defense Industry’s total on the same axis without that caveat.

This project’s matching logic started from an earlier campaign-finance tracker’s, then moved away from it in four places: PAC and individual-contribution filings are deduplicated on committee plus transaction ID rather than transaction ID alone, since a transaction ID is unique per filer and not globally, and independent expenditures go a step further, deduplicating on the specific filing an amendment supersedes rather than any transaction ID; every principal committee a candidate has registered is kept, with the one active in the most recent cycle used as primary, rather than whichever the API lists first; a contribution matching two rules counts once; and donor identities collapse through one canonical key.

Running the validate command checks two things against a committed reference fixture: a floor on one senator’s career pro-Israel total, and a ±10% band on each group’s portfolio-wide career total. Only Pro-Israel Lobby and Defense Industry are checked. The fixture came from that earlier tracker and holds figures for those two groups and no others, so a group added later has nothing to be compared against, and comparing it against zero would fail the check for no reason. Leadership PAC money is left out of the check for the same reason: the earlier tracker never counted it.

• No proprietary industry-categorization data beyond the FEC’s own filings.
• No family or household association matching.
• No state-level campaign finance records.
• No hand-entered donors. Every fingerprint comes from a $200-and-up FEC filing.
• No bills tracked by sponsorship alone. A bill appears here only once it has reached a roll-call vote, which is 64 bills so far. Trackers that also count cosponsorship cover more ground than this one does.

If a figure here does not match another tracker’s, the difference is usually one of the rules above rather than a missing filing. The committee lists, the employer patterns and the matching code are all in this project’s repository, and the JSON behind every page is served alongside it.

Interest groups · All members