How the funding receipt works.

How donor interests are classified.

On each CA-2026 primary candidate's receipt and in the "by interest" view of the primary field, we report which named interests fund the campaign — building trades, real estate, healthcare, tribal governments, and so on. The headline number is "X% from named interests, excluding self-funding". It's narrower than total dollars raised. Here's exactly what it counts and what it doesn't.

What "named interests" means.

Money from PACs, unions, trade associations, tribal governments, and firms whose industry is identifiable from the public record. Almost every large contribution to a California campaign comes from an entity whose name encodes its interest — "California Teamsters Public Affairs Council", "Consumer Attorneys of California PAC", "Sheet Metal Workers Local Union 104", "Pechanga Band of Luiseno Indians". Those classify cleanly into 19 sector categories using a published keyword ruleset.

What it doesn't capture, by design.

Individual donors — people writing personal checks — would in principle be classifiable by industry through their disclosed employer and occupation. In practice we see those records fail in two distinct ways, and we merge both into a single "unattributable individuals" row on the candidate receipt:

• Not disclosed. Roughly 85% of individual gifts in Cal-Access carry no employer or occupation. State law requires it for gifts above the itemization threshold, but in practice the field is left blank, marked "N/A", or filled with non-information like "Retired" or "Homemaker" that gives no industry signal.
• Disclosed but unmatched. The remaining ~15% do provide an employer or occupation, but it's a small business or job title with no recognizable sector — "Self-employed", "Consultant", a one-off LLC. The classifier won't guess.

Both cases sit in the same merged row on the receipt; the underlying derived JSON preserves the split for anyone who wants to audit it. Either way they're excluded from the "% from named interests" headline. This is the most important caveat: for a grassroots candidate whose money is mostly small individual gifts, the headline will be relatively low not because no interests are involved, but because the individual-donor industry data simply isn't there to attribute.

Self-funding is also excluded from the headline. When a candidate loans or donates millions to their own campaign, that money tells you about the candidate's wealth, not about which outside interests have a stake. We show self-funded dollars as their own row but compute the named-interest share as a percentage of non-self-funded money.

How the classification actually works.

Two passes over the donor name, employer, and occupation strings: first an ordered keyword ruleset (specific sub-sectors like "building trades" are checked before generic union keywords; teachers/nurses are their own buckets), then a curated override map for closely-held LLCs and individual "affiliated entities" we've classified by hand. The rules + overrides currently classify about 94% of identifiable organizational dollars across the 122 candidates. The residual is small businesses and shell entities where attribution would require per-name research.

Every dollar in the bar is real — the computed totals reconcile to each committee's filed total_raised at 100% across all 122 candidates. We restrict to primary candidate committees and apply the same cycle-attribution rules the underlying Cal-Access pipeline uses: 2026-cycle receipts for state-legislature committees, which persist across cycles; all receipts for "for Governor 2026" committees, which are cycle-locked by name.

Scope and what's next.

Coverage is currently the 122 California 2026 primary candidate committees (Governor + State Senate + State Assembly), built from the California Secretary of State's Cal-Access bulk export. The same classifier extends to federal FEC data and to outside / independent expenditure spending — those are the natural follow-ups. Outside spending is the bigger transparency gap: it often dwarfs direct contributions and isn't shown in this view yet.

Pipeline: docs/working/gov-interest-prototype/ · classifier classify_gov_donors.py + curated overrides.json · derived JSON data/california/derived/ca_2026_donor_interests.json.