CalBallot

Annual counts, decade composition, and election-cycle patterns show when California voters faced the heaviest measure load.

Method: active records by election year. Trend fits are descriptive and not coverage-adjusted.

County totals reflect local-government density and source coverage. They are not population adjusted.

Method: active local records grouped by normalized county name. Map geometry loads from the public us-atlas county topology.

Finance coverage is intentionally scoped to matched statewide propositions from CalAccess. Each campaign is keyed by election cycle, so receipts attributed to the 2020 PROP_16 are the 2020 measure’s, not the 2010 reuse of the same proposition number.

Money matters but isn’t decisive: across all reportable spending (direct receipts, in-kind, loans, and independent expenditures), the better-funded side wins about 65% of the time and loses the other 35%. And this is the visible top of the iceberg — California’s local ballot has tens of thousands of measures with no comparable donor data.

Method: totals now combine four scopes of reportable money: (1) itemized monetary contributions to recipient committees tagged with the prop’s CAL-ACCESS ballot number, (2) loans received by those committees, (3) in-kind contributions reported on Form 460 Schedule C, and (4) independent expenditures (Form 461 / 465 / S496 filings) advocating for or against the measure. Contributions to untagged side-committees and Schedule E party-passthrough expenditures are still excluded. Our totals now approximate (but typically run somewhat below) press citations like Ballotpedia, since methodology decisions on attribution remain conservative. Combined aggregates draw from finance_statewide_v2.db (monetary) and finance_statewide_v3.db (loans + in-kind + IE), with the v3 attribution layer applying field-specific ambiguity rejection (AG queue IDs, multi-prop separators, regional/local measures) and post-ingest cross-source dedup. The spending-arc chart offers two lenses: election-cycle (totals per measure’s actual election year, the substantive frame for ballot-measure campaigns) and calendar-year (totals per Monday-of-week bucket, useful for cash-flow timing). The calendar view groups boundary weeks by their week-start year, so a transaction in the week of 2007-12-31 is attributed to 2007 even though it’s for the 2008 cycle. Donor lists aggregate across all four receipt types per donor; sector labels are hand-curated for prominent visible donors. Donor canonicalization is partial; some entities still appear under multiple legal-entity variants across v2 and v3.