Methodology

Party signals

Torrance city races are nonpartisan by ballot, but voters reasonably want to know which way candidates lean. We report five sourced signals per candidate (voter registration, self-statement, media identification, prior partisan candidacy, coalition composition) and let the reader draw the picture. We do not append party letters to candidate names, and we do not compute a "leans D/R" verdict.

Why we publish these signals

Local races labeled nonpartisan still attract partisan coalitions, partisan endorsements, and partisan donors. Reading the partisan picture from public records is harder than reading a single party letter on a ballot, and most resources either skip the question or assign a label without showing their work.

This approach follows the framework Ballotpedia uses for partisan signals in nonpartisan races: report what is sourced and let the reader draw the picture. Each signal is one input. None of them is the answer on its own. Together they form a sourced, neutral basis for a reader's own judgment.

The five signals

Voter registration

  • Definition. The candidate's own party registration on file with the Los Angeles County Registrar.
  • Source. LA County Registrar voter file, queried on a laptop. The file does not leave the laptop.
  • Method. Deterministic name and Torrance ZIP match against the voter file. On a clean match, the registered party is reported. On no match or an ambiguous match, the row reads "Not registered to vote in California."
  • Limits. California voter registration is administrative, not declarative. Registration alone does not indicate ideology. We report it as one signal among five.

Self-statement

  • Definition. A public statement by the candidate identifying with a party, in candidate websites, interviews, social posts, or campaign materials.
  • Source. The candidate's own published material.
  • Method. Hand-curated. We surface a quoted phrase and a link to the source material.
  • Limits. We report only public, sourceable statements. A candidate who has not made one is reported as "No public self-statement on party identification on file." That is a research outcome, not an editorial judgment.

Media identification

  • Definition. A credible local outlet identifying the candidate with a party in their reporting.
  • Source. Outlets reviewed (current set, edited as research expands): The Daily Breeze, The Los Angeles Times, KCRW, Easy Reader, Torrance Tribune. Search window: announcement of candidacy through publication date of the candidate page.
  • Method. Hand-curated. We surface a quoted line and a link to the article.
  • Limits. Outlet selection is editorial. Pieces behind hard paywalls we cannot read end-to-end are excluded.

Prior partisan candidacy

  • Definition. A prior candidacy by this candidate for partisan office, running under a party label.
  • Source. California Secretary of State filings and Los Angeles County Registrar candidate filings.
  • Method. Hand-curated. We surface the fact and a link to the official filing.
  • Limits. Federal, state, and county partisan races only. Prior nonpartisan candidacies are not a partisan signal and are not reported here.

Coalition composition

  • Definition. The partisan composition of the candidate's endorsers and itemized donors.
  • Source. Endorsers come from the candidate page's endorsements section, party-tagged via curated lookup tables and manual review. Donors come from the public NetFile filings, party-matched against the LA County voter file.
  • Method, endorsers. Exact, fuzzy, or manual match against curated party-of-record tables. Rows that do not match are surfaced for manual review and never silently coded as nonpartisan.
  • Method, donors. Deterministic name and ZIP match against the voter file. Aggregate counts and dollars only, computed on the laptop.
  • Limits. We never publish individual donors' party registration. The histogram is surfaced for a candidate only when at least 15 of that candidate's itemized donors match the voter file — a floor set to keep the bar from depicting a coalition shape that isn't there yet. Below that floor we publish a sourced suppression sentence in the coalition cell instead of a bar; equal-treatment is preserved by rendering the sentence at the same typographic weight as a populated bar, never by hiding the cell. Above the floor we apply a privacy mask: any Democratic, Republican, or No Party Preference bucket with fewer than three donors folds into the "Other" segment before publishing, so the precise minor-party-or-mainstream split for a single registrant cannot be read off the bar. Minor-party registrations (American Independent, Green, Libertarian, Peace and Freedom) are already part of "Other," along with any unknown / no-status records, so the published bar always has at most four segments.
Equal-treatment commitment

The five signals above describe what the system can surface. The rendered section on each candidate's page shows only the signals where sourced data has been verified for at least one candidate in the race; rows where no sourced data exists across the candidates are hidden rather than presented as empty. As editorial research adds sourced statements, prior candidacies, or local-press attributions, the corresponding rows appear.

Within any visible row, every candidate in the race is shown side by side. If one candidate's signal is sourced and another's research came back empty, both cells render at the same typographic weight, with the empty cell carrying its honest "no record found" copy and a link back to this page.

The donor histogram is decided per candidate, not per race. A candidate who clears both suppression thresholds renders their bar; a candidate who does not renders the sourced suppression sentence in the same cell at the same typographic weight. The symmetry across the row comes from every cell rendering at the same weight with sourced copy in place of any missing bar — not from hiding a populated neighbor's bar.

Visualization

The endorser party tally and the donor histogram render as stacked horizontal bars. Each bar shows the candidate's coalition shape as proportional segments; the tally text directly above the bar gives the precise counts. The same bar appears in three places on the site: the partisan-signals coalition row (full size, with segment labels), the at-a-glance battlecard at the top of every race page (smaller, no labels), and the home-page race switcher card.

Bar segments use the conventional party colors readers already associate with each party: a muted navy for Democratic, a muted red for Republican, a medium gray for No Party Preference, and a light neutral for any other party. These tones are dialed back from the saturated red and blue used in national election graphics so the bar reads as a coalition shape, not a partisan flag, and so two candidate bars sitting side by side do not visually compete for attention. The conventional colors do the labeling work for the reader at a glance; the precise counts above the bar do the rest. Every segment clears WCAG 2.2 SC 1.4.11's three-to-one contrast ratio against the page background. The visualization follows the perceptual hierarchy described by Cleveland and McGill (1984), which places position on a common scale at the top of the accuracy ranking, so the reader's primary read is the segment proportions, with color carrying the party identity as a secondary cue.

Color is applied to the coalition aggregate, never to a candidate. We do not tint candidate names, photos, accents, or any other element by party; the bars are the only place partisan color appears on the site.

Privacy and source commitments

Endorser party affiliation, voter-file source. For endorsers who are individuals or current/former elected officials, party affiliation is sourced from the Los Angeles County voter file when a deterministic name match resolves cleanly within Torrance city limits. The voter file is a public record under California's Public Records Act, available to anyone via the LA County Registrar. Endorsers registered outside Los Angeles County, or under a name different from their canonical record on this site, appear as unmatched. The site's editor maintains a manual override (an opaque registrant identifier) for cases where the editor has identified the correct registration through external research.

Organizations. Clubs, political action committees, unions, and advocacy nonprofits are not in the voter file. Their partisan affiliation, where it applies, is sourced from the org's own published material, FEC filings, or the California Secretary of State's political-committee registry.

Individual donors, never published. Aggregate counts and dollars are surfaced as the donor histogram; individual donors' party affiliations are never published. This rule differs from the endorser rule because endorsers have voluntarily lent their name to a candidate's campaign in a public way and donors have not. The aggregate suppression thresholds for the donor histogram (described under Coalition composition above) protect against by-elimination inference.

PII handling. The voter file is queried only on a laptop and does not leave the laptop. No voter PII is committed to this repository or transmitted to any third party. Match reports under scripts/match/reports/ contain only aggregate counts, the canonical name and party of matched endorsers (already public via the public-record source), and the disambiguation candidates surfaced for editorial review of ambiguous matches. No donor name plus party pair is ever committed.

Last updated

Most recent computation: 2026-07-24.