← Footnote

Methodology & neutrality

We cite a candidate’s leaning. We never assign one.

This page is the contract behind everything else on the site. It spells out, in plain language, exactly how every fact, score, and quote you see is produced: what we never do, where every number comes from, and how to tell us when we’ve gotten something wrong. Rule of thumb: if a calculation or editorial choice isn’t disclosed here, it shouldn’t be on the site.

Status:Footnote hasn’t ingested real candidate data yet. What you see elsewhere on the site today is fictional sample data, built to test layout. This page describes the rules that will govern real candidate data the moment it ships, published ahead of time so the policy is public before any candidate appears.

What we don’t do

We do not rate, score, endorse, or editorialize. We aggregate and attribute. Every ideological signal on this site traces to an external, attributable source. If a feature would require us to be the arbiter of where a candidate falls on the spectrum, we don’t ship it. That includes:

A note on AI.AI tools were used in building this site, including writing code and this page. They are not used to generate, summarize, score, or otherwise determine anything about a candidate. Every candidate-facing fact, score, and quote comes only from the disclosed sources above, through the disclosed methods, never from a model’s judgment.

Three tiers, always with a source

Leaning is presented in three tiers, never blended into one score. Each tier has a different evidentiary basis, and we show you which one you’re looking at.

Tier 1: Party affiliation

Shown for every candidate. A fact pulled from official candidate filings: coarse, but unimpeachable. This is the always-on baseline filter; we don’t infer or second-guess it.

Tier 2: Academic ideology score

Shown for candidates with a voting record. Federal incumbents get DW-NOMINATE scores from Voteview; state legislators get Shor-McCarty scores from Harvard Dataverse. We display the number as published, link the methodology behind it, and filter on it. We never recompute, adjust, or smooth it.

Candidates with no legislative record (challengers, most state-executive candidates) don’t get a score. We never fill that gap with a guess. Instead they get a visible “No legislative record yet”marker, and the ideology filter simply doesn’t apply to them. The gap is information for the voter, not something to paper over.

Tier 3: Self-reported positions

Shown for everyone, especially challengers with no voting record. Pulled from campaign sites and candidate questionnaires (Vote Smart’s Political Courage Test). Always shown as a verbatim quote with its source, never paraphrased or characterized in our own words.

How the leaning buckets are calculated

Filters like “lean liberal,” “center,” and “lean conservative” are derived from the Tier 2 score at query time; we never store a hand-assigned bucket. For federal candidates, the current (v1) method is a fixed cutoff on the DW-NOMINATE scale, which runs from −1 (most liberal) to +1 (most conservative):

This cutoff is arbitrary by necessity (any line drawn on a continuous score is), but it is fixed and applied identically to every candidate, regardless of party or office, with no exceptions. If we ever revise it, the new method replaces this section publicly; it is never changed quietly. State legislators’ Shor-McCarty scores use a different scale, and the published thresholds for bucketing them will be added here once state coverage ships.

Positions and factual claims

A self-reported position often fuses two different kinds of statement: a stance (a value or policy preference; no truth value, we never adjudicate it) and an empirical claim (a checkable factual assertion; it has a truth value). Treating them as one thing is the trap: publish the fused quote whole and we launder any false premise; rate the premise ourselves and we become the arbiter we promised not to be.

Governing rule: evidence, never verdict.We never label a claim true or false, and we never attach a fact-checker’s rating. Where a position contains a checkable empirical claim, we may eventually attach a neutral pointer to the primary data source (the relevant agency, Census, CBO, or academic review) so you can compare the claim to the data yourself. The judgment is outsourced to a citable source, never made by us.

Today, every position on the site carries this standing disclaimer:

“Positions are candidates’ own statements, in their own words; we do not verify factual claims within them.”

The evidence-pointer layer described above is a deliberate later addition, not part of this disclaimer yet, but the data model already carries the hook so we never have to repaint history when it ships.

Residual judgment, stated honestly:we can’t reduce judgment to zero. Deciding what counts as “checkable,” which source is “authoritative,” and whether to annotate at all are themselves judgments. Any such rule we adopt will be rule-based, symmetric, and published here, applied to every candidate regardless of who said it or which way it cuts. The goal isn’t the absence of judgment; it’s judgment that is minimal, disclosed, and auditable.

A note on participation bias:when positions come from a survey candidates opt into, response rates skew. Vote Smart’s own program has seen large partisan gaps in some races. We pull from campaign sites in addition to Vote Smart, which softens this, but an empty Tier 3 for a candidate is never a neutral signal about that candidate. It’s a coverage gap, and we’d rather you know that than read silence as a statement.

Sources

Every Position and IdeologyScorein our data carries its own provenance: source name, source URL, and retrieval date. If a row can’t cite itself, it doesn’t render. These are the sources that policy draws on:

SourceProvidesNotes
FECFederal candidates, committees, campaign financeFederal only; the most reliable record of who is actually running, plus money.
Congress.govOfficial bills, members, roll-call votesThe official successor to the now-shut-down ProPublica Congress API.
congress-legislatorsLegislator ID crosswalk (bioguide, FEC, Vote Smart, ICPSR…)Public domain. Used to join every other federal source together.
VoteviewDW-NOMINATE ideology scores (federal)Academic (UCLA). The Tier 2 source for every member of Congress.
OpenStatesState legislators, bills, votesBackbone of the state tier. Strong on sitting legislators; thinner on challengers.
Shor-McCarty (Harvard Dataverse)State legislator ideology scoresAcademic. Released infrequently and lags the current cycle; vintage is always disclosed.
Vote SmartPolitical Courage Test, stated positions, interest-group ratingsCovers challengers and state candidates, critical for Tier 3. API access is gated; campaign-site quotes are the fallback.

Local, county, and judicial races, and richer ballot-access data, sit behind a paid Ballotpedia license, a possible upgrade path, not part of the current scope.

Known limitations

Corrections & contact

TODO: not live yet.A real correction process and contact path for candidates and the public belongs here, and we’re not willing to ship a fake one just to fill the section. This placeholder stays until that channel exists. If you’re reading this and need to reach us, none of the data here is real yet, which is exactly why.