OpenPolicy Labs
Request access

Candidate research you can check.

We build citation-backed intelligence on state legislative elections. Our platform tracks all 400+ candidates for the 2026 Pennsylvania General Assembly: every roll-call vote, every questionnaire answer, every dollar filed, every published endorsement, scored with methods anyone can reproduce.

Currently in private pilot with a statewide business organization for the 2026 Pennsylvania endorsement cycle.

Live evidence record, from the platform
HB 156 Regulatory − Workforce ·

Requires child care centers and family child care homes to install and maintain carbon monoxide alarms, establishing specific standards as part of state licensing requirements.

Floor vote · 164 yea · 33 nay Passed Committee · Apr 14, 2025
AI classifies the document · deterministic math scores it · source linked
403
Candidates tracked
54,000+
Cited evidence items
325
Bills classified & summarized
±2.5
Points mean error vs. published ratings

What the platform holds today

Built and operating for the 2026 Pennsylvania cycle. Every count below is live in the product, and every underlying record links to its public source.

Voting records

Every floor and committee roll call of the 2025–2026 session, weighted by how contested each vote actually was. Near-unanimous votes count less; deciding votes count more.

44,000+ votes · 6,000+ sponsorships

Candidates' own words

Questionnaires from Vote411 and Ballotpedia, campaign platform pages, press releases, and floor speeches, quoted with links to the original.

280+ questionnaire responses · 109 campaign platforms

Legislation intelligence

Session bills classified against policy priorities with plain-English summaries, committee dockets, hearing calendars, and per-bill vote rolls.

325 bills · 31 committees · live hearing schedule

Campaign finance

Current-cycle contributions from Pennsylvania Department of State filings, with donor organizations classified by industry and orientation.

$10M+ in 2026-cycle filings ingested

Endorsements

Published endorsement slates from business, labor, environmental, and education organizations, each cited to the organization's own announcement.

219 endorsements across 9 organizations

District context

Election history, voter registration, Census economic profiles, and calibrated win-probability estimates labeled as estimates, never as forecasts.

All 203 House districts covered

How the numbers are made

The platform is built on one rule: a reader should be able to check any number without trusting us.

Scores are arithmetic, not opinions
Language models classify documents: what a bill does, which way a vote points, what a statement claims. They never produce a rating. Every score is deterministic math over those classifications, reproducible from the same public records.
Validated against published ratings
Where established organizations publish their own legislator ratings, we show theirs verbatim and label the source. Our independent estimates cross-validate within ±2.5 points mean error of published ratings (0.88 rank correlation, five-fold cross-validation, n=187).
Provenance on every figure
A published rating, a modeled estimate, and a thin-evidence signal are visually distinct everywhere they appear. Confidence is shown next to the number, and low-evidence ratings are labeled provisional, not passed off as certain.
Disagreement is displayed
Where our data reads a candidate differently than established assessments, the gap is shown and explained on a public calibration page, not tuned away.
Human review before evidence
Public submissions, including our candidate questionnaire, enter a review queue. Nothing reaches scoring without human approval, and nothing is scored from a source that cannot be linked.

Public sources we build on

All primary data comes from public records and published documents.

Working on state legislative races?

The platform is in private pilot for the 2026 Pennsylvania cycle. We work with civic organizations, associations, and researchers. Reviewer access available on request.

jun@openpolicylabs.com