Research methods · Release 2026-07-28.3

External comparisons are checks, not replacements for official evidence

This ledger records exactly which outside observations were acquired, which fields can be compared, where values agree, and which planned comparators remain unavailable.

Three spot checks are not a comprehensive accuracy study

External aggregators are comparators, not authoritative replacements for official source artifacts. Three exact roll-call spot checks cannot establish comprehensive extraction accuracy.

Current external-validation state

What has actually been compared

Download comparison JSON
3selected roll calls compared
3agree on every comparable field
0spot checks with a disagreement
2planned comparator datasets not acquired

External observations were frozen on 2026-07-28T00:15:00-04:00. Only one of the three selected release records has a directly recorded local value for every external summary field. Missing local summary totals remain unavailable rather than being inferred.

Comparator inventory

Available and unavailable sources stay distinct

Selected public pages acquired

MichiganVotes

Three deliberately selected roll-call pages spanning 2021, 2023, and 2025. This is a spot check, not a comprehensive comparison.

Open source website
Account required · not acquired

Plural Open / Open States

The public bulk-data description advertises per-session bill and vote archives, but the current download route redirects anonymous users to login and the Michigan landing page returned HTTP 500 during the 2026-07-28 acquisition attempt.

Open source website
API access key required · not acquired

LegiScan

The bulk API requires an API key, dataset session ID, and access key. No credential or dataset snapshot is part of this release.

Open source website

Selected MichiganVotes records

Field-by-field comparison

2021 senate · Roll call 100

4 comparable fields agree

No observed disagreement
FieldMichiganVotesThis releaseState
OutcomepassedpassedAgree
Yeas2929Agree
Nays55Agree
Excused or not voting22Agree

External page SHA-256: 1ff1aa03e68f03421df867a0b1c4d75e1f6bfffa74aca182a1a81bfb78983331

2023 house · Roll call 200

3 comparable fields agree

No observed disagreement
FieldMichiganVotesThis releaseState
OutcomepassedpassedAgree
Yeas105105Agree
Nays44Agree
Excused or not voting1Not recordedLocal value not recorded

External page SHA-256: 5acd5e5f2dc30f020e95eff67324b6fc46c0837c4512caf0c106a259d7f79574

2025 house · Roll call 80

3 comparable fields agree

No observed disagreement
FieldMichiganVotesThis releaseState
OutcomepassedpassedAgree
Yeas105105Agree
Nays11Agree
Excused or not voting4Not recordedLocal value not recorded

External page SHA-256: 277fab8436797823b7fee93367680677d9568e46e6a06bb1dde6c46e7e8d5b4e

Remaining work

What would make this comprehensive

  • Acquire frozen Open States and LegiScan session datasets under documented access and reuse terms.
  • Match complete measure and roll-call universes, not only selected pages.
  • Compare member identities and individual choices after resolving cross-system identifiers.
  • Classify every unmatched record by scope, timing, source coverage, or extraction error.
  • Freeze the comparison tables and their hashes beside the manuscript release.