Shelter Governance Evidence Pack v0.1: Longitudinal Animal Shelter Outcome Records and PLAN Field Audit for Austin (2013–2025) and San Jose (FY2021–FY2026 partial)

This evidence pack supports empirical claims in the manuscript “When Reform Succeeds but Records Fail: Documentation Architecture and the Accountability Gap in Public Performance Systems.”

The pack contains the full analysis pipeline output for a longitudinal study of mandatory animal shelter outcome records from two US jurisdictions: Austin, Texas (N=173,775 records, 2013–2025) and San Jose, California (N=52,759 records, FY2021–FY2026 partial).

Key verified findings:- Austin euthanasia rate declined from 10.3% (2013) to 4.0% (2019); after a 2020 disruption (7.1%), it remained at or below 5.1% during 2021–2024.- F1 (actor identity): 0% present in both cities — no record identifies who made an euthanasia decision.- F6 (oversight gate): 0% present in both cities — no record identifies whether review occurred.- San Jose: all 5,379 euthanasia records have OutcomeSubtype missing (100%).- San Jose: 246 HEALTHY/NORMAL animals were euthanised; OutcomeSubtype is missing in all 246 cases (100%).- Austin: companion animal euthanasia subtype is missing in only 0.3% of records (31/10,833).

Pack contents:- analysis_log.txt — pipeline run log (20260505_130016) with SHA256 hashes for source files and computed outputs.- numbers_ledger_shelter.csv — 41 verified claims, all claimed == computed.- source_manifest.csv — source file registry with SHA256 prefixes.- table_A2_sj_by_year.csv — San Jose euthanasia rates by fiscal year and risk group.- table_B1_healthy_euth.csv — San Jose HEALTHY/NORMAL euthanasia cases (N=246).- table_C1_austin_by_year.csv — Austin euthanasia and live-release rates by year.- table_D1_cross_city.csv — cross-city comparison.- table_D2_plan_audit.csv — PLAN field audit.

Source data: publicly available open data from the City of Austin Open Data Portal and the City of San Jose Open Data Portal.

Pipeline: analysis v1.0, run 20260505_130016. All 41 numbers are script-reproducible from public source data. The Numbers Ledger confirms all claimed == computed values. SHA256 prefixes in analysis_log.txt enable independent source verification.

Programme context: This pack is part of the TBT/LFP-0/BTI research programme on decision-level reconstructability in mandatory reporting systems. Root Index: https://doi.org/10.5281/zenodo.17992916