# Four market evidence and calculation notes

This evidence supports TWO editorial packages: one coordinated New York Chicago Detroit mortgage-free affordability study, and one Honolulu required-fee study. Three local mortgage-free manuscripts are uncounted supplements. No article or pitch has been published or sent.

## Source vintage and scope

Use the 2025 ACS one-year detailed tables, released October 29, 2026, and the frozen 2024 one-year baseline. Data were retrieved October 29 for 2025 and October 24 for the original 2024 snapshot. The included source manifests preserve retrieval dates, official URLs, byte counts and SHA256 hashes. October 30 checks reconfirmed the release and current Census methodological notes. This is a re-use of validated frozen inputs, not a claim that every source file was downloaded afresh October 30.

Run `python evidence/reproduce.py` from the housing-existing directory, or invoke that file by any path. The script locates input files relative to itself and uses only Python's standard library. It writes `evidence/reproduced-estimates.json` and `evidence/chart-data.csv`, then invokes `reproduce_family.py` to regenerate the full12-test family in `family-reproduced-full12.json`. A new directory or ZIP extraction does not require the original workspace. Chart source code is `charts/build_charts.py` and requires matplotlib; numerical reproduction does not.

## Exact cell definitions

New York 310M700US35620; Chicago 310M700US16980; Detroit 310M700US19820. All are metropolitan areas, not central cities. B25091 numerator at least 30% = E019+E020+E021+E022; at least 50% = E022. Both denominator = E013−E023, excluding uncomputed ratios. The severe group is included in the broad group. Estimates are owner-occupied housing units, expressed as owner households; they are not counts of individual homeowners. Mortgage-free is not debt-free or necessarily retired. The table does not identify a cause of cost burden, equity, tax arrears, program eligibility or foreclosure.

Honolulu 310M700US46520 is the Urban Honolulu metropolitan area. B25142 any-required-fee numerator = E002; >=$500 per month numerator = E012+E013+E014+E015. The principal denominator for both = E001, all owner-occupied households. The extra descriptive 2025 fee-payer measure uses E002 as denominator. Both use the same 52,717 high-fee numerator. These are overlapping measures, not additive buckets. Required HOA/condo fees are combined. Regular reported monthly fees do not constitute a comprehensive special-assessment measure or evidence of reserve adequacy. Threshold is nominal. No statewide Hawaii inference.

## Margins of error and comparisons

All MOEs are approximate 90% MOEs. Component sum/difference MOEs are root-sum-of-squares. Percentage MOE =100*sqrt(MOE_num²−p²*MOE_den²)/den, p=num/den; if negative under the square root, use the additive ratio form. There is no fallback for any reported rate. The count/denominator MOEs and component labels are retained unrounded. Rounding occurs only for prose and charts.

Annual difference =2025%−2024%; approximate MOE =sqrt(MOE_2025²+MOE_2024²), treating years as independent without covariance. Individual significance criterion is abs(difference)>MOE. The original study examined12 comparisons across6 markets. All three raw tables, including B25014, are bundled for both years so the full family can be reproduced. `family-reference-full12.json` is the frozen October 29 original calculation, traced to the included official manifests;72 field checks compare the recomputed family against it. `family-sensitivity-full12.csv` preserves the complete original family, including the LA and Vegas comparisons that are NOT articles in this batch. The family is not reduced to these four markets after looking at results. The original Holm and Bonferroni corrections at family alpha0.10 leave neither NY increase distinguishable; Chicago, Detroit and Honolulu changes fail even individual tests. An inconclusive result is not proof of no change. No cross-metro hypothesis tests or national rank are claimed.

The extra Honolulu denominator calculation is a descriptive 2025 snapshot, not a newly added annual test:52.9515754794% ±3.0878511718pp. Its90% interval crosses50%; say approximately53% or about half, not a statistically certain majority.

## Census comparability warnings

The 2025 lapse suspended collection for six weeks and affected sample size/response/MOEs. Population controls changed to Vintage 2025 using the 2020 MARC base; household record swapping and selected legacy disclosure rules changed. These are material context, not grounds to claim a definition break in these selected cells. The2024 questionnaire added required HOA fees to condo-fee questions and selected owner costs; do not extend these comparisons to pre-2024 without reviewing the measurement change.

Primary guidance links:
- [2025 release](https://www.census.gov/programs-surveys/acs/news/data-releases/2025/release.html)
- [Collection disruption](https://www.census.gov/programs-surveys/acs/technical-documentation/user-notes/2026-04.html)
- [Population controls](https://www.census.gov/programs-surveys/acs/technical-documentation/user-notes/2026-05.html)
- [Disclosure processing](https://www.census.gov/programs-surveys/acs/technical-documentation/user-notes/2026-06.html)
- [Fee question change](https://www.census.gov/programs-surveys/acs/technical-documentation/user-notes/2025-07.html)
- [2025 definitions](https://www2.census.gov/programs-surveys/acs/tech_docs/subject_definitions/2025_ACSSubjectDefinitions.pdf)
- [Statistical testing](https://www.census.gov/programs-surveys/acs/guidance/statistical-testing-tool.html)

## Source access limitations

The October 30 web tool initially could not open selected API/group URLs or guessed errata/year URLs. Clicking the official user-notes index successfully retrieved notes 2026-04,05,06. The current errata index displayed an occupation-coding item 153; its detail URL failed, so no blanket no-errata claim is made. Frozen schemas and raw estimates were independently validated in the previous refresh and reproduced here. The source manifests provide direct download URLs if readers wish to refetch. Cook County Assessor and Detroit program pages returned 403 on direct-open attempts, but official-domain search extractions and corroborating county Treasurer/Detroit official application material supported the included local service facts. These limitations do not justify inventing missing claims.
