Methods / Draft rubric 0.1
A source can be official. A grade is still an interpretation. We keep those two things separate.
Every card names its population, unit, period, source and limitations. Missing observations stay null and visibly unavailable. A measured zero must remain zero. Source suppression, stale data, missing connections and withdrawn results need distinct labels when integrations are added.
There are no grades in this release. Before one appears, publish its thresholds, direction, weights, reference group, uncertainty policy and rubric version. No overall state score is selected. Incomplete inputs must not quietly become a lower grade or a reweighted score.
Future releases must retain source snapshots or lawful retrievable references, source versions, inputs, calculations and reviewer decisions. Distinguish the period measured from retrieval, validation and publication dates. A new rubric must not silently rewrite old grades.
Future updates should be staged and checked before release. A failed fetch is not a new observation. Keep a still-valid prior value with a stale notice when appropriate; withdraw a value if retracted or known wrong. New sources and methodology changes require review.
Living does not mean untraceable
The second edition connected New York unemployment; the third extended that coverage to all 50 states and D.C.; the fourth added 2024 state price levels from BEA. The fifth adds two ACS 1-Year 2024 measures—rental housing burden and uninsured residents under 65—for all 51 jurisdictions. New York’s unemployment remains 4.4%, unchanged. A new connection is not evidence that a measure rose from zero. Each measure keeps its own period: July 2026 for unemployment, 2024 for price levels and for both ACS measures. Reading them as one moment in time would be wrong.
Comparisons across rubric versions require a common recalculation or a visible break. Small changes should not be presented as meaningful when the source uncertainty does not support that interpretation.
Current implementation: 7 retained local editions, source-response evidence, validation and integrity hashes. The older New York capture retains an unavailable October 2025 value as missing—not zero. BEA’s annual reference is U.S. = 100; it is not inflation or a grocery-only index. The two ACS measures carry a margin of error at the 90% confidence level, taken from the Census Bureau’s published margins; that margin describes survey sampling error only and does not cover non-sampling error. Before these figures shipped, three gates were run and recorded with the capture: a release check that separates a confirmed absence from an unreachable endpoint, a check that the published DP04 denominator describes the same universe as B25070 in all 51 jurisdictions, and a scan of Census errata notes. Recurring collection, the other three measures, approval queues and unattended publication are not implemented.
Inspect the BLS source capture · Inspect the BEA source capture · Inspect the Census source capture
Source freshness / Checked 2026-09-23
Each measure below is a retained, dated estimate. Whether a newer release exists is a separate question, answered per measure. An unresolved check does not make a retained estimate wrong.
| Measure | Period covered | Current-release check |
|---|---|---|
| Unemployment rate | July 2026 | Newer release exists — not in this package |
| Relative price level | 2024 | Matches the source’s current release |
| Rental housing cost burden | 2024 | Current-release check unresolved |
| Uninsured residents under 65 | 2024 | Current-release check unresolved |
| Published corrections (errata) | Historical, 51 retained notes | Partial — not a current all-clear |
Seven definitions across 50 states and D.C. 204 observations across four measures; 153 unconnected slots. Each observation carries its own period, and the Census measures carry a 90% margin of error. No grades.
Download observations & availability JSONOfficial news references, data providers and the exact evidence behind unemployment, price-level and ACS coverage, including the gate results that justify the Census figures.
Download sources JSON7 local review snapshots and their content hashes. None is a public release or a scheduled update run.
Download change ledger JSONSecond correction pass, against the September 22 correction-checkpoint review. Errata scanning now fails closed with a recorded coverage inventory and rulings bound to reviewed bytes; release-proof validation now verifies table, vintage, geography, numeric cells and served-URL provenance, keeping unverifiable cases indeterminate; a versioned capture carries the retained insurance retrieval window and per-measure generation times, so no measure borrows another check’s timestamp. Previously: Selected-state explanations now follow the selected jurisdiction instead of a hard-coded New York; Census retrieval provenance is carried from the retained source bytes and kept separate from validation and calculation, with an unretained retrieval declared rather than substituted; confidence intervals are validated as the clipped estimate plus or minus the margin of error rather than merely bracketing it. Data, periods and geography are unchanged from edition 05. Previously: added two reviewed ACS 1-Year 2024 measures for all 50 states and D.C.: rental housing burden (B25070) and uninsured under 65 (S2701). Every percentage is recomputed from its own published numerator and denominator; each carries a 90% margin of error, using the published DP04_0136M denominator margin verified against B25070 for all 51 jurisdictions. July 2026 unemployment and 2024 price levels retained unchanged with their own periods. 204 observations; 153 slots unconnected; no grades or automatic updates.
Download this fixed editionCorrection pass against the September 22 acceptance review. Selected-state explanations now follow the selected jurisdiction instead of a hard-coded New York; Census retrieval provenance is carried from the retained source bytes and kept separate from validation and calculation, with an unretained retrieval declared rather than substituted; confidence intervals are validated as the clipped estimate plus or minus the margin of error rather than merely bracketing it. Data, periods and geography are unchanged from edition 05. Previously: added two reviewed ACS 1-Year 2024 measures for all 50 states and D.C.: rental housing burden (B25070) and uninsured under 65 (S2701). Every percentage is recomputed from its own published numerator and denominator; each carries a 90% margin of error, using the published DP04_0136M denominator margin verified against B25070 for all 51 jurisdictions. July 2026 unemployment and 2024 price levels retained unchanged with their own periods. 204 observations; 153 slots unconnected; no grades or automatic updates.
Download this fixed editionAdded two reviewed ACS 1-Year 2024 measures for all 50 states and D.C.: rental housing burden (B25070) and uninsured under 65 (S2701). Every percentage is recomputed from its own published numerator and denominator; each carries a 90% margin of error, using the published DP04_0136M denominator margin verified against B25070 for all 51 jurisdictions. July 2026 unemployment and 2024 price levels retained unchanged with their own periods. 204 observations; 153 slots unconnected; no grades or automatic updates.
Download this fixed editionAdded 2024 all-items Regional Price Parities for all 50 states and D.C.; all 51 API values reconcile to the official BEA download. July 2026 unemployment retained unchanged. 102 observations; 255 slots unconnected; no grades or automatic updates.
Download this fixed editionExtended checked unemployment to all 50 states and D.C.: July 2026, seasonally adjusted and preliminary. All 51 values reconciled to the official BLS table. Other 306 state/metric slots remain unconnected; no grade assigned.
Download this fixed editionAdded New York July 2026 unemployment, 4.4% preliminary and seasonally adjusted. All other state/metric slots remain unconnected; no grade assigned.
Download this fixed editionGeography and source-defined availability inventory added. No measured values or grades loaded. Earlier newsroom provenance preserved.
Download this fixed editionMap provenance
Navigation uses Census-derived 2017 cartographic boundaries packaged in us-atlas 3.0.1 and projected for this page. Alaska and Hawaii are inset; shading shows selection only. Territories are not included. D.C. is also available in the text selector.
Accuracy & reuse
Civitally shows a selected slice of public information, not a complete assessment of a place or person. Agencies may revise their records. Source links, dates, limitations and corrections matter more than a broad claim of accuracy.
Downloads make our methods inspectable. Linked source material remains subject to its own terms; availability is not blanket permission to republish third-party content.