Every dataset the Oracle holds, where it came from, and the standard every public sentence has to pass. Judge us by our sources: all public, all named, all datable.
Three rules govern everything published here. No adjectives of judgment — delete the word; if the sentence only loses emphasis, it never ships. Both columns, always — every analysis showing a benefit shows the corresponding cost at the same depth, and the reverse; balance is measured, not felt. Every number carries its source — publisher, table, and vintage, in the sentence, so screenshots keep the citation. Claims that are not ours are labeled: [CLAIM] for company or campaign statements, [ADV] for advocacy-commissioned research, [ARITH] for our own arithmetic with inputs stated. Forecasts ship with every assumption printed, labeled projection-not-prediction.
| Dataset | Publisher & table | What it provides |
|---|---|---|
| Personal income & transfers | BEA CAINC4 (Feb 2026 vintage), lines 10 & 47 | Income and government-transfer share, every county, 1969–2024 |
| GDP by area | BEA CAGDP2, 2001–2024 | County and metro GDP |
| Population | Census Bureau estimates (vintage 2024) + decennial | All counties and metro/micro areas, 2000–2024 |
| Demographics & housing | ACS 2020–2024 5-year (B15003, B17001, B19013, B19019, B20017, B23003, B23008, B23025, B25014, B25034, B25047, B25052, B25077) | Education, poverty, income, earnings, parents' labor force, housing age and condition |
| Employment & wages | BLS QCEW annual averages, per-area files | Establishments, employment, wages by area |
| Business starts & exits | Census Business Dynamics Statistics, 2023 release (county series 1978–2023) | Establishment births/deaths, job creation/destruction — from payroll-tax records, not surveys |
| Industry mix | Census County Business Patterns, 2013 & 2023 county files | Establishment counts by industry, every county — the arrivals-and-departures tables |
| Geography | OMB 2023 county–CBSA delineations; Census gazetteer; 2020 ZCTA–county crosswalk | All joins between counties, metros, and ZIP codes |
| State | Source | What it provides |
|---|---|---|
| Texas | HHSC Child Care Licensing (data.texas.gov, updated daily); TWC facility report 2024 Q2 (Texas Rising Star); TWC WARN notices | Licensed capacity, subsidy acceptance, inspection deficiencies, quality ratings, mass-layoff notices — 241 counties |
| Oklahoma | OKDHS Child Care Locator, per-provider public records (pulled 2026-08-24, 99.0% of 2,682 providers) | Capacity, 1–5 star ratings, subsidy acceptance, monitoring compliance — 77 counties |
| Kansas | KDHE licensing data via the Child Care Aware of Kansas county supply file (vintage: 2023–24 cycle; stated wherever used) | Licensed capacity and actual enrollment — 105 counties |
| New Mexico | ECECD Annual Outcomes Report FY25 | Children served by county; capacity published statewide only |
| Arkansas | No bulk file published; official search is an interactive application | Documented as a coverage gap, not silently omitted |
Weekly "most-discussed topics" per region come from Google News aggregation of each area's own outlets (last 7 days), scored against a fixed neutral taxonomy of eleven topics. This measures coverage volume, not merit. We do not collect from social platforms, never log in anywhere, and never post, reply, or seed discussions.
Care-needing children = children under 6 with every resident parent in the labor force (B23008 lines 4+10+13). Slots per care-needing child divides licensed capacity by that count — capacity is a licensed maximum, not enrollment, and some slots serve school-age children; the biases run in opposite directions and are stated wherever the ratio appears. Multi-county and multi-year rates are pooled sums, not averages of published annual rates. County-to-metro aggregation uses the OMB 2023 map. Where a measure is unavailable (e.g., BLS county unemployment files blocked to non-browser clients), the substitute (ACS 5-year) is labeled in place.
The correction policy and permanent log live on the corrections page. Every email is read by a person; an error report gets a correction at the top of the next issue and a permanent entry in the log.