Data sources

What’s live, how often it refreshes, which vintage you’re looking at, and where it doesn’t reach. Attribution and upstream terms: attributions. Vintages below are as-of 2026-07-09; /api/tracts/chicago carries the exact vintage block for the data it’s actually serving.

Current feeds

Source	Used for	Cadence	Vintage	Coverage
Cook County Assessor (Socrata)	live comp sales + characteristics	per request	rolling 6-yr window	Cook
Cook Assessor reassessments	tax-shock by township	each triennial mailing	latest triad; South triad reassessing 2026	Cook
Cook Clerk tax rates	effective property tax	annual	2023 rates	Cook munis only
Chicago Data Portal	permits (18 mo), licenses (12 mo), 311 water-in-basement (3 yr) + governance (12 mo), violations/vacancy; crime retained offline only	monthly	rolling windows	Chicago only
CFPB HMDA	mortgage volume, loan-size growth, FHA share; applicant income retained offline only	annual	flow 2022→2024; FHA 2023	Cook + Lake
Census LODES	workplace job growth	annual, lagged	2019→2022	Cook + Lake
Census ACS	pre-1980 housing stock (B25034)	annual	latest 5-yr	Cook + Lake
HUD SAFMR	rent anchor, voucher ceilings, S8 arbitrage	annual	FY2026	by ZIP
Zillow ZHVI/ZORI	prices, rents, gross yield	monthly	latest month	ZIP; ZORI ~288 IL ZIPs
RentCast API	licensed active sale listings	daily 13:00 UTC + monthly sweep	same-day	configured Cook + Lake search areas; plan/request limits apply
Redfin Data Center	sale-to-list, DOM, market heat	quarterly	latest quarter	ZIP
CTA ridership	L-station recovery vs 2019	monthly	≥ 2025-04 vs 2019 baseline	~0.75 mi of stations
FEMA NFHL	flood-zone flag	static	current NFHL	Cook + Lake
Lake County GIS	parcels (57k 1970–88 SFHs), flood, schools	static export	as exported	Lake
OSM / OSRM	amenity clusters, commute times	with refresh	OSM-dependent	variable

Known gaps

Gap	Consequence
Lake parcel-sale comp corpus is not yet calibrated for serving	Lake property ARV is suppressed; tax scenarios use the explicitly labeled county-median effective-rate fallback rather than pretending a parcel-specific bill was verified
Chicago-only feeds: permits, licenses, 311, violations/vacancy, distress, ridership	suburban tracts miss those signals entirely and lean on the county-wide feeds
Income/crime diagnostics	retained offline for fairness/data-quality review; stripped from public payloads, scores, risk, recommendations, and lenses
ZORI covers ~288 IL ZIPs, ZIP-blended	a hot block inherits its ZIP median; thin-ZORI areas fall back to SAFMR
`listed_at` falls back to first-seen	true days-on-market arrives with MLS data (P7)
Licensed inventory is not complete MLS coverage	some active listings may be absent until MRED/RESO access is licensed and enabled

Planned, in wiring priority order

DePaul IHS 2–4 unit indicators (quarterly), CCAO open assessment model + sales flat files, Board of Review appeals by PIN, FHFA tract-level HPI, Lake County recorder, OpenFEMA NFIP claims + National Risk Index, CCLBA/ChiBlockBuilder inventory, Fannie/Freddie REO feeds, EPA FRS/LUST, LiDAR/soils/imagery, Census Building Permits Survey, FRED/BLS cost indices, and licensed MRED/RESO inventory after contractual approval. Every adopted feed gets a data contract — schema, cadence, vintage, terms, coverage, freshness SLO, and null-rate SLO — before its signal ships.

Catalog governance

The repository-root DATA_CATALOG.md is the governed source inventory: source and dataset identifiers, cadence, coverage, serving status, loader, quality traps, and PII handling. Licensing decisions and remediation evidence live in COMPLIANCE_REGISTER.md. This consumer-facing table is a reviewed summary; CI and operator review must keep it aligned with those two registers. A feed without a catalog and compliance disposition does not ship.