Data sources

Public & licensed inputs, plus a live texture layer.

The forecast model runs on public and licensed data downloaded as published research files and public APIs. A separate Bright Data layer adds a current-state concession read where no clean public API exists. The model forecasts rent growth from the inputs; the live layer is texture, not a model input.

Model inputs
  • Rent (target): Zillow Observed Rent Index (ZORI), monthly.
  • Labor: BLS metro employment and state JOLTS quits/hires (via FRED/BLS).
  • Job postings: Indeed Hiring Lab postings index (via FRED).
  • Housing flow: Zillow Research for-sale inventory, new listings, and median days-to-pending — downloaded published CSVs, RegionID-matched, ~2018+. Used as internal model inputs, not redistributed.
Live concession layer · Bright Data

A live concession read from public rental-listing data, collected via Bright Data Web Unlocker and structured into a per-metro concession share. The share is page-depth de-biased (deeper organic pages pull it toward the true market rate, correcting first-page promotional inflation). This is current-state texture, not a forecast input or a skill number. As of 2026-06-04.

Most concessions
Phoenix, AZ82%
Dallas, TX79%
Houston, TX78%
Charlotte, NC75%
Washington, DC70%
Most constrained
San Francisco, CA9%
Detroit, MI15%
Pittsburgh, PA21%
Chicago, IL24%
New York, NY30%
Inventory

Every snapshot the site and the agent read.

Forecast-skill harness outputPre-registered rent-growth forecast vs fitted-AR + drift; dual-OOS, DM-HAC, BH-FDR, bootstrap CIs58 KB
Housing-flow inputsZillow Research CSVs (for-sale inventory · new listings · days-to-pending), downloaded, RegionID-matched3.6 MB
Job-postings indexIndeed Hiring Lab postings index via FRED, monthly904 KB
EmploymentBLS metro employment via FRED, monthly1.6 MB
JOLTS (quits/hires)State JOLTS via BLS, fanned to metros, monthly10.8 MB
Rent (ZORI)Zillow Observed Rent Index, monthly (forecast target)1.9 MB
Live concession layerBright Data Web Unlocker → structured extraction; page-depth de-biased concession share4.0 MB

Universe: 108 metros with full input coverage. All forecast numbers are reproducible from the committed harness output.

Built on public & licensed data — FRED · BLS · Zillow Research (rent, housing-flow) · Indeed Hiring Lab · with a Bright Data live concession layer from public listing data.
Forecast-skill results are pre-registered, out-of-sample, and reproducible from the committed harness.

Disclaimer. HomeStar is provided for informational and educational purposes only. Nothing on this site constitutes investment, financial, legal, or tax advice, or a recommendation, offer, or solicitation to buy, sell, or hold any security, property, or investment, or to engage any advisory service. The forecasts, signals, and data presented are model estimates subject to the limitations described herein, including a limited historical window and untested performance across market regimes. Backtested and out-of-sample results do not guarantee future performance. The information is provided "as is," without warranty of any kind as to accuracy, completeness, or fitness for any purpose. Do not rely on this site as the basis for any investment decision; consult your own qualified advisors.

This product uses the FRED® API but is not endorsed or certified by the Federal Reserve Bank of St. Louis. Job-postings data is derived from the Indeed Hiring Lab, used under CC BY 4.0.