Rent model
Rent model
Section titled “Rent model”Current (model p0a-2026.07)
Section titled “Current (model p0a-2026.07)”Per-unit market rent is built from HUD voucher ceilings, not hope:
- Anchor: HUD FY2026 Small Area FMR (2BR) for the ZIP.
- Bedroom ladder: map the unit’s bedroom count to HUD-typical ratios vs 2BR — 0BR 0.74, 1BR 0.86, 2BR 1.00, 3BR 1.30, 4BR 1.49. Per-unit beds = total beds ÷ units (default: 2BR).
- Blend: where Zillow ZORI exists for the ZIP, blend 55% SAFMR-ladder / 45% ZORI-ladder — SAFMR lags hot markets. Where SAFMR is missing, fall back to ZORI × ratio (last resort: a $1,500 base).
- Bounds: floor 0.7× and cap 1.6× the SAFMR-implied figure. No fantasy rents in either direction.
Renovated multiplier: 1.5×. Average-to-renovated lifts rent roughly 1.3–1.5×; a full gut of a distressed unit can reach ~1.7×, and that is the ceiling, not the default. The old 1.85× multiplier was retired as fantasy — the engine dropped it in p0-2026.06, and its last copy (the in-map calculator) died in the 2026-07 client/worker parity fix. DSCR and the income-approach ARV both run on these renovated rents.
Section 8 overlay
Section titled “Section 8 overlay”SAFMR (2BR voucher ceiling) minus ZIP market rent = the voucher-arbitrage signal. Positive means the voucher pays more than market. Most of Cook + Lake is negative as-of FY2026; the signal exists to find the green tail.
Limitations
Section titled “Limitations”ZIP-blended throughout — ZORI covers ~288 IL ZIPs, and a hot block inherits its ZIP’s median. No condition awareness: the 1.5× applies uniformly even though the true renovated premium varies by submarket. No per-building observation: this is a model of the ZIP, applied to your address.
Planned (P2)
Section titled “Planned (P2)”A trained rent model on scraped asking rents (Zumper, Apartments.com, Craigslist archives), HUD FMR/SAFMR, CHA payment standards, and ACS priors — all vintage-tagged — over unit mix, size, age/type, condition proxies, and submarket. The renovation premium gets learned from paired observations, expected in the +10–25% range by submarket, deleting the 1.5× constant. Output becomes a rent distribution per unit type, and DSCR becomes identical across every tool because its inputs can no longer differ.