FRC Research · Media

Press kit

Everything on this page may be quoted with attribution to FinanceRateCalc. Evidence packs with dated screenshots are available for every claim.

Boilerplate

FinanceRateCalc is an independent, one-person research institution computing FHA mortgage-denial statistics from the complete federal HMDA record (1.2 million credit decisions in 2025). Founded by independent researcher Ziya Yetiş, it publishes open data (CC BY 4.0), a peer methodology paper on SSRN, and the Mortgage AI Accuracy Index — an independent benchmark of what AI systems tell consumers about mortgage denial. No lender or AI vendor funds or previews its work.

Findings on the record

22.1% — the 2025 national FHA denial rate on decisioned applications; most of the web still cites 2023's 13.6%.
1.8% to 78.7% — the span of denial rates across major FHA lenders in the same federal program.
38% — the share of explainable denial variation attributable to lender identity, not the applicant's file (The Door Effect, 859,090 decisions).
73.7 points — the widest intra-metro lender spread (Cleveland, OH).
7 of 8 years — how often the softest door was the same lender, 2018–2025; 80.8% of panel variance sits between lenders (Persistent Doors, SSRN 7341481).
First AI citations measured — organic citation count in our frozen benchmark moved 0→4 in August 2026; Microsoft telemetry independently records 7 Copilot citations (Accuracy Index).

Chart — free to reuse (CC BY 4.0)

Bump chart of FHA denial-rate ranks 2018-2025 showing a nearly frozen lender hierarchy

Download SVG · attribution: FinanceRateCalc, CFPB HMDA 2018–2025

Quotable lines

“Which door you knock on moves your odds more than any single thing in your file that the public record can see.”
“One denial is one door's answer — not the market's answer.”
“Strict stays strict, lenient stays lenient: the hierarchy barely moved through a boom and a rate shock.”

What the data does NOT show

HMDA contains no credit scores. All figures are historical aggregates: consistent with lender overlays, channel mix, documentation practices and unobserved applicant risk. They do not establish misconduct or discrimination, and they are not predictions for any individual application. Please carry these boundaries with the numbers — we do.

Interviews & data requests

Ziya Yetiş is available for interviews (English/Turkish) on FHA denial patterns, lender variation, and AI accuracy in consumer finance. Custom data cuts for stories are free for working journalists. [email protected]

Citation formats

News: "…according to FinanceRateCalc, an independent analysis of federal HMDA data."
Academic: FinanceRateCalc (2026). A Public Benchmark for Consumer Mortgage AI Accuracy. SSRN 7156938.
Yetiş, Z. (2026). The Door Effect. SSRN 7309319, doi:10.2139/ssrn.7309319.
Yetiş, Z. (2026). Persistent Doors. SSRN 7341481, doi:10.2139/ssrn.7341481.