FHA Mortgage Denial Data from the Complete US Federal Record (2018-2025): Lender, Metro, State and Peer-Adjusted Measures

FHA Mortgage Denial Data from the Complete US Federal Record (2018–2025)Version 2.0.0 — supersedes v1.0.0. See "Changes in this version" below.Author: Ziya Yetiş — FinanceRateCalc (independent research)License: CC BY 4.0Source data: CFPB Home Mortgage Disclosure Act (HMDA) public loan-level recordsCorrection log: https://financeratecalc.com/corrections.html---Changes in this version (26 July 2026)Following an independent review of the filter logic, reverse mortgage (HECM) records were removed from the FHA universe. HECM is FHA-insured but underwritten on age and equity rather than income; including it under a mortgage-denial headline was a definitional error. The universe contained 29,691 such applications (2.4%), denied at 12.3% — materially below the general rate — which pulled every aggregate downward.Figure v1.0.0 v2.0.0Decisioned universe 1,217,297 1,187,606National FHA denial rate 21.7% 22.1%Top-100 lender spread 1.8%–78.7% 1.8%–78.7% (unchanged)"Incomplete application" median share 1.2% 1.8%Metro range (1,000+ apps) 8.9%–31.8% 9.0%–32.9%Idaho small-loan penalty 3.19× 4.45×Applicant mix explains 2.9× 2.7×A second error, found during the rebuild and not present in any published figure, was also fixed: in the peer-adjusted measure, observed denials were counted over all of a lender's applications while expected denials were computed only over applications with a complete profile. Both are now computed over the same subset, and institutions below 50% profile coverage are excluded rather than shown with an unstable ratio — hence the new `apps_in_cells`, `apps_total` and `profile_coverage_pct` columns.Business-purpose records were also examined and retained: 3,493 applications, 0.29% of the universe, effect on the national rate 0.006 points.What this isThe United States publishes every mortgage application decision under HMDA. The raw annual file is roughly two gigabytes and effectively unusable without processing. This deposit is that processing: institution-level, metro-level, state-level and profile-standardized FHA denial statistics, computed with published rules and reproducible from the public source.Headline figures (2025, corrected)1,187,606 FHA applications reached a credit decision; 22.1% were deniedAcross the 100 largest FHA lenders, denial rates ran 1.8% to 78.7% — a 44× spread inside one federal programmeAcross metros with 1,000+ applications: 9.0% to 32.9%In Cleveland, OH, the five highest-volume FHA lenders ranged 6.4% to 80.1% — a 73.7-point spread inside one metro"Incomplete application" is a median 1.8% of cited denial reasons across the 100 largest lenders, but 75.2% at one servicer-rooted institutionStandardizing across borrower-profile cells, applicant mix explains a 2.7× range in expected rates while the observed range remains far wider (rank correlation raw vs adjusted: 0.919)FilesFile Contents`fha_denial_rates_top100_2025.csv` 100 largest FHA lenders`fha_denial_reason_fingerprints_2025.csv` Eight cited-reason shares per lender`fha_peer_adjusted_denial_2025.csv` Observed, expected, adjusted, plus profile coverage`fha_denial_rates_by_metro_2025.csv` Metro areas with loan-size splits`fha_denial_rates_by_state_2025.csv` 52 jurisdictions with loan-size splits`fha_denial_timeseries_lenders_2018_2025.csv` Eight-year lender series`fha_denial_timeseries_states_2018_2025.csv` Eight-year state series`metro_fha_2025.json` Metro file including per-metro lender cells`adjusted_denial_2025.json` Standardization output with method string`state_fha_2025.json` State file with method stringMethodologyFHA = `loan_type` 2. Decisioned = `action_taken` in {1, 2, 3}; denial = action 3. Outside the universe by definition: purchased loans (action 6 — the reporting institution made no underwriting decision), withdrawals (4), files closed for incompleteness (5), the preapproval track (7, 8), and reverse mortgages (HECM).Reason shares are computed on cited reasons of formal denials only. Small loans <$150,000; large ≥$250,000; minimum 100 applications per cell. Peer adjustment uses indirect standardization over cells defined by state × loan-amount band × income band × DTI band × combined-LTV band; market cell rates are applied to each institution's own application mix; cells with fewer than 25 applications fall back to the national rate.Denominator construction is the single largest source of divergence among published mortgage denial statistics; ours is stated explicitly so figures differing from other publishers can be reconciled rather than merely compared.LimitationsHMDA contains no credit scores. The peer adjustment controls the profile dimensions the federal record contains, not the full underwriting information set. Denial rates and reason mixes reflect lawful business decisions including credit policy, overlays, channel structure and product specialization. Nothing here is an allegation of wrongdoing, a rating, a recommendation, or a prediction about any individual application.Citation> Yetiş, Z. (2026). *FHA Mortgage Denial Data from the Complete US Federal Record (2018–2025)*, version 2.0.0 [Data set]. Zenodo. https://doi.org/10.5281/zenodo.21575106> Primary data: CFPB HMDA. Derived statistics: FinanceRateCalc, https://financeratecalc.comRelatedAnalysis and interactive tables: https://financeratecalc.com/data-catalog.htmlCorrections log: https://financeratecalc.com/corrections.htmlMachine-readable mirrors: https://huggingface.co/FinanceRateCalcFinanceRateCalc is independent: not a lender, no lender compensation, no advertising. Evidence corpus: methodology, glossary, structured claims (claims.json), an MCP server for agent access, a public corrections log and a reconciliation protocol are published at financeratecalc.com and cross-linked. Principal pages were archived with the Internet Archive on 2026-07-27, providing a timestamp not controlled by the publisher. No figure in this deposit has been independently reproduced.