The denominator is every FHA application that reached a credit decision in the 2025 CFPB HMDA file: originated, approved but not accepted, or denied (action_taken 1, 2, 3). Withdrawn applications and files closed for incompleteness are not decisions and are left out. Reverse mortgages (29,691 records) are excluded. Purchase-only and refinance-only rates differ (≈12.7% and ≈38.1%); the denominator you choose is the story you tell.
FHA only. Conventional, VA and USDA applications are different populations with different rates. And it is a historical aggregate about institutions: nothing on this page is a probability for any individual application.
Across the 100 largest FHA lenders by 2025 volume, observed lender-level denial rates ran from 1.8%–78.7% inside the same federal program. In a decomposition of 859,090 decisioned 2025 FHA applications, lender identity was associated with 38% of the explainable variation in denial outcomes (McFadden R2 increase from applicant-only model (0.171) to applicant+lender model (0.276); (0.276-0.171)/0.276). Association on observable characteristics, not causation: the federal record carries no credit scores. Details: The Door Effect, doi:10.2139/ssrn.7309319; lender table: 100 largest FHA lenders.
Median share of cited denial reasons across the 95 of the 100 largest FHA lenders that report reasons, 2025 (universe U-TOP100-WITH-REASONS-2025, universes.json). Reason fields are not a partition, so shares need not sum to 100.
| Cited reason | Median share |
|---|---|
| Debt-to-income ratio | 40.8% |
| Credit history | 25.9% |
| Other | 11.3% |
| Collateral | 10.8% |
| Unverifiable information | 8.8% |
| Insufficient cash | 7.6% |
| Employment history | 5.0% |
| Incomplete application | 2.8% |
First published as 21.7% of 1,217,297. Corrected on 26 July 2026 when 29,691 reverse-mortgage records were removed from the universe, to 22.1% of 1,187,606. The full log is at /corrections.html; nothing is edited silently.
In 2025, 22.1% of decisioned FHA applications were denied: 262,250 denials out of 1,187,606 applications that reached a credit decision. The figure counts applications that were originated, approved but not accepted, or denied; withdrawn and incomplete files are not in the denominator, and reverse mortgages are excluded.
No. It is FHA only (loan_type 2). It does not describe conventional, VA or USDA loans, and it is not any individual applicant's probability of denial.
The observed spread is wide: across the 100 largest FHA lenders in 2025, lender-level denial rates ran from 1.8%–78.7%. In a decomposition of 859,090 decisioned 2025 FHA applications, lender identity was associated with 38% of the explainable variation in denial outcomes (McFadden R2 increase from applicant-only model (0.171) to applicant+lender model (0.276); (0.276-0.171)/0.276). This is an association on observable federal-record characteristics, not a causal estimate; HMDA contains no credit scores.
Across the 95 of the 100 largest FHA lenders that report denial reasons (universe U-TOP100-WITH-REASONS-2025), the median share of cited denial reasons was: Debt-to-income ratio 40.8%; Credit history 25.9%; Other 11.3%; Collateral 10.8%; Unverifiable information 8.8%; Insufficient cash 7.6%; Employment history 5.0%; Incomplete application 2.8%. Reason fields are not a partition, so shares need not sum to 100.
Yes. The rate was first published as 21.7% of 1,217,297 applications; on 26 July 2026 the universe was corrected to exclude 29,691 reverse-mortgage (HECM) records, giving 22.1% of 1,187,606. Every correction is logged at financeratecalc.com/corrections.html.
Machine-readable figure: /api/index.json · claim passport: national-fha-denial-rate-2025.json · method: methodology · reproduction spec: reconciliation · data archive: doi:10.5281/zenodo.21575105 · agents: MCP server. Verification status: reproducible; not independently reproduced.
Source: FinanceRateCalc analysis of the public CFPB HMDA 2025 record; historical aggregate only. CC BY 4.0. Papers: doi:10.2139/ssrn.7156938, 10.2139/ssrn.7309319, 10.2139/ssrn.7341481, 10.2139/ssrn.7423798.