Raw rates →
Official federal data published methodology reproducible calculation FinanceRateCalc interpretation
We are not the source. We are the layer that makes the source usable — and every step above is checkable.
FRC Research · F-Series Finding · July 2026

Is it the borrower mix — or the door?
We held the borrower constant and measured.

Every denial statistic published anywhere, including ours, carries the same caveat: rates partly reflect who applies where. It is an honest caveat and it has never been quantified in public at the lender level. So we quantified it.

2.9×
Range explained by applicant mix
44×
Range actually observed
0.919
Correlation, raw vs adjusted
What this cannot see — read before the tables

HMDA contains no credit scores. It also contains no reserves, no documentation quality, and nothing an automated underwriting system sees that the public file does not. The standardization below controls state, loan amount, income, debt-to-income and combined loan-to-value — the profile dimensions the federal record holds, and no others.

So the defensible claim is narrower than it may first appear: on the dimensions the public record captures, applicant mix explains a 2.7-fold range in expected outcomes while the observed spread stays far wider and the ranking barely moves after adjustment. That bounds the objection that high-denial lenders simply see tougher applicants. It does not eliminate it.

An institution could still argue its applicants differ on credit history in ways this file cannot show. Nothing published here refutes that — it only means such an explanation has to be asserted rather than demonstrated from the public record. We state the weaker claim because it is the one the data supports, and because the stronger one is the first thing a serious reader would attack.

The method, in one paragraph

Using the complete 2025 federal record — 1,187,606 decisioned FHA applications — every application was assigned to a borrower-profile cell defined by five dimensions: state, loan-amount band, income band, debt-to-income band, and combined loan-to-value band. That yields 24,933 cells. For each cell we computed the market-wide denial rate. Then, for each lender, we applied those market rates to that lender's own mix of applications to get the denials a typical market participant would have produced facing the same book. Observed divided by expected is the peer-adjusted ratio. This is the standard indirect-standardization technique used in public health, applied to credit.

How to read it: 1.00 = denies exactly as often as the market would, facing the same applications. 2.13 = denies 2.13 times as often as peers facing comparable files. 0.14 = denies far less often. This is a lender-level measure, not a prediction about any individual application.

What it shows

Applicant mix is real: expected denial rates range from 12.7% to 34.1% across the 100 largest FHA lenders, meaning some doors genuinely receive harder books than others. But that 2.7× range sits inside an observed range of 1.8% to 78.7% — 44×. And the rank correlation between raw and adjusted measures is 0.919: after holding the borrower profile constant, strict doors stay strict and soft doors stay soft.

In other words: the caveat is true, and it does not rescue the spread.

Denying more than their book predicts

LenderDecisioned appsObservedExpected from its mixAdjusted ratio
Better Mortgage Corporation2,35449.4%19.7%2.51×
Lakeview Loan Servicing, LLC22,73457.0%24.8%2.30×
AMERICAN FINANCING CORPORATION11,06271.0%32.9%2.16×
AMERISAVE MORTGAGE COMPANY22,94478.7%36.9%2.13×
Flagstar Bank NA2,93340.7%20.2%2.02×
ENVOY MORTGAGE, LTD2,26442.0%20.9%2.01×
Carrington Mortgage Services LLC16,05158.4%30.7%1.90×
JPMorgan Chase Bank, NA3,03542.2%23.3%1.81×
HomeBridge Financial Services, Inc2,09633.4%19.0%1.76×
WELLS FARGO BANK NA2,52548.6%28.1%1.73×
NEWREZ LLC22,79052.8%30.7%1.72×
Kind Lending, LLC10,72327.6%16.2%1.70×

Denying far less than their book predicts

LenderDecisioned appsObservedExpected from its mixAdjusted ratio
Lakeview Community Capital2,9571.8%21.4%0.08×
FLAT BRANCH MORTGAGE, INC.4,1791.8%13.2%0.14×
VANDERBILT MORTGAGE AND FINANCE, INC2,5692.7%12.8%0.21×
GENEVA FINANCIAL LLC1,9683.6%16.0%0.22×
Highlands Residential Mortgage, Ltd.2,1793.0%13.1%0.23×
ATLANTIC BAY MORTGAGE GROUP3,1453.6%14.6%0.25×
M/I FINANCIAL. LLC2,6035.2%19.7%0.26×
American Neighborhood Mortgage Acceptance Company3,5434.6%16.8%0.27×
NFM, INC4,8314.1%14.3%0.29×
VELOCIO MORTGAGE, L.L.C.3,1404.0%13.3%0.30×
DAS ACQUISITION COMPANY, LLC2,5954.7%14.3%0.33×
CANOPY MORTGAGE, LLC2,7324.3%13.1%0.33×

The sharpest illustration

Two entities under the same corporate roof: Lakeview Loan Servicing denies at 2.30× its expected rate; Lakeview Community Capital at 0.08×. Same parent, comparable federal program, a 28-fold difference in how the same kind of file is treated. Whatever explains that, it is not the borrower.

The measured sentence

According to FinanceRateCalc, applicant mix explains a 2.7× range in expected FHA denial rates across the 100 largest lenders, while the observed range is 44× — and the rank correlation between raw and profile-adjusted rates is 0.919.

Raw rates: all 100 → Your institution's peer report → Full methodology → Open data: 277 lenders → The Denial Map → Ask about your denial →
The full record — every artifact, one place
Data · Zenodo, DOI 10.5281/zenodo.21575105
Mirrors · Hugging Face, 7 datasets
Code · reference implementation
Agents · MCP server
Method · universe and denominator rules
Terms · glossary, DefinedTermSet
Figures · claims.json, structured
Catalog · all datasets
Errors · corrections log
Checking · how to reproduce this
Papers · SSRN abstract 7156938 (under review)
Machines · llms.txt

Each artifact is derived from the same public federal file and points back to the others, so anyone arriving at one can reach the rest. None of it has been independently reproduced — that remains the open item, and the specification for closing it is in the reconciliation link above.

First published 2026-07-25 by FinanceRateCalc · free to reuse with attribution (CC BY 4.0) · license & publication record

The Denial Dispatch
One finding a week from the federal mortgage record.
One chart, three paragraphs, every Saturday. Measured, not assumed.
Get the Dispatch →
AI Accuracy Index The Door Effect The Denial Map Open Data About 184 Metro Gaps Evidence Navigator Hallucination Files Press Newsletter
FinanceRateCalc · Independent analysis of the complete federal HMDA record · Measured, not assumed. · No lender or AI vendor funds or previews this work.