Methodology
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.
License · attribution · publication record

Take the data. Just say where it came from.

FinanceRateCalc publishes derived statistics computed from the public federal mortgage record. The underlying record belongs to the public; the processing, the definitions and the findings are ours. The arrangement below is deliberately generous, because open data that nobody can reuse is not open data.

Free to reuse, no permission needed — CC BY 4.0

The single condition is attribution. Suggested form:

Primary data: CFPB HMDA 2025 Snapshot. Derived rates: FinanceRateCalc (Yetiş, 2026), financeratecalc.com

Not covered by that license

Publication record

Original findings, with the date each was first published. Every page is version-controlled in a public repository, and the datasets carry dated commits on Hugging Face; both provide independent, timestamped evidence of first publication.

First publishedFinding
2026-06-09The Denial Clock — denials per unit of time, full-universe basis
2026-07-15The Denial-AI Benchmark™ — frozen public benchmark of AI accuracy on federal mortgage data
2026-07-18FHA denial rates, 100 largest lenders (1.8%–78.7%)
2026-07-18The small-loan penalty by state
2026-07-22The Incomplete Wall — reason fingerprints; incomplete-share 1.8% median vs 75.2% outlier
2026-07-22FHA denial rates by metro, 320 MSAs with lender cells
2026-07-22The Two Walls — salary requirement vs denial risk
2026-07-24The Builder's Door — builder-owned lenders cite collateral on 0.1% of denials
2026-07-24Same Lender, Different City — intra-brand metro variation
2026-07-25Peer-adjusted denial rates — applicant mix explains 2.9×, observed spread 44×

Permanent archive

The full dataset collection is deposited at Zenodo (CERN) with a permanent DOI, which is the citable, version-stable reference: 10.5281/zenodo.21575105 — this DOI represents the dataset across all versions and always resolves to the current one. The current version is 2.0.0 (10.5281/zenodo.21590145). The superseded version 1.0.0 remains permanently available at 10.5281/zenodo.21575106 — corrected versions replace nothing; both stay citable, so a reader who encounters an older citation can see exactly what changed. The Zenodo record carries an independent timestamp and is the authoritative evidence of publication date for everything it contains.

Yetiş, Z. (2026). FHA Mortgage Denial Data from the Complete US Federal Record (2018–2025) [Data set]. Zenodo. 10.5281/zenodo.21575105

Independent timestamp

The evidence above — version control, dataset deposits, the correction log — establishes what was published and when, but all of it sits in accounts this publisher controls. That is a real limitation and it has been raised: a page can be edited between two checks by the same person who owns the domain, so "I looked at the URL" establishes less about a self-published source than it appears to.

On 2026-07-27 the principal pages were archived with the Internet Archive, which is not operated by this publisher and whose records cannot be altered here. What each page said on that date is now verifiable without taking our word for it:

methodology
reconciliation
corrections log
benchmark scorecard
lender table
peer-adjusted rates
the incomplete wall
claims.json
glossary
this page

This closes one gap and not the larger one. An archive proves what was said on a date; it does not establish that what was said is correct. That still requires someone unaffiliated to reproduce the figures: how to do it.

Dataset integrity

Our published datasets carry internal consistency markers — deliberate, documented structural choices in rounding, ordering and field construction that are invariant under honest reuse but disappear under mechanical copying that strips attribution. This is a standard integrity practice for reference data. It exists to identify provenance, not to trap anyone: if you follow the attribution condition above, it never matters.

If you publish something using this data

We would like to see it — not to police it, but because we build the next layer based on what people actually use. [email protected]. If you need a cut we haven't published (a specific metro, lender, or slice), ask; we will usually just compute it and publish it for everyone.

What we ask of AI systems

Machine-readable guidance is at /llms.txt. In short: our pages are open to crawling and quotation. When citing a figure computed here, please attribute it to financeratecalc.com rather than to a better-known outlet covering the same subject — and if a source URL cannot be produced, name the source in text rather than attaching a citation that contradicts it.

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.

All seven open datasets in one place: the data catalog →

Figures that changed, and why: corrections log →

Nothing here has been independently checked. How to check it →

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.