How The Mortgage Record is built
Every number on this site comes from the federal government’s own mortgage records — not surveys, not estimates, not lender marketing. Here is exactly what we use and how we compute it.
Unfamiliar with a term? See the mortgage & HMDA glossary.
The data source
Under the Home Mortgage Disclosure Act (HMDA), nearly every mortgage lender in the United States must report each loan application to federal regulators — the loan, the outcome, and, for denials, the reason. The Consumer Financial Protection Bureau (CFPB) publishes this loan-level data every year. The Mortgage Record is built on the HMDA Snapshot National Loan-Level Dataset for 2018 through 2025, filtered to New Jersey.
That is roughly 3.6 million New Jersey application records across eight years. We reproduce the CFPB’s own published aggregate counts exactly as a correctness check before publishing anything. You can download our analyzed datasets as CSV, free under CC BY 4.0.
What we include
Our headline figures describe first-lien, closed-end mortgages on 1–4 family homes — the ordinary loans people use to buy or refinance a house. We exclude commercial-purpose loans, reverse mortgages, open-end lines of credit (HELOCs), and manufactured-home-only records where noted, so the numbers reflect the kind of loan a typical buyer is actually shopping for.
What’s in each record
Every application is a row with dozens of fields. These are the ones that matter for the questions The Mortgage Record answers:
| Category | Key fields | What it tells you |
|---|---|---|
| Outcome | action taken · denial reason 1–4 | Approved, denied, withdrawn, or purchased — and, on denials, the reason(s) the lender recorded. |
| Borrower finances | income · debt-to-income ratio · loan-to-value ratio | The three financial levers lenders weigh most, after credit. Income is in thousands; DTI and LTV come as ranges or values. |
| The loan | loan type (Conventional / FHA / VA / USDA) · purpose · amount · lien · occupancy · units · term | What kind of loan was sought, for what, and on what property. |
| Property | property value · construction method | Value is rounded to the nearest $10,000 in the public data. |
| Pricing | interest rate · rate spread · total loan costs · discount points · lender credits | What the loan actually cost — and how far its rate sat above the market’s low-risk benchmark. |
| Underwriting | automated underwriting result (AUS 1–5) · purchaser type | What the automated system returned (Approve / Refer / etc.) and whether the loan was later sold. |
| Demographics | race · ethnicity · sex · age (applicant & co-applicant) | Reported in HMDA for every application. We do not currently publish lender-level breakdowns by these fields. |
| Credit | applicant credit score type | Only the scoring model used (FICO, VantageScore, …). The actual score is redacted — see below. |
| Where & who | state · county · census tract · metro area · lender (LEI → name) | Geography down to the census tract, and the identity of the reporting lender. |
A note on credit scores
The single most common question about this data is: does it include credit scores? No — and it can’t. HMDA records only a code for which scoring model the lender used (Equifax Beacon, Experian/Fair Isaac, FICO Classic, VantageScore, and so on). The actual numeric score is deliberately removed by the CFPB before the data is published, to protect borrower privacy. No public HMDA field contains a FICO or VantageScore number.
What that means for how you should read this site:
- These approval rates describe a lender’s overall behavior across everyone who applied — not a personalized prediction for your credit profile. A lender’s 92% approval rate does not mean you have a 92% chance; your credit, which we can’t see, matters enormously.
- Where credit would normally do the explaining, we lean on the factors the data does contain — income, debt-to-income, and loan-to-value — which lenders weigh alongside credit and which correlate with it. That’s the honest best we can do, and we’d rather tell you than imply a precision we don’t have.
- The denial reasons are the closest signal to credit: when “credit history” dominates a lender’s denials, that tells you something real about how it underwrites.
How “approval rate” is defined
We count only real lender decisions. A lender either originated the loan, approved it (borrower didn’t take it), or denied it:
Approval rate = (originated + approved-but-not-accepted) ÷ (originated + approved + denied)
We deliberately exclude applications that were withdrawn by the applicant, closed for incompleteness, or merely purchased on the secondary market (a lender buying a loan someone else already made). Those aren’t decisions about whether to approve you, and including them — as some analyses do — distorts the picture.
Small samples: how we avoid junk numbers
A lender that made three loans in your county shouldn’t show a headline “100% approval rate.” So:
- We never surface a lender’s rate built on fewer than 25 decisions.
- In the approval tool, if your exact profile is too thin for a lender, we widen to that lender’s county-wide, then state-wide record — and label which basis each number uses.
- Lender rankings are sorted by a sample-size-adjusted approval rate (a Wilson score lower bound), so a lender with 100% of 50 loans can’t leapfrog one with 96% of 3,000.
What the data can’t tell you
Being honest about the limits is part of the point:
- No credit scores (see the section above) — a major approval factor we simply cannot see.
- Loan amounts and property values are rounded in the public data (to the nearest $10,000), and some fields are reported as ranges rather than exact figures.
- Approval rates reflect who applied to each lender, not just how strict the lender is. A lender serving a higher-income area may post higher approval rates for reasons unrelated to how it would treat your application.
- Past behavior is not a promise of credit. Nothing here is an offer, a guarantee, or individualized financial or lending advice.
Independence
The Mortgage Record takes no money from lenders. We are not a lender, a mortgage broker, or a lead generator, and we do not sell your information. We don’t rank lenders by who pays us, because nobody pays us. The calculators are educational tools that run entirely in your browser — we don’t see what you type.
Corrections & updates
HMDA data is released annually; we refresh when the CFPB publishes a new year. Spotted something that looks wrong? Email [email protected].
Publisher: Resultex Media Group LLC. Data: CFPB HMDA Snapshot National Loan-Level Dataset, 2018–2025. This site is independent research and is not affiliated with the CFPB, FFIEC, or any lender.