kvlak limn

Money & Accountability · CFPB HMDA · A fair-lending audit in eleven plates

Who Lends to Whom

Federal law makes the mortgage the best-documented yes-or-no decision in American life: since 1975, nearly every lender has had to put every application and every outcome on the public record. Read a year of that ledger and one pattern holds in nearly every state and metro - who gets the loan, and who hears no.

~15M mortgage applications a year
10.8 pp Black - White denial-rate gap
16.6% highest group denial rate
Illustrative CFPB HMDA · 2018-2024

The denial map

Plate 01Overall denial rate by state

Every state, shaded by the share of mortgage applications its lenders turned down. Darker is a harder market to get a yes in. The deepest grades pool across the Deep South - Mississippi, Louisiana, Georgia - while the Mountain West, the Northwest, and the upper Midwest stay palest. Geography is the first cut of the ledger; the sharper cut comes two plates down.

Alabama: 14.9% denied Alaska: 11.2% denied Arizona: 12.1% denied Colorado: 9.6% denied Florida: 14.2% denied Georgia: 15.1% denied Indiana: 12.9% denied Kansas: 11.4% denied Maine: 10.8% denied Massachusetts: 9.8% denied Minnesota: 9.3% denied New Jersey: 12.9% denied North Carolina: 13.3% denied North Dakota: 9.7% denied Oklahoma: 13.9% denied Pennsylvania: 12.1% denied South Dakota: 9.9% denied Texas: 15.4% denied Wyoming: 11.0% denied Connecticut: 11.8% denied Missouri: 12.2% denied West Virginia: 13.5% denied Illinois: 12.6% denied New Mexico: 13.4% denied Arkansas: 13.8% denied California: 10.4% denied Delaware: 12.3% denied District of Columbia: 10.9% denied Hawaii: 11.7% denied Iowa: 10.1% denied Kentucky: 13.2% denied Maryland: 12.7% denied Michigan: 12.4% denied Mississippi: 16.8% denied Montana: 10.6% denied New Hampshire: 9.5% denied New York: 13.6% denied Ohio: 12.7% denied Oregon: 9.4% denied Tennessee: 13.7% denied Utah: 9.2% denied Virginia: 12.0% denied Washington: 9.1% denied Wisconsin: 10.4% denied Nebraska: 10.3% denied South Carolina: 14.6% denied Idaho: 9.9% denied Nevada: 13.1% denied Vermont: 10.2% denied Louisiana: 16.3% denied Rhode Island: 11.9% denied
Denial rate
  • under 10.3%
  • 10.3% - 12.1%
  • 12.1% - 13.3%
  • over 13.3%
Illustrative
Denial rate by state - the extremes (table view is the source of truth)
Hardest to get a yesDenied EasiestDenied
Mississippi16.8% Washington9.1%
Louisiana16.3% Utah9.2%
Texas15.4% Minnesota9.3%
Georgia15.1% Oregon9.4%
Alabama14.9% New Hampshire9.5%

Fifty-one markets, one axis

Plate 02One dot per state

Flatten the map and the country's spread is a factor of nearly two: from Washington at 9.1% to Mississippi at 16.8%. Half the states sit within about a point of the 12.1% median - the tail to the right is the story.

9% 11% 13% 15% 17% US average 12.5% Minnesota: 9.3% denied Oregon: 9.4% denied Utah: 9.2% denied Washington: 9.1% denied Colorado: 9.6% denied Idaho: 9.9% denied Massachusetts: 9.8% denied New Hampshire: 9.5% denied North Dakota: 9.7% denied South Dakota: 9.9% denied California: 10.4% denied Iowa: 10.1% denied Nebraska: 10.3% denied Vermont: 10.2% denied Wisconsin: 10.4% denied District of Columbia: 10.9% denied Maine: 10.8% denied Montana: 10.6% denied Alaska: 11.2% denied Kansas: 11.4% denied Wyoming: 11.0% denied Connecticut: 11.8% denied Hawaii: 11.7% denied Rhode Island: 11.9% denied Arizona: 12.1% denied Delaware: 12.3% denied Michigan: 12.4% denied Missouri: 12.2% denied Pennsylvania: 12.1% denied Virginia: 12.0% denied Illinois: 12.6% denied Indiana: 12.9% denied Maryland: 12.7% denied New Jersey: 12.9% denied Ohio: 12.7% denied Kentucky: 13.2% denied Nevada: 13.1% denied New Mexico: 13.4% denied North Carolina: 13.3% denied Arkansas: 13.8% denied New York: 13.6% denied Oklahoma: 13.9% denied Tennessee: 13.7% denied West Virginia: 13.5% denied Florida: 14.2% denied Alabama: 14.9% denied South Carolina: 14.6% denied Georgia: 15.1% denied Texas: 15.4% denied Louisiana: 16.3% denied Mississippi: 16.8% denied WA 9.1 MS 16.8
Plate 02 · Illustrative Each dot is a state, shaded by the map's quartile grade. Quartile breaks fall at 10.3%, 12.1%, and 13.3%; the application-weighted US average is 12.5%. Stand-in figures - the extremes table above is the source-of-truth read.

The gap, one bar per group

Plate 03Denied · home-purchase slice

Now hold the loan constant. On the same product - conventional, first-lien, owner-occupied home purchase - the ledger still sorts its noes by race: Black applicants are turned down 2.9 times as often as White applicants. The dashed rule is the White baseline; everything to the right of it is the gap.

Plate 03 · Illustrative Scale runs 0 - 20% denied; 521,000 Black applicants sit in the slice. A denial rate is not a discrimination verdict - HMDA carries no credit score - but it is the count of who was turned away, and it moves the same direction every year it is measured.

Income doesn't buy parity

Plate 04Denied, by income band

The reflex explanation for the gap is money. HMDA lets you test it: hold income constant and the distance barely narrows. A Black applicant reporting over $150,000 a year is denied at 12.2% - within a point of the 13.4% rate for a White applicant earning under $50,000.

0% 10% 20% 30% Under $50k White (non-Hispanic), Under $50k: 13.4% denied Black, Under $50k: 26.3% denied 13.4 26.3 $50k - $75k White (non-Hispanic), $50k - $75k: 9.1% denied Black, $50k - $75k: 19.8% denied 9.1 19.8 $75k - $100k White (non-Hispanic), $75k - $100k: 7.2% denied Black, $75k - $100k: 16.1% denied 7.2 16.1 $100k - $150k White (non-Hispanic), $100k - $150k: 5.9% denied Black, $100k - $150k: 13.9% denied 5.9 13.9 Over $150k White (non-Hispanic), Over $150k: 4.9% denied Black, Over $150k: 12.2% denied 4.9 12.2
White applicants Black applicants % of applications denied · same home-purchase slice
Plate 04 · Illustrative Income is the gross annual figure the lender relied on, as reported in the HMDA record. Every dot carries its value; the gap runs 12.9 points at the bottom band and 7.3 at the top. Stand-in figures anchored to the published finding that high-income Black applicants are denied at roughly the rate of the lowest-income White applicants.

Where the gap is widest

Plate 05Top metros · Black - White gap

The national gap is an average of very different cities. Ranked here by the distance from the White denial rate (open dot) to the Black rate (filled) on the same home-purchase slice: Milwaukee leads by daylight, at 16.9 points, and the old industrial North - Chicago, Detroit, Cleveland, St. Louis - crowds the table alongside the Deep South.

# Metro White → Black denial rate Black White Gap
1 Milwaukee, WI
27.1% 10.2% +16.9
2 Memphis, TN
22.4% 9.1% +13.3
3 Jackson, MS
21.8% 8.9% +12.9
4 Baton Rouge, LA
20.9% 8.6% +12.3
5 Chicago, IL
19.7% 7.9% +11.8
6 Detroit, MI
19.2% 7.6% +11.6
7 Cleveland, OH
18.4% 7.3% +11.1
8 St. Louis, MO
18.1% 7.4% +10.7
9 Birmingham, AL
17.9% 7.5% +10.4
10 Philadelphia, PA
16.8% 6.7% +10.1
11 Atlanta, GA
16.2% 6.4% +9.8
12 Houston, TX
15.9% 6.5% +9.4
White applicants Black applicants axis 0 - 30% denied

Plate 05 · Illustrative Metros with enough applicants in both groups for the comparison to hold. In Milwaukee, WI, more than one Black application in four is turned down. Stand-in figures; the ranking echoes metros repeatedly flagged in published fair-lending reporting.

No metro passes the parity test

Plate 06One dot per metro

Put every metro on the same plot: the White denial rate across, the Black denial rate up. Even-handed lending would put a metro on the 1x line. All 21 sit above it - and all 21 clear the 2x line too. In this set there is no metro, North or South, cheap or expensive, where a Black applicant is less than twice as likely to be turned down; the narrowest ratio is 2.0x.

0% 10% 20% 0% 4% 8% 12% 1x parity 2x 3x Milwaukee, WI: Black 27.1%, White 10.2% denied Memphis, TN: Black 22.4%, White 9.1% denied Jackson, MS: Black 21.8%, White 8.9% denied Baton Rouge, LA: Black 20.9%, White 8.6% denied Chicago, IL: Black 19.7%, White 7.9% denied Detroit, MI: Black 19.2%, White 7.6% denied Cleveland, OH: Black 18.4%, White 7.3% denied St. Louis, MO: Black 18.1%, White 7.4% denied Birmingham, AL: Black 17.9%, White 7.5% denied Philadelphia, PA: Black 16.8%, White 6.7% denied Atlanta, GA: Black 16.2%, White 6.4% denied Houston, TX: Black 15.9%, White 6.5% denied Boston, MA: Black 14.1%, White 6.1% denied Dallas, TX: Black 17.1%, White 6.9% denied Los Angeles, CA: Black 15.4%, White 7.1% denied Miami, FL: Black 16.8%, White 8.3% denied Minneapolis, MN: Black 17.3%, White 6.2% denied New York, NY: Black 17.9%, White 7.4% denied Phoenix, AZ: Black 16.1%, White 7.2% denied Seattle, WA: Black 13.8%, White 6.3% denied Washington, DC: Black 15.6%, White 6.8% denied MilwaukeeMemphisChicagoSeattle Black denial rate ↑ White denial rate →
Plate 06 · Illustrative 21 metros from the leaderboard and compare set, same home-purchase slice. Milwaukee is the far corner: 27.1% against 10.2%. Exact rates for every dot are in the leaderboard table above and on the compare page. Stand-in figures.

Every metro tells its own version of the same story. Pull two of them side by side and read yours.

Compare two metros side by side →

Two mortgage markets

Plate 07Product mix of originations

The denials are only half the sorting. Among the loans that do close, 78% of White borrowers' home purchases ride conventional credit - the cheapest kind. For Black borrowers that falls to 55%, and government-backed FHA and VA loans, with their insurance premiums and fees, make up most of the difference.

0% 25% 50% 75% 100% Asian Asian - Conventional: 88% of originations Asian - FHA-insured: 8% of originations Asian - VA-guaranteed: 3% of originations Asian - RHS / other: 1% of originations White White (non-Hispanic) - Conventional: 78% of originations White (non-Hispanic) - FHA-insured: 12% of originations White (non-Hispanic) - VA-guaranteed: 7% of originations White (non-Hispanic) - RHS / other: 3% of originations Hispanic Hispanic or Latino - Conventional: 58% of originations Hispanic or Latino - FHA-insured: 31% of originations Hispanic or Latino - VA-guaranteed: 8% of originations Hispanic or Latino - RHS / other: 3% of originations Black Black or African American - Conventional: 55% of originations Black or African American - FHA-insured: 29% of originations Black or African American - VA-guaranteed: 13% of originations Black or African American - RHS / other: 3% of originations
  • Conventional
  • FHA-insured
  • VA-guaranteed
  • RHS / other
Product mix of home-purchase originations, % (table view is the source of truth)
Group ConventionalFHA-insuredVA-guaranteedRHS / other
Asian 88%8%3%1%
White (non-Hispanic) 78%12%7%3%
Hispanic or Latino 58%31%8%3%
Black or African American 55%29%13%3%
Plate 07 · Illustrative Home-purchase originations only, first-lien, owner-occupied, by HMDA loan type. FHA insurance is the market's fallback when conventional underwriting says no - so the mix is the gap's quieter twin. Stand-in shares anchored to the published FHA-reliance pattern.

Why the “no”

Plate 08Primary denial reason · share of denials

Every denial is filed with a reason. Debt-to-income leads by a wide margin - one no in three - then credit history and collateral. Read the list with care: these are the reasons lenders report, not an audit of them, and "collateral" in an appraisal-starved neighborhood is not a neutral word.

Plate 08 · Illustrative Primary reason only (HMDA allows up to four per denial); bars are scaled to the leader and every bar carries its share. Stand-in shares - the ordering matches published CFPB reporting, DTI first.

What happens to 15 million applications

Plate 09Action taken · all records

Ground the whole page in the raw outcomes. Of a year's ledger, roughly 8.7M applications become loans; millions more are withdrawn by the applicant, filed incomplete, or are loans bought and resold between lenders. The denied slice - about 2.0M decisions - is the one every other plate here reads.

  • Originated 8.7M 58.4%
  • Purchased / preapproval 1.7M 11.4%
  • Withdrawn 1.6M 11.0%
  • Denied 2.0M 13.8%
  • Approved, not taken 0.32M 2.1%
  • Closed incomplete 0.49M 3.3%

Plate 09 · Illustrative One proportional bar, 14.9M records wide. Purchased loans and preapproval requests are held out of every denial rate on this page, so no application is counted twice. Stand-in mix summing to the year's total.

The market that whipsawed

Plate 10Applications filed per year

The ledger itself is not steady. Pandemic-era rates swelled it to 23.3M filings in 2021; the 2022 rate shock cut it nearly in half. Keep that whiplash in mind for the next plate - the market's size moved violently, and the gap moved almost not at all.

0M 6M 12M 18M 24M 2018: 12,100,000 applications 12.1M 2018 2019: 15,100,000 applications 15.1M 2019 2020: 22,700,000 applications 22.7M 2020 2021: 23,300,000 applications 23.3M 2021 2022: 16,100,000 applications 16.1M 2022 2023: 12,400,000 applications 12.4M 2023 2024: 14,900,000 applications 14.9M 2024
Plate 10 · Illustrative All covered records filed under HMDA, every purpose and outcome - the universe the rest of this page samples from. Stand-in totals shaped to the published refi-boom-and-bust curve; the real series comes from counting rows per activity year.

The gap that won't close

Plate 11Denial rate by group · 2018 - 2024

The last plate is the stubborn one. Through the boom, the bust, and a five-point swing in mortgage rates, the oxblood Black line and the ink White line move in lockstep - the shaded distance between them was 10.4 points in 2018 and 10.7 points in 2024. Whatever moves this market, it is not closing the gap.

0% 5% 10% 15% 20% 2018201920202021202220232024 Black 16.8% Hispanic 12.3% Asian 9.2% White 6.1%
Plate 11 · Illustrative Denial rate on the comparable home-purchase slice, each line direct-labeled with its 2024 rate; the shaded band is the Black - White distance. Stand-in series shaped to the published year-over-year pattern.

Methodology

Notes on the Data

The Home Mortgage Disclosure Act (HMDA) requires most lenders to report every mortgage application they receive. The Consumer Financial Protection Bureau and FFIEC publish the pooled result each year as the CFPB HMDA Snapshot National Loan-Level Dataset (2024 activity year (snapshot published 2025-05-19)): roughly 15 million application records a year from about 4,900 filing institutions, one row per application, ~99 columns each.

What's real, what's a stand-in

Every number on this page is currently an Illustrative stand-in, badged as such at the top and on every plate. The magnitudes are anchored to published HMDA findings - for example, 2023 denial rates of roughly 16.6% for Black applicants versus 5.8% for non-Hispanic White applicants on comparable home-purchase loans, DTI as the leading reported denial reason, and the FHA-reliance and income-band patterns the CFPB and independent researchers have documented - but the exact state, metro, band, and reason figures here are hand-authored to shape the page, not ingested. The swap-point is documented in the repo's HANDOFF.md and src/lib/source.ts: drop the real snapshot CSV into data/raw/ and the same aggregations run against it. We never present curated numbers as real.

How the denial reading is framed

The by-group, income-band, metro-gap, and parity readings use a single comparable slice - first-lien, owner-occupied, conventional, closed-end home-purchase applications - so groups are read on like-for-like credit decisions rather than mixed against refinances and investment loans. Race and ethnicity use HMDA's derived_race / derived_ethnicity single-value fields; Hispanic or Latino is counted by ethnicity across races. Income is HMDA's income field - the gross annual figure the lender relied on, reported in thousands - which is self-limiting: it says nothing about wealth, savings, or a co-signer's balance sheet, only about pay.

The product-mix reading

The two-markets plate widens the slice on purpose: it keeps every loan type (conventional, FHA, VA, RHS/FSA) and reads only originated home-purchase loans, because the reading is which product the market hands each group, not whether it says yes. FHA insurance exists to extend credit that conventional underwriting refuses - which is exactly why its share is the denial gap's quieter twin, and why it costs its borrowers more in premiums.

What you're not seeing

A denial rate is not proof of discrimination. HMDA does not include credit scores, full debt-to-income detail, or loan-to-value at the record level published here, so a raw gap cannot control for creditworthiness the way a lender's underwriting can. What HMDA does show is who applies, who is turned away, and where - the starting point for a fair-lending question, not its verdict. Preapproval-only requests, purchased loans, and business-purpose records are held out of the denial reading.

Generated 2026-07-06 00:00 UTC. Source: CFPB HMDA, https://ffiec.cfpb.gov/data-publication/snapshot-national-loan-level-dataset/