United States/2025 ACS/October 22, 2026

The Vanishing Small Mortgage (2025)

Loans under $100,000 are both rare and the hardest to get approved. The cheapest homes are the hardest to finance.

Key findings

  • Mortgages under $100,000 made up just 9.5% of home-purchase applications in 2025, even though plenty of homes still sell for that little.
  • Those small loans were denied at 35.2%, about 2.9 times the 12.3% denial rate on loans of $100,000 or more.
  • The $50,000-to-$100,000 range fared worst of any size band, denied 44.0% of the time.
  • Cheap homes are the hardest to finance, a quiet barrier to ownership that rarely shows up in headline affordability numbers.

Housing affordability is usually framed as homes being too expensive. The mortgage data reveals the opposite problem at the bottom of the market: small loans are remarkably hard to get. This bins every 2025 home-purchase application by loan size and measures how often each band was denied. Small mortgages are not very profitable to originate, so lenders approve them less readily, which leaves buyers of modest homes squeezed out even when the price is within reach. HMDA is a near-census, so these are exact counts.

Small-dollar loans, 2025

9.5%
Share under $100k
35.2%
Denial rate under $100k
12.3%
Denial rate $100k and up
44.0%
Denial rate $50k to $100k

Denial rate by loan size

Loading chart…
Home-purchase denial rate by loan amount, HMDA 2025. Smaller loans are denied far more often.
CategoryDenial rate
Under $50k23.5%
$50k to $100k44%
$100k to $250k18.6%
$250k to $500k9.4%
$500k and up8.7%
Home-purchase denial rate by loan amount, HMDA 2025. Smaller loans are denied far more often.

By loan size

Loan amountPurchase decisionsShareDenial rate
Under $50k171,0744.1%23.5%
$50k to $100k229,8255.5%44.0%
$100k to $250k1,253,29429.8%18.6%
$250k to $500k1,721,76241.0%9.4%
$500k and up823,89619.6%8.7%

Covers home-purchase applications that were either originated or denied. HMDA rounds loan amounts to the nearest $10,000. Exact counts; no sampling margins of error.

Data Sources & Methodology

The figures on this page are computed directly from the source data below. Narrative text accompanies the figures but never generates them.

Primary data source

Home Mortgage Disclosure Act (HMDA) Loan/Application Register (LAR)

Vintage:
2025
Access:
BigQuery
Cadence:
annual

Loan-level records of U.S. mortgage applications and originations reported under the Home Mortgage Disclosure Act. Each row is one application or loan, with loan terms, action taken, property and census-tract geography, and privacy-protected applicant demographics. It is a near-census of covered mortgage activity, enabling analysis of lending, denials, and access to credit.

What the data is

HMDA requires most U.S. mortgage lenders to report every mortgage application and loan they act on. The annual snapshot file is a frozen national extract of these loan/application records (the Loan/Application Register, or LAR) for the reporting year.

Near-census, not a sample

The LAR is administrative data covering the large majority of U.S. mortgage activity, not a probability sample. Statistics are exact counts and shares over the reported records; there are no sampling weights and no sampling margins of error.

Privacy protections

To protect applicant privacy, the public file bins or rounds several fields (for example age, income, loan amount, and debt-to-income and loan-to-value ratios) and withholds some values. Analyses use these modified fields as published.

Geography

Each record carries state and county FIPS codes and a census tract, supporting state, county, and tract-level analysis; metropolitan areas can be derived from county.

Limitations
  • HMDA covers lenders above reporting thresholds, so it captures the large majority, but not literally all, of U.S. mortgage activity.
  • Several applicant fields are binned, rounded, or withheld for privacy, which limits precise computation on income, age, loan amount, and ratio fields.
  • The data reflects applications and lender actions, not loan performance; it says nothing about whether loans were later repaid.
  • Reported race, ethnicity, and sex follow HMDA collection rules and include "not provided" and institution-reported categories that require careful handling.

How these figures were produced

Estimate type:
Administrative near-census (exact counts)
Weighting:
None (not a weighted sample)

Suggested citation

Consumer Financial Protection Bureau, 2025 Home Mortgage Disclosure Act (HMDA) Loan/Application Register, via FFIEC. Analysis by Matt George, itsmattgeorge.com. Retrieved 2026-10-22.

Sources (2)
  1. Consumer Financial Protection Bureau: HMDA Snapshot National Loan-Level Dataset
  2. FFIEC: HMDA Public Data Documentation and Filing Instructions Guide