Premier Insights

2025 HMDA Data: More Volume, Better Aggregate Outcomes, Persistent Disparities

Written by Premier Insights | Aug 20, 2026, 2:48:24 PM

The 2025 Static HMDA Loan/Application Register tells a clear headline story: mortgage activity expanded again. Applications and originations both increased for a second consecutive year, aggregate denial rates declined, and pull-through improved.

The more useful story begins beneath those totals. Borrower and neighborhood reach did not rise with overall volume. Denial-rate differences remained pronounced across several demographic and geographic comparisons. Pricing differences moved in different directions depending on the population being compared.

These findings come from Premier Insights' new 2026 HMDA Market Intelligence research, available in national and market-level reports (or as a custom report tailored to a bank's specific assessment areas). The national analysis covers 13.54 million records in the 2025 Static LAR and compares them with the 2024 and 2023 Static LAR files.

Mortgage activity expanded for a second year

The 2025 Static LAR contains 11.77 million applications and 6.83 million originations. Compared with 2024, applications increased 9.4% and originations increased 10.6%.

The two-year movement is even more striking:

  • Applications increased from 9.99 million in 2023 to 11.77 million in 2025, a gain of 17.8%.
  • Originations increased from 5.69 million to 6.83 million, a gain of 20.0%.
  • Total reported LAR records reached 13.54 million in 2025, up 10.7% from 2024.

 

National HMDA applications and originations, 2023-2025 Static LAR files. Source: Public HMDA LAR data and Premier Insights analysis.

This was not simply more demand entering the top of the funnel. Completed production grew slightly faster than applications, which helped lift the national pull-through rate.

Aggregate outcomes improved

National pull-through increased from 57.0% in 2023 to 58.0% in 2025. Over the same period, the aggregate denial rate declined from 20.0% to 18.0%, including a 1.6 percentage-point decline in the latest year.

National pull-through and denial rates under the report's current approved universes and denominator conventions.

That improvement matters, but a national average can conceal very different operating environments. Institutions reporting at least 10,000 LAR records accounted for 73.2% of the latest national file. Their aggregate denial rate was 19.9%, compared with 11.7% among reporters with fewer than 1,000 records. Their pull-through rates were 54.4% and 72.3%, respectively.

Those differences should not be interpreted as simple performance rankings. Reporter size can be associated with different footprints, channels, products, customer populations, and operating models. It does show why national totals are most useful when they can be segmented.

One other segment deserves explicit attention: nonmetropolitan lending. State non-MSA areas accounted for 1.50 million applications and 857,400 originations in the 2025 file, or 12.8% and 12.6% of the national totals. Originations in these areas increased 9.0% from 2024. A national market view that excludes MSA/MD code 99999 would miss a meaningful part of the mortgage market.

Reach did not expand uniformly with volume

Overall activity grew, but several borrower and neighborhood shares declined in the latest year:

  • Black borrower share declined from 9.1% in 2024 to 8.4% in 2025.
  • Hispanic borrower share declined from 13.5% to 12.8%.
  • Majority-minority tract share declined from 27.3% to 26.4%.
  • Low- and moderate-income (LMI) borrower share increased from 26.8% to 27.7%, nearly returning to its 2023 level of 27.9%.

Borrower and neighborhood reach measures as shares of records under the report's current definitions.

These changes are not, by themselves, evidence of unmet demand or discriminatory treatment. Reach can move because of footprint, channel, product mix, applicant demand, reporting quality, and other factors. The practical takeaway is narrower: growth in total applications or originations does not necessarily mean that growth was distributed evenly across borrowers and communities.

Denial-rate differences remained persistent

The national screening results show sizeable target-control differences in denial rates. In the latest file:

  • The denial rate for Black applicants was 34.3%, compared with 19.1% for White non-Hispanic applicants, a difference of 15.3 percentage points.
  • The denial rate for LMI applicants was 33.2%, compared with 17.9% for non-LMI applicants, also a difference of 15.3 percentage points.
  • The denial rate for applications in majority-Black census tracts was 32.3%, compared with 22.1% outside those tracts, a difference of 10.2 percentage points.
  • The Hispanic-to-White non-Hispanic denial-rate difference was 7.7 percentage points.

The year-over-year direction was mixed. The LMI difference narrowed by 0.9 percentage point, while the Black-to-White non-Hispanic difference widened by 0.3 point and the majority-Black-tract difference widened by 0.4 point.

All seven configured national denial-rate comparisons met the report's current p<0.05 statistical screen. With samples this large, statistical detectability is expected and should not be confused with practical materiality, causation, or a finding of discrimination. These results identify where additional segmentation and validation may be warranted; they do not explain why a difference exists.

Pricing differences told a more complicated story

Pricing did not follow the same pattern as denials. Among originations with a numeric reported interest rate:

  • The mean reported rate in majority-Black census tracts was 18.3 basis points higher than the mean outside those tracts. That difference increased by 8.7 basis points from 2024.
  • The female-to-male difference was positive 7.5 basis points, up 3.8 basis points from 2024.
  • The Black-to-White non-Hispanic difference was negative 5.2 basis points, meaning the Black borrower mean was lower in this unadjusted national comparison. The difference moved 11.1 basis points toward zero from 2024.
  • The Hispanic-to-White non-Hispanic difference was negative 10.8 basis points and moved 9.1 basis points toward zero.

This is precisely why denial and pricing indicators should be reviewed separately. A population can show a higher aggregate denial rate without also showing a higher mean reported rate among originated loans. The people represented in a pricing comparison are, by definition, those whose loans reached origination, and unadjusted means do not control for product, purpose, geography, loan characteristics, credit profile, or other explanatory factors.

As with the denial-rate results, all seven configured pricing comparisons met the current p<0.05 statistical screen. They are screening indicators, not causal models or legal conclusions.

What should institutions take from the national results?

Three conclusions are useful at this stage:

  1. The market expanded, but headline growth is only the starting point. Volume, conversion, and denial trends improved nationally, while the distribution of activity varied by reporter scale, geography, and population.
  2. Reach and outcomes need separate attention. More applications and originations did not produce uniform gains in borrower or neighborhood representation.
  3. Fair-lending screens should create questions, not conclusions. Persistent denial-rate differences and mixed pricing results are most useful as a guide to deeper review by product, purpose, geography, channel, lender, and other relevant factors.

The national file provides the common frame. The next analytical step is usually local: determining whether the same patterns appear in a bank's markets, assessment areas, product lanes, and peer context.

These findings are drawn from Premier Insights' 2026 HMDA Market Intelligence release. The complete national analysis and market-level report library are available for institutions that need the supporting detail, lender context, and local comparisons.