2025 HMDA Data Deep Dive: Growth, Second Liens, Non-QM, and the Affordability Pressure Lenders Should Be Watching

Greg Oliven of Polygon Research grabbed the mic to walk us through what the newly released 2025 HMDA data actually says, and what lenders need to do with it. Be sure to check out the full Polygon Research here!

Watch the full episode below, or scroll for the highlights! Featured guests:

Bob Simpson — Daylight AML

Ray Snytsheuvel — Loan Risk Advisors

Dana Georgiuo — Lending Luminary

Greg Oliven — Polygon Research

Nathan Knottingham — MLO Force

TL;DR

The 2025 HMDA data landed at the end of March, and Polygon Research had already updated their full table structure, pivots, and analysis within two weeks. In this episode of Risk & Roll, Greg Oliven walked the panel through what the numbers are showing — from double-digit market growth to a surge in second liens, a climbing non-QM share, and early warning signs of delinquency pressure that lenders should not be ignoring.

The Market Was Up 10% in 2025 — Here's What That Means

After bottoming out in 2023, the mortgage market has been climbing. 2025 pushed that recovery into double-digit growth territory, with originations up 10% year over year. That's the headline number, but Greg was quick to point out that national figures only tell part of the story. The real strategic value in HMDA data is local, market by market, county by county, and lenders who aren't slicing the data at that level are missing where the actual opportunity lives.

Second Liens Were a Winning Strategy — One Player Saw It Coming

One of the most prominent trends in the 2025 data: the steady climb of subordinate lien lending. Second liens have been creeping up for several years, but 2025 made the story undeniable. Borrowers who locked in low rates on their first mortgages have been tapping home equity through seconds — and then refinancing those seconds as rates have softened slightly. One lender in particular, Rocket, was clearly ahead of this trend and leaned into it hard, especially in California where the concentration was striking.

Dana added another dimension: ADU construction. In markets with severe housing supply constraints, like greater Los Angeles, a meaningful portion of that HELOC and second lien activity is likely funding accessory dwelling unit builds, a way for homeowners to create value and potentially rental income without touching their low-rate first mortgage.

Cash-Out Refi and the Non-QM Surge: What the Data Shows

Breaking down loan purpose by lien status revealed another layer: cash-out refinances on second liens have been a notable part of the 2025 story, particularly in Southern California. The panel noted that purpose tracking in HMDA depends heavily on how well lenders are categorizing these loans at filing — a reminder that data quality at the source affects what anyone can learn from it downstream.

The non-QM picture was equally clear. Business-purpose non-QM originations jumped roughly 20% in the most recent year. Non-business non-QM climbed about 9%. Dana pointed out that DSCR lending — where qualification is based on whether rental income covers the payment, not on personal income documentation — has been one of the fastest-growing and consistently performing segments. Kiavi leads the private lending space in this category, and institutional capital has been moving in because the returns are there.

Greg noted that the definition of non-QM itself shifted around 2021, moving from DTI as the anchor to rate spread. That change, combined with rising rates, reshaped who was using non-QM and why — and the growth since has been steady and unmistakable.

Ray added the legal context: for a long time, lenders avoided non-QM because the liability lines weren't clear. DSCR lending, as business-purpose lending, operates outside CFPB disclosure requirements entirely — no LE, no CD — and that bright line made it easier for compliance teams to get comfortable.

The Delinquency Warning Sign Lenders Shouldn't Ignore

Delinquency rates have been ticking up quarter over quarter. They're still below pre-pandemic levels, but the direction matters. Bob raised a question the panel couldn't fully dismiss: borrowers who got fixed-rate loans at sub-4% are insulated from rate risk, but not from everything else. Taxes, insurance, cost of living, and home maintenance costs have all risen significantly. A borrower who closed at a 49% back-end DTI a few years ago had almost no financial buffer — and they may be feeling that now.

Dana called out that investment property loans — particularly in the RTL (residential transition loan) space, where securitizations are now rated — have actually held flat or declined slightly in delinquency year over year, bucking the trend. Asset-based underwriting, done right, appears to be performing.

The panel agreed that a more useful lens than DTI alone would be residual income — the VA loan framework that looks at what a borrower actually has left after all obligations. Nathan argued it's a better predictor of actual default risk and that the industry's comfort with DTI as the primary yardstick may be obscuring real vulnerability in parts of the current book.

Who's Moving Up the Leaderboard — and Who Isn't

The leaderboard data showed some notable shifts. CrossCountry Mortgage grew roughly 22-23% year over year in both 2024 and 2025, quietly building market share. PennyMac showed nearly 50% growth — partly through acquisitions of other brands. Bank of America, which ranked in the top five in 2024, dropped out of the top five depositories in 2025 as independent mortgage banks continued to dominate by unit count.

Dana connected CrossCountry's rise to a strategy of national reach with local execution — particularly around servicing. If you own the servicing book, HMDA data tells you exactly where to execute a recapture or retention play and where it's not worth the effort.

Fair Lending: Are We Actually Performing Better?

Ray brought the conversation back to what he sees as the core question the HMDA data should answer: not who originated the most, but are we doing a better job of serving the people who need access to homeownership most?

Greg walked through how HMDA data, combined with census data, covers all four major fair lending analysis pillars: redlining, underlining, pricing/steering, and marketing. The pricing data in particular has become much richer since 2018, allowing for detailed scatter-plot analysis of interest rate versus net borrower charges across lenders and markets.

Bob framed the underlying question clearly: how far would rates and prices have to fall to bring housing costs into a normal relationship with area median income? That's the actual affordability problem, and it's what that data, used well, should help the industry answer.

Key Takeaways

• The 2025 mortgage market grew 10%, but national numbers hide major local variation. Strategy has to be built at the market level.

• Second lien growth was one of 2025's defining trends. Lenders who anticipated it early captured significant share.

• Non-QM and DSCR lending are surging and performing well. Institutional capital has noticed.

• Delinquency rates are ticking up across traditional mortgage products. Fixed rates don't protect borrowers from rising insurance, taxes, and cost of living.

• DTI alone is an incomplete affordability measure. Residual income analysis gives a clearer picture of actual repayment risk.

• HMDA data is most powerful when used locally — and when combined with census data for fair lending analysis.

• The fundamental question the data should answer: are we getting better at serving the people who want to buy homes in their own communities?

Learn more about the latest HMDA data on the Polygon Research blog.

Risk & Roll covers mortgage compliance, risk, and industry strategy in every episode. Subscribe wherever you listen, and leave your feedback in the comments or at support@mloforce.com.