Roughly one in ten American mortgages no longer fits the official definition of a safe loan. New analysis by Polygon Research puts 2025 non-qualified mortgage volume at $239 billion across 697,605 loans, about 10 percent of originations by dollar volume and 10.2 percent by loan count. The number has been climbing for years, and rate-lock data from Optimal Blue shows the trend still accelerating in 2026, with non-QM exceeding 10 percent of monthly lock volume in July. None of this means subprime is back. What it means is that the Qualified Mortgage standard, the underwriting box built after the financial crisis, has become something narrower than a definition of creditworthiness. It has become a definition of paperwork, and the market is building an entire second system for people whose paperwork does not fit.
To understand what is growing, you have to understand what the box tests. The Qualified Mortgage rule, written into Regulation Z after the crisis, gave lenders a safe harbor: originate a loan that meets its terms and the lender enjoys a presumption of compliance with the ability-to-repay rule. Those terms are mostly about documentation and arithmetic. A 43 percent debt-to-income cap, limits on points and fees, no interest-only payments or negative amortization, and income verified the way an underwriter verifies a W-2 employee. The rule did not say who is creditworthy. It said which files are legally comfortable to approve. The distinction has been doing more work every year since.
The box tests paperwork, not risk
Consider what a conventional loan file looks like. Two years of W-2s, recent pay stubs, tax returns that match, an employer who answers the phone. For most salaried workers that is a description of a boring Tuesday. For a self-employed contractor it is a fiction, because the tax return is an artifact of deduction strategy rather than a statement of cash flow. For a real estate investor who owns six rentals it is irrelevant, because the income that repays the loan is the rent, not the investor's salary. For a retiree with a large portfolio and no paycheck it is absurd, because the borrower's ability to pay is obvious and the documentation of it does not exist. These are not marginal borrowers. They are a large and growing share of the American economy, and the Qualified Mortgage box was never designed to hold them.
The result is the fastest-growing product family in mortgage lending: bank statement loans, which underwrite business owners off 12 or 24 months of deposits; debt service coverage ratio loans, which underwrite investors off the property's rent; asset depletion loans, which convert portfolio balances into an income stream; and 1099 and profit-and-loss loans for the newly self-employed. Polygon Research classifies all of these as non-QM, and its count includes portfolio lending and private execution channels, not just securitizations, which is one reason the figure has been consistently underestimated. The other reason is that the market does not want the attention. Non-QM is a category defined by exclusion, and nobody markets a loan as the one the rules did not bless.
The borrowers who ended up in the second system
The share of Americans working for themselves has grown through the past decade, and the tax code gives them strong reasons to minimize reported income. The same person who qualifies for a conventional loan in a good year can fall out of the box the moment their accountant gets creative. The growth of short-term rentals and small landlord portfolios has created a parallel class of borrowers whose income is property-based, and whose financial sophistication often exceeds that of the salaried first-time buyer the box was designed for. And the demographic wave is structural: more workers are self-employed, more retirements are funded by assets rather than pensions, and more housing demand comes from investors. The box is rigid; the economy around it is not.
That mismatch shows up in the price. Non-QM loans carry higher rates and larger down payment requirements, partly because they are riskier to originate and partly because they sit outside the agency system that subsidizes conventional mortgages. A bank statement loan for a surgeon with irregular income is not subprime in any meaningful sense, but it is priced as if the lender has to be compensated for the absence of a GSE guarantee. The second system works. It is also more expensive than the first one, and its borrowers are paying a premium that is partly about risk and partly about the paperwork.
What 10 percent changes
The 10 percent line matters because it is a threshold of normalcy. At two or three percent, non-QM is a specialty. At ten percent of a $2 trillion-plus origination market, it is a structural feature, and its composition is changing. Optimal Blue's July 2026 lock data shows investor and DSCR loans at 33.5 percent of non-QM volume, bank statement loans at 30.6 percent, and other expanded-guideline products filling the rest. The investor share is the number to watch, because it ties non-QM's fate to the rental market. Investor loans are underwritten on rent, and rent is a market price that can soften. A concentrated investor portfolio financed with DSCR loans in a softening rental market is the scenario where the second system gets its first real stress test.
The comparison everyone reaches for is 2008, and it is mostly wrong. The subprime crisis was not caused by a documentation category. It was caused by loans whose payments were designed to jump, collateral values that were fraudulently inflated, and securities markets that priced none of it. Non-QM underwriting today is stricter than prime underwriting was in 2006, with real cash-flow analysis replacing stated income and significant down payments replacing nothing. The category called non-QM contains some of the most carefully underwritten loans in the market. What it shares with the old subprime world is the structural position: it is the part of the market that absorbs borrowers the mainstream system will not serve, and it grows fastest exactly when the mainstream system is most rigid.
The capital markets have voted too
The second system could not exist at this scale without someone funding it, and the funding channel is the quietest part of the story. Non-QM loans have to go somewhere, because most depositories do not want a large book of non-agency product on the balance sheet. So a securitization market has grown up around them, with private-label deals built from bank statement, DSCR, and investor collateral, rated by the same agencies that rate everything else, and sold to bond buyers who want the yield. A $239 billion annual origination flow implies a substantial and permanent capital market apparatus behind it, complete with its own underwriting standards, its own performance history, and its own vocabulary of risk. The apparatus is the real measure of maturity. When a category of lending develops a functioning secondary market, it stops being an experiment and becomes infrastructure, and infrastructure does not get wished away by changing the rules that created it.
That infrastructure has consequences for consumers that never show up in a rate sheet. Securitization imposes its own discipline, because deal buyers enforce underwriting guidelines as hard as any regulator, but it also concentrates information about borrower performance in the hands of investors rather than the public. The agency market publishes its data through HMDA and GSE disclosures. The private-label market does the minimum. The $239 billion estimate exists because Polygon Research reverse-engineered it from HMDA records and standards analysis, not because anyone publishes it. The second system is now large enough to move the national mortgage statistics, and opaque enough that measuring it requires forensic work.
The box shapes the market it refuses to see
The deepest lesson of the $239 billion is not about any loan. It is about how rules create markets. The Qualified Mortgage standard was written to make lending safer, and it did, by making the mainstream boring. But a rule that defines safety by documentation rather than by risk does not eliminate the riskier borrower. It relocates them, into a parallel system with thinner regulatory coverage, higher prices, and less agency support. That is the trade the system has made for a decade, and it was visible in the data long before anyone called it a milestone.
For the borrower standing in the second system, the trade is measured in points and months. The business owner who closes a bank statement loan pays a premium over what the same borrower's risk would command inside the box, and the premium is the price of the system's comfort rather than the borrower's default probability. That is not an argument for tearing the box down. It is a reminder that every rule has a price, and the price of this one is paid by precisely the people the box was never built to see.
The question for the next decade is whether the box gets rebuilt to fit the economy, or whether the second system keeps absorbing what the first one cannot see. The answer will be visible in the same numbers: the mix of investors and business owners, the spread between non-QM and agency pricing, and the default performance of DSCR portfolios when rents stop rising. One in ten mortgages now sits outside the box. The box is not going away, and neither is the pressure against its walls.
Primary sources
- Polygon Research for the 2025 non-QM market sizing, the $239 billion figure, the 697,605 loan count, and the methodology built on HMDA data.
- Tri-City Herald for the Stacker analysis of the one-in-ten finding and the Optimal Blue rate-lock data for 2026.
- Griffin Funding's publication of the Polygon Research findings for the non-QM borrower type breakdown, including bank statement, DSCR, and asset depletion loans.