Deteriorating non-qualified mortgage (non-QM) performance that first emerged in 2024 and accelerated through 2025 is likely to worsen further in 2026, a sector report published last week warns, linking concentrated weakness to specific non-QM products.

Impairment rates increased across all underwriting documentation types in May except for full-doc loans, dv01 reported, with impairments rates for certified public accountant-endorsed, profit-and-loss loans floating back up to 10% last month. Impairment rates reflect the portion of loans that are delinquent or under modification within a securitized loan pool.

“While dv01 initially highlighted a notable gap between self-employed and non-self-employed documentation types in 2026,” said the Fitch Solutions-owned data analytics firm, “it has become increasingly apparent that the divergence is actually between full doc and all other doc types.” Fitch Solutions is also the parent of ratings firm Fitch Ratings.

Non-QM loans are broadly distinguished from agency mortgages that satisfy Fannie Mae and Freddie Mac guidelines, and use alternative documentation to verify a borrower’s income and employment. The sector plugs mortgage credit access holes relied on by self-employed borrowers, small-business owners, gig workers, foreign nationals and investors.

“The non-QM market’s performance rapidly deteriorated in May,” said dv01, “continuing to diverge from overall mortgage market strength and mirroring trends seen in the weaker credit FHA market,” referring to government-insured mortgage programs from the Federal Housing Administration that require lower downpayments and cater to borrowers with lower credit scores.

Dv01 reported full-doc impairment rates have declined 100 basis points in 2026 to rest below summer 2022 lows. Debt-service coverage ratio investor loans have remained stable around 6% since the beginning of 2025.

“Conversely, impairment growth continues unabated among the self-employed/bank statement segment,” dv01 stated, underscoring the continuation of a trend it flagged earlier this year.

The report also said borrower FICO credit scores remain a “critical differentiator,” with impairment rates for borrowers with scores below 660 nearing 20%. Borrowers with scores under 700 accounted for over 80% of the monthly rise in impairments.

Dv01’s monthly report reflects the performance of roughly 125,000 active loans originated since 2018, with an aggregate outstanding balance of more than $74 billion. With an average balance of about $411,000, the loans have a weighted average FICO score of 741, loan-to-value ratio of 67.4%, mortgage rate of 5.6% and debt-to-income ratio of 31.8%.

Silver linings exist in the data, however. Total impairments rose 30 basis points to land at 6.7% in May, below January and February levels north of 7%. Monthly first-time new impairment rates are tracking “significantly lower than 2024 and 2025 levels,” said dv01.

First-time new impairments moved up 15 basis points from April, to around 0.6%, but roughly two-thirds of that deterioration was driven by weakness in a single non-QM security. The overall new impairment rate was 1.44%, up 20 basis points over the month, fueled entirely by a 20-basis-point rise in the dv01 delinquency rate to 6.1%.

“This disparity suggests that elevated impairment activity is primarily driven by strained borrowers returning to delinquency,” the report observed, “rather than weaker underwriting translating into higher delinquencies.” Alongside “persistent softness” in cure and made-payment rates, inconsistent payment histories undergird the broader trend.

Delinquency cure rates, which represent the share of delinquent loans returned to current status, declined 20 basis points to 23%. The made-payment rate in May, which indicates the portion of borrowers in a securitized pool who made their scheduled payment, increased 20 basis points to about 46%.

“The increase offers little reassurance, however, as they continue to hover at all-time lows, serving as a primary factor for the persistently high impairment rates,” dv01 concluded.