RiskSpan, a mortgage analytics firm, released Credit Model 7.1 specifically designed to assess credit risk in non-qualified mortgage (non-QM) loans. The model was built using data from $87 billion in unpaid principal balance across non-QM portfolios.
Non-QM loans serve borrowers who don't fit traditional lending criteria. These include self-employed individuals, gig workers, and applicants with irregular income patterns. Lenders use alternative documentation like bank statements and tax returns instead of W-2s and pay stubs. The market has expanded rapidly as borrowers sought flexibility and lenders pursued higher-margin products.
Q3 2025 issuance surged 97 percent year-over-year to $20.9 billion, signaling strong investor appetite and lender confidence in this segment. The jump reflects demand from both borrowers locked out of conventional financing and lenders chasing profitability in a competitive market.
RiskSpan's new model matters because non-QM loans carry different risk profiles than traditional mortgages. Borrowers have less predictable income streams. Default patterns diverge from standard QM loan performance. Existing credit models don't accurately capture these nuances, leaving lenders and investors vulnerable to mispriced risk.
Credit Model 7.1 uses $87 billion in historical non-QM performance data to better predict defaults, prepayments, and loss severity specific to this borrower pool. The larger training dataset improves accuracy. More accurate pricing protects lenders from underpricing risk and helps investors make informed securitization decisions.
For borrowers, better risk modeling could mean tighter underwriting or higher rates if lenders discover non-QM borrowers carry more risk than previously thought. Alternatively, if the model confirms non-QM loans perform better than feared, rates could