The Federal Housing Finance Agency’s introduction of "lender choice” – allowing mortgage originators to select between FICO and VantageScore 4.0 when delivering loans to Fannie Mae and Freddie Mac – represents one of the most sweeping overhauls to mortgage underwriting since automated systems debuted in 1996. Although this policy is framed as a move to foster competition, my previous research warned that it creates an inherent structural vulnerability: an incentive for lenders to score-shop, maximizing loan acceptance at the expense of long-term credit stability.
With the FHFA’s updated 2026 mandate allowing lenders to choose between FICO Score 10T and VantageScore 4.0, a critical question emerged for mortgage and mortgage-backed securities (MBS) investors: How does this next-generation, trended-data matchup affect credit and market risk?
To find out, I updated the earlier analysis utilizing the newly released historical GSE (government-sponsored enterprise) datasets. The data reveals that while FICO 10T provides powerful new capabilities to mitigate risk, the core flaw of lender choice remains an unchecked threat to market stability.
Our updated statistical modeling brings a definitive conclusion to light. FICO 10T exhibits notably greater power to distinguish between serious delinquent (D90+) and nondelinquent loans than VantageScore 4.0.
Dr. Clifford Rossi
This edge is largely driven by design. While both models utilize 24 months of trended data, FICO 10T requires a minimum of six months of credit history to generate a score, compared to just one month for VantageScore 4.0. Because longer credit history is inversely related to default, FICO 10T is significantly more adept at identifying higher relative credit risk among lower-scoring borrowers.
For example, our statistical modeling reveals that under Classic FICO, borrowers with credit scores below 660 are 8.57 times more likely to become seriously delinquent than those above 700. Crucially, our analysis shows that FICO 10T identifies much greater performance differences and higher relative credit risk among these lower-scoring groups than either Classic FICO or VantageScore 4.0, capturing a much steeper, more accurate risk gradient where it matters most.
The good news is that substituting FICO 10T for Classic FICO helps cushion the blow of lender score-shopping.
In a worst-case, 100% adverse selection scenario – where lenders systematically choose the credit score most favorable to loan delivery – the incremental increase in delinquency rates is reduced by nearly one-third when FICO 10T swaps out for Classic FICO and is the alternative to VantageScore 4.0. Applied to the GSEs' 2025 single-family loan purchases, the estimated incremental credit losses from full adverse selection drop from an average of $159 million (under Classic FICO vs. VantageScore 4.0) to $100 million (under FICO 10T vs. VantageScore 4.0).
However, "muted" risk is not "eliminated" risk. Leaving lender choice unchecked still exposes the GSEs and credit investors to millions in avoidable, structurally induced losses. Furthermore, our analysis indicates that the GSEs' decision to eliminate the minimum 620 credit score requirement was premature. The sheer volatility in risk assessment across these models warrants an immediate re-evaluation and the reinstatement of a low-side credit score overlay to protect the system.
Beyond credit defaults, lender choice injects an unprecedented layer of volatility into voluntary prepayment speeds, directly distorting Option-Adjusted Spread (OAS) modeling for MBS investors.
Prepayment models have been trained on point-in-time Classic FICO data for decades. Replacing or mixing these baselines with trended-data scores drastically flattens modeled prepayment expectations across different performance windows. For instance, across a 24-month horizon, borrowers with a Classic FICO score above 660 prepay nearly 1.4 times faster than those below 660. Under lender choice scenarios utilizing FICO 10T and VantageScore 4.0, that relationship compresses significantly.
Because investors cannot accurately predict how aggressively lenders will switch between scoring models from pool to pool, they face a double-blind on both voluntary (refinance) and involuntary (default) prepayment behavior. This structural ambiguity will almost certainly force the market to price in lender choice uncertainty premiums, unnecessarily raising execution costs.
The mortgage industry must remember the hard-learned lessons of the 2008 global financial crisis, where the credit rating "issuer-pay" model allowed asset issuers to shop around for the most lenient credit ratings. While originators are financially incentivized to maximize the volume of loans sold to the GSEs, credit investors require uncompromised, undegraded metrics to evaluate underlying risk.
"Credit score competition" is a misnomer; in practice, it is a vehicle for adverse selection. Without an effective, stabilizing response from the FHFA and the GSEs, lender choice will continue to compromise credit standards and destabilize MBS pricing.
Clifford Rossi is Executive-in-Residence, Johns Hopkins University, Carey Business School; and Professor-of-the-Practice Emeritus, University of Maryland – Robert H. Smith School of Business. Over a 25-year industry career spanning the S&L and 2008 financial crises, Dr. Rossi worked for both Fannie Mae and Freddie Mac, as well as some of the largest banks and nonbank institutions in various C-level risk management positions.