A new independent study is sharpening the debate over which credit scoring model will define the next era of mortgage underwriting.
A white paper by actuarial firm Milliman found that FICO’s FICO Score 10T outperforms VantageScore 4.0 in predicting mortgage default risk across all major loan types, including GSE and FHA originations.
The findings arrive just weeks after federal housing agencies signaled a shift toward allowing multiple credit models in mortgage lending.
Head-to-Head Results Favor FICO 10T
The Milliman study analyzed nearly 20 million U.S. mortgages spanning more than a decade, making it the first large-scale independent comparison of the two “modern” scores.
Among the key findings:
- FICO 10T delivered stronger predictive performance across all mortgage segments, including conforming and FHA loans.
- In FHA lending, performance improved by more than 8% versus VantageScore 4.0.
- The performance gap has widened over time, with a 7.4% advantage in 2023 GSE loans, more than double the margin seen in 2018.
- FICO 10T incorporates trended credit data, including payment behavior over time, and, when available, rental data, giving lenders a more dynamic view of borrower risk.
For LOs, that translates to sharper risk segmentation, particularly in marginal files where approval decisions are most sensitive.
Timing Matters: Credit Model Shake-Up Underway
The study lands at a pivotal moment.
As NMP recently reported, the Federal Housing Finance Agency and U.S. Department of Housing and Urban Development are moving toward a multi-model framework, allowing both FICO 10T and VantageScore 4.0 in mortgage underwriting.
That shift is expected to:
- Increase competition between scoring models
- Expand credit access using alternative data like rent
- Force lenders to evaluate performance, cost, and operational impact
It also creates a more complex decision-making environment for originators and capital markets teams.
What It Means
The takeaway isn’t just which score “wins” — it’s how lenders respond.
The study reinforces what early adopters have been signaling: more predictive models can reduce default risk while potentially expanding approvals, especially when trended data is incorporated.
But the broader market is moving toward choice, not standardization.
That creates several near-term implications:
- Dual-score workflows will likely become the norm as lenders test outcomes
- Pricing and AUS alignment will be critical as investors evaluate risk across models
- Borrower segmentation could widen, particularly between improving vs. deteriorating credit profiles
Bottom Line
The Milliman study gives FICO 10T a clear edge on predictive performance, but it doesn’t settle the question.
With regulators opening the door to multiple models, lenders aren’t choosing a single winner. They’re entering a side-by-side testing phase that will define how credit is evaluated — and who gets approved — in the next cycle.
For LOs, the shift is less about the score itself and more about what it enables: more granular risk decisions, greater outcome variability, and a more competitive credit landscape.
