New research from Prosperity Now analyzes 44.7 million mortgage loans to examine how credit scoring models function within the mortgage system, and what differences in model behavior mean for underwriting, pricing, and market stability.

‍

While both models consistently rank-order borrower risk, they differ in how risk is classified, segmented, and reflected in lending decisions. As the mortgage market moves toward a multi-score environment, these differences introduce new considerations for how risk is interpreted and applied across the system.

‍

Key Insights:

‍

• Both models consistently identify relative risk: Across all datasets and time periods, both scoring approaches reliably rank borrowers from higher to lower risk, providing a stable foundation for underwriting and pricing decisions.

• Differences emerge in how risk is classified and segmented: While relative risk ordering is consistent, borrowers may be placed into different score ranges depending on the model used, particularly near key eligibility and pricing thresholds.

• Model behavior diverges under stress conditions: During the COVID-19 period, both models maintained risk ordering, but differences in how borrowers were segmented became more visible, especially among lower-risk segments.

• Forbearance delayed, rather than removed, underlying risk: Temporary policy interventions suppressed observed defaults, but borrower outcomes over time reflected underlying risk differences across segments.

• Disagreement cases highlight where model differences matter most: A small share of loans receive different classifications across models. These cases provide insight into how differences in calibration translate into observed performance, though outcomes are not uniform across all segments.

• Pricing does not always move proportionally with observed risk: Across the current system, variation in interest rates is narrower than variation in default risk, reflecting structural features of how pricing is applied rather than the behavior of any single model.

‍

Why This Matters

‍

Credit scores are not used in isolation. They inform eligibility, pricing, servicing, and how risk is distributed across the mortgage system.

‍

As the market transitions from a single-score framework to one where multiple models may be used, the question is no longer just how risk is identified, but how it is interpreted and applied in practice.

‍

This report provides an empirical foundation for understanding those dynamics and what they may mean for lenders, investors, and borrowers as the system evolves.

‍