Agency MBS market faces repricing risk from shift to VantageScore 4.0

While both scoring models may yield similar numerical scores, they differ in their methodology. Classic FICO offers a point-in-time snapshot of credit utilization, whereas VantageScore 4.0 incorporates 24-month trended debt paydown behavior and factors in non-traditional payments like rent and utilities. This difference can lead to borrowers receiving higher scores under VS4, potentially creating adverse selection as lenders choose the model that optimizes borrower eligibility or loan pricing.

For the specified pool market, where investors pay premiums for predictable prepayment behavior, this model variance poses a challenge. Traditional econometric models have relied on Classic FICO scores to predict refinance propensity. The market currently lacks extensive empirical data on how pools with high concentrations of VS4 loans will perform across different interest rate cycles. This uncertainty could lead to a compression of the premiums investors are willing to pay for credit-based specified pools.

As market participants gain visibility into VS4 loan concentrations within specific pools, the market is expected to bifurcate. Trading screens may begin to show tiered pay-up quotes based on VS4 concentration limits. Pools composed entirely of Classic FICO loans may command a premium due to model continuity and historical performance data. Capital may also rotate towards structural specified pools, such as those with low loan balances or in high-cost states, where prepayment protection is tied to physical loan traits rather than credit scores.

Industry professionals are urged to adapt. Capital markets desks must monitor credit score model delivery in their pipelines to avoid inadvertently delivering blended loans into specified pools, which could negatively impact execution prices. Building algorithmic pooling strategies that account for VS4 concentrations is crucial. Buy-side investors and analysts are advised to update their risk engines to track loan-level credit model identifiers.