Another volley has been fired in what has become “The Great Credit Score War of 2025,” this time by a conservative think tank which says VantageScore’s claims that its 4.0 model is far more predictive than the Classic FICO score are “overstated.”
“While VantageScore 4.0 has a marginal advantage over Classic FICO in capturing high-risk loans within the bottom risk decile,” the study from the American Enterprise Institute (AEI) found, “the difference is relatively small.”
“Across the full sample, Classic FICO performs as well as, and in some cases better than, VantageScore 4.0 on several key predictive measures,” the report determined.
In a recent white paper, VantageScore claimed that its 4.0 model identifies 13.3% more defaults across all loan types than Classic FICO and delivers up to 3.8% more predictive lift — a credit scoring model's ability to predict a consumer's future creditworthiness accurately.
In July, the Federal Housing Finance Agency announced that lenders who sell mortgages to Fannie Mae and Freddie Mac can opt to use VantageScore 4.0 instead of FICO Classic credit scoring.
But the think tank’s analysis found the reported advantage comes from what it calls “methodological inconsistencies and selection bias.” If those issues are corrected, “the purported performance advantage of VantageScore 4.0 largely disappears,” the report argued.
In response, a VantageScore spokesperson said AEI’s study's conclusions are “false and should not be relied upon by industry participants.”
The spokesperson also claimed the AEI authors “do not have quantitative analytical educational backgrounds” and relied upon an “outdated and inaccurate” analytical tool used by FICO.
“Numerous independent analyses from large reputable organizations, including Bloomberg Intelligence, Bank of America, JPMorgan Chase, KBRA, and other experienced industry participants have unanimously found VantageScore 4.0 to be substantially more predictive of mortgage defaults than FICO Classic,” the VantageScore spokesperson said.
One of the irregularities AEI found in VantageScore’s white paper is that it compares a tri-merge 4.0 score with a Classic FICO score derived using the GSE’s current aggregation methodology as applied in Fannie Mae’s Loan Performance data. “The result is a fundamental mismatch that invalidates the comparison,” AEI maintained.
A second flaw in the VantageScore paper: It focused on a “narrowly drawn subgroup of loans,” according to AEI.
The VantageScore white paper’s sample consisted of loans originated in 2019 with a Classic FICO score of 720 or less — the goal being to assess each model’s predictive power for borrowers who are more vulnerable to default during a stressed event like the pandemic.
But AEI argued that by restricting that “narrow” cross-section of Classic FICO scores (from 620 to 720) while using the entire range of VantageScore 4.0 scores (from 383 to 850), the two scores are unevenly matched.
“This asymmetrical filtering creates an unfair comparison,” the think tank contended. After correcting for these methodologies, AEI said it found “significantly different” results than those reported by VantageScore.
VantageScore said 4.0 outperforms Classic FICO in predicting defaults pre-pandemic by 48.5%. But AEI found a 38.5% performance lift for Classic FICO vs. 37% for VantageScore 4.0. “This yields a 4.1% lift in favor of Classic FICO over VantageScore 4.0, reversing the conclusions from VantageScore’s white paper,” the think tank wrote.
Using a second statistical method and comparing all loans originated in 2019, AEI also found that Classic FICO delivers a better result, “again reversing the conclusions from VantageScore’s white paper.”
AEI concluded that both VantageScore 4.0 and Classic FICO are effective in identifying high-risk loans, with only marginal differences between the two.
“The reported advantages of VantageScore 4.0 largely disappear once two major methodological flaws are corrected,” AEI stated. “When both scores are compared on equal terms, Classic FICO performs as well as — or even better than — VantageScore 4.0 across key predictive metrics.”
