Credit scores similar whether from two or three bureaus S&P finds

The changes for credit score use proposed by the Federal Housing Finance Agency for the government-sponsored enterprises will only cause a small variation from the results produced by the current methodology employed, a Standard & Poor's analysis found.

FHFA Director Sandra Thompson announced a switch to use of both the FICO 10T and VantageScore 4.0 models from the current Classic FICO at the Mortgage Bankers Association's annual convention in October. The new system also calls for the use of the mean average of scores generated from two of the three credit bureaus versus the median of all three — Equifax, Experian and TransUnion.

However in the report, S&P only identified the bureaus as A, B or C.

More recently, at the MBA's Secondary and Capital Markets Conference in May, Deputy Director Naa Awaa Tagoe gave an update that put the deadline for the transition for Fannie Mae and Freddie Mac to update their underwriting systems at the end of 2025. The mortgage business is also dealing with a controversial pricing change by FICO that created volume discounts for large users.

S&P conducted an analysis from a sampling of the mortgages contained in securitization transactions it rated.

The change in credit scoring models is an improvement "because the new methods incorporate information such as rent, utilities and telecom payments, which could be used to infer payment behavior patterns, as well as trended data and presumably the identification of revolving versus transacting credit in order to acquire greater insight into borrower behavior patterns," said the S&P report, Residential Mortgage Credit Score Snapshot: When Three Bureaus Become Two.

The rating agency looked at the scores generated individually by each of the three bureaus, and found no "marked difference" between them. Furthermore, none of the three either consistently generated the highest or lowest score.

In the next step, S&P did the tri-merge versus bi-merge comparison; it also noted that the FHFA previously did its own analysis that found a minor variation between the median of the three bureau reports versus the average of the two.

"The difference between an average bi-merge and a median tri-merge is small (within about a point), and there was no systematic bias in terms of which was larger," the S&P report said. "Note, however, that these are aggregate statistics that do not capture the variability that might come about at a loan level within the sample."

The study found that nearly half of the time, the difference between the two methodologies was within 5 points, while approximately one-quarter was between 5 and 10, and another quarter was over 10.

The biggest differences in score were at the fringes of both the higher and lower ends, the analysis found.

"The average difference between high and low scores in a tri-merge is about 20 points for the 800-825 bucket, whereas for the 550-575 bucket the difference is about 45 points," S&P reported. "One explanation could be that data for certain adverse events may be restricted to fewer than three bureaus, due perhaps to specific geographies."

The other possibility is that higher credit scores will either compress or cluster when they are at the maximum.

This analysis also found the possibility of increased variation from those results at the loan level could arise in smaller securitization pools.

Overall, the switch to a bi-merge could lead to innovation and competition among the three credit bureaus, the report concluded.

The S&P report acknowledges many of the concerns raised regarding the potential negative consequences of the switch to using scores from two repositories versus the current three, said Joe Mellman, senior vice president, mortgage business for TransUnion.

He noted that 9.6 million consumers got a mortgage in 2022. "If a quarter of them see a 10-point difference or greater in their score, per the report, that means up to 2.4 million consumers potentially paying hundreds of dollars a month more in interest, or getting denied for a mortgage altogether," Mellman said. "Using less data to underwrite mortgages, particularly when we're starting to see delinquencies tick up, will ultimately hurt consumers and taxpayers, who are on the hook for bad mortgages."

Meanwhile, one of the other bureaus, Experian, said it was excited by the broad benefit from the use of expanded data — which can enhance financial inclusion — and more modern scorecards, it said in a statement.

"This is one of the reasons we've championed the use of expanded data sources, including consumer permissioned data obtained through open banking technologies," the Experian statement said. "When this information is leveraged with traditional credit bureau data, it has the power to open the door for more equitable access to fair and affordable credit."

As for the private label securitization market, it is unclear when, or even if, this segment will adopt the FHFA proposal — or some version of it.

"Because most originators of non-agency products typically focus on agency lending as well, the systems employed by these entities may include GSE-style conventional underwriting that may carry over to other mortgage products," the S&P report commented.