S&P says VantageScore 4.0 could work for MBS ratings

Standard & Poor's has found that VantageScore 4.0 and classic FICO produce similar enough results when used to assess credit risk that mapping between them could be done to incorporate the former into rating models based on the latter.

To test this, S&P modeled theoretical loss coverage and foreclosure frequency rates for a broad range of ratings and made comparisons between classic FICO and three different matches between scores: average, median and defaults within two years of origination.

Across this range, starting at the top investment-grade AAA and extending through to the lowest speculative-grade B rating, the results were closely clustered, according to S&P's report.

The AAA results illustrate the similar results found: classic FICO produced rates of 9.25% and 16.49%, respectively, while the three VantageScore-matched methods came in at 9.13%/16.30% (average), 8.97%/16.02% (median) and 9.31%/16.57% (default). The narrow spread across all four methods suggests the two are comparable enough to make adapting the model for VantageScore possible.

"The differences in loss output across these three mapping approaches relative to employing the actual FICO scores are minor," S&P wrote.

S&P Global Ratings also examined in its report how much variation there might be in loss output if the higher of the two scores was picked and found it was "about 10%."

Based on such findings, S&P said it may be possible for it to rate securitized mortgages with 4.0. That could pave the way for more use of the metric, which Fannie Mae and Freddie Mac recently rolled out to lenders. 

"We may apply a mapping approach to assess loans originated with VantageScore 4.0 within our existing credit model,"  the rating agency wrote.

S&P randomly ran 10,000 loans from each of the vintages across 2018-2022 through its model to estimate losses. The total set of data contained 50,000 loans. Mapping was based on summary statistics from 13 million government-sponsored enterprise loans.

The rating agency's model is largely calibrated to Great FInancial Crisis data, but available FICO classic and VantageScore information does not cover this period, S&P noted.

The two GSEs' standard securities do not typically require ratings because they have implicit government backing and are in conservatorship. However, there are exceptions, such as credit-risk transfer securities the enterprises sell to private investors.

The broader mortgage securities market tends to adopt some of the GSEs' practices, and VantageScore's use in rating could give home lenders more reason to consider 4.0 as an option it could use in a wider range of loan scenarios.

VantageScore 4.0 has been used in the private mortgage market but there are questions about whether it has broad acceptance as a standalone credit metric in that context. Vantagescore is more commonly used in other contexts, such as Fitch's unsecured consumer loan ratings.

Cherry picking and the tri-merge

S&P acknowledged that the GSEs have taken at least one step related to addressing risks related to cherry picking the highest score.

"Loans delivered cannot commingle scoring models at the borrower level," the rating agency noted. "Therefore, all borrowers on the loan would need to use the same scoring model, which would limit bias (at least for multi-borrower loans)."

However, the rating agency said that move does not assuage its concerns.

"It is not yet apparent what delivery/origination specifications are (or will be) in place to prevent market participants from cherry picking the best scores," S&P wrote.

A modifier could potentially be used to address this concern, according to the report.

The rating agency also said it was keeping an eye on the possibility that variations in the tri-merge for credit reports could resurface. 

"We will continue to assess the impact potential changes may have in light of the Mortgage Bankers Association's recent proposal to require only a single bureau score for loans delivered to the GSEs if the score is 700 or higher," S&P wrote.

S&P previously found that scores are similar whether they are from two or three credit bureaus when Fannie and Freddie's oversight agency considered allowing a bi-merge during the Biden administration. 

Some subsequent reports from the bureaus found otherwise, with the MBA subsequently challenging that notion by citing figures showing less variation for higher scores. The bureaus still oppose the idea based on the lack of data.

A recent response to a Housing Policy Council Freedom of Information Act request found the GSEs' oversight agency had found the bi-merge less risky than a single score. Others like the MBA have said the single score reduces the risk of gaming.

The GSEs' current oversight chief, Bill Pulte, decided to move ahead with adding score competition with the tri-merge due to complexity changing it, and possibly because it was not within the legislative mandate driving adoption of VantageScore and a more advanced FICO.

However, Pulte has also expressed concern about the lack of competition the bureaus face. The bureaus says differences in their reports show that competition does exist.