Panel Tackles FICO 10T, VantageScore 4.0 Differences, Implementation

(From left: David Battany, Ethan Dornhelm, Sasha Hewlett, Bob Johnson and Rikard Bandebo, by Anneliese Mahoney)

NEW YORK–Four years after the Federal Housing Finance Agency approved FICO Score 10T and VantageScore 4.0, a panel convened at the Mortgage Bankers Association’s Secondary and Capital Markets conference to discuss progress on the credit score modernization effort.

Featuring a representative each from FICO and VantageScore, there were some limited points of agreement. There are, after all, some similarities between the two new scores, the panelists said. Both FICO Score 10T and VantageScore 4.0, along with Classic FICO, are calculated using credit file data, are built to predict the likelihood of default over two years and provide a 300-850 score range.

But, they aren’t the same scales within that range, noted Rikard Bandebo, executive vice president, chief strategy officer and chief economist at VantageScore. “It’s not quite like Celsius from Fahrenheit, but it’s a good analogy,” Bandebo said, although panel moderator Sasha Hewlett, MBA associate vice president of secondary and capital markets, pointed out it’s not a direct ratio, but rather more dynamic.

Ethan Dornhelm, vice president and head of scores analytics at FICO, also shared his thoughts on some of the differences. For one, VantageScore uses the same algorithm across Equifax, Experian and TransUnion. That provides consistency, Bandebo said, but Dornhelm pointed to the logic behind FICO’s different algorithms across the three bureaus.

“Our standpoint is that it’s worth the investment in time and resources to actually build bespoke models for each bureau separately, because each bureau does not have the exact same data,” he stated. “By building separate models for each, we’re able to ensure we’ve tailored those models to maximize the signal that we can extract from that data.”

Bandebo highlighted that VantageScore has a larger scoreable population, with about 33 million more people–some of whom will be mortgage-eligible–kicking off a lively back and forth.

Dornhelm pointed to FICO’s exclusions as “safeguards,” cautioning that the many of the people who don’t fall under FICO’s scoring criteria, “are consumers who haven’t had an update to their credit file in six or more years, since pre-COVID.”

“We understand that a lot of folks in this room, especially in the secondary market, they only receive that three-digit score, they don’t get the credit data that goes along with it,” Dornhelm said. “So you have no way of knowing as an investor, is this based on a robust credit file?”

In response, Bandebo pointed out that “when you look at the newly scored population and you look at their default rates, and you compare them to those that are scored by both models, they overlap almost completely,” he said. “So that wouldn’t happen if they were more risky than the other population.”

Both scores use trended data–but Bandebo and Dornhelm disagreed on how big an impact that has. Dornhelm described it as providing a little more granularity and a “little sharper risk assessment than can be derived solely off of static data.”

Bandebo argued it’s quite a big difference, and that it can help improve the predictiveness of risk, giving the example of two consumers who might have similar moment-in-time snapshots, but one has balances and utilization trending upward and the other does not.

Dornhelm pointed out that in the development of FICO 10T, it was decided to build a version with trended data and one without, so users can decide.

There’s another important component to the credit score modernization effort, Hewlett pointed out. How are the early stages of use by lenders going?

David Battany, executive vice president of capital markets for Guild Mortgage Co., cautioned that lenders that are in the GSE’s limited rollout are not allowed to share certain information, so thoughts will have to be generic. “We look at both of these new models coming out as new, updated models that are very much worth evaluating, and the way that you do that is you collect data,” he said, as well as the need to conduct comparisons on the loans across the scores. “So we’re in the very early stages of collecting this data.”

“Taking a look at what you have in your existing portfolio–and we have 4 million or so loans within our servicing portfolio–and looking at how the VantageScore and the FICO score compare, and how the delinquency is trending over time,” described Bob Johnson, head of originations with Newrez. Newrez recently did a pilot with Freddie Mac to deliver loans that featured VantageScore.

There are challenges, though, said Johnson and Battany, including from a technology perspective.

And “there’s a whole ton of policy questions; how we as an industry do this,” Battany said. “So, I would just frame it as we’re very early in the history of learning how to compare the models to each other, pros and cons of each model, whether borrowers can benefit or not.”

“I would say for lenders in the room, our top focus should be to get the highest quality score possible, the score that’s the most predictive of default,” said Battany. “And second is to do this in a process that lowers costs for consumers.”