- 2 days ago
- #creditscoring
- #mortgagelending
- #fico10t
In this conversation with HousingWire’s Allison LaForgia, Ethan Dornhelm, Vice President of Scores and Predictive Analytics at FICO, explains why the mortgage industry's long-awaited mortgage credit modernization effort has entered a pivotal new phase.
From the release of GSE historical data to the advantages of FICO Score 10T, Dornhelm outlines why lenders now have the opportunity to evaluate the industry's next generation of credit scoring.
#CreditScoring #MortgageLending #FICO10T
From the release of GSE historical data to the advantages of FICO Score 10T, Dornhelm outlines why lenders now have the opportunity to evaluate the industry's next generation of credit scoring.
#CreditScoring #MortgageLending #FICO10T
Category
🗞
NewsTranscript
00:03From HousingWire in Dallas, Texas, I'm Alison LaForgia, sitting down with Ethan Dornhelm,
00:08Head of Scores Analytics at FICO. Ethan, thank you for joining me today.
00:13Thank you so much, Alison.
00:15So, Ethan, you have been at FICO for over 25 years, and the FICO score has long been considered
00:22the gold standard for risk decisioning and mortgage lending. Looking back over that time,
00:28how has credit scoring in the mortgage space evolved, and where do you see it heading next?
00:34Well, when I think about the last 25 years, I think what's most interesting is that the
00:40fundamental question that lenders are trying to answer with credit scores hasn't changed. It's
00:45what level of risk does the borrower represent? Are they likely to repay a mortgage as agreed?
00:53What has changed over time, I would say, what has evolved over time in credit scoring is that
00:58we've released successively new generations of the FICO score, culminating with our latest and
01:04greatest score, FICO score, 10T. And with respect to evolution in the mortgage space, that evolution
01:11is happening now. Now is the time when the industry is considering credit score modernization and the
01:17adoption of these new scores. So for decades now, classic FICO was underpinning the safety and soundness
01:23in the mortgage ecosystem, right? Providing that consistent, trusted metric to lenders and investors
01:30in the GSEs. And right now we are at the precipice of adoption of these more modern scores, including
01:36FICO 10T, which is again, our most powerful score to date. You just mentioned the FICO score 10T.
01:43The latest development in credit modernization effort was the release of the FICO score 10T historical
01:49data by the GSEs. Why is this release such an important milestone in the credit score modernization
01:56effort? Well, we're very excited. This release of this data has been a long time in the works.
02:03FICO submitted the FICO score 10T for consideration as part of the FHFA's credit score modernization
02:10effort back in 2020. The model was approved by the FHFA in 2022. And only now here in 2026 have
02:20we seen
02:20the release of this historical data on 50 million loans that were acquired by the GSEs over the past
02:3112 plus years, dating back to 2013. So it's a very rich source of real world GSE data. It's widely
02:39accessible and it's going to allow the industry to assess this historical data for themselves,
02:47not just from the standpoint of comparing classic FICO advantage score to FICO 10T on kind of a
02:52standalone basis, but also to examine how these new modern scores will impact other analytics,
03:00other decision rules that lenders and investors have. And not just evaluate the effectiveness of
03:07the models, but also understand how they need to maybe recalibrate those other analytics, those
03:12other rules to account for these new scores. So this is why this is such a significant milestone
03:18because it's such a large data set. It's the first of its kind that has all three of the credit
03:23scores
03:24being considered for use in mortgage on it. And it's large enough that it allows a lot of slicing
03:30and dicing of the data by different borrower characteristics, DTI, LTV, purchase versus
03:36refinance, first time home buyers, all of these different things with large enough counts that
03:41you can slice and dice and still have meaningful insights, statistically valid insights off of the
03:46data. Let's take a little bit more into, as people start working through the historical data set,
03:52what does that data set include and what nuances should they be aware of as they interpret findings
04:00on predictive performance? Right. Well, as I mentioned, this has some 50 million or so loans acquired by
04:08the GSEs dating back to 2013. It not only has characteristics about those borrowers, those applicants
04:16at the time of origination, but also includes their subsequent behavior payment on those mortgage
04:23accounts. Did they continue paying as agreed? Did they go delinquent? Did they prepay? And that, again, all that
04:29information together really allows for a very robust assessment of various scores that are being considered in
04:38the market and how well those help with the understanding of a given borrower's repayment risk. That said, there are
04:45some nuances in the data that need to be considered. And I think first and foremost, something we've been
04:51really trying to emphasize to the industry as they analyze this data is that this is a truncated data set.
04:58And
04:58it's truncated in the sense that for the majority of these 12 plus years of data that we have in
05:04this
05:04sample, the GSEs had a policy in effect that was effectively classic FICO scores below 620 were not being
05:13acquired by the GSEs. So as a result, there's virtually no loans below a classic FICO score of 620 in
05:21this data.
05:21And the reason that matters is, as I mentioned, one of the things we expect the industry to be using
05:26this
05:26data for is to calibrate, to understand, well, how does a 620 classic FICO correspond to risk of repayment with
05:34FICO-10-T, for example. And what we know to be true, because we've analyzed it on non-truncated data
05:41sets,
05:42is that the 620 to 629 band, for example, of FICO-10-T, about 30% of that population scores
05:50below 620
05:51with classic FICO. So what that means is in this GSE data, as important as it is, as rich as
05:58it is for
05:58analysis, it is true that you're missing part of the story when you're looking at that calibration
06:04analysis. Some 30%, in some instances, of the actual applicants who will come in at a certain FICO-10-T
06:11score band are simply not there in the data because they were below 620 in their classic FICO.
06:18I think one other thing we would mention with regards to nuances of the data that's worth calling
06:23out is the data goes back to 2013. It covers some different time periods, yes, but in terms of a
06:29true
06:29stress test time period, I think you could argue that COVID may represent a decent stress test. However,
06:38we know that was a pretty confounded time as well in terms of credit risk patterns because of the
06:43unprecedented amounts of intervention we saw both from public sources like the government stimulus,
06:50CARES Act mandated forbearance, as well as private, you know, sources as well, lenders were offering
06:56payment holidays and the like. So another caveat, you know, it's so important that these scores and
07:01we build the FICO score to hold up over time to be really robust and an accurate assessment of credit
07:08risk through the economic cycle. It's important to be mindful that at least on this data, this 50 million
07:13sample, we don't have a truly, truly stressed event like, for example, the Great Recession data. So
07:21that's just something to be mindful of as people analyze this data and consider how to how to consider
07:26how the scores are performing during stress periods. Ethan, let's talk about findings for a moment.
07:33As I know, there have been several notable analyses published recently. What does the data show us?
07:42Well, great question. And I think in sum, what it shows is that FICO 10T is the most powerful
07:49credit score available in the mortgage space today. One of the most recent analyses that have been
07:55published by trusted actuary firm Milliman concluded that they looked at over 46 million of the loans in
08:03this GSE data set. And they evaluated the effectiveness of the three scores in question,
08:09classic FICO Vantage and FICO 10T across the exact performance measures that the GSEs themselves
08:16articulated as the way to evaluate credit score accuracy as part of this modernization effort.
08:22And what they found across all of those performance measures, across all of the different vintages captured
08:28in this GSE data set across different borrower subpopulations, was that FICO 10T outperformed
08:34Vantage score 4.0 across the board. And interestingly enough, some of the most notable lift that they
08:43found in the Milliman study was on the most recent vintages, the vintage like 2023. That was where FICO
08:5010T was actually outperforming Vantage score 4 by an even larger margin. And I think that just speaks to,
08:55again, one of the core principles that we have at FICO is building these scores to last. We know,
09:01we've seen it with classic FICO, these scores can be used for decades in the mortgage space. So we
09:07really build them with an eye towards ensuring that they remain robust and reliable over time.
09:12And I think that's part of the reason why we're seeing the FICO 10T score hold up so well over
09:17time
09:17when other scores may be deteriorating and the gap between the two scores, therefore, is growing.
09:23So those are numbers. But what does this all mean in practice? Well, for lenders, a stronger
09:29predictive score supports more informed lending decisions, more informed risk management. For
09:34consumers, it can translate into increased credit access and more sustainable paths to homeownership.
09:40And for investors, it promotes confidence. It promotes liquidity and stability in the secondary
09:45mortgage market. So, again, we're really excited by these early results we're seeing. We know more people are
09:51going to do the analysis. They're going to publish those results. And we know it's going to continue
09:56to show that FICO 10T is the most predictive score out there. Now, there's a narrative out there that
10:03expanding credit access is primarily about scoring more consumers. How does FICO view the relationship
10:10between predictive accuracy and creating opportunities for more borrowers to access credit?
10:17Well, you know, first of all, I think one thing that is important to emphasize is it shouldn't be
10:24presented as a tradeoff, like the tradeoff between accuracy and scoring more borrowers. To us, actually,
10:30better access is driven by more accurate models. When lenders can assess risk more precisely,
10:37they have the confidence to approve borrowers that they otherwise may have missed. So when we talk about
10:44expanding credit access to us, it's not simply about scoring more people by lowering our scoring standards.
10:50It's about making better lending decisions for more people. And at FICO, we've always viewed accuracy as
10:58a foundation of financial inclusion. The more accurately a score predicts credit risk, the more
11:04confidence a lender has to approve borrowers. And I think it's also important to distinguish between having a score
11:10and qualifying for a mortgage. Those aren't the same thing. Simply delivering a score on a consumer doesn't
11:16by any means imply that they're going to qualify or that they're ready for a mortgage. Our competitor claims
11:23that they're able to score more consumers because they have access to more data. And that's simply not true.
11:30FICO 10T and Vantage Score 4 have access to and utilize the exact same traditional credit bureau data
11:38report. They leverage trended data, which is a step beyond what classic FICO had access to. They both
11:44leverage rental information when it's available in the file as well, giving consumers an opportunity
11:50to build their credit that way. But again, both Vantage Score and FICO 10T leverage the exact same data.
11:56So we agree that bringing more data to bear is the right path to financial inclusion, bringing more
12:04data to bear on these consumers that otherwise have such sparse, such little information in their credit
12:10file that they don't meet the classic FICO scoring standards. We're not going to lower those standards.
12:15What we've done over the last decade is invested a lot of time, a lot of resources into studying
12:22alternative data sources, data beyond what's found in the traditional credit file that we found was
12:28most predictive, most helpful in driving accurate assessments of credit risk on those borrowers
12:34who don't have enough of that traditional credit history. And so that's what we believe to be the safe
12:40and sound way to score more borrowers, to do a better job of scoring more borrowers accurately enough
12:47that the industry will embrace those scores, be able to use those scores to drive greater access
12:52to credit for millions of consumers. One last question for you, Ethan. As the industry begins
13:00reviewing and analyzing this data, what would you encourage lenders and other stakeholders to focus
13:06on? And what are the next steps in the transition toward even broader adoption of FICO Score 10T?
13:13Well, I think the next step is for lenders to move from awareness of this data to hands-on evaluation
13:21of the data. And that's exactly what the publication of the 10T data by the GSEs enables, right? Multiple
13:29pathways exist today to test, validate, and adopt FICO 10T. And we would just call out as they do that
13:35to be sure to keep in mind some of those nuances that I mentioned earlier, the truncation of the data,
13:41the presence or lack thereof of true stress test periods in the data. FICO, it would be delighted to
13:49talk to anyone in the mortgage industry who has questions or wants to discuss different ways to
13:55ensure that their testing and validation of these scores is robust. But beyond just analyzing that public
14:01GSE data, of course, lenders also should be pulling FICO Score 10T alongside of Classic FICO today. That way,
14:09they can understand how these models compare across their own portfolios, across their own borrower
14:14footprints. And FICO is supporting that effort with our FICO Score 10T free access program. And that
14:23enables lenders to receive FICO Score 10T alongside Classic FICO at no additional cost from FICO. And we
14:32we don't believe lenders should have to pay to evaluate new scores, which is exactly why the FICO 10T
14:38free access program was created. We've already had over 70 lenders sign up for the program, representing
14:44well over $500 billion worth of annual mortgage originations. And we're very confident that in
14:50the coming months, we're going to sign up many more, especially as people test the data that the GSEs
14:55have published and see how effective, how predictive, how powerful FICO 10T is. It's clear the industry
15:03is looking to modernize and take advantage of the best tools in the market. And we think FICO 10T
15:08is the natural next step along that path. Ethan, thank you so much for joining me today.
15:15And thank you for talking to me about FICO Score 10T. Thank you, Alison. It was a pleasure.
15:25Thank you, Alison.
Comments