Wednesday, July 8, 2026
7.8.26 Prepayment Data; FICO’s Ethan Dornhelm on Credit Score Modernization; Geopolitical Tension
0:00
0:00
Listen on
YouTube
Today’s episode includes a discussion on Agency mortgage-backed security (MBS) prepayment activity in the wake of QE4 lows. Plus, Robbie interviews FICO’s Ethan Dornhelm on how credit score modernization is giving lenders and market participants the ability to evaluate long-term predictive performance and expand responsible access to credit while preserving safety and soundness. And we close with the focus on geopolitical tensions versus inflation versus economic data.
This week’s podcasts are sponsored by FICO. As the industry's most predictive credit score, FICO Score 10T combines proven performance with deeper insight into borrower behavior to help support a stronger and more resilient housing finance system.
The Chrisman Commentary is your go-to daily mortgage news podcast, where industry insights meet expert analysis. Hosted by Robbie Chrisman, this podcast delivers the latest updates on mortgage rates, capital markets, and the forces shaping the housing finance landscape. Whether you're a seasoned professional or just looking to stay informed, you'll get clear, concise breakdowns of market trends and economic shifts that impact the mortgage world.
Full Transcript
Show full episode transcript
0:03) Welcome to the Chrisman Commentary Daily Mortgage News Podcast. I'm your host, Robbie Chrisman. (0:09) Topics on today's episode include agency MBS prepayment activity, why geopolitics are (0:15) driving rates higher again, and my interview with FICO's Ethan Dornhelm on how credit (0:20) score modernization is giving lenders and market participants the ability to evaluate (0:24) long-term predictive performance and expand responsible access to credit while preserving (0:29) safety and soundness.
Here, take a listen to a little preview. (0:33) Whatever company exists, with Chrisman LLC, our guiding light is to provide value to the (0:39) mortgage industry. I know a big part of your role, especially with FICO score 10T, is expanding (0:45) access to credit while maintaining safety and soundness.
And so can you talk about how (0:50) you're helping lenders achieve both of those objectives I just mentioned? (0:54) Yeah, absolutely, Robby. I mean, our focus is to expand access to credit without lowering (1:01) standards, maintaining the accuracy, the stability, the trust that investors rely on across economic (1:07) cycles. So in our view, the most predictive score promotes that safe and sound lending (1:12) and helps more consumers qualify for mortgages at the lowest interest rate.
And the way (1:17) that works is a more predictive score pushes more of the consumers who are likely to pay (1:24) the mortgage up the score spectrum. And those who are most likely to default are pushed (1:30) further down the score spectrum. So a more predictive score can be used to underwrite (1:35) more borrowers while still keeping the overall portfolio risk the same or even lowering it (1:41) a little.
So we've seen use cases in the mortgage space where relative to classic FICO, loan (1:46) approval rates could be expanded by 5 percent without taking on any additional risk simply (1:51) by virtue of this more predictive score tuned to more recent risk patterns, leveraging (1:57) additional data like trended credit bureau data and when available in the traditional (2:02) credit file rental payment history. It just gives lenders a more complete view of how (2:07) consumers are managing their credit. (2:11) In a market where every loan counts, FICO Score 10T lets lenders say yes to more borrowers (2:17) without added risk.
As the industry's most predictive credit score, FICO Score 10T combines (2:22) proven performance with deeper insight into borrower behavior to help support a stronger (2:26) and more resilient housing finance system. FICO has set the standard for decades and (2:31) I'm grateful for their support of today's podcast and the conversations that help bring (2:36) mortgage professionals together. (2:41) Bloomberg is reporting that Lennar is embroiled in a lawsuit with a Native American tribe (2:45) in Florida.
Lennar allegedly constructed homes on tribal land which are now uninhabitable. (2:50) The story has been around for almost a year, but apparently things are heating up and impacting (2:54) the stock price. (2:56) Before we get into politics, remember that mortgage prices are derived from supply and (3:00) demand.
Trading activity surged yesterday as heavy Class A dollar roll volume, pending (3:06) settlements and a wave of late-day panic locking following geopolitical headlines with attacks (3:11) near the Strait of Hormuz and subsequent U.S. retaliatory strikes drove elevated mortgage-backed (3:16) securities volume and hedged demand. (3:18) Agency mortgage-backed security prepayment activity has remained remarkably stable over (3:22) the past month despite meek seasonal support, corroborating the view that refinancing activity (3:27) is likely to stay subdued. Aggregate Fannie Mae 30-year speeds rose just 2% month over (3:32) month to an 8.4% one-month CPR, the first increase in three months and consistent with (3:37) seasonal patterns.
10-year, 15-year and 20-year mortgages posted somewhat larger gains. (3:43) Year over year, prepayment speeds continue to recover from their post-QE4 lows, but the (3:48) increase largely reflects normalization from exceptionally weak levels rather than a material (3:53) shift in borrower behavior. Deeply out-of-the-money coupons continue to show a slight pickup (4:01) in activity, reflecting a gradual thaw among borrowers locked into ultra-low pandemic-era (4:06) mortgage rates, while higher coupon refinanceable pools generally slowed.
With only about 7% (4:12) of outstanding 30-year borrowers currently holding a meaningful refinance incentive, (4:16) the opportunity set remains limited, but still provides a targeted source of production (4:20) for lenders. Modestly faster prepayments are expected this month primarily due to calendar (4:26) effects rather than any fundamental changes in refinancing dynamics. (4:30) Treasury's mortgage-backed security prices sold off sharply yesterday after oil prices (4:35) jumped late in the session, reinforced by the release of firmer June consumer inflation (4:39) expectations, sending the 10-year and 30-year Treasury yields to their highest levels in (4:43) nearly four weeks despite a well-received three-year Treasury auction.
While today's (4:48) FOMC minutes and ongoing dollar roll activity should keep markets active, yesterday's (4:52) late-day hedging may make today's results harder to interpret by blurring the impact (4:57) of normal loan activity with the effects of defensive hedge adjustments. (5:01) Even with geopolitical tensions briefly rattling markets, investors remain focused on next (5:06) week's consumer prices report, Treasury auctions, and the release of minutes from (5:11) Chair Warsh's first FOMC meeting, which will come out later today, and could offer (5:15) important insight into the Fed's evolving policy approach after removing much of its (5:19) forward guidance. With softer labor market data, easing energy inflation, and expectations (5:24) that the Fed will remain patient, expectations of the next realistic opportunity for a rate (5:31) move have shifted to September, leaving Treasury yields driven more by positioning—that's (5:35) the trades and risk exposures they're holding before new information arrives—supply, and (5:40) incoming inflation data than by geopolitical developments.
(5:46) For today's interview, I wanted to welcome to the show FICO's Ethan Dornhelm to talk (5:49) about how credit score modernization is giving lenders and market participants the ability (5:53) to evaluate long-term predictive performance and expand responsible access to credit while (5:58) preserving safety and soundness. (6:00) He's head of score analytics at FICO and leads the research and analytic development (6:03) of FICO scores globally. He's responsible for maintaining the predictive power of the (6:08) core FICO score product line, as well as research and development of new scoring products, analytic (6:12) methodologies, and alternative data sources.
Most recently, his team launched the newest (6:18) FICO Score 10 suite, as well as the alternative data-driven solutions FICO Score XD2 and the (6:24) Ultra FICO Score. Ethan's team also develops the analytic features available to millions (6:29) of consumers via FICO's B2C solutions, MyFICO.com, and FICO Score Open Access. (6:35) Obviously, all week listeners have heard me talk about how FICO Score 10T lets lenders (6:40) say yes to more borrowers without added risk.
You're head of scores analytics at FICO, (6:46) which to me is an incredibly interesting job title. And yes, I read off your bio, which (6:52) you probably heard a thousand times at conferences or in speaking. But how would you describe (6:56) your job, day-to-day, larger picture, and what do you enjoy so much about it? (7:02) Yeah, thanks, Robbie.
First of all, so glad to be talking with you today. The primary (7:06) focus of my job is to be ensuring that the score that we're supporting in the market, (7:14) we've provided as much insights as possible to our clients and investors and other stakeholders (7:20) as far as how they can most effectively use those scores to, as you say, say yes in a safe (7:26) and sound way to as many borrowers as possible. We're also always having that eye to the future, (7:32) looking at what's next, what is the next score we're going to build that's going to be even (7:36) more predictive, leverage even more additional data sources beyond just what's found in the (7:41) credit file.
So a lot of research and development coupled with a lot of product support of what's (7:46) already out there to make sure it's getting used in the best possible way for the financial ecosystem. (7:51) It's been really neat to see predictive scoring models get that much better. And so I guess I (7:56) want to thank you personally for doing that.
We've seen here recently FHFA and the GSEs, (8:02) they made more than a decade of historical FICO score 10T data available. And from your (8:09) perspective, why do you feel like this is such an important milestone in the credit (8:12) score modernization effort? We're really excited about this. You know, release of this data has (8:18) been a long time in the works.
We submitted FICO 10T for consideration as part of the FHFA's credit (8:25) score modernization effort back in 2020. And it was actually approved by the FHFA 10T was in late (8:32) 2022. So it's been a long time coming to get this data published.
The reason why we're so excited (8:37) about this data publishing is because it's really the richest, most widely available data (8:42) set in the conforming mortgage space that will really support the mortgage industry's transition (8:48) to modernized credit scoring. Access to this data allows the mortgage ecosystem to independently (8:54) evaluate the score's performance across 12 plus years of mortgage vintages dating back to 2013. (9:01) And I can tell you, Robbie, you know, I've been working at FICO for 25 years.
I've built a lot (9:06) of FICO scores. I've produced a lot of analyses comparing the relative effectiveness of a new score (9:11) to older scores. The industry values what FICO produces in terms of those collateral that (9:17) shows the score performance, but they always prefer getting to conduct their own analyses.
(9:23) And that's exactly what publication of this data enables. Any quantitative insights that you found (9:30) particularly interesting or worth sharing as you went back through these vintages and any data (9:35) points that popped out to you? Well, you know, the data just published, Robbie, late last week. (9:40) So we're continuing to work through the data.
And one of the pillars of FICO's philosophy is (9:46) absolute gold standard in the analytics we produce to market. So we're dotting I's and (9:51) crossing T's. But I can tell you with great confidence, we know what it's going to show.
(9:56) We know when practitioners and others in the industry analyze this data, they're going to (10:02) see that FICO score 10-T is the most predictive score among those currently available for use (10:08) or soon to be available for use in the mortgage space. Yeah, we've thrown out this phrase, most (10:13) predictive credit scoring model, a couple times now. And that's not just us saying it, that's (10:18) prior analyses that they've showed that FICO score 10-T is indeed that.
What will market participants (10:24) be able to learn about the score's performance from this historical data that you've referenced here? (10:29) You know, again, the data really enables lenders, investors, and other housing finance stakeholders (10:34) to evaluate the performance of 10-T using real world GSE loan level information. It's been (10:41) pretty tough sledding getting access to data that has both the Vantage score and the FICO score 10-T (10:46) on it. So to date, I would say folks have only had access to what's available in the GSE data, (10:53) which was classic FICO and Vantage score.
And while we build our FICO scores to last, (10:58) classic FICO was built over 20 years ago. And so we're eager to see 10-T data published (11:04) because that's our latest and greatest score version. It's our best horse in the race, (11:08) so to speak.
So we're eager for mortgage industry practitioners to have the chance to analyze the (11:13) improved accuracy of FICO 10-T, not only point in time, but over time, right, on Vantage as well (11:20) before, during, and after COVID, you know, to be able to analyze the scores across different key (11:25) segments of the population, such as first-time homebuyers or applicants with various levels of (11:31) LTV or debt-to-income, loan purpose, purchase versus refinance. Because this data set is (11:37) comprised of like 50 million loans between Fannie and Freddie combined, there's plenty of data to (11:43) enable that sort of slicing and dicing of this population and still, in most cases, come away (11:48) with statistically meaningful conclusions. Yeah.
So thank you for going through a little bit of (11:53) what the data set includes there. Any nuances for people that they need to be aware of? You mentioned (11:59) mortgage industry practitioners. Any nuances they need to be aware of? I'm sure there's a lot of (12:03) folks that are going to be digging into the data here.
Oh boy. Robbie, how much time do you have? (12:08) As much as you need, Ethan. This is where we can really geek out.
Again, the data set captures (12:14) some 50 million loans acquired by the GSEs since early 2013. It has information about those loans, (12:20) not only how the applicant looked at the time of origination and the associated term of the loan, (12:26) but also subsequent performance on those loans, whether they were paid as agreed or went delinquent, (12:33) whether that loan ultimately prepaid. So there's so many powerful ways and potential ways to (12:39) slice and dice the data, and we're really excited for the industry to start digging in.
(12:43) But I would call out a few watchouts with regards to this data that are certainly worth (12:48) highlighting. For one thing, there's the matter of what's called truncation. This is focused on (12:54) loans that were acquired by the GSEs over the last decade plus.
For the majority of that time, (13:00) the GSEs had this minimum credit score requirement in place, 620. So what that means is almost all (13:07) of these 50 million loans have classic FICO scores above 620. And that matters for score validation, (13:15) yes, but even more importantly for score calibration.
So let me give you an example, (13:20) Ravi. If you're trying to understand on this GSE data, what does the risk of a FICO 10T 620 to 629 (13:28) calibrate to going forward? Well, one thing you have to be mindful of, and we've done the analysis (13:34) on a non-truncated data set that found this, of that population, 30% of those actually have a (13:40) classic FICO below 620. But that 30%, you're not going to see in the GSE loan level data, right? (13:47) Because they wouldn't have been acquired by the GSEs because their classic FICO score was below (13:51) that minimum 620 cut.
So it's just a really important thing for, I think, practitioners to be (13:57) mindful of as they're using this data and using it to calibrate. What's the risk of new scores (14:02) relative to the legacy classic FICO? They need to be mindful of that significant percentage of (14:07) the population, especially in lower score ranges, that they simply can't see in the data. Now, that's (14:13) just one example.
Ravi, would you mind if I keep going? Please, go. I mean, I think another (14:20) important thing to be mindful of when doing analysis on this GSE data is just holding (14:25) things constant in any comparison. Maybe holding everything constant except for the score itself (14:32) that you want to compare the different scores to each other.
And what do I mean by holding (14:36) things constant? Well, for example, the window of repayment that you evaluate the score's accuracy (14:42) in predicting. If you think about it, the 2013 vintages, you have over 10 plus years of subsequent (14:49) repayment behavior on those borrowers. But on the 2023 vintage, you really only have (14:56) something like two years of repayment.
And so when we talk about holding things consistent, (15:00) if someone just throws the 2013 and 2023 vintages together with a measure of how (15:06) well did the score predict the borrower ever going default, you're mixing and matching two (15:10) very different outcome horizons. 2013, it's over 10 years. 2023, it's just two years.
So what we're (15:16) encouraging practitioners to do is make sure you use a consistent measure that will allow an apples (15:21) to apples comparison of the score's effectiveness on different vintages over time. And, you know, (15:26) the GSEs themselves recommended a two-year performance window, which we think is a very (15:31) reasonable definition, not only because we build the FICO score to a two-year definition as well. (15:38) I think the last thing, Ravi, I'd point out with respect to, you know, things to be mindful of in (15:43) this data is there isn't a wealth of stressed periods captured in the GSE data, right? In 2013 (15:48) to 2025, COVID certainly was stressed, but it was such an idiosyncratic period where both the onset (15:55) and the recovery of the economic downturn were sudden and dramatic.
And COVID was also sort of (16:04) affected by unprecedented levels of public and private sector intervention, right? We all are (16:09) aware of the CARES Act and the mandated forbearances in the mortgage space for government-backed loans. (16:15) It's easy to make the case that consumer behavior from this period is confounded and needs careful (16:20) interpretation. And so that's why we emphasize the importance of just being mindful of these things, (16:26) being mindful of truncation, being mindful of being as consistent as possible in applying your (16:32) score comparisons across vintages, and being mindful that there may not be a truly stressed (16:37) period represented in this data.
And as the industry begins reviewing and analyzing this data, (16:44) what would you encourage lenders and other stakeholders to focus on? What are the next steps (16:48) in the transition toward broader adoption of FICO score 10-T? Maybe another way of asking that. (16:54) Well, I think outside of analyzing the GSE data, which we're of course very excited for lenders (16:59) and other stakeholders to do, I think to the extent they're able, pulling both FICO 10-T and (17:06) classic FICO side by side is critical as the industry gears up to operationalize these modern (17:12) scores. So the idea here is it allows you to understand the differences in the difference (17:19) scoring algorithms on the specific portfolio or footprint or data that a given lender has.
And (17:26) you know, I should call out here, the way FICO is trying to enable this is with our FICO score 10-T (17:31) adopter program. So with that program, lenders are able to pull 10-T alongside classic FICO for free (17:38) on mortgage originations. And many are flowing the 10-T through to hedging analytics platforms, (17:43) which enables in turn the investor community to start seeing 10-T as an additional more predictive (17:49) data point.
At this point, we've signed over 60 lenders to this 10-T adopter program representing (17:56) over 500 billion in originations, over 1.5 trillion in servicing volume. And we just think that this (18:04) is important to provide this program, because the industry needs to move very thoughtfully (18:09) to assess these algorithms and understand credit risk differences between the scores. (18:13) Because if done right, this transition to modern scores can benefit the entire ecosystem.
(18:19) For people that want to see the decade plus of historical FICO score 10-T (18:24) data from the GSEs, where can they go to find it? (18:27) Oh, they can find that on Fannie Mae and Freddie Mac's website, simply googling (18:31) Fannie Mae, Freddie Mac, historical loan level data should take them right to the landing page (18:37) for which there will be a link for them to download the data. And if they have any questions, (18:42) Ravi, anything they'd like to discuss, any interesting findings, they shouldn't hesitate (18:46) to reach out to us at FICO. We've got a dedicated mortgage and capital markets team, (18:51) just standing ready and waiting to help with any insights, anything that can help lenders (18:55) adopt our new scores.
They can reach out to them at mortgageinfoatfico.com. (19:01) And finally, before I let you go, where does modernization go from here? How does it (19:07) continue to get better? How does it continue to push the envelope forward? (19:11) You know, it's a great question because as I mentioned, FICO score 10-T was submitted (19:16) to this effort, to this modernization effort in 2020. At this point, we're already six years (19:20) post that. There have been significant changes, of course, in the world since then.
Now, 10-T (19:25) remains a highly predictive score, Ravi, but where we're investing a lot of our time and energy (19:31) is in bringing in data, true alternative data, beyond what's found in the credit report. (19:37) Because we think that's a way for the millions of consumers who have relatively sparse traditional (19:43) credit experience, that's a way for us to gather enough information to responsibly and safely and (19:50) score them and bring them into the credit mainstream, including opportunities for a (19:57) mortgage. So, if you're asking me where does modernization go next, it's places like cash flow, (20:02) underwriting, and other sources of true alternative data.
That's the best way to expand access to (20:08) credit even further. Simply lowering existing scoring standards, as the competition has done, (20:14) that's not the answer. You know, what the answer is, is getting access to more data (20:18) beyond the traditional data that's leveraged by both FICO 10-T and other sports.
(20:24) Well said. And it's been very neat to see the embrace of data over the last several years (20:29) as we get toward a better scoring environment. So, Mr. Dorenhelm, I really appreciate the time.
(20:34) I enjoyed this thoroughly. And thank you very much, sir. (20:37) Thanks, Ravi.
(20:42) Today's economic calendar kicked off with mortgage applications from MBA, which fell 2.2% (20:46) for the weekend in July 3rd, with refinance applications down 4% and purchase applications (20:51) down 1% on a seasonally adjusted basis, reflecting relatively muted activity around the 4th of July (20:56) holiday. The average 30-year fixed mortgage rate edged up to 6.58%, while purchase applications (21:03) remained 5% higher and refinance applications 8% higher than the same week a year ago. Mortgage (21:09) rates have been essentially unchanged throughout the summer, and the MBA refinance index remains (21:13) near historically depressed levels, pointing to little prospect of a meaningful refinance wave.
(21:19) Later today brings May wholesale inventories, weekly crude oil inventories, (21:23) the aforementioned June FOMC minutes, May consumer credit, and a treasury auction of (21:27) $39 billion of 10-year notes. We begin the day with agency MBS prices worse about an eighth (21:32) from Tuesday's close, the two-year yielding 4.20, and the 10-year yielding 4.56 after closing (21:37) yesterday at 4.53%. Let's wrap up with a joke and some housekeeping. Here are some great truths (21:48) that I've learned as an adult.
Raising teenagers is like nailing jello to a tree. Wrinkles don't hurt. (21:56) Families are like fudge, mostly sweet, with a few nuts.
Today's mighty oak is just yesterday's nut (22:02) that held its ground. Laughing is good exercise. It's like jogging on the inside.
(22:08) Middle age is when you choose your cereal for fiber, not the toy. (22:16) Thanks again for FICO for sponsoring this week's podcasts. As the industry's most (22:21) predictive credit score, FICO score 10T combines proven performance with deeper (22:25) insight into borrower behavior to help support a stronger and more resilient housing finance system.