The number of loans originated in 2026 rose by 619k to 6,58 million. This was the highest figure since 2022, but the 4th-lowest number since 2000. The preponderance of the increase last year was due to a 512,000 increase in the origination of refinance loans, a 36% jump from 2024. Home improvement originations rose by about 22,000 (3.7%) while those for purchase loans edged up by 63,000 (1.9%). Elevated home prices and interest rates continue to be headwinds for new issuance.

        Ⅱ. Lender Type

The nonbank share of originations rose 1.6% to a record high of 72%. This is up by about 10% from 2019, just before the COVID-19 pandemic, and is almost double the 35.9% figure reported in 2011. Virtually all of the increase in loan production in the last two years has come from nonbanks. There were 1.1 million more loans originated in 2025 compared to 2023, of which only 0.7% came from banks.
​
        Ⅲ. Market concentration

In 2025, the two nonbanks, Rocket Mortgage and United Wholesale Mortgage, continued to dominate the origination market, with each holding shares above 6.0%. No other lender achieved a 2% share last year. The market became more concentrated, with the top 10 lender share rising to 23.9% in 2025 from 22.7% the prior year. Only two banks, JP Morgan and Bank of America, made the top 10 list in 2025, as US Bank, which was the number 9 lender the prior year, dropped out of the list in 2025.
​
All this is just scratching the surface of all the information available in the HMDA release. A final version containing updates and more information will come out in the summer of 2026.

The author notes that the FHA 90+ day delinquency rate rose sharply from 3.56% in September 2025 to 5.23% in January 2026 and attributed this increase to “how an FHA policy change is reported rather than increasing financial fragility among FHA borrowers.”
Recursion data is used throughout, including transition rates, Roll to Current and Early Buyout Rates.
As always, Recursion is pleased to be considered the top source for cutting edge data analytics across the mortgage industry ecosystem.
 
https://www.responsiblelending.org/research-publication/policy-related-reporting-change-not-increasing-financial-distress-drivin

​Recursion is proud to be recognized as the premier provider of agency MBS data across the mortgage ecosystem.

Recursion is always pleased to contribute the industry’s most reliable data and trusted analytic tools to top researchers to enhance the discussion of policy issues of importance across the mortgage landscape.

Recursion is proud to be recognized as the premier provider of agency MBS data across the mortgage ecosystem.

Trends in use of Valuation Methods
​
Given these changes, our area charts that break down the market of loans delivered to the GSE’s look like this:

The logic for “eligible” in the above charts reflects the new changes in criteria implemented this year. The main notable observation is that the difference in appraisal usage between Fannie Mae and Freddie Mac has swung substantially over the post-COVID period. Fannie Mae usage picked up sharply vs Freddie Mac over the 2020-2022 period. Subsequently the trend has more than reversed.

Interestingly, the preponderance of the recent decline in the Fannie Mae share of waiver usage is due to a decline in its intensity of usage of waivers among eligible loans compared to Freddie Mac, rather than to any significant change in the shares of deliveries that are waiver eligible on the part of either agency.
There is a lot more work that needs to be done to explain these behaviors.

The New Flexibilities
We can see from the charts above that the use of waivers and waiver plus property data collection approaches has risen during the course of 2025 in light of the expansion in eligibility criteria.
​
Next, we break out the changes in shares this year due to these recent policy adjustments. Below find the shares of waivers and waiver plus property data collection as a share of total issuance:

And below find the increase of shares broken down between new and old criteria over the period January 2025 – July 2025:

​For Fannie Mae, there was a 3.6% rise in the share of Waivers and AW+PD Approaches, of which about 40% was due to increased use of these products based on the old criteria. For Freddie Mac the increase was 7.0%, but this time 51% was due to increased utilization of products based on the old criteria.
Finally, the numbers are relatively small, but it is of interest to see which sellers have been quickest to enter this new market segment:

There are a couple of interesting observations to be made. First, nonbanks heavily dominate banks for both the waiver with CLTV>80&<90% category with an 85% share and for the CLTV>80% and AW+PDC category with a 92% share. Concentration is also greater for nonbanks across categories. For the waiver under new criteria category, the top 5 nonbanks account for 57% of total nonbanks, compared to 40% for banks. For the new AW+PDC with extended eligibility category, the top 5 nonbanks account for 89% of total nonbanks, compared to 63% for banks.

Appraisals and Credit Availability
​
Recapping our analysis so far, we face two key questions. First, with lack of affordability looming as problem number one in the housing market over the last several years, it is perhaps surprising that the take-up rate for eligible waivers and modern valuation approaches hasn’t been greater. The implication is that lenders and/or the agencies themselves have some reluctance to push this too far based on some perceived risk of eliminating or reducing the human element of the appraisal process. Second, there is a clear difference between the two agencies in terms of their perception of these risks and their likely costs. A deeper investigation that would shed light on the risk/return appetites of the various interested parties would be of great interest but is outside the scope of this note. Instead, we conclude by looking at the connection between property assessment choice and loan pricing for Fannie Mae and Freddie Mac:

These remarkable charts pull together the major themes of this note. We highlight three particular points:

• First, the yield gap between loans with regular appraisals that are not eligible for an appraisal and those with a waiver (the green bars) has been generally, but not uniformly, positive over the past five years. This is because the higher LTV associated with a loan that is not eligible for a waiver is perceived to be greater than the risk associated with the absence of a human appraiser in the valuation process. Note that the magnitude of this gap is different between Fannie Mae and Freddie Mac.

 

• Second, there was a very significant shift in the pricing of loans with appraisals that were eligible to get waivers vs those with waivers as the affordability crisis got fully underway starting in late 2022 (the purple bars). Note that the difference in LTVs in the first bullet above does not apply here as both categories in this case are waiver eligible. Essentially the direction of the gap flipped signs at that point, implying that loans with regular appraisals were subsidized relative to those with waivers starting in late 2022. In fact, houses simply became unaffordable at that point in time, and we surmise that in a soft market, lenders became willing to buy down loans with regular appraisals in order to reduce the risks associated with a growing share of loans with automated appraisals while keeping production pipelines as full as possible.

​

• Third, to validate this supposition, we look at the difference in the appraisal that were eligible to get waivers share of purchase loans between the two agencies. Start with Freddie Mac. It seems that pricing was adjusted to keep the waiver eligible appraisal share of new issuance in a steady range near 1/3. Fannie Mae, on the other hand, entered into a much more dynamic response, buying down the loans with appraisals to a much greater degree than Freddie Mac. The result was a surge in the waiver eligible appraisal share of their new loans from 28% to 35% over the past three years. Whether the differences in these strategies were due to different risk tolerances or to competitive dynamics remains an open question.

 
Conclusion
In this note, we argue that the role of collateral risk has come to play a key role in the pricing dynamics on the part of the GSE’s, and this factor needs to be considered along with more traditional factors such as underwriting characteristics when assessing their overall willingness to supply credit to mortgage borrowers. The exact nature of this role has not been spelled out here, but the tools and data to do so are available. A final supposition based on pure casual observation is that the changes seen in relative valuations of loans with different valuation techniques in the last few years correlate not just with affordability, but also with increasing awareness of growing climate risk. This is a future topic for these notes.