2025: Stabilization, Mix Shifts, and the Quiet Re-Sorting of Mortgage
If you spent 2025 waiting for the market to “come roaring back,” you probably left disappointed. This was not 2020 or 2021. It also was not 2022’s hangover. But it was not the free-fall mood of 2023, either.
2025 was the year the industry proved it can operate in a normal-rate environment, even if “normal” now means a 30-year fixed that starts with a 6. As of last week, Freddie Mac had the average 30-year fixed at 6.21%. (AP News)
The market did not need a miracle to improve. It needed stability, and it needed consumers to stop waiting for the perfect headline. That showed up first in the refinance channel. MBA data cited by AP showed refinance applications made up 59% of total applications, the highest share since September. (MBA via AP)
Now let’s talk production.
Because full-year “closed loan” totals vary by dataset and timing, I’m careful with absolutes. But for a credible directional read, the MBA’s July 17 Mortgage Finance Forecast projected total 1- to 4-family originations of $2.021 trillion in 2025, up from $1.779 trillion in 2024. (MBA)
The more important story is not simply “volume up.” The story is the type of volume.
In that same MBA forecast, purchase originations were projected at $1.357T in 2025 (vs $1.288T in 2024), while refinances were projected at $664B (vs $491B). (MBA) Purchases grew, but modestly. Refinances grew much faster, mostly because they were coming off the floor. 2025 was not “the refi year,” but it was the year refis became a real business line again.
If you want a reality check that purchase did not suddenly become easy, look at Q3. An ATTOM-based Q3 summary pegged total Q3 lending at 1,773,487 loans and $600.4B in dollar volume, up 3.1% year over year in dollars. (National Mortgage Professional) ATTOM’s Q3 2025 mortgage origination report adds the mix: purchase loans were 43.2% of all originations and 51.6% of total dollar volume, both down from Q2. (ATTOM)
Distribution mattered, too. According to a December MBA NewsLink summary of Inside Mortgage Finance data, brokers accounted for 19.8% of residential mortgage originations in Q3 2025, up from 19.2% in Q2. (MBA NewsLink) That is not “brokers took over,” but it is a clear signal: the broker channel held share and even gained a little when the pie was not expanding dramatically.
On licensing and headcount, social media narratives tend to get dramatic. Early in the year, National Mortgage News reported that 158,152 individuals had requested MLO license renewal as of January 1, 2025, versus 160,893 the year prior, citing CSBS/NMLS data. (National Mortgage News) That is not a mass exodus. What we do not have, at least from publicly summarized sources in a clean, comparable way, is a simple “net new versus unrenewed” figure for the full 2025 cycle, so I would avoid over-claiming that point in print.
Year in Numbers
30Y fixed: 6.21% (Freddie Mac, via AP News)
2025 total originations: $2.021T (MBA July 17 forecast)
2025 purchase / refi: $1.357T / $664B (MBA July 17 forecast)
Q3 purchase share: 43.2% of loans; 51.6% of dollar volume (ATTOM)
Broker share: 19.8% in Q3 (Inside Mortgage Finance via MBA NewsLink)
So what’s the clean way to summarize 2025?
Better performance without a rate tailwind.
Purchases were steady, not easy, still capped by affordability.
Refinances returned enough to matter, even without a wave.
Channel share kept rebalancing, with brokers drifting back toward 20%.
As we head into 2026, here are the practical things I’m watching in January:
Does refi momentum hold if rates stay around the low-6s, or was the late-year spike mostly seasonal mix?
Does broker share stay near 20%, or does retail regain ground as marketing budgets and lead funnels reset?
Does purchase re-accelerate if inventory loosens, or do we stay in the “move-up only” dynamic?
Who wins operationally, because in this market the difference between “surviving” and “growing” is often process, pull-through, and partner consistency, not a magic product.
If 2024 was the year of endurance, 2025 was the year of re-sorting. The winners were not the loudest. They were the ones who built a model that works when rates are not a cheat code.