themortgagellm^™

Written by The Mortgage LLM Team—a group of industry analysts leveraging our proprietary mortgage-domain language models to synthesize and decode housing data.

📌 Quick takeaways by role

• Loan officers & originators: The pipeline you can work right now is ~$289B across ~1.05M loans sitting ≥75 bps in the money — and it’s disproportionately government. FHA (6.0%) and VA (7.1%) are roughly twice as live as conventional (3.0%). Lead with FHA: many 2023–2024 buyers can shed life-of-loan mortgage insurance by refinancing to conventional at 80% LTV, a line home-price appreciation may have already crossed — no rate drop required. Jump to the warm-lead list.
• Sales managers & branch leaders: Staffing and prospecting are two different maps. By dollars the pool follows population — Texas ($32B), Florida ($29B), California ($22B) — so that’s where headcount goes. By density the odds flip to the interior: Oklahoma (7.6%), Ohio (6.4%), Indiana and Mississippi (6.2%). Staff by dollars; target campaigns by density. Jump to the map.
• Capital markets, secondary & servicing: The story isn’t the live 4% — it’s the locked 71%. $5.0T across 22.8M loans sits ≥100 bps below par: structural non-sellers, not refi candidates. Prepay speeds stay slow, MSR values hold, and the refi pool is too small to move aggregate speeds at current rates. Jump to the lock-in wall.

Trade coverage of the refi opportunity tends to wave at “millions of borrowers who could benefit.” We can do better than a vibe. Reading the active agency and government book loan by loan, about $289 billion across roughly 1.05 million loans is sitting at least 75 basis points in the money to refinance today — a refinanceable pipeline you can rank by state, product, and vintage. That is the number a loan officer can actually work.

It comes with a hard ceiling, though, and the two facts only make sense together. That $289 billion is 4% of a $7.1 trillion book — and 71% of that same book is locked in at note rates a full point or more below today’s par. The refinanceable pool is real and callable. It is also a sliver sitting on top of the most frozen mortgage market in modern memory. This is the map of both.

The pool is real, and it is concentrated

With the 30-year conforming rate at 6.51% (Freddie’s latest survey), FHA par near 6.27%, and VA near 5.75%, “in the money” means a borrower whose note rate clears par by enough to make a refinance pay after costs. At the standard 75-bps rule of thumb, the pool is $289 billion. Loosen the screen to 50 bps and it grows to $494 billion and 1.7 million loans; tighten it to a 100-bps cushion and it shrinks to $163 billion; demand a 150-bps incentive and only $20 billion survives. The pool is rate-sensitive at the margin — a quarter-point move in rates swings it by hundreds of billions — which is exactly why it is worth re-running as rates drift.

It is, disproportionately, a government story

The single most useful cut for an originator is by product. Only 3.0% of the conventional book (about $150 billion) is 75 bps in the money. The government books are roughly twice as live: 6.0% of FHA ($76 billion) and 7.1% of VA ($63 billion). The reason is vintage mix — the conventional book is dominated by the 2020–2021 sub-3% refinancing wave, while FHA and VA carry a heavier share of recent, higher-coupon purchase loans that never saw the trough.

FHA carries a second lever our rate-only screen doesn’t even count — and it may be the stronger pitch. FHA borrowers who put the minimum down can’t cancel their mortgage-insurance premium the way conventional borrowers can; on most loans it runs for the life of the loan, and the only exit is to refinance into a conventional loan at 80% LTV or below. The part a rate screen misses is that the 80% line can be reached on home-price appreciation, not just amortization. In the markets that kept climbing through 2025, equity has pulled some 2023–2024 FHA buyers toward that threshold years ahead of their payment schedule — and crossing it lets them shed a premium of well over a hundred dollars a month without rates falling at all. That reframes the FHA pitch from a rate play into a structural equity play, and it makes the FHA pool more live than its 6.0% rate-only share suggests. For an LO carrying government products, that is the warm-lead list.

Where the deals are — and where they’re densest

The two ways to read the map tell different stories, and both matter.

By raw dollars, the pool sits where the people are: Texas leads with $32 billion in 75-bps in-the-money balances (about 120,000 loans), followed by Florida ($29 billion) and California ($22 billion), with Georgia, North Carolina, Ohio, and Illinois each around $10–12 billion. That ranking is mostly a population map — useful for deciding where to staff, but not where the odds are best.

By density — the share of a state’s own active book that is in the money — the leaders flip to the lower-cost middle of the country: Oklahoma leads the pack at 7.6%, followed by Ohio at 6.4%, Indiana and Mississippi at 6.2% each, with Arkansas, Missouri, Kentucky, and New Mexico close behind. These are markets where borrowers bought more recently at higher coupons and never locked in a sub-3% rate, so a larger fraction of the book is refinanceable. A loan officer prospecting in Oklahoma City is fishing in water more than twice as dense as one in California.

Figure: The active fixed-rate book by note-rate band and product. The mass to the left of each product’s par marker is the lock-in wall; the shaded region to the right is the refinanceable pool. Interactive version — hover any band for loan count, balance, and average note rate; toggle products in the legend. A companion chart ranks states by in-the-money balances.

A barbell, not a bell curve

Stratify the pool by vintage and the picture sharpens to a point. Roughly 93% of the in-the-money balances were originated in 2023, 2024, or 2025 — the high-coupon cohorts written after rates left the floor. The 2023 book alone is 17% in the money. Step back to 2020 and 2021 and the share collapses to one-tenth of one percent: those two vintages hold $3.3 trillion of active balances at rates that today look like a gift, and essentially none of it will move.

That is the barbell. One end is a small, growing pool of recent borrowers who took a high rate on the bet they could refinance later — and now can. The other end is an enormous, immovable block of pandemic-era borrowers for whom refinancing would mean raising their rate. There is very little in between.

The lock-in wall is the real story for 2026

Flip the screen around and the structural read emerges. $5.0 trillion — 71% of the active agency/government book, across 22.8 million loans — sits at least 100 bps below par. These borrowers are not refinance candidates; they are non-sellers. Their below-market note is a financial reason to stay put, which is precisely why existing-home inventory remains starved and the purchase market stays rate-locked. For an originator, the lock-in wall is not background noise — it is the explanation for why purchase volume is hard to come by, and why the refinanceable pool, real as it is, cannot carry a pipeline on its own.

The takeaway for the desk is two-sided and honest. There is a genuine, nameable refi pool right now — concentrated in government loans, in 2023–2025 vintages, and densest across the lower-cost interior — and it is worth working before rates fall far enough that everyone sees it. But it is a pool, not a wave, and it sits on a book that overwhelmingly will not move until rates come down enough to crack the wall. The originators who win 2026 will be the ones who can find the live 4% without waiting for the frozen 71% to thaw.

---

Methodology. Loan-level balances and note rates sourced from the agency and government performance disclosures — Fannie Mae and Freddie Mac single-family loan-level data (conventional conforming) and Ginnie Mae loan-level data (FHA and VA) — consolidated via The Mortgage LLM’s analytics instance. The active book is each surface’s latest available monthly disclosure (Fannie December 2025, Freddie April 2026, Ginnie Mae March 2026), counting loans with positive current balance that are not in a terminal or paid-off state. Analysis is restricted to fixed-rate loans, for which the note rate is meaningful over the life of the loan; adjustable-rate loans (under 1% of the government book and absent from the GSE data) are excluded. “In the money” compares each loan’s note rate to the product-appropriate current par rate — the 30-year conforming rate from Freddie’s Primary Mortgage Market Survey for conventional loans, and prevailing FHA and VA 30-year rates, as of May 21, 2026 — and counts a loan as refinanceable when its rate exceeds par by the stated threshold (50, 75, 100, or 150 bps); 75 bps is featured as the standard after-cost rule of thumb. “Locked in” counts loans at least 100 bps below par. The universe is the agency and government book only; it excludes bank-portfolio, jumbo, non-QM, and private-label loans, and a small USDA/PIH government segment. The screen is rate-only: it does not net out closing costs, term reset, or mortgage-insurance changes (an FHA-to-conventional refinance can shed mortgage insurance, an additional incentive not captured here). Loan-level disclosures run one to two months behind, so the newest originations are under-represented by reporting lag. Informational, not advice.