Mortgage lock-in runs on two clocks
The number of mortgages below 4% fell at 6.2% a year from 2022 Q1 to 2026 Q1, even while the rate gap stayed open. That steady decline still leaves a long wait for lock-in to fade.
The estimated number of active US mortgages below 4% fell from 33.1m in 2022 Q1 to 25.7m in 2026 Q1. That was a decline of 7.4m loans, or 22.5%, equivalent to 6.2% compound annual runoff. Low-rate mortgages kept disappearing while the gap between their rates and the cost of a new loan remained open.
It's easy to see why homeowners would hold on to those loans. Replacing a 3% mortgage with one at 6.7% raises the rate on their debt, whether they refinance or borrow to buy another home. The incentive to stay put remained substantial: FHFA’s National Mortgage Database reported that 49.9% of active mortgages carried rates below 4% in 2026 Q1, and Freddie Mac reported a 6.66% new-origination rate on 30 July 2026.
Yet households don't make every decision around their mortgage. A job move, retirement, divorce or death can prompt a sale even when giving up the loan is costly. Mortgages also end through refinancing, default and maturity. For anyone forecasting existing-home transactions, the question is how quickly this continuing runoff reduces the stock exposed to lock-in.
The stock shrinks without waiting for lower rates
Lock-in runs on two clocks. A narrower rate gap can make moving or refinancing less costly; other reasons for ending a mortgage persist while that gap stays open. The national loan count shows how quickly the stock declined under those conditions, though it doesn't identify which reasons drove the exits.
The count comes from FHFA’s observed totals and rounded rate shares, weighted by loan count. It estimates the size of the sub-4% bucket at each date, rather than tracking the same loans throughout. Its decline is net of any loans entering the bucket. For a transaction forecaster, that count approximates how many mortgaged households remain exposed to lock-in.
Estimated active mortgages below 4%, by loan count. FHFA National Mortgage Database, 2022 Q1–2026 Q1.
The decline is substantial, but its pace is slow. If annual loan-count runoff stayed at 6.2%, roughly half the starting count would remain after 11 years. That is constant-rate arithmetic, not a forecast of when lock-in ends.
Payoffs continue, and the rate gap still matters
Freddie Mac’s direct payoff data offer a check on that slow pace in a narrower mortgage book. For low-rate cohorts originated in 2020 and 2021, calculations from observed monthly data put voluntary payoff at 3.64% in 2023, 3.78% in 2024 and 4.06% in 2025. Borrowers continued to pay off these loans, with only a modest increase across the three years.
Those rates and FHFA’s 6.2% measure different things. Freddie measures gross voluntary payoff, weighted by unpaid principal balance (UPB), for qualifying 30-year cohorts in the to-be-announced (TBA) securities market. FHFA measures national net runoff by loan count across all exit types, after inflows. Because coverage, weighting and construction differ, the gap between them can't be assigned to any single cause.
Within Freddie’s 2022 vintage, however, the comparison holds origination year and broad loan age constant. In 2025, cohorts below 4% paid off voluntarily at 4.22%, while those at 5.5% or above paid off at 8.26%. That difference is consistent with a response to the rate gap. It doesn't establish a causal rate effect, because borrower, balance and security characteristics may also differ.
Left: UPB-weighted gross voluntary payoff for low-rate Freddie cohorts originated in 2020–2021, shown for calendar years 2023–2025. Right: the same measure for 2022-vintage cohorts below 4% and at 5.5% or above in 2025. Freddie Mac Daily Prepayment Report, workbook dated August 5, 2026.
Together, these observations support separating continuing mortgage terminations from the response to a narrower rate gap. A single runoff input would miss the difference between the two payoff rates within the 2022 vintage.
Falling balances need a separate check
FHFA’s balance data show faster runoff: sub-4% UPB declined at 8.1% a year from 2022 Q1 to 2026 Q1. But borrowers reduce their balances through scheduled principal payments while remaining in their homes with the same mortgages. That makes balance runoff a poor direct measure of how quickly households leave lock-in.
Adjusting for a 2.5% annual scheduled-amortization proxy, derived from March 2022 Freddie 30-year TBA cohorts, gives implied exit-driven UPB runoff of 5.7% over the same period. This estimate and the count result are separately constructed from the same FHFA bucket; the balance estimate also depends on how well the Freddie proxy represents national amortization.
Loan terms matter to that adjustment. At the proxy’s mortgage rate, a new 15-year mortgage amortizes 5.31% of principal in its first year, compared with 2.04% for a new 30-year mortgage. If national amortization exceeds the proxy, implied exit-driven runoff is lower. The methods note gives the stock construction, multiplicative amortization adjustment, Freddie reconciliation and data vintages.
A slow pace can still change
Even runoff that continues with an open rate gap needn't stay constant. NAR’s 2025 Profile reports that sellers in its July 2024 to June 2025 transaction window had owned their homes for a median 11 years, compared with six years during 2000–08. Census counts likewise imply that the mover rate among people in owner-occupied units fell from 9.1% in 2000 to 4.1% in 2023.
These are directional indicators of changing turnover, not mortgage exit rates. Neither calibrates a runoff assumption, but both argue against treating any historical pace as permanent.
A range makes the consequence of the assumption easier to see. At constant annual loan-count runoff of 4% or 6%, the shares of the starting count remaining would be:
| Assumed annual loan-count runoff | After five years | After ten years |
|---|---|---|
| 4% | 81.5% | 66.5% |
| 6% | 73.4% | 53.9% |
These are illustrative scenarios; the observed 6.2% rate sits just above the range. Both leave more than half the starting count after a decade. If the rate gap changes, the response to it changes too, and a constant-runoff path no longer captures the whole process.
An existing-home transaction forecast therefore needs separate inputs for continuing stock runoff and the response to a narrower rate gap. Mortgage exits don't translate one-for-one into sales or net inventory: a payoff can be a refinancing, and a seller may also buy. What the evidence establishes is that the low-rate stock can shrink while the financial incentive to keep those loans persists. At the observed pace, that erosion takes more than a decade to halve the count. It is a meaningful source of gradual change, with a large stock still exposed to lock-in along the way.