EstateVision

Market

Mortgage lock-in may help explain why new US houses now sell below existing ones

The median new US house sold for $400,400 in September 2026, $17,100 below the $417,500 median for existing homes. Rate lock-in is one plausible reason builders supply more of the marginal house, but no release isolates it, and part of the price gap is mix.

105110115120125130111Sep 2025Nov 2025Jan 2026Mar 2026May 2026Jul 2026Sep 2026thousands of houses
New houses for sale at end of month, completed homes only (seasonally adjusted). Source: census.gov · United States.

The median new house sold in the United States in September 2026 went for $400,400, according to the US Census Bureau and the Department of Housing and Urban Development. That is $17,100 below the $417,500 median for existing homes reported by the National Association of Realtors (NAR), so the usual new-home premium has gone. Two cautions apply at once. The medians come from different surveys covering different homes, and the new-house price is 3.9% below September 2025 with a margin of error of ±8.9%, so Census cannot call that change significant.

The more reliable story sits underneath the price: who is able to sell at a 7% mortgage rate.

About half of US mortgages sit below 4%, which weighs on owners' willingness to sell

Freddie Mac's survey put the 30-year fixed mortgage at 7.18% on 5 November 2026, down from 7.24% the week before and the third weekly decline in a row. An owner who sells and buys again swaps an old loan for that rate.

The size of the swap shows in the loan book. Using the Federal Housing Finance Agency's National Mortgage Database, as reported by the financial site Wolf Street, 50.6% of outstanding US mortgages carried a rate below 4% in Q4 2025, and 19.7% were below 3%. At the Q1 2022 peak, over 65% were below 4%. The share is falling slowly. I could read only the Q4 2025 reading through that secondary report, so it is older than the rate quoted above. For about half of borrowers, moving means a rate more than three percentage points higher.

Resale supply is thin. NAR counted 1.60 million existing homes for sale in September 2026, a 4.8-month supply, with sales at a seasonally adjusted annual rate of 3.99 million. Lock-in is one plausible cause among several, and none of these releases isolates it. It would, if it holds, make the house that has to find a buyer more often a builder's.

Builders carry an 8.4-month supply, and the price gap is partly mix

Census counted 476,000 new houses for sale at the end of September 2026, an 8.4-month supply at the current sales pace of 676,000 a year. A year earlier it was 8.2 months. That is nearly twice the existing-home figure.

New houses are also taking longer to sell once finished. In Census Table 3, the median completed house had been on the market 3.0 months in September 2026, against 2.4 months in September 2025. The median peaked at 3.5 months in March, April and May. Builders' holding costs are not measured in these releases, so the pressure to cut price, shrink the house or add incentives is an inference from time on market, and Census does not capture incentives at all.

The price comparison needs the same care. The average new-house price fell 6.7% year on year to $472,800, a change Census flags as significant, while the median change is not. Census says price changes reflect the regional and size mix as well as like-for-like prices. The mix did move. Houses priced $300,000 to $399,999 were 31% of September sales against 23% a year earlier, and houses at $1 million and over were 4% against 6%. A smaller high-end share pulls the average down without any builder cutting a price.

Completed stock fell to 111,000, and lower completions probably explain it

The composition of the for-sale stock changed in the direction builders can control. Completed houses for sale stood at 111,000 in September 2026, seasonally adjusted, against 122,000 a year earlier. The line chart "New houses for sale at end of month, completed homes only" shows it mostly declining from 128,000 in January, with one small uptick in July (113,000 to 114,000). Houses for sale not yet started rose to 115,000 from 103,000. Those under construction slipped to 250,000 from 262,000.

The rise in unstarted houses should not be read as presale or as rate locks agreed in advance. Census counts only unsold houses in this series. A higher not-started figure means more unsold, unbuilt stock is on offer, which is a different signal from builders having already found buyers.

Completions are the likelier explanation for the fall in finished stock. Sales of completed houses were 380,000 at an annual rate in September 2026, against 428,000 a year earlier. Completed stock still fell, so fewer finished houses must have arrived. Total housing completions were 1,192,000 at an annual rate, 14.0% below September 2025's 1,386,000, a decline Census calls statistically significant. Single-family completions were 840,000. Total completions include multifamily, so this is a proxy for houses built for sale, and the link is my inference. I did not find a single-family completions series set against single-family completed sales. The direction is consistent with an inflow that has been cut, but it is not proof.

The South holds the pipeline, the West is cutting faster, and two releases test the reading

Building permits ran at a seasonally adjusted annual rate of 1,400,000 in September 2026, 3.0% below September 2025. Single-family permits were 845,000, down 4.0% from August. Total housing starts were 1,221,000, 7.4% below a year earlier, though inside the ±11.5% margin of error.

The South issued 759,000 of those permits, 1.8% fewer than a year earlier, and 413,000 of the 676,000 annual-rate new-house sales were in the South. In the West, permits were 283,000, down 11.3% from September 2025, and single-family permits were 170,000, down 7.6%. New-house sales in the West were 17.4% lower, though with a ±29.0% margin, so only the direction is informative.

The summary release I read does not report regional median prices. Census publishes a separate quarterly regional series of median and average new-house prices on its historical time series page, which I did not read, so the regional pressure described here rests on supply data, not on prices. For lenders, the data show finished stock falling because starts and completions were cut, not because demand recovered.

Census publishes October new residential construction on 18 November 2026 and new residential sales on 25 November 2026. The reading gains support if completed stock keeps falling and months' supply moves down from 8.4 toward the existing-home 4.8, with the median completed house selling in under 3.0 months. It weakens if completed stock rebounds above 120,000 while the Freddie Mac rate stays near 7%. That would mean builders are again building ahead of demand.

Questions

Why are new homes priced below existing homes in the US right now?
The Census median for new houses was $400,400 in September 2026 against $417,500 for existing homes from NAR. Part of the gap is mix, since the surveys cover different homes. Rate lock-in may also contribute: about half of mortgages carried rates below 4% in Q4 2025 (FHFA data as reported), against 7.18% for a new 30-year loan on 5 November 2026, which could leave owners selling less and builders supplying more of the marginal house. No release isolates that effect.
Is US housing supply recovering?
Not quickly. Permits were 1,400,000 at an annual rate in September 2026, 3.0% below a year earlier, and total completions were 14.0% lower. Existing-home inventory was 1.60 million, a 4.8-month supply, against 8.4 months for new houses.
What would change this view?
Census releases October construction data on 18 November and new home sales on 25 November 2026. Completed new houses for sale rebounding above 120,000 with rates near 7% would weaken it.

Sources

  1. US Census Bureau and HUD: New Residential Sales, September 2026 (released 27 October 2026), Tables 1 to 3
  2. US Census Bureau: New Residential Sales historical time series (houses sold and for sale by stage of construction; regional prices)
  3. US Census Bureau and HUD: New Residential Construction, September 2026 (released 20 October 2026)
  4. NAR: Existing-Home Sales Increased 1.5% in September
  5. Freddie Mac: Mortgage Rates Decline for Third Consecutive Week to 7.18% (5 November 2026)
  6. Wolf Street: Update on the Lock-in Effect in the Housing Market (FHFA National Mortgage Database, Q4 2025), 27 March 2026

How we write: each note starts from published data and the sources listed above, is drafted with AI and passes automatic checks on every figure, link and claim before it goes out. Analysis, not investment advice.

More research notes