Nobody is short of new single-family homes. Builders are carrying them, much as a market maker carries a book nobody else will take, and they are paying a hidden carry to keep that book moving. The "housing shortage" crowd misreads this. So does the "housing bust" crowd. Both are watching the inventory number. The number that decides the outcome is the cost of moving that inventory, and it does not appear in the median price.
Think of a bookmaker who keeps the window open in a thin market. He does not stop quoting because the punters have gone home. He shades the odds, pays for the business, and lets his margin absorb what the posted price will not admit. The new-home market has worked this way since 2023. To read it you have to look past the tape at the flow.
Where Is the Shortage?
There isn't one in the new-home segment, and the Census Bureau's September figures settle it. Inventory of new single-family homes for sale at all stages of construction stood at 480,000, not seasonally adjusted. That is 49% above September 2019 and 68% above September 2020. At the current pace of sales, builders hold 8.6 months of supply. Since mid-2024, this inventory has run roughly 45% to 55% above 2019 levels. It sits at its highest since the 2005-2007 peak that opened the Housing Bust.
The regional detail is worse. In the South, the largest new-home market, inventory was 287,000 homes. That is down 5.3% year over year but up nearly 70% from September 2019, while sales were roughly flat over the same span, which leaves 8.4 months of supply. Those levels roughly match the Housing Bust peak, and sales are a third lower than they were then. The Midwest holds 57,000 homes, up 7.5% year over year and 50% above 2019, the highest since 2008. The West has worked inventory down 14% year over year to 103,000. Its sales dropped 14% too, so months of supply did not improve and sits at 8.6.
The industry's claim that scarcity is propping up prices does not survive these numbers. Wolf Richter calls it a manufactured sense of scarcity, and in the new-construction segment the data supports that word.
A shortage does not come with 8.6 months of supply.
Who Is Holding the Inventory?
The shortage claim is wrong, but the bears have their own error. They treat 480,000 homes as one undifferentiated glut. It is three separate positions, and they carry very different risks.
The arithmetic splits as follows. Of the 480,000 homes, 253,000 are under construction, down 5.9% year over year. Another 111,000 are completed and unsold. The remainder, about 116,000 by subtraction, are homes for sale where construction has not started. The not-started pile costs little to hold. A builder can slow starts, stretch timelines, or let lot options lapse. The under-construction pile is partly committed but can still be managed. The completed pile can only be sold. A finished, unsold house is a fully funded position. The capital is spent, the carrying costs accrue every month, and the asset loses appeal the longer it sits as a spec on a half-built street.
The completed pile is the one that matters. At 111,000 homes, completed inventory is 48% above September 2019 and 227% above September 2021. Against September sales of 56,000, finished homes alone represent roughly two months of total sales volume, and those homes need buyers now rather than at closing a year out. Builders have made progress. Completed inventory has edged down from last year's very high levels. The stock remains heavy.
The flow comes from this position. A builder with a large book of finished homes is a forced seller, and Richter states the constraint plainly. Large builders cannot pack up when the market turns. They have stockholders, bondholders, and credit ratings to protect, and they need volume to keep land pipelines, subcontractor crews, and overhead absorption intact. That is a seller under compulsion, the most important participant in any market. Price is not set by the median buyer's wishes. It is set by what the forced seller will accept to clear the next unit.
The completed spec is the position that sets the clearing price.
What Is the Real Price?
It is lower than the reported one, and the gap is the subsidy.
The reported number is not bad on its own terms. The three-month average median price of new single-family homes sold fell to $393,300, the lowest since August 2021. That is 4.6% lower year over year and 11% below the late-2022 peak. The decline is real and has been running for nearly three years. But the median reflects only the price written into the sales contract. It leaves out the incentives, the closing-cost credits, and above all the mortgage-rate buydowns that builders pay to close the sale.
A buydown is a price cut moved to a different line. The builder pays the lender to give the buyer a rate below the market rate, which cuts the buyer's monthly payment without touching the contract price. To the buyer it works like a cheaper house. On the builder's income statement it is a cost against revenue. To the Census Bureau it is invisible. The contract price that reaches the statistics is therefore the asking price with the discount removed. Comparable sales, appraisals, and the public's sense of "what houses cost" all anchor to a number the builders keep high on purpose.
We do not have a national series measuring incentive spending, and we will not invent one. The source points to where the evidence shows up instead: homebuilder gross margins have plunged, net profits have plunged, and builder shares have plunged. Margin is the honest price series in this market. Every buydown dollar comes out of gross margin, so the margin trajectory records the true price path that the median hides. The 11% decline in the median is the minimum true price decline. The actual cut is that figure plus whatever compression shows up in the builders' own filings.
That changes how to read the "falling prices" in the title. Prices are falling faster than reported. Builders are accepting that loss by choice because they have chosen volume over price, and so far they have managed it.
The median price understates the cut, and builder margins absorb the difference.
Why Have New-Home Sales Held Up?
Builders have taken over the market from frozen existing-home sellers, and the lock-in effect protects that position.
Look at the sales comparison carefully. New single-family home sales in September were 56,000, down 3.4% year over year and flat with September 2019. Over the same period, existing-home sales fell about 40%. Compared with the bubble-era Septembers of 2005 through 2008, new-home sales are down more than 40%, so this is not a boom. They are middling sales on heavy inventory. But they have not collapsed, while the resale market has.
The explanation sits in the mortgage book. Millions of homeowners hold loans written below 3% in 2020 and 2021. Selling means giving up that rate and borrowing again at today's market rate, so they stay put, list nothing, and wait for a price the market will not pay. Richter describes the result bluntly: builders are running circles around them. The existing-home seller sets a list price and waits. The builder cuts, buys down, credits, and closes.
The structural read follows from this, and it is the part most of the builder bull case has not priced. The builders' share gain depends on rates. Part of their sales resilience comes from having little competition, since the resale inventory that would normally compete with every spec house is locked behind low-rate mortgages. The builder is not just discounting into a weak market. It is the only seller in many submarkets willing to pay down the buyer's financing cost. The lock-in is effectively a moat the Federal Reserve built and the builders did not have to pay for.
What Breaks the Dealer?
The consensus trade on homebuilders, insofar as one exists, is simple: rates fall, affordability improves, demand returns, and ITB, XHB, and D.R. Horton (DHI) rerate. That trade is partly right. Lower market mortgage rates cut the cost of each buydown directly. If the builder no longer has to pay as much to reach an advertised rate, margin recovers without any change in volume. Falling rates would help builders in the near term, and they would do it through the subsidy line before they do it through demand.
The second-order effect runs the other way. A rate decline large enough to matter would also start unlocking existing-home supply. Some low-rate owners who have waited for years would list. Every resale house listed in a Texas or Florida exurb competes directly with a builder's completed spec down the road, and the resale seller does not carry a builder's capital costs or need to hit a quarterly volume target. The moat that kept builder share high would leak. In the South, with 8.4 months of new-home supply and inventory near bust-era levels, a resale thaw would land on the most crowded part of the builders' book. This link is an inference, and we flag it as one. The size and timing of any resale unlock depends on how far rates fall and how many owners are actually ready to sell. But the direction is not ambiguous. Lower rates shrink the subsidy and remove the moat at the same time, and the market tends to celebrate the first effect long before it registers the second.
The other way the dealer breaks is simpler. Sales fail to hold at roughly 2019 levels, the completed pile stops shrinking, and the subsidy has to grow just to keep volume steady. Builders would then be paying more for the same flow, with balance sheets carrying finished homes on which interest and taxes keep accruing. A market maker that has to widen its concession every month to keep the book moving is not providing liquidity. It is being run over slowly.
On positioning, we should state the limits plainly. Absent live positioning data on the builder shares themselves, we will not claim the group is crowded on either side. The only systematic read available, based on available consensus data, concerns the broad index. In S&P 500 futures, commercial hedgers are net long about 6% of open interest while speculators lean short. That is a statement about equities in general, not about lot inventory in Houston, and stretching it to cover builders would be careless. The closest macro precedent is the 2022 rate shock, which repriced long-duration assets savagely and broke the 2021 completed-home shortage within a few quarters. It shows how fast this market can move. It does not show which way it moves next, since that cycle started from far higher inflation.
The position follows from the mechanism, not the mood. The contrarian error would be shorting builders because inventory is high. Inventory is high because builders are doing their job, and much of it is under construction or not started, which builders can manage. The opposite error would be buying ITB or XHB on the shortage narrative, which the Census data refutes. The tradeable variable is gross margin, which is the subsidy expressed as a share of revenue. A long in DHI or the builder ETFs is a bet that falling rates relieve the subsidy faster than a resale thaw erodes the moat. That race can be won, but it is not the frictionless rate-cut story the sector usually gets. The risk concentrates in the South, in completed specs, and in any builder whose margin has already absorbed several rounds of buydowns without stabilizing.
Joseph stored grain in the fat years so Egypt could eat in the lean ones. America's homebuilders did the reverse. They sold every finished house they had in 2021 and 2022, when buyers would pay anything, and they are now holding the warehouse in the lean years, paying out of margin to empty it a few units at a time. The shortage is fiction. So is the bust. What is real is a forced seller with a large book and a subsidy that shows up only in its earnings.
This housing market's true price is in the builders' margins, not in the median.
