The headline says vacant housing is at its highest since 2013. The more important detail is where that vacancy sits. Over the past year, vacant units for rent rose by 319,000. Vacant units for sale rose by 8,000. The for-sale market is not clearing. Owners who refuse to cut prices are converting listings into rental listings, and that moves the adjustment from home prices to rents. Anyone looking at stable for-sale inventory as evidence that home prices are holding is looking at the wrong ledger.
A Stock Growing Faster Than the People to Fill It
The Census Bureau's Q3 2026 Housing Vacancies and Homeownership release puts the net US housing stock at 149.8 million units. That is up 358,000 in the quarter and 1.512 million over four quarters. These are net figures, new construction minus demolitions, and they exclude abandoned or uninhabitable units.
Demand moved much more slowly. The Census Bureau's separate population estimates show the US adding only 757,000 people in the 12 months through July 2026, a sharp slowdown driven by the immigration crackdown and tighter legal inflows. At an average household size of about 2.5 people, that population growth supports roughly 300,000 new households. The supply side added more than five times as many units.
The comparison is not exact. Households also form from the existing population as young adults move out, divorces split one household into two, and an aging population lowers average household size. Those effects can absorb some of the gap. They cannot plausibly absorb a gap of more than a million units a year, and the vacancy data confirms they did not. Total vacant units rose 591,000 year over year to 15.81 million. Of those, 12.34 million are year-round vacant, meaning intended for occupancy at any time rather than seasonal use. That equals 8.2% of the entire stock.
The Disclosure Inside the Headline
The 4.93 million vacant units on the market for sale or rent, up 7.1% year over year, is the figure that made headlines. Split by tenure, it tells a different story:
- Vacant for rent: 3.87 million, up 319,000 year over year (+9.0%) and 139,000 in the quarter. That is the highest count since 2011.
- Vacant for sale: 1.06 million, up 8,000 year over year (+0.8%), though up 45,000 in the quarter.
Almost all of the annual increase in on-market vacancy is rental. Over five years the split is even more lopsided: vacant for-rent units rose by 1.14 million, against 341,000 for vacant for-sale units.
The mechanism is owner behavior. A seller lists at a price the market will not pay and refuses to cut. The home sits. Instead of lowering the price, the owner tries to rent it out. If a tenant signs, the unit moves into occupied stock and the owner becomes an accidental landlord. If no tenant signs because the asking rent is also too high, the unit stays vacant but changes category, from for-sale to for-rent.
That reclassification is why the for-sale vacancy count looks contained. Part of the for-sale glut no longer appears in the for-sale column. The sale-price data then reflects a pool of listings from which the most stubborn sellers have stepped aside. Prices look steadier than the underlying supply would justify, because the excess units are being marketed somewhere else.
Where the Pressure Lands Instead
If sellers avoid price discovery by becoming landlords, the cost shows up in the rental market. Each converted single-family home or condo competes for the same tenants as purpose-built multifamily, which has been delivering its own heavy pipeline. A rental vacancy count at a 15-year high is the result.
Two other lines in the release strengthen this reading.
First, units "rented or sold but not yet occupied" fell by 92,000 year over year and by 104,000 in the quarter, to 901,000. This category is the absorption pipeline: deals signed but not yet moved into. A shrinking pipeline while on-market vacancy rises means fewer transactions are closing relative to the supply being offered. Inventory is growing and turnover is slowing at the same time.
Second, vacant units "held off the market" rose by 227,000 year over year, to 6.51 million. This is the shadow inventory: units held for occasional use, temporarily occupied by people who live elsewhere, or vacant for reasons such as estate settlement. For five years this pool shrank as units came onto the market, so shadow inventory was turning into actual inventory. That process has now reversed modestly. Some owners who tried to sell, then tried to rent, appear to be withdrawing units entirely. A withdrawn unit is not gone. It is supply that will return to the market when the owner decides to try again.
The three year-round categories add up as reported: 4.93 million on the market, 0.90 million in the absorption pipeline, and 6.51 million held off, for a total of 12.34 million. The source text also mentions a 12.84 million figure, but the components support 12.34 million.
The implication is that rents are where this supply overhang will be tested first, and sale prices second. Sale prices will feel it only when withdrawn and unrented owners run out of patience or carrying capacity. Owners with low fixed-rate mortgages can wait a long time. That patience delays the price adjustment. It does not eliminate it.
Listed housing equities are a second-order read, and the link is inferred rather than shown in this data. Homebuilders compete directly with resale and rental supply, and they have more room to cut prices or offer incentives than individual sellers do. Single-family rental operators and apartment owners are most exposed to a rising rental vacancy count. The release says nothing directly about any listed company's margins, so the equity link is a hypothesis to test against company disclosures, not a conclusion.
The Case Against Reading This as a Glut
The strongest counterargument starts with the survey itself. The Housing Vacancies and Homeownership survey has wide sampling error, and quarterly moves of 100,000 to 200,000 units in a category of several million can fall within that noise. One quarter's jump in rental vacancy should not anchor a forecast.
The 2013 comparison also cuts both ways. In 2013 vacancy was falling from the post-bust peak, and the market was absorbing excess supply. Matching that level from below, after years of unusually tight vacancy, could mean a return to normal rather than the start of an oversupply cycle. Rental market data from other sources has also shown rents re-accelerating in some segments and metros. That is hard to square with an overhang that is pressing broadly on rents.
The population figure is an estimate and may be revised. Household formation can also diverge from population growth for long stretches. If headship rates rise, more households form from the same number of people, and the gap between supply and demand narrows without any change in prices.
These objections limit how strongly the conclusion can be stated. They do not explain away the tenure split. Even if every quarterly figure carries meaningful error, a 319,000 year-over-year increase in vacant rentals against an 8,000 increase in vacant for-sale units is too large an asymmetry to be noise. The direction of the shift, from sale to rent, holds up even if the size of the shift is uncertain.
The Line to Watch in Q4
The read fails if the next release shows vacant for-rent units falling while vacant for-sale units hold steady. That would mean accidental landlords are finding tenants at their asking rents and the rental market is absorbing the conversions without strain. It also fails if the "rented or sold, not yet occupied" pipeline recovers, which would show transactions catching up with supply.
The read is confirmed by three movements together: rental vacancy rising again, held-off-the-market units continuing to grow, and the absorption pipeline shrinking further. That combination would mean owners are running out of ways to avoid cutting prices. Sale-price weakness should follow, starting in the metros where conversions have been heaviest. Until then, flat for-sale vacancy says little about pricing power. It mostly shows that sellers have found somewhere else to wait.
