August 2026 Housing Completions: What the Decline Shows

August 2026 housing completions for U.S. privately owned housing fell 11.9% from revised July estimates to a seasonally adjusted annual rate of 1,128,000 units, according to the Census/HUD report released September 17, 2026. The monthly total and single-family declines were statistically significant; the decline in buildings with five or more units was not.

For builders and developers, the release shows a lower estimated pace of completed construction. It does not, by itself, establish that projects took longer, that fewer homes were available or that rental demand strengthened.

Single-family and multifamily completions

The figures below are seasonally adjusted annual rates. July is revised; August is preliminary. Changes include 90% confidence margins, expressed in percentage points. Annual comparisons use August 2025. Census/HUD, Table 5a.

The five-or-more category excludes smaller multifamily buildings. Census withholds the separate seasonally adjusted 2–4-unit series because it fails seasonal reliability standards. Table 5a notes.

Which declines exceed sampling uncertainty?

A change clears Census’s significance test when its confidence interval excludes zero. All three annual declines do. Monthly, the five-or-more-unit interval includes both a decrease and an increase. These intervals account for sampling variability but exclude nonsampling errors such as reporting problems. Release explanatory notes.

Statistical significance helps establish the direction of a change; it does not identify its cause. Evidence that fewer units reached completion is insufficient to conclude that the construction process became slower.

Completions and the stock under construction

The release also measures units still under construction at month-end. None of the monthly changes below was statistically significant. Stock estimates are seasonally adjusted; the changes include 90% confidence margins in percentage points. Census/HUD, Table 4a.

These figures use a different time basis from completions. An annualized completion rate expresses a month’s seasonally adjusted activity multiplied by 12. Dividing it by 12 gives the adjusted monthly pace; it does not recover the unadjusted number completed that month. Census explains this convention in its seasonal-adjustment guidance.

Under construction means work has started but has not reached completion. It is an inventory at a point in time, rather than activity accumulated during a month. Census distinguishes these measures in its construction statistics presentation.

A delivery forecast therefore needs to distinguish how much work remains underway from how quickly units are reaching completion. Subtracting an annual completion rate from a month-end stock answers neither question. The earlier permits-and-starts distinction addresses another stage of the same sequence.

A completion decline does not establish delays

The monthly totals do not connect each completed project to its own start date or original schedule.

Consider a hypothetical case in which fewer projects started in an earlier period, but every project still takes exactly the same time to build. Completions can subsequently decline without any project being delayed. Conversely, new starts can replenish the unfinished stock while older projects finish. These are possible explanations of aggregate movements, not findings about August.

Dividing the stock by the annualized completion pace produces a quantity measured in years. Calling that result an average construction time would require assumptions that these tables do not establish. The numerator mixes projects at different stages; the denominator reflects completion activity in one month.

To investigate delays, compare actual start-to-completion times for similar buildings and examine unfinished projects against their schedules. Distinguish elapsed construction time from time remaining: a recently started building and one awaiting final work can both sit in the under-construction total while presenting very different delivery risks.

Completed units are not available inventory

Census generally uses installed finished flooring as the single-family completion marker. For multifamily buildings, its building-level rule counts all units once half are available for occupancy. The completion count therefore is not a unit-by-unit tally of vacant apartments ready for immediate leasing. Census construction definitions, slide 12.

The ownership distinction matters too. Single-family construction includes owner-built houses, contractor-built houses on an owner’s land and homes built for rent. Census’s new-home sales measure instead covers houses built for sale, and its for-sale inventory excludes homes already under sales contract. A completion therefore need not add a home that another buyer can purchase. Census construction and sales definitions, slides 16–17.

Rental absorption requires separate evidence. Census’s Survey of Market Absorption follows sampled buildings containing five or more units, measuring units taken off the market and conducting follow-up interviews after completion. Those observations connect completed construction to leasing or sales activity.

Apply the release to a local delivery forecast

Before changing a project’s lease-up assumptions or a builder’s sales budget, compare national residential construction activity with the actual competing properties. Other housing data and reports can add context; project records supply the timing and availability needed for a specific decision.

Record expected completion dates, units still being finished, units already committed and units currently offered to customers. Keep those counts separate so that a physical delivery forecast does not silently become an available-inventory forecast.

For an apartment developer, the useful follow-up is whether nearby buildings are approaching delivery and how many units remain unleased. For a homebuilder, it is the number of competing completed homes still available at the relevant price point.