July 2026 New Home Sales: Read the Uncertainty

July 2026 new home sales fell an estimated 10.5% from June, but the decline was not statistically significant at the 90% confidence level, according to Census and HUD’s initial release. That uncertainty makes the weaker reading a reason to examine local orders before cutting a builder’s demand forecast.

What July’s confidence interval establishes

Released August 25, 2026, the report estimated July sales at a seasonally adjusted annual rate of 607,000, compared with a revised June rate of 678,000. The monthly change carries a margin of ±14.0 percentage points, giving a calculated interval of −24.5% to +3.5%. Because that range includes zero, the direction of the national change remains uncertain. July release

The lack of statistical significance does not establish that demand was unchanged. A substantial decline remains within the interval, alongside a modest increase. Census’s intervals address sampling variability; they do not capture every possible reporting or coverage error. Release explanatory notes

The regional results differ. The Midwest’s estimated monthly decline was 42.7%, with a ±13.4-percentage-point margin that excludes zero. That change was statistically significant at the 90% confidence level. Regional table For a builder concentrated there, the result warrants closer scrutiny of local orders, though it cannot establish an individual community’s sales trajectory.

What Census counts as a new home sale

Census records a sale when a sales agreement is signed or a deposit accepted. The house can be unstarted, under construction or completed. July release and construction-stage table

The measure covers new single-family houses built for sale. It excludes homes built for rent and homes a contractor builds on an owner’s land. Census’s June 2022 scope presentation, slide 17

The annual rate is the seasonally adjusted monthly estimate multiplied by 12. It describes July’s pace rather than predicting the year’s total. Census annual-rate explanation

For sales planning, treat the count as evidence of purchase commitments. A delivery forecast requires another judgment: which commitments will close, and when. When reviewing new-construction activity, keep those commitments distinct from expected deliveries and from the separate permits-and-starts pipeline.

How canceled contracts affect the count

Census does not subtract canceled sales contracts. After recording a sale, it stops asking about the home’s sale status, removes it from for-sale inventory and does not count a subsequent purchase by another buyer. The count therefore does not verify that the original transaction closed. Census cancellation explanation, revised April 22, 2020

Consider a hypothetical home first contracted in July, canceled in August and contracted to another buyer in September. Under that method, it remains a July sale: the cancellation does not reverse it, and the second contract does not create another counted sale.

Census explains that high cancellations in a worsening market can temporarily overstate sales; later purchases of those homes can produce understatement as conditions improve. It says these timing effects do not create a long-run overcount or undercount. This does not establish July’s cancellation rate or justify subtracting an assumed percentage from the estimate. Cancellation methodology

Why the release date matters

Revisions have already changed the comparison. The July 24 release put June sales at a 628,000 annual rate. The August 25 report raised that to 678,000 and marked April and May revised as well. July’s reported decline uses that updated June figure.

Census’s sample is drawn primarily from building permits, so it must estimate sales agreed before permits are issued. Those preliminary estimates are subject to revision. July explanatory notes

Keep the release date attached to every comparison in the forecast memo, including comparisons with other housing data and reports.

What would justify changing a builder’s forecast?

A forecast review should identify which assumption has weakened: the pace of new orders, the share of contracts expected to close, or the timing of deliveries. Each calls for different evidence.

• New orders: Compare contracts per active selling community across several periods, using comparable communities and similar seasonal windows. Record changes in available lots, product mix, prices and incentives. Stable orders obtained with larger concessions deserve a different margin assumption from stable orders at unchanged terms.
• Contract survival: Reconcile gross orders, cancellations and net orders. Track cancellations by original contract month and reason, as well as by the month they occur. This helps separate problems in older backlog from those affecting recent contracts.
• Expected closings: Review financing progress and construction readiness for homes scheduled to close. A delivery delay may change cash timing without changing the assumption about new buyer demand.

For a hypothetical builder that defines net orders as gross orders minus cancellations, 20 new contracts and two cancellations yield 18 net orders. If new contracts remain at 20 but cancellations rise to five, net orders fall to 15. The ledger identifies deterioration that a count of newly signed contracts alone would miss. This calculation is not an adjustment to Census data.

Repeated weakness across comparable communities, supported by deteriorating contract survival or heavier concessions, would provide a stronger basis for revising assumptions than the national headline alone. If operating evidence remains steady, a downside scenario may be more proportionate than an immediate cut to the base forecast.

Before the next review, specify which changes in community orders, cancellations or closing schedules would trigger a forecast revision.