July 2026 Permits and Starts: Reading the Split

July 2026 permits and starts diverged: authorizations rose 5.0% from June, while starts fell 12.4%. The total starts decline was statistically significant; the reported single-family monthly decline was not. Those figures show more approvals alongside a slower pace of construction beginning, without establishing when homes will become available. Census/HUD release

The original July estimates

Census and HUD published release CB26-127 on August 18, 2026. July estimates were preliminary; the June comparisons below use revisions from that same release. Keep the release date attached when comparing housing data and reports, because later downloads can incorporate revisions.

All levels are seasonally adjusted annual rates, from the original August 18 report.

An annual rate expresses the seasonally adjusted month's pace multiplied by 12. It does not count homes built during July or forecast the next year's output. Dividing by 12 yields an adjusted monthly figure, not the unadjusted estimate. Census annual-rate explanation

The unadjusted July estimates—activity during the month—were 129,900 units authorized, 111,200 started and 105,100 completed. Original monthly tables

Why permits can rise while starts fall

A permit records local authorization. A start occurs when excavation begins for footings or foundations. For a single-family house, completion generally means finished flooring is installed, though earlier occupancy also qualifies. In buildings with two or more units, all units count as completed when at least half are occupied or available for occupancy. Census definitions

The counts measure housing units, not buildings. A hypothetical 100-unit apartment building contributes 100 starts when foundation excavation begins. A few large projects can therefore move multifamily results. Census counting rules

These stages need not involve the same projects in a given month. July starts can follow earlier permits, and July completions can follow much earlier starts. Projects can also be abandoned or redesigned after authorization. Census explanation of the relationships

The inventory of units authorized but not started provides a better cross-check than subtracting the headline rates. It reached a seasonally adjusted 279,000 at July's end, up from 268,000 in June. The 4.1% increase, with a 90% margin of ±2.5 percentage points, was statistically significant. These are end-of-month units, not annual rates. Table 2

That supports an interpretation of more units waiting between approval and construction. It does not identify the cause or establish a delivery schedule for new-construction activity.

Which changes were statistically significant?

Census reports 90% confidence intervals for starts and completions. A change is statistically distinguishable from zero when its interval excludes zero. The intervals cover sampling variability; reporting and coverage errors remain outside them. Survey methodology

Margins in the table are percentage points. The reported estimates and margins support a total starts decline, but leave the direction of the other monthly changes uncertain. Lack of statistical significance does not establish that activity was unchanged.

Compared with July 2025, total starts fell 13.5% ±11.0 percentage points and completions fell 16.8% ±13.7 points; both declines were statistically significant. Permit changes have no comparable sampling interval in this release and remain subject to nonsampling error. Release notes

Building type and region change the picture

Single-family permits rose 2.5% to an annual rate of 894,000. Starts were 808,000 and completions 878,000. While the monthly starts decline was uncertain, the decline from July 2025—15.7% ±8.6 percentage points—was statistically significant. That provides stronger evidence of weakness relative to a year earlier than of a new July turning point. Tables 1, 3 and 5

For buildings with five or more units, permits rose 9.1% to 490,000, while starts fell 15.6% ±20.2 percentage points to 421,000. That starts interval includes zero. Completions were 329,000. These annual rates exclude the separate two-to-four-unit category. Original tables

Building size does not establish tenure: multifamily construction can include condominiums, so five-plus-unit authorizations cannot all be treated as future rental apartments. Census definitions

Regional results also differed. Midwest starts fell 27.6% ±16.5 percentage points, a statistically significant monthly decline. The Northeast's estimated 17.1% increase had a ±58.4-point margin, leaving its direction uncertain. Regional starts table

What the pipeline establishes for a development decision

Even completed units do not all become unsold homes available to buyers. Construction includes homes built for rent, owner-built homes and custom homes, as well as homes built for sale. Census comparison of construction and sales A home built for sale may already be under contract at completion. Census sales definitions That is why completed for-sale inventory is a separate measure.

For land and development pipelines, the national release frames questions that need local answers: which permits remain active, which sites have started, and which competing units have credible completion dates and remain available. It cannot establish competition at a particular price or whether a housing shortage has disappeared. Before changing a budget or negotiating a lot purchase, verify the local projects behind the assumed supply and compare available homes with demand.