An annual pace, not one month's spending
Census estimates September construction at a $2,186.1 billion seasonally adjusted annual rate, against revised August's $2,227.4 billion. That is an annual-rate comparison, not a claim that $2.186 trillion was spent in September. The figures are dollar values, not inflation-adjusted quantities. (Sources 1–2)
New homes are only part of the residential line
Table 1 puts the private new-single-family monthly change at 0.0% and new multifamily at 0.1%: nearly flat point estimates, not proof of a housing boom. Its residential total includes improvements, so the separate 5.5% decline cannot be read as a 5.5% drop in new-home construction. Private nonresidential work moved the other way, with a 1.1% spending increase. The national decline was not uniform across categories. (Sources 2)
A small public move is not a clear decline
Public spending's estimated 0.1% decrease has a 90% confidence interval of plus or minus 1.3 percentage points. Because that range includes zero, Census says the direction of the change is uncertain. The larger total-construction decrease is 1.9%, plus or minus 0.7 percentage point on the same basis. These intervals cover sampling variability, not every source of error. (Sources 1–2)
The estimates are preliminary and can be revised. Census cautions that identifying a trend can take two months for the total and up to eight for individual categories. This release records a lower monthly spending estimate, not the cause of it, a payroll count or a verdict on the whole economy. The category labels deserve as much attention as the headline. (Sources 2)
What this record cannot settle
Preliminary September2026 U.S. construction-dollar estimates, seasonally adjusted at annual rates and not adjusted for prices. Small point changes are not necessarily statistically established. This release does not measure completed homes, payroll jobs, AI-only investment or a policy's causal effect.