U.S. CONSTRUCTION · AUGUST 2026 DATA
Did U.S. Construction Rebound? Read the August Error Bars
The August construction-spending estimate increased, but the total’s uncertainty range included zero. That makes the direction of the total monthly change uncertain at the release’s 90% confidence level. Private construction had clearer evidence of a rise because its interval stayed above zero.
A single percentage hides the range of outcomes supported by a sample estimate. Plotting that range helps distinguish a positive headline from a more firmly established change.
Put the error bars back into the headline
On a narrow screen, scroll the chart sideways or use the complete text and table below.
| Spending | Estimate | 90% interval | Includes zero? |
|---|---|---|---|
| Total | +0.9% | −0.1% to +1.9% | Yes |
| Private | +1.1% | +0.6% to +1.6% | No |
| Public | +0.2% | −1.8% to +2.2% | Yes |
The total and public estimates sit to the right of zero, yet their intervals extend to the left. Private construction is the exception in this comparison: both its estimate and its entire displayed interval are positive. These observations do not prove that private and public growth rates differ statistically; testing that difference requires its own uncertainty calculation.
Open the interval arithmetic
Each lower endpoint is the point estimate minus its published margin; each upper endpoint is the estimate plus its margin. Total: 0.9 − 1.0 = −0.1 and 0.9 + 1.0 = 1.9. Private: 1.1 ± 0.5 gives 0.6 to 1.6. Public: 0.2 ± 2.0 gives −1.8 to 2.2.
The margins are measured in percentage points around a percentage change. The resulting endpoints are percentage changes, not dollar amounts. All inputs are rounded as published, so the endpoints inherit that precision.
Three periods tell three different stories
Census estimated total August spending at a $2,203.1 billion seasonally adjusted annual rate, against July’s revised $2,184.5 billion. August was 1.7% below August 2025. Spending over January–August totaled $1,450.4 billion, down 3.1% from the same months of 2025.
These comparisons can coexist. A series can turn upward during one month while remaining below its level a year earlier, and a recent improvement may be too small to offset earlier weakness in a year-to-date total. Before calling a rebound, specify which comparison has improved.
Does “annual rate” mean $2.2 trillion was spent in August?
No. An annual rate expresses the adjusted monthly pace on a yearly scale. As an arithmetic illustration, $2,203.1 billion ÷ 12 is about $183.6 billion. That is a monthly equivalent of the annualized rate, not the survey’s unadjusted August total and not a forecast of full-year spending.
By contrast, the January–August figure is the accumulated spending over those eight months. Comparing an eight-month total directly with an annualized monthly pace mixes units.
What the confidence interval can and cannot settle
The release defines its ranges as 90% confidence intervals and says they account only for sampling variability. Current-month estimates are preliminary and may be revised. The amounts are not adjusted for price changes.
“Includes zero” does not mean the actual change was exactly zero. It means the estimate is not precise enough, at the stated confidence level, to separate the change from zero. It also does not mean that every number inside the interval is equally likely.
A confidence interval is about the estimation procedure: under the statistical assumptions, repeated comparable samples and interval calculations would cover the fixed underlying value at the stated rate. It is not a guarantee that a future revision will remain inside the current interval.
For a careful summary, keep the estimate and its uncertainty together: August’s total spending estimate rose, with an interval that crossed zero; private construction showed a positive change whose interval stayed above zero.
Related questions
Does higher spending necessarily mean more construction output?
No. A dollar-spending measure can change with prices as well as the amount of work. To isolate a volume change, additional suitable price adjustment would be needed. The figures here should not be labeled real construction growth.
Can the private and public error margins simply be added?
No. Combining uncertainty requires the appropriate statistical treatment, including how estimates move together. Use the published total interval rather than creating one by adding sector margins.
Sources, dates and reuse
Primary source: U.S. Census Bureau, Monthly Construction Spending, August 2026, CB26-158, released 1 October 2026 at 10 a.m. EDT. The release PDF supplies the 90% confidence-interval definition and revision cautions. The Census economic-indicator notes explain that estimates are not price-adjusted.
Observation: August 2026, with July and year-earlier comparisons as specified. Checked 9 October 2026 UTC. The “current” page and PDF are rolling URLs: this article uses their 1 October release vintage, not a promise that those URLs will always display August data.
Original chart, arithmetic and explanation using U.S. government statistics. No source photograph or logo is reproduced. Federal-government works are generally outside U.S. copyright protection; see the U.S. Copyright Office overview. No agency endorsement is implied.
