Akshay DineshA monthly construction report can appear to offer a simple verdict on the economy: spending rose or spending fell. The figure is useful, but it is easy to read too much into it. Construction spending measures the value of work completed during a period, not the number of projects announced, permits issued, contracts signed, or buildings finished.
The distinction became especially visible in the U.S. Census Bureau’s September 2026 release. Total construction spending was estimated at a seasonally adjusted annual rate of $2.186 trillion, down 1.9 percent from the revised August rate. Yet the parts moved in different directions: private residential construction fell 5.5 percent, private nonresidential construction rose 1.1 percent, and public construction was nearly unchanged. A good reading begins below the headline.
What construction spending actually counts
The Census Bureau calls the measure value of construction put in place. It estimates the value of materials, labor, equipment rental, architectural and engineering work, contractor profit, and certain overhead costs attached to work installed or erected at a project site during the month. If a three-year hospital project is one-third complete, the report does not count its entire price. It assigns value as work progresses.
That makes construction spending different from a payment record. A developer may pay a bill before or after the corresponding work is performed. The survey aims to place the value in the period when the work occurred, although some categories use payments or historical progress patterns when direct monthly reports are not available.
The coverage is wider than new buildings. It includes additions, major renovations, roads, schools, water systems, power facilities, transit projects, and other fixed structures. Ordinary maintenance and repair are excluded, as are land purchases and most production machinery. Buying an empty parcel does not add to construction spending; installing structural steel on it does.
Why the headline is an annualized rate
The number most often quoted is a seasonally adjusted annual rate, or SAAR. First, statisticians adjust the monthly estimate for recurring seasonal patterns. Construction commonly slows in winter, school projects often follow a calendar, and the number of working days varies. Seasonal adjustment helps separate those expected patterns from unusual movement.
Then the adjusted monthly level is multiplied by 12. A $2.186 trillion annual rate does not mean that $2.186 trillion was spent in September. It means that if September’s adjusted pace continued for 12 months, the annual total would be about $2.186 trillion. The annualized format makes the scale easier to compare with annual measures such as gross domestic product, but it can mislead readers who treat it as cash already spent.
The release also provides unadjusted year-to-date figures. Those answer a different question. In the first nine months of 2026, actual construction spending was estimated at $1.650 trillion, 2.9 percent below the same period in 2025. For month-to-month momentum, the adjusted rate is usually more useful. For how much activity has accumulated during the calendar year, the unadjusted total is the better choice.
How one national estimate is assembled
No single ledger records every construction project in the country. The Census Bureau combines several sources. Privately owned nonresidential projects and government projects are sampled from project lists, while multifamily projects are identified through the Survey of Construction. New single-family estimates also draw on that survey. Owner-occupied home improvements rely partly on the Consumer Expenditure Survey, and specialized categories use information from other agencies and industry sources.
Selected projects report total cost, start and completion dates, and the value of work completed during the month. Survey weights expand those observations to represent the larger population. The system also adjusts for projects missing from the sampling frame, duplicate listings, unusual observations, and costs such as architectural work that may not appear in a basic site report.
Some estimates require imputation when responses are late or missing. Preliminary figures may also use forecasts for components whose source data arrive later. That is why revisions are a normal part of the process, not proof that the original report was careless. When August 2026 was first reported, its annualized total was $2.203 trillion. The September release revised August to $2.227 trillion, changing the base used to calculate September’s monthly decline.
How to read a monthly release step by step
Start with the total, then separate private from public construction. Private activity accounted for about three-quarters of the September 2026 annualized total. Within private construction, residential activity fell sharply while nonresidential activity increased. That split matters because apartment buildings, factories, data centers, offices, roads, and schools respond to different financing conditions and public decisions.
Next, compare both the previous month and the same month a year earlier. September’s total was down 1.9 percent from August and 2.7 percent from September 2025. The monthly comparison is sensitive to short-term changes and revisions. The annual comparison reduces some seasonal concern but can hide turns that happened during the intervening months. Neither comparison should stand alone.
Then read the uncertainty printed beside each percentage. The September total decline of 1.9 percent had a margin of error of plus or minus 0.7 percentage point, so the evidence supported a real decrease. Public construction’s reported 0.1 percent decline had a margin of error of plus or minus 1.3 points and carried an asterisk. In Census releases, that mark indicates there is not enough statistical evidence to conclude that the change differs from zero.
Finally, inspect the detailed categories and the revision history. A large movement in one sector can dominate the total, while smaller but broad changes may tell a different story. Avoid declaring a trend from one month. Two or three releases, viewed alongside employment, permits, housing starts, interest rates, and material prices, give a more stable picture.
What the indicator reveals, and what it cannot
Construction spending shows the pace at which planned projects are becoming physical structures. That gives it economic weight. Work put in place supports demand for labor, concrete, steel, engineering, transport, and finance. Public categories also reveal when approved infrastructure budgets are turning into roads, schools, water systems, and other assets.
It is not a pure measure of the amount of construction. A rise can reflect higher prices as well as more work, because the headline is reported in current dollars rather than inflation-adjusted units. It also arrives after permits, contracts, or groundbreaking decisions, so it is usually better at describing current production than predicting the earliest turn in the building cycle.
The categories have limits too. National estimates may hide major regional differences. A surge in manufacturing plants can coexist with weak homebuilding; a highway program can lift public spending while office construction contracts. The report describes ownership and project type during construction, not whether a project will be profitable, affordable, completed on schedule, or socially valuable.
Read carefully, the monthly release is neither a scoreboard nor a forecast. It is a disciplined estimate of work becoming real during a particular period. The most reliable interpretation combines the adjusted pace, actual year-to-date spending, sector details, margins of error, and revisions. That fuller view turns a dramatic percentage into something more useful: evidence about where building activity is actually happening.
