What Census Manufacturing Expense Tables Can and Cannot Tell a Buyer

Census released the full 2024 AIES dataset in September 2026, and its release announcement calls it a single, comprehensive annual snapshot of U.S. employer businesses, with data at the national, regional and divisional levels and state-level estimates for selected sectors. The same announcement puts food manufacturing at an estimated value of shipments of $1,054.3 billion. Industry figures that large invite a familiar shortcut in sourcing business cases: the industry spends a given share on a line, we spend more, so the difference is the saving. This article sets out what the expense tables measure, how their lines map to a category tree, and which checks come before any figure enters a plan.

Which 2024 AIES tables report manufacturing operating expenses?

The survey is new, while its manufacturing content has a long history. In the Census Bureau's description of the AIES, the survey replaced and integrated seven annual business surveys into one survey in 2023. The Federal Register notice that proposed it lists the Annual Survey of Manufactures among the collections it replaces and says that survey had been conducted annually since 1948 to collect revenues, expenses, capital expenditures, fuels and electric energy used, and inventories in the manufacturing sector.

Two 2024 tables matter most to a manufacturing buyer. Selected Sectors: Detailed Operating Expenses for Employer Firms, table AIES00EXP02, splits operating expenses into lines and is shown at the U.S. level only, at select 2- through 6-digit 2017 NAICS code levels. The manufacturing summary statistics, table AIES31BASIC01, are shown at the 2- through 6-digit levels for the manufacturing sector and carry summary measures such as total cost of materials. The panel above draws on the detailed expense file and shows each line's share of an industry's operating expenses, with the lines Census does not report for manufacturing listed beneath it.

Who does the survey count, and what does each line include?

The survey counts companies, and it reports their activity by industry. The AIES FAQ defines a company as an organization that owns or operates one or more locations nationwide and an industry as a group of locations providing similar products, services or activities, and its main survey content is asked by individual location or for groups of locations based on industry, so one company's locations can be reported under more than one industry. According to the 2024 release, a sample of approximately 385,000 businesses is selected based on geographic location and industry, and the estimates are not adjusted for price changes.

Line names are shorter than their definitions, and the definitions decide where a line belongs in a category tree. In the manufacturing table's item definitions, purchased professional and technical services include janitorial, security and landscape services alongside consulting, legal and engineering work. For manufacturing, the definition of the repairs line for machinery, vehicles and equipment also takes in repairs to integral parts of buildings, such as elevators and heating systems, while excluding janitorial and grounds maintenance. The materials, parts and supplies line includes materials used in repairs, office and janitorial supplies, small tools and packaging, and for manufacturers it also includes motor fuels. Purchased electricity counts only electricity paid directly to the utility company, not electricity within lease or rental payments.

Which expense lines can a buyer map to a category tree?

The two tables overlap, and the overlap is where double counting starts. The manufacturing summary table defines total cost of materials as the cost of materials, packaging and similar items used, plus resales, purchased fuels consumed, purchased electricity and contract work, and it defines contract work as work done by others on materials furnished by the manufacturing establishment. The detailed expense table carries each of those parts as a line of its own, so adding the summary total to the expense lines counts the same dollars twice. The first finding panel in this section checks that arithmetic against the companion summary file AIES31BASIC02, which carries total cost of materials at the two- to four-digit levels, for every industry where the summary file publishes the total and the expense table publishes all five parts.

A line can scope a category only after its definition has been matched to the category's own boundary. The register below is our proposal for doing that once per line, with the internal record that answers the question the Census line cannot.

What do the flags and coefficients of variation limit?

Every cell in the expense table carries a status, and a missing figure has a reason. Under the expense table's notes, data rows with fewer than three contributing firms or three contributing establishments are not presented at all. Among the rows that are shown, the table's symbols mark a value withheld to avoid disclosing data for individual companies with D, though it is included in higher level totals; a value not available or not comparable with N; an estimate that does not meet publication standards because of high sampling variability, poor response quality or other concerns about its quality with S; and a value that rounds to zero with Z. The S definition adds that unpublished estimates derived from the table by subtraction carry the same limitations and should not be attributed to the Census Bureau, so a figure backed out of a total is the team's own estimate, not a Census one. The second finding panel counts these cells line by line.

A published figure is still an estimate from a sample. The same notes state that each estimate is subject to sampling variability and may differ from the results of a complete census, and that estimates with large coefficients of variation should be used with caution. The notes mark a coefficient of variation between 30.0% and 60.8% with v, as a sign of high sampling error, and one that exceeds 60.8% with w, which means the estimate is not significantly different from zero at a 90% confidence level.

Sampling is not the only source of error. The 2024 release says the estimates may be subject to sampling variability and other sources of error, and imputation is one that has been measured in earlier manufacturing data. In a peer-reviewed study of Census manufacturing data, published in 2018 in the Review of Economics and Statistics, researchers found that 79% and 73% of observations in the 2002 and 2007 Censuses of Manufactures had imputed data for at least one variable used to compute total factor productivity, and that the mean-imputation methods the bureau primarily used can reduce the variance of the imputed variables. That study describes earlier censuses, not the AIES, so it gives no imputation rate for any 2024 line. Its lesson for a buyer is narrower: a published figure can rest partly on imputed values, so a decision about one plant belongs on that plant's own records.

What does a buyer still need from internal records?

The survey describes industries, and a sourcing decision needs one company's figures. Before a Census line moves a category up the plan, start from the general ledger accounts that match it, then cut the accounts payable and purchase-order detail by supplier, site and commodity code, including non-PO invoices and card spend, with the contracts, volumes and price terms behind it. Map each plant to its own six-digit NAICS industry rather than the company's headline code, and where that cell is withheld or flagged, use the four- or five-digit level and say so. Before reading anything into a gap between your share and the industry's, check the choices that move it, such as making versus buying, outsourced versus in-house maintenance, leasing versus owning, product mix and energy intensity. Where the question concerns a supplier rather than your own spend, start from records the supplier can share, as set out in our guide to supplier risk assessment methods.

The register also works in the other direction. When a supplier cites an industry figure to support a price change, ask which table, line, NAICS level and flag it comes from. An AIES line is an annual dollar total for the supplier's whole industry, so it cannot show a change in a unit cost: take any cost weight from the supplier's verified cost breakdown and any price movement from the index the contract names, which gives supplier relationship reviews a shared fact base.

Where does a public expense table belong in category planning?

For the next category review, leave out materials, payroll, fringe benefits and the catch-all other operating expenses, take the three largest remaining lines for the industry your main plant sits in, which are often cost of resale, machinery repairs, leases or electricity, and check each one's flag and coefficient of variation in the panel above. Pull the plant's spend on the same lines from the spend cube or accounts payable, ask the plant controller to confirm where the plant books items such as janitorial services or inbound freight, and divide each line by the plant's own operating expenses, built the way the table's total adds up for manufacturing, as the sum of all its published lines with payroll, materials and resale included. A gap tells you where to look next, not what you can save.