US Census Bureau — Economic Census, ASM & County Business Patterns (NAICS 3231)

It is the federal government's complete accounting of American print, delivered as typed rows. Three linked programs sit behind one record. County Business Patterns is the yearly layer: number of establishments, employment during the week of March 12, first-quarter payroll and annual payroll, every shop sorted into one of nine employee-size classes, tabulated for states, metropolitan areas, counties and ZIP Codes. The Economic Census is the five-year benchmark - reference years 2017 and 2022 - carrying product-line detail under NAPCS that the yearly files lack, and serving as the standard other federal series reconcile to, GDP and producer-price work included. The Annual Survey of Manufactures, now folded into the Annual Integrated Economic Survey, keeps the manufacturing dollar measures - shipments, value added, capital expenditures, inventories - running between benchmarks.

Filtered to NAICS 32311, the 2023 reference year counts 21,354 commercial printing establishments employing about 369,691 people for roughly $21.2 billion in annual payroll, and the parent sector 323 rolls up to 22,301 establishments in the identical layout. Within Datadory's catalog of 1,744 datasets across 159 viable industries, this record scores 9/10 against a cross-catalog mean of 7.81 - one of twenty records in the commercial printing slice. Get a sample of this dataset cut to your geographies, or read the evidence below first.

The 2023 reference year, straight from the NAICS 32311 slice:

# US totals, NAICS 32311 Printing - reference year 2023
fipstate : 98         US total
naics    : 32311/     Printing   (trailing slash = sub-industry aggregate)
est      : 21354      establishments
emp      : 369691     week-of-March-12 employees   nf G (<2% noise)
qp1      : 5307016    first-quarter payroll $k     nf G (<2% noise)
ap       : 21245723   annual payroll $k            nf G (<2% noise)

# one small county inside the same file
fipstate : 01   fipscty : 001    Autauga County, AL
est      : 4    emp : 34    qp1 : 310    ap : 1290    lfo : J

# the parent sector rolls up in the identical layout
naics    : 323///    Printing and related support activities
est      : 22301     emp : 386248    ap : 22169465

# manufacturing dollar measures join from the ASM/AIES side
valadd   : value added $k        shipdt : value of shipments $k

Three things worth noticing. First, the honesty machinery: the national row carries nf G on employment and both payroll measures, meaning under 2% disclosure noise was applied - the flags ship with the numbers so you know exactly how much smoothing happened before publication. Second, scale collapses cleanly: Autauga County's four shops employed 34 people for about $1.29 million in annual payroll inside the same columns that hold the national total, so a fifty-county rollup is a grouping exercise, not a re-modeling project. Third, the trailing slash on 32311/ marks an aggregate across sub-industries - read it as the printed sign on the door telling you which level of the hierarchy you are holding.

Thirteen entries cover the slice, and they divide into four jobs. Address: FIPSTATE and FIPSCTY locate the row to a state and county, while NAICS names the industry - 32311 isolates commercial printing, 323 its parent sector, and a trailing slash signals aggregation across sub-industries. Measures: EST counts establishments, EMP reports week-of-March-12 employment, and QP1 and AP carry first-quarter and annual payroll in thousands of dollars. Disclosure flags: EMP_NF, QP1_NF and AP_NF grade the applied protection on a G/H/J scale - G under 2%, H 2 to under 5%, J 5% or more. Structure and dollars: the N5_9 through N1000_4 columns distribute each establishment count across nine employee-size classes, and VALADD and SHIPDT bring value added and value of shipments in thousands of dollars from the manufacturing tables.

The habit worth forming early: filter on the noise flags before aggregating. A statewide payroll total built blindly over J-flagged counties blends heavily protected cells with clean ones; built flag-aware, it stays defensible.

The thirteen-entry spine covers every row; the extensions below fold under additional fields on request because they depend on vintage and geography level:

1. Legal-form-of-organization splits. Recent county vintages carry an lfo code - the sample row's lfo : J is exactly that - delivered with the full code sheet as readable labels.
2. Fine size splits. The n1000_1 through n1000_4 columns break the largest employee band apart where a vintage provides them, so the biggest print plants stop disappearing into '1,000 or more'.
3. Extra geography levels. ZIP Code industry detail, metro and micro areas, combined statistical areas and congressional districts beyond the base county dictionary.
4. Product-line detail. The 2022 Economic Census NAPCS tables, resolved to whatever digit level actually breaks out printing product classes - confirmed against the release rather than assumed.
5. Successor-survey mapping. Which tables in the Annual Integrated Economic Survey continue the manufacturing series for NAICS 32311 after the ASM's run ended, settled against live records.
6. Derived views. Payroll per employee by county, county share of state printing employment, or multi-year panels rebuilt to your footprint.

Name what your models need when you request the sample; column naming locks against live records at that point.

• Geography: seven levels deep on the patterns side - national, state, county, metropolitan and micropolitan areas, combined statistical areas, congressional districts and ZIP Codes - plus national and state cuts on the manufacturing surveys. One code system spans all of it, so a national trend traces down to a single county without changing sources.
• Temporal: the patterns archive reaches reference years 1986 through 2023; the Economic Census contributes the 2017 and 2022 benchmarks, with its product-line, firm-size and geographic releases published across 2024 through 2026; the manufacturing surveys run 2013-2021 plus a 2022 benchmark, and their successor's first main release covers reference year 2023. Newer reference years enter the archive on the Bureau's own calendar; deliveries move at the cadence you set.
• Granularity: one row per six-digit NAICS x geography x reference year, establishments spread across nine employee-size classes, payroll in thousands of dollars. There is no quarterly view and no firm-level microdata anywhere in the family - anonymity is the price of universal counting, and it is what makes the numbers usable as neutral denominators.

API, files, or your warehouse. Daily, weekly, or hourly.

Pick the channel and set the cadence - a one-time pull of the full historical depth lands as cleanly as a recurring feed of the newest reference year. Every delivery ships the field dictionary above attached, with noise flags preserved rather than smoothed away, so aggregation logic in your warehouse can weight low-noise and high-noise cells differently. Say which states, counties or metros matter when you get a sample of this dataset; the sample goes out first, and the schema in the sample is the schema you ship against.

• Territory design - rank counties by printer count and payroll so sales capacity follows where the shops actually are; the bread-and-butter move for sales growth teams.
• Market sizing - rebuild the national 21,354-establishment total bottom-up from county cells to test any narrower definition of 'the print market'; methods continue on our market sizing page.
• Demand modeling for suppliers - paper merchants and ink makers read printer concentration as a route-density map before committing distribution capacity.
• Quant inputs - multi-decade establishment and payroll series give investors and quants a structural feature for media and packaging exposures (quant backtesting).
• Research and journalism - local print-employment claims cited from a continuous federal series, the standard move for journalists and academics (citation-grade research).
• Product plumbing - builders stand dashboards on keyed county-year extracts whose layout repeats year after year.

Market researchers and consultants get the canonical print baseline - counted denominators instead of stitched trade estimates. Sales and growth teams get territory scores that survive scrutiny because no competitor is named and no estimate is spun. Data scientists and ML engineers inherit a tidy panel with stable FIPS-and-NAICS keys, typed numerics and flags that double as model features. Supply-chain and procurement desks read printer concentration as demand-side context when negotiating paper and consumables. Investors and quants read print-sector structure as an exogenous signal for advertising, media and packaging exposure with decades of history behind it. Journalists, academics and students cite the government's own counted series, which is what a claim about local print employment needs to survive an editor.

Scope note - three programs, three grains. Counts and payroll arrive every reference year; benchmark structure arrives five-yearly; manufacturing dollars arrive on the survey cycle. Mixing them in one model without labeling the grain is the classic error - deflating a shipment total with a count gets you confident nonsense.

Completeness note - the definitions above were verified against the published record layouts during the August 2026 research pass. Which successor-survey tables carry NAICS 32311 detail, and at what digit level the 2022 product-line tables break out printing, get confirmed against live records when your sample is prepared rather than claimed here.

Where to go next - the rail below collects the price layer, the value-added layer, the Canadian monthly counterpart and the glossary entries that decode the program names.