Skip to content

United States · Book publishers · 2024

What public records reveal about book publishing’s county footprint

The public county-coded employment footprint exceeds its establishment counterpart at fixed largest-record counts, even under a conditional one-unit allowance. Yet 1,318 establishments have undefined geography: the result is not a verdict on true physical concentration.

Published
Share
Share this page
EmailCite & reuse this study
Key findings as text
Reusable imagesWhat public records reveal about book publishing’s county footprint — 888: Published genuine county-coded records. Private Book publishers, 2024; 166 disclose employment and 722 suppress it.Download sharing image (JPG) ↓

Key findings

888

Published genuine county-coded records

Private Book publishers, 2024; 166 disclose employment and 722 suppress it.

29.13%

Unresolved national employment pool

15,596 of 53,534 annual average jobs; not assigned to individual counties.

1,318

Establishments in undefined location buckets

Retained in national denominator, outside genuine county top-k selections.

31.78%–61.33%

Conditional employment envelope for ten records

National-share outer envelope under ±1 assumption; not confidence or physical concentration.

A strong contrast, but not a map of every workplace

Public employment data can support a clear comparison without revealing every confidential county count. In the 2024 private book-publishing records examined here, the employment represented by the largest published county-coded records occupies a substantially greater national share than the corresponding largest establishment records. That statement survives a deliberately generous one-unit sensitivity allowance around published integers. It does not establish that actual physical jobs are more geographically concentrated than actual physical establishments. The difference between those statements is the central finding, rather than a qualification to hide after a map.

The distinction matters because geographic coding is uneven across the two measures. The national establishment count is 5,426, but 1,318 establishments appear in unknown or undefined location buckets rather than genuine county-coded records. Those buckets represent approximately one quarter of the national establishment denominator. Employment also has undefined geography, while hundreds of county employment cells are withheld. A comparison that deletes these records, changes the denominator, or labels their values zero can appear precise while answering a different question.

This study therefore asks a narrower, reproducible question: what national shares can the largest one, five and ten published genuine county-coded records represent, given the information actually released? The answer is expressed as conservative outer envelopes, not estimates of confidential cells. It measures a public reporting footprint. It does not rank named employers, reconstruct suppressed employment, or assign undefined records to imagined counties. The resulting limits are useful precisely because they show both a defensible comparison and the stronger geographic interpretation that the evidence cannot support.

One industry, one annual frame, several different units

The source is the Bureau of Labor Statistics Quarterly Census of Employment and Wages, using its 2024 annual industry slice for NAICS 2022 code 513130, Book publishers. Only private ownership is retained. The national denominator excludes Puerto Rico and the Virgin Islands, as specified in the BLS aggregation-code documentation; their county-level records are consequently excluded from this analysis. State, metropolitan and national aggregate records are not added to county records. They describe overlapping geographic layers, not additional jobs.

The annual employment variable averages twelve monthly employment levels. The annual establishment variable averages four quarterly establishment counts. They are different annual summaries, not a headcount and a synchronized photograph of businesses on one date. Covered payroll jobs are not unique people: a person can hold more than one job. Establishments are not necessarily independent firms, publishing imprints or corporate headquarters. Neither measure counts authors as a population, establishes employee residence, or identifies where a remote worker spends a working day. QCEW excludes unincorporated self-employed workers and unpaid family workers; it does not cover every person producing or self-publishing books.

Industry scope also needs care. The 2022 definition includes book publishing in print, electronic and audio forms, including exclusively Internet publishing. It should not be treated as a cosmetic renaming of the earlier 511130 category. A temporal comparison would need to address the classification change rather than join the two codes automatically. This study avoids that problem by examining one complete annual vintage. It makes no assertion about employment recovery, future hiring, industry health or the effect of digital publishing.

Three kinds of incomplete geography

After filtering ownership, industry, year and domestic geography, the slice contains 930 county-level records. Of these, 888 have genuine county codes and 42 are unknown or undefined location buckets. Among the genuine county-coded records, 166 disclose employment and 722 suppress it. The undefined group contains 35 disclosed employment values and seven suppressed values. These are counts of published records in this particular slice, not a claim that every county in the country has an observed industry record.

Suppressed, undefined and absent are different conditions. Suppression means that a published record exists but its employment value is withheld. The raw file uses a zero placeholder together with the disclosure flag for those cells; the placeholder is not observed zero employment. Undefined geography means that the record belongs to a special location bucket. Absence means that no corresponding county record is observed in the selected file. This analysis neither converts absent counties into zeros nor treats the three conditions as interchangeable missing data.

BLS explains that unknown or undefined location can describe a worksite without a defined physical address, or an aggregate of employees without a physical location specific to their economic activity. It is therefore inaccurate to describe every undefined record as a misplaced street address waiting to be repaired. Allocating all such establishments across physical counties would introduce a geographic model that the source does not supply. Keeping them outside the county top-k calculation, while retaining the national denominator, makes that information boundary visible.

Start with the national accounting identity

National annual average employment is 53,534. Disclosed genuine county-coded employment sums to 32,298, and disclosed undefined employment sums to 5,640. Subtracting both known components from the national total leaves an unresolved pool of 15,596, or 29.13% of national employment. This pool is an accounting remainder, not an estimate of any particular suppressed county. Some of it may belong to suppressed undefined records, and the calculation does not attempt to determine which ones.

Establishment counts are published even for records whose employment is suppressed. Genuine county-coded establishment values sum to 4,098, and undefined establishment values sum to 1,318. Their sum, 5,416, is ten below the national value of 5,426. BLS warns that rounding can create arithmetic discrepancies, so the gap is compatible with rounding. Its cause is not established here. No balancing county or additional establishment category is invented, and the observed values are not altered to force agreement.

Forty genuine county-coded records display zero annual average establishments. They remain in the data. A displayed annual integer zero is not proof of no activity, just as a positive establishment average is not sufficient justification for imposing a positive lower bound on a suppressed annual employment cell. The calculations use nonnegativity but do not add undocumented minimum staffing rules. This conservative treatment avoids extracting confidential information from assumptions that look intuitive but have not been established for the published measure.

How an outer envelope works

For a chosen k, add the k largest disclosed genuine county employment values. That sum supplies a lower bound on the employment represented by the largest k records in the published genuine county-coded universe: adding undisclosed nonnegative values cannot make the largest-k total smaller. Divide by national employment to retain the intended denominator. The statement concerns the existence of a set of k records with at least that share. It does not certify that the disclosed records are the actual largest records after confidential values are considered.

For an upper envelope, add the entire unresolved national pool to the same disclosed largest-k sum, then cap the result at the national total. This is deliberately permissive. It allows unresolved employment to contribute even when some actually belongs to undefined locations, and does not fully exploit how many records can fit within k. The result is therefore a conservative outer envelope, not the narrowest mathematically attainable interval. Its width honestly reflects both confidentiality and the decision not to reconstruct suppressed cells.

The establishment comparison is calculated separately from its own largest k genuine county-coded values and its own national denominator. The employment-selected and establishment-selected sets need not contain the same counties. This is not a comparison of the staffing levels of a common selected group. Nor is an establishment share a share of independent employers. These distinctions let the calculation compare reporting footprints without silently turning it into a study of firm ownership, employer concentration or local labor-market power.

What the displayed values support

Using the displayed integers directly, the employment envelope for one record is 16.89% to 46.03% of national employment. For five records it is 25.21% to 54.34%, and for ten it is 31.80% to 60.93%. The corresponding displayed establishment shares are 3.19%, 8.39% and 12.73% of national establishments. The employment envelopes are wide, but the lower ends already exceed the establishment shares at each prespecified value of k.

That result does not require guessing any of the 722 suppressed genuine county employment values. The known largest disclosed employment value alone is 9,044; the largest five disclosed values sum to 13,495, and the largest ten to 17,023. The corresponding establishment sums, selected independently, are 173, 455 and 691. These aggregate inputs permit reproduction without publishing a named county leaderboard or suggesting that confidential employers have been identified.

There are two different messages in the width and separation of the envelopes. Their width says that the exact largest-k employment share is not identified by this exercise. Their separation from the establishment values says that a specific ordering of national reporting shares can still be supported. A dataset can be inadequate for an exact geographic ranking while remaining informative about a limited aggregate comparison. Neither message cancels the other, and neither justifies describing the upper endpoint as the most likely value.

A conditional allowance for published integers

The main arithmetic treats published integers as displayed values. A second calculation asks whether the contrast survives allowing every published count, including each national denominator, to differ by up to one unit in either direction, with lower endpoints truncated at zero. This is an explicit sensitivity assumption. Although BLS documents rounding discrepancies, the inspected documentation does not certify this one-unit allowance as an unconditional error limit. The resulting intervals are not confidence intervals and do not quantify sampling probability, source error or coverage uncertainty.

Under that conditional allowance, employment spans 16.89% to 46.41% for one record, 25.20% to 54.73% for five, and 31.78% to 61.33% for ten. Establishment envelopes are 3.17% to 3.21%, 8.29% to 8.48%, and 12.55% to 12.92%, respectively. At every fixed k, the employment lower endpoint remains above the establishment upper endpoint. The separation is consequently not an artifact of ignoring a one-unit perturbation of the displayed counts within this specified calculation.

The sensitivity upper endpoint is intentionally conservative in another way: it combines national, disclosed county and disclosed undefined limits in the direction that permits a larger unresolved contribution. These combinations need not correspond to one uniquely realizable allocation. Calling the interval an outer envelope is important. A reader should not infer that every point inside it is achievable, that its extremes are equally plausible, or that an optimization model has proved the limits sharp.

Why the result is not physical geographic dominance

The large undefined establishment component prevents a stronger conclusion about actual physical geography. The 1,318 undefined establishments remain in the national denominator but cannot enter the largest-k genuine county numerator. That reduces the county-coded establishment share relative to a hypothetical dataset with complete county allocation. Employment has its own disclosure and location issues. Comparing the two observed reporting footprints is valid as defined, but their different coding coverage means the comparison does not isolate physical concentration alone.

One might propose assigning every undefined establishment to counties and recomputing concentration. Without evidence about the appropriate allocation, that would be a different, model-dependent exercise. Moreover, some undefined records do not describe a single specific economic-activity location. The present study therefore does not manufacture a physical map from administrative uncertainty. It reports the undefined component alongside the primary comparison so readers can see why the strongest tempting headline is not warranted.

The national denominator is equally important. Dividing only by disclosed genuine county employment would remove both undefined employment and the unresolved pool. It would describe concentration within a selected disclosed subset, not a national employment share. Such a calculation can produce larger percentages without discovering any additional geographic concentration. This paper retains national denominators throughout its primary tables and figures, and does not present a disclosed-only normalization as an alternative estimate of the same target.

What this adds to established publishing geography

The existence of major publishing centers is not a new discovery. New York City's official publishing-industry economic-impact study already examines geographic clusters and compares cities. Its industry framework and much of its data construction differ from this single-industry QCEW exercise: it uses a broader custom publishing sector and incorporates additional employment sources for much of its analysis. Its book-publishing comparisons are relevant prior work, not a directly comparable series to splice into the present figures.

The contribution here is a measurement question rather than another cluster map. It separates confidentiality from undefined geography, retains the national denominator, and shows exactly which comparison survives those constraints. It also makes an important asymmetry visible: establishments are numerically published, yet many lack genuine county coding. Complete numeric disclosure of one variable does not automatically imply complete geographic information. That lesson would be lost in a conventional chart of the largest visible employment locations.

The calculation uses established bounding logic and makes no claim to a new statistical theory or the first study of publishing concentration. Its practical value is an auditable route from released fields to a restrained conclusion. Readers assessing geographic claims can ask whether withheld cells were treated as zero, whether undefined records disappeared, whether the denominator changed, and whether rankings were asserted despite missing values. Those questions are answerable from the accompanying inputs and code rather than from a persuasive map alone.

Reproduction, limitations and a usable conclusion

The accompanying selected-input file retains the national private record and the domestic county-level records needed for this calculation, with disclosure flags preserved. Suppressed employment is represented as unavailable in that analytical file rather than as an observed zero. The original frozen source fingerprint and source URL are supplied separately. A standard-library Python script reproduces the accounting totals, the displayed-value envelopes and the conditional one-unit calculations from explicit input and output paths; it does not contact employers or recover confidential cells.

Source construction remains a limitation distinct from this analysis. BLS can estimate or impute late employer reports; the claim here is only that this analysis does not impute suppressed cells. Coverage exclusions, revisions, reporting error and geographic coding are not bounded by the one-unit exercise. The fixed-year scope also avoids treating county-equivalent changes or industry reclassification as economic movement. Protocol and scope refinements were recorded during source qualification, but this is exploratory secondary analysis, not an externally preregistered experiment.

The usable conclusion is deliberately two-part. Public records support a separated comparison between county-coded employment and establishment national-share envelopes at one, five and ten records. They do not establish the corresponding ordering of true physical geographic concentration. Keeping both parts together gives readers something stronger than a generic warning about missing data: a reproducible statement of what remains identifiable, alongside an explicit boundary around what the source cannot settle.

Published county codes do not cover the same national shares

United States · BLS QCEW 2024 · Private book publishing, NAICS 513130

Full chart and source
Published county codes do not cover the same national shares. Employment: 32,298 disclosed county-coded + 5,640 disclosed undefined + 15,596 unresolved.
Establishments: 4,098 county-coded + 1,318 undefined = 5,416; national total 5,426.
The −10 difference has unknown cause; no balancing category or county allocation is invented.
The unresolved pool includes suppressed county-coded AND undefined employment.
Published reporting footprint, not complete physical geography. Absent county rows are not asserted zero.

Scroll horizontally to inspect the chart. ↔

Chart and values

Scroll horizontally to see all columns.

National reporting coverage and reconciliation
MeasureNational totalDisclosed genuine county-codedDisclosed undefinedUnresolved employmentCounty-level sum minus national
Annual average employment53,53432,2985,64015,596Not applicable
Annual average establishments5,4264,0981,318Not applicable−10

Employment unresolved is a remainder, not an observed geographic category. Establishment gap has unknown cause; no balancing category added.

Employment: 32,298 disclosed county-coded + 5,640 disclosed undefined + 15,596 unresolved. Establishments: 4,098 county-coded + 1,318 undefined = 5,416; national total 5,426. The −10 difference has unknown cause; no balancing category or county allocation is invented. The unresolved pool includes suppressed county-coded AND undefined employment. Published reporting footprint, not complete physical geography. Absent county rows are not asserted zero. Source: U.S. Bureau of Labor Statistics · Public-domain data · Analysis: Alisson Bento

Download chart (SVG) ↓ · Download chart (PNG) ↓

Conditional outer envelopes for county-coded top-k shares

United States · BLS QCEW 2024 · Private book publishing, NAICS 513130

Full chart and source
Conditional outer envelopes for county-coded top-k shares. Top-k sets are selected separately for each measure among published county-coded records.
Conditional ±1 sensitivity per published annual integer, including the national denominator.
Outer accounting envelopes, not confidence intervals, certified error limits or sharp bounds.
Undefined establishments (1,318) stay outside the top-k numerator. No physical-concentration conclusion.

Scroll horizontally to inspect the chart. ↔

Chart and values

Scroll horizontally to see all columns.

Conditional one-unit national-share envelopes
k recordsEmployment lower (%)Employment upper (%)Establishments lower (%)Establishments upper (%)
116.8946.413.173.21
525.2054.738.298.48
1031.7861.3312.5512.92

Percent of the respective national total. Conditional ±1 per published integer; not confidence intervals or physical geographic concentration. Separate largest-k selections within 888 published genuine county-coded records.

Top-k sets are selected separately for each measure among published county-coded records. Conditional ±1 sensitivity per published annual integer, including the national denominator. Outer accounting envelopes, not confidence intervals, certified error limits or sharp bounds. Undefined establishments (1,318) stay outside the top-k numerator. No physical-concentration conclusion. Source: U.S. Bureau of Labor Statistics · Public-domain data · Analysis: Alisson Bento

Download chart (SVG) ↓ · Download chart (PNG) ↓

Data & downloads

Scroll horizontally to see all columns.

Conditional one-unit national-share envelopes
k recordsEmployment lower (%)Employment upper (%)Establishments lower (%)Establishments upper (%)
116.8946.413.173.21
525.2054.738.298.48
1031.7861.3312.5512.92

Percent of the respective national total. Conditional ±1 per published integer; not confidence intervals or physical geographic concentration. Separate largest-k selections within 888 published genuine county-coded records.

Scroll horizontally to see all columns.

National reporting coverage and reconciliation
MeasureNational totalDisclosed genuine county-codedDisclosed undefinedUnresolved employmentCounty-level sum minus national
Annual average employment53,53432,2985,64015,596Not applicable
Annual average establishments5,4264,0981,318Not applicable−10

Employment unresolved is a remainder, not an observed geographic category. Establishment gap has unknown cause; no balancing category added.

Scroll horizontally to see all columns.

Employment disclosure by record type
Record typeEmployment disclosedEmployment suppressedTotal records
Genuine county-coded166722888
Unknown or undefined35742

Domestic published county-level private records only. Absent county rows are unobserved, not zero.

Evidence & method

How we calculated it

  1. Exploratory secondary analysis of 2024 annual QCEW NAICS 2022 513130, private ownership 5.
  2. National denominator excludes Puerto Rico/Virgin Islands; only domestic county aggregation 78 enters selected record universe.
  3. Fixed k=1,5,10; nonnegative suppressed employment, no cell imputation. Conservative outer envelopes are not sharp.
  4. Conditional ±1 sensitivity is an assumption, not an official rounding guarantee or confidence interval.
  5. Original source and selected inputs pinned; standard-library reproduction with explicit paths.

What the data cannot tell us

  • County-coded national reporting footprints are not complete physical geography. Undefined establishment coverage asymmetry is central.
  • Absent county rows are unobserved, not zero. Employment suppression flags are preserved.
  • Ten-establishment reconciliation gap has unknown cause; no repair or invented balancing record.
  • Source may impute late reports; our analysis does not impute suppressed cells.
  • Annual covered jobs are not unique people, firms, authors, residences or remote-work locations. No causal or temporal claims.

Original sources

  1. BLS QCEW 2024 annual Book publishers industry slice

    Frozen annual NAICS 2022 513130 data; private ownership 5, national aggregation 18 and domestic county aggregation 78. Suppressed employment is unavailable, not zero.

    Accessed October 30, 2026

  2. BLS QCEW 2024 annual methods

    Coverage, annual averaging, disclosure, rounding discrepancies and source estimation. Employment averages 12 months; establishments 4 quarters.

    Accessed October 30, 2026

  3. BLS QCEW area guide

    Unknown or undefined locations can lack a defined address or represent activity without a specific physical location; these are not genuine counties.

    Accessed October 30, 2026

  4. BLS QCEW aggregation codes

    Codes 18 and 78 distinguish national from county totals; note(b) excludes Puerto Rico and Virgin Islands from national aggregations.

    Accessed October 30, 2026

  5. BLS QCEW CSV data slices

    Documented annual industry slice. Absence of a county record is not converted to an observed zero.

    Accessed October 30, 2026

  6. US Census Bureau 2022 NAICS manual

    Book publishers 513130 includes print, electronic and audio, including exclusively Internet. Earlier 511130 is not silently spliced into this frame.

    Accessed October 30, 2026

  7. New York City publishing-industry economic-impact study

    Prior geographic analysis; broader custom sector and mixed data construction. No first-ever cluster claim or direct historical numerical comparison.

    Accessed October 30, 2026

  8. BLS public-domain reuse policy

    Government numeric data are public domain with attribution requested. No BLS emblem or third-party report imagery reproduced.

    Accessed October 30, 2026

Cite & reuse this study

Alisson Bento. What public records reveal about book publishing’s county footprint. Secondary analysis of BLS QCEW 2024 annual private Book publishers records. Sources accessed October 30, 2026. https://alissonbento.com/research/us-book-publishing-county-reporting-2024/

BLS government numeric data are public domain; source credit: US Bureau of Labor Statistics, QCEW 2024. Original calculations and explanation: Alisson Bento. No official endorsement, agency emblem or third-party report artwork.

Corrections or editorial questions

Pass it on.

What public records reveal about book publishing’s county footprint | Alisson Bentoalissonbento.com

WhatsAppFacebookXTelegramRedditThreadsEmailSMS