A smaller employer-establishment total, not a general recovery
The United States had fewer employer establishments classified as Book Stores in 2023 than in 2019 in the Census Bureau's County Business Patterns data. The national total declined from 6,045 to 5,629, a difference of 416 establishments, or 6.88%. Within that smaller total, the number assigned to the under-five-employee class was almost unchanged: 2,799 in 2019 and 2,805 in 2023. The central finding is therefore a change in the composition of a classified employer sector, not evidence of a nationwide expansion in bookstore establishments.
The under-five group's share increased from 46.30% to 49.83%, or 3.53 percentage points. Its count increased by six, while the overall count declined. A larger share does not necessarily mean a much larger number. Reading the counts and shares together prevents a misleading account in which the smallest class appears to have expanded dramatically simply because it occupied more of a smaller denominator. These figures also do not identify individual establishments moving between classes: the data describe two cross-sections, not a tracked set of businesses.
The population is specific. It includes employer establishments assigned to 2017 NAICS industry 451211, Book Stores, in the 50 states and District of Columbia. It does not represent every business selling books, every independent bookshop, or every place where readers obtain books. The reference years are 2019 and 2023, not present-day conditions. The accompanying state tables retain the complete geographic coverage rather than selecting only places where counts increased. Employment comparisons are secondary because the released employment figures carry disclosure-protection noise.
What an establishment and the smallest size class mean
An establishment is a business location, not necessarily an independent company. One company can operate several establishments, and a company-level change cannot be inferred from a location count. County Business Patterns counts locations with paid employees at some time during the reference year. The annual-active employer denominator is important: it need not describe exactly the set of locations operating with employees during March. The analysis consistently calls this an employer-establishment count rather than a count of independent owners or currently open shops.
Employment, by contrast, concerns employees on payroll during the pay period including March 12. The Census definition includes full-time and part-time employees and certain paid absences. It excludes sole proprietors and partners of unincorporated businesses. The released total is not a count of full-time-equivalent positions, working hours, distinct individuals across employers, or all people working in the book trade. These distinctions remain relevant even when an employment value is published without a high-noise flag.
The under-five class deserves particular care. It includes establishments with no paid employees during the March reference period if they paid wages to an employee at another point in the year. It therefore cannot be renamed the one-to-four-worker class. Nor can it be interpreted as a permanent measure of a shop's normal staffing. Size assignment uses employment before disclosure noise is applied. The class counts answer a limited but useful question about how annual-active employer establishments were distributed around the March employment reference, without measuring annual hours, ownership arrangements, or customer-facing capacity.
The classification is consistent across these endpoints
A classification audit is necessary before comparing bookstore statistics. The newer 2022 NAICS system combines book retailing with news dealers in industry 459210 and changes how some retail activities are organized. Substituting that category for the earlier Book Stores code would not create a comparable series. A code crosswalk by itself cannot allocate establishments, employees, or size classes among the old and new categories, especially where multiple activities feed into a new industry.
That substitution is not required here. Census explicitly identifies the 2017 NAICS classification as the basis for County Business Patterns reference years 2017 through 2023. The downloaded 2019 and 2023 files both contain industry 451211. This study selects that same code at both endpoints. It does not combine 451211 with 459210 or use national conversion factors to adjust state observations. The consistency claim concerns the published industry framework; it is not a promise that every underlying administrative record is classified without error.
The official quantitative Economic Census bridge tables provide national information about changed industries, but they are a different product and do not supply a state-by-size-class bridge for this study. Their existence does not justify splicing unlike categories. Here the same-vintage CBP comparison is sufficient. More generally, a year printed on a dataset should not be mistaken for the vintage of its industrial classification. Checking the program's documentation avoids both false comparability and the opposite mistake of rejecting a comparison because a newer classification exists elsewhere.
Nearly flat smallest-class counts gained share
The national composition result is straightforward arithmetic. Divide 2,799 under-five establishments by 6,045 total establishments for 2019, and 2,805 by 5,629 for 2023. The resulting shares are 46.30% and 49.83%. Their difference is 3.53 percentage points. This is a change in a proportion, not a 3.53% change in the number of establishments. Keeping percentage points separate from percentage changes matters whenever readers compare distributions across years.
Both national counts and both national shares are reported because they answer different questions. The count describes the number of establishments assigned to the class. The share describes that class's weight within the selected industry. The six-establishment increase is small relative to the under-five total, while the 416-establishment decline in the overall total changes the denominator appreciably. It is thus reasonable to describe the smallest class as approximately stable in count and larger in share, without describing a broad surge in small bookstores.
This does not show which establishments became smaller, which entered the industry, or which ceased to be classified there. A location could change employment class, change industry assignment, operate during only part of a year, enter or leave employer status, or be affected by other changes in the business register. Two aggregate snapshots do not distinguish those processes. The finding is about the resulting composition, not its mechanism. It is also not an assessment of whether a smaller employer structure is commercially stronger, more vulnerable, or better suited to readers' needs.
Larger size bands add context, but unavailable cells remain unavailable
At the national level, the published number of establishments in the five-to-nine class declined from 1,388 to 1,220. The ten-to-nineteen class declined from 901 to 801, the twenty-to-forty-nine class from 822 to 725, and the fifty-to-ninety-nine class from 118 to 60. The one-hundred-to-249 class contained 16 published establishments at each endpoint. These observations provide context for the shift toward the under-five class without requiring employment totals as the headline evidence.
The national cells for 250–499, 500–999, and 1,000 or more employees are marked unavailable or not comparable. Those marks are preserved. The analysis does not replace them with zeros or subtract the available classes from the total to infer their contents. A published total and a set of incomplete component cells are not a license to reconstruct information the source has withheld. The same rule applies when a magnitude field is encoded as zero alongside an unavailable flag.
At state level, completeness differs across bands. The under-five count is numeric in all 51 jurisdictions at both endpoints. Other size bands have unavailable cells, sometimes in many jurisdictions, so they cannot support an equally complete state distribution without additional assumptions. The downloadable selected-cell table retains every original size-class entry, including unavailable markers. Readers can inspect the available evidence directly rather than relying on an apparently complete distribution created through zero filling. The main composition comparison uses the under-five class precisely because it is observed consistently across the full geographic set.
The complete state picture avoids a national-growth assumption
The state and District of Columbia results are presented as a complete set, not as a league table of apparent success. Eleven jurisdictions had a higher published establishment count in 2023 than in 2019. This means the national decline did not occur identically everywhere, but it does not reverse the national result. A count of jurisdictions also gives each jurisdiction equal weight, whereas the national total weights them by their number of establishments. These are different summaries and should not be substituted for one another.
The structural table shows both establishment counts and both under-five counts for every jurisdiction, with the corresponding shares. It is possible for a jurisdiction's smallest-class share to rise while its smallest-class count falls, because the total can fall faster. Conversely, a count can rise without a rising share. The table is designed to make these relationships visible without identifying a change in share as evidence that particular establishments reduced employment.
No population denominator or geographic-access measure is applied. A larger state may have more establishments without offering better access to any particular reader. State totals cannot show travel time, neighborhood availability, hours of operation, genre selection, or whether a location serves consumers in person. Nor does the table identify independent versus chain ownership. Legal-form categories in the source are not a substitute for that distinction. For each geography this study uses only the all-establishment legal-form total, avoiding double counting totals together with their organizational subcategories.
Employment results are released, perturbed numbers
Published March payroll employment declined nationally from 61,068 in 2019 to 48,997 in 2023. Both values carry the J high-noise flag. Dividing those released figures by annual-active employer establishments yields 10.10 and 8.70 published March jobs per annual-active establishment. These calculations describe the published dataset. They must not be read as precise underlying employment changes or as staffing per simultaneously operating store, because the employment and establishment reference periods differ and employment has been perturbed.
The source's historical noise flags describe how much disclosure protection changed a cell. G means less than 2%; H means at least 2% but less than 5%; J means at least 5%. J is open-ended, not a fixed 5% adjustment. These flags are not standard errors, confidence intervals, or measures of all possible data error. They do not establish statistical significance. It would be inappropriate to place conventional uncertainty bars around these values using the flags alone or claim an exact underlying direction by treating J as a known maximum distortion.
Thirty-four of the 51 jurisdictions have a J employment flag at one or both endpoints. The employment table therefore retains both flags next to every pair of values and ratios. It is not ordered by employment change, and no selected high-noise jurisdiction is used to anchor the conclusion. Readers should interpret the employment evidence alongside the establishment composition findings and the flag-restricted sensitivity summary, not as a separate ranking of workforce performance.
All-state employment comparisons and the G/H sensitivity belong together
Across all 51 jurisdictions, 42 have a lower ratio of published March jobs to annual-active establishments in 2023 than in 2019. Of the 11 jurisdictions with increased establishment counts, ten have a lower published ratio. These are counts of arithmetic relationships in released numbers, not counts of statistically established reductions in staffing. Reporting them without the flags would give the appearance of precision that the disclosure-protection information does not support.
Seventeen jurisdictions have only G or H employment flags at both endpoints. Within this restricted set, 16 have a lower published ratio; five have increased establishment counts, and all five also have a lower published ratio. The full-set and restricted-set summaries are displayed together. The restriction changes which jurisdictions enter the comparison. It is not a random sample of the country, and agreement between the two descriptions does not convert either into a significance test or establish the true direction of every underlying change.
The restricted set's establishment total is 1,925 in 2019 and 1,828 in 2023; its under-five count is 926 and 919. Its published employment sum is 19,547 and 15,488. These sums describe only that fixed roster, not a substitute national series. The downloadable sensitivity file identifies the included jurisdictions explicitly. No extrapolation is made from them to the excluded jurisdictions, and no employment observation is corrected by guessing a noise multiplier. The purpose is transparency about how a readily observable quality restriction changes the scope of the comparison.
Reproduction starts with selected cells and explicit arithmetic
The underlying inputs are the Census Bureau's public state and national County Business Patterns files for 2019 and 2023. The selection uses industry 451211 and the all-establishment legal-form marker, a hyphen in these files. National records are retained as a reference, not appended to the states when calculating an all-state total. The state extract contains exactly the 50 states and District of Columbia at both endpoints, with no territorial records mixed into that set.
For establishment changes, the calculation is the later count divided by the earlier count, minus one, multiplied by 100. The under-five share is the under-five count divided by the total establishment count, multiplied by 100. Share change is the later share minus the earlier share, in percentage points. The published employment ratio divides March payroll jobs by annual-active employer establishments. The code retains unavailable symbols and employment flags rather than silently converting source strings into numeric zeros.
A separate implementation reselected the original records, recalculated all state ratios, and compared national records with state sums. Establishments, employment, and under-five counts matched the independent national totals at both endpoints. This is a useful consistency check, not proof that administrative coverage or classification is perfect. Public downloads contain selected observations, derived results, source URLs, methods, and verification code rather than the complete bulk source archives. They allow the reported calculations to be checked while keeping the scope and unavailable-cell rules visible.
What this historical comparison can and cannot establish
The clearest conclusion is limited but substantive: between 2019 and 2023, the national count of employer establishments classified as Book Stores fell, while the under-five class stayed almost unchanged in count and increased in share. That statement is stronger than an unqualified recovery narrative because it follows the observed counts instead of imposing a preferred direction. It is also different from a study of book-publishing jobs, book sales, reading frequency, or the fortunes of independent booksellers.
The data cannot identify the causes of the change. The endpoints span years with substantial economic disruption, but this design contains no counterfactual, control group, or establishment-level history capable of attributing results to the pandemic or any other event. It cannot distinguish closures from changes in classification or employer status, measure profitability, or determine whether stores became more efficient. Repeated aggregate snapshots are not a business-demography panel. The employment ratios also cannot measure labor productivity without output and hours data.
Finally, this is an explicitly historical comparison. The latest reference year on the official CBP data page at the source check was 2023, and that page did not provide a planned release date for 2024. Later statistical releases or changes in disclosure protection should be evaluated on their own documentation before extending the series. For the present endpoints, the same 2017 industry classification and complete under-five geography make a careful composition study possible. The released employment figures add context only when their perturbation flags, mixed-period denominator, and restricted sensitivity set remain attached.
Bookstores: fewer establishments, a larger under-five share
United States · Employer establishments active during the year · 2019 and 2023
Full chart and source
Scroll horizontally to inspect the chart. ↔
Chart and values
Scroll horizontally to see all columns.
| Year | All establishments | Under-five establishments | Under-five share (%) |
|---|---|---|---|
| 2019 | 6,045 | 2,799 | 46.30 |
| 2023 | 5,629 | 2,805 | 49.83 |
Annual-active employer establishments. Under-five includes zero March jobs if wages were paid during another part of the year. Under-five is a subset of total, not additive. Shares calculated from exact counts.
Data & downloads
Scroll horizontally to see all columns.
| Year | All establishments | Under-five establishments | Under-five share (%) |
|---|---|---|---|
| 2019 | 6,045 | 2,799 | 46.30 |
| 2023 | 5,629 | 2,805 | 49.83 |
Annual-active employer establishments. Under-five includes zero March jobs if wages were paid during another part of the year. Under-five is a subset of total, not additive. Shares calculated from exact counts.
Scroll horizontally to see all columns.
| Jurisdiction | Establishments 2019 | Establishments 2023 | Under-five 2019 | Under-five 2023 | Under-five share 2019 (%) | Under-five share 2023 (%) |
|---|---|---|---|---|---|---|
| Alabama | 106 | 93 | 44 | 42 | 41.51 | 45.16 |
| Alaska | 23 | 26 | 15 | 13 | 65.22 | 50.00 |
| Arizona | 123 | 102 | 57 | 54 | 46.34 | 52.94 |
| Arkansas | 86 | 70 | 43 | 42 | 50.00 | 60.00 |
| California | 605 | 576 | 265 | 271 | 43.80 | 47.05 |
| Colorado | 111 | 112 | 46 | 50 | 41.44 | 44.64 |
| Connecticut | 78 | 75 | 32 | 31 | 41.03 | 41.33 |
| Delaware | 18 | 19 | 6 | 12 | 33.33 | 63.16 |
| District of Columbia | 28 | 27 | 13 | 9 | 46.43 | 33.33 |
| Florida | 338 | 317 | 156 | 162 | 46.15 | 51.10 |
| Georgia | 171 | 159 | 74 | 72 | 43.27 | 45.28 |
| Hawaii | 14 | 17 | 8 | 10 | 57.14 | 58.82 |
| Idaho | 29 | 24 | 10 | 12 | 34.48 | 50.00 |
| Illinois | 192 | 176 | 81 | 85 | 42.19 | 48.30 |
| Indiana | 158 | 126 | 89 | 62 | 56.33 | 49.21 |
| Iowa | 66 | 64 | 32 | 35 | 48.48 | 54.69 |
| Kansas | 46 | 42 | 25 | 22 | 54.35 | 52.38 |
| Kentucky | 92 | 83 | 47 | 46 | 51.09 | 55.42 |
| Louisiana | 74 | 65 | 35 | 28 | 47.30 | 43.08 |
| Maine | 37 | 48 | 20 | 22 | 54.05 | 45.83 |
| Maryland | 100 | 105 | 46 | 57 | 46.00 | 54.29 |
| Massachusetts | 191 | 187 | 85 | 97 | 44.50 | 51.87 |
| Michigan | 173 | 176 | 86 | 90 | 49.71 | 51.14 |
| Minnesota | 80 | 87 | 35 | 44 | 43.75 | 50.57 |
| Mississippi | 50 | 46 | 15 | 21 | 30.00 | 45.65 |
| Missouri | 102 | 84 | 41 | 40 | 40.20 | 47.62 |
| Montana | 32 | 31 | 17 | 12 | 53.12 | 38.71 |
| Nebraska | 49 | 44 | 25 | 27 | 51.02 | 61.36 |
| Nevada | 28 | 33 | 10 | 18 | 35.71 | 54.55 |
| New Hampshire | 47 | 39 | 22 | 15 | 46.81 | 38.46 |
| New Jersey | 147 | 141 | 69 | 73 | 46.94 | 51.77 |
| New Mexico | 50 | 46 | 38 | 32 | 76.00 | 69.57 |
| New York | 415 | 403 | 195 | 218 | 46.99 | 54.09 |
| North Carolina | 220 | 216 | 110 | 119 | 50.00 | 55.09 |
| North Dakota | 14 | 13 | 7 | 5 | 50.00 | 38.46 |
| Ohio | 189 | 160 | 81 | 67 | 42.86 | 41.88 |
| Oklahoma | 70 | 59 | 31 | 32 | 44.29 | 54.24 |
| Oregon | 104 | 95 | 52 | 44 | 50.00 | 46.32 |
| Pennsylvania | 261 | 234 | 122 | 130 | 46.74 | 55.56 |
| Rhode Island | 23 | 24 | 9 | 8 | 39.13 | 33.33 |
| South Carolina | 98 | 87 | 41 | 41 | 41.84 | 47.13 |
| South Dakota | 12 | 8 | 6 | 3 | 50.00 | 37.50 |
| Tennessee | 147 | 137 | 64 | 75 | 43.54 | 54.74 |
| Texas | 393 | 352 | 184 | 163 | 46.82 | 46.31 |
| Utah | 85 | 75 | 27 | 27 | 31.76 | 36.00 |
| Vermont | 28 | 30 | 13 | 18 | 46.43 | 60.00 |
| Virginia | 178 | 176 | 82 | 85 | 46.07 | 48.30 |
| Washington | 163 | 149 | 72 | 64 | 44.17 | 42.95 |
| West Virginia | 55 | 51 | 27 | 27 | 49.09 | 52.94 |
| Wisconsin | 117 | 93 | 69 | 53 | 58.97 | 56.99 |
| Wyoming | 29 | 27 | 20 | 20 | 68.97 | 74.07 |
2017 NAICS 451211; annual-active employers. Under-five class includes zero March jobs. No independent/chain ownership identification.
Scroll horizontally to see all columns.
| Year | Published March jobs | Noise flag | Published jobs / annual-active establishment |
|---|---|---|---|
| 2019 | 61,068 | J | 10.10 |
| 2023 | 48,997 | J | 8.70 |
Both employment endpoints are J (at least 5% perturbation, open-ended). Ratios divide published March jobs by annual-active employer establishments, not staffing per simultaneously open store.
Scroll horizontally to see all columns.
| Jurisdiction | Published jobs 2019 | Flag 2019 | Published jobs 2023 | Flag 2023 | Ratio 2019 | Ratio 2023 | G/H both |
|---|---|---|---|---|---|---|---|
| Alabama | 842 | J | 695 | J | 7.94 | 7.47 | No |
| Alaska | 139 | H | 147 | H | 6.04 | 5.65 | Yes |
| Arizona | 1,253 | J | 819 | J | 10.19 | 8.03 | No |
| Arkansas | 689 | H | 473 | J | 8.01 | 6.76 | No |
| California | 7,763 | J | 5,728 | H | 12.83 | 9.94 | No |
| Colorado | 1,181 | H | 959 | H | 10.64 | 8.56 | Yes |
| Connecticut | 930 | J | 881 | J | 11.92 | 11.75 | No |
| Delaware | 210 | J | 182 | G | 11.67 | 9.58 | No |
| District of Columbia | 443 | G | 412 | H | 15.82 | 15.26 | Yes |
| Florida | 3,374 | J | 2,647 | J | 9.98 | 8.35 | No |
| Georgia | 1,678 | H | 1,386 | J | 9.81 | 8.72 | No |
| Hawaii | 153 | H | 99 | J | 10.93 | 5.82 | No |
| Idaho | 321 | J | 220 | J | 11.07 | 9.17 | No |
| Illinois | 2,027 | J | 1,426 | H | 10.56 | 8.10 | No |
| Indiana | 1,346 | J | 1,117 | J | 8.52 | 8.87 | No |
| Iowa | 587 | H | 534 | J | 8.89 | 8.34 | No |
| Kansas | 601 | H | 510 | G | 13.07 | 12.14 | Yes |
| Kentucky | 721 | J | 727 | J | 7.84 | 8.76 | No |
| Louisiana | 616 | J | 594 | J | 8.32 | 9.14 | No |
| Maine | 280 | H | 272 | H | 7.57 | 5.67 | Yes |
| Maryland | 890 | J | 920 | J | 8.90 | 8.76 | No |
| Massachusetts | 1,926 | J | 1,682 | H | 10.08 | 8.99 | No |
| Michigan | 1,674 | J | 1,393 | J | 9.68 | 7.91 | No |
| Minnesota | 967 | H | 774 | H | 12.09 | 8.90 | Yes |
| Mississippi | 504 | H | 467 | H | 10.08 | 10.15 | Yes |
| Missouri | 930 | H | 1,007 | J | 9.12 | 11.99 | No |
| Montana | 322 | H | 252 | H | 10.06 | 8.13 | Yes |
| Nebraska | 452 | H | 303 | H | 9.22 | 6.89 | Yes |
| Nevada | 351 | J | 327 | J | 12.54 | 9.91 | No |
| New Hampshire | 400 | J | 346 | H | 8.51 | 8.87 | No |
| New Jersey | 1,537 | J | 1,233 | J | 10.46 | 8.74 | No |
| New Mexico | 295 | J | 257 | H | 5.90 | 5.59 | No |
| New York | 4,486 | H | 3,216 | H | 10.81 | 7.98 | Yes |
| North Carolina | 1,946 | H | 1,788 | H | 8.85 | 8.28 | Yes |
| North Dakota | 171 | J | 124 | H | 12.21 | 9.54 | No |
| Ohio | 1,889 | H | 1,449 | J | 9.99 | 9.06 | No |
| Oklahoma | 561 | H | 351 | J | 8.01 | 5.95 | No |
| Oregon | 1,209 | G | 922 | H | 11.62 | 9.71 | Yes |
| Pennsylvania | 2,756 | J | 2,133 | J | 10.56 | 9.12 | No |
| Rhode Island | 207 | H | 220 | J | 9.00 | 9.17 | No |
| South Carolina | 830 | H | 802 | J | 8.47 | 9.22 | No |
| South Dakota | 99 | J | 79 | H | 8.25 | 9.88 | No |
| Tennessee | 1,271 | H | 989 | J | 8.65 | 7.22 | No |
| Texas | 4,041 | H | 3,085 | H | 10.28 | 8.76 | Yes |
| Utah | 936 | J | 703 | J | 11.01 | 9.37 | No |
| Vermont | 225 | H | 199 | G | 8.04 | 6.63 | Yes |
| Virginia | 1,945 | J | 1,659 | J | 10.93 | 9.43 | No |
| Washington | 1,681 | H | 1,397 | H | 10.31 | 9.38 | Yes |
| West Virginia | 343 | J | 307 | J | 6.24 | 6.02 | No |
| Wisconsin | 925 | H | 679 | H | 7.91 | 7.30 | Yes |
| Wyoming | 145 | G | 106 | G | 5.00 | 3.93 | Yes |
G: <2% perturbation; H: 2–<5%; J: ≥5%, open-ended. Not standard errors or confidence intervals. Ratios use annual-active employer establishments. Read with adjacent sensitivity table; no true-direction ranking.
Scroll horizontally to see all columns.
| Set | Jurisdictions | Establishment increases | Lower published ratios | Both establishment increase and lower published ratio |
|---|---|---|---|---|
| All 50 states and DC | 51 | 11 | 42 | 10 |
| G/H at both endpoints | 17 | 5 | 16 | 5 |
Counts describe published values, not statistically established true changes. G/H subset is a fixed, nonrepresentative roster, not a confidence analysis. Thirty-four jurisdictions have J at one or both endpoints.
Scroll horizontally to see all columns.
| Size class | 2019 | 2023 |
|---|---|---|
| Under 5 | 2,799 | 2,805 |
| 5–9 | 1,388 | 1,220 |
| 10–19 | 901 | 801 |
| 20–49 | 822 | 725 |
| 50–99 | 118 | 60 |
| 100–249 | 16 | 16 |
| 250–499 | N | N |
| 500–999 | N | N |
| 1,000+ | N | N |
N means unavailable/not comparable, never zero. Do not reconstruct residual cells or sum an incomplete distribution. Classes refer to pre-noise March employment.
Evidence & method
How we calculated it
- Compare CBP 2019 and 2023, same 2017 NAICS 451211 Book Stores, LFO hyphen total only; 50 states plus DC and separate national reference.
- Establishment percentage change = 100×(E2023/E2019−1); under-five share = 100×U/E; share change in percentage points. Employment ratio = published March jobs / annual-active employer establishments.
- Preserve all G/H/J/N flags and unavailable size cells. No zero fill, imputation, residual reconstruction or code bridge.
- Display all 51 employment records alongside the fixed 17-jurisdiction G/H-at-both-endpoints sensitivity. No significance tests or claims of true underlying directions.
- Separate raw-file calculation check reproduces all 51 ratios and reconciles national establishment, employment and under-five totals with state sums.
What the data cannot tell us
- Annual-active establishments versus March payroll employment: ratio is not staffing per simultaneously open store; under-five can include zero March workers.
- Employment values are perturbed; J is open-ended ≥5%, not an SE/CI. G/H restriction does not establish statistical significance or representativeness.
- Employer establishments only; not all booksellers, independent shops, companies, nonemployers, FTEs, unique workers, or geographic consumer access.
- Two cross-sections do not identify establishment births/deaths or movements between classes; no profitability, productivity, sales or causal pandemic claims.
- Other size classes contain unavailable cells; preserve N. Historical 2019–2023, not current operating conditions.
Original sources
- Census CBP 2019 state file
Census CBP 2019 state file: State original public CSV archive; select 2017 NAICS 451211 and LFO hyphen total only. Employment noise flags retained.
Accessed October 30, 2026
- Census CBP 2019 national file
Census CBP 2019 national file: National original public CSV archive; select 2017 NAICS 451211 and LFO hyphen total only. Employment noise flags retained.
Accessed October 30, 2026
- Census CBP 2023 state file
Census CBP 2023 state file: State original public CSV archive; select 2017 NAICS 451211 and LFO hyphen total only. Employment noise flags retained.
Accessed October 30, 2026
- Census CBP 2023 national file
Census CBP 2023 national file: National original public CSV archive; select 2017 NAICS 451211 and LFO hyphen total only. Employment noise flags retained.
Accessed October 30, 2026
- Census CBP industry references
Census CBP industry references: 2017 NAICS applies to reference years 2017–2023; no 2022 code splice.
Accessed October 30, 2026
- Census CBP state layouts
Census CBP state layouts: 2018 layout covers 2018–2019; 2020 layout covers 2020–2023; flag and cell definitions.
Accessed October 30, 2026
- Census CBP glossary
Census CBP glossary: Annual-active employer denominator, March payroll employment, pre-noise classes, zero-March members of under-five, and nonemployer exclusion.
Accessed October 30, 2026
- Census CBP methodology
Census CBP methodology: Historical coverage and disclosure protection; current-methodology warning does not remove flags on historical files.
Accessed October 30, 2026
- Census CBP current data
Census CBP current data: At access, 2023 latest reference year; no planned 2024 release date.
Accessed October 30, 2026
- Census Economic Census bridge release
Census Economic Census bridge release: Quantitative national bridges exist, revised April 24, 2025; not used to harmonize state CBP.
Accessed October 30, 2026
- Census research transparency policy
Census research transparency policy: Government-produced data/works generally not copyright-protected in United States; no endorsement implied.
Accessed October 30, 2026
Cite & reuse this study
Alisson Bento. U.S. bookstore employer establishments declined as the under-five class gained share, 2019–2023. Original secondary analysis of U.S. Census Bureau County Business Patterns, 2019 and 2023, 2017 NAICS 451211. Sources accessed October 30, 2026. https://alissonbento.com/research/us-bookstore-employer-establishment-composition/
Source statistics: U.S. Census Bureau. Government-produced data generally not copyright-protected in the United States; source attribution retained. Original selection, calculations and explanation: Alisson Bento. No Census endorsement or rights to its logo implied.
Corrections or editorial questions