What the headline measures
The source is the national 2024 ACS one-year person file, covering the 50 states and the District of Columbia. We read all 3,422,888 released person records in bounded chunks and retained the columns needed for the calculation. The primary comparison includes resident adults aged 18 or older who were civilian employed and actually worked during the reference week, reported a valid transportation category and belonged to one of the three employment-class groups defined below. The final sample contains 2,558 writers and authors and 5,915 people in the comparison basket. These are sample records, not the estimated total populations.
The Census occupation code for writers and authors is 2850. It is a broad occupational category. It does not identify literary genre, distinguish self-published from traditionally published authors, enumerate everyone who writes on the side or establish whether someone earns income from books. The comparison basket contains eight separate codes: broadcast announcers and radio disc jockeys; news analysts, reporters and journalists; public relations specialists; editors; technical writers; interpreters and translators; court reporters and simultaneous captioners; and other media and communication workers. It excludes writers and authors, photographers and media-equipment occupations. This is our stated comparison basket, rather than a claim that every member works in publishing or performs comparable tasks.
The age threshold is also deliberate. Official journey-to-work statistics commonly describe workers aged 16 or older. Our estimates describe adults aged 18 or older, so they should not be presented as reproductions of an official 16-plus table. Similarly, these are residents classified by occupation, not jobs assigned to the county containing an employer's office. Neither a publisher's business address nor a national count of publisher employees identifies where writers live or perform their main work.
A home-work answer is narrower than a remote-work arrangement
The ACS transportation question asks how a person usually got to work last week. A respondent chooses one category. People using different modes on different days identify the mode used most often; when a work trip uses several modes, the longest-distance part matters. “Worked from home” is one available answer. The variable therefore describes a principal pattern during a particular reference week, rather than all occasions on which work happened at home.
A person who works from home on some days can fall into another transportation category. Conversely, the home category does not demonstrate that every working hour occurred there, that an employer offers a permanent fully remote contract or that the person would accept a distant employer. We cannot recover the number of hybrid days from this field. The continuously collected survey also combines people answering about different weeks throughout 2024; it does not photograph the entire workforce during one common week.
Occupation and employment class refer to the main job, generally the job accounting for the greatest number of hours during the preceding week. Journey-to-work questions also aim at the person's main work activity. Multi-job and multi-location situations remain measurement limitations, but there is no basis for asserting that these records actually attach the occupation and transportation responses to different jobs. The useful conclusion stays close to the survey: a principal home-work pattern was more prevalent in the writer group within each of the three classes in this comparison.
Compare employees with employees before combining business owners
We first distinguish three groups. The employee-only group combines class-of-worker codes 1 through 5: private for-profit, private nonprofit and local, state and federal government employees. Code 6 identifies people working in their own unincorporated businesses, professional practices or farms. Code 7 identifies owners of incorporated businesses, professional practices or farms. The latter two categories describe people, not counts of businesses, and neither is automatically equivalent to a freelance book author.
Among employee-only writers, the estimated home-mode share was 44.1%, with an approximate 95% interval of 40.8%–47.5%. For employees in the comparison basket, the estimate was 32.0%, with an interval of 30.5%–33.5%. The directly calculated difference was 12.2 percentage points, with an interval of 8.5–15.8 points. This comparison does not establish that becoming a writer causes a home-based arrangement: the occupational groups can differ in tasks, employers, location, age and other characteristics that we have not standardized.
The corresponding estimates for unincorporated business owners were 72.0% among writers and 55.0% in the comparison basket. Incorporated owners had estimates of 72.2% and 46.6%, respectively. The owner samples are smaller, the incorporated ones especially: 228 writer records and 283 comparison records, compared with 712 and 647 for unincorporated owners and 1,618 and 4,985 in the employee-only cells. Their wider uncertainty intervals deserve attention. A high point estimate is not a guarantee about the situation of a particular writer or business owner.
The Census classification convention adds an important distinction. Owners working in their own incorporated businesses are included with wage and salary workers in Census tabulations because they are paid employees of their companies. Our employee-only comparison deliberately keeps code 7 separate, and must not be renamed the official combined wage-and-salary category. Combining codes 1–5 and 7 produces a writer home-mode estimate of 47.6%, rather than 44.1%. The difference comes from changing the population being combined, not from discovering additional home workers or correcting the source.
What changes when employment-class composition is held common
The overall primary-population estimates were 54.3% for writers and 35.2% for the comparison basket. Their observed difference was 19.1 percentage points, with an approximate 95% interval of 15.9–22.2 points. This difference combines two features: different home-mode rates within employment classes and different proportions of people in those classes. Reporting only the overall figures hides that distinction.
For a direct standardized comparison, we calculate one target class distribution from the pooled estimated population of both groups. We then apply that same target distribution to each group's three class-specific home-mode rates. The resulting difference is 13.8 percentage points, with an interval of 10.6–17.0 points. This answers a constructed descriptive question: what difference would these measured class-specific rates produce under one common employment-class mix? It does not predict what would happen if real people changed employment status.
The target is estimated from the person-weighted pooled population. It is neither an equal weighting of the two occupational groups nor an externally fixed standard population. We recalculate its weights during each replicate calculation, so uncertainty in the target is retained. The standardized difference also leaves all other group differences in place. It should not be described as an adjusted causal effect or as a comparison between otherwise identical jobs.
A separate symmetric decomposition provides an exact arithmetic account of the observed difference. It assigns 5.1 percentage points to the class-composition component and 14.0 points to the within-class-rate component; their sum equals the 19.1-point observed difference. These components use the average of the two groups' class proportions and rates. Consequently, the 14.0-point within-class component is not the same estimand as the 13.8-point direct standardized gap, which uses the pooled-person target. The downloadable calculation exposes both formulas rather than allowing their similar values to conceal the different questions.
The pooled target assigns approximately 78.16% to employees, 15.57% to unincorporated owners and 6.27% to incorporated owners. It reflects the combined estimated population, in which the comparison basket is larger. Multiplying the three class-specific percentage-point differences by those common proportions and adding the products reproduces the standardized gap. Use the unrounded downloadable values for an exact result; the rounded chart labels support reading, rather than serving as the calculation's inputs. Applying a different external standard would answer a different question and should be labeled accordingly.
The decomposition table also shows why individual components can be negative. Writers have a smaller employee share than the comparison basket, so the employee row contributes a negative class-composition term. Other class rows offset it. Such a negative term is part of the accounting identity, not evidence that employees reduce anyone's ability to work at home. Nor should the three class-specific prevalence percentages be added: each has its own denominator. Population shares supply the weights that turn those separate rates into an overall rate.
The allocation sensitivity changes the sample substantially
The released ACS contains edited and allocated values as part of its processing. We retain those released values for the main analysis. We also run a prespecified sensitivity calculation restricted to records whose occupation, class-of-worker and transportation allocation flags are all explicitly zero. A blank or invalid flag would not qualify for that restricted group. This is a second analysis population, not an identification of respondents who told the truth or a correction of the main estimate.
The restricted common-class sample contains 2,032 writers and 4,737 comparison records. Its estimated home-mode shares are 63.2% and 40.2%, respectively, compared with 54.3% and 35.2% in the full released-data analysis. The restricted observed difference is 23.0 percentage points, and its direct standardized difference is 16.0 points. Both remain positive in the calculated intervals, but the levels change materially. A presentation that selected only the larger restricted percentage would misrepresent the source and the predeclared main analysis.
These differences show that the inclusion rule matters. They do not isolate the effect of allocation, reveal which allocated responses were originally missing or demonstrate that allocation created an erroneous home-work answer. Removing flagged records changes the composition of the people represented. We therefore show the main and sensitivity results together and identify the retained sample sizes. Readers should carry this sensitivity into any interpretation of a seemingly precise national percentage.
Denominators, small cells and uncertainty
People temporarily absent from their jobs have an inapplicable transportation answer. The source contains 54 employed writer records and 104 employed comparison records in that situation; they are excluded from the at-work comparison. They are not counted as people who worked outside the home. Among the remaining at-work adults in these two groups, there were no missing or invalid transportation categories in this release. Observing none in the sample does not prove a universal population zero.
The three-class comparison also excludes unpaid family-business workers, class 8. There were eight such writer records and four comparison records with valid transportation responses. These small groups appear in the coverage audit, but we do not estimate or promote a precise home-work prevalence for them. There were no invalid or missing class codes among the otherwise eligible records. The common-class populations should therefore be described explicitly, rather than silently treated as every employed writer or every comparison worker.
Before examining the outcomes, we required every primary group-by-class cell to contain at least 50 sample records, including at least ten home and ten non-home observations. Any failure would withhold dependent standardization and decomposition headlines. All six primary cells, and all six corresponding restricted-sample sensitivity cells, passed that rule. The rule is an editorial safeguard, not an official Census reliability designation, and passing it does not remove uncertainty or measurement limitations.
Uncertainty uses the released person weight and all 80 person replicate weights. For each replicate, we recalculate every applicable denominator, class share, target weight, rate, contrast and decomposition component. The successive-difference-replication variance uses the factor 4/80. We report pointwise normal-approximation 95% intervals using 1.96 standard errors, rather than silently calling the source's commonly used 90% level a 95% interval. Replicates are not discarded or renormalized. These intervals address sampling uncertainty under this calculation, not every possible reporting, coding or conceptual error. They are not simultaneous confidence statements across all displayed comparisons.
The record counts in the tables describe how much sample evidence supports each calculation. The estimated person counts and percentages instead use the released weights, because sample records do not all represent the same number of residents. A reader reproducing the work should start with the exact national release, confirm its fingerprint, apply the stated eligibility rules and calculate the weighted quantities before rounding. The source-file receipt, code definitions and derived tables make those steps inspectable without distributing a new person-level extract. Reusing the table's printed percentages as raw data would lose precision and would not reproduce the replicate uncertainty.
How to use the findings without promising a career outcome
For a prospective writer, the useful finding is that employment class and observed work location are separate questions. This national sample includes writers who are employees and writers operating businesses. The employee-only home-mode percentage is substantial, yet it still describes an existing broad occupational population during a reference week. It cannot establish how many suitable jobs are available, what qualifications employers require, whether a specific contract permits home working or what an individual could earn.
For an editor, the useful angle is the distinction between an overall difference and a within-class comparison. A responsible short description can identify the adult at-work population, the principal-home-mode definition and the employee-only or common-class denominator in the same sentence. It can link the tables and method so that a reader can see the comparison basket, uncertainty and allocation sensitivity. It should not replace “writers and authors” with “novelists,” “home-mode” with “fully remote jobs” or a descriptive difference with a causal explanation.
This work builds on established research rather than presenting home-based work as a discovery. The NEA's current artist indicators already examine occupational populations, residence and earnings. Census researchers have explained why work-from-home measures differ across household and employer surveys. Our narrower contribution is the reproducible joint class-and-occupation calculation and its two explicitly different ways of separating composition from within-class rates. The national source, code definitions, sample audit, formulas, derived tables and original charts are provided together so that those distinctions survive beyond the headline.
Compare employment classes before the overall average
Compare employment classes before the overall average. United States · Civilian residents 18+ actually at work · ACS 2024 one-year PUMS. Primary COW 1–7 population. Broad writer occupation; comparator is an explicit eight-occupation basket. Principal reference-week home mode does not mean all hybrid work or fully remote jobs.
Full chart and source
Scroll horizontally to inspect the chart. ↔
Chart and values
Scroll horizontally to see all columns.
| Occupational group | Employment class | Sample n | Home sample n | Principal home-mode share (%) | Approximate 95% interval |
|---|---|---|---|---|---|
| Writers and authors (2850) | Employees: COW 1–5 | 1,618 | 719 | 44.15 | 40.84–47.45 |
| Defined other media/communication basket | Employees: COW 1–5 | 4,985 | 1,673 | 31.98 | 30.50–33.45 |
| Writers and authors (2850) | Own unincorporated business: COW 6 | 712 | 532 | 71.97 | 67.54–76.40 |
| Defined other media/communication basket | Own unincorporated business: COW 6 | 647 | 370 | 54.96 | 49.79–60.12 |
| Writers and authors (2850) | Own incorporated business: COW 7 | 228 | 169 | 72.18 | 64.64–79.72 |
| Defined other media/communication basket | Own incorporated business: COW 7 | 283 | 146 | 46.64 | 39.40–53.89 |
US civilian residents aged 18+ at work in the reference week; common classes COW 1–7. Home means JWTRNS 11, usual principal transportation/home mode. Not any hybrid work or fully remote jobs. Counts are unweighted; percentages are weighted. Approximate pointwise 95% SDR intervals use 80 replicate weights.
Class standardization leaves a different descriptive gap
Class standardization leaves a different descriptive gap. United States · Civilian residents 18+ actually at work · ACS 2024 one-year PUMS. Primary COW 1–7 population: broad writers versus the defined eight-occupation basket. Outcome: JWTRNS 11, usual principal home-work mode during the reference week, not all hybrid work or remote vacancies. The estimated pooled-person class target is recalculated in each replicate. Class standardization leaves occupation mix and other differences uncontrolled; this is not a causal effect.
Full chart and source
Scroll horizontally to inspect the chart. ↔
Chart and values
Scroll horizontally to see all columns.
| Contrast | Percentage points | Approximate 95% interval |
|---|---|---|
| Observed difference | 19.09 | 15.94–22.24 |
| Standardized to estimated pooled-person class mix | 13.76 | 10.57–16.96 |
| Kitagawa class-composition component | 5.06 | 4.14–5.97 |
| Kitagawa within-class-rate component | 14.03 | 10.84–17.22 |
| Observed minus standardized difference | 5.33 | 4.31–6.34 |
| Difference within employees (COW 1–5) | 12.17 | 8.54–15.80 |
| Difference within unincorporated owners (COW 6) | 17.02 | 9.99–24.04 |
| Difference within incorporated owners (COW 7) | 25.54 | 14.67–36.41 |
Primary full released-data population; writer minus the explicitly defined eight-occupation comparison basket. All values and intervals are percentage points. The pooled-person standardized gap differs from the symmetric Kitagawa within-class component. Composition is descriptive accounting, not a cause.
Allocation restriction changes the population and the level
Allocation restriction changes the population and the level. United States · Civilian residents 18+ actually at work · ACS 2024 one-year PUMS. Full released data is primary. Strict sensitivity retains only occupation, class and transport flags explicitly zero. These are different populations, not corrected truth or a respondent-honesty comparison. Home means the usual principal mode during the reference week, not all home working.
Full chart and source
Scroll horizontally to inspect the chart. ↔
Chart and values
Scroll horizontally to see all columns.
| Analysis population | Occupational group | Sample n | Principal home-mode share (%) | Approximate 95% interval |
|---|---|---|---|---|
| Full released-data population | Writers and authors (2850) | 2,558 | 54.26 | 51.67–56.85 |
| Full released-data population | Defined other media/communication basket | 5,915 | 35.17 | 33.72–36.62 |
| All three allocation flags explicitly zero | Writers and authors (2850) | 2,032 | 63.24 | 60.60–65.88 |
| All three allocation flags explicitly zero | Defined other media/communication basket | 4,737 | 40.25 | 38.60–41.89 |
The full released-data population is primary. Requiring all three allocation flags to be zero removes records and changes the represented population; it is not corrected truth. Home is principal reference-week mode, not all home work. Approximate 95% SDR intervals are pointwise.
Data & downloads
Scroll horizontally to see all columns.
| Occupational group | Employment class | Sample n | Home sample n | Principal home-mode share (%) | Approximate 95% interval |
|---|---|---|---|---|---|
| Writers and authors (2850) | Employees: COW 1–5 | 1,618 | 719 | 44.15 | 40.84–47.45 |
| Defined other media/communication basket | Employees: COW 1–5 | 4,985 | 1,673 | 31.98 | 30.50–33.45 |
| Writers and authors (2850) | Own unincorporated business: COW 6 | 712 | 532 | 71.97 | 67.54–76.40 |
| Defined other media/communication basket | Own unincorporated business: COW 6 | 647 | 370 | 54.96 | 49.79–60.12 |
| Writers and authors (2850) | Own incorporated business: COW 7 | 228 | 169 | 72.18 | 64.64–79.72 |
| Defined other media/communication basket | Own incorporated business: COW 7 | 283 | 146 | 46.64 | 39.40–53.89 |
US civilian residents aged 18+ at work in the reference week; common classes COW 1–7. Home means JWTRNS 11, usual principal transportation/home mode. Not any hybrid work or fully remote jobs. Counts are unweighted; percentages are weighted. Approximate pointwise 95% SDR intervals use 80 replicate weights.
Scroll horizontally to see all columns.
| Contrast | Percentage points | Approximate 95% interval |
|---|---|---|
| Observed difference | 19.09 | 15.94–22.24 |
| Standardized to estimated pooled-person class mix | 13.76 | 10.57–16.96 |
| Kitagawa class-composition component | 5.06 | 4.14–5.97 |
| Kitagawa within-class-rate component | 14.03 | 10.84–17.22 |
| Observed minus standardized difference | 5.33 | 4.31–6.34 |
| Difference within employees (COW 1–5) | 12.17 | 8.54–15.80 |
| Difference within unincorporated owners (COW 6) | 17.02 | 9.99–24.04 |
| Difference within incorporated owners (COW 7) | 25.54 | 14.67–36.41 |
Primary full released-data population; writer minus the explicitly defined eight-occupation comparison basket. All values and intervals are percentage points. The pooled-person standardized gap differs from the symmetric Kitagawa within-class component. Composition is descriptive accounting, not a cause.
Scroll horizontally to see all columns.
| Employment class | Writer class share (%) | Comparison class share (%) | Pooled target share (%) | Composition component (pp) | Within-class component (pp) |
|---|---|---|---|---|---|
| Employees: COW 1–5 | 63.72 | 84.22 | 78.16 | −7.80 | 9.00 |
| Own unincorporated business: COW 6 | 27.37 | 10.61 | 15.57 | 10.63 | 3.23 |
| Own incorporated business: COW 7 | 8.91 | 5.16 | 6.27 | 2.23 | 1.80 |
Primary population. Class shares and estimated pooled-person target are percentages; decomposition contributions are percentage points. Negative contributions remain negative. Symmetric components use equal-group average proportions/rates, not the person-pooled standardization target. Class contributions add to the observed gap.
Scroll horizontally to see all columns.
| Analysis population | Occupational group | Sample n | Principal home-mode share (%) | Approximate 95% interval |
|---|---|---|---|---|
| Full released-data population | Writers and authors (2850) | 2,558 | 54.26 | 51.67–56.85 |
| Full released-data population | Defined other media/communication basket | 5,915 | 35.17 | 33.72–36.62 |
| All three allocation flags explicitly zero | Writers and authors (2850) | 2,032 | 63.24 | 60.60–65.88 |
| All three allocation flags explicitly zero | Defined other media/communication basket | 4,737 | 40.25 | 38.60–41.89 |
The full released-data population is primary. Requiring all three allocation flags to be zero removes records and changes the represented population; it is not corrected truth. Home is principal reference-week mode, not all home work. Approximate 95% SDR intervals are pointwise.
Scroll horizontally to see all columns.
| Occupational group | Eligibility / exclusion | Sample n | Estimated people |
|---|---|---|---|
| Writers and authors (2850) | Employed adults: at work or temporarily absent | 2,620 | 251,819 |
| Writers and authors (2850) | Temporarily absent: ESR 2 | 54 | 4,338 |
| Writers and authors (2850) | Actually at work: ESR 1 | 2,566 | 247,481 |
| Writers and authors (2850) | At work, invalid/missing transportation mode | 0 | 0 |
| Writers and authors (2850) | Valid mode, unpaid family business: COW 8 | 8 | 982 |
| Writers and authors (2850) | Valid mode, invalid/missing class | 0 | 0 |
| Writers and authors (2850) | Primary common classes: COW 1–7 | 2,558 | 246,499 |
| Writers and authors (2850) | Primary records with all three flags explicitly zero | 2,032 | 195,267 |
| Defined other media/communication basket | Employed adults: at work or temporarily absent | 6,023 | 596,842 |
| Defined other media/communication basket | Temporarily absent: ESR 2 | 104 | 9,492 |
| Defined other media/communication basket | Actually at work: ESR 1 | 5,919 | 587,350 |
| Defined other media/communication basket | At work, invalid/missing transportation mode | 0 | 0 |
| Defined other media/communication basket | Valid mode, unpaid family business: COW 8 | 4 | 297 |
| Defined other media/communication basket | Valid mode, invalid/missing class | 0 | 0 |
| Defined other media/communication basket | Primary common classes: COW 1–7 | 5,915 | 587,053 |
| Defined other media/communication basket | Primary records with all three flags explicitly zero | 4,737 | 464,132 |
Resident civilian employed adults, 18+; the audit begins with ESR 1 or 2, but the primary population requires actually at work, valid mode and COW 1–7. Absence is inapplicable, not non-home. COW 8 is unpaid family-business work, not unpaid family care. Sparse counts and observed zeros do not establish population prevalence or population zero.
Scroll horizontally to see all columns.
| Occupational group | Class aggregation | Sample n | Principal home-mode share (%) | Approximate 95% interval |
|---|---|---|---|---|
| Writers and authors (2850) | Employee-only COW 1–5 | 1,618 | 44.15 | 40.84–47.45 |
| Writers and authors (2850) | Census wage/salary COW 1–5 + 7 | 1,846 | 47.59 | 44.40–50.78 |
| Defined other media/communication basket | Employee-only COW 1–5 | 4,985 | 31.98 | 30.50–33.45 |
| Defined other media/communication basket | Census wage/salary COW 1–5 + 7 | 5,268 | 32.82 | 31.36–34.28 |
Primary population. Census wage-and-salary grouping includes incorporated business owners, COW 7. Employee-only COW 1–5 keeps those owners separate. Different aggregation of people, not new workers, changed behavior or a correction. Approximate 95% SDR intervals use 80 replicate weights.
Evidence & method
How we calculated it
- Read the exact 2024 national ACS one-year person ZIP; combine its USA and USB members once, confirm all 50 states and DC and selected-person uniqueness. Stream 92 required columns in 50,000-record chunks. The primary population is resident adults aged 18+ who are civilian employed and at work, ESR 1; audit ESR 2 absence separately.
- OCCP 2850 defines broad writers and authors. The explicit eight-code comparator basket is 2805, 2810, 2825, 2830, 2840, 2861, 2862 and 2865; there is no book-industry or novelist restriction. JWTRNS 11 denotes the usual principal worked-from-home mode during the reference week.
- Primary common classes are employee-only COW 1–5, own unincorporated business COW 6 and own incorporated business COW 7. Audit unpaid family-business COW 8 separately and exclude it from the common-class comparison. Census wage/salary sensitivity combines 1–5 + 7, without renaming that aggregation employee-only.
- Compute class rates, observed gap and directly standardized gap using estimated pooled-person class shares. Recalculate the pooled target and rates in all 80 replicates. Separately compute exact symmetric Kitagawa composition and within-class components, whose equal-group-average weights differ from the person-pooled target.
- Use PWGTP and all 80 person replicate weights with SDR variance factor 4/80 and a 1.96 normal multiplier for approximate pointwise 95% intervals. Preserve zero and signed replicate weights; do not drop bad replicates or clip intervals. Each of the six cells requires n >= 50 and home/non-home counts >= 10 for dependent contrasts.
- The main analysis includes released edited and allocated values. Sensitivity requires FOCCP = 0 AND FCOWP = 0 AND FJWTRNSP = 0 explicitly; missing or invalid flags never qualify. The restriction changes population; it is not a correction or respondent-truth check.
What the data cannot tell us
- US residents in the 50 states and DC, aged 18+ and at work: not global authors, all employed people, the official 16+ universe, publisher worksites, KDP users or novelists. Broad occupations can contain multiple tasks and industries.
- JWTRNS is one principal transportation/home answer about the preceding week, not all hybrid days or hours, permanent fully remote arrangements, job vacancies or an individual career forecast. Survey reference weeks vary across 2024.
- The comparator is an explicitly defined basket of eight occupations, not all nonwriters or only publishing staff. Standardization changes employment-class mix only; internal occupation mix, tasks, demographics and other differences remain.
- The full released sample is primary. Excluding allocated occupation/class/transport records shifts estimated levels materially and changes the represented population; neither subset is independently validated truth.
- COW 7 incorporated owners are included in Census wage/salary grouping, but separated from employee-only COW 1–5 here. COW 8 samples of eight writers and four comparison workers are audited, not promoted as precise prevalence estimates.
- Pointwise normal-approximation 95% SDR intervals address sampling uncertainty, not all measurement, editing or disclosure errors, simultaneous coverage or causation. Observed-zero audit cells do not prove a population zero.
Original sources
- 2024 ACS one-year national person PUMS ZIP
The exact national 2024 one-year person ZIP contains 3,422,888 released person records. Both national CSV members are read once; the public download provides original aggregates, the source receipt and its exact fingerprint, not a new person-level extract.
Accessed October 31, 2026
- 2024 ACS PUMS dictionary
The official 2024 PUMS dictionary defines STATE, OCCP 2850 writers/authors, the eight comparator occupation codes, employment class COW, principal transportation/home mode JWTRNS and the three allocation flags. Broad writer occupation does not identify novelists or book income.
Accessed October 31, 2026
- 2024 ACS subject definitions
The official 2024 ACS subject definitions establish the at-work reference-week universe and usual principal transportation mode. They distinguish employee-only COW 1–5 from the Census wage/salary grouping that also includes incorporated owners, COW 7.
Accessed October 31, 2026
- 2024 one-year PUMS accuracy
The official 2024 one-year PUMS accuracy document supports person weights and all 80 person replicate weights with SDR variance factor 4/80. The explicit normal-approximation 95% convention is this study’s choice, rather than silently using the commonly published ACS 90% level.
Accessed October 31, 2026
- 2024 one-year PUMS user guide
The official 2024 one-year PUMS user guide explains the two national person members, public-use sample and interpretation limitations. Our adult 18+ restriction differs from the official journey-to-work 16+ universe; source confidentiality measures and rounding remain.
Accessed October 31, 2026
- NEA NASERC A.1, May 2026
NEA NASERC Indicator A.1, published in May 2026, already describes the 2024 artist workforce and broad occupational populations. This study credits that work and does not claim the first count of writers or artists.
Accessed October 31, 2026
- NEA NASERC A.3, May 2026
NEA NASERC Indicator A.3, published in May 2026, examines 2024 artist residence by state and selected counties. Residence is distinct from publisher worksites, individual job vacancies and a principal home-work arrangement.
Accessed October 31, 2026
- NEA NASERC A.7, June 2026
NEA NASERC Indicator A.7, published in June 2026, already examines 2024 artist earnings. The outcome in this separate study is reference-week principal home-work mode, not earnings, royalties or book-author income.
Accessed October 31, 2026
- Census commuting-measures working paper, June 2026
The June 2026 Census staff working paper on commuting measures compares SIPP and ACS, including the distinction between principal home-based mode and any home work. It provides measurement context; the authors’ conclusions are not an agency endorsement.
Accessed October 31, 2026
- Census CES work-from-home paper, June 2025
The June 2025 Census CES work-from-home paper documents existing general and sector research and differences among household and business measures. It is cited, not republished: no report text, tables or graphics are included. The source paper’s conclusions are its authors’ views.
Accessed October 31, 2026
Cite & reuse this study
Alisson Bento. Writers at home: separating employment type from work location. 2026. Original analysis of U.S. Census Bureau, 2024 ACS one-year national person PUMS, accessed 31 October 2026. https://alissonbento.com/research/us-writer-home-work-employment-class-2024/
Original analysis, tables and charts by Alisson Bento. Reuse the original derived material with attribution and source year, adult at-work population, principal-mode definition and sensitivity limits retained. Credit the U.S. Census Bureau for the underlying public-use data. No Census or NEA endorsement is implied; source reports and person files are linked, not redistributed.
Corrections or editorial questions