The measurement in brief
The distinction remains when the people are held constant. Across the same 259 sample records, the weighted mean difference between the two income measures is approximately $10,200, with a replicate-based 95% interval of about $6,200 to $14,100. Selecting the 164 records that also meet a past-year full-time, year-round definition produces another figure: approximately $105,800 in mean wages plus net self-employment income. That last change selects a different group. It does not measure a pay increase caused by working longer.
This study makes those choices inspectable. It connects income construction, employment classification, population selection and reference periods in one bounded case, with the complete aggregate tables and reproduction program available for checking. Its purpose is to help readers evaluate a headline about writers’ earnings before comparing it with a book-income survey or using it as a career expectation.
The population comes before the amount
The source is the Census Bureau’s 2024 one-year Public Use Microdata Sample, or PUMS, for New York State. The file contains 210,150 person records. The analysis first selects people aged at least eighteen who were classified as civilian employed, either at work or with a job but temporarily not at work, and who have a positive person weight. Within that population, 259 records have occupation code 2850, writers and authors.
That occupational category is broader than published novelists. It does not identify Kindle Direct Publishing accounts, literary genre, publication history, books sold, advances, rights income or whether a person considers writing an artistic vocation. Someone who publishes a book while reporting teaching as their primary occupation is outside this particular writer group. A person whose main paid work is classified as writing can be inside it without being a book author in the sense used by a publishing survey.
The geography is residence in New York State. It is not New York City, a map of publishing offices, or a count of jobs located inside the state. Residents can work elsewhere. The public file also includes people in group quarters, subject to its survey and confidentiality procedures. There is no attempt to reconstruct an employer address or identify a respondent.
Those boundaries are part of the finding. A number attached to the phrase “writers in New York” can change when the analyst means residents, workplace jobs, publishing employees, members of an association or people who have earned money from a book. This article uses one stated population throughout rather than allowing the label to shift between comparisons.
A publishing industry is a different set of people
The same file permits a useful coverage check. Industry code 6481 combines periodical, book, directory and mailing-list publishers with other publishers. It is a mixed industry category, not a book-publishing-only classification. In the selected employed-resident population, 300 sample records belong to that industry. Only fifteen records belong both to that industry and to the writer occupation.
The resulting weighted intersection is approximately 4.9% of the writer group and 4.3% of the mixed publishing-industry group. The approximate 95% intervals are wide: about 2.1–7.7% and 1.7–6.9%, respectively. With only fifteen observations in the intersection, these figures are contextual checks, not a precise market-share result. They should not be used to rank subgroups or describe a small local market.
The two percentages have the same numerator but different denominators. More importantly, neither population is the number of people who make a living from books. A publisher employs people in many occupations; a writer can work in many industries. The data do not establish which particular books, services or clients generated an individual’s income.
This comparison therefore cannot be added to an establishment employment series as though both counted the same thing. It helps explain why a report on publishing jobs and a report on writer occupations may move differently without either being erroneous. They define different sets, locations and units. The small intersection here demonstrates that distinction in the selected public-use sample; it is not a census of the literary community.
Owning a company does not automatically mean exclusion from wage work
Class-of-worker definitions introduce another important choice. The public file separately identifies employees, owners of unincorporated businesses, owners of incorporated businesses and unpaid family workers. Among the writer records, the weighted shares are approximately 64.7% in employee categories one through five, 23.9% in an owned unincorporated business and 11.3% in an owned incorporated business. The remaining unpaid-family category has one observation and is too sparse for a substantive population interpretation.
It would be misleading to combine both kinds of owner and announce that they are all outside wage-and-salary employment. The Census subject definitions include incorporated business owners with private wage-and-salary workers because they are paid employees of their companies. Applying that published classification changes the writer wage-and-salary share from the literal employee-category count of 64.7% to approximately 76.0%.
This is a classification comparison on the same 259 records. The 11.3-point difference is the incorporated-owner category itself. It is not newly discovered employment, a change over time, or an estimate of workers omitted by a particular employer survey. An actual comparison with an establishment survey would also require its definitions, geography, job coverage and reference period.
For a reader assessing an employment claim, the practical question is which version of self-employment is meant. A broad everyday description of working for oneself can differ from the statistical grouping used in a published table. Keeping the underlying categories visible makes the comparison more useful than treating “employee” and “self-employed” as universally interchangeable labels.
The income comparison holds the people constant first
The primary income comparison uses two measures for each writer record. The first is wage or salary income. The second adds net self-employment income, retaining zero values and reported losses. Both use the source’s adjustment factor to express the amounts in constant 2024 dollars. The program also verifies that the sum agrees with the released person-earnings variable for every included writer record.
On this identical population, mean wage or salary income is approximately $75,020 and mean wages plus net self-employment income is approximately $85,185. Their difference is approximately $10,165. The comparison changes the income measure rather than changing who is in the group. The difference of the means equals the weighted mean of the net self-employment component, as it must under this construction.
The uncertainty calculation also keeps that relationship intact. The analysis computes the income-measure difference within every replicate, rather than subtracting two standard errors as though the estimates came from unrelated samples. Its approximate 95% interval is $6,196–$14,135. That interval describes sampling uncertainty under the released weights and stated calculation; it does not cover every possible reporting, coding or measurement error.
Approximately 17.8% of the weighted writer group has zero wage income in these records. All forty-nine such sample records have positive net self-employment income. This helps explain why dropping zero wages before calculating an average would change the population as well as the measure. A zero in one income component does not mean that the person had no earnings, and an earnings figure does not reveal the revenue from a book.
Selecting full-time, year-round workers changes the question
The next comparison requires at least thirty-five usual weekly work hours and at least fifty weeks worked over the preceding twelve months. This selects 164 of the 259 writer records. The weighted mean of wages plus net self-employment income for that selected group is approximately $105,788, compared with $85,185 for the full current-writer population.
The two groups overlap, so the second amount should not be described as what an individual would earn by moving from part-time to full-time writing. Hours and weeks concern all jobs during the past year. The occupation identifies the current main job. The file does not establish that every selected hour or week was spent writing, or that the writer occupation was held for the entire income reference period.
Full-time, year-round selection can be useful when comparing similar work-intensity definitions. It can also obscure the experience of people with interrupted or less intensive employment if a headline presents the selected amount as typical of everyone in an occupation. Here approximately 66.7% of the weighted current-writer group meets the stated past-year threshold. The remaining records stay in the full-population results.
The article therefore presents both bases and their unweighted sample sizes. It does not describe the remainder as hobbyists, unsuccessful authors or people who chose not to work. Those interpretations require information that is not established by this selection rule.
Mean and median answer different questions too
The downloadable tables include a second summary for readers who want to inspect distributional sensitivity. Using an explicitly defined weighted empirical median, the wages-only point is approximately $62,946 and the wages-plus-net-self-employment point approximately $70,052 for the full writer group. In the full-time, year-round subset the corresponding points are approximately $81,220 and $88,327.
These medians retain zeros and losses. The program selects the smallest observed value at which cumulative person weight reaches at least half the group’s total weight. This is an auditable convention for the released records. It is not the grouped, interpolated procedure used for Census published median tables, so exact agreement with an official published median is not claimed.
No sampling interval is established here for those median points. They provide a descriptive comparison, not a basis for asserting that a difference in population medians is statistically established. The figures emphasize means with the separately computed replicate intervals. The full tables keep the median convention and its limitation beside the values.
Nor should the gap between a mean and a median be turned into a claim about a few bestselling authors. The occupation is broad, income includes other work, and confidential-source protections alter extreme values. The public file cannot identify a sales success, an advance or the shape of the original uncensored upper tail.
Book royalties cannot be recovered from these fields
The source’s definition of earnings adds wage or salary income to net self-employment income. A separate income category combines interest, dividends, net rental income, net royalties and income from estates and trusts. The public-use variable for that combined category does not isolate book royalties.
Consequently, adding that whole category to the earnings measure would not create a valid book-income estimate. It would add potentially unrelated interest, investment and property receipts as well. Conversely, the file does not reveal whether particular writing-related receipts were reported as business income or in another income component. Those reporting and business circumstances are not identified at the book or contract level.
This is why the study makes no estimate of the proportion receiving book royalties, mean KDP income or returns from self-publishing. A person’s occupation does not assign all of their income to that occupation. A combined income component does not become book income because its definition mentions royalties.
The distinction matters when comparing this case with author-income surveys. Those surveys can ask directly about books, writing-related work, publishing routes and career experience. A general household survey serves a different purpose. Its broader coverage does not automatically make it a substitute for a source designed to measure book earnings.
What the public-use records can and cannot verify
The ACS public-use file is a sample with confidentiality edits, not the complete confidential survey. Income fields are rounded and can be top- or bottom-coded. Negative self-employment income remains in the calculation, including the two writer records with reported losses. Such sparse audit cells do not support a detailed prevalence claim, even though their values matter when reproducing the aggregate.
Responses can also be edited or allocated during survey processing. In the selected writer group, fifty-one records have an occupation-allocation flag, fifty-two an industry-allocation flag and forty-seven a class-of-worker allocation flag. Wage and self-employment income allocation flags appear on fifty-four and forty-four records. These flags are not proof that a respondent supplied incorrect information. They identify a processing feature that a clean-looking result table should not hide.
The source describes consistency edits between income and employment answers as well as imputation where responses are absent. Cross-field agreement in processed records therefore cannot validate the underlying reports by itself. Equally, a current job class and a past-year income component need not align because they refer to different periods and potentially different jobs. Neither agreement nor disagreement establishes undeclared work, fraud or reporting quality in an individual case.
Person weights and all eighty replicate person weights are used for the reported sampling calculations. The study uses a 95% normal approximation and states it explicitly; it does not silently substitute the 90% convention common in ACS products. Approximation intervals for very sparse audit cells are withheld from the public display rather than presenting negative probabilities as plausible population ranges.
Existing research remains the starting point
The National Endowment for the Arts already publishes extensive work on artist occupations, geographic distribution and earnings. Its 2026 releases include 2024 figures and explain that earnings can include wages and net self-employment income, with primary-occupation and work-intensity definitions. This study does not claim to discover those definitions, provide the first writer-earnings estimate or supersede the national indicators.
The Authors Guild’s published 2023 income-survey findings concern participating authors and calendar-year 2022 book and author-related income. Those measures are informative for their stated population. They are not directly comparable with the all-job household-survey measures in this New York case. Different dates, recruitment, occupational coverage and income questions prevent a simple subtraction or ranking from becoming an explanation of authors’ livelihoods.
The original contribution here is the joined, reproducible audit: the same records show how an occupation differs from a mixed industry, how incorporated ownership changes a class grouping, how an income component changes a mean, and how a work-intensity condition selects another population. Each change is kept separate so its meaning remains visible.
Before reusing a writer-income statistic, identify the people counted, the period, the components included and the selection rule. Then inspect whether the comparison holds those choices constant. The tables and charts here make that check possible for one exact source file. They supply a measurement reference, not a promise about what a book or a writing career will earn.
Income measure and population both matter
Income is from all jobs in the preceding 12 months, not book royalties or the current writer job alone. Selected group: 35+ usual weekly hours and 50+ weeks in the preceding 12 months; not necessarily all writing.
Full chart and source
Scroll horizontally to inspect the chart. ↔
Chart and values
Scroll horizontally to see all columns.
| Population | Measure | Sample n | Mean (2024 USD) | 95% interval |
|---|---|---|---|---|
| All 259 current writer records | Wages only | 259 | 75,020 | 61,989–88,051 |
| All 259 current writer records | Wages + net self-employment income | 259 | 85,185 | 71,822–98,548 |
| All 259 current writer records | Paired measure difference: earnings − wages | 259 | 10,165 | 6,196–14,135 |
| 164 with at least 35 usual weekly hours and at least 50 weeks in the preceding 12 months | Wages only | 164 | 95,961 | 78,448–113,473 |
| 164 with at least 35 usual weekly hours and at least 50 weeks in the preceding 12 months | Wages + net self-employment income | 164 | 105,788 | 88,199–123,377 |
| 164 with at least 35 usual weekly hours and at least 50 weeks in the preceding 12 months | Paired measure difference: earnings − wages | 164 | 9,827 | 4,456–15,198 |
Same-record means first, then a selected population. All-job income in the preceding 12 months, not book royalties or income from the current writer job alone. Approximate 95% SDR intervals use 80 replicates.
Three overlapping coverage conditions
These conditions overlap. Do not add the bars; net self-employment losses remain in the calculations. Income is from all jobs in the preceding 12 months, not book royalties or the current writer job alone.
Full chart and source
Scroll horizontally to inspect the chart. ↔
Chart and values
Scroll horizontally to see all columns.
| Measure | Numerator n | Denominator n | Share (%) | 95% interval |
|---|---|---|---|---|
| Zero wage income | 49 | 259 | 17.81 | 12.58–23.05 |
| Positive net self-employment income | 77 | 259 | 30.89 | 23.19–38.59 |
| Negative net self-employment income | 2 | 259 | 0.82 | Withheld: sparse diagnostic cell |
| Nonzero net self-employment income | 79 | 259 | 31.71 | 23.93–39.49 |
| Zero wages and positive net self-employment income | 49 | 259 | 17.81 | 12.58–23.05 |
| Positive combined INTP income: interest, dividends, net rent, net royalties, estate and trust income; cannot isolate book royalties | 67 | 259 | 23.84 | 17.37–30.31 |
| At least 35 usual weekly hours and at least 50 weeks in the preceding 12 months | 164 | 259 | 66.70 | 59.59–73.81 |
| Current owner of an unincorporated business (COW 6), with zero net self-employment income in the preceding 12 months | 0 | 259 | 0.00 | Withheld: sparse diagnostic cell |
| Other current class (outside COW 6), with nonzero net self-employment income in the preceding 12 months | 16 | 259 | 7.85 | 2.88–12.83 |
All 259 writer records; income is from all jobs in the preceding 12 months. Characteristics overlap and must not be added. Sparse-cell population intervals are withheld; raw normal diagnostics remain in downloads. Zero observed records do not establish zero population prevalence. INTP cannot isolate book royalties.
Incorporated owners change the statistical grouping
The 11.3-point difference is incorporated business owners, already included in the second group. A classification comparison on the same 259 records, not new jobs or an estimate of payroll-survey omissions.
Full chart and source
Scroll horizontally to inspect the chart. ↔
Chart and values
Scroll horizontally to see all columns.
| Measure | Numerator n | Denominator n | Share (%) | 95% interval |
|---|---|---|---|---|
| Mixed publisher industry among writers | 15 | 259 | 4.91 | 2.09–7.73 |
| Writer occupation among mixed publisher industry | 15 | 300 | 4.28 | 1.65–6.90 |
| Owners of unincorporated businesses among writers | 63 | 259 | 23.86 | 16.96–30.75 |
| Owners of incorporated businesses among writers | 24 | 259 | 11.31 | 5.46–17.15 |
| Literal employee categories COW 1–5 among writers | 171 | 259 | 64.70 | 56.02–73.38 |
| Census wage/salary group COW 1–5 + 7 among writers | 195 | 259 | 76.01 | 69.11–82.90 |
Denominators vary: 259 writer records or 300 mixed-industry records. Counts are unweighted; percentages are weighted. Same numerator does not mean same percentage.
Data & downloads
Scroll horizontally to see all columns.
| Population | Measure | Sample n | Mean (2024 USD) | 95% interval |
|---|---|---|---|---|
| All 259 current writer records | Wages only | 259 | 75,020 | 61,989–88,051 |
| All 259 current writer records | Wages + net self-employment income | 259 | 85,185 | 71,822–98,548 |
| All 259 current writer records | Paired measure difference: earnings − wages | 259 | 10,165 | 6,196–14,135 |
| 164 with at least 35 usual weekly hours and at least 50 weeks in the preceding 12 months | Wages only | 164 | 95,961 | 78,448–113,473 |
| 164 with at least 35 usual weekly hours and at least 50 weeks in the preceding 12 months | Wages + net self-employment income | 164 | 105,788 | 88,199–123,377 |
| 164 with at least 35 usual weekly hours and at least 50 weeks in the preceding 12 months | Paired measure difference: earnings − wages | 164 | 9,827 | 4,456–15,198 |
Same-record means first, then a selected population. All-job income in the preceding 12 months, not book royalties or income from the current writer job alone. Approximate 95% SDR intervals use 80 replicates.
Scroll horizontally to see all columns.
| Measure | Numerator n | Denominator n | Share (%) | 95% interval |
|---|---|---|---|---|
| Zero wage income | 49 | 259 | 17.81 | 12.58–23.05 |
| Positive net self-employment income | 77 | 259 | 30.89 | 23.19–38.59 |
| Negative net self-employment income | 2 | 259 | 0.82 | Withheld: sparse diagnostic cell |
| Nonzero net self-employment income | 79 | 259 | 31.71 | 23.93–39.49 |
| Zero wages and positive net self-employment income | 49 | 259 | 17.81 | 12.58–23.05 |
| Positive combined INTP income: interest, dividends, net rent, net royalties, estate and trust income; cannot isolate book royalties | 67 | 259 | 23.84 | 17.37–30.31 |
| At least 35 usual weekly hours and at least 50 weeks in the preceding 12 months | 164 | 259 | 66.70 | 59.59–73.81 |
| Current owner of an unincorporated business (COW 6), with zero net self-employment income in the preceding 12 months | 0 | 259 | 0.00 | Withheld: sparse diagnostic cell |
| Other current class (outside COW 6), with nonzero net self-employment income in the preceding 12 months | 16 | 259 | 7.85 | 2.88–12.83 |
All 259 writer records; income is from all jobs in the preceding 12 months. Characteristics overlap and must not be added. Sparse-cell population intervals are withheld; raw normal diagnostics remain in downloads. Zero observed records do not establish zero population prevalence. INTP cannot isolate book royalties.
Scroll horizontally to see all columns.
| Measure | Numerator n | Denominator n | Share (%) | 95% interval |
|---|---|---|---|---|
| Employee categories COW 1–5 | 171 | 259 | 64.70 | 56.02–73.38 |
| Owners of unincorporated businesses, COW 6 | 63 | 259 | 23.86 | 16.96–30.75 |
| Owners of incorporated businesses under the U.S. classification, COW 7 | 24 | 259 | 11.31 | 5.46–17.15 |
| Unpaid family business/farm work, COW 8 | 1 | 259 | 0.14 | Withheld: sparse diagnostic cell |
Same 259 writer records. COW 7 incorporated owners are included with wage-and-salary workers in the Census grouping; the one-record unpaid-family cell is diagnostic.
Scroll horizontally to see all columns.
| Population | Sample n | Estimated people | 95% interval |
|---|---|---|---|
| Writer and mixed publisher industry | 15 | 1,306 | 522–2,090 |
| Writer, other industry | 244 | 25,292 | 21,598–28,986 |
| Other occupation, mixed publisher industry | 285 | 29,222 | 24,793–33,651 |
| Neither group | 97,016 | 9,593,333 | 9,552,599–9,634,067 |
Four rows partition 97,560 selected New York State civilian-employed resident records aged 18+. Industry 6481 is mixed, not book-only. The 15-record intersection is supporting context with wide uncertainty.
Scroll horizontally to see all columns.
| Measure | Numerator n | Denominator n | Share (%) | 95% interval |
|---|---|---|---|---|
| Mixed publisher industry among writers | 15 | 259 | 4.91 | 2.09–7.73 |
| Writer occupation among mixed publisher industry | 15 | 300 | 4.28 | 1.65–6.90 |
| Owners of unincorporated businesses among writers | 63 | 259 | 23.86 | 16.96–30.75 |
| Owners of incorporated businesses among writers | 24 | 259 | 11.31 | 5.46–17.15 |
| Literal employee categories COW 1–5 among writers | 171 | 259 | 64.70 | 56.02–73.38 |
| Census wage/salary group COW 1–5 + 7 among writers | 195 | 259 | 76.01 | 69.11–82.90 |
Denominators vary: 259 writer records or 300 mixed-industry records. Counts are unweighted; percentages are weighted. Same numerator does not mean same percentage.
Scroll horizontally to see all columns.
| Population | Measure | Sample n | Median point (2024 USD) | Uncertainty |
|---|---|---|---|---|
| All 259 current writer records | Wages only | 259 | 62,946 | Point only; interval not established |
| All 259 current writer records | Wages + net self-employment income | 259 | 70,052 | Point only; interval not established |
| 164 with at least 35 usual weekly hours and at least 50 weeks in the preceding 12 months | Wages only | 164 | 81,220 | Point only; interval not established |
| 164 with at least 35 usual weekly hours and at least 50 weeks in the preceding 12 months | Wages + net self-employment income | 164 | 88,327 | Point only; interval not established |
Point estimates only; no median sampling intervals are established. Smallest observed value reaching half cumulative weight; not Census grouped/interpolated medians. Zeros and losses retained.
Evidence & method
How we calculated it
- Read the exact 2024 New York person ZIP; select residents aged 18+, civilian-employed ESR 1 or 2, positive PWGTP, and writers/authors OCCP 2850. Mixed publishing industry INDP 6481 is not book-only.
- Compute WAGP and WAGP+SEMP with ADJINC 1.015250; verify the released PERNP identity; retain zeros and losses. Current main-job occupation and past-year all-job income have different reference periods.
- Use PWGTP and 80 person replicate weights; variance equals 4/80 of the sum of squared replicate deviations. The 95% normal approximation is explicitly chosen; paired measure contrasts are calculated within replicates.
- The full-time/year-round subset requires WKHP>=35 and WKWN>=50. Weighted empirical median points use the first value reaching half cumulative person weight, not Census grouped interpolation; no median interval is established.
What the data cannot tell us
- New York STATE residents, not New York City or workplace/employer jobs. The broad writer occupation does not identify fiction authors, KDP accounts or book income.
- Only 15 unweighted records are in the occupation/industry intersection: supporting context with wide uncertainty, not a precise market-share estimate.
- Sparse prevalence cells with numerator or complement below 10 withhold displayed SE/intervals. Raw unbounded normal diagnostics are not admissible prevalence ranges.
- PUMS rounding, top/bottom-coding, allocation and imputation remain. Source editing can create cross-field consistency; no individual report validation or fraud claim is made.
- Incorporated owners (COW 7) are wage-and-salary workers in Census grouping. INTP combines interest, dividends, net rent, net royalties and trusts, and cannot isolate book royalties.
- Work-intensity selection and mean/median choices define different quantities, not causal effects, pay rises, writer-job income or career forecasts.
Original sources
- 2024 ACS one-year PUMS, New York person ZIP
The U.S. Census Bureau 2024 one-year New York person ZIP supplies 210,150 person records; its exact ZIP fingerprint is checked. This analysis exports aggregate tables only.
Accessed October 31, 2026
- 2024 ACS PUMS Data Dictionary
The 2024 PUMS dictionary defines OCCP 2850, INDP 6481, COW, income/work-intensity variables and allocation flags. The CSV state field is STATE, rather than the older ST.
Accessed October 31, 2026
- 2024 ACS Subject Definitions
The 2024 ACS subject definitions distinguish the current primary job from past-year all-job income, include incorporated owners with wage workers, and combine net royalties with other INTP income.
Accessed October 31, 2026
- 2024 one-year PUMS Accuracy
The 2024 PUMS accuracy document and user guide support PWGTP/80-replicate SDR 4/80 calculations and describe confidentiality and rounding limits. The 95% interval convention is this study’s explicit choice.
Accessed October 31, 2026
- 2024 PUMS Top and Bottom Coded Values
The 2024 top/bottom-code documents and state values identify censoring; source values are not reconstructed.
Accessed October 31, 2026
- NEA NASERC Indicator A.7, June 2026; A.1 and A.3, May 2026
NEA NASERC indicators A.1/A.3 (May 2026) and A.7 (June 2026) already describe 2024 artist occupations, locations and earnings. This case audits joined measurement choices rather than claiming the first writer-earnings estimate.
Accessed October 31, 2026
- Authors Guild 2023 Author Income Survey, key takeaways
The Authors Guild 2023 findings measure 2022 book/author-related income among participating authors, a different population, year and income construct from this case.
Accessed October 31, 2026
Cite & reuse this study
Alisson Bento. What a writer’s earnings figure includes: a New York measurement case. Original analysis of U.S. Census Bureau, 2024 ACS one-year Public Use Microdata Sample, New York person file; sources accessed 31 October 2026. https://alissonbento.com/research/new-york-writer-workforce-and-earnings-measurement-2024/
Original analysis and charts by Alisson Bento. Reuse the original derived charts and tables with attribution, source year, population and measurement limits retained. Credit the U.S. Census Bureau for the underlying public-use data. No Census or NEA endorsement is implied; original microdata and source documents are linked rather than redistributed.
Corrections or editorial questions