Still Working at 75: The Oldest Workers Are Nearly Three Times as Likely to Be Self-Employed (2024)
Almost 2 million Americans are still working past 75, a quarter of them for themselves, far above the prime-age rate. The deep tail of the labor force, from Census microdata.
Key findings
- In 2024 1-Year, about 1,942,634 Americans aged 75 or older were still working (±25,178), roughly 7.57% of the 25,645,796 people that age.
- The oldest workforce runs its own show: 26.0% of workers 75 and older are self-employed, versus 9.3% of workers aged 25 to 54, about 2.8 times the prime-age rate.
- It is not all light duty: 40.1% of these workers usually put in 35 hours a week or more.
- They earn less for it: median earnings of $26,397, against $55,839 for prime-age workers.
- Work does not stop at 80: about 428,282 workers are 80 to 84, and 169,299 are 85 or older.
This counts people 75 and older who are still working, using the 2024 1-Year American Community Survey Public Use Microdata Sample. "Working" is the Census employment status recode (ESR) for civilian employed or employed but not at work; "self-employed" combines incorporated and unincorporated self-employment (class of worker, COW 6 and 7). Estimates are weighted with the person weight (PWGTP) and carry 90% margins of error from the 80 replicate weights. Self-employment shares by age band, the full-time share, median earnings, and the age-band counts are point estimates of the weighted distribution; earnings are in constant 2024 dollars (Census factor ADJINC).
Workers 75 and older, 2024 1-Year
The older the worker, the more likely to be self-employed
| Category | Self-employed share of workers |
|---|---|
| 25 to 54 | 9.3% |
| 55 to 64 | 13% |
| 65 to 74 | 18.7% |
| 75 and older | 26% |
Self-employment combines incorporated and unincorporated self-employment (COW 6 and 7).
Still on the clock, by age
| Category | Workers |
|---|---|
| 75 to 79 | 1,345,053 |
| 80 to 84 | 428,282 |
| 85 and older | 169,299 |
Point estimates of the weighted distribution.
States with the most workers 75 and older
| State | Workers 75+ | Share of 75+ working | Sample |
|---|---|---|---|
| California | 215,297 ± 7,038 | 7.79% | 2,748 |
| Florida | 163,686 ± 7,228 | 6.99% | 1,948 |
| New York | 135,529 ± 7,523 | 8.33% | 1,683 |
| Texas | 133,151 ± 5,150 | 7.66% | 1,981 |
| Pennsylvania | 86,269 ± 5,467 | 7.62% | 1,117 |
| Ohio | 69,000 ± 5,509 | 7.39% | 919 |
| Illinois | 67,912 ± 5,100 | 7.16% | 896 |
| New Jersey | 62,694 ± 4,102 | 8.65% | 796 |
| North Carolina | 60,182 ± 4,025 | 7.38% | 843 |
| Virginia | 55,395 ± 3,612 | 8.52% | 703 |
| Massachusetts | 52,810 ± 4,757 | 9.33% | 689 |
| Georgia | 51,022 ± 4,004 | 7.18% | 727 |
Margins of error are at the 90% confidence level, from 80 replicate weights. Sample is the unweighted number of such workers surveyed in each state.
Data Sources & Methodology
The figures on this page are computed directly from the source data below. Narrative text accompanies the figures but never generates them.
American Community Survey (ACS) Public Use Microdata Sample (PUMS)
- Publisher:
- U.S. Census Bureau
- Vintage:
- 2024 1-Year
- Access:
- BigQuery
- Cadence:
- annual
De-identified individual person and housing records from the American Community Survey, the Census Bureau’s largest household survey. The 1-year PUMS file lets analysts compute custom estimates for demographic, social, economic, and housing characteristics down to the Public Use Microdata Area (PUMA) level.
The ACS is a continuous survey that samples households throughout the year. The 2024 1-year PUMS file is a sample of de-identified person and housing records collected over calendar year 2024, representing the resident population of the United States.
Every record carries a sampling weight (PWGTP for persons, WGTP for housing units) that expands the sample to population and housing-unit totals. All statistics on this page are weighted estimates; any unweighted record counts are shown only to convey sample size.
Margins of error are computed from the 80 replicate weights supplied with the PUMS file using the Census Bureau’s successive-difference replication method, and are reported at the 90% confidence level unless stated otherwise.
Sub-state estimates use Public Use Microdata Areas (PUMAs), statistical areas of at least 100,000 residents. The 2020-vintage PUMA boundaries apply to 2022 and later ACS data.
- PUMS is a sample of roughly one percent of the population, so every estimate carries sampling error; small subgroups and small geographies have wider margins of error.
- Figures are survey estimates, not a complete count, and can differ from administrative records or the full ACS tabulations.
- Dollar amounts reflect the survey year and are not adjusted for inflation across vintages unless explicitly noted.
- PUMA boundaries are redrawn after each decennial census, which breaks strict geographic comparability across the boundary change.
How these figures were produced
- Estimate type:
- Weighted survey estimate (person level)
- Weighting:
- PWGTP (person weight)
- Confidence level:
- 90%
- Error estimation:
- Successive-difference replication over 80 replicate weights
Suggested citation
U.S. Census Bureau, 2024 1-Year American Community Survey Public Use Microdata Sample (PUMS). Analysis by Matt George, itsmattgeorge.com. Retrieved 2026-10-15.