What Is Work?

In 1930, John Maynard Keynes predicted that the standard of life in progressive countries a century later would be "between four and eight times as high" as it was then. He expected people to need far less work: "three-hour shifts or a fifteen-hour week may put off the problem for a great while."^[1]

His target year was 2030. The first prediction has held up. In BEA data, US real GDP per person in 2024 was about 7.8 times its 1930 level.^[2]

The second has not. Valerie Ramey and Neville Francis built a century of US time-use data. Leisure per person rose by only about four to five hours a week over the century. Prime-age adults work about as many hours as they did around 1900, because women's added market hours offset the decline in men's.^[3]

Living standards rose about as much as Keynes expected. Hours of work did not fall to match.

Labor, Work, and Jobs

Work is purposeful productive activity, paid or unpaid. John Budd's fuller definition is "purposeful human activity involving physical or mental exertion that is not undertaken solely for pleasure and that has economic or symbolic value."^[4] It includes household production, care, subsistence farming, and volunteer activity as well as paid employment.

Labor is the part of work organized as a market transaction. Three features set it apart:

• Abstraction: the activity is reduced to units of time. An hour of legal analysis and an hour of data entry each count as one hour of labor.
• Exchange: the worker sells labor-power, the capacity to work, for a wage before the work begins.
• Direction: the employer who buys the labor-power directs how the hours are used.^[5]

The line between work and labor can move while the activity stays the same. Claudia Goldin traced how cooking, cleaning, and child care moved from unpaid household production into paid market labor over the twentieth century, and with that move began to be counted in GDP.^[6]

Hannah Arendt drew an older line: the products of labor are consumed almost as fast as they are made, while work makes durable things.^[7] The clay tablets of Uruk recorded grain that was eaten every season, and the tablets have lasted five thousand years.

Marx separated concrete labor, the specific activity such as weaving or carpentry, from abstract labor, the same activity reduced to hours that can be priced and compared across trades.^[5] Workers sell abstract labor. When a machine takes over a task, the abstract labor leaves the market, and the concrete activity continues in the machine.

A job is a bundle of tasks organized into an employment relationship at a particular point in time. In 1966, the National Commission on Technology, Automation, and Economic Progress put it this way:

By around 3000 BCE, the temple households of Uruk employed trained record-keepers who kept clay-tablet accounts of grain, livestock, land, and labor. A thousand years later, under the Third Dynasty of Ur, scribes were paid in rations of barley, oil, and cloth.^[9]

The scribal job is gone. People still record transactions, keep institutional records, and coordinate economic activity at scale.

Those tasks were later regrouped into jobs such as the clerk, the bookkeeper, and the accountant.

Productivity Connects Work and Labor

Robert Solow measured how that ratio changed in the United States between 1909 and 1949. Gross output per man-hour doubled. About one-eighth of the increase came from more capital per worker. Solow attributed the other seven-eighths to "technical change," which he defined as "a short-hand expression for any kind of shift in the production function," including "improvements in the education of the labor force."^[10] Most of the growth came from changes in how the work was done.

Technology changes the ratio in two ways:

• it changes how much labor an existing task needs, through automation and augmentation
• it changes which tasks exist, by creating new work

When a machine performs a task, the purpose behind the work is still a person's. Someone wanted the call connected, the cloth woven, or the claim paid.

The vocabulary also helps with "virtual employees." An LLM agent can perform work. It cannot sell labor-power, enter an employment contract, or carry liability for its output. An agent doing a role's tasks is capital working inside a job that a person still holds and answers for.

The Emergent Nature of Work

What counts as work, and what that work is worth, depends on the social and economic context, and that context keeps changing.

How much work a society does also depends on what gets counted. Marshall Sahlins estimated that foragers spent three to five hours a day getting food.^[11] Once tool-making and housework such as food preparation, firewood, and water are counted, Richard Lee's estimates for the Dobe !Kung rise to about 40 to 45 hours a week, as David Kaplan's review shows.^[12]

Foragers had no demand for scribes, administrators, or lawyers, because the institutions that needed them did not exist. Agricultural surplus supported settled communities, and settled communities needed records.

Record-keeping started before writing: small clay tokens, each standing for a measure of grain or a head of livestock, appear in Near Eastern villages around the time farming began.^[13] Writing appeared in Uruk in the late fourth millennium BCE. Most of the earliest tablets are bookkeeping records, and most of the rest are word lists used to train scribes.^[14]

This is the sense in which work is emergent: new kinds of work arise from needs that earlier work created. Farming produced stores and obligations that had to be tracked, and tracking them became a trade with its own training.

Work grows as societies become more complex, because complexity creates needs that did not exist before.

Émile Durkheim stated a general version of this in 1893: "The division of labor varies in direct ratio with the volume and density of societies." By volume he meant the number of people, and by density how closely they interact. In his account, growth and density "necessitate a greater division of labor."^[15]

Modern data show something similar. Jeffrey Lin found that workers are more likely to be in new work, measured by new titles in revised occupation classifications, in cities that started with more college graduates and a wider variety of industries.^[16] Autor, Chin, Salomons, and Seegmiller estimate that about 60 percent of US employment in 2018 was in job titles the Census Bureau added after 1940.^[17]

The Cycle That Part 5 Follows

Each article in Part 5 takes up one part of a cycle:

• The Task Structure of Work: a job is a bundle of tasks, each task is a sequence of steps, and automation acts on steps.
• Physical Labor and Cognitive Labor: some steps change matter and some change information, and they are automated differently.
• The Value of Human Labor: the labor in each step has a price, set largely by scarcity.
• Labor Automation: automation moves a step from labor to capital.
• Demand: cheaper output changes how much people buy, and freed income goes to new goods that need new work.
• The Evolution of Work: human wants are relative, so they rise as they are met.
• Engines of Creation: tools keep lowering the cost of meeting wants, and each turn of the cycle creates new work.

Telephone switching went through one full turn. Dial service took over the work of connecting calls in the 1920s and 1930s and displaced the operators' labor, while later cohorts of young women found work as typists, secretaries, and service workers.^[18]

What This Means for Knowledge Work

The cycle ran first on physical labor and then on routine cognitive labor such as calculation and record-keeping. LLMs now reach cognitive steps that resemble judgment.

Whether the cycle holds depends on which of those steps machines can perform reliably, whether demand for cognitive output grows as its cost falls, and what people do once the routine steps are gone.

---

References

1. Keynes, J. M. (1930). Economic possibilities for our grandchildren. The Nation and Athenæum, 11 and 18 October 1930. Reprinted in Essays in Persuasion (1931). Macmillan.
2. U.S. Bureau of Economic Analysis, real gross domestic product (chained 2017 dollars, FRED series GDPCA) and population (FRED series B230RC0A052NBEA). Real GDP per person was about $8,850 in 1930 and $68,790 in 2024. (The Maddison Project's 2023 estimates give a smaller ratio, about 5.5 times, for 1930 to 2022.)
3. Ramey, V. A., & Francis, N. (2009). A century of work and leisure. American Economic Journal: Macroeconomics, 1(2), 189–224. doi:10.1257/mac.1.2.189
4. Budd, J. W. (2011). The Thought of Work. ILR Press, an imprint of Cornell University Press. [definition of work, Introduction, p. 2]
5. Marx, K. (1867). Capital: A Critique of Political Economy, Vol. I. [concrete and abstract labor, ch. 1; labor-power, ch. 6; the labor process, ch. 7]
6. Goldin, C. (1990). Understanding the Gender Gap: An Economic History of American Women. Oxford University Press.
7. Arendt, H. (1958). The Human Condition. University of Chicago Press. [labor, ch. III; work, ch. IV]
8. National Commission on Technology, Automation, and Economic Progress [H. R. Bowen, Chairman]. (1966). Technology and the American Economy: Report of the National Commission on Technology, Automation, and Economic Progress (Vol. I). U.S. Government Printing Office, p. 9. Quoted in: Autor, D. H. (2015). Why are there still so many jobs? The history and future of workplace automation. Journal of Economic Perspectives, 29(3), 3–30, at p. 4. doi:10.1257/jep.29.3.3
9. Nissen, H. J., Damerow, P., & Englund, R. K. (1993). Archaic Bookkeeping: Early Writing and Techniques of Economic Administration in the Ancient Near East (P. Larsen, Trans.). University of Chicago Press.
10. Solow, R. M. (1957). Technical change and the aggregate production function. Review of Economics and Statistics, 39(3), 312–320. doi:10.2307/1926047
11. Sahlins, M. (1972). Stone Age Economics. Aldine-Atherton. ["a mean of three to five hours per adult worker per day in food production," p. 34]
12. Kaplan, D. (2000). The darker side of the "original affluent society." Journal of Anthropological Research, 56(3), 301–324, at p. 308. doi:10.1086/jar.56.3.3631086 Citing Lee, R. B. (1984). The Dobe !Kung. Holt, Rinehart and Winston, pp. 51–52.
13. Schmandt-Besserat, D. (1992). Before Writing, Vol. 1: From Counting to Cuneiform. University of Texas Press.
14. Damerow, P. (2006). The origins of writing as a problem of historical epistemology. Cuneiform Digital Library Journal, 2006:1. [bookkeeping documents and school texts, §5; administrative use, §8.7] cdli.earth/articles/cdlj/2006-1
15. Durkheim, É. (1893). The Division of Labor in Society (G. Simpson, Trans., 1933). Macmillan; reprinted by The Free Press of Glencoe. [Book II, ch. 2, p. 262]
16. Lin, J. (2011). Technological adaptation, cities, and new work. Review of Economics and Statistics, 93(2), 554–574. doi:10.1162/REST_a_00079
17. Autor, D., Chin, C., Salomons, A., & Seegmiller, B. (2024). New frontiers: The origins and content of new work, 1940–2018. Quarterly Journal of Economics, 139(3), 1399–1465. doi:10.1093/qje/qjae008
18. Feigenbaum, J., & Gross, D. P. (2024). Answering the call of automation: How the labor market adjusted to mechanizing telephone operation. Quarterly Journal of Economics, 139(3), 1879–1939. doi:10.1093/qje/qjae005