The Half That Works

John Rector · November 4, 2026 · Economics The Half That Works Employment, the 1,800-hour year, and the productive system behind a $100 trillion world. A civilization does not need everyone on a payroll to keep everyone alive. It needs a productive system capable of supporting everyone—and a way for people to gain access to what that system produces. For generations, we have treated the full-time worker as the basic unit of economic life. Count the workers. Multiply by their hours. Watch the economy grow. There is real information in that picture. There is also a profound mistake if we turn the picture into a theory of where prosperity comes from. The historical pattern is compelling, but its wording matters. “Employment has stayed at 45–48 percent of the population since World War II” is too precise and, as a universal historical claim, incorrect. The better-supported observation is that measured employment involves a large minority of humanity, often roughly two people in five to one in two, while economic output can change enormously. The worker share moves within that broad scale. What the productive system accomplishes with each hour moves much more. Understanding that distinction requires four separate ledgers: people, hours, production, and ownership. Confuse them and a familiar employment statistic becomes an explanation it cannot possibly provide. 01 / People 02 / Hours 03 / Output 04 / Ownership 01 / People Start with the denominator. Imagine a town with 100 residents: children, students, employed adults, people seeking work, caregivers, and retirees. Forty-seven residents have jobs. That gives us 47 employed people per 100 residents. It does not mean that 53 percent are unemployed. Unemployment has a narrower meaning. Someone must be without work, available for work, and actively seeking it, subject to specific survey rules. The labor force combines employed people and unemployed jobseekers. People outside that group are outside the labor force, not automatically unemployed. The standard U.S. employment-population ratio uses the civilian noninstitutional population aged 16 and older; the international series used here starts at 15. Neither denominator includes every human being. [1] [2] Employed / all residents The share of everyone counted as employed, including children in the denominator. Employed / working-age people The standard employment ratio, with a specified age and population boundary. Labor force / working-age people Participation: employed people plus unemployed jobseekers. Unemployed / labor force The unemployment rate. A different numerator and denominator. A population can have low unemployment while fewer than half its residents are employed. That is an ordinary consequence of childhood, education, retirement, care, and the definitions themselves. What the postwar U.S. record actually says The United States supplies a long, regularly measured postwar record. Dividing annual civilian employment by the broader annual population gives about 39.8 percent in 1948, 38.8 percent in 1950, and 36.4 percent in 1960. The ratio subsequently rose: 43.6 percent in 1980, 47.5 percent in 1990, and 48.5 percent in 2000. In 2024 it was 47.4 percent. These are calculated all-ages ratios, not the official BLS employment-population rate. [3] [4] [5] So the 45–48 percent description fits much of the recent American experience, with exceptions. It does not fit the entire period since World War II. That correction strengthens the argument: even a society that profoundly changed who worked and how it worked never approached universal paid employment. United States / employed civilians per 100 people of all ages 1948 39.8% 1950 38.8% 1960 36.4% 1970 38.4% 1980 43.6% 1990 47.5% 2000 48.5% 2010 44.9% 2019 47.7% 2020 44.5% 2024 47.4% Common scale: 0–100%. Each full track represents the entire population. BLS civilian employment divided by BEA all-ages population. Sources [3–5]. Read the chart as a data table Year Employed / all people 1948 39.8% 1950 38.8% 1960 36.4% 1970 38.4% 1980 43.6% 1990 47.5% 2000 48.5% 2010 44.9% 2019 47.7% 2020 44.5% 2024 47.4% The stability is relative, not literal. A change of a few percentage points represents millions of people, and a recession can devastate families even when the national ratio still looks broadly familiar. The 2020 decline is visible here. So is the much larger transformation between the early postwar decades and the end of the twentieth century. Changes in the age structure alter the denominator. Women’s movement into paid employment changes the numerator. Longer education, retirement, disability, immigration, and recessions also matter. BLS documents the long rise in women’s participation alongside declining male participation; within the prime ages of 25–54, women’s participation reached a survey high in 2023. Aggregate continuity can contain an enormous social rearrangement. [6] The global pattern is closer to the low forties. We must switch datasets explicitly when moving from America to the world. The World Bank’s current modeled employment-ratio series begins in 1991. Using its employment ratio for people aged 15 and older and the United Nations age structure, we can estimate employed people aged 15 and older as a share of the entire population. That calculation gives about 41.6 percent in 1991, 42.1 percent in 2000, 42.6 percent in 2019, and 43.8 percent in 2024. [7] [8] World / approximate employed aged 15+ per 100 people of all ages 1991 41.6% 2000 42.1% 2010 42.6% 2019 42.6% 2020 41.3% 2024 43.8% Common scale: 0–100%. Each full track represents the entire population. Approximation from ILO modeled employment ratios and UN age shares; excludes employment below 15. Sources [7–8]. Read the chart as a data table Year Employed / all people 1991 41.6% 2000 42.1% 2010 42.6% 2019 42.6% 2020 41.3% 2024 43.8% This is a striking degree of continuity across three decades of industrialization, globalization, computerization, and population growth. It is not evidence of an unbroken 45–48 percent global band since 1945. Older international labor-force reconstructions exist: the ILO describes an estimates-and-projections program whose first edition covered 1950–1985. But labor force is not employment, and those historical exercises are not one seamless, unchanged global employment survey. We cannot manufacture an eighty-year statistical constant by stitching unlike measures together. [9] The defensible teaching is therefore precise: a large minority of the population works in measured employment, and that share can remain broadly similar through enormous economic change. The evidence for this is strongest globally from 1991 onward and in the separate U.S. postwar record. The other residents are part of the economy. A child is a future adult. A retiree carries a history of contribution. Someone caring for a parent may make another person’s paid employment possible. A student is building capacity that will matter later. Treating every person outside employment as a passive dependent mistakes a statistical boundary for a judgment about human usefulness. Even “employed” does not mean “salaried full-time employee.” The ILO’s March 2026 review estimates that 57.9 percent of global employment was informal in 2025. The working world includes self-employment, family enterprises, irregular work, and livelihoods far removed from a corporate timecard. [10] Measured employment connects to the rest of society through several routes. Workers produce and maintain services; earnings support household purchases; taxes and contributions help fund public provision; saving and investment finance future capacity. Ownership income, public transfers, credit, and shared household resources also carry purchasing power to people who are not currently employed. These routes overlap. A retiree can own shares, receive a pension, and provide unpaid childcare in the same week. The economy is sustained by a circulating system of production, claims, and spending. The labor force is central to that system. It is not a separate caste producing everything for an otherwise inert population. 02 / Hours The 1,800-hour worker is a unit of accounting. Let us define one full-time equivalent, or FTE, as 1,800 working hours per year. That is a useful convention: 40 hours for 45 weeks, or 37.5 hours for 48 weeks. It is not a universal legal definition of full-time work, nor a measured global annual average. An FTE measures a quantity of hours. It does not necessarily identify one person. Two people working 900 hours each supply one such FTE. One person working 2,700 hours supplies 1.5. Employment headcount counts people; FTE counts standardized time. Actual hours differ across workers and years. OECD hours-worked statistics include full-time, part-time, and part-year workers, additional jobs, and overtime, while excluding leave and other time not worked. The OECD also warns that national source differences limit simple comparisons of annual levels across countries. We should not quietly assign 1,800 hours to every employed person on Earth and call the result observed labor input. [11] A deliberately simplified $100 trillion world 3.5 billion assumed workers × 1,800 assumed hours per worker per year = 6.3 trillion assumed annual working hours $100 trillion ÷ 6.3 trillion hours = about $15.87 of output per assumed hour. $100 trillion ÷ 3.5 billion workers = about $28,571 of output per assumed worker. Illustrative assumptions, not a global hours estimate, wage, or worker’s personal contribution. The arithmetic is correct. The causal inference people often attach to it is not. Dividing a system’s output by the people in it does not establish that each person independently created an equal slice. Dividing by hours does not establish that hours alone produced the value. The $100 trillion scale is also a historical round number, not a current exact total. The World Bank reports world GDP at approximately $111.7 trillion in 2024 and $118.4 trillion in 2025 in current U.S. dollars. These are annual flows valued at market exchange rates, not global wealth and not purchasing-power-parity totals. [12] The FTE became a convenient managerial shorthand for staffing and cost. It never supplied a complete theory of economic production. A timecard tells you how much human time entered an organization. It cannot tell you what machinery, knowledge, energy, infrastructure, or coordination made that time effective. 03 / Output A similar worker share can support a different civilization. Consider an illustrative pair of farms. Each employs ten people, each working the same annual hours. One has hand tools and unreliable water. The other has irrigation, powered equipment, improved seeds, weather forecasts, refrigerated transport, and dependable market access. Equal headcounts and equal hours will not make their outputs equal. Now move that reasoning to a hospital, an electrical grid, a software company, or a port. Today’s work activates a system accumulated over generations. A nurse’s hour depends partly on training, diagnostic equipment, medicines, clean water, scheduling, and colleagues. A programmer’s hour can draw on libraries and tools created by thousands of other people. The present workforce does not start civilization over every Monday. U.S. nonfarm business output per hour was about 5.1 times its 1947 level in 2024 , calculated from annual averages of the BLS productivity index. That is real output per hour, adjusted for price changes, in a specified sector—not nominal GDP and not a global productivity figure. [13] 1947 1.0× 2024 5.1× U.S. nonfarm business real output per hour; 1947 annual average = 1. Bars start at zero. Calculated from BLS data via FRED. BLS explicitly explains that labor productivity reflects technology, capital investment, capacity use, energy and materials, production organization, managerial skill, and workforce characteristics and effort. “Labor productivity” names a ratio with labor hours in its denominator. It does not attribute every improvement to the worker alone. [14] We can express the accounting relationship without pretending it is a causal model: Real output = population × employed share × hours per employed person × real output per hour The terms must cover the same economy, population, and period. Once they do, the expression is an identity: rearranged definitions. It reveals why a relatively stable employed share does not imply a stagnant economy. Population, working time, and especially the productive capacity available per hour can all change. It also reveals why “correlation, not causation” needs care. Labor is a causal input. Stop essential maintenance, logistics, nursing, or food distribution and consequences arrive quickly. What the evidence rejects is the claim that the number of workers or their hours alone determines the level of prosperity. An accounting relationship cannot identify the independent contribution of each jointly necessary input. GDP is a production ledger. GDP measures current production within an economy. The production approach adds value created at each stage, avoiding repeated counting of the same intermediate inputs. The expenditure approach adds final spending; the income approach records the incomes generated by production. These describe the same activity from different sides. [15] In an illustrative bakery, flour purchases are inputs. Bread sales include the value of those inputs plus value added by the bakery. Adding every supplier’s gross sales to every customer’s purchases would count the same production repeatedly. Likewise, a worker spending a wage does not create a second independent pile of output equal to the wage. Income accounting distributes production across employee compensation, operating surplus and mixed income, taxes less subsidies, and related accounting categories. Gross measures also retain the consumption of fixed capital: depreciation is part of what annual production must cover. An owner’s receipts, a company’s revenue, GDP, and household wealth are different quantities. [16] For a longer global comparison, World Bank real GDP in constant 2015 U.S. dollars in 2024 was about 2.65 times its 1991 level. Over those years, the approximate employed share of all people moved from 41.6 to 43.8 percent. This is a much stronger demonstration than comparing nominal dollars across decades, because inflation and exchange-rate changes can enlarge a current-dollar total without equivalent real growth. [17] The lesson is visible: the share of people employed changed modestly relative to total real production. That observation invites an explanation involving population and productivity. It does not establish a law that employment must remain near half forever. 04 / Ownership Financialization makes ownership visible. It does not turn asset trades into GDP. A company’s shares can rise in value while its employee count barely changes. A property owner can receive rent without personally supplying a fresh hour of labor for every payment. Software can serve another customer with little additional staff time. These examples expose a real gap between human hours, ownership income, and market valuation. But we need to preserve the accounting boundaries. Buying an existing share transfers a financial claim. A rise in its price is a capital gain. Neither is itself current production. The U.K. Office for National Statistics separates asset transactions and holding gains from the fees associated with arranging a transaction; those service fees can enter GDP. [18] BEA makes the same basic distinction in its explanation of Federal Reserve banks: loans and securities transactions belong in financial accounts, while capital gains reflect changes in existing asset values rather than current production. A large financial market therefore cannot be read as a pile of GDP created without workers. [19] Choose a transaction / follow the accounting Which ledger changes? A share rises A broker charges A home shelters A parent cares An existing share doubles in price. Wealth valuation changes. The price gain itself adds no GDP. It can affect later spending or financing decisions, but it is not current output. A broker charges a fee to arrange a trade. A current service is supplied. The service contributes to production; the face value of the security being traded is not its value added. Intermediate-service accounting still applies. A homeowner lives in a home they own. Housing services enter GDP through imputed rent. No cash rent payment or new work shift is needed for the home to supply shelter during the year. A parent provides unpaid care at home. Real care is supplied. Most unpaid household services fall outside conventional GDP, even though they support human life and paid work elsewhere. Accounting explanations: ONS on capital gains, BEA on imputations and the GDP production boundary. [18] [20] [16] Housing provides a particularly useful example. BEA includes an estimated rental value for owner-occupied homes so that the same shelter service does not disappear from GDP when a tenant becomes the owner. The building supplies a flow of services. Construction, maintenance, utilities, and management still require resources and work; current output simply cannot be reduced to current payroll hours. [20] My interpretation is that a financialized economy makes the distribution problem especially visible. Control over productive assets and the claims on their future income can matter enormously to a person’s economic position. That is an argument about ownership, power, and access. It does not demonstrate that most production has become unrelated to labor. Income shares do not identify independent causes. The ILO’s March 2026 review estimates the global labor income share at 52.6 percent of GDP in 2025, down from 53.0 percent in 2015. This is income allocated to labor, including an estimated labor component for the self-employed. It is a different statistic from the percentage of people employed. [10] That estimate does not support the assertion that most GDP has nothing to do with labor. Nor would a smaller labor share establish that proposition. A division of income is not an experiment that removes labor from production. Capital and labor often work together; the fact that one receives income does not show that the other contributed nothing. Think of an automated warehouse that still needs electricians, technicians, software, energy, transport, and customers. Automation can reduce direct hours for a task and change who receives the gains. It does not make the whole production chain immaterial. Conversely, paying a technician a wage does not mean the technician’s hands alone created all warehouse output. The question the timecard cannot answer The familiar worker share conceals two distinct problems. The first is productive capacity: can society reliably provide food, shelter, healthcare, energy, transport, and everything else people need? The second is access: who has the income, ownership, public entitlement, or shared resources needed to obtain those things? A system can become more productive while access becomes more unequal. It can create jobs whose pay does not cover necessities. It can raise asset prices without proportionately improving everyday life. Employment totals are indispensable evidence about work and income, but they do not settle any of these questions by themselves. As AI and automation change the work inside firms, I expect this distinction to become harder to ignore. The practical test will be what the coordinated system can deliver, at what resource cost, and how its benefits reach people. That is my inference from the argument, not a forecast derived from a supposedly permanent employment percentage. Since World War II, America has moved from fewer than two employed civilians per five residents toward nearly one per two. Since 1991, the modern global reconstruction has stayed closer to the low forties. Across these different histories, productive capacity has changed far more than those population shares suggest. The 1,800-hour FTE measures human time. Prosperity depends on what we build around that time—and how the resulting capacity serves human lives. Sources & calculation notes Researched November 4, 2026. Charts use 2024 as the latest common full-year comparison. The U.S. series divides BLS annual civilian employment by BEA annual population, including armed forces overseas; it deliberately uses a broader denominator than the official BLS rate. Military personnel are absent from its employment numerator. It measures civilian employment per total population, not every form of employment. The world series approximates employed people aged 15+ per all people as EPR15+ × (1 − population share aged 0–14). It excludes employment below age 15 and combines modeled labor statistics with demographic estimates. It is not an official all-ages employment rate. Rounded values support the broad pattern; differing aggregate coverage and model vintages discourage false precision. As a cross-check, reconstructing 2024 employment from World Bank labor-force totals and modeled unemployment yields about 43.2 percent, compared with 43.8 percent using the age-share method. Both place the global ratio in the low forties. Labor force ; unemployment ; total population . U.S. productivity uses the mean of four quarterly OPHNFB observations for each year: 22.519 in 1947 and 115.35425 in 2024, with 2017 = 100. Ratio: 5.12. World real GDP uses one constant-dollar World Bank series in both years. Neither comparison estimates labor’s separate causal contribution. Original conceptual artwork generated with AI; it is not a historical image or a data visualization. BLS: Concepts and Definitions (CPS) Employment, unemployment, labor force, age boundaries, and headcount. ILOSTAT: Statistics on employment International employment concepts and definitions. BLS: Employment status, 1948 to date (2018 archive) Early postwar annual employment, including 1948 and 1950. BLS: Employment status, last 70 years Annual civilian employment and standard employment-population ratios. BEA via FRED: Population (B230RC0A052NBEA) Annual population including armed forces overseas; downloaded data table. BLS: Women in the labor force, 2023 Long-term participation changes and prime-age participation. World Bank / ILO: Employment to population ratio, 15+, total WDI SL.EMP.TOTL.SP.ZS; modeled series, world aggregate. World Bank / UN: Population ages 0–14 WDI SP.POP.0014.TO.ZS; used to convert the denominator approximately. ILO: Labour force estimates and projections, methodological description 2017 edition; history of international reconstructions and their scope. ILOSTAT: The world of work and the 2030 Agenda: a ten-year review March 19, 2026; 2025 informality and labor income share. Labor share methods: ILOEST. OECD: Hours worked Actual annual hours and limits of cross-country level comparisons. World Bank: GDP (current US$), World WDI NY.GDP.MKTP.CD; 2024 and 2025 annual nominal totals. BLS via FRED: Nonfarm business labor productivity (OPHNFB) Quarterly real output per hour; annual means calculated for 1947 and 2024. BLS: Productivity and Costs technical notes Meaning, coverage, and joint influences on output per hour. OECD: Annual National Accounts frequently asked questions GDP approaches, value added, and national-account concepts. BEA: Measuring the Economy: A Primer on GDP and the NIPAs Income components, depreciation, and the production boundary. World Bank: GDP (constant 2015 US$), World WDI NY.GDP.MKTP.KD; comparison of 1991 and 2024 real output. ONS: Capital Gains contribution to Gross Domestic Product January 11, 2021; securities, asset sales, holding gains, and associated fees. BEA: How do the GDP accounts treat the Federal Reserve banks? Financial transactions versus production and income accounting. BEA: Why does GDP include imputations? Housing-service flows and owner-occupied housing. More writing by John Rector →

The Half That Works

The Half
That Works

Start with the denominator.

What the postwar U.S. record actually says

The global pattern is closer to the low forties.

The other residents are part of the economy.

The 1,800-hour worker is a unit of accounting.

A similar worker share can support a different civilization.

GDP is a production ledger.

Financialization makes ownership visible. It does not turn asset trades into GDP.

An existing share doubles in price.

A broker charges a fee to arrange a trade.

A homeowner lives in a home they own.

A parent provides unpaid care at home.

Income shares do not identify independent causes.

The question the timecard cannot answer

Sources & calculation notes

Like this:

Leave a Reply