Where the Hours Went

Essay · economics · September 2026

The first part of this essay followed the hours: a century of productivity, taken mostly as income rather than time. This second part follows the income and the work. Who received the gains, what they bought, how much of the work goes into things we need and how much into selling us things we do not, what the unpaid work at home and the hours in front of screens add to the picture, and what the long hours cost people and the planet.

Reading time ≈ 29 min Galbraith 1958 → Ramey 2009 → FIGARO 2026 Part II of two

From the hours to the income

In 1930 Keynes predicted that his grandchildren would be four to eight times richer and would work fifteen hours a week. The Fifteen-Hour Week, the first part of this essay, found that the first half of the prediction came true and the second did not. Output per hour in the rich countries rose about sevenfold, but the working week of the average worker fell by only about a quarter, and in the United States total hours worked per head of population are about where they were in 1929. Counted over whole careers, much of Europe took close to a third of its gains as time, much of it as retirement; America took far less. Either way most of the productivity went into income.

This part asks what that sentence hides. Income for whom? Spent on what? And how much of the work that produced it went into things anyone needed? It also counts two kinds of time that the usual statistics leave out, the unpaid work done at home and the leisure spent in front of screens, and it asks what the long hours cost, to the people who work them and to the planet that pays for the output.

+13%US median full-time wage, 1979–2025, against +139% output per hour

32.5%of US household wealth held by the richest 1% in 2026, up from 23% in 1989

≈½of the hours behind European consumption go to food, housing, energy, transport, health and education

9.3 ha day of paid and unpaid work done by Portuguese women in 1999, against 7.8 for men

Whose income

The income went somewhere, but not to everyone

“Took the income” is an average, and averages are exactly where this story hides its most important fact. GDP per head divides everything the economy produces by everyone in it. It says nothing about who receives it. For the first half of the post-war period that did not matter much, because the gains were spread fairly evenly. Since about 1980, in the United States above all, it matters a great deal.

Figure 1 is the American version, from the Bureau of Labor Statistics.[1] From 1948 to 1973, output per hour in the business sector roughly doubled and average hourly pay, after inflation, rose by almost exactly as much: the two lines move together. Then they part. Between 1979 and 2025 output per hour rose by 139 per cent. Average hourly compensation, which includes the pay of chief executives and the cost of health insurance, rose by about half. The median full-time weekly wage rose by 13 per cent, or 20 per cent with the price index the BLS itself prefers. The real hourly wage of production and nonsupervisory workers, the four-fifths of the private workforce who do not manage anyone, fell for twenty years after 1973 and did not get back above its 1973 level until 2019.

Fig. 1 — US nonfarm business sector, index 1979 = 100. Output per hour; average hourly compensation, deflated by CPI-U (dashed); median usual weekly earnings of full-time workers, deflated by CPI-U; real average hourly earnings of production and nonsupervisory workers (dotted; from 1964, to 2024). Data: BLS Productivity and Costs, Current Population Survey, Current Employment Statistics, CPI.

So the common impression that incomes stopped rising in the late 1970s is right, but it is the typical wage that stopped, not income as a whole. That matters for the Keynes question, because it changes who was offered his deal. A worker whose real wage did not rise for thirty-five years was never offered the 1977 living standard in fifteen hours a week. They were offered the 1977 living standard in 1977 hours.

The gap between the top line and the bottom two is argued over, and the argument is worth knowing, because it is partly about measurement. Economists who have decomposed it, from the Economic Policy Institute on the left to Martin Feldstein on the right, find three pieces.[2] Doing the same sum on the data behind Figure 1 gives roughly this. About a fifth of the gap is a fall in the labour share: the part of business income paid to workers at all, rather than to owners, fell by about thirteen per cent. About two-fifths is prices. Output is valued at the prices of what workers make, which include computers whose prices have collapsed; wages are deflated by the prices of what workers buy, which include rent, medicine and tuition, which have not. Some of that is a real loss for workers, and some is a price index that overstates inflation and misses the value of new goods. And about two-fifths is inequality among workers: the average pulled away from the median as pay at the top rose much faster than pay in the middle, and as more of the median worker’s compensation went into health insurance premiums rather than wages. Lawrence Summers and Anna Stansbury, who looked hardest for evidence that the link between productivity and pay was broken, concluded that productivity still lifts pay almost one for one, but that since 1980 other forces, falling union membership and bargaining power among them, have held the median back about as hard as productivity pushed it up.[3]

Household incomes tell the same story in more detail, and they also show that it is not the same story everywhere. The Luxembourg Income Study measures median disposable income, after taxes and transfers and adjusted for household size.[4] In the United States it rose by 9 per cent between 1979 and 2014, while GDP per head rose by 77 per cent. It then rose quickly, by a quarter in a decade, helped by the tight labour markets of the late 2010s and the pandemic transfers; over the whole period from 1979 to 2024 the median rose by 36 per cent against 112 for GDP per head. Britain is the counter-example: its median disposable income more than doubled between 1979 and 2024, faster than GDP per head. France’s median rose by 43 per cent against 70 for GDP per head between 1979 and 2022; Germany’s by 22 against 43 since reunification; Italy’s not at all. An OECD study of two dozen countries found that median wages fell behind productivity in most of them after the mid-1990s, but by much less than in the United States.[5] GDP per head always grows faster than household income, for boring reasons as well as interesting ones, since it includes depreciation, government spending and profits retained by firms. But the size of the gap is a choice, and it is widest where the least was done about it.

Figure 2 shows where the difference went. The share of pre-tax national income received by the richest one per cent fell in every rich country from the 1920s to the 1970s, the period economists call the Great Compression. Since 1980 it has climbed back in the United States from about 10 to about 21 per cent, the level of the late 1920s, while in France and Sweden it has risen much less, to about 12 and 10.[6] Wealth is more concentrated than income and has concentrated faster. By the Federal Reserve’s own accounts, the richest one per cent of American households owned 23 per cent of all net worth in 1989 and 32.5 per cent in 2026; the richest tenth of that one per cent went from under 9 per cent to 15. The bottom half of all households, some sixty-six million of them, own 2.3 per cent.[7]

Fig. 2 — Pre-tax national income share of the top 1% (WID, via Our World in Data), and share of US household net worth held by the top 1% by wealth (Federal Reserve Distributional Financial Accounts, quarterly from 1989).

Put this back into Keynes’s terms and the fifteen-hour week looks different. The productivity to pay for it was produced. In the United States a large part of it went to people who were already rich, and much of what they did not spend became wealth, which is income from owning rather than working. You cannot take as leisure a raise you did not get.

There is also a ceiling on the other side, and it makes concentration worse for leisure than for income. Time is the one thing that cannot be concentrated: a week has 168 hours for a billionaire as for anyone else, and nobody can take more than all of them off. Wealth has no such limit. Income beyond what a person can use as free time can only be saved or spent, and at the top both happened. The richest one per cent of Americans saved a rising share of their income from the 1980s onward, so much that their savings were largely lent back to the rest of the country as mortgages and consumer debt.[8] What is not lent goes largely into owning things that already exist, shares, land and buildings, rather than into paying for new work, and that bids up their prices. Across fourteen rich countries, rising land prices, not the cost of building, explain about 80 per cent of the rise in house prices since the Second World War:[9] the scarce flat in the city with the jobs is exactly the kind of asset that surplus wealth chases. The money does not leave the economy. It moves from the part that employs people to the part that trades ownership, and comes back to everyone else as higher rents and larger mortgages. Atif Mian, Ludwig Straub and Amir Sufi argue that this pattern, savers at the top and borrowers below, itself weakens spending and pushes interest rates down, so that demand comes to depend on ever more debt.[10] What they spent went disproportionately on services that are other people’s hours: cleaning, childcare, restaurants, personal care. One study estimates that rising demand for such services from better-paid households explains a third of the growth of low-skill service jobs in the United States in the 1990s.[11] And the best-paid themselves work the longest hours.[12] A productivity gain that goes to the top is lost as leisure twice over. The people who receive it cannot turn it into time, and when they spend it they buy the time of others, or lend it back as debt that others work to repay. Leisure spread evenly is the one form of the gains that everyone could share; wealth concentrated becomes a claim on other people’s hours.

Better off?

What the extra income bought

The next question is whether the quality of life rose with the income. Some of it plainly did. A median household today has things no king had in 1930: antibiotics, vaccines, cheap flights, a telephone that is also a library. Life expectancy in the rich countries rose from about sixty in 1930 to over eighty. Some of the flatness in real wages is a measurement artefact running the other way: the official price indices were found in 1996 to overstate American inflation by about a percentage point a year, largely because they missed improvements in quality and the arrival of new goods, and William Nordhaus showed with the price of light that the error compounds over a century.[13] A worker whose measured real wage is flat is still better off than those figures say.

But the goods got cheaper and the things that anchor a life did not. Figure 3 shows American consumer prices by category from 2000 to 2025. The hourly wage of production and nonsupervisory workers rose by 124 per cent, faster than prices in general at 87. Televisions fell by 98 per cent, computers by 92, and clothes did not get dearer at all. College tuition rose by 188 per cent and childcare by 147. Medical care and rents rose about as fast as wages, so they took the same share of the pay packet, and the index for housing measures rents, not the price of buying a home, which rose much faster in most cities.

Fig. 3 — Change in US consumer prices by category, 2000–2025, before adjusting for inflation. Dashed: average hourly earnings of production and nonsupervisory workers. Dotted: all items, CPI-U. Pink bars rose faster than wages. “Housing” is the CPI shelter index, which tracks rents and not house prices. Data: BLS.

The pattern is Baumol’s again, with a twist that Keynes would have recognised. What got cheap is what machines make. What got dear is made of people’s time, like teaching and caring, or of scarcity, like land in the places where the jobs are. Many of the dear things are also positional goods, what Keynes called relative needs: the house near the good school, the degree that sorts you into the better-paid job. When those rise faster than wages, a household cannot keep its place by working the same hours. It adds hours, most often by adding a second earner. Elizabeth Warren and Amelia Tyagi called this the two-income trap: families that sent a second parent to work in the 1980s and 1990s found that much of the second income went on bidding up the price of houses in good school districts against other families doing the same.[14] Nobody chose that outcome. Each household chose sensibly, and together they produced a result that none of them wanted.

Pointless work is a smaller part of the answer than it seems. David Graeber argued that the fifteen-hour week was absorbed by pointless jobs, roles that even the people doing them privately think should not exist, and put them at 37 to 60 per cent of all work.[15] An earlier essay on this blog took the argument further: the bigger problem, it suggested, is not whole jobs that are pointless but real jobs slowly filled up with pointless tasks. The survey evidence treats these two claims very differently. On whole jobs it is hard on Graeber. Across the EU, 4.8 per cent of workers said in 2015 that they rarely or never felt they were doing useful work, down from 7.8 per cent in 2005, and the feeling was more common among manual workers than in finance.[16] Across 47 countries about 8 per cent of workers think their job socially useless and another 17 per cent are unsure, with no upward trend.[17] A 2015 American survey found higher figures, 19 per cent, and did find Graeber’s occupations, finance, sales and management, over-represented.[18] On any of these measures the share is well below Graeber’s estimate, and far too small to explain why hours stopped falling. The task version is harder to measure and more plausible, since time spent on reporting, compliance and email has certainly grown. But it explains why output per hour did not rise even faster, not why working hours stayed long: the productivity that was achieved was real, and it went into income. The rest of the answer lies in what the work is for.

Sixty kinds of crisps. The supermarket aisle makes the point in a more everyday way. The average American supermarket stocked just over 14,000 different items in 1980 and about 51,000 at the peak in 2008; the number has since fallen back to around 33,000 as smaller stores spread.[19] The world made roughly twice as many garments in 2015 as in 2000, more than 100 billion a year, while the number of times each one was worn before being thrown away fell by 36 per cent, and almost three-quarters of the material ends up in landfill or an incinerator.[20] About a fifth of the food available to consumers, 1.05 billion tonnes in 2022, was wasted.[21] Behind each of these numbers is work: designing, making, packaging, advertising, shipping, stocking, selling and finally disposing of things that add little or nothing to anyone’s life. The sixtieth brand of crisps is not made by one person in a kitchen. It needs a marketing plan, a supply chain, shelf space and a campaign to persuade someone that it is the one they want.

Economists will object that variety has value, and they have measured some: Christian Broda and David Weinstein put the gain from the growing number of imported varieties at about 2.6 per cent of American GDP between 1972 and 2001.[22] The psychologists’ counter-claim, that too much choice makes people miserable, turned out weaker than its famous jam experiment suggested: shoppers offered 24 jams were a tenth as likely to buy as those offered six, but across fifty such experiments the average effect was close to zero.[23][24] Neither result touches the real question, which is not whether the sixtieth variety is worth something to somebody but whether it is worth the hours. That question never gets asked, because nobody is in a position to ask it: each firm has every reason to add a variety, and each worker is paid to make, market or move it.

This is the consumption side of Graeber’s argument, and it is stronger than the jobs side. Few workers think their own job is pointless. But a great many of them work, usefully and diligently, to produce, sell and move things that nobody would miss. John Kenneth Galbraith made the point in 1958: in a rich economy many of the wants that production satisfies are created by production itself, through advertising and emulation, so the fact that people buy something is no proof that they needed it.[25] It is also Keynes’s distinction between absolute needs, which can be met, and relative ones, which cannot. The largest items in household budgets are still housing, transport and health, about 33, 17 and 8 per cent of American spending,[26] where the problem is price and position rather than variety. But around them has grown an economy of things that exist because they can be sold, and it takes hours. Part of the answer to the fifteen-hour week is that the income it would have cost was spent, a good deal of it on things we were persuaded to want.

Where the hours actually go. The same question can be asked of the whole economy: what share of all the hours worked goes into the things a society plainly needs? Eurostat’s national accounts count the hours worked in 64 industries, and they can be sorted into three groups, following the idea of a “foundational economy” of everyday essentials.[27] Essential: farming and fishing, food manufacturing, energy, water and waste, construction, transport and post, telecoms, education, and health and care. Mixed, because they serve needs and wants alike: trade, other manufacturing, restaurants and hotels, administrative and support services, public administration and other personal services. Other: finance, real estate, professional services and consultancy, advertising and market research, publishing, media and software, and recreation and gambling. Figure 4 shows the result.[28]

Fig. 4 — Share of all hours worked in essential, mixed and other industries. Essential: NACE A, C10–C12, D, E, F, H, J61, P, Q. Mixed: B, rest of C, G, I, N, O, S, T. Other: rest of J, K, L, M, R, U. Data: Eurostat, national accounts employment by industry (nama_10_a64_e), hours worked by all employed persons.

The share of hours in essential industries has not grown with prosperity. It has fallen: from 41 to 35 per cent in Portugal and from 41 to 37 per cent across the EU between 1995 and 2023, and from 42 to 38 per cent in France since 1978. Inside it, the work changed more than the total did. Farming collapsed as machines replaced people, in Portugal from 13 to 4 per cent of all hours, while health and care grew, from 5.5 to 8.7 per cent in Portugal and from 9 to 13 per cent in Germany. The group that grew in every country is the third. Finance, property, consultancy, advertising, media and recreation took 7 per cent of Portuguese hours in 1995 and 13 per cent in 2023; 10 and 15 per cent across the EU; 9 per cent of French hours in 1978 and 18 per cent now. Advertising and market research alone doubled in Portugal, to about 1 per cent of all hours worked.

The classification is rough, and the caveats matter. Industry figures show where people work, not what their work is for: the food industry makes the sixtieth brand of crisps as well as bread, and the third group includes the software this essay was written with. Trade moves hours abroad, so most of the clothes Europeans wear are sewn in Asia and do not appear in these figures at all. But the direction is the one the argument predicts. As productivity rose, the hours freed from growing food and making things did not go into leisure, and went only partly into health and care. A growing share went into finance, property, consultancy and selling.

The hours behind what we consume. The industry figures count where Europeans work, including work done for export and for other firms. The more direct question is how many hours, anywhere in the world, it takes to produce what people actually consume. Eurostat’s FIGARO tables trace every euro of final spending back through the supply chains of fifty countries and regions; combined with the hours worked in each industry of each country, they give the hours embodied in consumption, including those of the seamstress in Bangladesh and the soya farmer in Brazil.[29] Counting what households buy together with the health, education and social care that governments provide, Figure 5 shows the result for 2010 and 2022, the span the tables cover.

Fig. 5 — Hours worked anywhere in the world to produce the consumption of households and non-profit institutions plus government spending on health, education, social care and culture. Essential: food and farm products, health and social care, education, energy, water and waste, housing and construction, transport and post. Other: finance and insurance, business services, recreation and personal services. Data: Eurostat FIGARO 2026 edition; hours from Eurostat, OECD and ILO; computed by scripts/io_hours.py.

About half the hours behind European consumption go to the essentials: food, housing, energy, water, transport, health and education. That share has not fallen. Between 2010 and 2022 it rose a little, from 50 to 52 per cent across the EU and from 49 to 54 per cent in Germany, almost entirely because health and care took more; in Portugal it held at about 47 per cent. The rest is where the argument of this section lives. Finance, business services and recreation, bought directly, take about a tenth. The largest single item in the remainder is selling: the margins of wholesale and retail trade, the work of moving goods from factory to shelf to customer, take 12 per cent of the hours behind EU consumption and 17 per cent in Portugal, as many hours as the food itself. Clothes, cars, gadgets and other manufactured goods take another 14 per cent in the EU, and restaurants and hotels 8 per cent, 14 in Portugal, where the figure also includes what tourists spend.

The two measures tell one story from two sides. Productivity made the necessities cheap in hours, and their share of the work behind consumption has held steady rather than grown. The industry view shows where the freed hours went: into finance, property, consultancy and selling. The consumption view shows that much of that work does not reach us as a separate product at all; it is embedded along the supply chain in the price of everything we buy, the essentials included. What neither can do is separate the sixtieth brand of crisps from the first, or the garment worn forty times from the one worn four. That is the real limit of any statistic of this kind: the waste the argument is about lives inside the categories, not between them.

Two broader measures make the same point. Charles Jones and Peter Klenow built a measure of welfare that counts consumption, leisure, life expectancy and inequality, and found that Western European countries whose income per head is a third lower than America’s come out close to it, because their people work less, live longer and share the income more evenly.[30] And the United States, which took the least of its productivity as time and let the most of it go to the top, has fallen behind in the most basic measure of all. In 1980 Americans could expect to live about a year less than the average of eight other rich countries. By 2023 the gap was three and a half years: 79.3 against 82.9.[31] Anne Case and Angus Deaton traced much of it to what they called deaths of despair, from drugs, alcohol and suicide, concentrated among Americans without a university degree, whose wages were the ones in the bottom lines of Figure 1.[32]

On happiness the evidence is honestly mixed. Richard Easterlin found in 1974 that average happiness in the United States did not rise as income rose, which fits Keynes’s relative needs exactly: if what makes you happy is having more than others, a country cannot get happier by all getting richer. Later work with better data finds that richer people and richer countries do report higher life satisfaction, and that it keeps rising with income.[33] Both can be true. Income buys wellbeing, but a given rise buys less when everyone else gets it too, and much less when it arrives as more hours.

The uncounted shift

The work that is not paid

Every figure in the first part of this essay counts paid work. A large share of the work any society does is not paid: cooking, cleaning, shopping, caring for children and for the old. For most of the twentieth century it was done mainly by women, and a good part of the story of hours is that some of it moved into the market, as women took paid jobs and households bought what they had once made. Valerie Ramey reconstructed the American record from old surveys. Prime-age women spent about 47 hours a week on home production in 1900, 41 in 1965 and 29 in 2005; prime-age men went the other way, from about 4 hours to 17.[34] Adding paid and unpaid work together, and counting everyone, she and Neville Francis found that leisure per person rose by only four or five hours a week over the whole twentieth century, and that for people aged 25 to 54 it was back at its 1900 level by 2005.[35] Measured by time diaries, men and women both gained some leisure between 1965 and 2003, the less educated more than the rest.[36] But the great gain in free time that the hours-per-worker series suggest is much smaller once the unpaid shift is counted.

Figure 6 shows paid and unpaid work together, from the most recent time-use survey in each country. Women do more unpaid work everywhere, and in most countries more work in total. Portugal stands out. In its 1999 survey, Portuguese women aged 15 to 64 did about 9.3 hours of paid and unpaid work a day, against 7.8 for men, the largest total and the largest gap in the group, and had about 3.3 hours of leisure against 4.8 for men.[37] A national survey in 2015, asked differently and not directly comparable, found the same pattern among people in paid work: women put in more than an hour a day more than men once the work at home is counted.[38] A country that took so little of its productivity as time took least of all for the people doing both shifts.

Fig. 6 — Paid work (including study and commuting) and unpaid work (housework, shopping, care, volunteering), hours per day, ages 15–64, from each country’s latest survey in the OECD Time Use database; survey years differ and several countries’ definitions differ, so comparisons are approximate. Data: OECD.

The time we did get

Where the leisure goes

Keynes worried that people freed from work would not know what to do with themselves. The time-use surveys give a blunt answer. In the United States in 2025, watching television took 2.6 hours a day, half of all the 5.2 hours of leisure the average adult had; games and leisure computer use took another 37 minutes, up from 25 a decade earlier, while time spent socialising fell from 41 minutes a day to 35.[39] European surveys around 2000 found television taking between about 30 and 55 per cent of free time, depending on the country.[40] Surveys by the advertising industry, which have every reason to count generously, now put the time adult internet users spend online at more than six and a half hours a day.

This is the leisure side of the argument about the sixtieth brand of crisps. Much of the free time that productivity did deliver has been captured by industries whose business is to hold attention and sell it on, a process this blog examined in The Accelerant. Time spent in front of a screen is not time wasted by definition. But a large part of it is designed, like the variety in the supermarket, to be wanted rather than chosen, and it is the hours of attention, not the price of the product, that pay for it.

The price of the hours

What long hours cost

The last piece is the one the ledger of Part I leaves out. Hours of work are not just a subtraction from leisure. Past a point they do damage, and they are not always hours that anyone wanted.

Health. A pooled analysis of prospective studies covering more than six hundred thousand people found that working fifty-five hours a week or more, compared with thirty-five to forty, raised the risk of stroke by about a third and of coronary heart disease by about an eighth.[41] The World Health Organization and the International Labour Organization estimated from that and similar evidence that long hours caused about 745,000 deaths from stroke and heart disease worldwide in 2016, and that 488 million people, about one worker in eleven, were working that long.[42] Long hours are also linked to depression, sleep loss and injury; the evidence there is less certain, but it points the same way.

Output. The extra hours are worth less than they look. John Pencavel went back to records of British munitions workers in the First World War, a rare case where hours and output were both measured carefully, and found that output rose in proportion to hours up to about fifty a week and then flattened: at seventy hours a week, workers produced hardly more than at fifty-six.[43] Modern knowledge work is harder to measure, but nobody has found that it is exempt.

Munitions workers at the Chilwell shell-filling factory, Nottinghamshire, July 1917: the industry, and the war, whose careful records of hours and output let Pencavel measure what the extra hours were worth. Photograph by Horace Nicholls, Imperial War Museums (Q 30040); public domain, via Wikimedia Commons.

Time squeeze. The fall in hours per worker hides a rise in hours per household. In the United States, Jerry Jacobs and Kathleen Gerson found that the combined paid hours of married couples rose by about ten hours a week between 1970 and 2000, as dual-earner couples went from the exception to the rule; and the long hours were concentrated among professionals, while other workers wanted more hours than they could get.[44] Daniel Hamermesh and Jungmin Lee found that feeling pressed for time rises with income across countries, which says the squeeze is not only a problem of the poor.[45] The same household that earns more has less time in which to spend it, and more of what it earns goes on buying back time: childcare, cleaning, ready meals, deliveries.

The journey. Commuting is not counted as working time in these figures, and it is not leisure either. The average American commute, one way, rose from 21.7 minutes in 1980 to 27.2 in 2024; across the EU it was about 25 minutes in 2019, and 21 in Portugal.[46] Twice a day, five days a week, that is another four hours a week that belong to the job without appearing in any of its statistics.

Choice. This is where “took the income” needs the most salt. Economists have known since the 1980s that most workers cannot choose their hours. Jobs come as packages; the choice on offer is usually forty hours or none, sometimes fifty or none, and a worker who wants thirty must usually change job and take a cut in hourly pay to get it.[47] When surveys ask, large minorities of workers in every rich country say they would rather work fewer hours for less money, and rather fewer say the reverse. The split between income and time is not a decision each person made. It is a decision made by employers, laws, bargaining and habit, and ratified, one job offer at a time, by people with rent to pay.

Work people want. None of this means that work is simply a cost that people would shed if they could. When Swedish and Dutch lottery winners were followed for years after their windfall, most kept working: a large win reduced earnings only modestly, by around one per cent of the prize a year, and hardly changed whether people worked at all.[48] And losing a job hurts far more than the lost income explains; in German panel data, the non-financial cost of unemployment to well-being is much larger than the financial one.[49] Work gives structure, company, status and a sense of use. Keynes knew this, and worried about what people would do without it. The argument of this essay is not that work is bad but that its quantity, and much of its direction, were decided for people rather than by them.

So the honest version of the ledger is this. Productivity rose sevenfold. In aggregate, rich societies took four-fifths of it as income and one-fifth as time, but that aggregate was not a choice anyone made. In the countries where the income was shared, where hours were limited by law and bargaining, and where the dear things, health, education, care, were provided collectively, people got more time, longer lives and incomes that rose in the middle as well as at the top. Where they were not, the median worker got little of the income and almost none of the time, and in the United States, on the measure that matters most, a shorter life than their peers abroad.

The planet’s share

Income, time and emissions

There is one more reason the split between income and time matters, and Keynes could not have known it. Productivity taken as income is spent, and spending is production: more goods, more transport, more energy, more emissions. Productivity taken as time is not. Studies that try to measure the difference agree on its direction if not its size. A Swedish study of household consumption estimated that cutting working time by 1 per cent would cut energy use by about 0.7 per cent and greenhouse gas emissions by about 0.8 per cent, almost entirely because people with less income buy less; what they do with the extra free time adds back less than a tenth of the saving.[50] A panel study of OECD countries from 1970 to 2007 found that countries working shorter hours had smaller ecological and carbon footprints, mostly through lower output.[51] A systematic review in 2021 warned that much of this literature is weaker than its headlines, with some studies unable to separate the effect of hours from everything else that differs between rich countries.[52] The careful statement is that shorter hours reduce emissions to the extent that they reduce consumption, which is to say to the extent that the productivity is taken as time rather than income.

That turns Keynes’s question into a modern one. The climate series on this blog argues that the pace of warming is itself accelerating. An economy that takes each year’s productivity gain as more output must decarbonise that extra output as well as the old; one that takes part of it as time has less to decarbonise. The fifteen-hour week was offered in 1930 as a reward. It may yet be needed as a constraint.

Coda

Where the hours went

The two parts of this essay together give a fuller answer to Keynes than either alone. The productivity came, as he said it would. The rich countries took some of it as time, more in Europe than in America and much of it at the end of life, and most of it as income. That income went disproportionately to the top in the United States since 1980, and there it could not become leisure, because nobody can take more than all of their time off; it became wealth, which bid up the price of homes and came back to everyone else as debt. What the rest of the income bought was real, in health above all, but a large part of the work behind it now goes into selling, into finance and business services embedded in every price, and into a churn of goods and variety that nobody would miss, while a second, unpaid shift of work at home, done mostly by women, never appeared in the accounts at all. The leisure that did arrive has been bid for by industries that sell attention.

None of this was a law of nature, and none of it was a free choice. It was the result of rules: about how pay is set, how wealth is taxed, how long the working week may be, what a pension age is tied to, what may be advertised to whom. Keynes thought the economic problem would solve itself and leave humanity with its real problem, how to live well. The first part came close to true. The second is still waiting, and it will not be solved by more productivity alone. It will be solved, if it is, by deciding again, together, what the productivity is for.

Open threads

Where this could go

A Portuguese time-use survey. Portugal’s last full diary survey is from 1999. A new one, comparable with the European rounds, would show whether the second shift has shrunk as women’s paid work and the minimum wage have risen, and how much of the country’s leisure is now spent on screens.

Variety inside the categories. The input-output figures cannot separate the sixtieth brand from the first. Scanner data on the number of products sold, and the hours of design, marketing and logistics behind each, would put a number on the part of the argument this essay could only describe.

On method and tools

This piece was written collaboratively with Claude Opus 5.5 (Anthropic): human specification, editorial direction and critical review; machine data analysis, research and drafting. The numbers are computed by scripts/keynes.py and scripts/io_hours.py from public data: the BLS public API for pay, productivity and prices; the Federal Reserve’s Distributional Financial Accounts; the WID and LIS series published by Our World in Data; Eurostat’s hours worked by industry; the FIGARO input-output tables with a worldwide hours account assembled from Eurostat, OECD and ILO data; and the OECD Time Use database. Results are in docs/keynes-results.json and the figures in docs/keynes-figures.html. The decomposition of the gap between productivity and the median wage splits the log difference, 1979–2025, into the change in the labour share, the gap between the output deflator and consumer prices, and the gap between average compensation and the median weekly wage, computed with two price indices. Hours worked outside Europe are estimates; time-use surveys differ in year and definitions between countries.

The cover photograph is the snacks aisle of a supermarket in Bergen, Norway, in 2017; photograph by Wolfmann, via Wikimedia Commons, licensed CC BY-SA 4.0, cropped.

Authored by: Luis Matos Ferreira — Physicist, Developer, Writer

Related essays on this blog

1. The Fifteen-Hour Week — Part I: Keynes’s prediction, productivity and the hours.
2. The Two-Thirds Country — the same questions for Europe and Portugal.
3. The Long Retirement — exit ages, healthy years and pensions.
4. Bullshit Jobs — pointless jobs, and the slower colonisation of real jobs by pointless tasks.
5. The Accelerant — social media, attention and the second accelerant.
6. The Second Derivative — why climate change is speeding up.

Sources

1. US Bureau of Labor Statistics: Productivity and Costs, nonfarm business sector (output per hour, hourly compensation, real hourly compensation, implicit price deflator, labour share); Current Population Survey, median usual weekly earnings of full-time wage and salary workers; Current Employment Statistics, average hourly earnings of production and nonsupervisory employees, current and 1982–84 dollars; Consumer Price Index for All Urban Consumers, all items and categories. All via the BLS public API.
2. Bivens & Mishel, “Understanding the Historic Divergence Between Productivity and a Typical Worker’s Pay”, Economic Policy Institute Briefing Paper 406 (2015); Feldstein, “Did wages reflect growth in productivity?”, Journal of Policy Modeling 30, 591 (2008).
3. Stansbury & Summers, “Productivity and Pay: Is the link broken?”, NBER Working Paper 24165 (2017).
4. Luxembourg Income Study Database, median equivalised disposable household income, via Our World in Data.
5. Schwellnus, Kappeler & Pionnier, “Decoupling of wages from productivity: Macro-level facts”, OECD Economics Department Working Papers 1373 (2017).
6. Piketty & Saez, “Income Inequality in the United States, 1913–1998”, Quarterly Journal of Economics 118, 1 (2003); World Inequality Database, pre-tax national income shares, via Our World in Data.
7. Board of Governors of the Federal Reserve System, Distributional Financial Accounts, net worth by wealth percentile group, 1989:Q3–2026:Q2.
8. Mian, Straub & Sufi, “The Saving Glut of the Rich”, NBER Working Paper 26941 (2020).
9. Knoll, Schularick & Steger, “No Price Like Home: Global House Prices, 1870–2012”, American Economic Review 107, 331 (2017).
10. Mian, Straub & Sufi, “Indebted Demand”, Quarterly Journal of Economics 136, 2243 (2021).
11. Mazzolari & Ragusa, “Spillovers from High-Skill Consumption to Low-Skill Labor Markets”, Review of Economics and Statistics 95, 74 (2013).
12. Costa, “The Wage and the Length of the Work Day: From the 1890s to 1991”, Journal of Labor Economics 18, 156 (2000).
13. Boskin, Dulberger, Gordon, Griliches & Jorgenson, Toward a More Accurate Measure of the Cost of Living, final report to the US Senate Finance Committee, December 1996; Nordhaus, “Do Real-Output and Real-Wage Measures Capture Reality? The History of Lighting Suggests Not”, in Bresnahan & Gordon (eds.), The Economics of New Goods, University of Chicago Press, 1997.
14. Warren & Tyagi, The Two-Income Trap: Why Middle-Class Parents Are Going Broke, Basic Books, 2003.
15. Graeber, Bullshit Jobs: A Theory, Simon & Schuster, 2018.
16. Soffia, Wood & Burchell, “Alienation Is Not ‘Bullshit’: An Empirical Critique of Graeber’s Theory of BS Jobs”, Work, Employment and Society 36, 816 (2022).
17. Dur & van Lent, “Socially Useless Jobs”, Industrial Relations 58, 3 (2019), using the International Social Survey Programme, 47 countries, about 100,000 workers.
18. Walo, “‘Bullshit’ After All? Why People Consider Their Jobs Socially Useless”, Work, Employment and Society 37, 1123 (2023).
19. Food Marketing Institute annual surveys of stock-keeping units per store, as reported in Bureau of Labor Statistics, “Checking out productivity in grocery stores”, Beyond the Numbers 8 (2019).
20. Ellen MacArthur Foundation, A New Textiles Economy: Redesigning Fashion’s Future, 2017, using Euromonitor apparel volumes 2000–2015.
21. United Nations Environment Programme, Food Waste Index Report 2024, March 2024.
22. Broda & Weinstein, “Globalization and the Gains from Variety”, Quarterly Journal of Economics 121, 541 (2006); Hausman, “Valuation of New Goods under Perfect and Imperfect Competition”, in Bresnahan & Gordon (eds.), The Economics of New Goods, University of Chicago Press, 1997.
23. Iyengar & Lepper, “When Choice is Demotivating: Can One Desire Too Much of a Good Thing?”, Journal of Personality and Social Psychology 79, 995 (2000); Schwartz, The Paradox of Choice, Ecco, 2004.
24. Scheibehenne, Greifeneder & Todd, “Can There Ever Be Too Many Options? A Meta-Analytic Review of Choice Overload”, Journal of Consumer Research 37, 409 (2010).
25. Galbraith, The Affluent Society, Houghton Mifflin, 1958, ch. 11, “The Dependence Effect”.
26. Bureau of Labor Statistics, Consumer Expenditures — 2023, news release, September 2024: shares of average annual expenditures of $77,280.
27. Foundational Economy Collective, Foundational Economy: The Infrastructure of Everyday Life, Manchester University Press, 2018.
28. Eurostat, national accounts employment data by industry, up to NACE A*64 (nama_10_a64_e), hours worked by total employment, 1995–2023 (France from 1978). Shares computed by scripts/keynes.py.
29. Eurostat, FIGARO inter-country input-output tables, industry by industry, 2026 edition (2010–2022 used); hours worked by industry from Eurostat (nama_10_a64_e) for European countries, and for other countries employment from the OECD’s Trade in Employment database (2025) times ILO modelled weekly hours; rest of the world from ILO modelled estimates. Hours outside Europe are therefore estimates, and in Portugal the household column includes tourists’ spending.
30. Jones & Klenow, “Beyond GDP? Welfare across Countries and Time”, American Economic Review 106, 2426 (2016).
31. UN World Population Prospects and Human Mortality Database, period life expectancy at birth, via Our World in Data. The peer average is the unweighted mean of the UK, France, Germany, Italy, Spain, the Netherlands, Sweden and Japan.
32. Case & Deaton, Deaths of Despair and the Future of Capitalism, Princeton University Press, 2020.
33. Easterlin, “Does Economic Growth Improve the Human Lot? Some Empirical Evidence”, in David & Reder (eds.), Nations and Households in Economic Growth, Academic Press, 1974; Stevenson & Wolfers, “Economic Growth and Subjective Well-Being: Reassessing the Easterlin Paradox”, Brookings Papers on Economic Activity 2008(1), 1; Killingsworth, “Experienced well-being rises with income, even above $75,000 per year”, PNAS 118, e2016976118 (2021).
34. Ramey, “Time Spent in Home Production in the Twentieth-Century United States: New Estimates from Old Data”, Journal of Economic History 69, 1 (2009), tables 6A and 7, prime-age (18–64) women and men.
35. Ramey & Francis, “A Century of Work and Leisure”, American Economic Journal: Macroeconomics 1(2), 189 (2009).
36. Aguiar & Hurst, “Measuring Trends in Leisure: The Allocation of Time over Five Decades”, Quarterly Journal of Economics 122, 969 (2007).
37. OECD Time Use database (OECD.WISE.INE, DSD_TIME_USE@DF_TIME_USE), minutes per day by sex, ages 15–64, latest survey per country (Portugal 1999).
38. CESIS, Os Usos do Tempo de Homens e de Mulheres em Portugal, policy brief of the Inquérito Nacional aos Usos do Tempo 2015 (questionnaire on the last working day, not a diary survey).
39. Bureau of Labor Statistics, American Time Use Survey — 2025 Results, news release, 25 June 2026.
40. Eurostat, How Europeans Spend Their Time: Everyday Life of Women and Men, 2004 edition (time-use surveys 1998–2002), tables 7.1–7.2.
41. Kivimäki et al., “Long working hours and risk of coronary heart disease and stroke: a systematic review and meta-analysis of published and unpublished data for 603 838 individuals”, The Lancet 386, 1739 (2015).
42. Pega et al., “Global, regional, and national burdens of ischemic heart disease and stroke attributable to exposure to long working hours for 194 countries, 2000–2016”, Environment International 154, 106595 (2021).
43. Pencavel, “The Productivity of Working Hours”, Economic Journal 125, 2052 (2015).
44. Jacobs & Gerson, The Time Divide: Work, Family, and Gender Inequality, Harvard University Press, 2004.
45. Hamermesh & Lee, “Stressed Out on Four Continents: Time Crunch or Yuppie Kvetch?”, Review of Economics and Statistics 89, 374 (2007).
46. US Census Bureau, American Community Survey, table S0801 (2024) and Census 2000 brief C2KBR-33 (1980); Eurostat, Labour Force Survey 2019 ad hoc module on work organisation, commuting time.
47. Altonji & Paxson, “Labor Supply Preferences, Hours Constraints, and Hours-Wage Trade-offs”, Journal of Labor Economics 6, 254 (1988); Reynolds, “You Can’t Always Get the Hours You Want: Mismatches between Actual and Preferred Work Hours in the U.S.”, Social Forces 81, 1171 (2003).
48. Cesarini, Lindqvist, Notowidigdo & Östling, “The Effect of Wealth on Individual and Household Labor Supply: Evidence from Swedish Lotteries”, American Economic Review 107, 3917 (2017); Picchio, Suetens & van Ours, “Labour Supply Effects of Winning a Lottery”, Economic Journal 128, 1700 (2018).
49. Winkelmann & Winkelmann, “Why Are the Unemployed So Unhappy? Evidence from Panel Data”, Economica 65, 1 (1998).
50. Nässén & Larsson, “Would shorter working time reduce greenhouse gas emissions? An analysis of time use and consumption in Swedish households”, Environment and Planning C 33, 726 (2015).
51. Knight, Rosa & Schor, “Could working less reduce pressures on the environment? A cross-national panel analysis of OECD countries, 1970–2007”, Global Environmental Change 23, 691 (2013).
52. Antal, Plank, Mokos & Wiedenhofer, “Is working less really good for the environment? A systematic review of the empirical evidence for resource use, greenhouse gas emissions and the ecological footprint”, Environmental Research Letters 16, 013002 (2021).