In 1920 the Cambridge economist Arthur Pigou set down a defect in the way nations count themselves. "If a man marries his housekeeper or his cook," he wrote in The Economics of Welfare, "the national dividend is diminished." He filed it under paradoxes. Nothing has happened to the house. The same floors are scrubbed, the same meals cooked, the same hours spent by the same person. The only thing that changed is the label on her and, with the label, the money. The country's measured output falls because a wage stopped being paid for work that never stopped being done.

A century on, the defect is in every set of national accounts on earth and it is not a defect. It is the definition. Work, for the purposes of counting it, ranking it, taxing it and arguing about it, is whatever we decide to call work and pay for. It is not a natural category. Nobody dug it out of the ground. Every generation has drawn the line between work and not-work somewhere different, always with a straight face, always as if the line were a discovery rather than a choice. That history matters now for a reason Pigou could not have seen coming. A machine is arriving that may take a great deal of what currently sits on the paid side of the line. Maybe not your job. Maybe. Either way the line is about to move and the only question worth asking is who holds the pen. Whose pen are you holding, who holds yours?

The line

The pen has changed hands many times and every holder was certain. The Physiocrats of eighteenth-century France held that the land alone produced anything. Quesnay's Tableau économique files farming, mining and fishing under "productive Expenditures" and handicrafts, housing, clothing, servants and commerce under "sterile Expenses". Adam Smith, a generation later, drew the line somewhere else entirely. His unproductive labourers included "the sovereign, for example, with all the officers both of justice and war who serve under him". Then "churchmen, lawyers, physicians, men of letters of all kinds; players, buffoons, musicians, opera-singers, opera-dancers", whose work "perishes in the very instant of its production". He allowed that such labour might be "honourable", "useful" or "necessary". It still produced nothing. Two founding schools of the discipline, within a generation of each other, disagreed about which half of the economy was idle.

The Soviet Union drew it differently again. Its Material Product System counted only the "material sphere" and, in the words of the United Nations handbook written to help the successor states abandon it, "regards the production of non-material services as non-productive activities". The doctor, the teacher and the bookkeeper produced nothing, by definition, until the 1990s. When Simon Kuznets built the first modern national income estimates for the United States Senate in 1934, he drew the line through the front door of every house in the country and said so. The services of housewives "must be imposing indeed" when totalled across thirty million families, he wrote. They were "large enough to affect materially any estimate of national income". They were left out anyway, because they belonged to "family life at large, rather than of the specifically business life of the nation" and because "no reliable basis is available for estimating their value". Two pages later comes the sentence everyone quotes and nobody acts on: "The welfare of a nation can, therefore, scarcely be inferred from a measurement of national income as defined above." The United Nations System of National Accounts inherited the exclusion and the world's statisticians have defended it since, most famously against Marilyn Waring, whose 1988 book If Women Counted documented what a production boundary drawn around the market does to women and to the natural world.

Each school looked at the same human activity and called part of it work and part of it nothing. The parts changed; the confidence somehow did not!

The line moves under your feet within a lifetime too, quietly. Nobody calls it the end of anything. An Irish worker put in around 3,108 hours a year in 1870 and 1,654 in 2023. Half the hours, incomparably more output. Anyone standing in 1870 would have called the 2023 arrangement the abolition of work. We call it Tuesday. Or take the largest single redrawing of the Irish line in living memory: the female labour force participation rate was 44.0% in 1998 and 61.3% in the second quarter of this year, with over 1.3 million women in employment, up almost 90% since 2000. The activity did not appear from nowhere. Much of it walked across the line from the unpaid side to the paid side and the moment it crossed, it started to count.

The line, in the State's own hand

Ireland wrote its definition into the founding document, in two sentences that should be read as a pair. Article 41.2.1: "the State recognises that by her life within the home, woman gives to the State a support without which the common good cannot be achieved." Article 41.2.2: "The State shall, therefore, endeavour to ensure that mothers shall not be obliged by economic necessity to engage in labour to the neglect of their duties in the home."

The first sentence says the work is indispensable. The second calls it "duties" and reserves the word "labour" for what happens outside the door. Indispensable, unpaid and by definition not work. That is the line, drawn in 1937 and still in force.

Still in force because in March 2024 the State asked to redraw it and the country declined. The Fortieth Amendment would have deleted 41.2 and inserted a new Article 42B: "The State recognises that the provision of care, by members of a family to one another by reason of the bonds that exist among them, gives to Society a support without which the common good cannot be achieved, and shall strive to support such provision." It was defeated by 74% to 26% on 8 March 2024. Voters had many reasons and this piece is not about them. Look only at what was on offer. The old sentence had the care propping up the State. The new one had it propping up Society and committed the State to strive. Readers of The Will will recognise the verb. Even the redrawn line kept care on the far side of work.

Count who is over there. Census 2022 found 299,128 people providing regular unpaid care, 6% of the population, up by more than half since 2016. Sixty-one per cent are women! Some 86,972 of them give 43 hours or more a week, which is a full-time job plus overtime by any employer's reckoning no matter the angle from which you might view such things. Fifty-seven per cent of carers were also at work, holding one job on the paid side of the line and another on the unpaid side.

What is it worth? The State has never produced a satellite account to find out would you believe? I have no trouble entertaining the thought I'll not lie to you, this is Ireland and so are we 'governed'! Its last diary-based time-use survey was taken in 2005: women did five hours and six minutes of unpaid work on an average day, men one hour and fifty-four. The only peer-reviewed valuation, Hanly and Sheerin in the Economic and Social Review, put Census-defined informal care at €2.1 billion to €5.5 billion a year on 2011 data, when there were a hundred thousand fewer carers. Family Carers Ireland puts the wider figure above €20 billion a year at a €20 replacement rate; Oxfam put women's unpaid care at €24 billion. That's worth repeating so I will; €24 billion! €24 billion!!

The spread is itself the finding. A State that has been measuring the price of land since 1973 and how to pass it on and protect it has never bothered to measure this at all. My entire life, not once!

What it pays instead tells you it knows. Carer's Allowance is, in the Department's own words, "a payment for people on low incomes who are caring for a person who needs full-time care and attention". Income support, means-tested, €270 a week. Not a wage. It comes with a condition: the carer must not be "employed, self-employed, doing voluntary work, training, or any education courses for a combined total of more than 18.5 hours a week". Please, sit with that rule for a moment!
The State knows the caring is full-time, because it rations the carer's remaining hours to less than half a working week. It knows the caring is work, because it caps the other work. It declines only the word and, with the word, crucially the wage. There were 107,900 people on the payment in July, at a cost of €1.48 billion this year across Carer's Allowance and Carer's Benefit together.

Now watch the same State draw the line the other way when it suits. Community Employment, Tús and the Rural Social Scheme had 19,641, 4,407 and 2,739 participants respectively in June, at a cost of over €565 million this year, their purpose in the Minister's words "an opportunity to engage in useful work within their communities". No market asked for that work. The State decided it was work, paid for it and so it is? The Basic Income for the Arts paid 2,000 artists €325 a week and the State's own evaluation found eleven more hours a week on their practice, three and a half fewer in other sectors, depression down fifteen points, anxiety down sixteen and €1.39 back for every euro. It has been renewed to 2029. This site has already argued what that result means for welfare design in A Decent Gap. The point here is narrower in that the State decided that making art is work, paid for it and got work. One last line, from the same rulebook: a professional artist on Jobseeker's Allowance is exempt from labour-market activation for the first year out of work. The carer gets 18.5 hours. Nobody found these lines in nature. Somebody drew them.

The machine

Which brings the virtual pen to the present with the forecasts first because these set the temperature so probably best. The IMF's January 2024 staff note: "almost 40 percent of global employment is exposed to AI", about 60 per cent in advanced economies and, of those, "about half may be negatively affected". Goldman Sachs, March 2023: the equivalent of 300 million full-time jobs exposed. Daron Acemoglu, May 2024, running the same technology through actual macroeconomics: "no more than a 0.66% increase in total factor productivity over 10 years". Between those numbers there is room for every prophecy on sale.

Then what has actually shown up, three years in. The Yale Budget Lab's tracker, updated in August on July data, "does not provide clear evidence of labor market disruption associated with AI"; churn across occupations remains flat or inside historical ranges. Danish administrative data gives "precise null effects on earnings and recorded hours", ruling out anything larger than 2 per cent two years after ChatGPT. The Dallas Fed worked out that if every young worker who failed to get into an exposed occupation had gone straight onto the unemployment rolls, the American unemployment rate would be about a tenth of a point higher. The aggregate, so far, is quiet.

The canaries are not. Brynjolfsson, Chandar and Chen at Stanford whose August revision runs to June, find that employment of 22 to 25 year olds in the most exposed occupations "now stands about 19% below where it would be if it had kept pace" with less-exposed peers. Ouch! The mechanism "operates primarily through reduced hiring of young workers rather than increased separations". They add, in the same breath, "we do not see widespread, economy-wide job displacement". Indeed's American postings show the shape: entry-level down 7.5 per cent year on year, senior up 14.7. In Britain graduate postings are the lowest for the time of year since 2020. Two things are true at once. The economy is not shedding jobs. The bottom rung of the ladder is being sawn through. The people who feel it first are the ones who have not yet been allowed to start.

Ireland is not just hurling from the ditches here, watching this from the stands. The Department of Finance's own analysis has 63 per cent of Irish employment exposed against an advanced-economy average of 60, with 30 per cent in occupations where the technology is likelier to substitute than assist. Women are in highly exposed roles at 76 per cent against 51 for men. By this February the Department could report the canary in Irish data: employment of 15 to 29 year olds in the ICT sector "has fallen by over 20 per cent" between 2023 and 2025 "while it grew by 12 per cent for prime-age workers". The Central Bank puts 32 per cent of employment in roles highly vulnerable to automation with limited scope for collaboration. The ESRI and the Department, in April, simulated a central scenario in which "around 7 per cent of current jobs could be displaced" and found that income inequality "rises moderately in every scenario". All of this against a labour market at 2,839,300 employed and 5.1 per cent unemployed in the second quarter, with youth unemployment at 13.2. Record employment and a sawn rung, both at once, in the same release. Square that circle!

Now the counter-frames, because the doom register has a track record and it really isn't good. Keynes, in 1930, named "a new disease of which some readers may not yet have heard the name", technological unemployment. He suggested that "a fifteen-hour week may put off the problem for a great while". In 1964 a committee including Linus Pauling and Gunnar Myrdal wrote to President Johnson that "the traditional link between jobs and incomes is being broken" and that the economy could now sustain all citizens "whether or not they engage in what is commonly reckoned as work". In 1983 Wassily Leontief, a Nobel laureate, told the National Academy of Engineering that the role of humans as the most important factor of production was "bound to diminish", "in the same way that the role of horses in agricultural production was first diminished and then eliminated by the introduction of tractors". Each was right about the technology and wrong about the timing, or right about the timing and wrong about what people would do with it. Economists have a name for the error of assuming a fixed quantity of work to be shared out. They call it the lump of labour, after an 1891 essay on why workmen dislike piece-work. They deploy the name to close conversations about working time as often as to open them. And there is the frame the Anglophone press mostly cannot see, which is that for most of the rich world the problem runs the other way. The United Nations counts 63 countries, holding 28 per cent of humanity, whose populations peaked before 2024: China, Germany, Japan and Russia among them. Japan's Recruit Works Institute projects a labour shortfall of over 11 million workers by 2040 and names "an extensive shift to mechanization and automation" as the first remedy. In half the world the machine is the housekeeper nobody can hire.

So let's be honest, the position on the employment number is that nobody actually has it. Anyone selling you one is selling something else which to be fair is not an uncommon practice. Two things are certain regardless! The first is that the bite starts at the bottom rung and with the young, maybe your young. It will be visible there long before any aggregate admits it. The second is that the distribution question does not wait for the employment question. Whoever owns the machine owns what it makes, it has always been thus! The share of corporate income paid to labour across the world fell from roughly 64 to 59 per cent between 1975 and 2012 and the ILO has it at 52.3 per cent and flat. Ireland's own share, on the ESRI's honest measure using modified national income has been stable since 1998, which is worth saying plainly because it does cut against the easy story. Stable is the pre-machine reading. The post-machine reading is whatever the pen decides. This site's standing axis applies with more force than usual. A tax system built on work is a tax system built on the thing being cut. Tax wealth, not work stops being a slogan and becomes arithmetic.

Upon review

Now the part that is about you! Maybe the machine takes your job. Maybe it takes the job you would have had at 23 and never tells you. Maybe it takes your neighbour's and you keep yours at a wage that stops moving. Whichever it is, a review is coming and for once you are the reviewer. Be that!

There are two ways to fail it. The first is to roll over: to accept that you were the job, that your worth was the wage and to go down with it. That is the outcome a century of counting has been training you for but the ultimate decision will remain yours once the grief passes or until it's arrival if you'd like to head it off atany given pass. The second is to fight the machine for the old line: to demand that whatever counted as work in 2019 be preserved as work, at any cost, forever. That was never even the Luddites' position. They were skilled cloth workers who broke the frames being used to cut their wages and hollow out their trade. Byron told the Lords in February 1812 that "one man performed the work of many, and the superfluous labourers were thrown out of employment". He added that "these men never destroyed their looms till they were become useless, worse than useless; till they were become actual impediments to their exertions in obtaining their daily bread". Parliament's answer was the Frame-Breaking Act, which made breaking a frame a hanging offence. The machine was never the enemy. The pen was.

The third answer is the one this piece has been circling. Redefine!! Not just yourself! Upon review, redefine what you count as valuable, which means going one level down and redefining the value of what you value. The oldest thinking on the subject already did this and we have spent two centuries reversing it. Aristotle, in the Ethics: "we are busy that we may have leisure, and make war that we may live in peace." The Greeks had no positive word for work at all. Their word was ascholiasxoli/a), not-leisure. Rome's word for business was negotium, not-otium, the same negative. We still negotiate with it. Hannah Arendt, in 1958, split the active life in three: labour, the body's endless round of necessity; work, the making of a durable world of things; action, "the only activity that goes on directly between men without the intermediary of things or matter". The accounts count the first two when a wage is attached and the third never. For most of human history work was the thing you did so that you could do the thing that mattered. We inverted it, then built the accounts to measure the inversion, then taught our children the accounts.

Btw, you already know what you value even if it's become a little obscured by the daily. The census counted 299,128 of you doing it for nothing. When twelve Irish companies ran a four-day week for six months in 2022, staff spent the fifth day on sleep, on hobbies and on "community and voluntary work". Nine of the twelve kept the schedule. When the State gave two thousand artists a wage, they made art. Shocker!! Nobody had to be told what to do with the time. When YouGov asked British workers in 2015, 37 per cent said their job made no meaningful contribution to the world, the number David Graeber built a book on. When three sociologists tested his theory on European survey data they found the share who felt their work was rarely or never useful was 4.8 per cent and falling. Read the two together and they say the same thing. Most people do not think their job is pointless. A great many think it is not the point.

The pen

If work is a decision, the advent of the machine is the moment to take it in public. Ireland has always taken it in private: in a 1937 clause, in a means-test disregard, in an 18.5-hour rule, in a €325 cheque to a painter. Five things follow and none of them requires believing any forecast.

Count it. A household satellite account and a time-use survey. The last diary was taken in 2005. You cannot move a line you refuse to measure and you cannot argue with a number that does not exist.

Pay for care as work. The 2024 wording failed. It did! Whatever the next wording, the test is whether the State "shall support" rather than strive. The second test is whether Carer's Allowance becomes a wage for work done rather than a means-tested apology with a cap on the carer's other work.

Use the template the State already built. The arts scheme is the State's own proof that paying for unpriced work produces it. The occupations its own tables mark as first exposed are the place to run the next one.

Tax wealth, not work. If the machine thins the link between work and income, the income tax base thins with it. The base that remains is the one Ireland taxes least and the one that will own the machine.

Watch the pen. The Oireachtas committee on artificial intelligence noted in its first report that "22 of 25 AI commitments in the Programme for Government are coming under an economic theme". Its second, in June, asked that "as much emphasis should be placed on issues such as the environment, human rights, worker's rights, women's rights, child protection and equality as on the economy". That is a committee noticing who is holding the pen. This site has already set out what the State has wagered on the machine paying off. The wager on what happens to work is the same wager from the other side of the table.

Pigou's housekeeper never stopped working. She married, vanished from the accounts and every economy since has been counting around her. The machine will not decide what work is. It cannot. It has no pen. We will decide, the way we always have, in constitutions, in disregards, in cheques to painters. The only novelty on offer is doing it on purpose.

Sources

The counting. A. C. Pigou, The Economics of Welfare, Part I, chapter III (Macmillan, 1920; 4th ed. 1932 text at econlib.org; 1920 scan at archive.org). François Quesnay, Tableau économique, 3rd ed. 1759, trans. Monroe, at marxists.org. Adam Smith, An Inquiry into the Nature and Causes of the Wealth of Nations, 1776, Book II, chapter III, at econlib.org. United Nations, Handbook of National Accounting: Use of the System of National Accounts in Economies in Transition, Series F No. 66, 1996, at unstats.un.org. National Income, 1929-32, Senate Document No. 124, 73rd Congress, 1934, estimates by Simon Kuznets, pp. 4-7, at fraser.stlouisfed.org. Saunders and Dalziel, "Twenty-Five Years of Counting for Nothing: Waring's Critique of National Accounts", Feminist Economics 23(2), 2017, at ideas.repec.org.

Hours and participation. Our World in Data, annual working hours per worker, Ireland, from Huberman and Minns (2005) and the Penn World Table, at ourworldindata.org. CSO, Statistical Yearbook of Ireland 2007, chapter 2, at cso.ie. CSO, Labour Force Survey Q2 2026, labour force and key findings, at cso.ie and cso.ie. CSO, "Employment trends in Ireland: what CSO data tells us", 31 August 2026, at cso.ie.

The Irish line. Constitution of Ireland, Articles 41 and 45, at irishstatutebook.ie. Fortieth Amendment of the Constitution (Care) Bill 2023, as initiated, at oireachtas.ie; result at referendum.ie. CSO, Census 2022 Profile 4, Carers, at cso.ie. McGinnity and Russell, Gender Inequalities in Time Use, ESRI and Equality Authority, 2008, at esri.ie. Hanly and Sheerin, "Valuing Informal Care in Ireland: Beyond the Traditional Production Boundary", Economic and Social Review 48(3), 2017, at esr.ie. Family Carers Ireland, Pre-Budget Submission 2026, at familycarers.ie. Oxfam Ireland, 31 January 2020, at oxfamireland.org. Department of Social Protection, Carer's Allowance, at gov.ie; means-test press release, 1 July 2026, at gov.ie. Dáil written answer, Minister for Social Protection to Tony McCormack TD, 18 June 2026, at oireachtas.ie. Citizens Information, Jobseeker's Allowance, at citizensinformation.ie. Basic Income for the Arts, Impact Assessment (2 years), September 2025, at gov.ie; cost-benefit press release, September 2025, at gov.ie; successor scheme, 10 February 2026, at gov.ie.

The machine. IMF Staff Discussion Note SDN/2024/001, Gen-AI: Artificial Intelligence and the Future of Work, January 2024, at imf.org. Goldman Sachs, "Generative AI could raise global GDP by 7%", 5 April 2023, at goldmansachs.com. Acemoglu, "The Simple Macroeconomics of AI", NBER Working Paper 32487, May 2024, at nber.org. Yale Budget Lab, "Tracking the Impact of AI on the Labor Market", updated 19 August 2026, at budgetlab.yale.edu. Humlum and Vestergaard, "Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI", NBER Working Paper 33777, revised March 2026, at nber.org. Atkinson and Yamco, Federal Reserve Bank of Dallas, 6 January 2026, at dallasfed.org. Brynjolfsson, Chandar and Chen, "Canaries in the Coal Mine", Stanford Digital Economy Lab, August 2026 update, at digitaleconomy.stanford.edu. Indeed Hiring Lab, "The Labor Market Is Tilting Toward Seniority", 23 July 2026, at hiringlab.indeed.com; UK mid-year report, 3 August 2026, at hiringlab.indeed.com. Department of Finance and Department of Enterprise, Artificial Intelligence: Friend or Foe?, 11 June 2024, at gov.ie and Box 5 at gov.ie. Department of Finance, Economic Insights Volume 1 2026, February 2026, at gov.ie. Yadav and McIndoe-Calder, "Digital Transformation: Challenges and Opportunities for the Irish Labour Market", Central Bank of Ireland Quarterly Bulletin Q1 2026, at centralbank.ie. Doorley, O'Connor, O'Shea and Tuda, Artificial intelligence and income inequality in Ireland, ESRI and Department of Finance, April 2026, at esri.ie. Keynes, "Economic Possibilities for our Grandchildren", 1930, at marxists.org. Ad Hoc Committee, The Triple Revolution, 1964, reprinted in International Socialist Review, Summer 1964, at marxists.org. Leontief, "National Perspective: The Definition of Problems and Opportunities", in The Long-Term Impact of Technology on Employment and Unemployment, National Academy Press, 1983, pp. 3-4, at nap.nationalacademies.org. Schloss, "Why Working-Men Dislike Piece-Work", Economic Review, April 1891, at timeworkweb.com. United Nations, World Population Prospects 2024: Summary of Results, at un.org. Recruit Works Institute, Future Predictions 2040, 2023, at works-i.com. Karabarbounis and Neiman, "The Global Decline of the Labor Share", NBER Working Paper 19136, 2013, at nber.org. ILO, World Employment and Social Outlook, September 2024 Update, at ilo.org. O'Shea, "Estimating Ireland's Labour Share", ESRI Research Note, December 2024, at esri.ie.

Upon review. Byron, House of Lords, Frame Work Bill, 27 February 1812, at api.parliament.uk. Destruction of Stocking Frames, etc. Act 1812, at wikipedia.org. Merchant, Blood in the Machine, 2023, author's introduction at bloodinthemachine.com. Aristotle, Nicomachean Ethics X.7, trans. Ross, at classics.mit.edu. Lewis and Short, A Latin Dictionary, negotium, at perseus.tufts.edu; Liddell, Scott and Jones, ascholia, at perseus.tufts.edusxoli/a). Pieper, Leisure: The Basis of Culture, 1948, trans. Malsbary 1998, at archive.org. Arendt, The Human Condition, 1958, chapter I, at archive.org. YouGov, "37% of British workers think their jobs are meaningless", 12 August 2015, at yougov.co.uk. Soffia, Wood and Burchell, "Alienation Is Not 'Bullshit': An Empirical Critique of Graeber's Theory of BS Jobs", Work, Employment and Society 36(5), 2022, at archive.org. UCD, "Four-day week trials brought 'significant improvements' to Irish companies and workers", 1 December 2022, at ucd.ie.

The pen. Joint Committee on Artificial Intelligence, First Interim Report, 16 December 2025, at oireachtas.ie; Second Interim Report, 9 June 2026, at oireachtas.ie. On this site: The Will, A Decent Gap, The Wealth Tax Ireland Already Has, The Future Ireland Already Spent.

Note on AI use: I used an AI system as a research and drafting assistant on this piece. I directed the work. I checked every factual claim in it against the primary sources listed above. I decided what went in and what stayed out. The judgements here are mine. So are the errors. So is the responsibility for publishing it. This disclosure is made in line with Article 50(4) of the EU AI Act, whose transparency obligations have applied since 2 August 2026.