Go back to your sister from three chapters ago, the one whose pay pulled away from the men beside her the year her daughter was born. Follow the care itself this time rather than the paycheck. In the months after the birth she did the feeding, the night waking, the doctor’s visits, the endless low-grade management of a small life, and none of it registered anywhere in the national accounts. Had she instead gone back to her desk and paid a nanny to do the same feeding and waking and ferrying, the nanny’s wages would have entered the economy’s official size, counted as output, added to the number the news reports each quarter as how the country is doing. Same tasks, same hours, same child. The only thing that decides whether the work shows up in the economy’s headline measure is whether a payment was made for it. This chapter is about that line: where it sits, why it was drawn there, how much sits on the far side of it, and what follows from a gauge that cannot see the largest kind of work most people do.
The question is not a trick, and the answer is not that the accountants were careless. The line is deliberate, old, and defended, and there are real reasons a national statistician draws it where they do. But a line drawn for good reasons still has consequences, and the one this chapter is about is a habit of mind: because the unpaid work is invisible to the gauge, it becomes easy to treat as costless, and a thing treated as costless is easy for policy to pass over.
The line where money changes hands
Gross domestic product is a measure of production, and the rulebook that defines it, the international System of National Accounts, has to decide what counts as production. It cannot simply count all useful activity, because almost everything a person does is useful in some sense, and a measure with no boundary measures nothing. So the accounts draw what they call a production boundary, and for household work they draw it at the market. Services that a household produces and consumes itself, cooking its own dinners, cleaning its own rooms, minding its own children, sit outside the boundary; the identical services, bought from someone else, sit inside it. How the gauge is built and what it is for is the subject of the chapter on measuring whether a country is doing well, which owns that machinery; the point here is only the edge of it, the seam where the same task changes status depending on who is paid.
A familiar illustration is an old economists’ puzzle. A man who employs a housekeeper is paying for cleaning and cooking that count in GDP. If he marries her, and she goes on doing exactly the same work now as a spouse rather than an employee, the payment stops, and measured GDP falls, even though not one room is dirtier and not one meal is missed. Nothing real has changed except the presence of a wage, and the wage is the only thing the boundary can see. The boundary is not tracking whether work is done or whether it has value. It is tracking whether the work passed through a market, because a market transaction leaves a price behind, and a price is the thing national accounts are built to add up. A production measure needs prices it can observe, which is the accounts’ stated reason for drawing the line at the market, and it is also exactly why the measure goes blind at the household door.
How much sits on the far side
If the unpaid work were small, the blind spot would be a curiosity. It is not small. The United States Bureau of Economic Analysis keeps a satellite account, a set of books it runs alongside the official ones precisely to size what the official ones leave out, and in its newest estimates the value of household production, the cooking and cleaning and caring that households do for themselves, came to about 5.7 trillion dollars in 2024. Set against the extended measure of output that includes it, that is roughly 16 percent; laid on top of conventional GDP, it would enlarge the figure by something close to a fifth. This is not a fringe estimate from an advocacy group. It is the national statistical agency’s own accounting of the work its headline number omits.
Two cautions have to travel with a figure that large, because the number is more fragile than its size suggests. The first is that it depends heavily on the wage you choose to value the hours at, and the choice is not obvious. The BEA prices an hour of unpaid household work at the wage of a general-purpose domestic worker, a housekeeper, and it states in its own methodology that this is meant as a reasonable lower bound. The alternative it discusses is to price each task at the wage of the specialist who would do it in the market, a chef for the cooking, a driver for the ferrying, a nurse for the nursing, which would raise the total considerably. A different tradition prices the hour instead at what the person doing it gives up to do it, their own foregone wage, which for a high earner can be higher still. The BEA does not publish that opportunity-cost version at all, so any single dollar figure for unpaid work is a figure at a chosen wage, and the choices span a wide range. The second caution is that the number has just moved. The agency changed its method in 2026, which lowered the headline share by a few points from the roughly quarter-of-GDP figure that had circulated for years, so the old percentage and the new dollar levels belong to different vintages and should not be mixed. What survives both cautions is the shape of the thing: even valued at a deliberate lower bound, the uncounted work runs to trillions.
The second question a large number invites is who does it, and here the gender split is measured, not asserted. The American Time Use Survey asks a large sample of people to account for their day hour by hour, and the figure below sets the paid work and the unpaid work side by side for men and women, then attaches the lower-bound replacement wage so the uncounted hours can be read in dollars as well as time.
The picture is the same one the pay-gap chapter arrived at from the other direction. There the earnings of mothers and fathers split at the first birth and never rejoined, and much of the gap ran through hours and the jobs that bend around a home. This is the labour on the other side of that split, the hours a career gives up, seen directly and counted in time rather than inferred from a wage. A woman doing more total work than a man while less of it registers in the economy’s measure is the population-wide counterpart of a mother’s pay falling behind; the averages here cover everyone, parent or not, so the two are related readings rather than one fact measured twice. The two chapters describe one arrangement: care is real work, it falls unequally, and the part of it that is done for love rather than wages is the part the gauge was built not to see.
What the gauge cannot see, policy tends not to fund
None of this is an argument that GDP is measuring the wrong thing. As a gauge of market production, the boundary is doing its job, and the people who built it were not confused about what they left out; the satellite account exists precisely because they knew. The trouble lies not in the measure; it lies in how the measure gets used. A headline number, repeated every quarter and treated as the score, comes to stand in for the economy itself, and what it omits comes to feel like it does not exist. Unpaid care is the largest thing it omits, and a resource that never registers in the score is easy to treat as free: something the economy can draw on without limit and without cost, because no line item ever falls when it is drawn on. That is an inference about attention rather than a measured effect, offered as this chapter’s own reading of where the boundary’s consequences run. Childcare policy, family leave, the support of people caring for aging parents, all of it concerns work that the primary gauge scores at zero, and a cost the gauge scores at zero is easier to defer as a soft concern than as the multi-trillion-dollar activity the satellite account says it is. The boundary was drawn to size market production, and asking it to double as a verdict on what work is worth supporting asks a ruler to do a job it was never cut for.
A stay-at-home parent doesn’t contribute to the economy.
Backwards Moderate confidence
This one is not merely overstated; it points the wrong way, and it does so by mistaking a measuring instrument for the thing it measures. The claim survives only on a confusion between the economy and the particular gauge, GDP, that scores part of it. It is true, and only trivially so, that a stay-at-home parent’s work does not enter GDP, because GDP’s production boundary counts market transactions and there is no wage for a parent’s own care. But that is a fact about where a line was drawn, not about whether the work is economic or whether it contributes. The same care bought from a nanny would count in full, and the work is identical on both sides of the line, so it is real work producing the same services either way, and only the payment differs. Measured directly, that work is large. Even averaged over all women, employed or not, the unpaid hours run to several a day and are worth on the order of twenty thousand dollars a year at the lowest defensible wage; a parent at home full time does more of those hours, and a realistic wage prices them higher. In aggregate, the unpaid household and care work the economy runs on comes to trillions. A claim that this amounts to no contribution has the direction of the truth reversed. The correction is not that the parent contributes a little, or contributes in some non-economic way; it is that they perform a large quantity of real economic work that the standard gauge was built not to record. What caps the confidence at moderate rather than high is that “the economy” is a definitional term whose edge is genuinely contested, and that the dollar size of the contribution depends on a valuation method that ranges across a wide band. The existence and direction of the contribution are not in doubt; its exact magnitude is.
Sources
- The boundary: the System of National Accounts places own-account household services outside the production boundary, so the same care work counts in GDP when purchased and not when done unpaid by a household member. The mechanics of the gauge are treated in Countries, on measuring whether a country is doing well, which this chapter cites rather than restates.
- The magnitude: US Bureau of Economic Analysis, household-production satellite account (Bridgman, BEA Working Paper 2026-15) — household production value added was about 5.7 trillion dollars in 2024, some 16 percent of an extended GDP that includes it, or close to a fifth on top of conventional GDP. The valuation uses the wage of a general-purpose domestic worker, which the BEA’s own methodology (Survey of Current Business, 2012 and 2022) calls a reasonable lower bound; specialist-cost and opportunity-cost methods run higher, and the BEA publishes no opportunity-cost variant. The 2026 method revision lowered the share from the earlier roughly one-quarter figure, so vintages should not be mixed.
- The hours and their split: BLS American Time Use Survey, 2025 annual averages, Table 1 — women average about 3.9 hours a day of household and care work to men’s 2.7, and about 2.8 hours of work and work-related activity to men’s 3.8, so women average more work in total while less of it is counted. Valued at the May 2025 median housekeeper wage of $17.07 an hour, the uncounted hours come to roughly 24,400 dollars a year for women and 16,800 for men. The survey records primary activities only, so care is understated. The child-penalty half of this story, where a mother’s measured earnings fall at the first birth, is in the chapter on the male–female pay gap.
- Confidence is moderate under the rubric, which scores the weaker of evidence directness and construct match. Evidence directness is strong: the hours are measured from time-use diaries and the aggregate from the national accountant’s own satellite books. The binding, weaker leg is construct match. “Contribute to the economy” turns on what counts as “the economy,” a definitional edge the whole chapter is about, and the dollar value of the contribution depends on a replacement-cost method that is a stated lower bound with higher alternatives. The direction is not in question, which is why the ruling is Backwards and not merely Oversimplified; the size is, which is why the confidence is moderate.
The edge of work
This chapter marks a turn in the volume. The questions before it were all about work that is paid, and how the pay is set, divided, and disputed. This one steps to the edge of paid work and looks at the vast unpaid remainder that surrounds it, holds it up, and makes it possible, and that the economy measures at nothing. The care your own upbringing was built from sat on the uncounted side of that line, as most care does, and its invisibility in the accounts is a feature of a particular ruler, not a statement about its worth. Keeping the ruler and its blind spot both in view is the whole task: GDP is a good measure of what it measures, and a poor definition of what matters, and the distance between those two is where a great deal of real work lives.
Having reached the edge of what counts as work, the volume turns next to the shape of the paid work that does count, and asks whether even that is as fixed as it feels. The forty-hour week arrives looking like a law of nature and turns out to be a settlement that has moved before and could move again.