Where We Left Off
Before we begin
Parts Six and Seven built each side’s best case and scored it. Part Six found that the strongest traditionalist arguments were about household stability, partnership, male behaviour and women’s economic security — and that not one of them reached the rules they were supposed to defend. Part Seven found six feminist claims supported, four contested and five failed, and that the strongest supported claim of all points away from the metric the movement’s institutions use.
Both parts scored claims. This one counts results.
Sixty years have passed since the legal architecture began coming down in most of the world. Long enough for the effects to be visible, and short enough that most of the people who made the arguments are still alive to see them. What actually happened?
The answer has four parts and they do not point the same way.
Some changes are enormous and went further than anybody predicted. The reversal in education is the clearest: women now out-graduate men at every level across most of the developed world, and the gap is still widening. Nobody forecast that. Nobody campaigned for it. It happened anyway.
Some changes stalled halfway. Earnings narrowed sharply for two decades and then largely stopped narrowing. Employment rose and then plateaued. Part Seven identified the mechanism and this part measures what is left of it.
The largest gain of all is medical, and it is almost never counted on this ledger at all. Chapter Five is about it and it dwarfs everything in the political column.
And one measure went the wrong way. Across the decades of greatest gain, women’s own reports of their happiness fell. That finding is real, it has been checked, nobody predicted it, and Chapter Six goes through every explanation on offer and reports that none is established.
How to Read the Boxes
The notation, in case this is where you started
Six kinds of box run through the series. One live example of each.
A ledger is a record with two columns: what came in and what went out. The word is chosen deliberately for this part.
Almost all writing about this subject reports one column. Advocates list gains; critics list costs. Both are accurate and neither is a ledger, because a ledger requires that the same person count both sides using the same units.
Why it matters here: this part attempts both columns. Chapter Seven is the second one, and a reader who finds it uncomfortable after six chapters of the first should notice that the discomfort is what a ledger is for.
A Word Box appears the first time a hard word does — never later, never only in a glossary.
One number, before any argument, because it is the largest thing on the ledger and it is almost never on it.
In 1990, for every hundred thousand babies born alive in India, somewhere in the region of five hundred and fifty women died in the process of giving birth.
By the end of the 2010s that figure was under a hundred.
Put it in a district. Take a town where two thousand children are born each year. Three decades ago, that town buried around eleven women a year for having them. Now it buries about two.
Nine women a year, in one town, every year, who used to die and now do not.
No change in employment, earnings, representation or law comes close to that number, and it appears on almost no list of what changed for women. Chapter Five asks why not.
A figure that large should be checked before it is used, and the checking is the job of the next box.
A ledger is only as good as its measurements, and the ones in this part vary enormously in quality. It is worth knowing the ranking before the numbers start.
Strongest: administrative counts. Degrees awarded, births registered, deaths recorded, seats in a legislature. These are counted because an institution needed to count them, and nobody is reporting a feeling.
Strong: large national surveys with stable instruments over time — labour force surveys, time use surveys, health surveys.
Weaker: anything measured by asking people to rate something on a scale. Chapter Six rests entirely on this and its limitations are the reason that chapter cannot be resolved.
And a limitation running through all of it: the further back you go and the poorer the country, the worse the data. Several figures for India in the 1950s and 1960s are estimates rather than counts, and the improvement in the numbers partly reflects improvement in the counting.
With the measurements ranked, the argument that governs how to read every one of them can be stated.
The question that governs every number in this part.
The legal barriers were removed by campaigns that took decades and met fierce resistance. Property, franchise, divorce, employment, credit, education — none of these fell on their own, and in each case there is a documented record of who pushed and who opposed. Attributing the result to impersonal forces erases the people who did the work.
The washing machine, the contraceptive pill, antibiotics, the shift from manual to service employment, and the collapse in child mortality would each have transformed women’s lives with no politics at all. Countries with no feminist movement worth the name show many of the same trends. The movement may have arrived alongside the change rather than causing it.
And underneath that argument sits something both sides need and neither examines.
Both sides above assume that there is a baseline — a year to count from, against which everything is measured.
Almost everybody uses the 1960s. It is a strange choice, and it is strange in a way that flatters one side.
Because the mid-twentieth century was not a stable ancient condition. Part Three found that Indian women’s position on several measures had deteriorated across the colonial period, and Part Seven found a government committee reporting in 1974 that things had got worse since independence. Measuring from a trough produces a rising line.
Choose 1900 and the slope changes. Choose 1800 and it changes again. Choose the eleventh century, when Part Three found publicly carved erotic temples and Part Two found property passing through women in Kerala, and the shape of the graph is not obvious at all.
The general form: choosing a baseline that produces the desired slope. Every ledger in every field has this problem and almost none of them state which year they started from and why.
This is the signature box. There are five in this part.
Every chapter closes with one of these, restating it in the plainest words available, key terms in bold.
Read only these and you should still finish holding the whole argument.
A Note on This Part
Read this before Chapter One
This part is mostly numbers and it is the least argumentative in the series. That is deliberate. Six parts of argument have established what each side claims; this one asks what happened. Where a figure is disputed I have given a range, and where a trend is contested I have said so rather than picking the version that fits.
Chapter Seven is the costs column and it exists because a ledger requires one. A reader who has found the previous chapters congenial may find it unwelcome. It contains no claim that the project was not worth it — that is a value judgement and Chapter Nine declines to make it for you.
And a warning about Chapter Six. The declining happiness finding is quoted constantly by people who have not read what it does and does not establish. It does not establish that women were made worse off. It establishes something narrower and stranger, and both camps consistently overstate it in opposite directions.
1How To Read A Ledger
Three questions have to be answered before a single number means anything. Compared to what. Caused by what. And for whom. Almost every published account of what changed for women skips at least two of them.
1.1 — Compared to what
The front matter set this out and it needs saying once more in the chapter, because it governs everything that follows.
A number is a change only against a baseline, and the baseline is chosen. Nearly everybody writing about this counts from somewhere in the middle of the twentieth century, without saying why, and the choice is not innocent.
The mid-twentieth century was in several respects a low point rather than a stable condition. Part Three found that Indian women’s position on several measures had deteriorated across the colonial period. Part Seven found a committee of the Indian government reporting in 1974 that things had got worse since independence on the sex ratio, on women’s paid work and on political participation.
Measuring from a trough produces a rising line. That is arithmetic rather than achievement, and a ledger that does not state its starting year is not reporting a result. It is reporting a choice.
What this part does about it: where the starting point matters to the conclusion, I give it, and where a longer baseline changes the picture I say so. Chapter Five is the clearest case — the medical gains are enormous from any baseline you choose, which is one reason they are the most solid entry on the whole ledger.
1.2 — Caused by what
The second question is harder and it is where both camps do their most convenient reasoning.
Consider what changed in the world between 1950 and today, entirely independently of anybody’s politics.
Antibiotics became general. Childbirth stopped being dangerous. Child mortality collapsed, so a woman no longer needed six pregnancies to raise three children. Reliable contraception was invented. Piped water, gas cooking, refrigeration and washing machines arrived in ordinary houses. And the economy shifted from work requiring physical strength — where Part Four found the largest measured sex difference — to work requiring literacy, where Part Four found essentially none.
Each of those changes women’s lives enormously. None of them requires a movement.
Attribution: the problem of deciding what caused an observed change, when several things were happening at once and none can be switched off to check.
It is the central difficulty in evaluating any social movement, and it is not solvable in general. What can be done is to look for places where the causes separate — a country that got the technology and not the movement, or a law that changed on a particular date while everything else stayed still.
Why it matters here: both camps attribute confidently and in opposite directions. Supporters credit the movement for the century’s work. Opponents credit the century for the movement’s work. Neither has the evidence for the confidence, and the honest answer is a split that nobody has calculated.
There is one useful complication, and it cuts against the simple technological story.
Household technology was expected to reduce women’s domestic hours dramatically. Careful historical work found something different: as machines arrived, standards rose. Clothes that had been washed weekly were washed daily. Floors that had been swept were now expected to be clean. The hours did not fall as predicted, because the work expanded to fill the capacity.
Which is a warning against the tidy version of either account. Technology changes what is possible. It does not decide what is expected, and what is expected is social.
1.3 — For whom
The third question is the one this part takes most seriously, because it is where the honest answer is least comfortable.
“Women gained” is a sentence about an average, and Part Four established what an average conceals. The gains of the last sixty years were not distributed evenly. They went overwhelmingly to women with education, in cities, in rich countries, in advantaged castes and classes — and Chapter Eight sets out how stark the pattern is.
For a substantial number of women the ledger has barely moved. For some, in specific respects, it has moved backwards.
Any account reporting a single figure for what changed for women is averaging across those groups, and Part Four, Chapter One disposed of what an average can tell you about a person.
Three questions before any number means anything. Compared to what — the mid-twentieth century was a trough on several measures, and measuring from a trough produces a rising line by arithmetic rather than achievement.
Caused by what — antibiotics, safe childbirth, collapsing child mortality, contraception, piped water and the shift from physical to literate work each transformed women’s lives and none required a movement. Supporters credit the movement for the century’s work; opponents credit the century for the movement’s. Neither has the evidence.
With one complication against the tidy technological story: household machines did not cut domestic hours as predicted, because standards rose to fill the capacity. Technology changes what is possible and does not decide what is expected.
And for whom. The gains went overwhelmingly to educated women in cities in rich countries. For a substantial number the ledger has barely moved. Any single figure for “what changed for women” is averaging across those groups.
2The Reversal
The largest single change of the period was not planned, not campaigned for, and is barely discussed. Women now out-graduate men at every level across most of the developed world, and the gap is still widening.
2.1 — The crossover
In the United States, women passed men in the award of first degrees in the early 1980s. They have not looked back. Women now take somewhere around three in five bachelor’s degrees, a larger share of master’s degrees, and a majority of doctorates.
The pattern is not American. Women are a majority of university graduates in nearly every country in Europe, across most of Latin America, and in much of East Asia.
Put a graduating class of a hundred on a stage in three different decades.
1960: roughly sixty-five men and thirty-five women.
1982: fifty and fifty. The crossover.
Today: roughly forty men and sixty women.
Now the fact that makes this chapter necessary. The gap today, favouring women, is larger than the gap favouring men was in the early 1970s — the gap that produced a generation of campaigning, legislation and public concern.
The first gap was a national issue with a name. The second is roughly the same size, in the other direction, and most people reading this have never seen it stated as a number.
Before going further it is worth pausing on how these particular figures are obtained, because they are the most reliable in the series and the reason is instructive.
This is the most reliable measurement in the entire series and it is worth saying why.
What the evidence is: universities record every degree they award, with the recipient’s sex, because they are required to. Governments compile the totals. The series run continuously for decades in every country with a functioning education ministry.
Why that is unusually strong: nobody is sampled, nobody is asked how they feel, and no definition is in dispute. A degree was awarded or it was not. There is no reporting bias, no recall problem, no scale that drifts.
What it still cannot show: what the degrees are worth. A count of qualifications is not a count of skills, and if the same credential means something different than it did in 1970, a stable-looking series can conceal a changing thing. That criticism applies to both sexes equally, so it does not touch the comparison — which is what this chapter is about.
2.2 — And India
The Indian version arrived later and moved faster, and it is not widely known even in India.
Female enrolment in Indian higher education has risen steeply for three decades. Women are now around half of all students in higher education, and on the standard measure of what share of the relevant age group is enrolled, women have drawn level with men and in recent years edged ahead.
Part Five added the detail that makes this remarkable: women are around forty-three per cent of Indian STEM graduates, ahead of the United States, Britain, France and Germany.
So India has produced, within one generation, an educational transformation of women that most of its public argument has not registered — while, as Part Five established, employing almost none of the result.
Everybody in this argument, for two centuries, assumed that education was the bottleneck — that once women were educated, the rest would follow.
It was a reasonable assumption. Education was genuinely blocked, unblocking it was genuinely hard, and it is genuinely a precondition for almost everything else.
The assumption has now been tested, and it failed.
Women out-graduate men across the developed world and the earnings gap stalled decades ago. India produces a higher share of female science graduates than Britain and employs a fraction of them. The education arrived. The rest did not follow.
The general form: mistaking a necessary condition for the binding one. A thing can be genuinely required and still not be what is holding the outcome back — and a movement organised around removing it will keep removing it long after it has stopped being the constraint, because that is the machinery it has.
Notice which chapter this points at. If education is no longer binding, the binding constraint is in Chapter Three, and Chapter Three is about what happens after a first child.
2.3 — The boys nobody was watching
A reversal has two sides and the other one is now the larger problem.
Part Four established that girls outperform boys in reading in every country where it has been measured, in every cycle, and that the writing gap is around half an effect size — roughly ten times the mathematics gap and largely absent from public discussion.
Downstream of that: boys are more likely to be excluded from school, more likely to leave without qualifications, less likely to apply to university, and less likely to complete a course once enrolled.
This is not a small effect at the margins. It is the education system producing, at scale, a growing population of young men with no credential in an economy that increasingly requires one — and Part Six, Chapter Six established what happens to the earnings, offending, drinking and mortality of men who do not form stable attachments, while Part Five established that women’s partner preferences track education.
Those findings interlock. Fewer credentialed men, in a marriage market where credentials matter, produces both the unmatched women of Part Six, Chapter Five and the unattached men of Part Six, Chapter Six. It is the same arithmetic seen from two ends.
A live dispute with several serious candidates and no settled answer.
Boys mature later on the specific capacities schooling rewards — sitting still, sustained attention, self-regulation, organised written work. A system that assesses continuously from age five, on exactly those capacities, will systematically disadvantage the later-developing sex. The remedy on this account is structural: later school entry for boys, or assessment that does not compound early differences.
Boys respond to incentives. When manufacturing and manual employment offered a decent living without a degree, staying in school had a visible cost. Those jobs went, and the incentive to be educated rose — but the belief that a man can earn without qualifications lags the labour market by a generation, and the boys following it are following advice that used to be correct.
Part Six, Chapter Three found that the effects of family instability fall harder on boys, and family instability rose most exactly where educational outcomes fell most. On this account the education gap is a symptom and the household is the cause.
Boys’ absolute attainment has risen throughout. What changed is that girls’ rose faster, because girls were starting from an artificially suppressed position and released. There is no boy crisis — there is a girl recovery, and describing it as a crisis is a way of resenting it.
What would settle it: the school-entry-age question is directly testable, and where it has been examined the results are suggestive rather than conclusive. The economic-returns account is testable against regional labour market shocks and this has been done partially.
Why people care so much: because the answer determines whether this is somebody’s problem to fix. If it is development, schools must change. If it is returns, nothing needs fixing and the market will do it. If it is family structure, it belongs to Part Six. And Part Fifteen counts what happens to these men if none of it is fixed.
What is not disputed is the direction, and the direction has now run for four decades.
Women passed men in first degrees in the United States in the early 1980s and now take around three in five, plus a majority of master’s and doctoral degrees. The pattern holds across nearly all of Europe, most of Latin America and much of East Asia.
The gap today favouring women is larger than the gap favouring men was in the early 1970s — the gap that produced a generation of campaigning and legislation. Most readers have never seen the second one stated as a number.
India’s version arrived later and moved faster. Women are now around half of higher education enrolment, have drawn level or ahead on enrolment ratios, and are around 43 per cent of STEM graduates — while, as Part Five found, being employed at a fraction of that rate.
Everybody assumed education was the bottleneck. That assumption has now been tested and it failed. The education arrived and the rest did not follow. A necessary condition is not the same as the binding one, and a movement organised around removing it will keep removing it long after it stops being the constraint.
And the other side of the reversal is now the larger problem. Boys read worse in every country measured, are excluded more, complete less and apply less — producing men without credentials in a market that requires them, and in a partner market where Part Five found credentials matter. The unmatched women of Part Six and the unattached men of Part Six are the same arithmetic seen from two ends.
3Work
Employment rose enormously and then stopped. Earnings narrowed sharply and then stopped. Both plateaus arrived around the same time and for the same reason, and one profession accidentally demonstrated what would fix it.
3.1 — The rise, and where it stopped
Across the developed world, the share of women in paid work rose steeply from the middle of the twentieth century. In the United States it went from around a third of working-age women to around three fifths.
Labour force participation rate: the share of working-age people who are either in paid work or looking for it.
Two things it excludes are worth holding. Somebody working unpaid on a family farm or in a family business may or may not be counted, depending on how the survey is written — which is the heart of the Indian dispute in Chapter Ten. And somebody who has stopped looking because there is nothing to find drops out of the numerator entirely, so the rate can improve when conditions worsen.
Why it matters here: this is the headline measure of whether women are working, and it is considerably less straightforward than it looks.
Then, around the turn of the century, it stopped. American women’s participation peaked and has drifted slightly downwards since. Several European countries continued rising, some steeply, and the Nordic countries reached the highest levels recorded anywhere.
The earnings figures show the same shape. The raw gap in the United States narrowed from women earning around sixty per cent of men’s median to around three quarters by 1990 — a very large change in twenty years. Since then it has moved to somewhere in the low eighties and has been close to flat for a decade.
Two decades of rapid convergence, then three of very little. That is the shape of this chapter and every explanation has to account for it.
3.2 — What the remaining gap is made of
Part Seven scored the “same work” framing as failed and identified the mechanism. Here is the arithmetic.
Take the raw gap and remove hours worked, occupation, industry and years of experience. What is left is small — commonly estimated at somewhere between five and eight per cent — and it is not zero.
But the removed portions are not innocent, and Part Seven’s evidence box on this technique said why: if women are steered into lower-paying fields, the method files that under “occupation” and thereby removes it from view. What the decomposition establishes is that the gap does not operate through unequal pay for identical work. It does not establish that the rest is fair.
And underneath all of the removed portions sits one event.
Everybody calls it the gender pay gap, and the name is wrong.
Look at what happens when the data is split by parenthood rather than by sex. In several countries where this has been done carefully, the earnings of women without children track men’s closely — far more closely than they track the earnings of mothers. The gap between mothers and childless women is larger than the gap between childless women and men.
Which means the phenomenon is not primarily a gap between men and women. It is a gap between parents who absorb the interruption and everybody else, and it appears as a gender gap because of who does the absorbing.
The general form: naming a phenomenon after the wrong variable. And the naming has consequences, because a thing called a gender gap attracts gender remedies — representation targets, unconscious bias training, pipeline programmes — none of which touch the mechanism. Part Seven found that the movement’s strongest finding is invisible to its own metrics. This is why.
Notice also who is missing from the conversation entirely. If the penalty is for absorbing the interruption, then a father who absorbs it should show the same pattern. Where this has been measured, he does. Almost nobody measures it, because the question is not asked.
3.3 — And the same shape again
One further measure moves in exactly the pattern of the other two, and it is worth adding because three measures with the same shape are more informative than one.
Ask how separated men’s and women’s occupations are — not what they earn, simply whether they are doing the same jobs. That separation fell substantially through the 1970s and 1980s as women entered fields that had been closed, and then it largely stopped falling.
Part Five found the end state of that process: the Nordic countries, with the strongest equality institutions anywhere, are among the most occupationally separated economies in the developed world, with very high female employment concentrated in public health, care and education.
So the same shape appears three times. Employment: steep rise, then plateau. Earnings: rapid convergence, then stall. Occupational separation: sharp fall, then flat.
Three measures stopping at roughly the same time is unlikely to be three coincidences. Whatever the binding constraint became around 1990, it was the same constraint for all three — and §3.4 is the best available account of what it is.
3.4 — Greedy work
The most useful account of why the convergence stalled comes from an economic historian who was awarded the Nobel prize for it, and it is specific enough to be actionable.
Greedy work: a job that pays disproportionately more for long, inflexible and unpredictable hours. Working sixty hours earns not one and a half times a forty-hour salary but two or three times it.
The opposite is substitutable work, where one qualified person can hand a task to another without loss, so hours can be arranged and nobody is irreplaceable at short notice.
Why it matters here: where work is greedy, a couple maximises household income by having one person take all of it and the other absorb everything else. That is not discrimination and it is not a preference about careers. It is arithmetic, performed by the household, and it produces the specialisation Part Six described.
The account explains the stall precisely. The gap closed rapidly while the barriers were the binding constraint. Once they were gone, what remained was the structure of high-paying work itself — and no amount of legal equality touches that, because there is nothing illegal about a job that pays more for being available at eleven at night.
It also predicts where the gap should be largest, and it is right. Law, finance, consulting and senior corporate management are the greediest occupations and have the widest gaps. Fields where work is substitutable have narrower ones.
3.5 — The profession that fixed itself by accident
Pharmacy used to be a small business. A pharmacist owned a shop, was known to the customers, kept the records, and could not be replaced on a Tuesday. Long hours, personal client relationships, irreplaceability — a greedy occupation in every respect.
Then the industry consolidated. Pharmacists became employees of chains and hospitals. Records went into shared systems. Any qualified pharmacist could pick up where another left off.
Nobody did this for equality. It was consolidation and computerisation, driven by cost.
The result is one of the smallest gender earnings gaps of any high-paying profession. The penalty for working part-time nearly disappeared, because two pharmacists working half a week became genuinely equivalent to one working the whole of it. Women are now a majority of pharmacists, and they earn close to what the men earn.
Set that beside a generation of pipeline programmes, mentoring schemes and representation targets. The thing that closed the gap in pharmacy was making the work substitutable — and it happened by accident, for reasons that had nothing to do with women at all.
Which leaves the question of what the part of the gap that pharmacy did not eliminate is actually evidence of.
Everybody agrees a residue exists after adjustment. What it means is the dispute.
A residue that survives every measured control is what unmeasured discrimination looks like. Audit studies have demonstrated differential treatment directly, and the fact that a gap shrinks when you control for occupation does not exonerate the process that sorted people into occupations.
The residue is small and shrinking, and the large explained portions are not mysterious — they are hours, continuity and occupational choice, all downstream of who absorbs children. No employer is deciding anything. The greedy-work account predicts the pattern, predicts the stall, and predicts where the gap is largest, and it does all three correctly.
Discrimination is demonstrated to exist and is not large enough to account for the observed gap. Structure accounts for most of it and does not account for all. Nobody has apportioned them and the honest range is wide.
What would settle it: more cases like pharmacy. Every time an occupation becomes substitutable for unrelated reasons, the experiment runs again, and the structural account predicts the gap should fall without any equality intervention.
Why people care so much: because the remedies are completely different. Discrimination calls for enforcement. Structure calls for redesigning how high-paying work is organised, which is expensive, benefits men who want to see their children too, and requires nobody to be accused of anything.
Which is a pattern this series has now met several times: the remedy that would work is the one nobody can campaign on.
Women’s employment rose steeply for four decades and stopped around 2000. The raw earnings gap narrowed fast to about 1990 and has been close to flat since. Two decades of convergence, three of very little — and every explanation has to account for that shape.
Remove hours, occupation, industry and experience and the residue is five to eight per cent. That establishes the gap does not work through unequal pay for identical work. It does not establish that the removed portions are fair.
And the name is wrong. Split by parenthood rather than sex and childless women’s earnings track men’s closely. The gap between mothers and childless women is larger than the gap between childless women and men. It is a parenthood gap appearing as a gender gap because of who absorbs the interruption — which is why gender remedies do not touch it.
Greedy work explains the stall: jobs paying disproportionately for long unpredictable hours make it arithmetically rational for a couple to specialise. Law and finance are greediest and have the widest gaps.
And pharmacy proves the point by accident. Consolidation and computerisation made pharmacists substitutable — for cost reasons, with no thought of women — and it now has one of the smallest gender earnings gaps of any high-paying profession. What closed it was redesigning the work.
4Power
Representation in the world’s parliaments has more than doubled in thirty years and is still nowhere near half. Where quotas were imposed the numbers moved immediately — and in one well-studied case, moved nothing else at all.
4.1 — Parliaments
In the mid-1990s, women held roughly one seat in nine across the world’s national legislatures. Today it is somewhere above one in four.
That is a real change and it is also, after three decades, still a long way from anything proportional. On the current trajectory the arithmetic does not reach parity within the working lives of anybody reading this.
Put a national legislature in a room of a hundred people.
1995: eleven women, eighty-nine men.
Now, world average: around twenty-six women.
Rwanda: above sixty — the highest anywhere, and the reason it outranks most of western Europe on the index Part Five took apart.
India: around fourteen. In the lower house of the world’s largest democracy, roughly seven seats in fifty are held by women — which places India below the world average and well below several of its neighbours.
That last figure sits beside Part Seven’s randomised evidence that women in Indian village councils changed spending and closed the gap in girls’ schooling. The country with the strongest experimental evidence that female representation works has among the weakest representation at national level.
4.2 — India’s law, and why it is not yet doing anything
India passed a constitutional amendment in 2023 reserving one third of seats in the lower house and in state assemblies for women.
It is not yet in operation. Implementation was made conditional on a census and a subsequent redrawing of constituency boundaries, neither of which had been completed as this is written. So the law exists, the reservation does not, and the gap between those two states is itself worth noting — it is the same structure Part Two found in the dowry ban and Part Seven found in the marital rape recommendation. The provision that would change a household’s arrangements is the one that does not commence.
Whether it will work when it does commence is a genuine question, and India has unusually good evidence bearing on it, because the village council reservation described in Part Seven is the same instrument at a smaller scale — and it worked.
4.3 — Boards, and a quota that did exactly one thing
The corporate story is where the honest ledger gets least comfortable for the case that representation is self-propagating.
Norway required listed companies to have at least forty per cent of board seats held by women, with real penalties. The requirement was met. Board composition changed almost overnight, and several countries followed with similar rules.
Then researchers looked at what else changed.
The Norwegian quota is an unusually clean case to study, because it applied to one category of company on a fixed date, leaving comparable companies unaffected — which supplies a comparison group.
What changed: the boards, immediately and completely. The women appointed were more qualified on measurable criteria than the men they joined.
What did not change: almost everything else. Analysis found little effect on women in the senior positions just below the board, little effect on the gender gap in earnings within these firms, and little sign of the change propagating downwards through the organisation.
What it shows: that a quota reliably changes the thing it is applied to, and that the assumption of a knock-on effect through the rest of the organisation is not supported in this case.
What it cannot show: that no such effect exists anywhere. One country, one policy, a period of about a decade. And Part Seven’s Indian evidence found effects that did propagate — on spending, and on what girls in those villages expected of themselves. Both results are solid and they differ.
Set the two side by side, because the contrast is the most useful thing in this chapter.
In Indian village councils, putting women in charge changed what was built and changed what girls expected. In Norwegian boardrooms, putting women on boards changed the boards.
The most plausible difference is what the position controls. A village council head decides where the money goes, on things — water, roads — where local women’s stated priorities differed from men’s. A board member sits on a committee overseeing a company whose operations she does not run.
Which suggests the useful question is not how many women are present but whether the position they hold decides anything that a differently-composed group would decide differently. Counting seats does not ask that.
4.4 — Underneath the national number
India’s fourteen per cent is a national figure and it conceals two things worth separating.
State assemblies are generally lower than parliament, not higher. Several large state legislatures sit in single figures. So the low national number is not an artefact of one chamber; it runs through the system.
And underneath both sits the tier where the reservation already operates. Following the amendment described in Part Seven, women hold a third or more of village council seats across the country — in some states well above a third, because the reservation is a floor rather than a ceiling.
Which produces an unusual structure. India has, simultaneously, one of the world’s largest bodies of women in elected office and one of the lower rates of women in its national legislature. Hundreds of thousands of women decide where a village’s money goes; a much smaller share decide anything at state or national level.
Part Seven established that the village tier works — randomised evidence, changed spending, closed attainment gaps. So India has run the experiment, obtained a positive result, and not yet applied it upwards. The 2023 amendment is the attempt to do so and it awaits a census.
4.5 — The top
At the very top the numbers remain small. Women run somewhere around one in ten of the largest American companies, a share that has risen slowly. Heads of government remain a small minority worldwide, and the count fluctuates by a handful each year.
Part Four’s arithmetic is worth recalling here without being over-applied. The very top of any distribution is where small differences in the middle produce large imbalances — and it is also where the greedy-work structure of Chapter Three operates most severely, since a chief executive’s job is the least substitutable there is.
Both mechanisms predict a persistent gap at the summit even under complete fairness. Neither predicts its current size, and neither is evidence that the current size is right.
Women held roughly one seat in nine in the world’s parliaments in the mid-1990s and somewhere above one in four today. Real, and after thirty years still far from proportional.
India is at about fourteen per cent — below the world average — while holding the strongest experimental evidence anywhere that female representation changes outcomes. India passed a one-third reservation for parliament and state assemblies in 2023 and it is not yet operating, being conditional on a census and boundary redrawing. Same structure as the dowry ban and the marital rape recommendation: the provision touching households is the one that does not commence.
Norway’s board quota worked and did nothing else. Boards changed immediately and completely; the women appointed were more qualified than the men they joined. Effects on senior women below the board, on the internal earnings gap, and on the rest of the organisation were slight.
Set that against India’s village councils, where putting women in charge changed what was built and what girls expected. The likely difference is what the position controls — a council head decides where money goes; a board member oversees a company she does not run. So the useful question is not how many women are present but whether their position decides anything a differently-composed group would decide differently. Counting seats does not ask that.
5The Body
The largest improvement in women’s lives over this period is not on anybody’s ledger. It is not employment, earnings or representation. It is that childbirth stopped killing them, and the numbers are not close.
5.1 — Dying in childbirth, and then not
Part Six, Chapter Two priced a single sexual act in a world without antibiotics, contraception or safe delivery, and found a real chance of pregnancy, a real chance of dying from it, and a real chance of an incurable disease. That was the world the rules were written for.
Here is what happened to the first of those dangers in India, within one lifetime.
Maternal mortality ratio: the number of women who die from causes related to pregnancy and childbirth, for every hundred thousand babies born alive.
It is expressed per hundred thousand rather than per hundred because, in a country with functioning hospitals, the figure per hundred would be a decimal too small to compare. That fact is itself the achievement being measured.
Why it matters here: it is the single cleanest measure of whether a society keeps its women alive through the most dangerous ordinary thing most of them will do.
In 1990, the number of women dying for every hundred thousand live births was somewhere above five hundred. By the end of the 2010s it was below a hundred. It has continued falling since.
That is a reduction of more than four fifths in thirty years, in a country of a billion people, in the single most dangerous ordinary event in a woman’s life.
Globally the direction is the same and the change is smaller, because much of the world started lower.
5.2 — The crossover India took forty years to reach
In almost every country in the world, women outlive men. Part Four established this and found the global gap runs to around five years.
India was, for most of the twentieth century, one of the very few exceptions. Indian women died younger than Indian men.
That is not a small anomaly. It is a measurable statement that something was being done to women, at scale, across a population — because there is no biological reason for it. Part Four found that male infants die at higher rates from the first days of life, everywhere. For a population to invert that requires the deliberate allocation of food, medicine, attention and care.
The crossover happened in the early 1980s. Indian women now outlive Indian men by around three years.
Put that in a household. For most of the century, a girl born in India could expect fewer years than her brother — not because of her body, but because of what would be spent on her. Within one generation that reversed.
It is among the largest changes in this entire series and it appears on no political ledger, because neither camp can claim it and neither wants to explain it.
I want to be careful about what that reversal does and does not mean. It does not mean the discrimination that produced the anomaly has ended — Part Twelve counts what is still happening before birth, and the sex ratio at birth remains skewed. It means that the excess mortality of girls and women after birth, which used to be large enough to invert a biological constant, has largely gone.
5.3 — Control
The third medical entry is the one that changes what a life can contain.
India’s total fertility rate — the number of children an average woman has across her life — has fallen below replacement. Modern contraception is used by a majority of married women. Both of those would have been unimaginable to the people writing the rules in Parts Two and Three, and both arrived within about fifty years.
Part Two, Chapter Nine scored this problem as “technically solved, practically not,” because access in India is uneven and much of the burden falls on female sterilisation after childbearing rather than on preventing unwanted pregnancy. That criticism stands. It sits on top of a change that is nonetheless enormous.
5.4 — And the entry that goes the other way
The medical column is not uniformly positive, and the exception is the most uncomfortable finding in this part because it comes from the same source as the gains.
The technology that made pregnancy safe also made it visible. Ultrasound imaging arrived in India through the 1980s and spread rapidly. It is the reason obstructed labour can be anticipated, complications detected and lives saved.
It is also the reason a family can find out the sex of a foetus.
India’s sex ratio at birth — the number of girls born for every thousand boys — worsened across exactly the decades when maternal and child mortality were collapsing. Determining the sex of a foetus for this purpose was prohibited by law in 1994 and the prohibition has been extensively documented as poorly enforced. Part Twelve counts the totals and they are very large.
So the same medical advance appears in both columns. It stopped women dying in childbirth and it supplied the instrument by which daughters are not born.
Which is worth holding against the whole of Chapter Eight’s argument. A gain that arrives at the door reaches everybody — including the households that will use it for something else. Public health delivers a capability, and what a family does with a capability is decided by the arithmetic in Part Two, Chapter Nine: no pension, a son as the only instrument, a daughter as a transfer to another household.
5.5 — Violence, counted honestly
The last entry in this chapter is the one where the ledger is genuinely mixed and the reporting is worst.
In the United States and several comparable countries, intimate partner violence against women fell sharply from the mid-1990s — by a large fraction over about two decades, on the national victimisation surveys. Killings of women by partners fell alongside it. This is one of the clearest improvements on the whole ledger and it is almost never mentioned by either camp.
In India the change is much smaller. National health surveys ask ever-married women directly about spousal violence, and the share reporting it has moved from roughly three in ten to slightly under three in ten across the most recent survey rounds. That is a decline. It is not a transformation, and Part Twelve counts what it means at population scale.
And a caution that applies to both. Survey prevalence and reported crime measure different things, Part Seven scored the undercounting claim as supported, and a fall in reported incidents can reflect a fall in incidents or a change in what gets reported. The Indian surveys ask women directly, which is why they are the better instrument, and they still depend on what a woman will say to a stranger in her own house.
Both camps assume that the important changes are the ones being argued about.
Sixty years of public argument about women has been conducted almost entirely over employment, earnings, representation and law. Those are the columns everybody counts.
Now compare magnitudes. The earnings gap moved from roughly sixty to roughly eighty-two cents in the dollar. Parliamentary representation moved from one in nine to one in four. Real changes, decades of work.
And maternal deaths in India fell by more than four fifths. And women went from dying younger than men to outliving them by three years. Nothing in the political column is the same order of magnitude as either.
Why are they not on the ledger? Because neither camp can claim them. They were produced by antibiotics, antenatal care, trained birth attendants, blood transfusion and vaccination — by public health systems, largely staffed by people with no position on any of this. A gain nobody can take credit for generates no argument, and a change that generates no argument does not get counted.
The general form: arguing about the column that produces argument rather than the one with the largest number. It is not a feature of this subject. It is a feature of every subject where the accounting is done by the participants.
The next chapter is the one entry on this ledger that everybody counts, and it is the only one that goes the wrong way.
Indian maternal deaths fell from above five hundred per hundred thousand live births in 1990 to below a hundred by the end of the 2010s. More than four fifths, in thirty years, in the most dangerous ordinary event of a woman’s life.
And India was one of the very few countries where women died younger than men — an anomaly with no biological basis, requiring the deliberate allocation of food, medicine and care. The crossover came in the early 1980s. Indian women now outlive Indian men by around three years.
India’s fertility has fallen below replacement and a majority of married women use modern contraception — on top of the access problems Part Two recorded, not instead of them.
Violence is mixed. Intimate partner violence against women fell sharply in the United States and comparable countries from the mid-1990s. In India it has moved from roughly three in ten to slightly under.
None of the medical gains appears on anybody’s ledger, and the reason is that neither camp can claim them. They came from antibiotics, antenatal care, trained birth attendants and vaccination — from people with no position on any of this. A gain nobody can take credit for generates no argument, and what generates no argument does not get counted.
6The Happiness Problem
Across the decades of greatest gain, women’s own ratings of their happiness fell. The finding is real, nobody predicted it, and it is quoted constantly by people who have not checked what it establishes — which is narrower and stranger than either camp wants.
6.1 — The finding
Large national surveys have asked people the same question about their own happiness, in the same words, for decades. That consistency is what makes a long comparison possible.
Subjective wellbeing: what a person says when asked to rate their own life or happiness, usually on a short scale with three or five options.
Three properties matter for everything in this chapter. The scale has no fixed meaning — one person’s "very happy" is another’s "pretty happy," and there is no external anchor. It is ordinal, so the distance between options is unknown and cannot be averaged in the way it routinely is. And what a person is willing to say depends on what is socially acceptable to admit, which changes across decades.
Why it matters here: this is the only measure in this entire part that is not a count of something, and it is the only one that moved in the unexpected direction. Those two facts have to be held together.
Researchers examining the American series across roughly thirty-five years found that women’s average reported happiness declined — both in absolute terms and relative to men, whose ratings held steadier. Examining European data, they found the same direction in most of the countries where comparable series existed.
The period covered is the one in Chapters Two to Five. Education, earnings, employment, legal rights, physical safety within marriage, control over childbearing and survival of childbirth all improved substantially, several of them enormously.
Nobody predicted this. Part Seven established that the movement expected the opposite and its opponents expected something else entirely — social disorder rather than a quiet decline in self-reported wellbeing among the people who had gained most.
6.2 — Seven explanations
Every one of these has been proposed by serious people. None is established.
One: expectations rose faster than conditions. A person rates their life against what they thought it would be. If the horizon expanded faster than the reality, the gap widens even as the reality improves.
Two: total work increased. Part Seven’s second shift. If paid work was added and unpaid work was not correspondingly removed, the result is more hours and less rest, which would plausibly show up in a wellbeing measure.
Three: the comparison group changed. Part Five identified this mechanism precisely. In more gender-integrated societies people compare themselves across sex rather than within it. A woman comparing her life to a mixed population rather than to other women is measuring against a different standard, and the rating can fall with nothing else changing.
Four: the people being measured changed. As more women entered paid work, the population answering the survey shifted in composition. The average woman in 1975 and the average woman in 2005 are not the same sample in the ways that matter for a wellbeing question.
Five: reporting norms changed. If it became more acceptable for a woman to say she was not happy, the measured decline is a decline in the pressure to say she was. This would be a change in candour, not in wellbeing, and the two are indistinguishable in the data.
Six: the scale is the problem. Answers are on a short ordinal scale with no fixed meaning, and the meaning of the middle option can drift across decades. This is not a marginal objection; it applies to every long comparison of self-rated anything.
Seven: it is real. Something in the changes made women’s lives worse in a way the objective measures do not capture.
Here is the fact that reframes the whole chapter, and it comes from outside this subject entirely.
Self-reported happiness barely moves for anybody, in response to almost anything.
Rich countries got several times richer across the second half of the twentieth century. Average reported happiness in those countries was roughly flat. This is one of the most replicated findings in the study of wellbeing and it has its own long literature.
So a group’s happiness line failing to rise during a period of enormous objective gain is not an anomaly requiring explanation. It is what these measures do. Nothing has ever made a national happiness average go up much — not wealth, not health, not peace.
Which means the expectation that women’s happiness would rise with their gains was an expectation that has never been met by any improvement, for any group, anywhere.
The genuine puzzle is therefore narrower than it is usually stated. It is not that happiness failed to rise. It is that it fell, and fell relative to men. That is a smaller finding and it is still a real one.
Which is where the dispute about what to do with it begins.
Quoted in every direction, usually by people who have read the headline.
The people who received the benefits are reporting lower wellbeing than the people who did not. That is the most direct evidence available about whether a change was good for its intended beneficiaries, and dismissing it because it is inconvenient is exactly the moralistic fallacy Part One named.
Self-rated happiness is flat against everything, moves with reporting norms, has an unstable scale and is measured on a shifting population. And the same period produced women living years longer, dying in childbirth at a fraction of the rate, and being beaten by partners far less. If a wellbeing measure disagrees with survival, health and safety, the question is what the measure is tracking.
The relative decline against men is a genuine result for the period examined, and its continuation into more recent data is disputed. Seven explanations fit and none has been established. It is a fact in search of a cause and it should be held that way.
What would settle it: the same question asked with better instruments — measures anchored to something stable rather than to a respondent’s own scale. Some exist and the series are too short.
Why people care so much: because it is the only number in this entire part that goes the wrong way, and a single contrary number in a long ledger attracts attention out of all proportion to its weight. That is worth noticing about the number and also about the attention.
Part Nine returns to one of the seven explanations with much better data behind it, because postponement leaves a trace that a happiness scale does not.
Across roughly thirty-five years covering the greatest gains in women’s education, earnings, rights, safety and survival, women’s reported happiness fell — absolutely and relative to men — in the United States and in most European countries with comparable data. Nobody predicted it.
Seven explanations have been offered: rising expectations, more total work, comparison across sex rather than within it, a changed sample, changed reporting norms, an unstable scale, and genuine decline. None is established.
And the reframing that matters comes from outside the subject. Self-reported happiness barely moves for anybody in response to anything. Rich countries got several times richer and their happiness lines were flat. So the expectation that women’s happiness would rise with their gains was an expectation never met by any improvement for any group.
Which narrows the puzzle. It is not that happiness failed to rise — that is what these measures do. It is that it fell relative to men. Smaller, and still real.
What you are entitled to say: measured female happiness fell over that period, nobody predicted it, nobody has explained it. What you are not entitled to say: that the changes made women worse off — because that requires ruling out six explanations nobody has ruled out.
7The Bill
A ledger needs a second column. Some of the entries usually placed in it do not survive checking. The ones that do are not the ones most often named.
7.1 — The claim that does not survive
The most frequently stated cost is that women took on paid work without putting down unpaid work, and therefore ended up doing more in total than men.
It has a name, it entered the language, and it described something real when it was first written about in the 1980s.
A time use survey asks people to account for a whole day — usually in a diary, in intervals, recording what they were doing at each point.
It is a much better instrument than asking "how many hours a week do you spend on housework," because people are poor at estimating totals and systematically over-report activities they think they should be doing. A diary of yesterday is harder to get wrong.
Why it matters here: nearly everything in this chapter rests on these surveys, including India’s own, and they are the reason the second-shift claim can be checked at all rather than merely asserted.
The current evidence is more awkward than that.
Add up everything a person does that counts as work — paid employment, plus cooking, cleaning, childcare, shopping, repairs, care of the old. Then compare the totals.
In rich countries, men and women do roughly the same total. The composition differs sharply — women do far more of the unpaid, men more of the paid — but the sum comes out close to level. This has been found across many wealthy countries.
In poorer countries it does not. There, women’s total work substantially exceeds men’s, and India is well inside that group. Indian women report roughly five hours a day of unpaid domestic and care work against about an hour and a half for men, and the paid hours do not compensate.
So the second-shift claim, as a general statement about women everywhere, is not supported. As a statement about Indian women it is supported, and by India’s own national survey.
Which is a pattern worth noticing across this entire part. Several claims that have become weak in rich countries remain accurate here, and a reader importing the rich-country version of either camp’s argument will get India wrong.
7.2 — What is genuinely on the bill
Three entries survive checking, and they are not the ones most often named.
Leisure, not child time. The expectation was that working parents would spend less time with their children. Measured, the opposite happened: parental time with children rose in rich countries across exactly the decades when both parents took paid work. What fell was everything else — sleep, rest, time alone, unstructured hours. The cost of the transition was not paid by children. It was paid out of the parents’ own remaining hours, and it was paid disproportionately by mothers.
Postponement, and the gap it leaves. The age at which women have a first child rose sharply across the developed world and is rising in urban India. One consequence is measurable and is genuinely a cost: surveys in rich countries consistently find that women end up with fewer children than they say they wanted. Not fewer than a demographer would like — fewer than they themselves reported wanting. That gap between intended and achieved family size is a real deficit and it is one of the clearest entries on this side of the ledger. Part Nine takes up what it does at population scale.
And the care that was moved rather than removed. When women in rich countries entered paid work, a large part of the domestic and care work did not disappear and was not redistributed to men. It was purchased — from other women, frequently poorer, frequently migrants, frequently leaving their own children behind to do it.
That last entry deserves emphasis because it is the one that most complicates the ledger. A gain recorded for one woman was in part a transfer from another, and the second woman appears in nobody’s accounting because she is in a different country’s statistics.
Everything in this chapter assumes that unpaid work is a cost.
Look at the accounting. Hours of childcare, cooking and care of the old are entered as burden — something to be reduced, redistributed, outsourced or compensated. Both camps do this. The traditional side says women should bear it willingly; the reformist side says it should be shared or paid. Neither questions that it is a debit.
But a substantial part of what is being counted is the raising of children and the tending of the dying. Those are not chores that happen to fall on somebody. For a great many people they are the content of a life, and the reason the rest of it is for anything.
Part Seven, Chapter One identified a whole strand of feminism built on precisely this objection: that the remedy is not to escape this work but to stop treating it as the absence of achievement. That strand loses the argument automatically the moment the accounting begins, because a ledger has no column for it.
The general form: adopting the accounting system of the thing you are criticising. If only paid work counts as achievement, then a movement measuring its success in paid work has conceded the premise it started by attacking — and every hour of care becomes a loss, including the hours somebody would not have given up.
I am not going to resolve that, and the next chapter proceeds with the accounting anyway, which is itself the concession the box describes.
The most-quoted cost — that women took on paid work without putting down unpaid work — does not survive checking as a general claim. In rich countries, men’s and women’s total work is roughly level; the composition differs sharply and the sum does not.
In India it does survive. Women report roughly five hours of unpaid work a day against about an hour and a half, and paid hours do not compensate. Several claims that have gone weak in rich countries remain accurate here.
Three entries survive. Leisure, not child time — parental time with children rose across the transition, and what fell was sleep, rest and unstructured hours, paid disproportionately by mothers. Postponement — women in rich countries end up with fewer children than they themselves said they wanted. And care that was moved rather than removed — purchased from poorer women, often migrants, who appear in nobody’s accounting because they are in another country’s statistics.
And the whole chapter assumes unpaid work is a cost. Much of what is counted is raising children and tending the dying — for many people the content of a life rather than a chore. A ledger has no column for that, which means the strand of feminism built on exactly this objection loses automatically the moment the accounting starts.
8Who Actually Gained
“Women gained” is a sentence about an average, and the distribution behind it is stark. The gains went overwhelmingly to one group — with a single exception, and the exception is the most important thing in this chapter.
8.1 — By education
Take the changes in Chapters Two, Three and Four and ask which women received them.
A woman with a university degree gained on nearly every measure. Earnings rose substantially. Occupational range expanded. Autonomy within marriage increased. Her marriage, as Part Six established, is more likely to be stable than a marriage in any other educational group.
A woman without one received a much smaller share. Her earnings rose less. The jobs that opened were mostly not open to her. And — this is the entry that matters most — Part Six, Chapter Three found that the collapse in marriage was concentrated in exactly her group, which means she is far more likely to be raising children alone, in the household type with the least money and the fewest adults.
So the same period produced, for one group of women, more income and a stable partnership, and for another, slightly more income and a substantially higher chance of raising children without one.
8.2 — By country, and by caste
The second and third cuts run the same way.
By country: the gains recorded in this part are concentrated in rich countries. In much of sub-Saharan Africa and South Asia the political and economic columns moved far less, and Part Five’s paradox complicates any simple ranking.
By caste, in India: the gaps are large and documented in the national health and education surveys. Literacy, maternal outcomes, age at marriage, exposure to violence and schooling all differ substantially between caste categories, and the differences have narrowed less than the headline national figures suggest.
With one complication that inverts the expected pattern and is worth stating carefully. Dalit and Adivasi women’s participation in paid work is higher than that of women in higher-status groups — for the reason Part Three identified. Withdrawing a woman from work is a purchase, and only a household with a surplus can make it. So a measure that reads as advancement in one context reads as necessity in another, and Part Five, Chapter Six found the same inversion at country level.
Two Indian women, the same age, the same year, the same country.
The first has a degree, lives in a city, married at twenty-eight, has one child, works in an office, has a bank account in her own name and a phone nobody else uses. On almost every measure in this part, her life differs from her grandmother’s beyond recognition.
The second left school before finishing, married at nineteen in a village, has three children, works in fields or in somebody’s house for wages that are not recorded anywhere, and has no independent claim on any asset. Her life differs from her grandmother’s in some respects and not in most.
Both are covered by the sentence “Indian women gained.”
And there is exactly one column where their gains are similar, which is the subject of §8.3.
Claims about distribution are harder to establish than headline averages, so it is worth setting out what makes these ones possible.
Distributional claims are harder to establish than headline averages, because they require the data to be broken down without the subgroups becoming too small to measure.
What makes it possible here: India’s national health surveys interview several hundred thousand households and record caste category, education, wealth quintile and location for every respondent. That sample is large enough that a figure for, say, maternal care among rural Dalit women in a particular state is a real measurement rather than a handful of cases.
What it shows: gaps between groups on literacy, age at marriage, antenatal care, exposure to violence and schooling, and — the finding that matters most here — that the health indicators have converged between groups considerably more than the economic and educational ones.
What it cannot show: why a household did what it did. A survey records that a woman had antenatal care or did not. It does not record whether she was prevented, could not travel, was not told, or chose otherwise — which are four different problems requiring four different remedies, and the aggregate cannot distinguish them.
8.3 — The exception
Chapter Five’s medical gains are the exception, and it is the most important finding in this chapter.
Maternal mortality did not fall only for educated urban women. It fell across the population. The collapse in child mortality reached villages. Vaccination reached villages. The life expectancy crossover happened for Indian women as a group, not for a segment of them.
The reason is structural. Public health reaches people who are not asking for it. A vaccination campaign, an antenatal programme, a trained birth attendant scheme and a supply of antibiotics deliver to whoever is there, and do not require the recipient to have education, income, mobility, confidence, a supportive family or a functioning labour market.
Every other gain on this ledger required something of the woman receiving it. She had to be able to attend, to apply, to travel, to be permitted, to compete. Which means every other gain was filtered through exactly the constraints Parts Two and Three described, and reached the women who had least of them.
The gains that arrived without being asked for are the gains that arrived for everybody.
Everybody agrees the gains were concentrated. What follows is disputed.
A project designed by educated women delivered to educated women, and the women who needed it most got least. The priorities that were pursued — professional entry, boardrooms, representation — were the priorities of the people setting them, which is exactly the criticism Part Seven recorded from the intersectional and Dalit critique, made from inside and largely not acted on.
Nothing arrives evenly. Literacy, electricity, medicine, the vote and the telephone all reached the advantaged first and spread. Judging a sixty-year process by its distribution at year sixty mistakes a stage for an outcome, and the alternative — refusing gains to anybody until they are available to everybody — helps nobody.
Both of the above are about blame. The useful reading is diagnostic. The gains that required something of the recipient reached the women who had most; the gains delivered by a system reached everybody. That is not a fact about intentions. It is a fact about mechanisms, and it says what to build.
What would settle it: comparing the distributional reach of delivered programmes against opt-in programmes, on the same population, over the same period. India has run enough of both to make this answerable and I have not found it done.
Why people care so much: because the first two positions are about who is at fault and the third is about what to do, and the first two generate more conversation.
Chapter Nine takes that diagnostic reading and asks what it says about the whole ledger.
“Women gained” is an average. A woman with a degree gained on nearly every measure — earnings, occupation, autonomy, and a marriage more stable than in any other group. A woman without one gained less on all of them, and Part Six found the collapse in marriage was concentrated in exactly her group, so she is far more likely to be raising children alone.
The same cuts run by country and by caste. And with one inversion: Dalit and Adivasi women’s paid work participation is higher than higher-status women’s, because withdrawal is a purchase only a household with a surplus can make — the same inversion Part Five found at country level.
There is one exception and it is the most important thing here. The medical gains in Chapter Five reached everybody — maternal mortality, child mortality, vaccination, the life expectancy crossover. Because public health reaches people who are not asking for it.
Every other gain required something of the woman receiving it: to attend, apply, travel, be permitted, compete. So every other gain was filtered through exactly the constraints Parts Two and Three described, and reached the women who had least of them. The gains that arrived without being asked for are the gains that arrived for everybody — and that is a fact about mechanisms rather than intentions, which means it says what to build.
9Scoring The Ledger
Eight chapters of counting. The remaining question is what the leftover gaps are evidence of — unfinished business, or a tool that reached the limit of what it can do. The answer differs by country and depends on a goal nobody chose.
9.1 — The shape of the result
Put the columns together and a specific pattern appears, and it is not the pattern either camp describes.
The medical column is enormous, broadly distributed, and uncontested. Maternal deaths down by more than four fifths in India. A life expectancy crossover that reversed an anomaly with no biological basis. Child mortality collapsed. Fertility below replacement and under control. Nothing else on the ledger is the same size, and it reached everybody because it did not require anybody to come and ask.
The educational column went further than anybody predicted and then stopped mattering as much as expected. Women out-graduate men across the developed world, India has drawn level, and the assumption that this would unlock the rest turned out to be wrong.
The economic column moved fast for twenty years and then stalled. And the stall has an identified cause that is not legal and not attitudinal: the structure of high-paying work, and who absorbs a first child.
The political column moved slowly and steadily and is nowhere near proportional. Where quotas were imposed the numbers changed immediately, and whether anything else changed depended on whether the position decided anything.
The costs column has three real entries — leisure rather than child time, a gap between children wanted and children had, and care purchased from poorer women — and one widely repeated entry that does not survive checking in rich countries and does survive in India.
And one measure went the wrong way, in a way nobody predicted, that nobody has explained, and that neither camp is entitled to use the way they use it.
9.2 — Unfinished, or the edge of the instrument?
The question that decides how the rest of this series reads.
Sixty years is nothing against several thousand. The gaps that remain are the residue of a system still operating — in hiring, in expectation, in who is asked to leave work when a child arrives. Every previous generation was told the remaining gap was natural and every previous generation was wrong, which is a reason to distrust the claim now.
The instrument was removing legal and formal barriers, and it worked spectacularly — Part Seven scored that as supported. But its returns fell as it ran out of barriers. What remains is not a barrier: it is the structure of greedy work, the allocation of children within couples, and the preference differences of Parts Four and Five. None of those is illegal and none responds to the tool that produced every previous gain.
Both of the above are answers about rich countries. In India the barrier-removal instrument has an enormous amount left to do — Part Three’s machine is intact, the marital rape exception stands, Part Five found 43 per cent of science graduates and a fraction employed, and Chapter Four found national representation below the world average. The question “is the project finished” has one answer in Norway and another here.
What would settle it: whether the structural remedies work. If redesigning work reduces the gap where legal remedies stopped — as Chapter Three’s pharmacy case suggests — the second position is right about the instrument and wrong about the ceiling.
Why people care so much: because “unfinished” licenses continued effort and “finished” licenses stopping, and both conclusions are being drawn from a ledger that supports neither.
9.3 — What the ledger does to the rules
Every part of this series ends by asking what its findings do to the rules in Parts Two and Three. A ledger of results has a specific answer and it is sharper than the earlier ones.
The rules in this series exist to constrain women’s behaviour — who they marry, whether they remarry, what they wear, where they go, what is said about them. Sixty years of counting produces four findings that bear on them, and every one runs the same way.
The gains that arrived reached women through capability, not through permission. Chapter Eight’s finding was that what was delivered reached everybody and what had to be claimed reached the least constrained. Every rule in Parts Two and Three operates on the second channel — by determining what a woman is permitted to go and get. Which means the rules did not stop the medical gains and did throttle almost everything else.
The bottleneck moved and the rules did not. Chapter Two found that education was unblocked and stopped being the constraint. Chapter Three found the constraint is now the structure of work and the allocation of children. No rule in Parts Two and Three addresses either. They were built for a world in which a woman’s problem was access, and access is no longer the binding problem in the places where they are most strictly enforced.
The one entry that went the wrong way in the medical column was produced by the rules, not by the reforms. Section 5.4 found that ultrasound gave families a capability, and Part Two, Chapter Nine established what a family does with it: no pension, a son as the only instrument, a daughter as a transfer. That is the machine in Parts Two and Three, operating on a modern technology, producing a modern harm.
And the strongest positive result in the whole part is a case of the rules being suspended. India’s village council reservation works by removing, for a third of positions, the ordinary operation of who gets to hold power. It is a rule about women, imposed against the existing rules about women, and it produced measurable change in spending and in what girls expected of themselves.
So the ledger does not merely fail to support the rules. It records that where they were suspended, outcomes moved, and where they operate, the gains that require permission do not arrive.
9.4 — And the question underneath
Every attempt to score this ledger assumes that the project had a goal.
You cannot say whether something succeeded without knowing what it was for. And Part Seven, Chapter Eight established that this particular thing had two aims which conflict: equal freedom and equal outcomes. Nobody chose between them. The philosophical case rested on the first; the institutions measure the second; and the drift happened because one is countable and the other is not.
So the ledger gives two different verdicts and both are correct.
Measured against freedom — can a woman own, earn, vote, leave, choose, refuse — the project in rich countries has substantially succeeded, and in India substantially has not. That is a clear answer and it is the answer to the question the movement originally asked.
Measured against outcomes — proportional presence in fields, earnings and leadership — it stalled decades ago and Part Five suggests the target may not be reachable by these means.
The general form: scoring against a target chosen after the fact. Whoever gets to name the goal determines the verdict, and in a dispute where nobody chose the goal, everybody names the one that produces the answer they wanted.
9.5 — What a ledger cannot do
One closing limitation, stated plainly because this part is the most numerical in the series and numbers invite more confidence than they earn.
A ledger records magnitudes. It cannot weigh them against each other, because the entries are in different units and no exchange rate exists.
How many percentage points of representation is a maternal death worth? How many hours of lost sleep against the ability to leave a marriage? How does a child a woman wanted and did not have compare against the years she gained by not dying in childbirth?
These are not hard questions. They are questions with no answer, because they require a common unit that does not exist — and Part Two’s front matter identified the same problem in its first pages, when it noted that “was this good for women” requires adding together experiences that cannot be added.
What a ledger can do is stop people asserting magnitudes that are not there. It establishes that the medical column is larger than the political one, that the education reversal is larger than anybody says, that the earnings gap is not what it was described as, and that one number went the wrong way. Those are all real constraints on what a person may claim.
Anybody who tells you the ledger comes out positive or negative is adding across units. That is not a calculation they performed. It is a preference they held before they started.
The medical column is enormous, broadly distributed and uncontested, and reached everybody because it did not require anybody to ask. The educational column went further than predicted and mattered less than expected. The economic column moved fast for twenty years and stalled on a cause that is neither legal nor attitudinal. The political column moved slowly and is nowhere near proportional. The costs column has three real entries. And one measure went the wrong way.
The remaining gaps are evidence of different things in different places. The barrier-removal instrument has largely exhausted itself where the barriers are gone, and has barely started where they are not — and almost all the writing on this is produced in the first kind of country and read in the second.
And you cannot score a project without knowing its goal. Measured against freedom, it has substantially succeeded in rich countries and substantially not in India. Measured against outcomes, it stalled decades ago. Both verdicts are correct, nobody chose between the goals, and whoever names the goal determines the answer.
Finally, a ledger records magnitudes and cannot weigh them, because the entries are in different units. How many points of representation is a maternal death worth? Anybody telling you the ledger comes out positive or negative is adding across units — which is not a calculation they performed but a preference they held before they started.
10An Honest List Of What We Do Not Know
Two lists, no hedging in either. What is genuinely unknown about the last sixty years, what is solid enough to build on, and the columns nobody keeps.
10.1 — Genuinely unknown
The unknowns in a ledger are of a particular kind: not what happened, but why, and what would have happened otherwise.
How much of this the movement caused. Chapter One set out why this is unresolvable in general. Antibiotics, contraception, collapsing child mortality, piped water and the shift from physical to literate work each transformed women’s lives without requiring any politics. Nobody has apportioned them and both camps apportion confidently in opposite directions.
Why women’s reported happiness fell. Seven explanations fit. None is established. Whether the decline continued past the period originally studied is disputed.
Why the economic convergence stalled when it did. The greedy-work account fits well and makes correct predictions. How much of the residue is structure and how much is discrimination has not been apportioned, and the honest range is wide.
Whether the education reversal will correct itself. Boys’ relative decline has continued for four decades. Four accounts compete and none dominates. Nobody knows whether it stabilises, continues, or reverses.
What India’s employment numbers actually mean. Part Five flagged this and it is unresolved. Recent Indian surveys report a substantial rise in women’s workforce participation, and economists disagree about how much reflects women entering work against a change in how unpaid family labour is counted. Part Twelve takes it up and will not settle it either.
And whether the structural remedies work. The pharmacy case is one occupation that became substitutable by accident. Whether the effect generalises — whether deliberately redesigning work reduces the gap where legal remedies stopped — has not been tested at scale anywhere.
10.2 — Solid
That maternal mortality in India fell by more than four fifths in thirty years, from above five hundred per hundred thousand live births to below a hundred. Registered deaths against registered births, in a national system.
That Indian women went from dying younger than Indian men to outliving them. The crossover happened in the early 1980s. The anomaly it corrected had no biological basis, which means it recorded a deliberate allocation of food, medicine and care — and its reversal records the end of most of that.
That women out-graduate men at every level across most of the developed world, and that the gap favouring women now exceeds the gap that favoured men in the early 1970s. Administrative counts of degrees awarded.
That the assumption education was the bottleneck has been tested and failed. The education arrived and the rest did not follow — most starkly in India, which produces 43 per cent female STEM graduates and employs a fraction of them.
That earnings convergence stalled around 1990 in the United States and comparable countries, after two decades of rapid movement.
That the residue after adjustment is small — commonly five to eight per cent — which establishes the gap does not operate through unequal pay for identical work, and establishes nothing about whether the adjusted-away portions are fair.
That parliamentary representation rose from about one in nine to above one in four globally, and stands at about fourteen per cent in India.
That Norway’s board quota changed boards and little else, while India’s village council reservation changed spending and girls’ attainment. Both findings are solid and they differ, and the likely difference is what the position controls.
That women’s reported happiness declined absolutely and relative to men over the period studied, and that self-reported happiness barely moves for anybody in response to anything — which narrows the puzzle to the relative decline rather than the failure to rise.
That total work is roughly level between the sexes in rich countries and substantially higher for women in India.
And that the gains were distributed unevenly, with one exception. Everything requiring the recipient to attend, apply, travel or compete reached the women who were least constrained. The medical gains, which required nothing of her, reached everybody.
10.3 — The columns nobody keeps
One closing observation, following the pattern of the previous parts.
Look at what this part could be built from. Degrees awarded. Earnings recorded for tax. Seats counted in a chamber. Births and deaths registered. Hours reported to a time use survey.
Every one of those exists because an institution needed it for a purpose unconnected to this argument. Which is exactly what makes them reliable — and also what determines what could be counted at all.
So notice what has no column. Whether a woman was consulted about her own marriage. Whether she can leave a room without asking. Whether her opinion is solicited when money is spent. Whether the work she does is regarded by the people around her as work. Whether she is afraid in her own house.
Some of these are asked in specialised surveys and none is counted routinely, over decades, in a series that could show a trend. Which means the entire ledger in this part is a record of the things that governments and employers had a reason to write down — and Part Two, Chapter Ten reached the same conclusion about a very different archive four thousand years earlier.
The instruments changed. What they are pointed at did not.
Genuinely unknown: how much of this the movement caused rather than the century. Why reported happiness fell. Why economic convergence stalled when it did, and how much residue is structure against discrimination. Whether the education reversal corrects itself. What India’s employment numbers actually mean. And whether structural remedies generalise beyond the one occupation that ran the experiment by accident.
Solid: Indian maternal mortality down more than four fifths. Indian women went from dying younger than men to outliving them, crossing over in the early 1980s. Women out-graduate men everywhere in the developed world, by a margin now exceeding the one that favoured men in the 1970s. Education was not the bottleneck. Earnings convergence stalled around 1990 and the adjusted residue is five to eight per cent. Representation rose to above one in four globally and stands at fourteen per cent in India. Norway’s quota changed boards and little else while India’s changed spending and girls’ attainment. Reported happiness fell relative to men. Total work is level in rich countries and higher for women in India. And the gains were uneven except where they were delivered rather than claimed.
And notice what has no column. Whether a woman was consulted about her own marriage. Whether she can leave a room without asking. Whether her opinion is solicited when money is spent. Whether she is afraid in her own house. This ledger is a record of what governments and employers had a reason to write down — which is exactly what Part Two concluded about an archive four thousand years older. The instruments changed. What they are pointed at did not.
The Ledger
Sixty years on one sheet. Direction, rough magnitude, and — the column almost never included — which women it reached.
| Measure | Direction and size | Reached whom |
|---|---|---|
| Maternal deaths (India) | Fell from above 500 per 100,000 live births to below 100. More than four fifths, in thirty years. | Everybody. Delivered rather than claimed. |
| Female against male life expectancy (India) | Crossed over in the early 1980s. Women now outlive men by around three years. | Everybody. |
| Child mortality (India) | Fell by roughly three quarters since 1990. | Everybody. |
| Fertility and contraception (India) | Total fertility below replacement. Majority use of modern contraception. | Broad, with the access problems Part Two recorded. |
| Higher education | Reversed. Women a majority of graduates across the developed world; India level or ahead on enrolment. | Educated and urban first; now broad in India. |
| STEM graduation (India) | Around 43 per cent women — ahead of Britain, France, Germany, the United States. | Urban and educated. |
| Employment (rich countries) | Rose steeply for four decades, then plateaued around 2000. | Educated women most. |
| Employment (India) | Flat or falling for decades; a recent reported rise whose interpretation is disputed. | Barely, and contested. |
| Earnings gap | Narrowed fast to about 1990, then close to flat. Residue after adjustment five to eight per cent. | Educated women most. |
| Parliamentary representation | From about 1 in 9 globally in 1995 to above 1 in 4. India at about 14 per cent. | Not a personal gain; an institutional one. |
| Corporate boards under quota | Changed immediately and completely where required. Little propagation below board level. | A very small number of women. |
| Intimate partner violence (US and comparable) | Fell sharply from the mid-1990s. Partner killings of women fell with it. | Broad. |
| Spousal violence (India) | From roughly three in ten ever-married women to slightly under. A decline, not a transformation. | Broad and small. |
| Total work, paid plus unpaid | Roughly level between the sexes in rich countries. Substantially higher for women in India. | — |
| Time with children | Rose across the transition. What fell was sleep, rest and unstructured hours. | Cost paid disproportionately by mothers. |
| Children wanted against children had | A persistent deficit in rich countries. Women end with fewer than they reported wanting. | Rich countries; arriving in urban India. |
| Marriage and partnership | Fell sharply, and the fall was concentrated among the less educated. | Negative, and concentrated on those with least. |
| Reported happiness | Fell — absolutely and relative to men — over the period studied. Unpredicted and unexplained. | Measured almost entirely in rich countries. |
Two things are visible in that sheet that are not visible in any single chapter.
The largest entries are at the top and they are medical. Nothing in the political or economic rows is the same order of magnitude as maternal mortality falling by four fifths, and none of the medical rows appears on the ledgers either camp publishes.
And the right-hand column sorts almost perfectly. Everything that required a woman to attend, apply, travel, compete or be permitted reached the women who were least constrained. Everything delivered to her door reached everybody. That is the most useful line in this part and it is a fact about mechanisms, not about intentions.
Sources & further reading — Part 8
Glossary
Every hard word used in this part, in plain English. Each was explained where it first appeared.
| Term | Plain meaning |
|---|---|
| Attribution | Deciding what caused an observed change when several things were happening at once and none can be switched off to check. Unsolvable in general; both camps do it confidently in opposite directions. |
| Baseline | The year a comparison counts from. Chosen rather than given, and choosing a trough produces a rising line by arithmetic rather than achievement. |
| Greedy work | A job that pays disproportionately more for long, inflexible and unpredictable hours. Where work is greedy, a couple maximises income by having one person absorb all of it — which produces specialisation without anybody discriminating. |
| Ledger | A record with two columns. Advocates publish gains and critics publish costs; neither is a ledger, because a ledger requires the same person to count both sides. |
| Maternal mortality ratio | Deaths of women from causes related to pregnancy and childbirth, per hundred thousand live births. India’s fell from above 500 to below 100 within thirty years. |
| Substitutable work | Work where one qualified person can take over from another without loss, so hours can be arranged and nobody is irreplaceable at short notice. Pharmacy became this by accident and its gender earnings gap nearly closed. |
| Total fertility rate | The number of children an average woman has across her lifetime. India’s has fallen below the level required to replace the population. |
| Total work | Paid employment plus unpaid domestic and care work added together. Roughly level between the sexes in rich countries; substantially higher for women in India. |