Before We Begin — Where We Left Off
Part Eight was the least argumentative thing in this series. It was a ledger. I stopped arguing about whether the changes in women's lives over the last century were good or bad, and simply counted what had actually changed, and then added a second column that most people leave out: who it reached.
The counting showed something uncomfortable in both directions. A great deal really did change. Girls' schooling, women in paid work, laws on property and violence and divorce, deaths in childbirth, seats in legislatures — on almost every line the number moved, and on most of them it moved a long way. But the second column showed that the movement was not spread evenly. Many of the biggest gains landed hardest on women who were already urban, already schooled, already from families with money or standing. For a woman in a poor rural district, several lines on the ledger had barely moved at all. And on a few lines — unpaid work at home, safety, the amount of the day a woman gets to decide for herself — the number had hardly moved for anybody.
That part ended with one entry left blank. Every ledger has a column for what a thing cost, and I said I was not going to fill it in there, because the cost of the changes in Part Eight is not paid by the women in the ledger. It is paid, if it is paid at all, by people who do not exist yet.
This part fills that column in. It is about the one consequence of the last century that nobody planned, nobody voted for, and almost nobody predicted: across most of the world, and now across most of India, women are having fewer children than are needed to replace the people already here. That fact is now being used, very loudly, as an argument. It is being used to say the old rules were right after all. Before anybody can judge that argument, we have to know what the number actually is, what it measures, why it fell, and what happens next. That is this part.
How this document is built
If you have read the earlier parts you already know the notation. If this is the first one you have picked up, here is the whole system. There are six kinds of box. Each does one job, and each looks different so you can see at a glance what you are about to read.
The first kind explains a hard word the moment it turns up, so you never have to carry an unexplained term forward.
Demography: the study of populations by counting them — how many people are born, how many die, how many move, and how old everybody is. It is one of the least glamorous fields in the social sciences and one of the most reliable, because most of what it studies can be counted rather than argued about.
Why it matters here: nearly every claim in this document rests on a count, which means you can usually check who did the counting and how.
The second kind takes a number too big to picture and turns it into something with a body — a room, a wage, a walk, a week.
India's population is around 1.46 billion people. That number means nothing to anybody. Try it this way instead: if you stood at a window and counted one Indian per second, without sleeping, without stopping, you would finish in about forty-six years.
The third kind shows you the actual evidence behind a claim, and then tells you what that evidence cannot show. It is the box that lets you decide how much to trust a sentence.
India does not learn its birth rate from the census. It learns it from the Sample Registration System, a survey that has watched the same set of about 8,800 sample areas continuously since the 1970s, recording every birth and death in them twice over — once by a local recorder living there, once by a visiting official — and then comparing the two lists.
What it is good at: consistency. The same method, year after year, so a change over time is a real change and not a change of ruler. What it is bad at: small groups. It does not report by caste, income or religion, so several arguments in this document have to be settled with other data or not settled at all.
The fourth kind is for places where informed people genuinely disagree. Each side gets its best case, not a version I find easy to knock down.
Every one of these has the same shape. A question, then the sides, then a verdict that does not pretend to more certainty than exists.
Its strongest case, put the way its best advocate would put it, with the evidence it actually has.
The same, with equal care. If a position sounds stupid here, I have failed to understand it, not proved it wrong.
What would settle it: the evidence that would decide it. Sometimes the honest answer is that nothing available would, and saying so tells you something.
The fifth kind is the one this series is built around. It does not argue with either side. It digs out the thing both sides are quietly assuming without noticing.
These are not caveats and not disclaimers. They are unexamined premises sitting underneath an argument everybody is having. There are five of them in this part. One of them is turned on this document itself, near the end.
The sixth kind closes every chapter. It restates the whole chapter in the plainest words available. If you read only these, you should still have the entire argument of this part.
Six boxes. Word explains. In Real Terms converts. How We Actually Know This shows the evidence. The Argument gives every side its best shot. The Hidden Assumption digs underneath. Remember This closes the chapter.
The rule I hold myself to: simple words, serious content, nothing left out.
A warning specific to this part
Two, actually.
The first is about me. I am an Indian man, in my twenties, unmarried, with no children. This part is about how many children women have. I have never been on the receiving end of a single one of the costs I am about to describe, and I never will be. That does not disqualify me from writing it — but it does mean there is one particular mistake I am likely to make, which is to treat the number of babies as an interesting arithmetic problem rather than as several hundred million individual pregnancies happening inside individual bodies. Where I catch myself doing it, I will say so in the text. Chapter Nine is where this matters most, and I flag it again there.
The second is about the material. Fertility numbers are now political ammunition in India, in Europe, in America and across East Asia. They are used to argue about immigration, about religion, about feminism, about which states get how many seats in a parliament. That means a great deal of what is published about them is written to reach a conclusion. I am going to give you the numbers, tell you who produced them, and tell you where the honest disagreements are. Where a figure is contested I will give the range and explain why the ends of it differ. Where I do not know, I will say so, and Chapter Ten is nothing but that.
One more thing before we start. This part contains no argument that any woman should have more children, and no argument that any woman should have fewer. That is not modesty. It is the method. My job is to show you what is actually happening and what each side actually has, and then get out of the way.
1The Number That Decides Everything
One figure is doing all the work in this argument. Almost nobody who quotes it knows what it measures. It is not a count of anybody's children, and it is not a record of what any woman decided.
1.1 — A rate is not a count
You will have seen the number. Someone says India's fertility rate has fallen to 1.9. Someone else says South Korea's is 0.8. A third person says that below 2.1 a country is finished. The number gets thrown around in newspaper headlines, in speeches by chief ministers, in arguments online about feminism and immigration and the end of the West.
Almost everyone quoting it thinks it means something like "the average woman is having 1.9 children." That is close enough to be useful and wrong enough to matter. Before we go any further we have to be exact about what is being measured, because half the arguments in the rest of this document turn on the difference.
Total Fertility Rate (TFR): take every woman in a country in a single year. Sort them by age. For each age, work out what share of the women that age gave birth this year. Then add all those shares together, from age fifteen to age forty-nine.
The answer is the number of children one imaginary woman would end up with if she lived her whole life through, and if at every age she behaved exactly like the real women of that age did this year.
Why it matters: that woman does not exist. She is a construction. The TFR is not a count of anybody's family. It is a summary of one year, dressed up to look like a life.
Read that last line twice, because it is the source of most of the confusion in this field. When a newspaper says the fertility rate is 1.9, no woman anywhere has had 1.9 children. What has happened is that this year's births, spread across this year's women of each age, add up to a pattern which — if frozen and run through a whole lifetime — would produce 1.9 children.
The figure comes from adding up age-specific fertility rates. In India's Sample Registration System for 2023, the highest of those rates is among women aged 25 to 29, at about 137 births per thousand women that age in that year. The next highest is the 20-to-24 group, at about 108.
What that tells you: roughly fourteen out of every hundred women aged 25 to 29 gave birth during 2023. What it cannot tell you: whether those were the same women who gave birth two years earlier, whether they intend to have more, or whether the women who did not give birth are childless or simply not this year.
Independent confirmation exists. Hospital delivery records, immunisation registers and the household surveys run every few years by the health ministry produce the same broad picture. When a survey, a hospital ledger and a village recorder all agree, the number is real.
So the number is a snapshot, not a biography. Hold on to that. Now to the line everybody quotes.
1.2 — Why the line sits at 2.1, and not at 2
Two people make a child. So you might think that two children per woman would keep a population exactly level. It does not, for two reasons, and both of them are arithmetic rather than opinion.
The first is that slightly more boys are born than girls. Everywhere in the world, without anybody arranging it, about 105 boys are born for every 100 girls. So to replace one woman with one woman, you need a bit more than two births.
The second is that not every girl born survives to have children of her own. In a country where almost every child lives, that adds only a small amount. In a country where children still die, it adds a lot.
Replacement level: the fertility rate at which each generation exactly replaces the one before it, ignoring migration.
In a rich country with very low child mortality it is about 2.1. In a country where children still die in significant numbers it can be 2.3, 2.5, or higher. The commonly quoted 2.1 is not a law of nature. It is a figure that assumes almost every girl born will live to adulthood.
This matters more than it sounds. When someone says India crossed below replacement, they are using 2.1 as the line. Fifty years ago, India's true replacement level would have been noticeably higher than 2.1, because more Indian children died. The line has moved down towards us at the same time as the number has fallen towards it. Both halves of that are the same underlying story: children stopped dying.
1.3 — The snapshot problem
Now the hardest idea in this chapter, and one you need in order to read Chapter Six honestly.
Suppose every woman in a country decides to have exactly two children, but the whole country decides to start five years later than their mothers did. In the years while everyone is postponing, very few births happen. The TFR collapses — perhaps to 1.2. Newspapers announce a crisis. Then the postponement finishes, those women start having their two children each, and the TFR climbs back up.
Nobody changed their mind about how many children to have. Only the timing changed. But the yearly number swung wildly, because the yearly number is a snapshot of a moving thing.
Period fertility: the snapshot. All the births in one calendar year, across women of every age. This is what the headline TFR is.
Cohort fertility: the real thing. Take all the women born in, say, 1975. Wait until they are fifty. Count how many children they actually had. That is a true average, and it has one enormous drawback: you have to wait thirty-five years to learn it.
Tempo effect: the distortion the snapshot suffers when people are shifting the timing of births. Postponement pushes the period number down below the truth. A catch-up pushes it up above the truth.
Demographers have known this since the 1990s, and there are standard adjustments for it. The most widely used, developed by John Bongaarts and Griffith Feeney in 1998, strips out the timing shift and estimates what the number would be if nobody were postponing. Almost always, the adjusted figure is higher than the headline figure — usually by a tenth to three tenths of a child.
Across Europe, women born in the 1970s are finishing their reproductive lives with a real, completed average of roughly 1.7 to 1.8 children each. During the same decades, the headline period rate for those countries was often quoted at 1.4 or 1.5.
The gap between those two figures is roughly one extra child for every four women. That is not a rounding error. It is the difference between a slow decline and a collapse, and it is entirely produced by the fact that women were having their children later, not by their having fewer.
Now the honest complication. Tempo adjustment does not make the problem go away, and it can be misused in the other direction. Postponement is not free. A woman who plans two children at thirty-eight does not always get two children at thirty-eight. Some of the postponed births are never made up. So the truth sits between the two figures: the headline number overstates the fall, and the adjusted number understates it.
This is the first real disagreement in this document, and it sits underneath everything that follows. When Korea reports 0.72, or India reports 1.9, how much of that is a genuine drop in how many children women will end up with, and how much is an artefact of a moving clock?
The period rate is what it is, and it is the number that determines how many babies are actually born this year. Schools, hospitals, pension funds and labour markets do not respond to a woman's eventual completed family — they respond to how many children are actually in the country. Postponed births are also births that are not happening now, and a birth delayed twelve years is a generation lengthened by twelve years, which itself slows population growth even if the total is unchanged. And in the very low countries the postponement explanation has run out of room: Korean women have been postponing for thirty years, and at some point a permanent delay is simply a decline.
Almost every dramatic fertility headline of the last twenty years has been at least partly a timing illusion, and the illusion runs in a predictable direction. Look at what happened in Central Europe: the celebrated fertility rises of the 2010s largely vanish once you adjust for the slowing of postponement, which means the rises were not real either. If the adjustment cuts against your side sometimes and for it other times, it is a measurement correction and not a debating trick. Judge a country by the completed families of women who have finished, not by a synthetic woman who has never lived.
What would settle it: waiting. In about fifteen years we will know the completed family size of women born in the 1990s, and the argument will be over for that generation. There is no way to get the answer sooner, which is precisely why the argument persists.
Why people care so much: because the size of the number decides whether governments are facing a manageable adjustment or an emergency, and emergencies justify measures that adjustments do not.
That argument is about measurement. There is a bigger thing being assumed underneath it, by both sides, and neither of them says it out loud.
Listen to how the fertility rate gets discussed, by anyone, on any side. "Women are choosing to have fewer children." "Young people have decided family is not for them." "She is choosing a career instead." Even the people who disagree furiously about whether this is good or bad agree completely that what the number records is a choice.
But the TFR does not measure choices. It measures outcomes. Between the choosing and the outcome sit a great many things that are not choices at all: whether you found a partner, whether the partner wanted the same thing, whether either of you could afford a flat, whether the pregnancy took, whether the pregnancy held, whether the job survived it, whether anybody was available to mind the child, whether your body was still fertile by the time the rest of your life had arranged itself.
The surveys make the gap visible, and it runs both ways. People in low-fertility countries typically say they want about two children and end up with roughly half a child fewer. In India, about one woman in five reports having had more children than she intended. Chapter Nine goes through that evidence properly.
The general form: reading a result as an intention. The same mistake appears everywhere. Exam results are read as effort. Wealth is read as choice. A number that emerges from thousands of constraints is handed back to the person as though they had picked it off a menu.
Why this one matters so much: if the number is a choice, then the correct response is persuasion or pressure, and both are aimed at women. If the number is an outcome, then the correct response is to look at the constraints. Every argument in Chapters Six and Nine is really a fight about which of those two it is.
1.4 — Why the population keeps growing anyway
One last piece of arithmetic before we look at what actually happened, because it explains something that confuses almost everybody.
India's fertility rate has been at or below replacement for several years. India's population is still growing, and is projected to keep growing for decades, peaking somewhere around 1.7 billion. How can both be true?
Population momentum: the tendency of a population to keep growing after fertility has fallen to replacement, because of the shape of its age pyramid.
Picture it with a family. Suppose your grandmother had six children, and each of them had two. There are now far more women of childbearing age in your family than there were in hers, even though every one of them is having a small family. The total number of babies being born in the family goes up, not down, for a whole generation.
India is that family. It has a very large number of young women, because their mothers' generation was large. Even at 1.9 children each, that produces a great many births.
Momentum is why the two commonest claims about India — "we have far too many people" and "we are about to run out of people" — are both being made at the same time by serious people, and why both can point at real numbers. One is describing the size of the population. The other is describing its direction. They are different facts.
It also means something else, which matters for everything in Chapter Seven. Momentum runs in reverse too. Once the large generations pass through, a country with few young women has few births even if each of them has three children. By the time a shrinking population becomes obvious, the decisions that caused it were taken thirty years earlier. There is no fast lever. That is not a political claim; it is a property of how generations work.
The Total Fertility Rate is not a count of anybody's children. It is a snapshot of one year, arranged to look like a lifetime. The imaginary woman it describes does not exist.
The replacement line sits at about 2.1 rather than 2.0 because slightly more boys are born than girls, and because not every girl survives to adulthood. In poorer countries the true line is higher.
The snapshot is distorted by timing. When people postpone childbearing, the yearly number falls below the truth. Adjusting for that usually adds a tenth to three tenths of a child. It does not make the decline disappear, but it does make it smaller than the headlines say.
Momentum means a population keeps growing for decades after fertility falls, and keeps shrinking for decades after it rises. Nothing here responds quickly to anything.
The number everybody is arguing about is not a record of what women decided. It is a record of what happened.
2What Actually Happened
In one human lifetime the world's fertility rate has roughly halved. It has happened in rich countries and poor ones, in Catholic countries and Muslim ones, under communism and under capitalism, and almost nobody predicted the speed.
2.1 — The whole world, in one paragraph
In the early 1950s, the average woman on Earth had about five children. As of the mid-2020s she has about 2.25. That is the entire story in two numbers, and it is one of the largest changes in human behaviour ever recorded.
It did not happen evenly. It started in Europe in the nineteenth century and moved slowly. It reached the rest of the world in the second half of the twentieth century and moved fast. In some countries the fall from six children to two took a hundred and fifty years. In others it took twenty-five.
Take a woman in Iran born in 1955. Her mother had six or seven children. Her granddaughter is likely to have one, or none. Three generations — grandmother, mother, daughter — and the family has gone from a crowded house to an empty one.
Now put a number on the speed. Iran's fertility rate fell from about 6.5 in the mid-1980s to about 2.0 by the year 2000. That is the same distance England travelled between roughly 1850 and 1970, done in fifteen years.
The United Nations estimates that one in four people on Earth now lives in a country whose population has already peaked and begun to fall. That group includes China, Japan, Germany, Italy and Russia. The world total is still rising, and on the UN's central projection will keep rising to a peak of about 10.3 billion in the middle of the 2080s, from about 8.2 billion in 2024. After that it goes down.
There are two kinds of evidence behind these numbers and they have different weaknesses.
Vital registration. Rich countries record every birth as a legal event, because a birth certificate is needed for school, passport and inheritance. These numbers are close to exact. Korea can tell you how many babies were born last September.
Sample surveys. Poorer countries estimate from surveys — the Demographic and Health Surveys, and in India the Sample Registration System and the National Family Health Survey. These have real error bars, particularly for small states and short time periods.
What both are bad at: the very recent past. Anything from the last twelve months is provisional and gets revised. What they are good at: direction. When registration data, survey data and school enrolment numbers all point the same way for a decade, the direction is not in doubt even if the second decimal place is.
2.2 — The bottom of the table
The lowest fertility rates ever recorded in peacetime are being recorded now, and almost all of them are in East Asia.
South Korea is the extreme case and deserves its own paragraph, because it has now done both things — fallen further than anyone and then started to come back. Its rate fell to 0.72 in 2023, the lowest national figure ever recorded anywhere. Then it rose to 0.75 in 2024 and to about 0.80 in 2025, on provisional figures. In Seoul, the capital, the 2025 figure was 0.63. Chapter Six deals with what that rebound does and does not mean; for now, note only that even after two years of rising, Korea sits at roughly a third of replacement.
Around it: Taiwan and Hong Kong are below 0.9. China is somewhere around 1.0, and its population has been falling since 2022. Japan is at about 1.2 and has been below replacement since 1974 — half a century. Singapore, which has run pronatalist policy since 1987, fell below 1.0 for the first time in 2023.
Europe is less extreme but the direction has surprised people who thought the Nordic countries had solved it. Finland's rate fell from about 1.87 in 2010 to below 1.3. Norway fell from about 1.98 to around 1.4. These are countries with a year of paid parental leave, cheap universal childcare and among the highest rates of women in paid work in the world. Whatever the answer is, generous family policy alone is not it. That fact is inconvenient for two different political camps at once, which is a good sign it is worth taking seriously.
Lowest-low fertility: a term coined by demographers in the year 2000 for a total fertility rate at or below 1.3. At the time it was thought of as a rare, temporary state that a few countries had stumbled into.
Why it matters: at 1.3, each generation is about 40 per cent smaller than the one before it. Two generations of that and the number of young people has roughly halved. When the term was invented, a handful of countries were there. Several are now well below it, and 1.3 has stopped being the floor of the conversation and become the middle of it.
The United States is at about 1.6, France at about 1.6 to 1.7 after decades as the highest in Western Europe, Italy and Spain around 1.2. None of these is the anomaly. Israel is the anomaly, at about 2.9, and it is the only rich democracy that has never fallen below replacement. Chapter Six comes back to it, because it is the single most important test case anyone has.
2.3 — The middle of the world caught up faster than anyone expected
The part that broke the old textbooks is what happened in countries that were still poor.
Thailand, a middle-income country with no coercive population policy, is now around 1.0 to 1.2, lower than most of Europe. Iran, an Islamic republic that once actively encouraged large families, is around 1.6. Brazil, Turkey, Chile, Colombia and Tunisia are all below replacement. Bangladesh, the standard example of hopeless overpopulation in the 1970s, is at about 2.1.
This mattered intellectually. The old model said fertility falls when a country gets rich. These countries fell before they got rich, so whatever drives this is not simply money.
The demographic transition: the standard model of how populations change. Stage one: many births, many deaths, population flat. Stage two: deaths fall first (because of clean water, vaccines and food), births stay high, population explodes. Stage three: births fall too. Stage four: few births, few deaths, population flat again at a much larger size.
Why it matters: every country that has gone through it has followed the same sequence. What the model got wrong was the timing of stage three, which has arrived earlier and moved faster almost everywhere than the people who built the model expected. It also assumed stage four was a resting place. It is now clear that many countries do not stop at replacement. They go straight through it.
2.4 — India is fourteen different countries
Now to India, where the national number hides more than it shows.
The Sample Registration System reported India's total fertility rate at 1.9 in 2023, down from 2.0 in 2021 and 2022. By Indian official estimates the country crossed below replacement around 2019. Urban India crossed in 2004 — twenty years earlier. In 2023, for the first time, rural India reached 2.1 as well.
Underneath that national 1.9 the range is enormous. Delhi reported 1.2. West Bengal and Tamil Nadu reported 1.3. Maharashtra 1.4. Andhra Pradesh and Telangana are around 1.6 to 1.7. Kerala has been below replacement since the late 1980s. At the other end, Bihar reported 2.8. Eighteen states and union territories are now below the replacement line, and a handful of northern states are holding the national average up almost single-handedly.
Tamil Nadu at 1.3 is at roughly the same fertility level as Italy. Delhi at 1.2 is at roughly the same level as Spain. Bihar at 2.8 is at roughly the level of Egypt.
Put it another way. Two women meet on a train between Patna and Chennai. Statistically, the first will have about twice as many children as the second. They hold the same passport, vote in the same national election, and are governed by the same central laws on marriage and inheritance. Almost nothing else about their reproductive lives is the same.
The urban–rural split says the same thing from another angle: urban India is at 1.5, rural India at 2.1. And the strongest single predictor of all — stronger than religion, stronger than income — is how long the mother stayed in school. Chapter Five takes that apart properly.
There is one more Indian number that belongs here, because it is the hinge of Chapter Seven. Those below-replacement states are also the ones that are ageing fastest. In Kerala, about fifteen in every hundred people are already over sixty. Nationally the figure is under ten. The parts of India that succeeded at family planning are the parts that will hit the pension problem first, and they will hit it while the country as a whole still thinks of itself as young.
2.5 — The part of the world that has not done it yet
Sub-Saharan Africa is the exception, and its numbers are the biggest single source of uncertainty in every long-run projection on Earth. The region's average is around 4.3. Niger is close to 6. The Democratic Republic of Congo and several neighbours are above 5. Nigeria, with more than 200 million people, is around 4.5.
Fertility there is falling, but slowly, and the pace of the fall is genuinely disputed.
This sounds like a technical question for demographers. It is not. Whether world population peaks at 9 billion in 2060 or 10.3 billion in 2085 depends almost entirely on the answer, and so does every argument in Chapter Eight about whether humanity is about to shrink.
Every region that has begun this transition has completed it more quickly than the region before it, because the tools arrive ready-made. Contraception, girls' schooling, mobile phones and urban living all arrive in Africa simultaneously rather than over a century. Iran did it in fifteen years. Bangladesh did it while poor. Recent survey rounds in several African countries have come in below what the UN projected, and the UN has revised its global peak down by about 700 million people compared with what it was saying a decade ago. The revisions have all gone one way.
The countries with the highest fertility are also the ones with the weakest schooling systems, the least urbanisation, the youngest age at marriage and the highest child mortality — and child mortality is the strongest single brake, because parents who expect to bury children have more of them. Desired family size in surveys across much of West and Central Africa remains genuinely high, at four, five or six children, which is not a supply problem that contraception fixes. Projecting a fast African decline mostly assumes that the future will not look like the past twenty years there, and that is an assumption, not a finding.
What would settle it: the next three rounds of Demographic and Health Surveys, over roughly the next decade. This is one of the rare arguments in this document where the settling evidence is actually coming, on a known schedule.
Why people care so much: because one answer produces a story about a crowded, young, African-majority future and the other produces a story about global depopulation. Both stories are already being used in immigration politics in Europe, and both sides reach for the projection that suits them.
One honest note about all projections, including the ones I have just quoted. Demographers have a poor record on fertility and a good one on mortality. Deaths are predictable; people age at a fixed rate. Births are not. Almost every major projection of the last seventy years has been wrong about fertility, usually by assuming the current trend would flatten out when it did not. Treat the population peak numbers in this document as the best available guesses and not as facts.
In one lifetime, the average number of children per woman worldwide has fallen from about five to about 2.25. This is one of the biggest behavioural changes ever recorded, and it happened almost everywhere.
The lowest figures are in East Asia. Korea hit 0.72 in 2023, the lowest ever recorded anywhere, and has since risen to about 0.80. Even the generous Nordic countries have fallen sharply, which means money and childcare alone are not the answer.
It happened in poor countries too, and fast. Iran fell from 6.5 to 2.0 in fifteen years. Thailand and Brazil are below most of Europe. The old idea that fertility falls only when a country gets rich is dead.
India is not one country on this measure. Delhi is at 1.2 and Bihar at 2.8. Eighteen states are below replacement. Urban India crossed the line in 2004; the nation crossed around 2019.
Sub-Saharan Africa is the only region still well above replacement, and how fast it falls decides what the whole world's population does this century.
3Why It Fell
Seven explanations are offered, and they are usually offered as rivals. Most of them are true. The interesting question is not which one is right but which one arrives first, because that determines what you would have to undo.
3.1 — Children stopped dying
This is the first cause, the biggest, and the one people leave out of political arguments because it is not politically useful to anybody.
For almost all of human history, a very large share of children died before they were five. In many places it was one in three. A woman who wanted two grown-up children had to give birth four, five or six times, and bury the difference. High birth rates were not a sign of enthusiasm. They were a response to death.
Then, over roughly a century, that stopped. Clean water, sewers, vaccines, antibiotics, oral rehydration for diarrhoea, more food. India's infant mortality rate has fallen to about 25 deaths per thousand live births, from more than 140 in the 1970s.
An Indian woman giving birth in 1970 could expect roughly one in seven of her babies to die before their first birthday. Today it is about one in forty.
Turn that around into the decision she faces. In 1970, to be reasonably confident of ending up with two adult children, she needed to give birth three times. Today she needs to give birth twice. The same want, the same worry, produces one fewer pregnancy. Not one woman in that story changed her mind about anything.
This is why the demographic transition happens in that order everywhere: deaths fall, then births fall, with a lag of a generation or two. The lag is the whole problem — it is the population explosion. It is also why a country's fertility decline usually cannot be stopped by persuasion. What is being adjusted to is a change in survival, and nobody wants to undo that.
The strongest evidence is not a correlation, it is a set of long-running field studies. The best known is at Matlab in Bangladesh, where since 1966 a research station has recorded every birth, death and move in a set of villages, house by house, for sixty years. In 1977 half the villages were given an intensive family planning and maternal health programme and half were not, and both halves have been followed ever since.
What it shows: fertility fell in the treated villages faster than in the untreated ones, so services matter. But it also fell substantially in the untreated villages, so services are not the whole story. And in both halves, the fall followed improvements in child survival rather than leading them.
What it cannot show: whether the same would happen in a country with a different family structure. One district in Bangladesh is one district in Bangladesh. Its great value is that it is real, long, and not reconstructed from memory.
3.2 — The child stopped being an asset
On a farm, a child of eight can weed, carry water, mind goats and watch a younger sibling. By twelve a child is close to an adult worker. Children cost food and produce labour, and past a certain age the labour is worth more than the food.
In a city, with a school, a child is a pure cost for twenty years. Fees, books, uniforms, tuition, a room, and — this is the part people underestimate — an adult's attention, every day, for two decades.
The quantity–quality trade-off: an idea from the economist Gary Becker in the 1960s. Parents are not choosing only how many children to have. They are choosing how much to invest in each one. Once investing more in each child starts to pay — because schooling leads to jobs — parents shift towards fewer children with more spent on each.
Why it matters: this predicts that fertility falls fastest exactly where education starts to pay off, which is what the data shows. It also predicts something less comfortable: the shift is not a retreat from family. It is a family strategy. Fewer children is what caring about your children looks like once schooling is the route out.
You can watch this happen inside a single Indian family across two generations. The grandfather has five children and no fees to pay. The son has two children and pays for coaching classes for both. Ask either of them whether they love their children and you get the same answer. The arithmetic underneath is completely different.
For an urban household with one modest salary, a private school place, books, uniform, transport and a couple of tuition subjects can absorb most of what is left after rent and food. Two children can be carried. A fourth cannot, unless all four get the cheaper version.
So the trade-off is not between children and no children. It is between two children with coaching and four children without. Almost every Indian family making that calculation now makes it one way, their parents made it the other way, and neither generation was being foolish.
3.3 — Schooling, and what it actually does
Across almost every dataset ever assembled, the strongest single predictor of how many children a woman has is how long she stayed in school. It beats income. It usually beats religion. In India's own health surveys the gap between women with no schooling and women who finished twelve years is often more than a full child.
The interesting question is why, because at least four different mechanisms are hiding inside "education", and they have different implications.
Schooling delays marriage, which shortens the years available for childbearing. It teaches literacy, which changes what health information a woman can get and act on. It changes what a woman believes she is allowed to want, and what she can argue for at home. And it raises what her hour of time is worth on the open market, which is the mechanism the economists care about most.
Opportunity cost: what you give up in order to do a thing. Not what it costs in money — what it costs in the next-best option you had to drop.
Applied here: for a woman with no schooling and no paid work available, the opportunity cost of a pregnancy is close to zero, because there is nothing else she would have been paid for. For a woman with a degree and a salary, a pregnancy may cost several years of earnings, a promotion, and a permanent step down in what she is paid for the rest of her working life.
Why it matters: this is the single most important idea in the chapter. Nothing about her desire for children has changed. The price has changed. And the price rose precisely because her options improved.
That last line is uncomfortable for both camps in the argument this series is about, which is why I want it stated plainly. If you believe expanding women's options was right, you have to accept that the fertility decline is partly a direct result of it, and not an unrelated accident. If you believe the fertility decline is a disaster, you have to accept that reversing it means either lowering the price of children or lowering the value of a woman's alternatives. Chapter Nine is where that fork gets examined properly, and there is no third road that anybody has found.
3.4 — Contraception: necessary, not sufficient
The pill, the intrauterine device, sterilisation, the condom. Obviously these matter. A woman cannot act on a decision to stop having children if she has no means of stopping.
But contraception on its own explains much less than people expect, and there are two clean pieces of evidence for that. First, fertility fell dramatically in nineteenth-century France and in parts of Europe long before modern contraception existed, using withdrawal, abstinence and later marriage. Second, in the Matlab study above, fertility fell in the villages that got no programme too.
The honest way to put it: contraception is the tool, not the motive. Where people want fewer children and have no tools, they find worse tools and use them, at real cost to women's health. Where people want more children, handing out contraception does very little. India's own experience proves the point from the ugly side — the coercive sterilisation drive of 1975 to 1977 produced millions of operations and a political catastrophe, and India's fertility decline continued afterwards at roughly the pace it would have anyway.
3.5 — The pension that used to have a face
Here is a cause that gets almost no attention outside demography, and it is the one that connects this chapter to Chapter Seven.
For most of human history, children were the retirement system. There was no other. You had sons so that somebody would feed you at seventy, and you had several because some would die and some would leave. In much of India this is still literally true, and it is still enforced — India has a law, the Maintenance and Welfare of Parents and Senior Citizens Act of 2007, that obliges children to maintain their elderly parents and gives parents a tribunal to go to if they do not.
When a state pension arrives, that motive weakens. You no longer need a son to survive old age. Several careful studies across different countries have found that the introduction or expansion of public pensions is followed by measurable falls in fertility.
Note what this means. A society that builds an old-age safety net reduces the reason to have children, which reduces the number of workers available to fund the safety net. That is not an argument against pensions. It is a genuine circularity sitting inside the design of every modern welfare state, and almost nobody planned for it.
Everything above is agreed to matter. The dispute is about which cause does the heavy lifting, because that determines what a government could even in principle change.
Fertility falls when child mortality falls, cities grow, schooling spreads and children become expensive. The mechanism is material, it is the same everywhere, and the sequence is remarkably consistent across a hundred and fifty years and every continent. The best evidence is the sheer regularity: countries as different as Sweden, South Korea, Iran and Brazil have followed the same curve at different times. That is not a coincidence of culture. It is a response to conditions.
Development cannot explain the speed. Iran did in fifteen years what England did in a hundred and twenty, without a comparable rise in income. Fertility fell in Thailand and Bangladesh while both were still poor. What actually spreads is the idea that a small family is normal, respectable and achievable — carried by radio, then television, then the phone. The classic finding here is that fertility decline in nineteenth-century Europe followed language and cultural boundaries more closely than it followed income. It spread the way a fashion spreads, not the way a price responds.
Both of the above are describing the same underlying thing badly. What changed is that women acquired alternatives. Every factor listed — schooling, cities, jobs, contraception, later marriage — raises what a woman gives up by having a child. On this account fertility did not fall because families got richer or because ideas travelled, but because the cost of childbearing shifted onto a person who now had something else to do with her twenties. The strongest evidence is that within every country, at every income level, the women with the most alternatives have the fewest children.
What would settle it: nothing clean, and this is worth being honest about. You cannot run the experiment. The closest thing available is careful comparison of places where one factor moved and the others did not — a state that expanded schooling without industrialising, a region that got television before it got roads — and that work exists but is patchy and contested.
Why people care so much: because each answer implies a different lever. If it is development, nothing can be done and nothing should be. If it is ideas, then propaganda and cultural revival might work, which is what religious and nationalist movements believe. If it is the price of a woman's time, then raising fertility means changing something about women's options or about who bears the cost of children — and those are very different policies with very different politics.
Underneath that whole argument, all three positions share a picture of who is doing the deciding, and it is worth stopping on.
Every model above imagines a woman, or a woman and a man, weighing costs and benefits and arriving at a family size. Economists model it as a household budget. Sociologists model it as a couple negotiating. Traditionalists model it as a couple failing in their duty. Feminists model it as a woman finally being allowed to decide. All four are assuming the same thing: that the decision is made by the people who will raise the child.
For most of history, and for a very large number of Indian households right now, it is not. In a joint family, whether and when a young wife becomes pregnant has often been decided, or heavily steered, by her mother-in-law and by her husband's family — for reasons to do with property, labour, the succession of the household, and the standing of the family in the village. Studies of contraceptive use in north India repeatedly find that the mother-in-law's view predicts the daughter-in-law's behaviour better than the daughter-in-law's own stated preference does. The United Nations survey mentioned in Chapter One found one Indian woman in five reporting more children than she intended, and the most common reason given was what her community expected.
So when a country urbanises and young couples move away into their own flats, something happens that has nothing to do with income, schooling or contraception: the decision changes hands. A whole layer of people who wanted more grandchildren is removed from the room. Some part of the fertility decline is not a change of mind at all. It is a change of who holds the pen.
The general form: assuming that the modern decision-unit was always the decision-unit. It appears everywhere — in economics, where the household is treated as a single mind; in law, where the couple is treated as the family; in policy, where a cash incentive is aimed at a person who may not be the one deciding.
What follows is practical. A payment offered to a young couple for a third child is being offered into a negotiation whose other participants are invisible to the policy — which may be part of why such payments work as poorly as Chapter Six shows.
Put all seven causes together and they point one way. Every single one of them made a child more expensive — in money, in years, in attention, or in what somebody had to give up to have one. Nobody legislated that. It is what a century of falling child mortality, spreading schools and growing cities does to the arithmetic of a family.
Children stopped dying. That is the first and biggest cause. When a woman no longer has to give birth five times to raise three adults, she gives birth fewer times without changing a single thing about what she wants.
Children stopped being workers and became investments. On a farm a child earns; in a city with a school a child costs for twenty years. Families shifted from many children to few children with more spent on each. That is a family strategy, not a rejection of family.
Schooling is the strongest single predictor — stronger than income, usually stronger than religion. It works by delaying marriage, changing what a woman knows, changing what she can argue for, and raising what her time is worth elsewhere.
Contraception is the tool, not the motive. Fertility fell in Europe before modern contraception existed, and India's coercive sterilisation drive of the 1970s did not change the underlying trend.
Every explanation on the table describes the same event from a different angle: having a child got more expensive, and the person it got most expensive for was the mother.
4The Curve That Was Supposed to Come Back Up
For about a decade there was a comforting answer: development pushes fertility down, and then, past a certain point, pushes it back up again. The comfort has largely gone. What replaced it is more interesting and much harder to act on.
4.1 — The paper that gave everyone hope
In August 2009 three demographers — Mikko Myrskylä, Hans-Peter Kohler and Francesco Billari — published a short paper in the journal Nature that changed the conversation. They plotted every country's fertility rate against a broad measure of its development, and found something nobody expected.
Fertility fell as development rose, exactly as the textbooks said. But past a very high level of development, the line turned. In the most developed countries, further development was associated with fertility going back up. Drawn on a chart, the relationship looked like the letter J lying on its side: a long fall, a bottom, and a lift at the far end.
Human Development Index (HDI): a single score between 0 and 1 that combines three things — how long people live, how much schooling they get, and how much income they have. Invented by the United Nations in 1990 so that countries could be compared on something other than money alone.
Why it matters here: the entire J-curve finding was expressed in terms of the HDI. The turning point was put at about 0.85 to 0.86. That is a level reached by countries like Sweden, the Netherlands and the United States, and not by India, which sits closer to 0.64.
You can see why this landed. It said the fertility problem was a phase. Push development far enough, keep going, and the number comes back. Every government facing a birth-rate panic had a reason to be patient. The paper was cited thousands of times and quoted in newspapers all over the world.
4.2 — What happened to it
Two things went wrong, one technical and one simply factual.
The technical problem came first. In 2014 Kenneth Harttgen and Sebastian Vollmer showed that the upturn was highly sensitive to how you built the index. The HDI's own formula had been changed in 2010. Rebuild the measure a slightly different way, or use a longer run of years, and the lift at the far end shrank or disappeared. A finding that depends on which version of a composite index you use is not a law of nature. Others pointed out that much of the apparent upturn was the tempo effect from Chapter One: several very developed countries had finished a long period of postponement, so their headline numbers rose without their completed families rising.
The factual problem came later and was decisive. After about 2010, and sharply after 2015, fertility in the most developed countries did not carry on up. It fell. Finland, one of the most developed societies on Earth by any measure, went from about 1.87 to below 1.3 in little more than a decade. Norway fell by around half a child. The United States drifted from about 2.1 down to about 1.6. The countries that were supposed to demonstrate the recovery instead demonstrated the opposite.
The evidence in the original paper was a cross-country scatter plot: one dot per country, development on one axis, fertility on the other. This is the most common kind of evidence in this whole field and it is worth understanding what it can and cannot do.
What it is good at: showing that a relationship exists across many places at once, using data everybody can check.
What it is bad at: telling you about people. A pattern between countries does not have to hold inside them. Rich countries may have higher fertility than middle-income countries while, inside every one of those countries, rich people have fewer children than poor people. Both statements can be true simultaneously. Demographers call the mistake of confusing the two the ecological fallacy, and it is the single most common error in popular writing about fertility.
The decisive evidence against the J-curve was not a better scatter plot. It was time passing.
To be fair to the original authors, they described a correlation in the data as it then stood and were careful about it, and Myrskylä himself later published work qualifying it. What was oversold was not the paper but the reading of it — worth remembering whenever a single striking chart becomes a reason not to worry.
4.3 — The theory that replaced it
If development alone does not bring fertility back, what determines where a country lands? The most influential answer came from the Australian demographer Peter McDonald in the year 2000, and it is more subtle than the J-curve.
McDonald's argument was that you should not ask how equal a society is. You should ask whether it is equal in the same way in both of the places a woman lives.
Modern societies have granted women near-equal treatment in what he called individual-oriented institutions — schools, universities, the labour market, the law. A girl gets the same education as her brother; a woman applies for the same job. But in family-oriented institutions — the household, the division of housework, the assumption about who leaves work when a child is ill, the expectations of in-laws — many of those same societies changed far less.
The result is a trap. A woman is educated as an equal and employed as an equal, and then, on becoming a mother, is expected to carry a load her husband is not. She is not choosing between a traditional life and a modern one. She is being offered a modern life on the condition that she also does the traditional one.
Time-use surveys across many countries find the same shape. In India, one of the widest gaps ever measured, women spend several hours a day on unpaid domestic and care work while men spend well under an hour. In Italy, Spain, Japan and Korea — all countries with very low fertility — the gap is narrower than India's but still large.
Put it in a working week. If a woman does a full day of paid work and then four hours of unpaid work at home, and her husband does a full day of paid work and forty minutes at home, she is working roughly a six-day week to his five, every week, for as long as the children are small.
The theory's claim in one sentence: a woman looking at that arithmetic before her first child can see exactly what a second one costs, and she is not wrong.
This theory has one great strength. It explains the pattern the J-curve could not: why the lowest fertility in the developed world is not in the least equal countries but in the ones that are highly equal in public life and stubbornly unequal at home. Italy, Spain, Japan, Korea. And it explained, for a while, why the Nordic countries — the most equal at home as well as at work — had the highest fertility in Europe.
4.4 — The revolution in two halves
A group of researchers led by Frances Goldscheider extended this in 2015 into a framing that is easier to hold in the head. They described the change in women's lives as a gender revolution in two halves.
The first half is women moving into the public world: school, university, paid work, public life. That half is largely done in rich countries and well advanced in urban India. It reduces fertility, because it raises the cost of a woman's time without reducing what is asked of her at home.
The second half is men moving into the private world: nappies, school runs, sick days, the emotional management of a household. On this account, the second half raises fertility again, because it spreads the cost of a child across two people instead of loading it onto one.
So the low-fertility countries are not the ones that went too far. They are the ones that stopped halfway.
4.5 — And then the Nordic countries fell
This is where honesty is required, because the theory has taken a serious blow and its supporters do not always say so.
If the two-halves account is right, the countries furthest through the second half should have held up best. They did not. Finland's collapse from 1.87 to below 1.3 happened during a period when Finnish men's share of domestic work was still rising, not falling. Norway and Sweden fell too. The Nordic exception, which was the single strongest piece of evidence for the gender equity theory, stopped being an exception.
The question matters far beyond academia, because it is the only theory on offer that says a country could raise its birth rate without asking women to give anything up. If it is wrong, that option may not exist.
Nothing in the theory ever promised that equality at home would push fertility back to two. It said unequal loading depresses fertility. Remove that and other things — housing costs, insecure work, the collapse of early partnering — can still push it down. Within countries, the micro evidence still supports it: studies in the Nordic countries and elsewhere find that couples who share domestic work more equally are more likely to go on to a second and third child. Finland's fall coincided with a severe and prolonged economic slump and a steep fall in the number of young people forming couples at all. Judge the theory on whether unequal households have fewer children, which they do, not on whether it can single-handedly hold a national rate up against everything else.
A theory that predicts a rise, and gets a fall in exactly the places where the mechanism was strongest, has failed. The micro evidence is weak protection here: couples who share housework equally may simply be couples who were more committed to each other and to having children in the first place, which makes the correlation a description of the couples rather than an effect of the sharing. And the crucial point is the size of the thing. Even in the most equal societies on Earth, no combination of equal housework, universal childcare and a year of paid leave has held a national rate at replacement. If the strongest version of the treatment does not work, the treatment is not the answer.
What would settle it: a rich country that reaches genuine equality in domestic work — not policy on paper, but measured hours — and holds it for twenty years, while housing and job security stay stable. No such country exists, which is why this argument will not be settled soon. The honest position is that the strongest form of the test has never been run.
Why people care so much: because this is the last place to look for a solution that costs women nothing. Every other lever on the table involves either large sums of public money, which has been tried, or a change in what is expected of women, which is the subject of this whole series. If domestic equality does not do it, the menu gets much uglier.
One more thing has changed since these theories were written, and it may end up mattering more than either of them. In much of East Asia and increasingly in the West, the fertility decline is now largely a partnership decline. People are not having fewer children inside their relationships; they are not forming the relationships. In Korea, births outside marriage remain rare, and the entire recent rise in births tracked a rise in marriages one to two years earlier. If people are not pairing up, no amount of childcare policy reaches them, because policy for parents arrives after the point where the number is being decided. That is a different problem with a different shape, and it is taken up later in this series.
In 2009 a famous paper found a J-curve: fertility falls as countries develop, then rises again at very high development. It was widely read as proof that the problem would fix itself.
It did not hold. The finding was sensitive to how the development index was built, some of the upturn was a timing illusion, and after 2015 the most developed countries fell instead of rising. Finland went from 1.87 to below 1.3.
The theory that replaced it says the problem is uneven equality: women are treated as equals at school and at work, and not at home. A woman carrying a six-day week to her husband's five can see exactly what a second child costs.
The two-halves version of this says the first half of the revolution — women entering public life — lowers fertility, and the second half — men entering domestic life — should raise it again.
But the countries furthest through the second half fell too. Nobody has yet found a way to hold a rich country at replacement, and the honest position is that the strongest version of the cure has never actually been tested.
5Who Is Having the Children
A national average tells you nothing about who is inside it. Once you look, you find that the people having children and the people not having them differ in ways that are being used, right now, as political ammunition in India.
5.1 — The average hides the argument
A country with a fertility rate of 1.9 could be a country where every woman has about two children. It could also be a country where half the women have three and half have one. These are completely different societies with the same headline number, and which one you are living in matters enormously for what happens next.
Differential fertility: the fact that different groups within the same country have different numbers of children — sorted by education, income, religion, region, caste, or anything else you care to measure.
Why it matters: over generations, the composition of a population shifts towards whichever group is having more children, even if nobody's individual behaviour changes. This is arithmetic, not ideology. What is fiercely contested is how strong the effect is and whether it survives contact with the real world, which is what most of this chapter is about.
5.2 — Schooling beats everything else
Start with the biggest differential, because it is the one that gets the least political attention despite being the largest.
In India's health surveys, the fertility gap between women with no schooling and women who completed twelve years or more has consistently run at more than a full child. That is a bigger gap than the one between any two religions, bigger than the gap between most income groups, and roughly the same size as the gap between Bihar and Kerala.
It is also the gap that has been closing fastest, because schooling has spread fastest. Much of India's fertility decline is simply this: the share of Indian women who never went to school has collapsed, and those women are the ones who used to have five children.
India has two big fertility datasets and they answer different questions. The Sample Registration System gives the best year-by-year national and state numbers, but it does not break results down by religion, caste or income. The National Family Health Survey, run every few years across several hundred thousand households, does.
What that means in practice: every claim in this chapter about religion or education comes from the NFHS, and the most recent full round covered 2019 to 2021. So the religious comparisons are a few years older than the state comparisons in Chapter Two, and cannot simply be updated by taking the newer national figure and assuming the gaps stayed the same.
Why the survey is trustworthy on this: it asks women directly about every birth in their lives, and it has repeated the same questions across five rounds since 1992, which makes the trend more reliable than any single round's level.
5.3 — The Indian religion gap, and how fast it is closing
Now the number that gets shouted about in India. I am going to give it to you plainly, with the trend, because the trend is the whole point and it is almost always left out.
In the 2019–21 survey, Hindu fertility was about 1.94 and Muslim fertility about 2.36. So Muslim fertility was higher. That is true, and pretending otherwise helps nobody.
Now the part that is usually omitted. In the survey of 1992–93, the same gap was about 1.1 children. By 2019–21 it was about 0.42. Muslim fertility in India has fallen faster than Hindu fertility over the last three decades — it has fallen further, from a higher starting point, in less time. And in the most recent rounds, Muslim fertility itself has come very close to the replacement line.
Think of two runners. In 1992 one was 1.1 kilometres behind. By 2021 the gap was 0.42 kilometres. The runner behind is running faster, and both are approaching the same finish line.
Here is the fact that settles most of the argument on its own: Muslim fertility in India's southern states is lower than Hindu fertility in India's northern states. A Muslim woman in Tamil Nadu is likely to have fewer children than a Hindu woman in Bihar. If religion were the driver, that could not happen. Where you live, how long you were in school and whether your children survive predict your family size far better than which building you pray in.
This is a place where I should say what I am doing. The Hindu–Muslim fertility gap in India is not primarily a demographic subject. It is a political weapon, used to argue that one community will outnumber another. I am reporting the numbers because a series that claims to face facts cannot skip the ones that are being misused. The trend is the fact that matters: the gap has narrowed by more than half in thirty years and is still narrowing. Anyone quoting the level without the trend is either uninformed or is choosing not to tell you.
5.4 — The inheritance argument
The differential fertility idea has a serious version, and it is worth taking seriously because it is not obviously wrong.
It was set out most fully by the political scientist Eric Kaufmann in a 2010 book asking whether the religious will inherit the Earth. The argument runs like this. Secular, liberal, highly educated people have fewer children than devout, traditional people, in nearly every country where it has been measured. Compound that over several generations and the share of the population raised in devout households rises, regardless of whether anybody converts anybody. The future belongs not to whoever wins the argument but to whoever shows up.
The examples are real. In Israel, the ultra-Orthodox community has fertility far above the national average and its share of the population has risen steadily for decades. In the United States, the Amish population has roughly doubled every twenty years for a century. Across Europe and America, regular religious attendance is one of the few individual characteristics that still predicts having more children.
Retention rate: the share of children raised in a group who still belong to that group as adults.
Why it matters: it is the hinge of the whole inheritance argument. A group with high fertility and low retention grows slowly or not at all, because it is losing at the back door what it gains at the front. A group with high fertility and near-total retention compounds. Almost every fight about the argument below is really a fight about this one number.
With that word in hand, the argument becomes tractable. Both sides below accept the same multiplication. What they disagree about is whether the groups doing the multiplying stay intact long enough for it to matter.
The mathematics is not in dispute. If group A averages three children and group B averages 1.5, then after four generations group A's descendants outnumber group B's by roughly sixteen to one, starting from equal size. The dispute is whether groups stay groups.
The arithmetic is relentless and it does not need anyone to change their mind. Israel's ultra-Orthodox share has risen from a tiny minority to a substantial one within living memory, and it is now large enough to shape national politics, school curricula and the military draft. The same compounding is visible among conservative religious communities in the United States and Europe. Critically, the groups that hold their children are exactly the groups that build separate institutions — their own schools, their own marriage markets, their own neighbourhoods — and those institutions are getting easier to maintain, not harder, because a community can now educate and entertain its children entirely inside its own network. The secular assumption that everyone eventually drifts to the centre was a feature of twentieth-century mass media, and mass media is gone.
Every historical case of high-fertility religious growth has been accompanied by defection, and defection scales with contact. Quebec had among the highest fertility in the Western world and became one of the most secular societies on Earth within a single generation. Iran's fertility collapsed under a religious government that wanted the opposite. The high-retention communities that get cited — the Amish, the ultra-Orthodox — are small, unusual, and maintain their retention through a degree of separation from the wider economy that does not scale. And there is a simpler point: the trait being inherited is not a gene. It is a way of life, and it is transmitted only as long as the children want it. As soon as a community becomes big enough to need the outside economy, its retention rate starts to fall, and the compounding stops.
What would settle it: long-run retention data on large high-fertility communities as they urbanise and enter the wider labour market. Some of this exists for Israel and is genuinely contested, with each side reading the same numbers differently. There is no clean answer available yet.
Why people care so much: because in India, Europe and Israel alike, this argument is the respectable form of a demographic anxiety about who a country will belong to. That is worth naming. The evidential question — how strongly does group fertility compound — is real. But it is being fought at a volume that only makes sense if the actual stake is something else.
5.5 — The other differential, which nobody campaigns about
There is a differential inside every country that gets far less attention than religion, and it is larger.
In almost every society ever measured, the people with the most education and the highest incomes have had the fewest children. This was true in nineteenth-century Britain and it is true in India today. It is the pattern that alarmed the eugenicists a century ago, and their alarm is one reason the topic is handled so nervously now.
Two things have changed recently and they cut in opposite directions. In several rich countries the pattern at the very top has partially reversed: the highest-earning men and, increasingly, the highest-earning couples now have slightly more children than those just below them, because they can buy the labour that makes a third child possible. At the same time, the sharpest fertility falls in rich countries have been among the young and less secure — people without stable jobs or affordable housing.
Put those together and something is happening that the political argument has not caught up with. In several countries, having children is quietly becoming a thing that the secure can afford and the insecure postpone until it is too late. That is a class story, not a culture war story, and it is much less useful to everybody's politics, which may be why you hear about it less.
A national average hides who is inside it. Differential fertility — different groups having different numbers of children — shifts a population's composition over generations without anyone changing their mind.
In India, the biggest fertility gap is education, not religion. Women with no schooling have had more than a full child more than women who finished school.
The Hindu–Muslim gap is real and shrinking fast: about 1.1 children in 1992–93, about 0.42 in 2019–21. Muslim fertility in south India is lower than Hindu fertility in north India. Geography and schooling beat religion.
The inheritance argument — that the devout will out-reproduce the secular — is arithmetically sound and empirically shaky, because it assumes children keep their parents' way of life. That assumption is called the retention rate, and it is the whole ballgame.
The differential nobody campaigns about is the largest one: in country after country, children are becoming something the secure can afford and the insecure postpone until it is too late.
6Has Anyone Ever Turned It Round
Governments have been trying to raise birth rates for a hundred years, and several are spending enormous sums on it right now. This is the record. It is not encouraging, and the one clear exception is not an exception anybody can copy.
6.1 — What "it worked" would have to mean
Before looking at the cases, we need a test, because otherwise every government can claim victory. Almost all of them do.
Pronatalism: any deliberate effort by a state to increase the number of births. It runs from cash payments and tax breaks at one end to bans on contraception and abortion at the other.
Quantum and tempo: two different things a policy can change. Quantum is how many children a woman ends up with in total. Tempo is when she has them. A payment that persuades a couple to have their second child at twenty-nine instead of thirty-two changes tempo. A payment that persuades a couple to have a third child they were never going to have changes quantum.
Why it matters: only quantum changes the long-run size of a population. But tempo is what shows up in the headline rate immediately, and it is what governments announce. Almost every claimed pronatalist success in the last twenty years has turned out, on inspection, to be mostly tempo.
So the test is: did completed family size rise? Not the headline number, not the number of births in one good year — the total that a generation of women ended up with. That test is demanding and slow. It is also the only one that means anything.
6.2 — Hungary: the most-watched attempt in the world
Since 2010, Hungary's government has run the most ambitious pronatalist programme in any democracy. Zero-interest loans for couples expecting a child, with portions of the debt forgiven as more children arrive. Income tax exemption for mothers of four or more children. Housing subsidies tied to family size. Spending on family support of around five per cent of national income — among the highest in the world by that measure, and several times what most rich countries spend.
For a decade it looked like it was working. Hungary's rate rose from about 1.23 in 2011 to about 1.59 in 2021. Conservative politicians across Europe and America pointed at it. Italy's government announced it would copy the approach.
Then it went into reverse. By 2024 the figure was about 1.38. By 2025 it was about 1.31 — not much above where the programme started. Meanwhile, across the eleven years between two censuses, Hungary's population fell by about 334,000 people, because a rise in the rate from 1.23 to 1.59 does not produce population growth. It produces a slightly slower decline.
The most informative analysis of the Hungarian case did not use the headline rate at all. It used the tempo-adjusted rate from Chapter One — the estimate of what the number would be if nobody were shifting the timing of their births.
What it showed: most of the celebrated fertility rises across Central Europe in the 2010s disappear once timing is stripped out. The rises were largely the tail end of a long postponement finishing, not a change in how many children women were having. And on the adjusted measure, Hungary and Poland stand out as the two countries whose underlying fertility genuinely did not recover.
Why this is strong evidence: the adjustment was not invented to embarrass Hungary. It is a standard tool applied across all of Europe at once, and it makes some countries look better and others worse. A correction that cuts both ways is a measurement, not an argument.
The demographer Tomas Sobotka of the Vienna Institute of Demography put it about as bluntly as a specialist ever does: judged against what the policies set out to achieve, this is clearly a failure. Hungary's own government continues to expand the programme.
6.3 — The rest of the case file
Poland. In 2016 Poland introduced a large monthly child benefit, generous by regional standards and paid to a very wide range of families. The rate rose from about 1.29 to about 1.45 within two years. Then it fell, and kept falling, to the lowest levels in Polish history — below where it had been before the benefit existed.
Russia. In 2007 Russia introduced "maternity capital", a large lump sum for a second child, usable for housing or education. The rate rose from about 1.3 to about 1.78 by 2015 — one of the largest rises anywhere. Then it fell back towards 1.4. The most careful studies of the scheme found that a substantial part of what it did was bring forward births that would have happened anyway, and shift them towards the second child specifically, which is what the money was attached to.
Singapore. Pronatalist since 1987, after a famously successful antinatalist campaign before that. Cash bonuses, tax rebates, housing priority, matched savings accounts for children, state-supported matchmaking. The rate has fallen throughout, and dropped below 1.0 for the first time in 2023.
France. Often cited as the success story, and it is the best of them. Decades of consistent family policy — universal childcare places, a tax system that explicitly rewards larger families, long paid leave — kept France at or near the top of Western Europe for forty years, at about 1.9 to 2.0. It has now drifted down to about 1.6. The fair reading is that France bought itself several tenths of a child and forty years, which is more than anybody else managed, and did not stop the direction.
What does a tenth of a child cost? Take a country spending an extra one per cent of national income on family support. Studies that find a positive effect usually put it somewhere between a few hundredths and about two tenths of a child per woman.
Turn that into households. If a country of ten million women of childbearing age raises the rate by 0.1, that is roughly one additional child for every ten women — spread across everybody, over a generation. The spending, meanwhile, goes to every family, including the very large majority who were going to have that child anyway.
This is why estimates of the cost of one additional birth — one that would not otherwise have happened — routinely run into the hundreds of thousands of pounds or dollars in rich countries. The money is not wasted; it goes to families and it makes their lives easier. But as a way of buying babies it is close to the most expensive method available.
6.4 — Korea, and what a rebound looks like from inside
Korea deserves careful handling, because it is the live case and it is being read badly by both sides right now.
The facts first. Korea has spent, by its own government's accounting, more than 280 trillion won on low-fertility policy over about sixteen years — something over two hundred billion US dollars. Cash grants, extended parental leave, childcare, housing support, plan after plan. Through all of it the rate fell, from about 1.24 in 2015 to 0.72 in 2023, the lowest national figure ever recorded anywhere on Earth.
Then it turned. The rate rose to 0.75 in 2024 and to about 0.80 in 2025. Births rose 6.8 per cent in 2025 to about 254,500, the largest annual increase since 2007, and the second consecutive annual rise after eight straight years of decline.
Two readings are available and the difference between them matters.
The tempting reading is that the money finally worked. The evidence points elsewhere. In Korea, births outside marriage remain uncommon, so marriages are a reliable leading indicator of births one to two years later — and marriages jumped about 14.8 per cent in 2024, the biggest rise since records began in 1970, followed by a further 8.1 per cent in 2025. A large part of that was weddings postponed during the pandemic finally happening. On top of that, the women currently in their early thirties are from a relatively large birth cohort. The cohorts behind them are much smaller, which puts a ceiling on how long any rebound can run.
In other words: this looks, so far, mostly like tempo. Korean researchers have made the point directly — the level has not recovered to where it was before the pandemic, and an echo is not a reversal. It may become more than that. Two years is not enough to know, and anyone telling you confidently in either direction is telling you their politics.
6.5 — The one country that never fell
Israel is the only rich democracy that has never gone below replacement. Its rate is around 2.9, higher than several developing countries, in a society with high female education, high female employment and a large technology sector. It is the single most important case in this chapter, and the most misused.
It is misused because people assume the answer is the ultra-Orthodox, whose fertility is very high. That is part of it, but it is not the interesting part. The striking fact is that secular Israeli Jews are also above replacement, at roughly 2.2. Take the religious communities out entirely and Israel still has higher fertility than any other rich country.
What is going on there appears to be several things at once, and none of them is a policy you can order off a shelf. Almost everybody marries, and marries relatively young. Having children is treated as a normal part of adult life rather than a project to be attempted once other things are secure. The state funds fertility treatment more generously than anywhere else on Earth, with treatment cycles covered up to two children for essentially any woman who wants them. Extended family childcare is widely available because families live close together. And underneath all of it sits a collective story — about a small population, about survival, about a national future — that gives childbearing a public meaning it has lost almost everywhere else.
That last item is the one nobody knows how to reproduce. It is also, uncomfortably, closely bound up with a permanent security threat. The honest summary of the Israeli case is that it shows high fertility is compatible with a rich, modern, educated society — which is genuinely important, because it refutes the claim that development makes low fertility inevitable. What it does not show is any transferable method.
Every government facing this now has to decide whether to spend. The record above is not ambiguous about the size of the effect, but it is genuinely ambiguous about what to conclude from it.
Not one country has returned to replacement through policy. Hungary spent five per cent of national income and ended roughly where it started. Korea spent over two hundred billion dollars and set a world record for the lowest rate ever measured. Poland's benefit was followed by the lowest fertility in its history. The measured effects are small, mostly on timing, and cost a fortune per additional birth. The reason is not that governments picked the wrong policy. It is that the things actually driving the decision — whether you found a partner, whether you can see a stable ten years ahead, what a child will cost you in career and in hours — are not things a monthly payment reaches.
The right question is not "did the rate return to two" but "what would the rate have been without the policy". Fertility fell across all of Europe over this period, including in countries that spent nothing. Hungary rose while its neighbours fell and then fell alongside them during an inflation shock; that is a policy holding a line, not a policy failing. France's forty years at the top of Western Europe cost money and bought real children. And the effects that are measured are almost all measured on cash transfers, which are the crudest instrument available. The evidence on services — actual childcare places, actual job protection, actual housing — is stronger than the evidence on cash, and most countries have not seriously tried the expensive version.
What would settle it: a country that sustains a serious programme for thirty years without a recession or a political reversal, and reports completed cohort fertility at the end of it. France is the closest thing we have to that trial, and its result is "a few tenths, then decline anyway".
Why people care so much: because if policy cannot raise the number, the conversation moves on to things that might — immigration, or changing what is expected of women, or accepting a smaller population. All three of those are far more politically explosive than writing cheques, and writing cheques lets a government be seen to act while the harder argument is postponed.
6.6 — And now India is trying it
This is no longer a foreign story. India, the country that ran one of the world's largest and most coercive family-limitation programmes, has begun to reverse.
In 2024 the Chief Minister of Andhra Pradesh, N. Chandrababu Naidu, publicly urged families to have more children, citing his state's fertility rate of around 1.6 to 1.7 and an elderly population heading from about eleven per cent towards nineteen per cent by 2047. His government repealed the long-standing rule barring people with more than two children from contesting local body elections, and floated a law that would allow only those with more than two children to stand. The state has since announced cash payments for third and fourth births. The Chief Minister of Tamil Nadu, M. K. Stalin, asked publicly why Tamil families should not have sixteen children. In 2026 the Telangana assembly passed a bill abolishing its own two-child norm, citing a rural fertility rate of 1.7.
At the same time, in the north, the opposite argument continues. Uttar Pradesh has drafted population control legislation proposing to bar people with more than two children from government jobs and benefits.
So one country is now running pronatalist policy in the south and antinatalist proposals in the north, simultaneously. That is not incoherence. It is two different states responding to two genuinely different demographic situations — and, underneath, to a fight about parliamentary seats that Chapter Seven has to explain.
The test of a pronatalist policy is whether it changes quantum (how many children a woman ends up with) or only tempo (when she has them). Almost every claimed success has turned out to be mostly tempo.
Hungary spent about five per cent of national income for fifteen years. The rate rose from 1.23 to 1.59, then fell back to about 1.31. The population fell by 334,000 between two censuses anyway.
Korea spent over two hundred billion dollars and hit 0.72 in 2023, the lowest ever recorded anywhere. Its rise to 0.80 by 2025 tracks a surge in postponed weddings and a temporarily large cohort of women in their thirties — an echo, not yet a reversal.
Israel is the only rich democracy never to fall below replacement, and even its secular population is above it. The reasons are near-universal early marriage, heavily funded fertility treatment, close family networks and a collective story about the future. None of that can be bought.
No country has ever restored replacement fertility through policy. The effects that exist are real, small, and cost a fortune per additional child. India has just started spending anyway.
7The Bill Nobody Has Costed
Every society has a system for keeping old people alive after they stop working. In India that system is called "your children". This chapter is about what happens to it when there are fewer of them.
7.1 — Ageing is not what people think it is
When people hear that a country is ageing they picture more old people. That is half of it, and it is the harmless half.
More old people means people are living longer, which is the single greatest achievement of the last two centuries and nobody wants it reversed. The problem is the other half, which is fewer young people. A population ages from the bottom as much as from the top, and the bottom is where the workers, the taxpayers and the carers come from.
Old-age support ratio: how many working-age people there are for each person over sixty-five. It is the number that decides whether an ageing society is comfortable or strained.
India today has roughly ten working-age people for every person over sixty-five. On current projections that falls to somewhere around four by 2050. Japan is already at about two. Korea is heading for less than that.
Why it matters: everything an old person needs — a pension, a doctor, a nurse, a hand getting out of a chair — is provided by someone of working age, whether it is paid for through taxes or provided directly by a family member. When ten become four, either each of the four does more, or each old person gets less. There is no third option, and this is true regardless of how the system is financed.
7.2 — Three ways to pay for old age, and only three
Every society that has ever existed has used some combination of the same three methods.
Children. Your grown children feed and house you. This is the oldest system and still the most common one worldwide. Its advantage is that it needs no institutions. Its disadvantage is that it needs children, and it collapses if they move away, if they are poor, or if you did not have any.
Pay-as-you-go pensions. The state taxes today's workers and pays today's pensioners. Most rich countries do this.
Pay-as-you-go (PAYG): a pension system in which the money paid to today's pensioners comes from today's workers, not from a pot saved up earlier.
The thing to understand: despite the language of "contributions", nothing is stored. What a worker buys is a promise that the next generation will do the same for them. A PAYG pension is therefore the children system, run through a government instead of a family. It depends on the next generation existing in the same way a family does. It just hides the dependency inside an institution.
Saving. You put money aside and live off it. This looks as though it escapes the problem, and it does not entirely: when you retire you sell your assets, and you sell them to younger people. Fewer younger buyers with less money means lower prices. Saving softens the blow without removing the fact that what a retired person consumes must be produced, that year, by somebody working.
That is the hard core of this chapter, and it applies to every country regardless of ideology: you cannot store labour. The care an eighty-year-old needs in 2055 must be delivered by a person alive and working in 2055.
7.3 — India's system is children, and there is a law about it
India has not built a general old-age pension. What it has is a patchwork covering a minority, and underneath it, an explicit legal instruction to families.
The Maintenance and Welfare of Parents and Senior Citizens Act of 2007 makes it a legal obligation for children to maintain their elderly parents, and gives parents a tribunal to apply to if they do not. Read that again in the context of this document. India's principal old-age security policy is a statute telling adult children to pay. That is not a criticism of the law, which was passed because the informal system was already failing some people. It is a description of what the system is.
The best evidence on how India's elderly actually live comes from the Longitudinal Ageing Study in India, which since 2017 has interviewed a nationally representative sample of more than thirty thousand people aged sixty and over — asking about income, health, work, living arrangements and who supports whom.
What it shows: only about one in five people over sixty is covered by any central or state pension programme, and only about two in a hundred by a private or employer pension. Financial transfers from adult children to parents are about three times as common as transfers from parents to children. And of those elderly who spent their working lives in the organised sector — the minority most likely to be covered — around 78 per cent receive no pension at all. For women who worked in that sector the figure is starker: fewer than one in twelve receives a pension, against roughly one in four of the men.
What it cannot show: what people would do under different arrangements, or how much support is given informally in kind — a room, food, medicine — rather than in cash. Some of what looks like independence in the data is a widow living in a son's house and counted separately.
There is a state pension for the very poor, the Indira Gandhi National Old Age Pension Scheme. The central government's contribution to it has been ₹200 per month for those aged sixty to seventy-nine. States add their own top-ups, so what an individual actually receives varies enormously — some states pay a genuinely useful amount, others barely more than the central share.
₹200 a month is about ₹6.60 a day. At recent prices that is somewhere around a kilogram of rice, or a single cup of tea and a snack, or roughly one-fifth of a day's wage for unskilled work in most of the country.
Put it against the thing it is meant to cover. Basic groceries for one person, plus routine medicine for the ordinary conditions of old age, do not come to ₹6.60 a day anywhere in India.
The central share has not been raised since 2007. So a scheme designed to keep the poorest old people alive is now worth, in what it will actually buy, a fraction of what it was worth when it was set. India's public old-age pension for the poorest is not small. It is close to symbolic, and the actual system is the family.
7.4 — Growing old before growing rich
Here is why India's situation is not simply "what happened to Japan, arriving later".
Japan's fertility fell below replacement in 1974. Korea's in the mid-1980s. China's in the early 1990s. India's around 2019. Now rank those four countries by how wealthy the average person was at the moment their country crossed the line. India comes last, by a very wide margin — poorer at the crossing point than Japan was fifty years earlier, and poorer than China was in the 1990s.
The comparison needs care, because comparing incomes across fifty years requires adjusting for inflation and for what money buys in different places, and different adjustments give different multiples. But every reasonable method gives the same ranking and the same qualitative answer: India has begun ageing at a much lower level of income than any large country that went before it.
The practical meaning of that is simple. Japan built its pension system, its hospitals and its nursing homes while it was rich and still had a young population paying for them. India has to build the same things while it is much poorer, and it has perhaps two decades to do it in.
India's population aged sixty and over is projected to grow from about 149 million in 2022 to roughly 347 million by 2050. That is an increase of about 198 million people over 28 years — an average of about seven million more elderly people every single year.
Seven million a year is, roughly, adding the entire population of a large Indian city to the retired population annually, for a quarter of a century, without stopping.
Kerala is already there: about fifteen in every hundred Keralites are over sixty, against under ten nationally. The southern states will hit this a full generation before the north.
None of that is in dispute. What is in dispute is whether India has enough time, and the disagreement is sharp enough that the two sides produce completely different pictures of the 2040s from exactly the same population figures.
This is the single most consequential open question in Indian economics, and it is genuinely open.
India still has one of the youngest large populations on Earth, with a quarter of its people between ten and twenty-four. The working-age share is still rising and will keep rising into the 2040s. That is a two-decade window in which a large young workforce supports a small number of dependants at both ends — exactly the configuration that powered East Asia's growth. Growth of six or seven per cent a year compounds fast: at seven per cent, income per head doubles in ten years. Two doublings before the ageing bites would put India in a completely different position, and the ageing itself will arrive unevenly, with the north still young while the south ages, which buys internal time through migration between states.
A young population is not a dividend. It is a potential dividend, and it converts only if the young are employed in productive work. India's rate of women in paid work is among the lowest in the world for a country at its income level, which means half the potential workforce is not in the calculation at all. The large majority of Indian workers are in informal employment with no contract, no pension and no social insurance, so even a growing workforce does not automatically build a contributory pension base. Meanwhile the south, which is the productive core, is ageing now — and it is doing so with a per-head income far below what Japan or Korea had at the same stage. The window is not twenty years everywhere. In Kerala and Tamil Nadu it has already closed.
What would settle it: the formal employment numbers over the next ten years, and the female labour force participation rate. Those are measurable, they are published, and they will answer this question in public and in real time.
Why people care so much: because if the answer is no, India will be the first country in history to have a very large elderly population and no general pension system — and the people who absorb that are not governments. They are families, and inside families, overwhelmingly, they are daughters and daughters-in-law.
That last sentence is doing more work than it looks, and it points at something both sides of the whole fertility argument are quietly assuming.
Read any projection of population ageing, from any direction. The alarmed version says the ratio of workers to pensioners is collapsing and the state cannot pay. The relaxed version says productivity, automation and immigration will cover the gap. Both are arguing about money. Both are assuming that the physical work of care — the bathing, the feeding, the lifting, the sitting with someone through a long afternoon — will get done.
It does not get done by itself, and in almost every society on Earth it is done, unpaid, by women. It is the largest single block of unpaid labour in the world economy and it appears in no national accounts anywhere. When a government says its pension system is sustainable, what it usually means is that its cash obligations are sustainable. The hours are not on the balance sheet, because the hours have never been priced.
This produces a specific and rarely stated trap. Ageing increases the hours of care needed. Low fertility means fewer daughters and daughters-in-law to provide them. Rising female employment — which everybody, including every government worried about growth, is actively trying to increase — removes those same women from availability during working hours. All three trends push in the same direction at once, and no country's plan accounts for the third one.
The general form: an input treated as free because it has never been priced. It runs through economics everywhere: household labour, environmental services, the unpaid work of raising the workers themselves. When something has always been supplied for nothing, its disappearance is not forecast, because it was never counted.
The uncomfortable implication for this document: a large part of what is called "the fertility crisis" is really a care crisis wearing a demographic costume. And a society that cannot find carers has exactly two options — pay for them, which nobody has costed, or lean harder on the women it already has, which is a rule about women arriving by the back door.
7.5 — Why India's fertility argument is really about seats
One last thing, and it is the piece that explains why Indian politicians have suddenly started talking about birth rates.
Seats in India's national parliament are allocated between states according to population. But that allocation was frozen in 1976, using the 1971 census, and then frozen again until 2026. The freeze existed precisely so that states which succeeded at family planning would not be punished for it by losing representation.
That freeze is now expiring. When seats are redistributed on current populations, the states that reduced their fertility fastest — Tamil Nadu, Kerala, Andhra Pradesh, Karnataka — stand to lose weight relative to the states that did not. From the southern point of view: we were told to have fewer children, we did it, and the reward is fewer seats.
So when an Andhra chief minister asks families to have more children, and a Tamil Nadu chief minister jokes about sixteen, they are making a demographic argument and a constitutional one at the same time. It is worth naming plainly what is happening, because it explains the volume: a question about how many children women should have is being fought as a question about political power between regions. The women in the middle of it are the mechanism, not the subject.
Ageing is not mainly about more old people. It is about fewer young ones. India has about ten working-age people per person over sixty-five today, heading for roughly four by 2050.
There are only three ways to fund old age: children, taxing today's workers, or saving. All three depend on the next generation existing, because you cannot store labour. Care needed in 2055 must be delivered by a person working in 2055.
India's system is children, plus a law telling them to pay. About one in five people over sixty has any pension at all. The central share of the pension for the poorest has been ₹200 a month since 2007 — around ₹6.60 a day.
India is ageing at a much lower income than any large country before it, and its elderly population is growing by roughly seven million people a year. Kerala is already where the rest of the country will be in twenty years.
Underneath the whole argument sits an unpriced assumption: that the hours of care will simply be there. In India, and everywhere else, those hours belong to women — and nobody's plan has costed them.
8What Follows From Fewer People
One camp says a shrinking population is the end of progress. The other says it is the correction the planet needed. Both have real arguments, and the thing they are actually disagreeing about is not the arithmetic.
8.1 — The strongest case for alarm
Let me put this side as well as I can, because it is usually caricatured as nostalgia for large families and it is not.
The first argument is about compounding, and it is the one people underrate. A population at replacement is stable. A population below replacement does not settle at a smaller size; it keeps shrinking, generation after generation, for as long as the rate stays below the line. There is no floor. The only thing that stops it is fertility returning to about 2.1, and no country has managed that once it fell.
Suppose a country's fertility settles at 1.3 and stays there. Each generation is roughly 40 per cent smaller than the one before.
Start with a hundred people. The next generation is about sixty. Then thirty-six. Then twenty-two. Then thirteen. In five generations — about a hundred and fifty years — a hundred people have become thirteen.
Now hold that against a single family. If your line runs at 1.3 for five generations, you have roughly one great-great-great-grandchild for every eight you would have had at replacement. This is what the alarmed side means when it says the situation is not a plateau. Below-replacement fertility is not a smaller population. It is a continuous decline that only stops when the rate changes.
The second argument is about ideas. Knowledge, unlike land or oil, is not used up when it is shared: a vaccine invented once can be used by everyone forever. That means the number of people is directly linked to the rate of discovery, because inventors are drawn from the population. Halve the number of young people and you halve the draws from the tail of the distribution where the rare inventors are. This is not a claim that crowded countries are cleverer. It is a claim about sample size, and it is the core of the case made by the economists Dean Spears and Michael Geruso, who argue that a permanently shrinking world would mean permanently slowing progress.
The third argument is fiscal and physical. Public debt is a claim on future taxpayers, and it was issued assuming there would be more of them. Pensions and health systems are promises to be paid by a generation that is now smaller than the promise assumed. And beyond money, there is the plain arithmetic of Chapter Seven: someone has to actually do the caring.
The fourth is about the character of an old society. A society whose median voter is sixty behaves differently from one whose median voter is thirty. It is more protective of asset values and less tolerant of disruption, because its wealth is in houses and pensions rather than in future earnings. Whether you call that stability or stagnation is a value judgement. That it is different is not.
The fifth is concentration. National averages hide the fact that decline is not spread evenly. It empties the periphery first. Japan has millions of abandoned houses. Whole districts of rural Spain and Italy have lost their schools, their doctors and their bus routes. The last person to leave a village is not experiencing a gentle national adjustment.
8.2 — The strongest case for calm
Now the other side, also at full strength.
The first argument is the track record of population panic, and it is a serious argument rather than a debating point.
In 1968 the biologist Paul Ehrlich published The Population Bomb, which opened by stating that the battle to feed humanity was over and predicted that hundreds of millions would starve in the 1970s. It was enormously influential. It shaped aid policy, and it shaped India's own coercive sterilisation programme of the mid-1970s.
What actually happened: world population roughly doubled, and the share of people who are undernourished fell substantially. Food production rose faster than population, largely through the crop varieties of the Green Revolution.
What this evidence establishes: that confident projections about population and catastrophe have a poor record, and that acting on them has done real harm to real people — in India's case, millions of coerced operations.
What it does not establish: that today's forecasts are wrong. Ehrlich was wrong about famine because of an innovation nobody had predicted. That is a reason for humility in both directions, not a reason to assume every worry is unfounded.
The second argument is that income per person does not depend on the number of people. There is no relationship in the data between a country's population size and how well off its citizens are. Switzerland, Norway and Singapore are small and rich. If the concern is living standards rather than national bulk, a smaller population is not obviously worse.
The third is that scarce labour has historically been good for workers. After the Black Death killed a third of Europe, wages for ordinary labourers rose sharply and stayed high for a century. Smaller generations tend to face less competition for jobs, housing and university places. If you are a young person in a shrinking country, some of the arithmetic runs in your favour.
The fourth is the environment, and this is the one that made depopulation a hopeful idea for a generation of people who are now being told to panic about it. Fewer people means less land cleared, less water drawn, fewer emissions. Every serious climate scenario is easier at nine billion than at twelve.
The fifth is that Japan already ran the experiment. Japan has had below-replacement fertility for over fifty years and a falling population since around 2010. It has not collapsed. Unemployment is low, life expectancy is the highest in the world, the streets are safe and income per person has continued to rise. Adjustment has been expensive and slow, and rural Japan has genuinely hollowed out. But the catastrophe has not arrived, and it has had five decades to.
8.3 — The distinction that resolves half the argument
A great deal of this dispute dissolves once you separate two things that are constantly confused.
A smaller population is not a problem. A country of forty million can be as prosperous, healthy and pleasant as a country of eighty million. Nothing in economics says otherwise.
The transition to a smaller population is the problem, because during it the age structure is lopsided: many old, few young. That is when the support ratio is worst and the bills fall due.
Now the sting. If fertility fell to 1.6 and then rose back to 2.1, the transition would be a bad thirty years followed by a stable, slightly smaller country. Difficult, survivable, over. If fertility stays below replacement indefinitely, the transition never ends. Every generation is the lopsided one. That is why the honest version of the alarmed case is not "the population will be small" but "the adjustment period has no exit unless the rate comes back up" — and Chapter Six showed how hard it is to bring back up.
8.4 — What immigration can and cannot do
The obvious answer to a shortage of young workers is to import some, and it works. Immigration is the single most effective tool any individual country has for fixing its own age structure. It is fast, it is proven, and it delivers exactly the demographic profile a shrinking country needs — people arriving already grown, already trained, at the start of their working lives.
It has three limits, and they are all real.
The first is political. In most rich democracies immigration at the scale required is not achievable, whatever anyone thinks it should be. The scale is genuinely large: to hold a support ratio constant purely through migration usually requires inflows far above anything currently politically survivable.
The second is arithmetic. Immigrants age too. A migrant who arrives at twenty-five is a pensioner in forty years, so migration postpones the problem rather than solving it, unless the flow keeps growing.
The third is the one that changes everything, and it is new. The sending countries are running out of young people as well. India, the world's largest source of migrants, is now below replacement. Mexico, the main source for the United States, is below replacement. The global pool of surplus young workers is shrinking, and within a few decades most of what remains will be in sub-Saharan Africa. Countries are now beginning to compete for migrants rather than merely to control them, and that competition will get sharper.
This is the argument the whole chapter has been building to, and I want to be clear that I am not going to settle it, because it is not fully settleable by evidence.
Nothing else on the policy agenda compounds like this. Every other problem — climate, debt, disease — is a problem for people, and this one is a problem about whether there are people. A permanently below-replacement world means a permanently shrinking one, and shrinking is not a state you settle into; it is a direction you keep travelling. Progress in medicine, technology and living standards has been driven by more people building on each other's work, and there is no example in history of sustained progress in a shrinking population. Add to that the fact that nobody has ever reversed it, and the honest conclusion is that we are running an experiment with no known exit and no precedent.
Every previous population panic was wrong, and each one licensed coercion — forced sterilisation in India, the one-child policy in China, abortion bans in Romania. The current panic is already producing proposals about what women should do with their bodies, which is a predictable cost and should be weighed against a speculative benefit. Meanwhile the change is slow enough to adapt to: a population falling one per cent a year gives a country decades. Japan proves a rich society can shrink for fifty years and remain one of the best places on Earth to live. And the environmental gains are real and immediate, while the innovation losses are theoretical and centuries away.
What would settle it: nothing, and I want to say that plainly rather than pretend otherwise. There is no measurement that tells you how much a future person's existence is worth. That is a moral question wearing a statistical costume, and it will still be a moral question when all the data is in.
Why people care so much: because the answer determines whether governments are entitled to intervene in reproduction at all, and everyone on both sides knows it. The evidential dispute is a proxy. What is being fought over is permission.
There is one more thing both sides have in common, and neither of them mentions it.
Every fertility statistic in this document is national. Korea's rate. India's rate. Hungary's rate. The alarm is national — "our" population is falling. The reassurance is national — "our" economy will adapt. Even the migration argument is framed as what migration does for the receiving country.
But people are not contained by borders, and neither is the care they provide. Right now the global picture looks like this: rich ageing countries recruit nurses and care workers from poorer, younger countries. Those workers arrive already raised and already trained, at the cost of the country that raised and trained them. An Indian nurse working in Italy improves Italy's support ratio and worsens Kerala's, and appears in Italy's national accounts as a gain and in nobody's as a loss.
So when a country announces a demographic crisis, the first question worth asking is: crisis for which container? A world with eight billion people, many of them young, has no shortage of humans. What some countries have is a shortage of humans who are theirs. Those are extremely different problems, and only one of them is about fertility.
The general form: choosing the boundary that produces the crisis. The same move appears in every field. A firm reports a labour shortage at the wage it wants to pay. A city reports a housing shortage inside its own limits. A country reports a population crisis while the planet has more people than ever. The boundary is doing the work, and the boundary is a choice.
Why it matters for this series specifically: the national framing is what turns a private question into a public duty. Nobody can be asked to have a child for the world. People can be, and are, asked to have children for a nation. The container is what makes the demand sayable.
Keep that in mind through the next chapter, which is where the demand gets made explicitly, and where we finally look at what it asks of the people who would have to satisfy it.
The alarmed case at its strongest: below-replacement fertility compounds. At 1.3, a hundred people become thirteen in five generations. Fewer people also means fewer inventors, unpayable pension promises, and a society whose politics is organised around protecting what already exists.
The calm case at its strongest: every previous population panic was wrong, income per person does not depend on population size, scarce labour has historically helped workers, fewer people helps the climate, and Japan has shrunk for fifty years while remaining one of the best places on Earth to live.
The distinction that resolves half of it: a smaller population is not the problem; the transition is. And if fertility never returns to replacement, the transition never ends.
Immigration works for an individual country and cannot work for the world. It is fast and proven and limited by politics, by the fact that migrants also age, and by the new fact that the sending countries are below replacement too.
What the two sides actually disagree about is not the data. It is how much a person who does not yet exist is worth, and how much coercion is acceptable in pursuit of them.
9Who Pays For The Reversal
Suppose you accept the alarmed case. Suppose you want the number to go back up. This chapter is about what that actually requires, from whom, and what has happened every previous time somebody tried to require it.
9.1 — The question, stated without decoration
A fertility rate is not raised by a government. It is raised by women being pregnant more often. Every policy, every campaign, every appeal to national duty ends at the same place: a specific woman, in a specific body, having a specific pregnancy she would not otherwise have had.
I will keep saying it, because this subject floats off into the abstract very easily. When a chief minister says the state needs a higher birth rate, he is not describing an economic target. He is describing something he wants several million individual women to do with their bodies over the next decade.
That does not make the wish illegitimate — societies ask things of people all the time, including taxes, jury service and conscription. It does mean the request has to be examined the way we examine any claim one group makes on another group’s bodies, and not the way we examine a budget line.
9.2 — The gap runs in both directions
Here is the fact that makes this chapter more complicated than either side wants it to be.
Ask people in low-fertility countries how many children they would ideally like. The answer, consistently, across Europe, East Asia and America, is around two — sometimes a little more. Then look at what they end up with: often half a child less. There is a persistent gap between the family people say they want and the family they get.
That gap is the strongest card in the pronatalist hand, and it deserves to be played properly. It means the low numbers are not, mostly, a story of people who do not want children. They are a story of people who wanted children and ran out of time, money, partner or health. On that reading, raising fertility does not require persuading anybody of anything. It requires removing obstacles between people and something they already said they wanted.
The most useful recent evidence is the United Nations Population Fund's 2025 report on fertility, which surveyed fourteen thousand adults across fourteen countries covering a large share of the world's population — and asked, unusually, about failure in both directions.
What it found: large numbers of people had fewer children than they wanted, mostly for economic reasons and lack of a partner. And large numbers had more than they wanted. In India, about one woman in five reported having had more children than she intended, most often because of what her community expected. Around three in ten Indian women reported feeling pressured to continue a pregnancy they did not want, and roughly one in seven said pressure from health workers had affected their reproductive goals.
What it is good at: showing that the standard framing — women choosing fewer children — misses half the picture. What it is bad at: stated preferences are shaped by what people think they are supposed to want. A woman in a society that expects three children will often report wanting three. The gap is real; the ideal number it is measured against is not a fixed fact about human nature.
So the honest position is this. Some of the shortfall is genuinely unmet demand, and closing it costs women nothing and would help them. And some of it is not — some of it is that people, given the actual choice, want fewer children than the state wants them to have. No policy can tell the two apart in advance, and every government facing this problem has to decide how much of the gap it believes is the first kind.
9.3 — Three roads, and there is no fourth
Strip away the rhetoric and there are exactly three ways to raise a birth rate. Every scheme ever tried is one of these, or a mixture.
Road one: make children cheaper. Cash, tax breaks, childcare, housing, paid leave, job protection, fertility treatment, and men doing half the work at home. This is the humane road. It asks nothing of women that they have not already said they want. Chapter Six is the record of what it achieves: real effects, small effects, expensive effects, and no country back to replacement.
Road two: make the alternatives worse. This is never announced in these words, but it is what much of history consists of. If a woman cannot own property, hold a job, get an education, travel alone, or survive outside a marriage, then marriage and motherhood are not one option among several. They are the only option. Fertility is reliably high under those conditions. It is the most effective pronatalist technology ever devised, and every society that has had high fertility has had some version of it.
Road three: remove the exit. Restrict or ban contraception and abortion. Make divorce difficult. This does not change what anybody wants; it changes what they can do about it.
Almost every public argument about fertility is an argument about which of these three roads is being proposed while pretending to propose road one. That is worth watching for, in both directions — the pronatalist who talks about family values while meaning road two, and the opponent who treats every childcare subsidy as though it were road three.
9.4 — Romania, 1966, and what road three actually did
There is one clean, large, well-documented experiment on road three, and it is worth going through in detail because it is the closest thing this subject has to proof.
In October 1966 the government of Nicolae Ceaușescu, worried about a falling population, issued Decree 770. Abortion, which had been legal and was the main method of family limitation in Romania, was banned for most women. Contraceptive imports were cut off. Divorce was made harder. Childless adults were taxed. Women were subjected to regular workplace gynaecological examinations to detect and monitor pregnancies.
It worked, immediately and dramatically. The fertility rate roughly doubled within a year, from about 1.9 to about 3.7. The number of births in 1967 was nearly twice that of 1966. Measured against its own stated goal, Decree 770 is the single most successful pronatalist policy in recorded history.
Now the rest of it.
Within a few years the rate began falling again as people found their way around the ban, and within about a decade it was back down close to where it had started, despite the law remaining in force for twenty-three years. Illegal abortion became widespread and dangerous. Estimates of the number of Romanian women who died from unsafe abortions during the decree's lifetime run to around ten thousand, and maternal mortality rose to the highest in Europe by a wide margin. Many of the children born into families that could not support them ended up in state institutions, which is the origin of the Romanian orphanage crisis discovered by the world in 1989.
Ten thousand deaths over twenty-three years is roughly one Romanian woman dying every twenty hours, for a generation, from a procedure that had been safe and legal the year before.
And the payoff for that: a fertility rate that ended the period roughly where it began. The policy did not even deliver its own objective in the long run. It delivered one enormous cohort — the generation born in 1967, who went through overcrowded schools, then overcrowded universities, then a labour market with no room in it — and then a return to the previous trend, at that cost.
China provides the same lesson from the opposite direction. The one-child policy was the most aggressive antinatalist programme ever attempted, running from 1980 to 2015 with fines, forced insertions of contraceptive devices, and coerced abortions and sterilisations. When the state reversed course — two children permitted in 2016, three in 2021, and effectively no limit after that, with cash incentives attached — fertility carried on falling. Coercion could push the number down. Money and permission could not pull it back up.
That asymmetry is one of the most robust findings in this entire document, and it deserves stating on its own line. Stopping births is easy and cheap. Causing births is hard and expensive. Force works in one direction only.
9.5 — Where I have to be careful
I said in the front matter that I would flag this at the point where it applies, so here it is.
I am an Indian man in my twenties with no children. Everything in this chapter is a description of a cost I will never pay. There is a specific failure available to someone in my position, and it is not bias in the ordinary sense — it is that road two can be written down as a third bullet point on a list of options, in the same typeface as the other two, when for the people it applies to it is not an option among others but the removal of a life.
I have tried to guard against it by writing what each road requires in physical terms rather than policy terms. Whether that worked is not something I can judge from here. What I can do is tell you which way the error would run, so you can correct for it as you read.
There is a second thing. In India, a man writing about how many children women should have is writing inside a live political fight in which women are almost entirely the object and almost never the speaker. The chief ministers in Chapter Six are men. The seat calculation in Chapter Seven is being run by men. The community pressure recorded in the survey above is applied largely to daughters-in-law. I am one more male voice in a conversation that has never lacked one, and my only defence is to keep pointing at what the women in the data actually said — which is that many of them are already having a different number of children than they wanted, in both directions.
This is the question the whole series has been walking towards. I am not going to answer it, and I want to be clear that this is a refusal, not an oversight.
Nobody has actually tried road one properly. What has been tried is cash — the cheapest, laziest instrument available — plus childcare in some countries and paid leave in others, almost never all at once, almost never for thirty years, and almost never including the one thing that matters most, which is a labour market that does not punish a woman permanently for taking two years out. Add secure housing for people in their twenties, genuine job protection, fully funded fertility treatment, and men taking half the domestic load, and you are addressing the actual reasons people give in surveys for not having the children they said they wanted. The gap between wanted and achieved is real and it is roughly half a child. Closing it would get most countries most of the way back, and it would cost women nothing — it would give them something.
Look at what has actually happened where road one was tried hardest. The Nordic countries built the most generous version of it in existence and their fertility fell by half a child anyway. Hungary spent five per cent of national income and ended where it started. If the strongest available version of the humane road produces a fifth of a child at enormous cost, then a country that genuinely needs another half a child cannot get there that way — and a government that genuinely believes it faces an existential crisis will not stop at policies that do not work. Historically it never has. The high-fertility societies of the past were not societies with better childcare. They were societies where women had fewer options, married younger, and could not leave. That is not a coincidence or a slander; it is what the record shows, and everyone arguing about this knows it.
Both sides above assume that some number is correct and the job is to reach it. Perhaps there is no correct number. Perhaps a fertility rate is simply the sum of several hundred million private lives and is not a policy variable at all, any more than the national average height is. On this view the right response to ageing is to build the institutions that ageing requires — pensions, care systems, immigration, automation — and to stop treating the birth rate as a lever, because treating it as a lever is what produces Decree 770.
What would settle it: a rich country that spends at Hungarian levels, on services rather than cash, with equal domestic labour and secure housing for the under-thirties, sustained for thirty years, and then reports its completed cohort fertility. Until somebody runs that, both sides are arguing from the same absence.
Why people care so much: because if the answer is no — if fertility cannot be raised without taking something back from women — then a society facing demographic decline has to choose between two things it says it values. Everyone involved can see that fork coming, and the argument is fought at this volume because both sides would rather win it before the choice becomes explicit.
One last thing, and it is aimed at this document rather than at anybody else's.
I have now written nine chapters treating the number of children born in a country as a thing that can be evaluated, worried about, projected and, if necessary, managed. I have weighed the alarmed case against the calm case as though the question were open and interesting. I have used phrases like "the bill", "the reversal" and "what would settle it".
Every one of those framings assumes something I never argued for: that the sum of several hundred million private reproductive lives is a legitimate object of public policy, and that there exists a "we" with standing to have a view about it.
Notice that both sides of every argument in this part share that premise. The pronatalist wants the number up. The relaxed side says the number is fine. The environmentalist wants it down. All three are treating the aggregate as something a society is entitled to hold an opinion about. The only position that does not is one almost nobody defends in public: that the number is simply what it is, that it belongs to the people who produced it, and that having a national view about it is already the beginning of the problem.
I do not know whether that position is right. I notice that I did not consider it until I had written eighty pages, and that the entire structure of this part — its chapters, its arguments, its arithmetic — only makes sense if it is wrong. That is exactly the kind of assumption this series exists to dig out, and I found it underneath my own document rather than underneath somebody else's.
The general form: treating the sum of private decisions as a lever. It appears wherever an aggregate has a name — the birth rate, the savings rate, the marriage rate, the crime rate. Naming a total makes it feel like a thing that someone is responsible for, and once a total has a custodian, the individuals inside it become inputs.
I am leaving this part standing as written rather than rewriting it around this, because the point of the box is to show you the floor you have been walking on for eighty pages. But read the last eight chapters again with it in mind. Several of the arguments look different.
That is where this part ends, and it ends without a verdict, which I know is unsatisfying. The next chapter is the standing accounting: what in all of this is genuinely unknown, and what is solid enough to build on.
A birth rate is not raised by governments. It is raised by specific women having specific pregnancies. Every policy in this part ends there.
The gap between the children people say they want and the children they have is real, roughly half a child, and runs in both directions — in India, about one woman in five reports having had more children than she intended.
There are exactly three roads: make children cheaper, make the alternatives worse, or remove the exit. Most public arguments are about roads two or three while claiming to be about road one.
Romania's Decree 770 doubled the birth rate in a year — and within a decade it was back near where it started, at a cost of roughly ten thousand women's lives. China's reversal shows the same thing from the other side: force can push births down and cannot pull them up.
Nobody has yet shown that a birth rate can be raised substantially by any method that costs women nothing. That is the honest state of the evidence, and it is why this argument is being fought so hard.
10An Honest List of What We Do Not Know
Every part of this series ends the same way. Two lists: what is genuinely unknown, and what is solid enough to build on. This is the most useful chapter in the document and the one almost nobody else writes.
10.1 — Genuinely unknown
The reason a thing is unknown is usually more interesting than the gap itself. Each item below comes with its reason.
Where fertility settles
Nobody knows whether the very low rates in East Asia are a floor, a passing low, or a station on the way further down. Every projection published in the last thirty years has assumed some floor and most have been wrong. The reason we do not know: we have never observed a society at 0.8 for a full generation, so there is no case to reason from. Korea will be the first, and it will take twenty years to read.
How much of the decline is timing and how much is total
Chapter One set this out. The period rate overstates the fall; the tempo-adjusted rate understates it. We will know the answer for women born in the 1990s in about fifteen years and not before. The reason we do not know: you cannot measure a completed family before it is completed. This is the only item on this list with a guaranteed answer and a known delivery date.
Whether a serious version of the humane road would work
The great absence in this whole field. No country has ever run the full package — generous services rather than cash, secure housing for people in their twenties, genuine job protection through childbearing, and equal domestic labour — sustained for a generation, without a recession or a change of government. Every claim that it would work and every claim that it would not is therefore an extrapolation. The reason we do not know: the experiment is expensive, slow, and politically impossible to hold steady for thirty years.
How fast African fertility will fall
The single largest source of uncertainty in every world population projection. The credible range for when sub-Saharan Africa reaches replacement spans about forty years. The reason we do not know: the region is in the early part of its transition, and every previous region has moved faster than forecast, but the countries with the highest rates also have the weakest schooling and the highest child mortality. Genuine data is coming, on a known schedule.
Whether the recent Korean rebound is real
Two years of rising births after eight years of falling. It tracks a surge in postponed weddings and a temporarily large cohort of women in their thirties, both of which will pass. Whether anything underneath it has changed is not knowable yet. The reason we do not know: two data points do not distinguish an echo from a turn, and no method exists that can.
What the retention rates of high-fertility religious communities will be
The inheritance argument in Chapter Five stands or falls on this and the data is thin, contested, and mostly about small communities. The reason we do not know: retention only becomes measurable when a generation has grown up and chosen, and the communities in question have only recently become large enough for the question to matter.
Whether ageing societies actually innovate less
The claim that fewer people means slower progress is theoretically well grounded and empirically almost untested, because no large society has yet been through a sustained population decline in a modern economy. Japan is the closest case and its record is genuinely mixed. The reason we do not know: the sample size is one, and it is only fifty years in.
What people actually want
Every survey of desired family size is measuring an answer given inside a set of expectations. A woman who says she wants three in a society that expects three, and a woman who says she wants one in a society that expects one, may or may not want different things. The reason we do not know: there is no way to ask a person what they would want in a society they have never lived in.
One absence on the list above is worth treating as evidence rather than as a gap.
India has run a Sample Registration System since the 1970s and a National Family Health Survey since 1992. It has counted births, deaths, infant mortality and fertility by state, by residence and by education, continuously, for fifty years. It has one of the better demographic statistical systems in the developing world.
What it has not done is publish a decennial census since 2011. The census due in 2021 was postponed, and the delimitation of parliamentary seats — the fight described in Chapter Seven — depends on it.
When a state has the capacity to count something and the count is delayed while a political question rides on it, the delay is a finding rather than an oversight. It does not tell you what the numbers would show. It does tell you that somebody understands what they would trigger.
10.2 — Solid
Now the other list. These things can be built on. Each comes with the reason it is solid, which is usually that independent sources with no shared interest agree.
The fall happened, and it is very large
World fertility has roughly halved in seventy years, from about five to about 2.25. This is confirmed by vital registration in rich countries, household surveys in poor ones, school enrolment numbers, and the age structures visible in every census on Earth. Nobody disputes it in any direction.
India is below replacement, and its states differ enormously
The Sample Registration System put India at 1.9 in 2023, with urban India at 1.5 and rural India at 2.1. Delhi is at 1.2, Bihar at 2.8, eighteen states and union territories below the replacement line. Urban India crossed in 2004. These figures come from a system that has used the same dual-recording method for fifty years, which makes the trend particularly reliable.
Child survival came first
Everywhere it has been studied, the fall in child mortality preceded the fall in fertility. This is visible in European parish registers, in Indian survey data, and in the sixty-year field study at Matlab in Bangladesh. Three completely different kinds of record, in three centuries, agreeing.
Education is the strongest single predictor
Across countries and within them, how long a woman stayed in school predicts how many children she has better than her income or her religion. In India the gap between no schooling and completed schooling has consistently exceeded a full child. This has held across five rounds of the National Family Health Survey over thirty years.
The Hindu–Muslim fertility gap in India is real and has more than halved
About 1.1 children in 1992–93; about 0.42 in 2019–21. Muslim fertility in the southern states is lower than Hindu fertility in the northern states. Same survey, same questions, five rounds, one direction.
No country has restored replacement fertility through policy
This is solid because it is a negative claim about a well-documented set of attempts. Hungary, Poland, Russia, Singapore, Japan, Korea and France have all tried, several at enormous expense, and their completed fertility figures are published. The effects that exist are small. The claim is not that policy does nothing; it is that nothing has got a country back to 2.1.
Force works in one direction only
Coercive antinatalism has repeatedly and rapidly reduced births — China's one-child policy, India's sterilisation drive of 1975–77. Coercive pronatalism produces a large one-off spike and then decays, as Romania's Decree 770 showed across twenty-three years. And China's post-2016 reversal, with permission and money attached, did not raise fertility at all. Different countries, different systems, same asymmetry.
Ageing is arithmetic, not opinion
India's old-age support ratio falls from roughly ten working-age people per person over sixty-five to roughly four by 2050. That is not a forecast about behaviour. The people who will be sixty-five in 2050 are alive now and can be counted. Fertility projections are unreliable; near-term ageing projections are close to certain, because everyone involved has already been born.
India's old-age provision is thin, and the system is the family
About one in five people over sixty has any pension. The central share of the pension for the poorest has been ₹200 a month since 2007. A 2007 statute obliges children to maintain their parents. Survey data, administrative data and the law itself all say the same thing.
Unknown: where fertility settles, how much of the fall is timing, whether a serious version of the humane road would work, how fast Africa falls, whether Korea's rebound is real, and what people would want in a society that expected something different from them.
Solid: the fall happened and is very large; India is below replacement with enormous variation between states; child survival came first; education predicts fertility better than religion or income; the Hindu–Muslim gap is real and has more than halved; no country has policied its way back to replacement; force works only downwards; and near-term ageing is arithmetic rather than forecast.
One absence is itself informative. India can count and has counted for fifty years, and the census that would trigger the redistribution of parliamentary seats has not been held since 2011.
The most important thing on either list is the third item: the humane road has never been properly tried, so the central question of this part is genuinely open — and everybody arguing about it is arguing from the same missing experiment.
Sources & further reading — Part 9
Timeline
Two hundred and thirty years of counting people, and of governments deciding what the count meant.
| Year | What happened |
|---|---|
| 1798 | Thomas Malthus publishes his essay arguing that population grows faster than food. Wrong about the future, enormously influential about it. |
| 1877 | The Bradlaugh–Besant trial in London puts birth control information in front of a jury and on the front pages. British fertility begins its long fall shortly afterwards. |
| 1934 | Alva and Gunnar Myrdal publish Crisis in the Population Question in Sweden, arguing that the answer to falling births is a welfare state rather than a ban. The first modern statement of road one. |
| 1952 | India becomes the first country in the world to adopt a national family planning programme. |
| 1960 | The contraceptive pill is approved in the United States. |
| 1966 | Romania issues Decree 770. Abortion is banned, contraception cut off, divorce restricted. Births nearly double within a year. |
| 1968 | Paul Ehrlich publishes The Population Bomb, predicting famines that do not arrive. |
| 1974 | Japan's fertility falls below replacement. At Bucharest, the world population conference hears the argument that development is the best contraceptive. |
| 1975–77 | India's Emergency. A mass sterilisation campaign produces millions of operations, lasting public distrust, and no change in the underlying trend. |
| 1980 | China's one-child policy begins. |
| 1987 | Singapore reverses from antinatalist to pronatalist policy. Its fertility continues to fall for the next thirty-six years. |
| 1992–93 | India's first National Family Health Survey. Hindu–Muslim fertility gap measured at about 1.1 children. |
| 1994 | The Cairo conference shifts global policy from population targets towards reproductive rights and women's health. |
| 2000 | Peter McDonald publishes the gender equity theory. The term "lowest-low fertility" is coined for rates at or below 1.3. |
| 2004 | Urban India falls below replacement fertility, fifteen years before the country does. |
| 2007 | Russia introduces maternity capital. India passes the Maintenance and Welfare of Parents and Senior Citizens Act, obliging children to support their parents, and sets the central old-age pension share at ₹200 a month. |
| 2009 | The J-curve paper appears in Nature, suggesting fertility rises again at very high development. |
| 2010 | Hungary begins the most ambitious pronatalist programme in any democracy. |
| 2015 | China ends the one-child policy. Fertility keeps falling. |
| 2016 | Poland introduces a large universal child benefit. A two-year rise is followed by the lowest fertility in Polish history. |
| 2019 | India's fertility rate falls below the replacement line by official Indian estimates. |
| 2022 | China's population falls for the first time in six decades. |
| 2023 | South Korea records 0.72, the lowest national fertility rate ever measured anywhere. Singapore falls below 1.0. India's Sample Registration System reports rural India at replacement for the first time. |
| 2024 | Andhra Pradesh repeals its two-child bar on contesting local elections and urges larger families. The United Nations projects a world population peak of about 10.3 billion in the 2080s. |
| 2025 | Korea's rate rises to about 0.80 on a surge of postponed marriages. The United Nations Population Fund reports that people are missing their fertility goals in both directions. |
| 2026 | Telangana abolishes its own two-child norm. India's freeze on the redistribution of parliamentary seats reaches its expiry. |
| Next | Part Ten starts here. |
Glossary
Every hard word used in this part, in plain English.
| Term | What it means |
|---|---|
| Age-specific fertility rate | The share of women of one particular age who gave birth in a given year. Add these up across all ages and you get the total fertility rate. |
| Antinatalism | Policy aimed at reducing births. India's sterilisation drive and China's one-child policy are the two largest examples in history. |
| Cohort fertility | The real average number of children had by women born in the same year, counted once they have finished. Accurate, and available only decades late. |
| Crude birth rate | Births per thousand people in a year. Easier to measure than fertility but misleading, because it depends on how many young women a country happens to have. |
| Decree 770 | Romania's 1966 law banning abortion for most women. The most successful pronatalist policy ever measured, and the most costly. |
| Delimitation | Redrawing constituency boundaries and reallocating parliamentary seats between Indian states according to population. Frozen since 1976; the freeze expires in 2026. |
| Demography | The study of populations by counting them — births, deaths, moves and ages. |
| Demographic dividend | The temporary economic advantage a country gets when it has many working-age people and few dependants. It is a window, not a gift, and it converts into growth only if the workers have work. |
| Demographic transition | The standard sequence: deaths fall, then births fall, with a population explosion in between. |
| Differential fertility | Different groups within a country having different numbers of children, which shifts the population's composition over generations. |
| Ecological fallacy | Assuming that a pattern between countries also holds inside them. The commonest mistake in popular writing about fertility. |
| IGNOAPS | The Indira Gandhi National Old Age Pension Scheme, India's pension for the very poor. The central government's share has been ₹200 a month since 2007. |
| Human Development Index | A single score combining life expectancy, schooling and income, used to compare countries on something other than money. |
| LASI | The Longitudinal Ageing Study in India, a large repeated survey of Indians aged sixty and over, covering income, health and who supports whom. |
| Lowest-low fertility | A total fertility rate at or below 1.3. Named in 2000, when it was thought to be rare and temporary. |
| NFHS | The National Family Health Survey, India's large household survey, run in five rounds since 1992. The only Indian source that breaks fertility down by religion, caste and education. |
| Old-age support ratio | How many working-age people there are for each person over sixty-five. India is at roughly ten and heading for about four. |
| Opportunity cost | What you give up in order to do something — not the money it costs, but the next-best option you had to drop. |
| Pay-as-you-go | A pension system that pays today's pensioners out of today's workers' taxes. Nothing is stored; it is the children system run through a government. |
| Period fertility | The snapshot version of the fertility rate: all births in one year, across women of every age. This is what the headlines quote. |
| Population momentum | The tendency of a population to keep growing for decades after fertility falls to replacement, because of the shape of its age pyramid. It also runs in reverse. |
| Pronatalism | Policy aimed at increasing births, from cash payments at one end to bans on contraception at the other. |
| Quantity–quality trade-off | The idea that parents choose not only how many children to have but how much to invest in each, and shift towards fewer children once investing pays. |
| Quantum | How many children a woman ends up with in total, as opposed to when she has them. |
| Replacement level | The fertility rate at which each generation exactly replaces the last. About 2.1 where almost all children survive; higher where they do not. |
| Retention rate | The share of children raised in a group who still belong to it as adults. The hinge of the whole inheritance argument. |
| SRS | The Sample Registration System, India's continuous dual-recording survey of births and deaths, running since the 1970s. |
| Tempo effect | The distortion in the yearly fertility rate caused by people shifting the timing of births. Postponement pushes the number below the truth. |
| Total Fertility Rate (TFR) | A summary of one year's births, arranged to look like a lifetime. Not a count of anybody's actual children. |