What Changed and What Did Not

The last lesson corrected the baseline. This one is about the changes that are real — and they are enormous, measurable, and international.

The largest of them is not about attitudes or values. It is that the biological shape of an adult life was reorganised within about a century, and that almost everything else follows from it.

Three women, one country, a hundred and fifty years.

Martha, born 1845. She marries at twenty-five. Over the next seventeen years she is pregnant eight times. Five children survive infancy. Her last child is born when she is forty-two. Her eldest is still at home when the youngest arrives. She dies at fifty-five, with her youngest child aged thirteen.

Count her adult life. From marriage to death: thirty years, of which she spends roughly seventeen either pregnant or nursing, and all thirty with dependent children in the house. There is no period of her adult life in which she is not raising someone.

Elizabeth, born 1925. She marries at twenty-two. She has three children between twenty-three and twenty-eight. Her youngest starts school when she is thirty-three, and leaves home when she is forty-six. She lives to seventy-eight, and is widowed at seventy-one.

Count hers. Active childrearing occupies about twenty-three years of a fifty-six-year adult life. She has thirty years after her last child leaves.

That phase did not exist before. Not as a minority experience — as a phase of life. Her mother's generation did not have it, because women died with children still at home. The "empty nest" and long widowhood are twentieth-century inventions, produced by the combination of fewer births, closer together, and much longer lives.

Sarah, born 1990. She has her first child at thirty-two, possibly a second at thirty-five. She may not marry, or may marry after the children. She expects to live to her late eighties.

Count hers. Active childrearing occupies perhaps twenty years of a seventy-year adult life. She will spend more of her adulthood without dependent children than with them, and she is likely to spend a substantial period caring for a parent rather than a child.

One more comparison, which is the sharpest.

In Martha's world, the gap between a girl reaching reproductive maturity and having her first child was around five to eight years. In Sarah's it is closer to twenty.

Two decades of adult sexual life before childbearing is a new thing in human history, and almost everything in this lesson — contraception, cohabitation, later marriage, the changed meaning of marriage itself — is downstream of it.

Which changes are real, and which are artefacts of the baseline?

9.2.1 established that comparing to the mid-twentieth century inflates apparent change. So the discipline required is to ask, of each change, whether it is a departure from a long-run pattern or a return to one.

Some are genuine long-run transformations: the fall in mortality, the fall in fertility, the lengthening of life, and the reorganisation of the life course they produced. Nothing like these had happened before.

Some look like departures from tradition and are partly returns. Household size, non-marital childbearing rates, step-families and multi-generational co-residence have all been higher in the past than the mid-century baseline suggests (see 9.2.1).

And some are genuinely new but smaller than their public prominence — which requires knowing the numbers rather than the coverage.

Two measurement points before any of it, because both change the story.

Period and cohort fertility are different quantities. The commonly quoted total fertility rate is a period measure: the number of children a woman would have if she experienced this year's age-specific rates throughout her life. If women are postponing births, that measure falls even when completed family size does not change, because the births have moved later rather than disappeared. This is the tempo effect, and it means period fertility overstates decline during any period of postponement — and then overstates recovery afterwards.

And "family structure" measured at a moment differs from a childhood experienced. A cross-sectional count of children living with two parents understates how many children experience a change in family structure at some point in their childhood (see 9.2.1's developmental cycle).

The demographic transition

The single largest change, and the order in which it happened.

Every society that has industrialised has passed through the same sequence, at different dates and different speeds.

Stage one: high mortality, high fertility, slow growth. Stage two: mortality falls — through nutrition, sanitation, clean water and eventually medicine — while fertility remains high, producing rapid population growth. Stage three: fertility falls. Stage four: both low, growth near zero. And what some call a fifth: fertility persistently below replacement, with population decline and ageing.

The lag between the two falls is the whole of the population growth of the modern era.

Europe took roughly two centuries to complete the sequence. Much of Asia and Latin America compressed it into fifty years. Sub-Saharan Africa is in the middle of it, at varying speeds.

Why fertility fell is the more contested question, and five accounts compete.

Children changed from an asset to a cost. In an agrarian household children work early and support parents later; in an urban wage economy they are expensive, are educated for years, and do not transfer resources back. The direction of the flow of wealth between generations reverses.

Mortality decline meant fewer births were needed to end with a given number of surviving children — and the response lags because expectations adjust slowly.

The opportunity cost of women's time rose as paid work outside the household became available and better paid, making each birth more expensive in forgone earnings.

Contraception and its diffusion.

And the finding that complicated the economic accounts. A large historical project mapping fertility decline across hundreds of European provinces found that decline began at similar dates within linguistic and cultural regions, across provinces with very different levels of urbanisation, industrialisation and income — and at different dates in adjacent provinces separated by a language boundary rather than by economic conditions.

That pattern is much easier to explain by the diffusion of ideas and practices than by economic thresholds, and it is why the field takes cultural transmission seriously rather than treating fertility as a pure calculation.

The single strongest predictor globally, across countries and within them, is women's education. It operates through several of the above at once — later marriage, higher opportunity cost, better information, more autonomy in the household — which is why it is a robust predictor and a poor discriminator between the theories.

And the relationship that reversed, which is the most instructive finding in this lesson.

In 1980, across rich countries, the correlation between women's labour force participation and total fertility was negative. Countries where more women worked had fewer children. The interpretation was obvious and everyone made it: work and childbearing compete.

By around 2000 the correlation across the same set of countries had turned positive. Countries where more women work now have higher fertility.

Nothing about the biology changed. What changed was the institutions.

Where childcare is available and affordable, where leave is designed so that taking it does not end a career, and where men do a larger share of domestic work, the conflict between employment and childbearing is reduced — and both are higher. Where those things are absent, women face a sharper choice, and increasingly resolve it by having fewer children or none.

Which is 8.5.1's argument arriving as a demographic fact, and it has been formalised as an account of a two-part transformation: the first half moved women into paid work and depressed fertility; the second half — men moving into domestic and care work — raises it. Countries that completed only the first half have the lowest fertility in the world.

The evidence for the framework is correlational and the pattern is striking, and it makes a prediction that can be checked: fertility should be lowest where women's employment is high and the domestic division of labour has not moved. That is broadly where the lowest fertility rates are found.

Marriage and its transformation

From an institution to a relationship to an achievement.

Marriage was for most of history an arrangement between families concerning property, alliance, labour and legitimate descent. Who married whom was substantially a matter for kin, and the emotional content was not the point of it.

The companionate marriage of the nineteenth and twentieth centuries made the relationship itself the purpose — mutual affection, companionship, a shared domestic life — while retaining a strong division of labour and a clear set of obligations.

And what has happened since has been described as deinstitutionalisation: the weakening of the shared norms specifying what married people are supposed to do. Where the rules are unclear, everything must be negotiated — who works, who cares, who moves for whose job, whose surname, what fidelity means — which makes marriage more individually tailored and more effortful, and removes the default answers.

The measurable changes are consistent across rich countries.

Age at first marriage has risen substantially, commonly by five to eight years since the mid-century.

Cohabitation has become the majority first union in many countries, ranging from a stage before marriage to a functionally permanent alternative depending on the country.

Non-marital childbearing has risen to around or above half of all births in a number of European countries — and in several of them the parents are stably cohabiting, which is why the statistic means very different things in different places.

And marriage has become, in some societies, a capstone rather than a foundation — something entered after achieving stable employment, savings and often after having children, rather than the event that begins adult life. Which has a distributional consequence discussed below.

Divorce: the rise, the plateau, and what the causal evidence shows.

Divorce rates rose sharply in most rich countries from the 1960s, and in many they peaked between the 1980s and the 2000s and have since fallen.

Part of the fall is compositional. Fewer people marry; those who do marry later, at higher levels of education, and after longer cohabitation — so marriage is increasingly selected on the characteristics that predict stability. A falling divorce rate among a more selected married population is not the same as more stable relationships overall.

On the causal question — did legal reform cause the rise? — there is a clean piece of work worth knowing.

Unilateral divorce laws, permitting one spouse to end a marriage without the other's consent, were introduced across American states at different dates. That staggered adoption is a natural experiment (see 7.4.3).

An earlier analysis found a persistent increase in divorce. A later reanalysis, using a specification that allowed the effect to vary over time since adoption, found that the reform raised divorce rates for roughly a decade and that the effect then dissipated — the rates in reforming and non-reforming states converged.

Which suggests the law changed the timing and the process rather than the quantity of marital breakdown, and it is a good example of a well-identified study correcting a well-identified study by asking a more careful question about dynamics.

Consequences for children: the association, the effect, and the gap between them.

Children who experience parental separation do worse on average — in attainment, in mental health, in their own later relationship stability. The associations are real and replicated.

The causal estimates are substantially smaller than the associations, and the reason is 7.6.2's.

Families that separate differ beforehand — in income, conflict, mental health, employment stability and the characteristics that predict both separation and children's outcomes. Designs that hold family constant — comparing siblings differently exposed by age at separation, or using fixed effects — consistently find smaller effects than cross-sectional comparisons.

Three further findings that the headline association conceals.

Conflict matters more than structure. Children in high-conflict intact households frequently do worse than children whose parents separated, and separation can improve outcomes where conflict was severe.

Resources carry a substantial share of the effect. A large part of the association operates through the fall in household income and in parental time following separation — which is a fact about the economics of two households, not about family form.

And instability matters more than any single transition. Repeated changes in household composition predict worse outcomes than a single separation, which is the finding most relevant to what follows.

The class divergence, which is the most consequential current pattern.

Family formation has become steeply stratified by education, and the divergence is recent.

Among the university-educated: marriage is later and more common, divorce rates have fallen, childbearing occurs within stable unions, and both parents are typically in employment. The mid-century arrangement, in a two-earner form, is increasingly a graduate arrangement.

Among those without degrees: marriage rates have fallen much further, childbearing occurs earlier and more often outside stable unions, and family transitions are more frequent.

And the consequence compounds everything in Part 8. Children of educated parents increasingly receive two adult incomes, two adults' time, and stability. Children of less-educated parents increasingly receive fewer of all three. The resources flowing to children have diverged, and the divergence operates through exactly 8.3.3's channels.

Why this happened is contested and three explanations have support. The disappearance of stable employment for men without degrees, which removed the economic basis of the older arrangement (see 8.6.2). The capstone shift, which makes marriage something achieved after stability rather than a means to it — so those least likely to achieve stability are least likely to marry. And changes in norms, which removed the near-compulsion to marry before childbearing.

All three are consistent with the pattern, and the first has the clearest supporting evidence — the timing of the divergence tracks the deterioration of the labour market for men without degrees closely.

Because the largest change is one nobody argues about, and the changes people do argue about are smaller and differently shaped than the argument assumes.

What actually changed. Mortality, fertility, and therefore the shape of a life. Two decades between reproductive maturity and first birth; thirty years of adulthood after childrearing; and a period of caring for parents that is now longer, for many people, than the period of caring for young children.

What changed less than it appears. Kin contact and support remain extensive — grandparents provide very large volumes of childcare in every country where it has been measured. Most people still form partnerships and most still have children. And the domestic division of labour narrowed and did not close (see 8.5.1).

What is genuinely new and consequential: the class divergence in family stability, which is a mechanism of inequality rather than a moral matter, and which the evidence ties closely to labour market change.

Four questions.

Period or cohort? Postponement makes period fertility fall without completed family size falling as much.

Association or effect? Family structure effects on children shrink substantially under designs that hold family background constant.

Structure or resources or conflict or instability? These are four different claims and only the last two survive the better designs well.

And compared to when? 9.2.1's question, and it still applies.

One closing observation about the reversed correlation. In 1980 the evidence said work and children compete. In 2000 the same measurement across the same countries said the opposite — because institutions had changed in some places and not others.

That is as clear a demonstration as this course contains of a relationship that looks natural being institutional, and it is why 8.8.1's conclusion holds here too: the arrangement is not given, and the countries that changed it can be identified.

Three reproductive lives show what changed. A woman born in 1845 spent essentially her whole adult life with dependent children and died before her youngest grew up. A woman born in 1925 had thirty years after her last child left — a phase of life that had not previously existed. And the gap between reproductive maturity and first birth widened from around five to eight years to close to twenty, which is new in human history and is what everything else is downstream of.

Two measurement points. Period fertility falls during postponement even when completed family size does not — the tempo effect. And cross-sectional family structure understates how many children experience a change during childhood.

The demographic transition: mortality falls first, fertility follows with a lag, and the gap is the population growth of the modern era — two centuries in Europe, fifty years where compressed. Five explanations for fertility decline — children shifting from asset to cost, mortality decline reducing the births needed, the rising opportunity cost of women's time, contraception, and diffusion. The historical project mapping European provinces found decline beginning at similar dates within linguistic and cultural regions across very different economic conditions, which favours diffusion. Women's education is the strongest predictor and a poor discriminator between theories.

And the correlation between women's employment and fertility across rich countries reversed — negative in 1980, positive by 2000 — because childcare, leave design and the domestic division of labour changed in some countries and not others. The lowest fertility is where women's employment rose and the domestic division did not move.

Marriage moved from an arrangement between families to a companionate relationship to a deinstitutionalised one in which the defaults must be negotiated — with later marriage, cohabitation as a majority first union, non-marital childbearing above half of births in several countries, and marriage as a capstone achieved after stability rather than a foundation.

Divorce rose, peaked and fell, partly compositionally. And the staggered adoption of unilateral divorce laws shows an increase lasting roughly a decade that then dissipated — timing and process changed rather than quantity.

On children: associations are real, causal estimates are substantially smaller, conflict matters more than structure, resources carry a large share, and instability matters more than any single transition.

And the class divergence in family formation is the most consequential current pattern — stable two-earner families increasingly concentrated among the educated, with the timing tracking the deterioration of the labour market for men without degrees.

Demographic transition — the sequence in which mortality falls, fertility follows, and the lag produces population growth.

Second demographic transition — below-replacement fertility with postponement, cohabitation and non-marital childbearing.

Tempo effect — the depression of period fertility by postponement, independent of completed family size.

Period versus cohort fertility — a synthetic measure from this year's rates; the actual completed family size of a birth cohort.

Wealth flows — the reversal of intergenerational resource transfer as children shift from asset to cost.

Gender revolution in two halves — women entering paid work, then men entering domestic work; fertility lowest where only the first has occurred.

Companionate marriage / deinstitutionalisation / capstone marriage — marriage as relationship, as a set of weakened norms requiring negotiation, and as an achievement following stability.

Unilateral divorce — reform permitting one spouse to end a marriage; effects concentrated in the decade after adoption.

Family instability — repeated changes in household composition, which predict worse outcomes than a single transition.

Diverging destinies — the growing gap in the adult time, income and stability available to children by parental education.

One — do the three lives. Write out the reproductive and life-course timeline of a woman in your family in each of three generations. Mark how many years of adulthood each spent with dependent children at home.

Two — find the tempo effect. Look up your country's period total fertility rate and its completed cohort fertility for a cohort that has finished childbearing. The gap is postponement.

Three — check the reversal. Plot female employment against fertility for ten rich countries. Then look up the same plot for 1980.

Four — separate the four claims. Take a statement about family structure and children's outcomes and decide whether it is about structure, resources, conflict or instability. The evidence supports the last two much better.

Five — locate the divergence. For your country, find marriage rates and non-marital birth rates by education level. The gap between the top and the bottom is the pattern this lesson says matters most.