The Global Fertility Decline

Total fertility rate is the average number of children a woman would have over her lifetime if she were exposed to current age-specific birth rates throughout her reproductive years. It is not the crude birth rate, which counts births per thousand population in a given year and is heavily distorted by a country's age structure. It is also not completed cohort fertility, which measures what women born in a given year actually produced over their lifetimes, a number that cannot be known with finality until those women reach 49. TFR is a snapshot, taken in the present and projected as a lifetime. It is the cleanest single metric for asking how many children a society is currently producing per woman of childbearing age. Roughly 2.1 children per woman is the level at which a population replaces itself across generations, absent migration and given current mortality.

Seventy-five years of data makes one thing plain. Every region of the world is now at or falling toward sub-replacement fertility. Africa is the last exception, and even there the trajectory is downward. As of 2024, 131 of 237 countries have fallen below the replacement line. Those countries contain 68% of the world's population. This is no longer a story about Europe or East Asia. It is the global majority condition.

The global rate has more than halved since the 1960s. In 1970 it stood at 4.8 children per woman. By 1990 it was 3.31. The 2024 reading is 2.25. The decline did not happen everywhere at the same time. Europe and North America led, falling below replacement by the mid-1970s. East Asia followed, with Japan crossing the line in 1974 and South Korea in 1983. Latin America transitioned through the 1990s and 2000s. India crossed below replacement in 2020. Today only Sub-Saharan Africa, with a regional rate of around 4.3, sits meaningfully above the replacement line, and even there the trajectory is downward. Nigeria, the continent's largest country, has fallen from 6.0 children per woman in 1990 to around 4.8 today.

Three forces have organised this transition. The first is the entry of women into the paid workforce, combined with the contraceptive revolution and rising age at first marriage. This drove the Western inflection of the late 1960s and 1970s. The second is the East Asian transition that began with Japan and spread through Korea, Taiwan, and China as those countries industrialised and urbanised rapidly. China's one-child policy is often credited with the country's fertility collapse, but the steepest decline began before the policy was enacted in 1980 and continued after it was relaxed in 2016. The third is the recent acceleration. From roughly 2015 onward, the ultra-low cohort emerged in East Asia, the United States dropped well below replacement after the financial crisis, and the post-COVID period produced the steepest annual declines on record across most developed economies. In 2024 and 2025, record lows appeared simultaneously in France, Australia, South Korea, China, Japan, the United States, and Hungary.

// Epoch history

The temporal view gives the arc. The spatial view gives the spread. The world today divides into roughly three fertility bands, and the boundaries between them are sharper than the historical curve suggests.

At the high end sits Sub-Saharan Africa. The population-weighted regional rate is around 4.0. Niger, Somalia, and Chad sit above 5.5. Nigeria, the largest country in the region, sits at roughly 4.8 per its most recent demographic survey. Pakistan (3.5), Afghanistan, and a handful of others outside the region still sit above replacement, but the rest of the developing world has already crossed below. Latin America, most of Asia, and North Africa are now sub-replacement countries.

In the middle sits the sub-replacement cluster, with rates between 1.5 and 2.1. This is where most of the Anglosphere now lives. The United States (1.58 in 2025), Australia (1.48 in 2024), the United Kingdom (1.55), and France (1.56) cluster within a fraction of a child of each other. So do most of Latin America and Northern Europe. India crossed into this band in 2020.

At the low end is the ultra-low cluster, with rates below 1.5. East Asia dominates it. Japan (1.15 in 2024) and South Korea (0.80 in 2025) have been in the ultra-low band for years. China joined them recently, with a 2025 reading at or just below 1.0. Italy (1.2) and Spain (1.16) sit near the same level. Germany (1.35) and Hungary (1.38) sit at the upper end. South Korea's 0.80 remains the lowest recorded fertility rate of any country in history.

// What the map shows

The developed world has split. The Anglosphere and Western Europe sit in the high 1.5s. Southern and Eastern Europe sit in the low 1.2s to mid-1.3s. East Asia sits below 1.2. These are not minor differences. A society with a sustained TFR of 1.5 sees each generation shrink by roughly 30%. A society at 1.0 sees each generation shrink by half.

Israel is the OECD outlier. It is the only member country still above replacement, at around 3.0 in 2024. The drivers are unusual enough to warrant their own treatment in Section VI, but the headline matters here. High fertility in a developed country is not impossible, only rare.

The recent moves are fast. The pace of the post-2015 declines is the part the historical chart in Section I obscures. China's TFR was around 1.5 a decade ago and is now near 1.0. Korea fell from 1.24 in 2015 to 0.72 in 2023. France slid from 2.02 in 2010 to 1.56 in 2025, its lowest since 1918.

The total fertility rate is the headline, but it is not the whole picture. A period TFR can fall sharply even if women are not, on average, having fewer children over their lifetimes. To understand what the fertility data is actually telling us, two further distinctions matter.

The first is the distinction between quantum and tempo. Quantum is the question of how many children women ultimately have over their reproductive lives. Tempo is the question of when, within that span, they have them. Period TFR conflates the two. If the average age at first birth rises from 28 to 32 within a decade, the period TFR will fall during that decade even if the cohort of women involved ultimately have exactly as many children as their mothers did. The births did not vanish. They were postponed. Bongaarts and Feeney formalised this in 1998 and proposed a tempo-adjusted TFR that attempts to isolate the quantum component.

The second is the distinction between period and cohort measures. Period TFR is a snapshot of one year, asking what the lifetime fertility would be if current age-specific rates persisted. Completed cohort fertility (CCF) is the actual lifetime fertility of a group of women born in the same year, knowable only when those women reach the end of their reproductive span. Cohort fertility is the truer measure of what a society produced. Period fertility is what we can measure now.

In some countries, much of the decline is tempo

Australia is the cleanest example. The headline period TFR was 1.48 in 2024, a record low. But the completed cohort fertility for women born in 1975, who reached the end of their reproductive years in 2024, was 2.03. Almost at replacement. The gap between 1.48 and 2.03 is largely accounted for by postponement. The median age of mothers giving birth rose to 32.1 in 2024, a record high. France is in a similar position, where INED has argued that births will likely rise modestly in the coming decade as large cohorts enter their childbearing years.

In other countries, the decline is quantum

East Asia is where the story changes. Korean cohort fertility is genuinely collapsing, not merely postponed. The cohort of Korean women born in 1980, who turned 44 in 2024, have completed fertility well below 1.5. The cohort born in 1990, now approaching 35, is on track for even less. China shows the same pattern. The decline is not a story of births delayed and recovered later in life. It is a story of births that are not happening at all, in cohort after cohort, with no recovery on the horizon.

The childlessness component

A useful decomposition that has emerged in recent demographic work is TFR = TMR × CPM. Total Maternity Rate is the share of women who become mothers at all. Children Per Mother is the average number of children among those who do. Korea's decline has been driven heavily by rising childlessness rather than by mothers having fewer children. Korean mothers, when they have children, still average close to two. The collapse is happening because a growing share of Korean women are not becoming mothers in the first place. Japan shows a similar pattern. Australia and France do not, at least not yet.

Why is fertility falling? The question has political weight that has often distorted the answers. The honest reading of the literature is that no single driver explains the trend, that several drivers compound in ways that make their individual contributions hard to isolate, and that the evidence is stronger for some than for others. The strongest cases are economic and educational. The cultural case is real but harder to measure. The biological case is the most contested.

// Female educational attainment

The single most robust correlation in the cross-country fertility literature is the negative relationship between female education, particularly tertiary education, and fertility. It holds across regions, across time periods, and across natural experiments. Where school construction has been studied as a quasi-random shock, the causal direction is clear. More years of female schooling produce fewer children per woman, on average, and the effect is large. The mechanism is not mysterious. Education raises the opportunity cost of childbearing, increases age at marriage and first birth, expands the set of life paths a woman might pursue, and reshapes the social environment in which decisions about family formation happen.

// Housing costs

The case for housing as a major driver has strengthened considerably in the last five years of empirical work. The strongest single finding comes from Couillard's 2025 work at the University of Toronto, which attributes roughly 51% of the US fertility decline between the 2000s and 2010s to rising housing costs. A cross-country panel covering 1870 to 2012 found that a 10% increase in real house prices reduced fertility by 0.01 to 0.03 births per woman, an effect comparable in magnitude to a year of additional female education. HSBC's analysis of Australia found a 10% house price rise associated with a 1.3% drop in the birth rate. Korean research finds a 0.02 fall in TFR per 10% housing price rise. Chinese research from 2025 finds a 0.45% reduction in the probability of childbirth per 10% increase in the price-to-income ratio.

The mechanism is partly direct (children require space, families require larger dwellings) and partly indirect (high housing costs push back marriage and first birth, which feeds the tempo effect from Section III). In countries where housing costs have outpaced wages most aggressively, the fertility effect appears largest. The Anglosphere, Korea, and urban China are the clearest cases.

// Other economic drivers

Beyond housing, the economic literature consistently identifies several other contributors. The opportunity cost of children rises as labour market opportunities expand for women. Childcare costs interact with this in obvious ways. In the United States, where childcare frequently exceeds $10,000 per year per child, the cost burden is large enough to be decisive at the margin for many families. In France and the Nordic countries, where public childcare is widely available, the cost burden is smaller and fertility has held up correspondingly better. Economic insecurity, particularly youth unemployment and unstable employment, also depresses fertility. The post-2008 decline in Southern European fertility tracks the unemployment crisis closely.

// Cultural drivers

The cultural case is harder to measure than the economic one, but is not therefore weaker. Secularisation correlates with lower fertility almost everywhere it has been measured. Religious women, on average, have more children than non-religious women within the same country. Israel is the most striking example, with ultra-Orthodox fertility around 6.4 and secular Israeli Jewish fertility still above replacement at about 2.7. Marriage decline is a related driver. In China, marriage registrations are at record lows. In Hungary, marriages fell from 72,000 in 2021 to about 50,000 in 2024. In Japan and Korea, where extramarital births remain rare (around 2%), declining marriage essentially translates directly into declining births.

The smartphone hypothesis sits awkwardly in this section. The proposition that smartphones have accelerated the post-2010 decline by reducing teen socialisation and partnership formation has been heavily amplified in 2025 and 2026. The evidence is mixed. The underlying 4G study the journalism relies on found a measurable effect on teen fertility but no detrended response in the 25-plus population that accounts for roughly 80% of women of reproductive age. The more measured reading is that smartphones likely contributed to a faster decline in teen pregnancy, which is largely a positive outcome, while the effect on adult fertility is unproven.

// Biological drivers

The biological case is the most contested. The meta-analyses on sperm count trends, claiming a roughly 50% decline since the 1970s, have been criticised on methodological grounds and the magnitude is contested. Rising age at first birth pushes more conceptions into less-fertile years, but this is best understood as a downstream effect of social and economic drivers, not an independent biological force. What seems most defensible is that the biological component, if it exists, is small compared to the social and economic components. The biological story is more often invoked in popular discourse than in demographic literature.

The data establishes what is happening. It does not settle what it means. The debate over the consequences of fertility decline is genuinely live, and the most amplified positions on both sides are often louder than the evidence behind them. Six argument pairs follow, each presenting the strongest form of the position, the strongest form of the counter, and where the evidence currently sits. Swipe or use the arrows to move between them.

The historical record divides fertility outcomes into bands that behave qualitatively differently. Above replacement, populations grow naturally. Between 1.5 and 2.1, populations shrink modestly across generations and can be stabilised by migration. Below 1.5, populations shrink fast enough that recovery becomes empirically rare. Below 1.0, generational halving sets in.

// The recovery question

No country that has fallen into ultra-low fertility (below 1.5) has durably climbed back above 1.5 and sustained it for a decade. Several have approached the threshold. France crossed back briefly in the 2000s and is now well below it again. Sweden moved up to 1.99 in 2010 before retracing to 1.45 by 2024. The Czech Republic moved from 1.13 in 1999 to 1.71 in 2008, then fell back. The pattern is consistent. Brief partial recoveries happen. Durable sustained recoveries do not.

This is the empirical claim that does the most analytical work in the report. It is also the claim with the smallest sample size and the youngest data. Ultra-low fertility as a sustained phenomenon is only about 30 years old in most countries. What can be said with confidence is that durable recovery has not yet been observed, that the burden of proof for its possibility sits with those claiming it, and that the policy record so far provides little reason for confidence.

// Israel

Israel is the developed-world anomaly. TFR has hovered around 3.0 for decades, the highest of any OECD country by a substantial margin. Ultra-Orthodox Jewish women have fertility around 6.4. Religious Jewish women around 3.8. Secular Jewish Israeli women average around 2.7, still above replacement. Israeli Arab fertility has converged toward Jewish fertility at around 2.9.

The drivers are unusual. Religious adherence remains high across most segments of Israeli society, even among the secular. Familism is culturally central. Compulsory military service interrupts but does not delay family formation in the way that university tracks do in most Western contexts. The combination of cultural, religious, and demographic-political factors does not transplant easily to other contexts.

The Israel case matters because it refutes the strong form of the claim that above-replacement fertility is impossible in a developed economy. It does not refute the weaker form, that above-replacement fertility is rare and that the conditions producing it in Israel are not policy levers available to other countries.

The eight focus countries represent four broad policy archetypes: no coherent national policy (USA, Australia, Nigeria), universalist family support (France), conservative pro-natalism (Hungary), and maximalist East Asian intervention (Japan, Korea, China). The empirical record is informative.

// The maximalist case: South Korea

South Korea has spent more than $270 billion over 18 years on pro-natalist policy. The measures have included direct cash payments per child, extended parental leave (now three years shared between parents), housing incentives for young families, childcare subsidies, fertility treatment coverage, and at various points government-organised dating events. The TFR in 2006, when serious spending began, was 1.21. The TFR in 2023 was 0.72. The TFR in 2025 was 0.80. This is the canonical case that policy spending alone does not move TFR.

// The conservative pro-natalist case: Hungary

Hungary has run the most aggressive non-East Asian pro-natalist programme. Spending peaked at around 6% of GDP on family benefits. The instruments include large tax credits scaling with the number of children, housing loans forgiven proportionally with each child born, marriage incentives, and substantial cash subsidies. The TFR rose from 1.23 in 2011 to 1.61 in 2021. The TFR fell back to 1.38 by 2024 and remains there in 2025. The optimistic reading is that the rise was real. The pessimistic reading, most clearly articulated by AEI, is that the rise was substantially a tempo effect, and the retracement to 1.38 supports the tempo interpretation.

// The universalist case: France

France has the longest-running comprehensive family policy regime among the eight focus countries. Public spending on family benefits is around 3.5 to 4% of GDP, distributed across childcare provision, parental leave, tax benefits, and direct payments. The approach is universalist and is integrated with strong public childcare infrastructure. France's TFR in 2025 is 1.56, its lowest since 1918. The French case suggests that even the gold-standard model has its limits, and that the universalist model may be more useful as a brake than as a reversal mechanism.

// Japan and China

Japan has been pulling family policy levers at modest intensity for two decades. The 2024 TFR was 1.15. The Semafor projection for full-year 2025 births is under 670,000, lower than Japan's own demographic institute had projected for 2041. China expanded family policy aggressively from 2021 onward, including the move to a three-child policy, extended maternity leave, cash subsidies for births, and tax incentives. The 2025 TFR is at or just below 1.0. Both cases show maximalist policy effort producing minimal observable effect.

Three major institutions produce long-range population forecasts. They disagree by amounts that translate, by 2100, into billions of people.

The United Nations World Population Prospects 2024 (medium variant) projects global population to peak at around 10.3 billion in 2084, then decline modestly to about 10.2 billion by 2100. The implied global TFR trajectory falls to 2.1 by the late 2040s and to 1.8 by 2100.

The Wittgenstein Centre at IIASA produces alternative projections under the Shared Socioeconomic Pathways framework. The middle scenario, SSP2, projects peak global population around 9.7 billion in the 2070s.

The IHME/Lancet projection, published in 2024, is the most pessimistic of the three. It projects global TFR falling to 1.83 by 2050 and 1.59 by 2100. Population peaks at 9.7 billion in 2064 and declines to 8.8 billion by 2100. By 2100, IHME projects 97% of countries to be below replacement.

// Why they disagree

The largest single source of disagreement is the assumption about the floor for TFR. The UN model assumes that as countries reach very low TFRs, they recover partially toward higher levels. The IHME model assumes much less recovery, with TFR continuing to drift downward in countries that have reached the ultra-low band. The difference between assuming a 1.4 floor and a 1.0 floor compounds into billions of people across a 75-year projection.

The UN's projections have been revised downward across recent revisions. The 2019 projection had global population peaking at 10.9 billion in 2100. The 2022 projection moved the peak to 10.4 billion in 2086. The 2024 projection moved it to 10.3 billion in 2084. Each revision has moved the peak earlier and lower. IHME argues this reflects systematic optimism about TFR recovery that the data does not support.

Three trajectories applied to the eight focus countries, with population in 2075 as the output. The three scenarios are not equally probable. Scenario 1 is the central estimate. Scenario 2 is a plausible downside if the East Asian pattern proves transmissible. Scenario 3 is the upside that policy actors are nominally pursuing.

Scenario 1: Current course. All countries follow the UN WPP 2024 medium variant. Global population rises to 10.3 billion by 2084, with most of the growth in Africa and most of the decline in East Asia and Eastern Europe.

Scenario 2: Korean convergence. All sub-replacement countries' TFRs converge to 0.8 by 2050. This is the pessimistic case, assuming the East Asian ultra-low pattern spreads. Global population peaks earlier and at a lower level. The eight focus countries' projected populations in 2075 are substantially lower than the UN medium variant suggests.

Scenario 3: Partial recovery. All ultra-low countries recover to 1.6 by 2050 and hold. This is the optimistic case, assuming policy and conditions combine to halt and partly reverse the ultra-low pattern. This scenario has no empirical precedent. No country has yet demonstrated the policy or social transformation required.

A single table assembling the key numbers for each focus country. The cells should be read alongside the source notes, which document where each value comes from. The figures will go stale on different timelines. TFR values revise annually. CCF for the 1975 cohort is now final.

Source: National statistics offices for TFR; Human Fertility Database for CCF; UN WPP 2024 for population and projections; OECD Family Database for spending; World Bank for tertiary enrolment.

// Source discipline

Every TFR figure in this report comes from a national statistics office or from the UN World Population Prospects 2024 release directly. Aggregator sites including Macrotrends and Statista publish TFR figures for these countries that differ from the official numbers, sometimes by as much as 0.3 children per woman. Those aggregator sites appear to publish UN WPP medium-variant smoothed projections presented as current data. This report treats only national statistics office releases and the UN's own data tables as authoritative for current TFR. Readers checking the report's numbers against aggregator sites will find discrepancies. The aggregator numbers are the discrepancy.

The eight focus countries are sourced as follows: Australia from the Australian Bureau of Statistics Births release (October 2025); China from the National Bureau of Statistics, supplemented by academic estimates where official figures are contested; France from INSEE Demographic Report 2025 (January 2026); Hungary from the Hungarian Central Statistical Office (KSH); Japan from the Ministry of Health, Labour and Welfare Vital Statistics; Nigeria from the Nigeria Demographic and Health Survey 2024; South Korea from KOSIS; USA from the CDC National Center for Health Statistics.

// Period versus cohort

The report uses period TFR as the headline measure throughout, with completed cohort fertility introduced as a corrective in Section III. Where the two diverge meaningfully, the report distinguishes between the period and cohort interpretations explicitly. Readers who want to follow the academic distinction more deeply should consult Bongaarts and Feeney (1998), "On the quantum and tempo of fertility," Population and Development Review 24(2): 271–291.

// Projection uncertainty

All projections beyond approximately 2040 should be read as ranges, not point estimates. The wider the time horizon, the wider the uncertainty. The UN's own 95% prediction interval for global population in 2100 spans roughly 2 billion people. The disagreement between the UN, IIASA, and IHME forecasts represents a further layer of model uncertainty. Confident point projections for 2075 or 2100 should be treated with corresponding scepticism.

// Known limitations

China's data has reporting opacity. The National Bureau of Statistics has revised historical fertility figures at various points, and some recent academic estimates suggest the actual TFR may be slightly lower than the official figure. The report uses the official figure where available and notes where academic estimates differ.

Nigeria's NDHS 2024 figure of 4.8 differs from the UN WPP figure of 4.3 for the same year. The discrepancy reflects differences in survey methodology and the timing of data collection. The report uses the NDHS figure as the primary value and notes the UN figure for comparison.

The smartphone hypothesis literature is too new for a settled academic consensus to have formed. The report presents the strongest current evidence in both directions without resolving the question.

Three things the data shows, with high confidence. Three things the data does not settle.

// What the data shows

First, sub-replacement fertility is the global majority condition. It is no longer a Western or East Asian anomaly but the dominant pattern for the majority of the world's population. As of 2024, 131 of 237 countries are below replacement, containing 68% of the world's population. The transition is broad and deep.

Second, no advanced economy that has fallen into ultra-low fertility (below 1.5) has durably recovered to sub-replacement and held it. Brief partial recoveries have happened. Sustained recoveries have not. This is the empirical finding that bears the most weight on policy discussions. The burden of proof for confident pro-natalist claims sits with those who claim recovery is achievable.

Third, maximalist pro-natalist policy has not reversed the trend in any of the countries that have attempted it. South Korea has spent more than $270 billion over 18 years and remains the lowest-fertility country in the world. Hungary's substantial policy expansion produced a decade-long rise that has now retraced most of its gains. China's recent pro-natalist push has not slowed the decline. The honest reading of the policy record is that intervention can affect the timing of births and may produce modest sustained effects in some contexts, but has not been shown to durably move completed lifetime fertility at the population level.

// What the data does not settle

First, whether a 2100 global population of 8.8 billion or 10.3 billion is the better central estimate. The methodological disagreement between the UN and IHME is substantive, both teams have credible cases, and the difference compounds into more than a billion people over a 75-year horizon.

Second, whether the consequences of population decline are net negative or net positive on balance. The economic, environmental, and social effects pull in different directions. The strongest case for serious concern is grounded in workforce shrinkage, fiscal pressure from inverted dependency ratios, and the political difficulty of migration as a buffer. The strongest case for equanimity is grounded in productivity offsets, environmental benefits, and the recognition that lower fertility reflects expanded choice for women. Both cases have empirical support and they are not mutually exclusive.

Third, whether the constraints driving the decline are constraints or preferences. The desire-reality gap data supports the constraint reading. The cross-country pattern of higher female education and labour force participation correlating with lower fertility supports the preference reading. Disentangling the two is hard, and the policy implications of the answer are large. If most of the decline is constraint, then in principle policy can move the trend. If most of the decline is preference, policy can engage with it only through measures that almost no one would endorse explicitly.