Global Fertility Has Fallen Below Replacement, With Population Peaking Near 9 Billion

Jesús Fernández-Villaverde argues that the world had likely crossed a demographic threshold by 2023 or 2024: humanity as a whole was below replacement fertility. Replacement fertility is the average number of children women need to have over their lifetimes to keep a population constant. The claim is not that the population is already shrinking everywhere. Population momentum—the continued growth produced by large younger generations—can keep the total rising for roughly another 30 years. Fernández-Villaverde puts the global peak at about 9 billion people around 2054 or 2055, with the precise number depending in part on how quickly life expectancy rises.

He distinguishes this process from earlier population collapses. Wars, epidemics and social breakdown can produce more deaths than births, as he said happened in the Roman Empire from the end of the second century onward. But fertility in those cases remained above replacement. The new condition, in his account, is that births themselves are too few to replace the population. China already records about 3.3 million more deaths than births each year, which he compared to losing a city the size of Los Angeles annually.

This projection differs sharply from the United Nations’ 2024 medium variant, which puts the global peak at about 10.3 billion in 2084. Fernández-Villaverde says his simulations are close to the UN’s low variant, but rejects the medium variant as an unreliable forecast. His criticism is that the UN’s estimates and projections do not, in his view, adequately reflect how quickly fertility is falling. His claims about UN estimates of births concern particular country comparisons; his separate discussion of China concerns claims that China’s own reported births may be too high.

The more surprising part of his argument is where the decline is concentrated. It is not confined to wealthy countries. Mexico’s total fertility rate was 1.2 in 2024, he said, citing the country’s official statistics. Tunisia’s was 1.45. He put Mexico’s GDP per capita at purchasing-power parity at 28 percent of the U.S. level, and Tunisia’s at 18 percent. In his account, middle-income countries are now falling below rich countries in fertility—a reversal of a longstanding pattern.

The figures in his presentation came from different years. Fernández-Villaverde’s point was that the pattern did not depend on a single example.

In the United States, he reported, fertility among Hispanic women was around 1.7 and falling quickly; African American fertility was around 1.45–1.46, and Asian American fertility around 1.3. He expected Hispanic fertility to converge with white fertility within a few years.

That pattern complicates familiar explanations based on national wealth or ethnic composition. Fernández-Villaverde noted that cross-country fertility and income had been negatively correlated for centuries: richer places tended to have fewer children. Across OECD countries, he said, that correlation began changing sign in the 1980s. Within the United States, the historical relationship between income and fertility also reversed sometime between 1980 and 2010: richer people now have more children than poorer people. He cited a shift in relative fertility by race as another indication, while acknowledging an audience member’s point that the decline in teenage pregnancy may account for part of the U.S. pattern. His reply was that the same broad reversal is visible in other countries where teenage fertility is not the main explanation.

One consequence is that the standard expectation—that poorer countries will have more children and eventually converge toward the rich world’s fertility level—may no longer hold. Fernández-Villaverde entertained the possibility that, by 2060, the United States could be unusually fertile because it is richer than other countries. He offered this as an illustration of how unfamiliar the new cross-country pattern is, not as a settled prediction.

For most of human history, Fernández-Villaverde said, people lived with both high fertility and high mortality. Depending on the region, women had an average of roughly 5.5 to 6.5 children, while life expectancy was about 35 to 40 years. Fertility exceeded mortality slightly, allowing slow population growth over a very long period.

The first demographic transition began when mortality fell, first in Western Europe and then elsewhere, while fertility stayed high for another 80 to 90 years. That gap produced rapid population growth. Fertility began falling around 1870. Economists have well-developed explanations for this transition, he said, particularly the trade-off between the number of children and the resources invested in each child. When child mortality was high, families had more children partly to ensure that some survived. In rural and small-scale economies, children could also contribute labor. As survival improved and economies became urban and industrial, both incentives weakened.

That framework helped economists understand the decline from five or six children per woman toward the fertility levels of the late twentieth century. It also encouraged the expectation that fertility would settle around replacement, or only modestly below it. Demographers looking at completed fertility—the number of children women have had by the end of their childbearing years—saw apparent stability in some European countries around 1.8 or 1.9. In the United States, women born through 1985 were close to two children on average, Fernández-Villaverde said.

He contrasts completed fertility with the total fertility rate, a real-time estimate based on current age-specific birth rates. The latter can move more sharply and does not necessarily tell us how many children a particular generation will ultimately have. That difference, in his account, helped make the longer decline harder to see. The postwar baby boom, he noted, did not necessarily translate into a comparable rise in completed fertility.

Japan and Italy were early exceptions, with fertility well below replacement, but were often treated as special cases. Over time, Spain, Greece, South Korea and China joined them at what demographers call “lowest-low” fertility, below 1.3. More recently, Fernández-Villaverde said, Chile, Colombia and much of the Caribbean have moved into similarly low territory; even the United States and France have fallen below the old 1.8 benchmark.

He offered two reasons the broader pattern was missed. First, U.S. researchers often saw the world through a country that had remained unusually fertile among advanced economies. He attributed that in part to later secularization, relatively accessible suburban housing and higher fertility among African American and Hispanic women. Second, standard datasets lag the changes. When fertility is falling quickly, delayed or smoothed estimates can obscure how far the trend has moved.

The resulting transition is more than the final stage of the familiar story. In Fernández-Villaverde’s account, births have fallen below deaths in an increasing number of places, while population momentum postpones the point at which the population itself begins to contract. He described this as a second demographic transition: not simply the shift from high to low fertility, but a widening gap in which fertility remains below mortality as population ages.

A striking part of the presentation was the “unpleasant integer arithmetic of replacement.” Children come in whole numbers, so a small change in the distribution of family sizes can move the average substantially. Fernández-Villaverde compared a hypothetical cohort with completed fertility of 1.8 to one at 1.3. In the first, 15 percent of women have no children, 10 percent have one, 60 percent have two, 10 percent have three and 5 percent have four. In the second, the shares with no children and one child rise to 26 and 23 percent, the share with two falls to 46 percent, and larger families become rarer.

The difference in average fertility is large, but it does not require a world in which almost nobody has children. In the lower-fertility cohort, nearly half still have two. What has receded, he argued, is the number of three- and four-child families, alongside an increase in childlessness and one-child families. That helps explain why the decline can appear modest in individual decisions yet produce a major change in population replacement.

For an advanced economy, he put replacement fertility at about 2.1. Reaching it requires a distribution with enough families having three or more children to offset those with none or one. He argued that a modest rise in childlessness and one-child families can make that arithmetic difficult to sustain, even if two-child families remain common. When an audience member asked whether a small number of very large families could make up the difference, he replied that the important change was the disappearance of many three- and four-child families, not a loss of exceptionally large families.

The sharpest disagreement in the seminar concerned the United Nations’ population data. Fernández-Villaverde said UN estimates of births in a number of countries exceed local registry counts, including in countries whose civil registries the UN itself rates as at least 90 percent complete. He showed figures for Colombia, Egypt and Türkiye in 2022 and 2023 that together amounted to more than one million excess births in the UN estimates. These were comparisons of estimates of births that had already occurred, not comparisons of future projections.

For 37 countries with populations above one million and registries considered at least 90 percent complete, he compared 2024 registry counts with UN estimates. The UN figure was 9.6 percent higher overall; only four countries came in above the projection. The comparison included a total of about 16.3 million registry births against about 18.1 million in the UN figures.

He also showed a Colombia series in which the UN revised its estimates for past births between the 2022 and 2024 releases, while Colombia’s statistical agency reported lower numbers. He said he had met with the agency and asked whether its figures were reliable; according to him, its staff stood by them. He said he did not know the source of the large discrepancy. He also questioned why the UN had revised estimates for births in the late 1980s, calling that a warning sign.

The argument extends from estimates of past births to the UN’s projections of future fertility. Fernández-Villaverde and co-author Patrick Norrick worked through the UN methodology and concluded that it assumes a return toward a mean. In his description, countries move between a fertility-transition regime and a low-fertility regime; once in the latter, the model is mean-reverting. The effect is that fertility begins to rise in the projection even when the recent data show a continuing fall. In the UN’s series, he noted, South Korea rises from 0.72 in 2023 to 1.30 in 2100, while China rises from 1.00 to 1.35.

The UN’s stated rationale, quoted on the presentation slide, is that a rebound in low-fertility countries is consistent with continued progress toward gender equality, women’s empowerment and improved social and economic opportunities. Fernández-Villaverde did not object in principle to modelling a rebound, or to acknowledging that future conditions could change. His objection was that the assumption is easy to mistake for a data-driven forecast. In the discussion, he said his complaint was not that the UN used mean reversion, but that the assumption was buried in technical documentation rather than made central to the projection. He agreed with an audience member that uncertainty about the distant future is real and that a recovery could happen.

This distinction matters to how the forecast should be read. A projection that assumes fertility rises under specified social and economic conditions is not identical to a forecast that says the observed decline will probably reverse. The UN’s methodology report, as quoted in the presentation, also acknowledged that the rebound is not universal. Fernández-Villaverde said he did not find the explanation convincing, but the disagreement is partly about what a projection is meant to represent: a conditional scenario, or a best estimate of future births.

The audience also raised doubts about China’s official birth counts, including claims based on BCG vaccination data. Fernández-Villaverde rejected the idea that this evidence implied Chinese fertility as low as 0.5 or 0.6. He argued that even a substantial overstatement of births since 2005 would change the total population by tens of millions, not the roughly 200 million suggested by some claims. His distinction was that births are a flow and population a stock: changing the stock substantially requires large, sustained changes in the flows. Tyler Goodspeed agreed that the lowest proposed fertility estimates sounded exceptionally low, but said the vaccine-based evidence was hard to dismiss outright. Fernández-Villaverde remained skeptical, in part because he said fertility among ethnic Chinese populations in mainland China, Taiwan, Malaysia and Singapore looks remarkably similar.

He argued that methodological choices have consequences beyond demography. Population assumptions feed into policy and climate models, including models used to estimate future emissions. He also described a political constraint: analysts in international institutions may know that local figures conflict with UN projections but find it difficult to publish an alternative series under their own institution’s name. In his account, reliance on a common reference dataset can persist even when researchers inside institutions have reservations about it.

Sub-Saharan Africa is the region where Fernández-Villaverde said the evidence is weakest. He sharply criticized UN fertility and birth estimates there, arguing that many are closer to guesses than reliable counts. The criticism matters to his global projection: if fertility in the region remains high, the timing and scale of world population decline would differ. But he said the available signs point toward a faster transition than standard projections assume.

One relatively strong example, in his account, is the Republic of Congo, where a household survey put fertility at 3.5 in 2023–25, down from 5.1 in 2011–12. He acknowledged that 3.5 remains high and that a simple extrapolation is not a forecast. Still, the speed of the fall resembles transitions elsewhere. Guatemala, he said, went from about 4.5 children per woman to 1.8 over roughly 20 years. An audience member cautioned that falling mortality could affect population growth even as fertility declines; Fernández-Villaverde replied that the total population might still rise for a time.

He also cited survey data from Nairobi, Maputo and Dakar, all of which he said were already below replacement fertility. Large metropolitan areas often lead demographic change, so he did not treat those observations as proof that the whole region had crossed the threshold. His broader expectation was that sub-Saharan Africa could be below replacement by 2050, a claim he explicitly called uncertain and invited listeners to revisit.

Israel is the clear exception in his account. Its total fertility rate was 2.91 in 1991 and 2.86 in 2024, using the boundaries and definitions of Israel’s statistical authorities. But the stable national average conceals a change in composition: Jewish fertility has held relatively steady while Muslim fertility has fallen. Since 2020, he said, Muslim fertility has been below Jewish fertility for the first time since 1948. Within the Jewish population, secular mothers’ fertility was about 1.9; among Muslim women, he put the figure at about 1.8 for non-religious women, 2.2 for religious women and 2.6 for very religious women.

These examples underscore a distinction that recurred in the discussion: national averages can conceal differences across groups, and the future depends partly on whether fertility patterns persist across generations. An audience member raised the possibility that groups with above-replacement fertility could eventually grow to dominate. Fernández-Villaverde agreed that persistence matters, but cautioned that the traits associated with small groups may weaken as the groups expand. He used Mennonites in Bolivia as an illustrative case: he said they made up about 2 percent of the population but 7 percent of births. A projection that simply holds those shares and fertility rates constant would produce implausible results. He said the relevant question is not only how many children a group has, but how likely its defining characteristics are to persist as it grows.

Fernández-Villaverde said the standard economic explanations help account for the long decline but do not, by themselves, explain its speed. The familiar quality–quantity trade-off links rising income and education to greater investment in each child and fewer children overall. Models incorporating the spread of skill-intensive technology can reproduce the timing of demographic transitions across countries, he said. They can help explain why transitions happen at different times and why they accelerate as technology spreads. But he doubted that plausible changes in the skill premium could explain a sudden fall such as the 35 percent decline in births over ten years that he cited for Colombia.

That is the boundary he draws around the established models: they explain much of the first demographic transition, and may account for part of the current decline, but do not yet explain its speed or its reach into middle-income countries. He did not reject the models. He argued that they leave an unexplained component, and that a single economic mechanism is unlikely to account for the whole pattern.

The price of parents’ time may help explain why the income–fertility relationship has changed. Research he discussed argues that wealthier households can buy childcare and domestic services, making it easier for high-income parents to have children. But participants questioned how broadly this applies: informal childcare once supplied by women with fewer outside options has become more expensive as those women gain better opportunities. Fernández-Villaverde agreed that the effect is uneven. High-income households can now outsource more tasks—meal preparation, laundry and other domestic work—but that does not mean the services are affordable to everyone. The audience also pointed out that earlier household technologies, such as the washing machine, did not produce an obvious sudden rise in fertility; Fernández-Villaverde responded that such changes may instead have increased women’s labor supply.

Housing also cuts in opposite directions. Higher prices make homeowners wealthier, but make it harder for renters to buy a home or move to a larger one. The presentation cited research finding that local price rises increased births among homeowners while lowering births among non-homeowners. Fernández-Villaverde said the net effect was not obviously large enough to explain the wider decline, particularly in places where housing is relatively cheap compared with income. The slide also noted that housing prices may matter differently in places where real estate is exceptionally expensive.

He was similarly cautious about public policy as a master explanation. China’s one-child policy is a major intervention, but he compared China with Taiwan, where fertility also fell sharply despite the absence of the same policy. Taiwan’s fertility fell below China’s by the time China lifted the policy, though he acknowledged that the comparison is not airtight: Taiwan developed earlier and had its own, milder policies, and the one-child rule had exemptions and uneven enforcement. His broader point was that fertility among ethnic Chinese populations in several countries looks similar enough to suggest that policy alone cannot explain the East Asian pattern.

Policies intended to raise fertility can have measurable effects, according to the studies summarized in the presentation. Cash benefits in Quebec, Israel and Spain, and family benefits in France and Germany, have been associated with more births or modest increases in completed fertility. But two uncertainties remain: some policies may bring forward births that would otherwise have happened later, and the cost may exceed the benefits. Even small lasting effects—perhaps 0.1 or 0.2 children per woman—could accumulate into substantial population differences over decades. Fernández-Villaverde presented these interventions as positive but limited evidence, not a policy lever known to restore replacement fertility.

Changes in preferences and social norms are another candidate. The presentation listed individualism, secularization, changing views of marriage and descendants, and a more demanding standard of parenting. Research on the United States has also ruled out several potential explanations for the recent decline, including unemployment, housing costs, childcare, contraception and student debt. What remains, in one study’s phrase, is a “Solow residual of fertility”: a measured change that researchers can observe but do not yet explain. Fernández-Villaverde said his account also relies more on elimination than on a settled causal model.

He sees smartphones as an accelerant rather than the origin of the shift. Research discussed in the presentation attributed a substantial share of the U.S. fertility decline among women aged 15 to 44 to the iPhone’s introduction, and another study found a similar effect on teen conceptions in England and Wales. But Italy was already at 1.19 in 1995, before smartphones, and was at 1.14 in 2025. For Fernández-Villaverde, the technology may have made norms and behaviors spread faster: a transition that once took 25 years might now take 10. It did not create the underlying direction.

Fernández-Villaverde’s more speculative explanation is that modernity itself is poorly compatible with high fertility. By modernity, he means a society organized around formal institutions, specialized expertise and large-scale coordination—not simply a capitalist economy. Such societies require standardized knowledge and credentials, and the path to functioning in them involves years of investment. Children are no longer useful household workers; labor-saving technologies have replaced their contribution to farms and workshops. Careers run on schedules set by large organizations, often colliding with the years when people are most likely to have children. At the same time, pensions and health care replace some of the support once provided by adult children, while an impersonal society imposes little social penalty for having none.

The resulting imbalance, in his formulation, is that modern society makes childlessness cheap and a third child expensive. That may help explain why the fertility decline is not limited to the rich. The United States, in his example, has settled around 1.57 among non-Hispanic white women, while countries such as Guatemala, Chile and Colombia have fallen toward 1. His conjecture is that modern institutions and expectations can produce a deeper decline when they arrive in countries with fewer resources.

He linked this argument to the shift from kin networks to formal institutions, including the welfare state. Where relatives once shared childcare and adult children supported parents, public pensions and health care reduce the private cost of having no children. Meanwhile, equipping each child to succeed in a world of formal credentials takes time and money.

The room offered concrete challenges to the idea of what makes children costly. Car-seat rules were raised as one reason a third child can require a larger vehicle. Fernández-Villaverde also pointed to zoning and the shift toward high-rise housing, which he argued makes raising children harder than a suburban home with a yard. Others noted that more access to domestic help could support fertility, while he replied that any increase among richer households must be weighed against the fiscal cost of the additional workers in a welfare state.

He said he was trying to develop a formal model of the conjecture. The seminar did not settle whether modernity produces the same fertility outcome everywhere; his examples—from the United States to Guatemala, Chile and Colombia—were part of his proposed explanation for why similar social changes might lead to different levels of fertility depending on income.

The audience debated whether shrinking populations would necessarily cause a large fall in material living standards. Fernández-Villaverde responded that GDP per capita may not capture the main costs. In small villages in the Cotswolds, he said, declining population can mean the closure of the local pub—not simply a place to drink, but a center of social life. In Philadelphia, smaller school cohorts have led to school closures; he argued that a school can also serve as a community hub, hosting local events and activities.

Others countered that people can move to places with better opportunities. Fernández-Villaverde agreed that population decline will not be evenly distributed, but said this creates “sponge cities”: some locations continue to attract residents while other places lose them rapidly. He cited Japan, where population was falling outside Tokyo and some northern communities had experienced dramatic contractions. Moving may ease the problems in a growing city, but it carries a welfare cost for people attached to the places where they grew up.

A related concern is whether a smaller population can preserve specialized knowledge and institutions. Fernández-Villaverde is working on a model of the minimum scale needed to sustain modern science and technology. He used medical specialization as one example: Spain’s division of its national health system into 17 regional systems, he said, left some smaller regions with too few oncologists experienced in rare cancers. He reported that mortality differences between Madrid and smaller regions, previously similar, had widened. The broader question is whether a smaller population can maintain the people, expertise and institutions needed to keep complex services functioning.

The discussion also touched on the age structure of the population. One participant described the population pyramid turning into a chimney: workers in the middle supporting both children below them and retirees above. Others raised the possibility that longer, healthier lives could change the age at which people remain fertile. Fernández-Villaverde accepted that a longer reproductive window could matter, but argued that longer life expectancy alone would delay rather than reverse population decline if fertility remained below replacement. Even extremely long lives, he said, would not permanently offset fertility below replacement; they would change how long the decline takes.

Fernández-Villaverde rejected the idea that immigration automatically solves the fiscal effects of low fertility, while the audience pushed back on the breadth of that claim. His calculation for Spain considered a hypothetical Colombian immigrant arriving at age 18 without a STEM degree, working continuously until retirement, never drawing unemployment benefits or committing a crime, and then receiving retirement benefits. Including the present value of taxes and welfare-state spending, and allowing for general-equilibrium effects, he said, the person’s net present value for the Spanish government would be minus €250,000.

He argued that the issue is the fiscal structure: in a welfare state, the average immigrant may receive more in public spending over a lifetime than they contribute. He cited Spain’s large immigration inflows over the previous two decades and said relatively few arrivals were net contributors. He also said research from the Netherlands and Denmark found negative fiscal effects in the second generation, which he connected to the persistence of educational attainment. He acknowledged that he did not know the Canadian case as well.

Participants challenged the assumptions behind the argument, including the decision to hold the welfare state fixed. They distinguished between high-skilled immigration, which they saw as beneficial, and lower-skilled migration, which might supply domestic care workers or other services. Fernández-Villaverde agreed that the United States has greater capacity to attract STEM workers than Spain, and accepted the point that policy should distinguish between skill groups. He also said that a country might want a fixed number of highly skilled immigrants but struggle to find enough people with the relevant skills. His objection was to treating large inflows of low-skilled immigrants as a general remedy for population decline in countries with developed welfare states.