A new NBER paper by Daron Acemoglu and co-authors reports that over the past seven decades lower birth rates went together with faster growth in output per worker and no loss of aggregate GD. The scarcity of young workers redirects innovation towards labour-saving technology. Today’s situation is different, as the fertility decline of the past decades may not have the same consequences.
For half a century, economists had a tidy story about why rich countries have fewer children. Following Becker (1981), the two workhorse ideas were the quantity–quality trade-off — parents trade family size for investment per child — and the opportunity cost of a mother’s time. Both predicted the same thing: as incomes and female wages rise, births fall.
This story, which can handily describe fertility patterns of the last centuries, no longer fits the data of the last few decades. In their survey of the field, Doepke et al. (2023) show that the old regularities have weakened or reversed outright. Within high-income countries the income–fertility relationship is now roughly flat; across them it has turned positive. The cross-country correlation between female labour force participation and fertility, once firmly negative, is now positive: the countries where most women work are the countries where most children are born. What drives fertility today, they argue, is the compatibility of career and family — childcare provision, cooperative fathers, permissive social norms, flexible labour markets. And they close by flagging the question the literature has barely begun to answer: what ultra-low fertility and rapid population decline actually do to an economy.
The optimistic case
In “Baby Busts and Growth Booms,” Acemoglu (fresh off his Nobel prize) and coauthors (Acemoglu et al., 2026) marshal seven decades of data — a large cross-section of countries and, separately, US commuting zones — to ask whether the demographic pessimism is borne out by the historical record, looking at birth cohorts up to 1980. Their answer is a flat no.
The headline result is a strong negative relationship between birth rates and subsequent growth in output per working-age adult. Across countries, a birth rate one percentage point lower in 1950 predicts GDP per worker roughly 27 per cent higher over 1970–2020. Similar results are found for US commuting zones. Lower fertility, in other words, went with faster growth in income per worker, not slower.
More striking still is what does not happen to the aggregate. One might expect the per-worker gain to be arithmetic — fewer workers, same output, higher ratio — with total GDP falling. It doesn’t. The estimated effect of 1950 birth rates on aggregate GDP growth is essentially zero, and aggregate earnings across commuting zones are likewise unaffected. The per-worker gain is large enough to fully offset the smaller workforce.
This rules out the textbook explanation. A neoclassical model does predict a temporary rise in income per worker when population growth slows, because the capital–labour ratio rises — but it also predicts lower aggregate output, a falling capital stock, and dissipation of the per-worker effect over time. The authors find none of these; capital stocks rise. They then work through the obvious alternatives and dispatch them one by one: rising female participation, educational upgrading, and the shift out of agriculture, all with limited to no support.
What is left is the mechanism the authors favour, drawn from Acemoglu’s (2010) work on labour scarcity and directed technical change: when young workers become scarce, firms and inventors redirect effort towards technologies that economise on them. The supporting evidence is broad. Lower birth rates predict more labour-saving and automation patenting, rising high-tech export shares, employment shifts towards high-tech industries, and — crucially — higher total factor productivity. A one-percentage-point lower birth rate twenty years earlier is associated with total factor productivity (TFP) about 15 log points higher thirty years on, comfortably exceeding the 9.6 log points of direct drag from a smaller workforce.
The cleverest piece is a test leveraging historical data. Using cross-country variation in Second World War deaths, the authors separate civilian deaths, which shrink the population without much changing its age structure, from military deaths, which do both. Total war deaths predict lower subsequent GDP per worker; military deaths predict higher. It is the scarcity of the young, not the loss of population as such, that triggers the productivity response — a pattern which echoes Bergeaud et al.’s (2025) findings on French regions after the First World War.
If this is right, the alarm sounded by the IMF, the OECD and a long line of macroeconomists is misplaced. Depopulation is not a growth problem; it is an innovation stimulus.
Why this may not be the case anymore
The main issue with this paper is the timeframe of the data used. We are now in a different world from the post-WWII era, in which baby-boomers lived in increasing prosperity and opted for smaller families and stronger gender equality.
In 1950, the country with the lowest birth rate in the sample was Austria — with a total fertility rate of 2.09, essentially replacement. Even in the most recent vintage the authors can use, 1980, the extreme case is Germany at a TFR of 1.56, with Denmark just below. South Korea today is at 0.75. These are not points on the same line; they are different regimes.
The arithmetic makes the point. A TFR of 1.56 means each generation is about 76 per cent of the last — roughly a 0.9 per cent annual decline, leaving some 40 per cent of the initial population after a century. A TFR of 0.75 means each generation is 37 per cent of the last: about 3.35 per cent a year, leaving around 3.5 per cent after a century. Even if both may be coded as “low fertility”, the story they describe is very different, and no data point in their analysis has fertility at comparable levels to the ones observed today in South Korea.
Three further issues are closely related to the main one. First, migration. The low-birth-rate observations are precisely the countries that absorbed large immigration flows — Germany above all — while the high-fertility comparators, such as Mexico, sent emigrants. The coefficient therefore bundles in an endogenous migration response.
Second, timing. The engineers powering South Korea’s technological vitality today were born in the 1980s and 1990s, when fertility was far higher. The children not born after 2000 would still be in school. Their absence is not yet visible in any productivity series, and will not be for two decades.
Third, and most fundamentally, the ideas channel runs the other way. Ideas are produced by people and their human capital, and fewer people eventually mean fewer ideas, especially in rich countries where education levels have been very high for decades and there is little scope for further improvement.
Here the distinction that matters is between the direction and the level of innovation. Acemoglu et al. (2026) provide good evidence on direction: scarcity bends research towards labour-saving techniques.
On the level, the direct evidence points the opposite way. Wong, Man and Li (2022) match US birth cohorts from 1931–84 to patents per thousand residents in 1996–2012. They find that a one-standard-deviation larger cohort (about 70,600 births) raises per-capita patents by 0.23, roughly a quarter of the outcome’s standard deviation. The paper has limitations, but the sign is consistent with the story we have recounted. More births, more inventors, more ideas — the very level effect that the labour-scarcity mechanism must outrun.
What follows
Acemoglu et al. (2026) establish something worth establishing: the transition from high fertility to around replacement was not growth-destroying, and labour scarcity genuinely does redirect technology. That should temper the crudest doom-mongering, and it is a useful corrective to projections that assume mechanically that fewer workers means proportionally less output.
What remains to be assessed is what happens when fertility falls from 1.5 to 0.7. The evidence for that proposition does not exist, because the experiment has never been run.
Which leaves us with an asymmetry. If the pessimists are wrong, we will have over-invested in childcare, family policy and pension reform — and Doepke et al. (2023) tell us fairly precisely where such policies bite. That is a modest cost, and much of it is worth paying on its own terms. If the optimists are wrong, we will find out around 2050.
References
- Acemoglu, D. (2010). “When Does Labor Scarcity Encourage Innovation?” Journal of Political Economy 118(6): 1037–1078.
- Acemoglu, D., D. Autor, K. Beirne and A. Scott (2026). “Baby Busts and Growth Booms: Demographic Change and the Macroeconomy.” NBER Working Paper No. 35401, National Bureau of Economic Research, July.
- Becker, Gary S., Treatise on the Family, Cambridge: Harvard University Press, 1981; Enlarged edition, 1991.
- Bergeaud, A, J Chaniot and C Malgouyres (2025), ‘DP20492 Escaping Labor Scarcity: Innovation and Human Capital after WW1 in France’, CEPR Discussion Paper No. 20492. CEPR Press, Paris & London. https://cepr.org/publications/dp20492
- Doepke, M., A. Hannusch, F. Kindermann and M. Tertilt (2023). “The Economics of Fertility: A New Era.” In S. Lundberg and A. Voena (eds.), Handbook of the Economics of the Family, Vol. 1, Amsterdam: North-Holland, pp. 151–254.
- Fernández-Villaverde, J., G. Ventura and W. Yao (2025). “The Wealth of Working Nations.” European Economic Review; earlier version NBER Working Paper No. 31914 (2023).
- Wong, J. T. H., M. Hei Man and A. Li Cheuk Hung (2022). “Population and Technological Growth: Evidence from Roe v. Wade.” arXiv preprint arXiv:2211.00410.
