The number is 5.6, and it arrived on Monday the way these numbers always arrive: as a headline on a phone propped against a cereal box, read sideways by somebody halfway through getting out of the house. China’s birth rate, the lowest since 1949. Below it, the sentences that always follow, about a country not having enough children and about what that will do to it.
The number is accurate. It is also not measuring the thing almost everyone reading it will assume it measures, and the gap between those two facts is worth ten minutes, because it changes what the headline is about.
What a crude birth rate counts
Five point six is births per thousand people. Not per thousand women, not per thousand women of an age to have children. Per thousand people, all of them, including every retired person in the country.
That denominator is the whole problem. It means the rate can fall without a single person changing her mind about anything, simply because the composition of the population underneath it changed.
And it changed enormously. The number of Chinese women of childbearing age fell by sixteen million between 2020 and 2025. Sixteen million, in five years, which is not a change in behavior but a change in who is alive and how old they are, and which traces back to a policy that limited births decades ago and thereby limited the number of people available to have children now. The one child rule ended in 2015. Its arithmetic did not.
So a good part of the fall in 5.6 is a fact about the shape of a population, arriving on schedule, decades after the decision that shaped it. Nobody decided 5.6. It was set, largely, by people who are now in their seventies.
The number that is about choices
If you want the figure that describes what people are actually doing, it is the total fertility rate, and China’s for 2025 comes out around 1.02. That is the average number of children a woman would have across her life at current rates, and it is calculated in a way that holds the age structure still, which is exactly what the crude rate fails to do.
1.02 is a startling figure. Roughly one child per woman, against the 2.1 at which a population replaces itself without migration. It is low by any standard, including the standards of other places with very low fertility.
It is also a period measure, which carries its own distortion. It answers the question “if this year lasted forever, how many children would a woman have,” and if a generation is simply starting later, a period rate will understate what they finish with. Some of 1.02 is probably delay rather than absence. Nobody yet knows how much, and the only way to find out is to wait until the women in question are in their forties and count.
What moves it, then
Here is where I expected to find a clear answer and did not.
The instinct of most governments facing this is to pay people. The evidence on whether that works is genuinely divided, and the division is technical enough that both sides get quoted confidently by people who have not read either.
One side says the payments move timing, not totals. Work on early South Korean schemes found that more than seventy four percent of the money went to births that would have happened anyway, which is to say the state paid for children already on their way. On this account a bonus persuades a couple to have in 2026 the child they were going to have in 2028, the rate rises, and then it falls back as the borrowed years arrive empty.
The other side has better data than it used to. A study following Australia’s baby bonus tracked the women it applied to all the way out to 2022 and found no spacing effect at all, but 6.8 percent more children completed. That is a real change in totals, not a rearrangement of dates, and it is the strongest evidence anyone has that money can do more than shuffle the calendar.
Broader Korean work sits in between, putting the effect at a fraction of a percent of births for every ten percent added to a transfer. Small, positive, and nowhere near the scale of the fall it is meant to answer.
The things that are not money
What runs through the research more consistently than any payment is the cost of the room.
Not the cost of the child, which people are famously bad at estimating in advance and which they absorb when it arrives. The cost of the space to put the child in, and of the hours someone has to be in the room with it. Housing price relative to income, the availability of childcare and what it costs, whether a job survives a year away from it, whether the hours are compatible with collecting somebody at four o’clock.
Those associations are consistent across countries and they are also, awkwardly, associations. A society that makes childcare cheap is usually a society that has decided several other things too, and disentangling the childcare from the rest of it is not something the data can do cleanly.
And the single most reliable relationship is the plainest one. The later the first child, the fewer the eventual children, because the window is not negotiable and everything that pushes the first birth later shortens it.
What nobody has done
No country has raised its fertility rate substantially and kept it raised. Several have slowed a decline. A few have produced a bump that flattened once the policy stopped being new. Nobody has reversed it.
That is a statement about sixty years of attempts across a great many countries with very different politics, and it should probably make everyone more modest, including the people confident that the right combination of payments and leave and nurseries has simply not been tried yet. It might not have been. It also might not be the kind of thing that answers to a policy.
What is left is the part that no rate describes. Somewhere behind 5.6 and 1.02 is a very large number of ordinary decisions, made in kitchens, about whether this is the year, made by people weighing a flat, a job, a parent who is getting older, and a feeling about the next ten years that nobody has yet worked out how to measure. The statisticians will keep publishing the rate every January because it is what they can count.
