The Year the Lines Crossed: How Japan Became a Country of More Deaths Than Births

In 2005, for the first time since modern vital statistics began, more Japanese people died than were born. Twenty years later the annual gap is not 21,000. It is 918,000. This is a reading of that crossing, and of the marriage and divorce totals that follow, against the official record through 2025.

The year the chart changed was not dramatic on the ground.

Shops opened. Trains ran. A little more than a million Japanese children were born, which still sounded like a large number if you said it quickly. What changed was a relationship that had held, with a three-year hole for the war, since the Ministry began counting in 1899: births no longer exceeded deaths.

In 2005, 1,062,530 Japanese children were born in Japan and 1,083,796 Japanese people died. The difference — shizen zōgen (自然増減), natural increase — was −21,266. The Ministry of Health, Labour and Welfare recorded it as the first natural decrease in the history of the survey, setting aside the missing years 1944–1946. (MHLW, 2005 confirmed Vital Statistics)

Twenty years later the same subtraction produces a different kind of number.

In 2024, the confirmed count is 686,173 births and 1,605,378 deaths. Natural change: −919,205. In 2025, still an approximate annual total, the figures are 671,236 and 1,589,489. Natural change: −918,253 — the nineteenth consecutive year of deficit. (MHLW, 2024 confirmed; MHLW, 2025 approximate annual total)

The crossing was not a moment. It was a hinge. Everything on the chart after it is the sound of a door that did not close.

How to read this chart

Terracotta is births. Steel blue is deaths. Both are the Ministry’s Vital Statistics of Japan (人口動態統計) for Japanese nationals in Japan, 1947–2025. The 2005 marker is the first year deaths exceeded births. 2006 is the one-year uncross — a surplus of 8,223. The shaded gap is the continuous deficit from 2007. The dashed tips are 2025 approximate figures, not yet the confirmed count. 1947–1972 exclude Okinawa, as the source tables do.

That series has rules, and they are easy to break.

It counts Japanese nationals in Japan. Foreign residents who give birth or die in Japan, and Japanese nationals who do so abroad, are tabulated separately. The census, by contrast, counts everyone usually resident. The 2025 census population of 123.05 million is therefore a different object from the Ministry’s birth and death totals. Mixing them in one sentence as if they were the same population is the most common error in writing about this subject. (Statistics Bureau of Japan, 2025 Census preliminary)

Confirmed and approximate are not interchangeable. The Ministry publishes a monthly bulletin of approximate figures (gaisū 概数), an annual total of those figures each June, and a confirmed count (kakuteisū 確定数) the following September. This article uses 2024 confirmed (released 16 September 2025) and 2025 approximate (released 3 June 2026). The 2025 total fertility rate of 1.14 will be recalculated against the 2025 census population and may move.

2005 was the first crossing, not the start of an unbroken decline. 2006 returned to a small natural surplus. The continuous run of deficits begins in 2007. The Ministry’s 2025 overview calls 2025 the nineteenth consecutive negative year. That arithmetic only works if 2007 is year one. The chart marks both facts: the 2005 crossing, and the gap that opens only after 2006.

A monthly flash is not the same series. The Ministry’s sokuhō (速報) counts registration forms, including foreigners in Japan and events from previous years. It is useful for spotting a turn. It is not a number you set next to 671,236 and call a trend. This article does not use 2026 monthly flashes as if they were Japanese-national annual totals.

These rules are the difference between reading the figure and decorating a narrative with it.

The latest official totals

The 2024 confirmed release is the one to learn first, because nothing in it is still moving. Births were the fewest since the survey began in 1899. The fertility rate of 1.15 undercut the previous year’s 1.20. Deaths were the most on record. The natural deficit of 919,205 was the largest confirmed gap the Ministry has ever published. Marriages rose. Divorces rose slightly. (MHLW, 2024 confirmed overview)

2025 did something the chart-watchers had stopped expecting: deaths fell. Not by enough to close anything — they still exceeded births by 918,253 — but by 15,889, the first decline in five years. Births fell again, to 671,236. Marriages rose for a second year, to 489,119. Divorces fell, to 179,068. Mean age at first marriage: 31.0 for husbands, 29.7 for wives. (MHLW, 2025 approximate overview)

Two years of more weddings and two years of fewer babies is the sentence most coverage dropped. It is the sentence the rest of this article is about.

The birth line: a country that used to replace itself

On the opening figure, the birth line is not a smooth decline. It is a mountain range.

The first peak is the postwar baby boom. 1949 remains the record year: 2,696,638 births. The second peak is 1971–1974, the children of that boom having children. 1973 is the summit of that range: 2,091,983. The Ministry’s 2025 overview still prints both numbers as the two markers a reader is meant to carry. From 1975 the line falls, with brief recoveries, and from 2016 it has fallen every year.

Between the two mountains there is a notch so sharp it looks like a printing error.

1966 was a Hinoeuma (丙午) year — the Fire Horse in the sixty-year sexagenary cycle. A superstition that girls born that year would be difficult wives, with roots in the Edo-period legend of Yaoya Oshichi, produced the largest one-year collapse in the modern series. Births fell from 1,823,697 in 1965 (TFR 2.14) to 1,360,974 (TFR 1.58), then rebounded to 1,935,647 in 1967 (TFR 2.23). That is a 25 percent drop and a recovery that overshot the year before. Demographers have treated it less as a lost generation than as a three-year timing shift: couples who could wait, waited. The National Institute of Population and Social Security Research still names the 1966 drop as one of the irregularities that give Japan’s population pyramid its shape. (IPSS, 2023 projection summary)

2026 is the next Fire Horse year — the same sixty-year combination explained in Galloping into 2026: Understanding the Year of the Horse in Japan. There is no official forecast that the superstition will repeat, and this article does not offer one. The 1966 notch was a decision available to a married generation that was still having two children. The married generation of 2026 is smaller, older at first birth, and already below one-and-a-half. A superstition can shift a peak. It cannot invent a peak that is not there. The Ministry’s 2026 confirmed count will be published in September 2027. Until then, anyone selling a Fire Horse baby bust is selling a story the data have not written.

What the data have written is slower and harder.

The total fertility rate — the number of children a woman would have if she lived her life at this year’s age-specific rates — was 1.57 in 1989, the so-called 1.57 shock, because it undercut the 1966 Fire Horse rate of 1.58 and made the postwar floor look like a ceiling. It was 1.26 in 2005, the year of the crossing. It was 1.15 in 2024 and 1.14 in 2025. Replacement is about 2.1. Japan has not been near it in a generation.

The national rate is also a lie of averages. In 2024, Tokyo’s fertility rate was 0.96. Okinawa’s was 1.54. Both are official. Neither replaces a population. (Tokyo Metropolitan Government, 2024 vital statistics; MHLW 2024 confirmed prefecture ranking)

A fertility rate is not a birth count. The rate can stabilize while births keep falling, because the number of women aged 15–49 is itself the echo of earlier declines. That is why 2025 can produce 671,236 births at 1.14 and still be a new low on both measures at once. The birth line is not only about how many children people want. It is about how many people are left to want them.

The death line: the country getting older in public

Deaths are the easier line to misunderstand, because they look like a verdict on health. They are mostly a verdict on age.

On that same figure, the death line falls as medicine and nutrition do their work, to a modern low of 670,342 in 1966 — the same year as the Fire Horse notch, which is coincidence of the calendar, not of the cause. From the late 1970s it turns up, not because Japanese people started dying younger, but because the first baby boom started arriving at the ages at which people die. The Ministry’s 2025 overview says deaths exceeded one million in 2003 and then rose, with 2024 the record, and 2025 the first decline in five years.

That decline is real and should not be talked away. It is also not a rejuvenation. People aged 75 and over have accounted for more than 80 percent of all deaths since 2024. The leading causes in 2025 were the causes of an old country: malignant neoplasms 378,812 (23.8 percent), heart disease excluding hypertensive disease 220,447 (13.9 percent), senility 214,711 (13.5 percent). (MHLW, 2025 approximate overview)

IPSS’s 2023 revision is blunt about what happens next. The number of people aged 65 and over keeps rising to a peak of 39.53 million in 2043. The share aged 65 and over keeps rising after that, from 28.6 percent in 2020 to 38.7 percent in 2070 under the medium-fertility, medium-mortality path, because the working-age population shrinks faster than the elderly one. The old-age dependency ratio — people 65+ per 100 people aged 15–64 — goes from 48.0 in 2020 (2.1 working-age adults per older person) to 74.2 in 2070 (1.3). (IPSS, 2023 revision)

The death line is not a pandemic story, and it is not a failure of hospitals. It is the first baby boom, leaving.

The marriage gate: almost no one walks around

In most rich countries births and marriages have drifted apart. In Japan they have not, because the law and the culture still treat them as nearly the same event.

In 2024, 18,093 children — 2.6 percent of all births — were born outside marriage. The official category is chakushutsu de nai ko (嫡出でない子). The other 97.4 percent were legitimate births. (MHLW Vital Statistics 2024, table 4-30)

That is the load-bearing fact of the marriage count. A country in which two or three births in a hundred occur outside marriage cannot offset a collapse in weddings with a rise in unmarried parenthood. In much of Europe the two series have decoupled. In Japan they have not. The marriage count is not a lifestyle series. It is a leading indicator of the birth count, with a lag of one to three years and a remainder that never arrives.

The marriage peak is 1972: 1,099,984 couples, the second baby boom pairing off. The 2025 approximate count is 489,119 — a second consecutive rise, after 485,092 in 2024, and still less than half the peak. Mean age at first marriage in 2025 is 31.0 and 29.7. The people who marry are older, and there are fewer of them.

They still say they will. The Institute’s 16th National Fertility Survey, taken in 2021, found that 81.4 percent of unmarried men and 84.3 percent of unmarried women aged 18–34 intended to marry eventually — down from 85.7 and 89.3 five years earlier, and still a large majority. Those who said they never would: 17.3 percent of men, 14.6 percent of women. Among those who intended to marry, the average number of children hoped for was 1.82 for men and 1.79 for women, the first time the women’s figure has fallen below two. Married couples still named an ideal of 2.25 and a planned 2.01. The finished number among wives 45–49 was 1.81. The most common reason for having fewer children than the ideal remained the cost of raising and educating them. (IPSS, 16th National Fertility Survey)

The gap between “I will marry” and “I have married” is now a demographic statistic of its own. IPSS’s Population Statistics of Japan puts the 2020 proportion never married at age 50 — the old “lifetime unmarried rate,” which is an average of the 45–49 and 50–54 unmarried shares, not a claim about the whole of life — at 28.25 percent for men and 17.81 percent for women. (IPSS, Population Statistics, table 6-23)

So the 2024–2025 marriage uptick is worth taking seriously and not worth mistaking for a baby boom. The Ministry has not published a causal breakdown. A rebound among couples who waited through the pandemic is a plausible part of it. None of it enlarges the cohort of women now in their late twenties and early thirties. A gate that admits more people in a given year still admits them from a shrinking line.

Divorces: a smaller institution produces a smaller aftermath

Divorces are the quietest official count and the easiest to moralize. The official series does not moralize.

The count stays low through the 1960s, rises with the marriage boom, and peaks in 2002 at 289,836 couples. It has declined, with interruptions, ever since. 2024 confirmed: 185,904. 2025 approximate: 179,068. A falling divorce count in a country with falling marriages is partly a smaller stock of marriages to dissolve. It is not, by itself, evidence that marriages have become happier, or more brittle, or anything else the divorce count is often asked to prove.

What the divorce count does confirm is the scale of the institution. At the 1972 peak, Japan recorded 1,099,984 marriages and 108,382 divorces. At the 2002 peak it recorded 289,836 divorces against 757,331 marriages. In 2025 it recorded 489,119 and 179,068. Marriages and divorces have come down together because they are attached to the same thing: a smaller married population.

What a crossing feels like, twenty years on

Natural decrease is an accounting identity. A country feels it as empty classrooms, shuttered shopping streets, hospitals that cannot staff night shifts, and municipal offices that no longer have enough children to keep a school open and enough elders to keep a clinic closed.

The 2025 census, which counts everyone usually resident, not only Japanese nationals, put the population at 123.05 million on 1 October 2025 — 3.097 million below 2020, a 2.5 percent decline, faster than the previous five years. Population fell in 1,558 of 1,719 municipalities, 90.6 percent of the country. Households rose to 57.125 million because people are living in smaller groups: 2.15 persons per household. (Statistics Bureau of Japan)

The houses remain after the households thin. The 2023 Housing and Land Survey found 9,002,000 vacant dwellings, 13.8 percent of the housing stock, both records. Of those, 3,856,000 were vacant and not listed for rent, sale, or second-home use — the akiya (空き家) that sit in the landscape as a physical form of natural decrease. (Statistics Bureau of Japan, 2023 Housing and Land Survey) How foreign buyers are being steered toward that stock is a separate question: Navigating Japan’s Akiya Opportunity.

Labour is the other surface. Employers reported 2,571,037 foreign workers as of 31 October 2025, a record in a series that began in 2007. Foreign residents at the end of 2025 were 4,125,395. Those figures belong to a different article — Japan Wanted Foreigners. Now It Is Deciding How Many Is Too Many — and they are not a rebuttal of the crossing. They are what a labour market does when the crossing has been running for twenty years and the working-age population is the second baby boom, now in its fifties.

IPSS already builds immigration into the official future. Under the medium-fertility, medium-mortality path, total population falls from 126.15 million in 2020 to about 87 million in 2070, dropping below 100 million in 2056. The high-fertility variant (TFR 1.64) still yields 95.49 million. The low-fertility variant (TFR 1.13) yields 80.24 million. A projection limited to Japanese nationals reaches 77.61 million, with 40.9 percent aged 65 and over. The 87-million path is not a closed country. It is a country that takes in people and still shrinks.

The projection above is what the births-and-deaths figure cannot show: the future already in the age structure. Everyone who will be 65 in 2070 has already been born. Everyone who will be 45 has already been born. The next twenty years of the death line are, to a first approximation, a schedule.

What the chart does not decide

A fertility rate of 1.14 is not a moral failure and it is not a policy success. It is a year.

The Institute’s own survey says most unmarried adults still intend to marry and still want something close to two children. The Ministry’s own totals say they are not doing so at replacement, and that the number of people who could is falling. The Children and Families Agency, created in April 2023, and the child-allowance expansion that followed, exist inside that gap. This article does not award them a result they have not claimed. The 2025 birth count is a new low. That is the result that is on the page.

Nor is Japan a unique ruin. Other rich countries in East Asia have fertility rates at or below Japan’s. What is distinctive is the combination: a non-marital birth share of 2.6 percent, a death line driven by a baby boom that was once 2.7 million children in a single year, and a political argument that has begun to treat foreign residents as a quantity to be managed. The crossing made all three of those arguments inevitable. It did not decide how they end.

If you came to this chart for a single number, the honest one is not 671,236, and it is not 1.14. It is the relationship that changed in 2005 and, after a one-year pause, did not change back: more deaths than births, every year, for nineteen years, while the generation that could reverse it got smaller.

The lines are still crossing. They just did it a long time ago.

Figures are from the Ministry of Health, Labour and Welfare Vital Statistics (Japanese nationals in Japan), the Statistics Bureau of Japan, and the National Institute of Population and Social Security Research. 2025 vital totals are approximate annual figures pending the confirmed release. A full source ledger is in `docs/cultural-sources.md`.