How Many People Die Alone in Japan?

Quick answer

Japanese police handled 204,184 bodies in 2024. Of those, 76,020 were people who had been living alone and died at home.

That is not the answer to the question, though it is the number most often given. Japan’s government does not treat all 76,020 as dying alone in the sense the phrase carries. In 2025 the Cabinet Office settled on a measure for the first time and applied it to the same data. Counting only those found eight or more days after death, the figure is 21,856. Counting from four days, 31,843.

So there are three numbers, and they differ by a factor of three and a half. The one Japan now stands behind is 21,856.

Inside that number, two things sit badly with the way the subject is usually described. 17,364 of the 21,856 were men, close to four in five. And the largest single age band was not the very old but 70 to 74, with people aged 85 and over accounting for 1,705, under eight per cent of the total.

The figure the English-speaking web repeats

The English Wikipedia entry on kodokushi offers a national figure of 32,000 elderly people in 2009, attributed to NHK, alongside 1,049 lonely deaths in Tokyo in 1994 and around 2,200 in Tokyo in 2008. Its most recent national number is 37,227 for the first half of 2024, described as roughly 70 per cent aged 65 and over.

Two of those are worth checking against the police release they come from. The 37,227 is real: it is the provisional count the National Police Agency published on 21 August 2024, the first time this series appeared. The 70 per cent is not. In the full-year 2024 table, 58,044 of the 76,020 were aged 65 and over, which is 76.4 per cent.

The larger problem is age, not arithmetic. The 32,000 figure describes 2009 and counts only the elderly. It has been carried forward for fifteen years into sentences written in the present tense. It is the same shape as the claim that a hundred thousand people vanish in Japan every year: a figure with an origin far in the past, repeated until it became the answer. It is also how the sentence about foreign advisers taking 33.7 per cent of Japan’s national budget has survived for decades, uncited, in an encyclopaedia entry nobody divided by the budget the Ministry of Finance still publishes.

The same isolation this series measures also turns up in a place few English-language accounts look for it. In the studies that asked why elderly people offend, the circumstance recorded most often was not hunger or destitution but the absence of family, which is the finding behind the number of elderly people in Japanese prisons.

What the police actually counted

The Japanese title of the statistic is 警察取扱死体のうち、自宅において死亡した一人暮らしの者, which translates as “among bodies handled by police, persons who died at home while living alone.” The word kodokushi does not appear in it. Neither does the word for isolation.

That wording is exact about what the data can support. A body enters this count when the police handle it, which happens when a death is unattended or its cause is not immediately clear. It records where the person died, that they lived alone, their age band, and how many days passed between death and discovery. It does not record whether anyone had been in contact with them, whether they had family, or whether they wanted to be found.

The count also begins recently. The first release covered the first half of 2024 and appeared in August of that year. The first full year, 2024, was published on 1 April 2025. Before that, national figures for this were estimates built by private research institutes from municipal data, mostly Tokyo’s, which is why the numbers in circulation vary so widely and cite so many different years.

The word had no definition until 2025

A working group convened by the Cabinet Office met for the first time on 29 August 2023 to answer a narrower question than it sounds: what, for the purpose of counting, is an isolated death?

Its report notes at the outset that some local governments compiled statistics on people who died at home, that private organisations had built estimates from them, and that “the definitions and the underlying reasoning vary.” Japan had a word in common use, municipal data of differing shapes, and no national measure.

The report settles two definitions. The conceptual one is: dying with nobody present, and the body being discovered only after a period of time has passed. The operational one, the one that can be applied to a table, takes the police data as its base and specifies that the place of death is the home, that the household type is a person living alone, that suicides are not excluded because isolated death is not a question of cause, and that there is no reason to restrict the count to the elderly.

The remaining problem was the period of time. The report is direct about it: it is not possible to define an isolated death as one discovered after a fixed number of days. What it does instead is set a threshold as a measuring device, choosing eight days or more as the point at which prior social isolation can be strongly inferred, and offering four days or more as a secondary reference. It then adds, in its own cautions, that this “is not a statement that a particular number of days elapsed after death constitutes an isolated death.”

Japanese has two words for it, and the government chose the other one

Before the working group could decide what to count, it had to decide which word it was counting. The report sets out the distinction plainly.

孤独 (kodoku) is a subjective concept. It refers to the mental state of feeling alone. 孤立 (koritsu) is an objective one. It refers to having no connection to society, or very little, and no help, or very little.

From there the reasoning is short. A death accompanied by kodoku, a subjective state, cannot be established after the fact. A death accompanied by koritsu, an objective state, can be inferred from evidence that outlives the person. So the working group discarded the first word and proceeded with the second. What Japan now counts is 孤立死 (koritsushi), isolated death, not 孤独死 (kodokushi), lonely death.

The English-language term in general use is “lonely death,” a translation of the word the Japanese government set aside precisely because nothing about it can be measured. Japan’s own legislation keeps both: the Act on Promotion of Policy for Loneliness and Isolation (Act No. 45 of 2023) defines a single combined state of “loneliness and isolation” and declines to separate them, because for the purpose of helping people the distinction does not matter. For the purpose of counting the dead, it is the whole problem.

The estimate is 21,856

Applying that threshold to the 2024 police table gives 21,856 deaths discovered eight or more days after they occurred. At four days, 31,843.

The full distribution is where the shape of the thing becomes visible:

28,756 people, 37.8 per cent of the total, were found within a day. Most of these deaths are discovered quickly, by a neighbour, a carer, a delivery driver, a relative who calls. The image the word carries, of a body undisturbed for months, describes the tail of this distribution: 6,945 people were found after more than a month, and 253 after more than a year.

Both halves of that are true at once, and neither is the whole answer. That is the reason the government needed a threshold before it could publish a number at all.

Four in five were men

Of the 21,856, 17,364 were men and 4,466 were women, with 26 unrecorded. Set against the age bands, the imbalance is not spread evenly:

In the 55 to 59 band there are seven and a half men for every woman. By 85 and over the two columns have nearly converged, at 882 and 823.

The gap closes because the population changes shape: women outlive men, so the oldest brackets contain far more women to begin with. What the table shows is that in middle age, where men and women are present in roughly equal numbers, it is overwhelmingly men who die at home without anyone noticing for a week. A man in his late fifties living alone in Japan is in the demographic this statistic is really about.

The connection to a subject Japan has been counting for longer is hard to miss. When the Cabinet Office first surveyed social withdrawal in the 40-to-64 age bracket in 2018, that bracket turned out to be larger than the youth bracket officials had been measuring for years. The government’s current estimate is about 1.46 million people aged 15 to 64 in social withdrawal, roughly one in fifty. These are different measurements of different things, and no official source links them. They describe adjacent conditions in the same households.

The oldest are found first

Compare the two tables by age and something inverts.

Among all 76,020 who died alone at home, the largest band is 85 and over, at 14,658. Among the 21,856 found after eight days or more, the same band supplies only 1,705.

Expressed as a share of each age band, the pattern is consistent:

Nearly four in ten deaths among people in their fifties and sixties go undiscovered for over a week. Among those aged 85 and over, roughly one in nine does.

The oldest are the most likely to die alone at home and the least likely to lie there. Japan’s long-term care insurance system, home-visit services and municipal watch programmes are built around old age, and the people inside them are seen on a schedule. The police table cannot show why the shares fall this way, and no official source draws the causal line. What it does show is that the state’s attention is distributed by age, and that the gap it leaves is in middle age.

Fukuoka 50.3 per cent, Fukui 23.5 per cent

The police publish the same 2024 count broken down by prefecture, alongside the total number of bodies each force handled. Dividing one by the other gives the share of police-handled deaths that were people living alone at home, and that share is not close to uniform.

The top of the table is more than double the bottom. Tokyo’s Metropolitan Police, handling by far the largest volume at 25,667 bodies, sit at 40.4 per cent, above the national figure but below five prefectures with a fraction of its population. Osaka is at 40.0 per cent.

The two ends do not sort into city and countryside. Tottori, Yamagata and Fukui are rural and sit at the bottom; Yamaguchi and Kagoshima are also rural and sit near the top. Tokyo is above the national line, Ibaraki and Tochigi are well below it. Whatever the table is measuring, it is not population density.

The police release carries no analysis with these figures, and we have not found an official one. Anyone reaching for an explanation will reach for household structure, and there is a plausible version of that argument, but testing it would require joining this table to census data on household composition, which nobody has published. The numbers are set out here as they were released.

Something else recorded in this same police series varies by prefecture too, and by a wider margin: whether the body is autopsied at all, which ran from 2.5 per cent to 23.0 per cent across the 47 forces in 2025. That is a different year from the table above, and the two spreads do not line up with each other.

2025 is running higher

Provisional figures for January to June 2025, published on 8 August 2025, give 40,913 deaths at home among people living alone, out of 111,036 bodies handled. The comparable half-year figure for 2024 was 37,227, so the count is running about ten per cent higher.

Applying the eight-day threshold to the 2025 half-year table gives 11,669. The full-year 2024 figure was 21,856, so the first half of 2025 is already tracking above half of it.

The direction is not surprising. The National Institute of Population and Social Security Research projects one-person households to keep growing for another two decades, and the Comprehensive Survey of Living Conditions counted 9.031 million single-person households headed by someone aged 65 or over in 2024. Against that denominator, the 58,044 people aged 65 and over who died alone at home in 2024 work out to roughly one in 156. On the narrower measure, the 15,630 in that age group found after eight days or more are about one in 578.

Which number answers the question

If somebody asks how many people die alone in Japan, the honest answer is that the country published a national count of any kind for the first time in August 2024, and agreed on what to count only in 2025.

Given that, the three numbers are answers to three different questions:

• 76,020 is how many people died at home in 2024 while living alone, and were handled by police. Most were found within days.
• 31,843 is how many were found after four days or more, the government’s secondary reference point.
• 21,856 is how many were found after eight days or more. This is what the Cabinet Office calls the estimate of isolated deaths, and it is the figure to use.

None of the three is 32,000, and none of them describes only the elderly. The count that Japan now stands behind is smaller than the one in circulation, more male, and younger.

Where else we looked, and what we could not verify

• The NHK figure of 32,000 for 2009. We have cited it as it appears in the English Wikipedia entry, because that is where English-language repetition of it begins. We could not locate the original NHK broadcast or an archived page for it, so it is quoted here as a circulating figure rather than as evidence.
• A definitive figure for total single-person households. The Comprehensive Survey of Living Conditions gives the number for households headed by someone aged 65 and over, which is what the age-specific rates above use. Whole-population household totals from the census and from the Institute’s projections are compiled on different bases and different dates, so we have not combined them into a single rate.
• Causes of the age gradient. The falling share of delayed discovery with age is visible in the police table. The reasons for it are not in that table, and we have not found an official analysis that tests them, so the paragraph above describes the pattern and names the systems that exist without asserting that they produce it.
• The prefectural spread. The 2.1-fold gap between Fukuoka and Fukui is in the police release. The explanation is not. Three-generation households are more common in some of the low-ranking prefectures, and that is the obvious hypothesis, but the household-composition figures we could find at prefectural level are private compilations of census data rather than a published official table, and joining them to this series would be our own analysis rather than a reported finding. We have left it as an open question.
• Municipal series. Tokyo’s medical examiner’s office has published counts of unattended deaths for far longer than the national series exists. We have not used them here, because the national table now covers the same ground on a consistent basis and mixing the two would repeat the problem that made the older estimates so inconsistent.