What Did It Mean to Count Everyone in Japan?

A strip of paper from 702 records people one line at a time. Each entry identifies a person's relationship to the head of a household, name, age, age category, and other information. At the end of a household came totals: people, and the area of state-allotted land associated with them. The document survives only as a fragment of the household register for Kawabe village in Chikuzen Province, in present-day Fukuoka.^[1]

It looks, at first glance, like the beginning of a census. Officials were certainly counting people. Yet the register had not been created simply because a government wanted to know how many human beings lived within its reach. The early statutory state needed information it could use to allocate land and impose obligations. A person became legible through a household and through categories that mattered to administration.

More than twelve centuries later, Japan's first national census, taken on 1 October 1920, tried to do something importantly different. Instead of carrying forward a population from household-registration records by adding reported births and subtracting reported deaths, enumerators were to investigate people directly where they actually were. The Statistics Bureau later described the distinction as a break between records produced in the course of administration and a modern statistical survey designed to observe the population itself.^[2]

Between those two documents lies no simple march from bad counting to good counting. Governments repeatedly learned how to turn people into records. The harder question was what the record was for. Who had to appear? Which household or locality was responsible for them? Did an official need names, categories, or only totals? What happened when people moved faster than the paperwork?

Counting everyone has therefore meant different things in Japanese history. The method changed when the purpose of the count changed.

The Kawabe fragment belonged to the ritsuryō order, the system of laws and administrative institutions consolidated around the seventh and eighth centuries. Household registers were to be compiled at six-year intervals in the provinces and transmitted to the center. The Nara National Museum identifies their purposes as including taxation and labor obligations, military recruitment, and the allocation of kubunden, plots distributed under the land-allotment system.^[1]

That combination explains the document's structure. A government concerned only with a head count might be satisfied with a total. A government distributing land according to legally relevant categories needed more. Age mattered. Relationship to the household head mattered. The household itself mattered because it connected an individual to a unit that officials could locate and revisit.

The lines on the paper therefore did two kinds of work at once. They described people, and they placed those people inside an administrative scheme. A name was not merely evidence that someone existed. It was attached to information that could alter what the state expected to provide or demand.

This is why an old population register should not be imagined as an early version of a spreadsheet waiting to be modernized. The categories were part of the governing system. If land entitlement depended on age and status, then recording age and status was not extra demographic curiosity. The classification helped make the entitlement operable.

It also created problems that every later registration system would face. A register was periodic, while life was continuous. People were born, died, married, moved, fled, or changed status between compilations. Local officials and household heads stood between lived circumstances and the central record. A document could be meticulously ruled and sealed while still depending on information gathered through human relationships.

The surviving Kawabe fragment makes that dependence visible in another way. The register was not preserved forever because it was a sacred census of the population. After its official retention period had passed, its reverse side was reused in 748 for a record connected with copying the Lotus Sutra.^[3] Administrative memory had a life cycle. Information that once helped govern a household could later become useful as paper.

That reuse is a useful warning. Modern historians treasure the fragment because it lets us see individuals in the early eighth century. The officials who made it valued it for a different reason. To understand what their counting meant, we have to ask what the list allowed them to do.

A millennium later, local records again placed people inside households, but the institutional problem had changed. The shūmon aratame, or religious investigation, developed in the seventeenth century as part of Tokugawa efforts to suppress Christianity. The resulting shūmon aratame chō are often translated as religious investigation registers. Their later value to historical demographers is enormous, but they were not invented for demographic research.^[4]

Susan B. Hanley's study of Fujito village in Okayama shows how a system aimed at religious control could become a population record. She notes that investigations were conducted in some places by the 1620s and became nationwide after the Shimabara Rebellion of 1637–38. Over time, the form of the religious registers came to resemble regional population surveys closely enough that they could serve as a basis for wider population counting.^[4]

In Fujito, a copy went to the domain government while a working copy remained in the village. That local copy could be amended as life changed. A death might be marked by pasting a slip of paper onto the deceased person's entry. A birth or a person marrying into a family could be added. Household heads affixed seals to their entries.^[4]

This is a different temporal logic from a register remade only at long intervals. The village copy could become a running account. The official question—who belonged to which household, under what recognized religious and social status?—produced data from which later scholars can reconstruct marriages, births, deaths, household composition, and migration.

But the richness of the source should not make us confuse the historian's question with the administrator's. Hanley found that registration practices varied by domain. In some places children were not entered immediately; in Fujito, very young infants who died before registration could disappear from the demographic record. Later in the nineteenth century, some marriages also went unreported.^[4] A register could be extraordinarily detailed and still fail to capture every event equally.

The imperfection is not a reason to dismiss the source. It is part of what the source teaches. Counting is never only arithmetic. Someone has to decide when a person becomes recordable, which change must be reported, which household should claim an entrant, and what officials will do with an omission.

The Tokugawa village register could therefore know people in a way the ancient central government could not: as members of households whose changes were followed year after year. Yet its categories emerged from a political and religious order of its own. The person on the page was not an abstract resident. He or she appeared through a household, a locality, and an administrative requirement.

In 1721, the Tokugawa shogunate undertook the first of a series of population surveys that historians can follow at national scale. The Statistics Bureau lists twelve Tokugawa-era survey years for which province-by-province tables survive: 1721, 1750, 1756, 1786, 1792, 1798, 1804, 1822, 1828, 1834, 1840, and 1846.^[5]

A sequence like that looks reassuringly modern. Put the totals in columns, compare one year with another, and a population history appears.

Yet the national figure was assembled from an administrative world in which information originated locally and the rules of registration were not identical everywhere. The shogunate did not send a corps of statistically trained enumerators to knock simultaneously on every door. It relied on existing governing structures, including registers whose purposes and coverage had developed before the national aggregation.

This difference matters because a total inherits the decisions inside its components. If a domain registered young children differently from its neighbor, or if people absent without formal permission remained attached to one household while living elsewhere, addition could not remove the discrepancy. Arithmetic can combine categories without making them equivalent.

The national surveys nonetheless represented a major change in scale. Information that had been useful for managing villages and domains could be collected upward to describe the realm. The government could compare provinces and successive dates. People who never encountered one another became part of the same table.

That shift—from a named local population to an aggregate population—changes what rulers can ask. A household register can answer, “Who is this person's head?” A table can answer, “How many people are reported in this province?” The second question does not require the central government to know every name. It requires local institutions to produce numbers in a form that can travel.

This is one reason the phrase “count everyone” is slippery. An authority may seek a total without creating a direct encounter between the state and every individual. The people can be counted through institutions that already claim them.

For historians, the Tokugawa counts and registers are powerful precisely when their institutional origins remain visible. They support arguments about long-term demographic patterns, household structure, marriage, fertility, mortality, and mobility. They do not become stronger evidence when described as a modern census centuries ahead of its time. Their strength comes from understanding how the count was made.

The Meiji state inherited local records but reorganized the relationship between registration and national administration. The Family Registration Law was promulgated in 1871. A nationwide population investigation under the new system followed in 1872, producing the register commonly called the Jinshin koseki. The resulting Nihon zenkoku kosekihyō—the national household-register tables—was the earliest nationwide population statistical compilation of the Meiji period.^[6]

The Statistics Bureau reports that this registration produced a national total of about 33.11 million people.^[7] That number did more than tell officials how large the country was. The new state was centralizing institutions that had previously been divided among domains. A common register helped connect people to a national legal-administrative order.

Again, however, “population” depended on what the underlying record represented. The household registration system tracked legal registration. It could be updated through notices of birth, death, departure, arrival, and other changes. From the 1872 base, officials estimated later population by adding and subtracting reported changes.^[7]

This was efficient. It reused information the government already required for administration. It also seemed to make an expensive separate census unnecessary. If the register contained the population and every change was reported, why send enumerators to count everyone again?

The answer appeared in movement.

Industrialization, urban growth, schooling, military service, employment, and new transport networks all increased the practical importance of knowing where people were living. A person's legal home and actual residence could diverge. If someone moved without the paperwork being fully updated, the registration system could preserve a person in one place while daily life placed that person somewhere else.

The resulting problem was not simply that a clerk had made an arithmetic mistake. It arose because the record had been built to maintain a legal-administrative relationship and was being asked to answer a statistical question about current presence.

The Meiji system thus exposes a distinction already latent in earlier periods. A government can ask, “Where does this person belong in our records?” or “Where is this person now?” Those questions often overlap. They are not the same.

By the late Meiji period, the weakness of registration-based population estimates had become difficult to ignore. The Statistics Bureau notes that city surveys in Tokyo and Kobe showed substantial duplication in populations estimated from household-registration records.^[2]

The problem had a simple mechanism. An estimate beginning from an old registered population remains accurate only if subsequent changes are reported accurately. Births that are reported can be added. Deaths that are reported can be subtracted. Moves that are reported can shift a person from one place to another. Missing notifications leave old information alive.

A dead person can therefore remain statistically present. A migrant can continue to be counted where the register remembers him while also appearing in a survey where he actually lives. What looks like a population problem is partly a problem of administrative time: the record has failed to catch up with the life it describes.

There is a deep continuity here with the 702 household register, though the institutions are otherwise very different. The ancient register was periodic; the Meiji register was designed to be updated through reporting. Both depended on events in households becoming events on paper.

The response could not be to make officials care more about arithmetic. It required a different way of observing.

Statistical reformers had long advocated direct population surveys. A notable experiment took place in Kai Province, present-day Yamanashi, in 1879. Under Sugi Kōji's direction, some 2,000 enumerators participated in a survey that recorded 397,416 people. The Statistics Bureau presents the Kai no kuni genzai ninbetsu shirabe as a trial for a modern census.^[8]

The word genzai—present or current—is revealing. The survey was interested in people as a population existing at a particular moment, rather than only as names maintained in a register. But a national census still did not follow immediately. The household register was already there, and building a new statistical apparatus was expensive and politically demanding.

The old system persisted not because officials were incapable of imagining another method, but because an existing information system had institutional weight. A count can survive after its limitations are known if it remains useful for other purposes.

Japan's first national population census took place on 1 October 1920. The Statistics Bureau describes it as the beginning of modern statistical investigation in Japan. Its key methodological difference was direct inquiry: the survey investigated the target population itself and combined multiple attributes rather than merely editing administrative records created for other purposes.^[2]

The census did not abolish household registration. The two systems answered related questions and continued to coexist. That coexistence is the point.

The 1920 results exposed the accumulated problem in the older estimates. The registered-home population and estimates based on it contained substantial overcounting because deaths, removals from registers, and movements had not all been reported. The Statistics Bureau notes the paradox that the population tied to permanent registers could exceed the population actually found by the census in ways that showed the presence of fictitious or duplicated entries.^[7]

A direct census made absence visible.

That statement sounds strange: how can counting people reveal people who are not there? It does so by changing the reference point. A register starts with an established legal record and asks what officially reported changes have happened to it. A census starts with a date and asks who is present in the surveyed population under its rules. When the results differ, the discrepancy becomes evidence about the record system itself.

The first census also changed the relationship between individuals and aggregate knowledge. Earlier national totals had been assembled from administrative units. The modern census made a coordinated observation campaign into a national event. Enumerators, forms, instructions, publicity, and a common reference date were all parts of the counting mechanism.

This did not produce perfect knowledge. No census does. People can be missed, misunderstood, or placed in categories they would not choose for themselves. But the intended object had changed. The primary goal was no longer to infer the population from a register maintained for another administrative purpose. It was to measure the population as a statistical population.

That distinction gave the count a different authority. It could be compared with the register instead of merely inheriting it.

The 702 Kawabe register, the Tokugawa religious investigation registers, the national population tables of the eighteenth and nineteenth centuries, the Meiji household-registration statistics, and the 1920 census all put people into numbers. They did not define “the population” in the same way.

The ancient state needed households it could tax, assign obligations to, and use in land administration. Tokugawa authorities made religious and household belonging recordable, then turned local information into larger population totals. The Meiji state created a more uniform national registration system and tried to keep a running population from reported changes. The census finally separated a statistical investigation from the legal register strongly enough to test one against the other.

Each method solved a problem left by an earlier one while creating its own conditions of visibility.

A frequently updated household register can tell us more about family changes than an occasional census. A direct census can locate people whom a legal register has left attached to an old home. A province-level aggregate can reveal the scale of a realm without giving the center a file on every individual. No method is simply “more information” in every direction.

This is why counting everyone is a historical question rather than an arithmetic one. Before the addition begins, someone has already decided what counts as a person present, a household member, a resident, a subject of obligation, or a case that must be reported.

The ruled lines on the Kawabe fragment made those decisions visible in 702. The census forms of 1920 made different ones. Between them, Japan did not gradually discover that people could be counted. Governments repeatedly changed what they wanted a count to mean.