Who Was Counted in Japan's Population?

In 1869, Sugi Kōji began a population survey in Suruga, in what is now Shizuoka Prefecture. The undertaking encountered an objection that had little to do with arithmetic. After the domain's return of its lands and people to the imperial court, senior officials argued, it was improper for the domain to conduct such an investigation when the court had not done so. The survey was curtailed; tables for Numazu and Hara were completed.^[1]

Sugi was trying to establish how many people could be found. His opponents were concerned with who could undertake that inquiry. The people themselves had not disappeared during the disagreement. What failed was the arrangement for turning their presence into an accepted account.

Japan had kept records of people long before Sugi's experiment. Ancient registers organized households and their obligations. Tokugawa village books recorded religious affiliation and changing membership. Twentieth-century censuses coordinated enumeration across a much larger territory. These were substantial achievements, but their numbers answered different questions.

To understand who was counted, it helps to follow the decisions made before anyone added up the names: which household a person belonged to, when a life became visible to the record, and which place could claim the person in its total.

A surviving fragment of a register from 702 gives each person a line. It concerns a community in Chikuzen Province, in northern Kyushu. Members are arranged by kinship, and the end of a household's entry records the total land allotted to it. The account belongs to the ancient state's system of household registration, prepared at six-year intervals for purposes including taxation, service, and land allotment.^[2]

The arrangement reveals what the list was expected to accomplish. Each individual appeared within a relationship that made the information usable. A name supplied an identity; placement within the household connected that identity to other people and to an administrative unit. The land total then attached a material consequence to the grouping.

Counting in this setting was therefore already a sophisticated act of classification. An official did not simply need to know how many bodies occupied the countryside. The government wanted an account through which demands and allocations could be organized. The household supplied a way of reaching people without treating every person as an entirely separate administrative undertaking.

That choice also determined what the document left outside its immediate concern. The record could preserve a person's position in an allocation system without describing everything that person did during the day. A household entry was an account prepared for governing, rather than a miniature biography of everyone included.

The interval between compilations introduced a further consideration. A register supplied a dated description that had to remain usable while people grew older and households changed. Whatever other records or local knowledge helped officials between compilations, the surviving sheet cannot be treated as a continuously refreshed view of the community.

Even its survival reflects an administrative decision. After its retention period had passed, the reverse of this particular paper was used for a Buddhist scripture-checking record dated 748.^[2] The earlier names survived because the sheet acquired another use. For the officials who no longer needed the register, its useful life as a population record had ended. For historians, the reuse helped preserve evidence of how that population had once been organized.

The object thus carries two selections: the ancient administration's decision about whom and what to record, and the later circumstances that allowed this portion of the record to remain available.

Almost a thousand years later, a register from the Hidamori district in present-day Niigata began with an assurance about religion. Its 1672 preamble declared that the community contained no Christians, including among its women, men, old people, young people, and servants. Niigata's prefectural historical account identifies it as a record from the developing system of religious and population inspection.^[3]

The declaration explains why the compilers needed more than a village total. To answer for the absence of prohibited affiliation, they had to identify the people for whom the assurance was being made. Household membership and temple affiliation helped construct that answer.

One grouping under a headman named Denbei contained twenty-two people, including relatives, servants, and dependent cultivators with their families. The curator also notes that twelve houses, including a temple, were reported, while only six household groups appeared. The suggested explanation is that some dependents occupied separate dwellings without being recognized as independent household units.^[3]

A modern reader who turns every group into one roof would consequently produce a misleading settlement. Administrative dependence could draw people together on paper even when domestic arrangements separated them on the ground. Counting houses, counting recognized households, and counting individuals were three different operations.

There was inequality within the entries too. The same archival account notes that women could be identified as someone's wife, mother, or widow without a personal name.^[3] Such a woman might contribute to the population total while remaining harder for a later historian to follow as a named individual. Numerical inclusion did not require equal documentary individuality.

The books also acquired a capacity to follow movement. In the Kuroda village records described by Toyota City's historical dictionary, surviving intermittently from 1774 onward, entries concern marriage, service away from home, births, and deaths. More generally, recognized moves between communities required corresponding changes in the registers.^[4]

This made the work recurrent. A marriage could alter which household answered for a woman; employment could require an account of someone absent; a death could remove an obligation attached to a living person. Village officers needed information from people whose lives were changing, while those people needed their changed circumstances recognized.

For historical demographers, the repeated entries offer an exceptional resource. Kitō Hiroshi emphasizes that sufficiently long runs allow researchers to connect the histories of individuals and families, rather than merely compare isolated totals.^[5] A succession of lists can reveal movement that one list conceals. Its usefulness comes from the ability to recognize the same person again—and to investigate why that person ceases to appear.

Annual registration created a particular blind spot. A child could be born after one compilation and die before the next. When that happened without a surviving supplementary entry, neither event necessarily appeared in the sequence of household lists.

Demographer Kinoshita Futoshi uses an illustrative case to explain the problem: a child born in August 1800 and dying that December could pass entirely between two annual registrations. This is his analytical example, not an identified infant whose family history he reconstructs. It shows why births and infant deaths are liable to underregistration in these sources.^[6]

The omission is especially consequential because it selects by survival. Children who live long enough to appear in a later register have a better chance of entering the historian's reconstructed population. A calculation based only on those visible children begins after some of the most vulnerable lives have already dropped out of the evidence.

Moving the annual date would change which intervals fell on either side of it. It would not, by itself, convert a periodic list into a continuous record of events. To capture a birth and a death occurring between compilations, someone must record those events when they happen, or recover them afterward through another source.

Kinoshita therefore distinguishes the information obtainable from a child's first registered year from a direct measurement of mortality during the first year after birth. His study investigates the relationship through additional demographic evidence and modeling.^[6] The important lesson for reading the register does not depend on adopting one universal correction percentage. The denominator has to include the lives exposed to the risk being measured.

Read alongside Kitō's account of long register sequences, Kinoshita's argument changes how the evidence should be used. The books can follow a marriage or years of household change with considerable detail. Reconstructing the births from which a generation emerged requires an additional inquiry into lives that never reached a surviving list.

Nor can the absence of a name, by itself, tell us how a family valued a child. Before attributing the silence to parental feeling, a historian must examine the interval between registrations and the survival of the records. The book may have missed a death that mattered intensely to the people around it.

The Meiji government did not begin with an empty page. Population estimates were compiled from 1872 using a population defined through permanent domicile registration. Subsequent annual estimates added registered births and other entries, subtracted registered deaths and removals, and made adjustments concerning Japanese nationals overseas.^[7]

This was an economical way to use an existing administrative structure. Once a baseline had been established, recorded changes could update the account. It also made the quality of each later estimate depend on the completeness and timing of those changes. A total could be newly calculated while still inheriting the limitations of its starting list.

Permanent domicile was a registered attachment. It need not coincide with the place where a person slept on a particular night. For some purposes, retaining that attachment was precisely the point: an administrative relationship should not have to be recreated every time someone traveled. For a question about the population actually present in a locality, however, the same continuity could obscure movement.

Sugi's next major experiment tackled that difference. The Kai survey of 1879 was designed around people actually living there, rather than an inspection whose purpose was to verify the people assigned to a household register. The reference moment was midnight at the close of December 31.^[1]

Kai, corresponding to Yamanashi Prefecture, was also a deliberately manageable testing ground. It was near Tokyo, contained a relatively small population, and had comparatively limited movement. The survey was intended to establish practical methods, likely difficulties, and approximate costs before a national undertaking.^[1]

The choice of a less mobile population is revealing. It made the experiment easier, but mobility was among the difficulties a wider survey would have to confront. The people most readily located in a trial were not necessarily those who would most challenge its methods elsewhere.

The contrast with 1869 goes beyond the success or failure of one statistician. His earlier investigation had encountered a dispute about authority. The later project treated enumeration as a method to be tested and extended. Permission, organization, and a workable definition of the people being sought had to come together.

Nor did presence simply supersede registration as the correct answer to every question. A list maintaining recognized membership and a survey measuring actual location could both be useful. Their coexistence made comparison more demanding: a discrepancy might disclose a failure to record a change, or it might arise because the two systems deliberately attached the same person to different places.

A national census required local enumerations to refer to the same moment. The decree issued in September 1918 fixed the first census for midnight at the beginning of October 1, 1920. An earlier law, enacted in 1902, had envisaged an initial census in 1905; war with Russia and subsequent financial difficulties postponed it.^[8]

The delays put a material limit around the ambition. Agreement that the country should be counted did not supply the funds or personnel required to count it. A national inquiry needed preparations extending well beyond the night to which its answers would refer.

The 1920 operation counted people at their location at the reference time, using paper questionnaires completed by respondents. Its archival catalogue distinguishes this actual-presence basis from the usual-residence basis of later censuses.^[9] The common moment allowed forms completed in different places to describe one intended population rather than a rolling procession of arrivals and departures.

The midnight rule did not require every answer to be written at midnight. Its purpose was to make later answers refer back to a common instant. Without such an instruction, a traveler might be included where a household remembered having him and again where an enumerator subsequently found him. The rule supplied a way to resolve that problem, although it still depended on people understanding and applying it.

Hiroshima's surviving example form from 1920 makes the instructions concrete. Its eight groups of questions included household position, sex, date of birth, marital relationship, occupation and occupational position, birthplace, and registration or nationality categories. Occupation was divided into principal and subsidiary work. A separate line emphasized the precise reference time.^[10]

The subsidiary-work field is worth noticing. A person could be counted once while being described through more than one economic activity. Recording a second occupation could alter the account of local livelihoods without adding another inhabitant. A headcount and a description of the population were being assembled from the same form.

Public cooperation was part of the undertaking. Hiroshima's archive preserves posters and notices intended to explain the date, purpose, and method and to reduce suspicion or uncertainty. One proclaimed that the census's accuracy depended on the people's cooperation.^[10]

This was a different public relationship from the 1672 declaration that a village contained no Christians. The older document answered a demand for religious assurance through an organized community. The census publicity asked inhabitants to help produce an accurate statistical description. Both needed local knowledge, but they explained the request in different terms.

Neither a reassuring poster nor a common form could make every answer reliable. Their significance lies in the work they were designed to perform: creating enough shared understanding for millions of separate statements to be combined.

The census definition changed again in 1950. Earlier censuses had generally counted the population actually present; the new basis assigned people according to usual residence. The historical account maintained by the Japan Institute for Labour Policy and Training connects the change with the uses to which the results would be put.^[11]

The difference can be explained without making either method absurd. In an illustrative case, someone spends the census night away from home. An actual-presence count associates that person with the place of the visit. A usual-residence count seeks the place where the person ordinarily lives. Both can be correctly executed while producing different local totals.

The second approach is useful when the inquiry concerns the continuing population of a community rather than everyone temporarily there. But it also requires more information. An observer can ask where a person was on a specified night. Establishing usual residence requires rules for interpreting a pattern of living, including absence and the possibility of more than one address.

This was not a return to the ancient household register. Usual residence concerns the person's ordinary location; permanent domicile registration maintains an administrative attachment. They may coincide, but they are not interchangeable definitions. Confusing them would erase the very distinction that Sugi's nineteenth-century experiment had made important.

A later change shows how even accurate local totals can answer a misleading comparative question. For the 1960 census, densely inhabited districts were introduced after large-scale municipal mergers had blurred the older distinction between urban and rural areas. A municipality called a city could now encompass extensive countryside.^[12]

Adding up its inhabitants correctly did not establish how many lived in a densely settled urban environment. The administrative boundary and the settlement pattern had diverged. The response was to define another area for statistical analysis, rather than assume that an inherited label still described the same kind of place.

The historical movement was therefore not simply toward larger and more accurate totals. It also involved reconsidering what the totals should describe. A person might belong to a registered household, be present in a town, usually reside in a village, or live within a densely settled part of a much larger municipality. Choosing among those relationships changed the result before the addition began.

The population series now used to study the years before 1920 were not all produced by a succession of identical contemporary surveys. The Statistics Bureau describes a retrospective series for 1899 onward built from the 1920 census count of Japanese nationals in Japan proper. Registered demographic events and overseas migration were used to work backward; further adjustments linked the results with the earlier estimates.^[7] The nationality and territorial limits belong to the definition of this series. It is not a count of everyone throughout the Japanese empire.

A smooth line across the nineteenth century can consequently conceal a change in the evidence beneath it. The line is an analytical achievement: records made for different purposes have been related to one another. It is not a surviving observer who counted every year by the same procedure.

This returns the question of inclusion to the reader. Before comparing two numbers, we need to know whether their compilers sought registered members, people present, or ordinary residents; whether short lives could fall between observations; and whether the geographical units remained comparable. These questions make comparison possible rather than forbidding it.

Japan's population was counted through changing arrangements for recognizing people. Household ties could make a person visible to an official. A recurring inspection could preserve a long life while missing a brief one. A national reference night could locate someone away from home; a later rule could place that person back at the usual residence for statistical purposes.

Sugi's abandoned survey exposed the authority needed to ask. The later history exposes the decisions needed to answer. A population total becomes intelligible when we recover both: who was entitled to collect the information, and what relationship made each person one of the people being counted.