What is Civilisation | How a Society Counts Itself — Census, Population Data, Registers and Official Statistics

45. Identity systems and statistical systems solve different problems

It is tempting to treat national identity systems, civil registration, population registers and census statistics as interchangeable because all of them contain information about people. They are not interchangeable. An identity system is primarily concerned with establishing that a particular person is who they claim to be, usually so that rights, services, transactions or legal responsibilities can be attached to the correct individual. A statistical system is concerned with producing aggregate descriptions of a population. The same person may appear in both systems, but the purpose changes the architecture.

This distinction matters for data minimisation. An identity credential may need only the attributes necessary to authenticate a transaction. A statistical programme may need age, household structure or housing conditions but should not automatically expose those variables in an identity service. When systems are linked, purpose boundaries should remain legible. Interoperability should not become permission for every institution to see everything another institution knows.

The distinction also matters when records disagree. A census response saying that a person usually lives at one address does not necessarily mean an identity register must immediately overwrite its legal address field. Different systems can use different reference concepts. Reconciliation therefore needs rules rather than an assumption that one database is universally authoritative.

Civilisation becomes safer when it knows what each register is authoritative for. A birth register can be authoritative for a registered birth event. A land registry can be authoritative for registered title under the relevant law. A statistical register can be authoritative for a statistical population frame. Clarity about authority prevents convenient data from silently becoming universal truth.

46. Base registers are the quiet scaffolding behind modern administration

Many advanced administrative systems rely on a small number of foundational or base registers: people, businesses, addresses, buildings, land parcels and sometimes public organisations. The exact design differs by country, but the idea is powerful. Instead of every ministry inventing its own version of an address or business identity, the system establishes reference sources that other services can reuse under defined governance.

The benefit is not simply efficiency. Shared reference data reduces semantic drift. If a school system, tax authority and emergency service all use different spellings and identifiers for the same street, integration becomes expensive and error-prone. A governed address register gives those systems a common coordinate. The same principle applies to enterprise identifiers and geographic codes.

Yet base registers create concentration risk. A flawed reference record can propagate widely. That means the foundational layer needs unusually strong quality assurance, correction channels, version histories and service-level expectations. Downstream systems should know when a record changed and whether they need to refresh dependent data. Critical reference systems also need continuity arrangements because an outage can affect many services simultaneously.

This is a recurring pattern in civilisation: a small set of apparently boring standards carries a disproportionate share of coordination. The more systems reuse a foundation, the more carefully that foundation must be maintained. Reuse is not the end of maintenance; it is the reason maintenance becomes important.

47. Business registers let civilisation count organisations as well as people

A society does not run only through households. Firms, charities, schools, hospitals, associations and public bodies also need to be counted and classified. Statistical business registers provide a frame for measuring enterprises and establishments. They help economic surveys know which organisations exist, what industries they operate in, how large they are and where activity occurs.

This is harder than it sounds. A legal company can own many operating establishments. A multinational can contain subsidiaries across sectors. A small business can start or close without immediate statistical visibility. Informal enterprises may not appear in tax or company records. Corporate restructuring can make one continuing activity look like multiple births and deaths in the data.

The distinction between an enterprise and an establishment matters because economic questions differ. If we want to know corporate ownership, the legal enterprise may be the relevant unit. If we want to know how many workers are employed in manufacturing in a particular town, the local establishment matters more. Statistical systems therefore maintain relationships among units rather than pretending one identifier answers every question.

Business registers complete the civilisation picture. Population statistics describe people and households; business statistics describe organised production. Together they let societies study employment, productivity, industrial change, regional economies and resilience. A civilisation that can count only its people but not its productive organisations is still partially blind.

48. Economic censuses and population censuses are cousins, not twins

Some countries conduct economic censuses or establishment censuses that collect detailed information across businesses at multi-year intervals. These operations resemble population censuses in scale and logistical complexity but serve different statistical universes. The unit may be a firm, establishment, farm or production site rather than a person or household.

Economic censuses can benchmark national accounts, update business registers, reveal industrial structure and provide detailed geographic information that sample surveys cannot support. They are especially useful when economic transformation is rapid and administrative sources do not fully capture informal or newly emerging activities.

The same design lessons recur: definitions must be clear, frames must be complete, response burden must be controlled, confidentiality must be protected and results must be released fast enough to remain useful. Business data can also be commercially sensitive, so disclosure protection may need to prevent identification of dominant firms in small industries or locations.

Seen together, population and economic censuses reveal what civilisation is trying to do when it counts. It is building a periodic map of agents and capacities: people, households, dwellings, organisations, farms, factories and services. The map is never the economy or society itself, but it allows coordinated institutions to reason about scale.

49. Occupation and industry classifications turn millions of jobs into comparable structure

If a census simply recorded free-text job descriptions, the result would be a linguistic ocean. One person writes “teacher”, another “primary school educator”, another “classroom teacher”, and another describes the employer instead of the job. Statistical classification turns these responses into structured categories that can be aggregated and compared.

Occupation and industry are different concepts. Occupation describes the kind of work a person does; industry describes the primary activity of the establishment or organisation where the work occurs. An accountant can work in a hospital, a bank, a factory or a university. The occupation may remain accounting while the industry changes. Confusing these dimensions produces misleading labour statistics.

Classification systems require maintenance because economies change. New technical roles appear. Old industries shrink or merge. Platform work blurs traditional categories. Green-transition activities cut across existing sectors. Statistical offices need rules for coding ambiguous responses and methods for mapping revisions so long-term series remain interpretable.

This is another example of civilisation manufacturing a common language. The classification is not nature. It is an engineered vocabulary that makes coordination possible. Its legitimacy comes from transparency, consistency and fitness for analytical purpose.

50. The cost of a census is really the cost of temporary national coordination

Censuses are expensive because they concentrate work that usually occurs in smaller streams. Millions of households may need contact within a limited period. Temporary workforces must be recruited, trained and supervised. Devices or paper forms must be procured. Maps must be updated. Advertising must reach almost everyone. Call centres, cybersecurity, translation, logistics and quality assurance all peak around the same operation.

Judging a census only by its headline budget can therefore be misleading. The relevant question is cost per unit of durable public information and the degree to which investments strengthen the wider statistical system. A modern address frame, improved geospatial data, reusable collection platforms and trained staff can support later surveys and administrative work if designed deliberately.

Cost control also requires resisting questionnaire inflation. Every additional question increases respondent burden, testing, processing and sometimes field time. Some topics are better studied through sample surveys. Others can be derived from administrative data. Good governance forces each proposed item to justify its place in the most expensive statistical operation many countries undertake.

The civilisational principle is portfolio thinking. Do not ask the census to become every dataset. Build a family of instruments and allocate each question to the tool that can answer it best.

51. Procurement can determine statistical quality long before data collection begins

Large censuses depend on procurement: devices, connectivity, software, printing, transport, secure hosting, scanning, temporary facilities and specialised services. Procurement failure can therefore become statistical failure. Devices that arrive late compress training. A poorly specified software contract can lock an agency into inflexible workflows. Weak vendor exit clauses can leave critical systems unsupported.

Strong census procurement begins with architecture, not shopping. Agencies need to know which components are critical, which can be standard products, which require custom development, how systems will interoperate, what data will leave government-controlled environments, and what happens if a supplier fails. Testing should include scale, offline conditions and security rather than polished demonstrations alone.

Vendor dependence is particularly important in a once-a-decade operation because institutional memory can weaken between rounds. Documentation, source-code arrangements where appropriate, data portability, training and handover clauses help prevent knowledge from disappearing when contracts end.

This may feel far removed from the philosophical question “What is civilisation?” It is not. Civilisation becomes real through procurement, maintenance and handover. Grand public functions fail when mundane ownership is vague.

52. A temporary census workforce is a management challenge at national scale

Many census programmes recruit tens of thousands of temporary enumerators, supervisors and support staff. The statistical office must transform a short-term workforce into a reasonably consistent national measurement instrument. That requires role design, recruitment standards, training, identity verification, payment systems, performance monitoring and support.

Training cannot rely on people memorising a manual. Staff need scenario practice: a locked building, a respondent who speaks another language, a household with multiple residences, an institutional dwelling, a safety concern, a device that fails, a person unsure whom to include. Training should teach escalation routes so frontline staff know when not to improvise.

Payment systems matter more than outsiders might expect. Delayed or incorrect payment damages morale and can disrupt fieldwork. Workforce data must therefore be accurate too: who was hired, where they were assigned, which training they completed, which devices they received and which workloads they finished. The census creates a miniature administrative state inside the statistical agency.

Good management recognises that temporary does not mean disposable. Enumerators are the public face of the operation and often the first line of data quality. Their working conditions affect the quality of the national picture.

53. Public communication is an operational control, not advertising decoration

A census asks an entire population to do something unusual at roughly the same time. Communication is therefore part of operations. People need to know when the census occurs, why participation matters, which channels are legitimate, how confidentiality works, where to seek help and how to identify official staff.

One national slogan is rarely enough. Different groups need different explanations and channels. Community organisations may reach populations that national advertising misses. Employers, schools, housing providers and local governments can reinforce awareness. Accessible formats are needed for people with disabilities. Minority-language media can reduce misunderstanding. Rumours need rapid response before they harden into non-response.

Communication should also prepare the public for what the census does not do. If people expect immediate benefits, individual casework or direct changes to local services, disappointment can undermine trust. Statistical agencies should explain that the census provides aggregate evidence used by many downstream institutions rather than personally allocating every resource.

This is civilisation communicating with itself about its own measurement. Clarity improves both participation and legitimacy.

54. Scam resistance becomes part of census design

Any large public data collection creates an opportunity for impersonation. Criminals can exploit public awareness by sending fake links, requesting payments or pretending to be enumerators. A digital census must therefore design not only a response channel but an authentication story ordinary people can understand.

Official domains, publicised phone numbers, verifiable staff identification and consistent guidance help. Agencies should make clear what they will never ask for, such as payment or unrelated banking credentials. Communications should explain how to report suspicious contacts. Help centres need access to up-to-date operational information so legitimate interactions can be confirmed quickly.

Cybersecurity teams also need monitoring for lookalike domains, phishing campaigns and compromised accounts. Incident communications should be prepared in advance because delay creates a vacuum that rumours fill. Security is partly technical and partly social: the public must have a simple mental model of what legitimate census contact looks like.

Civilisation increasingly depends on trust signals. As more public services become digital, authentication itself becomes infrastructure.

55. Rural and remote enumeration reveals the geography of administrative cost

Counting a dense apartment block and counting a remote mountain settlement are not equivalent tasks. Travel time, weather, roads, river crossings, seasonal accessibility, power and connectivity can make the cost per household vastly different. A uniform operational model can therefore create systematic gaps.

Remote enumeration often requires earlier mapping, local guides, flexible transport, offline digital systems, longer field periods and contingency supplies. In some regions, community calendars matter: agricultural cycles, festivals, migration or severe weather can make a nominally convenient census date operationally poor.

Remote communities may also have languages, household structures or land-use patterns that challenge standard categories. Working with local institutions can improve interpretation without surrendering statistical consistency. The goal is equivalence of meaning, not identical logistics.

A civilisation that serves only places where administration is cheap will gradually confuse convenience with universality. Census operations make that bias measurable because every blank area on the map asks why the system could not reach it.

56. Disability statistics show why functional questions matter

Disability is difficult to measure because medical diagnoses alone do not capture how people function in everyday environments. Statistical approaches increasingly use questions about seeing, hearing, mobility, cognition, self-care or communication, often with graded response categories rather than a single yes-or-no label.

The wording matters greatly. Asking “Are you disabled?” can produce different results from asking whether a person has difficulty performing specified activities. Cultural attitudes, stigma, access to diagnosis and age all affect self-identification. Functional questions can improve comparability, though they still simplify complex lived experience.

Why include such measures? Because accessibility planning needs denominators. Transport, schools, digital services, workplaces and emergency systems need to understand how many people may face barriers. A census can provide broad geographic patterns that specialised surveys then investigate in greater depth.

This illustrates a general statistical design principle: when a label is socially or administratively unstable, measure the underlying function needed for the decision. Civilisation improves when categories are connected to practical jobs rather than treated as identities for their own sake.

57. Sensitive identity variables require unusually careful governance

Some censuses collect information on ethnicity, language, religion, Indigenous identity or related characteristics. Others do not. These choices reflect national history, law, policy needs and risk. Such variables can illuminate inequality, cultural change and service requirements, but they can also become sensitive in societies with histories of discrimination or conflict.

There is no universally correct questionnaire. The civilisational requirement is that purpose, legal basis, consultation, classification and confidentiality receive serious attention. Categories should not be created casually by administrators who assume social identities are simple. Self-identification, multiple identities and changing terminology may need accommodation.

Publication requires caution because small geographic cells can make individuals identifiable. Analysts also need to distinguish description from causation: a statistical disparity associated with a group does not by itself explain why the disparity exists.

Good official statistics make sensitive differences visible when there is a legitimate public need while resisting the transformation of descriptive categories into deterministic stories about people. The measuring system should reveal conditions, not manufacture stereotypes.

58. Sex, gender and household relationships expose changing classification needs

Population statistics must remain usable across time while societies and legal frameworks evolve. Questions about sex, gender, marital status and household relationships illustrate the difficulty. Historical series may rely on older categories; contemporary policy needs may require additional distinctions. Statistical offices must decide which concepts they are measuring and explain them precisely.

The key is to avoid asking one field to perform multiple conceptual jobs. Biological sex, gender identity, legal sex markers and household roles may be relevant to different analyses. Conflating them can create ambiguity. Separating them can increase questionnaire burden or sensitivity. Testing and consultation therefore matter.

Long-term comparability should be managed explicitly. When categories change, agencies can publish bridge tables, metadata and caveats rather than pretending an unbroken series exists. Historical data should not be retroactively reinterpreted beyond what the original question measured.

This is classification maintenance under social change. Civilisation needs both continuity and the ability to describe realities that previous vocabularies overlooked.

59. Institutional populations require a second map of society

Most people live in private households, but not everyone does. Students in dormitories, residents of care homes, prisoners, patients in long-stay institutions, military personnel, religious communities and workers in collective accommodation may live under arrangements that standard household questionnaires do not fit.

Census systems therefore distinguish private households from collective or institutional living quarters. Operational procedures may rely on facility administrators for rosters while preserving individual confidentiality. Definitions of usual residence must determine whether a person is counted at the institution or another address.

These populations matter disproportionately for certain services. Care-home residents influence ageing and health planning. Student populations reshape transport and housing demand in university towns. Prison populations matter for justice statistics. Military populations can affect small localities. If they are mishandled, local counts can be distorted even when national totals barely change.

The civilisational lesson is that edge cases are often entire institutions. A system built only for the modal household can fail thousands of people at once.

60. Urbanisation is not one number

News reports often say that a country is “60 percent urban” as though urban status were a natural property. In reality, countries define urban areas in different ways: administrative status, population thresholds, density, built-up form, economic activity or combinations. International comparison therefore requires care.

Urbanisation also has multiple spatial scales. A metropolitan labour market can extend far beyond the legal city boundary. Peri-urban settlements may look rural in administrative classification while functioning as part of an urban economy. Small towns can provide urban services to surrounding rural populations. Satellite imagery can detect built-up land but cannot by itself define social or administrative urbanity.

Census geography helps because it provides fine-grained population distribution that can be combined with density and commuting. More harmonised spatial concepts can then support international analysis while national definitions remain available for legal and planning purposes.

Again, the lesson is not to seek one perfect category. It is to know which definition is being used and whether it fits the question.

61. Commuting creates a daytime civilisation and a nighttime civilisation

Residential population describes where people usually live. Many infrastructure systems experience people where they are during the day. Central business districts can multiply their population each morning. University campuses, industrial zones and tourist areas can experience large temporary concentrations. Residential counts alone cannot describe those loads.

Journey-to-work and journey-to-study data help map functional relationships among places. Transport planners use flows rather than only origins. Emergency managers need to know how many people may be present at different times. Retail and service planning also depends on catchment movement.

Mobile-device and transport smart-card data can add timeliness, but these sources contain coverage and privacy limitations. They should be calibrated against population benchmarks rather than assumed to represent everyone equally. Children, older adults and people with multiple devices can distort naive interpretations.

Civilisation is dynamic. A population map frozen at midnight captures only one state of a moving system.

62. Seasonal populations challenge the idea of one correct count

Tourist towns, agricultural regions, pilgrimage sites, mining camps and university cities can experience dramatic seasonal population changes. The census still needs a defined reference population, but service systems may need additional measures of presence and demand.

Water use, waste generation, emergency calls and transport loads can peak far above resident population. A town of 20,000 residents may host 100,000 visitors during a major event. Designing infrastructure only around usual residents can therefore understate capacity needs.

Administrative records, accommodation statistics, mobility data and event calendars can supplement census baselines. The important discipline is to label each measure clearly: resident population, de facto population, visitor nights, daytime population or service load are not interchangeable.

Once again, civilisation needs multiple maps because different systems experience the same place differently.

63. Environment and population data become powerful when layered carefully

Environmental risk becomes socially meaningful when hazards are connected to populations. Air pollution, heat, flood exposure, drought, wildfire smoke, industrial hazards and access to green space all vary geographically. Census and housing data provide demographic context for understanding who may be exposed.

Overlay analysis must be done carefully. A person living in a census area is not necessarily exposed uniformly to every condition measured there. Environmental data may have different spatial resolution and time periods. Ecological fallacy can arise when analysts infer individual characteristics from area averages.

Nevertheless, population-environment integration is invaluable for screening and planning. It can identify communities that deserve more detailed monitoring, guide cooling-centre placement, estimate populations in flood zones and support environmental-health research.

The civilisation mechanism is integration without overclaiming. Different datasets become more useful when joined, but linkage does not erase the limitations of each source.

64. Satellite imagery can see buildings, not households

Remote sensing has transformed mapping. High-resolution imagery can reveal new settlements, road networks, building footprints and land-cover change. Night-time lights can provide broad signals of electrification or economic activity. These tools are especially valuable where maps age quickly.

But an image of a roof does not tell us how many people live underneath it. A building can be vacant, commercial, institutional or subdivided. Household size varies. Informal shelters may be hard to detect. Dense vertical housing can contain many more residents than a simple footprint suggests.

The strongest use of remote sensing is therefore complementary. Imagery can improve enumeration-area design, detect growth, identify possible missing structures and support spatial modelling. Ground data or reliable administrative records remain necessary to convert structures into human populations.

This is a useful warning for the age of abundant sensors: seeing the physical world is not the same as understanding the social world occupying it.

65. Mobile-phone data is rich, fast and systematically incomplete

Aggregated mobile-network data can reveal mobility patterns at a scale that traditional surveys cannot match. During disasters or transport disruptions, it can show changes in movement. For urban planning, it can illuminate commuting and activity centres. But subscriber data is not population data without adjustment.

Some people have no phone. Others have several SIM cards. Devices may be shared. Network market share varies geographically. Children are underrepresented. Roaming and machine-to-machine devices complicate counts. Location is inferred through network interactions rather than continuous human presence.

Privacy risks are substantial because mobility traces can be identifying even when obvious personal fields are removed. Responsible use therefore requires aggregation, legal safeguards, technical controls and careful assessment of whether the public benefit justifies the data handling.

Census benchmarks can help calibrate such alternative sources. The old and new instruments are not enemies. A civilisation becomes more observant when slow, representative baselines and fast, partial signals are used together.

66. Synthetic data may widen access, but it is not the census

Synthetic datasets are generated to resemble important statistical properties of real data without directly reproducing individual records. They can support software testing, teaching, exploratory analysis and some forms of research while reducing disclosure risk.

The danger is overconfidence. Synthetic data reflects the model that generated it. Rare relationships can be distorted. Small subpopulations may be poorly represented. Analyses that depend on subtle correlations can produce different results from secure access to real microdata. Users therefore need clear labels and validation information.

A layered access model works well: open aggregate tables for everyone, synthetic or public-use samples for exploration, and secure research environments for approved projects requiring greater detail. Each layer trades fidelity against disclosure risk differently.

Civilisation does not solve privacy by making all data open or all data closed. It builds differentiated channels matched to legitimate uses.

67. Open data becomes valuable only when metadata travels with it

Publishing a CSV file is not the same as creating open statistical infrastructure. Users need variable definitions, geographic codes, reference dates, suppression rules, quality notes and revision histories. Without those, machine-readable data can spread misunderstanding faster than a printed table ever could.

Stable identifiers and versioned APIs help developers build tools without scraping changing web pages. Geographic boundary files should match statistical codes. Data licences should state permissible reuse. Release calendars help users know when updates are expected.

Accessibility also matters. Not every user is a programmer. Good dissemination offers tables, maps, narrative summaries and APIs rather than forcing everyone into one interface. Education materials can show how to interpret rates, margins and classifications.

The best open-data system therefore behaves like a public library: organised, catalogued, searchable and documented, not merely unlocked.

68. Secure research access lets society learn without publishing everyone

Many important questions require individual-level records because aggregate tables cannot capture complex relationships. Researchers may need to study how education, migration, housing and labour outcomes interact. Publishing identifiable microdata would be unacceptable, so statistical systems create controlled access mechanisms.

Secure research environments can require project approval, accredited researchers, safe settings, restricted exports and disclosure review of outputs. Some systems provide remote access where data never leaves a controlled environment. The principle is to move analysis to the data rather than moving sensitive data everywhere.

Governance should be proportionate. Excessive barriers can prevent socially useful research; weak barriers can damage confidentiality. Transparent criteria, audit trails and sanctions for misuse help maintain the balance.

This is another civilisational compromise engineered into procedure. Trust does not require refusing all access. It requires making access accountable.

69. Federal and decentralised systems must reconcile many statistical owners

In federal or strongly decentralised countries, population data may be produced and used across national, regional and local governments with distinct legal powers. Local registers can be richer than national systems, but definitions may differ. Data-sharing authority may be fragmented. Boundaries and service responsibilities can change independently.

Coordination requires common standards, data-exchange agreements, reference classifications and governance forums. Local expertise should not be erased; it often improves quality because municipalities understand addresses and settlement changes first. The national system’s job is to create interoperability and comparability across local realities.

Funding also matters. If local authorities are expected to maintain national-quality registers without resources, data quality will diverge. Shared infrastructure can reduce duplicated costs, but ownership of correction and maintenance must remain clear.

Civilisation at scale is frequently a federation of partial views. The art lies in making them add up without demanding that every institution become identical.

70. Conflict can destroy the paperwork that proves continuity

War and displacement attack statistical systems in two ways. Populations move rapidly, making old residence data obsolete, while registries, archives and local offices may be destroyed or become inaccessible. People can lose identity documents exactly when they most need proof of birth, family relationships, qualifications or property claims.

Post-conflict population estimation therefore relies on multiple imperfect sources: displacement registrations, humanitarian records, surveys, satellite imagery, local lists and renewed civil registration. None is automatically complete. Duplicates and omissions are common because people move between systems.

Rebuilding civil registration and identity continuity becomes a foundational recovery task. Procedures need to handle lost documents and contested records without making fraud trivial. Historical archives, backup systems and distributed copies can reduce catastrophic information loss.

A civilisation is partly the continuity of its records across disruption. When documentation disappears, people can become administratively displaced even after physical safety returns.

71. A pandemic shows why population systems need both baselines and speed

Public-health emergencies expose the temporal mismatch between censuses and crises. A census gives a rich baseline but may be years old. An outbreak requires daily or weekly information. Health systems therefore combine census denominators with current surveillance, laboratory reports, hospital data, vaccination records and mobility indicators.

Population structure shapes risk. Age distributions affect expected severity. Household size affects transmission opportunities. Occupation and commuting influence exposure. Housing crowding and institutional living can create concentrated vulnerability. Without a reliable baseline, fast data can be difficult to interpret.

Emergency data collection also creates privacy pressure because speed encourages broad linkage. Governance should specify which extraordinary uses are time-limited, what data will be retained and how emergency access ends when the emergency does.

The civilisational architecture is two-speed: slow systems create trusted reference frames; fast systems detect change. Resilience comes from connecting them before the crisis rather than improvising the relationship during one.

72. Population projections are promises about assumptions, not prophecies

Planners often ask for a single future population number because budgets and infrastructure plans need something concrete. Demographers know the future is conditional. Fertility, mortality and migration can change. Economic shocks alter movement. Policies influence family formation. Unexpected events can break trends.

A responsible projection therefore states its assumptions and often provides variants. Short-term projections may be relatively stable for cohorts already born, while long-term projections become increasingly sensitive to fertility and migration. Local projections can be more volatile than national ones because a housing project or employer closure can shift population sharply.

Projection error should be studied after the fact. If a region consistently grows faster than projected, the method or assumptions need review. Planning systems should also use scenarios rather than one point estimate when infrastructure has high irreversibility or long lead times.

Civilisation cannot eliminate uncertainty about the future. It can make uncertainty explicit enough to design buffers.

73. International comparison requires harmonisation without erasing difference

Global statistics invite comparison: population growth, urbanisation, household size, education, employment and migration. But national systems differ in definitions, administrative sources and census timing. International organisations therefore develop concepts and classifications that improve comparability while documenting national deviations.

Harmonisation is not the same as forcing every country into one operational model. A register-based census and a traditional census can both produce comparable population concepts if methods are designed carefully. Countries may retain national categories while mapping them to broader international classifications for comparison.

Users should be cautious with league tables built from indicators that are technically comparable but contextually different. Similar numbers can arise from different structures. Statistical comparison is strongest when it opens a question rather than pretending to close one.

The civilisational value of international standards is shared language. They let different societies learn from one another without claiming that one model fits all.

74. The 2030 census round points toward integrated population data systems

The United Nations’ 2030 World Population and Housing Census Programme covers the period from 2025 to 2034 and encourages countries to conduct a census while strengthening administrative registers, alternative data sources, geospatial integration and quality standards. The direction is significant: census-taking is increasingly understood as part of a wider population-data ecosystem.

This shift responds to practical pressures. Traditional full enumeration is expensive. Populations are more mobile. Users expect faster data. Administrative systems have improved in many countries. Digital technology makes integration possible. At the same time, privacy, cybersecurity and public trust concerns become more important as data linkage expands.

The future is therefore not simply “more data”. It is better-governed relationships among sources. Census programmes will need stronger metadata, provenance and quality frameworks precisely because users may no longer be able to point to one questionnaire as the origin of every statistic.

Civilisation is moving from periodic counting toward continuous statistical sensing. The engineering challenge is to gain timeliness without losing representativeness or trust.

75. A civilisation should know the cost of not counting well

The price of a census is visible because it appears in a public budget. The price of weak statistics is dispersed. A school is built too late. A hospital catchment is underestimated. A water network is undersized. A vulnerable group remains invisible. Business investment uses poor market estimates. Emergency plans rely on stale population maps. Each error appears in another department’s ledger.

This creates a structural incentive problem. Statistical infrastructure is a public good whose benefits are spread across government, business, research and civil society. The agency paying for data production may capture only a small fraction of the value. Underinvestment therefore becomes easy even when the social return is high.

Good governance treats national statistical capacity as enabling infrastructure, similar to mapping, standards or digital identity. It is not glamorous, but many high-level systems quietly assume it exists.

The absence of reliable statistics does not produce an empty decision. It produces a decision made with guesswork, lobbying, outdated proxies or whoever has the loudest anecdote. Civilisation always uses information; the choice is whether that information is disciplined.

76. The final operating model: observe, reconcile, explain, correct

A robust population-data system can be summarised in four verbs. Observe: collect information through census, registration, surveys, administrative systems and geospatial tools. Reconcile: compare sources, investigate mismatches and quantify error. Explain: publish definitions, methods, quality notes and accessible outputs. Correct: repair records, revise estimates and improve the next cycle.

Each verb guards against a different failure. Observation without reconciliation produces parallel databases that disagree silently. Reconciliation without explanation produces technically polished numbers the public cannot trust. Explanation without correction turns transparency into ritual. Correction without preserved history makes change impossible to audit.

The cycle never ends because society never stops moving. New dwellings are built. People migrate. Families change. Businesses open and close. Boundaries shift. Technologies alter behaviour. Climate changes settlement risk. A population system must therefore be maintained as a living public capability rather than rebuilt only when a crisis exposes its absence.

That is the deeper meaning of a civilisation counting itself. It is not an obsession with numbers. It is the discipline of keeping a shared map of human reality accurate enough that strangers can coordinate across distance, time and institutions.

77. A practical checklist for readers, planners and students

When you encounter a population statistic, ask: What population does this number describe? What is the reference date? What geographic boundary is used? Is the source a census, survey, register, estimate or model? What important groups might be undercovered? Have definitions changed? Is the figure adjusted? How uncertain is it? Can the source explain its method in ordinary language?

When you encounter a proposal to link datasets, ask: What is the legitimate purpose? Which source is authoritative for each field? How are mismatches handled? Who can access identifiable data? Can individuals correct errors? How long are data retained? What audit trail exists? What happens when the linkage model makes a mistake?

When you encounter a planning claim, ask whether the denominator matches the decision. A service count without a population denominator may measure workload rather than need. A national average may hide local shortages. A residential population may not reflect daytime demand. A historical trend may cross a methodological break.

These questions turn statistics from passive facts into inspectable machinery. That habit is useful far beyond census data. It is part of learning how civilisation works.

78. Counting well is a form of respect

At its best, population statistics begins from a modest moral premise: people should not disappear merely because they are inconvenient to count. A child without documents, a person without a home, a remote village, a migrant household, an older person living alone and a resident of an institution all remain part of the civilisation the statistic claims to describe.

That does not mean every characteristic should be collected. Respect also means privacy, restraint and purpose limitation. The system should know enough to fulfil legitimate public jobs without treating the population as raw material for unlimited surveillance.

The balance is difficult because both blindness and overreach can harm. Civilisation needs legibility to coordinate, and human beings need protected spaces beyond institutional legibility. Strong statistical design acknowledges both truths instead of pretending one cancels the other.

When a census works, millions of individual responses are transformed into public knowledge while the individuals themselves recede from view. That transformation—from person to pattern without erasing dignity—is one of the quiet achievements of modern statistical civilisation.

79. What this mechanism contributes to the larger idea of civilisation

Civilisation is often described through monuments, inventions or political eras because those are easy to see. Population statistics reveals a different kind of achievement: the capacity to coordinate abstractly with people we will never meet. A planner can design a school for children not yet born because birth statistics and projections exist. A health agency can estimate mortality because deaths are registered. A researcher can study inequality because households were classified consistently enough to compare.

None of this work is spectacular. It is cumulative. Standards improve. Maps become cleaner. Registries become more complete. Questionnaires become more intelligible. Privacy controls become more sophisticated. Data releases become faster. A civilisation gains the ability to inspect itself with gradually better instruments.

That ability changes what is possible. Problems that once appeared as vague impressions become measurable distributions. Claims can be tested against denominators. Policy can be evaluated against baselines. Future needs can be estimated rather than guessed. Disagreements do not disappear, but some disagreements can at least begin from a shared map.

Counting is therefore not beneath civilisation. It is one of the mechanisms that allows civilisation to become conscious of its own scale.

80. Conclusion: civilisation must keep its map close to the territory

A population census is not civilisation, and a population register is not society. They are models. Their value comes from remaining close enough to reality to support competent action while staying honest about what they cannot see. That requires constant maintenance because the territory moves.

The strongest population-data systems do not worship a single database. They combine census benchmarks, civil registration, administrative registers, surveys, geospatial information and carefully governed new data sources. They reconcile contradictions. They publish metadata. They measure coverage. They protect confidentiality. They let people correct errors. They preserve historical comparability. They know that speed, precision and completeness are different virtues.

The deepest civilisational function is not the final number announced on television. It is the institutional chain that makes the number worthy of attention: definitions, law, mapping, collection, validation, professional independence, disclosure control, publication and revision. That chain allows strangers to share a statistical description of the world even when they disagree about what should happen next.

A civilisation counts itself because it has responsibilities too large for memory and relationships too numerous for personal knowledge. It needs a disciplined way to know who is here, where life is changing and where capacity must move. Keep the map close to the territory, and the rest of civilisation can navigate with fewer illusions.

81. Reference dates prevent a moving population from becoming an impossible count

A population never holds still for the convenience of a census. People are born and die while fieldwork continues. Travellers cross borders. Families move house. Students change residence. A census therefore needs a precise reference moment or reference period: the conceptual instant to which answers should relate, even if the questionnaire is completed days or weeks later.

This rule sounds technical until two households interpret it differently. One may report everyone physically present when the form is completed; another may report usual residents on census night. Without a shared reference, the national total becomes a mixture of incompatible clocks. Training, questionnaire wording and processing rules all exist partly to keep millions of responses attached to the same temporal frame.

Reference dates also matter when census data are compared with school enrolment, tax records, births, deaths or labour statistics. Two perfectly accurate sources can disagree because they describe different dates. Reconciliation begins by aligning time before searching for deeper explanations.

Civilisation repeatedly solves moving-world problems by agreeing on reference frames. Time zones, accounting periods, school years and census dates all perform this quiet function: they let dispersed institutions talk about the same state of a changing system.

82. Usual residence and actual presence answer different questions

Population statistics commonly distinguish between a person’s usual residence and the place where that person happens to be at a particular moment. The difference matters in societies with commuting, tourism, seasonal work, boarding schools, hospitals, prisons, military service and frequent travel. A person can be physically present somewhere without belonging to that place’s resident population.

A usual-residence concept is often better for long-term service planning because schools, housing and local health systems generally need to know where people ordinarily live. A present-population concept can be useful for emergency management, event planning or short-term infrastructure loads. Neither measure is universally superior; each is built for a different job.

The difficulty comes when users treat one measure as if it answers the other question. A city centre can have a modest resident population and an enormous daytime population. A resort can have small permanent population and extreme seasonal demand. A university district can change radically between term and holiday periods.

Good statistical systems therefore resist the seduction of one “true” population number. Civilisation needs multiple valid views of the same human geography because different services experience different versions of presence.

83. Fertility statistics connect today’s births to tomorrow’s infrastructure

Birth counts are among the earliest signals of future demand. A cohort born today reaches childcare, primary school, secondary school, higher education and the labour market on a roughly predictable sequence. That does not make the future deterministic, because migration and participation alter each stage, but it gives planners a moving wave they can observe years in advance.

Fertility measurement is more subtle than simply counting births. Demographers relate births to the population of women at relevant ages and distinguish period measures from completed family size. Age-specific patterns matter because the same total number of births can arise from different timing of parenthood. Delayed births can create temporary troughs and rebounds that are easy to misread.

Reliable birth registration, census age structures and surveys can be used together to understand fertility change. Where registration is incomplete, estimates require additional modelling and uncertainty. The quality of tomorrow’s school projections may therefore depend on today’s ability to record events that happen far from an education ministry.

This is how civilisation becomes longitudinal. One institution’s routine record becomes another institution’s future planning input. Good systems are connected not because everything belongs in one database, but because consequences travel across time.

84. Mortality statistics tell civilisation where life is being lost

Death registration closes an individual legal record while opening a statistical signal. When deaths are recorded consistently with age, sex, place and medically meaningful cause information, society can see patterns that anecdotes cannot reveal: rising chronic disease, road injury, infectious outbreaks, occupational hazards, maternal mortality or unusual excess deaths.

Cause-of-death statistics require more than a death certificate existing somewhere. Certification practices need standards, medical staff need training, coding systems need consistency, and ambiguous causes need quality review. Where many deaths occur outside medical facilities, verbal autopsy or other methods may be used to improve population-level understanding while acknowledging uncertainty.

Mortality rates also need denominators. Ten deaths in a small community and ten in a large city imply different levels of risk. Age standardisation may be necessary when populations have different age structures. A raw count can therefore be true and still answer the wrong question.

A civilisation that records deaths well is not being morbid. It is learning where prevention, care and safety systems are failing. The statistical record becomes a feedback mechanism through which lost lives can alter future design.

85. Household formation can move faster than population growth

Housing demand is often discussed as if it follows population totals directly. Household formation breaks that shortcut. If average household size falls, the same population requires more dwellings. Divorce, delayed marriage, longer life expectancy, student independence, migration and changing intergenerational living patterns can all alter the number of households without proportional change in total residents.

This is why census questions about household relationships and dwelling occupancy matter to planners. A city with slow population growth can still face acute housing pressure if one-person households increase quickly or if many existing dwellings are unavailable to ordinary residents. Conversely, rapid population growth can sometimes be absorbed temporarily through larger household sizes, though crowding may rise.

Household projections therefore deserve their own assumptions. Age structure, partnership patterns, migration and housing costs influence who forms a separate household. Forecasting only population can leave water, electricity and housing systems underprepared because service connections often attach to dwellings rather than individuals.

Civilisation becomes more accurate when it models the unit that actually drives the infrastructure. Sometimes that unit is a person. Sometimes it is a household, a building, a trip, a firm or a hectare. The right denominator depends on the mechanism.

86. Sampling can exist inside a census without making the census a survey

Some census designs use a short form for most households and a longer set of questions for a sample, or combine complete enumeration of core variables with sample-based detail. This can reduce burden while preserving broad geographic coverage. The concept can seem contradictory only if a census is imagined as one identical questionnaire asked of everyone.

The design must be explicit because sampled variables have different precision from fully enumerated counts. Estimates require weights and measures of sampling uncertainty. Small geographic areas may not support reliable estimates for rare characteristics. Dissemination systems need to signal these differences so users do not treat every table cell as equally exact.

Sampling also shows why the census is best understood as a statistical programme rather than a ritual. The job is to produce high-quality population information, not to preserve one historical collection method regardless of cost or analytical need.

A capable civilisation chooses instruments deliberately. Complete enumeration is powerful where universality matters; sampling is powerful where depth matters. Combining them intelligently can produce more information with less burden than insisting on a single method everywhere.

87. Longitudinal linkage lets civilisation study pathways, not only snapshots

A census is traditionally a cross-sectional picture: what the population looks like at one reference time. When legally and ethically governed records can be linked across time, researchers can study pathways—how education relates to later work, how neighbourhood change relates to mobility, or how demographic transitions unfold within cohorts.

Longitudinal data can reveal patterns that repeated snapshots cannot. Two censuses may show the same proportion of people in a category while hiding intense movement into and out of it. Following cohorts can distinguish persistence from turnover. This matters for designing interventions because a chronic condition and a rapidly rotating condition require different responses.

Linkage also raises privacy stakes. Persistent identifiers make records analytically powerful precisely because they connect events through time. Governance therefore needs strict purpose controls, secure environments and methods that keep published findings aggregate. The analytical gain cannot be separated from the responsibility created by the connection.

Civilisation learns more when it can see trajectories, but trajectory data should not become a permanent dossier available for any purpose. The distinction between research value and administrative power must remain visible.

88. Data lineage answers the question: where did this number come from?

As population statistics integrate more sources, provenance becomes essential. A published estimate may combine census responses, civil registration, migration records, imputation, geographic adjustments and modelled components. Without data lineage, even internal experts can struggle to reconstruct how a figure was produced after staff or software changes.

Data lineage documents transformations from source to output: which files were used, what versions, which rules were applied, which records were excluded, what code ran, and what revisions occurred. This is statistical reproducibility translated into operational practice. It turns a number from an opaque product into an auditable chain.

Lineage is especially important when errors are discovered. Investigators need to know which published tables inherited the problem, which downstream datasets were affected and whether a correction requires rerunning one stage or the entire pipeline. Without lineage, correction becomes archaeological work.

The wider civilisation lesson is that complex systems need memory of transformation, not only storage of final states. Knowing what exists now is useful; knowing how it became that way is what makes repair possible.

89. Reproducible statistics protects knowledge from staff turnover

Statistical offices accumulate expertise in people, but people retire, transfer and leave. If a critical estimate depends on one analyst remembering a sequence of manual spreadsheet steps, the method is not truly institutional. Reproducible workflows convert tacit practice into documented procedures, code, tests and versioned inputs that another competent team can rerun.

This does not mean every decision can be automated. Expert judgment remains necessary when classifications change, anomalies appear or methods are redesigned. The goal is to make judgment points visible and recorded rather than burying them inside an undocumented chain.

Reproducibility also makes quality review easier. Independent analysts can test whether outputs follow the stated method. Automated regression tests can detect unexpected changes when software is updated. Historical releases can be regenerated if source corrections require revision.

Civilisation survives generations by converting personal mastery into transmissible institutional capability. Statistical reproducibility is one small, concrete example of that larger mechanism.

90. Timeliness and accuracy are a production trade-off, not a moral contest

Users want population data quickly because stale evidence loses value. Statistical agencies want accuracy because premature release can spread errors. The two goals can conflict. Waiting until every uncertainty is resolved may make data irrelevant; publishing instantly may force large revisions and damage trust.

Staged releases are one solution. Preliminary totals can be published with clear labels, followed by validated detailed tables and later revisions. Some indicators can be produced quickly because they rely on mature administrative systems; others need longer processing. A release calendar lets users plan around the sequence.

The important governance question is whether the trade-off is explicit. Users should know what “provisional” means, what quality checks remain, and when final estimates are expected. Speed should not be achieved by quietly lowering standards, and precision should not become an excuse for indefinite delay.

Civilisation often works through managed trade-offs rather than perfect optimisation. Reliability comes from making the compromise visible, repeatable and reviewable.

91. Statistical misinformation can begin with a correct number

A false claim does not always require fabricated data. A correct population number can mislead when detached from its denominator, geography, reference period or definition. A percentage increase can look dramatic from a tiny baseline. A national average can conceal regional divergence. A revised classification can be mistaken for sudden social change.

Statistical agencies therefore have a communication responsibility beyond publishing tables. Plain-language explanations, charts with appropriate scales, methodological notes and rapid correction of misinterpretations can help. This does not mean policing every opinion. It means ensuring that the underlying statistical product is difficult to misuse accidentally.

Journalists, teachers and researchers become part of the dissemination system. Statistical literacy education can teach readers to ask about denominators, uncertainty and comparability. The most resilient defence against misleading claims is not a central authority deciding what everyone may say, but a population better equipped to inspect evidence.

A civilisation that produces statistics but not statistical literacy has built an instrument panel without teaching people how to read the gauges.

92. Local statistical capacity determines whether national data reaches the street

National statistical offices may produce excellent census datasets, but many decisions occur locally. Municipalities decide where to place services, maintain roads, manage waste, plan housing and prepare for emergencies. If local agencies lack analysts, geographic tools or usable data interfaces, national statistical sophistication may never reach daily administration.

Capacity building therefore includes more than central methodology. Local governments need stable geographic codes, training, access to current data and the ability to feed corrections back upstream. A municipality often notices a new settlement or demolished building before a national agency does. Two-way information improves both levels.

Small jurisdictions face a scale problem because they may not justify specialist teams. Shared regional services, standard dashboards and national technical support can spread capability without forcing every town to build a full statistical office.

Civilisation works when high-level knowledge can descend into operational decisions and local observations can climb back into the national picture. A one-way data pipeline eventually becomes stale.

93. Archives keep statistical civilisation from losing its own past

Census records are valuable long after the immediate planning cycle. Historical tables reveal urbanisation, migration, fertility, language change and economic transformation across generations. Researchers can reconstruct how neighbourhoods changed. Families may eventually use released historical records for genealogy where law permits. Institutional historians can understand earlier administrative choices.

Preservation requires deliberate formats and metadata. Software becomes obsolete. Storage media degrade. Proprietary file formats can become unreadable. Geographic boundaries change. Codebooks get separated from datasets. Digital preservation therefore needs migration plans, checksums, redundant storage and documentation that future users can understand.

Confidentiality periods also matter. Detailed historical records may remain closed for decades before legal release, while aggregate outputs are public immediately. Archives must preserve protected material securely enough that future access remains possible without premature disclosure.

A civilisation does not merely generate information; it decides which information survives. Statistical archives are part of the bridge through which future generations can inspect the societies that produced them.

94. Climate mobility will make static population assumptions more fragile

Sea-level rise, extreme heat, drought, wildfire and repeated flooding can alter where people live, sometimes gradually and sometimes through abrupt displacement. Population systems will need to distinguish temporary evacuation, seasonal adaptation, permanent migration and planned relocation because each has different implications for services and infrastructure.

Census baselines can identify exposed populations, but climate mobility unfolds between census rounds. Property records, school enrolment, utility connections, address changes and surveys can provide timelier signals. Models should remain cautious because exposure does not mechanically determine movement; income, insurance, family ties and policy shape whether people stay or leave.

Receiving communities also need attention. A place that gains residents after repeated disasters elsewhere may face housing and infrastructure pressure without being directly damaged itself. Population statistics must therefore see both origin and destination.

Civilisation’s map will need to update faster as environmental conditions change. The core job remains the same: keep institutions aligned with where people actually are rather than where old assumptions say they should be.

95. Humanitarian registration and national statistics should cooperate without becoming the same system

During displacement or disaster, humanitarian organisations may register people quickly to deliver food, shelter, cash or medical assistance. These operational lists are invaluable, but they are created for service delivery under crisis conditions, not as complete national population registers. Coverage follows need and access rather than a statistical sampling frame.

National statistical agencies can still learn from humanitarian data when methods, consent and legal frameworks permit. Duplicate registrations, mobility and changing household composition need careful treatment. Aggregate information may help estimate displaced populations, while independent surveys can assess people outside assistance systems.

Purpose separation remains important because vulnerable people may provide information in exchange for urgent help under conditions that do not resemble ordinary administrative choice. Reuse should not be automatic merely because the data exists.

A mature civilisation can coordinate across institutional boundaries without erasing them. Cooperation is strongest when each system’s purpose, authority and limitations remain explicit.

96. Statistical capacity is a development capability, not a luxury purchased at the end

It is easy to imagine that poorer countries should spend scarce resources only on visible services and improve statistics later. The problem is circular: without reliable population, health, education, agricultural and economic data, it becomes harder to allocate scarce resources intelligently or measure whether programmes work.

Statistical capacity therefore grows alongside service capacity. Civil registration improves when health systems record births and deaths. School data improves when enrolment administration improves. Business statistics improve when enterprise registration and tax systems mature. Census investment can strengthen maps and frames used by many programmes.

International support can help finance technology, training and methodology, but sustainable capability requires domestic ownership. A system dependent on a temporary project team can collapse when funding ends. Long-term staffing, legal mandates and maintenance budgets matter as much as initial equipment.

Civilisation is not built by choosing between services and information about services. It is built when feedback lets each improve the other.

97. Artificial intelligence can assist statistical production, but responsibility cannot be outsourced

Machine-learning systems can help code occupation descriptions, detect anomalies, match records, translate text and support respondent help services. These tasks can reduce manual workload when models are validated against representative data and monitored for drift. The value is practical: faster processing and more consistent handling of repetitive cases.

The risks are equally practical. A classification model may perform poorly on minority languages or unusual occupations. A record-linkage model can create false matches. A conversational interface can give an incorrect answer about whom to include in a household. Because errors can become systematic at scale, human review and measurable error thresholds remain necessary.

Statistical agencies should document where models are used, what they were tested on, which decisions remain human and how failures are detected. Automation should make provenance clearer, not more mysterious. If nobody can explain why a model altered an official statistic, governance has become weaker even if processing became faster.

Civilisation gains from tools when tools enlarge accountable human capability. It loses when responsibility disappears into the tool.

98. Differential privacy and related methods formalise a difficult bargain

Modern disclosure-control techniques can add calibrated statistical noise or constrain outputs so that the presence of any one individual has limited influence on what is released. Differential privacy is one prominent framework. Its attraction is that privacy protection can be expressed mathematically rather than relying only on informal judgments about whether tables “look safe”.

The trade-off is utility. Stronger privacy can reduce accuracy, particularly for small geographic areas or rare groups. The system therefore needs an explicit privacy budget or equivalent decision about where precision matters most. Users need documentation because noisy counts can behave differently from conventional suppression.

No technical method resolves the ethical question by itself. Privacy parameters encode a social choice about acceptable disclosure risk and analytical loss. Statistical agencies need public reasoning, testing and transparent explanation rather than presenting mathematics as if it eliminated judgment.

This is a recurring civilisation pattern: technical machinery can discipline a trade-off, but it cannot decide the values behind the trade-off.

99. Quality has several dimensions, and improving one can damage another

Statistical quality is not one score. Relevant dimensions include accuracy, timeliness, coherence, comparability, accessibility, interpretability and credibility. A dataset can be highly accurate but released too late. It can be timely but poorly documented. It can be internally consistent but impossible to compare across years because definitions changed.

This multidimensional view prevents simplistic performance claims. A digital census that shortens processing time may improve timeliness while introducing new cybersecurity or coverage risks. A richer questionnaire may improve relevance for some users while reducing response rates. A privacy measure may protect confidentiality while reducing small-area accuracy.

Quality frameworks help agencies state these tensions explicitly and measure them over time. User needs matter too: a researcher, emergency planner and local councillor may prioritise different dimensions. The best product is not universally maximal; it is fit for the decisions it supports.

Civilisation improves when performance becomes plural enough to reflect the real job. One headline metric is rarely sufficient for a complex public system.

100. The final civilisation test is whether counting improves the world it describes

A census can be technically brilliant and still fail its wider purpose if the knowledge never reaches decisions. The final value of population statistics appears when schools are placed more intelligently, health services see changing need, infrastructure anticipates growth, vulnerable populations become visible, research improves understanding and public debate gains a more reliable factual floor.

The chain is long. Good collection does not guarantee good policy; statistics cannot substitute for judgment, resources or values. But poor information weakens every later stage. A civilisation that sees itself badly spends more energy correcting avoidable mistakes and arguing about basic conditions.

The purpose of counting is therefore not to celebrate the count. It is to reduce the distance between the civilisation people imagine and the civilisation people actually inhabit. That requires evidence capable of surprising institutions when their assumptions are wrong.

When population statistics can do that—reliably, transparently and with respect for the people behind the rows—the counting system has become what it should be: a quiet instrument through which a large society learns to notice itself.