A population projection is a calculation of the future size and composition of a population based on a starting population and explicit assumptions about demographic change. A central tool of demography, it describes how a population would develop if specified patterns of births, deaths, and migration occurred. Projections may cover the world, a country, or a smaller area and may distinguish people by age, sex, place of residence, or other characteristics. Their defining feature is their conditional nature: a projection is not a statement that a particular future must occur.
Projections, forecasts, and estimates
Population projections are related to, but conceptually distinct from, population estimates and forecasts.
- An estimate measures a population at a past or present date, often by combining incomplete observations.
- A projection calculates a future population under stated assumptions, whether or not those assumptions are judged the most likely.
- A forecast aims to describe the future judged most likely, sometimes through a probability distribution rather than a single number.
Terminology varies among statistical agencies, and a central projection may function as a forecast in practice. Nevertheless, a scenario designed to explore the consequences of zero migration, for example, need not imply that zero migration is expected.
A projection has a base date, a projection horizon, and a schedule of time steps, commonly annual or five-year intervals. Its results depend both on the accuracy of the base population and on the assumptions applied after that date.
Demographic accounting
The accounting foundation is the population balancing equation:
where is the population at the beginning of an interval, and are births and deaths during that interval, and and are immigration and emigration. The difference is natural increase; is net migration.
For the world as a whole, international migration does not change the total population. For countries and smaller areas, migration can be a major source of change. Actual statistical accounts may also include adjustments for boundary changes, coverage errors, and inconsistencies between data sources.
The balancing equation is an accounting identity rather than a complete forecasting model. A projection requires methods for generating its future components.
Principal methods
Cohort-component method
The cohort-component method is the standard framework for detailed official projections. It divides the base population into age and sex groups, then advances each group through time while accounting for survival, migration, and births.
Its main steps are:
- Survive and age the existing population. Age-specific survival probabilities determine how many people remain alive as each cohort moves into the next age group.
- Apply migration assumptions. Immigration and emigration are incorporated by age and sex, either separately or as net migration.
- Calculate births. Age-specific fertility rates are applied to the projected female population exposed to childbearing.
- Add the new birth cohort. Births are divided by sex and adjusted for survival between birth and the next projection date.
For a one-year age group and interval, an illustrative relationship is:
where is the probability of surviving from age to age , and represents the contribution of migration after allowing for its timing and survival. Newborns and the oldest open-ended age group require separate treatment.
Births can be expressed as:
where is the fertility rate at age and is female exposure, measured in person-years, at that age during the interval.
This framework makes age structure explicit. Two populations of equal size can follow very different trajectories if one contains many young adults and the other contains a high proportion of older people.
Extrapolation and aggregate models
Simpler methods extend an observed trend in total population. Constant exponential growth, for example, gives:
where is a constant growth rate and is elapsed time.
Linear extrapolation, logistic curves, and time-series models are also used. These approaches can be useful when detailed data are unavailable or for short-term comparisons. However, they do not necessarily represent the demographic processes underlying change. A constant growth rate cannot by itself explain how changing fertility or age structure alters future growth.
Multiregional and multistate models
Multiregional models represent movements among several places simultaneously, so that migration out of one region can be linked to migration into another. This improves consistency across geographic units.
Multistate models additionally distinguish characteristics such as educational attainment, partnership status, or labor-force participation. They incorporate transitions between states as well as demographic events. These models offer richer detail but require more data and more assumptions.
Small-area projections may combine demographic models with housing development, land availability, or local administrative information. Their results are often constrained to sum to a larger regional or national total.
Data and assumptions
Base population
The starting population commonly draws on a population census, population registers, household surveys, and administrative records. Birth and death data often come from civil registration systems.
Before projection, statisticians may adjust for undercounting, age misreporting, incomplete registration, or inconsistencies among sources. Residence definitions also matter: a population defined by usual residence is not necessarily the same as one based on citizenship or physical presence on a particular date.
Fertility
Assumptions about fertility concern both the number of births and their timing. The total fertility rate summarizes a period’s age-specific fertility schedule, but it does not reveal every feature needed for projection. Different age patterns can produce the same total fertility rate while generating different birth trajectories.
Long-term assumptions may reflect historical trends and theories of the demographic transition. They remain uncertain, particularly where birth patterns are changing rapidly.
Mortality
Mortality assumptions usually specify age- and sex-specific death rates or survival probabilities, often organized through a life table. Life expectancy is a useful summary, but different mortality schedules can produce similar life expectancies.
Changes in mortality at older ages are especially important for projections of population ageing. Epidemics, conflict, and other disruptions may require adjustments that cannot be inferred from smooth historical trends alone.
Migration
Migration assumptions may distinguish immigration from emigration, internal from international movement, and different ages or categories of migrants. Migration can respond quickly to economic conditions, policy changes, conflict, and family networks, making it particularly difficult to project.
Net migration alone can conceal large flows in both directions. Models that use only a net total may therefore lose information relevant to age composition and subsequent demographic behavior.
Scenarios and uncertainty
Deterministic scenarios
A deterministic projection applies a specified set of assumptions and produces one resulting trajectory. Agencies often publish a central scenario alongside alternatives with higher or lower fertility, different mortality improvements, or different migration levels.
Such scenarios explore sensitivity; they do not automatically define a probability range. “High” and “low” variants are not necessarily upper and lower bounds, and observed outcomes can fall outside them. A scenario combining several extreme assumptions may also be less plausible than its individual assumptions considered separately.
Probabilistic projections
Probabilistic projections use statistical models to generate distributions of future demographic outcomes. These may incorporate uncertainty in future fertility, mortality, migration, and sometimes the base population.
Results can include medians, prediction intervals, and probabilities of events such as population decline or exceeding a specified size. Their interpretation depends on the model, its calibration, and its treatment of dependence across countries, ages, and demographic components. For example, national outcomes cannot generally be aggregated reliably by assuming that all countries’ uncertainties are independent.
Probability intervals express uncertainty within a modeling framework. They do not guarantee that every possible structural change or exceptional event has been represented.
Horizon and demographic momentum
Uncertainty generally increases with the projection horizon, but not equally for every group. Many people who will be older adults in twenty years are already alive, whereas most people who will be young adults several decades from now have not yet been born.
Population momentum explains why population growth may continue after fertility reaches replacement level: a youthful age structure can produce many births as large cohorts enter reproductive ages. Conversely, an older age structure can sustain decline even after fertility rises. Replacement-level fertility is therefore not a condition for immediate zero population growth.
Historical development and institutional practice
Modern projections developed from population accounting, life tables, and age-structured demographic analysis. During the twentieth century, cohort-component methods became central to official statistical practice, while computing made repeated calculations for detailed populations increasingly feasible.
National statistical offices produce projections for domestic planning. The United Nations, through its Population Division, publishes internationally comparable population estimates and projections in World Population Prospects. Later methodological developments have included probabilistic modeling, more detailed migration analysis, and projections differentiated by characteristics such as education.
Each release is tied to a particular data base and set of assumptions. A revised projection may differ because new evidence changes the estimated current population, because future assumptions change, or both. Comparing successive releases therefore requires separating revisions to the historical baseline from revisions to the projected trajectory.
Applications
Population projections provide demographic inputs for many planning exercises:
- Education: estimating the future number of school-age children and demand for teaching capacity.
- Health and care: assessing how population size and age composition may affect demand on a health system.
- Employment and pensions: examining potential changes in working-age populations and retirement-age groups.
- Housing and infrastructure: informing urban planning, transport, utilities, and housing requirements.
- Environmental analysis: combining population scenarios with assumptions about consumption, technology, land use, and climate change.
Population totals alone do not determine these outcomes. A projection of working-age people is not a forecast of employment; a population projection is not automatically a projection of households; and a rising older population does not translate mechanically into a fixed level of care expenditure. Additional behavioral, institutional, and economic assumptions are required.
Limitations and interpretation
A projection can be arithmetically precise without being predictively certain. Its reliability depends on data quality, the geographic scale, the horizon, and the stability of the processes modeled. Small-area results may be especially sensitive to housing construction, institutional openings or closures, and a relatively small number of migrants.
Aggregate accuracy can also conceal compositional error. A projection may approximate the total population while misrepresenting its age structure or geographic distribution. Evaluation therefore examines both totals and components, compares results with simple benchmarks, and assesses whether probabilistic intervals achieve their intended coverage.
Interpretation requires attention to the base date, residence definition, age grouping, geographic boundaries, and assumptions. A projected peak population, date of decline, or future population share belongs to a particular projection release and scenario, not to an independently observed future fact.