Life expectancy is a summary measure of mortality that expresses the average number of additional years a person of a given age could expect to live if current age-specific death rates remained unchanged. It is calculated from a life table, a statistical model that converts observed mortality rates into survival probabilities across the lifespan. Life expectancy is a methodological construct describing a hypothetical population, not a prediction for any individual.
Because it condenses a population’s entire mortality experience into a single, comparable number, life expectancy is one of the most widely cited indicators in demography and public health. Understanding it correctly means understanding the life table that produces it and the data that feed that table.
How a life table works
A life table is the engine behind life expectancy. It takes age-specific mortality rates for a population and translates them into a hypothetical cohort, conventionally 100,000 people, that is followed from birth to death. At each age interval the table records the probability of dying, the number surviving, the number of deaths and the years of life lived within the interval. Summing the years lived above each age and dividing by the survivors at that age yields life expectancy for that age.
The key inputs are age-specific death rates, usually derived by dividing recorded deaths in an age band by the corresponding population at risk. These observed rates are converted into probabilities of dying for each interval, and those probabilities drive the survivorship column. Because the method aggregates the entire age structure, life expectancy at birth is sensitive to mortality at every age, not only old age; a fall in deaths among the very young, for example, can raise life expectancy at birth substantially. Statistical agencies publish the full life table so the calculation is transparent and reproducible, allowing analysts to inspect every column rather than trusting a single headline figure. The intervals are usually single years of age or five-year age groups, and an open-ended final interval covers the oldest ages, where special methods are used because deaths there are few and the population is small. Abridged life tables, which use grouped age intervals, are common when detailed single-year data are unavailable, and they yield very similar results to full tables while being simpler to compute and publish.
Period versus cohort life expectancy
The most important distinction is between two ways of assembling the death rates that feed the table.
| Measure | Death rates used | Interpretation |
|---|---|---|
| Period life expectancy | Age-specific rates observed in a single reference period | A snapshot assuming current mortality conditions hold for life |
| Cohort life expectancy | Rates a real birth cohort actually experiences as it ages | Reflects mortality change over the cohort’s lifetime, partly projected |
Period life expectancy is the figure most national statistics offices report routinely, because it requires only recent data: it asks what would happen to a hypothetical group exposed for life to the death rates of a single period. Cohort life expectancy instead tracks an actual generation, using the rates each age group truly experiences as the years pass. It therefore accounts for expected future improvements or deteriorations in mortality, but it depends on projections for the years not yet lived, which introduces assumptions. The two measures can differ substantially, especially when mortality is changing quickly, which is one reason cross-source comparisons require care and explicit labelling of which measure is being used.
Data sources
Life-table construction draws on two foundational data streams. The numerator is mortality records from civil or vital registration systems, ideally capturing every death with age and, where available, cause. The denominator is the population at risk, estimated from a census or a continuously maintained population register and updated for births, deaths and migration between census points.
The quality of life expectancy therefore depends on the completeness of death registration and the accuracy of population estimates. Where registration is incomplete, or where population estimates drift between census years, statisticians apply documented adjustment methods rather than leaving the gaps unaddressed. This makes data provenance central: a life expectancy figure is only as trustworthy as the death counts and population denominators behind it. Treating data lineage as first-class metadata is consistent with the principles in the CASRAI dictionary and with broader work on data infrastructure.
Healthy life expectancy and related measures
Life expectancy is sometimes extended into measures that combine length of life with quality of life. Healthy life expectancy partitions the years a population can expect to live into those spent in good health and those spent in poor health, by combining the life table with survey data on health status. This requires an additional data source describing the prevalence of ill health by age, layered onto the same mortality-based life table. The result answers a different question from life expectancy alone: not only how long people live, but how much of that time is lived in good health. Because such measures depend on subjective or survey-based definitions of health, the definitions used must be documented even more carefully than for the underlying life table, and they are not directly comparable across populations that defined health differently.
Common misinterpretations
Several misreadings recur and are worth naming explicitly. Period life expectancy at birth is not a forecast that today’s newborns will live exactly that long; it describes a hypothetical cohort under fixed current rates that no real generation will ever experience unchanged. A rise or fall between periods reflects changes in age-specific mortality across the population, not a guaranteed individual outcome. Historical life expectancy figures that look low are also often misread, because they were heavily pulled down by mortality in infancy and early childhood rather than implying that few people reached older ages.
Comparing life expectancy across populations requires consistent age-specific inputs, because differences in data collection, registration completeness or population estimation can produce artefactual gaps that have nothing to do with real differences in survival. Because life expectancy depends on the population at risk, it connects directly to census and population data sources, and because it is built from mortality rates it is closely related to death rate and mortality statistics. Researchers reporting derived measures should document their methods and data sources clearly, as encouraged by guidance for authors, so that readers can see exactly which life table and which assumptions produced a given figure.
Frequently asked questions
Is life expectancy a prediction for an individual?
No. It is a population-level average produced by a life table under stated assumptions. It does not forecast how long any specific person will live, because individual longevity depends on many factors the model does not represent, and because the period version assumes rates that will not stay constant.
Why do period and cohort figures differ?
Period life expectancy assumes today’s death rates persist unchanged, while cohort life expectancy follows a real generation and incorporates expected future changes in mortality. When mortality is improving over time, cohort figures are typically higher than period figures for the same reference year.
What data are needed to calculate life expectancy?
Two streams are required: age-specific mortality records from vital registration, and population-at-risk estimates from a census or population register. The ratio of deaths to population at each age produces the death rates that drive the life table, so both streams must be complete and consistent.