A life table is a structured way of tracking how a population dies off over time, age by age, starting from a group of newborns (or any starting age) and following them until the last survivor is gone. At each age, the table records how many individuals are still alive, how many die before reaching the next age, and what the probability of dying is during that interval. From these columns, you can calculate life expectancy at any age, not just at birth. Life tables show up in surprisingly different fields, from pension planning and insurance pricing to wildlife conservation and cancer research, but the underlying logic is always the same: count who survives, count who doesn’t, and see what the pattern looks like.
Where Life Tables Came From
The idea of systematically recording deaths by age and doing math with the results goes back over three centuries. Edmond Halley, better known for predicting the return of a comet, published what is widely considered the first life table based on actual population data in 1693. He used birth and death records from the city of Breslau (now Wrocław, Poland) to construct a table showing survival at each age.1Journal of the Royal Statistical Society Series A: Statistics in Society. A New Look at Halley’s Life Table Before Halley, a few people had made crude estimates of mortality, but nobody had used real demographic data in a systematic way. His table allowed, for the first time, a rational basis for pricing annuities, which were financial products the English government was selling to raise money. The gap between “roughly guessing how long people live” and “building a mathematical table from real records” turned out to be one of the most consequential leaps in the history of statistics.
What the Columns Actually Tell You
A standard life table starts with a hypothetical group, often 100,000 people born in the same year. This is called the radix. At each age (or age interval), the table provides several quantities. The most fundamental is the probability of dying during that interval. From that single column, everything else flows: the number of survivors at each age, the number of deaths in each interval, the person-years lived within each interval, and the total person-years remaining for everyone still alive. Dividing that last quantity by the number of survivors gives you the life expectancy at that age.
Life expectancy at birth often gets misunderstood. When someone says “life expectancy was 40 in the 1800s,” people picture adults routinely dying at 40. In reality, high infant and childhood mortality dragged down the average enormously. A person who survived to age 20 in the 1800s could expect to live considerably longer than 40. Life tables make this crystal clear because they show life expectancy at every age, not just at birth. Researchers have developed formulas that express death rates, survival probability, and life expectancy at various ages with close agreement to reported national data, making these tables practical tools for epidemiological evaluation.2International Journal of Bio-Medical Computing. Formulas expressing life expectancy, survival probability and death rate in life table at various ages in US adults
There are two main flavors. A “period” life table uses death rates observed in a single calendar year (or short span) and applies them to a hypothetical cohort, as if those rates stayed fixed for a lifetime. This is what government agencies typically publish. A “cohort” life table follows an actual birth cohort through real time, recording what actually happened to people born in, say, 1920, as they aged through the decades. Cohort tables are more accurate for the generations they describe but can only be completed once nearly everyone in the cohort has died, which makes them useless for anyone currently alive. Period tables are a snapshot; cohort tables are a documentary.
The Shape of Mortality With Age
One of the striking features that life tables reveal is how mortality risk changes across a human lifetime. It follows a rough bathtub shape: relatively high in infancy, dropping to its lowest point somewhere around age 10, then rising slowly through early adulthood and accelerating steeply in middle and old age. The steep upward curve after around age 30 is so regular that it was formalized in the nineteenth century as what demographers call the Gompertz law: after a certain point, the risk of dying roughly doubles at a fixed interval (about every eight years in humans). A recent study applying this framework to surgical patients found that the exponential rise in mortality risk becomes apparent around age 30 and holds across subgroups defined by sex, surgical urgency, cancer status, and ethnicity.3PubMed. Biodemography of Human Aging (Gompertz-Makeham Law) Applied to Surgical Mortality Modeling: A Retrospective National Cohort Study
What happens at the far end of life is more contentious. A team analyzing Italian semi-supercentenarians reported essentially flat hazard curves beyond age 105, suggesting that mortality risk stops climbing and hits a plateau.4PubMed Central. The plateau of human mortality: Demography of longevity pioneers That finding stirred considerable debate. Analysis of data from multiple countries found that the risk of dying at age 105 sits around 0.6 to 0.7 per six-month interval for women, rising only slowly toward about 0.8 by age 110.5PLoS ONE. Regularities in human mortality after age 105 That looks like a gentle plateau. But a competing analysis argued that demographic errors, including misreported ages and blending of slightly different birth cohorts, can fully account for the apparent leveling-off. Error rates as low as one in ten thousand were shown to be enough to produce artificial plateaus, and correcting for known error rates in birth and death registration eliminated the deceleration entirely.6PLOS Biology. Errors as a primary cause of late-life mortality deceleration and plateaus This matters beyond academic curiosity: whether mortality truly plateaus at extreme ages has implications for projections of maximum human lifespan and for pension systems that need to estimate how many centenarians they will be paying.
Forecasting Future Mortality
Period life tables are snapshots, but governments and insurers need to plan decades ahead. That means forecasting how death rates will change over time. The dominant method for doing this was introduced in 1992 and captures the pattern of mortality change using a single time-varying index. Thirty years of retrospective evaluation have shown that it works surprisingly well for many populations, and it spawned an entire family of extensions and refinements.7International Journal of Forecasting. Thirty years on: A review of the Lee–Carter method for forecasting mortality The basic idea is to decompose mortality into a fixed age pattern and a trend that shifts over time. Because improvements in mortality have been remarkably steady in many developed countries, this linear trend captures most of the action.
The approach has known blind spots, though. It assumes the age pattern of mortality improvement stays constant, which can go wrong when a shock hits one age group harder than others, as happened with the opioid crisis in the United States or the HIV epidemic in sub-Saharan Africa. Extensions of the method try to address this by incorporating additional components or by smoothing the input data to reduce noise.8PubMed Central. Mortality Forecasting with the Lee-Carter Method: Adjusting for Smoothing and Lifespan Disparity Still, the core method remains the most widely used starting point for probabilistic mortality forecasts. When you hear a government actuary say something like “Social Security trust funds will be depleted by year X,” the death-rate projections feeding that estimate almost certainly owe something to this modeling tradition.
Life Tables Beyond Humans
Ecologists have been building life tables for plants and animals since at least the 1940s. The logic is identical: track a cohort, record survival at each age or stage, and see how mortality is distributed across the lifespan. But the patterns look radically different from the human bathtub curve. Biologists classify survivorship curves into broad types. A Type I curve, characteristic of humans and other large mammals, concentrates mortality in old age. Type II shows roughly constant mortality at all ages, common in some bird species. Type III, typical of many marine invertebrates, fish, and plants, involves massive early mortality with high survival among the lucky few who make it past the juvenile stage.9PubMed Central. Measuring the shape of mortality across animals and plants: Alternatives to H entropy metrics reveal hidden type IV survivorship curves and associations with parental care at macro-ecological scales
Plants illustrate the extremes. Studies of perennial grassland species have found that grasses strongly follow Type III survivorship, meaning most individuals die young, while many forbs (herbaceous flowering plants) show something closer to Type II, with more constant survival rates across ages.10Journal of Ecology. Demography of perennial grassland plants: survival, life expectancy and life span A separate study across six semi-arid ecosystems confirmed this distinction, finding both forbs and grasses leaned Type III but that forbs sat closer to the Type II boundary.11Journal of Vegetation Science. Life form influences survivorship patterns for 109 herbaceous perennials from six semi‐arid ecosystems These differences matter for conservation: a Type III species that loses nursery habitat may see catastrophic population decline, while a Type I species is more sensitive to threats affecting adults.
For organisms whose age is hard to determine, or whose biology involves multiple distinct life stages (think caterpillar to butterfly, or seedling to mature tree), ecologists use stage-structured models instead of age-based tables. These models track transitions between stages rather than between birthdays. Researchers have shown that all the standard age-based life history measures, like generation time and population growth rate, can still be derived from stage-structured models.12Ecological Monographs. Simple Methods for Calculating Age‐Based Life History Parameters for Stage‐Structured Populations The key is getting the transition rates right: how fast individuals move from one stage to the next, and how many die along the way. Converting between life-table data and stage-structured models introduces some bias, so the method of estimation matters.13PubMed Central. Constructing stage-structured matrix population models from life tables: comparison of methods
Measuring Not Just How Long, but How Well
Knowing that life expectancy at birth is 78 or 82 years tells you something, but it doesn’t tell you how many of those years are spent in good health. That question led to the development of disability-free life expectancy, which merges a standard life table with survey data on disability prevalence at each age. The most common approach, Sullivan’s method, has been the go-to technique for over 30 years. It works by weighting each age interval’s person-years by the proportion of the population at that age who are disability-free.14PubMed Central. On the Estimation of Disability-Free Life Expectancy: Sullivan’ Method and Its Extension The result splits total life expectancy into years lived with and without disability.
Tracking this measure over time reveals whether gains in longevity are being matched by gains in health. Researchers have applied Sullivan’s method to US data spanning four decades, from 1970 to 2010, using mortality rates from vital statistics and disability prevalence from national health surveys.15PubMed Central. Trends Over 4 Decades in Disability-Free Life Expectancy in the United States Whether the extra years people gained over those decades were healthy years or years spent managing chronic conditions has significant implications for healthcare spending, workforce planning, and the personal experience of aging.
Cause-Deleted Life Tables and What Kills Us
A standard life table treats death as a single event. Cause-deleted life tables ask a more targeted question: what would happen to life expectancy if we could eliminate one specific cause of death? If nobody died of heart disease, how much longer would people live on average? The answer isn’t as straightforward as you might think, because people who don’t die of heart disease don’t become immortal. They remain exposed to cancer, stroke, accidents, and everything else. The gain in life expectancy from eliminating any single cause is always less than you’d guess by looking at how many deaths it currently accounts for.
Researchers have developed integrated approaches that combine cause-deleted analysis with decomposition methods, attributing changes in life expectancy over time to specific causes of death.16PubMed Central. An integrated approach to cause-of-death analysis: cause-deleted life tables and decompositions of life expectancy This kind of analysis is how public health officials can say things like “reductions in cardiovascular mortality contributed X years to the increase in life expectancy between 1990 and 2020.” It also reveals competing risks: eliminating one cause sometimes has a smaller effect than expected because the people saved go on to die relatively soon of something else. The concept of competing risks is essential whenever a life table is used to evaluate an intervention that targets only one disease.
Why Women Outlive Men in Almost Every Life Table
One of the most consistent patterns visible across nearly all human life tables is that women live longer than men. The gap varies by country and era, but the direction is remarkably stable. Explaining why has turned out to be harder than documenting it. Proposed biological mechanisms include hormonal effects on inflammatory and immune responses and greater resistance to oxidative damage, but current support for these mechanisms is weak.17PubMed Central. Sex Differences in Lifespan
The story has a historical twist. The female advantage in life expectancy widened dramatically in the early twentieth century, and the reasons go beyond declining maternal mortality and falling fertility. Research using cause-of-death data from Massachusetts going back to 1887 showed that females between the ages of 5 and 25 were disproportionately affected by infectious diseases. As the burden of infectious disease fell, both sexes benefited, but women benefited more.18PubMed. XX > XY?: The changing female advantage in life expectancy This is a useful reminder that life table patterns are not just biology; they are shaped by the diseases, behaviors, and environments of each era. The combination of biological and socio-behavioral factors, along with historical and lifestyle changes, continues to evolve.19Genus. Gender differences in survival across the ages of life: an introduction
Where the Data Come From
Building a reliable life table requires good data on deaths and on the population at risk, broken down by age and sex. In countries with well-functioning vital registration systems, this is straightforward: civil authorities record births and deaths, and a census counts the living. But data quality varies enormously around the world, and even high-income countries have imperfections in their records, especially at extreme ages.
The most important international repository for high-quality life table data is the Human Mortality Database, which applies uniform methods for constructing life tables and calculating death rates across all the countries it covers, currently including detailed data for 38 countries or areas. The project emphasizes comparability across countries and over time, maximum data quality, and free public access.20PubMed Central. Data Resource Profile: The Human Mortality Database (HMD) For populations with incomplete data, particularly in low-income countries where vital registration is patchy, demographers rely on model life tables. These are templates based on the age patterns of mortality observed in populations with good data. Given just one or two pieces of information, such as child mortality alone or child and adult mortality together, a model life table can estimate the full age pattern of death rates for a population that lacks complete records.21PubMed Central. A flexible two-dimensional mortality model for use in indirect estimation
This is how global estimates of life expectancy, including those published by the United Nations and the World Health Organization, are produced for countries that cannot construct proper life tables from their own data. The model life tables are not arbitrary guesses; they are constrained by the regularities that mortality patterns follow across human populations. But they are still models, and their accuracy depends on how well the target population matches the historical patterns encoded in the model. Populations experiencing unusual shocks, like wars or epidemics, can deviate from model predictions in ways that only become apparent after the fact.
Evolutionary Perspectives on Aging Patterns
Life tables don’t just describe mortality; they raise the question of why organisms age the way they do. The disposable soma theory proposes that aging arises from an evolutionary balancing act between investing resources in reproduction and investing in bodily repair. Organisms that pour everything into reproduction at the expense of cellular maintenance age faster; those that invest more in repair can live longer but reproduce less.22PubMed. Evolution of aging: individual life history trade-offs and population heterogeneity account for mortality patterns across species Mathematical models of this trade-off predict specific shapes for the mortality-rate-versus-age curve, connecting evolutionary theory directly to the kind of data life tables provide.23PubMed. OPTIMALITY THEORY, GOMPERTZ’ LAW, AND THE DISPOSABLE SOMA THEORY OF SENESCENCE
Species with high external mortality from predation or environmental hazards tend to invest less in repair and age faster, because the odds of surviving long enough to benefit from a well-maintained body are low. Species in protected environments, such as turtles with shells or birds that can fly from predators, tend to age more slowly. Life tables across the animal kingdom confirm these broad patterns, though there are plenty of surprises. Some organisms, like certain tortoises and rockfish, show negligible senescence over decades, while others, like Pacific salmon, die almost immediately after reproducing. The diversity of survivorship curves across the tree of life is one of the richest datasets for testing evolutionary theories of aging, and life tables are the tool that makes those comparisons possible.
Working Life Tables and Economic Applications
Life tables built for economic purposes track not just survival but labor force participation. A working life table follows a cohort and records, at each age, the probability of being employed, becoming unemployed, re-entering the workforce, retiring, or dying. These tables are used in legal proceedings to calculate the economic value of a life lost to injury or wrongful death: how many more years would the person have worked, and how much would they have earned? They are also used by pension actuaries to estimate how long retirees will collect benefits after leaving the workforce. The modeling approach can incorporate transitions between multiple states, including employment, disability, and retirement, rather than just alive and dead.
Insurance pricing relies heavily on life tables as well. Premiums for life insurance and annuities are fundamentally bets on how long policyholders will live, and the mortality rates from life tables are the inputs that determine those bets. Insurers maintain their own proprietary tables that separate policyholders by risk factors like smoking status and health history, producing more granular mortality estimates than the general-population tables published by governments. The gap between the general-population table and the insured-population table can be substantial, because people who buy life insurance tend to be healthier than average (a selection effect insurers account for in pricing).