What Is a Life Table? Components and Applications

A life table is a structured summary of how a population dies off over time, tracking how many individuals survive from one age to the next and how long they can expect to live at each point. It is one of the oldest tools in statistics, with roots stretching back to the 1600s, and it remains central to fields as different as insurance pricing, public health planning, ecology, and clinical medicine. The basic idea is simple, but the applications and the nuances behind the numbers are surprisingly rich.

The Basic Idea

A life table starts with a hypothetical group of people, usually 100,000 newborns. It then applies the observed death rates at each age to that group, showing how many survive to age 1, age 2, age 3, and so on until no one is left. The result is a step-by-step portrait of mortality for a given population and time period. You can read it from top to bottom and see, at a glance, where the heaviest losses happen and how rapidly survival declines with age.

The concept dates to seventeenth-century England. John Graunt, a London merchant, published a table in 1662 that tracked the survivors of 100 conceptions at intervals through age 86, drawing on London’s “Bills of Mortality” (weekly parish records of burials). Edmond Halley, the astronomer better known for his comet, refined the approach a few decades later using data from the city of Breslau. The question of who deserves more credit has been debated ever since, but together they gave the world a tool that has been in continuous use for over three centuries.¹

The Columns You Will See

Life tables come in different formats depending on the field, but human demographic tables almost always include the same handful of columns. Each one tells you something distinct about what is happening to the population at a given age.

• Age interval (x): The starting age for each row. Tables may use single years or broader brackets like five-year intervals.
• Death rate (m_x or n_Mx): The observed rate of death during that age interval, drawn from real census and vital registration data.
• Probability of dying (q_x): The chance that someone who reaches the start of the interval will die before reaching the next one. This is the column that does the heaviest lifting, since it converts raw death rates into the conditional probabilities the rest of the table depends on.
• Survivors (l_x): The number of people, out of the original hypothetical cohort, still alive at the start of each age interval. The first row is always the starting number (commonly 100,000), and each subsequent row is lower.
• Deaths (d_x): How many from the cohort die during each interval. This is simply the difference between successive l_x values.
• Person-years lived (L_x): The total time lived by the cohort within that interval. People who die partway through an interval contribute a fraction of a year rather than a full year.
• Total person-years remaining (T_x): All the years of life the survivors from age x onward will collectively live. It is the running sum of L_x from the bottom of the table up to age x.
• Life expectancy (e_x): The average number of additional years a person at age x can expect to live, calculated by dividing T_x by l_x. Life expectancy at birth (e₀) is the figure most often quoted in the news, but the column exists for every age.

Life expectancy at birth is frequently misunderstood. When historical tables report an e₀ of 35 or 40, people sometimes assume no one lived past middle age. That is rarely accurate. High infant and childhood mortality drags the average down dramatically, while adults who survived those early years often lived well into their sixties or seventies. Life expectancy at age 5, or at age 20, tells a very different story than life expectancy at birth.

Period Tables Versus Cohort Tables

There are two fundamentally different ways to build a life table, and the distinction matters more than it might seem.

A period life table uses death rates from a single calendar period, often a single year or a span of a few years. It answers the question: if someone experienced today’s age-specific death rates for their entire life, how long would they live? This is a useful snapshot, but it is a hypothetical. No real person will experience 2024’s death rates at age 10, then again at age 50, then again at age 90, because mortality rates change over time. Period life expectancy tends to underestimate how long people actually live, because in most countries mortality has been falling decade after decade.

A cohort life table follows a real group of people born in the same year and records what actually happened to them. It is more accurate in hindsight, but you cannot complete one until nearly everyone in the cohort has died, which makes it useless for planning anything about the living. Most government-published tables and the figures you see in news reports are period tables, supplemented occasionally by projected cohort estimates.

The gap between these two approaches is not trivial. A sudden improvement in medical care, for example, can postpone large numbers of deaths in a short period. When that happens, the period table can exaggerate the longevity gain because it treats the avoided deaths as if they represent a permanent shift, not a one-time postponement. Demographers call this a “tempo effect,” and it can distort the interpretation of trends in period life expectancy.²

Life Tables in Insurance and Pensions

Life tables were arguably built for insurance. Pricing a life insurance policy or an annuity requires knowing the probability that a person of a given age will die within the next year. That probability, the q_x column, is the foundation of every premium calculation in the industry.

Insurers do not just use one national table. They build tables tailored to their policyholders, adjusting for the fact that the people who buy life insurance tend to be healthier than the general population (a phenomenon called selection bias). Separate tables exist for smokers and non-smokers, for men and women, and for people who pass medical underwriting versus those in simplified-issue products. Pension funds use their own tables, often with different assumptions about how fast mortality will improve in the future, since underestimating future longevity means the fund runs out of money.

Because mortality keeps falling, actuaries need to project death rates forward. One of the most widely used forecasting methods is the Lee-Carter model, which separates the overall mortality pattern by age from a time trend that captures how quickly death rates are declining. The time trend is then extended into the future using standard forecasting techniques.³ Variations on this model are used around the world to build the mortality tables that underpin pension funding and insurance reserves.⁴

Public Health and Cause-of-Death Analysis

Governments publish life tables as a standard measure of population health. The life expectancy figures you hear reported annually come straight from these tables. But the real analytical power shows up when researchers start taking the table apart by cause of death.

A cause-deleted life table asks: what would life expectancy be if we could eliminate a specific cause of death entirely? If you remove all deaths from heart disease, for instance, how much would life expectancy rise? This is not a prediction of what would happen in practice (eliminating heart disease would shift some of those people into other causes of death), but it quantifies how much each disease contributes to the overall mortality burden. Researchers combine cause-deleted tables with decomposition methods that attribute observed changes in life expectancy to specific diseases, making it possible to see exactly which conditions drove a country’s gains or losses over a given period.⁵

This kind of analysis was critical during and after COVID-19, when countries needed to understand how much the pandemic had reduced life expectancy relative to trends, and which age groups bore the brunt. Life table decomposition turns a single headline number into a detailed diagnostic.

Healthy Life Expectancy

Raw life expectancy tells you how long people live, but it says nothing about how well they live. An extra five years of life spent entirely with severe disability is a very different thing from five healthy years. That gap motivated the development of health-adjusted life expectancy, often called HALE or healthy life expectancy.

The basic approach, known as Sullivan’s method, takes a conventional life table and layers in data on the prevalence of disability or disease at each age. Years lived with disability are discounted, so the resulting figure reflects the average number of years a person can expect to live in good health.⁶ Calculating HALE requires three inputs: probabilities of death at each age (from the standard life table), prevalence of diseases or disabilities, and a set of disability weights that capture how much each condition reduces quality of life.⁷

The disability weights are where things get contentious. How much to discount a year lived with, say, mild back pain versus severe dementia is partly a value judgment. Different methods for handling people who have multiple conditions at once also produce meaningfully different HALE estimates. One study found that switching the adjustment method for comorbidity lowered HALE estimates by roughly one to two years, depending on sex.⁸ That may sound small, but in policy terms a one-year shift in national HALE is a significant change.

More sophisticated versions go beyond Sullivan’s snapshot approach and use multistate life tables, which model people moving back and forth between healthy and disabled states over time rather than treating disability prevalence as fixed.⁹ The multistate approach can show, for example, that a 65-year-old woman is likely to spend a certain number of her remaining years with limitations in daily activities like bathing or dressing, then potentially recover to an independent state for a period, before declining again.¹⁰ These transitions between states are invisible in a standard life table but matter enormously for health-care planning and long-term care policy.

Clinical Survival Analysis

If you have ever looked at a medical study reporting how long patients survive after a diagnosis or treatment, you have seen a clinical cousin of the life table. In medicine, the actuarial life table method divides follow-up time into intervals and calculates survival probability within each one, much like a demographic life table does with age groups. The Kaplan-Meier method, which is more commonly used in modern clinical research, takes the same basic idea but recalculates at every individual event (each death or withdrawal from the study) rather than at fixed intervals. Both are tools for estimating survival when you do not have complete data on everyone, which is the norm in clinical trials where patients drop out, move away, or are still alive when the study ends.¹¹

The connection to life tables is more than cosmetic. The logic is identical: at each time point, you calculate the conditional probability of surviving to the next time point, given that you have survived to this one, and then multiply those probabilities together to get the cumulative survival curve. A Kaplan-Meier curve stepping downward over months after surgery is, at its mathematical core, the same object as a life table’s l_x column stepping downward over years of age.

Life Tables Outside the Human World

Ecologists adopted life tables decades ago to study animal and plant populations. The structure is the same: track a cohort, record survival at each age or stage, and calculate life expectancy and reproductive rates. But the questions are different. Instead of pricing insurance, ecologists want to know whether a species’ population is growing, shrinking, or stable, and which life stage is the bottleneck.

A key concept borrowed from life table analysis is the survivorship curve, which plots the l_x column on a logarithmic scale. Three idealized types exist. Type I curves, where most individuals survive to old age and then die relatively quickly, describe humans in wealthy countries and many large mammals. Type II curves show a roughly constant death rate at every age, characteristic of some birds and reptiles. Type III curves describe species in which massive early mortality is followed by high survival among the few that make it through, typical of many fish, invertebrates, and plants. Research on perennial grassland plants, for example, found that grasses tend strongly toward Type III survivorship (huge losses among seedlings), while many forbs had more constant survival rates with age, placing them closer to Type II.¹² A broader study spanning six semi-arid ecosystems confirmed that both grasses and forbs followed Type III curves overall, though forbs were closer to Type II than grasses.¹³

Entomologists build life and fertility tables for insects, combining survival data with age-specific reproduction to calculate population growth parameters. A study on the predatory stink bug Podisus nigrispinus, for instance, used life tables to determine that a single female produced an average of about 32 female offspring during her lifetime and that the population’s generation time was roughly 47 days.¹⁴ This kind of information is directly useful for biological pest control programs, where you need to know how fast a beneficial predator population can grow relative to the pest it is supposed to suppress.

Conservation biologists use life tables to model endangered species and estimate how quickly a population can recover from decline. Age-specific survival and fertility rates feed into population projection models that help managers decide where to focus resources: improving juvenile survival, protecting breeding adults, or some combination. The translation from life table parameters into stage-structured population models has its own methodological challenges, since real organisms often do not fit neatly into the age-based framework a life table assumes.¹⁵

Limitations and Common Pitfalls

Life tables are powerful, but they rest on assumptions that can quietly go wrong. The most fundamental is the stationarity assumption: a period life table implicitly assumes that the population is neither growing nor shrinking and that the age-specific death rates remain constant. Real populations violate this constantly. When mortality is falling, as it has been in most countries for decades, a period life table underestimates how long today’s newborns will actually live. The tempo effects mentioned earlier are one specific version of this problem.

Data quality is another vulnerability. Life tables are only as good as the death and population counts that feed them. In countries with incomplete vital registration, or in historical populations where records are sparse, the resulting tables can be misleading. Paleodemographers who try to build life tables from archaeological skeletal remains face even steeper challenges: age estimation from bones is imprecise, especially for older adults, and the assumption that the excavated skeletons represent a stationary population is often unrealistic.¹⁶

Even in modern populations with excellent data, small errors compound at the oldest ages. One analysis showed that age-estimation and cohort-blending errors at rates as low as one in ten thousand are enough to create the appearance of a mortality plateau in very old age, where death rates seem to level off rather than continue rising. The researchers argued that this late-life mortality deceleration, which had spawned its own subfield of biological and evolutionary theories, is largely an artifact of data errors rather than a real biological phenomenon.¹⁷ Others have offered biological explanations for the same pattern, attributing it to population heterogeneity: frailer individuals die first, leaving a progressively hardier group whose mortality rate rises more slowly.¹⁸ The debate is unresolved, but it is a good illustration of how much analytical weight rests on the accuracy of a life table’s oldest rows.

Life Tables for Ancient and Prehistoric Populations

Archaeologists and anthropologists have tried for decades to reconstruct the demographic profiles of past populations using life tables built from cemetery data. The appeal is obvious: if you can estimate the age at death for a collection of skeletons, you can apply life table methods to infer mortality patterns, average lifespans, and population structure. In practice, the exercise is fraught with problems that go beyond ordinary data quality issues.

Age estimation from skeletal remains is reliable for children and young adults but becomes increasingly imprecise after about age 40. Many methods tend to compress older individuals into a narrow range, making it look as though few people survived past 50 when in fact the methods simply cannot distinguish a 55-year-old skeleton from a 70-year-old one. The stationarity assumption is even harder to defend here than in modern populations, since ancient communities experienced famines, epidemics, migrations, and warfare that made stable population sizes the exception rather than the rule.¹⁶ There is also the problem of selective burial: not everyone in a community ends up in the excavated cemetery. Infants may be buried elsewhere, elites may have separate tombs, and some individuals may not be buried at all. Each of these biases distorts the age-at-death distribution in ways that a life table cannot correct for on its own.

None of this means paleodemographic life tables are useless. They remain one of the few quantitative windows into the mortality experience of past populations. But the results need to be read as rough sketches rather than precise portraits, and the assumptions behind them deserve as much attention as the numbers they produce.