Different Approaches to Life Tables: Cohort vs. Period Tables

Life tables sit quietly behind some of the most important numbers we hear about a population, from the average life expectancy reported in the news to the premium an insurance company charges for a policy. But not every life table is built the same way. Demographers, public health researchers, and actuaries use different approaches depending on the question they want to answer. The two big distinctions you will keep running into are cohort vs. period life tables, and single-decrement vs. multiple-decrement life tables. Understanding what separates them is the difference between reading a number correctly and misinterpreting it entirely.

What a life table actually tries to do 🔗

At its simplest, a life table follows a hypothetical or real group of people and records how many of them die at each age. From these age-specific death probabilities, demographers derive familiar measures like survival rates and life expectancy at birth. As the U.S. Social Security Administration explains, a standard life table starts with a hypothetical cohort of 100,000 births and tracks the survivors, deaths, and probabilities of death (qx) at each integer age. The structure looks the same across most tables; what changes is the kind of data feeding into it and the kind of attrition being counted.

That is where the two big classifications come in. The first asks whose mortality is being measured: a real generation followed over time, or a snapshot of everyone alive in a given year. The second asks what kind of exit is being counted: death alone, or a combination of events like marriage, divorce, contraceptive discontinuation, or specific causes of death.

Cohort life tables: following a generation through life 🔗

A cohort life table, sometimes called a generation life table, follows a real group of people born within the same time interval and records their actual mortality experience as the years pass. To build a true cohort life table for everyone born in 1950, for instance, you would need death data for that group from 1950 onwards until the last member dies. The Office for National Statistics describes a cohort life expectancy as the average number of additional years a person would live considering assumed future changes in mortality for their cohort over the remainder of their life.

The advantage is realism. A cohort table captures what people of that generation actually experienced, including improvements in medicine, sanitation, and nutrition that unfolded across their lifetime. Wikipedia notes that a cohort life table is more frequently used when researchers want to project expected changes in mortality and analyze long-run patterns.

The catch: cohort tables take a lifetime 🔗

The obvious problem is time. If you are studying a cohort born in 1980, you cannot finish their life table until the last one of them is gone, roughly a century from now. So pure cohort tables for living generations are always partly projected, using assumptions about how mortality will change in the future. For older generations whose members have all died, the table is purely historical, but no longer useful for predicting the present.

When cohort tables are most useful 🔗

Cohort tables become essential for long-term financial planning. Pension funds, annuity providers, and long-term care insurers cannot afford to ignore that today’s 30-year-olds will probably live longer than today’s 60-year-olds did at the same age. Using a static snapshot of current mortality would underestimate how long retirees will need to be supported. Cohort tables are also irreplaceable for generational studies in epidemiology, like understanding how exposure to a famine, a war, or a particular pollutant in early life affected the death rates of an entire birth cohort decades later.

Period life tables: a snapshot of mortality in a given year 🔗

A period life table, also called a current or static life table, takes a different shortcut. Instead of following a real generation, it takes the age-specific death rates observed in a single year (or a short window of years) and applies them to a synthetic cohort of 100,000 hypothetical newborns. The question it answers is, “If the mortality rates of this year stayed frozen forever, how long would a person born today live on average?”

The Demography subwiki defines a period life table as one where the survivorship probabilities for all ages are based on data collected in a single period, usually a year. So a 2023 period life table would use the probability of a 40-year-old dying in 2023, the probability of a 60-year-old dying in 2023, and so on, even though those people belong to entirely different birth cohorts.

Why most official life expectancy figures are period figures 🔗

Almost every life expectancy figure quoted in newspapers, government reports, or health policy documents is a period figure. The reason is practical: period tables can be produced every year from current vital statistics, while cohort tables require either a century of data or a forecast. The ONS notes that period life expectancies provide an objective way of comparing trends in mortality over time, between regions, and across countries, which is why official life tables for past years are generally period life tables.

In India, the Office of the Registrar General publishes SRS-based abridged life tables, with the most recent release covering 2019-2023. These are period life tables built from the Sample Registration System, a continuous demographic survey set up in 1970 that records births and deaths in around 8,861 sample units across the country, covering about 0.5 percent of the population. As one assessment of India’s vital registration system explains, the SRS was created as an interim source of vital statistics until the Civil Registration System became reliable, and it has since become the optimal data source for mortality indicators in India.

Limitations of the period approach 🔗

The trade-off for timeliness is that a period table describes a fiction. Nobody actually lives through a single year’s mortality conditions for their entire life. Period tables also get distorted by short-lived shocks. A bad influenza year, a heat wave, or a pandemic like COVID-19 can pull period life expectancy down sharply, even though it tells us little about the long-term prospects of anyone alive. They also underestimate the lifespan of younger generations when mortality is steadily improving, since they assume today’s rates will hold forever.

Single-decrement life tables: one way out 🔗

The cohort/period split is about time. The single/multiple decrement split is about causes of exit. A “decrement” is simply a way a member can leave the cohort being tracked. In a single-decrement life table, there is just one way out.

For an ordinary mortality life table, that single decrement is death, regardless of cause. It does not matter whether a person dies from cancer, a road accident, or old age; they are simply counted as having left the living cohort. This is the kind of table used to compute the standard life expectancy figures published by the SRS and most national statistical offices.

Single-decrement tables are not limited to mortality, though. In family-planning research, a single-decrement table might track discontinuation of contraceptive pill use without distinguishing why the user stopped. In marriage studies, it might track “marital disruption” as a single attrition factor, combining death of a spouse and divorce. The MEASURE Evaluation tutorial points out that a single-decrement table treats divorce and widowhood together as one combined exit, rather than separating them.

Multiple-decrement life tables: several competing exits 🔗

A multiple-decrement life table extends the idea by recognising that members of a cohort can leave through more than one route, and by analysing each route separately. The classic example in formal demography is cause-of-death analysis, where deaths are broken down into competing causes such as cardiovascular disease, cancer, infectious disease, injuries, and so on. Each cause is a separate decrement, and the table computes the probability of dying from each cause at each age, along with the combined probability of dying from any cause.

The IIASA Decrement Life Tables paper describes how the multiple-decrement model was originally designed to recognize different causes of death but is now also used to analyse demographic events that can be viewed in cohort terms, such as marriage and divorce. So a multiple-decrement table for ever-married women, for example, might track women leaving a marriage cohort separately by divorce and by widowhood, instead of lumping them together.

Cause-deleted life tables: a powerful extension 🔗

Once you have a multiple-decrement table, you can ask a useful counterfactual question: how much would life expectancy improve if one cause of death were eliminated? This is done through what demographers call cause-deleted or associated single-decrement life tables. The IUSSP demography training material explains that a cause-deleted life table represents the mortality experience of a population if one cause of death were removed, allowing researchers to estimate the hypothetical gain in life expectancy from eliminating, say, malaria or tuberculosis. These calculations are heavily used in public health priority-setting, because they help quantify how much disease control programmes could extend average life.

Where multiple-decrement tables really earn their keep 🔗

Multiple-decrement tables are workhorses in a wide range of fields. In actuarial science, life insurance and pension models use them to handle the fact that a policyholder can exit a plan through death, disability, surrender, retirement, or lapse. In labour economics, they underlie working-life tables that track entry to and exit from the workforce. In family-planning evaluation, they help compute method-specific discontinuation rates, separating women who stop using a contraceptive because of side effects from those who stop because they want to conceive. The ScienceDirect demography reference notes that multiple-decrement life tables recognise the possibility of attrition through different causes, making them a fundamental extension of the basic life table in demography, public health, and sociology.

Putting the four approaches together 🔗

The two distinctions are not alternatives; they combine. You can have a period single-decrement table (a standard annual life expectancy table for a country), a cohort single-decrement table (a generation’s actual lifespan distribution), a period multiple-decrement table (cause-of-death analysis for a given year), or a cohort multiple-decrement table (a real group’s exits from marriage, employment, or contraceptive use over time). The right combination depends on the question.

If a public health official wants to know whether infant mortality has improved compared to last year, a period single-decrement table works perfectly. If a pension fund wants to know how long today’s 35-year-olds will draw a pension, a cohort table with projected mortality is far more appropriate. If a state health department wants to know whether tackling road-traffic injuries or non-communicable disease would buy more years of life, multiple-decrement and cause-deleted tables are the right tools. The flexibility of the life-table framework is precisely why it has survived as a core method of demography for more than three centuries.

What do you think? If you had to advise an Indian state government on whether to publish cohort or period life expectancy figures for planning a pension scheme, which approach would you recommend, and why? And in which area of population health do you think a multiple-decrement table would reveal the most surprising story right now?

References

1. https://www.ssa.gov/oact/NOTES/as116/as116_I_II_III.html
2. https://www.ons.gov.uk/peoplepopulationandcommunity/healthandsocialcare/healthandlifeexpectancies/methodologies/guidetointerpretingpastandprojectedperiodandcohortlifetables
3. https://en.wikipedia.org/wiki/Life_table
4. https://demography.subwiki.org/wiki/Life_table
5. https://www.ons.gov.uk/peoplepopulationandcommunity/birthsdeathsandmarriages/lifeexpectancies/methodologies/periodandcohortlifeexpectancyexplained
6. https://censusindia.gov.in/nada/index.php/catalog/46175
7. https://www.medrxiv.org/content/10.1101/2020.04.03.20052894.full.pdf
8. https://www.measureevaluation.org/resources/training/online-courses-and-resources/non-certificate-courses-and-mini-tutorials/multiple-decrement-life-tables/lesson-7.html
9. https://pure.iiasa.ac.at/id/eprint/972/7/RM-78-025.pdf
10. https://papp.iussp.org/sessions/papp104_s03/PAPP104_s03_090_010.html
11. https://www.sciencedirect.com/topics/computer-science/decrement-life-table