Skip to content
Back to search results

Abridged Life Table

Introduced by
Chin Long Chiang; Samuel Preston, Patrick Heuveline & Michel Guillot
Year
1984
Type
Demographic estimation pipeline for mortality and survivorship
Family
Demography / actuarial life tables
Font size

What does it do?

The abridged life table is the workhorse of demography and population health for summarizing the mortality experience of a population in a single, age-grouped table.

Instead of a single-year (complete) life table, it works on broad age intervals — typically <1, 1-4, then five-year groups up to an open-ended oldest interval — which makes it robust when deaths or populations in single years of age are sparse or noisy. The construction propagates a small set of inputs, the age-specific death rates nMx, through a chain of columns: the probability of dying nqx, the survivors lx, the deaths ndx, the person-years lived nLx and Tx, and finally life expectancy ex. Chiang's 1984 treatment supplied the standard estimator and the fraction-of-interval term ax that controls how person-years are allocated within each interval, while Preston, Heuveline and Guillot's 2001 textbook codified the modern pipeline used across demography and epidemiology.

The intuition behind the method

Imagine following an imaginary cohort of 100,000 newborns and asking, at each age, what fraction die before reaching the next age band. The life table is just bookkeeping on that hypothetical cohort, but the death rates fed in come from a real population in a real period. The single trick that makes everything work is translating a rate (deaths per person-year of exposure, which can exceed values you would expect from a probability) into a probability (the chance an individual alive at the start of the interval dies before its end). Once you have those probabilities, you simply multiply survivors down the table, count up the person-years everyone lived, and divide to get the average years of life remaining at each age. The 'abridged' part means you do this on chunky five-year bands rather than single years, trading a little resolution for far more stable estimates — which is exactly what you want for small areas, subgroups, or historical data where single-age counts are too thin to trust.

History

The life table descends from John Graunt and Edmond Halley in the seventeenth century and matured into a rigorous statistical object in the twentieth. Chin Long Chiang's 1984 monograph, The Life Table and Its Applications, gave the abridged table its modern statistical foundation, formalizing the rate-to-probability conversion through the fraction-lived term ax and deriving the sampling variances of life-table functions. The 2001 textbook by Samuel Preston, Patrick Heuveline and Michel Guillot, Demography: Measuring and Modeling Population Processes, consolidated the construction into the standard teaching and research pipeline now used throughout demography and population health, including the handling of the open-ended oldest interval and the linkage to standardization and decomposition methods.

When to use it

Use an abridged life table whenever you need a period summary of mortality — life expectancy, survival probabilities, or person-years — and either your data are tabulated in age groups or single-year counts are too sparse to be stable. It is the default for national and subnational mortality reporting, for comparing mortality across countries, regions, sexes, or social groups, and for producing the lx and nLx inputs that feed decomposition, standardization, and health-expectancy methods. Prefer a complete (single-year) life table when you have abundant single-age data and need fine resolution, for example in actuarial work on a specific cohort. Be cautious in very small populations where even grouped death counts are tiny; there, model life tables or smoothing may be needed, and the open interval and infant nax assumptions deserve particular scrutiny.

Limitations and considerations

  • Coarser age resolution than a complete life table, which can obscure features within broad intervals.
  • Results in the open-ended oldest interval depend on a stable-hazard assumption and on the quality of old-age death and population counts, which are often poor.
  • Period life tables describe a synthetic cohort under current rates, not the actual experience of any real birth cohort, so e0 can be misread as a forecast.
  • Sensitive to denominator errors and age misreporting (heaping, census undercount), especially at the oldest ages where rates are highest.

Source conditions do not replace independent scientific review or validation by an analysis engine.

Strengths

  • Stable and parsimonious: requires only age-grouped death rates yet yields the full suite of survivorship, person-years, and life-expectancy measures.
  • Robust to sparse data because five-year intervals smooth the noise that destabilizes single-year tables for small areas and subgroups.
  • Chiang's nax correction handles the non-uniform timing of deaths, giving accurate person-years even where deaths cluster (infancy, old age).
  • Produces standardized lx and nLx columns that serve as direct inputs to decomposition, standardization, and healthy-life-expectancy pipelines.

Applications

  • Computing national and subnational life expectancy at birth and at older ages for vital-statistics reporting.
  • Comparing mortality and survival across countries, sexes, regions, or socioeconomic groups on a common scale.
  • Producing the lx and nLx columns that feed life-expectancy decomposition, age standardization, and healthy-life-expectancy estimation.
  • Monitoring mortality shocks (epidemics, crises) by recomputing period tables and tracking changes in ex.
  • Constructing actuarial survival probabilities and median ages at death for population health and insurance applications.

How it works

How it works

1

The pipeline begins with the central death rate for each age interval, nMx, defined as deaths to people in that age band divided by the person-years of exposure they contributed during the period — usually approximated by the mid-year (mid-period) population. These rates are the only mortality input the table needs. Getting them right means aligning the numerator (deaths classified by age at death within the interval) and the denominator (population exposed in that interval) to the same period, and handling the youngest interval carefully because infant deaths are concentrated near birth.

Formula
{}_{n}M_x = \frac{{}_{n}D_x}{{}_{n}P_x}
2

The central rate is then converted into the conditional probability of dying within the interval, nqx, for someone alive at its start. Chiang's estimator does this with the term nax, the average number of years lived within the interval by those who die in it. Because deaths are not spread uniformly — they cluster early in infancy and late in old age — nax corrects the rate-to-probability conversion so the table reflects the true timing of deaths. For the youngest ages nax is set from empirical regularities (it is well below n/2), while for most adult bands the midpoint assumption nax = n/2 is a reasonable default.

Formula
{}_{n}q_x = \frac{n \cdot {}_{n}M_x}{1 + (n - {}_{n}a_x)\,{}_{n}M_x}
3

With the probabilities in hand, the synthetic cohort is propagated. Starting from a radix (conventionally l0 = 100,000), the survivors entering the next interval are the current survivors times the probability of surviving the interval, and the deaths in the interval are the difference. Repeating this multiplication down the age axis produces the full lx survivorship curve and the ndx column of deaths by interval, which together describe the shape of mortality across the lifespan for the synthetic cohort.

Formula
l_{x+n} = l_x\,(1 - {}_{n}q_x),\qquad {}_{n}d_x = l_x - l_{x+n}
4

Next the table accumulates person-years lived within each interval, nLx. Survivors who live the whole interval contribute n years each; those who die contribute on average nax years, captured by the same fraction used in the probability conversion. The open-ended final interval is special: with no upper bound, its person-years are obtained by dividing the survivors entering it by the mortality rate prevailing there, which assumes a stable hazard at the oldest ages. This person-years column is what turns a survivorship curve into a measure of time lived.

Formula
{}_{n}L_x = n\,l_{x+n} + {}_{n}a_x\,{}_{n}d_x,\qquad L_{\infty} = \frac{l_{\omega}}{{}_{\infty}M_{\omega}}
5

Finally, cumulating person-years from each age upward to the end of life gives Tx, the total future person-years lived by survivors at age x. Dividing Tx by the survivors lx yields ex, the expectation of life at age x — the headline output of the table and the basis for life expectancy at birth (e0), at age 65, and for downstream tools such as decomposition and healthy-life-expectancy analysis. The full set of columns also supports survival probabilities, median age at death, and actuarial calculations.

Formula
T_x = \sum_{i \ge x} {}_{n}L_i,\qquad e_x = \frac{T_x}{l_x}

Common pitfalls

  • Treating the central death rate nMx as if it were the probability nqx, skipping the rate-to-probability conversion and biasing high-mortality intervals.
  • Using the midpoint assumption nax = n/2 for the infant interval, which overstates infant person-years because infant deaths concentrate near birth.
  • Mishandling the open final interval by applying the n-year person-years formula instead of dividing survivors by the oldest-age rate.
  • Misaligning numerator and denominator periods or age definitions, so deaths and exposure refer to inconsistent populations.

Frequently asked questions

What is the difference between an abridged and a complete life table?

A complete life table uses single years of age, while an abridged table uses broad intervals — typically under 1, 1-4, and five-year groups up to an open oldest interval. The methods are identical in logic; the abridged table simply works on grouped ages, which makes it far more stable when single-year death or population counts are small. For most population-health and cross-national work the abridged table is preferred, and Chiang and Preston, Heuveline and Guillot both present it as the standard form.

Why convert the death rate nMx into a probability nqx at all?

Because survivorship is built by multiplying probabilities of surviving each interval, and a rate is not a probability. A rate has person-years in its denominator and can take values that a probability cannot. Chiang's conversion uses nax, the average years lived in the interval by those who die, to translate the central rate into the conditional probability of dying given survival to the start of the interval. Skipping this step and treating nMx as nqx biases the table, most severely in high-mortality intervals like infancy and old age.

How is the open-ended final age interval handled?

The oldest interval has no upper bound, so the usual person-years formula does not apply. The standard solution, given in Preston, Heuveline and Guillot, sets the person-years in the open interval equal to the survivors entering it divided by the mortality rate prevailing there. This is equivalent to assuming a constant hazard above the cutoff. Because old-age death and population data are often unreliable, results in this interval — and therefore e0 to a small degree — are sensitive to this assumption and to data quality.

Related methods

Sources

  1. Chiang, C. L. (1984). The Life Table and Its Applications. Malabar, FL: Robert E. Krieger Publishing.

    ISBN: 9780898745702
    Copied.
  2. Preston, S. H., Heuveline, P., & Guillot, M. (2001). Demography: Measuring and Modeling Population Processes. Oxford: Blackwell Publishers.

    ISBN: 9781557862143
    Copied.