Historical Life Table Construction
- Introduced by
- John Graunt (origins); Coale, Demeny, Ledermann (model life tables); applied by Wrigley-Schofield and Bengtsson
- Year
- 1662
- Type
- Mortality estimation and survivorship modelling
- Family
- Mortality analysis
What does it do?
Historical life table construction is the craft of converting the patchy mortality evidence of the past—burial registers, family genealogies, monastic obituaries, even skeletal age-at-death distributions—into the formal apparatus of the life table: age-specific death rates, the probability of dying within each age interval, the number of survivors to each age, and expectation of life.
The life table descends from John Graunt's 1662 reading of London's Bills of Mortality and Halley's Breslau table, but applying it to historical populations demands special care, since exposures are rarely known and deaths are often recorded without reliable ages. Historians therefore lean heavily on families of model life tables to smooth noisy data and fill missing age bands. Whether built as period tables capturing a single era's mortality or cohort tables following one birth-year group through life, these reconstructions are the indispensable summary of how, and how long, people lived in the past.
The intuition behind the method
A life table is a thought experiment: take a hypothetical cohort of, say, 100,000 newborns and march them through life, killing off the right fraction at each age according to observed mortality, until none remain. The table records how many survive to each birthday and, from that, how many years of life remain on average at any age. The difficulty in historical work is that we usually know the dead far better than the living—burials are recorded, but the population at risk is hazy. So the historian works from the distribution of ages at death, often supplementing it with assumptions about the population's growth and structure, and leans on model life tables, which encode the regular ways mortality usually varies with age across thousands of real populations. By matching the fragmentary data to the closest model and level, the historian fills the gaps and produces a smooth, internally consistent picture of survival from cradle to grave.
History
The life table originates with John Graunt's 1662 Natural and Political Observations on the London Bills of Mortality and Edmond Halley's 1693 Breslau table, the first to use proper age data. Actuaries and demographers refined the apparatus over the following centuries. Its application to fragmentary historical populations matured in the twentieth century with the creation of model life-table systems—Coale and Demeny's regional tables (1966) and Ledermann's—which gave historians a principled way to handle sparse data. Wrigley and Schofield embedded model tables in their English reconstruction, while the Eurasian Population and Family History Project, reported by Bengtsson and colleagues, derived mortality directly from rich individual-level micro-data, and paleodemographers adapted the tools to skeletal samples.
When to use it
Build a historical life table when you need a coherent summary of a past population's mortality—life expectancy, survivorship, age-specific death rates—and you have at least an age-at-death distribution, ideally with some measure of exposure. It is appropriate for reconstituted parish populations, genealogically documented elites, monastic communities, and skeletal samples in paleodemography. Use model life tables to discipline noisy or incomplete data. It is the natural output stage for family reconstitution and the input mortality structure for inverse projection. Be cautious where ages at death are unreliable, where the population at risk cannot be estimated, or where selective survival of records makes the sample unrepresentative of the living population.
Limitations and considerations
- Exposure is often unknown, forcing reliance on auxiliary structural assumptions
- Age-at-death data suffer heaping, misreporting and infant under-recording
- Model life table fitting imposes a borrowed age pattern that may not fit
- Selective record survival can make samples unrepresentative of the living
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Distils fragmentary mortality data into a single coherent, comparable summary
- Model life tables smooth noise and supply missing age bands
- Applicable across diverse sources from burials to genealogies to skeletons
- Yields life expectancy, the standard cross-period standard-of-living indicator
Applications
- Estimating life expectancy of pre-industrial parish populations
- Constructing mortality tables for ducal and aristocratic genealogies
- Paleodemographic reconstruction of mortality from skeletal age-at-death data
- Comparing mortality regimes across regions and centuries
- Supplying age-specific mortality structures for inverse projection
How it works
How it works
The starting point is a distribution of deaths by age interval, drawn from burial registers that state age, from reconstituted family histories, from genealogical compilations, or from osteological age estimates of skeletal samples. Each interval's death count must be assembled and the inevitable heaping of ages at round numbers, and the under-recording of infant deaths, recognised. The quality and completeness of this age-at-death distribution governs everything that follows in the table.
Formula
{}_nD_x = \text{deaths in age interval } [x, x+n)Central death rates divide the deaths in each age interval by the person-years of exposure lived in that interval. In well-documented settings, such as reconstituted populations or the Eurasian micro-data, exposure is measured directly. Where only deaths are known, exposure must be inferred from an assumed population structure or growth rate, a step that introduces dependence on auxiliary assumptions and is the chief vulnerability of historical mortality work.
Formula
{}_nm_x = \frac{{}_nD_x}{{}_nE_x}Central rates are converted into the probability of dying within each interval using a separation factor that accounts for when, on average, deaths occur within the interval—crucially in infancy, where deaths cluster in the first days and weeks. This conversion turns occurrence-exposure rates into the conditional probabilities of death that drive the survivorship column, and getting the infant separation factor right is essential to a credible table.
Formula
{}_nq_x = \frac{n\cdot{}_nm_x}{1 + (n - {}_na_x)\,{}_nm_x}Beginning with a notional radix, the survivorship column is built by successively applying the probabilities of dying: survivors to the next age equal current survivors times the probability of surviving the interval. This column traces the attrition of the synthetic cohort and is the table's intuitive heart, showing directly the proportion of a birth cohort expected to reach each age under the estimated mortality regime.
Formula
l_{x+n} = l_x (1 - {}_nq_x)Person-years lived in each interval are accumulated from the survivors, and summing them above any age and dividing by survivors at that age yields expectation of remaining life. Life expectancy at birth, the headline summary, is highly sensitive to infant and child mortality; small errors in early-age estimates swing it substantially, which is why historical e-zero figures must be reported with their uncertainty.
Formula
e_x = \frac{\sum_{y \geq x} {}_nL_y}{l_x}Where data are sparse or implausible, the observed schedule is matched to a family of model life tables—Coale-Demeny, Ledermann, or Princeton regional patterns—by finding the mortality level minimising the discrepancy with the data. The fitted model smooths noise, supplies missing age bands, and lends the table the regular age pattern of mortality observed across many real populations, at the cost of imposing that borrowed shape.
Formula
\min_{\text{level}} \sum_x (q_x^{obs} - q_x^{model})^2Common pitfalls
- Treating the dead's age distribution as the living population's, ignoring growth and structure
- Mishandling the infant separation factor, badly biasing life expectancy at birth
- Forcing a model life-table family whose age pattern contradicts the evidence
- Reporting point estimates of e-zero without acknowledging their wide uncertainty
Frequently asked questions
Period or cohort table—which should I build?
A period table summarises mortality prevailing across all ages in one time span, as if a cohort experienced today's rates at every age; it answers what conditions were like in that era. A cohort table follows one birth-year group through its actual life. Historical data more often support period tables; cohort tables require tracking a generation across many decades of records.
Why rely on model life tables?
Historical age-at-death data are noisy, gap-ridden and distorted by age heaping and infant under-recording. Model life tables encode the regular ways mortality varies with age, distilled from hundreds of real populations, indexed by a level parameter. Fitting the data to the nearest model smooths irregularities and supplies missing age bands, producing a plausible complete schedule—though at the cost of imposing the model's age pattern.
Why is infant mortality so important?
Life expectancy at birth is dominated by deaths in the first year, when mortality is highest and deaths cluster in the earliest days. If infant deaths are under-recorded or the separation factor is wrong, e-zero is badly biased. Historical infant under-registration is pervasive, so credible tables must correct for it and report the resulting uncertainty in early-age mortality.
Related methods
Sources
Wrigley, E. A., & Schofield, R. S. (1981). The Population History of England 1541-1871: A Reconstruction. Edward Arnold / Harvard University Press.
ISBN: 9780674690073Copied.Bengtsson, T., Campbell, C., & Lee, J. Z. (2004). Life under Pressure: Mortality and Living Standards in Europe and Asia, 1700-1900. MIT Press.
ISBN: 9780262025515Copied.