Skip to content
Back to search results

Pollard Decomposition

Introduced by
John H. Pollard
Year
1982
Type
Age-specific decomposition of a difference in life expectancy
Family
Standardization & decomposition
Font size

What does it do?

Pollard's decomposition breaks a difference in life expectancy between two populations into additive contributions from each age, showing exactly how much of the gap is due to mortality differences at infancy, in midlife, or in old age.

John Pollard derived a continuous-age formula expressing the life-expectancy difference as an integral of the age-specific mortality-rate difference weighted by life-table functions. Because the contributions sum exactly to the total gap and can be further split by cause of death, the method is a standard tool for explaining why one population outlives another.

The intuition behind the method

When population A lives longer than population B, the natural question is where in the life course that advantage comes from — is it fewer infant deaths, fewer midlife heart attacks, or better survival in old age? Pollard's insight is that the total gap in life expectancy can be unbundled age by age. At each age he compares the two populations' death rates; the difference is then weighted by how much that age matters for life expectancy (how many people reach it and how many years of life ride on surviving it). Adding these weighted age-specific differences across the whole life span reconstructs the total gap exactly, so you can point to the ages — and, with cause-of-death data, the causes — responsible for the longevity difference.

History

John Pollard, an Australian actuary and demographer, published his continuous formulation in 1982 in the Journal of the Institute of Actuaries, deriving the relationship between the expectation of life and the underlying force of mortality and showing how a difference in expectancy decomposes by age. His approach paralleled and complemented the discrete stepwise method developed by Eduardo Arriaga in the same era, and the connections among decomposition methods were later unified by Vaupel and Canudas-Romo. Pollard's decomposition became a standard tool in comparative mortality analysis and remains widely used for sex, regional, and cause-of-death gaps in life expectancy.

When to use it

Use Pollard's decomposition when you have two life tables and want to explain a difference in life expectancy — between countries, sexes, periods, or social groups — in terms of the ages (and, with cause data, the causes) driving it. It is the appropriate continuous-age complement to Arriaga's discrete stepwise decomposition. Assumptions: accurate age-specific mortality for both populations, and that decomposing into a smooth age profile of mortality differences is the goal. Do NOT use it to attribute the gap to changes in population age structure — it decomposes the mortality (life-table) difference only; and be aware that, like all such methods, the exact age contributions depend slightly on the chosen weighting convention, so report which variant was used.

Limitations and considerations

  • It decomposes only the mortality component and does not address differences arising from population age structure.
  • The precise age contributions depend on the weighting convention chosen, so different variants give slightly different splits.
  • Cause-specific decomposition requires reliable cause-of-death coding, which is weak in many settings.
  • Like other one-dimensional decompositions, it describes rather than explains causes, offering no causal mechanism.

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

Strengths

  • Decomposes a life-expectancy difference into age-specific contributions that sum exactly to the total gap.
  • Extends naturally to a simultaneous decomposition by cause of death, attributing the gap to ages and causes together.
  • Provides a continuous-age formulation with a clear analytic weight function, complementing discrete stepwise methods.
  • Widely applicable to differences between countries, sexes, periods, or socioeconomic groups using only two life tables.

Applications

  • Explaining the male-female gap in life expectancy by age and cause of death.
  • Attributing cross-national differences in longevity to specific ages and causes.
  • Decomposing changes in life expectancy over time into the ages where mortality improved or worsened.
  • Quantifying which ages drive socioeconomic differences in life expectancy.
  • Identifying the age-specific mortality components behind regional health inequalities.

How it works

How it works

1

The difference in life expectancy at birth between two populations is written as an integral over age of the difference in their age-specific forces of mortality, μ₁(x) − μ₂(x), multiplied by a weight w(x). This recasts a single summary gap as an accumulation of age-specific mortality differences.

Formula
e_0^{(2)} - e_0^{(1)} = \int_0^{\infty} w(x)\,\bigl[\mu_1(x) - \mu_2(x)\bigr]\, dx
2

The weight w(x) combines life-table functions from the two populations — typically survivorship in one population times remaining life expectancy in the other, or a symmetric average across the two — so that a mortality difference at age x is valued by how many people are exposed at that age and how many future years depend on it. This weighting is what makes the age contributions sum to the true gap.

Formula
w(x) = \ell_1(x)\, e_2(x)\ \text{(or a symmetric average of the two life tables)}
3

Multiplying the mortality-rate difference by the weight at each age gives that age's contribution Δₓ to the life-expectancy gap. Positive contributions are ages where the longer-lived population has lower mortality; negative contributions are ages where it has higher mortality. By construction these contributions sum across all ages to the total difference in life expectancy.

Formula
\Delta_x = w(x)\,\bigl[\mu_1(x) - \mu_2(x)\bigr],\qquad \sum_x \Delta_x = e_0^{(2)} - e_0^{(1)}
4

Because the decomposition is exactly additive, the age contributions can be aggregated into broad age groups, and when mortality is split by cause the contribution at each age further partitions into cause-specific pieces that also sum to the total. This lets analysts attribute the life-expectancy gap simultaneously to ages and to causes of death.

Formula
e_0^{(2)} - e_0^{(1)} = \sum_{\text{age groups}} \Delta_x = \sum_{\text{causes}} \Delta^{c}

Common pitfalls

  • Comparing age contributions computed under different weighting conventions as if they were identical.
  • Interpreting the decomposition as including a composition (age-structure) effect, which it does not contain.
  • Over-trusting cause-specific contributions where cause-of-death attribution is incomplete or inconsistent.
  • Aggregating age contributions into groups without preserving the exact additivity, leading to apparent residuals.

Frequently asked questions

How does Pollard's decomposition differ from Arriaga's?

Both attribute a life-expectancy difference to ages, but Pollard's is a continuous-age, integral formulation with an explicit analytic weight function, while Arriaga's is a discrete stepwise method that separates direct, indirect, and interaction effects across life-table age intervals. The two give very similar results and are often used interchangeably; the choice is largely one of formulation and convenience.

Can Pollard's method decompose by cause of death?

Yes. Because each age contribution is proportional to the difference in the total force of mortality at that age, and the force of mortality is the sum of cause-specific forces, the age contribution splits additively into cause-specific pieces. This yields a simultaneous decomposition of the life-expectancy gap by age and by cause, with all pieces summing to the total.

Why does the choice of weighting matter?

The weight function combines life-table functions from the two populations, and there is more than one symmetric way to do this (for instance which population's survivorship and which one's remaining expectancy to use). Different conventions distribute the gap slightly differently across ages, though they all reproduce the same total. Reporting the convention ensures comparability across studies.

Related methods

Sources

  1. Pollard, J. H. (1982). The expectation of life and its relationship to mortality. Journal of the Institute of Actuaries, 109(2), 225–240.

    Open DOI reference
    Copied.
  2. Preston, S. H., Heuveline, P., & Guillot, M. (2001). Demography: Measuring and Modeling Population Processes. Blackwell.

    ISBN: 9781557864512
    Copied.