Lifespan Inequality
- Introduced by
- Lifespan-variation literature; life disparity formalized by Vaupel & Canudas-Romo
- Year
- 2003
- Type
- Measures of variability in the age-at-death distribution
- Family
- Lifespan variation
What does it do?
Lifespan inequality measures how unequally length of life is distributed within a population — the spread of the life-table ages at death, not just their average.
Two populations can share the same life expectancy yet differ sharply in how predictable death is: in one nearly everyone reaches old age, in the other deaths are scattered across all ages. A family of measures — life disparity (e†), the standard deviation of age at death, the life-table Gini coefficient, and Keyfitz entropy — quantifies this dispersion, complementing life expectancy with a measure of how fairly survival is shared.
The intuition behind the method
Life expectancy is just the mean of the age-at-death distribution, and like any mean it hides the spread. Lifespan inequality looks at that spread. Picture two life tables with identical average length of life: in a low-inequality population the ages at death cluster tightly around old age (a near-rectangular survival curve), so an individual's lifespan is highly predictable; in a high-inequality population deaths occur at all ages from infancy to extreme old age, so lifespan is a lottery. Lifespan inequality measures put a number on this. Most measures fall as mortality is reduced at young ages — saving an infant adds many years and pulls the death distribution toward older, more uniform ages — so historically lifespan inequality declined hand in hand with rising life expectancy.
History
Interest in the variability of the age at death goes back to the early life-table tradition, but it was long overshadowed by the focus on life expectancy. Late-20th-century work, notably by James Vaupel and colleagues at the Max Planck Institute for Demographic Research, revived it by formalizing life disparity e† and by demonstrating, in the 2003 'bouquet of formulas' honoring Nathan Keyfitz, how change in life expectancy decomposes into mortality improvement and changes in its age pattern. This linked lifespan inequality to Keyfitz entropy and established it as a standard complement to life expectancy in studies of mortality and health equity.
When to use it
Use lifespan-inequality measures whenever you report life expectancy and want to characterize how equally that longevity is shared, or to study whether mortality improvements have made age at death more or less predictable. They are central to research on health inequality, the compression of mortality, and the relationship between longevity and equality. Assumptions: a complete, accurate life table and a clear choice between absolute (years) and relative (scale-free) measures. Do NOT report a single inequality measure as definitive — different measures (e†, standard deviation, Gini, entropy) can rank populations differently, especially absolute versus relative ones; and do NOT confuse lifespan inequality, which is variation within a single life table, with socioeconomic mortality differences between groups, which require between-group decomposition.
Limitations and considerations
- Different inequality measures can rank the same populations differently, so conclusions can hinge on the chosen index.
- Absolute and relative measures can move in opposite directions, complicating interpretation of trends.
- Measures are sensitive to how the open-ended oldest age interval and infant mortality are handled in the life table.
- They describe variation within a single life table and do not by themselves capture inequalities between social groups.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Adds a second dimension — equality of survival — to the conventional focus on average length of life.
- Reveals that populations with equal life expectancy can differ markedly in the predictability of lifespan.
- Most measures share a threshold age below which saving lives reduces inequality and raises life expectancy together, giving a unified policy lens.
- Computable directly from standard life-table columns, with a coherent family of absolute and relative variants.
Applications
- Reporting lifespan inequality alongside life expectancy to assess the equality of survival across countries.
- Studying the historical compression of mortality as deaths concentrated at older ages.
- Linking longevity gains to reductions in lifespan inequality through the threshold-age framework.
- Comparing the predictability of lifespan across socioeconomic groups within a population.
- Decomposing changes in life expectancy into level and dispersion components for mortality analysis.
How it works
How it works
Begin with the life-table distribution of deaths dₓ, the number of the synthetic cohort dying in each age interval. This distribution sums to the radix and is the object whose dispersion every lifespan-inequality measure summarizes; its mean is life expectancy at birth.
Formula
d_x = l_x - l_{x+n},\qquad \sum_x d_x = l_0Life disparity e† is the average remaining life expectancy at the ages where deaths occur: each death is weighted by how many further years that person could have expected to live. A death in infancy contributes a large remaining expectancy, a death in old age a small one, so e† captures the total potential life lost and is large when many people die young.
Formula
e^{\dagger} = \frac{1}{l_0}\sum_x d_x\, e_xThe standard deviation of age at death measures dispersion in the familiar root-mean-square sense, weighting squared deviations of age at death from life expectancy by the death distribution. It and related measures such as the interquartile range describe the absolute width of the lifespan distribution in years.
Formula
\sigma = \sqrt{\frac{1}{l_0}\sum_x d_x\,(x + a_x - e_0)^2}Dividing absolute measures by life expectancy gives relative, scale-free indices: Keyfitz entropy H = e†/e₀ and the coefficient of variation CV = σ/e₀. Relative measures compare how unequal lifespans are net of differences in average length, while absolute measures answer how many years of spread there are; the two can move in opposite directions, so both are reported.
Formula
H = \frac{e^{\dagger}}{e_0},\qquad CV = \frac{\sigma}{e_0}Common pitfalls
- Reporting only one measure and treating it as the definitive picture when others might rank populations differently.
- Conflating absolute lifespan variation (in years) with relative inequality, which behave differently over time.
- Mishandling the open age interval or infant deaths, which strongly influence dispersion measures.
- Mistaking within-population lifespan inequality for between-group mortality disparities, which require different methods.
Frequently asked questions
What is life disparity (e†)?
Life disparity is the average number of remaining life-years lost at the ages where deaths occur — each death weighted by the remaining life expectancy at that age. It is large when many people die young (losing many potential years) and small when deaths cluster at old ages. It is the absolute lifespan-inequality measure most directly tied to Keyfitz entropy, since entropy equals e†/e₀.
Can two populations have the same life expectancy but different lifespan inequality?
Yes, and this is the whole point of the measure. Life expectancy is the mean of the age-at-death distribution and says nothing about its spread. One population may have nearly everyone dying around the same old age (low inequality) while another with the identical mean has deaths scattered from infancy to extreme age (high inequality).
What is the threshold age?
The threshold age is the age below which reducing mortality lowers lifespan inequality and above which reducing mortality raises it. Saving lives at young ages pulls deaths toward older, more uniform ages and compresses the distribution, whereas postponing deaths that already occur at old ages spreads the distribution out. The threshold typically lies in middle to older adulthood and unifies the level and inequality effects of mortality change.
Related methods
Sources
Vaupel, J. W., & Canudas-Romo, V. (2003). Decomposing change in life expectancy: A bouquet of formulas in honor of Nathan Keyfitz's 90th birthday. Demography, 40(2), 201–216.
Open DOI referenceCopied.Preston, S. H., Heuveline, P., & Guillot, M. (2001). Demography: Measuring and Modeling Population Processes. Blackwell.
ISBN: 9781557864512Copied.