Skip to main content

Lifespan Inequalities: Why the age-at-death distribution varies between countries and socioeconomic groups

Periodic Reporting for period 2 - LIFEINEQ (Lifespan Inequalities: Why the age-at-death distribution varies between countries and socioeconomic groups)

Reporting period: 2019-01-01 to 2020-06-30

Statistical agencies and health researchers require summary metrics of mortality to make sense of the fine-grained information collected at the individual level. These metrics serve four key roles: to set health targets, to compare populations, to uncover emerging threats and to evaluate policy outcomes.
Life expectancy is the most commonly used metric of mortality. It represents the hypothetical average age at death in a population for a given year if death rates were to remain unchanged.

Overlooked in this approach are inequalities in mortality within groups, which are both substantial and changing over time. Lifespan inequality (also known as age-at-death variation) is a metric of mortality difference in age at death between individuals. At the population level, lifespan variation indicates the heterogeneity in population health. This heterogeneity is important to consider when designing health and welfare policies including equitable pension schemes. At the individual level, lifespan variation measures the uncertainty in the timing of death. Economic models have shown that individuals are highly risk averse when it comes to survival, and would prefer to live in a society with a lower life expectancy if they could increase the certainty that they would survive to such an age. While life expectancy captures the magnitude of survival improvement, lifespan inequality is its complement, capturing the equality in survival improvement. A full picture of population health requires us to monitor both.

The LIFEINEQ project is the most comprehensive inquiry to date into the development and causes of lifespan inequality. Specifically LIFEINEQ has four main objectives:

1. To track and forecast the relationship between life expectancy and lifespan inequality in national populations,
2. To determine the ages and causes of death that drive outlying age patterns of mortality,
3. To analyze the development of lifespan inequality by socioeconomic groups, and,
4. To assess the impact of individual differences in behaviour on lifespan inequality.

All projects have the potential to uncover novel results with important policy implications. The first objective is descriptive and addresses the degree to which we need to worry about lifespan inequality. The second and third objectives address how populations and socioeconomic groups differ in lifespan inequality--
this comparative perspective allows us to identify best practices in reducing inequalities across populations. The fourth objective identifies the reasons why populations differ in lifespan inequality.
LIFEINEQ has made considerable progress in documenting and understanding the trends and causes of lifespan inequality. The most important findings to date are the following:

- Lifespan inequalities are stagnant or increasing among individuals with lower socioeconomic status, even when life expectancy itself is increasing. This has been uncovered across many parts of Europe and the United States, and seems to hold for different measures of socioeconomic position including education, occupation, income, and area-level deprivation.
- A number of populations experienced sustained periods of increasing lifespan inequality in the past few decades, including the United States, Central and Eastern European countries (CEE), and several Latin American countries (LAC). Typically the causes of these increases related to midlife mortality crises such as the ongoing opioid crisis in the United States, periods of economic crises and hardship in CEE, and periods of increasing homicide mortality in LAC.
- The empirically negative correlation between life expectancy and lifespan inequality is weakening across high income countries, particularly when absolute rather than relative inequality is measured. It does not appear to hold at all for subgroups with low socioeconomic status. The correlation is also weak or non-existent among populations undergoing midlife mortality crises.
- In the United States, many of the determinants of mortality from extrinsic causes of death, for instance influenza, drug overdose, alcohol abuse, HIV/AIDS, Hepatitis C, and suicide show strong patterning by birth cohort, with trailing-edge baby boomers fairing poorly compared to those born before and afterward.
- There is a disconnect between population-level ranking of lifespan inequality depending on the temporal frame examined, i.e. period (year), cohort (birth year), or cross-sectional average (average exposure to ages at death over the lifetime of those present). [Findings under review]
- Using the nationally published data on old age mortality or educational attainment estimated from censuses in less developed country settings leads to highly erroneous estimates of lifespan inequality. [Findings under review]

As of end of February 2020, LIFEINEQ members have published 15 peer-reviewed articles, 2 book or encyclopedia chapters, 3 working papers, and 2 PhD dissertations (which were carried out in part while working on the project). The findings above appeared in leading journals of general science, demography, and public health/epidemiology such as Science, Demography, International Journal of Epidemiology and Social Science and Medicine.

In addition to our scientific output, LIFEINEQ members have organized a 2-week training course for PhD students and postdocs held at the Max Planck Institute for Demographic Research on the methods for calculating and decomposing lifetable metrics including lifespan inequality.
In the first 2.5 years the LIFEINEQ project has produced robust scientific findings that underscore the policy relevance of monitoring lifespan inequality, which is currently not undertaken by any major statistical agency in the world. In addition, the LIFEINEQ project has made substantial methodological contributions that help us to understand changing mortality patterning across the different temporal dimensions of age, period and cohort, as well as deriving the mathematical properties of several indices of lifespan inequality.

To date, most empirical findings have been based on period (calendar year) data, under the assumption that period mortality rates are fixed throughout an individual's lifetime (the classic period assumption). Between now and the end of the project we will accumulate more evidence on trends in lifespan inequality along the birth cohort dimension, by forecasting mortality (completing the age-at-death schedule) up to the 2000 birth cohort. We will also look more closely at lifespan inequalities relating to the retirement lifespan using advanced multistate methodologies.

Finally, we are currently developing an R-package to calculate indices of lifespan inequality, currently available on my GitHub ( and which we expect will be available on CRAN before the project end.
Full press release available here: