LAUSR

Individuals with low socioeconomic status have, on average, a shorter lifespan than individuals with high socioeconomic status. These inequalities are found across the globe; they are large; and they are persistent over time. Socioeconomic inequalities in life expectancy are a key challenge of modern societies. They are both a cause and a consequence of unequal opportunities, and they have farreaching consequences at both the individual and the societal level. For instance, they are a serious concern for the equitable design of pension policies, as individuals with high socioeconomic status spend more time in retirement and receive pensions for a longer time. Inequalities in life expectancy by education have long been a focus of epidemiological and demographic research, as education is comparatively straightforward to measure, it is fixed from early adulthood and precedes other markers of attained socioeconomic status such as income or wealth, and it is a strong predictor of the overall socioeconomic status. Trends in educational inequalities in mortality are heterogeneous across countries. The gap between the lowest and highest educated has increased in some countries and decreased in others. The paper by ZazuetaBorboa et al provides an important contribution to our understanding of these changing inequalities. Using highquality register data from three European countries (England/Wales, Finland and Italy represented with data from Turin), it assesses longterm trends in the gap in remaining life expectancy at age 30 between the high and low educated from 1971/1972 to 2017/2019. A key finding is that all countries experienced both narrowing and widening gaps in life expectancy across educational groups, or what they refer to as ‘reversals of inequalities’. For Finnish men inequality first increased over time, but recently started to decrease; and for English men the opposite holds. Using decomposition techniques, the authors show these reversals have largely been driven by the mortality dynamics of the low educated aged 30–54. This accords with research arguing that differences in the institutional, economic and environmental contexts matter most for disadvantaged groups. Monitoring inequalities in life expectancy is important, and it comes with challenges. For many of these challenges, there is no easy solution. We argue that researchers need to be aware of them when studying, and reading about, educational inequalities in mortality. We outline some of these challenges below. Most of them are not restricted to educational inequalities in mortality, but apply more broadly to socioeconomic inequalities. Specifically, we see three main challenges: measures, mapping and meaning. These challenges relate to three questions. How do we measure inequalities? How do results map to policy measures? And what meaning do the results have in a broader context of mortality change? First, there are many measures available to assess mortality inequalities, and our assessment might depend on the measure we use. ZazuetaBorboa et al used the range in life expectancy, which ignores the impact of intermediate groups on the gradient. Measures also vary in their response to mortality change at different ages, as well as in their substantive interpretation. Life expectancy is equal to the average length of lifespans in a life table population. In contrast, measures of lifespan variation move away from the average and aim to capture the variability in lifespans. This includes, for instance, the variance of lifespans. The variance of lifespans can be decomposed into the variance between educational groups and the variance within educational groups. Empirically, the variance between educational groups is much smaller than the variance within groups, and education explains only little of the overall variance, usually less than 10%. While low explanatory power is not uncommon for socioeconomic variables, it puts educational inequalities into perspective, and there are many other factors determining the length of lifespans. Recent methodological developments try to reconcile inconsistent findings from different measures with new approaches which take into account both the overall variability of lifespans and the relative performance of different socioeconomic groups. Second, is the question of how measures are mapping to policy recommendations and interventions. Measures based on life tables—including life expectancy and measures of lifespan variation—do not use the actual age structure of populations, but the artificial age structure implied by the life table. This has the benefit of removing the impact of the age structure from group comparisons, making these comparisons easier to interpret. However, policy interventions are not applied to artificial but to real populations. For instance, consider two groups which have the same agespecific mortality rates, but one group is very young and the other group is very old. When comparing life tables, both groups will appear to be equal. However, if mortality rates increase with age, the older group will have more deaths. When the older group is also the more disadvantaged (as in the case of the low educated) reducing inequalities at older compared with younger ages would have a larger absolute impact on the total inequalities, even if relative inequalities in rates are smaller at these ages. Finally, the changing magnitude of mortality inequality only has meaning in a broader historical context of intertwined social, economic and political developments. What do inequality trend reversals really mean? By analysing these reversals, ZazuetaBorboa et al implicitly suggest that inequalities in life expectancy should remain stable. Why should we expect this? Life expectancy regularly converges and diverges between populations; diseases rise and fall. Inequalities widen when medical innovation is first adopted by the most advantaged, and narrow as technologies become widely available. New healthy and unhealthy behaviours also spread through social diffusion processes. The myriad of pathways through which socioeconomic status acts as a fundamental cause of disease cause both continuities and change in mortality gradients. These more distal longrunning patterns occur simultaneously, and on top of the more proximate policy changes singled out by ZazuetaBorboa et al in the discussion. While it is certainly plausible that changing alcohol patterns (Finland) and austerity measures (UK) are driving some of these MaxPlanckInstitute for Demographic Research, Rostock, Germany Federal Institute for Population Research, Wiesbaden, Germany Max Planck – University of Helsinki Center for Social Inequalities in Population Health, Helsinki, Finland