Revisiting within-cohort compositional change to understand mortality inequalities

The distribution of social characteristics changes over the life course of cohorts. These cohort compositional changes vary in magnitude across time, across populations and across socioeconomic groups. While this is evident intuitively, it is only rarely made explicit in demographic studies of mortality. In this debate piece, we argue that the classic demographic study of compositional change has become increasingly neglected in the field, as we have shifted towards causal inference of the individual-level determinants of demographic change. Using examples from the USA and Finland, we demonstrate how within-cohort compositional change operates for social characteristics such as education and divorce – a change driven by background and socially selective mortality, migration and changes in the characteristics themselves as the cohorts age. These cohort compositional changes produce non-linear age patterns of difference in the distribution of these social characteristics between Finland and the USA, and across socioeconomic groups, and may thus pose a challenge for the analyses of differential mortality. Ultimately, to understand aggregate, not individual, inequalities in mortality we need to more explicitly investigate how these covariates of mortality are changing in size and importance, and how they interact with indicators of social position over the life of the cohorts.