Getting Better All the Time? Selective Mortality, Attrition, and Compositional Changes in Longitudinal Studies on Ageing

By default, life-course studies on ageing include selection processes. Selective mortality changes the composition of samples as people age. In addition, in longitudinal studies, this selection process may be compounded by selective attrition. Consequently, the distribution of important social attributes, such as education and social class, is likely to change over time in any given study cohort.

The aim of this study is to explore changes in the distribution of baseline sample characteristics due to i) selective mortality and ii) attrition in an ageing panel sample (born 1914–1934) throughout different follow-up waves (1968, 1974, 1981, 1991/1992, 2000/2002, 2011). The study is based on nationally representative longitudinal survey data from Sweden (Swedish level of living survey – LNU and The Swedish panel study of living conditions of the oldest old – SWEOLD).

After 43 years of follow-up, 67% of the total sample (N=2335) had died, and an additional 16% had been excluded because of non-response in at least one follow-up wave. The proportion women in the sample increased from 50% to 60% because of selective mortality. The mean level of education increased among both men and women. Similarly, changes in class structure were observed for both sexes. In particular, the proportion of manual workers decreased and the proportion of non-manual workers in the sample increased over time. In both women and men, the changes in class structure were compounded by attrition. In men but not women, selective attrition also compounded changes in education. Finally, the proportion of the sample that experienced poor childhood conditions, financial hardship, or had received social benefits at baseline was successively reduced during follow-up.

The impact of selection processes throughout the life course is commonly discussed in ageing research. However, few studies have explored how these processes change the social composition of longitudinal samples over time.