Sequencing the real time of the elderly: Evidence from South Africa

Abstract

BACKGROUNDUnderstanding how the elderly in developing countries spend their time has received little attention. Moreover, the potential of time use data to discern variation in activity patterns has not been fully realized by methods which use a mean added time approach. OBJECTIVESTo uncover patterns of time use among the elderly (60 years and older) in South Africa by applying an innovative methodology that incorporates the timing, duration, and frequency of activities in the analysis. METHODSWe use sequence analysis, which treats the daily series of activities of each individual as a sequence, and cluster analysis, to group these sequences into common clusters of time use behaviour. We then estimate multinomial logit regressions to identify the characteristics of the elderly which predict cluster membership. RESULTSWe find that the time use behaviour of the elderly in South Africa can be divided into five distinct clusters, according to the relative importance in their day of personal care, household maintenance, work, mass media, and social or cultural activities. In comparison to men, women are overrepresented in the cluster where household work dominates, while they are underrepresented in the cluster of the elderly who engage in production work. A range of other individual and household characteristics are also important in predicting cluster membership.

Gender Divisions in the Real Time of the Elderly in South Africa

“…3. In a previous study, we identified clusters of time use behavior among the pooled sample of the elderly and we then compared the membership of elderly women and men in these clusters (Grapsa and Posel 2016). In this study, as we explain in the later text, we estimate the clusters separately by gender.…”

“…Of these, only 21,188 (or 8.7%) include multiple activities, the majority of which (66.7%) are sequential. The inclusion of these multiple activities makes little difference to the mean time allocations of the elderly across activity types, or to the gender differences in these allocations (Grapsa and Posel 2016), and given the computational requirements of our empirical methods, we focus the analysis on the first activity reported per time slot.…”

“…OM provides a measure of the similarity (or distance) between sequences, by calculating the minimum number of operations that would be needed to transform one sequence into another. These operations include the replacement or insertion/deletion of sequence states (see also Grapsa and Posel 2016).…”

“…Each cluster is represented by its mean or centroid, and the distance between clusters is measured by the change in the error sum of squares 5 (SSE) that occurs after merging two clusters. Ward's method seeks to minimize the sum of squared distances of points from their cluster centers (see also Grapsa and Posel 2016).…”

“…Most of the elderly in South Africa do not live alone and many live in extended or multi-generational households (Case and Deaton 1998;Grapsa and Posel 2016). With persistently high unemployment rates and high rates of adult mortality related to HIV/AIDS, the receipt of the social pension has meant that the elderly are often the primary providers of support in their household, rather than economic dependents.…”