Selection and mortality differentials.

The Office of Population Censuses and Surveys Longitudinal Study provides reliable mortality data by a much wider range of characteristics than are available for other national sources. Although it is based on only a 1% sample of the population, it broadens the scope of mortality analysis and permits study of changes in relationships using different aspects of the time dimension. Data from this study have made us increasingly aware of the importance of selection to the interpretation and understanding of observed mortality differentials. Here we focus on that aspect of selection called "health-related mobility," which is associated with the relative health of people acquiring or losing individual characteristics. It is suggested that, for characteristics affected by health-related mobility, mortality differentials would narrow or widen with increased duration of follow-up. One of the basis of this argument the contribution of health-related mobility to mortality differentials by economic position and social class, to regional differentials, and to family and household differentials is investigated. Selection can thus be shown to operate when people change economic position, when they migrate, or when they change marital status. While the effects of these selection processes can be shown to contribute to social class gradients they do not explain regional differentials and contribute only to a limited degree to differentials by marital status. Differentials by household circumstances also reflect the product of selection processes.