Research note: revisiting the classification of household composition among elderly people.

The note updates comments written a decade and a half ago on our need for a standard schemata for comparing the living arrangements of elderly people around the world. Perforce limiting our view to populations living in private households, the note considers different schema beginning with an examination of household size and head/nonhead data and proceeding to schemes using ever more detailed data on "relation to household head" information and data on marital status. The best scheme will of course depend on the data available and on the study's theoretical ideas, but if the ideas have to do with the nature of the family then it would seem that one is best off using a combination of information on people's relation to the household head and information on marital status to indicate family household composition. The six-category scheme (alone, spouse only, with married child[ren], with unmarried child[ren], with other relatives, with non-kin) might best be collapsed into a bivariate or trivariate categorical dependent variable for purposes of multivariate regression.