This paper uses a new data set to study the choice of living arrangements of some 3000 Massachusetts elderly between 1982 and 1986. The data have a number of unique features; they are longitudinal and combine detailed information on health with information on economic status and family relations. This paper considers the influence on living arrangements of alternative measures of health (subjective versus functional abilities versus diagnosed condition), incomes and marital status of parents, and the number and sexes of children. It also examines the extent to which changes in health and the death of a spouse trigger changes in living arrangements and how rapidly such changes occur. The main findings of the paper are: Functional ability indices are very good predictors of living arrangements. Subjective health reports are poor predictors of living arrangements. The probability of institutionalization declines rapidly with the income of the elderly. In the cases of the older old daughters are much more likely than sons to share living quarters. Living arrangements are fairly stable. When changes in living arrangements occur they are often triggered by changes in health status or the death of a spouse. When deterioration in health status or the death of a spouse leads to a change in living arrangements, such changes typically occur within a year of the triggering event.