The single-person household has become the modal household category for unmarried men and women over the age of 65. The large numbers living alone have been linked to some of the most significant demographic and social changes of the past several decades, such as the increasing "nuclearization" of the modern family, the premium placed on personal privacy, and the continued divergence of men's and women's expectations of life. As a result of these trends, the elderly person living alone had become a common occurrence by 1980, when over half of all unmarried persons aged 65 and over lived alone, compared to less than 40 percent in 1960 (U.S. Bureau of the Census, 1963; 1983). Yet, despite the overwhelming focus in the literature on this status, not all elderly live alone. A significant share are either unable or unwilling to form single-person households, and are found instead in multi-person households, either as household head or as an added member.

The distribution of elderly individuals across households having different compositional characteristics is reflective of the economic, social, and normative environment within which housing choices are made. For example, an individuals standard of living is strongly conditioned by living arrangements, as income depends partially on household size and composition. Thus, analyzing household composition characteristics can help us evaluate the distribution of resources within the elderly population. Furthermore, eligibility for public transfer programs is based in part on household composition and size. As a result, understanding living arrangements helps scholars and policy makers to better evaluate the economic and social well-being of the elderly (Schwartz et al.. 1984). We also suggest that elderly living arrangements may be viewed as behavioral outcomes through which the dynamics of the normative systems relating to the aging process-are displayed. Norms regarding appropriate or expected behavior and preferences of older versus younger or male versus female elders. for example, may be reflected in living arrangements.

The purpose of the present paper is to explore the degree to which demographic characteristics, economic resources, and health determine the choice of living arrangements among the unmarried elderly population of the United States. A model is generated and tested, making use of data from the Survey of Income and Program Participation (U.S: Bureau of the Census, 1987). These data are particularly useful for the present project because, unlike many extant data sources. they include information on a wide variety of relevant predictors of household composition for a large cross-section of the population, thereby permitting a more careful consideration of the process leading to an individuals being located in a particular household type.