The variables found to be significant correlates of functional dependency suggest some interesting implications. They confirm the strong relationship reported by other researchers between dependency and age, as well as the variation in age-specific rates of dependency between men and women, and whites and nonwhites. Explication of the underlying determinants of these variations are beyond the scope of this paper but reconfirming their importance suggests the need for policies and research agendas sensitive to these relationships and variations. Of particular importance is the quadratic relationship between age and dependency, meaning that with each passing five year interval rates of dependency increase at an increasing rate--a sobering prospect given the rapid expansion of the oldest old population.
Introduction of a contextual variable into the multivariate regression model may be unique in this analysis but appears overdue. The results here, which are consistent with other researchers' work, suggest that just as poverty is a strong correlate of many unwanted problems in youth and adulthood, so, too, its sequela are present in old age, manifesting themselves as higher dependency rates. Poverty rates among the elderly are known to correlate with a number of important health care system variables including the nursing home bed supply and use rates, Medicaid generosity, and the poor population's life styles, educational levels and occupational experiences.
The estimates produced here are likely to be most useful as initial building blocks for estimating long-term care service demand. A major barrier to cost-effective home and community care has been poor estimates of the rates of enrollment in such programs. Often, the result has been lower-than-expected attendance and, consequently, higher unit costs associated with operating below capacity. While functional dependency estimates at the small area level will not translate directly to demand for service, previous research has shown that utilization of health care services is closely related to need (Andersen et al. 1983; Hulka and Wheat 1985). They may also enhance understanding of some of the variation in the supply of long-term institutional care settings from region to region, state to state, and county to county. While many of the determinants of variation in both demand and supply are likely to defy measurement, either because they are stochastic (e.g. disease onset) or they are difficult to measure (e.g. political preferences of legislators and regulators in the case of supply), "need" estimates provide a useful starting point for planning.
Finally, it should be noted that while the data support the use of these equations to produce estimates of functional dependency among the noninstitutionalized elderly population, the quality of the small area estimates produced by them still needs to be evaluated in future research.