Housing Demand for the Elderly

Housing Demand for the Elderly


by Robert H. Edelstein and Allan J. Lacayo


We explain elderly housing tenure choices using personal and household location, demographic, sociological, and economic factors: models including some but not all of these factors do not explain tenure choices as well. Our preliminary analysis of variance and nested logit estimates reveal a preference for independent living over at home care, and for the latter over group care choices. These estimates also identify significant age and gender differences in tenure choice. Our study represents a fruitful exercise for policy analysis and real estate market participants because it outlines factors that determine the demand for desired present and future elderly housing services.


Elderly Americans constitute the fastest growing sector of our population between now and the year 2020. For this reason it is important to study the behavior and motives of elderly Americans so as to understand their pattern of needs and preferences in order to identify adequate economic and policy strategies.

In this paper we specify a model of housing tenure choice specific to the elderly and the housing choices available to them. What makes this research different from traditional housing tenure choice models is that the nature of housing services demanded by the elderly include specialized nontraditional services related to “needs” or “wants” not commonly observed in younger age cohorts of the population at large, but more frequently observed in older age cohorts of our population. “Needs” for nontraditional housing services are related to health status or lifestyle changes that occur as individuals age, whereas “wants” for nontraditional housing services are rooted in the cultural, social and economic issues that relate exclusively to life style changes (e.g., retirement), also highly correlated with advancing age.

In specifying the model used in this paper great care is used in classifying housing choices in terms of the extent to which individuals in such housing arrangements need care, want nontraditional housing services, or choose to live (rent or own) independently (alone, with a spouse or relative).

To estimate the model specification employed in this paper we use Public Use Micro-data Samples (PUMS)[1] from the 1990 U.S. Bureau of the Census. These samples yield sufficiently large samples over all relevant age cohorts, geographic regions, and levels of urbanization. They also provide us with samples representative of the elderly population at large in terms of the population variation in marital and family status of the Census respondents (single, married, married but separated, divorced, widowed, living alone, living with children, children living with elder, etc.), age and gender of the respondents, their health and disability status, as well as their economic characteristics.

I.ATaxonomy of the Housing Choice

We set up the available housing choices by classifying respondents according to the source of our sample data: the 1990 Bureau of the Census public use micro-data samples (PUMS) survey. The housing choice alternatives included in the PUMS survey are:

  • Nursing Home (NH): A group quarters institution or facility housing elderly with physical and/or mental limitations. Ownership of the facility may be public (Fed, State, County, City) or private (for-profit, non-profit), and the set of services that residents have to their avail vary significantly from home to home.
  • Assisted Living (AL): People choose Assisted Living arrangements when they have personal care physical limitations or mobility limitations that do not enable them to perform activities for daily living, such as minimal hygiene and personal care functions, without assistance.
  • Congregate Care / Continuing Care Retirement Communities (CC/CCRC): Are group facilities where amenities such as meals, transportation, security, medical and other services are provided to the tenants. The individual households in the community may have kitchenettes in their units and be self-sufficient.
  • “Stay at Home” Care Services (SHCS): people requiring Assisted Living services but do not abandon their place of residence (most of these people own their residences, but some are renters, or live with relatives or friends).
  • Independent Living, Rent (NCRO): Are renters who live on their own, who reportedly purchase no special care services of the aforementioned kind.
  • Independent Living, Own (NCOO): Are elderly individuals who own a home, whose spouses own a home, or who live with a younger relative who owns a home. The home may be single or multifamily, attached or detached. In addition, the elderly household members do not report the purchase of special services as listed above.

We translate the observed housing choices as a two step process represented in figure 1. First, we identify whether the respondent wants or needs specialized care services particular to the elderly. If we observe a need or desire for elderly housing services, we observe establish whether such care is needed critically and chronically (NH), chronically (AL), wanted more than needed (CC/CCRC), or whether the services are received by the elderly in their homes (SHCS). We also account for people who neither need nor want any of the care services that conform elderly housing services, identifying independent living home owners and renters[2].

There are sampled observations for which it is difficult to distinguish whether the observed housing choice is at-home care services, congregate care/CCRC, or Assisted Living facilities. This is because survey responses do not always clarify whether the source of elderly housing services (meals, security, common areas, etc.) coincides with the owner of the housing unit itself. We opted to censor observations that we could not assign to any of the six categories of housing choices with certainty. This censoring biases our sampling because we eliminate observations from AL, CC/CCRC and SHCS in greater proportion than from NH, or the two independent living alternatives (Rent, Own).

Another criticism that may be levied against our tenure choice space is our employment of the intensity of the need or want for critical care services and for services with activities of daily living (ADL) as the principal classification criteria. An alternative arrangement of the housing choices that the elderly can select from is described in Schwartz, Danziger and Smolensky (1984). Their paper distinguishes housing choices by the number of people living with the respondent in the sample: alone versus living with spouse, or relatives or friends. OF course, there are major differences in their study and this one: first, their data source is not P.U.M.S. but Retirement History Surveys conducted in the late 60’s through the 70’s. In addition, their focus was not to evaluate the relationship between income an health and how this relationship affected the observed living arrangement. They could not estimate the demand for housing because the Retirement History Surveys do not contain information about elderly living in group quarters or Nursing homes. Thus, even though such a classification of housing choices as in Schwartz et. al. (1984) can be used with P.U.M.S., it would not enable us to examine in detail the underlying factors that motivate elderly individuals to select living arrangements in group quarters or institutions.

Thus, Figure 1 describes the multinomial logit model specification needed to identify the determinants of housing demand by the elderly. In estimating the logit specifications of this paper we are able to examine some interesting claims about observed choices, related to gender differentials in housing choices and the emergence of at home care services as an important tenure choice.

It is commonly asserted that the proportion of elderly females living independently, is significantly lower than that of elderly males, females more readily select institutional housing arrangements that include assistance in ADL, including at-home care services. This hypothesis implies that there are factors that make mobility tendencies different for elderly males and females. Males are more inclined to live alone or with relatives rather than move to congregate or institutional facilities, whereas females favor independent living arrangements in households that include family, or group care situations[3]. Our estimation confirms this hypothesis for the older elderly (85+) but finds it more difficult to not reject this assertion in the young elderly group (65-74).[4]

We also use a specification test to verify the claim that the “stay at home” care options provide the main competition to traditional housing services for the elderly –as opposed to no-service independent living options. preliminary test results indicate that there is a definite trade-off between at-home care services and independent non-institutional living which may indicate support for the hypothesis that these services are selling based on the opportunity cost of providing care services for elderly family members to younger ones (children of the elderly), enabling the children to focus on the relationship with their elderly relatives as opposed to spending their time in helping with ADLs. The test result for this hypothesis is obtained directly from the specification test for the SHCS alternative in our logit demand specification.

II.Literature Review

Some commonly help perceptions about housing arrangements selected by the elderly are that the preference for {independent living or sharing a home} over that of {living in an institution or group quarters facility} diminishes with functional (physical and mental) limitations and (to a lesser degree) age. Also, such preference intensifies with increasing income and wealth (see Golant, (1992), and Schwartz, Danziger and Smolensky (1984)). In addition, new research shows that females are more likely than males to prefer shared independent living arrangements over living alone or in group quarters (institutionalized living), regardless of marital status (Borsch-Supan, Hajivassiliou, Kotlikoff and Norris (1992))[5].

We take advantage of these results and others from earlier research to formulate logit model specifications that allow us to arrive at a model that explains how elderly Americans choose housing arrangements.

[a lot more editing from here on…]

II.AEconomic Factors

Two main streams of research in the economics of aging can be identified. The first one is interested in examining consumption and savings patterns among the elderly and whether or not they conform to traditional economic theory. The second current of literature is more normative and tends to address impacts or consequences of policy changes in social security, medicare, welfare, etc., by tracking the behavior of the elderly. The purchase of housing services by the elderly is employed in both strands of research because housing represents a large share of the expenditures. While positive research usually attempts to verify the life cycle theory using elderly consumption behavior, the second literature focuses on the economic welfare implications of policy alternatives and their impact on observed patterns of consumption and savings implied by elderly tenure choice, household arrangements (where and with whom they choose to live with), and retirement decision[6].

II.A.1Life-Cycle Issues: Retirement, Work/Leisure, Savings/Wealth

[Jones, Larry, Testing the Central Prediction of Housing Tenure Transition Models (1994)

Theme: Housing Tenure Choice depends centrally on a Household’s liquefiable wealth, controlling for the asset price of housing (transitory and permanent incomes, and measures of wealth and its liquidity), household demographic (age, race) and social characteristics (marital status of the head, and composition of the household).]

[Feinstein, Jonathan and Daniel McFadden, The Dynamics of Housing Demand by the Elderly: Wealth, Cash Flow and Demographic Effects

Using a life cycle model, the authors focus on identifying the main factors in the decision of elderly households to move and to “downsize” conditional on moving. They find that wealthier households are less likely to move, controlling for demographic factors and also find little evidence of housing market imperfections -inefficiencies would catalyze downsizing if income constraints are binding and prevent downsizing from being exercised if capital markets imposed constraints differentially on households with varying wealth, neither case is statistically significant.]

A common framework in which the above issues is typically worked out is the Life Cycle theory of consumption and savings behavior. Life cycle models postulate that individuals dissave in the early and late years of life while accumulating savings in the middle years when the individual can accumulate wealth by saving income from labor and other sources and using the compounded savings to repay their dissaving in their early years and to finance their consumption after they choose to retire.

II.A.2Labor Market Issues: Ability to participate and Participation

Cuba and Longino (1991) Regional Retirement Migration

Hazelrigg and Hardy (1995) Migration to Sunbelt

Hogan and Steinnes (1994) Elderly seasonal migration

Kallan Elderly Migration

- work leisure choice and linkages to migration and mobility (as signals of labor changes)

II.A.3Housing Affordability: Income effects see ref’s for II.A.1

Household composition changes affects household income and consumption and savings decisions of individuals. Housing arrangement selected directly relates to income and its sources. Mobility enters into the analysis because it is an indicator of how income may or may not be sufficient to maintain housing lifestyle.

II.BDemographic Factors

The employed demographic factors in our model include age, gender, mobility and migration tendencies of the respondent.


Age enters non-linearly into our model specification; we use age to segment our logit specification by age groups[7]. Separating our estimates by age cohorts enables other factors to better explain the observed housing choices.

Preliminary tabulations of our data confirms widely accepted notions about the importance of age in determining housing arrangements by the elderly. According to Nursing Home Surveys performed by the National Center for Health Statistics, while only 5% Americans over the age of 65 choose Nursing Homes as a temporary or permanent housing arrangement, the same figure for persons over 84 years of age is 22%, and the proportion for the age cohorts between 65 and 84 is ten to fifteen times smaller (1.3% for ages 65-74 and close to 3% for ages 75-84)[8].

The implication is that an explanation of the factors impacting demand for housing services by the elderly should account for gender and age, or an even more complex mechanism between these two variables and their link to the health status of elderly[9].

In most of the research about elderly housing demand, age enters the models directly as a linear and/or non-linear factor. Beyond tabulating sample proportions by age cohort, not much effort is expended to actually segment demand estimates by age-cohort. Differences in choice that arise by separating samples using age cohorts are accounted for by inferring the meaning of the estimated coefficient of the age variables. Rogers –the main reason is that survey samples employed by research in the literature tend to be small.

II.B.2Mobility [income constrained?] As a signal of changes in the economic status of the individual

II.B.3Migration [geographic and economic motives] As a signal of changes in lifestyle (work v. leisure)

II.CSociological Factors

Two issues are usually addressed in housing demand by the elderly topics: whether elderly choose independent living arrangements or require assistance in any of the ADL. The second issue is the family/household composition of the elderly, i.e., marital status, family composition and household size and composition.

II.C.1Marital Status

Write about the contribution to changing household arrangements due to changes in the marital status of sampled elderly.


Spouse Dying

Earlier separation or divorce

Living alone or in group

II.C.2Household Structure and Composition: Family Structure

Children moved in, or moved in with children

II.DVital/Health Factors

II.D.1Health and Work see ref’s in Life Cycle

Health factors more easily affect work/leisure choice than housing arrangement choice. Preference for independent living is dominant over dependent care living whether care is received by relatives, at home (SCHS) or at facilities with different levels of care.

II.D.2Health and Mobility Bradsher, Jackson, Longino and Zimmerman (1991,1992)

In the next section, we describe the model specification of our paper. We first review those factors that affect the selection of the observed housing arrangements and some of what the literature on elderly issues has contributed toward understanding how these factors enter into the selected, observed housing arrangement.

III.Modeling Housing Choices made by the Elderly

To facilitate this process we categorize the decision criteria into the distinct groups defined in section I.A: economic, demographic, sociological, and health/vital. We understand that no single factor stands alone in influencing the observed housing choice. Indeed, factors affecting observed housing choices may enter in more than one of the above categories for obvious reasons. For example, labor market decisions (an economic factor in the selected housing type) by sampled individuals are the outcome of health related factors, age, household composition, wealth, income/wages, et. al. Thence ability to work, in conjunction with the work status of the individual affect the observed housing choice; but many other factors indirectly enter into the chosen housing arrangement by affecting labor market participation. For our purposes, our discussion of the characteristics we choose in our logit estimations is compartmentalized by classification category, because this is sufficient to specify a model that determines the extent and factors entering into the housing demand decisions of the elderly.

III.A Specifying and Estimating the Demand Model

LOGIT RUNS: By State for Arizona, California, District of Columbia, Florida, Kansas, Nevada, New Jersey, and Texas (8)

Five sets of eight logit runs are first executed for the above states, with the purpose of illustrating how all four factors contribute to observed housing choices made by the elderly. The first four runs exclude one of the four categories of factors affecting the housing demand choices of the elderly and the fifth run includes all four categories of factors.