Beliefs about Levels of Care Responsibility for Older Adults in Need: Differences by Relationship and the Presence of Dementia

Abstract

INTRODUCTION:

Social norms prescribe family responsibility for care of older adults. However, it is unclear how beliefs about levels of responsibility may vary.

METHODS:

Using the 2024 RAND American Life Panel (ALP), we test whether beliefs vary by type of relationship, including for lifelong friends, whether the older adult has dementia, and by respondent’s caregiver status.

RESULTS:

Level of caregiving responsibility assigned is highest for spouses/partners, but biological children, stepchildren, and lifelong friends are believed to have some level of responsibility. Beliefs are relatively unchanged by impairment, except levels are higher for the family unit when the older adult has dementia. Caregivers and non-caregivers have similar beliefs with a few exceptions.

DISCUSSION:

An older adult’s social network members are perceived to have some caregiving responsibility, with greater levels of responsibility for nuclear family. The results lend support to the application of hierarchical compensatory theory to beliefs about care responsibility.

Background

In the United States, social norms and policies have traditionally placed primary responsibility for caring for older adults on families rather than public entities (1–3). Family members, especially partners and children, continue to provide the majority of caregiving in the United States, helping older adults with various impairments, including physical limitations and intensive care needs like dementia (4,5). Yet, the composition of older adults’ families is changing, with substantial demographic shifts in marriage, remarriage, and number of children (6). These shifts have, in turn, led to questions about changes in the potential care networks available to older adults with increasingly diverse family forms and beliefs about levels of responsibility for non-familial ties, like friends (7–11).

Currently, there is limited knowledge about whether public beliefs about levels of care responsibility vary by type of relationship to the older adult, as well as whether beliefs vary by the type of care needed by the older adult (11,12). Beliefs can be shaped by social and cultural norms, and can impact behavior (13). Research that has compared attitudes or beliefs for different social groups often uses umbrella categories of “family” or “relatives”, without further distinction of relationship types; these studies have found more endorsement for family to care over friends (12,14). Another set of more recent studies on caregiving responsibility beliefs suggest potential for variation by family-specific relationship types; however, this body of work focuses on step vs. biological ties only without comparisons to other types of relationships (15,16). Concerning the type of care needs that the older adult has, previous studies have found that the potential caregiver’s characteristics (e.g., gender, race/ethnicity) and the care situation (e.g., type of care tasks needed) may impact attitudes (12). We turn to evidence from a study on personal preferences for long-term care which shows that attitudes shift away from family-based help if the older adult were to develop dementia compared to if they only need help with household tasks (17). However, preferences are often different than other types of beliefs (e.g., I would prefer my spouse to care for me vs. I believe spouses should care for their partner) (18,19), so it is unclear how universal trends in preferences may be in their application to beliefs.

Finally, there is limited understanding of how beliefs about levels of care responsibility vary between caregivers and non-caregivers, given caregivers direct experience performing such activities. It is important to gauge any differences between these two groups because beliefs have the potential to shift overall public opinion as the number of family and other unpaid caregivers (including friends and others) of older adults has grown from 18 million in 2011 to 24 million in 2022 (5). Current knowledge around beliefs is often limited to national-level samples that infrequently gauge the respondents’ caregiver status (3), or studies of attitudes among caregivers only (1,11). Although a recent study found few differences between caregivers and non-caregivers in attitudes toward care economic policies (20), beliefs about levels of care responsibility may vary from attitudes toward policy solutions for care. In sum, the body of work that would allow us to understand beliefs about levels of care responsibility by relationship type, the older adult’s care needs, and the respondent’s caregiver status are currently disjointed. We extend and bridge together current work in these different areas.

We use new data from the October 2024 wave of the RAND American Life Panel (ALP) to examine beliefs about levels of caregiving responsibility for an older adult who needs care. The ALP is an online panel study with omnibus surveys that allow for new questions to be added and tested with a nationally representative sample of U.S. adults. We contribute to an understanding of beliefs about care responsibility by asking about responsibility for the family as a whole and by a series of relationship types (i.e., partner/spouse, biological child, stepchild, lifelong friend) across two types of older adult care needs: has physical limitations only and has dementia. These measures improve upon prior work that use only umbrella categories (e.g., family), focuses on responsibility of adult children only, or relies on a yes/no dichotomy that obscures potential variation in the level of responsibility (1,3,12,15). We first describe overall patterns of beliefs, then test for differences by relationship type and the older adult’s care needs (physical limitations only or dementia). We focus on variation by whether the respondent was a caregiver for an older adult at the time of the survey, distinguishing between those who care for an older adult with dementia and those who care for an older adult without dementia.

Beliefs About Care

Beliefs about who should provide care can include shared or global social norms (i.e., population-level public opinion) as well as personal norms or preferences (i.e., beliefs about one’s own care or care of a family member) (11,18,21–23). Here, we focus on national-level beliefs because of our interest in understanding prescriptive or evaluative attitudes around responsibility for care of older adults (22). This interest is driven by the growing population of both older adults in need of care as well as the growing number of family and unpaid caregivers (5,7). Tracking public opinions of caregiving beliefs is important as they can influence the individualized process of caregiving (23–25), as well as public policy (20,26,27).

Often, population-level care beliefs are driven by familism, the normative ideology of family responsibility for other members which prescribes certain behaviors like family-based care of older adults (1,28). Within the U.S. context, familism is the “default” in the absence of robust government supports for the provision of care for older adults, and thus a stronger driver of care behavior relative to other nations with more public supports (25). Familism is a widely held belief across various groups in the United States (1,29). These normative beliefs are not only present in public and personal opinion but also embedded in numerous social institutions, including medical systems, support organizations, and public policies, which assume that families will meet the care needs of older adults (30,31). Therefore, within a framework of familism, we expect that beliefs for the family as a whole to provide care for older adults will be strong.

Responsibility by Relationship Type

Existing theoretical frameworks and empirical findings with respect to caregiving trends suggest that beliefs about caregiving responsibility may vary systematically by type of relationship, including within families and for friends. Here, we first focus on the “nuclear family” - partners and adult children - who are the most likely to be caregivers of older adults compared to other social network members (5). Hierarchical compensatory theory was originally conceptualized as a way to understand older adult’s care preferences, which Cantor saw as ordered by the “primacy of the relationship”, meaning that preferences are strongest for family members than other social network members (32). Although beliefs may differ from preferences (e.g., believe that adult children should care, but prefer a spouse to), the theory has been expanded in its use to understand actual care behavior (33). We similarly propose extending the theory to understand societal-level beliefs about levels of care responsibility.

Within this theory, a partner or spouse is often considered the first-choice caregiver, if present. Beliefs that a spouse is responsible for care are implicit or seen as a routine practice within this type of relationship (34–36); that is, care is a part of the institutionalized norms of marriage. While the proportion of older adults who are married remains high (37), a growing portion are cohabiting with a partner (i.e., coresiding but not formally married) or living apart together (i.e., both married and unmarried partners who live separately) (38). Although research on these types of couples is limited, recent work shows growing acceptance of cohabitation among older adults and similarities between unmarried and married partners in relationship quality (39,40). Therefore, we use spouse and partner interchangeably, and, following hierarchical compensatory theory, we expect that beliefs about the spouse or partner’s level of responsibility will overwhelmingly be high.

Adult children are the next source of care after a partner or spouse within the compensatory theory. To date, work on global beliefs for adult children to provide care to their parents (i.e., filial obligations) has received the most attention in the caregiving attitudes scholarship among various relationship types (41). Adult children’s commitment to their parents is considered “strong and stable” regardless of the parent’s care needs or the potential consequences for the adult child (42), and public opinion on adult children’s responsibility to care for their parents has been shown to remain relatively high across time (41). The belief that children should provide care to an aging parent when needed is also consistent across different study populations, from samples of adults to professional home care workers, often driven by beliefs of reciprocity and moral duty (16,43).

Measures of beliefs about children’s responsibility often do not specify child type; yet, a growing segment of older adults have stepchildren, a trend projected to continue (44). Therefore, it is important to consider whether relationships between adult children and older adults are biological or through repartnering (i.e., stepties). Several studies comparing beliefs about caring for step versus biological parents find that filial obligations are greater for biological ties than for stepties (15,16). These studies suggest that stepchildren are not entirely exempt from responsibility, but rather, the level of responsibility varies by relationship type (21,45,46). Therefore, we expect that beliefs about adult children, both biological and step, and their levels of responsibility will be similarly high, but less than for a partner/spouse, and stepchildren will be believed to have lower levels of responsibility than biological children.

Finally, as people age they become more likely to anticipate support from their friends regardless of their family type (14), and friends do provide care in some older adults’ care networks (11,33,47). Within compensatory theory, friends are viewed as an “overlooked” but better source of care than more formal organizations or paid caregivers, though still less so compared to family (32). Socioemotional selectivity theory (SST) can be used in conjunction with compensatory theory to understand why friends may still be an important source of care, especially in later life or in cases of impairment. SST suggests that older adults’ friendship networks shrink with age, but that relationship quality and emotional satisfaction takes precedence over relationship quantity (48,49). Taken together, these two theories suggest that after winnowing networks later in life, older adults’ friends may be more likely to be “like family” or close enough to rely on for care (11,50).

Research on population-level beliefs about friends’ responsibility is limited (8). Scholarship that asks adults about their personal beliefs of who they expect to care or help when needed show greater overall levels of responsibility for family compared to friends, but there is still some preference for receiving some help from friends depending on the circumstances (12,14,51). Importantly, the type of friendship may also matter for beliefs. Lifelong or close friends may be similar to family members in their relationship quality and connection with the older adult compared to other, more casual or associational friends (11,50,52), and expectations for close friends often exceed those for casual friends (51). Therefore, we focus on lifelong friends and expect that beliefs about levels of responsibility for lifelong friends will be moderate but lower than those for family.

The Context of Dementia

There are well-documented differences in the intensity of care needs, size of care network, and the typical progression of impairment when caring for an older adult with dementia compared to caring for an older adult with physical limitations only (53,54). For instance, caregivers for an older adult with dementia spend on average 171 hours providing care per month, compared to 66 hours for an older adult without dementia (55). The care needs of older adults with dementia are also more intensive, and progress from initial support with household tasks to more complex tasks over time. For instance, 40% of older adults with dementia require help with three or more self-care activities (e.g., help with eating, bathing) compared to only 14% of older adults without dementia (54). The intensity of caring for an older adult with dementia also leads to higher reports of negative consequences or caregiver strain (54,56,57). Moreover, given the progressive nature of dementia and the different stages with different levels of impairments, caregiving often begins nearly a decade before death, whereas caregiving for an individual with primarily physical limitations starts closer to the end of life (58).

Despite these differences in care, a significant gap remains in our understanding of whether and how global caregiving beliefs differ by the type of impairment experienced by the older adult. Among studies addressing caregiving attitudes, although some have measured beliefs based on specific instances (e.g., a car accident versus financial help) (12), few rarely measure dementia-specific care relative to other types of care needs. Conversely, research explicitly addressing the context of dementia often lacks comparisons with caregiving for other types of limitations (59,60).

A set of measures and studies address attitudes toward someone with dementia specifically, but these studies often focus on living arrangement preferences (i.e., nursing home use), stigma, and general attitudes about comfort interacting with or caring (by medical staff) for a person with dementia (59,60). Therefore, research on population-level beliefs about levels of care responsibly for an older adult with dementia remains limited. We can, however, look to other work on personal beliefs to understand the influence of relationships on care. Aligned with compensatory theory, one survey of U.S. adults found that 45% expected their spouse to be their primary caregiver if they developed Alzheimer’s disease, followed by a child (12%), another relative (12%), or a friend (2%) (61). However, measures querying about a single primary caregiver can obscure the networked nature of care, particularly relevant when caring for a person with dementia, where multiple family and non-family caregivers often provide varying levels of support to the older adult (7,62).

In regards to comparisons between dementia and other forms of impairment, a handful of studies on similar topics suggest that there will be differences in perception of care responsibility by whether the older adult has dementia. For instance, one study explored attitudes toward older adults with different forms of impairment among nursing home employees. They found that employees had more positive attitudes toward older adults with physical impairments compared to those with dementia (63). Another study, on disabled older women currently receiving family or unpaid caregiving, inquired about the women’s preferences for different combinations of care and living arrangements. This study found increased support for living and receiving care within a nursing home if the person developed dementia compared to a situation where they only needed help with household tasks (non-dementia impairment implied but only measured by type of tasks) (17). Taken together, these results suggest that there may be differences in beliefs about level of responsibility when an older adult has dementia compared to when they do not.

Beliefs Among Those Currently Caregiving

Research on differences in beliefs about care responsibility among those currently (vs. not currently) providing care to an older adult are limited, despite the growing number of family and unpaid caregivers over the last decade and continued impending growth in this population (5,7). Estimates show that the likelihood of providing care at some point during the life course is common. For instance, estimates in the U.S. show that a 20-year-old adult can expect to spend on average 5.1 years (9% of remaining life) providing care to an older adult with an activity limitation, with half of that time spent providing for an older adult with severe limitations (26).

Caregiver beliefs about care may vary from those of non-caregivers. One study found that caregivers were more likely than non-caregivers to endorse family obligation to care for an older adult in need; however, this study was only among younger adults (64). Other research comparing caregivers to non-caregivers has found differences in care-related attitudes, but the differences vary by the outcome measured. For instance, caregivers have more negative views of dementia screenings than non-caregivers (65). Some work finds that caregivers are more likely to support policies that would support care of older adults by families (e.g., paid leave, paying caregivers) than non-caregivers (20,66), but other work finds few differences in attitudes toward care policies by caregivers status (20). Qualitative work among family and friend caregivers of older adults suggest few differences in beliefs about responsibility by whether or not the older adult has dementia relative to other impairments, even if patterns of actual care vary by impairment (11). Caregivers overwhelmingly view traditional family ties as most responsible, despite the array of adults in the actual care network, such as extended relatives and non-kin (1,11). Yet, few of these studies have comparisons to non-caregivers. Therefore, we expect variation in beliefs based on caregiver status and whether or not the caregiver is currently helping an older adult who has dementia or not.

Materials & Methods

This study uses newly collected data from the RAND American Life Panel (ALP) Omnibus Survey (Wave 18, October 2024) to examine beliefs about levels of caregiving responsibility among American adults aged 18 and over. The ALP is a probability-based panel consisting of approximately 11,000 active respondents who participate in online surveys (https://www.rand.org/education-and-labor/survey-panels/alp.html). To ensure representativeness, RAND provides internet access and computers to participants who lack these resources (67). RAND updates panel members’ demographic information on a quarterly basis.

We designed and implemented an ALP module in October 2024 to assess beliefs about levels of responsibility among adults of any age for the care of adults aged 65 or older in need of assistance. We defined the older adult’s need as having only physical limitations or having dementia; we asked separate questions for each type of need (see Table A1). We asked how much responsibility each of the following groups have for caring for an older adult: the family as a whole, a spouse or partner, an adult biological child, an adult stepchild, and a lifelong friend. Four response options were given for responsibility: none, some, most, and almost all. We designed the module to also ask whether the respondent provided care for an older adult in the last 12 months. Taken together, these questions enable respondents to express beliefs on a graduated scale across multiple family care sources (rather than only one or two), capturing greater variation in caregiving beliefs than possible in past surveys. Other information is collected quarterly from ALP panel members, including demographics (e.g., sex), household information (e.g., number of household members), employment-related information (e.g., work status), and health information (e.g., self-rated health).

Of 2,876 invited panel members, 2,002 completed the survey, yielding a completion rate of 69.6% (the ALP’s average response rate for probability samples is 61%) (67). Item-level missing data are low. The item with the greatest non-response—respondents’ ratings of stepchildren’s responsibility for caring an older adult with dementia—was missing for only 9 out of 2,002 respondents (0.4%). All other items had few missing responses. We removed an additional three total cases during data quality checks when respondents both sped through the survey (defined as spending less than two minutes) and answered all the same response for the responsibility items. The final analytic sample was restricted to respondents with complete data on all variables of interest (1,984 out of 2,002; 99.1%). Survey weights provided by ALP are applied to adjust for sampling selection and to generate nationally representative estimates (67). We used STATA version 18.5.

Measures

Dependent Variables

We use a continuous measure for beliefs about levels of responsibility to provide care to an older adult (age 65 or older). Respondents are introduced to the question series with the following wording: “We are interested in your opinion about who should take care of older adults who need help either because of physical limitations or dementia.” They are then asked: “How much responsibility do the following groups have for caring for an older adult with only physical limitations/ dementia?” (see Table A1 for a visual of the question series). We measure each outcome as a score ranging from 0 (“none”) to 3 (“almost all”) for each of the five relationship types. Within the question wording, we defined physical limitations as difficulty with walking, standing, reaching, or grasping. Dementia was defined as a memory or thinking problem that gets worse over time.

Independent Variable

We focus on a categorical measure of the respondents’ caregiver status. We measure whether the respondent provided care to or looked after a family member or friend aged 65 or older who was unable to care for themselves in the last 12 months. We explicitly included both “provided care” and “looked after” in the question, recognizing that many individuals who assist older adults do not self-identify as caregivers (68). Respondents could report caring for an older adult with dementia or an older adult without dementia.

Demographic Covariates

Our models control for multiple characteristics that may influence beliefs, including respondents’ family, economic, and demographic characteristics. The family-level covariates incorporated are marital status (married or cohabiting, separated or divorced, widowed or never married) and household size (1 person, 2 persons, and 3+ persons). The economic covariates measured are employment (currently employed or not employed), educational attainment (high school or less, some college or associate’s degree, or bachelor’s degree or higher), and household income (less than $25,000, $25–49,999, $50–74,999, $75–124,999, and $125,000+). The demographic covariates include dichotomized age (18–64 and 65+ years), sex (male, female), immigrant status (born in the U.S. or foreign-born), race/ethnicity (non-Hispanic White, non-Hispanic Black, Hispanic, and non-Hispanic other racial group), and rural residence (resides in a rural or small town with population under 50,000).

Analytic Strategy

We first provide weighted descriptive statistics for our sample, including beliefs about the levels of caregiving responsibility by relationship type and whether the older adult has dementia or not measured as a score (on a 0–3 scale). We then estimate adjusted OLS regression models using weights to account for complex survey design (svyset) to examine predictors of levels of caregiving responsibility for older adults with only physical limitations. Because respondents report on each relationship type, separate models are estimated for each of the five caregiving relationships. We then compare the separately estimated models using seemingly unrelated regression estimation (suest) to test whether the coefficients differ across relationship types. Next, we estimate an additional set of adjusted, weighted OLS regression models for levels of care responsibility if the older adult has dementia and again test for significant differences across relationship types. Finally, we use seemingly unrelated regressions (suest) to compare model estimates across limitation types to understand whether the predictors of caregiving responsibility differ by limitation type.

Results

Table 1 presents weighted descriptive statistics for the analytic sample (N =1,984). Just over one-fifth (21.8%) reported providing care to or looking after a family member or friend aged 65 or older in the last 12 months. A majority (59.7%) of respondents were married or cohabiting. The most common household sizes were two-person households (35.4%) and households with three or more people (49.5%). Most respondents were currently employed (60.5%). In terms of educational attainment, 38.1% had completed high school or less and a substantial portion has attended some college (26.4%) or obtained a bachelor’s degree or higher (35.5%). A majority (69.7%) have a household income of $50,000 or more. Less than a quarter (22.4%) of respondents were aged 65 or older. The sample includes respondents from diverse racial and ethnic backgrounds, as well as residents of rural areas, in line with national averages (67,69).

Table 1.

Weighted Sample Descriptive Statistics

	%
Caregiver Status
Not providing care	78.1%
Providing care for someone age 65 or older without dementia	15.3%
Providing care for someone age 65 or older with dementia	6.5%
Family characteristics
Marital status
Married or cohabiting	59.7%
Separated or divorced	12.5%
Widowed or never married	27.8%
Household size
1 person	15.1%
2 persons	35.4%
3+ persons	49.5%
Economic characteristics
Employed	60.5%
Education
High school or less	38.1%
Some college or associate’s degree	26.4%
Bachelor’s degree or higher	35.5%
Income
<$25,000	13.1%
$25–49,999	17.1%
$50–74,999	21.3%
$75–124,999	20.5%
$125,000 +	27.9%
Demographics
Age 65 or older	22.4%
Female	51.2%
Immigrant	11.1%
Race/ethnicity
Non-Hispanic White	61.5%
Non-Hispanic Black	12.0%
Hispanic	19.4%
Non-Hispanic Other Racial Group	7.1%
Rural residence	24.2%

Note: Weighted proportion estimates from RAND ALP Omnibus W18 (2024). N = 1,984.

Descriptive Figure 1 displays survey-weighted means for levels of caregiving responsibility (0 = “none,” 1 = “some,” 2 = “most”, 3 = “almost all”) by relationship type and limitation type. Respondents assign a high level of caregiving responsibility to an older adult’s spouse or partner (mean = 1.91 for physical limitation only and 1.93 for dementia), or nearly “most” (=2) of the responsibility. The family as a whole is also assigned a high level of responsibility (1.78 for physical, 1.87 for dementia). Levels drop for adult biological children to be more squarely in the “some” category (1.47, 1.50), and levels are even lower for adult stepchildren (1.08 for both limitation types—about one third lower than for biological children). Beliefs about level of responsibility are very low, but not zero, for lifelong friends (0.74 for both). Pairwise tests that compare the dementia and physical scores within each relationship type find a statistically significant difference only for the family as a whole comparison (p = .009); that is, beliefs about level of responsibility are higher for the family as a whole if the older adult has dementia compared to if they only have physical limitations. All other comparisons are not statistically significant.

Figure 1.

Mean Scores of Levels of Responsibility for an Older Adult by Limitation Type and Family Relationship Type (Unadjusted; Weighted)

Note: Unadjusted, weighted mean level of responsibility score from RAND ALP Omnibus W18 (2024). N = 1,984. Asterisks indicate statistically significant pairwise comparisons (***p < .001, **p < .01, *p < .05) in mean caregiving responsibility scores between dementia and only physical limitations within each relationship type. Estimates are based on survey-weighted means and linearized standard errors using Stata’s svy: mean and lincom.

Table 2 shows the adjusted, weighted OLS regression models for mean level of caregiving responsibility for older adults with physical limitations only, providing insight into the factors associated with differences in responsibility beliefs. Caring for someone aged 65 or older without dementia, relative to not being a caregiver, was associated with higher mean level of responsibility for the family as a whole (β = 0.25, p = .019), biological children (β = 0.24, p = .045), and adult stepchildren (β = 0.20, p = .035). The effect for each of these groups was similar in size, with being a caregiver for an older adult with dementia increasing the mean score by about a quarter on the scale. No other significant differences emerged for beliefs by whether the respondent was a caregiver or not. A few other respondent characteristics were associated with beliefs, however. Respondents with a household income of $50–74K compared to $25K or less believed in a slightly higher level of responsibility for the family (β = 0.34, p = .048), and respondents with an income of $75–124K believed in a slightly lower level of responsibility for lifelong friends (β = −0.40, p = .037). Female respondents compared to male respondents believed in lower levels of responsibility for lifelong friends, though the effect was very small (β = −0.16, p = .032). Residence in a rural area was associated with believing in higher levels of responsibility for stepchildren (β = 0.20, p = .012). Seemingly unrelated estimation tests of all pairwise comparisons of relationship types (e.g., adult biological child versus adult stepchild) and a joint test across all relationship types find statistically significant differences in coefficients (p < .001), indicating that caregiving beliefs vary meaningfully depending on the type of relationship specified.

Table 2.

Adjusted, weighted OLS regression models for level of caregiving responsibility for older adults with only physical limitations by relationship type

	Family as a Whole	Partner/ Spouse	Biological Child	Stepchild	Lifelong Friend
Caregiver Status (Ref = not a caregiver)
Someone without dementia	0.254*	0.227	0.237*	0.218*	0.055
	(0.108)	(0.123)	(0.118)	(0.103)	(0.096)
Someone with dementia	0.309	−0.110	0.057	−0.018	−0.016
	(0.189)	(0.187)	(0.157)	(0.112)	(0.126)
Marital Status (Ref = Married or cohabiting)
Separated or divorced	0.024	−0.302	−0.167	−0.130	−0.064
	(0.205)	(0.178)	(0.163)	(0.120)	(0.104)
Widowed or never married	−0.057	−0.196	−0.100	−0.050	−0.106
	(0.144)	(0.147)	(0.108)	(0.091)	(0.101)
Household size (Ref = 1 person)
2 persons	−0.263	−0.051	−0.104	−0.016	−0.183
	(0.140)	(0.149)	(0.129)	(0.111)	(0.120)
3+ persons	−0.262	−0.308	−0.110	0.010	−0.185
	(0.157)	(0.168)	(0.140)	(0.111)	(0.131)
Educational Attainment (Ref = High school or less)
Some college or Associate’s	0.009	0.215	0.035	−0.112	0.067
	(0.121)	(0.130)	(0.104)	(0.091)	(0.092)
Bachelor’s degree or higher	0.038	0.153	0.119	0.060	0.018
	(0.116)	(0.125)	(0.098)	(0.090)	(0.084)
Employed	0.113	0.123	0.112	0.069	0.042
	(0.130)	(0.137)	(0.108)	(0.092)	(0.098)
Household Income (under 25K omitted)
25–49K	−0.163	−0.230	−0.083	−0.082	−0.193
	(0.181)	(0.196)	(0.177)	(0.168)	(0.173)
50–74K	0.338*	0.239	0.214	0.116	−0.042
	(0.171)	(0.180)	(0.175)	(0.173)	(0.187)
75–124K	−0.071	−0.022	−0.033	−0.083	−0.399*
	(0.182)	(0.209)	(0.164)	(0.166)	(0.191)
125K+	0.075	−0.078	−0.079	−0.164	−0.276
	(0.217)	(0.224)	(0.190)	(0.179)	(0.188)
Female	−0.174	−0.128	−0.044	−0.080	−0.158*
	(0.090)	(0.089)	(0.079)	(0.073)	(0.074)
Age 65 or older	0.102	0.197	0.136	0.052	0.041
	(0.110)	(0.117)	(0.097)	(0.084)	(0.087)
Race/Ethnicity (Ref = Non-Hispanic White)
Non-Hispanic, Black	−0.114	−0.338	−0.028	−0.004	0.064
	(0.218)	(0.212)	(0.195)	(0.150)	(0.128)
Hispanic	0.177	−0.128	−0.059	0.010	0.023
	(0.126)	(0.118)	(0.106)	(0.112)	(0.110)
Non-Hispanic other racial group	0.056	−0.233	0.080	0.233	0.242
	(0.152)	(0.144)	(0.147)	(0.154)	(0.158)
Immigrant	0.110	0.174	0.163	0.080	0.116
	(0.144)	(0.148)	(0.144)	(0.160)	(0.160)
Rural residence	0.125	0.032	0.175	0.202*	0.144
	(0.102)	(0.110)	(0.094)	(0.080)	(0.079)
Constant	1.817***	2.054***	1.384***	1.054***	1.075***
	(0.238)	(0.245)	(0.220)	(0.184)	(0.214)
R-squared	0.072	0.105	0.052	0.053	0.071

Note: RAND ALP Omnibus W18 (2024); N = 1,984;

^***p<0.001,

^**p<0.01,

^*p<0.05

Table 3 displays the adjusted, weighted OLS regression models for mean level of caregiving responsibility for older adults with dementia. Respondents caring for someone without dementia, compared to non-caregivers, reported higher mean levels of responsibility for adult biological children (β = 0.29, p = .018). Respondents caring for someone with dementia, compared to non-caregivers, reported higher mean levels of responsibility for the family as a whole (β = 0.35, p = .026). Other respondent characteristics were associated with care beliefs. Being in a two-person or three-person household compared to a one-person household were associated with lower levels of responsibility assigned to the family as a whole when the older adult has dementia (β = −0.31, p = .031; β = −0.33, p = .040). Having a household income of $50–74K, relative to under $25K, was associated with moderately greater levels of responsibility for the family as a whole (β = 0.43, p = .018), while income of $75–124K was associated with moderately lower levels of responsibility for lifelong friends (β = −0.43, p = .026). Being female was associated with believing in lower mean level of responsibility for all family relationships, including the family as a whole, partner/spouse, adult biological children, and adult stepchildren with the effect size being largest for biological children (β = −0.20, p = .027; −0.21, p = .020; −0.21, p = .012; −0.17, p = .027; −0.19, p = .016), though the effects sizes were overall small. Seemingly unrelated estimations tests of all pairwise comparisons of relationship types indicate statistically significant differences in coefficients, indicating that caregiving beliefs vary meaningfully by relationship type (p = .000).

Table 3.

Adjusted, weighted OLS regression models for level of caregiving responsibility scores for older adults with dementia by relationship type

	Family as a Whole	Partner/ Spouse	Biological Child	Stepchild	Lifelong Friend
Caregiver Status (Ref = not a caregiver)
Someone without dementia	0.154	0.147	0.294*	0.132	0.092
	(0.122)	(0.128)	(0.125)	(0.106)	(0.098)
Someone with dementia	0.352*	0.039	0.185	0.066	0.025
	(0.158)	(0.169)	(0.178)	(0.132)	(0.145)
Marital Status (Ref = Married or cohabiting)
Separated or divorced	−0.097	−0.274	−0.127	−0.205	−0.144
	(0.211)	(0.177)	(0.163)	(0.123)	(0.110)
Widowed or never married	−0.111	−0.235	−0.143	−0.120	−0.073
	(0.148)	(0.150)	(0.112)	(0.094)	(0.110)
Household size (Ref = 1 person)
2 persons	−0.307*	−0.120	−0.174	−0.077	−0.196
	(0.142)	(0.147)	(0.138)	(0.117)	(0.136)
3+ persons	−0.328*	−0.263	−0.096	0.004	−0.195
	(0.159)	(0.169)	(0.151)	(0.120)	(0.135)
Educational Attainment (Ref = High school or less)
Some college or Associate’s	0.079	0.161	−0.016	−0.143	0.038
	(0.125)	(0.133)	(0.112)	(0.092)	(0.098)
Bachelor’s degree or higher	0.103	0.138	0.039	0.001	−0.027
	(0.120)	(0.130)	(0.116)	(0.094)	(0.089)
Employed	0.055	0.033	0.126	0.075	0.088
	(0.135)	(0.142)	(0.111)	(0.093)	(0.102)
Income (under 25K omitted)
25–49K	−0.083	−0.078	−0.071	−0.078	−0.153
	(0.196)	(0.201)	(0.180)	(0.177)	(0.180)
50–74K	0.432*	0.360	0.220	0.080	−0.030
	(0.182)	(0.186)	(0.171)	(0.179)	(0.190)
75–124K	0.007	0.014	−0.087	−0.163	−0.434*
	(0.203)	(0.222)	(0.169)	(0.175)	(0.194)
125K+	0.178	−0.004	−0.063	−0.110	−0.287
	(0.226)	(0.228)	(0.197)	(0.188)	(0.194)
Female	−0.203*	−0.207*	−0.213*	−0.171*	−0.188*
	(0.092)	(0.089)	(0.084)	(0.077)	(0.078)
Age 65 or older	0.083	0.174	0.145	0.092	−0.024
	(0.116)	(0.120)	(0.104)	(0.091)	(0.090)
Race/Ethnicity (Ref = Non-Hispanic White)
Non-Hispanic, Black	−0.102	−0.357	−0.063	−0.007	0.089
	(0.222)	(0.217)	(0.198)	(0.150)	(0.123)
Hispanic	0.219	0.010	−0.097	0.030	0.071
	(0.129)	(0.115)	(0.109)	(0.115)	(0.114)
Non-Hispanic other racial group	−0.023	−0.151	0.067	0.287	0.229
	(0.160)	(0.173)	(0.175)	(0.177)	(0.157)
Immigrant	0.100	0.062	0.147	0.032	0.158
	(0.144)	(0.149)	(0.143)	(0.160)	(0.160)
Rural residence	0.200*	0.078	0.143	0.123	0.158
	(0.102)	(0.107)	(0.097)	(0.082)	(0.088)
Constant	1.912***	2.098***	1.565***	1.205***	1.085***
	(0.248)	(0.241)	(0.228)	(0.204)	(0.223)
R-squared	0.078	0.089	0.063	0.047	0.087

Note: RAND ALP Omnibus W18 (2024); N = 1,984;

^***p<0.001,

^**p<0.01,

^*p<0.05

Finally, we compared the model estimates across the two limitation types. Seemingly unrelated estimations tests revealed statistically significant differences between the physical and dementia models for caregiving beliefs toward the family as a whole only (p = .000).

Sensitivity Analysis

We ran additional sensitivity tests by analyzing the series with categories and ordered logistic regressions (see Appendix Table 2, 3, and 4). Results were similar so we retain the mean measure in our models here for more parsimonious descriptions of results across the various combinations of relationship and impairment types. In addition, we tested controlling for whether the respondent took the survey on a desktop computer (compared to other devices) and found results to be robust.

Discussion

Family care of older adults is normative in the United States (1,3), and the size of the family and unpaid caregiver population is already growing as the number of older adults in need of care is increasing (5). However, our understanding of global beliefs about levels of caregiving responsibility has been limited by few distinctions among family relationship types, a lack of consideration of lifelong friends, as well as few comparisons by the types of limitation experienced by the older adult who needs care and the respondent’s caregiver status. By using a nationally representative sample of respondents from the 2024 RAND ALP, we were able to examine differences by these characteristics, contributing new understanding of different dimensions of care beliefs. We found that U.S. adults believe a partner/spouse or the family as a whole has the highest level of responsibility for ensuring that the care needs of older family members are met, but biological children, stepchildren, and lifelong friends still have some level of responsibility. Few differences in beliefs emerged by whether the older adult has dementia or only physical limitations, except for the family as a whole. Caregivers and non-caregivers generally hold similar beliefs about levels of responsibility, with a few exceptions.

Our results illustrate continued population-level adherence to the principles of familism, or the idea that families are responsible for the care of other family members (1,11), regardless of the older adult’s impairment. We contribute to the literature by showing the variation in beliefs about the level of responsibility not often captured in other studies. Overall, our findings support the application of the hierarchical compensatory theory – traditionally used to understand care preferences – to population-level beliefs about level of caregiving responsibility in the United States (32). Although preferences can be different than other types of care beliefs, we find that the same hierarchical ordering of family members found in preferences (32) can also applied to beliefs about levels of care responsibility for individuals in an older adult’s network. In particular, we find that beliefs about the level of responsibility are highest for a partner or spouse with other types of relationship having lower levels of assigned responsibility. These findings also align with other studies that see care as a normative part of partnerships. Mirroring other studies, we also find that beliefs about the level of responsibility are higher for biological children compared to stepchildren (15,45), and that adult children are believed to have the next highest level of responsibility behind partners as the hierarchical compensatory would predict.

Respondents believe lifelong friends have some responsibility for an older adults’ care, albeit less than family members and the lowest level of responsibility among all types of relationships measured, between “none” and “some” (12,14). This result contributes to the literature which has less frequently gauged beliefs about responsibility for friends in relation to family members. Given impending changes in family structure of older adults, friends may become a more important source of care in the future (18) as the number of older adults lacking close kin increases (70). With changing families of older adults, beliefs about responsibility may shift at the population-level. Our work shows that while beliefs about lifelong friends’ responsibility for providing care are lower than for other family members, they are still perceived as part of the caregiving team.

The presence of dementia does not greatly affect beliefs about care responsibility, with one exception. The family unit as a whole is perceived to have more responsibility in the case of an older adult with dementia compared to older adults with only physical limitations. These findings suggest that familism may be strengthened by more progressive care needs like living with dementia, and our study is one of the first to compare dementia to other impairments regarding these beliefs. The lack of differences in perceptions of responsibility for different individual relationship types beyond the family unit as a whole, however, parallel a recent study of caregivers of older adults with dementia that found that although caregivers perceived greater levels of needs, there were few changes in beliefs about who in particular should be responsible within the family (11). In addition, these results correspond with rates of actual care in the case of dementia, where a larger number of family members and others are helping the older adult compared to when the older adult does not have dementia (54).

The population of family and unpaid caregivers of older adults is growing (5), and our results suggest that there are slight variations in caregivers’ beliefs compared to non-caregivers, depending on whether the caregiver was caring for an older adult with dementia or not. Caregivers of an older adult without dementia were more likely to believe that the family as a whole, biological children, and stepchildren had higher levels of responsibility if the older adult had physical impairment only compared to non-caregivers. These caregivers also believed in greater levels of responsibility for biological children in the care of an older adult with dementia. Although caregivers for an older adult with dementia had few significant differences in beliefs compared to non-caregivers (potentially due to small sample size), they did believe in higher levels of responsibility for the family as a whole when an older adult has dementia. These findings lend further support to understanding important nuances in the beliefs about responsibility to care based on the family relationship (45,46). The finding among caregivers for an older adult with dementia may be driven by larger care networks that often accompany caring for someone with dementia (54).

Other results by different respondent characteristics suggest that beliefs about care responsibility are relatively universal among different groups except in a few cases. The most striking difference between different groups was the overall lower levels of responsibility that female respondents assigned when the older adult had dementia compared to male respondents (and for lifelong friends in the case of physical impairment). This result mirrors international work that has found women are less likely to endorse family care for older adults compared to government care provision (71). Women account for a larger proportion of caregivers to older adults, including those with dementia (5), and therefore may be less likely to endorse family care. Beliefs vary by household income, with those with moderate family income ($50–74K) endorsing higher levels of responsibility for the family as a whole in the case of both physical impairment and dementia, whereas those with income $75–125K endorsed lower levels of responsibility for lifelong friends in both cases. Finally, rural residents believed in greater levels of responsibility for stepchildren in the case of an older adult’s physical impairment, as well as greater responsibility for the family as a whole in the case of dementia. These attitudes may reflect several factors in addition to geographic dispersion, including reduced access to formal services and the presence of larger older adult populations, which might require relying on a wider array of care sources.

This study has limitations. First, we were limited in the number of relationships or alternative caregivers we could inquire about, omitting other potential variations (e.g., casual friends, government) or family specifics (e.g., number of remarriages, relationship quality, gender). Respondents may be considering different options, including access to long-term care, when responding regarding family responsibility. There is growing desire for government assistance for older adults (72), and families continue to diversify in their shape and connections (6), suggesting patterns may continue to change. Second, the measures used here are not task-specific. Litwak’s (73) task-specific model is often paired with compensatory theory and suggests that beliefs about care may vary based on the type of task. Third, although we advance our understanding of caregiver vs. non-caregiver beliefs, we did not measure whether respondents had ever been a caregiver beyond the last year or the type and intensity of care they provide (beyond whether the person they care for has dementia).

Although our study advances understanding of beliefs by investigating multiple types of relationships, including friends, and level of responsibility, there are many remaining questions for future research to consider. Future surveys could inquire about other care and relationship dimensions and dynamics not captured here. For instance, questions could be asked about variation in beliefs in relation to long-term care costs and supports, relationship quality, task type, geographic location, and sex (e.g., daughters vs. sons), as well as for an extended list of relationship types (e.g., grandchildren). These additional aspects could be used to further tease out whether the hierarchical compensatory theory can be applied beyond the set of relationships and dimensions presented here.

Our study builds upon prior research that has investigated beliefs about family care by investigating multiple dimensions including multiple relationship types, levels of responsibility, the older adult’s impairment, and the respondent’s caregiver status. In addition, our results empirically apply the hierarchical compensatory theory to beliefs about care, although the theory has historically been used to understand preferences and actual care. Our results suggest that this theory can be applied to a hierarchical set of beliefs about levels of responsibility for care of older adults, finding the greatest responsibility for the older adult’s partner, followed by adult children and lifelong friends. Beliefs about the family as a whole’s care responsibility are greater for an older adult with dementia compared to one with only physical limitations. This result suggests that the social norms of family-care are strengthened in relation to care for an older adult with dementia.

Data Availability Statement:

The RAND ALP data used in this study are available at https://alpdata.rand.org/index.php?page=data&p=showsurvey&syid=644.