Abstract
This study examines how unpaid caregiving histories shape older adults’ perceptions of whether their children will be willing and available to provide future care. Existing frameworks suggest caregiving reinforces expectations of future support, but less understood is how caregiving may also heighten awareness of its burdens and how this relationship varies across heterogeneous caregiving histories. Using life history data from the Health and Retirement Study and latent class analysis, we identified distinct caregiving histories based on the timing, duration, frequency, and overlap of caregiving roles. We then estimated logistic regression models comparing caregivers with non-caregivers and comparing each caregiving history group with non-caregivers. Caregivers were more likely than non-caregivers to anticipate support from children, but this pattern was driven by those with shorter, sequential caregiving histories rather than earlier or more prolonged caregiving experiences. These findings suggest caregiving may reinforce anticipated support when limited in duration but attenuate it otherwise.
Unpaid caregivers—individuals who provide day-to-day support to family members and friends with health-related limitations—form the backbone of the U.S. long-term care system. Between 2015 and 2025, the number of unpaid caregivers in the United States rose from 43.5 to 63 million, and this trend is projected to continue as the population ages (AARP & NAC, 2025). Although caregiving provides vital support to those in need, it is often physically taxing for caregivers, and its health consequences can persist long after caregiving responsibilities have ended (Caputo et al., 2016; Chin & Lin, 2026). Consequently, a growing share of older adults who will eventually require care will themselves have prior caregiving experience.
These first-hand experiences often provide caregivers with unique insights into the realities of aging and caregiving (Hu et al., 2023; Solway et al., 2017), which may, in turn, shape how they perceive the availability and desirability of receiving care from their own children in later life. Yet, this linkage remains largely underexamined in the literature. Prior research has documented correlates of anticipated support from adult children (Lin & Wu, 2014; Patterson et al., 2026; Pillemer & Suitor, 2006) but has rarely examined how individuals’ own caregiving experiences shape these perceptions. For some, caregiving may reinforce intergenerational norms of care provision. For others, exposure to the demands of caregiving may temper expectations of receiving future care or foster reluctance to place similar burdens on the next generation.
To examine these processes, we draw on two competing frameworks while also accounting for heterogeneity in caregiving experiences over the life course. One perspective suggests that providing unpaid care establishes a precedent for generalized reciprocal support, leading individuals to expect similar care from their own adult children in later life. Commonly referred to as the demonstration effect, this perspective predicts a positive association between caregiving experience and anticipated future care availability (Bergstrom & Stark, 1993; Cox & Stark, 1994; Wolff & Laferrère, 2006). In contrast, this study introduces the burden-informed care perspective, which proposes that exposure to caregiving strain may dampen individuals’ willingness or confidence to rely on their adult children for future care, implying a negative association between care experience and perceived future care availability. Importantly, these contrasting processes may vary by caregiving experiences that differ in timing, duration, and recurrence across the life course (Chin & Lin, 2026; Fast et al., 2021; Verbakel et al., 2024). More limited or contained caregiving experiences may reinforce expectations of future care, whereas more prolonged or complex caregiving may instead attenuate or offset such expectations.
Using newly available data from the Life History Mail Survey, a supplemental component of the Health and Retirement Study, this study examines how caregiving histories influence older adults’ perceptions of the availability of future care from their own children, assessing whether this association aligns with expectations derived from the demonstration effect or the burden-informed care perspective. In doing so, findings offer new insights into how caregiving history may either reinforce or disrupt older adults’ sense of available future support from adult children.
Background
Perceptions of future care availability constitute an important dimension of older adults’ well-being. Those who believe that someone will be willing and able to provide care if needs arise report better psychological outcomes than those who do not (Bisconti & Bergeman, 1999; Kim & Yoon, 2022; Yang, 2006), with adult children most often identified as the anticipated source of such care (Abrahamson et al., 2017; Lin & Wu, 2014; Roth et al., 2007). Despite its health relevance, there is limited research and theoretical guidance on what leads some older adults to anticipate care while others do not. Existing studies have largely focused on sociodemographic correlates of these expectations, including gender, race, age, marital status, and health (Lin & Wu, 2014; Roth et al., 2007), but these approaches offer limited insight into how these perceptions are shaped by life course experiences.
In this study, we conceptualize reports of future care availability as reflecting how individuals draw on prior experiences to evaluate the support they expect to receive later in life. Older adults are attuned to their social contexts, and their expectations tend to reflect what they see as realistically available and appropriate to rely on given their circumstances (Peek et al., 1998). In this sense, such reports capture not only preferences but also informed assessments of viable sources of support.
Among the experiences that may shape these assessments, one’s own caregiving experience is especially salient. Emerging research suggests that providing care can reshape how individuals view their own aging and future support needs. For example, caregiving has been linked to more negative self-perceptions of aging (Hu et al., 2023) and to greater consideration of one’s own potential need for care (Solway et al., 2017), indicating that firsthand care experience can heighten awareness of the demands and realities of caregiving. Building on these insights, we examine how caregiving histories inform anticipation of similar support from children.
Long-Term Reciprocity Framework
To situate the conceptual foundation of this study, it is important to begin with the long-term reciprocity framework, which has guided much of prior research on intergenerational caregiving but is limited in explaining the caregiving dynamics examined in the current study. The long-term reciprocity model conceptualizes mutual support as an ongoing process in which family members alternate between the roles of giver and receiver across the life course (Leopold & Raab, 2013; Silverstein et al., 2002). Parent-child exchanges offer a clear illustration of this model: Earlier parental support is often associated with stronger expectations of future care from adult children, as well as with children’s actual provision of time-based assistance in later years (Henretta et al., 1997; Leopold & Raab, 2013; Lin & Wu, 2014; Pillemer & Suitor, 2006). The underlying premise is that early parental investment functions as a form of implicit future care insurance, with adult children’s caregiving serving as repayment for the support they received earlier in life (Silverstein et al., 2002).
However, the long-term reciprocity framework often relies on the notion of heteromorphic exchange—that is, the exchange of different types of support over time, such as early emotional or financial assistance being repaid through time-intensive caregiving later in life (Leopold & Raab, 2013). By contrast, the present study conceptualizes time-based care as both the form of support provided, and the form of support anticipated to be received. Additionally, the framework generally assumes direct dyadic reciprocity, where the recipient of earlier support is expected to provide care to the original benefactor (Silverstein et al., 2002). This assumption rarely holds in the context of caregiving for aging or ill family members, who may no longer be able to reciprocate years later due to declining health or death. As a result, anticipation of future care may be redirected toward other family members, particularly adult children. Accordingly, this study draws on the demonstration effect, a framework that moves beyond heteromorphic and dyadic exchange, to examine how unpaid caregiving experiences shape perceptions of receiving future care from one’s own children.
Demonstration Effect
Unpaid caregivers assume and sustain their roles for a variety of reasons, including love, altruism, adherence to family norms, a sense of duty, and the desire to give back (Funk, 2012; Klimaviciute et al., 2017). Particularly relevant to this study is the demonstration effect, which posits that caregivers may also be motivated by a desire to model the support they hope to receive in the future (Bergstrom & Stark, 1993; Cox & Stark, 1994; Wolff & Laferrère, 2006). Even if anticipating future care is not an initial motivator, it may develop over the course of the caregiving experience (Hsu & Shyu, 2003). From this perspective, individuals with caregiving experiences may be more likely than non-caregivers to anticipate support from their own children in later life, resulting in a positive association between caregiving experience and perceived future care availability from children.
Empirical research on the demonstration effect remains limited, but existing studies generally support its relevance through multiple lenses, including direct reports of care expectations from children, the modeling of caregiving behaviors, and the actual receipt of upward intergenerational support over time. Regarding expectations, Cox and Stark (2007) found that individuals who provided financial assistance to aging parents were more likely to anticipate financial support from their own children. In terms of modeling behaviors, Wolff (2001) observed that parents of young children maintained more frequent contact with their aging parents, which may reflect an effort to model intergenerational support. Similarly, Pezzin and colleagues (2009) found that spouses with joint biological children provided more intensive care to their spouses with disabilities than those with only stepchildren, suggesting an implicit effort to reinforce caregiving norms among likely future caregivers. Evidence on actual support received further reinforces the demonstration effect. For instance, Jellal and Wolff (2002) found that parents who maintained strong ties with their own aging parents were more likely to receive visits from their adult children years later.
However, the assumption that caregiving invariably reinforces confidence in future support overlooks the possibility that such experiences may lead individuals to reconsider whether they can—or should—rely on their own children for care. To address this conceptual gap, we introduce an alternative perspective in which caregiving experience may have a negative association with expectations of future support from children.
Burden-Informed Care Perspective
Research shows that many individuals exhibit a strong desire to avoid becoming a burden to their own children in old age (Burton et al., 2024; Roberto et al., 2001; Van Leeuwen et al., 2019; Weaver et al., 2018). Self-sufficiency is often regarded as a cornerstone of dignity and successful aging, to the extent that some older adults actively decline help or co-residence with their children out of concern that it would disrupt their children’s lives or add to their responsibilities (Roberto et al., 2001; Van Leeuwen et al., 2019). When older adults do seek support, they typically prioritize socioemotional support over instrumental care (Burton et al., 2024). Indeed, receiving extensive hands-on care from children has been associated with elevated depressive symptoms, suggesting that family-provided assistance may sometimes be viewed as an unwanted necessity in later life (Djundeva et al., 2015; Funk, 2010; Lin & Wu, 2011; Silverstein et al., 1996). Taken together, these studies underscore the salience of burden aversion in shaping older adults’ expectations regarding care from adult children.
Our proposed burden-informed care perspective builds on these insights by suggesting that such sentiments may be especially pronounced among individuals who have provided care themselves. Navigating the emotional, physical, and logistical demands of caregiving can heighten awareness of its toll and foster a desire to shield their children from these strains. Many caregivers balance these responsibilities alongside paid work, parenting, and other obligations (Klimaviciute et al., 2017; Pope, 2013), creating a level of burden they may be reluctant to pass on to their children. Beyond protective desires, caregivers may also develop more realistic evaluations of their children’s capacity to provide care based on their own experiences (Peek et al., 1998). Many caregivers report feeling unprepared for the tasks themselves and are uncertain about their willingness or ability to assume the role again (Mroz et al., 2023). These lived realities may make it difficult for individuals with caregiving experience to envision their children taking on a comparable caregiving role in the future. As such, reflecting more cautious or constrained expectations, those with prior care experience may be less likely to report anticipating future care from their own children relative to non-caregivers.
Although this dynamic remains understudied, emerging evidence offers suggestive insights. For instance, Gorenko et al. (2021) found that caregivers were more likely than non-caregivers to plan for or enroll in long-term care services, potentially reflecting concerns highlighted by the burden-informed care perspective. However, such patterns may be influenced by national context. Gorenko and colleagues’ study was conducted in Canada, where publicly funded long-term care systems make formal options more accessible. In contrast, the U.S. care landscape is often characterized by high costs and limited coverage, which constrain the feasibility of formal care planning for many older adults (Roberto et al., 2001). Thus, even when caregivers in the U.S. become hesitant to rely on their children, viable alternatives such as enrollment in long-term care programs are often limited (Finkelstein et al., 2012). In such context, burden-informed care may be expressed less through formal care uptake and more through tempered perceptions of receiving support from one’s own children.
Caregiving Histories and Anticipated Care
Perceptions of future care may depend not only on whether individuals have provided care, but also on the intensity and complexity of their caregiving experiences over the life course. Yet much of the existing research on intergenerational support treat caregiving as a binary status, distinguishing simply between caregivers and non-caregivers. This approach overlooks the substantial variation in caregiving experiences. In this study, we adopt a lifespan perspective that considers caregiving experiences across the life course, including those undertaken prior to older adulthood, and how they may shape assessments of future support. We follow prior studies that have mapped lifespan caregiving histories by capturing differences in when caregiving began, how long individuals were actively caregiving, whether caregiving was a recurring role, and whether care responsibilities overlapped across multiple recipients (Chin & Lin, 2026; Fast et al., 2021; Verbakel et al., 2024). These dimensions of caregiving history are particularly important because they can reflect the degree of strain or disruption that caregivers have experienced (Keating et al., 2019), which may, in turn, shape how they evaluate future care availability.
For many, caregiving is relatively brief and contained—a compressed caregiving history reflecting normative kinship or marital obligations (Goldsteen et al., 2007; Karner & Bobbitt‐Zeher, 2005; Montgomery & Kosloski, 2012) that may reinforce expectations of future family-based care. At the same time, even relatively limited caregiving involvement may still be experienced as burdensome, potentially dampening expectations of relying on others and aligning more closely with the burden-informed care perspective. In contrast, for others, caregiving roles may begin early in life, extend over long periods, or recur multiple times, sometimes with overlapping responsibilities that can be more disruptive to personal and social lives (Chin & Lin, 2026; Fast et al., 2021; Verbakel et al., 2024). These more intensive experiences may either reinforce expectations of future care, particularly if sustained caregiving strengthens beliefs that receiving care in later life is appropriate following one’s own extensive caregiving contributions, or foster reluctance to impose similar burdens on one’s own children and reduce confidence in the availability of such support. Taken together, these competing perspectives suggest that caregiving experiences may be associated with either higher or lower anticipated care relative to non-caregivers, making the direction of these relationships an open empirical question.
The Present Study
Over the life course, many unpaid caregivers will eventually require the same support they once provided and must confront the question of whom they can rely on when that time comes. Unlike non-caregivers, caregivers approach this stage with firsthand knowledge of caregiving’s rewards, demands, and challenges—insights that may shape how they envision their own future care. This study investigates whether and how caregiving histories influence older adults’ perceptions of future care availability from adult children, using data from the retrospective Life History Mail Survey linked to the Health and Retirement Study. Two competing perspectives guide this investigation. The demonstration effect posits that caregiving reinforces intergenerational norms of reciprocity, leading individuals to expect similar care support from adult children (positive association). In contrast, our proposed alternative burden-informed care perspective suggests that caregiving experience may temper confidence in or willingness to rely on adult children for future care (negative association).
Several factors previously shown to be associated with caregiving and perceptions of future care availability are included as controls in the models to isolate the association between caregiving histories and anticipated care from children. These include sociodemographic characteristics such as gender, age, race-ethnicity, and marital status (Lin & Wu, 2014; Roth et al., 2007), which have been linked to variation in anticipated support such as higher expectations among women and Black individuals, and lower expectations among those without partners. Family characteristics are also crucial in shaping care perceptions. We therefore control for family structure, including the number and gender composition of children, as well as their proximity, as these factors shape the availability of potential caregivers (Lin & Wolf, 2020). Socioeconomic resources such as educational attainment, employment, wealth, and long-term care insurance are included given their association with access to formal care and potentially lower reliance on family support. Finally, health factors such as chronic conditions, depressive symptoms, and cognitive impairment are considered, as poorer health is often associated with lower expectations of care (Lin & Wu, 2014; Roth et al., 2007), likely reflecting a more realistic assessment of caregiving needs and demands.
Method
This study draws on data from the 2017 and 2019 waves of the Life History Mail Survey (LHMS), a supplemental component of the Health and Retirement Study (HRS). The HRS is sponsored by the National Institute on Aging (grant number NIA U01AG009740) and is conducted by the University of Michigan. The HRS is a nationally representative longitudinal panel survey that follows U.S. adults aged 51 and older. Launched in 1992 with an initial cohort of individuals aged 51–61, the study achieved full population representation in 1998 with the integration three additional birth cohorts. To retain representativeness, the HRS adds a refresher cohort of adults aged 51–56 every six years. Respondents are interviewed every other year, with off-cycle surveys such as the LHMS to enrich the data on specialized topics not covered in regular core interviews. The LHMS, fielded in 2015, 2017, and 2019, extended HRS’ life course reach by including calendar grids of important life events. We drew exclusively from the 2017 and 2019 LHMS, which included an expanded questionnaire covering unpaid caregiving histories. By the conclusion of the 2019 LHMS survey, nearly all eligible HRS participants had been offered the opportunity to complete the expanded LHMS version. The overall response rate was 60.8% (Larkina et al., 2025).
The LHMS is uniquely suited to the objectives of this study because it retrospectively captures caregiving experiences across adulthood, including episodes that occurred prior to respondents’ enrollment into the HRS. Additionally, in contrast to the HRS core surveys which collect prospective data on caregiving for parents and spouses only, the LHMS gathers detailed histories of care provided to a broader network of recipients, including extended family and nonrelatives. This broader scope allows for a more comprehensive construction of caregiving histories and supports the identification of varied patterns of caregiving over the life course.
To assess perceptions of future care availability, we linked the LHMS data to respondents’ records in subsequent waves of the HRS core survey. Specifically, for those who completed the 2017 LHMS, care expectations were drawn from the 2018 HRS interview. For the 2019 LHMS respondents, the relevant survey items were taken from the 2020 HRS interview. This design strengthens the study’s ability to approximate a temporal order between caregiving histories and subsequent future care perceptions.
Sample Selection
A total of 13,670 respondents participated in the 2017 and 2019 waves of the LHMS, including 10,209 non-caregivers and 3,461 caregivers. Among caregivers, 262 were excluded due to missing data on caregiving years and five for implausible reports (i.e., caregiving predating respondents’ birth year), resulting in 3,194 unpaid caregivers with sufficient information to construct caregiving histories. Because this study centers on perceptions of future care from children, the sample was further restricted to respondents with at least one living child at the time of HRS core interview, excluding 1,146 childless individuals. An additional 3,052 respondents were excluded due to an item skip for those already receiving care at the time of interview (N = 1,513) or nonresponse to the future care availability question (N = 1,539). The final analytic sample included 9,205 respondents: 6,943 non-caregivers and 2,262 caregivers.
Measures
The primary outcome was a binary measure indicating whether respondents anticipated receiving personal care from at least one of their children in the future. Specifically, respondents were asked whether they believed anyone (other than their spouse or partner) would be willing and able to assist them with basic personal care tasks such as dressing or eating for an extended period if the need arose. The HRS allows respondents to name multiple potential caregivers, including children, other family members, and nonkin. Consistent with the predictions of the demonstration effect and the burden-informed care perspectives, responses were coded 1 if respondents named any of their children as a potential source of support, and 0 if they named non-child sources only or reported no expected source of support. This question was asked only of respondents who were not receiving unpaid care assistance at the time of the interview.
Caregiving history was drawn from the LHMS, where respondents reported whether they had ever provided unpaid care (excluding childrearing) for at least six months to someone with illness, disability, or mental health issues. Care tasks included personal care, household help, financial management, service coordination, or regular check ins. Those who answered “yes” were classified as caregivers. Respondents could report up to five caregiving episodes, including start and end years for each. We constructed four indicators for inclusion in the latent class analysis: (1) age at first caregiving (in terciles: early, mid-, later-life), (2) total caregiving duration (in terciles: short, medium, long), (3) number of episodes (dichotomized: one vs. two or more), and (4) overlapping care (dichotomized: reported at least a year where they cared for multiple recipients simultaneously vs. no overlap) (for more details, see Chin & Lin, 2026).
Sociodemographic characteristics included gender, age, race-ethnicity, and marital status. Gender distinguished between respondents who identified as women or men. Age was measured continuously in years. Race-ethnicity was categorized as non-Hispanic White, non-Hispanic Black, Hispanic, or Other. Marital status distinguished between individuals who were married or partnered, divorced or separated, widowed, or never married.
Family characteristics encompassed the total number of children, accounting for both biological and stepchildren. Based on this, we calculated the proportion of children who were biological and the proportion who were daughters. Child proximity was measured as an indicator, coded 1 when at least one child lived within ten miles of the respondent or in the same household, and 0 otherwise.
Economic resources consisted of education, employment, total household wealth, and long-term care insurance. Educational attainment was measured in years of schooling completed. Employment status classified individuals as currently employed, retired, or not working. Total household wealth was calculated as assets minus liabilities, and a binary measure of long-term care insurance coverage, defined as having insurance that covered nursing home care for at least one year or any in-home personal or medical care.
Health included measures of physical, psychological, and cognitive function. Physical health was gauged using a count of eight physician-diagnosed chronic conditions: heart disease, hypertension, cancer, lung disease, arthritis, diabetes, stroke, and psychiatric disorders. Depressive symptoms were assessed using an eight-item version of the Center for Epidemiologic Studies Depression Scale, with higher scores indicating more symptoms. Cognitive function followed the Langa-Weir classification, distinguishing between normal cognition and cognitive impairment, which includes both mild cognitive decline and dementia (Crimmins et al., 2011).
Analytic Strategy
We conduct three sets of analyses. We first identified distinct patterns in caregiving experiences across the life course by conducting latent class analysis (LCA) among the 3,194 ever unpaid caregivers in the sample (for more details, see Chin & Lin, 2026). LCA is a person-centered approach that classifies individuals into subgroups (latent classes) based on similar response patterns across multiple categorical indicators (Collins & Lanza, 2009). We estimated models specifying between two and six latent classes. Each candidate model provides class distribution for each identified class and item response probabilities that guide the distinct characterization of each class. Selection of the optimal solution was informed by model fit indices, including the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), sample-size adjusted BIC (SABIC), bootstrapped likelihood ratio test (BLRT), entropy values for class separation, as well as the substantive interpretability of each class (Sinha et al., 2021; Weller et al., 2020). Respondents were assigned to the class for which they had the highest posterior probability of membership.
In the second analysis, with the identified latent caregiving classes, we estimated descriptive statistics for all study variables by caregiving status (caregivers vs. non-caregivers) and by class membership. In the last analysis, we estimated two multivariable logistic regression models to examine perceived future care availability. The first model compares caregivers to non-caregivers. The second model compares non-caregivers to each caregiving history group identified in the LCA. Multiple imputation addressed missing data on independent variables, which ranged from 1% to 3%. All models were estimated without weighting adjustments as LHMS-specific weights were unavailable at the time of analysis.
Results
Lifespan Caregiving Histories
Model fit indices and theoretical interpretability supported a four-class solution. As shown in Figure 1, statistical fit improved with additional classes, but gains diminished beyond the four-class solution. The six-class model did not significantly outperform the five-class model (BLRT p > .05, results not shown). Although the five-class solution demonstrated slightly better fit statistics and higher entropy (0.81 vs. 0.76 for the four-class, results not shown), these differences were modest when assessed against other considerations (Muthén, 2008). Moreover the additional class in the five-class solution represented only 2% of the sample—below the commonly recommended 5% threshold for class stability (Collins & Lanza, 2009)—and did not provide additional meaningful theoretical distinction. In contrast, the two- and three-class models offered insufficient differentiation to capture the heterogeneity of caregiving histories. Model fit statistics for latent class models.
Figure 2 displays the distribution of caregiving class membership among unpaid caregivers and, for each class, the probabilities for the four LCA indicators in the selected four-class solution. These probabilities provided the empirical basis for assigning class labels and characterizing the defining caregiving patterns of each profile. Posterior probabilities of caregiving indicators across lifespan caregiving profiles.
Compressed Caregivers (50%, N = 1,606) emerged as the most common caregiving profile. The majority of caregivers in this group assumed caregiving responsibilities later in adulthood, with 52% starting after age 54, and 87% starting after age 44. All individuals in this group provided care for only one person, and most did so for a period of 1 to 4 years. If caregiving lasted longer, it was typically for less than a decade. No one in this group reported overlapping caregiving roles.
Early Enduring Caregivers (22%, N = 685) began caregiving earlier in adulthood and often sustained this role for over a decade, typically for a single care recipient. Approximately 77% started before the age of 45, and no one had a first caregiving episode past age 56. Although some individuals in this group returned to caregiving after an initial episode (24%), no one reported overlapping responsibilities.
Recurrent Caregivers (17%, N = 520) provided care intermittently across the life course, cycling in and out of caregiving roles. Every caregiver in this group supported more than one care recipient, and 37% reported overlapping care responsibilities. Cumulative caregiving duration varied but hovered around moderate lengths. Around 58% accumulated 5 to 10 years of care, whereas the remaining caregivers provided 1 to 4 years or more than 10 years of care.
Immersed Caregivers (11%, N = 360) represent the smallest group but is arguably the most complex and intensive. Most caregivers in this group first entered the role before age 45 (70%). All provided support to multiple care recipients, remained active in the role for over a decade, and experienced overlapping responsibilities.
Descriptive Statistics
Descriptive Means or Percentages of Future Care Perceptions From Children and Covariates in the Analyses, by Caregiver Status and History
Source: Life History Mail Survey and Health and Retirement Study core interviews.
Note. Sample limited to respondents with at least one living child at the time of interview.
aSig. diff. from non-caregivers.
bSig. diff. from Compressed Caregivers.
cSig. diff. from Early Enduring Caregivers.
dSig. diff. from Recurrent Caregivers at p < .05 level; Mdn = median.
Women comprised approximately 73% of caregivers in the sample, with greater representation in the Recurrent and Immersed Caregivers profiles. Caregivers were slightly older than non-caregivers, although Immersed Caregivers represented the youngest group overall. Racial-ethnic composition slightly varied as well. Caregivers were more likely to be non-Hispanic White and less likely to be Hispanic or from other racial-ethnic groups compared to non-caregivers. However, racial-ethnic distributions across caregiving profiles were relatively similar. Marital status patterns also distinguished caregivers from non-caregivers. Caregivers were less often married or partnered, more frequently widowed, and less likely to have been never married than non-caregivers. Recurrent Caregivers were more likely to be widowed than Early Enduring Caregivers but less likely than Immersed Caregivers. Across all groups, respondents reported having about three children on average, and most children were biological offspring, although caregivers had a smaller proportion of children who were daughters than did non-caregivers. Approximately two-thirds of respondents, across caregiving and non-caregiving groups alike, had at least one child who was coresident or lived nearby, with Immersed Caregivers reporting the highest percentages.
With regards to economic resources, caregivers reported slightly higher educational attainment than non-caregivers, with Recurrent Caregivers showing higher educational levels than Early Enduring Caregivers. Caregivers were less likely to be employed and more likely to be retired relative to non-caregivers. Median household wealth was higher among caregivers compared to non-caregivers, although Early Enduring Caregivers appeared to have comparatively lower wealth. Having long-term care insurance coverage remained relatively low across the board at around 12–16%, with Recurrent Caregivers significantly more likely than non-caregivers to have such coverage.
Finally, for health factors, caregivers exhibited a higher burden of chronic conditions compared to non-caregivers, particularly Early Enduring and Immersed Caregivers. Caregivers and non-caregivers, however, were comparable in depressive symptom levels. By contrast, caregivers were significantly less likely than non-caregivers to meet criteria for cognitive impairment or dementia, with Compressed and Recurrent Caregivers showing the lowest prevalence. Although Early Enduring and Immersed Caregivers exhibited higher rates of cognitive impairment relative to other caregiving profiles, they remained less likely to have impairment than non-caregivers overall.
Results From Logistic Regression Models
Odds Ratios (OR) From Logistic Regression Models Predicting Future Care Perceptions From Children, by Caregiver Status and History
Source: Life History Mail Survey and Health and Retirement Study core interviews (HRS, 2025; LHMS, 2025).
Note. Sample limited to respondents with at least one living child at the time of interview.
*p < .05, **p < .01, ***p < .001.
aSig. diff. from Early Enduring Caregivers at p < .05 level.
Several characteristics were associated with perceived future care availability. Women, older adults, widowed individuals, and those with more children were all significantly more likely to anticipate support, as were individuals with a greater proportion of biological children or daughters, and a nearby or coresident child. These patterns reinforce the central role of family structure in shaping later-life support expectations. Conversely, Hispanic and other-race respondents were less likely than non-Hispanic Whites to anticipate future care. Notably, educational attainment and having long-term care insurance were not significantly related with future care perceptions, although greater household wealth linked with lower care expectations, as did those who reported disability or unemployment. Finally, individuals with higher levels of depressive symptoms or cognitive impairment were significantly less likely to believe their children would provide care in the future.
Discussion
This study adopts a life course perspective in conjunction with reciprocity frameworks to advance understanding of how caregiving histories shape individuals’ perceived availability of future care from adult children. The demonstration effect suggests that caregiving reinforces intergenerational norms of reciprocity, thereby increasing expectations of receiving similar support from adult children. In contrast, our proposed concept of burden-informed care perspective posits that caregiving experiences may curtail desires to rely on family for similar care. This conceptual innovation is bolstered by a life course approach to caregiving, one that moves beyond binary caregiver classifications to capture variation in caregiving onset, duration, recurrence, and complexity across the lifespan (Chin & Lin, 2026; Fast et al., 2021; Verbakel et al., 2024).
Although anticipated care from adult children was common, it was far from universal. Just over half of U.S. older adults not currently receiving care identified at least one child they believed would be willing and able to provide personal care support if needed. This variation underscores the importance of identifying potential sources of differences in later-life care expectations, including one’s own caregiving history.
On average, unpaid caregivers were more likely than non-caregivers to report that their children would provide care in the future, a pattern broadly consistent with prior evidence supporting the demonstration effect (Cox & Stark, 2007; Jellal & Wolff, 2002). However, a closer analysis of caregiving histories revealed important intricacies. Only Recurrent Caregivers—those who engaged in multiple but relatively brief caregiving episodes—were significantly more likely than non-caregivers to report such perceived availability. One plausible explanation is that these individuals may have encountered comparatively manageable care demands, given the episodic nature of their roles. Prior research suggests that caregiving experiences that are more contained or less continuous may allow individuals to develop familiarity and a sense of competence without sustained strain (Mroz et al., 2023). Relatedly, evidence also indicates that Recurrent Caregivers are not worse off in later-life health and, in some domains, may even exhibit advantages relative to non-caregivers (Chin & Lin, 2026). Such experiences may reinforce confidence in caregiving as a manageable role, thereby strengthening expectations that similar support would be available in the future.
Our findings also reveal a key contrast between Recurrent and Early Enduring Caregivers. While Recurrent Caregivers were more likely to anticipate future support, Early Enduring Caregivers—those with more prolonged and sustained caregiving experiences—were significantly less likely to report such expectations. Their extended exposure to caregiving responsibilities may have contributed to a more burdened outlook, consistent with the burden-informed care perspective. Supporting this interpretation, prior research shows that intensive caregivers were more likely to plan for alternative care arrangements, presumably to rely less on family support (e.g., Gorenko et al., 2021), suggesting that firsthand exposure to caregiving may shift expectations away from intergenerational reliance. Taken together, these findings suggest that the demonstration effect and burden-informed care perspective may not operate as strict opposites but rather lie on a continuum, where the extent to which caregiving reinforces expectations of support depends on the nature and demands of one’s own care experience.
Beyond this contrast, the lack of significant differences among Early Enduring and Immersed Caregivers relative to non-caregivers suggests that caregiving may operate through multiple, offsetting processes, whereby it strengthens expectations of support for some while leading others to become more reluctant to impose burden, resulting in no net difference at the group level. These dynamics reveals the limits of single-framework explanations, such as the demonstration effect, and underscore the need to consider other plausible explanations, including those emphasized in the burden-informed care perspective.
As a relatively understudied area, this study is not without limitations. First, although the data used in the current study offer a rare population-level insight into caregiving histories and perceptions of future care, the outcome measure may conflate perceived constraints or availability of future support with preferred sources of care. As such, respondents may report that their children would provide care even if they themselves would prefer a different care arrangement. Future work would benefit from distinguishing care preferences from perceived availability more explicitly. For instance, surveys could incorporate hypothetical questions such as, “If financial cost were not a factor, what type of care arrangement would you prefer if you needed personal care?” to better disentangle normative anticipation of intergenerational support from actual desires.
Second, our analyses exclusively focused on anticipated care from children, but anticipated care from non-child sources are equally meaningful and may be shaped by different processes not captured in our discussed frameworks. Relatedly, although gender differences in expectations are observed, this study does not examine gendered caregiving dynamics or spousal care processes that may also shape assessments of future support. Examining broader care networks is therefore an important direction for future research, particularly given the increasing diversity of support arrangements in later life.
Third, some findings may reflect selection rather than causal relationships. For instance, individuals with strong familial obligation norms may both be more prone to assume caregiving roles and to expect support from their children in the future. Because we do not directly measure such norms, we cannot fully disentangle these processes. That said, if selection alone explained the results, we would expect uniformly higher expectations of care among caregivers, which was not observed. To help clarify these causal pathways more explicitly, future research should use longitudinal studies that track changes in perceived care availability before, during, and after caregiving episodes, as well as qualitative interviews that uncover the mechanisms underlying these patterns across diverse populations.
Fourth, although our caregiving history measures capture the timing and duration of caregiving across the lifespan, they do not account for day-to-day caregiving intensity, such as hours of care, specific tasks, or care recipient health conditions. The absence of these measures may obscure additional heterogeneity within caregiving profiles and represents an important direction for future research. Finally, the process underlying both demonstration effect and burden-informed care frameworks may also unfold from the vantage point of children who observe their parents’ caregiving endeavors. However, because the data do not include children’s perceptions or expectations regarding future care provision, this study may understate the full intergenerational scope of these processes.
Despite these limitations, this study initiates an important line of inquiry into how caregiving experiences shape perceptions of future care for oneself. This underexplored dynamic has meaningful implications for well-being in later life. The absence of anticipated care from children may itself undermine well-being, particularly if former caregivers do not view family as a realistic or desirable source of support as they age. Our findings suggest that family-based care may remain perceived as viable or desirable when caregiving roles are manageable, but less so when responsibilities become prolonged and demanding. These patterns underscore the need for structural investments that reduce the burdens of caregiving and support a more balanced distribution of care responsibilities between families and formal systems in meeting the growing demands of long-term care in the United States.
Footnotes
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.