Keeping Us Young? Grandchild Caregiving and Older Adults’ Cognitive Functioning

Abstract

Objective:

This study investigates longitudinal associations between providing care to grandchildren and cognitive functioning. It also examines heterogeneity in these relationships.

Background:

Grandchild caregiving may support older adults’ cognitive functioning by providing social engagement and emotional meaning. However, studies caution that time- intensive or custodial grandchild caregiving can take a toll on grandparents. The cognitive health implications of grandchild caregiving may thus depend on contexts including time spent providing care and living arrangements. They may also vary across sociodemographic groups and have greater effects on older adults who are more vulnerable to cognitive decline.

Method:

Data came from the 1998–2016 waves of the Health and Retirement Study (HRS) and represented over 11,000 U.S. adults age 50+. Using linear growth curve and dynamic panel models, the analysis explored relationships between level of grandchild care and cognitive functioning over time and across sociodemographic, family, work and health characteristics.

Results:

Those providing 100–199, 200–499 or 500+ hours of care to grandchildren had better cognitive functioning than non-caregivers regardless of whether they lived with grandchildren. Positive links between grandchild caregiving and cognition were stronger for lower income, non-working, and unpartnered adults and grew with age and functional limitations.

Conclusion:

These findings suggest that providing care to minor grandchildren may help support cognitive functioning as adults age. They also support the hypothesis that more vulnerable or isolated groups of older adults may benefit the most from grandchild caregiving.

Introduction

Grandparenting is widely considered to be one of the most positive and fulfilling experiences of mid- to later-life. A variety of theory and research suggests that providing care to grandchildren may be linked to older adults’ health outcomes, including cognitive functioning. The “use it or lose it” hypothesis highlights that social engagement in older age stimulates intellectual activity and can help to maintain cognitive functioning through neurological pathways, particularly when social activities are complex (Hultsch et al., 1999; Seeman et al. 2001, 2011). Researchers cite the stimulation that childcare activities provide as well as their emotional benefits as explanations for positive cross-sectional and shorter-term links between grandchild caregiving and cognitive functioning (e.g., Arpino & Bordone, 2014; Jennings et al., 2021; Sneed & Schulz, 2019). However, as is emphasized in theory on caregiving and ambivalence in intergenerational relationships (Milkie et al. 2008; Pillemer et al., 2019), providing care to family members is a complex experience that involves stressors as well as rewards. Grandchild caregiving covers diverse forms and activities that may carry very different meanings for grandparents’ cognitive functioning, including occasional childcare, regular caregiving activities, and custodial care. Reflecting this diversity, studies also show that time-intensive and coresidential grandparenting can take a toll on older adults’ well-being, particularly their mental health (e.g., Baker & Silverstein, 2008; Hughes et al., 2007; Waldrop & Weber, 2001). These forms of grandchild caregiving may have similar negative impacts on cognitive functioning. It is thus important to consider how contexts including time spent in the role and coresidence with grandchildren shape the cognitive health implications of grandchild caregiving.

Implications of providing care to grandchildren for cognition may also vary across sociodemographic groups and social status characteristics. On the one hand, the stresses of grandchild caregiving may have greater negative impacts on cognitive functioning for racial and ethnic minority, less socially integrated, and more materially disadvantaged older adults, who are also more vulnerable to cognitive decline (Baker et al., 2008; Chen et al., 2015; Harrington Meyer & Kandic, 2017; Schwartz et al., 2004; Seeman et al., 2011). On the other hand, research finding that these groups of older adults especially need and benefit from social contact and ties suggest that they may reap greater cognitive rewards from time with grandchildren (Forster & Stoller, 1992; Schöllgen et al., 2011; Walen & Lachman, 2000).

This study uses data from the 1998–2016 waves of Health and Retirement Study (HRS) to examine longitudinal relationships between time spent providing care to grandchildren and cognitive functioning among U.S. adults age 50+. The analysis extends previous work by exploring whether the links between grandchild caregiving and cognition vary across grandparent-grandchild living arrangements and sociodemographic, family, work, and health characteristics. It also follows individuals over eighteen years and makes use of linear dynamic panel models, helping to account for selection into grandchild caregiving.

Background

One consequence of extending life expectancies is that many people spend an increasing proportion of their lives as grandparents (Margolis, 2016; Margolis and Wright, 2017). At the same time, the demand for grandparent childcare in the U.S. has grown as a side-effect of increases in the rates of mothers’ labor force participation and single parenthood, immigration, and parental incarceration and substance abuse disorders (Cherlin, 2010; Harrington Meyer & Kandic, 2017; Livingston & Parker, 2010). In 2008, 39 percent of grandparents age 65+ reported providing some care to grandchildren in the last 12 months (Livingston & Parker, 2010). Over 7 million grandparents shared a home with a minor grandchild in 2019, about a third of whom were the grandchild’s primary caregiver (U.S. Census Bureau, 2019).

Benefits of Grandchild Caregiving

Consistent with the widespread sentiment that grandparenting is one of life’s greatest joys, research shows that many grandparents find grandchild caregiving profoundly rewarding (Crowther et al., 2015; Hayslip & Kaminski, 2005; Villar et al., 2012; Waldrop & Weber, 2001). In qualitative work, custodial grandmothers age 42–75 (Crowther et al, 2015), custodial grandparents age 40–80 (Waldrop & Weber, 2001), and grandparents age 41–71 providing custodial, co-residential and regular supplementary care (Jendrek 1993) emphasized emotional benefits including affectionate relationships with their grandchildren, the ability to positively impact their development, and feeling connected to future generations. Grandchild caregiving can further benefit grandparents by enhancing opportunities for social interaction with other adults and family members, as indicated in studies of non-coresident grandparents age 50–80 (Arpino & Bordone, 2014), supplementary and coresidential grandparents age 50+ (Notter, 2022), and grandparents age 40+ providing any type of care (Quirke et al., 2021). Caregiving for grandchildren also predicted increased physical activity among grandmothers age 60+ who provided regular supplementary care to grandchildren (Villar et al., 2012) and grandparents age 40–90 providing any type of grandchild care (Jennings et al., 2021). Scholars draw on this compelling evidence that grandchild caregiving can be emotionally fulfilling, improve social integration, and increase physical activity to explain findings that it is positively associated with psychological well-being (di Gessa et al., 2016a, 2016b; Giarrusso et al., 1996; Notter, 2022; Xu et al., 2017) and physical health (di Gessa et al., 2016b; Hughes et al., 2007; Zeng et al., 2021)

Adults who provide care to grandchildren also appear to have better cognitive health than those who do not perform these care roles across diverse national contexts (Ahn & Choi, 2019; Arpino & Bordone, 2014; Jennings et al., 2021; Jun, 2015; Liao et al., 2021; McKay & Nadorff, 2021; Silverstein & Zuo, 2021; Sneed & Schulz, 2019; Zeng et al., 2021). Although most research documenting this pattern is cross-sectional or short-term, there is growing longitudinal evidence that caring for grandchildren predicts slower cognitive decline, including analyses spanning up to three waves of U.S.-based panel studies of adults providing non-custodial (Sneed & Schulz, 2019) and custodial care (McKay & Nadorff, 2021). These findings resonate with the “use it or lose it” hypothesis of cognitive aging, which highlights the explanation that participation in everyday social interactions stimulates cognitive ability through neurological pathways due to the intellectual engagement these activities demand, particularly if they are complex (Hultsch et al., 1999; Seeman et al. 2001, 2011). Emotional well-being and increased physical activity may also be pathways linking grandchild caregiving to slower cognitive decline(Cheng et al., 2014; Fratiglioni et al., 2004; Kåreholt et al., 2005 Kelly et al., 2014; Seeman et al., 2011). Although studies tend to group kinship and other networks together, some evidence suggests that social connections to family members are especially important sources of support for cognitive functioning (Béland et al., 2005; Zunzunegui et al., 2003)

Costs of Grandchild Caregiving

As highlighted in theory and research on intergenerational relationships and family caregiving, providing care to loved ones, including grandchildren, is a multifaceted experience that may be simultaneously rewarding and stressful (Dunifon et al., 2020; Milkie et al., 2008; Pillemer et al, 2019). In addition to reaping benefits from time spent caring for grandchildren, caregiving grandparents may find grandchild care exhausting, feel that it is non-normative, and have less time for other beneficial social activities and themselves (Crowther et al., 2015; Harrington Meyer & Kandic, 2017; Jendrek, 1993). Pointing to these potentially stressful aspects of grandchild caregiving, studies also find negative impacts on custodial (Baker & Silverstein, 2008) and coresidential (Hughes 2007) grandparents age 50+, particularly for mental health (see Danielsbacka et al., 2022 and Kelley et al., 2021 for two recent reviews).

Inconsistencies in findings about how grandchild caregiving is related to grandparents’ well-being can typically be attributed to differences in types and levels of care. Although supplementary or relatively low levels of grandchild care are often linked to benefits, custodial and time-intensive care more frequently takes a toll on grandparents (Chan et al., 2022; Chen et al., 2015; Danielsbacka et al., 2022; Notter, 2022). These forms of caregiving are often driven by stressful circumstances that prevent parents from looking after their children, such as health problems, incarceration, and divorce (Baker et al., 2008; Hayslip et al., 2019; Hughes et al., 2007; Jendrek, 1993). Grandparents with fewer resources to draw on are also more likely to be exposed to time intensive and custodial grandchild caregiving (Baker et al., 2008; Hayslip et al., 2019; Livingston & Parker, 2010). By contrast, supplementary grandchild care is likely to be more voluntary, and some studies suggest that grandparents who are better off materially are more likely to become non-custodial grandchild caregivers (di Gessa et al., 2016b; Notter, 2022).

The Importance of Grandchild Caregiving Contexts

These patterns highlight the need for researchers to consider the contexts in which care is given including level of care and living arrangements as well as to account for the selection of disadvantaged families and grandparents into more intensive care. However, most analyses of the relationship between grandchild caregiving and cognitive functioning are limited to one mode of care (e.g., Arpino & Bordone, 2014; Jun, 2015; McKay & Nadorff, 2021; Sneed & Schulz, 2019). In others, time spent in the role and living arrangements are unclear (e.g., Ahn & Choi, 2019; Jennings et al., 2021). Further, although the majority of studies find a positive link between cognitive functioning and grandchild care, most also adopt a cross-sectional or short-term longitudinal approach. They thus may not account for the possibility that grandparents who are cognitively healthier are more likely to become grandchild caregivers.

Variations Across Socodemographic Groups

The implications of grandchild caregiving for cognitive functioning may also vary across sociodemographic and social status characteristics. On the one hand, grandparents who are more vulnerable to cognitive decline, such as racial and ethnic minority, lower socioeconomic status, and less socially engaged adults (e.g., Fratiglioni et al., 2004; Schwartz et al., 2004; Seeman et al., 2011; Zhang et al., 2016) may be more exposed to and negatively impacted by the stresses associated with caring for grandchildren (Baker et al., 2008; Harrington Meyer & Kandic, 2017; Hayslip & Kaminski, 2005). Supporting this possibility, some research shows greater negative health impacts of grandchild caregiving on racial and ethnic minority and female grandparents than White and male counterparts (Chen et al., 2015; Choi, 2020; Szinovacz & Davey, 2006). On the other hand, evidence also suggests that minority, materially disadvantaged and less socially integrated adults are especially likely to need, draw on and benefit from social and family support (Burr & Mutchler, 1999; Forster & Stoller, 1992; Schöllgen et al., 2011; Silverstein & Bengtson, 1994; Walen & Lachman, 2000). The emotional and social rewards of grandchild caregiving may thus have a greater protective effect for these grandparents than for those with more resources and social roles. In line with this hypothesis, other studies find that minority grandparents and women experience better health outcomes in association with grandchild caregiving than White and male counterparts (di Gessa et al., 2016b; Goodman & Silverstein, 2006; Hughes et al., 2007; Notter, 2022). Prior research has yet to systematically investigate these competing hypotheses about heterogeneity in the relationship between grandchild caregiving and cognition.

Research Questions & Contributions

Building on this literature, we investigate the following questions in this study:

1. How are different levels of time spent providing care to grandchildren related to cognitive functioning over time among U.S. adults age 50+?

In response to our first research question, drawing on theory that social engagement supports cognitive functioning (Hultsch et al.,Seeman et al., 2001; 2011) and evidence that grandparents find time spent with grandchildren emotionally rewarding (Crowther et al., 2015; Hayslip & Kaminski, 2005; Villar et al., 2012; Waldrop & Weber, 2001), we expect that grandchild caregiving will predict higher cognitive functioning regardless of time spent in the role.

2. Is the longitudinal relationship between time spent providing care to grandchildren and cognitive functioning mediated by multigenerational living arrangements?

We expect that grandparents living in skipped-generation households, who are likely to be primary caregivers, will not experience benefits from caregiving, consistent with prior work (Chan et al., 2022; Chen et al., 2015; Danielsbacka et al., 2022; Kelley et al., 2021; Notter, 2022)

3. Does the relationship between grandchild caregiving and cognitive functioning vary across grandparents’ sociodemographic and status characteristics?

In line with theory and research highlighting that the benefits of social engagement and support may be especially profound among more vulnerable groups (Burr & Mutchler, 1999; Forster & Stoller, 1992; Schöllgen et al., 2011; Silverstein & Bengtson, 1994; Walen & Lachman, 2000), we expect that any increases in cognitive functioning associated with grandchild caregiving will be greater among minority, materially disadvantaged and less socially integrated adults.

We extend previous research exploring links between grandchild caregiving and cognitive functioning in three key ways. First, we examine both differences in time spent caregiving and multigenerational living arrangements. Second, by asking whether the grandchild caregiving-cognitive functioning relationship varies across sociodemographic and status characteristics, our analysis acknowledges diversity in the structure and meaning of grandchild caregiving in the U.S. Exploring these heterogeneities is of deepening importance as the demand for grandparent childcare grows and the population of grandparent caregivers becomes increasingly diverse. Last, our longitudinal analysis follows adults over of 18 years from mid- to later-life and incorporates linear dynamic panel modeling. Although other studies have adopted creative strategies to help account for the selection of healthier adults into grandchild caregiving, our longer-term approach allows us to observe how these associations unfold over many years and to consider reciprocal effects in the caregiving-cognition relationship.

Method

Data and Sample

Data for this study came from the HRS, an ongoing, biennial panel survey of U.S. adults age 50+ and their spouses supported by the National Institute on Aging and the Social Security Administration and housed at the University of Michigan (Juster & Suzman, 1995). The HRS began in 1992 with a national probability sample of 12,654 adults born from 1931–1941. Respondents from the Asset and Health Dynamics Among the Oldest Old study (b. <1924) were added in 1993. In 1998, two additional birth cohorts were added to extend the sample to all adults age 50+ (Children of the Great Depresion, b. 1924–1930 and War Babies, b. 1924–1947).

We utilized the longitudinal individual-level data file cleaned and prepared by RAND (Bugliari et al., 2021), to which we merged in raw data on household members and grandchild caregiving from each year. Our base sample included HRS respondents who had joined the study by 1998, were born before 1949, and who had completed at least three of the ten surveys from 1998–2016 (N=18,061). We omitted respondents whose cognitive functioning scores suggested that they had cognitive impairment or dementia at baseline in 1998 (N=2,667), or in 1996 for those missing cognitive functioning scores in 1998 (N=206). Respondents missing a cognitive functioning score in both 1996 and 1998 (N=1,728) and without any grandchildren at any study wave (N=1,527) were also excluded. Both respondents who were grandparents at baseline and who became grandparents during the observation period are included. Other omissions include respondents who resided in nursing homes a majority of the waves that they were interviewed (N=213), those without at least three waves of non-missing data on key time-varying study variables (N=298), and those who were missing data on any non-time-varying covariate (N=7). This resulted in an analytic sample size of 11,357 respondents.

Variables

Grandchild caregiving level.

The key independent variable was a time-varying indicator of time spent caring for grandchildren. At each wave respondents were asked whether they or their spouse spent 100 or more hours in the last two years taking care of grandchildren. Those responding “yes” were asked how many hours they themselves spent providing care, and were probed to answer if it was more or less than 200 and 500 hours. We created dummy variables for low (100–199 hours), medium (200–499 hours), and high grandchild caregiving (500+ hours). When averaged over the two year period, these caregiving levels translate to about 1 to 2, 2 to 5, and 5 or more hours a week. Those providing 0–99 hours of care were coded as non-grandchild caregivers (hereafter, “non-caregivers”). The HRS did not ask for other details about grandchild caregiving, such as frequency, grandchild characteristics, and whether the care was custodial.

Cognitive functioning.

Cognitive functioning is measured using an adapted version of the Telephone Interview for Cognitive Status, a reliable and valid survey tool for assessing cognitive impairment (TICS; Brandt et al. 1998, Crimmins et al. 2011; Ofstedal et al. 2005). TICS scores range from 0 to 27, and summarize a respondents’ ability to recall 10 nouns immediately after they are read and after a delay (0–20 points), to subtract 7 from a given starting number five times (0–5 points), and to count backwards from a given number (0–2 points). Normal cognition scores range from 27–12. Scores suggesting cognitive impairment without dementia range from 11–7, and scores indicating likely dementia range from 6–0 (Langa et al., 2005).

Multigenerational living arrangements.

Multigenerational living arrangements are included to provide further context about different types of grandchild caregiving, particularly since respondents living in skipped-generation households are likely to be custodial grandparents. We created three dummy variables identifying adult child and minor grandchild household members to capture multigenerational households and differentiate possible custodial grandparents: 1) adult child but no minor grandchildren, 2) three generation household (i.e., adult child and minor grandchildren present), and 3) skipped generation household (coresident minor grandchild but no adult children). Older adults who do not live with adult children or grandchildren may still provide care to grandchildren, and in 1998 32.2% of such respondents reported providing 100 or more hours of grandchild care in the previous two years.

Other time-varying characteristics.

Our analysis also identified the presence of a spouse/partner or other household member in each year, which may be independently linked to cognitive functioning. Because almost all married or partnered (and not separated) respondents coresided with their spouse/partner, we coded the .2 to 1.4 percent of respondents living apart from their spouse/partner at any given wave as missing on partnership status in that year. Thus, the marital/partnership status variable only includes individuals who lived with their partners. A separate dummy variable identified the presence of all other non-spouse, non-adult child and non-grandchild household members. The comparison category is respondents who lived alone in that year for analyses including multigenerational living arrangements, partnership status, and presence of other household members.

Other time-varying covariates were household income quartile, employment status, and number of grandchildren. We included a time-varying indicator of urban residence to help account for differences in multigenerational living arrangements between urban and rural older adults. We identify respondents who reported no living children at any given wave, since some may be residing in skipped-generation households due to a child’s death. We also accounted for whether one or more adult child lived within ten miles of the respondent at any given wave, since this shapes their risk for grandchild caregiving. Depressive symptoms scores (0–8; measured by the Center for Epidemiologic Studies-Depression scale) and activities of daily living (ADL) limitations scores (0–5) were included as possible mediators of grandchild caregiving-cognitive functioning relationships.

Non-time-varying characteristics.

Respondents’ age at baseline in 1998, race and ethnicity, nativity, gender, and years of education were also included as covariates.

Analytical Strategy

We employed linear growth curve models to assess relationships between low, medium, and high grandchild caregiving and cognition. Linear growth curves are a type of multilevel model that nests observations at different time points within individuals. The first level of these models predicted individual cognitive functioning scores at a given wave as a function of baseline cognition scores, change associated with time-varying predictors (including grandchild caregiving), and random error. At the second level they calculated individual baseline cognitive functioning scores as a function of an intercept, time-invariant predictors, and error. Individuals can contribute different numbers of observations, making the models robust to missing data. Because they account for change within individuals, they also help control for unobserved error. Study wave is the time scale for these analyses. Each individual ages two years across each successive wave, meaning that this approach to measuring time captures within-individual aging effects on cognitive functioning as well as potential cohort effects. Addressing our first research question, the main model for the linear growth curve analysis shows relationships between time-varying grandchild caregiving level and cognitive functioning over time, net of time-varying and non-time varying controls. We then added living arrangements to examine our second research question, which asks whether the presence of grandchildren in the household, with or without adult children, conditions the relationship between caregiving and cognitive functioning. We interacted all control characteristics with grandchild caregiving to address our third research question regarding heterogeneities in the caregiving-cognition relationship.

Although they are well-suited to our research questions, linear growth curve models examine relationships between outcomes and predictors contemporaneously, and thus cannot shed light on the time ordering of these relationships. To provide further context on causal ordering we also estimated linear dynamic panel models using the command xtdpdml in Stata (Williams et al., 2018). Like linear growth curve models, these models help account for unobserved confounders by focusing on change within individuals over time. They also add the possibility to include lagged effects of predictors and outcomes, helping to account for endogeneity, or the presence of variables that are partly determined by previous values of other variables in the model. We included one and two lags of cognitive functioning and grandchild caregiving in our linear dynamic panel model analyses to help account for likely reciprocity in the relationship. We do not add additional lags due to the likelihood for lost degrees of freedom and the introduction of multicollinearity. Stata’s full-information maximum likelihood (fiml) approach to dealing with missing data is compatible with xtdpdml, allowing us to preserve our full sample size in these analyses. However, fiml also dramatically increases the amount of time required to run these models. To reach model convergence, we simplified the linear dynamic panel models by including dichotomous measures for grandchild caregiving and the presence of non-spouse, non-adult child, non-grandchild household members.

Results

Table 1 shows descriptive statistics for the total sample and across grandchild caregiving groups at baseline in 1998. Grandchild caregivers had somewhat higher mean cognitive functioning scores than non-caregivers at baseline. Grandchild caregivers were on average younger than non-caregivers, and a greater proportion of high than non-caregivers belonged to racial/ethnic minority groups, were female, or were employed. High grandchild caregivers also had lower levels of education, while low and medium caregivers belonged to a higher average income quartile than non-caregivers. Low caregivers had lower ADL limitation scores than non-caregivers. Grandchild caregivers were more likely than non-caregivers to have living children at baseline, had more grandchildren, and were more likely to report that a child lived within ten miles of them. Grandchild caregivers were also more likely to live with a partner or grandchild and less likely to live alone than non-caregivers.

Table 1:

Sample Descriptive Statistics Across Grandchild Caregiving Level in 1998 in Percent or Mean

	Total sample (N=11,357)	Non-caregiver 1998 (N=7,330)	Low grandchild care 1998 (N=1,031)	Medium grandchild care 1998 (N=1,097)	High grandchild care 1998 (N=1,638)
Cognitive functioning score (12–27)	17.49 (3.31)	17.40 (3.31)	17.71* (3.30)	17.71* (3.31)	17.62* (3.31)
Age (50–96)	64.31 (8.73)	65.19 (9.31)	63.79* (7.47)	63.06* (7.04)	61.89* (7.13)
Race/ethnicity
White	81.4	83.0	84.8	82.5	73.6*
Black	10.6	9.4	8.4	9.7	17.0*
Hispanic	6.2	5.9	5.8	6.7	7.0*
Other	1.7	1.6	1.0	1.2	2.4*
Foreign-born	7.8	7.8	7.2	7.4	7.7
Female	60.9	60.3	60.8	60.8	65.5*
Years of education (0–17)	12.67 (2.79)	12.71 (2.83)	12.67 (2.69)	12.79 (2.70)	12.38* (2.71)
Household income quartile (1–4)	2.50 (1.12)	2.47 (1.13)	2.59* (1.10)	2.60* (1.09)	2.48 (1.08)
Employed	47.3	46.1	48.5	47.5	50.7*
Urban residence	42.7	42.2	42.6	41.2	43.5
Depressive symptoms (0–8)	1.37 (1.78)	1.37 (1.78)	1.32 (1.72)	1.36 (1.78)	1.41 (1.83)
ADL limitations (0–5)	.16 (.59)	.17 (.61)	.10* (.43)	.16 (.57)	.18 (.58)
No living children	.80	1.09	.1*	.2*	.4*
Number of grandchildren	5.30 (4.97)	4.92 (4.91)	6.18* (5.24)	5.93* (4.53)	6.06* (5.17)
1+ child lives within 10 miles	67.3	62.0	76.7*	76.0*	82.6*
Household members
Spouse/partner	74.5	71.8	81.0*	79.4*	78.2*
Adult child without grandchild	18.3	18.7	16.6	15.1*	19.4
Three generation household	4.2	1.7	3.3*	7.8*	13.1*
Skipped generation household	1.6	.4	1.9*	1.1*	6.9*
Any other household member	9.2	9.8	6.7*	7.5*	9.2
Lived alone	16.3	19.6	11.7*	11.9*	8.5*

^*= Significantly different from those who did not provide care to grandchildren during the study, p<=.05, two-tailed tests. Standard deviations in parentheses.

Results from linear growth curve models regressing cognitive functioning on grandchild caregiving are shown in Table 2. Coefficients for grandchild caregiving in Model 1 indicate that those who provided any level of grandchild care at any given wave had higher cognitive functioning scores than non-caregivers. Coefficients are very similar across level of care, but medium grandchild caregiving had the strongest association with cognitive functioning. Medium grandchild caregivers had cognitive functioning scores that were .186 points higher than those of non-caregivers at any given wave. Cognitive functioning scores decreased with time and age, were lower among men and racial/ethnic minority respondents than women and White respondents, and increased with years of education. Higher income predicted better cognitive functioning, and those who were working or lived in an urban area had better cognition scores than the non-working and rural. Number of grandchildren was positively linked to cognitive functioning scores. When compared to coefficients for other time-varying characteristics, the effects of grandchild caregiving on cognitive functioning are weaker than the effects of household income, comparable to the effects of being employed and residing in an urban location, and greater than the effects of being married or partnered.

Table 2:

Results from Linear Growth Models Regressing Cognitive Functioning on Grandchild Caregiving Level, 1998–2016 (N=11,357)

	Model 1	Model 2	Model 3
Low grandchild care^	.121**	.120**	.106**
	(.040)	(.040)	(.040)
Medium grandchild care^	.186***	.185***	.178***
	(.041)	(.041)	(.041)
High grandchild care^	.109**	.112**	.106**
	(.037)	(.038)	(.037)
Time	−.471***	−.473***	−.456***
	(.006)	(.006)	(.006)
Age	−.133***	−.135***	−.133***
	(.003)	(.003)	(.003)
Female	.856***	.852***	.882***
	(.048)	(.048)	(.048)
Race/ethnicity (Ref.: White
Black	−1.851***	−1.823***	−1.779***
	(.078)	(.078)	(.077)
Hispanic	−.732***	−.687***	−.652***
	(.111)	(.111)	(.110)
Other	−1.280***	−1.248***	−1.229***
	(.183)	(.183)	(.180)
Foreign-born	.155	.169	.166
	(.097)	(.097)	(.096)
Years of education	.380***	.378***	.365***
	(.009)	(.009)	(.009)
Household income quartile (Ref.: First)
Second quartile ^	.302***	.299***	.267***
	(.035)	(.035)	(.035)
Third quartile ^	.398***	.395***	.353***
	(.039)	(.039)	(.039)
Fourth quartile ^	.489***	.484***	.446***
	(.044)	(.044)	(.044)
Employed ^	.112***	.114***	.064*
	(.032)	(.032)	(.032)
Urban residence ^	.161***	.170***	.173***
	(.042)	(.042)	(.041)
Married/partnered ^	.061	.036	−.020
	(.039)	(.039)	(.039)
Number of grandchildren ^	.013*	.013***	.013***
	(.004)	(.004)	(.004)
No living children ^	−.123	−.144	−.095
	(.197)	(.198)	(.196)
1+ child lives within 10 miles ^		−.000	.005
		(.028)	(.028)
Multigenerational living arrangements ^
Adult child without grandchild		−.178***	−.149***
		(.038)	(.038)
Three generation household		−.139*	−.121
		(.070)	(.069)
Skipped generation household		−.017	.014
		(.106)	(.106)
Other household member		−.231***	−.207***
		(.044)	(.045)
Depressive symptoms^			−.079***
			(.007)
ADL limitations^			−.409***
			(.017)
Intercept variance	4.830	4.824	4.693
Rate of change	−.154	−.153	−.145

Notes: Standard errors in parentheses.

^*=p<.05,

^**=p<.01,

^***=p<.001.

^{^}= Time-varying covariate.

Model 2 added time-varying multigenerational living arrangements to account for the presence of grandchildren in the household in a skipped or three generation household along with variables to account for the proximity of adult children. Those who lived with an adult child only or in a three-generation household or with any other than non-spousal household member had lower cognitive functioning scores than their counterparts who lived alone. We interpret this finding as suggesting that these individuals may have had cognitive or other health needs requiring the coresidential help of another adult. However, coresiding with a minor grandchild in a skipped generation household (i.e., without the presence of an adult child) was not significantly related to cognitive functioning, and accounting for household members did not impact the relationships between grandchild caregiving and cognition. Additional analyses (available in Supplemental Table 1) further show that interactions between grandchild caregiving and living arrangements are also not significant. Last, depressive symptoms and ADL limitations had strong negative relationships to cognitive functioning in Model 3, but did little to explain the positive links between grandchild caregiving and cognitive functioning.

As adults approach the end of their lives, their cognitive function may decline more quickly (MacDonald et al., 2011). HRS respondents who died during the study period are right-censored, meaning that they are dropped from the analysis at that timepoint, but are not restricted from the analyses. To explore impacts this design may have on the findings, analyses limiting the sample to those who survived over the whole study period are available in Supplemental Table 2. They show that while low levels of grandchild caregiving are not a significant predictor of cognitive functioning in the reduced sample of respondents who survived over the entire study period, medium and high levels of caregiving remain positive predictors. We also conducted supplemental analyses including respondents who never reported grandchildren (Supplemental Tables 3 and 4). The results show patterns that are substantively identical to the Table 2 analyses that exclude these respondents, suggesting that relationships between grandchild caregiving and cognitive functioning are not driven by individuals who become grandparents.

To provide a visual representation of the longitudinal relationships between cognitive functioning and grandchild caregiving, we plotted the predicted odds of scoring within the normal range (i.e., of having TICs scores 12+) from 2000 to 2016 by grandchild caregiver group in Figure 1. Since grandchild caregiving is time-varying, the figure represents odds of scoring in the normal cognition range based on respondents’ level of grandchild care at the different years. Variations across caregiving level are modest. However, non-caregivers stably had the lowest and medium grandchild caregivers stably had the highest odds of having a score indicating normal cognition. It also shows that the gap between caregivers and non-caregivers increased over time. By 2016, non-caregivers had under 70 percent odds of scoring in the normal cognitive functioning range, compared to over 80 percent of medium caregivers.

Figure 1:

Predicted Odds of Normal Cognitive Functioning by Level of Grandchild Caregiving with 95% Confidence Intervals

Note: Figure 1 is presented on a scale of .7 to 1 to support the visualization of differences across caregiving groups.

Table 3 summarizes the significant interactions between covariates and grandchild caregiving. Because the differences across level of care were small, as demonstrated in Table 2 analyses, and to simplify the presentation of the results, we used a dichotomous measure identifying those who provided 100 or more hours of grandchild care for these analyses. Model 1 shows that the relationship between grandchild caregiving and cognitive functioning increased over time, a pattern reflected in the growing differences between the grandchild caregivers and non-caregivers that were shown in Figure 1. Model 2 indicates that the relationship between providing grandchild care and cognitive functioning also increased with age. Other interaction terms suggest that the positive relationship between grandchild care and cognition is stronger for native-born, lower income, non-working, unpartnered, and more functionally limited adults than for their foreign-born, higher income, employed, partnered and less limited peers. Interactions between grandchild caregiving and race/ethnicity, gender, years of education, and urban residence were not significant. Analyses interacting all levels of care with these characteristics are available in Supplementary Table 5 show slight variations in statistical significance across levels of care, which are likely attributable to differences in the sizes of different grandchild caregiving groups across the study years. However, the direction and size of the coefficients are very similar across caregiving level.

Table 3:

Abbreviated results from linear growth models regressing cognitive functioning on significant interactions between grandchild caregiving and predictors, 1998–2016 (N=11,357)

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7
Grandchild caregiving ^	−.052	−.090***	.147***	.251***	.173***	.275***	.084**
	(.042)	(.235)	(.029)	(.054)	(.033)	(.051)	(.028)
Grandchild caregiving ^ * Time	.054***
	(.010)
Grandchild caregiving ^ * Age		.017***
		(.004)
Grandchild caregiving ^ * Foreign-born			−.260**
			(.095)
Grandchild caregiving ^ * Second income quartile^				−.189**
				(.070)
Grandchild caregiving ^ * Third income quartile^				−.195**
				(.070)
Grandchild caregiving ^ * Fourth income quartile^				−.100
				(.071)
Grandchild caregiving ^ * Employed ^					−.130*
					(.051)
Grandchild caregiving ^ * Married/partnered						−.203**
						(.058)
Grandchild caregiving ^ * ADL limitations							.227***
							(.038)

Notes: Standard errors in parentheses. All models control for time, age, gender, race/ethnicity, nativity, education, and time-varying partnership status, income, employment status, presence of living children, number of grandchildren, presence of at least one adult child within 10 miles, non-spousal household members, and random effects for time.

^*=p<.05,

^**=p<.01,

^***=p<.001

^{^}= Time-varying covariate.

Last, Table 4 presents coefficients from the linear dynamic panel models regressing cognitive functioning on grandchild caregiving. Model 1 includes baseline lags for cognitive functioning and grandchild caregiving as predictors, and thus predicts cognitive functioning scores from 2000–2016. Coefficients of these models are constrained to be equal across time (Williams et al., 2018), and thus show average effects of lagged variables on cognitive functioning from 2002 to 2016. Model 1 indicates that cognitive functioning two years earlier (t-1) is a strong predictor of later cognitive functioning. Grandchild caregiving two years earlier is also positively linked to cognitive functioning scores from 2000–2016. Accounting for these two lagged measures, grandchild caregiving also predicted higher cognitive functioning contemporaneously. In Model 2, cognitive functioning scores both two and four years earlier are strong predictors of cognition from 2002–2016. Grandchild caregiving four years earlier in 1998 but not in two years earlier had modest positive relationships with cognitive functioning from 2002–2016 when accounting for two lags of caregiving as well as cognitive functioning. A contemporaneous positive link between grandchild caregiving and cognitive functioning also remained when controlling for these lags.

Table 4:

Coefficients from linear dynamic panel model regressions of cognitive functioning scores on grandchild caregiving and covariates ^a

	Model 1 (N=11,211)	Model 2 (N=11,357)
Lagged cognitive functioning score
t-1	.151***	.172***
	(.005)	(.006)
t−2		.148***
		(.005)
Lagged grandchild caregiving
t−1	.092**	.027
	(.034)	(.038)
t−2		.071*
		(.034)
Contemporaneous grandchild caregiving ^	.109**	.092*
	(.039)	(.046)
Age in 1998	−.146***	−.136***
	(.003)	(.003)
Female	.656***	.478***
	(.046)	(.042)
Race/ethnicity (Ref.: White)
Black	−1.738***	−1.424***
	(.073)	(.068)
Hispanic	−.655***	−.525***
	(.103)	(.089)
Other	−1.118***	−.875***
	(.167)	(.146)
Foreign-born	.255**	.214**
	(.091)	(.081)
Years of education	.356***	.291***
	(.009)	(.009)
Household income quartile ^	.063***	.060**
	(.017)	(.019)
Employed ^	−.068	−.073
	(.037)	(.042)
Urban residence ^	−.051	−.088
	(.077)	(.087)
Married/partnered ^	.122	.112
	(.064)	(.072)
Any non-spousal household member ^	−.101	−.092
	(.057)	(.062)
Number of grandchildren ^	.029***	.022***
	(.005)	(.006)
No living children ^	−.436	−.427
	(.235)	(.279)
1+ child lives within 10 miles ^	.065	.047
	(.034)	(.039)
Depressive symptoms^	−.054***	−.046***
	(.008)	(.009)
ADL limitations^	−.360***	−.277***
	(.019)	(.021)

Note: Standard errors in parentheses.

^aModel 1 models cognitive functioning from 2000–2016, and Model 2 models cognitive functioning from 2002–2016.

^*=p<.05,

^**=p<.01,

^***=p<.001

^{^}= Time-varying covariate.

Discussion

Following a nationally-representative sample of U.S. adults age 50+ from 1998 to 2016, we found that those who provided from 100 to 500 or more hours of care to grandchildren in the last two years had higher cognitive functioning scores during this period than their peers who did not perform this care work. Addressing our first research question, this finding is consistent with the “use it or lose it” hypothesis that participation in everyday forms of social engagement—which may cover many types of interactions and activities, including caregiving--stimulates intellectual activity, particularly if these interactions are complex or demanding, helping to maintain older adults’ cognitive functioning (Hultsch et al., 1999; Seeman et al. 2001, 2011). Grandchild caregivers who provided an average of 2–5 hours of care per week appeared to fare the best, but both low (100–199 hours of care over two years) and high levels of care (500+ hours of care over two years) predicted better cognitive functioning. It is possible that medium levels of care offer the greatest benefits to congnitive functioning because grandparents spending less time in this role reap fewer of its rewards, while the benefits of high levels of grandchild caregiving may be offset by caregiving stressors. In addition, although we account for baseline and lagged measures of cognitive functioning and examine within-person change in cognition, it is not possible to fully account for the possibility of selection into different caregiving levels. For example, grandparents undertaking high levels of care may have been pushed into these circumstances out of family need rather than choice. However, because the differences across caregiving level are small, we interpret the results as suggesting that grandchild caregiving may have a modest positive impact on cognitive functioning regardless of level of care.

In response to our second research question and less consistent with our expectations, the positive association between grandchild caregiving and cognitive functioning was not explained by respondents’ multigenerational living arrangements. Interactions between grandchild caregiving and living arrangements also were not significant. The results thus indicate that providing care to grandchildren predicted better cognition regardless of whether grandparents lived apart from grandchildren, in a three generation household with adult children and grandchildren, or were likely custodial grandparents living in a skipped generation household without the presence of adult children.

These findings do not mean that no type of grandchild caregiving has negative impacts on older adults’ well-being. However, we interpret them as suggesting that the rewards of grandchild caregiving may ultimately outweigh the costs--such as caregiving stress or loss of time spent in other rewarding roles--for many grandparents with respect to cognitive functioning, just as a variety of research indicates it does for other health indicators (e.g., di Gessa et al., 2016a, 2016b; Giarrusso et al., 1996; Notter, 2022; Xu et al., 2017). Although this later-life caregiving work does not have the profound effects on cognitive functioning that social determinants which are considered to be fundamental causes of health do (including race/ethnicity, gender, and socioeconomic status; Phelan et al. 2010), our findings suggest they have meaningful time-varying impacts that are comparable to the effects of maintaining employment and greater than those of being partnered. Note that because grandchild caregiving is time-varying, the results do not reflect the full impact of grandchild caregiving on cognitive functioning over the course of an individual’s life. Research exploring summary measures of later-life grandchild caregiving activities will offer further insight about the scope of their meaning for long-term cognitive functioning trajectories.

Explanations for the positive relationship between cognitive functioning and grandchild caregiving include that it benefits grandparents’ mental and physical health. However, grandchild caregiving remained a robust predictor of cognition in models adding depressive symptoms and ADL limitations as controls, suggesting that these two health indicators do not account for much of the cognition-caregiving relationship in this study. Further examinations of the mechanisms through which grandchild caregiving may help maintain cognitive functioning are an essential task for future work on this topic. Along with other health and health behavior indicators, the literature suggests that improved social networks may be an important explanation for why spending time caring for grandchildren predicts better cognitive functioning (Arpino & Bordone, 2014; Jendrek, 1993; Notter, 2022; Quirke et al., 2021; Szinovacz & Davey, 2006). Future studies should examine the role of other kinship and social ties in predicting cognitive functioning among different grandchild caregiving groups.

One contribution of our study is that in addition to focusing on a longitudinal window of eighteen years, we used linear dynamic panel models to help further adjust for the selection of cognitively healthier adults into grandchild caregiving. These analyses revealed that controlling for cognition scores in the wave prior, providing care to grandchildren was a significant predictor of higher cognitive functioning scores. Analyses including two lags of cognitive functioning and grandchild caregiving show that grandchild caregiving four years earlier predicted somewhat better cognitive functioning, taking into account differences in caregiver and non-caregiver cognition scores in the two prior waves. Caregiving in the most recent wave was not a significant predictor of cognitive functioning in models with two lags. Although the difference in lagged coefficients is small and we do not wish to over-interpret this finding, it is possible that the benefits can take some time to accumulate. Contemporaneous positive relationships between grandchild caregiving and cognitive functioning also persisted in the model with two lags. In other words, adults age 50 and up who spent at least 100 hours looking after grandchildren over a two-year period had better cognitive functioning both at the time they were providing care and two to four years down the road than their peers who did not engage in this care work. Although further research to clarify the causal direction of the relationship between cognitive functioning and grandchild caregiving is needed, these results further support the hypothesis that grandchild caregiving may help maintain cognitive functioning from mid- to later-life.

Variations Across Sociodemographic Groups

Addressing our third research question, we assessed whether the relationship between grandchild caregiving and cognitive functioning differed across adults’ sociodemographic, family, work, and health characteristics. Heterogeneity in the relationship between grandchild caregiving and cognition is underexplored in prior work, despite significant diversity in the structure and meaning of intergenerational family relationships across sociodemographic groups, including grandparenting. Unlike some prior studies (Chen et al., 2015; Sneed & Schulz, 2019), we did not find significant differences across race and ethnicity. However, our results indicated that the positive association between grandchild caregiving and cognitive functioning grew with age and the acquisition of ADL limitations, and that it was greater for lower income, unpartnered and not working respondents than for higher income, partnered, and employed counterparts. Overall, these findings provide support for our hypothesis that providing care to grandchildren may especially benefit groups of older adults who are more vulnerable to cognitive decline or who may have smaller social networks. They also run counter to the hypothesis that disadvantages are more likely to accumulate over the life course of disadvantaged groups (O’Rand, 1996). However, it is important to note that care work can also be accompanied by material and emotional burdens, and is more likely to fall on certain disadvantaged groups, including women and racial/ethnic minority older adults. For example, we also found that the link between cognitive functioning and grandchild caregiving was weaker for the foreign- than native-born individuals. While this pattern deserves further investigation, possible explanations include that foreign-born grandparents have fewer resources to draw on to cope with caregiving demands and that grandchild caregiving is more often driven by obligation or economic need among foreign- than native-born grandparents.

Limitations and Future Research

Although it advances prior work on this topic in several innovative ways, our study also has limitations. Because we use study wave as the time scale, it is not possible to distinguish the effects of aging on cognitive functioning from period and cohort effects. Future analyses using age as the time scale will offer additional insights about how grandchild care work may impact cognitive functioning as adults age over different historical time periods.

There are also limitations to the measurement of grandchild caregiving in the HRS. One consequence of the way that these questions are asked is that we cannot explore linear relationships between time spent caring for grandchildren and cognitive functioning. Grandparents providing less than 100 hours of childcare are considered to be non-caregivers, although they may be providing some care. In addition, the levels of caregiving built into the way these questions are asked translate to just about 1–2, 2–5, and 5+ hours of grandchild caregiving per week. These time variations are substantively quite small, which likely helps explain why we found few differences across levels of grandchild care. Five hours of care a week is also arguably still a low amount of time spent in this role. Because the HRS does not differentiate among those providing 500 or more hours of care, our study does not capture the distinct experiences of those providing regular caregiving that may amount to several days a week or daily care. Research leveraging other data sources should examine the cognitive functioning impacts of these very high levels of care, which may have different implications for grandparents’ well-being.

In addition, while we were able to identify multigenerational and skipped generation households through household rosters, grandchild caregiving questions were asked separately. We thus do not know for certain whether caregiving grandparents were looking after minor grandchildren they lived with or if those living in skipped generation households were custodial grandparents, and research shows that the well-being impacts of grandchild caregiving may be greatest for those living in three-generation households (Dunifon et al., 2020). We also do not have information about how many grandchildren respondents were caring for, how frequently care was provided, the ages of grandchildren that respondents were providing care to, or the kinds of activities they had in mind when responding to the question about “taking care” of grandchildren. The tasks that grandparents think of no doubt vary widely across grandchildrens’ ages and individual perceptions, and could range from watching television with teenage grandchildren to helping with homework to changing and feeding an infant. Future research may consider exploring grandparents’ interpretations and definitions of caregiving through cognitive interviews, and should examine whether the relationship between grandchild caregiving and cognitive functioning depends on grandchildrens’ ages. Another area for future research is to consider the effects of transitions into and out of grandchild caregiving, which are not captured with our time-varying measure.

Conclusion

Despite the need for further investigation, our study adds to a growing literature suggesting that some groups of U.S. grandparents may benefit from time spent looking after their grandchildren. As indicated by theory and prior research (e.g., Dunifon et al., 2020; Milkie et al. 2008; Pillemer et al., 2019), caregiving is a complex and multifaceted experience that is accompanied by costs as well as rewards. However, our analyses suggest that providing care to grandchildren predicts higher cognitive functioning regardless of living arrangements and whether care was provided an average of 100–199 to more than 500 hours over a two year period. Interpreting these results within the context of the larger research literature, which provides conflicting evidence about how grandchild caregiving impacts well-being, it is possible that although more time-intensive or custodial grandparenting can be stressful and increase psychological distress in the short-term, the multilevel emotional, social, and physiological rewards of this care work tend to outlast the costs for many grandparents when it comes to cognitive health. Additionally, our results suggest that the potential cognitive functioning benefits of grandchild caregiving may be the most profound for groups who are more vulnerable to cognitive decline and the negative impacts of social role loss, such as those who are lower income, older, retired, widowed, or declining physically. Together, these findings contribute to a growing body of research indicating that meaningful social interactions and relationships can act as a buffer against age-linked cognitive decline.