Choice and Constraint in the Negotiation of the Grandparent Role: A Mixed-Methods Study.

Few studies have examined how the allocation and consequences of grandchild care vary across different socioeconomic groups. We analyze qualitative data alongside data from The Irish Longitudinal Study on Ageing (TILDA), in a convergent mixed-methods approach. Regression models examined characteristics associated with grandchild care, and the relationship between grandchild care and depressive symptoms and well-being. Qualitative data shed light on processes and choices that explain patterns of grandchild care provision. Tertiary-educated grandparents provided less intensive grandchild care compared with primary educated. Qualitative data indicated that this pattern stems from early boundary-drawing among higher educated grandparents while lower socioeconomic groups were constrained and less able to say no. Intensive grandchild care was associated with more depressive symptoms and lower well-being and was moderated by participation in social activities and level of education attainment. The effect of grandchild care on well-being of grandparents depends on whether it is provided by choice or obligation.

Introduction

The provision of care by grandparents to grandchildren is widespread and varies by country, state provision of child care infrastructure, social provision of welfare payments, and cultural expectations (Di Gessa, Glaser, Price, Ribe, & Tinker, 2016; Hank & Buber, 2009; Igel & Szydlik, 2011; Timonen, 2018). European grandchild care is common, ranging from 37% to 59% and has been found to be highly associated with public expenditure on child care infrastructure and employment rates of both younger and older women (Di Gessa et al., 2016). Working mothers were particularly predictive of high intensity grandchild care and more common in countries with limited provision of child care (Di Gessa et al., 2016; Ko & Hank, 2014). This provision of informal grandchild care for working parents is also common in Asia and the United States ranging from 40% to 60% (Fuller-Thomson & Minkler, 2001; Ko & Hank, 2014; Thiele & Whelan, 2006, 2008; Yoon, 2005). The adoption of austerity measures in response to the Great Recession in Ireland placed further emphasis on individual and family responsibility (Dukelow & Considine, 2014). This has been marked in the area of child care, and child care costs in Ireland now constitute the highest proportion of average earned incomes in the European Union, leading to widespread use of informal care solutions including grandparents (Dukelow & Considine, 2014). Grandparents were an important resource for lower income families and younger mothers (McNally, Share, & Murray, 2014), who provided substantial intergenerational transfers of both money and time (McGarrigle, Cronin, & Kenny, 2014).

Grandparents’ time and resources are viewed as the ideal response to growing structural needs for child care, especially in the context of welfare states that have low provision of formal child care (Igel & Szydlik, 2011), and in family contexts where circumstances of the middle generation (e.g., lone parenthood, divorce, low income) render them less able to arrange (pay) for child care without assistance from extended family (McNally et al., 2014). In this context, grandparents have been branded as “child savers” (where they replace the parenting functions of the middle generation by becoming custodial grandparents); “mother savers” (where they enable the mother to work outside the home); “family savers” (where they step in occasionally, when, for example, the grandchild is sick and both parents need to work); and even as “family maximizers” (where they enable both parents to be absent for long periods of time to maximize the extended family’s income (Baker & Silverstein, 2012; Herlofson & Hagestad, 2012)). The consequences for grandparents themselves are usually measured in terms of a narrow range of their health outcomes, and well-being outcomes for the grandchildren (Tanskanen, 2013). The literature in modern welfare states focuses on grandparenting as an opportunity structure that is open to growing numbers of older people, for longer periods of time, with fewer grandchildren who can enjoy more attention from older family generations (Igel & Szydlik, 2011).

Grandchild care enables social engagement and an active lifestyle; the importance of both for sustaining good health and cognition in older people is well recognized (Burn & Szoeke, 2015; Grundy et al., 2012; Holtzman et al., 2004). However, the effects of grandparenting on health vary by cultural, ethnic, education, and socioeconomic group. Reports on noncustodial grandparental care have found both positive (Chen & Liu, 2012; Hughes, Waite, LaPierre, & Luo, 2007) and negative health effects (Chen & Liu, 2012; Hughes et al., 2007; Lee, Colditz, Berkman, & Kawachi, 2003; Tsai, Motamed, & Rougemont, 2013). A positive association between nonintensive grandparenting and better health outcomes has been found across 11 European countries (Glaser, Di Gessa, & Tinker, 2014), Australia (Burn & Szoeke, 2015), and in both China and Taiwan (Chen & Liu, 2012; Tsai et al., 2013). However high intensity grandchild care was associated with poorer health outcomes (Burn & Szoeke, 2015). There is relatively little literature on the mental health impact of lower intensity grandchild care; although it remains more common, one recent study suggesting that nonintensive grandchild care has a positive effect on grandparents health, independent of their previous health and socioeconomic group and childhood and adulthood experiences of advantage and disadvantage (Glaser et al., 2014). Furthermore, 4 or more hours of grandchild care a week was associated with improved well-being in grandfathers and lower depression scores in grandmothers in Chile (Grundy et al., 2012).

Overall, the literature and policy discourses appear to suggest that grandparental child care is potentially a win-win situation for the grandparents (who come to enjoy better health as a result of care inputs), the grandchildren (who benefit from the time and resources of two generations of adults), and welfare states (that save on formal care costs, and benefit from improved population health and well-being outcomes in both child and older populations). We hypothesize that this win–win situation may not always be perceived as such by grandparents themselves; many might end up providing grandchild care out of solidarity toward their children, even where it would not be their own preference to do so. It is also possible that the heaviest inputs into grandchild care are being made by the grandparents with the least resources (money, health, etc.), rendering the attempts to further tap into this “free” resource inequitable. There has been relatively little examination of the reasons why the allocation of grandparenting duties varies across different socioeconomic groups and its consequences for the health and well-being of older people delivering varying amounts of grandchild care. A systematic review found that none of the theoretical approaches reviewed explained the differences in well-being by duties that emerged (Kim, Kang, & Johnson-Motoyama, 2017). Although the typology seeks to be sensitive to different characteristics of grandparents (Cherlin & Furstenberg, 1986), we strive to further develop understanding of a relatively underresearched aspect, namely, socioeconomic group. Differences in the distribution of grandparenting duties by socioeconomic group are important and may represent differential expectations, constraints, and levels of choice in personal time use. Understanding such differences in grandparents’ characteristics and consequences of grandchild care is important also from the point of view of policy planning, not in the least because some might be under increasing pressure to provide grandchild care (against their own preferences), especially if policy instruments are used to incentivize such arrangements.

The study investigated the association between different levels of intensity (low: <60 hr and high: 60+ hr) of grandchild care provision and mental health (depressive symptoms) and well-being (quality of life). Using a convergent mixed-methods approach (Creswell, 2015), we compared and integrated findings from a large nationally representative study with qualitative data. It combined data from two recent qualitative projects conducted in Ireland (Grandparenting in Divorced and Separated Families; Changing Generations) with data from a large nationally representative study of the older population of Ireland, The Irish Longitudinal Study on Ageing (TILDA). Drawing on TILDA data, we asked three questions: Which socioeconomic group were most likely to provide intensive grandchild care? Is there a relationship between mental health and grandchild care provision? Do leisure and social activities moderate the effects on mental health and well-being outcomes? Drawing on the qualitative studies, we asked the following questions: Are there significant differences in the degree of choice that grandparents of different socioeconomic groups have exercised in the process of becoming involved, or not, in the care of their grandchildren? How do grandparents providing higher and lower levels of grandchild care view their capacity to engage in other pursuits, in particular social leisure activities? The mixed-method analysis asked the question, “What are the processes and choices from the qualitative data that help to provide a deeper understanding of the patterns and consequences of grandchild care provision on mental health and well-being seen in the quantitative data?”

Method

Quantitative Study

TILDA is a nationally representative cohort study of 8,504 people aged over 50 years in Ireland (www.tilda.ie). This study is based on Wave 1 (carried out: October 2009-March 2011). TILDA assessed the social, economic, and health characteristics; lifestyles; behaviors and attitudes of older adults resident in Ireland. The methodological details have been reported elsewhere (Kearney et al., 2011), but briefly, participants were interviewed by trained interviewers in their own home using computer-assisted personal interviewing (CAPI), in addition to a Self-Completion Questionnaire (SCQ) consisting of more sensitive questions, returned by mail. A multistage probability sample of addresses was chosen (Whelan, 1979). All household residents aged over 50 years, and their spouse/partner were eligible to participate. The response rate at Wave 1 of the survey was 62% and 84% returned the SCQ. This study was approved by the Faculty of Health Sciences Research Ethics Committee of Trinity College Dublin, and participants provided written informed consent prior to participation in the study.

Outcome variables

We defined grandparents as all respondents, reporting grandchildren. A grandchild care variable was created from numbers of hours of grandchild care reported in the past month in response to the following questions: “Have you spent at least 1 hour a week taking care of grandchildren or great-grandchildren (who live outside your own household)? About how many hours on average per month did you spend taking care of your grandchildren or great-grandchildren (who live outside your own household)?” Grandchild care was categorized as follows: no grandchild care, 1 to 59 hr (low intensity), and 60+ hr per month (high intensity). More than 60 hr was defined as high intensity based on previous research (Glaser, Evandrou, & Tomassini, 2005).

Mental health

Depressive symptomology was measured with the 20-item Center for Epidemiological Studies Depression (CES-D; Beekman et al., 1997) scale. Each item requests a frequency score from never to almost all the time (McGarrigle et al., 2014). A total score was calculated by summing responses across the 20 items (range = 0-60, Cronbach’s α = .94). Scores >16 indicate clinically significant depressive symptoms. Quality of life was measured using the Control, Autonomy, Self-realization, Pleasure-19 (CASP-19) scale, a brief (19-item) self-inventory designed to measure quality of life independent of the factors that might influence it such as health, social supports, and material circumstances (Hyde, Wiggins, Higgs, & Blane, 2003). Each statement requests a 4-point frequency scale from often to never and scores for each of four subscales: Control, Autonomy, Self-realization, and Pleasure were calculated. A total quality of life score was calculated by summing scores across the four subscales, a higher score indicating a higher quality of life (range = 0-57, Cronbach’s α = .88).

Covariates

A range of covariates associated with both grandchild care and mental health were included. These included sociodemographic variables such as age, gender, marital status, years of education (primary: 8 years, secondary: 14 years, or tertiary: 15+ years), household net income (quintiles), and employment status (employed, retired, and other—includes unemployed, permanently sick or disabled, looking after home or family, in education or training, and other). Education was used as a measure of socioeconomic group in the older population in Ireland (Winkleby, Jatulis, Frank, & Fortnum, 1992), as income for the adult children, which would drive need, was unknown. Disability was assessed by asking about the basic tasks of everyday life that pertain to personal care, activities of daily living (ADL). Health professionals often use these activities as a measure of disability or functional status. ADL limitations were a count of the number of six activities the respondent had difficulty performing: walking across a room, dressing, bathing, eating, getting in and out of bed, and using the toilet. Instrumental activities of daily living (IADL) included difficulty preparing a meal, doing household chores, shopping, making telephone calls, taking medications, and managing expenses. Disability was grouped into a categorical variable if participant responded yes to any ADL or IADL question (no disability, IADL disability only, ADL disability). Participants were asked whether they had participated in an active social activity outside the home in the last week, including a range of activities like going to the cinema, eating out, participating in sports activities. This was then generated into a binary variable (any active social participation in the last week: yes, no). Family structure included total numbers of children (continuous variable), and numbers of grandchildren (continuous variable).

Statistical analysis

Generalized ordered logistic regression models with grandchild care as a categorical outcome were used to examine sociodemographic, family, and health characteristics associated with intensity of grandchild care. Proportionality of odds was violated on ordinal logistic regression and using autofit, a partial proportional odds model was fitted, which allowed covariates that met the proportional odds assumption to affect different levels of intensity of grandchild care with the same magnitude while employment, IADL disability, and numbers of grandchildren (which violated the proportionality assumption) had different effects on levels of intensity in the model. To take account of possible nonindependence within households, we relaxed the independence assumption within households, while assuming independence between households, to produce robust standard errors using VCE (CLUSTER) command in STATA. We then examined mean CES-D by grandchild care, using linear regression, with CES-D as a continuous outcome, with robust standard errors adjusting for age, gender, education, employment, marital status, household income, and social activity. These analyses were repeated with quality of life (CASP-19) as a continuous outcome.

Interaction terms for Grandchild Care × Education, Grandchild Care × Age, Grandchild Care × Social Activity, and Grandchild Care × Disability were created to test whether any of these aspects of age, social activity, or disability moderated the relationship of grandchild care with mental health. To allow for the possibility that age effects might not be linear and moderated by sex, we included a quadratic term for age, and an interaction term for age, age2, and sex. A Wald test was used to determine whether the addition of interaction terms improved model fit over the main effects models. Significant interaction terms were examined and plotted using MARGINS and MARGINSPLOT commands in STATA, presenting the linear predicted values of CES-D and CASP-19, at each discrete level within each variable, for each interaction term. The results of the regression analyses were reported as β coefficients with 95% confidence intervals (CIs) errors. We weighted to account for noncompletion of the SCQ CASP-19 questions to be comparable with the CES-D analysis (collected on main questionnaire) using inverse probability weights, generated using logistic regression of the probability of having completed the SCQ adjusted for all covariates of interest. All analyses were conducted in STATA 12.

Qualitative Data

The qualitative analysis was based on two datasets. For the “Grandparents in Divorced and Separated Families” study conducted in 2008-2010, respondents were recruited through support groups and advice centers for divorced and separated adults, lone parents’ support groups, and older people’s groups (yielding 15 respondents); advertisements were also placed in the main national newspaper (The Irish Times) and newsletters of the above organizations (yielding 16 respondents). Salient factors identified in the literature (gender, lineage) guided the sampling and recruitment process. However, due to the difficulties in recruiting grandfathers, the majority of participants (23 out of 31) were grandmothers (Timonen, Doyle, & O’Dywer, 2010). As paternal grandparents are in a disadvantaged position in Ireland’s family law system, paternal grandparents were more motivated to participate, amounting to 19 of the 31 participants.

The second qualitative study, “Changing Generations,” investigated intergenerational solidarity in Ireland, and comprised 100 in-depth qualitative interviews with people aged 18 to 102, conducted in 2011-2013 (Conlon, Timonen, Carney, & Scharf, 2014; www.icsg.ie). Thirty-nine of the 44 older adults (aged 50-102) in the sample were grandparents. Seventeen of these 39 grandparents gave lengthy and detailed accounts of their grandparenting practices. In total, the qualitative data comprise 48 detailed accounts of grandparenting practices in contemporary Ireland. These were analyzed together to explicate the patterns identified in the quantitative data analysis presented above, using thematic coding (Flick, 2014). The data extracts that feature in the article are quotes from participants whose words best illustrate broader processes discovered inductively in the data.

Results

Quantitative Data

About 4,438 TILDA respondents (52.2%) were grandparents, and 59.0% (95% CI = [57.5, 60.4]) had taken care of grandchildren in the last month: 58.7% (95% CI = [57.1, 61.6]) of men and 58.7% (95% CI = [56.8, 60.5]) of women. The demographic, health, and family structure characteristics of grandparents and intensity of grandchild care given are presented in Table 1. The distribution of the numbers of hours spent caring for grandchildren in the past month are presented in Supplementary Figure 1. A quarter of the grandparents were in paid employment and one in five had a tertiary-level education. A high proportion of all grandparents had taken part in a social activity in the past week and the majority of grandparents had no disability.

Table 1.

Characteristics of Grandparents by Intensity of Care Provided for Grandchildren.

Characteristics	Total grandparents[a]	No grandchild care	Grandchild care1 to 59 hr per month	Grandchild care60+ hr per month
	% (95% CI)	% (95% CI)	% (95% CI)	% (95% CI)
%	100	41.8 [39.6, 42.5]	50.3 [48.9, 51.8]	8.6 [7.8, 9.5]
Base	4,536	1,821	2,234	383
Grandchild care hours
M (SD)			16.9 (13.5)	109.9 (67.3)
Age
M (SD)	66.8 (9.4)	70.1 (10.1)	65.0 (8.2)	61.8 (7.3)
Retirement age
<65	36.3 [34.8, 37.8]	26.2 [24.1, 28.5]	40.6 [38.5, 42.9]	56.7 [51.2, 62.1]
≥65	63.7 [62.2, 65.2]	73.8 [72.0, 75.9]	59.4 [57.1, 61.5]	43.3 [38.0, 48.8]
Gender
Male	41.1 [39.7, 42.5]	40.6 [38.4, 42.9]	41.7 [39.6, 43.7]	39.7 [34.9, 44.7]
Female	58.9 [57.5, 60.3]	59.4 [57.1, 61.6]	58.3 [56.3, 60.4]	60.3 [55.3, 65.1]
Education
Primary	38.1 [36.7, 39.5]	40.6 [38.4, 42.9]	36.7 [34.7, 38.8]	36.0 [31.4, 41.0]
Secondary	39.4 [38.0, 40.9]	36.1 [33.9, 38.4]	40.2 [38.2, 42.3]	48.0 [43.1, 53.1]
Tertiary	22.5 [21.3, 23.7]	23.3 [21.4, 25.2]	23.2 [21.4, 24.9]	15.9 [12.6, 19.9]
Employment
Employed	24.1 [22.9, 25.3]	18.3 [16.6, 20.2]	27.8 [25.9, 29.6]	27.9 [23.7, 32.7]
Retired	45.2 [43.7, 46.6]	51.9 [49.7, 54.2]	42.3 [40.3, 44.4]	30.3 [25.9, 35.1]
Other	30.7 [29.4, 32.1]	29.7 [27.7, 31.9]	29.8 [28.0, 31.7]	37.3 [32.2, 42.8]
Income quintile
Lowest	19.9 [18.7, 21.1]	24.4 [22.5, 26.5]	17.1 [15.6, 18.7]	16.4 [13.1, 20.5]
2nd	23.0 [21.8, 24.2]	23.7 [21.8, 25.7]	22.7 [21.0, 24.0]	23.5 [19.5, 28.0]
3rd	18.7 [17.6, 19.8]	18.6 [16.9, 20.5]	17.5 [16.0, 19.1]	23.2 [19.3, 27.7]
4th	20.6 [19.5, 21.8]	17.4 [15.7, 19.2]	22.9 [21.2, 24.7]	20.6 [16.9, 25.0]
Highest	8.2 [7.4, 9.0]	6.3 [5.3, 7.5]	9.8 [8.6, 11.1]	8.6 [6.2, 11.9]
Missing	9.6 [8.8, 10.6]	9.6 [8.3, 11.0]	10.1 [8.9, 11.4]	7.3 [5.1, 10.4]
Social/leisure activity[b]
Yes	88.0 [86.9, 89.4]	84.2 [82.3, 86.0]	89.5 [88.0, 90.8]	90.2 [86.5, 93.0]
Family structure
Total number of children, M (SD)	3.9 (1.9)	4.0 (2.0)	3.9 (1.8)	3.8 (1.8)
Total number of grandchildren, M (SD)	5.6 (5.0)	6.0 (5.6)	5.5 (4.6)	4.7 (4.3)
Health
Depression, CES-D, M (SD)	6.0 (7.3)	6.3 (7.2)	5.8 (7.3)	6.4 (8.0)
Quality of Life, CASP-19,[c] M (SD)	44.5 (7.6)	43.7 (7.8)	45.1 (7.2)	44.2 (8.1)
Disability
No disability	85.5 [84.4, 86.5]	80.4 [78.5, 82.2]	88.7 [87.3, 90.0]	88.8 [85.2, 91.6]
Instrumental activities of daily living	4.2 [3.6, 4.8]	6.4 [5.4, 7.6]	2.6 [2.0, 3.3]	3.7 [2.2, 6.1]
Any activities of daily living	10.3 [9.4, 11.2]	13.2 [11.7, 14.8]	8.7 [7.6, 10.0]	7.6 [5.3, 10.7]

Note. CI = confidence interval; CES-D = Center for Epidemiological Studies Depression; CASP-19 = Control, Autonomy, Self-realization, Pleasure.

Includes 98 who reported caring for grandchildren but answered not known for number of hours.

Any social or leisure activity outside the house in the past week.

Quality of life (Control, Autonomy, Self-realization, Pleasure) scale.

Table 2 shows the results from the unconstrained partial proportional odds ordinal regression analyses of the sociodemographic, family, and health variables associated with grandchild care. Factors associated with low intensity (<60 hr in the past month) and high intensity grandchild care (60+ hr in the past month), relative to no grandchild care, are presented. Provision of both low intensity and high intensity grandchild care was associated with being older, married, and lower educational attainment. High intensity grandchild care was associated with employment classified as “Other” (neither employed nor retired), relative to low intensity and no care. Low intensity grandchild care was associated with increased numbers of grandchildren and being healthier relative to providing no grandchild care.

Table 2.

ORs for Unconstrained Partial Proportional Odds Model of the Sociodemographic, Family, and Health Variables Associated With Grandchild Care[a]: The Irish Longitudinal Study on Ageing.

	No child care versus 1-59 hr	No child care or 1-59 hr versus 60+ hr
	OR (95% CI)	OR (95% CI)
Sociodemographics
Age (years)	1. 29 [1.17, 1.43]***	1.29 [1.17, 1.43]***
Age2 (years)	1.00 [1.00, 1.00]***	1.00 [1.00, 1.00]***
Women[b]	1.04 [0.94, 1.14]	1.04 [0.94, 1.14]
Not married[c]	0.55 [0.47, 0.65]***	0.55 [0.47, 0.65]***
Education[d]
Secondary	0.98 [0.84, 1.14]	0.98 [0.84, 1.14]
Tertiary	0.73 [0.60, 0.87]*	0.73 [0.60, 0.87]**
Employment[e]
Retired	1.13 [0.94, 1.37]	1.13 [0.93, 1.37]
Other	0.97 [0.80, 1.17]	1.51 [1.19, 1.94]**
Net income (quintile)[f]
2nd	1.02 [0.82, 1.26]	1.02 [0.82, 1.26]
3rd	0.97 [0.77, 1.23]	0.97 [0.77, 1.23]
4th	1.13 [0.89, 1.42]	1.13 [0.89, 1.42]
Highest	1.20 [0.90, 1.62]	1.20 [0.90, 1.62]
Not recorded	1.00 [0.78, 1.30]	1.00 [0.78, 1.30]
Family structure
Increasing numbers of children	0.97 [0.92, 1.02]	0.97 [0.92, 1.02]
Increasing numbers of grandchildren	1.05 [1.02, 1.07]***	1.00 [0.97, 1.04]
Health[g]
IADL disability only	0.53 [0.38, 0.73]***	0.96 [0.54, 1.72]
Any ADL disability	0.83 [0.67, 1.02]	0.83 [0.67, 1.02]
Constant component of odds ratios at cutoff points	−0.0061505 [0.00021, 0.17848]	−0.0003342 [0.000116, 0.009611]
Wald	437.39
n	4,431
R 2	.080

Note. SE adjusted for 3,280 clusters in household. OR = odds ratio; CI = confidence interval; IADL = instrumental activities of daily living; ADL = activities of daily living.

Reference: No grandchild care.

Reference: Men.

Reference: Married.

Reference: Primary.

Reference: Lowest quintile.

Reference: 1-2.

Reference: No disability.

The p values of generalized ordered logit model: *p < .05. **p < .01. ***p < .001.

Quality of Life by Grandchild Care

Linear regression models of quality of life measured through CASP-19 score showed that grandparents who provided higher levels of grandchild care had increased quality of life (see Supplementary Table 1 for tabulated terms from regression model); however, this effect was moderated by educational attainment and social and leisure activity, F = 4.02, p = .003. Linear predictions of CASP-19 scores for the three levels of grandchild care by education level and social activity are plotted in Figure 1.

Figure 1.

Regression line of predicted mean quality of life (CASP-19)a and intensity of grandchild care by educational attainment and social and leisure activity: The Irish Longitudinal Study on Ageing.

Note. Adjusted for age, age2, gender, marital status, employment, income, and disability; F(30, 2363) = 19.84, p < .001, R2 = .181. SE adjusted for 2,364 clusters in households and weighted for SCQ completion. SCQ = Self-Completion Questionnaire. CASP-19 = Quality of life (Control, Autonomy, Self-realization, Pleasure) scale.

aMarginal means from multivariate linear regression model.

Figure 1 shows that lower educated grandparents who provided high intensity grandchild care and did not participate in a social or leisure activity had lower quality of life compared with those providing either no or low intensity grandchild care. In contrast, tertiary-educated grandparents carrying out high intensity grandchild care with no other social activity had higher quality of life compared with both low intensity grandchild carers and noncarers. Quality of life increased for secondary-educated grandparents who provided low intensity grandchild care in the past month, compared with no grandchild care, but only for those grandparents who also took part in a social and leisure activity. These differences were not seen for those grandparents who had participated in a social or leisure activity in the past month in the other educational attainment groups.

Depressive Symptoms by Grandchild Care

Linear regression models of depressive symptoms measured through CES-D score showed that grandparents who provided higher levels of grandparent care experienced significantly more depressive symptoms (see Supplementary Table 2 for tabulated terms from regression model). This effect was moderated by participation in social or leisure activity in the past month and interacted with level of education attainment, F = 2.60, p = .035. Linear predictions of mean CES-D score from this three-way interaction are shown in Figure 2. The effect was similar for both men and women. Figure 2 shows that lower educated grandparents who provided high intensity grandchild care and did not participate in a social or leisure activity had increased depressive symptoms compared with lower educated grandparents who provided either low intensity or no grandchild care.

Figure 2.

Regression line of predicted mean depressive symptoms (CES-D)a and intensity of grandchild care by educational status and leisure and social activity status: The Irish Longitudinal Study on Ageing.

Note. Adjusted for age, age2, gender, educational attainment, marital status, employment, income, and disability; F(30, 2726) = 14.02, p < .001, R2 = .149. SE adjusted for 2,727 clusters in households. CES-D = Center for Epidemiological Studies Depression.

aMarginal means from multivariate linear regression model.

This difference in levels of depressive symptoms between high and low intensity grandparent care was not seen in grandparents who had higher educational attainment. There was no difference in CES-D score between the grandparents who had taken part in a leisure or social activity, across all educational attainment groups.

Grandparents from higher socioeconomic groups tend to be more strongly oriented to (and could afford) social and leisure (“third age”) activities that are incompatible with time-intensive grandparenting. (This position is partly facilitated because their adult children also tend to be higher earners and hence better able to pay for formal child care.). They engage in early “boundary drawing” around the extent of their involvement in grandchild care. Lorna, aged 79, the wife of a retired businessman, recounted how she had clearly signaled her preference for leisure, social, and educational pastimes to her children:

I couldn’t have [cared for grandchildren] because I had too many things going on in my own life at the time . . . My days are pretty well planned out that I wouldn’t have been available on long term. They [my children] would know that.

In response to a question about his availability for occasional grandchild care, Hugo (aged 86, retired manager) laughed when he indicated that this depends on his other activities: “Yes, if I am available because as you know—with my activities I am busy” (his activities included playing bridge, going to theater and concerts, acting as treasurer in a business club, voluntary work in a hospice, socializing and going to sports matches). Conor, a retired lawyer, expressed delight both at being able to spend time with his grandchildren over a Sunday lunch, and being able to say, “normally after about an hour and a half, ‘Well, that’s grand, time to leave now.’” Brendan (67, retired teacher), who offers occasional babysitting together with his wife, recounted how this places them in the “lucky” category of grandparents who are able to cultivate closeness at a suitable distance:

We are fortunate that they [adult children] are getting on with their lives. They are fortunate to be working. They have a good home. They have good careers and a nice standard of living but at the same time, they need a break. We are just involved with them in a family sense and it is good to be part of that and just to be supportive.

Rebecca, an entrepreneur in her late 50s, drew a pointed distinction between the grandparents whom she saw as “exploited” and her own careful boundary-drawing:

I can choose basically when I want to see the [grand]kids . . . and my daughter is great, she never imposes . . . I wouldn’t do [grandchild care] on a daily basis. I mean I would put my foot down, I wouldn’t do it . . . I’ve got my own life to lead, thanks.

Elsie, a grandmother in her early 60s and employed in administrative role, had at first become heavily involved in the care of her son’s children when his marriage broke up. Elsie had become the “bridge” between her son and his children, always facilitating their contact and providing the physical space (her home) and catering for the occasions (Sunday dinners, birthdays) around which the contact revolved. However, over time, Elsie had become tired of this role, and had started adopting strategies, with the view to limiting her role in the lives of her grandchildren, including the decision to allow contact with another grandchild to diminish radically:

I’m just sick of kids . . . I hate them now and that’s the truth [laughs] . . . I stepped in there with good intentions . . . because I was thinking of the good of the grandchild but . . . I won’t do that again . . . my other son split up . . . I don’t see that child. I certainly wouldn’t go to the same extremes to see her . . .

Here, we are witnessing a “learning process” where excessive involvement over time leads to withdrawal. The data contain several other examples of such “pulling back,” for instance, Maeve who stated that she had “compensated for lack of parenting” but is “not doing that any longer,” and Sinead who disclosed that “at the very beginning we [she and husband] did plan our lives around it . . . but you can only do that for so long . . . now we don’t plan our lives around it.” We identified this as another type of grandparental agency: calibration over time to levels that are more manageable and in line with competing preferences, including more time for oneself. Overall, therefore, the grandparents from higher socioeconomic groups derived enjoyment from the time they spent looking after their grandchildren; we theorize that this is in large part because they were in most cases able to choose the level of their inputs. Most higher educated grandparents felt they had complete control over their level of involvement in grandchild care, and most participants in this category had chosen relatively low levels of involvement. However, there was a subsection of highly educated grandparents who chose to be very involved and derived considerable satisfaction from their inputs, as can be seen in the example of Malachy and Joanne, retired teachers, who together minded their granddaughter from 7:00 a.m. until 7:00 p.m., 2 days a week:

We hated the idea of the little one being put into a crèche . . . it’s in our value system . . . and [the granddaughter’s] language is really good, developed.

She has activities here, our kitchen is like a playschool with toys.

We do art, we do reading, we do like to read books, we have fun, we go for walks, we go off on the train, we have chats.

In contrast to these patterns among grandparents from higher socioeconomic groups, lower socioeconomic groups are constrained by structural pressures to provide grandchild care and are less able to “say no.” Rose, aged 60, retired housekeeper, had become the full-time carer for her grandson when her daughter gave birth at 18 and continued to pursue education and work opportunities:

I remember saying to her one day “Do you think I needed a baby in my life at that time?” I mean I had four children to rear and I was on my own [after her husband left]. I really didn’t need another child in my life but I wasn’t left with a choice. (Author’s emphasis)

Moreover, Rose identified intensive grandparenting as a widespread practice in her working-class community, and was aware of the limitations this put on her and other grandparents, repeatedly using the word “tied” in relation to the nature of her and other grandparents’ involvement:

I thought at this stage in my life that it would be time for me to do things that I want to do and it hasn’t worked out that way because the way life has gone, I mean I am very tied with grandchildren and children and everything else and while it is nice and I love to have them but we don’t seem to have time anymore . . . there are an awful lot of grandparents now looking after children and who are tied.

Intensive grandparenting was often practiced despite severe health conditions. Eilis, aged 57, homemaker, recounted,

[I]f there was any more grandchildren that I mightn’t be able to be looking after them, you know, because even now we’ll say with the last one I used to find even putting on their nappy some days would be a problem [due to arthritis].

The stress and attendant health problems associated with intensive grandparenting (for herself and other grandparents) were vividly described by Michelle, who had given extensive help to her sons who became fathers in their teenage years and struggled with addictions, unemployment, and problems arising from neighborhood conflict and violence:

I go up to the school to collect my grandchildren and if you see what’s standing outside, it’s mostly grandparents . . . I live in a cul-de-sac and all you see is the grandchildren because their parents are out at work and the grandmothers—my neighbor, she has heart problem and she has five grandchildren dumped [in her house] . . . and she’s running and racing from schools and she’s absolutely wrecked. My own granddaughter . . . I was doing a course in the college and gave it up [because] I used to get up a 7 o’clock in the morning, I’d go over and mind her all day . . . I was just wrecked . . . So that’s being a granny nowadays, you know.

The capacity of these “heavy duty” grandparents to delimit their involvement over time, even where the pressure of grandchild care becomes excessive or their capacity to care declines, was limited, and they were usually unable to prioritize their own time and well-being. The contrast between the two “worlds of grandparenting,” differentiated by social, economic, and health resources, is therefore stark.

Discussion

Looking after grandchildren was associated with improved quality of life for grandparents, but this differed by educational attainment and participation in other social and leisure activities outside the home. Grandparents who had primary-level education and provided high intensity grandchild care, had a lower quality of life score compared with those providing low intensity grandchild care if they had no other social and leisure activities. However, tertiary-educated grandparents who provided high intensity grandchild care had higher quality of life compared with those who provided no care and had no other social activity. Integrating these quantitative finding with the insights about choice strongly supported through the qualitative data, we hypothesize that these differences arise from the extent to which intensive grandchild care is a choice versus an obligation that presents itself as a constraint, and generates further constraints on the grandparent’s ability to do other things (including looking after own health). This is supported by previous research that found voluntary choice affected health outcomes, and in particular choice in whether or not to provide grandchild care (Deci & Ryan, 1987).

The qualitative data also found that intensive provision of grandchild care was carried out despite severe health conditions and the attendant health problems were vividly described. These grandparents’ capacity to reduce their involvement was limited and they were usually unable to prioritize their own time and well-being. In contrast, the qualitative data indicate that for the higher educated grandparents who did provide intensive grandchild care, it is typically a choice that arises from motives such as having a positive impact on the grandchild’s developmental outcomes, and lessening the pressure of combining work and child rearing for their adult children. These impacts were experienced as highly validating by these grandparents.

The quantitative data found that providing high intensity grandchild care was also associated with increased depression in grandparents with lower educational attainment, who had no other social activity outside the home. Tertiary-educated grandparents were less likely to provide high intensity grandchild care compared with grandparents with primary education. Integrating this with the qualitative data, enriched our understanding of the phenomena and the combined evidence suggests that this pattern of provision of grandchild care stems from early boundary-drawing around their involvement in child care among higher educated grandparents, a form of agency that the lower educated grandparents felt they were not able to exercise.

Similar differences were found in Germany where relationships with grandchildren were associated with higher subjective well-being, but not among those with lower educational attainment. They conclude that the less educated grandparents may be more exposed to, and less able to cope with, the stressful aspects of grandparenthood (Mahne & Huxhold, 2014). Retirement age was found to be an important moderating factor in grandparents’ response to grandchild care in the United States, and this differed by gender. While grandfathers postretirement found freedom from grandchild care improved their sense of well-being, grandmothers found grandchild care as a role enhancement, although employment was protective against the more stressful aspects of extensive grandchild care (Szinovacz & Davey, 2006). Furthermore, switching to higher levels of grandchild care has been found to be associated with worsening physical health and increased stress over time (Musil et al., 2011).

Socioeconomic group was important in the provision of grandchild care but is also associated with both health and well-being (Griffin, Fuhrer, Stansfeld, & Marmot, 2002). Furthermore, the association between socioeconomic group and subjective health has also been found to be attenuated by social relationships (Chandola, Kuper, Singh-Manoux, Bartley, & Marmot, 2004; Mao & Zhao, 2012), particularly for women in the lower socioeconomic status groups (Vonneilich et al., 2012). Thus, an alternative explanation for the variances in quality of life and depression could be associated with social relationships. Our findings support this hypothesis, as increased depression was only seen in the lower educated grandparents who did not take part in a social or leisure activity. The qualitative data in our study suggested that lower socioeconomic groups are more constrained by structural pressures to provide grandparental care and are less able to “say no.” Intensive grandparenting was identified as a widespread practice in lower socioeconomic group communities, and grandparents articulated at length and in great detail the limitations it put on them and other grandparents.

Our study had some limitations. The quantitative data are cross-sectional, and thus causal inferences cannot be drawn. The association between social inequality and health should also be considered when interpreting these results. We controlled for both income and education, and found disparities were independent of household income. We did not collect information on the reasons for providing grandchild care in TILDA, which may have helped in differentiating between choice and necessity as motivations for provision of child care, and it is likely that these would have influenced well-being and health outcomes. In addition, grandparents who are socially active may have better well-being and be more able to provide grandchild care. However, we did integrate detailed accounts of grandparenting, from our qualitative studies with the quantitative findings. We found that an active social life when taking care of grandchildren was protective for both depression and quality of life. However, this effect may not be causal but simply reflect an unmeasured dimension of “personal resilience.” Research has shown that social support and participation is protective for health and mortality and a recent intervention found a decrease in functional disability following a social participation intervention (Hikichi et al., 2015). Although our study supports this finding, we will look at the longitudinal effect of grandchild care using lagged regression models when future waves of TILDA become available. Our study had many strengths; it provided a large nationally representative sample study from which detailed social, economic, and health characteristics associated with grandparenting were examined. In addition, the combination of this with the in-depth interviews with grandparents provided us with a unique perspective into the motivations and expectations surrounding the grandchild care experience of grandparents.

Conclusion

Our study found the effect of grandchild care on the mental health and well-being of grandparents depended on whether the frequency and intensity of grandchild care were through choice or obligation. These findings have policy relevance in view of increasing structural pressures on grandparents as a source of care and support for younger family generations, and the calls for supports for grandparents who provide grandchild care in many welfare states. Policy tends to view grandparents as a homogeneous group, with uniform levels of ability and interest in providing grandchild care. Our findings suggest that this is not the case. Incentivizing grandchild care could have deleterious consequences for those grandparents who have the lowest level of health and economic resources, and who feel less able to say “no” to demands for grandchild care especially if policy instruments are used to incentivize such arrangements. For instance, the mother of a low-income worker might come under heavy pressure to provide grandchild care in a welfare state context where public provision/support toward costs of formal child care is low, but grandparental inputs are supported through income transfers. The important contribution that grandparents make to support their families and to the wider economy by providing informal child care needs to be recognized and supported by government. However, public policies also need to address the continuing need for high-quality, affordable child care options from which parents can have a choice in the type of child care that suits both the needs of the children and the needs of their working and family life. Improved access to alternative, affordable, child care options can encourage greater labor market participation and reduce poverty. This is particularly important in supporting the economic engagement of all women, including those from lower socioeconomic groups. This would also ease the pressure on grandparents as the only viable form of child care available, and enable grandparents to have a choice in whether, and how often, they provide grandchild care.