Grandparental childcare, health and well-being in Europe: A within-individual investigation of longitudinal data

Abstract

Previous studies suggest grandparental childcare is associated with improved health and well-being of grandparents but limited information on the causal nature of this association exists. Here, we use the longitudinal Survey of Health, Ageing and Retirement in Europe (SHARE) of people aged 50 and above across 11 countries including follow-up waves between 2004 and 2015 (n = 41,713 person-observations from 24,787 unique persons of whom 11,102 had two or more measurement times). Between-person and within-person (or fixed-effect) regressions were applied, where between-person models show associations across participants and within-person models focus on each participant's variation over time. Health and well-being were measured according to self-rated health, difficulties with activities of daily living (ADLs), depressive symptoms, life satisfaction and meaning of life scores. Across all analyses, childcare assistance provided by older adults to their adult children, was associated with increased health and well-being of grandparents. However, these associations were almost completely due to between-person differences and did not hold in within-person analyses that compared the same participants over time. Fewer ADL limitations for grandparents who provided childcare assistance was the only association that remained in the within-individual analyses. These findings suggest that there might be only limited causal association between grandchild care and grandparental well-being and that it may be specific to physical rather than cognitive factors. The results are discussed with regard to evolutionary psychology assumptions of altruistic behavior and positive health outcomes for the helper.

Full text

Contents lists available at ScienceDirect
Social Science & Medicine
journal homepage: www.elsevier.com/locate/socscimed
Grandparental childcare, health and well-being in Europe: A within-
individual investigation of longitudinal data
Mirkka Danielsbacka
a,b,∗
, Antti O. Tanskanen
a,b,c
, David A. Coall
d,e
, Markus Jokela
f
a
University of Turku, Assistentinkatu 7, 20014, Finland
b
Population Research Institute, Helsinki, Finland
c
Faculty of Social Sciences, University of Helsinki, Helsinki, Finland
d
School of Medical and Health Sciences, Edith Cowan University, Joondalup, WA, 6027, Australia
e
School of Medicine, The University of Western Australia, Crawley, WA, 6009, Australia
f
Department of Psychology and Logopedics, Faculty of Medicine, 00014 University of Helsinki, Helsinki, Finland
ARTICLE INFO
Keywords:
Childcare
Fixed effect regression
Grandparents
Health
SHARE
ABSTRACT
Previous studies suggest grandparental childcare is associated with improved health and well-being of grand-
parents but limited information on the causal nature of this association exists. Here, we use the longitudinal
Survey of Health, Ageing and Retirement in Europe (SHARE) of people aged 50 and above across 11 countries
including follow-up waves between 2004 and 2015 (n = 41,713 person-observations from 24,787 unique per-
sons of whom 11,102 had two or more measurement times). Between-person and within-person (or fixed-effect)
regressions were applied, where between-person models show associations across participants and within-person
models focus on each participant's variation over time. Health and well-being were measured according to self-
rated health, difficulties with activities of daily living (ADLs), depressive symptoms, life satisfaction and
meaning of life scores. Across all analyses, childcare assistance provided by older adults to their adult children,
was associated with increased health and well-being of grandparents. However, these associations were almost
completely due to between-person differences and did not hold in within-person analyses that compared the
same participants over time. Fewer ADL limitations for grandparents who provided childcare assistance was the
only association that remained in the within-individual analyses. These findings suggest that there might be only
limited causal association between grandchild care and grandparental well-being and that it may be specific to
physical rather than cognitive factors. The results are discussed with regard to evolutionary psychology as-
sumptions of altruistic behavior and positive health outcomes for the helper.
1. Introduction
Grandparental care may provide many benefits for grandchildren
(Sear and Coall, 2011;Sear and Mace, 2008; but see Tanskanen and
Danielsbacka, 2017), but does it also produce health and well-being
benefits for grandparents, and if it does, is this association causal? In
contemporary aging societies, this question is salient (Coall and
Hertwig, 2010;Mare, 2011). An obvious reason for the growing interest
in grandparental well-being is the fact that life expectancy and the
number of elderly people are increasing in Western populations
(Bengtson, 2001;Glaser et al., 2010). The relevance of exploring
grandparental outcomes is derived from the potential health benefits to
the helpers, which provides a new perspective on older people who are
simultaneously providers of assistance to others and recipients of the
health benefits they may gain from these helping behaviors (Coall and
Hertwig, 2010;Tanskanen and Danielsbacka, 2019).
Previous studies have used either subjective (e.g., self-rated health;
Di Gessa et al., 2016a) or objective (e.g., cognitive tests; Arpino and
Bordone, 2014) measurements of grandparental health and well-being.
In this article, we measure grandparental health and well-being with
self-rated health, life satisfaction, meaning of life scores, depressive
symptoms and activities of daily living (ADL) limitations. All these
measurements are based on respondents’ own perceptions, and thus,
they may be defined as subjective measures of grandparental health and
well-being. However, depressive symptoms and ADL limitations are
both measured through a series of questions that have been validated in
previous studies (Katz et al., 1963;Prince et al., 1999) and they may be
considered also more “objective” measures than self-rated health, life
https://doi.org/10.1016/j.socscimed.2019.03.031
Received 9 December 2018; Received in revised form 18 February 2019; Accepted 20 March 2019
∗
Corresponding author. University of Turku, Assistentinkatu 7, 20014, Finland.
E-mail addresses: mirkka.danielsbacka@utu.fi (M. Danielsbacka), antti.tanskanen@utu.fi (A.O. Tanskanen), d.coall@ecu.edu.au (D.A. Coall),
markus.jokela@helsinki.fi (M. Jokela).
Social Science & Medicine 230 (2019) 194–203
Available online 24 March 2019
0277-9536/ © 2019 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/BY-NC-ND/4.0/).
T

satisfaction and meaning of life. In addition, self-reported ADL limita-
tions in the Survey of Health, Ageing and Retirement in Europe
(SHARE) correlate with objective measures of grip strength and walking
speed (Seidel et al., 2011).
According to evolutionary theory, grandparents' investments in
grandchildren may ultimately increase the number of offspring who
survive to reproductive age (Euler, 2011;Hawkes et al., 1997;Sear and
Coall, 2011). Inclusive fitness theory (Hamilton, 1964) suggests that a
selection pressure exists for altruism toward one's own kin, especially
toward close kin in descending order. Previous studies have produced
evidence for grandparental fitness benefits in historical populations
(Lahdenperä et al., 2004) and in contemporary societies, suggesting
that grandparental care might still be beneficial for grandchild well-
being (Sear and Coall, 2011;Tanskanen and Danielsbacka, 2012; but
see Tanskanen and Danielsbacka, 2017) and parental fertility (Kaptijn
et al., 2010;Tanskanen et al., 2014;Waynforth, 2011).
Evidence also suggests that grandparenthood and grandparental
childcare are associated with direct survival benefits for grandparents
themselves as measured by mortality (Christiansen, 2014;Hilbrand
et al., 2017a). One mechanism linking grandparental investment and
decreased mortality is that grandparental investment (e.g., looking after
one's grandchildren), potentially through increased activity levels or
cognitive stimulation, may improve a grandparent's health and well-
being, which in turn reduces mortality (Hilbrand et al., 2017b). Thus,
from this viewpoint, the evolutionary function of health benefits re-
ceived from helping is enhanced survival and longevity of the ego (Kim
et al., 2014).
Another potential explanation for positive grandparental health
outcomes is that positive emotions such as improved life satisfaction
and lower depressive symptoms function as proximate mechanisms (or
motives) that facilitate investments toward children and grandchildren.
Hence, the evolutionary function of positive emotions is that they en-
courage grandparents to act in an evolutionary beneficial way (i.e., to
invest in grandchildren). This function is not necessarily targeted at the
grandparent's own direct health and survival; however, the possible
well-being benefits may be a by-product (Buss, 2000;De Waal, 2008;
Euler, 2011). Thus, according to evolutionary theory, we may predict
that grandparental childcare is associated with improved subjective and
objective measures of grandparental health and well-being.
Importantly, the association between grandparental childcare and
the health of grandparents may not always be positive (Coall and
Hertwig, 2011;Glaser et al., 2010). Indeed, a negative effect may arise
when no grandparental investment exists at all (Drew and Silverstein,
2007) or when grandparental investment reaches its highest level (e.g.,
in the form of custodial care; Baker and Silverstein, 2008;Chen and Liu,
2012;Grinstead et al., 2003;Hughes et al., 2007;Taylor et al., 2016).
Therefore, the familial context of grandparental childcare is central to
understanding the health implications of helping. The context of
grandparenting (e.g., whether they are custodial or non-coresiding)
may cause possible selection effects (i.e., regarding health and well-
being custodial grandparents could have initially different character-
istics than non-coresiding grandparents), which in turn can explain
grandparental outcomes. If grandparental outcomes are due to selection
effects, the association between caregiving and health and well-being
should not be causal.
Previous evidence suggest, however, that providing moderate help
may have health benefits for the grandparents. Positive associations
have been found between active grandparenting (measured as contact
frequencies, emotional closeness or childcare), especially among
grandparents who do not live with their grandchildren, and grand-
parent's subjective well-being (Di Gessa et al., 2016a,2016b;Mahne
and Huxhold, 2015), increased life satisfaction, improved mental health
and lower risk of depression (Arpino and Bordone, 2014;Grundy et al.,
2012;Tsai et al., 2013; but see Brunello and Rocco, 2016) that may
ultimately increase survival (Hilbrand et al., 2017a).
These studies provide a valuable body of evidence that
grandparental childcare and health are associated among non-custodial
grandparents when compared to noncaregiving grandparents. Many of
the studies mentioned above (e.g. Di Gessa et al., 2016a,2016b;Tsai
et al., 2013) have utilized longitudinal data, meaning that they have
measurements for more than one time point and can thus investigate
whether different groups of grandparents (e.g., those who provide
childcare and those who do not) differ in their health outcomes at a
subsequent time point. These analyses, however, remain between-in-
dividuals which does not enable conclusions about the causal nature of
this association to be made (i.e., whether an increase or decrease in
childcare is influencing a particular grandparent's health over time).
These studies provide a between-person approach by comparing the
health and well-being of non-caring grandparents with that of car-
egiving grandparents. Longitudinal, within-individual analyses are
needed to better understand potential causality in this association.
Attempts to understand the causal and longitudinal nature of the
association between grandparental childcare and grandparental out-
comes have just begun. Using SHARE data, Arpino and Bordone (2014)
provide important evidence suggesting a causal association between
grandparental childcare and improved verbal fluency in grandparents
by utilizing an instrumental variable (IV) approach. To date, however,
only a handful within-individual analysis have investigated within-
person effects between childcare and grandparental health outcomes
(Ates, 2017;Ku et al., 2012,2013;Reinkowski, 2013;Sheppard and
Monden, 2018). Within-person effects show how an increase or de-
crease in grandchild care is associated with the health of a particular
grandparent over time. Only longitudinal analyses, such as within-
person regression, or other methods addressing the problem of en-
dogeneity, such as IV regression, can provide evidence of a possible
causal association between caregiving and grandparental outcome.
A preliminary study by Reinkowski (2013) used three waves of
SHARE data and utilizing different methods including within-person
models to detect causal associations did not found any within-person
associations between grandparental childcare and grandparental health
outcomes (measured with an index of physical health, cognitive func-
tioning and mental health) among grandmothers. To build upon
Reinkowski (2013) work, our analysis utilizes additional SHARE data
waves, we were able to extend the analysis to include grandfathers, and
to incorporate additional outcome measures that reflect grandparent's
life satisfaction and meaning of life. We have also tested different in-
teractions according to grandparent's gender, age and country group.
Thus, our analyses bring several additional results compared to
Reinkowski (2013) preliminary study. Sheppard and Monden (2018)
also tested with three waves of SHARE data whether looking after
grandchildren is associated with depressive symptoms, subjective life
expectancy and life satisfaction and found no within-person associa-
tions.
Using longitudinal Taiwanese data, Ku et al. (2012) found a small
within-individual effect between the provision of childcare and
grandparent's health. However, they did not distinguish the effect of
caregiving by grandparents' co-residence status. In their subsequent
study, when they separated the sample by co-residence status, the
within-person effect among non-coresiding grandparents attenuated
and was no longer statistically significant (Ku et al., 2013). Using
German longitudinal data, Ates (2017) found no within-person effects
supporting the association between supplementary childcare and
grandparent's self-rated health. These findings point to the possibility
that the associations found in the first mentioned set of studies may
reflect between-person effects and potential selection bias – healthier
grandparents with greater well-being provide more childcare than those
in poorer health conditions – or that an unobservable third factor as-
sociated with both childcare and health is producing the association
(Hilbrand et al., 2017b).
Here, we investigate whether the childcare-grandparental health/
well-being association reported previously can be found in a within-
person design (i.e., whether changes in grandchild care influences
M. Danielsbacka, et al. Social Science & Medicine 230 (2019) 194–203
195

grandparent health and well-being over time) with a large and multi-
national database. Our study is a valuable addition to the previous in-
vestigations that have tried to detect causal associations between
grandparental childcare and grandparental health and well-being be-
cause we are able to use multinational data, over longer follow-up
periods, incorporating several outcome measures and include interac-
tions according to grandparent's gender, age and country group.
We will examine within-person (or fixed-effect) associations in
longitudinal models that concentrate on within-person variation in
exposure and exclude between-person effects (Curran and Bauer, 2011).
This design will enable us to test a social causation hypothesis (i.e.,
whether changes in grandchild care frequencies are associated with
subsequent changes in grandparent health and well-being) and also an
alternative hypothesis that the association is due to social selection (i.e.,
changes in grandparental health and well-being are associated with
subsequent changes in the amount of grandchild care). To establish the
correct temporal ordering, grandparental childcare and other covariates
are measured one study wave before the outcome variables of grand-
parental health and well-being (for the social selection hypothesis,
childcare is the outcome variable and it is measured one study wave
after the independent variables i.e., grandparental health and well-
being, and covariates).
In addition, the association between grandparental childcare and
grandparental health may vary due to different demographic factors,
therefore we investigate several interactions. Based on previous results
examining grandparental childcare and grandparental health, in this
analysis we have included interactions between childcare and grand-
parent's gender, age and country group. The reason for including in-
teractions between grandparental childcare and gender is that it is well
known that grandmothers provide more childcare than grandfathers
(e.g., Coall and Hertwig, 2010) and for this reason they might also have
different outcomes associated with childcare provided. Grandmothers
may also experience active grandparenting more rewarding than
grandfathers, which may also produce different well-being outcomes
for grandmothers compared to grandfathers (e.g., Danielsbacka and
Tanskanen, 2016). Grandparental age can be an important determinant
of health and that is why it is relevant to see whether childcare is dif-
ferentially related to grandparental health outcomes among different
age groups. Previous results show (Di Gessa et al., 2016c;Hank and
Buber, 2009) that in Europe the need for grandparental childcare as
well as intensive childcare is associated with country of residence and
that is why we include also interaction between grandparental child-
care and country groups in our analyses. We have grouped the countries
into four categories (Southern Europe: Italy and Spain; Central Europe:
Switzerland, France, Germany, Austria and Belgium; Northern Europe:
Netherlands, Sweden and Denmark; Eastern Europe: the Czech Re-
public) based on family policy regimes (Leitner, 2003;Reinkowski,
2013). In addition to interaction results, we further examine the dif-
ferences in gender, age and country groups by running separate ana-
lyses for each group.
We examine these questions across five waves of the longitudinal
SHARE data that examined respondents aged 50 and above from 11
European countries between 2004 and 2015. The fixed-effect procedure
used here provides a test of the causality in the associations between
grandparental childcare and health and well-being.
2. Material and methods
2.1. Data
We used data from the SHARE, which was designed to collect
longitudinal data on the aging process of Europeans (see www.share-
project.org). The target population consists of people aged 50 and
above who speak the official language of their country and did not live
abroad or in an institution during the fieldwork period. The SHARE
data collection procedure was based on a computer-assisted personal
interview. Here, we use the first (data collection in 2004 and 2005
(Börsch-Supan, 2018a;Börsch-Supan and Jürges, 2005)), second (2006
and 2007 (Börsch-Supan, 2018b;Börsch-Supan et al., 2008)), fourth
(2011 and 2012 (Börsch-Supan, 2018c;Malter and Börsch-Supan,
2013)), fifth (2013 (Börsch-Supan, 2018d;Malter and Börsch-Supan,
2015)) and sixth (2015 (Börsch-Supan, 2018e;Malter and Börsch-
Supan, 2017)) wave data collected from 11 European countries: Aus-
tria, Germany, Sweden, Netherlands, Spain, Italy, France, Denmark,
Switzerland, Belgium and Czech Republic. The third wave of SHARE
was a retrospective life history data collection round (SHARELIFE) with
different questionnaires and was thus excluded from this study. In the
total and between-individual analyses, we included respondents who
had at least one grandchild, and answered all of the variables used in
this study. Our final sample included 41,713 person-observations from
24,787 unique people (aged 50–89 years) over the 11-year follow-up
period. In within-person analyses only those respondents who partici-
pated in at least two waves, who had at least one grandchild, and an-
swered all of the variables used in this study are included. This resulted
in within-person models with 28,028 person-observations from 11,102
unique people.
2.2. Outcome variables
SHARE contains several measures of grandparental health and well-
being including self-rated health, life satisfaction, meaning of life
scores, depressive symptoms and limitations with activities of daily
living (ADL limitations) that we explored in this study (see Table 1 for
descriptive statistics). Respondents reported self-rated health in all study
waves on a scale ranging from 1 = excellent to 5 = poor. We inverted
the scale so that a higher score reflected improved health and treated
the variable as continuous. Life satisfaction and meaning of life questions
were asked only in waves 2, 4, 5 and 6. Life satisfaction was measured
via self-report on a scale ranging from 0 (completely dissatisfied) to 10
(completely satisfied). For the meaning of life measure, respondents were
asked to answer five questions on a 4-point scale (ranging from
1 = never to 4 = often). These questions were “How often do you look
forward to each day?” “How often do you feel that your life has meaning?”
“How often do you feel full of energy these days?” “How often do you feel
that life is full of opportunities?” and “How often do you feel that the future
looks good for you?” A mean sum score was created from these items
(Cronbach's alpha = 0.80). The standardized sum scale ranged between
1 and 4, where higher numbers denoted greater feelings of meaning of
life.
Depressive symptoms were measured in the SHARE questionnaire
(waves 1, 2, 4, 5 and 6), using the EURO-D 12-item scale with estab-
lished validity and reliability (see Prince et al., 1999). Respondents
were asked whether, in the last month prior to the interview, they had
experienced any depressive symptoms such as feeling depressed or sad,
feeling that they would rather be dead, and whether they had lost in-
terest in things, had trouble sleeping recently. The response options
were either yes/no or whether or not the respondent had experienced
any of these feelings. Approximately 78% of all person-observations
reported having depressive symptoms.
In waves 1, 2, 4, 5 and 6, respondents were also asked whether they
had experienced any difficulties with basic activities of daily living
(ADL limitations) that they expected to last at least three months. In the
questionnaire, such difficulties included walking 100 m, getting up
from a chair after sitting for long periods, dressing (including putting on
shoes and socks), shopping for groceries up to a total of 23 different
difficulties. Approximately 54% of all person observations included at
least one difficulty. The reason that such a high proportion of people
experienced a limitation (c.f., Di Gessa et al., 2016a) was that we in-
cluded all possible limitations mentioned (including, for example,
problems with instrumental activities, mobility, fine motor skills), not
only the most noticeable limitations such as bathing, showering, or
walking across the room. By including all possible limitations, we are
M. Danielsbacka, et al. Social Science & Medicine 230 (2019) 194–203
196

able to gain more accurate information regarding the possible changes
in ADL functioning over time.
2.3. Independent variable
Grandchild care was the main independent variable in this study.
The SHARE includes questions whether respondents provided childcare
without the presence of the parents during the 12 months prior to the
interview and, if so, how often (ranging from 1 = almost daily to
4 = less than almost every month). Only those respondents who had
grandchildren were asked this question regarding up to 20 children,
separately (in all waves). We calculated the mean grandchild care
variable by summing and averaging the answers for all children, pro-
ducing a scale where 0 = no care (48% of all person observations),
1 = less than almost every month (15%), 2 = almost every month (13%),
3 = almost every week (17%), 4 = almost daily (7%). For instance, if a
grandparent has grandchildren via three children and he/she looks
after one child's children less often than every month and second and
third child's children almost every month the mean childcare is thus
almost every month ((1 + 2+2)/3 = 1.7). We have treated the
childcare variable as continuous. For sensitivity purposes we also for-
mulated a variable that accounted for approximate total care days
(which might be higher than mean care because a grandparent might
provide care less frequently but more in total for several children's
children; thus, the total care days might be higher than the mean care
score indicates). Sensitivity analyses with this variable did not change
the results.
2.4. Control variables
Several potential confounds, which were assessed at baseline (i.e.,
one study wave before the outcome measure), were adjusted for. These
covariates included respondent gender, age at interview, partnership
status, employment status, years of education, smoking status, heart
attack status, cancer status, and number of children and grandchildren.
We also included the fixed-effect of country in the analyses. These
confounders were included because in previous studies they have been
shown to associate with grandparental health, well-being and/or
childcare provided (Tanskanen and Danielsbacka, 2019). In within-
person models, those covariates whose values do not change between
the waves (i.e., time-invariant factors) were omitted from the analyses
(i.e., respondents gender, years of education, country). In addition to
other covariates, we controlled for the time period (in months) between
the baseline and outcome measure interview. Covariates, whose values
might change between waves, were modeled as time-varying variables
(e.g., age at interview, partnership status, employment status). To avoid
a drop in the number of observations, the age of the youngest grand-
children was not controlled for in the basic analyses because the SHARE
only collected this information systematically with regard to the re-
spondents’ four oldest children. However, we conducted sensitivity
analyses in which this variable was accounted for with similar results to
those found in the main analyses. Descriptive statistics are presented in
Table 1.
2.5. Methods
We analyzed the SHARE data using random-intercept multilevel
regression and, in the case of depressive symptoms and ADL limitations,
multilevel Poisson regression analyses where the repeated measures
(i.e., person-observations) are nested within responding persons. We
conducted both between- and within-person (or fixed-effect) models,
where the between-person effects represented the results across in-
dividuals, and the within-person effects represented individual varia-
tion over time. In practice, the between-person models provide mean
scores for respondents whereas for the within-person models, the ob-
served grandparents served as their own controls (other studies using
this approach see e.g., Jokela et al., 2018). Therefore, within-person
models eliminate all time-invariant components (Allison, 2009) such as
ethnic background, numerous genetic factors, and other selection ef-
fects. In all analyses the outcome variable was measured as time-lagged,
i.e., one study wave after the independent and control variables.
Within-person regression models have several strengths, but they
Table 1
Descriptive statistics of the 41,713 person observations from 24,787 people
across waves 1, 2, 4, 5 and 6 of the SHARE.
Total no. No. of
persons
% Mean (SD) Within
person SD
Self-rated health
Poor 3391 2748 8.1
Fair 10983 8396 26.3
Good 16087 12124 38.5
Very good 7660 6136 18.3
Excellent 3648 2916 8.7
Life satisfaction
a
34,696 22,099 7.7 (1.74) 0.74
Meaning of life
a
34,805 22,146 3.3 (0.62) 0.24
Depressive symptoms 41,775 24,827 2.3 (2.18) 0.96
ADL limitations 41,775 24,827 2.1 (3.29) 1.27
Sex
Men 16,658 10,243 39.9
Women 25,111 14,580 60.1
Age at interview 41,769 24,823 67.9 (9.01) 2.0
Partnership status
Living with spouse/
partner
26,655 16,874 63.8
Living without spouse/
partner
15,114 8496 36.2
Employment status
Working 7350 5360 17.6
Not working 34,419 20,748 82.4
Years of education 41,769 24,823 10.4 (4.35) 0
Smoking
Never smoke 23,005 13,531 55.1
Ever smoke 11,749 7123 28.1
Currently smoking 7015 4227 16.8
Heart attack
Never had heart attack 36,411 22,346 87.2
Ever had heart attack 5358 4053 12.8
Cancer
Never had cancer 39,517 23,673 94.6
Ever had cancer 2252 1935 5.4
Number of children 41,763 24,819 2.6 (1.28) 0.2
Number of grandchildren 41,769 24,823 3.9 (2.91) 0.6
Child care
No care 19,896 13,687 47.7
Less often than almost
every month
6217 5042 14.9
Almost every month 5578 4692 13.4
Almost every week 7104 5647 17.0
Almost daily 2924 2347 7.0
Country
Austria 3388 2059 8.1
Germany 3081 2143 7.4
Sweden 4029 2391 9.7
Netherlands 2243 1378 5.4
Spain 3979 2566 9.5
Italy 3553 1993 8.5
France 4699 2691 11.3
Denmark 3528 1986 8.5
Switzerland 2426 1312 5.8
Belgium 5582 3020 13.4
Czech Republic 5261 3284 12.6
Notes. Total no. = Number of total person-observations; No. of
persons = Number of unique person.
SD = Overall standard deviation; Within-person SD = Within-person standard
deviation.
Notes. Total no. = Number of total person-observations; No. of
people = Number of unique people; SD = Overall standard deviation; Within-
person SD = Within-person standard deviation.
a
questions were asked only in waves 2, 4, 5 and 6.
M. Danielsbacka, et al. Social Science & Medicine 230 (2019) 194–203
197

also have some limitations. One is that these models cannot account for
the time-variant unobserved characteristics. Fixed-effect models may
also exacerbate measurement errors which means that it is important to
avoid over-interpretation (Angrist and Pischke, 2008). In addition, the
sample size may be reduced in within-person models because of a small
number of participants who experience a change in the case of both
outcome and main independent factors. However, in the present study
this is not an issue because we have enough observations in the within-
person models, as mentioned above. Despite the limitations, within-
person regressions may provide a test for causality in the association
between active grandparenting and grandparental well-being (see also
Ates, 2017).
3. Results
3.1. Descriptive results
The descriptive results of the childcare transitions of respondents
who provided within-person data and were thus included in the fixed-
effect models are shown in Table 2. The majority of grandparents
maintained the same childcare level between waves. The childcare
transition frequencies indicated that more transition existed between
categories close to each other than those further from each other, and
more transition to less frequent rather than more frequent care was
found during the study period which may reflect ageing of grandparents
and grandchildren. Stability and change as measured by the intraclass
correlation coefficient reporting the correlation of the person-observa-
tions within a person over time for childcare was 0.67. The patterns
were fairly similar for both women (0.67) and men (0.66). The in-
traclass correlations for ADL limitations, self-rated health, meaning of
life, life satisfaction, and depressive symptoms were 0.73
(women = 0.73, men = 0.77), 0.70 (women = 0.70, men = 0.70), 0.69
(women = 0.70, men = 0.68), 0.66 (women = 0.65, men = 0.66), and
0.66 (women = 0.65, men = 0.67), respectively. These correlations
indicate high stability over time especially with regard to ADL limita-
tions and self-rated health.
3.2. Analytical results from the multilevel models
We investigate the associations between childcare provision and
ADL limitations, self-rated health, life satisfaction, meaning of life and
depressive symptoms. The magnitudes of the total, between-person, and
within-person regression coefficients of the multilevel models are illu-
strated in Fig. 1. Total regression coefficients consist of both between-
person and within-person effects. Between-person coefficients illustrate
variation across individuals (i.e., between those grandparents who
provide more childcare and those who provide less). We are here
mainly interested in within-person coefficients, which represent an in-
dividual's variation over time (i.e., whether changes in the amount of
childcare are associated with subsequent changes in grandparent's
health and well-being).
The total regression coefficients show that across all health
outcomes, grandchild care was associated with more positive outcomes
(Fig. 1). The majority of these effects, however, were found in the be-
tween-person analyses and thus they represent mainly variation be-
tween individuals. In the between-person analysis, childcare fre-
quencies were associated with increased self-rated health, life
satisfaction and higher meaning of life scores (Fig. 1; self-rated health
β = 0.06, p < .001; life satisfaction β = 0.08, p < .001; meaning of
life β = 0.04, p < .001). Similarly, grandchild care was associated
with fewer depressive symptoms and fewer ADL limitations (Fig. 1;
depressive symptoms β = −0.03, p < .001; ADL limitations
β = −0.14, p < .001).
In the within-person analyses, however, the majority of these as-
sociations did not remain. This means that those grandparents who
experienced a change in their childcare frequencies did not experience
any corresponding change in the subsequent wave of their self-rated
health, life satisfaction, meaning of life or depressive symptoms that
would be associated with the childcare provided (Fig. 1; self-rated
health β = 0.004, p = .532; life satisfaction β = 0.01, p = .222;
meaning of life β = −0.01, p = .131; depressive symptoms
β = −0.002, p = .635). The ADL limitations outcome variable was the
only exception. ADL limitations decreased when childcare provision
increased according to both between-person and within-person models
(Fig. 1; ADL limitations β = −0.01, p = .039). Although within-person
association was small and the between-person effect was much
stronger, these results nevertheless suggest that grandparents may gain
specific health benefits from active grandparenting over time.
3.3. Interactions
Because grandparental childcare might have different effects on
grandmothers and grandfathers as well as grandparents of different
ages, we re-conducted all within-person models with interaction terms
between grandparent age (categorized and centred at 60 years old) and
childcare and grandparent gender and childcare. To examine potential
cultural differences, we grouped the countries into four categories
(Southern Europe: Italy and Spain; Central Europe: Switzerland, France,
Germany, Austria and Belgium; Northern Europe: Netherlands, Sweden
and Denmark; Eastern Europe: the Czech Republic) and included the
interaction term between country group and childcare in within-person
models. Country groups were classified based on type of family policy
regimes (Danielsbacka et al., 2011;Haavio-Mannila and Rotkirch,
2009;Leitner, 2003;Reinkowski, 2013), and were used instead of
specific countries to achieve sufficient statistical power. In addition, we
split the data according to gender, age and country groups to examine
further the possible differences between groups.
We found an interaction effect in within-person models between
childcare and the grandparents’ age at interview for life satisfaction and
meaning of life (Table 3). Analyses with split data confirmed that for
both of these outcomes, the results were similar: Increases in childcare
decreased life satisfaction and meaning of life among the youngest
grandparents and conversely, improved these outcomes among the
oldest grandparents, although in the split models the associations were
Table 2
Transitions in childcare between consecutive study waves (5 study waves in total). All respondents who experienced change.
Childcare Childcare
No care Less often than almost every month Almost every month Almost every week Almost daily Total n
No care 5711 605 363 421 177 7277
Less often than almost every month 956 1012 424 240 40 2672
Almost every month 470 447 725 442 83 2167
Almost every week 497 240 515 1282 253 2787
Almost daily 261 58 80 276 505 11,780
Total n 7895 2362 2107 2661 1058 16,083
Total % 49.1 14.7 13.1 16.6 6.6
M. Danielsbacka, et al. Social Science & Medicine 230 (2019) 194–203
198

significant in the case of life satisfaction only among oldest age group
women (over 80 years-old) (β = 0.20, p = .042, n = 1949 person ob-
servations from 1124 persons) and in the case of meaning of life only
among youngest age group women (under 60 years-old) (β = −0.02,
p = .043, n = 3396 person observations from 1939 persons). Other
interactions were non-significant except in the case of ADL limitations:
Childcare reduced the ADL limitations of Northern Europeans (Table 3;
split analyses for Northern Europeans: β = −0.07, p < .001, n = 4136
person observations from 1468 persons). Therefore, the effect of
childcare on grandparent health does appear to vary by age and
childcare regime. Interestingly, we did not found any differences by
gender either in interaction models or separate models for grand-
mothers and grandfathers (results from split models not shown in the
tables).
3.4. Sensitivity analyses
Because all outcome variables are not available in all waves, we
have conducted robustness checks with the data waves 2, 4, 5 and 6 for
outcomes self-rated health, depressive symptoms and ADL limitations.
These sensitivity analyses did not change the main results except in the
case of within-person analysis for association between childcare and
ADL limitations which cease to be significant when first wave re-
spondents are excluded (β = −0.01, p = .260, n = 8774 person ob-
servations from 4020 persons). This highlights the need for long follow-
up period and sufficient amount of data to detect possible within-person
associations.
In addition, we considered that the health outcomes associated with
grandparental childcare might require long exposure periods as re-
flected in more than two measurements within a person. As such, we
reanalyzed the basic models mentioned previously by including only
the person-observations of participants who took part in all five waves
(in the case of life satisfaction and meaning of life, four waves; total and
between models n = 12,420 person-observations from 4481 people;
within-person models 6752 person observations from 1688 people).
These sensitivity analyses did not considerably change the main results.
Due to the data structure, age of a grandchild as well as co-residence
of a grandparent and grandchild could not be included in the basic
analyses without a large decrease in number of observations. To de-
termine whether these limitations make a difference, we performed
sensitivity analyses in which we restricted the sample first only to those
participants who provided systematic information regarding whether
they had at least one 14-year-old or younger grandchild (total and
between models 27,168 person observations from 16,680 people;
within-person models 17,411 person observations from 7010 people)
and second only to those grandparents whose living arrangements were
known and excluded those grandparents who lived with their 14-year-
old or younger grandchildren in the same household (n = 671). Either
of these sensitivity analyses did not change the main results sub-
stantially. Only when taking into account grandchild age, meaning of
life scores decreased slightly in the within-person models when child-
care increased (β = −0.01, p = .006). In addition, within-person
models taking into account grandchild age for the association between
childcare and ADL limitations cease to be significant which again
highlights the need for sufficient statistical power for within-person
analyses (β = −0.006, p = .348; within-person models 12,196 person
observations from 4844 people).
3.5. Social selection
Finally, we studied whether these results may be due to social se-
lection. Specifically, whether the association was reversed and it is
actually grandparental health and well-being at baseline that predicts
subsequent childcare. In most cases, no within-person effects were
found in these models. Thus, these results indicate that the association
between grandparental childcare and grandparent health is not ex-
plained completely through social selection (Appendix Table A1). Only
in the case of depressive symptoms were significant negative within-
person effects observed, which indicates that increase in depressive
symptoms in baseline predicted less childcare in the subsequent wave
(Appendix Table A1). Based on these results, and consistent with the
findings of Hilbrand et al. (2017b), the between-person associations
found for childcare and health may be due to an unobserved third
variable that affects both health and childcare.
Fig. 1. Associations between childcare and outcome variables based on total (blue bars), between-person (red bars), and within-person (grey bars) regressions using
four repeated measurements of childcare and outcomes (41,713 person observations from 24,787 people, of whom 11,102 had two or more measurement times) from
the Survey of Health, Ageing and Retirement in Europe, 2004–2015. The bars illustrate the magnitude of the linear regression coefficients (self-rated health, life
satisfaction, and life meaning) and Poisson regression coefficients (depressive symptoms and ADL limitations). Bars = 95% CIs. (For interpretation of the references
to colour in this figure legend, the reader is referred to the Web version of this article.)
M. Danielsbacka, et al. Social Science & Medicine 230 (2019) 194–203
199

4. Discussion
Here, we investigated whether grandparental childcare produces
direct benefits for grandparents over time. Overall, we found that
grandparental childcare was associated with better health and well-
being, decreased depressive symptoms and fewer ADL limitations
among European grandparents. However, these associations were lar-
gely due to between-person effects and, except for ADL limitations,
were absent when comparing the same individuals over time. That is,
grandparents who looked after their grandchildren more were healthier
but within-individual changes in childcare, in the majority, were not
associated with corresponding changes in health and well-being of the
grandparents over time.
The analyses where within-person effects were detected suggests
that childcare by grandparents may be causally associated specifically
with health benefits identified through physical health measures. The
only significant within-person effect in the main analyses showed that
the number of ADL limitations decreased over time among grand-
parents whose childcare frequencies increased over the previous wave.
No such within-person associations were found in the case of other
health and well-being measurements. Thus, changes in childcare fre-
quencies are associated with corresponding changes in grandparent
ADL limitations. Moreover, this was independent of the associations
that exist when comparing across different grandparents. It is important
to note that if grandparents have severe ADL limitations they are un-
likely to look after their grandchildren at all, meaning that in our
analyses the changes in ADL limitations most likely occur between
those older people who have only few or none limitations in the first
Table 3
Within-person associations between child care and self-rated health, life satisfaction, meaning of life, depressive symptoms and ADL limitations including interaction
terms between child care and age, child care and gender, and child care and county of origin of the respondent.
Model 1 Model 2 Model 3
β SE 95%Cl β SE 95%Cl β SE 95%Cl
lower upper lower upper lower upper
Self-rated health Child care 0.001 0.007 −0.01 0.02
Age at interview (centred to 60
years)
−0.02 0.002 −0.03 −0.02
Child care x Age at interview 0.0003 0.001 −0.001 0.002
Child care −0.002 0.018 −0.04 0.03
Child care x Gender 0.003 0.011 −0.02 0.02
Child care 0.01 0.008 −0.01 0.02
Child care x Country group −0.005 0.005 −0.02 0.01
Life satisfaction Child care −0.01 0.01 −0.04 0.01
Age at interview (centred to 60
years)
−0.02 0.004 −0.03 −0.01
Child care x Age at interview 0.003** 0.001 0.001 0.01
Child care −0.03 0.03 −0.10 0.03
Child care x Gender 0.03 0.02 −0.01 0.07
Child care 0.03 0.01 −0.003 0.05
Child care x Country group −0.01 0.01 −0.03 0.01
Meaning of life Child care −0.01 0.004 −0.02 −0.01
Age at interview (centred to 60
years)
−0.01 0.001 −0.01 −0.01
Child care x Age at interview 0.001*** 0.0004 0.001 0.002
Child care −0.01 0.01 −0.03 0.01
Child care x Gender 0.004 0.01 −0.01 0.02
Child care −0.005 0.005 −0.01 0.005
Child care x Country group −0.0001 0.003 −0.01 0.01
Depressive symptoms Child care −0.004 0.01 −0.02 0.01
Age at interview (centred to 60
years)
0.02 0.002 0.01 0.02
Child care x Age at interview 0.0002 0.001 −0.001 0.001
Child care 0.01 0.02 −0.03 0.05
Child care x Gender −0.01 0.01 −0.03 0.01
Child care −0.004 0.01 −0.02 0.010
Child care x Country group 0.001 0.005 −0.01 0.01
ADL limitations Child care −0.0005 0.008 −0.02 0.01
Age at interview (centred to 60
years)
0.07 0.002 0.06 0.07
Child care x Age at interview −0.001 0.001 −0.002 0.00004
Child care −0.04 0.02 −0.08 0.002
Child care x Gender 0.02 0.01 −0.01 0.04
Child care 0.002 0.007 −0.01 0.02
Child care x Country group −0.01** 0.005 −0.02 −0.004
*p < .05, **p < .01, ***p < .001.
Note: All models control for respondent's partnership status, employment status, years of education, smoking, ever had a heart attack, ever had a cancer, number of
children and grandchildren and time period between waves.
Note: self-rated health n = 28,032 person observations from 11,104 persons; life satisfaction n =27,404 person observations from 10,861 persons; meaning of life
n = 27,557 person observations from 10,917 persons; depressive symptoms n = 26,153 person observations from 10,283 persons; ADL limitations n = 20,751
person observations from 8066 persons.
M. Danielsbacka, et al. Social Science & Medicine 230 (2019) 194–203
200

place. It could be also that panel attrition might influence the results
related to ADL limitations. Selective panel attrition is possible in the
SHARE data, because people with initial better health could be more
likely to participate in the follow-up waves than their worse-off coun-
terparts (Bergmann et al., 2017;Börsch-Supan et al., 2013). Finally, it
must be noted that the between-person effect of ADL limitations re-
mained stronger than the within-person effect, indicating that most of
the total variance in association between childcare and ADL limitations
is due to between-person variation.
To examine the possibility that for different subgroups of grand-
parents childcare effects may be positive or negative, interactions be-
tween childcare and grandparent age, gender, and country group were
conducted. Previous research has shown that grandparental childcare
may differ according to these variables and thus its association with
grandparental health and well-being may also differ according to dif-
ferent gender, age and country categories (Tanskanen and
Danielsbacka, 2019). Interestingly, these analyses revealed that an in-
crease in childcare was associated with decreased life satisfaction and
meaning of life scores among the youngest grandparents but on the
contrary it was associated with an increase in both these outcomes
among the oldest grandparents. This finding is consistent with the idea
that not solely the childcare, but also the grandparents’ life stage may
impact their health outcomes.
Additional interaction results added further support to the associa-
tion between childcare and a decrease in ADL limitations. Specifically,
it was found that providing more frequent childcare significantly re-
duced ADL limitations especially among Northern European grand-
parents. In Northern European countries with generous public benefits
for families and the availability of formal child care services, grand-
parents look after their grandchildren at lower intensities compared to
grandparents from other parts of Europe and grandparental child care
could be rather complementary than the main source of support for
adult children (Hank and Buber, 2009). Moreover, the pension systems
as well as health care and social services for older people are most
generous in Northern European countries. In these circumstances
grandparents are not so often “forced” to look after their grandchildren
but they can provide child care when their own condition is favourable.
Thus, in Northern European countries it may not be too burdensome for
older people to look after grandchildren, which can explain why pro-
viding child care support can have health benefits for them. Previous
studies that have also used an outcome variable similar to our func-
tional limitations measure, have either not looked at interactions (Di
Gessa et al., 2016a;Hughes et al., 2007) or have not found differences
between subgroups such as between different European regions
(Reinkowski, 2013). Therefore, to our knowledge this may be a unique
finding that makes sense in terms of levels of care but needs replication.
The current analysis provides a more nuanced picture of the asso-
ciations between childcare and grandparental health benefits than has
been available previously. Specifically, we have statistically separated
the between- and within-person effects to continue the process of es-
tablishing the causal nature, if any, of this association. Our main find-
ings show that the positive associations between active grandparenting
and grandparent subjective well-being (Di Gessa et al., 2016a,2016b;
Mahne and Huxhold, 2015), better life satisfaction, better mental health
and lower risk of depression (Grundy et al., 2012) that may ultimately
increase survival (Hilbrand et al., 2017a) may not be as general an
effect as previously thought. For most health outcomes it is likely the
positive associations reflect between-person variation (but see Arpino
and Bordone, 2014). According to our analyses, only in the case of ADL
limitations can we observe a within-person effect. The effect was small
and is possible it is a chance finding. However, it will benefit from
further investigation. Our results are in most part consistent with those
of the preliminary study by Reinkowski (2013) and Sheppard and
Monden (2018) who found hardly any within-person associations using
three waves of SHARE data and Ates (2017) who found no causal as-
sociation in Germany between supplementary grandchild care and
grandparents' self-rated health. In comparison with previous studies we
were able to utilize more SHARE data waves and outcome variables
than Sheppard and Monden (2018), extend the analysis to include
grandfathers and incorporate additional outcome measures that reflect
grandparent life satisfaction and meaning of life than Reinkowski
(2013) and compared to Ates (2017) study of 1875 person-observations
from 625 people from Germany, we were able to use a multinational
sample and a broader range of outcome variables.
The risk of potential reverse causation, the possibility that there is
social selection and healthier grandparents increased their childcare
over time was directly examined and excluded. Social selection with
regard to the association between health and childcare arises when a
grandparent whose health or well-being status changes providing a
different amount of childcare over subsequent study waves. However,
only one small but significant within-person effect was found with re-
gard to these analyses, and it concerned the association between de-
pressive symptoms and childcare. This finding indicated that when
depressive symptoms increase, childcare frequencies decrease over
time, which makes sense as depression may reduce social activities
more generally.
Overall, neither social causation nor social selection appear to be
likely explanations for most of the associations between grandparental
childcare and grandparental health and well-being. Based on our re-
sults, we conclude that simultaneous variation in health and childcare
are most likely predominantly due to a third, unobserved, factor that
covaries with health and childcare. We call for future studies to in-
vestigate likely mediating factors in more detail.
In addition to these methodological considerations, our results also
yielded some theoretical implications. According to evolutionary
theory, we have predicted that grandparental childcare may be asso-
ciated with better grandparental health and well-being. This could be
caused through either the helping behavior itself, a general effect that
might improve individual's own direct survival. Alternatively, improved
grandparental well-being may be a by-product of acting in a fitness
enhancing way (i.e., helping one's own descendants to survive and re-
produce). Whether the association between caregiving and grand-
parental health is or should be causal is another question.
Although it could be assumed that caregiving offered a selective
advantage in our evolutionary past (Hawkes and Coxworth, 2013), and
contemporary humans may have a genetic propensity towards helping
behavior (Brown et al., 2011), this does not necessarily mean that
changes in caregiving behavior should have within-person effects in an
individual's health over time. If selective advantages exist in higher
caregiving tendencies, then the differences should in fact exist between
people, not within-individuals. Natural selection operates through se-
lective survival and reproduction and requires variation in a popula-
tion. This means that in the case of genetic propensity towards helping
behavior and its association with health and longevity, there should be
variation between individuals. Whether caregiving contributes proxi-
mately to grandparental health and well-being in within-individual
observations in contemporary society or, for that matter, previous so-
cieties over time might not be the point of interest from an evolutionary
point of view. The strong between-person associations found in this
study support this view.
Conversely, if we consider health and well-being as benefits that
serve as proximate mechanisms to facilitate and encourage grand-
parents to invest in their offspring (or did in our past), then this mo-
tivating effect should be especially visible in the case of measurements
that are likely to increase with increases in childcare over time (e.g., life
satisfaction or meaning of life). Such a within-individual effect was only
found in reduction of ADL limitations (improved physical health).
Investing in theoretical explanations on the nature of the association
between caregiving and caregivers’ health will help to define more
precisely in which questions we actually are predicting causal asso-
ciations and in which we are not.
Our results may also offer new insights into the wider conversation
M. Danielsbacka, et al. Social Science & Medicine 230 (2019) 194–203
201

concerning altruistic behavior and positive health outcomes for the
helper. In light of previous studies that have found associations between
altruistic behavior and positive health outcomes (e.g., Brown et al.,
2005;Brown et al., 2003;Morrow-Howell et al., 2003;Post, 2005),
some researchers have interpreted the results as causal: Doing good
makes you feel good (e.g., Post, 2005). The same applies, however, to
these associations in the case of potential health benefits of active
grandparenting (Hilbrand et al., 2017a,2017b): We do not know
whether these associations are causal (i.e., within-person effects) or
whether they reflect between-person differences.
The present study has several methodological strengths. Using re-
peated-measure data, we were able to separate between-person and
within-person associations from one another. The use of representative
and cross-national data makes our results more generalizable compared
with single-country studies and those with non-representative samples.
However, it should be noted that to more formally address the issue of
causality, one would ideally carry out an experimental study. In the
absence of such experiments, a quasi-experimental study design such as
detecting within-person variation over time with longitudinal survey
data can be used to gather hints of causality. The limitations of the
present study include that the SHARE does not have any measurement
of grandparental investment other than childcare. Positive within-
person effects might be found using other grandparental investment
measurements such as financial help that may not require the local
investment of time and effort associated with childcare. Future studies
are needed to explore the within-person effects of, for example, contact
frequencies and relationship quality with grandchildren on grand-
parental health and well-being.
To conclude, the present study found that active grandparenting is
associated with improved health and subjective well-being among
grandparents in between-person models (those that present the results
across individuals). In most cases, however, these associations did not
hold for the within-person models that analyzed an individual grand-
parent's variation over time. These results also have implications for
policy because it is important to know how the health and well-being of
older adults in ageing societies may be improved. It is also valuable to
know that the childcare assistance provided by grandparents is not
associated with decreased health and well-being of grandparents over
time.
Acknowledgements
The present study was supported by the Academy of Finland (grant
number 316229 and 317808 and 320162) and the Kone Foundation.
This paper uses data from SHARE Waves 1, 2, 4, 5 and 6 (DOIs:
10.6103/SHARE.w1.600, 10.6103/SHARE.w2.600, 10.6103/
SHARE.w4.600, 10.6103/SHARE.w5.600, 10.6103/SHARE.w6.600),
see Börsch-Supan et al. (2013) for methodological details. The SHARE
data collection has been primarily funded by the European Commission
through FP5 (QLK6-CT-2001-00360), FP6 (SHARE-I3: RII-CT-2006-
062193, COMPARE: CIT5-CT-2005-028857, SHARELIFE: CIT4-CT-
2006-028812) and FP7 (SHARE-PREP: N°211909, SHARE-LEAP:
N°227822, SHARE M4: N°261982). Additional funding from the
German Ministry of Education and Research, the Max Planck Society for
the Advancement of Science, the U.S. National Institute on Aging
(U01_AG09740-13S2, P01_AG005842, P01_AG08291, P30_AG12815,
R21_AG025169, Y1-AG-4553-01, IAG_BSR06-11, OGHA_04–064,
HHSN271201300071C) and from various national funding sources is
gratefully acknowledged (see www.share-project.org).
Appendix
Table A1
Total, between and within effects between self-rated health, life satisfaction, meaning of life, depressive symptoms and ADL limitations and child care according to
five multilevel linear regression models.
Total effect Between effect Within effect
β SE 95%Cl β SE 95%Cl β SE 95%Cl
lower upper lower upper lower upper
Self-rated health 0.05*** 0.01 0.04 0.07 0.10*** 0.01 0.08 0.11 −0.01 0.01 −0.04 0.01
Life satisfaction 0.03*** 0.004 0.02 0.04 0.05*** 0.01 0.04 0.07 −0.01 0.01 −0.02 0.003
Life meaning 0.14*** 0.01 0.12 0.17 0.21*** 0.02 0.18 0.24 0.03 0.02 −0.01 0.07
Depressive symptoms −0.02*** 0.003 −0.03 −0.02 −0.03*** 0.004 −0.04 −0.03 −0.01* 0.006 −0.02 −0.001
ADL limitations −0.02*** 0.002 −0.03 −0.02 −0.03*** 0.003 −0.04 −0.03 0.01 0.005 −0.02 0.0003
Note: All models control for respondent's partnership status, employment status, years of education, smoking, ever had a heart attack, ever had a cancer, number of
children and grandchildren and time period between waves.
Note: within-person models: self-rated health n = 23,606 person observations from 9292 persons; life satisfaction n = 18,765 person observations from 8065
persons; meaning of life n = 18,857 person observations from 8100 persons; depressive symptoms n = 14,863 person observations from 5724 persons; ADL lim-
itations n = 14,863 person observations from 5724 persons.
References
Allison, P.D., 2009. Fixed Effects Regression Models. Sage, Los Angeles.
Angrist, J.D., Pischke, J.S., 2008. Mostly Harmless Econometrics: an Empiricist's
Companion. Princeton University Press, Princeton.
Arpino, B., Bordone, V., 2014. Does grandparenting pay off? The effect of child care on
grandparents' cognitive functioning. J. Marriage Fam. 76, 337–351.
Ates, M., 2017. Does grandchild care influence grandparents' self-rated health? Evidence
from a fixed effects approach. Soc. Sci. Med. 190, 67–74.
Baker, L.A., Silverstein, M., 2008. Preventive health behaviors among grandmothers
raising grandchildren. J. Gerontol. Ser. B Psychol. Sci. Soc. Sci. 63, S304–S311.
Bengtson, V.L., 2001. Beyond the nuclear family: the increasing importance of multi-
generational bonds. J. Marriage Fam. 63, 1–16.
Bergmann, Michael, Kneip, Thorsten, De Luca, Giuseppe, Scherpenzeel, Annette, 2017.
Survey Participation in the Survey of Health, Ageing and Retirement in Europe
(SHARE), Wave 1–6. Based on Release 6.0.0. SHARE Working Paper Series 31-2017.
Munich Center for the Economics of Aging (MEA), Munich.
Brown, W.M., Consedine, N.S., Magai, C., 2005. Altruism relates to health in an ethnically
diverse sample of older adults. J. Gerontol. 60B, 143–152.
Brown, S.L., Nesse, R.M., Vinokur, A.D., Smith, D.M., 2003. Providing social support may
Be more beneficial than receiving it: results from a prospective study of mortality.
Psychol. Sci. 14, 320–327.
Brown, S.L., Brown, R.M., Preston, S.D., 2011. The human caregiving system: a neu-
roscience model of compassionate motivation and behavior. In: Brown, S.L., Brown,
R.M., Penner, L.A. (Eds.), Moving beyond Self-Interest: Perspectives from
Evolutionary Biology, Neuroscience, and the Social Sciences. Oxford University Press,
New York, NY, pp. 75–88. https://doi.org/10.1093/acprof:oso/9780195388107.001.
0001.
Brunello, G., Rocco, L., 2016. Is childcare bad for the mental health of grandparents?
Evidence from SHARE. IZA Discussion Paper Series No. 10022.
Börsch-Supan, A., 2018a. Survey of Health, Ageing and Retirement in Europe (SHARE)
Wave 1. Release version: 6.1.1. SHARE-ERIC. http://doi.org/10.6103/SHARE.w1.
611.
M. Danielsbacka, et al. Social Science & Medicine 230 (2019) 194–203
202

Börsch-Supan, A., 2018b. Survey of Health, Ageing and Retirement in Europe (SHARE)
Wave 2. Release version: 6.1.1. SHARE-ERIC. http://doi.org/10.6103/SHARE.w2.
611.
Börsch-Supan, A., 2018c. Survey of Health, Ageing and Retirement in Europe (SHARE)
Wave 4. Release version: 6.1.1. SHARE-ERIC. http://doi.org/10.6103/SHARE.w4.
611.
Börsch-Supan, A., 2018d. Survey of Health, Ageing and Retirement in Europe (SHARE)
Wave 5. Release version: 6.1.1. SHARE-ERIC. http://doi.org/10.6103/SHARE.w5.
611.
Börsch-Supan, A., 2018e. Survey of Health, Ageing and Retirement in Europe (SHARE)
Wave 6. Release version: 6.1.1. SHARE-ERIC. http://doi.org/10.6103/SHARE.w6.
611.
Börsch-Supan, A., Jürges, H. (Eds.), 2005. The Survey of Health, Ageing and Retirement
in Europe – Methodology. Mannheim Research Institute for the Economics of Aging
(MEA), Mannheim.
Börsch-Supan, A., Brugiavini, A., Jürges, H., Kapteyn, A., Mackenbach, J., Siegrist, J.,
Weber, G. (Eds.), 2008. First Results from the Survey of Health, Ageing and
Retirement in Europe (2004-2007). Starting the Longitudinal Dimension. Mannheim
Research Institute for the Economics of Aging (MEA), Mannheim.
Börsch-Supan, A., Brandt, M., Hunkler, C., Kneip, T., Korbmacher, J., Malter, F., Zuber, S.,
2013. Data resource profile: the survey of health, ageing and retirement in Europe
(SHARE). Int. J. Epidemiol. 42 (4), 992–1001.
Buss, D., 2000. The evolution of happiness. Am. Psychol. 55, 15–23.
Chen, F., Liu, G., 2012. The health implications of grandparents caring for grandchildren
in China. J. Gerontol. Ser. B Psychol. Sci. Soc. Sci. 67B (1). https://doi.org/10.1093/
geronb/gbr132. 99–367 112.
Christiansen, S.G., 2014. The association between grandparenthood and mortality. Soc.
Sci. Med. 118, 89–96. https://doi.org/10.1016/j.socscimed.2014.07.061.
Coall, D.A., Hertwig, R., 2010. Grandparental investment: past, present, and future.
Behav. Brain Sci. 33, 1–59.
Coall, D.A., Hertwig, R., 2011. Grandparental investment: a relic of the past or a resource
for the future? Curr. Dir. Psychol. Sci. 20 (2), 93–98.
Curran, P.J., Bauer, D.J., 2011. The disaggregation of within-person and between-person
effects in longitudinal models of change. Annu. Rev. Psychol. 62, 583–619.
Danielsbacka, M., Tanskanen, A.O., 2016. The association between grandparental in-
vestment and grandparents’ happiness in Finland. Pers. Relat. 23 (4), 787–800.
Danielsbacka, M., Tanskanen, A.O., Jokela, M., Rotkirch, A., 2011. Grandparental child
care in Europe. Evidence for preferential investment in more certain kin. Evol.
Psychol. 9, 3–24.
De Waal, F.B.M., 2008. Putting the altruism back into altruism: the evolution of empathy.
Annu. Rev. Psychol. 59, 279–300.
Di Gessa, G., Glaser, K., Tinker, A., 2016a. The health impact of intensive and non-
intensive grandchild care in Europe: new evidence from SHARE. J. Gerontol. Ser. B
Psychol. Sci. Soc. Sci. 71 (5), 867–879. https://doi.org/10.1093/geronb/gbv055.
Di Gessa, G., Glaser, K., Tinker, A., 2016b. The impact of caring for grandchildren on the
health of grandparents in Europe: a lifecourse approach. Soc. Sci. Med. 152, 166–175.
https://doi.org/10.1016/j.socscimed.2016.01.041.
Di Gessa, G., Glaser, K., Price, D., Ribe, E., Tinker, A., 2016c. What drives national dif-
ferences in intensive grandparental childcare in Europe? J. Gerontol. Ser. B Psychol.
Sci. Soc. Sci. 71, 141–153.
Drew, L.M., Silverstein, M., 2007. Grandparents' psychological well-being after loss of
contact with their grandchildren. J. Fam. Psychol. 21, 372–379. https://doi.org/10.
1037/0893-380 3200.21.3.372.
Euler, H.A., 2011. Grandparents and extended kin. In: Salmon, C.A., Shackelford, T.K.
(Eds.), The Oxford Handbook of Evolutionary Family Psychology. Oxford University
Press, New York, pp. 181–210.
Glaser, K., Ribé, E., Montserrat, Waginger, U., Price, D., Stuchbury, R., Tinker, A., 2010.
Grandparenting in Europe. Grandparents Plus, London.
Grinstead, L.N., Leder, S., Jensen, S., Bond, L., 2003. Review of research on the health of
caregiving grandparents. J. Adv. Nurs. 44, 318–326.
Grundy, E.M., Albalab, C., Allena, E., Dangoura, A.D., Elbournea, D., Uauya, R., 2012.
Grandparenting and psychosocial health among older Chileans: a longitudinal ana-
lysis. Aging Ment. Health 16 (8), 1047–1057.
Haavio-Mannila, E., Rotkirch, A., 2009. Sexuality and family formation. In: Immerfall, S.,
Therborn, G. (Eds.), Handbook of European Societies: Social Transformations in the
21st Century. Springer, New York, pp. 465–498.
Hamilton, W.D., 1964. The genetical evolution of social behaviour (I and II). J. Theor.
Biol. 7, 1–52.
Hank, K., Buber, I., 2009. Grandparents caring for their grandchildren: findings from the
2004 survey of health, ageing, and retirement in Europe. J. Fam. Issues 30, 53–73.
Hawkes, K., Coxworth, J.E., 2013. Grandmothers and the evolution of human longevity: a
review of findings and future directions. Evol. Anthropol. 22 (6), 294–302. https://
doi.org/10.1002/evan.21382.
Hawkes, K., O'Connell, J.F., Blurton Jones, N.G., 1997. Hadza women's time allocation,
offspring provisioning, and the evolution of post-menopausal lifespans. Curr.
Anthropol. 38, 551–578. https://doi.org/10.1086/204646.
Hilbrand, S., Coall, D., Gerstorf, D., Hertwig, R., 2017a. Caregiving within and beyond the
family is associated with lower mortality for the caregiver: a prospective study. Evol.
Hum. Behav. 38, 397–403.
Hilbrand, S., Coall, D.A., Meyer, A.H., Gerstorf, D., Hertwig, R., 2017b. A prospective
study of associations among helping, health, and longevity. Soc. Sci. Med. 187,
109–117.
Hughes, M.E., Waite, L.J., LaPierre, T.A., Luo, Y., 2007. All in the family: the impact of
caring for grandchildren on grandparents' health. J. Gerontol. Ser. B Psychol. Sci. Soc.
Sci. 62, S108–S119.
Jokela, M., Airaksinen, J., Kivimäki, M., Hakulinen, C., 2018. Is within-individual var-
iation in personality traits associated with changes in health behaviours? Analysis of
seven longitudinal cohort studies. Eur. J. Personal. 32, 642–652.
Kaptijn, R., Thomese, F., van Tilburg, T.G., Liefbroer, A.C., 2010. How grandparents
matter. Support for the cooperative breeding hypothesis in a contemporary Dutch
population. Hum. Nat. 21, 393–405.
Katz, S., Ford, A.B., Moskowitz, R.W., Jackson, B.A., Jaffe, M.W., 1963. Studies of illness
in the aged. The index of ADL: a standardized measure of biological and psychosocial
function. J. Am. Med. Assoc. 185, 914–919. https://doi.org/10.1001/jama.1963.
03060120024016.
Kim, P.S., McQueen, J.S., Coxworth, J.E., Hawkes, K., 2014. Grandmothering drives the
evolution of longevity in a probabilistic model. J. Theor. Biol. 353, 84–94. https://
doi.org/10.1016/j.jtbi.2014.03.011.
Ku, L.-J.E., Stearns, S.C., van Houtven, C.H., Holmes, G.M., 2012. The health effects of
caregiving by grandparents in Taiwan: an instrumental variable estimation. Rev.
Econ. Househ. 10, 521–540. http://doi.org/10.1007/s11150-012-9154-9.
Ku, L.-J.E., Stearns, S.C., van Houtven, C.H., Lee, S.-Y.D., Dilworth-Anderson, P., Konrad,
T.R., 2013. Impact of caring for grandchildren on the health of grandparents in
Taiwan. J. Gerontol. Ser. B Psychol. Sci. Soc. Sci. 68, 1009–1021. http://doi.org/10.
1093/geronb/gbt090.
Lahdenperä, M., Lummaa, V., Helle, S., Tremblay, M., Russell, A.F., 2004. Fitness benefits
of prolonged post-reproductive lifespan in women. Nature 428, 178–181.
Leitner, S., 2003. Varieties of familialism: the caring function of the family in comparative
perspective. Eur. Soc. 5, 353–375.
Mahne, K., Huxhold, O., 2015. Grandparenthood and subjective well-being: moderating
effects of educational level. J. Gerontol. 70, 782–792. https://doi.org/10.1093/
geronb/gbu147.
Malter, F., Börsch-Supan, A. (Eds.), 2013. SHARE Wave 4: Innovations & Methodology.
MEA, Max Planck Institute for Social Law and Social Policy, Munich.
Malter, F., Börsch-Supan, A. (Eds.), 2015. SHARE Wave 5: Innovations & Methodology.
MEA, Max Planck Institute for Social Law and Social Policy, Munich.
Malter, F., Börsch-Supan, A. (Eds.), 2017. SHARE Wave 6: Panel Innovations
andCollecting Dried Blood Spots. Munich Center for the Economics of Aging (MEA),
Munich.
Mare, R.D., 2011. A multigenerational view of inequality. Demography 48, 1–23. https://
doi.org/10.1007/s13524-011-0014-7.
Morrow-Howell, N., Hinterlong, J., Rozario, P.A., Tang, F., 2003. Effects of volunteering
on the well-being of older adults. J. Gerontol. Ser. B Psychol. Sci. Soc. Sci. 58 (3),
S137–S145. https://doi.org/10.1093/geronb/58.3.s137.
Post, S.G., 2005. Altruism, happiness, and health: it's good to be good. Int. J. Behav. Med.
12, 66–77.
Prince, M., Reischies, F., Beekman, A., Fuhrer, R., Jonker, C., Kivela, S., et al., 1999.
Development of the EURO-D scale: a European, Union initiative to compare symp-
toms of depression in 14 European centres. Br. J. Psychiatry 174, 330–338. https://
doi.org/10.1192/bjp.174.4.330.
Reinkowski, J., 2013. Should We Care that They Care? Grandchild Care and its Impact on
Grandparent Health. Ifo Working Paper No. 165.
Sear, R., Coall, D., 2011. How much does family matter? Cooperative breeding and the
demographic transition. Popul. Dev. Rev. 37, 81–112. https://doi.org/10.1111/j.
1728-4457.2011.00379.x.
Sear, R., Mace, R., 2008. Who keeps children alive? A review of the effects of kin on child
survival. Evol. Hum. Behav. 29, 1–18.
Seidel, D., Brayne, C., Jagger, C., 2011. Limitations in physical functioning among older
people as a predictor of subsequent disability in instrumental activities of daily living.
Age Ageing 40, 463–469. https://doi.org/10.1093/ageing/afr054.
Sheppard, P., Monden, C., 2018. Becoming a first-time grandparent and subjective well-
being: a fixed effects approach. J. Marriage Fam. 80 (in press).
Tanskanen, A.O., Danielsbacka, M., 2012. Beneficial effects of grandparental involvement
vary by lineage in the UK. Pers. Indiv. Differ. 53, 985–988.
Tanskanen, A.O., Danielsbacka, M., 2017. Multigenerational effects on children's cogni-
tive and socioemotional outcomes: a within-child investigation. Child Dev. 88.
https://doi.org/10.1111/cdev.12968.
Tanskanen, A.O., Danielsbacka, M., 2019. Intergenerational Family Relations. An
Evolutionary Social Science Approach. Routledge, New York.
Tanskanen, A.O., Jokela, M., Danielsbacka, M., Rotkirch, A., 2014. Grandparental effects
on fertility vary by lineage in the United Kingdom. Hum. Nat. 25, 269–284.
Taylor, M.F., Marquis, R., Batten, R., Coall, D.A., 2016. Understanding the mental health
travails of custodial grandparents. Occup. Ther. Ment. Health 32 (3), 259–280.
Tsai, F.-J., Motamed, S., Rougemont, A., 2013. The protective effect of taking care of
grandchildren on elders' mental health? Associations between changing patterns of
intergenerational exchanges and the reduction of elders' loneliness and depression
between 1993 and 2007 in Taiwan. (Report). BMC Public Health 13 (1), 1–9.
Waynforth, D., 2011. Grandparental investment and reproductive decisions in the long-
itudinal 1970 British cohort study. Proc. R. Soc. Lond. Ser. B Biol. Sci. 279,
1155–1160.
M. Danielsbacka, et al. Social Science & Medicine 230 (2019) 194–203
203