1 – The survey sample

10Our analysis concerns family care arrangements for dependent older adults living at home and having at least one child – a situation where informal care plays a major role and where various care arrangements can emerge. For each respondent, as well as a set of individual characteristics, we know the main kinds of formal and informal care they are receiving. Respondents were also asked about their families. We therefore have a set of sociodemographic data on each of their children (sex, age, family situation, employment situation, whether living with or near the respondent), whether they are involved with care and, if so, in what way (helping or not, type of help, frequency, time allocation). This information on each child is essential for addressing the determinants of their caregiving decisions.

11Studying older adults living at home involves making a selection from the total population of dependent older adults: although those living at home are a majority in numerical terms, they are more often women, are less dependent, and have larger family networks than those in residential care (Désesquelles and Brouard, 2003; Duée and Rebillard, 2006). Our analysis of care arrangements should therefore be read as partial, conditional on the older person’s decision to live at home.

12In our analysis “older” adults are all those aged 60 or over. This threshold was chosen by reference to French law: 60 is the age at which French people become eligible for senior citizens’ benefits (Tenand, 2016). [4] We regard as “dependent” any person who reported needing human help to perform at least one activity of daily living, whether a basic activity of daily living (ADL) – eating and drinking, dressing, going to the toilet, going to bed and getting up, sitting down and standing up – or an instrumental activity of daily living (IADL) – preparing meals, shopping, doing housework, managing medicines, going outside, taking public transport, using a computer or telephone. Under this broad definition, which goes beyond personal care and help with domestic chores, 10.6% people aged 60 or over (2,226 individuals in the sample) were dependent in 2008. [5]

13Table 1 shows the distribution of dependent older adults by conjugal situation and number of living children. The description of care arrangements between siblings in the next section concerns only the 1,883 dependent older adults who had at least one child. The analysis of the structural determinants and interactions that follows concerns only those dependent older adults who had precisely two children; after excluding twin sibships, the final sample comprised 479 sibships. [6]

Table 1

Dependent older adults by conjugal situation and number of children

Number of living children of the dependent older adult All No spouse With spouse Number of observations in the sample Frequency in the population (%) Number of observations in the sample Frequency in the population (%) Number of observations in the sample Frequency in the population (%) No children 1 child 2 children 3 children 4+ children 307 14.9 425 20.0 492 24.8 351 16.1 615 24.2 218 19.1 237 21.5 241 22.5 173 14.9 327 22.0 89 9.3 188 18.1 251 27.6 178 17.8 288 27.2 All 2,190 100 1,196 100 994 100

Dependent older adults by conjugal situation and number of children

Note: Weighted frequencies.
Sample: 2,190 dependant older adults living in ordinary households, with no minor children.

2 – Definition of caregiving child

14In the survey, informal care corresponds to regular care received in a context of dependence. Respondents reported it by answering the question “Are there people (family, friends etc.) other than paid carers who regularly help you perform certain daily life tasks, who help you financially or materially or who give you emotional support owing to a health problem or impairment, including people who live with you?” Respondents receiving care could list up to ten informal caregivers. The caregivers we consider here are the children included in that list. [7] Our definition of care or help is thus based on the parents’ responses. This is not a neutral measure. In particular, depending on the characteristics of the respondents and their families, there is the risk that some care is not reported (e.g. if the parent takes their child’s help for granted and does not mention it) and a risk of over-reporting (e.g. if the respondent wants to give the impression of being well cared for by their children) (Beliard et al., 2012). Also, this measure tells us nothing about how freely the caregivers decided to provide their support, or the nature or extent of the help given.

15Some children living with their parent were not reported as caregivers. [8] We assume that co-residence leads in practice to caregiving, at least for daily life tasks, and we therefore classed as caregivers all children living with their parent. [9] Children living with their parent but not reported as caregivers represented 10% of the children we considered as caregivers [10].

16The definition used in the survey covers various kinds of help. It includes not only help with daily life tasks (personal care, dressing and housework) but also material or financial help and emotional support. This diversity matches the broad range of activities used for defining dependence. The questionnaire distinguishes between types of help which are often provided simultaneously by the children reported as caregivers: [11] nearly 85% gave at least emotional support, 59% helped with daily life tasks and 18% gave financial or material help (Figure 1). Few children gave financial or material help without also giving another kind of help, so their inclusion only marginally alters the definition of care. The inclusion of emotional support as a form of caregiving has more impact, because for one-third of children reported as caregivers it is the only form of help provided.

Figure 1

Types of care provided by caregiving children to a dependent elderly parent

Types of care provided by caregiving children to a dependent elderly parent

Notes: Weighted frequencies. A child is considered a caregiver here only if he/she is reported as such by the parent (all types of care). The 157 children in the sample who were living with their parent and not reported as caregivers are not included.
Interpretation: 12.8% of children caring for a dependent elderly parent provide help only with daily life tasks. All in all, 58.6% (12.8% + 36.6% + 0.3% + 8.9%) of caregiving children provide help at least with daily life tasks.
Sample: 1,341 children of a dependent elderly adult who were reported as caregivers by their parent.

17The results presented are initially based on a broad definition of care covering any or all of the three types of care. However, because the question of coordination among potentially caregiving siblings probably arises more often in relation to daily life tasks, a narrower definition was used for the analysis of endogenous interactions, taking into account only children reported as providing this particular type of help.

1 – Identifying interactions: methodological approach

25To analyse the extent to which a child’s involvement depends on the involvement of their brother or sister, we must use a statistical model that simultaneously considers the reported caregiving behaviour of both siblings. Four care arrangements are possible: neither sibling is a caregiver, only the elder sibling is a caregiver, only the younger is a caregiver, or both siblings are caregivers.

26To isolate endogenous interactions from other factors determining involvement in care, our statistical model [13] assumes that the family care arrangement is a Nash equilibrium. Each sibling adopts what they believe to be the best behavioural option in light of the behaviour of their brother or sister, but also according to their preferences and the situation in which those preferences are expressed. The observed arrangements are therefore assumed to be “stable” in that neither sibling wants, at this stage in the process, to change their behaviour, given the behaviour of the other.

27The main value of this approach is that the probability of each sibling’s involvement depends directly on the involvement of their brother or sister, once structural factors (the individual characteristics of the child, their parent and their sibship) have been controlled for. This makes it possible to test for the presence and, if applicable, the extent of endogenous interactions. The model also allows us to differentiate behaviours by birth order: the estimated effects are assumed to be different for elder siblings and younger siblings.

28Some of the characteristics that determine children’s involvement in caregiving may be unobserved and correlated within a sibship. We therefore make the assumption that the random terms summarizing these unobserved characteristics follow a normal bivariate distribution, and estimate the model using maximum likelihood.

2 – Choice of structural determinants

29We assume that the probability of a child being a caregiver depends partly on the caregiving behaviour of their sibling and partly on their own individual characteristics and those of their parent and the sibship (see description of the sample in Appendix Table A.1).

30As well as the child’s age, [14] we take into account family and occupational situations liable to affect their availability for care (Le Bihan-Youinou et Martin, 2006; Peyrache and Ogg, 2017). A caregiver’s own children require time and attention; they can be regarded as a competing element in time allocation decisions. The effect of conjugal situation is, in principle, more ambiguous: being in a couple may reduce the caregiver’s availability or enable them to shift the care of the dependent elderly parent to their spouse, the parent’s child-in-law (negative effect expected). Or the couple may share tasks, freeing up time for the child to be a caregiver. The employment variable distinguishes the economically active (in work or unemployed) from those in other situations (retired, student, homemaker, or other [15]). We assume that individuals apply different rationales of time allocation and that the opportunity cost of caregiving is greater for the economically active. We also isolated those cases where the parent was unable to report on the child’s employment status. It was especially important to take individual characteristics into account because our definition of caregivers includes children living with their parent(s); these children have particular family and occupational profiles (Ogg et al., 2015).

31For the parent’s sociodemographic characteristics, we studied gender, as well as age, to see if this has an effect on the level and type of care provided. Women more often report receiving regular help (Bonnet et al., 2013) and the help they receive is more often of a financial or material nature, whereas men more often report receiving help with daily life tasks (Soullier, 2012; Soullier and Weber, 2011). Although all the parents in our sample were dependent, we isolated the most disabled individuals, defined as those who reported needing help for at least one basic activity of daily living. As the literature has shown the importance of the spouse as main caregiver, with children in second place (Fontaine et al., 2007; Weber, 2011) – a fact also highlighted by our descriptions of family arrangements – we included the parent’s conjugal situation among the determinants of care. We also tested the effect of the parent’s educational level (by whether or not they had an upper secondary qualification), and monthly income, isolating those who were in the top quartile of our sample (monthly income of at least €1,800). We assume that these variables, which correlate with an individual’s social position, influence family elder care norms and modify the likelihood of recourse to paid carers (Weber et al., 2014).

32Lastly, we studied the variables relating to the sibship’s characteristics. The employment status of the other sibling was included in the determinants, using the distinction between economically active and other situations. Instead of taking the sex of each sibling individually as a factor for involvement, we used the sex composition of the sibship. [16] This variable, rarely used in the literature, allowed us to test for gender-dependent contextual interactions. Knowing that women are more routinely caregivers than men (Bonnet et al., 2013; Membrado, 2013), to what extent does being a woman affect caregiving behaviour differently according to the sex of the other sibling? Lastly, we included a dichotomous variable for whether or not the age gap between the two siblings was greater than ten years.

33Our estimations used almost all the information on the children provided by the survey. Two variables were not used directly: the children’s socioeconomic category (which correlates strongly with employment status) and how far the children lived from the parent’s place of residence, which is a potentially endogenous variable. Distance from the parent’s home may be a result of the caregiving behaviour rather than the reverse, but it can also be a potential source of collinearity: living in a different region to the parent correlates closely with the child’s family situation and employment status and the parent’s conjugal situation and level of education. [17] Including that variable might therefore have biased the estimations and have decreased their precision. The robustness tests, however, discuss the results of estimations that include these two explanatory variables.

1 – Differences in behaviour by birth order and type of care

35Table 6 shows the results of the model’s estimations, [18] with two different definitions of care. In the first estimation we studied the determinants of caregiving by the elder sibling (column 1) and the younger sibling (column 2), considering as caregivers all those either living with the parent or reported as caregivers, regardless of the type of help. For the second estimation (columns 4 and 5), we restricted the status of caregiver to children living with their parent and those reported as giving help with daily life tasks.

36As regards the effect of individual characteristics, differences between elder and younger siblings emerge when the broader definition of care is applied. Younger siblings’ involvement depends significantly on their not having a partner, not having children and not being economically active. Elder children’s involvement is affected by age and whether they have a partner. When the definition of care is limited to giving help with daily life tasks, the behaviour of elder and younger children is more similar. Both, in this case, are sensitive to the fact of having children and the fact of being economically active. [19]

37The parent’s characteristics seem to have similar effects on the behaviour of both siblings. Children are more involved in caring for older parents, those living without a spouse and those without an upper secondary qualification. [20] Elder and younger siblings alike are more often involved when the parent is a mother, but that effect is only significant with the broader definition of care; this suggests that the dependent parent’s gender makes a difference mainly for financial help and emotional support. However, with the broader definition of care, younger siblings’ behaviour differs from that of elder siblings in two respects. First, the greater the parent’s difficulties with ADL, the more frequently they help, whatever the definition of care; second, they help significantly less frequently if the parent reports an income of over

38€1,800 per month, but as this applies only with the broader definition of care, the distinction probably concerns financial help. Beyond the differences in behaviour by birth order, these results reinforce the impression, already perceptible for individual determinants, that this sharper difference when care is defined more broadly is due to the elder siblings’ relatively less flexible behaviour.

39As to the variables concerning the sibship, our results shed new light on the gender effect. They show that it is not that women in general are more likely to be caregivers. For both elder siblings and younger siblings, with the narrow definition of care, only a woman with a brother is more frequently a caregiver than any other child of the same birth order, all other things being equal. Taking the broad definition of care, being a woman only has an impact if she is the younger sibling and the elder sibling is a man; in other words, in sibships where the elder is a woman, younger brothers and younger sisters are equally likely to help, all other things being equal. This effect could not have been shown with an analysis treating gender at the individual level, since it concerns the gender of the other sibling. [21]

40Elder and younger siblings are also both insensitive to whether their sibling is economically active or not, whichever definition of care is used. A wide age gap between the siblings significantly reduces the probability of the younger sibling providing care, whichever definition of care is used; for the elder, a wide age gap increases the likelihood of being a caregiver, but only with the narrow definition of care. This type of arrangement may reflect particular situations where the elder child started helping in the home early on (helping their mother with housework and taking care of their younger sibling), and continues to help in a new way (caring for the parent and protecting the younger sibling). Elder children much older than their siblings may also behave like only children, and thus be routinely designated as caregivers (Weber, 2011).

Table 6

Determinants of care provided to a dependent elderly parent by elder and younger children in sibships of two

Child reported as a caregiver by the parent (all types of aid) or co-resident with parent Child reported as a caregiver by the parent for daily life tasks or co-resident with parent Elder younger Equality of coefficients Elder younger Equality of coefficients Constant – 0.364 – 0.364 (0.563) (0.563) Yes – 0.464 – 0.464 (0.605) (0.605) Yes Child’s characteristics Age Has a partner Ref.: does not have a partner Has children Ref.: has no children Active (in work or unemployed) Parent does not know child’s occupation Ref.: retired, student, homemaker or other – 0.036*** – 0.010 (0.011) (0.010) – 0.492*** – 0.490*** (0.124) (0.124) – 0.174 – 0.550*** (0.188) (0.174) – 0.177 – 0.641*** (0.158) (0.158) – 1.434*** – 1.434*** (0.376) (0.376) No Yes No No Yes – 0.032*** – 0.016 (0.011) (0.010) – 0.733*** – 0.733*** (0.126) (0.126) – 0.347** – 0.347** (0.140) (0.140) – 0.308* – 0.823*** (0.160) (0.159) – 2.027*** – 2.027*** (0.469) (0.469) No Yes Yes No Yes Parent’s characteristics 0.026** 0.026** (0.011) (0.011) 0.159 0.159 (0.128) (0.128) 0.104 0.468*** (0.178) (0.157) – 0.707*** – 0.707*** (0.219) (0.219) – 0.123 – 0.123 (0.114) (0.114) 0.271** 0.271** (0.124) (0.124) Yes Yes No Yes Yes Yes Age Woman Ref.: man Has ADL limitations Ref.: has only IADL limitations Has upper secondary qualification Ref.: does not have upper secondary qualification Monthly income ≥ €1,800 Ref.: monthly income < €1,800 Living without spouse Ref.: living with spouse 0.026*** 0.026*** (0.010) (0.010) 0.225** 0.225** (0.116) (0.116) – 0.073 0.368** (0.165) (0.151) – 0.519*** – 0.519*** (0.191) (0.191) – 0.116 –0.155*** (0.159) (0.149) 0.394*** 0.394*** (0.125) (0.125) Yes Yes No Yes No Yes Child reported as a caregiver by the parent (all types of aid) or co-resident with parent Child reported as a caregiver by the parent for daily life tasks or co-resident with parent Elder younger Equality of coefficients Elder younger Equality of coefficients Sibship characteristics Other child active – 0.132 – 0.132 Yes – 0.076 – 0.076 Yes Ref.: other child retired. (0.129) (0.129) (0.138) (0.138) student. homemaker or other. or parent does not know other child’s occupation Ego = woman. – 0.399** – 0.215 – 0.244* – 0.244* alter = woman Ref.: ego = woman. (0.175) (0.176) No (0.138) (0.138) Yes alter = man Ego = man. alter = man – 0.459*** – 0.459*** – 0.481*** – 0.481*** Ref.: ego = woman. (0.144) (0.144) Yes (0.161) (0.161) Yes alter = man Ego = man. – 0.700*** – 0.122 – 0.402*** – 0.402*** alter = woman Ref.: ego = woman. (0.190) (0.199) No (0.145) (0.145) Yes alter = man Age gap between elder 0.377 – 0.720*** 0.433* – 0.514* and younger > 10 yrs (0.236) (0.265) No (0.242) (0.280) No Ref.: age gap < 10 yrs Interactions Other child is involved(a) 1.197*** (0.360) – 0.476 (0.524) No 1.004*** (0.367) – 0.373 (0.428) No Correlation coefficient – 0.090 – 0.015 between the residuals (0.361) (0.322) Log likelihood – 486.795 – 403.442 Number of observations 479 479

Determinants of care provided to a dependent elderly parent by elder and younger children in sibships of two

Significance: * p < 0.10; ** p < 0.05; *** p < 0.01.
Notes: Standard deviations in brackets. In the estimates shown in columns 1 and 2, a child is a caregiver if he/ she is reported as such by the parent (all types of help) or if he/she lives with the parent. In the estimates shown in columns 4 and 5, a child is a caregiver if he/she is reported as such by the parent and helps with daily life tasks, or if he/she lives with the parent. Columns 3 and 6 show which coefficients have been forced to be equal in the estimation process, i.e. coefficients for which the difference between the elder and younger child’s coefficients in the unconstrained model were not significant at the 20% level.
(a) As the likelihood function was not differentiable at the point where the interaction coefficients were simultaneously equal to 0, it was necessary to run a special significance test, which is formally described in Fontaine et al. (2009).
Sample: 492 dependent older adults living in ordinary households and having two adult children.

41In terms of endogenous interactions, we see an asymmetry of behaviours, whether care is defined as helping with daily life tasks or also includes financial help and emotional support. All other things being equal, elder siblings are significantly more frequently involved in care if the younger one is a caregiver; conversely, younger siblings tend not to adjust their behaviour according to the elder’s involvement, and are perhaps less likely to help if their elder sibling is a caregiver (negative coefficient, but not significantly different from zero).

2 – Alternative specifications

42To check the robustness of the results presented above, several alternative estimations were made using the broader definition of care. [22]

43The estimation of a model that included distance from the parent’s home, separating out those children living in a different region to their parent, gave the classic results found in the literature (see e.g. Mulder and van der Meer, 2009). Living outside the region reduces the probability of helping the parent. The results of the previous section are robust to the inclusion of the distance variable and the endogenous interaction coefficient becomes significant for the younger child. [23] The model was also estimated with a variable for the parent’s living environment – whether “living in a rural area” or not. Living in a rural area did not significantly affect the probability of receiving care from the children, and the results for the other determinants are robust.

44We then included indicators of child’s occupational category, based on INSEE’s list (six categories). Elder children who are manual workers or in intermediate occupations are more likely to be caregivers than those in higher-level occupations, as are younger children who are farmers. To facilitate interpretation, we then tested the effect of a difference between the two siblings’ occupational categories, isolating cases where one of them is in one of the top three INSEE categories (farmers; artisans, traders and small business owners; managers and higher-level occupations) and the other is in one of the other categories (technicians and other intermediate occupations; clerical workers; manual workers). When the elder is in one of the top classes and the younger in one of the others, the probability of providing care is lower for both siblings. In the converse case, there is no impact on the probability of caregiving.

45Other variables were introduced that did not seem to alter the family balance observed so far. When the parent is not living with a spouse, there may be a sharing of the care given to separated or divorced parents, particularly if the second parent is also dependent. The initial model was therefore estimated with an indicator for the parent’s marital status, distinguishing the divorced and separated from the widowed and unmarried. This estimation did not show that marital status has a significant effect on the care received.

46Last, to take into account formal care and care from other members of the dependent person’s close circle, we estimated the model with, first, an indicator isolating those respondents who received formal care and, second, an indicator isolating those who received help from another member of the close circle (friend, neighbour or family member other than spouse or children). These estimations suggest that neither formal care nor other informal care have any impact on the help provided by the children, whether elder or younger. In all these models, the results for the other explanatory variables remain robust to the inclusion of the added indicators.