Gender socialisation and care-giving

Women, particularly from earlier birth cohorts, have been socialised to fulfil nurturing roles, deriving a part of their identity and social integration from caring. For women, caring for their dependent husbands has often been described as a continuation of earlier care-giving tasks (e.g. child-rearing). Starting to care for a frail spouse should, thus, require fewer adaptations for women than for men who newly transition into a ‘career’ of care-giving and eventually also housekeeping (Russell, Reference Russell2007). Men have more legitimate ‘excuses’, such as lacking skills and routine, if they do not wish to take over personal care (Connidis and McMullin, Reference Connidis and McMullin2002; Henz, Reference Henz2010). We could thus expect that men are involved in spousal care-giving less often than women. This is also reflected in the literature on adult children's care-giving to their parents (Brandt et al., Reference Brandt, Haberkern and Szydlik2009; Henz, Reference Henz2009).

On the other hand, spousal care can be described as ‘care by default’, arising from lifelong co-habitation and a close emotional bond (Arber and Ginn, Reference Arber and Ginn1994). Men, too, might choose deliberately to care for their sick spouses as they might otherwise ‘lose’ their lifelong companion. Caring for a sick or frail spouse is often seen as an extension of the love and support that has been exchanged for many years (Arber and Ginn, Reference Arber and Ginn1994; Corden and Hirst, Reference Corden and Hirst2011; Hong and Coogle, Reference Hong and Coogle2016). Caring for one's frail spouse is often culturally framed as a spousal duty. Moreover, spouses possess valuable knowledge about the frail elderly person's medical history, preferences and health behaviour, which facilitates care-giving. Gender might thus not only be an important factor when it comes to an individual's propensity to provide care for a spouse, but also with respect to the choice of care-giving arrangement. For instance, female care-givers spend on average more hours on care-giving than male care-givers (Noël-Miller, Reference Noël-Miller2011), who do so often as secondary care-giver or support to the female main care-giver (Gerstel and Gallagher, Reference Gerstel and Gallagher2001; Henz, Reference Henz2009). An important question which was addressed by Feld et al. (Reference Feld, Dunkle, Schroepfer and Shen2010) is to what extent care given to a spouse is performed alone (‘solo care-giving’), or with either informal or formal support (‘shared care-giving’). The authors found that men are less likely than women to be the sole care-givers, but more likely than women to share care-giving tasks. The distinction between solo and shared care seems thus an especially important aspect when describing male care-giving, as men have been shown to more readily accept help with caring tasks (Mutchler and Bullers, Reference Mutchler and Bullers1994; Bond et al., Reference Bond, Farrow, Gregson, Bamford, Buck, McNamee and Wright1999; Henz, Reference Henz2009). Moreover, men have been described as embracing a more delegating or ‘managerial’ approach to organising care (Deufert, Reference Deufert2013; Gallo and Scrinzi, Reference Gallo and Scrinzi2016). We thus expect men to exhibit a higher likelihood of both outsourcing and sharing the care of their spouses, while women are expected to be more likely to act as solo care-givers.

• • Hypothesis 1: Men are less likely than women to act as solo spousal care-givers.

Economic explanations for care-giving

Besides gender role socialisation, economic considerations are involved in an individual's decision regarding the choice of a care-giving arrangement. The household economy model (Becker, Reference Becker1965) assumes that partners specialise by dividing paid and unpaid labour to maximise the household's utility and minimise costs. Men, and particularly male breadwinners, might encounter higher costs by reducing work hours or giving up employment to care for their spouse than by hiring a care-giver paid a lower salary than their own (Ettner, Reference Ettner1996). Economic considerations might also be at work when comparing the care-giving involvement of employed persons with persons who are not employed, or between income groups (Arber and Ginn, Reference Arber and Ginn1992; Sarkisian and Gerstel, Reference Sarkisian and Gerstel2004). Employed persons face higher opportunity costs, both time- and moneywise, and might thus be more likely to share or outsource care. The link between employment and elder-care is not free from issues of endogeneity. As previous research has shown, it is not the circumstance of care-giving, but rather the high intensity of the care, that leads to care-givers (gradually) dropping out of the labour market (Henz, Reference Henz2009, Reference Henz2010). This indicates that care-givers develop strategies to remain in the labour market. One of these strategies could be to outsource care completely, or to share it with a formal or informal source of support. As employees have a higher income than retired persons or individuals not in the labour market, they should also display a higher propensity to hire a formal care-giving service to supplement their care-giving responsibilities entirely. We can thus formulate two hypotheses:

• • Hypothesis 2a: Employed persons are more likely to outsource care-giving to a third party than non-employed persons.

• • Hypothesis 2b: Employed persons are more likely to share care-giving with a formal source of support than non-employed persons.

As pointed out earlier, since economic considerations are likely to play a role when deciding on a care-giving arrangement, solo and shared care-giving might not only vary depending on employment status, but also with respect to socio-economic status which entails greater financial resources such as income or home-ownership. Individuals with higher income might face considerably higher opportunity costs of engaging as (solo) care-givers than persons with lower income. They could thus be expected to be more likely to share or outsource care-giving. Moreover, the source of care support – formal services or informal helpers – with whom spousal care-giving is shared or to whom care is outsourced might vary, too, with socio-economic status. Care receivers with a higher socio-economic status have been shown to be less dependent on informal care (Arber and Ginn, Reference Arber and Ginn1992; Blomgren et al., Reference Blomgren, Breeze, Koskinen and Martikainen2011), as they are more likely to be able to pay for formal care services.

In many European countries, even the publicly provided formal care partly needs to be paid ‘out of pocket’ by the care receiver or the number of care hours guaranteed are not sufficient for covering the care need. Since households with greater economic resources can rather afford to supplement the publicly provided care with their own means, households with fewer resources are more likely to rely (exclusively) on informal care-giving. We thus expect a higher propensity of informally shared or outsourced care-giving arrangements among individuals with lower socio-economic status, and a higher likelihood of sharing or outsourcing care-giving formally among persons with higher socio-economic status. Empirical evidence supports these two hypotheses. Informal family-based care-giving is often found among lower-income groups, while higher-income groups are more likely to choose formal home care (McAuley and Arling, Reference McAuley and Arling1984; Broese van Groenou et al., Reference Broese van Groenou, Glaser, Tomassini and Jacobs2006). Besides income, home-ownership might play a role, too (Broese van Groenou et al., Reference Broese van Groenou, Glaser, Tomassini and Jacobs2006; Corden and Hirst, Reference Corden and Hirst2011). Having no costs for rent frees up financial resources for hiring a paid care-giver, which should increase the likelihood of sharing care-giving formally. We can thus formulate two hypotheses:

• • Hypothesis 3a: Persons with a higher socio-economic status (i.e. with higher income/home-owners) are more likely than persons with lower socio-economic status to share or outsource care-giving with a formal service provider.

• • Hypothesis 3b: Persons with a lower socio-economic status (i.e. with lower income/not home-owners) are more likely than persons with higher socio-economic status to share or outsource care-giving informally.

Welfare states and family care

Over the past two decades, elder-care has received increasing attention from welfare researchers (Pfau-Effinger, Reference Pfau-Effinger2005; Saraceno and Keck, Reference Saraceno and Keck2010; Suanet et al., Reference Suanet, Broese van Groenou and Van Tilburg2012). As the baby-boomers approach old age, many countries face the challenge of providing care infrastructure for a growing number of frail elderly people. Over the past decade, expenditures for elder-care have risen in most countries (Eurostat, 2017), however their financing has remained a topic of public debate. There are four ways in which states can organise elder-care: first, by providing formal care in institutions; second, by providing publicly funded home care services (the so-called ‘Care-in-Kind’ services); third, by enabling family members or other private persons to act as informal care-givers for a frail person (through care-giver allowances, the so-called ‘Cash-for-Care’ schemes); and fourth, by obliging family members to take responsibility for the care of their kin (Pfau-Effinger, Reference Pfau-Effinger2005).

However, these four ways constitute strictly ideal types. Most welfare states provide a mix of several strategies, whereby Cash-for-Care and Care-in-Kind have become the most important pillars. Cash allowances, by providing wages for informal care-givers, promote informal care-giving arrangements. They may, however, in some countries also be used to purchase services on a privatised market (Frericks et al., Reference Frericks, Jensen and Pfau-Effinger2014). Care-in-Kind schemes rather promote formal care-giving. By providing formal services for ADL and IADL delivered to the home of the care recipient, they enable receivers to remain in their own homes as long as possible. For instance, in Southern and Eastern European countries, more informal care-giving occurs due to a lack of public care facilities. In Northern European countries and the Netherlands, on the other hand, where formal home care services are more developed, less informal and family care-giving is provided (Alber and Kohler, Reference Alber and Kohler2004).

Several approaches have been undertaken to formulate welfare state classifications based on the organisation of elder-care (Bettio and Plantenga, Reference Bettio and Plantenga2004; Haberkern and Szydlik, Reference Haberkern and Szydlik2010; Saraceno and Keck, Reference Saraceno and Keck2010). In their comparison of the explanatory power of these different typologies for elderly people's participation in informal care-giving, Strauss and Trommer (Reference Strauss and Trommer2017) yield support for the typology advanced by Saraceno and Keck (Reference Saraceno and Keck2010) who point out that defamilised welfare states generally reduce the care load for the family. As the overall burden of care work diminishes with higher expenditures on elder-care infrastructure, children's involvement in the physical care of their parents might be ‘crowded out’ (Brandt et al., Reference Brandt, Haberkern and Szydlik2009). For co-resident spouses, on the other hand, the availability of professional home care support might reduce the actual workload, but not the propensity of being involved in care-giving. This is in line with the study by Hank (Reference Hank2011), who concludes that the probability of helping and caring individuals rises with increasing social government spending.

Besides impacting on the choice of care-giving arrangement, different care policies also have the power to either increase or reduce gender inequalities in care-giving (Orloff, Reference Orloff1996; Bettio and Plantenga, Reference Bettio and Plantenga2004; Schmid et al., Reference Schmid, Brandt and Haberkern2012). Generally, increasing female participation in the labour market is regarded as the key to reducing gender inequalities (Frericks et al., Reference Frericks, Jensen and Pfau-Effinger2014). By promoting formal care arrangements, Care-in-Kind facilitates reconciling employment and care-giving, and enables both male and female spousal care-givers to remain in gainful employment. Higher expenditures on Care-in-Kind schemes could thus be expected to have the same impact on men's and women's choices of care-giving arrangements. On the other hand, Cash-for-Care schemes promote informal care-giving by paying so-called ‘care-giver wages’. However, as these wages are rarely competitive with labour market incomes, they are particularly attractive for individuals employed part-time or earning only low wages (mostly women, see Frericks et al., Reference Frericks, Jensen and Pfau-Effinger2014). Higher expenditures on Cash-for-Care schemes are thus likely to influence male and female engagement in spousal care-giving differently. With regard to shared and outsourced care as well as formal and informal care support, one would expect women to be particularly likely to act as solo spousal care-givers and men to be more likely to share or outsource care informally under Cash-for-Care schemes. This can be explained, respectively, with the financial incentives to remain in the labour market (for men) and to give up paid work (for female spousal care-givers as well as for potential female informal care-givers, e.g. daughters):

• • Hypothesis 4: Higher expenditures on Care-in-Kind services do not lead to differences in the choice of care-giving arrangements between men and women. Cash-for-Care programmes promote solo care-giving for women and informally shared or informally outsourced care-giving for men.

Besides gendered impacts, welfare policies and their influence on care-giving arrangements can also be expected to vary with socio-economic status, as empirical research suggests. Comparing Finland and England, Blomgren et al. (Reference Blomgren, Breeze, Koskinen and Martikainen2011) showed that in market-based care systems like England, children provide informal care when parents have a low socio-economic status and thus limited access to formal care services. In social-democratic Finland, by contrast, children are not only more likely to supplement formal care services than in England, they also do so irrespective of parents’ socio-economic status as the receipt of formal care does not depend on financial resources. Based on data from four European countries, Broese van Groenou et al. (Reference Broese van Groenou, Glaser, Tomassini and Jacobs2006) showed that the use of informal and formal care has a social gradient. While frail elderly people with lower socio-economic status are more likely to receive informal family-based care, individuals with a higher socio-economic status are more likely to receive formal care. This strengthens the argument that, in contexts where formal care-giving is promoted, financial resources play a smaller role for the choice of care-giving arrangement whereas in contexts where informal care-giving is promoted, the opportunity to purchase formal care depends largely on one's own financial resources.

• • Hypothesis 5: Higher expenditures for Care-in-Kind services do not lead to differences in the choice of care-giving arrangement between individuals with higher or lower socio-economic status. Cash-for-Care programmes increase the likelihood of individuals with low socio-economic status choosing an informally shared or informally outsourced care-giving arrangement and increase the likelihood of individuals with high socio-economic status choosing a formally shared or formally outsourced care-giving arrangement.

Control variables

Two theoretical models are frequently referred to in the literature on caring: the Andersen–Newman model (Andersen and Newman, Reference Andersen and Newman1973) and the task-specific model (Messeri et al., Reference Messeri, Silverstein and Litwak1993). Both models address the choice of the ‘optimal’ care-giver from the perspective of the care receiver. Drawing on these models, we can identify a range of important need, enabling and predisposing factors to explain the choice of care-giving arrangement. Besides age, the number and severity of the care receiver's health limitations are the strongest predictors of a person's need for help (McAuley and Arling, Reference McAuley and Arling1984). Disability as well as mobility and ADL limitations have been found to determine the degree of care needed (Balia and Brau, Reference Balia and Brau2014). On the other hand, the health of the care-giver plays a role, too (Corden and Hirst, Reference Corden and Hirst2011; Shen et al., Reference Shen, Feld, Dunkle, Schroepfer and Lehning2015). Our models thus contain indicators on the functional health of both respondent and spouse. Moreover, we can assume that if both spouses are limited functionally, they are most likely to delegate care to a third party. We thus also include an interaction term between the respondent's and spouse's functional health limitations.

Whether and how the care of one's spouse is shared depends also on the availability and characteristics of informal helpers, particularly, the family network. Children, next to spouses, are the most important source of informal care-giving to the elderly. Several comparative studies have investigated children's care-giving to their elderly and needy parents. Daughters are more frequently involved in parental care-giving than sons. Moreover, the further away a child lives, the more unlikely he or she is to provide care for his or her parents (Brandt et al., Reference Brandt, Haberkern and Szydlik2009; Schmid et al., Reference Schmid, Brandt and Haberkern2012). Men are particularly likely to share care-giving with informal helpers when there are daughters living nearby (Feld et al., Reference Feld, Dunkle, Schroepfer and Shen2010). In our models, we thus include information on the number of the care receiver's daughters (as in cases where the couple is re-married or re-partnered, information on the children of the care receiver is favourable) as well as on the number of children of the care receiver who live within a 25 kilometre radius (Suanet et al., Reference Suanet, Broese van Groenou and Van Tilburg2012).

Dependent variable

Our analyses compare solo care-givers with men and women who share the care of their spouse or outsource it to a third person. To generate our dependent variable, we use three items. The first item assesses the respondent's spousal care-giving. It is phrased as follows: ‘Is there someone living in this household who you have helped regularly during the last 12 months with personal care, such as washing, getting out of bed or dressing?’, followed by a list of persons from which to choose the care receiver. The second and third items are taken from the spouse's interview. They assess whether the spouse has received help from a third person. They were worded as follows: ‘Which (other) family member outside the household, friend or neighbour has helped you in the last 12 months?’ for informal care receipt, and ‘During the last 12 months did you receive any professional or paid help listed on this card due to a physical, mental, emotional or memory problem?’ for formal care receipt. If answered with ‘yes’, both items were followed up with a specification of the type of help received (personal care, domestic help, help with paperwork), whereby we only respected the receipt of personal care. Since the sixth wave of SHARE was the only wave which collected each of these three items, our analyses rely on that wave exclusively.

If the respondent reported giving care to his or her spouse, but the care receiver did not report any further care by a third person, the case was coded as ‘solo care-giving’. ‘Informally shared care-giving’ applies to respondents giving care and their spouse reporting additional care received from children, other relatives, friends or neighbours, i.e. informal sources of care support exclusively. ‘Formally shared’ care-giving arrangements pertain to cases where the respondent reports spousal care-giving and the spouse reports receiving any formal care (including cases where informal plus formal care support was received). Outsourcing pertains to all cases where the respondent does not report providing care, but the spouse reports receiving care from a third party. It was distinguished likewise between only informal care-givers (‘outsourced, informally’) and care-giving outsourced to professional services or paid helpers, including care-giving arrangements where formal and informal care-givers are involved (‘outsourced, formally’).

As our dependent variable features five unordered categories, we apply multinomial logistic models with country dummies, and robust standard errors clustered at the household level.Footnote ² Cross-sectional analyses often face problems of endogeneity. To check for the robustness of our results, we also calculate fixed-effects logistic regression models using SHARE Waves 1, 2, 4, 5 and 6, with a dummy variable (no/yes) capturing the occurrence of spousal care-giving in the respective wave as a dependent variable (see Table 3 in the online supplementary material). Since the focus of our paper is, however, on the various care-giving arrangements of spousal care-givers, these models can only serve as additional information to address the question of endogeneity.

Independent variables

As independent variables, we use information both from the respondent as well as from the respondent's spouse living in the same household (for an overview, see Table 1 in the online supplementary material). Respondent characteristics included in the models are: gender, age (which must be 50 years or older), the number of functional limitations, which was calculated as the sum of the self-reported number of mobility limitations plus the self-reported number of ADL limitations (ranging from 0 to 15), as well as a dichotomous variable which takes on the value ‘1’ if the respondent is employed or self-employed. To account for the socio-economic status of our subjects, we use a dummy variable for home-ownership, as well as country-specific weighted quintiles on household income. The income quintiles were obtained using the imputed variable on the net household income, which was then weighted by the number of persons living in the household (where the first household member counted as a full person and every additional person counted as 0.5, with households of more than seven members being treated as if they contained seven members). In a second step, we aggregated the thus weighted income variable by country into quintiles to place the respondents and their households according to their socio-economic status within their national contexts. For the interaction terms between socio-economic status and the two macro indicators, we created a dummy variable which takes on the value ‘1’ if the respondent was placed in the lowest or second-lowest quintile of the thus generated income variable.

Information retrieved from the spouse includes the age of the spouse, the number of functional limitations (calculated the same way as for the respondent), the number of daughters, the number of children who are living within a 25 kilometre radius, as well as the receipt of practical help (i.e. household help but not care) from outside the household.

At the macro-level, we insert dummy variables for each country, with Germany taken as the reference category. For the models featuring macro indicators, we look at the relative size of government spending on Cash-for-Care and Care-in-Kind schemes as their share as a percentage of the Gross Domestic Product (Organisation for Economic Co-operation and Development, 2017). These macro indicators are lagged by one year prior to the year in which the SHARE Wave 6 survey took place. To check the robustness of the results, we also checked a range of further indicators, among others the percentage of employees in the social sector as an indicator for an extensive welfare state (see e.g. Haberkern and Szydlik, Reference Haberkern and Szydlik2010; Hank, Reference Hank2011). These findings are not displayed; however, the coefficients indicate robustness across the models. To account for gender-specific or socio-economic status-specific impacts of welfare spending in care-giving arrangements, we calculate four separate models with cross-level interactions between the two macro indicators and gender and socio-economic status, respectively.

Descriptive findings

Our dependent variable distinguishes between solo and shared care-giving arrangements, or care outsourced to a third care-giver. Moreover, we separate informal from formal care support. The focus lies on individuals who care for a spouse, or individuals living with their care-receiving spouse in the same household. We are interested in both gender-specific (see Figure 1) and country-specific differences in care-giving arrangements (see Figure 2). Women are more likely to act as solo care-givers, while men are both more prone to share care-giving and to outsource care-giving completely. These are bivariate findings, however, not controlling for age or actual health status, which is one of the main reasons behind women's higher involvement in spousal care (Feld et al., Reference Feld, Dunkle, Schroepfer and Shen2010).

Notes: N = 4,012 respondents aged 50 years or older, living in a common household with partner or spouse who has also completed the interview.Source: Survey of Health, Ageing and Retirement in Europe (SHARE) Release 6.0.0, Wave 6 (own calculations, weighted results).

Figure 1. Female and male solo, shared and outsourced care-giving.

Notes: N = 4,012 respondents aged 50 years or older, living in a common household with partner or spouse who has also completed the interview.Source: Survey of Health, Ageing and Retirement in Europe (SHARE) Release 6.0.0, Wave 6 (own calculations, weighted results).

Figure 2. Solo, shared and outsourced care-giving by country.

Figure 2 displays country differences. Common to all countries is the high share of solo spousal care-giving, which depicts the most popular ‘option’ (the darkest bars on the left-hand side). In all but four countries, namely Luxembourg, Spain, Poland and Germany, more than half of the individuals who live together with a spouse who receives care provide that care themselves without any further support. An overall high share of care outsourced to a third person (light grey bars on the right-hand side) can be found in Portugal, Italy, Spain and Poland; whereas informal outsourcing (often to a family member) occurs more often in Mediterranean and Eastern European countries, formal outsourcing is comparatively likely to occur in Switzerland, Belgium, Luxembourg and Portugal. Informally shared care-giving (dark grey bars, second from left) is most prominent in Eastern European and Balkan countries, namely Poland, Slovenia, Estonia and the Czech Republic. Formally shared care-giving (medium grey bars, middle) is most common in Luxembourg and Germany, followed by France, Sweden and Belgium. Contradicting common assumptions about the ‘Scandinavian welfare model’, formally shared care-giving is relatively rare in Denmark and less common in Sweden than in Germany or Luxembourg. Again, these findings are of a bivariate nature and will be replicated in a multivariate model.

Multivariate findings: individual-level factors

In a second step, we thus compare solo care-givers to persons who share care, either with other informal helpers or a home care service, and persons who outsource care completely to a third (formal or informal) care-giver (see Table 1). To illustrate our main effects (for gender, socio-economic status, and the interaction between welfare policy and socio-economic status), we also present figures with average marginal effects below.

Table 1. Solo versus shared versus outsourced care-giving

Notes: Multinomial logistic regressions with standard errors in parentheses. Respondents aged 50 years or older, living together with partner or spouse who has also answered questionnaire. Ref.: reference category.

Source: Survey of Health, Ageing and Retirement in Europe (SHARE) Release 6.0.0, Wave 6 (own calculations, unweighted results).

Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

As expected (Hypothesis 1), men are less likely to act as solo care-givers than women (for average marginal effects, see Figure 3). Men are more likely than women to both outsource care-giving and to share care-giving. However, the latter is only true for informally shared care-giving, i.e. care-giving shared with one's children or other family members, friends or neighbours. With respect to the odds of choosing formally shared care-giving over solo care-giving, men and women do not differ. With respect to outsourcing the care of their dependent spouse to a third party, men are more likely to choose both formally and informally outsourced care-giving arrangements than women. For the most part, the data primarily support our first hypothesis, as men are more likely to outsource the care of their frail spouses completely and to share care informally, albeit not formally.

Notes: Capped spikes represent 95 per cent confidence intervals. Reference category: female. Calculations based on models from Table 1.Source: Survey of Health, Ageing and Retirement in Europe (SHARE) Release 6.0.0, Wave 6.

Figure 3. Average marginal effects of gender (male).

Employed individuals are generally less likely to act as solo care-givers than individuals who are not employed. They have significantly greater odds of choosing informally shared care-giving over being the solo care-giver. Being employed is also positively associated with outsourcing, which is true for both types of outsourcing (formally and informally). These findings support Hypothesis 2a, insofar as being employed is positively associated with the likelihood of choosing informal care support, but not Hypothesis 2b, from which we would have expected employed care-givers to be more likely to choose formal care support. Employment entails both time restrictions and financial opportunity costs which might not only hinder solo care-giving but also influence the choice of care support. For employed persons whose spouse is in need for care, sharing care responsibilities with informal helpers might be chosen over sharing care with formal support, because informal helpers are more flexible at providing care-giving around one's work schedule. The current findings seem to suggest that the availability of time is a more important factor than opportunity costs when it comes to choosing care-giving arrangements.

Our third hypothesis addressed the role of socio-economic status, represented by country-specific income quintiles and home-ownership. With respect to income, no clear picture emerges. Contrary to our expectations (Hypothesis 3a), higher income is not associated with a higher propensity to share or outsource care-giving formally, just as lower income is not significantly associated with a higher propensity to share care-giving informally. Further bivariate analyses (not shown) revealed that even without controlling for potential confounders, the expected association between income and choice of care support cannot be corroborated. One reason for this might be that income inequalities diminish after retirement age, as pensions are often subject to redistribution. A second reason might be that the effect of financial resources on choice of care-giving arrangement depends on the welfare policy context (see Broese van Groenou et al., Reference Broese van Groenou, Glaser, Tomassini and Jacobs2006).

To check the consistency of this finding, we dichotomised the income variable with the bottom two quintiles taking on the value ‘1’ in the newly created dummy variable. Figure 4 depicts the average marginal effects for the dichotomous income variable as well as the dummy variable pertaining to home-ownership. There is a noticeable tendency among people with higher earnings to care alone (left panel of Figure 4), as well as a significantly higher propensity towards informal outsourcing in the lower ranks of the respective national income distributions, which is in line with our hypothesis (Hypothesis 3b) and corresponds to the literature (Blomgren et al., Reference Blomgren, Breeze, Koskinen and Martikainen2011; Brandt et al., Reference Brandt, Haberkern and Szydlik2009). Our second indicator for socio-economic status is home-ownership (right panel of Figure 4). For home-owners, we find a significantly higher propensity to share care-giving formally, which is in line with our expectations (Hypothesis 3a). Living rent-free frees up financial resources, e.g. pension income, that can be spent on formal care-giving services without negatively impacting the household's standard of living. Owning living space might, moreover, allow hiring a live-in care-giver, who can be accommodated easier the larger the property.

Notes: Capped spikes represent 95 per cent confidence intervals. Calculations based on models from Table 1. Lower household income: first and second income quintile. N = 4,012 individuals.Source: Survey of Health, Ageing and Retirement in Europe (SHARE) Release 6.0.0, Wave 6.

Figure 4. Average marginal effects of income and home-ownership.

Further factors influence the decision for solo versus shared care-giving. They include the age and functional health of both the respondent and his or her spouse and the help that is available from networks. Formally shared and formally outsourced care-giving occur more frequently when both spouses are old, and when the care-receiving spouse has many health limitations. Informally outsourced care-giving occurs more frequently the more daughters there are. When couples already receive IADL help from outside the household, there is a markedly higher propensity for sharing and outsourcing care.

Multivariate findings: macro indicators

As we have already seen in Figure 2, countries differ considerably in the distribution of care-giving arrangements. These country differences remain significant even when controlling for a range of relevant independent variables (see Table 1). We can thus assume that the mechanisms behind these differences are not (only) located at an individual level or due to ‘composition effects’, such as country variation in income, living conditions or health. Rather, the significant coefficients and large effect sizes indicate that there are profound differences at the national level. We are interested in the link between public expenditures on two different elder-care schemes and the frequency of the occurrence of informally and formally shared and outsourced care-giving arrangements (see Figures 1 and 2 in the online supplementary material). We contrast two welfare state characteristics, namely Cash-for-Care and Care-in-Kind. Above-average expenditures for Care-in-Kind can be found in Sweden and Denmark. Above-average expenditures for Cash-for-Care are found in Italy, Greece, Portugal, Slovenia, France and Austria. In Luxembourg, Estonia and Switzerland, expenditures for both schemes are comparably low (for an overview, see Table 2 in the online supplementary material). In the next step, we therefore calculate our models including macro indicators containing information on the respective elder-care systems in each country (see Table 2).

Table 2. Cross-level interactions: gender and socio-economic status

Notes: Multinomial logistic regression models with standard errors in parentheses. Respondents aged 50 years or older, living together with partner or spouse who has also answered questionnaire. N = 4,012 in 17 countries. Separate models for each indicator and interaction, controlling for individual-level variables from Table 1. Pseudo-R ² = 0.14.

Source: Survey of Health, Ageing and Retirement in Europe (SHARE), Release 6.0.0, Wave 6; Organisation for Economic Co-operation and Development (2017) (own calculations, unweighted results).

Significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

Table 2 shows the effects of the macro indicators and their respective interaction terms with gender and socio-economic status introduced into the models of Table 1 (separate models for each interaction term). First, we look at the ‘pure’ effects without any interaction terms. Consistent with the respective policy goals, we find that Cash-for-Care schemes increase informally outsourced care-giving arrangements (a third person is paid to provide care). Contrary, the likelihood of sharing or outsourcing the care of one's spouse informally decreases when larger sums are committed to Care-in-Kind services, which is, too, in line with the associated policy goals such as unburdening non-resident family members, particularly children. However, the odds of choosing to share or outsource care formally over caring alone are not significantly higher in contexts with high expenditures on Care-in-Kind schemes. Care-in-Kind can thus be interpreted as an inclusive measure which does not ‘crowd out’ spouses as care-givers.

To test our fourth and fifth hypotheses, we estimated four separate models for each cross-level interaction. We assumed the influence of Care-in-Kind expenditures to be more gender-neutral, i.e. that men and women do not differ with respect to their choice of care-giving arrangements, than the influence of Cash-for-Care expenditures (Hypothesis 4). Moreover, we expected the effect of Care-in-Kind expenditures to be independent of individuals’ socio-economic status (Hypothesis 5), while we expected the influence of Cash-for-Care expenditures to differ between individuals of lower and higher socio-economic status. As the interaction terms between gender and the macro indicators are not significant at the 0.05 level, this indicates that the effects of both schemes on the choice of care-giving arrangements mostly do not differ between men and women. This was expected for Care-in-Kind, albeit Cash-for-Care seems to be gender-neutral, too. At a first glance, this might contradict the findings by Schmid et al. (Reference Schmid, Brandt and Haberkern2012), where welfare state expenditures display gendered impacts on adult children's propensity to care for their parents. Spousal care-giving might, however, constitute an exception to this. As spousal care already features a relatively high share of male participation, it is plausible that care policies may reach out to care-giving husbands and wives in a comparable way.

With respect to differences between individuals of lower and higher socio-economic status, a number of significant interaction terms indicate that these expenditures have a different impact on low-income and high-income households, as was expected from Hypothesis 5. To ease interpretation, we display predicted probabilities for sharing informally and formally (see Figure 5). In higher-income households, increased cash allowances for care-givers do not lead to higher rates of sharing care formally. In contrast, the likelihood of spending the cash allowances received on formal care (either through sharing or through outsourcing) is significantly lower for households with low income (top right panel). Care-in-Kind, as expected, does not seem to impact differently on lower- and higher-income households (lower panels). With regard to outsourcing care, higher Cash-for-Care expenditures are associated with a higher probability of wealthier households outsourcing care formally (see Figure 3 in the online supplementary material). Together with the findings pertaining to sharing care, this indicates that there are indeed differences between higher and lower socio-economic groups regarding the impact of welfare policy on the choice of a care-giving arrangement.

Notes: Capped spikes represent 95 per cent confidence intervals. The x-axis indicates the percentage of expenditure for the respective scheme as a share of total Gross Domestic Product; the y-axis indicates the prevalence of the care-giving arrangement. Calculations are based on separate models with cross-level interactions from Table 2. Lower household income: first and second income quintile; higher household income: third to fifth income quintile. N = 4,012 individuals.Source: Survey of Health, Ageing and Retirement in Europe (SHARE), Release 6.0.0, Wave 6; Organisation for Economic Co-operation and Development (2017).

Figure 5. Predicted probabilities of shared care-giving arrangements over Cash-for-Care and Care-in-Kind policies, by household income.