Largely due to population aging and the associated greater need for long-term care (Organisation for Economic Co-operation and Development [OECD], 2011), policy arrangements in many developed countries are being reconsidered (Pavolini & Ranci, Reference Pavolini and Ranci2008). Normative beliefs about what is appropriate with regard to caregiving are crucial in the care policy debate (Hochschild, Reference Hochschild1995), in addition to concerns about public expenditure and quality of care (Mot, Reference Mot2010). Public attitudes largely determine whether planned policy reforms achieve intended effects (Svallfors, Reference Svallfors, Castles, Leibfried, Lewis, Obinger and Pierson2010a). Mau (Reference Mau2004) has argued that “any reform attempt will be more likely to be successful if it possesses a good deal of moral plausibility, that is, if it responds to people’s moral assumptions of how societal contingencies should be collectively dealt with and how burdens and benefits should be distributed” (p. 69, italics added).
Scholars have examined specific aspects of normative care beliefs, such as filial responsibility (i.e., the generalized expectation that children should support their older parents when they are in need) (Dykstra & Fokkema, Reference Dykstra and Fokkema2012; Gans & Silverstein, Reference Gans and Silverstein2006), or the extent to which individuals perceive the state as responsible for financing care for the frail old (Deeming & Keen, Reference Deeming and Keen2003). These different aspects of normative care beliefs are largely discrete. Daatland and Herlofson (Reference Daatland and Herlofson2003) found, for instance, that norms regarding filial obligations and welfare state orientations are only weakly associated. Hence, a multifaceted approach is required to fully grasp normative care beliefs.
The first aim of the study is to distinguish care ideals in an attempt to capture multiple dimensions of normative care beliefs simultaneously. Drawing on an essay by Hochschild (Reference Hochschild1995), we distinguish three key dimensions along which care ideals differ: a state dimension, a family dimension, and a gender dimension. In a large sample of the Dutch population, we aim to identify care ideals with characteristic patterns for the three dimensions. The second aim of this study is to assess shifts in these care ideals over time in the early twenty-first century. With women’s engagement in paid work increasingly becoming the norm, a shift towards care ideals in which men and women have similar roles, and in which family members have a restricted caring role, is to be expected. In what follows, we argue that such a shift might weaken the moral plausibility of Dutch long-term care (LTC) policy, given ongoing policy changes.
In many developed countries, LTC arrangements are being reconsidered, with policy makers increasingly seeking ways to activate and maintain family members as caregivers (Chappell, Reference Chappell1993; Österle & Rothgang, Reference Österle, Rothgang, Castles, Leibfried, Lewis, Obinger and Pierson2010; Pavolini & Ranci, Reference Pavolini and Ranci2008). The Netherlands is exemplary with regard to this development. The country has historically had generous LTC arrangements, but in the past two decades these arrangements have been reformed in order to contain costs. Concomitantly, Dutch policy makers have been encouraging family members to take on an active role in caring for dependent relatives (Morée, van der Zee, & Struijs, Reference Morée, van der Zee and Struijs2007; van den Broek, Reference van den Broek2013). Sustained moral plausibility of Dutch LTC policy calls for shifts in care ideals away from rather than towards care ideals in which men and women have similar roles and in which family members have a restricted caring role. In this study, we assess how care ideals have effectively shifted in the Netherlands in the early twenty-first century to gain insight into the moral plausibility of the ongoing developments in Dutch LTC policy.
Background
Hochschild’s Cultural Ideals of Care
Hochschild’s (Reference Hochschild1995) typology of cultural ideals of care is unique in that it addresses multiple dimensions of normative care beliefs simultaneously. Unfortunately, her typology lacks an empirical assessment of its validity. Hochschild distinguished four care ideals: traditional, cold-modern, warm-modern, and postmodern. She applied these care ideals not just to care for the frail old but also to care for young children. Three key dimensions can be distinguished in Hochschild’s ornate descriptions of her four cultural ideals of care: (a) the extent to which the state is deemed responsible for the provision of care; (b) the extent to which the family is deemed responsible for the provision of care; and (c) whether or not men and women are deemed to be equally involved in family caregiving.
Hochschild’s (Reference Hochschild1995) typology has received criticism for being normative a priori (e.g., Kremer, Reference Kremer2006). Hochschild seems to have had most affinity with the warm-modern care ideal. We do not disagree with this criticism, but believe that it does not disqualify the typology of care ideals as a promising theoretical starting point for the study of multiple aspects of normative care beliefs in conjunction. In the following, we will therefore build on Hochschild’s labels, including their elements that may be considered value-laden: for example, “cold” and “warm”. A simple schematic overview of Hochschild’s four care ideals is presented in Table 1.
Schematic overview of Hochschild’s four cultural ideals of care
Individuals adhering to a traditional care ideal believe that the family carries the principal responsibility regarding care for those in need. They embrace the male breadwinner model and feel that women should stay at home and provide unpaid care to family members with care needs. In this care ideal, the state’s responsibility for care provision is limited: family members – more specifically, female family members – are the main providers of care, and the state’s role is only to enable and support family caregiving.
Diametrically opposed to the traditional care ideal is the cold-modern care ideal. Individuals adhering to a cold-modern care ideal believe that providing care to those in need is primarily the responsibility of the state rather than that of the family. They also feel that men and women alike should be in the workforce rather than take on care tasks. The substitution of family caregiving by publicly supported LTC services is often spoken of in terms with negative connotations, such as “crowding out” (Schlesinger, Reference Schlesinger2012). However, as Greene (Reference Greene1983) argued, whether this kind of substitution is negative or positive depends on what one perceives to be the goal of state involvement in caregiving. Individuals adhering to a cold-modern care ideal believe reduction of family involvement in caregiving is desirable. To illustrate the cold-modern care ideal with regard to care for the frail old, Hochschild (Reference Hochschild1995) stated that “[how] much of [an] older person’s life is to be spent in institutional care is a matter of degree, but the cold-modern position presses for maximum hours and institutional control” (p. 340).
State involvement in care is also crucial in the warm-modern care ideal. Unlike those adhering to a cold-modern ideal, however, individuals adhering to a warm-modern care ideal believe that care responsibility should be shared between the family and the state. They also value equal involvement of men and women in family caregiving and that both should be enabled to combine caregiving with participation in the labor market. In a warm-modern care ideal, family involvement in caregiving is considered important because it assures a level of warmness in the care provided to those in need. Individuals with a warm-modern care ideal believe that the state is responsible for a share of the care tasks so that, for family members, the burdens associated with caregiving are limited. Realization of this care ideal manifests itself in forms of care in which family members and formal caregivers jointly engage, such as when relatives and home care professionals share the care for community-dwelling older adults with functional limitations (Sims-Gould & Martin-Matthews, Reference Sims-Gould and Martin-Matthews2010). This division of labour between family and state is consistent with task-specific theory (Brandt, Haberkern, & Szydlik, Reference Brandt, Haberkern and Szydlik2009; Litwak, Reference Litwak1985), which holds that joint engagement of family and state enables a division of labor that allows different kinds of caregivers to provide the type of support for which they are best equipped.
In a postmodern care ideal, Footnote 1 the responsibility of adequately arranging care first and foremost rests with those in need themselves, rather than with the family or the state. Individuals adhering to this care ideal expect neither women nor men to participate strongly in unpaid caregiving. Neither do they perceive the state as an entity with large caring responsibilities. Manifestation of the postmodern care ideal would thus result in very low aggregate levels of care provision. The cognitive dissonance that women, in particular, experience between the demands of a career in paid work and the feeling of responsibility towards relatives in need of care (Aronson, Reference Aronson1990) is reduced by downplaying the latter (Hochschild, Reference Hochschild1994). This can, for instance, be done by portraying frail older adults “as ‘content on their own’” (Hochschild, Reference Hochschild1995, p. 339).
Dutch Context: Rising Female Labor Participation and Re-familialization
Life course theory holds that changing times tend to be reflected in persons’ lives (Elder, Reference Elder1994; Hagestad & Neugarten, Reference Hagestad, Neugarten, Binstock and Shanas1985). As argued previously, we expect shifts in care ideals over time given the decline of the male breadwinner model in which men engage in paid work and women take on the role of homemaker and unpaid caregiver. The Netherlands has seen a dramatic rise in women’s participation in paid work in the second half of the twentieth century and the first decade of the twenty-first century (Janssen & Portegijs, Reference Janssen, Portegijs, Merens, van den Brakel, Hartgers and Hermans2011; van Doorne-Huiskes & Schippers, Reference van Doorne-Huiskes and Schippers2010). The labour force participation rate of Dutch women is today above the European Union average (European Institute for Gender Equality [EIGE], 2013), but Dutch women tend to work part-time. In the early twenty-first century, the average number of hours worked has slightly increased (Janssen & Portegijs, Reference Janssen, Portegijs, Merens, van den Brakel, Hartgers and Hermans2011).
Women’s rising engagement in paid work is arguably reflecting modernization of social norms regarding gender relations and the division of labor between men and women (Pfau-Effinger, Reference Pfau-Effinger1998). According to Vlasblom and Schippers (Reference Vlasblom and Schippers2004), in the Netherlands as well as in other European countries, norms and values have changed in such a way that women are increasingly expected to participate in the labor market even when they are married or have children. Dutch women value paid labor as a means for self-development to the same degree as their male counterparts (Janssen & Portegijs, Reference Janssen, Portegijs, Merens, van den Brakel, Hartgers and Hermans2011). A 2008 survey conducted by the Netherlands Institute for Social Research indicated that young women, in particular, are ambitious with regard to paid work. The survey found that 6 out of 10 Dutch women, younger than 26 years old, seek promotion to a higher rank or position in their organization, and 8 out of 10 pursue a wage increase.
When women are more and more expected, by themselves and by others, to focus on a career in paid work, this is likely to be accompanied with an increasingly negative stance towards family caregiving. Although informal caregiving can be a positive experience for caregivers – for instance, by giving a sense of satisfaction or through learning new skills (Cohen, Colantonio, & Vernich, Reference Cohen, Colantonio and Vernich2002; Reinhard, Levine, & Samis, Reference Reinhard, Levine and Samis2012) – research suggests that providing informal care to a dependent relative hampers a career in paid work. Longitudinal studies conducted in Europe (Kotsadam, Reference Kotsadam2011), Australia (Berecki-Gisolf, Lucke, Hockey, & Dobson, Reference Berecki-Gisolf, Lucke, Hockey and Dobson2008), and the United States (Lee, Tang, Kim, & Albert, Reference Lee, Tang, Kim and Albert2015; Pavalko & Artis, Reference Pavalko and Artis1997) indicated that working women providing informal care are more likely to reduce working hours or to leave their jobs altogether. The changing social norms regarding female labor market participation and the risen ambitions in paid work of, in particular, young women can thus be expected to be accompanied with a shift towards care ideals in which men and women have similar roles and in which family members have only a limited caring role. In Hochschild’s terms, this is a shift towards cold-modern care ideals.
When men and women alike increasingly focus on paid work and wish to be freed from caregiving responsibilities, so-called “decommodified defamilialization” of care is called for. This entails widely available, affordable, publicly supported LTC services (Esping-Andersen, Reference Esping-Andersen1990; Reference Esping-Andersen1999; Lister, Reference Lister, Baldwin and Falkingham1994; Saraceno & Keck, Reference Saraceno and Keck2011). Along with the Nordic countries, LTC in the Netherlands has traditionally been characterized by a high degree of decommodified defamilialization (Saraceno & Keck). As noted previously, the Netherlands, along with many other European countries (Österle & Rothgang, Reference Österle, Rothgang, Castles, Leibfried, Lewis, Obinger and Pierson2010; Pavolini & Ranci, Reference Pavolini and Ranci2008), has been reforming its historically generous public LTC arrangements since the late 1990s and continuing in the early twenty-first century in order to contain costs (van Hooren & Becker, Reference van Hooren and Becker2012). Meanwhile, the substantial increase in the number of informal caregivers has become a formal policy goal (Mot, Reference Mot2010). Increased pressure on family members to provide care to dependent relatives is, for instance, evident in definitions of certain forms of care as “usual care”: that is, “normal, daily care that nuclear family members or other people who share a household can be expected to provide to one another” (Centrum Indicatiestelling Zorg [CIZ], 2012, p. 9; authors’ translation). The usual care concept is formalized in a protocol, with the explicit intention to limit formal care provision (Morée et al., Reference Morée, van der Zee and Struijs2007). For individuals sharing a household, it restricts the entitlement to benefits under the Exceptional Medical Expenses Act (Dutch: Algemene Wet Bijzondere Ziektekosten, AWBZ), which aims to provide a general insurance covering the Dutch population against exceptional health care needs. Individuals sharing a household also have only limited access to publicly provided domestic assistance because municipalities entrusted with the execution of the Social Support Act (Dutch: Wet Maatschappelijke Ondersteuning, WMO) – which aims to offer support to people who need it to sustain their independence and participation in society – mostly use a protocol similar to the usual care protocol when assessing eligibility (Tuynman & Marangos, Reference Tuynman, Marangos, de Klerk, Gilsling and Timmermans2010).
The developments in Dutch LTC policy outlined here can be seen as manifestations of “re-familialization” rather than decommodified defamilialization. Given that family caregivers tend to be women, the LTC policy arrangements work out differently by gender (Saraceno, Reference Saraceno2010; Saraceno & Keck, Reference Saraceno and Keck2011). Policy makers encouraging family members to take on a caring role are effectively addressing women (Schenk, Reference Schenk2013; van den Broek, Reference van den Broek2013). This is particularly the case with regard to intensive forms of support (Schmid, Brandt, & Haberkern, Reference Schmid, Brandt and Haberkern2012). Assuming that normative care beliefs show changes that match policy changes (Raven, Achterberg, van der Veen, & Yerkes, Reference Raven, Achterberg, van der Veen and Yerkes2011; Svallfors, Reference Svallfors2010b) and thus that the moral plausibility of LTC policy is sustained regardless of policy changes, we might expect to find a shift away from rather than towards cold-modern care ideals in which men and women have similar roles and in which family members have a restricted caring role.
Socio-demographic Predictors
Apart from examining shifts over time, we assess whether key predictors based on previous research distinguished specific care ideals. Gender is such a predictor. American findings show that women have stronger norms of filial obligation than men (Gans & Silverstein, Reference Gans and Silverstein2006), but studies conducted in Western European countries tend to find the opposite pattern (Daatland & Herlofson, Reference Daatland and Herlofson2003; Daatland, Herlofson, & Lima, Reference Daatland, Herlofson and Lima2011; Dykstra & Fokkema, Reference Dykstra and Fokkema2012; Herlofson, Hagestad, Slagsvold, & Sørensen, Reference Herlofson, Hagestad, Slagsvold and Sørensen2011). Dykstra and Fokkema (Reference Dykstra and Fokkema2012) argued that men may find it important that children care for aging parents merely in a theoretical sense. Daughters are more often expected to provide burdensome care tasks than sons (Finch & Mason, Reference Finch and Mason1991). For daughters rather than sons, valuing family caregiving is therefore more likely to imply a perceived personal duty to take on demanding care tasks. This leads us to expect that women are less likely than men to adhere to a traditional care ideal in which family members have the principal caring responsibility.
Life course experiences can also be expected to shape attitudes (Gans & Silverstein, Reference Gans and Silverstein2006; Poortman & van Tilburg, Reference Poortman and van Tilburg2005). The employed and the higher educated arguably have relatively strong feelings of autonomy, making them more likely to adhere to a care ideal in which the principal care responsibility rests with the individual rather than the family or the state (Daatland et al., Reference Daatland, Herlofson and Lima2011). In Hochschild’s terms, we expect them to be relatively likely to adhere to a postmodern care ideal. Research has shown that people with severe care needs tend to prefer receiving care from a professional, rather than from a family member (Daatland & Herlofson, Reference Daatland and Herlofson2003; Wielink, Huijsman, & McDonnell, Reference Wielink, Huijsman and McDonnell1997). Therefore, we expect them to be relatively unlikely to adhere to a traditional care ideal in which the family is considered principally responsible for the provision of care to the frail old.
A person’s family situation is likely to shape specific care ideals. For instance, divorce is detrimental for feelings of family obligations (Ganong & Coleman, Reference Ganong and Coleman1999), which may make divorced persons relatively likely to adhere to a care ideal in which the family carries a restricted caring responsibility. Parents may also be more likely than childless persons to adhere to such care ideals. Research has shown that older parents tend to have relatively weak feelings of filial obligations (Daatland et al., Reference Daatland, Herlofson and Lima2011; Herlofson et al., Reference Herlofson, Hagestad, Slagsvold and Sørensen2011), possibly because parents do not want to burden their own children with demanding care tasks (Cahill, Lewis, Barg, & Bogner, Reference Cahill, Lewis, Barg and Bogner2009). Gans and Silverstein (Reference Gans and Silverstein2006) have found that the death of the last living parent is associated with a substantial weakening of filial norms, presumably because the possibility of being a care recipient becomes more real following generational succession. In sum, we expect to find the divorced, those with children, and those who no longer have living parents to be particularly likely to adhere to care ideals in which family members have a restricted caring responsibility, such as a cold-modern care ideal.
Methods
To distinguish care ideals among the Dutch population and to identify shifts over time in care ideals, in our study we estimated latent class regression models with co-variates. Latent class analysis (LCA) enables the empirical identification of a multidimensional discrete latent variable from a cross-classification of two or more observed (or “manifest”) categorical variables (Hagenaars & Halman, Reference Hagenaars and Halman1989; McCutcheon, Reference McCutcheon1987). It distinguishes a set of mutually exclusive latent classes that account for the distribution of cases across all scores on the joint observed discrete variables. An alternative method for categorization into classes is cluster analysis, but, contrary to LCA, this method relies on arbitrarily predetermined cut-off points, and the choice of these cut-off points greatly influences results (Fonseca, Reference Fonseca2013). In LCA, the relationship between the latent variable and the observed variables is furthermore probabilistic, rather than deterministic (Hagenaars & Halman, Reference Hagenaars and Halman1989).
To predict class membership, we estimated latent class regression models that allow the prior probabilities of belonging to various latent classes to vary as a function of a set of observed co-variates (Linzer & Lewis, Reference Linzer and Lewis2010). Rather than calculating the predicted scores on the latent variables and subsequently treating these as observed dependent variables in a regression model as is commonly done, we estimated the coefficients on the co-variates simultaneously as part of the latent class model. The advantage of this approach is a reduction of bias in the coefficient estimates (Bolck, Croon, & Hagenaars, Reference Bolck, Croon and Hagenaars2004). However, a downside is that – depending on the number of latent classes, the number of manifest variables, and the number of these variables’ categories – only a limited number of co-variates can be included before models become unidentified (Linzer & Lewis, Reference Linzer and Lewis2010).
We used the poLCA package in R (Linzer & Lewis, Reference Linzer and Lewis2010), which uses the expectation-maximization (EM) algorithm to estimate the latent class model by maximizing the log-likelihood function (see Dempster, Laird, & Rubin, Reference Dempster, Laird and Rubin1977). The iterative nature of the EM algorithm allows poLCA to estimate LCA-models with missing observations on manifest variables (Linzer & Lewis, Reference Linzer and Lewis2010). A known problem of the EM algorithm is that, depending upon the initial parameter values chosen in the first iteration, it may find only a local rather than the global maximum of the log-likelihood function (McLachlan & Krishnan, Reference McLachlan and Krishnan1997). In order to locate the estimated model parameters that correspond to the model with the global maximum, rather than a local maximum, we estimated each model 500 times with different starting values.
We began with a model with two classes and subsequently kept adding classes until an additional class no longer improved the model fit. Given the number of manifest variables, the number of these variables’ categories, and the number of co-variates we wanted to include in our model, a model with five or more classes would be unidentified (see Linzer & Lewis, Reference Linzer and Lewis2010). For that reason, we estimated models with up to four classes. To determine whether a model with an added class had a better model fit than the model with one class fewer, we compared the Bayesian information criteria of both models (Schwarz, Reference Schwarz1978). When this procedure indicated that a four-class model fit our data better than a model with fewer classes, we estimated LCA-models without co-variates, starting with a four-class model and, again, kept adding classes until an additional class no longer improved the model fit. Given the number of manifest variables and the number of these variables’ categories, it was possible to estimate LCA-models without co-variates with up to seven classes before the model became unidentified (see Linzer & Lewis, Reference Linzer and Lewis2010).
Data
Our data came from the public release file of the first and third wave of the Netherlands Kinship Panel Study (NKPS). We did not use second-wave data because the indicator for the gender dimension of care ideals was not available in the second wave. In the first wave, 8,161 men and women, aged 18–80, and living in private households, were interviewed between 2002 and 2004 (Dykstra et al., Reference Dykstra, Kalmijn, Knijn, Komter, Liefbroer and Mulder2005). The overall response rate in Wave 1 was 45 per cent, which is lower than rates obtained in other countries, but comparable to that of other large-scale family surveys in the Netherlands.
We restricted our sample to the 4,390 respondents who were still present in the panel during Wave 3. Data collection for this wave took place in 2010 and 2011. The Wave 3 sample significantly differs from the Dutch population at large with respect to important socio-demographic characteristics (see Merz et al., Reference Merz, Dykstra, Hogerbrugge, Komter, Liefbroer and Mulder2012). Women are overrepresented, with about 60 per cent of respondents being female. The distribution of age ranges is skewed, with those in the middle age ranges overly likely to be included. The married and those living with children (except for single fathers) are over-represented in the NKPS data, while those living alone or with their parents are under-represented. The degree of urbanization of the respondents’ living environment matches that of the Dutch population at large. The distribution by region is also quite representative for the Dutch population, with the East slightly over-represented and the West somewhat under-represented.
We excluded respondents with missing values on any of our models’ co-variates in either of the waves, leaving a final sample of 4,186. We randomly selected one observation per respondent, effectively turning our panel data into a repeated cross-section. By doing so, we avoided violating the assumption of non-independence underlying our analyses. To check the robustness of our findings, we repeated the procedure of randomly selecting one observation per respondent five times and estimated our models on each of the five additional samples. Footnote 2
Measures
Manifest Variables
We aimed to identify latent classes underlying the responses to four survey questions. As an indicator for the extent to which the state is deemed responsible for the provision of care, we used the question that sought to determine whether the respondent considered care for the elderly more of a task for the family or more of a task for the government. Answering categories were “primarily a task for the government”, “(somewhat) more a task for the government”, “(somewhat) more a task for the family”, and “primarily a task for the family”. A similar question has been used in earlier studies on the relative responsibility of state and family for the care of older persons (see Herlofson et al., Reference Herlofson, Hagestad, Slagsvold and Sørensen2011).
The state dimension was measured only relative to the extent to which the family was deemed responsible for the provision of care. Proper interpretation, therefore, required also taking the family dimension into account. Respondents were asked to what extent they agreed with two statements: “Children should look after their sick parents” and “In old age, parents must be able to live in with their children”. Similar questions have been used in earlier studies on filial responsibility (Dykstra & Fokkema, Reference Dykstra and Fokkema2012; Gans & Silverstein, Reference Gans and Silverstein2006; Herlofson et al., Reference Herlofson, Hagestad, Slagsvold and Sørensen2011). For both statements, the response categories were as follows: “strongly agreed”, “agreed”, “neither agreed, nor disagreed”, “disagreed”, and “strongly disagreed”. To ensure a manageable number of cells in our data matrix, we collapsed response categories (Hagenaars & Halman, Reference Hagenaars and Halman1989; Hogan, Eggebeen, & Clogg, Reference Hogan, Eggebeen and Clogg1993). Though collapsing categories implies a loss of information, using all answer categories would produce unacceptably sparse data (van Gaalen & Dykstra, Reference van Gaalen and Dykstra2006). For each statement, we created a categorical variable with three categories instead of the original five. Respondents who “disagreed” or “strongly disagreed” with the statement were assigned to the first category, those who “neither agreed, nor disagreed” were assigned to the second category and those who “agreed” or “strongly agreed” were assigned to the third category.
Given the non-availability of a direct measure of gender attitudes with regard to caregiving, we used a measure indicating whether a respondent believed that, within the family, it was the man’s task to provide income to capture the gender dimension of care ideals. We did so because ideas about gender and caregiving mirror ideas about gender and work in all four of Hochschild’s (Reference Hochschild1995) cultural ideals of care. Respondents were asked who – in a family made up of a father, a mother, and children – should carry out the task “earning money”. The answering categories were “primarily the father”, “both equally”, and “primarily the mother”. Again, we collapsed categories, merging “both equally” and “primarily the mother”. It should be noted that less than one per cent of our respondents indicated that, within a family such as described, it was primarily the mother’s task to earn money. For Wave 1, this question was directed only to a subsample of 1,369 NKPS respondents, who served as control group for the Social Position and Use of Welfare Provisions by Migrants survey (Dutch: Sociale Positie en Voorzieningengebruik van Allochtonen, SPVA), which was commissioned by the Dutch Minorities Integration Policy Department and conducted in 2002 and 2003 (Dykstra et al., Reference Dykstra, Kalmijn, Knijn, Komter, Liefbroer and Mulder2005).
Co-variates
To estimate whether dispositions for specific care ideals varied between the period 2002–2004 (Wave 1) and the period 2010–2011 (Wave 3), we included a dummy variable coded as 1 when observations were from Wave 3, and as 0 when observations were from Wave 1. Given that our data are derived from a panel, our respondents are older in Wave 3 than in Wave 1. By statistically controlling for respondents’ age in our models, we avoided estimating a time period effect that effectively captured an age effect.
We included a dummy variable for gender, coded 1 for women and 0 for men. Employment status was measured with a dummy variable indicating whether the respondent was employed. We coded it 1 for those who indicated that the status “working” applied most to their personal situation, and 0 for those who picked any of the alternative statuses: “unemployed or job seeking”, “homemaking”, “prolonged sick leave or occupationally disabled”, “studying, at school”, “retired (early)”, or “other”. An additional dummy variable was included to capture whether or not the respondent was higher educated. We coded it 1 for those with higher vocational, university, or post-graduate degrees, and 0 for those with lower levels of education. Another dummy variable was included to measure whether the respondent reported coping with a disability and/or a chronic disease. Those indicating that they had one or more prolonged illnesses, health disorders, or handicaps and that this restricted them lightly or severely in their daily activities were coded 1. Those indicating they had no prolonged illnesses, health disorders, or handicaps, or that they did not feel restricted in their daily activities despite their health issues were coded 0.
Marital disruption was measured with a dummy variable coded as 1 for those indicating that they were divorced and 0 for those who were either married, never married, or widowed. The presence of children was measured with a dummy variable coded as 1 for those with at least one child and 0 for those who were childless. We finally included a dummy variable indicating whether both parents had passed away (coded as 1) versus whether at least one parent was still alive (coded as 0).
Results
Descriptive information on the respondents is presented in Table 2. A comparison of Bayesian information criteria indicated that the model fit of our latent class regression model with four classes is better than that of the models with two or three classes. Estimation of LCA-models without co-variates but with a greater number of classes further indicated that adding a fifth class did not improve the model fit. Footnote 3
Descriptive statistics of respondent characteristics
Source: Netherlands Kinship Panel Study; n = 4,186.
SD = standard deviation
Estimated conditional probabilities of scores on our manifest variables on normative care beliefs are presented in Table 3. The most prevalent latent class is the one with a response pattern that is consistent with a warm-modern care ideal, which values joint engagement of family and state in caregiving. Members of this class are relatively unlikely to be outspoken about either state or family carrying the principal responsibility for care provision to the frail old as they have low probabilities on each of the two most extreme responses. They have a relatively high probability (62%) to regard care for the frail old as somewhat more of a task for the government than for the family, but this does not imply that they do not also perceive the family as carrying responsibility. They have a very low probability (0%) to disagree with the statement that adult children should care for sick parents. They have a high probability (71%) to have an undecided or neutral stance towards this statement and have a probability of 29 per cent to outspokenly agree.
Estimated class-conditional response probabilities
Source: Netherlands Kinship Panel Study.
Despite this moderately receptive stance towards family involvement in care for the frail old, warm-modern individuals tend to have strong reservations regarding children’s obligation to let frail old parents move in with them. This suggests that they believe that family members can only be expected to engage in forms of caregiving that do not excessively impact the privacy and the personal life of the family caregiver. The probability of outspoken agreement with the statement that parents should be able to live with their children is extremely low (0%). Members of the most prevalent class have a relatively low probability (8%) to perceive earning money as a task for men rather than for women. This suggests that they tend to believe that men and women should be equally involved in family caregiving.
The response pattern of the second latent class is consistent with a cold-modern care ideal, in which state involvement in caregiving is greatly valued and family caregiving is not. In this class, the probability to regard care for the frail old as primarily a task for the government is relatively high (36%), and the probability to perceive it as primarily (1%) or somewhat more a task for the family (11%) is low. The probability to disagree with the statements that children should care for sick parents (92%) and that parents must be able to live with their children (96%) is very high. The probability to perceive earning money as a task for men rather than for women is low (6%). Arguably, members of this class believe that male as well as female family members should be in the workforce, with the state taking full responsibility for the provision of care.
The response pattern of the third latent class is consistent with a traditional care ideal. Here, female family members are deemed responsible for the provision of care, and the state is perceived as an entity with only few caring responsibilities. Members of this class have a relatively high probability to regard care for the frail old as primarily (6%) or somewhat more a task (33%) for the family. They have a high probability of agreeing with the statement that children should care for sick parents (87%) and also tend to have a neutral (44%) or positive (39%) stance towards the idea that older parents should be able to move in with their children. They have a moderate probability of regarding earning money as a task for men rather than for women (28%), suggesting that they may also be relatively unsupportive of equal involvement of men and women in family caregiving.
The fourth latent class shows a response pattern that does not fully fit with any of Hochschild’s four cultural ideals of care. Somewhat consistent with a cold-modern care ideal, members of this class have a relatively high probability to regard care for the frail old as primarily a task for the government rather than for the family (35%) and a high probability of disagreeing with the statement that parents must be able to live with their children (91%). Inconsistent with a cold-modern care ideal, however, they are unlikely to have a clearly negative stance towards family involvement in caregiving for the frail old. The probability of agreeing with the statement that children should care for sick parents (30%) is about as high as the probability of disagreeing (33%) or of being undecided (37%). Remarkably, they have by far the highest probability of all classes to perceive earning money as a task for men rather than for women (82%). We label this class cold-traditional. Individuals adhering to this care ideal are not likely to greatly value family involvement in the provision of care to the frail old, making this care ideal, in Hochschild’s terms, cold. They tend to have traditional ideas regarding gender roles.
Results of the latent class regression model predicting class membership are presented in Table 4. Controlling for all other characteristics, the model predicts that in 2010–2011, when Wave 3 data were collected, the odds of having a warm-modern care ideal relative to a cold-modern care ideal were lower than in 2002–2004 during the data collection for Wave 1 (OR: .729, p < .05).
Results of latent class regression analysis predicting class membership
Source: Netherlands Kinship Panel Study; Reference category: cold-modern.
*p < .05; **p < .01; ***p < .001.
B = coefficient
SE = standard error
Furthermore, the model predicts that when controlling for all other characteristics, every year increase in age is associated with decreases in the odds of having a warm-modern (OR: .984, p < .05) or a traditional (OR: .941, p < .001) care ideal relative to a cold-modern care ideal. Compared to men, women have lower odds of having a warm-modern (OR: .760, p < .05), traditional (OR: .318, p < .001), or cold-traditional care ideal (OR: .438, p < .01) relative to a cold-modern care ideal. Those with a higher education degree are less likely than those without a higher education degree to have a traditional (OR: .651, p < .05) or a cold-traditional care ideal (OR: .228, p < .001) relative to a cold-modern care ideal. The employed are less likely than the jobless to have a traditional (OR: .600, p < .01) or a cold-traditional care ideal (OR: .482, p < .05) relative to a cold-modern care ideal. The odds of having a cold-traditional care ideal relative to a cold-modern care ideal are a factor 2.103 (p < .05) higher for parents than for childless individuals. Divorced individuals are less likely to have a cold-traditional care ideal relative to a cold-modern care ideal than the non-divorced (OR: .247, p < .05). We did not find that coping with a disability or chronic illness or that no longer having living parents were associated with a disposition for specific care ideals.
For easier interpretation of the findings, we calculated predicted probabilities for a “typical” woman or man in waves 1 and 3 to have each of the distinguished care ideals. We performed separate calculations for those with and those without a higher education degree. Age was set to the mean and the categorical predictor variables were set to the mode. Footnote 4 These predictions are depicted in Figure 1. Model predictions suggest that, there has been a shift away from the warm-modern care ideal and towards a cold-modern care ideal in the first decade of the twenty-first century. The predicted probability for a typical woman with (without) a higher education degree of adhering to a warm-modern care ideal was about 53 per cent (47%) in Wave 1, and it declined to a predicted probability of about 46 per cent (40%) in Wave 3. For a typical man with (without) a higher education degree, the predicted probability of having a warm-modern care ideal declined from 49 per cent (39%) to 42 per cent (32%) over the same time period. Concomitantly, the predicted probability of adhering to a cold-modern care ideal increased. The magnitude of this increase varied from two percentage points for men without a higher education degree to six percentage points for women with a higher education degree.
Predicted probabilities class membership
Discussion
Scholars have examined specific aspects of normative care beliefs, such as filial responsibility or the extent to which the state is deemed responsible for financing care for the frail old. We have argued that a multifaceted approach is required to fully grasp normative care beliefs. The first aim of the current study was to distinguish care ideals in an attempt to capture multiple dimensions of normative care beliefs simultaneously. The second aim was to assess how care ideals have shifted in the Netherlands in the early twenty-first century. The Netherlands is exemplary for the development in long-term care policy that has been taking place in many Western countries, in the sense that an ever stronger appeal to the family is made to take on a caring role.
Our analyses indicate that four care ideals can be distinguished in the Netherlands, three of which are consistent with the cultural ideals of care presented by Hochschild (Reference Hochschild1995). The most prevalent latent class is consistent with a warm-modern care ideal. Individuals adhering to this care ideal value joint engagement of family and state in caregiving and tend to have egalitarian gender roles. We also distinguish care ideals consistent with, respectively, a cold-modern and a traditional care ideal. Individuals adhering to the former care ideal believe that women and men should be in the workforce, with the state taking full responsibility for the provision of care for the frail old. Those adhering to the latter care ideal believe that female family members are responsible for the provision of care and that the state is an entity with only few caring responsibilities. Furthermore, we find a care ideal not described in Hochschild’s typology. We label this care ideal “cold-traditional”. Those adhering to this care ideal are traditional with regard to gender roles. They also believe that the state is primarily responsible for the provision of care to the frail old and that the family has only a restricted caring role. The cold-traditional care ideal is consistent with what Havenaar (Reference Havenaar2006) calls new conservatism. A central aspect of new conservatism is the great value ascribed to maintaining a generous welfare state. Havenaar argues that new conservatism is a continental European phenomenon. This may explain why Hochschild – who wrote about the American context – did not describe a care ideal similar to our cold-traditional one in her typology.
We did not find a latent class consistent with Hochschild’s postmodern care ideal. In a postmodern care ideal, neither state nor family but rather the persons in need themselves carry the principal responsibility of adequately arranging care. Individuals adhering to such a care ideal would therefore find it difficult to identify either state or family as carrier of the principal responsibility for care provision to the frail old. This would result in a response pattern with low probabilities on the two most outspoken responses on the manifest state-versus-family item. A postmodern response pattern would further be characterized by high probabilities of disagreement with the statements that children should care for sick parents and that parents must be able to live with their children. Individuals with a postmodern care ideal would also be unlikely to regard earning money as a task for men rather than for women. The fact that we did not find a postmodern care ideal among the Dutch population suggests that gloomy presentations of contemporary society as a society where individualism thrives and solidarity has disappeared are exaggerated.
We expected the rising female labor market participation in the Netherlands to be accompanied with shifts in care ideals. Consistent with our expectations, our findings indicate that, in the early twenty-first century, a shift has taken place towards a cold-modern care ideal in which state involvement in caregiving is greatly valued and family caregiving is not. An alternative explanation for this shift is that concerns among the population about the growing demands placed by the government on family members regarding the provision of care to relatives (de Klerk, de Boer, Kooiker, Plaisier, & Schyns, Reference de Klerk, de Boer, Kooiker, Plaisier and Schyns2014) has led the Dutch to emphasize the value of state involvement in care provision more strongly and to underline that there are limits to what can be demanded from family members (cf. Pierson, Reference Pierson1996).
The shift towards a cold-modern care ideal suggests a discrepancy between Dutch LTC policy and people’s normative care beliefs. Dutch policy makers increasingly aim to activate and maintain family members as caregivers, but our findings show a trend away from rather than towards care ideals in which the family has a prominent caring role. In Mau’s (Reference Mau2004) terminology, the moral plausibility of Dutch LTC policy may thus be declining. This apparent discrepancy should not be exaggerated, however. After all, Dutch older adults are still largely protected against unmet needs for care despite the LTC policy changes that have taken place over the past decades (Smits, van den Beld, Aartsen, & Schroots, Reference Smits, van den Beld, Aartsen and Schroots2014).
People who were younger than 18 or older than 80 when Wave 1 data were collected were not present in the Netherlands Kinship Panel Study. Particularly the absence of the last group should be kept in mind when researchers interpret the findings of the current study. The age 80-and-older segment of the Dutch population is growing. It is expected that by 2025 almost one million inhabitants of the Netherlands will be in this age range, of a total population of 17.5 million (Wobma, Reference Wobma2011). With our findings indicating that it becomes more unlikely to adhere to a warm-modern, relative to a cold-modern, care ideal with increasing age, the absence of this growing group of oldest-old in our sample implies that the shift away from warm-modern care ideals and towards cold-modern care ideals may be stronger than presented here. Future research, using broader age groups, is needed to assess whether this is the case.
The current study roughly pertained to changes in the first decade of the twenty-first century. Given that established normative beliefs can be hard to change (Svallfors, Reference Svallfors, Castles, Leibfried, Lewis, Obinger and Pierson2010a), this is a rather short period to study shifts. Furthermore, additional reforms in LTC policy have been implemented after the studied period, and more reforms are on their way. In 2015, the Exceptional Medical Expenses Act will be replaced by the Long-Term Care Act (Dutch: Wet Langdurige Zorg, WLZ). The aim of the Long-Term Care Act is to provide care to people who are in need of care 24 hours per day. Lighter forms of nursing care and most personal care services will be transferred from the Exceptional Medical Expenses Act to the Health Insurance Act (Dutch: Zorgverzekeringswet, ZVW). A small share of personal care services will be transferred to the Social Support Act. Municipalities will be responsible for the organization of support to inhabitants coping with limitations while performing activities of daily living. As a result, this form of care will no longer be a right to which those in need are entitled, but a social provision. Before taking on caring responsibilities, municipalities will first require individuals in need – provided that they have the financial means – to buy services on the market and to turn to family members and others in their personal networks for support.
Future studies should address the moral plausibility of the planned reforms. To gain proper insight, additional indicators on normative care ideals are needed. In this study, we interpreted low scores on the family and state dimensions of our care ideals as a sign that the respondent believed that the principal responsibility of adequately arranging care rested with those in need themselves. This operationalization of individual responsibility does not allow capturing combinations of individual, family, and state responsibility, and such a combination is exactly the direction in which Dutch policy is moving. Assessing the moral plausibility of the planned policy reforms will only be possible when data sets become available that not only include measures of the extent to which family and state are deemed responsible for care provision but also measures of the extent to which individuals in need themselves are deemed responsible.
While our results suggest a discrepancy between the Netherlands’ LTC policy and normative care beliefs of the Dutch population, it is important to acknowledge that public opinion tends to support more individual responsibility when care for the deserving and needy is guaranteed (van der Veen, Achterberg, & Raven, Reference van der Veen, Achterberg, Raven, van der Veen, Yerkes and Achterberg2012). The state still tends to be held responsible for the protection of individuals in need, but those in need are increasingly expected to reciprocate and to organize the fulfillment of their care needs themselves. Taking responsibility will be more straightforward for older adults in need of lighter forms of care than for those in need of more demanding forms of care. Concomitantly, the latter are more likely to be perceived as deserving than the former. Normative beliefs about lighter forms of care may therefore differ from normative beliefs about more demanding forms of care. Data that capture the intricacies of state, family, and individual responsibilities, in relation to gender as well as to deservingness and need, will enable future researchers to extend our approach and possibly further refine the care ideals distinguished in this study.
