Introduction
The goal of this paper is to examine family obligations, family support behaviour, and the links between them in samples drawn from two countries with dramatically different social welfare policy regimes. Though a number of studies use comparative European data to address related issues (Albertini, Kohli and Vogel Reference Albertini, Kohli and Vogel2007; Attias-Donfut, Ogg and Wolff Reference Attias-Donfut, Ogg and Wolff2005; Daatland and Lowenstein Reference Daatland and Lowenstein2005), limited recent work contrasts European and American families (seeGrundy and Henretta Reference Grundy and Henretta2006). We thus capitalise on comparable data from two fairly recent national surveys to contrast patterns of family support behaviour, family obligations, and their connection for a subsample of middle-generation adults in the United States of America (USA) and the Netherlands who have both adult offspring and aged parents. We compare the Dutch and American samples regarding their espoused obligations to support adult offspring and aged parents, their actual support behaviour, and the association between obligations and support.
Basically, there are two ways of doing comparative research of this type. The first is to treat findings from different countries as repeated confirmations of a theoretical model because it is assumed that the model holds across countries. The second is to start with the assumption of cultural specificity and to focus on ways in which policies or other macro-level indicators influence elements of a particular theoretical model. This paper takes the latter approach as we explicitly develop hypotheses regarding differences between the USA and the Netherlands in terms of family obligation, family support, and their connection to one another. Our rationale rests largely on the differences that exist in the social welfare systems of the two countries.
We draw upon Saraceno's (Reference Saraceno2010) classification of the ways in which legal norms and public provisions frame financial and caring obligations in families. A novel element of her approach is consideration of obligations both up and down family lines. Saraceno distinguishes three patterns. Familialism by default exists when there are few publicly provided alternatives to family care and financial support. Supported familialism is when policies, usually through financial transfers, augment families in maintaining their financial and caring responsibilities. De-familialisation applies when the individualisation of social rights (e.g., with regard to minimum income provision, unemployment benefits for the young, entitlement to higher education or to receiving care) limits family responsibilities and dependencies.
The Netherlands leans toward what Saraceno labels the de-familialisation pattern. The Dutch government offers a minimum pension for senior citizens, grants and loans to students, and, compared to other European Union nations, fairly high coverage of institutional and at-home care for frail older adults (Saraceno and Keck Reference Saraceno and Keck2010). Though obligations to provide financial support to parents and adult children exist in law, enforcement is rare (Millar and Warman Reference Millar and Warman1996). The obligations specified in the civil code have been ‘overtaken’ by individual and couple-based social security laws.
The USA, in contrast, is viewed as a welfare state ‘laggard’ (Quadagno and Street Reference Quadagno and Street2006) with its ‘relatively restricted range of social protections and services, meager income benefits, and few programs as a right of citizenship or residence’ (Olsen Reference Olsen2007: 145). The USA lacks an adequate income support system for families and the elderly, and has extreme income inequality compared to other developed nations (Smeeding Reference Smeeding2005). Most of the limited social programmes in the USA are means-tested, including Medicaid, which offers the only publicly funded long-term care for the frail elderly (Salganicoff et al. Reference Salganicoff, Cubanski, Ranji and Neuman2009) and TANF (Temporary Assistance to Needy Families). TANF also has time limits, resulting in high poverty rates for single mothers and their children (Polakow Reference Polakow1997). Finally, the USA has a scarcity of education benefits: universal pre-school is non-existent (Schutz, Ursprung and Woßmann Reference Schutz, Ursprung and Woßmann2008) and limited public support for post-secondary education means that college attendance correlates strongly with family income (Huang et al. Reference Huang, Guo, Kim and Sherraden2010). The USA clearly models Saraceno's familialism by default social welfare system.
Given these sharp cross-national differences in social welfare policy for the Netherlands and the USA, we pose three questions for comparative study of Dutch and American adults and their families:
1. Do adults in the USA and the Netherlands differ in their expressed views regarding responsibilities to ageing parents and adult offspring?
2. Do American and Dutch adults differ in the actual support they provide these family members?
3. Does the connection between espoused family obligations and support behaviour differ for adults in the two countries?
Below we review theoretical formulations and empirical evidence that shape our hypotheses regarding these three questions.
Literature review
Family obligations and policy context
Family obligations are generalised expectations regarding family members' responsibilities for each other, which provide guidelines for family behaviour (Finch and Mason Reference Finch and Mason1990). In referring specifically to filial obligations, Finley, Roberts and Banahan claimed that they are ‘a product of the social and structural world in which a person lives’ (1988: 77). As such, the views individuals possess regarding support to family members reflect the legal and care systems of their countries. Indeed, support for norms regarding family obligation tends to be lower in generous welfare states (Daatland and Herlofson Reference Daatland and Herlofson2003; Dykstra Reference Dykstra2010).
Haberkern and Szydlik (Reference Haberkern and Szydlik2010) contend that policy provisions are not only consistent with the values and norms of a society, but that they can also have an effect on them. Their study, which classified the social organisation of care based on data on state care provision, legal obligations and opinions about family care from 11 Western European nations, revealed three clusters of countries. Scandinavian countries and the Netherlands have strong public care systems, where the state is largely responsible for providing care for ageing family members. Mediterranean countries, Germany and Austria have family-based care systems; few people perceive the state to be primarily responsible for older adult care. France and Belgium have arrangements falling between the public-based and family-based systems. Switzerland, on the other hand, does not clearly fit any of the clusters. It has extensive state-funded care, but the cultural norm is that family has primary responsibility for older adult care. A shortcoming of Haberkern and Szydlik's study is their sole focus on care for older family members. A more comprehensive test of cultural differences in family obligations would assess obligations to both older and young generations in the family, as the current study does. Additionally, Haberkern and Szydlik's work is limited to European nations, like most of the recent comparative work on family support patterns and obligations. This US–Netherlands study thus expands on such research.
Based on the Haberkern and Szydlik findings and Saraceno's discussion of the link between social welfare systems and family obligations, our first hypothesis – referred to as familialism by default, is that family obligations are stronger in the USA than in the Netherlands. Americans are expected to subscribe more strongly than Dutch adults to the belief that it is important to give support to family members in need in correspondence with the more limited public support available in the USA than in the Netherlands. Note that the direction of causality in this association is unclear; we cannot say whether the stronger feelings of family obligation are a consequence of welfare state arrangements or whether the cultural norm sets the context for limited availability of publicly funded care.
Family support behaviour and policy context
Our second research question asks whether American and Dutch adults differ in the actual support they provide to adult offspring and ageing parents. Competing views exist regarding the connection between family support patterns and the extensiveness of welfare systems. One view, labelled the ‘substitution hypothesis’, is that public assistance threatens family members' interest in and willingness to give support to those in need (Attias-Donfut and Arber Reference Attias-Donfut, Arber, Arber and Attias-Donfut2000). State supports supposedly ‘crowd out’ family support, resulting in less support from family members (Cox and Jakubson Reference Cox and Jakubson1995; Künemund Reference Künemund and Saraceno2008). An alternative view offered by Attias-Donfut and Wolff (Reference Attias-Donfut, Wolff, Arber and Attias-Donfut2000) is that public welfare systems do not necessarily interfere with family solidarity and may even promote it – what is known as the ‘complementarity hypothesis’. They note, for example, that when ageing individuals are assured a state pension, they may be compelled to assist their offspring rather than using and saving their resources to meet their own needs. Some scholars posit that when public support is generous, family members are more able to redistribute resources to assist those in need (Kohli et al. Reference Kohli, Künemund, Motel, Szydlik, Arber and Attias-Donfut2000), and more willingly share responsibilities or perform certain support tasks they feel especially capable of doing well (Brandt, Haberkern and Szydlik Reference Brandt, Haberkern and Szydlik2009; Künemund and Rein Reference Künemund and Rein1999; Lowenstein and Daatland Reference Lowenstein and Daatland2006).
Empirical evidence fails to support the substitution hypothesis, yet the notion that public welfare systems have a positive impact on family support is not fully supported by the data either. One study contrasting Japan and several Western countries revealed a positive association between levels of state support and family support (Künemund and Rein Reference Künemund and Rein1999). Daatland and Lowenstein's (Reference Daatland and Lowenstein2005) analysis of OASIS (Old Age and Autonomy: the role of service systems and intergenerational family solidarity) data from five European countries found that in countries like Norway, with high levels of public support, family members played a substantial role in care provision, although the authors noted that ‘the family is dominant when services are not available’ (2005: 181, our emphasis), as is the case in Spain. Thus, ‘generous welfare state services have not crowded-out the family, but may have reduced dependence on the family’ (2005: 181, our emphasis). Using data from 11 European countries, Brandt, Haberkern and Szydlik (Reference Brandt, Haberkern and Szydlik2009) revealed that public provisions allow a specialisation of help, with professional providers taking on time-consuming care and family members giving less intensive practical help. In countries with low levels of government-funded support, family members tend to assume physical care activities, leaving them less time for other types of help. Finally, Grundy and Henretta (Reference Grundy and Henretta2006) reported mixed findings in examining family support to both adult offspring and aged parents in the USA and Great Britain. They found that Americans were less likely than the British to provide help to ageing parents, but were slightly more likely than their British counterparts to assist adult offspring.
Following the observation that family support is more dominant when publicly funded services are not widely available, we offer the hypothesis that adults in the USA will be more supportive of family members than in the Netherlands. Yet, before testing this idea, which we label the family-steps-in hypothesis, it is essential to consider differences in geographic proximity as a potentially critical influence. Because the Netherlands is dramatically smaller than the USA and has greater population density, family support may be more easily provided in the Netherlands than in the USA as those who live nearby have more opportunity to offer support (Hank Reference Hank2007). For research question 2, we thus amend our family-steps-in hypothesis to state that American adults will be more likely to engage in family support behaviour than Dutch adults, but only after differences in geographic proximity have been controlled.
The connection between family obligations and support behaviour
Although family obligations and family support behaviour are distinct, an important question is to what extent individuals' espoused family obligations predict their family support behaviour. Research question 3 addresses whether responsiveness to stated family obligations differs for adults in the USA and the Netherlands. Only a few studies have considered the obligations–support connection, and then only in reference to respondents from a single country. Lee, Netzer and Coward (Reference Lee, Netzer and Coward1994) failed to demonstrate an association between obligations and support behaviour in a sample of Americans aged 65 and older. They found no connection between the support ageing parents received from their adult offspring and their views of filial responsibility. Yet, a more appropriate test of the question of the obligation–support link would have correlated levels of support received by the ageing parents with the adult offspring's espoused level of filial obligation.
More recent studies in both the USA and the Netherlands used this latter approach and found a significant connection between individuals' stated obligations and support behaviour. Analysing data from a sample of Dutch adults aged 55–89 and their adult offspring, Klein Ikkink, van Tilburg and Knipscheer (Reference Klein Ikkink, van Tilburg and Knipscheer1999) found that expressed obligations to parents positively predicted levels of support provided to parents. Longitudinal analyses of US data conducted by Silverstein, Gans and Yang (Reference Silverstein, Gans and Yang2006) also found that adults espousing stronger filial norms gave significantly more support to their parents, but only in the case of their mothers, not their fathers. Because these two studies used different measures and analytic models, and were conducted at quite different points in time (five to eight years separated data collections) they provide a weaker post hoc comparison of the USA and the Netherlands than the current study, at least in terms of the obligations–support connection. Further, neither study addressed support behaviour targeted at adult offspring in the family. Thus, whether family obligations differentially predict a broader set of family support behaviours in the Netherlands than the USA remains to be addressed.
Competing hypotheses are therefore formulated for research question 3. One possibility is that the link between obligation and support behaviour is stronger in the USA than in the Netherlands. Because of the country's limited public support system, US families are considered the first line of defence when needs arise. Consequently, compared to the Dutch, Americans may see it as more critical to act upon beliefs about family support because of the more severe consequences that may occur if one fails to follow through on obligation norms in the USA. This reasoning is consistent with the family-steps-in hypothesis. Alternatively, the connection between obligations and support behaviour may be weaker in the USA than in the Netherlands. This view, labelled the no choice hypothesis, is based on the reasoning that no matter what the circumstances are, the limited public support system in the USA gives Americans no choice but to assist family members in need. Because of fewer alternatives to family support in the USA than in the Netherlands, the link between support behaviour and family obligations is thus weaker in the USA.
Data and methods
This study draws on national data sets from the USA and the Netherlands. The US data are from the third wave of the National Survey of Families and Households (NSFH) (Sweet and Bumpass Reference Sweet and Bumpass2002), conducted from 2001 to 2003 (N=4,600 main respondents). The original NSFH study, completed in the mid-1980s, included main respondents (aged 19 and older) drawn from approximately 13,000 households in the contiguous USA. Budget constraints limited the third follow-up to original main respondents who had reached age 45 or older, or those whose target child (identified in wave one) was currently between the ages of 18 and 33. These selection factors are not problematic for this study given the criteria used to address our specific research questions (see below). The NSFH research team completed wave three interviews with 63 per cent of the eligible respondents. Sampling weights were not developed by the NSFH staff for wave three therefore the analyses herein are based on unweighted data.
The Dutch data are from the first wave of the Netherlands Kinship Panel Study (NKPS), conducted from 2002 to 2004 (N=8,161). The main respondents, who were aged 18–80, were drawn from a random sample of private addresses in the Netherlands. The overall response rate was 45 per cent, which is comparable to other family surveys in the Netherlands (Dykstra et al. Reference Dykstra, Kalmijn, Knijn, Komter, Liefbroer and Mulder2005) where response rates are generally low and appear to be declining over time (De Leeuw and De Heer Reference De Leeuw, De Heer, Groves, Dillman, Eltinge and Little2001; Stoop Reference Stoop2005). Adults under age 30, women residing alone and young adults living at home are under-represented in the sample. Primary respondents selected for the NKPS sample completed face-to-face interviews and a self-enumerated questionnaire. The latter had a 92 per cent return rate. Respondents also were asked for contacts for a randomly selected parent and two randomly selected children aged 15 and over. Co-operation with this request was lower (40 per cent of respondents refused to give that parent information, and 28.7 per cent refused such information on their children); these refusals resulted in substantially more missing data in the Dutch than American data set because some information on aged parents and adult offspring used in the analyses was drawn directly from the reports of these family members.
Although these two surveys were not conducted as part of a planned comparative study, they are a good choice for comparative research. The NKPS study fashioned many of its questions off the NSFH study so the two study protocols contain several similar questions regarding family obligations, support behaviour and characteristics of family members. Data collections for the two surveys also occurred close in time (both between 2001 and 2004), eliminating the possibility that period effects would confound the national comparisons.
Several selection factors were used to identify the analytic samples from these two studies. First, to study family support given up and down generational lines, only respondents with at least one parent and one adult child (age 19 or older) living outside the household were included. The analytic samples for both countries were also limited to main respondents between the ages of 40 and 79 to capture middle-generation adults of approximately the same ages in both countries. (Because sampling weights were not used, this criterion ensures a clear age group in each population to whom we can generalise our findings.) These selection criteria resulted in 1,232 cases for the US sample and 792 cases for the Netherlands sample who had complete data regarding provision of support to both aged parents and to adult children – our dependent variables.
Measures
These two data sets included three items pertaining to family obligations that were worded similarly enough to constitute comparable items for a measure of family obligation: two items referred to helping adult offspring (providing financial support and letting adult children live at home if they have financial problems) and one item from each data set addressed co-residence with ageing parents who can no longer live on their own. Having more than a single item addressing obligation to ageing parents would be preferable. Yet, we have confidence in the single item available to us because Kalmijn and Saraceno's (Reference Kalmijn and Saraceno2008) research that included a Dutch sample revealed that a similar co-residence item correlated highly with attitudinal items pertaining to looking after ageing parents (r=0.81) and paying for elderly parents' care (r=0.70). Our three items asked respondents the extent to which they agreed or disagreed (five-point scale) with the statements on family obligation. The two items referring to adult offspring were weighted by 0.5 so that views about adult offspring contributed equally to a summed index of family obligation as the one item regarding ageing parents. Index scores thus range from 2 to 10, with high scores reflecting stronger obligation to help family members. The Cronbach's alpha coefficients for the scales in the US and Dutch samples were 0.41 and 0.47, respectively. Though somewhat low, these reliability scores are comparable to those of other brief scales of family obligation used in the literature (e.g. Ward's 2001 four-item scale had a Cronbach's alpha=0.44). Moreover, all items used in each of the scales met the standard set by Kline (Reference Kline1986) that item-total scale correlations should be at least 0.20. Finally, the reliabilities are comparable for the two samples, thus eliminating confounds due to measurement differences.
Support to aged parents and adult offspring were each examined with three items addressing financial support and instrumental support (e.g. errands, transportation, house and yard help). The US survey asked respondents about providing support to each of their surviving parents and to each adult child during the past month. In the Netherlands study, one living parent of a respondent and up to two adult children were the foci of the support questions pertaining to transfers in the past three months. To institute comparability in the two data sets we used a random number generator to select one living parent (if more than one alive) and one adult child as the focus of the analyses. Though the timeframe used for considering support provision was shorter in the USA than the Netherlands survey, this is not highly problematic given methodological studies that show that when asked to report about interactional events that are not rare – in our data such events as helping around the house or yard, or providing transportation – people are biased to reporting the usual pattern of occurrence, making the timeframe relatively unimportant (Freeman and Romney Reference Freeman and Romney1987). As much as possible, we worked to make measures comparable. For example, in the US survey respondents reported financial help over $200 to either parents or adult offspring. If over that amount, they were asked a follow-up on the amount of money provided. In the Netherlands survey, respondents reported any support exceeding 500 Euros. Thus, when coding US data to indicate financial help, we only included those cases where the reported gift or loan was over $500 (in 2002, the Euro was equal to approximately $1.09). Finally, we examine support to ageing parents and to adult children with dichotomous dummy variables denoting ‘any support given’ to minimise comparison problems due to different timeframes and response choices in the two surveys.
Control variables
Comparing family obligations and support behaviour and their connection in the USA and the Netherlands requires consideration of other factors that may influence this association. As noted earlier, the influence of geographic proximity is a critical variable given marked differences in the geographic size and population density of these two countries. Several other structural factors may also differ between the samples for the two countries, which may influence the resulting association between obligations and behaviour, given their connection to support provision. Whether middle-aged respondents act on feelings of obligation may depend on the needs or demands of the receiving generation. Thus, we include in our analysis the ages of the adult child and aged parent, the sex (1=male, 0=female) of each, the aged parent's health status (1–5 scale, 5=excellent), and whether the parent lived alone (1=yes, 0=no). For adult offspring, other controls included their education level, full-time work status (1=35 or more hours/week, 0=fewer hours/week), and marital status (1=married, 0=unmarried). Preliminary analyses considered other forms of relationship status (cohabiting or married versus single) and employment status (e.g. any work versus not working; part-time, full-time, not working) for offspring with results showing that the selected options were most predictive of support outcomes. To control for proximity, distance of the adult child and aged parent from the respondent also were included; US distances were transformed from miles to kilometres for comparison purposes. These variables were logged to eliminate negative skew. Finally, a single-item rating of relationship quality between the respondent and parent was included because preliminary analyses revealed a higher likelihood of missing data for the aged parent in the Dutch sample in the event of a lower quality relationship. By including relationship quality in the models we control for this non-random element of selectivity in the missing data that we eventually impute (see below).
Several controls pertaining to respondent constraints that could influence support-giving were also included in the analyses. Among this set of variables were the respondent's education level, sex (1=female, 0=male), disability status (1=long-term illness, physical or mental condition, 0=none), age, net household income (converted into Euros and logged to reduce skew) and relationship status (1=live-in partner or spouse, 0=living outside a relationship). A variable indexing number of siblings was also included, as was a count of adult children (logged to eliminate skew). A few other possible control variables were assessed in preliminary analyses, including other types of marital status (e.g. respondent divorced) and respondent's work and retirement statuses. These variables did not contribute significantly to the analyses and were eliminated as analyses progressed.
Analytic plan
To address our questions the analyses start with bivariate comparisons of the American and Dutch samples on the espoused obligation scale and on likelihood of various support behaviours. These analyses address research questions 1 and 2, as to whether the two countries differ in terms of family obligations and support behaviour, respectively. Additionally, a set of descriptive analyses is conducted with controls for geographic proximity to eliminate its potential confound because the USA and the Netherlands vary dramatically in geographic dispersion and size. This analysis allows for testing of the conditional hypotheses in question 2. We then conduct multivariate analyses that permit us to test the influence of family obligations on family support behaviour, controlling for the demand factors that adult children and aged parents may possess (noted above), along with constraint factors (noted above) that may affect the ability of respondents to offer assistance to either ageing parents or adult offspring, or both. Multinomial logistic regression analyses are used; with this approach we examine the association between family obligations and other control factors and the likelihood or odds that respondents provide each of three different patterns of family support – all three of which are compared to the ‘provides no family support’ pattern. Specifically, the multivariate models test the likelihood of providing: (a) support to adult child only; (b) support to ageing parent only; and (c) support to both generations relative to the likelihood of providing no support to either generation. These analyses allow us to assess whether the link between family obligations and support behaviours differs for the American and Dutch samples.
Before conducting these analyses, the ICE programme for multiple imputation (Royston Reference Royston2005) in STATA was used to impute values for the missing data. Based on Acock's (Reference Acock2005) recommendations regarding the amount of missing data and suggested number of imputations, we generated five imputed data sets for the US data set and ten for the Netherlands data set. The multinomial logistic regression models presented below are based on the parameter estimates obtained from the multiple imputation and estimation process.
Results
We start by comparing background characteristics for the two analytic samples, as well as recipient demand factors and respondent constraint factors that may influence support behaviour (Table 1). Samples for the two countries are quite similar. Respondents in both samples are in their early fifties on average, though the Dutch sample is about one year older on average (54 versus 53). The majority of respondents in both samples are female and reports about parents are primarily about mothers, but more so in the Netherlands (77%) than the US sample (68%). Offspring considered in the study are more evenly split along male–female lines, but again the Dutch sample is slightly over-represented by females compared to the US sample. Most respondents from both samples are partnered (about three-quarters), have on average 3.5 siblings and two children. The US sample has a markedly higher household income than the Dutch sample, which is at least partially attributable to the higher tax rate in the Netherlands and the lower prevalence of dual-earner couples. Approximately one-quarter of both samples report a physical or mental disability that limits their activity. Parent's average age is 78 in the US sample and 82 in the Dutch sample. A substantial difference exists in the percentage of ageing parents that live alone, with it being substantially higher in the Dutch than American sample (77 versus 58%). Parents' health status and relationship quality are statistically different in the two countries, with somewhat higher reported relationship quality in the US sample, but better health among older parents in the Dutch sample. As expected, ageing parents and adult offspring live much further from the respondent in the USA than the Netherlands. Finally, the mean age of the third-generation offspring averages 28 in both samples. More offspring are married in the US sample than in the Dutch sample, and working full-time. The adult offspring in the US study also report higher educational attainment, on average, than those in the Netherlands survey. All of these variables are controlled in the multivariate analyses to eliminate potential confounds in the comparison of the obligation–support link for the two samples.
Sample comparisons on key study variables
Note: SD: standard deviation.
Significance levels: * p⩽0.05, ** p⩽0.01, *** p⩽0.001.
Research questions 1 and 2 address differences in stated family obligations and support provision for the Dutch and American respondents. Table 2 shows that the American respondents espouse much stronger obligations to support ageing parents and adult offspring, consistent with our familialism by default hypothesis. The average family obligation score for the US sample (6.77) is approximately one standard deviation (SD) higher than the mean score for the Dutch sample (5.42, SD=1.33).
Family obligation and support provision to aged parents and adult offspring, by country
Significance level: ** p⩽0.01.
Despite this marked difference in stated obligations to help family members, the likelihood that respondents in the Dutch sample reported assisting family members, either instrumentally or financially, in the recent past is markedly higher than it is for the American survey respondents. Table 2 reveals that this holds true regardless of whether we consider support to the younger or older generation, or both. Well over three-quarters of the Dutch respondents reported having helped either their ageing parent or the designated adult child recently, compared to just over half of American respondents. Interestingly, in both samples, the likelihood of helping one's aged parent is approximately equal to helping one's launched adult child. Yet, the Dutch respondents are nearly twice as likely (68%) to report such recent support behaviour as are the American respondents (35–36%). Even more striking is the difference in reported support provision to both generations in one's family; half of the respondents in the Netherlands sample reported this pattern of recent support, compared to just one-eighth of the US sample.
However, to test fully our family-steps-in hypothesis, which states that the limited public support available in the USA relative to the Netherlands puts more demand on families to help out, geographic proximity must be controlled. Therefore, in Table 3, analyses are restricted to respondents from each country who live nearby the specific family member in question – nearby here defined as within ten kilometres, which is the median distance between respondents and designated family members in the Dutch sample. By only looking at nearby family members in each sample, we are accounting for country differences in proximity, which could explain differences in support behaviour. In this analysis we report the proportion of Dutch and American respondents who provided financial, instrumental and any support to adult children and parents because differences in types of support may exist once geographic proximity was controlled. For example, respondents may reserve financial support for those family members who they cannot assist instrumentally because of distance. The figures presented in Table 3 reflect a similar pattern to those in Table 2. Except for the likelihood of providing financial support to family members – particularly adult offspring, where we see significant differences favouring the American respondents, Dutch respondents are more likely than their American counterparts to be engaged in support behaviour, even when geographic distance is controlled. Compared to American respondents, Dutch respondents are about 50 per cent more likely to be providing any type of help to adult offspring who live nearby, and about 20 per cent more likely to be involved in support behaviour with aged parents who are in close proximity. Chi-square analyses (not shown here) revealed that geographic proximity is a significant predictor of help to offspring in the USA but not the Netherlands, whereas it is highly predictive of support to aged parents in both countries. To highlight, Table 3 shows that controlling for proximity does not fully eliminate the Dutch–American differences in support-giving to family members. Thus, the family-steps-in hypothesis is not supported by Table 2, nor is it upheld even when geographic proximity is conditioned in Table 3. Yet, these results make it evident that geographic distance must be controlled in subsequent multivariate analyses addressing question 3.Footnote 1
Financial and instrumental support to aged parents and adult offspring for respondents living nearby1 family members, by country
Note: 1. Defined as within ten kilometres of the respondent's home.
Significance level: ** p⩽0.01.
The results of the multinomial logistic regression analyses presented next, in Table 4, permit a test of our competing hypotheses regarding the association between family obligations and support behaviour in the American and Dutch samples. Table 4 presents the odds ratios for predictors of the three different support patterns for each country, relative to a pattern of providing no support to either generation. An odds ratio less than 1.00 indicates a reduced likelihood of the event, whereas an odds ratio greater than 1.00 indicates that an increased likelihood of the event is associated with the independent variable. In Model I, the odds of respondents reporting support to adult offspring only, but not parents, is shown; Model II indicates the odds associated with each predictor as they relate to respondents providing support to ageing parents only; and Model III shows the odds associated with each independent variable in predicting support to both the ageing parent and adult child generations relative to providing no support to either generation. The models in Table 4 reveal differences in the association between family obligations and support behaviour for the USA and the Netherlands, accounting for between-country variations in geographic distance, as well as family background and situational factors that both influence demand for and constraints on family support.
Multinomial logistic regression predicting family support behaviour, by country
Notes: 1. Odds ratio predicting support to child only, relative to providing no support to either generation. 2. Odds ratio predicting support to parent only, relative to providing no support to either generation. 3. Odds ratio predicting support to parent and child, relative to providing no support to either generation.
Significance levels: * p⩽0.05, ** p⩽0.01, *** p⩽0.001.
The three left-hand columns in Table 4 indicate the odds of the three types of support patterns for the Dutch respondents, relative to providing no support to either adult child or parent. The most important finding for these three models is the lack of significance for the odds ratio pertaining to family obligation. Therefore, family obligation fails to predict support behaviour directed toward adult offspring, ageing parents, or both generations in the Dutch data. In contrast, in one of the models for the US sample shown in the right-most three columns we see a significant association between family obligations and support behaviour. Specifically, Model II reveals that a one-point change on the ten-point family obligation scale is associated with a 17 per cent increase in the likelihood that American respondents provide support to their ageing parents, versus neither generation. However, family obligation does not predict an increased likelihood of support to adult children only (Model I), or to both generations (Model III) in the American sample.
The other clear difference between the models for the Dutch and American samples is the overall number of factors that are significantly associated with the likelihood of family support behaviour. In the Dutch sample, many fewer factors are associated with provision of support by the middle generation than in the American sample. Child demand factors, specifically the adult child's age and education level, are the most consistent factors to significantly impact support-giving in the Dutch sample. When offspring are older, there is a significantly reduced likelihood that respondents have recently provided support to them, or to both generations. When the adult child is more highly educated, there is an increased likelihood of the respondent reporting recent support to the child generation, or to both the child and parent generations. Only geographic distance produces significant effects in the model predicting support to aged Dutch parents, with greater distance markedly reducing the odds of support-giving to ageing parents.
In the models for the American sample, more predictors reveal a significant association with the support outcomes. The significant factors associated with altering the odds of the adult child receiving support (Model I) are all in the child-demand domain. Specifically, increased distance between the respondent and adult child is associated with reduced odds of support, as is increased age of the child and full-time employment status. In Model II, factors influencing the odds of ageing American parents receiving support are concentrated in the parent-demand domain. In addition to the respondent's espoused family obligations being associated with an increased chance of the respondent assisting the aged parent, so too are the parent's increased age, his or her living alone, and a positive relationship quality between the respondent and aged parent. Reduced odds of assisting an ageing parent are associated with distance, the parent being a father rather than mother, and better parental health in the American sample. Finally, these same factors, plus a few more in each domain, contribute to the odds of American respondents assisting both the parent and adult child generations. The child's age is not significant in this final model, though all other child factors are. Male offspring and those who are married and working full-time face reduced chances of receiving parental help, while those with more education face improved chances of getting help. In terms of demand factors for ageing parents, all have similar effects as in Model II, with the exception of lone residence, which is no longer significant in predicting support to both generations. Finally, in the American sample respondent constraint factors are significantly associated with the likelihood that support is provided to both child and ageing parent generations (Model III). American respondents who are female, more educated, and have higher incomes are more likely to report giving support to ageing parents and adult offspring. In contrast, reduced chances of such support are found for respondents with more adult children, who report a disability, and who are in a relationship.
Discussion
This research considered how families living in dramatically different social welfare regimes of the USA and the Netherlands enact intergenerational support behaviour in relation to their espoused feelings of family obligation. This study builds on a body of work focusing largely on European comparisons or comparisons between the USA and Great Britain. Three questions were raised and hypotheses were presented for each.
We first addressed whether differences exist between American and Dutch middle-generation adults in terms of their espoused family obligations. Using Saraceno's discussion of social welfare systems and intergenerational family obligations, we expected that the American sample would report stronger family obligations than the Dutch sample. Consistent with this familialism by default hypothesis, we found that these Americans reported stronger feelings of obligation to support ageing parents and adult offspring than did Dutch respondents. This finding adds to other data that suggest that individuals feel more strongly about providing for family members in need in countries with limited public assistance, which is the case in the USA compared to the Netherlands. Haberkern and Szydlik's (Reference Haberkern and Szydlik2010) 11-country European study used a measure of obligation toward ageing parents to compare countries with different welfare regimes. They found that the Netherlands aligned with northern European countries (e.g. Sweden) in offering low support for such views, whereas Southern European countries like Italy and Spain – with less generous care and financial support arrangements – strongly favoured filial norms. Our findings thus contribute to this body of work indicating a link between norms and welfare regimes by expanding the evidence with a measure of family norms that includes views about adult offspring as well as ageing parents.
Our second question addressed differences in actual support between the two countries. The family-steps-in hypothesis posed that family support is more dominant when publicly funded care and financial services are not widely available. Contrary to this hypothesis, and controlling for differences in geographic proximity of family members in the two countries, we found a greater likelihood of supporting either the older parent or the adult offspring generation, and of supporting both generations, in the Dutch than American sample. Interestingly, Dutch adults were much more likely to report giving instrumental support to their ageing parents and their adult offspring than were Americans, even after controlling for distance, along with other sample differences. Perhaps because of the proximity that characterises most Dutch families, and that has for generations, turning to family members for help is a more engrained reaction to need among Dutch than American individuals. In contrast, because many Americans live a greater distance from family members, which makes interaction and support more difficult (Hank Reference Hank2007), there may be more openness to and acceptance of alternative supports (e.g. asking friends or neighbours for help) among Americans than the Dutch, even when family members live close by.
There was less difference between countries in the likelihood of monetary transfers by middle-generation adults, with distance being of limited importance for this type of exchange. Consistent with other studies (Wong, Capoferro and Soldo Reference Wong, Capoferro and Soldo1999), few adults in either sample provided monetary assistance to ageing parents, and in both the Dutch and American samples adult children were more often recipients of middle-generation financial support than were ageing parents. Wong, Capoferro and Soldo (Reference Wong, Capoferro and Soldo1999) question whether adults exaggerate financial support to adult offspring because of social desirability. They also speculate that adults may more often provide monetary support for adult offspring than parents because of enhanced feelings of financial responsibility for one's children. Our American respondents were slightly more likely to give money to adult offspring than were the Dutch respondents, regardless of distance. This finding makes sense because monetary support is critical for the pursuit of higher education by offspring, especially in the USA, and the US offspring had higher education levels in these samples. Finally, the greater likelihood of providing financial support to adult offspring than to aged parents in both countries may reflect a form of indirect support that some respondents were providing to grandchildren (offspring of the adult child in question). Hagestad (Reference Hagestad and Settersten2003)argues that sometimes the support exchanges in a family benefit more than just the direct recipient. When the middle-generation parent provides financial support to an adult child, not only is that grown child's situation improved, but so too is the lifestyle of any children s/he may currently have, or have in the future. This dispersal of benefits across multiple generations may provide significant motivation for respondents to be more financially supportive of their offspring than their ageing parents.
Results showing a high likelihood of intergenerational support to both ageing parents and adult offspring in the Netherlands are important to stress, given oft-heard warnings that strong social welfare programmes dampen family supportiveness (Cox and Jakubson Reference Cox and Jakubson1995). Functional solidarity has not been abandoned in Dutch families in the presence of the relatively generous public provisions, and the likelihood of Dutch adults acting as supports to both their ageing parents and adult offspring actually surpasses such behaviour in the USA where expectations for family assistance are stronger. This finding could be important evidence for any future US policy debates that push for enhanced public assistance for families.
Our third research question considered the connection between family obligations and family support behaviour and whether differences in this linkage exist for the Dutch and American samples we studied. We found no support for the no choice hypothesis that posed that the connection between obligations and support would be weaker in the USA than in the Netherlands because American adults have no options other than to assist family members in need. Limited support was found for the family-steps-in hypothesis that proposed that obligations and behaviour would be more strongly associated in the USA than in the Netherlands because of the relative lack of public provisions for individuals and families in the former. We expected that adults who possess strong obligations to support family members would step in and do just that when living in a welfare regime like that of the USA with few other support options.
Only in the case of Americans' support for ageing parents, did we find a significant association between family obligations and helping behaviour. Connections between these factors were not apparent in regard to either Dutch or American adults' relationships with adult offspring or help to ageing parents among the Dutch. We know from decades of research that intergenerational support more often flows from parents to offspring, across most of adulthood (Albertini, Kohli and Vogel Reference Albertini, Kohli and Vogel2007; Attias-Donfut, Ogg and Wolff Reference Attias-Donfut, Ogg and Wolff2005; Hill et al. Reference Hill, Foote, Aldous, Carlson and MacDonald1970; Kohli Reference Kohli1999; Troll, Miller and Atchley Reference Troll, Miller and Atchley1979). As the dominant family exchange pattern, this particular exchange appears to be governed by a desire to maintain independence and to continue to provide for children (Lye Reference Lye, Hagan and Cook1996). In the USA, feelings of obligation toward parents may be a necessary condition for adults to assist ageing parents, as Silverstein, Gans and Yang (Reference Silverstein, Gans and Yang2006) have argued. Moreover, consistent with our data, these authors found that obligations work in tandem with increasing parental need, leading to intensified support over time, especially in situations involving assistance to mothers. Although we did not conduct our analyses by parent sex, both the Dutch and American samples involved mostly ageing mothers, thus our findings for the US sample appear consistent with the results and arguments of Silverstein and his colleagues.
Finding no significant connection between family obligations and support behaviour to either generation in the Netherlands sample suggests that factors other than norms may shape intergenerational exchanges in Dutch culture. Moreover, the finding that the structural factors included as needs and constraints in our multivariate models were less influential in the Netherlands than in the USA suggests that personal or family-based preferences and routines perhaps contribute more to support patterns in Dutch than American families. Certainly, the greater geographic proximity of Dutch than American families may afford them more opportunities to develop, and over time to reinforce, a family culture of togetherness and interdependence. This proximity advantage may result in individuals who choose to engage in high levels of support with family members, not only in times of need, but at other times too, in demonstration of family solidarity. Regardless of the underlying motives for support, recent qualitative findings from a Dutch study support this pattern of relatively weak social prescriptions regarding filial obligations, yet substantial parental care. Interestingly, this study revealed much stronger personal motives to provide care for ageing parents than generalised obligations (Stuifbergen et al. Reference Stuifbergen, Dykstra, Lanting and Van Delden2010).
Of central interest in framing this study were the dramatically different social and family policies and supports available in the Netherlands and the USA. Though not explicitly tested, recognition of the vastly different contexts in which our samples live contributed to the development of our hypotheses and interpretation of our results. It is important to note, however, that more contextual variation may exist within the US sample than the Dutch sample due to state differences in policy application and services take-up. Thus, statements about US families and services availability and use must be made somewhat cautiously. Yet, research shows that much of the variability in benefits access and use across states is due to demographic and family characteristics of state populations (Shen and Zuckerman Reference Shen and Zuckerman2003), rather than actual policy differences (Bansak and Raphael Reference Bansak and Raphael2007). Because several family characteristics are included in this study's analyses, much of this variation may thus have been controlled.
Although our analysis did not directly test the impact of social welfare programmes or family processes on support behaviour in the USA and the Netherlands, it does offer evidence that adults in these two countries hold different beliefs about responsibility to family members and engage in different patterns of family assistance. Consistent with other research using different samples and different obligation and support measures (Haberkern and Szydlik Reference Haberkern and Szydlik2010; Kalmijn and Saraceno Reference Kalmijn and Saraceno2008; Lowenstein and Daatland Reference Lowenstein and Daatland2006), we conclude that Dutch individuals are highly individualistic; they actively engage in support to family members in response to personal or family-based preferences and routines, rather than in reaction to general norms of obligation, the needs of family members, or their own available resources. Living in an individualistic welfare regime that offers a relatively high level of support for its citizens seems to allow the Dutch to act on their individual preferences. In contrast, we conclude that American adults are more influenced by obligatory feelings to family members – especially their ageing parents. Moreover, although American adults are generally less likely to provide support to family members than are the Dutch, their support behaviour is more contingent and predictable when either their adult offspring or ageing parents encounter pressing need for help. Yet, when it comes to providing support to both ageing parents and adult offspring, it is the personal resources available to middle-generation adults and their own constraints that become highly salient in this decision. Given these differences not only in the patterns but also the predictors of intergenerational support for the Dutch and American middle-generation respondents we studied, future work is needed to examine whether the provision of such supports plays a differential role in the wellbeing of middle-generation adults in different welfare regimes.
Acknowledgements
This research was initiated in 2007 when Dr Cooney was a visiting scholar at The Netherlands Interdisciplinary Demographic Institute, where Dr Dykstra held a position at the time. Dr Cooney's leave was funded by the University of Missouri's Faculty Development Leave Program. Earlier versions of the paper were presented at the International Association for Relationship Research Conference, Providence, Rhode Island, 17–20 July 2008, and the International Population Conference, Marrakech, Morocco, 27 September to 2 October 2009. NSFH3 was jointly funded by NICHD (Center for Population Research of the National Institute of Child Health and Human Development) grant no. HD21009 and by NIA (National Institute on Aging), grant no. AG10266. The Netherlands Kinship Panel Study (NKPS) was financially supported by the ‘Major Investments Fund’ of the Netherlands Organisation for Scientific Research (NWO), under grant 480-10-009. Financial institutional support also was provided by The Netherlands Interdisciplinary Demographic Institute (NIDI), the Faculty of Social Sciences (Utrecht University), the Faculty of Social and Behavioral Sciences (University of Amsterdam) and the Faculty of Social Science (Tilburg University).
