North American and other developed countries are currently witnessing the aging of the population together with significant changes in family structure and relationships. As a result of recent declines in marriage and childbearing – as well as increases in cohabitation, divorce, remarriage, and step- and lone-parenthood – we are seeing an increasing diversification of family structures (Chappell & Funk, Reference Chappell and Funk2011; Glaser, Stuchbury, Tomassini, & Askham, Reference Glaser, Stuchbury, Tomassini and Askham2008; Milan, Vezina, & Wells, Reference Milan, Vezina and Wells2007). For example, from 2001 to 2013, the proportion of Canadians aged 65 and older who were currently married increased only slightly (from 54.5% to 55.7%). In contrast, the percentage living in common-law relationships almost doubled (from 1.8% to 3.5%), and the proportion who reported being separated or divorced also increased significantly (6.8% to 9.9%) (Statistics Canada, 2013). Among incoming senior cohorts (those aged 45–64 in 2013), the proportions were even higher: 10.7 per cent were living as common-law, whereas 14.5 per cent were divorced or separated. In addition, they were more than twice as likely to have never married (11.0% compared to only 5.2% of those aged 65+; Statistics Canada, 2013). Currently, from 12 to 13 per cent of Canadian cohorts born between 1925 and 1944 are childless (Dykstra, Reference Dykstra and Uhlenberg2009; Ravanera & Beaujot, Reference Ravanera and Beaujot2014)Footnote 1. However, whereas all but the oldest old (aged 85+) today belong to cohorts with historically low levels of childlessness in Canada as well as many other countries (Dykstra, Reference Dykstra and Uhlenberg2009; Rowland, Reference Rowland2007), the proportion of older adults that is childless is expected to increase significantly in coming years, particularly over the longer term as baby boomers and their children age (Lin & Brown, Reference Lin and Brown2012). For example, Carrière et al. (Reference Carrière, Keefe, Légaré, Lin, Rowe, Martel and Rajbhandary2008) projected that the proportion of Canadian women aged 65 and older without any surviving children will increase from 16 per cent in 2001 to a high of 30 per cent in 2051.
The rapid demographic growth of an increasingly diverse older adult population raises important questions about their current and future well-being, including the continued ability of informal support networks – family, friends, and others – to provide high levels of support (Carrière et al., Reference Carrière, Keefe, Légaré, Lin, Rowe, Martel and Rajbhandary2008). On the one hand, it is commonly believed that decreasing availability and increasing rejection of traditional family roles and relationships mean that families will become less supportive of one another (Glaser et al., Reference Glaser, Stuchbury, Tomassini and Askham2008). Consistent with this view, being unmarried and/or childless have been linked to a number of negative outcomes, including social isolation and disruptions to social support (Dykstra, van Tilburg, & de Jong Gierveld, 2005; Victor, Scambler, Bowling, & Bond, Reference Victor, Scambler, Bowling and Bond2005). However, other researchers are critical of continued reliance on normative assumptions regarding the primacy of the nuclear family (e.g., Cotterill, Reference Cotterill1994), contending that older adults actively manage their social ties and that with increasing acceptance of divorce and of diverse family forms, the negative implications of such changes for support in old age may be disappearing (Glaser et al., Reference Glaser, Stuchbury, Tomassini and Askham2008, p. 330; Thornton & Young-DeMarco, Reference Thornton and Young-DeMarco2001). Consistent with this latter view, there is also evidence to suggest that unmarried and/or childless individuals fare better than traditionally assumed (e.g., Connidis & McMullin, Reference Connidis and McMullin1994).
However, despite theoretical and empirical reasons to question the validity of longstanding assumptions, to date, limited research attention has been directed to the joint implications of marital and parental status for social support. For example, never married and childless older adults have been described as invisible within social science literature (Dykstra & Hagestad, Reference Dykstra and Hagestad2007a). Yet even less appears to be known regarding previously married (i.e., widowed, divorced) childless older adults. This study addresses these gaps in our knowledge about the growing unmarried and childless populations, focusing on the impact of intersections involving marital and parental status for informal support from individuals living outside the household in later life.
Review of the Literature
Marital Status and Social Support
A review of the empirical literature supports the view that marriage is central to social support and the receipt of care in later life. It is well-documented, for example, that older people prefer to remain living independently in the community as they age, and that spousal relationships play a major role in facilitating this, particularly in terms of the provision of informal support and care as health declines (Walker & Luszcz, Reference Walker and Luszcz2009). Spouses, if available, are reported to be the most likely to provide instrumental forms of support and assistance and to do so during periods of greater illness and disability than any other support provider (Feld, Dunkle, Schroepfer, & Shen, Reference Feld, Dunkle, Schroepfer and Shen2006; Lima, Allen, Goldscheider, & Intrator, Reference Lima, Allen, Goldscheider and Intrator2008; Walker & Luszcz, Reference Walker and Luszcz2009). In addition, married individuals are likely to name their spouse as a confidante or source of emotional support, particularly among men (Chappell, McDonald, & Stones, Reference Chappell, McDonald and Stones2010).
Given the widely acknowledged importance of the spouse for access to and the receipt of support, it is frequently assumed that never married as well as previously married (widowed, divorced) older adults are likely to be disadvantaged relative to those who are married (Chappell & Funk, Reference Chappell and Funk2011; Keith, Reference Keith2003; Keith, Kim, & Schafer, Reference Keith, Kim and Schafer2000). Whether this is actually the case, however, remains unclear. To date, most of the research attention has been on comparisons of the structural aspects of social and/or helping network ties – including network size and composition – rather than on differences in functional support. In addition, studies comparing currently married to never-married and/or previously married (widowed, divorced) older adults with regard to functional dimensions of support generally focus on the support received from children. Consequently, little is known regarding the comparative implications of being currently married versus never married and/or previously married (widowed, divorced) without children. Finally, available results are inconsistent. Whereas most report a negative association between parental divorce or widowhood and assistance received (e.g., see Curran, McLanahan, & Knab, Reference Curran, McLanahan and Knab2003; Pezzin & Schone, Reference Pezzin and Schone1999), some report a positive relationship, especially with regard to the relationship between widowhood (compared to marriage) and the receipt of assistance from children (Ha, Carr, Utz, & Nesse., Reference Ha, Carr, Utz and Nesse2006), particularly among women (Glaser et al., Reference Glaser, Stuchbury, Tomassini and Askham2008).
Parental Status and Social Support
Despite some indications that marital status may be more influential than parental status when it comes to social support (Connidis & McMullin, Reference Connidis and McMullin1994; Wenger, Dykstra, Melkas, & Knipscheer, Reference Wenger, Dykstra, Melkas and Knipscheer2007), the primary importance of children for the provision of support in later life is widely noted (Basten, Reference Basten2009; DeOllos & Kapinus, Reference DeOllos and Kapinus2002; Wenger, Scott, & Patterson, Reference Wenger, Scott and Patterson2000; Zhang & Hayward, Reference Zhang and Hayward2001). Childless individuals are generally assumed to fare less well at older ages than parents, presumably because they lack the critical social support represented by the presence of children in the informal network (Choi, Reference Choi1994).
To a considerable degree, empirical literature appears to support to this view, particularly with regard to structural dimensions of support (e.g., see Grundy & Read, Reference Grundy and Read2012; Koropeckyj-Cox, Reference Koropeckyj-Cox1998; McMullin & Marshall, Reference McMullin and Marshall1996) such as support network size (Dykstra, Reference Dykstra2006, Reference Dykstra and Uhlenberg2009; Dykstra & Wagner, Reference Dykstra and Wagner2007; Wenger et al., Reference Wenger, Scott and Patterson2000). Fewer studies have directly compared older parents and childless adults with regard to functional dimensions of support (Silverstein & Giarrusso, Reference Silverstein and Giarrusso2010). Some evidence suggests that childless elders may be able to compensate for the disadvantage of having smaller familial networks by having better access to other resources, including “social alternatives to children” (Silverstein & Giarrusso, Reference Silverstein and Giarrusso2010, p. 1043; also see Dykstra & Wagner, Reference Dykstra and Wagner2007). For example, it has been reported that childless elderly individuals often have higher levels of involvement with siblings, friends, and other age peers (Campbell, Connidis, & Davies, Reference Campbell, Connidis and Davies1999; Connidis & Davies, Reference Connidis and Davies1990; Connidis & McMullin, Reference Connidis and McMullin1992; Cornwell, Laumann, & Schumm, Reference Cornwell, Laumann and Schumm2008; Dykstra & Hagestad, Reference Dykstra and Hagestad2007a; Wenger et al., Reference Wenger, Dykstra, Melkas and Knipscheer2007). Similarly, non-kin have been noted to be more likely to help out childless older adults than they are those with children available (Wenger, Reference Wenger1984). However, others report findings indicating that the absence of either a partner or a child relationship tends not to be “fully compensated by the (instrumental) support received from other types of network members” (Broese van Groenou & van Tilburg, Reference Broese van Groenou and van Tilburg1997, p. 39). As a result, despite alternative resources for care, childless older adults remain disadvantaged relative to older parents (e.g., see Grundy & Read, Reference Grundy and Read2012; Larsson & Silverstein, Reference Larsson and Silverstein2004).
Marital Status, Parental Status, and Social Support
Whereas most research has tended to focus on the comparative advantages and disadvantages of having or not having children, others have suggested that the meaning, experience, and consequences of childlessness are likely to vary (Umberson, Pudrovska, & Reczek, Reference Umberson, Pudrovska and Reczek2010), and thus, it is important to acknowledge the heterogeneity of the childless experience in old age (Dykstra & Wagner, Reference Dykstra and Wagner2007; Wenger et al., Reference Wenger, Dykstra, Melkas and Knipscheer2007). According to Dykstra (Reference Dykstra and Uhlenberg2009), for example, it is the intersection of parental and marital status that is especially important to consider.
Dykstra (Reference Dykstra and Uhlenberg2009) suggested that being childless is a source of vulnerability particularly in the absence of a partner (also see Keith, Reference Keith2003; Wu & Pollard, Reference Wu and Pollard1998). To date, however, empirical support for this view remains limited. Despite some supportive evidence (Albertini & Mencarini, Reference Albertini and Mencarini2014; Dykstra & Hagestad, Reference Dykstra and Hagestad2007b), other findings suggest that unmarried childless elders generally fare better than those who are married and childless. According to an early study by Johnson and Catalano (Reference Johnson and Catalano1981), for example, those who were not married (i.e., never or previously married) were found to be “more resourceful in using a long-term accumulation of social resources to meet their needs” (p. 610), whereas married childless individuals (especially men) tended to be more isolated and to rely primarily on each other (see also Wenger et al., Reference Wenger, Scott and Patterson2000, Reference Wenger, Dykstra, Melkas and Knipscheer2007). Similarly, Wenger (Reference Wenger1984) reported findings indicating that, unlike married individuals, single childless elderly women appear more likely to receive support from a wide range of network members and particularly from siblings and their siblings’ families (see also Campbell et al., Reference Campbell, Connidis and Davies1999; Connidis & McMullin, Reference Connidis and McMullin1994; Larsson & Silverstein, Reference Larsson and Silverstein2004; Wenger et al., Reference Wenger, Dykstra, Melkas and Knipscheer2007). Older people who are not married or who have few or no children also appear more likely than the married to have non-kin in their support networks (Keating, Otfinowski, Wenger, Fast, & Derksen, Reference Keating, Otfinowski, Wenger, Fast and Derksen2003).
Important differences, however, may exist among those who are not married. For example, some researchers point to the particular vulnerability of the never married (e.g., Keith, Kim, & Schafer, Reference Keith, Kim and Schafer2000). On the other hand, evidence also suggests that never-married childless adults, particularly women, tend to be socially active and are more likely to belong to social groups compared to older married women with children (Cwikel, Gramotnev, & Lee, Reference Cwikel, Gramotnev and Lee2006; Dykstra, Reference Dykstra and Uhlenberg2009; Johnson & Catalano, Reference Johnson and Catalano1981; Wenger et al., Reference Wenger, Scott and Patterson2000; Wenger et al., Reference Wenger, Dykstra, Melkas and Knipscheer2007). Wenger (Reference Wenger2009) contended that findings indicating the greater assistance coming from non-kin to childless older adults may reflect the greater self-sufficiency of those who never married. They also appear among the most likely to receive support from siblings (Wenger et al., Reference Wenger, Scott and Patterson2000; Reference Wenger, Dykstra, Melkas and Knipscheer2007). In contrast, other researchers point to the greater vulnerability of previously married (widowed, divorced/separated) than of never-married childless individuals. According to Wenger et al. (Reference Wenger, Scott and Patterson2000), this ultimately stems from the greater vulnerability attached to being married than unmarried and childless: “childless married couples tend to have local self-contained or private restricted support networks, which are vulnerable when a partner dies or both become frail” (p. 179). However, others attribute it to the effects of divorce or widowhood in eroding the support network (Connidis, Reference Connidis2010). Finally, still other researchers point to similarities between previously married and never-married childless individuals, suggesting that they tend to develop and benefit from similar patterns of interaction with relatives and friends (e.g., Johnson & Catalano, Reference Johnson and Catalano1981; also see Larsson & Silverstein, Reference Larsson and Silverstein2004; Wu & Pollard, Reference Wu and Pollard1998).
Present Study
The aforementioned review suggests a need to focus attention on the implications of marital and parental status intersections for the receipt of social support in later life. Research conducted to date reveals considerable support for the importance of a spouse and children for support received in later life. However, studies that focus directly on the joint implications of marital and parental status on social support are limited (Wu & Pollard, Reference Wu and Pollard1998) and, where available, rarely consider intersections across all major marital/parental status groups simultaneously. As a result, the comparative advantages and disadvantages of having or not having children among currently married, never-married, and previously married (widowed or divorced) older adults remains unclear. Studies frequently focus on structural rather than functional dimensions of support. In addition, where functional aspects of support have been addressed, limited attention has been focused on the receipt of various types of support (instrumental, emotional).
To address these gaps, this study drew on national survey data to examine the effects of marital and parental status intersections on extra-household social support received by older adults living independently in the community. Two research questions were examined: (1) What impact does parental status (childless, not childless) have on the functional social support received from non-household members across marital status groups? (2) Does this vary depending on the type of functional support involved? In particular, does it differ with regard to various types of instrumental support (e.g., help with household tasks, transportation) that tend to be required as health declines and emotional support that tends to be less clearly associated with declining health?
In asking these questions, we were interested not only in contributing to empirical literature in the area but also to theoretical debates regarding the importance of marital and parental status for social support in later life. In particular, we were interested in assessing support for the implications of two competing theoretical positions. The first, exemplified by Cantor’s (Reference Cantor1975, Reference Cantor1979) hierarchical compensatory or social care (Cantor, Reference Cantor1991; Cantor & Little, Reference Cantor, Little, Binstock and Shanas1985) perspective, considers unmarried and/or childless older adults to be disadvantaged when it comes to the receipt of social support because they lack access to some of the most important and normatively preferred sources of support. For example, several decades ago, Cantor argued that older adults’ preferences for support tend to follow a normatively defined sequential hierarchy based on the primacy of their relationship with potential support providers. Within this hierarchy, kin and, especially, the nuclear family unit (i.e., spouse, followed by children, siblings, and other relatives) were said to represent the most preferred sources of support, regardless of the type of support involved, followed by friends, neighbours, and finally, by formal organizations.
If and when more preferred sources of support were unavailable, ties lower in the hierarchy were said to substitute for the missing relationships. However, substitution was not considered inevitable but, rather, dependent upon the prior existence of a “functional” relationship (i.e., one characterized by close proximity, frequent/regular contact, ability to provide the support required). Consequently, not only were spouseless and/or childless elders considered likely to represent potentially at-risk groups, but also siblings, friends, and neighbours were viewed as not being able to be counted on to provide the same types and levels of support and care provided by kin, particularly within the context of the nuclear family structure (Cantor & Little, Reference Cantor, Little, Binstock and Shanas1985).
In contrast with assumptions of normatively ordered substitution and the primacy of the nuclear family, more recent theoretical perspectives have argued that older adults actively manage their social ties in order to meet the challenges of aging. For example, the functional specificity model, as outlined by Simons (1983–84), suggests that social relationships are negotiated over time and that they tend to be functionally specific. However, functions are not specific to social ties (Connidis, 1994). Along somewhat similar lines, Kahn and Antonucci’s (Reference Kahn, Antonucci, Baltes and Brim1980) convoy model conceptualizes individuals as being embedded within a convoy with whom they exchange instrumental and emotional support – one that is established early on but that is dynamic and continually changing – over the life course (e.g., see Antonucci, Reference Antonucci, Binstock and Shanas1985, Reference Antonucci1986, Reference Antonucci, Binstock and George1990; Antonucci & Akiyama, Reference Antonucci and Akiyama1987). The convoy model, too, “makes no a priori assumptions regarding the specific relations comprising an individual’s social network” (Levitt, Reference Levitt2005). Instead, the structure and supportive functions of the convoy at any given time are thought to be the product of personal and situational characteristics that affect the individual’s need for support. Thus, social network substitution occurs throughout the life course, as individuals lose network members (due to death, residential relocation, etc.) and add new network members in response to such losses (Antonucci, Reference Antonucci, Binstock and Shanas1985).
The implications of the two theoretical positions for an understanding of the impact of marital and parental status intersections differ. For example, whereas the hierarchical compensatory model suggests that those without a spouse and /or children available will tend to be disadvantaged when it comes to the receipt of social support, models founded on the view that social roles are flexible in function and that older individuals actively negotiate/construct their support networks to meet changing needs suggest a more positive scenario. That scenario is one implying that over the life course, unmarried childless older adults – regardless of whether they be never married, widowed, or divorced – will actively construct their support convoys to ensure the availability of support when needed. Consequently, if the former view is correct, we would expect to find that older adults with a spouse and children available will fare better when it comes to social support than those without a spouse or children available (i.e., spouse available/no children, no spouse/children available). However, one of the reasons they will do better is because they have access to spousal (thus intra-household) support. When it comes to support from non-household members only, this view nevertheless suggests that childless individuals regardless of marital status will likely be disadvantaged relative to those with children. Those with neither a spouse nor children available will do less well although there appears to be little or no basis for predicting differences among never-married and previously married (separated/divorced, widowed) individuals without children.
Conversely, if the latter view is correct, it is the composition of the support network rather than the likelihood of support itself that is likely to differ, and we would therefore expect to find few if any differences in the availability of support to older adults within any of the marital/parental status groups.
Data and Methods
To address these issues, we constructed multivariate regression models estimating the effects of parental status on support within differing marital status groups. Thus, our study population included all elderly persons, regardless of their marital or parental status.
Data Source and Sample
The study used data from the 2007 Canadian General Social Survey, Cycle 21 (GSS-21), conducted by Statistics Canada. The GSS is an annual national (cross-sectional) survey that collects individual- and household-level data on Canadian adults to monitor social conditions and the well-being of Canadians and to provide information on social policy issues of current or emerging interest (Statistics Canada, 2009). Aside from collecting basic demographic and socioeconomic data, each cycle of the GSS has a specific thematic focus, such as family, time use, or victimization. The thematic focus of the GSS-21 was social support and aging. It collected detailed data on social support, health conditions, family history, retirement planning and experience, as well as standard demographic and socioeconomic variables.
The target population of the GSS-21 included Canadians aged 45 and older living in all 10 provinces, excluding the northern territories as well as full-time residents of institutions. The survey was conducted through telephone interviews. Households without telephones were excluded, but represented only 0.9 per cent of the target population (Statistics Canada, 2009). Households with cellular phone service only were also excluded; they represented 6.4 per cent of the target population. The exclusion of the households that did not have landline phone service is a limitation of the study. However, research has shown that the cellular-phone-only household is more common among low-income and/or young adult households (Blumberg & Luke, Reference Blumberg and Luke2007, Reference Blumberg and Luke2008). Given our study population, this exclusion is unlikely to significantly bias our regression estimates. Nevertheless, caution is called for when generalizing the results of our study to the entire study population.
The GSS-21 included a nationally representative sample of 23,404 Canadians aged 45 and older, with an overall response rate of 57.7 per cent. To study social support among older adults, we limited our study sample to individuals who were age 60 or older. As the study focused on marital and parental status, we removed cases where we could not identify the respondents’ marital or parental status (n = 50). In unreported analysis, we found that those who were missing data on marital or parental status did not differ in the response variables from the rest of the sample. With these restrictions, the final study sample included 11,494 respondents who were age 60 or older living in private households at the time of the survey.
Measures
The study focused on social support. Social support is a broad and complex concept that has been variously defined and operationalized (Cohen, Gottlieb, & Underwood, Reference Cohen, Gottlieb, Underwood, Cohen, Underwood and Gottlieb2000; Gottlieb & Bergen, Reference Gottlieb and Bergen2010; Haber, Cohen, Lucas, & Baltes, Reference Haber, Cohen, Lucas and Baltes2007). Here we used a functional definition of social support that emphasized “the social resources that persons perceive to be available or that are actually provided to them by nonprofessionals in the context of … informal helping relationships” (Cohen et al., Reference Cohen, Gottlieb, Underwood, Cohen, Underwood and Gottlieb2000, p. 4). Specifically, we focused on two dimensions of support perceived or received from non-household members: instrumental and emotional support. We created three dummy variables to indicate whether respondents reported support in the areas of domestic help, transportation assistance, and emotional support.
To measure domestic support, we drew on responses to the following question: “In the past 12 months, did anyone help you by doing domestic work, home maintenance, or outdoor work?” Respondents were instructed to focus on unpaid help and to exclude both help provided by those living with them and help provided by organizations. Taking into account whether respondents needed assistance is also important (Taylor, 1990). Therefore, respondents who provided a negative response (indicating that they did not need help) were then asked, “(In the past 12 months), if you had needed help (with these activities), would you have had someone to turn to for help with (domestic work, home maintenance, or outdoor work)?” Using the responses to both of these questions, we then created a dummy variable, contrasting those who provided a positive response to either question to those who did not.
The measure of “providing transportation or running errands” was based on similar questions in the GSS-21. We used the responses to these questions, which were identical to those for domestic work, to create a dummy variable, indicating whether or not the respondent received such help in the past 12 months (including those who did not receive such help but had someone to turn to if such help was needed).
The measure of emotional support was based on responses to a single question: “In the past 12 months, did anyone help you by giving you emotional support?” Again, we created a dummy variable to indicate receipt or non-receipt of emotional support in the past year. Also, respondents were once again instructed to focus on unpaid help and to exclude help provided by those living with them as well as that provided by organizations.
The extent of non-response (missing data) to the social support questions was generally low (2.7% for domestic assistance, 2.2% for transportation assistance, and 1.1% for emotional support). In unreported analysis, we found that the likelihood of missing data on each of the response variables was unrelated to marital and parental status, our primary independent variables. Cases with missing values on response variables were removed from the regression analyses.
The primary independent variables were marital and parental status. Marital status at the time of interview was measured as a four-level categorical variable: (a) married or cohabiting, (b) separated or divorced, (c) widowed, and (d) never married. Prior research demonstrates that cohabitation has become a common path of entry into conjugal relationships (Kennedy & Bumpass, Reference Kennedy and Bumpass2008; Kerr, Moyser, & Beaujot, Reference Kerr, Moyser and Beaujot2006), although cohabitation often appears indistinguishable from marriage when it comes to the receipt of social support and other outcomes (Penning & Wu, Reference Penning and Wu2013; Schimmele & Wu, Reference Schimmele and Wu2011; Wu & Hart, Reference Wu and Hart2002). Due to small cell counts, it was not feasible to estimate separate regression models for cohabiters. We therefore included a dummy indicator for cohabiting status (1 = yes) in the models for married and cohabiters. Similarly, parental status is also a dummy variable, contrasting those who were childless (1 = yes) to those who had ever had or raised a child.
Throughout the analyses we also controlled for other demographic factors (i.e., age, gender, number of siblings, visible minority status, length of residence), socioeconomic indicators (educational attainment, employment status, household income, home ownership), and health status variables (self-reported health, activity limitations, chronic illness) known to influence social support (e.g., Albertini & Mencarini, Reference Albertini and Mencarini2014; Broese van Groenou & van Tilburg, 1997; Cantor & Little, Reference Cantor, Little, Binstock and Shanas1985; Feld et al., Reference Feld, Dunkle, Schroepfer and Shen2006; Glaser et al., Reference Glaser, Stuchbury, Tomassini and Askham2008; Grundy & Read, Reference Grundy and Read2012; Gurung, Taylor, & Seeman, Reference Gurung, Taylor and Seeman2003; Keith et al., Reference Keith, Kim and Schafer2000; Larsson & Silverstein, Reference Larsson and Silverstein2004; Wu & Pollard, Reference Wu and Pollard1998). We considered three socio-demographic control variables. Respondents’ gender was a dummy indicator (1 = female). Age was measured as a three-level categorical variable: 60–64, 65–74, and 75 and older. Visible minority status was also included as a dummy variable (1 = visible minority), indicating whether the respondent belonged to a visible minority grouping (e.g., Black, Chinese, Filipino, Japanese, Korean, South Asian-East Indian, Southeast Asian, non-White West Asian, North African or Arab, non-White Latin American, person of mixed origin, and other visible minority groups). Finally, four variables were used to assess the availability of social ties: number of siblings (coded 0 through 4+), living arrangements (1 = alone, 0 = not alone), whether or not the respondent had most of their relatives living in the same city/region (1 = yes), and whether or not they had most of their friends living in the same city/region (1 = yes).
We considered five socioeconomic variables. Educational attainment was a 5-level categorical variable, ranging from less than high school education to bachelor’s degree or higher. Employment was coded into three categories: employed outside the home, retired, and other (i.e., looking for work, caring for children, or engaged in household work). Household income was also a four-level categorical variable, ranging from less than $30,000 to $100,000 or more. Home ownership was a dummy variable (1 = yes), and length of residence was measured as years of living in the current residence (ranging from 1 [less than 6 months] to 6 [10 years or more]).
Finally, we included three health status indicators. Self-reported health is known to be a robust indicator of general health for the general and elderly population (e.g., Idler & Benyamini, Reference Idler and Benyamini1997). It was measured as an ordinal variable ranging from poor (1) to excellent (5). Activity limitation was a dummy variable, indicating whether the respondent was limited in the amount/kind of regular activity at home, work, or in other activities due to a physical or mental condition, or health problem (1 = yes). The presence of chronic conditions was also a dummy variable (1 = yes), indicating the presence of any chronic condition (e.g., arthritis or rheumatism, back problems, diabetes, Alzheimer’s disease, heart disease, or cancer).
Missing data on the control variables were minimal (less than 2.5%), except for household income where data were missing for 30 per cent of the study sample. To avoid substantial reduction in sample size, we created a dummy variable for missing data on income. In our analyses, we removed cases with missing data on the other control variables.
Statistical Models
Our empirical analysis began with investigating the issue of endogeneity of parental status, our main independent variable. To our knowledge, no prior studies have taken into consideration such a potential selection bias in the examination of the effects of marital and parental status on social support. It is well-known, however, that parenthood is endogenous because not everyone chooses to become a parent (e.g., Veevers, Reference Veevers1980). Older persons who are childless are even more selective (e.g., Rubinstein, Alexander, Goodman, & Luborsky, Reference Rubinstein, Alexander, Goodman and Luborsky1991; Wu & Pollard, Reference Wu and Pollard1998). If the decision to remain childless is correlated with social support, then the effect of parental status on social support may be biased (see Greene, Reference Greene2012). For instance, if individuals who choose to become a parent are more outgoing and sociable, and these attributes are also associated with an increased likelihood of receiving social support when needed, then the potential positive effect of being a parent may be overestimated. By the same token, if people who remain childless tend to be more introspective and solitary and develop a limited circle of relationships, then the potential negative effect of being childless can also be overstated.
To correct for the potential selection bias, using the maximum likelihood method, we estimated two simultaneous (seemingly unrelated bivariate) probit models (an “outcome” model and a “selection” model) that allow for a correlation of the error terms from the two models (Maddala, Reference Maddala1983). Such models typically assume that there exists an underlying relationship for the outcome variable (y 1)
$$y_{1i}^* = {x_{1i}}{\beta _1} + {u_{1i}}$$
$${y_{1i}} = 1\;\;if\,\;y_{1i}^* > 0$$
$${y_{1i}} = 0\;\;{\rm{otherwise}}$$where 
$y_1^*$ is a latent dependent variable (receipt of support); x 1 is a vector of co-variates; β 1 is a vector of regression coefficients associated with x 1; and u 1 is an error term. There is a similar setup for the selection (into non-marriage and childlessness):
$$y_{2i}^* = {x_{2i}}{\beta _2} + {u_{2i}}$$
$${y_{2i}} = 1\;\;if\,\;y_{2i}^* > 0$$
$${y_{2i}} = 0\;\;{\rm{otherwise}}\,.$$From Equations (1) and (2), the error terms u 1 and u 2 are assumed to be jointly normally distributed with a mean of zero, variance of one, and a correlation of ρ. When ρ = 0, the single outcome equation is unbiased. When ρ ≠ 0, the regression estimate on the treatment (childlessness) is likely biased (Greene, Reference Greene2012). When ρ > 0, the estimated effect of childlessness from a standard single-equation model is generally biased away from zero. The converse is true when ρ < 0.
In Equation (1), x 1 included the independent variables shown in Table 1. In Equation (2), x 2 comprised a somewhat different set of co-variates, including gender (1 = female), age (as a continuous variable), minority status, education (as a 10-level ordinal variable), religion (in 4 levels), and region (1 = Quebec, 0 = elsewhere in Canada), as well as an interaction term of gender and education (see, e.g., Goldscheider & Waite, Reference Goldscheider and Waite1986; Pollard & Wu, Reference Pollard and Wu1998; Thornton, Axinn, & Xie, Reference Thornton, Axinn and Xie2007). Although not necessarily required, choosing a somewhat different set of co-variates for the selection equation helps identify the effect of the “treatment” variable (parental status) in the outcome equation (Amemiya, Reference Amemiya1985; Greene, Reference Greene2012). There was clear evidence confirming the selection effect of childlessness in modeling social support (see Tables 2–4). We present the regression estimates from the outcome models in Tables 2–4 and the regression estimates from the selection models in Appendix A. To examine model adequacy, we also investigated various model assumptions (e.g., outliers, multicollinearity) and found no evidence of violation of the assumptions (see Appendices B and C). Correlations among the explanatory variables are low (none exceeds 0.40), and the variance inflation factors are acceptable (the highest is 5.03 for low income).
Table 1: Descriptive Statistics of the Variables Used in the Regression Models: Older Canadians (Age 60+)
a See text for details.
Note: Weighted means or percentages, unweighted N.
Source: The 2007 Canadian General Social Survey.
Table 2: Probit Models of Receiving Domestic Assistance on Marital/Parental Status Accounting for Selection into Marriage and Parenthood: Older Canadians (Age 60+), 2007
*** p < 0.001; **p < 0.01; *p < 0.05 (two-tailed test)
a Reference group.
Note: All models include a dummy variable for missing household income. See Appendix A for the covariates in the selection equation.
Source: The 2007 Canadian General Social Survey.
Table 3: Probit Models of Receiving Transportation Assistance on Marital/Parental Status Accounting for Selection into Marriage and Parenthood: Older Canadians (Age 60+), 2007
*** p < 0.001; **p < 0.01; *p < 0.05 (two-tailed test)
a Reference group.
Note: All models include a dummy variable for missing household income. See Appendix A for the covariates in the selection equation.
Source: The 2007 Canadian General Social Survey.
Table 4: Probit Models of Receiving Emotional Support on Marital/Parental Status Accounting for Selection into Marriage and Parenthood: Older Canadians (Age 60+), 2007
*** p < 0.001; **p < 0.01; *p < 0.05 (two-tailed test)
a Reference group.
Note: All models include a dummy variable for missing household income. See Appendix A for the covariates in the selection equation.
Source: The 2007 Canadian General Social Survey.
Results
Table 1 reports sample characteristics, comparing the prevalence of each across the four marital status groups. Compared to those in other groups, individuals who were married were the least likely to be female, had the highest income levels, were the most likely to own their homes, reported the longest length of residence, had the best self-reported health, and were the least likely to report experiencing activity limitations. In contrast, those who were separated/divorced were the youngest, had the most siblings, were the most likely to live alone, were the most likely to be employed, had the lowest household incomes, were the least likely to own their homes, and had the shortest length of residence. Widowed respondents were the most likely to be female and older, had the fewest siblings, were the least likely to be visible minorities, had the lowest levels of education, were the most likely to be missing information on income, reported the poorest self-reported health, and had the greatest proportion with activity limitations and chronic illness. Finally, never-married respondents were the least likely of the unmarried groups to be living alone, were the most likely to be members of a visible minority, and had the highest levels of education.
With regard to the dependent variables, more than two-thirds of those in the study sample reported that they had received domestic support or had such support available to them from individuals living outside their household if needed in the past 12 months. An even greater proportion (over 80%) reported that they had received transportation assistance or had such support available if needed during this same period. Somewhat fewer respondents (from 35% to 50%) reported having received emotional support in the past 12 months: this differential reflects the fact that respondents were only asked about whether they had received such support but not whether they felt such support would have been available if needed.
Bivariate comparisons of the proportion reporting having received each of the three types of support by marital status revealed significant differences across the groups. Specifically, married/cohabiting individuals were the most likely to report receiving domestic assistance (70.7%), whereas those who were separated/divorced (66.4%) were the least likely to report such support from non-household members. In contrast, few differences were evident across the groups with regard to transportation assistance. For emotional support, married individuals were the least likely (34.9%) to report emotional support from others living outside the household, whereas widowed individuals were the most likely (49.7%).
When marital and parental status were addressed simultaneously (see Figure 1), the findings changed somewhat. In this case, whereas married/cohabiting individuals with children were the most likely to report receiving domestic assistance from non-household members (71%), those who were both separated/divorced or widowed and childless (59%, 59%) emerged as the least likely to report such support. A somewhat different pattern was evident with respect to transportation assistance: in this domain, those who were married and who had children were once again among the most likely to report receiving assistance (85%) along with those who were not currently married, whether separated/divorced or widowed, but who had children (85%). In contrast, those who had never married but who had children (72%) were the least likely to report such support. For emotional support, yet another pattern emerged: those who had never married but who had children were the most likely to report receiving such support (54%), whereas those who were married/cohabiting and childless (33%) were the least likely to report such support, followed closely by those who were married/cohabiting and who had children (35%) and then by never-married or widowed and childless individuals (42%, 42%).
Figure 1: Social Support by Childlessness and Marital Status
Table 2 presents regression estimates obtained for models in which extra-household domestic support was regressed on marital/parental status with selection into marriage/parenthood taken into account (see Appendix A for the regression estimates in the selection models), and with demographic, socioeconomic, and health factors controlled for. Prior to the introduction of these control variables, the results (not shown) revealed that among those who were either currently married/cohabiting or separated/divorced, being childless was associated with a greater likelihood of domestic support from individuals outside the household. In comparison, among those who were widowed, childlessness was associated with a reduced likelihood of domestic support from individuals outside the household, whereas among those who had never married, childlessness had no significant impact. In the multivariate analyses, the impact of childlessness on domestic support remained the same with the exception of never-married adults. For this group only, the impact of childlessness on support became significantly negative following the introduction of the control variables, suggesting that one or more control variables (e.g., home ownership) may have suppressed an otherwise negative association between childlessness and domestic support within this marital status group.
Turning to the control variables, we found that gender and living arrangements were not significantly associated with domestic support among any of the four marital status groups. Among married individuals specifically, those who were older (aged 75+), members of visible minority groups, and who reported longer length of residence were less likely to report having received domestic support from others outside the household, whereas those who reported having relatives and friends available were more likely to have received domestic support from others outside the household. With regard to socioeconomic factors, it was married individuals with low to moderate levels of education and moderate incomes who reported more domestic support. Finally, with regard to health, our findings showed that better self-reported health was associated with a greater likelihood of support from others outside the household, whereas the presence of activity limitations reduced the likelihood of support from others.
Some of the same factors (i.e., having friends, having relatives, length of residence, self-reported health, activity limitations) emerged as significantly related among separated or divorced individuals. However, in contrast with those who were currently married, divorced or separated individuals benefited from having more siblings available but were disadvantaged if they were looking for work, caring for children, or engaged in household work rather than being either employed in the labour market or retired. Those who were widowed were also similar to those who were married with regard to the negative implications of visible minority status and activity limitations as well as the positive implications of having relatives nearby, low to moderate levels of education, and excellent self-reported health. In addition, however, they also reported negative implications of chronic illness. Finally, fewer factors emerged as significantly related among those who had never married: those in the middle of the three age groups (65–74) were more likely to report receiving domestic support than those who were either younger or older as were those who owned their own homes, while those reporting longer length of residence reported less domestic support.
Table 3 presents the results of similar analyses conducted with assistance with transportation as a dependent variable. At the bivariate level (results not shown), childlessness was significantly and positively associated with extra-household transportation assistance among married/cohabiting, separated/divorced, as well as never-married individuals but was negatively related among widowed older adults. However, once control variables were introduced, the results shifted, with the impact of childlessness becoming non-significant among both widowed and never-married adults. Thus, differences in the likelihood of living alone, having relatives or friends available, or in other factors likely accounted for the previously negative associations.
With regard to the control variables themselves, in many cases, the results were similar to those obtained for domestic support. However, several notable differences also emerged. Among those who were married, being female was positively associated with the extra-household assistance with transportation. In contrast, being employed or engaged in other employment-related activities (i.e., looking for work, caring for children, or engaged in household work) was associated with reduced support relative to being retired. Home ownership was also associated with reduced likelihood of support. Among those who were separated or divorced, gender also emerged as a significant (and positive) correlate as did low to moderate rather than high income levels. Among those who were widowed, living alone increased the likelihood of support in this domain as did having friends in relatively close proximity. The importance of access to friends was also evident among those who had never married.
The results of the regression of emotional support on marital and parental status are reported in Table 4. Before the introduction of control variables, being childless was associated with reduced support among all four marital status groups (results not shown). However, in multivariate analyses, being childless increased the likelihood of support from extra-household members among those who were married and separated or divorced but had no impact on those who were widowed or never married. Thus, it appears that the control variables served to suppress the otherwise positive impact of childlessness on extra-household support among the former two groups, whereas they explained the negative impact of childlessness on emotional support among the latter two groups.
The impact of the control variables on emotional support differed somewhat from that evident with regard to both instrumental support domains. Whereas gender was not significantly associated with domestic support and only significant among those who were married and separated or divorced when it came to transportation assistance, when it came to emotional support, being female was associated with greater perceived support in all four marital status groups. Among those who were married, older individuals reported being less likely to receive emotional support than those in the reference group (aged 60–64). Similarly, visible minority group members, those with lower levels of education, who were not employed for pay, who did not have friends nearby, who had lived in their residence for a longer period of time, and who had better perceived health but who also had chronic illness and activity limitations, were less likely to report receiving emotional support from others outside the household.
Among those who were separated/divorced, some of the same factors emerged as significantly related: those who were older, who did not have friends nearby, who had lived in their residence for a longer period of time, and who had chronic illness and activity limitations were more likely to report receiving emotional support from others. Among widowed and never-married individuals, age was no longer significant. Instead, lower education was associated with a reduced likelihood of support. Among widowed individuals only, having friends nearby, home ownership, and activity limitations were associated with an increased likelihood of support, whereas being engaged in other employment-related activities (i.e., looking for work, caring for children, or engaged in household work) and longer length of residence were associated with reduced emotional support compared to those who were retired.
Discussion and Conclusions
In this study, we set out to examine the implications of marital and parental status intersections for social support from non-household members in later life, focusing specifically on the comparative advantages and disadvantages of having or not having children among currently married, previously married (widowed, separated/divorced) and never-married individuals. As one of the few studies available to address this issue, it has a number of important theoretical and empirical implications.
First, in contrast with literature suggesting that married individuals and, in particular, married individuals with children, will tend to fare better than those in other marital/parental status groups when it comes to accessing social support, our findings suggest that this is not uniformly the case, at least not when it comes to support from non-household members. Whereas initial bivariate analyses revealed that married individuals and, in particular, married individuals with children were more likely than others to report receiving domestic assistance from those outside the household, this was not the case with regard to transportation assistance or emotional support. Nor was the advantaged position of married individuals with children preserved once selection biases and other factors associated with social support were taken into account. Instead, our results revealed that among those who were married, childless individuals were more likely than those with children to be receiving or reporting access to instrumental sources of support (domestic and transportation assistance) as well as to be receiving emotional support from people living outside the household.
Insofar as childlessness appeared to represent an advantage rather than disadvantage when it came to non-household support for those who were married, our findings would seem to contradict prior studies suggesting that it is married childless individuals who tend to be the most vulnerable insofar as they tend to rely more or less exclusively on one another, whereas those who are unmarried and childless draw from a wider network of social resources.
Our findings would also appear to provide only limited support for prior theoretical and empirical work suggesting that that never married and/or previously married childless older adults are likely to be comparatively disadvantaged when it comes to accessing social support – in other word, that childlessness serves as a particular source of vulnerability in the absence of a spouse. Instead, as noted, given similar social, economic, and health characteristics, childlessness was associated with a reduced likelihood of extra-household support only among never-married and widowed individuals and only then with respect to domestic support. Conversely, there was no evidence to suggest that those who were childless fared worse than those with children in terms of domestic support among those who were separated/divorced; nor was there evidence indicating that, within any of the three unmarried groups, those who were childless fared worse than those with children when it came to help from those outside the household with transportation or emotional support. Instead, those who were unmarried and childless fared either as well (i.e., among widowed and childless, never married and childless) or better (i.e., among separated/divorced and childless) than those with children available.
Although our findings do little to support assumptions regarding the generally negative implications of being never married and childless for extra-household support, neither do they appear to provide clear support for the opposing view that never-married older adults are comparatively advantaged in later life: that is, that unlike other older adults, never-married childless individuals tend to be socially active, resourceful, and, therefore, to reap greater benefit from close ties with selected kin (e.g., siblings and their families, cousins, nieces, nephews) as well as to non-kin (e.g., friends). Instead, among never-married individuals, those without children were found to be less likely to report receiving support from others living outside the household with domestic household tasks but equally likely to be receiving the other two forms of support studied. Therefore, although it may well be that never-married and childless individuals benefit from sibling and other social ties, this does not appear to confer an overall support advantage in later life.
Our findings, such as those indicating that married individuals with children do not necessarily fare better than those in other marital/parental status groups when it comes to accessing social support from others living outside the household, appear to provide little support for normative assumptions embedded within the hierarchical compensatory model and other theoretical perspectives that assume the primacy of the nuclear family for the provision of support. However, does the lack of support found for assumptions regarding the negative implications of being unmarried and/or childless suggest support for the alternative view? In other words, does it suggest that regardless of marital/parental status, older individuals will actively construct their social support networks to ensure their access to the social support they need and value? To the extent that this view suggests little or no difference in the support received across marital/parental status groups, the answer would seem to be no. In fact, systematic differences were evident across the groups with regard to both instrumental (domestic, transportation) and emotional forms of support. Perhaps, as noted by Rook (Reference Rook2009), whereas the view of older adults “as proactively managing their social lives is a valuable antidote to earlier views of older adults as passive victims of societal rejection …, it would be an oversight to ignore the changing life circumstances that can cause the loss or disruption of older adults’ social relationships or that can reveal limitations of their intact social support resources” (p. 104).
In the present study, it would seem that those who were most disadvantaged when it came to accessing support from outside the household were those who were either currently married or separated/divorced but who had children available and those who were never married or widowed but also childless. This suggests that, in some circumstances, having children may serve as a barrier to accessing social support from non-household members (e.g., when one is married and in need of assistance with domestic activities, transportation, emotional support), whereas in other circumstances, not having children may serve as a barrier to support (e.g., when one is widowed and in need of assistance with domestic activities).
Although our findings would therefore provide little support for prevailing generalizations regarding the joint implications of marital and parental status, nevertheless, they do support claims regarding the importance of acknowledging the heterogeneity of the childless experience in old age (Dykstra & Wagner, Reference Dykstra and Wagner2007) and attending to marital status and parental status intersections when addressing issues of support (Dykstra, Reference Dykstra and Uhlenberg2009). In doing so, they also suggest a need to consider the specific types of support involved. Not only did the impact of childlessness differ when comparing currently married to previously married and never-married individuals, but differences also were evident among those childless individuals who were never married, separated/divorced, or widowed. For example, in contrast with frequent suggestions to the effect that childlessness in general has negative implications for support, our findings revealed that this was only the case among widowed and never-married individuals and only with regard to extra-household domestic support. As a result, arguments regarding the greater vulnerability of never-married and childless individuals in general were only partially supported.
Overall, these findings would seem to point to considerable variability and complexity when it comes to issues of support from non-household members among older adults. The likelihood of receiving social support appears to vary depending on older adults’ marital histories, the presence of children, and the type of support involved. That is, findings suggest that it is not simply the presence or absence of a spouse or children that matters but, rather, their presence or absence considered in the context of the marital/family history within which these relationships are embedded.
A major strength of our analyses was the ability to rule out selection effects. In previous studies using cross-sectional data, differences between parents and childless individuals may not be attributable to parenthood per se but rather to the characteristics of those likely to have children. However, our study also has a number of limitations. For example, unmarried and childless individuals are over-represented in institutional facilities, yet our data included only those living in the community. Limitations on sample size were also evident within specific marital and parental status groups (e.g., never married with children). As a consequence, the role of gender and other factors (e.g., health, socioeconomic status) in influencing the impact of marital and parental status on extra-household social support were not considered. Yet it has been suggested that “gender interacts with marital status to influence the composition of caregiving networks” (Barrett & Lynch, Reference Barrett and Lynch1999, p. 695). Furthermore, in some situations, small cell counts may also have contributed to non-significant results. Overcoming such limitations will be essential for future research on these and related questions.
Given the dataset employed for the study, it was also necessary to limit our analysis to two dimensions of social support – instrumental and emotional – as well as to specific indicators of each. Also, as we have noted, our measures of instrumental support combined support received during the past year with that projected to have been available if needed. Potential discrepancies between perceived and received support may therefore be important to consider when researchers compare our results with those reported by others. In general, studies tend to focus on one or the other, with studies of received support (and especially received instrumental support), generally limited to subsamples of older adults with functional impairments or disabilities. However, this latter approach unnecessarily confines investigations of instrumental support to those with disabilities/needs for care and by implication, manages to equate support with care. Yet social support is a broader concept and not confined to those who have disabilities of one sort or another. Further research is needed to address such issues.
Finally, it should be noted that our data do not speak to the overall levels or volume of support received by older adults but, rather, only to support from others living outside the individual’s household. Yet whereas almost all of those who are currently married lived with at least one other person, the vast majority of those who were never married or previously married and childless lived alone. This restriction to extra-household support represents a limitation of the study and, despite our inclusion of living arrangements (alone versus with others) as a control variable, points to the need for research to be conducted that includes both intra- and extra-household support. Such research would help to clarify the importance of children, and particularly of the spouse, for the provision of both instrumental and emotional support. As it stands, because our measures of support do not include intra-household support, support may have been underestimated for those who were married or who lived with children and/or others (i.e., because they would receive more intra-household support but perhaps less extra-household support). By the same token, receipt of support by those who were never married and childless could have been overestimated because they are more likely to live alone (86% were living alone in our study) and have limited intra-household support available to them. In other words, the potential disadvantage of being never married and childless could be understated in this study.
These and other limitations call for further research to be conducted. This includes a need to confirm our findings in different contexts and using different measures of instrumental and emotional support. There is also a need for research to include a focus on who is providing various types of support both inside and outside of the household. However, despite these limitations, this study provides an important update to our understanding of the implications of marital and parental status intersections for the receipt of social support in later life. The findings indicate that, contrary to common assumptions, such structures are not uniformly positive, neutral, or negative in terms of their implications for the receipt of social support. Instead, both advantages and disadvantages appear to be associated with various marital and parental status combinations in later life. Overall, such findings suggest the need for future theoretical and empirical work to address the complexities of these relationships in order to enhance our understanding of these increasingly prevalent family structures.
Appendix A Probit (Selection) Equations of Childlessness: Older Canadians (age 60+), 2007
Appendix B. Zero-Order Correlations of the Explanatory Variables Used in the Regression Analysis
Appendix C. Variance Inflation Factors (VIFs) of the Explanatory Variables Used in the Regression Analysis
