Introduction

An imbalance between work and family demands adversely affects both work and non-work attitudes and elicits negative outcomes (Abendroth & Dulk, 2011; Sirgy & Lee, 2018). These negative outcomes include stress, exhaustion, low productivity, and poor interpersonal relations (Aziz & Cunningham, 2008; Hämmig & Bauer, 2009; Hyman et al., 2003; Makabe et al., 2015). The extant literature on work-family balance has rather thoroughly investigated whether the presence of children leads to work-family conflicts and a lower parental wellbeing (Ford et al., 2007; Kelliher et al., 2019; Milkie et al., 2010; Voydanoff, 1988). But, thus far, the scholarship has paid little attention as to whether the children are living in a stable two-parent household or belong to a “fragile family.” The extant literature defines fragile families as households where parents have never lived together, are widowed, divorced, separated, or are cohabiting without being married (Carlson et al., 2004; McLanahan, 2009). In the past few decades, several concomitant trends have led to a rise in proportion of fragile families, across the globe (Lesthaeghe, 1995; Surkyn and Lesthaeghe, 2004; Cherlin, 2017). While this higher prevalence justifies a greater scholarly attention, the investigation of the effects of raising children in fragile families on parental work-life balance satisfaction has, surprisingly, been neglected in the scholarship. As explained in below, Canada provides a unique context for the investigation of this question.

In Canada, marriage was the normative choice well into the 1960s, as above 90% of the population married over the course of their lives (Boulet & Le Bourdais, 2017; Le Bourdais & Lapierre-Adamcyk, 2004; Mayer & Le Bourdais, 2019). With the gradual decline of marriage, currently, just above half of Canadians are expected to marry (Fostik & Le Bourdais, 2020). Consequently, it is projected that 50% of Canadian children born in the 2010s will spend time living in a single parent or stepfamily household (Vézina, 2015). In the francophone province of Quebec, the institution of marriage has declined substantially more than in the rest of the country, where just about one third of Quebecers are expected to marry in their lifetimes (Cherlin, 2010, 2017; Le Bourdais & Lapierre-Adamcyk, 2004). Marriage has also become more unstable in most countries, including Canada (Cherlin, 2010, 2017). In the 1970s, the Canadian divorce rate was around 10%. However, more recent statistics suggest that about 40% of married couples are expected to divorce (Milan, 2011). In Quebec, in spite of a much lower rate of marriage, nearly half of those who marry will see their unions dissolve (Lapierre-Adamcyk & Le Bourdais, 2004). In fact, while the francophone province of Quebec is very similar to Sweden regarding family arrangements, conjugal life in the Rest of Canada is reminiscent of the United States (Fussell et al., 2007; Le Bourdais & Lapierre-Adamcyk, 2004). Particularly, in Quebec, unmarried cohabitation is almost as prevalent as marriage as a way to start a family; and currently, the majority of Quebec children are born into cohabiting unions (Boulet & Le Bourdais, 2017; Laplante, 2016; Le Bourdais et al., 2016; Mathieu et al., 2020). In contrast, unmarried cohabitation is still regarded as a ‘childless phase of conjugal life’ in the Rest of Canada (Dilmaghani, 2019a; Le Bourdais & Lapierre-Adamcyk, 2004).

Aside from the above mentioned cultural trend differences, to date, Quebec is the only Canadian province which provides parents with universal childcare. In 1997, Quebec implemented a heavily subsidized, $5-a-day, universal childcare program for all parents, regardless of their labour market status and financial needs. The 1997 version of the program covered all children aged 4, expanding to those aged 3 and 2 in 1998 and 1999, respectively. Lefebvre and Merrigan (2008) attributed a 13% increase in female labour force participation and a 22% rise in female labour supply to this policy. In addition, in 2006, Quebec introduced a non-transferable paternity leave to fathers, while only maternity and parental leave remain available elsewhere in Canada (Mayer & Le Bourdais, 2019; Wray, 2020). According to some, the expansion of subsidized childcare and parental leave affordances, and the subsequent improvements in female employment, may transform the work and family ideals of the populations, and bring about a greater gender equality in attitudes (Boulet & Le Bourdais, 2017; Hochschild, 1995; Hochschild & Machung, 1989; Mathieu et al., 2020; Wray, 2020). Hence, the province of Quebec is expected to differ from the Rest of Canada in significant ways. This difference allows for gaining insights about the impact of sociocultural contexts and public policies on the association of family arrangement and parental work-life balance satisfaction.

Using the Canadian General Social Surveys of 2010 to 2016, we examine how belonging to a fragile family associates with parental work-life balance satisfaction. The contribution of this study to the literature is twofold. First, we examine whether there are differences in work-life balance satisfaction between married parents and parents in fragile families, considering also the effects of children’s age and the gender of the parent. Second, we compare the francophone province of Quebec with the Rest of Canada in this regard. The remainder of this study is organized as follows. The next section provides the literature background and lays out the hypotheses. Section III presents the data and the methodology. The results are reported in Section IV. A discussion and the concluding remarks follow.

Literature Background and Hypotheses

In the literature, work-life balance assessment is posited to capture a harmonious interface between different life domains (Abendroth & Dulk, 2011; Hobfoll, 1989). To conceptualize how satisfaction with work-life balance is affected by household composition and family arrangement, we draw on the management literature concerned with work-family conflict and work-family enrichment. In this literature, the Conservation of Resources theory postulates that the time demands from either the work or the family make conflict between the two domains hard to avoid (Allen et al., 2000; Greenhaus & Beutell, 1985; Hobfoll, 1989). A conflict between two domains is associated with lower productivity, higher rates of absenteeism, and lower job satisfaction (Georgellis et al., 2012; Nomaguchi & Fettro, 2018; White et al., 2003). Conversely, as Greenhaus and Powell (2006) argue, work and life domain experiences can enrich each other, and positive experiences from one domain can buffer against the negative effects of the other domain (Ruderman et al., 2002; Carlson et al., 2006; Hanson et al., 2006; Van Steenbergen et al., 2007; Masuda et al., 2012). The Job Demands and Resources framework proposes compatible patterns (Bakker & Demerouti, 2017; Demerouti et al., 2001). Hence, as noted in several meta-analytical studies (Amstad et al., 2011; Ford et al., 2007; French et al., 2018), a satisfactory work-family balance generates a slew of positive individual, organizational, and societal outcomes. The present paper is built on the premise that family arrangement is an important determinant of work-life balance satisfaction of parents; at the same time, the effects of family arrangement on this outcome depends on the social policy and cultural context in which the family is imbedded. In the lines below, this premise is put in the context of the extant literature.

A growing literature is dedicated to the outcomes of both parents and children in fragile families, i.e. households other than those composed of two parents in a stable union (Salganik et al., 2019). A stable union has a positive effect on family income by creating economies of scale and increasing efficiency through specialization of spouses (Becker, 1981). With higher incomes, these families usually move to more desirable neighborhoods and experience a greater residential stability as well (Pinderhughes et al., 2001). Hence, the members of these families generally have a higher social capital and can access greater support from their enlarged social networks (Bowles & Gintis, 2002; Coleman, 1988). A stable union is also reported to improve parents’ mental health through the mutual emotional supports of spouses (Eggebeen, 2005; Gove et al., 1983).

Depending on the social context, a stable union can refer to both marriage and non-marital cohabitation, or legal marriage only. In Canada and other Western countries, the decline of marriage has been largely compensated by the growth of non-marital cohabitation (Bumpass & Lu, 2000; Cherlin, 2004, 2017; Le Bourdais & Lapierre-Adamcyk, 2004; Mathieu et al., 2020; Smock et al., 2005; Wray, 2020). However, in spite of the apparent normalization of cohabitation in the West, differences between marriage and cohabitation, in various outcomes, appear to persist (Dilmaghani, 2019b; Kiernan, 2001; Pelletier, 2016; Thomson et al., 2019). For instance, using Canadian data, Pelletier (2016) shows that while childbearing within cohabitation has increased in Canada, cohabiting families have remained less stable than married ones. Revisiting the question and comparing 15 European countries, Canada, and the USA, Pelletier and Schnor (2017) suggest that the rise in non-marital births is associated with an increase in family disruptions. They also note that the stability gap between married and unmarried couples depends on the prevalence of non-marital childbearing in the social context, where a greater prevalence is associated with a smaller gap. Along the same lines, Thomson et al. (2019), using data from the UK, Italy, Sweden, and Norway, find that, while many among the younger cohorts had their first birth in non-marital cohabitation, those who had children after marriage had the most stable relationships. Thomson et al. (2019) conclude that cohabiting unions, still, remain somewhat less stable than legal marriages. Finally, using Canadian data, Dilmaghani (2019b) finds life satisfaction gaps between married and cohabiting couples in favour of the married, among both heterosexuals and gay males.

These persistent differences between legal marriage and non-marital cohabitation are due to a multiplicity of factors, most of which confound with the reasons for self-selection into either of marriage or non-marital cohabitation. First, it has been reported that pregnancies are less likely to be planned in cohabitation than in marriage (Bouchard, 2005; Finer & Henshaw, 2006; Lachance-Grzela & Bouchard, 2009; Manning, 1995). Second, in cohabitation, the share of household chores is reported to be more equal between spouses (Bardasi & Taylor, 2008; Boulet & Le Bourdais, 2017; Davis et al., 2007; Hamplová et al., 2014; Hersch, 2009). Third, cohabiting couples are reported to have a lower socioeconomic status (Kerr et al., 2006; McLanahan, 2009; Seltzer, 2000, 2004), lower psychological wellbeing and life satisfaction (Dilmaghani, 2019b; Mikucka, 2016; Stavrova & Fetchenhauer, 2015; Wu et al., 2003), higher levels of depression and substance abuse (DeKlyen et al., 2006), and more frequent incidents of domestic violence than married partners (McLanahan, 2009). Finally, there are differences directly associated with self-selection into cohabitation as opposed to marriage; namely, cohabiting partners are generally less committed to their relationships than their married counterparts (Brines & Joyner, 1999; Brown, 2003; Dilmaghani, 2019b; Reed, 2006). Regarding children, those living with cohabiting partners are reported to experience more behavioural and emotional problems than children who reside with two biological (married) parents (Brown, 2003; King, 2018). These children are also more likely to live with a single parent later in their lives (Lachance-Grzela & Bouchard, 2009; Manning et al., 2004; Osborne & McLanahan, 2007; Raley & Wildsmith, 2004).

Single parenthood results from either widowhood, union dissolution, or childbearing of a non-partnered parent. Single parenthood ensuing an unintended pregnancy is associated with lower educational attainment, earnings, and subjective wellbeing of the parent (Fostik & Le Bourdais, 2020; Lachance-Grzela & Bouchard, 2009; Musick, 2007; Smock & Greenland, 2010). There are also significant differences in social support between stable and fragile families, especially those headed by a single parent (Ambrey et al., 2017; Cowan & Cowan, 1992; Eggebeen, 1992). Research finds that single mothers have a diminished access to financial aids and quality housing than married parents (McLanahan, 2009), who usually live outside of the inner city, and in neighborhoods with a higher median income (Turney & Harknett, 2010). There is some evidence that children of a single parent more frequently experience behavioural problems which require a greater time investment (King, 2018; McLanahan, 2009; Osborne & McLanahan, 2007). Single parents themselves are reported to have lower financial security (DeKlyen et al., 2006; McLanahan, 2009) and labour market attainment (Ahearn & Brand, 2019; Harknett et al., 2001). They are also more likely to be employed in occupations with poorer working conditions, a lower schedule flexibility, and a greater time demand (Almquist & Angrist, 1970; Beaudry & Lemieux, 1999; Harmon et al., 2003; Lemieux, 2014). Some researchers posit that it is the labour market conditions that affect marital status, while other researchers have found that it is the labour market attainment which is determined by marital status and family structure (Ahearn & Brand, 2019; Bruns, 2017; Harknett et al., 2001). As evidence of the latter proposition, Noonan et al. (2005) find that having a young child in poor health reduces the likelihood of an unmarried father being employed. Similarly, Richard et al. (2014) report a negative relationship between children’s emotional and behavioural problems and single mothers’ labour supply. Richard (2016) further finds that single mothers of children with emotional and behavioural problems have a significantly lower wage rate than otherwise comparable mothers.

One factor, consequential for the socioeconomic disparities associated with marital status and household composition, is the public provision of childcare. Namely, public provision of childcare can act as a substitute for resources provided by the partner, grandparents, and extended family or friends, where fragile families likely face important deficits (Ambrey et al., 2017; Eggebeen, 1992). In recent years, the provision of state-subsidized childcare has expanded in many western countries, including Canada. In Europe, this expansion was partly in response to the Barcelona targets of the European Union, which aimed at improving parental work-life balance (Schober & Schmitt, 2017), and bringing about a greater gender equality by weakening the male breadwinner model (Crompton, 1999). A growing interdisciplinary literature has been concerned with the impacts of the availability and affordability of childcare on maternal employment and wellbeing. The assessments generally suggest small positive effects on mothers’ labour supply (Haan & Wrohlich, 2011; Havnes & Mogstad, 2011). But, the empirical findings on the relationship between childcare availability and parental wellbeing have been mixed (Baker et al., 2008; Brodeur & Connolly, 2013; Stier et al., 2012). Regarding work-life balance, the evaluations indicate both positive associations (Chung, 2011; Stier et al., 2012; Strandh & Nordenmark, 2006), and an absence of association (Steiber, 2009; Van der Lippe et al., 2006).

The theoretical frameworks accommodating the effects of childcare policies on parental wellbeing are (i) the Demands and Resources approach (Bakker & Demerouti, 2017; Demerouti et al., 2001, 2010; Voydanoff, 2005); and (ii) the Social Production Function theory (Lindenberg, 1986; Ormel et al., 1999). Using the person-environment fit framework, Voydanoff (2005) asserts that work-family balance relies on a global assessment that work and family resources are sufficient to satisfactorily meet both work and family demands. Hence, within the Voydanoff (2005) framework, subsidized childcare is a boundary-spanning resource, which allows parents to better meet structural or psychological demands in both work and family domains (Desrochers & Sargent, 2004; Schober & Schmitt, 2017). Rather similarly, the social production function framework (Lindenberg, 1986; Ormel et al., 1999), implies that the individuals’ goal is to maximize their own physical and social wellbeing. The simultaneous demands of work and family, which conflict with the available limited resources, impedes the ability to achieve this goal. Since subsidized childcare helps with meeting the competing demands of work and family, it improves subjective wellbeing, inclusive of work-life balance satisfaction. In couples, while working mothers are more likely to benefit from subsidized childcare, the positive effects on maternal wellbeing can spillover to their (ex) partners (Lourel et al., 2009; Schober & Schmitt, 2017; Schober & Stahl, 2014). Therefore, a greater childcare availability is likely to be positively associated with the wellbeing of fathers as well, even if they are not the primary caregiver. This spillover can also elicit more desirable work outcomes for both parents (Hanson et al., 2006; Masuda et al., 2012).

In fact, as work and family are interrelated domains, various approaches have been developed to conceptualize their linkages (Voydanoff, 2001). For instance, ecological systems theory (Voydanoff, 2001), boundary theory (Desrochers & Sargent, 2004), and spillover theory (Edwards & Rothbard, 2000; Georgellis et al., 2012) have been applied to the question. Boundary theory suggests that when the boundaries between the work and family are permeable, work and family domains influence each other (Ashforth et al., 2000). Acknowledging the interaction between family and work demands, Voydanoff (2001) proposes two linking mechanisms: perceived work-family fit and perceived work-family balance. In this framework, stress arises from the lack of fit between the person and the environment rather than independently from one domain (Voydanoff, 2001). According to Voydanoff (2001), “fit” has two variants: (i) demands-abilities fit; and (ii) needs-supplies fit. The provision of subsidized childcare is an additional environmental “supply,” which is expected to improve “needs-supplies fit” for parents. Rather similarly, the social production function theory suggests that subsidized childcare, supplied through a social production function, allows parents to produce a greater wellbeing (Ormel et al., 1999; Schober & Schmitt, 2017).

Empirical investigations of the effects of socially-provisioned childcare are growing. Examining the effects of childcare availability in Germany, Schober and Schmitt (2017) find evidence for a positive effect only for single mothers, and for partnered mothers in West Germany. Their results suggest that the differences in work-family culture between East and West Germany moderate the relationship between childcare availability and subjective wellbeing of mothers; as with its legacy of state-socialism, East Germany has a greater cultural acceptance for maternal employment and outsourced childcare (Hašková & Klenner, 2010; Schober & Stahl, 2014, 2016; Schober & Schmitt, 2017). Accordingly, Treas et al. (2011), comparing 28 countries, found that married working women were less happy than those who did not work, especially in countries with a less developed female labour force participation. These findings are in line with the seminal contribution of Goode (1993), suggesting that in societies where legal marriage and the male breadwinner model are destabilized, union dissolution and alternative family arrangements become institutionalized and the society creates laws, regulations, and norms which mitigate the negative effects of this cultural evolution.

Since 1997, the provision of subsidized universal childcare is an important factor distinguishing Quebec from the Rest of Canada. The availability of childcare services in Quebec has been shown to be positively correlated with maternal labour supply (Haeck et al., 2015; Lefebvre & Merrigan, 2008). At the same time, research has also found that this policy has caused a fall in parents’ life satisfaction, especially among the more educated (Baker et al., 2008; Brodeur & Connolly, 2013). While the latter consequence of subsidized childcare in Quebec may appear surprising, it has previously been found that childcare availability can lower mothers’ wellbeing, if it induces them to enter the labour force and still bear the brunt of household chores and childcare (Blanchflower & Oswald, 2004; Hochschild, 1995; Stevenson & Wolfers, 2009). In addition, owing to the relatively low quality of the Quebec childcare policy (Japel et al., 2005), negative causal effects on child outcomes have been consistently and overwhelmingly reported (Baker et al., 2008, 2019; Baker, 2011; Kottelenberg & Lehrer, 2013, 2014, 2017, 2018; Haeck et al., 2015). Particularly, Baker et al. (2019) find that, among the cohorts exposed to the Quebec universal childcare, the negative effects persisted, with worse health and life satisfaction outcomes as well as higher crime rates later in life. These negative effects can spillover to parents.

In light of the literature reviewed in the above, we start the study with the rather intuitive hypothesis below.

Hypothesis 1

Satisfaction with work-life balance is the highest among married partners, followed by cohabiting parents, and those without a domestic partner.

Next, owing to the disruption costs of a union dissolution, to both parents and children (Dilmaghani, 2019a; McLanahan, 2009; Meadows, 2009; Meadows et al., 2007), we propose:

Hypothesis 2

Divorced and separated parents have a lower satisfaction with their work-life balance than parents who never had a domestic partner.

In general, fathers are found to be less involved in childrearing than mothers, even when in a union with the child’s mother. For the separated and divorced fathers, substantially lowered emotional and financial support to their children have been reported (Carlson et al., 2008; Nepomnyaschy & Garfinkel, 2010). Thus, we propose the hypothesis below.

Hypothesis 3

Work-life balance satisfaction disadvantages of mothers in fragile families, when compared with married mothers, are larger than the disadvantages of fathers in fragile families, when compared with married fathers.

As argued in Goode (1993) and reiterated in Kiernan (2001), in steadily high divorce societies, union dissolution becomes institutionalized and the society creates alternative norms to replace the familial security of a stable marriage. This argument, highlighting the role of culture and institutions on family outcomes, has been echoed in the empirical works previously reviewed (Boulet & Le Bourdais, 2017; Laplante, 2016; Le Bourdais et al., 2016; Mathieu et al., 2020; Wray, 2020). For instance, Pelletier (2016) finds that the effects of cohabiting union dissolution on children depend on how common cohabitation is in their area of residence. This context-dependence is also echoed in Schober and Schmitt (2017), documenting the differences between East and West Germany. In light of these assertions, and notwithstanding the reported negative consequences of some of Quebec’s social policies (Baker et al., 2008, 2019), we propose the following final hypothesis.

Hypothesis 4

The work-life balance satisfaction gaps between married and other types of parents are smaller in Quebec than in the Rest of Canada.

Data and Methodology

The data used in this paper are from the Canadian General Social Survey (GSS). The GSS datasets are nationally representative probability samples of the non-institutionalized population of Canada, 15 years of age and older. These data are collected by phone interviews, and the respondents are reached through ‘Random Digit Dialing’ of the phone numbers registered as ‘in service for residential use’ in Statistics Canada’s administrative sources (Statistics Canada, 2017). Statistics Canada started conducting the Canadian General Social Surveys in 1985. While the GSS is conducted yearly since 1985, each cycle relies on a new random sample. Hence, even if several cycles are used, the data structure remains cross-sectional. Only a few of the GSS cycles include a question on respondents’ satisfaction with their work-life balance (WLB). The most recent available cycles which have surveyed the respondents on their WLB satisfaction were conducted in 2010, 2011, 2012, 2015, and 2016. These five cycles are pooled to achieve a large enough sample size by marital status and province of residence. Given the goal of the paper, the sample is restricted to those who were gainfully employed in the year prior to the GSS interview, and had at least one dependent child. Among the working respondents with dependent children, the number of the widowed was too small to warrant a meaningful separate analysis. And, given the emotional and financial complications arising from losing a spouse, widowed respondents could not be reasonably pooled with the divorced and separated. Hence, they are excluded from the sample. The missing observations are assumed ‘missing at random,’ and are dealt with through list-wise deletion. Excluding the missing observations, and implementing the sample restrictions, the total number of observations remains as large as 15,870 men and women.

The WLB question is identically formulated in all the five GSS cycles pooled and the responses are uniformly recorded on a 5-item scale. More precisely, the GSS asks the respondents “How satisfied are you with the current balance between your job and home life?,” invariably across the cycles. The response options offered are Very satisfied, Satisfied, Neither Satisfied nor dissatisfied, Dissatisfied, and Very dissatisfied. These GSS cycles also contain basic sociodemographic information such as age, marital status, labour market status, income, and the highest degree attained by the respondents. In the analyses, separated individuals are grouped with the divorced, as the number of observations in the former category was too small for a separate analysis (referred to as Divorced/Separated in the tables). The category recorded in the GSS under “Single, Never married” refers to the respondents who have never lived with a domestic partner, either as a married or a common-law couple (referred to as Single/Never married in the tables).

Table 1 shows the descriptive statistics, by marital status, gender and region of residence. Within each gender, t-tests of mean differences by region of residence have been conducted, and statistically significant differences are noted by asterisks. As shown in Table 1, the distribution of marital status substantially varies between Quebec and the Rest of Canada (ROC). For women, as reported in the top panel, outside Quebec, 75.22% of working women with at least one child are married against 43.87% in Quebec. Conversely, 9.12% of women are cohabiting in the ROC, versus 40.43% of Quebec women. Remarkably, in spite of the substantially greater decline of marriage in Quebec, the share of women with a domestic partner is almost identical in this province (84.30%) with the Rest of Canada (84.34%). The WLB statistics are computed as the percentage who report being very satisfied or satisfied with their work-life balance. In the ROC, the WLB is the highest for married mothers, at a non-negligible distance with other categories. The WLB differences are small among cohabiting, divorced/separated, and single/never married mothers in the ROC. The pattern differs for Quebec mothers, as smaller WLB gaps related to marital status are observed.

Table 1 Descriptive statistics
Full size table

The educational attainment appears to be the highest among the married, invariably across Canada. In Quebec, the fertility rates of married and cohabiting women are almost identical, while cohabiting women have slightly fewer children than the married in the Rest of Canada (ROC). One interesting point to note is that educational attainment is statistically significantly higher among cohabiting and never married mothers in Quebec than comparable mothers in the ROC. While cohabiting parents are found to generally have a lower socioeconomic status than the married (Kerr et al., 2006; McLanahan, 2009; Seltzer, 2000, 2004), this statistically significant difference suggests the relatively higher socioeconomic status of cohabiting and single mothers of Quebec than comparable ROC mothers. Also, fertility is statistically significantly higher among cohabiting Quebec mothers than their ROC counterparts, while the opposite holds for divorced mothers. The statistically significant higher fertility of cohabiting Quebec mothers indicates the previously documented greater normalization of unmarried cohabitation in Quebec, as a pathway to start a family (Boulet & Le Bourdais, 2017; Le Bourdais & Lapierre-Adamcyk, 2004).

The lower panel of Table 1 reports the same statistics for males. Interestingly, among men, married and divorced fathers have very close WLB satisfaction levels in the ROC, followed by cohabiting fathers, and the single/never married at a distance. In Quebec, cohabiting fathers have the highest level of WLB satisfaction, followed by the married. The patterns regarding education and fertility levels do not differ much compared with those observed among women. Similarly to the statistics for females, cohabiting Quebec fathers are statistically significantly more likely to have a university degree than their ROC counterparts. Again, this pattern suggests the higher socioeconomic status of cohabiting parents in Quebec than in the ROC. Also, the number of children of cohabiting and never married fathers is statistically significantly higher in Quebec than in the ROC. Taken together, the higher educational attainment and fertility of cohabiting and never married parents in Quebec, reaffirms the greater normalization of non-marital family formation in Quebec, compared with the ROC.

Figures 1, 2, 3 and 4 provide some visuals for the statistics of Table 1. Figures 1 and 2 clearly illustrate the large differences in marital status of parents between Quebec and the ROC. Figures 3 and 4 show the generally higher WLB satisfaction of Quebec parents, compared with parents residing in the ROC.

Fig. 1
figure 1

Marital status across Canada, mothers

Full size image
Fig. 2
figure 2

Marital status across Canada, fathers

Full size image
Fig. 3
figure 3

Work-life balance satisfaction across Canada, mothers

Full size image
Fig. 4
figure 4

Work-life balance satisfaction across Canada, fathers

Full size image

The descriptive WLB statistics, reported in Table 1 and Figs. 3 and 4, do not account for the confounding factors. Multivariate regression, conducted using Probit, is implemented to investigate whether the marital status-related discrepancies in WLB satisfaction subside after controlling for the confounding factors. The generic form of the estimated equation is as follows:

$$\Pr \left( {{\text{WLB}} = 1|Z} \right)\; = \; \Phi \left( {\delta_{0} \; + \;\mathop \sum \limits_{i = 1}^{3} \delta_{i} {\text{Mar}} {\text{Stat}}_{i} \; + \;X\beta \; + \; \varepsilon } \right)$$

The dependent variable, WLB, is a dummy taking the value of 1 for those who have reported being Very Satisfied or Satisfied with their work-life balance. The variables of interest Mar Stati capture the three categories of (i) unmarried cohabitation; (ii) divorced/separated; and (iii) never married/single. Hence, married respondents are the reference category.

The \(\mathrm{Vector} X\) includes the confounding variables, selected to reflect the validated determinants of subjective wellbeing (Helliwell, 2003). Namely, age and its squared form are included to allow for the well documented u-shaped relationship between age and wellbeing outcomes (Blanchflower & Oswald, 2004). Education is also controlled for, nested through dummies for below high school, some college, and having a bachelor’s degree or above, leaving those with a high school diploma as the reference category. Annual income and the number of annual hours worked are also controlled for, after log-transforming them to allow for nonlinear patterns. Income is not adjusted for inflation as the time-span is short and the regressions control for a vector of survey cycle dummies as well as a time trend. The number of dependent children is also controlled for. While the dataset did not include the exact ages of dependent children, rather detailed information about their age groups was available. Using this information, three dummies control for when all the dependent children were (i) under the age of 5; (ii) between 5 and 12 years of age; (iii) older than 13 years of age. Those with mixed-age children are the reference category. Dummies for immigrant and visible minority status are also included in the regressions. Finally, in the subsample of the ROC (containing nine Canadian provinces), dummies control for eight provinces, leaving Ontario as the reference category. In addition to sample partitioning between Quebec and the Rest of Canada (ROC), the sample is further restricted by children’s age groups. The implemented sample restrictions allow the confounders to have varying effects on the outcome of interest (WLB satisfaction). All the estimations are made using sample weights. Sample weights are adjusted before data pooling. The robust standard errors are clustered by province of residence and GSS cycle. The equations are estimated by Probit. In the results tables, the marginal effects of the Probit coefficients are reported.

The methodological choice of dichotomizing the dependent variable is based on several considerations. Most importantly, since the responses to the surveys’ wellbeing questions are qualitative, a choice has to be made regarding the treatment of the data. Some scholars treat the wellbeing responses as cardinal values and use OLS as the estimation methodology (Ng, 1997). Alternatively, Ordered Logit can be used to preserve the ordinal meaning of the response items. But, Ordered Logit method creates some difficulties in the interpretation of the results (Ferrer-i-Carbonell & Frijters, 2004). The third option of dichotomizing the dependent variable is often employed to preserve the ordinal nature of the responses, and at the same time simplify the interpretation of the results (Chakraborty et al., 2011; Dilmaghani, 2018a). In their in-depth examination of this question, Ferrer‐i‐Carbonell and Frijters (2004) conclude that assuming cardinality or interpersonal ordinality of wellbeing scores makes little difference to the conclusions. The disadvantage of dichotomization is the loss of some information incorporated in the data. Here, to simplify the interpretation of the results and to preserve the ordinal nature of the responses, a dichotomized version of the dependent variable is used, with Probit methodology. The effects of this choice are addressed in the robustness tests, discussed later in the paper.

Before presenting the results, the limitations of this study must be noted. First, subjective wellbeing assessments are partly determined by stable individual characteristics (Diener et al., 2003). These stable traits simultaneously impinge on marital status, education, labour market outcomes, and the like. Only datasets with a longitudinal structure allow to address this endogeneity issue by using individual fixed effects (Georgellis et al., 2012). It was not the case with this dataset. Second, establishing the direction of causation has nontrivial data requirements. The cross-sectional structure of the GSS data did not allow any means to examine the direction of causation. In light of this limitation, the results must be interpreted as ceteris paribus correlations, rather than causal relationships. Third, it is most sensible to make a distinction across cohabiting couples by the duration of their unions, since long-term non-marital partnerships are likely as stable as a marriage (Boulet & Le Bourdais, 2017; Dilmaghani, 2019a). But, the GSS cycles used in this paper were mute about the duration of cohabitation, except for the cycle conducted in year 2011. Unfortunately, a single GSS cycle cannot provide an adequate number of observations for meaningful analyses. Therefore, we could not further break down cohabiting couples by the duration of unions. Finally, some important confounders, such as the degree of job flexibility, could not be included as controls in the regressions, due to data limitations.

Results

Table 2 reports the marginal effects of four Probit estimations. The sample is restricted to working women with at least one dependent child who reside in the Rest of Canada (i.e. all Canadian provinces aside from Quebec). The dependent variable is a dummy taking the value of 1 for those who report being Very Satisfied or Satisfied with their WLB. Column (1) does not employ any control. The confounders are sequentially added to the equation from Column (2) to Column (4). Column (2) adds demographic variables, such as the respondent age and age squared as well as the number and age groups of the children. Column (3) further includes human capital variables, i.e., education and ethnic background. Finally, Column (4) adds the natural logarithm of annual income, the natural logarithm of total annual hours worked, time trend and a vector of survey cycle dummies, and dummies for the province of residence. The control variables, not all shown in the table to save space, are exhaustively listed as note to the table.

Table 2 Mothers’ WLB satisfaction: Canada, excluding Quebec (ROC)
Full size table

As shown in Column (1), when no covariate is accounted for, cohabiting mothers are 6.2% points less likely to be very satisfied or satisfied with their WLB than married mothers, while divorced/separated mothers are 6.5% points less likely to report this outcome, in the ROC. There is also a weakly statistically significant coefficient for the “single/never married,” indicating that these mothers are 5.8% points less likely to be satisfied with their WLB than married mothers. The gradual addition of controls shows the effects of the confounders on these zero-order gaps. More specifically, with the Column (2) inclusion of demographic variables, such as children’s age groups, the disadvantage of divorced/separated (cohabiting) mothers marginally grows (shrinks). In this column, as could be expected, the coefficient for the variable indicating all children are younger than 5 years of age (older than 13 years of age) is negative (positive). Hence, the Column (2) changes in the gaps likely reflect the fact that the odds of both non-marital cohabitation and divorce are correlated with children’s age. Namely, some cohabiting couples are likely to transition to marriage within a few years of a childbirth while some couples may part ways when their children become “old enough” (Raley & Sweeney, 2020; Wright, 2019).

As shown in Column (3), higher educational attainment predicts a lower WLB satisfaction. Comparable patterns have been reported regarding job satisfaction, and have been attributed to the greater demands of high-skilled jobs (Clark, 1997). The Column (3) inclusion of human capital variables slightly shrinks the gap of cohabiting mothers with those who are married. This pattern is congruent with the previously reported lower socioeconomic status of cohabiting couples (Kerr et al., 2006; McLanahan, 2009). In contrast, the disadvantage of divorced/separated mothers is enlarged with this set of controls. Given the negative association of education with odds of divorce (Jalovaara, 2003), this pattern is not surprising. Finally, the Column (4) inclusion of hours worked and income (both in natural logarithm) further shrinks the gap between cohabiting mothers and the married, with rather negligible effects on the other coefficients of interest (dummies for Divorced/Separated and Never married/Single). This finding further underscores the role of socioeconomic status in explaining the differential WLB outcomes of married and cohabiting mothers of the ROC.

In sum, as shown in Column (4), a rather large gap, 8.2% points lower likelihood of being very satisfied or satisfied with one’s WLB, is found between married and divorced/separated mothers who reside in the ROC. In addition, as Column (4) shows, the coefficient for single/never married women is not statistically significant, while the coefficient for cohabiting mothers indicates a 3.6% points WLB satisfaction disadvantage compared with married mothers. Regarding the controls, as noted in Column (4), having all children above 13 years of age remains a positive predictor of WLB satisfaction. The other indicators of children’s age are no longer statistically significant. Having a university degree is a negative predictor of WLB satisfaction. But, interestingly, immigrants and visible minority mothers are found to be more satisfied with their WLB than the white Canadian-born. A possible explanation for this pattern, worthy of further exploration, is differential expectations (Wang and Jing, 2018). Finally, the number of hours worked (in natural logarithm) is a large and negative predictor of WLB satisfaction for working mothers in the ROC.

Table 3 follows the same approach as Table 2, but restricts the sample to Quebec mothers. As the table shows, there are zero-order gaps for cohabiting and single/never married mothers which eventually lose their statistical significance. As shown in Column (2), with the inclusion of variables pertaining to children’s age, the coefficient for single (cohabiting) mothers increases (shrinks). This pattern is compatible with the association of children age with marital status transitions among unmarried parents (Wright, 2019). With Column (3) inclusion of human capital variables, the WLB disadvantages of cohabiting and single/never married Quebec mothers gain momentum. But, they ultimately subside in Column (4), where income and labour supply variables are added to the equation. As exhibited in Column (4), accounting for all the confounders, there is no WLB satisfaction gap associated with marital status for Quebec mothers. With respect to the controls, having all children aged older than 13 has a positive and quite sizable coefficient, indicating 14.2% points higher likelihood of WLB satisfaction. Regarding education and labour supply, like elsewhere in Canada, having a university degree and the number of hours worked are negatively associated with the outcome, with sizable coefficients.

Table 3 Mothers’ WLB satisfaction: Quebec
Full size table

Table 4 focuses on fathers residing in the ROC. The same approach of sequential addition of controls is employed. As shown in the table, there is no gap associated with marital status for fathers residing in the ROC, in any of the specifications. Among the controls, having all children older than 13 years of age, an education level below high school diploma, and being a visible minority are found to positively associate with the outcome. In contrast, similarly to females, the number of hours worked is found to be a potent negative predictor of WLB satisfaction among fathers residing in the ROC.

Table 4 Fathers’ WLB satisfaction: Canada, excluding Quebec (ROC)
Full size table

Table 5 examines the WLB outcomes of Quebec fathers. The structure of this table is identical to Table 4. As shown in Table 5, similarly to fathers residing in the ROC, there is no gap associated with marital status for Quebec fathers either. Among the controls, having all children younger than 5 years of age, having a university level education, and the number of hours worked are found to negatively associate with the outcome.

Table 5 Fathers’ WLB satisfaction: Quebec
Full size table

Table 6, partitions the sample by the age group of children and region of residence. This investigation is guided by the statistically significant coefficients consistently found for the age group of children in Tables 2, 3, 4 and 5. These additional regressions allow to ensure that the overall patterns are not driven by parents with children of a specific age group. Since with this partitioning, the sample sizes become relatively small, the results are tentative. The coefficients for mothers are presented in Panel A, while Panel B focuses on fathers. Moreover, the left panel presents the results for the ROC while the right panel uses the subsample of Quebec parents. As the coefficients remained largely similar to those reported in the previous tables, the controls are all suppressed.

Table 6 WLB satisfaction by children age
Full size table

In Column (1), the sample is restricted to mothers residing in the ROC whose children are all younger than 5 years of age. As shown in the table, divorced/separated women are 24.4% points less likely to be satisfied with their WLB than those who are married. For women with children between 5 and 12, there is no WLB satisfaction gap by marital status. When the sample is restricted to women with dependent children of 13 years of age and older, as shown in Column (3), there are large gaps related to marital status regarding WLB satisfaction in the ROC. Divorced/separated mothers are 14.4% points less likely to be satisfied with their WLB than married mothers while single/never married mothers are 21.1% points less likely to report this outcome than their married and cohabiting counterparts. In contrast, in Quebec, the only weakly statistically significant coefficient belongs to divorced/separated mothers of children between 5 and 12 years of age. No other coefficient is statistically significant for Quebec mothers. Panel B of Table 6 restricts the sample to fathers. In contrast to mothers residing in the ROC, there is no gap associated with marital status for fathers in the Rest of Canada. For Quebec fathers, there are two weakly statistically significant gaps, indicating higher (lower) WLB satisfactions for cohabiting (divorced/separated) fathers of children aged between 5 and 12 (older than 13) years of age. One point of caution is the small samples for Quebec, which have likely made the results sensitive to the presence of outliers. Therefore, as previously noted, the investigations of Table 6 are tentative, especially for Quebec parents.

Numerous sensitivity tests have been conducted. Wherever the samples did not become too small, it was made sure that the patterns are not driven by a single GSS cycle. To do so, the estimations are re-made excluding one GSS cycle at a time. The results proved very robust. Moreover, using OLS or Ordered Logit did not make a difference to the estimates. Finally, employing the raw WLB scores as cardinal values, instead of dichotomizing the dependent variable, led to similar conclusions. Some of these regressions are shown in the Appendix. The rest of the sensitivity tests are available upon request.

Discussion and Conclusion

The contributions of this paper to the literature can be summarized as follows. First, the analysis shows that in the francophone province of Quebec, married mothers and their counterparts in “fragile families” have rather comparable work-life balance satisfaction levels. But, for mothers living in the Rest of Canada, marital status was found to strongly associate with work-life balance outcomes, with advantages for married mothers. Taken together with the generally higher work-life balance satisfaction of Quebec parents, shown in Figs. 3 and 4, it appears that the association of marital status with work-life balance satisfaction is sensitive to the cultural and policy contexts. Second, comparing divorced and separated mothers with those who have never been in a couple showed significant work-life balance satisfaction disadvantages for the latter group, outside of Quebec. These patterns indicate that perhaps, “fragile families” should not be defined in the same way in Quebec and in the Rest of Canada. Finally, the analysis produced evidence that marital status is generally not a consistent predictor of fathers’ work-life balance satisfaction in Canada as a whole. More formally, the Hypotheses 1, 2 and 3 are largely confirmed for the Rest of Canada and rejected for Quebec, while the Hypothesis 4, which is concerned with the Quebec difference, is confirmed. In the lines below, the highlights of these findings are further discussed.

As Lapierre-Adamcyk and colleagues report, the majority of Canadians prefer ‘to have a lasting relationship as a couple (Lapierre-Adamcyk et al., 1999). Yet, the trends of decoupling have been strong in Canada, much like in comparable Western countries (Cherlin, 2017; Pelletier & Schnor, 2017). Marital dissolution generally leaves one parent, usually the mother, with a greater responsibility in childrearing. Although, the imbalance in childrearing responsibilities is likely smaller in more gender-equal societies such as Quebec (Boulet & Le Bourdais, 2017; Crompton, 1999). In the past literature, numerous consequences of family disruptions have been investigated (King, 2018; Mathieu et al., 2020; McLanahan, 2009; Osborne & McLanahan, 2007). Partially confirming Hypotheses 1 and 3, the present paper showed that parenting in a “fragile family” is consequential for work-life balance satisfaction of mothers (and not fathers), in Canada outside of Quebec (i.e. the Rest of Canada). While this result, by itself, is not surprising, it is rather interesting when put into the perspective of two other findings of this paper. First, as noted in Hypothesis 2, slighter work-life balance satisfaction disadvantages are found for working mothers who never were in a couple. Second, as proposed in Hypothesis 4, the results indicate that this pattern does not hold for the province of Quebec.

According to Kiernan (2001), the societal evolution of cohabitation is a four phase process. In the first stage, it is an “avant-garde” phenomena, as it was in the early 1970s in the West. In the second stage, nonmarital cohabitation is normalized. In the Rest of Canada, cohabitation seems to have reached this second stage (Boulet & Le Bourdais, 2017; Le Bourdais & Lapierre-Adamcyk, 2004). In the third stage, cohabitation is an alternative to marriage, while in the fourth stage, it is a perfect substitute for legal marriage. In Quebec, given its social policy and cultural differences with the Rest of Canada, non-marital cohabitation can be safely regarded as a real alternative to marriage. Even, arguably, the Quebec society can be regarded to be in the fourth stage (Kiernan, 2001; Le Bourdais & Lapierre-Adamcyk, 2004; Laplante, 2016; Pelletier, 2016). According to Goode (1993), with the prevalence of alternative forms of marital arrangement and higher rates of marital dissolution, societies develop ways to cope and mitigate the negative effects of these changes. It appears that in Quebec, these developments have taken place, as evident by its universal childcare and non-transferable paternity leave policies (Baker et al., 2008, 2019; Brodeur & Connolly, 2013; Mayer & Le Bourdais, 2019; Wray, 2020). In fact, Quebec’s welfare provision policies are qualified as more ‘social-democratic’ than ‘liberal,’ where the latter label applies to the Rest of Canada (Mathieu et al., 2020).

There are also more general cultural differences between Quebec and the Rest of Canada, such as differences in religious orientation and gender role attitudes (Dilmaghani, 2018a, 2018b; Le Bourdais et al., 2016). Particularly, up to the 1960s, the Catholic Church had a stronghold on Quebec institutions and personal lives. The Quiet Revolution of the 1960s, led to an effective secularization of Quebec institutions and the creation of a welfare state (Gauvreau, 2005). Quebec women enthusiastically embraced the cultural changes of the Quiet Revolution (Le Bourdais & Lapierre-Adamcyk, 2004). Given the association of the institution of marriage with the Catholic Church, the decline of marriage has been much stronger in Quebec than in the Rest of Canada. Specifically, the fertility rates, previously higher in Quebec than in the Rest of Canada, strongly fell in the province (Dilmaghani, 2018a, 2018b, 2019a), and the divorce rates rose to one in two marriages (Lapierre-Adamcyk & Le Bourdais, 2004).

The present investigation showed that, unlike the Rest of Canada, Quebec mothers in “fragile families” appear as likely to be satisfied with their work-life balance as those who are legally married. Therefore, it seems that Quebec society has culturally adapted to the new schemes of conjugal life and household composition (Goode, 1993; Kiernan, 2001). This cultural difference is reminiscent of the difference between the former East and West Germany in attitudes towards maternal labour supply and outsourced formal childcare (Hašková and Klenner, 2010; Schober & Stahl, 2014, 2016; Schober & Schmitt, 2017). Institutionally as well, Quebec social policies are consistent with this differential adaptation, when compared with the Rest of Canada. In fact, Quebec is generally considered more supportive of work-family balance than elsewhere in Canada (Campbell, 2006; Maillé, 2018; Tremblay, 2010, 2012). Examples of social policies in line with this greater support are the previously reviewed universal childcare of 1997 and the non-transferable paternity leave of 2006, unique to Quebec in the Canadian context (Baker et al., 2008, 2019; Brodeur & Connolly, 2013; Mayer & Le Bourdais, 2019; Wray, 2020). Our results are compatible with the view that such policies are effective in helping mothers meet the demands of both their jobs and their personal lives in a satisfactory manner. Further, the results indicate that the definition of “fragile family” is context-dependent, and must be adjusted to reflect both social policies and cultural attitudes regarding expected family arrangements and gender norms.

In fact, another possible explanation for the Quebec difference relies on gender role attitudes within the family. Particularly, it is possible that ensuing a marital dissolution, Quebec mothers are able to negotiate a greater contribution by fathers to childrearing tasks than mothers in the Rest of Canada. Evidence congruent with this conjecture is presented in Boulet and Le Bourdais (2017), who studied the effects of the division of household chores between spouses on work-life balance satisfaction in Quebec and elsewhere in Canada. Yet, detailed comparative evidence on the division of childcare between mothers and fathers by marital status is still lacking. Hence, the present study showcases several literature gaps in the investigation of the consequences of the Quebec difference in the Canadian context (Mathieu et al., 2020; Pelletier, 2016). Namely, the application of the household specialization theory (Becker, 1981; Foster & Stratton, 2018), the relative resource theory (Brines, 1994; Bianchi et al., 2000; Álvarez & Miles, 2003), and the literature on “doing gender” (Gupta, 1999; West & Zimmerman, 1987) can shed light on the other implications of the differences between Quebec and the Rest of Canada. These frameworks, of interest to economists, sociologists, and demographers, have not yet been empirically applied to compare Quebec with the Rest of Canada.

This paper’s final contribution is the finding of a lower work-life balance satisfaction disadvantage for mothers who have never been with a domestic partner, compared with divorced/separated mothers. Particularly, while a rather large work-life balance satisfaction gap was found for divorced/separated mothers in the Rest of Canada, especially for those with young children, no such disadvantage was found for single mothers who did not experience a union dissolution (single/never married category). This difference, perhaps, reflects the disruption of household organization following a union dissolution (Frank et al., 1987; Pelletier & Schnor, 2017; Thomson et al., 2019). Such disruption is not relevant to the parents who have never lived together as a couple. Family disruption is documented to have long and short run negative effects on children (Baker et al., 2019; Haeck et al., 2015; Hawkins & Eggebeen, 1991; Thomson et al., 2019). The results reported here indicate that such disruptions are also detrimental to the mothers’ work-life balance satisfaction, an outcome never studied before. Future research should establish whether this pattern is specific to the Canadian context, or holds for comparable countries.