1 Introduction

Achieving a good level of work–life balance (WLB) is an important goal for many people (Newman 2011; Sturges and Guest 2004). Experiencing WLB provides a greater sense of completeness in life (Greenhaus and Powell 2017) as, through the participation in multiple roles and realms, individuals have more chances to fulfil a variety of significant needs and thrive (ten Brummelhuis 2016). In support of these claims, research has shown that WLB is positively related to work-related, family-related and health-related outcomes, including well-being (Gropel and Kuhl 2009; Haar et al. 2014; Lyness and Judiesch 2014; Wheatley 2017).

A critical point when examining WLB concerns its conceptualization. Albeit in the past scholars have mostly assumed that WLB coincides with equal satisfaction and effectiveness in both work and non-work domains (Valcour 2007), the absence of work–family conflict, and/or the presence of work–family enrichment (Guest 2002), there is now consensus that WLB depends on the perception of fit between personal values, preferences and aspirations and the demands associated with work and non-work roles (Greenhaus and Allen 2011). In other words, people tend to perceive greater WLB when they feel that the different parts of their lives that matter the most to them fit together well (Greenhaus and Powell 2017).

Drawing on this definition, in this paper we investigate, at the individual level, the relationship between, on the one hand, the perceived fit between the number of hours worked and the hours spent engaging in other commitments outside the work (hereafter perceived work–life fit) and, on the other hand, subjective well-being. Working hours are a predictor of WLB (Milkie et al. 2010; OECD 2013; Unger et al. 2015). Nonetheless, subjective well-being is mainly influenced by the perception of fit between working hours and personal preferences and desires regarding working hours, rather than by the actual hours worked (Bardasi and Francesconi 2004; McKee-Ryan and Harvey 2011; Wooden et al. 2009).

That said, a key issue in current research investigating how employees assess their work–life interface based on what they consider is most appropriate for them (e.g. Sturges 2012; Thompson and Bunderson 2001) is to understand the real freedom that anyone has to live in ways that are meaningful and valuable (Hobson 2011, 2014). Following the capability approach (Nussbaum 2000; Sen 1992), the real freedom to attain WLB is a function of both individual resources and contextual factors, which may either inhibit or enable preferences, aspirations, and goals. Tensions, which result in lower well-being, may arise when expectations about WLB are constrained and individuals are not able to achieve the life (and WLB) they truly value (e.g. Hobson and Fahlen 2009). The extent to which resources and options for WLB are available is indeed context-specific: comparative studies suggest that there is substantial cross-country variation in the influence of the work–life interface on subjective well-being (e.g. Drobnic and Guillen 2011; Drobnic et al. 2010). Nevertheless, there is still a paucity of research investigating country-level variables as predictors or moderators of work–life issues at individual level (for a review see, e.g. Annor 2016; Shockley et al. 2017a). Accordingly, we also aim to examine, in a multilevel framework, whether and to what extent national-level variables interact with the relationship under investigation. More in detail, specific attention is paid to the moderating effect of the national gender context (e.g. Hagqvist et al. 2017; Powell et al. 2009).

The institutional and cultural components of the gender context are closely and mutually entwined. On the one hand, national and local institution, by means of policies, provide a solid framework for organizing social relations, allocating responsibilities and resources, forming attitudes and preferences, and constructing personal identities (Crompton 2006; Lewis 2009). On the other hand, prevailing gender culture is likely to have an impact, which is mediated by financial and political forces (e.g. Esping-Andersen 1990), on the main features of the welfare state as well as on work–life arrangements (e.g. Pfau-Effinger 1998). As a result, cultural and institutional frameworks may generate different opportunities or constraints for men and women (e.g. Connell 1985; Scott et al. 2010) to achieve the preferred level of WLB. In effect, a substantial body of research has documented the weight of institutional and cultural variables (e.g. Abendroth and Den Dulk 2011; Lyness and Judiesch 2014; Van der Lippe et al. 2006) on the individual experience of the work–life interface and its outcomes (Hagqvist et al. 2017; for a review see, e.g. Greenhaus and Powell 2017; Ollier-Malaterre and Foucreault 2017; Shockley et al. 2017a).

However, individual preferences, aspirations and choices about WLB are situated in everyday life and interaction, where material constraints may add up to cultural and institutional factors and possibly inhibit individual capabilities and agency (e.g. Connell 2002; Crompton 2006). Should opportunities emerge to question existing gendered arrangements, men and women may explore viable alternatives, depending on the degree of access and control they have to relevant resources such as income and power. Nonetheless, the material aspects of the gender context are seldom accounted for in the work–life field of research (Emslie and Hunt 2009; for a literature review see, e.g. Ollier-Malaterre and Foucreault 2017). Hence, in order to address this shortcoming, we employ the Gender Inequality Index developed by the United Nations Development Programme. It estimates gender disparities in health, empowerment and economic status. It can be understood as a loss in potential or achievements in key aspects of human development. Therefore, the GII goes beyond existing measures of the gender context and contributes to better depicting how  gender inequality may enhance or restrain the freedom that individuals have to lead the kind of life they desire to live. Said differently, the GII may better situate individual agency in a truly multidimensional context encompassing the multiple structural barriers that men and women may encounter (Hobson 2011; Lyness and Kropf 2005; Lyness and Judiesch 2014).

We believe that this article may advance knowledge in the following ways (see, e.g. Bianchi and Milkie 2010; Ollier-Malaterre et al. 2013; Rajadhyaksha et al. 2015; Shockley et al. 2017a). First, although academics and practitioners agree that the “fit perspective” has become pivotal (Greenhaus and Allen 2011; Voydanoff 2005), only a few studies have already provided empirical evidence on the impact of the perceived fit between work and non-work commitments on individual outcomes (e.g. Kreiner et al. 2009; Gropel and Kuhl 2009); most literature is still grounded on the “conflict perspective” (e.g. Hagqvist et al. 2017; Steiber 2009). Second, this paper addresses the recent recommendations to frame research within the national context (Ollier-Malaterre and Foucreault 2017; Ollier-Malaterre et al. 2013) and adopt a multilevel perspective when examining the work–life interface (Greenhaus and Powell 2017). In this respect, while cross-national research has been largely limited to a small number of countries (e.g. Beham et al. 2017; Gronlund and Oun 2010) or clusters of countries (e.g. Lunau et al. 2014; Notten et al. 2017), we address the call for a more diversified sample (Williams et al. 2016) and investigate employees in 31 European countries, some of which have been rarely considered. Third, this study deepens our understanding of subjective well-being by using a single overarching index. Work–life research has predominantly concentrated on few facets/dimensions or single-item measures of well-being, such as life satisfaction and mental health (Haar et al. 2014), quality of life (Greenhaus et al. 2003), and self-assessed health (Lunau et al. 2014). Nonetheless, subjective well-being is a complex and multidimensional concept (e.g. Diener 2009; OECD 2013; Stiglitz et al. 2009). Accordingly, the aggregation of a set of items into a synthetic and multifaceted construct that encompasses life evaluations, good mental functioning and measures of affect may offer a better knowledge of the experiences of well-being (for a review see, e.g. Fisher 2014): indeed, such a composite measure, which is easy to interpret and communicate, reflects what really matters for people, beyond the mere material aspects of life (e.g. Stiglitz et al. 2009). Furthermore, there is a broad consensus among scholars that constructing aggregate metrics reduces measurement errors (OECD 2013: 190–194). Hence, the clarity and statistical quality of measures of subjective well-being may both benefit from this study. Finally, gender differences in the outcomes of the work–life interference, an issue that deserves more attention (Ollier-Malaterre and Foucreault 2017; Shockley et al. 2017b), are investigated.

2 Theoretical Background

2.1 Subjective Well-Being and Perceived Work–Life Fit

Prior research has produced mixed results regarding the relationship between working hours and subjective well-being (Angrave and Charlwood 2015). Drawing on the job demands-resources model (Bakker and Demerouti 2007), some studies found that working long hours has a negative impact on subjective well-being (Burke and Cooper 2008), as work demands increase whereas available resources (such as time) reduce. Accordingly, there is abundant evidence that long working days bring about negative work-to-family spill-overs (White et al. 2003) and work–family conflict (Cousins and Tang 2004; Sturges and Guest 2004) and negatively affect work–life balance (Valcour 2007). However, other studies have found no significant association between total number of hours actually worked and subjective well-being (Angrave and Charlwood 2015). In response to these mixed results, scholars contend that the degree of control over work hours (Ford and Collinson 2011; Kossek et al. 2012), the flexibility and predictability of the working schedule (Bourbonnais et al. 2006; Eby et al. 2005), and the fit between preferred and actual working hours (Angrave and Charlwood 2015) are more relevant predictors of employee well-being.

The fit and balance perspective is grounded on the person-environment fit theory. It suggests that people are more stressed and report poorer well-being when they perceive a misfit between their preferences and their social environment (Kristof-Brown et al. 2005; Ten Brummelhuis and Bakker 2012; Voydanoff 2005). Drawing on this theory, Voydanoff coined the term “work–family fit” and defined it as “a form of interrole congruence in which the resources associated with one role are sufficient to meet the demands of another role such that participation in the second role can be effective” (Voydanoff 2005, p. 825). In this paper, we use the term work–life fit, rather than work–family fit, as long as we intend to grasp a wider understanding of the non-work domain. As a matter of fact, employees may engage, be interested, find satisfaction, realise their potential and flourish in other domains beyond just work and family, such as leisure, sport, community, social life, and politics. Available literature suggests that when work demands fit individual preferences regarding the work–life interface, employees tend to score more positive outcomes (Kreiner et al. 2009). Conversely, when employees feel a misfit between the demands associated with their multiple roles and their personal preferences, they tend to report lower levels of well-being (Angrave and Charlwood 2015). That may hold true for both men and women. In fact, a mounting body of research has found that both men and women are exposed to work and family demands and aspire to be meaningfully engaged in multiple roles (e.g. Lyness and Kropf 2005; Lyness and Judiesch 2014; Steiber 2009; Van der Lippe et al. 2006). Accordingly, we hypothesize the following:

Hypothesis 1

perceived work–life fit is positively associated with subjective well-being for both male and female employees.

2.2 Gender Differences in the Effect of Perceived Work–Life Fit

Research on gender differences with regard to the intersection of work and non-work domains and its relationships with well-being or satisfaction with life and/or specific life domains has produced mixed results. Even if some papers have not found gender differences, several studies have shown that the work–life interface is more salient for women; however, some recent research has found that men are believed to experience similar or even stronger outcomes, in terms of well-being and life satisfaction, compared to women (for a literature review see, e.g. Beham et al. 2017; Hagqvist et al. 2017; Rajadhyaksha et al. 2015; Shockley et al. 2017a, b). Based on this overall inconsistency in findings, we will also investigate whether or not there are gender-related effects of perceived work–life fit on subjective well-being.

2.3 The Moderating Role of the Gender Context

Based on the capability approach (Nussbaum 2000; Sen 1992), Hobson (2011, 2014) argues that there is an increasing divide, termed agency gap, between preferences and actual choices concerning WLB. The chances to exercise genuine choices are framed by the gender context—encompassing institutional, cultural and material aspects (Connell 2002; Crompton 2006)—which bounds preferences and options about the work–life interface (Beham et al. 2017; Hagqvist et al. 2017). More specifically, due to the influence of the gender context, individuals may lack the freedom to choose the course of action that could lead them to attain a gainful engagement in both work and non-work domains and fulfil their desires, needs and expectations (e.g. Back-Wiklund et al. 2011; Greenhaus and Powell 2017; Hobson and Fahlen 2009; Lyness and Kropf 2005; Powell et al. 2009; Smithson and Stokoe 2005). Consistent with this line of reasoning, we postulate that the positive effects of perceived work–life fit on subjective well-being at the individual level are moderated by the gender context, which acts as a societal constraint.

Following the call of Powell et al. (2009), most prior research has investigated the influence of the gender context on WLB and its outcomes by using national-level variables or constructs measuring social, cultural or institutional characteristics (for a literature review see, e.g. Greenhaus and Powell 2017; Ollier-Malaterre and Foucreault 2017; Shockley et al. 2017a). However, the use of a culture-sensitive or institutional approach, while authoritative and fruitful (Rajadhyaksha et al. 2015), has not allowed capturing the full array of factors involved in the reproduction of gender inequality (Beham et al. 2017; Scott et al. 2010). There are numerous features of the gender context that frame the opportunities of men and women for flourishing in work and non-work domains. Some of these, still overlooked in current research, concern the material aspects of life and, namely, the gendered structure of the labour market or the nature of decision-making (Beham et al. 2017; Emslie and Hunt 2009). Hence, in order to gain a deeper understanding of the mechanisms affecting work–life trajectories of both men and women, several scholars have urged to pay more attention to multifaceted and objective measures of gender inequality, which could better assess the different access to and control over resources such as health, work, standard of living, power (Hobson 2011; Lyness and Kropf 2005). Coherently, we believe that the gender context may be better depicted and operationalised by the Gender Inequality Index (GII), developed by the United Nations (UNDP 2013). It measures the loss in potential human development due to gender disparities in health, labour market participation and uneven engagement in decision-making processes. It contributes to uncover how the work–life interface is embedded and constructed not only in a given cultural and/or institutional context but also in a specific material context, which affects the capabilities to claim for WLB.

Against this background, we contend that the positive effects of perceived work–life fit on subjective well-being are stronger when the GII is comparatively lower. As for gender differences, literature suggests that it is not just women who are constrained and disadvantaged by gender inequalities; men can be limited in the pursuit of their ideal work–life balance by the dominant gender order (Connell 2002). In effect, in contexts where male breadwinning is highly valued, men may find it extremely difficult to deviate from gender expectations concerning the work–life interface (Allen and Eby 2016; Wall and O’Brien 2017). Accordingly, we contend that in presence of lower societal gender constraints, both men and women are encouraged and supported to accomplish their intimate aspirations and goals as regards the intersection of work, family and community lives, with possible positive repercussions on their subjective well-being (Hobson and Fahlen 2009). In this regard, conservation of resources theory (Hobfoll 1989) indicates that gaining valuable resources in personal and/or professional domains may have relevant generative effects for both men and women, produce a virtuous cycle, a so-called resource gain spiral (Halbesleben et al. 2014), and improve overall individual functioning (ten Brummelhuis and Bakker 2012). Therefore, we predict the following:

Hypothesis 2

The relationship between subjective well-being and work–life fit is moderated by GII such that higher levels of GII result in a weaker relationship for both male and female employees.

GII has rarely been used as a moderator variable in work–life literature exploring the individual experience of the work–life interface. To the best of our knowledge, it has just been employed to inspect gender differences in work–family enrichment (e.g. Beham et al. 2017). Other available articles that  investigated the moderating effect of GII (e.g. Lyness and Kropf 2005; Lyness and Judiesch 2014) have not addressed the nature and extent of gender differences, which we intend to test.

3 Data, Measures and Methods

3.1 Sample

The data used in this study are taken from the European Quality of Life Survey (EQLS) 2012 (Eurofound 2014), which was conducted between September 2011 and August 2012 on a random sample of residents (aged 18 or older) of 34 European countries (27 EU member states and 7 non-EU countries at the time of the survey). To fit the scope of this study, the sample (N = 43,636) was restricted to those who currently had a paid job (N = 19,651) and eventually limited to 31 countries (N = 18,361): residents of Serbia, Montenegro, and Kosovo were excluded from the analyses due to the lack of comparable country-level variables. Observations with missing values for any relevant variables were omitted from the estimation sample, too. The resulting sample (N = 15,835) is almost equally divided between women (51.3%) and men (48.7%). As for countries, sample sizes range from N = 259 in Bulgaria to N = 1083 in Germany. There is quite a good mix of employees in different life-course stages: 28.9% are in the 18–34 age group, 42.5% are aged 35–49, and 28.6% are aged 50 or over. In terms of household structure, 34.6% of respondents are married or co-habiting and have dependent children, 22.8% are married or co-habiting but do not have dependent children, 14.9% live alone, 4.6% are single parents with dependent children, and 23.1% belong to other household types. In terms of occupational classification (Isco-08 major groups), 26.0% of respondents are in managerial or professional positions, 11.2% are technicians or junior professionals, 13.9% are clerical support workers, 23.4% are service or sales workers, 2.2% are skilled agricultural workers, 11.4% are craft and related trades workers, 3.5% are plant and machine operators, 7.9% are in elementary occupations, and 0.6% are in the armed forces.

3.2 Measures

3.2.1 Dependent Variable: Subjective Well-Being

The EQLS provides information on several measures of subjective well-being in its main aspects—that is, hedonic, eudemonic and evaluative well-being. Much of the current literature suggests that subjective well-being is best conceptualized as a multifaceted construct (e.g. Diener 2006; OECD 2013). Accordingly, we computed a composite measure of well-being encompassing the items listed in Table 1. The first six items provide an assessment of evaluative well-being, in terms of domain (job, family life, health and social life) and life satisfactions and happiness. Items 7 through 11 make up the World Health Organization’s Mental Well-Being Index (WHO-5) and measure a mix of eudaimonic and hedonic well-being. The three remaining items provide an assessment of both positive and negative short-term feelings (i.e. hedonic well-being). Some items were recoded so that higher scores indicated a higher degree of subjective well-being. We conducted a factor analysis for data reduction purposes (see Tables 4, 5 in “Appendix”) and retained one factor. In this solution, the 14 items measuring subjective well-being are highly correlated (Cronbach’s α = 0.866), and the first factor accounts for most of the variance (84.1%). Factor loadings are displayed in Table 1.

Table 1 Items measuring subjective well-being
Full size table

3.2.2 Independent Variable: Perceived Work–Life Fit

The following question was used in the EQLS to assess perceived work–life fit: “In general, do your working hours fit in with your family or social commitments outside work?”. Response categories were dummy-recoded as follows: 0 = not very well/not at all well and 1 = very well/fairly well.

3.2.3 Country-Level Independent Variable: Gender Inequality

As previously stated, this study examines the moderating effect of gender inequality on the relationship between subjective well-being and perceived work–life fit. As a measure of gender inequality, we used the Gender Inequality Index (GII), which is calculated by the United Nations. GII captures the gender gap in human development (i.e. on a combined loss in achievements) in three domains: reproductive health (measured by maternal mortality ratio and adolescent birth rate), empowerment (measured by the female and male percentages of the population aged 25 and older with at least a secondary education and by the female and male shares of parliamentary seats), and labour market (labour force participation rate of female and male individuals aged 15 and older). The GII ranges between 0 (perfect gender equality) and 1 (perfect gender equality). The scores of the 2011 index (UNDP 2011), which were used in this article (Table 6 in “Appendix”), indicate that the lowest levels of gender inequality are found in Sweden (GII = 0.049), the Netherlands (GII = 0.052), and Denmark (GII = 0.060). Conversely, the European countries that suffer the highest loss in potential human development due to gender inequality are Malta (GII = 0.272), Romania (GII = 0.333), and Turkey (GII = 0.443).

3.2.4 Covariates

Building on prior research (e.g. Allen and Eby 2016; Chen and Cooper 2014; Landry and Cooper 2014), we included a wide set of control variables because of their potential relationships with both subjective well-being and perceived work–life fit. Age was coded in the following categories: 1 = 18–24 years; 2 = 25–34 years; 3 = 35–49 years; 4 = 50 years and older. Household structure was measured as follows: 1 = single; 2 = couple; 3 = single parent; 4 = couple with children; 5 = other household types. Employment contract was measured with four categories: 1 = unlimited permanent contract; 2 = fixed-term contract; 3 = other types (including temporary employment agency contract and apprenticeship or other training scheme); 4 = no written contract. Occupation was measured using the International Standard Classification of Occupations (ISCO-08), with ten categories. Education was measured with three categories: 1 = primary or less; 2 = secondary; 3 = tertiary. Weekly work hours were recoded into five categories: 1 =less than 30 h; 2 = 30–39 h; 3 = 40 h; 4 = 41–49 h; 5 = 50 h and more. Entitlement to flexible work hours was coded as a dummy variable (0 = no; 1 = yes). Perceived job insecurity was measured through the following item: “How likely or unlikely do you think is it that you might lose your job in the next 6 months?” and coded into three categories (1 = quite or very unlikely; 2 = neither likely nor unlikely; 3 = very or quite likely). Respondents were also asked to evaluate their financial situation in comparison to most people in their country. That was coded in the following categories: 1 = much/somewhat worse; 2 = neither worse nor better; 3 = somewhat/much better. Area of residence was dummy coded (0 = countryside or village and 1 = town or city). In order to test whether and to what extent the sign and magnitude of the estimated parameter of GII (i.e. the national-level main explanatory variable) behave after including other national-level variables, we added Gross Domestic Product (GDP) per capita in the models. Previous research (e.g. Drobnic et al. 2010; Steiber 2009) proves that this country-level measure of economic development may affect the relationship between the work–life interface, subjective well-being and the quality of life. In particular, we used World Bank GDP per capita data based on purchasing power parity (PPP). Data, which are in current international dollars based on the 2011 ICP round, are reported by country in Table 6 in “Appendix”. Briefly, GDP per capita ranges from 11,611.85 (in Macedonia, FYR) to 92,005.02 (in Luxembourg), with an average value of 33,989.96 and a standard deviation of 13,852.02.

3.3 Analytical Strategy

Sampled data have a two-level structure, with employees at level 1 and countries at level 2. To predict well-being by perceived work–life fit, we estimated a multilevel regression model for employees nested within countries. We examined simultaneously the effect of individual (level 1) and country-level (level 2) predictor variables, as well as their interaction, on the outcome variable (subjective well-being). GII and GDP per capita (i.e. the level-2 variables) were centred on their overall mean in order to both remove high correlations and help with the interpretation of parameter estimates (Hox 2010). Consistent with research hypotheses, we fitted separate models for male and female employees. Subsequently, we fitted a combined model that included both male and female employees and tested the statistical significance of gender differences.

4 Findings

4.1 Subjective Well-Being

Cross-tabulation analysis (see Table 7 in “Appendix”) reveals that female employees report comparatively lower levels of well-being than their male counterparts. For both male and female samples, average subjective well-being scores are comparatively higher for employees who report good work–life fit, have a tertiary education, have a better financial situation than most people in their country and are entitled to flexible working hours. Conversely, male and female employees with higher levels of perceived job insecurity report lower levels of subjective well-being. In addition, GII and subjective well-being are significantly correlated with each other in the expected direction for both men (r = − 0.141) and women (r = − 0.147) (see Tables 8, 9 in “Appendix”).

4.2 Impact of Perceived Work–Life Fit on Subjective Well-Being

In Model 1, which has no predictor variables (intercept-only model), we considered the variance components and assessed the amount of variability in subjective well-being due to each level of analysis. Parameter estimates reported in Tables 2 and 3 show that the intraclass correlation coefficient is 0.0572 for men and 0.0645 for women, which means that, respectively, 5.72 and 6.45% of the variance of well-being scores is accounted for at the country level. Therefore, even if subjective well-being appears to be much more an individual issue, with low variance explained at the country level, multilevel regression analysis seems appropriate (Kahn 2011).

Table 2 Results of multilevel modelling analysis (males)
Full size table
Table 3 Results of multilevel modelling analysis (Females)
Full size table

Model 2 (i.e. random intercept and random slope model) includes perceived work–life fit as an explanatory variable. Parameter estimates for the main predictor are positive and significant. Accordingly, on average, well-being scores increase as perceived work–life fit increases. The difference in well-being scores for those who report a good work–life fit (i.e. very well/fairly well) versus those whose work–life fit is poor (i.e. not very well/not at all well) is 0.456 points (p < 0.001) for men, with a confidence interval of 0.410–0.501, and 0.467 points (p < 0.001) for women, with a confidence interval of 0.419–0.516.

Model 3 includes potential confounding variables measured at individual level that may correlate with both the main predictor and the outcome variable. As expected, the variance components have considerably decreased for both male and female employees. Most of the confounders considered in the analysis have a significant effect in the expected direction on the outcome variable in both samples: subjective well-being is higher for younger, more educated, and better-off men and women, as well as for male and female employees with more secure jobs and respondents living with a partner. Parameter estimates for work–life fit have slightly diminished: from 0.456 to 0.392 (p < 0.001) for male employees and from 0.467 to 0.411 (p < 0.001) for female employees. After controlling for a broad set of confounders, we can assume that Hypothesis 1 is supported: employees whose perceive that their working hours fit very well or fairly well with their family or social commitments report significantly higher subjective well-being compared to those who perceive comparatively lower levels of work–life fit.

In Model 4, GII (i.e. the main level-2 explanatory variable) was added. Results indicate that for every one-unit increase in mean-centred GII scores, the well-being scores of male respondents decrease by 1.250 points (p < 0.001) whereas female well-being scores reduce by 1.419 points (p < 0.001). Comparison with Model 3 indicates that GII contributes to explain variation in intercepts. Based on the likelihood ratio test (χ2 = 10.38; p < 0.01 for the male sample and χ2 = 12.60; p < 0.001 for the female sample), we reject the null hypothesis that there is no significant difference between Model 4 and Model 3 and conclude that the addition of GII enables a significant increase in the explanatory power of our models.

Subsequently we expanded our models and included GDP per capita (Model 5), a country-level variable that has been found to affect subjective well-being. Contrary to expectations, parameter estimates indicate that GDP per capita has no significant effect on male or female subjective well-being scores. We can see that the slope for GII does not change that much—i.e. it reduced to − 1.205 (p < 0.01) for male employees and to − 1.403 (p < 0.01) for female employees—as we added GDP per capita in the model.

In Model 6, we estimated the cross-level interaction effect in order to test Hypothesis 2, i.e. whether GII moderates the relationship between perceived work–life fit and subjective well-being. Model 6 fits the data significantly better than Model 5, as proven by the likelihood ratio test (χ2 = 8.06; p < 0.01 for male employees and χ2 = 4.17; p < 0.05 for female employees). The interaction term is negative and significant for both men (− 0.751; p < 0.01) and women (− 0.729; p < 0.05).

In the process of interpreting the results (Williams 2012), we computed, using the “margins” command in Stata 13, the adjusted means of subjective well-being as a function of perceived work–life fit and mean-centred GII scores, ranging from − 0.11 (which equals to GII = 0.049, i.e. the GII scores reported in Sweden) to 0.28 (which equals to GII = 0.443, i.e. the GII scores reported in Turkey). Figure 1 provides a graphical display of the substantive meaning of the moderating effect of GII on the relationship under investigation, given all confounding variables, with separate lines to represent different levels of perceived work–life fit. More in detail, it shows the work–life fit slope for GII scores ranging from − 11 (i.e. 11 percentage points below the average) to 28 (i.e. 28 percentage points above the average) in three percentage point increments; this is the range of values for GII in selected countries. Hypothesis 2 is supported: the slope of the relationship between perceived work–life fit and subjective well-being decreases as a function of GII. Said differently, higher levels of societal gender inequality result in a weaker relationship between perceived work–life fit and subjective well-being for both men and women. In particular, in a country with average GII and GDP per capita scores:

Fig. 1
figure 1

Subjective well-being by perceived work–life fit and Gender Inequality Index scores (95% confidence intervals in grey) for males and females

Full size image

  • For average male employees, subjective well-being is expected to increase by 0.178 points when working hours fit very well/fairly well in with family or social commitments outside work. When the GII is three percentage points below the average, subjective well-being is expected to increase by 0.206; conversely, when the GII is three percentage points higher than the average, subjective  well-being is expected to increase by 0.150;

  • For average female employees, subjective well-being is predicted to growth by 0.052 points when working hours fit very well/fairly well in with family or social commitments outside work. When the GII is three percentage points below the average, subjective well-being is expected to increase by 0.088; however, when the GII is three percentage points higher than the average, subjective well-being is expected to rise by 0.015.

Next, we fitted a common model for males and females (Table 10) so as to test whether: (a) the difference in slope comparing those with a good work–life fit and those with a poor work–life fit is significantly different for males versus females; (b) the moderating effect of GII on the relationship between perceived work–life fit and subjective well-being is significantly different for males versus females. We used a separate intercept and separate slopes coding system in Stata 13 and compared the coefficients using the “contrast” command. Results indicate that: (a) the comparison of the perceived work–life fit slope for females versus males is not significant (F = 1.31, p = 0.252); the moderating effect of GII on the relationship under investigation is not significantly different for females than for males (F = 0.34, p = 0.562).

5 Discussion and Conclusion

This article has examined, at the individual level, whether and to what extent the perceived fit between working hours and family or social commitments outside work influenced subjective well-being in a large and cross-national sample of European employees. As previously discussed, well-being is a multidimensional concept, which encompasses hedonic, eudaimonic and evaluative components (e.g. Diener 2009; OECD 2013; Stiglitz et al. 2009). Thus, in this study it has been operationalized through a multiple-item construct, measuring flourishing and mental health, experience of positive and negative affect, and cognitive evaluations of satisfaction with life and specific life domains (e.g. Diener 2009). It is worth mentioning that such a composite indicator is a self-reported measure of well-being. According to the capability approach, research that relies on self-assessments may be problematic. Social norms and expectations and gendered institutions may influence individual perceptions and evaluations, thus leading to biased results (Sen 2000). Consequently, other composite and objective indicators of well-being, which would not reflect individual assessments potentially conditioned by societal or material constraints, could be employed in future research.

Despite this potential limit, the analyses have provided robust empirical evidence that, after controlling for a wide set of confounding variables, better perceived work–life fit is associated with greater subjective well-being. This is a novel and important finding. In fact, even though work–family scholars agree on the importance of considering the fit between personal preferences and environmental demands in order to better understand work–life dynamics (Greenhaus and Allen 2011), very few studies have demonstrated the relevance of this approach (see, e.g. Kreiner et al. 2009). Hence, we have contributed to fill a gap by proving that perceived work–life fit is a significant predictor of subjective well-being. This holds true for both men and women and tests for gender differences returned no significant results. Nonetheless, future research could benefit from the availability of longitudinal datasets. Cross-section data, which have been employed in this paper, have several limits indeed. As is well known, they do not allow establishing a true cause-effect relationship. Besides, important influences on attitudes, perceptions and behaviour may not be observed in the dataset: this may be the case of personality traits, whose omission is likely to significantly inflate the coefficients and may lead the regression models to suffer from endogeneity bias.

Furthermore, this manuscript has tested, in a multilevel framework, the moderating effect of the Gender Inequality Index, which identifies gender inequalities in capabilities, livelihoods, and agency (United Nations Development Programme 2013), on the relationship between perceived work–life fit and subjective well-being. Most of the extant literature has used socio-cultural items and constructs, such as gender egalitarianism (e.g. Steiber 2009; Lyness and Judiesch 2014), or institutional variables, referring to welfare regimes or policy models (e.g. Lunau et al. 2014; Chung 2011), to investigate WLB cross-nationally. Conversely, a few scholars have tested the influence of structural factors (notable exceptions are, e.g. Beham et al. 2017; Hagqvist et al. 2017; Lyness and Judiesch 2014). Thus, several calls have been issued (e.g. Lyness and Kropf 2005; Ollier-Malaterre et al. 2013; Shockley et al. 2017a) to broaden the analytical framework and include material aspects of the gender context in work–life research.

Building on the capability and agency framework (Hobson 2011, 2014; Sen 1992), we contended that the relationship between perceived work–life fit and subjective well-being could be moderated by societal gender inequality, such that high levels of gender inequality would result in a weaker relationship for both men and women. Findings, which further emphasize the importance of multi-level research in the work–life field (Greenhaus and Powell 2017; Ollier-Malaterre et al. 2013), reveal that variations in a multi-layered national setting including economic, educational, health and power factors can make perceived work–life fit more or less salient for subjective well-being, for both male and female employees. Once again, no significant gender differences were found. A plausible explanation for these results is that when gender inequality is acute the individual experience of the work–life interface may not be a key determinant of subjective well-being. If women have limited access to key functionings, have few chances to provide for themselves or their families, and have little or no voice in decision-making processes, work–life fit may not be an important life aspiration, a crucial element of well-being. Indeed, if there are major barriers to female human development, women may first aspire to more equality in terms of essential functionings, such as education and health, as well as in the ability to exercise their preferences and shape their life course through economic and political participation. However, where there are systematic gender inequalities in terms of opportunities and outcomes in multiple domains, men may be much more interested in aligning, or pushed by prevailing social norms to align (Ridgeway 2011), with the male-breadwinner prescription and prioritize work over the family in order to maintain a prominent place in society. Hence, control over material resources in work and life may matter more for their subjective well-being than the possibility of effectively combining work with engagement in other life domains. Conversely, in more egalitarian contexts, work–life fit is a critical factor shaping subjective well-being, since both men and women may have greater aspirations and expectations to be both earners and carers and, more in general, participate simultaneously in paid employment, family and community life (Hobson and Fahlen 2009). Based on these outcomes, we invite scholars not to focus just on sociocultural norms and constructs at the country level, but rather to include material variables in their analyses in order to gain new insights on the predictors and moderators of the work–life interface.