Explaining job satisfaction through the use of work–familybenefits and their impact on the employee’s family context

Abstract

Purpose

This study analyses the relationship between the use of work–family benefits and job satisfaction (JS). Furthermore, it proposes that work-to-family conflict (WFC) and work-to-family enrichment (WFE) play a mediating role in this relationship. The purpose of this paper is to address these issues.

Design/methodology/approach

Data are gathered from 1,051 employees of Colombian organisations. Partial least squares path modelling is used.

Findings

The results show that the perception of WFE to a greater extent and the WFC perception, to a lesser extent, are significant mediators in the relationship between the use of benefits and JS.

Practical implications

This study justifies investments and initiatives on the adoption and promotion of work–family benefits. Moreover, it provides practical clues on how to boost JS: WFC and WFE are variables to be considered.

Originality/value

This study proposes a multiple mediation model to analyse the relationship between the actual use of work–family benefits and JS from a family perspective. It contributes to the literature in examining antecedents of JS, highlighting the role of WFE.

Citation

Gutiérrez Vargas, L.M., Alegre, J. and Pasamar, S. (2023), "Explaining job satisfaction through the use of work–familybenefits and their impact on the employee’s family context", International Journal of Manpower, Vol. 44 No. 9, pp. 20-36. https://doi.org/10.1108/IJM-01-2022-0055

1. Introduction

Today, work–life balance (WLB) is a major concern in many countries (Powell et al., 2019). Several demographic changes have drawn society’s attention to this phenomenon, resulting in increased academic and managerial interest (Alegre and Pasamar, 2018; Cegarra-Leiva et al., 2012; Poelmans et al., 2003) and justifying the need for further research (Moore et al., 2007). Previous studies have raised awareness of context and its vital role in empirical research. Contextual differences are significant in the behavioural sciences (Davison and Martinsons, 2016), and it has been stated how political, legal and cultural changes have an effect on the WLB institutionalisation (Pasamar and Valle, 2011). Moreover, institutions such as the International Labour Organisation (ILO) or the 2030 Agenda for Sustainable Development have been encouraging strategies for reconciling work, family and personal life in South America. These strategies imply the design and the adoption of work–family benefits (Osterman, 1995; Scheibl and Dex, 1998). The flexibility offered by an organisation with this type of benefit allows employees to enjoy a better family life (Greenhaus and Powell, 2006). This could create competitive advantages through attracting and retaining the best employees (Allen, 2001; Carlson et al., 2010; Poelmans et al., 2003).

Even though work–family benefits are generally associated with positive organisational outcomes, it is not so clear how this occurs (Konrad and Mangel, 2000; Talukder et al., 2018). More precisely, Baltes et al. (1999) suggest including work-to-family conflict (WFC) and work-to-family enrichment (WFE) when examining the positive consequences of work–family benefits.

Work–family benefits (Baral and Bhargava, 2010) are resources that help employees fulfil both their work and their family responsibilities (Voydanoff, 2004). They minimise WFC, promote WFE and improve employees’ functioning and performance both at work and at home (Carlson et al., 2010). However, further analysis of formal work–family benefits and their connection with both WFC and WFE as well as their influence on job satisfaction (JS) is needed (Baral and Bhargava, 2010; Quade et al., 2021).

Drawing on the study by Carlson et al. (2010), Baral and Bhargava’s (2010) and Martinez-Sanchez et al. (2018) this study aims to analyse the joint effect of different work–family benefits that are available to workers in an insufficiently researched context such as Colombia. The universalist approach to research do not consider cultural, institutional or other environmental differences (Davison and Martinsons, 2016). Since much of the previous literature on WLB has focussed on the Anglo-Saxon perspective and eras of economic prosperity (Chang et al., 2010), there is a call for studies which pay more attention to other contexts (Pasamar and Valle, 2011; Peters and Heusinkveld, 2010).

The work–family benefits are seen as contextual characteristics that provide resources for both enrichment (Lapierre et al., 2018) and conflict (Michel et al., 2011). Hence, the purpose of this study is to develop a better understanding of the mechanisms linking the use of work–family benefits to both work–family interface (conflict and enrichment) and an important work outcome such as JS in Colombia. It is therefore argued that employees who use work–family benefits will experience greater JS than those who do not. It is further argued that this relationship is mediated by WFC and WFE, hence shedding light on the mediating role of the work–family interface.

2. Theoretical framework and hypotheses

2.1 An integrative model

Voydanoff (2002) proposed several linking mechanisms in the processes through which work and family characteristics are linked to the individual, family and work outcomes; two of these mechanisms are WFC and work–family facilitation, also known as WFE. A WFC is a form of role conflict in which the role pressures of the work and family domains are mutually incompatible in some respect (Greenhaus and Beutell, 1985). Nevertheless, WFE has a more positive impact as experiences in one role improve the quality of life in the other role (Greenhaus and Powell, 2006). For this reason, WFC and WFE are conceptually and empirically distinct (Carlson et al., 2010).

The model advanced by Voydanoff (2002, 2004, 2005) states that work–family outcomes are mainly a function of the work–family demands and resources available for people to solve those demands, introducing the concept of boundary-spanning resources. Accordingly, Voydanoff (2004) defines demands as “structural or psychological claims associated with role requirements, expectations and norms to which individuals must respond or adapt by exerting physical or mental effort” while resources are “structural or psychological assets that may be used to facilitate performance, reduce demands, or generate additional resources.” Resources can refer to how work and family are interconnected, meeting the demands that emerge in another life domain. Within the boundary-spanning resources in the work domain, Voydanoff (2005) includes a series of work and family supports such as flexible schedule, dependent care, part-time work or other work–family benefits (Lapierre et al., 2018).

In order to complement the models mentioned above, this paper specifically looks at how WFC and WFE can affect the relationship between work–family benefits and JS. From the perspective of positive emotions (Fredrickson, 1998, 2001), it is posited that certain positive emotions, such as joy and contentment, share the ability to broaden people’s momentary thought and action repertoires and build their personal resources. Consequently, positive emotions stimulate outward-oriented thoughts and actions that contribute to deepening social relationships and acquiring additional skill sets. By contrast, negative emotions prompt a momentary narrowing in the way people think and act on the world, specifically enacting an inwards or self-focussed survival response (Carlson et al., 2010). This would suggest that domain-specific positive outcomes would be inhibited by accumulated WFC and enhanced by accumulated WFE.

On the other hand, previous research suggests that both WFE and WFC orientations should be examined separately (Carlson et al., 2006; Greenhaus and Powell, 2006; Wayne et al., 2007) in a particular work domain (Premchandran and Priyadarshi, 2020).

2.2 Work–family benefits

In order to manage work, family and life demands, organisations offer employees flexible working conditions through a series of flexible benefits, policies and work arrangements. Flexibility is thus broadly seen in terms of formal flexibility policies set by human resources (Pasamar, 2020) or informal arrangements regarding flexibility in the organisational setting (Carlson et al., 2010).

Prior studies support the notion that the availability and use of family-friendly policies are associated with various positive individual and organisational outcomes. These outcomes include lower levels of WFC (Hornung et al., 2008; Michel et al., 2011), higher levels of WFE (Lapierre et al., 2018), satisfaction with WFE (Ezra and Deckman, 1996), better physical and mental well-being, greater organisational and affective commitment, greater satisfaction with life and work and eventually higher performance and productivity (Allen, 2001; Van Dyne et al., 2007; Hughes and Bozionelos, 2007; Kossek et al., 2006). Although it is generally assumed employees who individually obtain benefits when these policies are available (Allen, 2001), the mere provision of benefits does not guarantee their use by workers (Pasamar and Alegre, 2015), and the advantages may end up being lost due to lack of use (Beauregard and Henry, 2009). Therefore, family-friendly benefits can be considered formal organisational support measures, and their use is likely to contribute to positive outcomes (Allen, 2001; Wayne et al., 2006).

2.3 Work–family benefits and policies and job satisfaction

Several studies have confirmed the relationship between work–family policies and positive work outcomes (Dalton and Mesch, 1990; Kopelman et al., 2006; Lambert, 2000; Thomas and Ganster, 1995). Job satisfaction is defined as “a pleasurable or positive emotional state resulting from the appraisal of one’s job or job experiences” (Locke, 1969, p. 316). Taking into account that employees’ benefits’ use generates more positive work attitudes (Allen, 2001; Thomas and Ganster, 1995) and based on our integration of the work–family interface models with the broaden-and-build theory of positive emotions, we argue that the use of work–family benefits provides a boundary-spanning resource for individuals in an organisation, so effective outcomes are improved. Therefore, we hypothesise the following:

H1.

The use of work–family benefits is positively related to JS.

2.4 The mediating role of the work–family interface (conflict and enrichment)

The broaden-and-build theory of positive emotions (Fredrickson, 1998, 2001) suggests that contentment, as a discrete positive emotion, has the ability to broaden people’s momentary thought and action repertoires and build their enduring personal resources, i.e. psychological resources. We suggest that both WFC and WFE play a mediating role in the link between the use of work–family benefits and JS. Based on our integration of the work–family interface models with the broaden-and-build theory of positive emotions, it is argued that the use of work–family benefits provides a boundary-spanning resource for individuals in an organisation and one of the ways in which this resource is utilised is via reducing WFC and developing WFE, so effective outcomes are improved, namely, JS (see Figure 1). Therefore, we hypothesise the following:

H2a.

The relationship between the use of work–family benefits and JS is mediated by WFC.

H2b.

The relationship between the use of work–family benefits and JS is mediated by WFE.

3. Methodology

3.1 Sample and procedure

Data were gathered in September 2017 through an on-line self-administered survey to employees belonging to a representative sample of Colombian organisations in terms of industries, size and corporate governance through an intentional sampling method applied on Fundación Carolina’s database on grants’ recipients (7,000 persons). Fundación Carolina is an international foundation which has the mission to encourage and promote cultural relationships and cooperation in educational and scientific issues between Spain and Latin American countries.

As for the respondents’ demographics, 47.1% of the sample was males and 52.9% females, with 82.3% of them ranging between 25 and 50 years of age. At the time of the survey, 65.7% reported to be married or cohabiting with a partner, 42.2% to have dependent children and 46.5% to have dependent adults. In total 60.6% occupy intermediate or management positions, 62.3% are under an indefinite-term employment contract and 42.2% have seniority in the firm from 1 to 5 years. We obtained 1,051 valid responses. Our response rate was 15.01%.

3.2 Measurements and instruments

Measurements were chosen or adapted from previous the literature giving priority to their fit to the concepts’ definitions we use in this study. An additional choice criterion had to do with their relevance in the previous literature: we have chosen measurements coming from highly cited previous studies.

Work–family benefits: Overall, 46 work–family benefits were used. With the purpose of having a comprehensive representation, adapted to the Latin American context, we drew from the general classification of benefits and policies proposed by Chinchilla and León (2007) and Chinchilla et al. (2006), to which others were added (Allen, 2001; Pasamar, 2015). They were grouped into five categories: job flexibility, support, training and development, family services and fringe benefits. For each benefit, the respondents were asked to choose one of four answers: 1) exists and I have used it, 2) exists but I haven’t used it, 3) I don’t know if it exists and 4) It definitely doesn’t exist. To obtain the use of benefits score, all occurrences of answer (1) across the 46 benefits were added, creating the USOB variable, according to the procedure by O’Driscoll et al. (2003).

WFC: Nine items of the WFC scale developed by Carlson et al. (2000) were used, in the direction in which work interferes with family. A sample item is, “My work keeps me from my family activities more than I would like”. A seven-point Likert scale was used, ranging from 1. “Totally in disagreement” to 7. “Totally in agreement”. Cronbach’s alpha for this scale was 0.905.

WFE: Nine items of the WFE scale developed by Carlson et al. (2006) were used, in the direction in which work enriches family. A sample item is, “My involvement in my work helps me to understands [sic.] different viewpoints and this helps me be a better family member”. A seven-point Likert scale was used, ranging from 1. “Totally in disagreement” to 7. “Totally in agreement “. Cronbach’s alpha for this scale was 0.949.

JS: The JS scale proposed by Agho et al. (1992) was used, which corresponds to a reduced Brayfield and Rothe's (1951) scale made up of six items. A sample item is, “I feel quite satisfied with my job.” A seven-point Likert scale was used, ranging from 1. “Totally in disagreement” to 7. “Totally in agreement.” Cronbach’s alpha for this scale was 0.93.

Control variables: Control variables commonly used in the WFC and WFE literature were included such as gender, having a spouse or partner, having dependent children and having dependent adults.

3.3 Data analysis

The study uses the partial least squares (PLS) path modelling, a variance-based structural equations modelling (SEM) technique. The SmartPLS 3 software version 3.2.8 was employed.

4. Results

4.1 Measurement model results

The reflective measurement model in PLS is evaluated following three criteria: reliability, convergent validity and discriminant validity, resulting in a completely satisfactory measurement model (see Tables 1 and 2).

Construct reliability or internal consistency is adequate. It was verified that all reflective constructs presented a Cronbach’s alpha (α) and composite reliability (ρc: Dillon–Goldstein index) greater than 0.7 (Nunnally and Bernstein, 1994). Additionally, Table 1 shows the new construct reliability indicator (ρA: Dijkstra–Henseler index), which was confirmed to be above 0.7 (Dijkstra and Henseler, 2015).

The individual reliability of most of the items is adequate. According to Hair et al. (2017), the external loadings of the indicator must be equal to or greater than 0.707, meaning that the variance shared between the construct and its indicators is greater than the error variance. It is considered that, for initial stages of scale development, and when the scales are applied to different contexts, weak indicators are occasionally retained on the basis of their contribution to content validity. In this situation, the inclusion of weak items would be helpful to create a better scoring of the latent variable. Hence, those items lower than 0.707 and greater than 0.4 were kept (Hair et al., 2011) (See Table 1).

Convergent validity guarantees that a set of indicators represents the same underlying construct. The average variance extracted (AVE) value of at least 0.5 means that construction can explain on average more than half of the variance of its indicators. Table 1 shows that the latent variables reach convergent validity because their AVE measurements exceed 0.5 (Fornell and Larcker, 1981).

Discriminant validity assesses that a construct is different from the others. Table 2 shows that all variables present discriminant validity according to the Fornell–Larcker and heterotrait–monotrait (HTMT) criteria. For satisfactory discriminant validity, the diagonal elements (AVE in bold) must be significantly higher than the elements outside the diagonal in the corresponding rows and columns (correlations) (Fornell and Larcker, 1981). On the other hand, the HTMT correlation ratio evaluates the average of the heterotrait–heteromethod correlations (Henseler et al., 2015). In a well-adjusted model, the heterotrait correlations should be smaller than the monotrait correlations, which implies that the HTMT ratio should be below 1. We confirmed that all HTMT values were below the threshold of 0.85 and 0.9.

4.2 Treatment of common method variance (CMV)

To avoid the problem of common method variance (CMV), we followed the recommendations of Podsakoff et al. (2003), such as hiring a professional interpreter for the inverse translation of the scales, the careful review of the scales, the application of a pre-test or the mixing of items from different scales.

Besides, Harman’s single-factor test (Podsakoff et al., 2003; Podsakoff and Organ, 1986) was used to detect whether CMV had a significant influence on the data and to treat the potential of social desirability of the answers. The test was conducted with SPSS (Statistical Package for the Social Sciences) and EQS (Bentler, 2006). The results of the test with SPSS showed that the single factor explains 42.09% of the total variance. The results of the CFA performed with EQS with all the indicators loading in a single factor (χ²₍₂₅₂₎ = 11,602.34, p = 0.000; d.g. (degrees of freedom) = 252; χ²/g.l. = 46.04; Bentler Bonnet normed fit index (BBNFI) = 0.514; CFI = 0.519; goodness of fit index (GFI) = 0.41; and root mean square error of approximation (RMSEA) = 0.207) showed poor fit, suggesting that the factor does not take into account all the variation in the data. Therefore, these results indicate it is very unlikely that the data present the problem of CMV (Podsakoff et al., 2003).

4.3 Structural model results

For the structural model analysis, we followed Henseler et al. (2009)’s bootstrapping procedure, which generates standard errors and t statistics to evaluate the statistical significance of the path coefficients. In the structural model evaluation (Tables 3 and 4), the criteria to be considered include the assessment of possible collinearity problems through the variance inflation factor (VIF) test, the determination coefficient (R²), the blindfolding-based cross-validated redundancy measure (Q²), as well as the statistical significance and the relevance of the path coefficients (Chin, 1998; Hair et al., 2019).

Initially, the non-existence of multicollinearity between the antecedent variables of each endogenous construct was verified (Table 3), for which VIF values should be close to 3 and lower (Hair et al., 2019). The values and significance levels of the path coefficients, as well as the R² coefficients for each one of the endogenous constructions, are shown in Figure 2. It is observed that all direct effects are significant. The bootstrap percentile in a confidence interval of 95% also has this same result in Table 3, which shows that the path USOB → JOB SATISFACTION (β = 0.14 and t value = 5.47) is significant and positive, and its confidence interval does not contain zero. Hence, these results support Hypothesis H1. On the other hand, the control variables in the hypothesis model are mostly not significant.

The R² value of the latent dependent variable was used to determine the amount of variance accounted for (VAF) by the model. R² measures the VAF in each one of the endogenous constructions and, therefore, is a measure of the explanatory power of the model (Hair et al., 2019). In Table 4, the R²_{JOB SATISFACTION} index of the JS variable indicates that the theoretical model explains 51.4% of the construct’s variance, which can be considered moderate and close to substantial, according to the criteria established by Chin (1998) for R² values (0.19: weak, 0.33: moderate and 0.67: substantial). Therefore, it can be concluded that the model has adequate explanatory power for the JS variable.

Table 3 also shows the effect size (f ²) of the predictor constructs. This assesses the degree to which an exogenous construct contributes in explaining a given endogenous construct in terms of R², where values of 0.02, 0.15 and 0.35 represent small, medium and large effect sizes, respectively (Cohen, 1988). A large effect size of WFE on JS (f ² _{WORK-TO-FAMILY ENRICHMENT → JOB SATISFACTION} = 0.558) is observed, in contrast to the small effect sizes for the other predictors.

Another evaluation criterion of the structural model involves the prediction capacity of the model. The prevailing measure of predictive relevance is the Stone-Geisser Q² statistic (Hair et al., 2017), which can be measured through blindfolding procedures. If this value for a given latent endogenous variable is greater than zero, its explanatory variables provide predictive relevance (Henseler et al., 2009). Q² values above 0, 0.25 and 0.5 represent small, medium and large predictive relevance of the PLS-path model (Hair et al., 2019). As can be seen in Table 4, the values for the Stone-Geisser Q² statistic presented for the three endogenous constructs are above zero. Specifically, JS has the highest value, Q²_{JOB SATISFACTION} = 0.325, for a medium level of predictive relevance. In consequence, it can be concluded that the model has satisfactory predictive relevance for the JS variable.

4.4 Mediating effects analysis of work-to-family conflict and enrichment

We adopted the procedure developed by Nitzl et al. (2016) to test the mediation effect in PLS-SEM and define, at the same time, the type of mediation found. This procedure proposes two steps:

• Step 1: The importance of indirect effects: Table 5 exhibits the total effect (c) of the use of work–family benefits on JS as the sum of the direct effects (c′) and specific indirect effects (a₁ × b₁) and (a₂ × b₂). The estimation of the latter uses the product of the path coefficients for each one of the paths in the mediation chain: c = c′ + a₁ × b₁ + a₂ × b₂ (4.1).

According to Nitzl et al. (2016), the indirect effect (a × b) must be significant to establish a mediation effect. In our case, both specific indirect effects (a₁ × b₁ and a₂ × b₂) are significant. Neither the path of the mediating variable M_{1 USOB → WORK-TO-FAMILY CONFLICT → JOB SATISFACTION} (a₁ × b₁ = 0.031 and t value = 3.931) nor the path of the mediating variable M_{2 USOB → WORK-TO-FAMILY ENRICHMENT → JOB SATISFACTION} (a₂ × b₂ = 0.141 and t value = 7.59) includes zero in their respective 95% confidence intervals, which indicates the existence of significant indirect effects in both. Therefore, H4a and H4b are supported. The existence of mediation through the two mediating variables, WFC and WFE, is thus confirmed.

• Step 2: the type of effect and/or mediation: The significance of the direct effect (c′) must be verified in this step. In case it is not significant; there is total mediation; the opposite would indicate partial mediation (Nitzl et al., 2016). Table 5 shows the point estimate of the direct effect (c′), the specific indirect effects (a₁ × b₁, a₂ × b₂) and the total indirect effect [(a₁ × b₁) + (a₂ × b₂)]. Given the direct effect is significant (H1: c′ = 0.14; and t = 5.7), it provides support to H1. The specific indirect effects and the total indirect effect are significant, so a partial mediation can be claimed for both the mediating variable M₁ (WFC) and the mediating variable M₂ (WFE) towards JS. Moreover, the (VAF) test was conducted which permits to evaluate the magnitude of each mediation: 50.13% of the total effect is due jointly to two mediation effects. Since the VAF is below 80%, this implies an additional argument for partial mediation (Hair et al., 2017; Nitzl et al., 2016). On the other hand, a complementary partial mediation can be observed both for the WFC variable and the WFE variable, since both the direct effect (c′) and the specific indirect effects (a₁ × b₁ and a₂ × b₂) point towards the same direction (Baron and Kenny, 1986; Nitzl et al., 2016), positive, in this case.

Comparison of mediation effects: Following the guidelines provided by Chin et al. (2013) and Nitzl et al. (2016), the potential statistical difference between (a₁ × b₁) and (a₂ × b₂) was evaluated. Since we have not posited any hypotheses on the differential impact of both indirect effects, a bilateral test (95% confidence interval (CI)) will be conducted (see Table 6). The test shows there is a differential impact between the specific indirect effects M₁ and M₂ since neither CI contains the zero value. Therefore, it can be claimed that the WFE variable (M₂) has a stronger mediating effect than that of the WFC (M₁) variable.

5. Discussion

This study aims to contribute a better understanding of the mechanisms linking the use of work–family benefits to the work–family interface (conflict and enrichment) and work outcomes (JS) associated with them, from the perspective of the work domain. We argue that the use of work–family benefits contributes to improving JS by minimising WFC and promoting WFE. Furthermore, our contributions are found in the context of an emerging country in the field of WLB.

The predictions of our integrated research model have been supported in all cases. As expected, WFC and WFE mediate the relationship between the use of work-family benefits and JS. This allows us to infer that both WFC and WFE increase the magnitude of the existing relationship between the use of work–family benefits and JS, the effect magnitude being higher through WFE.

The mediating role is consistent with previous research focused on WFC (Anderson, 2002), WFE (Baral and Bhargava, 2010), or both (Carlson et al., 2010). It is also consistent with a growing body of research suggesting that conflict and enrichment are distinct concepts (Byron, 2005; Carlson et al., 2006, 2010; Greenhaus and Powell, 2006; Wayne et al., 2007).

Regarding the relationships between the antecedents of the work–family interface, such as the boundary-spanning resources, work–family benefits in our case, significant relationships were found, both with WFC and WFE. The relationships between the use of work–family benefits and WFC and WFE were observed to be significant, although their effects were small. The relationship was a little higher on the enrichment side. These results contrast with the ones observed by Wayne et al. (2006) and Baral and Bhargava (2010), for whom the use of work–family benefits did not keep a significant relationship with WFE. Hence, our findings represent an important contribution to this line. On the other hand, the literature suggests a negative and significant association of WFC with the use of work–family benefits (O’Driscoll et al., 2003), which is consistent with our findings. In the same vein, Lapierre et al. (2018) argued that the provision of resources seem to relate more strongly (positively) with enrichment than with conflict and, conversely, the deployment of resources seem to relate more strongly (negatively) with conflict than (positively) with enrichment.

As for the relationships with the results of the work–family interface such as JS, we found significant relationships, both with conflict and enrichment, as with work–family benefits. Baral and Bhargava (2010) did not find significant relationships between the use of work–family benefits and job outcomes such as JS, which contrasts with our findings. These results are consistent with previous literature results (Carlson et al., 2010). Our results suggest that WFE is different from WFC and that enrichments can play a more fundamental role in the configuration of the organisational outcomes expected.

6. Conclusions and theoretical and practical implications

This study makes valuable contributions not only to the literature but also provides some implications for practice are suggested by our results. First, our conceptual model proposes the integration of several work–family interface models with the broaden-and-build theory of positive emotions, drawing on suggestions from previous works (Carlson et al., 2010). Our results help to strengthen these conceptual developments and pose challenges for future integrative analyses.

Second, our model provides evidence of the conceptualisation of boundary-spanning resources (Voydanoff, 2004, 2005) and much-needed research on the positive side of the work–family interface. Third, by adopting a double perspective (conflict and enrichment), it contributes to understanding the relationship of WLB measures with the work–family interface seeing there is no consensus in the literature on how conflict and enrichment are affected by different measures. In addition, it contributes to the understanding of the mediation processes linking the use of work–family benefits to affective outcomes (JS), through the work–family interface.

Fourth but connected to the previously mentioned double perspective (conflict and enrichment), this study model provides a tool for firms’ decision-making; it justifies managers to undertake initiatives on the adoption and promotion of work–family benefits. Moreover, it fosters the identification of the levels of satisfaction associated with the perception of both WFC and WFE to establish strategies for retaining firms’ most valuable employees. Our results confirm that the effect of the enrichment on JS is higher than the impact that conflict may have. Therefore, it seems that if companies want to maximise the influence on JS, helping their employees to avoid the conflict has only a limited effect. Practitioners should focus on a more efficient approach, aiming to have an impact on the WFE to increase the JS.

Fifth, this study focusses on the insufficiently researched Latin American context, where interest in reconciling work and personal life is progressively increasing at a personal, business and governmental level. Political, legal and cultural elements are vital to explain the relevance of WLB in different contexts (Pasamar and Valle, 2011). In this regard, it has been possible to build a broader and more specific group of policies for the Latin American context that can be a referent for researchers, organisations and government agencies. Furthermore, by analysing the joint effect of 46 work–family benefits, we are taking into account the synergistic effect suggested by Carlson et al. (2010), Baral and Bhargava (2010) and Martinez-Sanchez et al. (2018) to continue analysing the use of various formal policies together. Finally, by choosing a large population with individuals who worked for firms from different sectors, sizes and capital origins, it was possible to broaden the scope of our results’ generalisation.

Our work has faced some limitations. First, the focus of this research revolved around the work domain adopting a source attribution perspective. However, some researchers advise it is important to bear in mind cross-domain effects when analysing the antecedents and outcomes of both conflict and enrichment (Amstad et al., 2011; Lapierre et al., 2018; Zhang et al., 2018).

Second, some limitations have to do with the survey design. The questionnaire addresses a single respondent, and there was no time lapse between the data collection for the dependent and independent variables. We assume that the dependent variables behave consistently over time. Some additional control variables such as age or occupational status could be included in future research initiatives. Then, the problem of CMV has been assessed properly and has been found to have no relevant implications.

Third, employees are not randomly assigned into workplaces. This may bias the estimates of our independent variables on JS. Employees’ previous performance at the workplace could be an important antecedent of JS. Böckerman et al. (2012) used employees’ work and sickness absence histories to solve this bias problem. Because our database did not include this information, we could not perform this additional analysis. However, we suggest information on work histories could be taken into account when designing further empirical studies on JS.

Future research could overcome these limitations and include additional variables connected with both conflict and enrichment such as psychological job demands (Huang et al., 2019), job autonomy (Huang et al., 2019; Wang et al., 2020), financial well-being (Huang et al., 2019) and turnover intention (Aboobaker and Edward, 2020; Wang et al., 2020). Future research could also examine work–family benefits and JS from a multilevel approach (Hernaus et al., 2020).

Figures

Figure 1

Research model

Figure 2

Structural model (multiple mediation model)

Table 1

Measurement model: loadings, construct reliability and convergent validity

Construct	Item	Loading	t value	α	ρA	ρc	AVE
Work-to-family conflict (WFC) (Common or reflective factor)	CTWF1	0.711***	15.890	0.905	0.923	0.906	0.527
	CTWF2	0.700***	14.587
	CTWF3	0.739***	16.759
	CSWF7	0.830***	19.368
	CSWF8	0.869***	22.700
	CSWF9	0.927***	23.112
	CBWF13	0.473***	8.107
	CBWF14	0.610***	10.711
	CBWF15	0.547***	9.721
Work-to-family enrichment (WFC) (Common or reflective factor)	EWFD1	0.613***	16.662	0.949	0.961	0.948	0.676
	EWFD2	0.626***	17.545
	EWFD3	0.596***	16.377
	EWFA4	0.833***	36.186
	EWFA5	0.951***	56.365
	EWFA6	0.904***	44.487
	EWFC7	0.973***	53.942
	EWFC8	0.920***	51.998
	EWFC9	0.867***	38.593
Job satisfaction (JS) (Common or reflective factor)	JS1	0.877***	42.015	0.930	0.940	0.932	0.700
	JS2	0.804***	29.960
	JS3	0.838***	37.830
	JS4	0.627***	19.171
	JS5	0.924***	56.818
	JS6	0.914***	61.735

Note(s): t value for 5,000 subsamples; α: Cronbach’s alpha; (ρA: Rho): Dijkstra–Henseler indicator; (ρc): composite reliability (Dillon–Goldstein index); and AVE: average variance extracted. *p < 0.05, **p < 0.01, ***p < 0.001

Table 2

Measurement model: discriminant validity

Fornell–Larcker criterion				HTMT
	WFC	WFE	JS	WFC	WFE	JS
Work-to-family conflict (WFC)	0.726
Work-to-family enrichment (WFE)	−0.423	0.822		0.430
Job satisfaction (JS)	−0.414	0.685	0.837	0.410	0.673

Note(s): WFC: work-to-family conflict; WFE: work-to-family enrichment; and JS: job satisfaction. Fornell–Larcker criterion: the diagonal elements (in italics) are the square root of the variance shared between the constructions and their measurements (AVE). The elements outside the diagonal are the correlations between constructs. For discriminant validity, the diagonal elements must be larger than the elements outside the diagonal. HTMT: heterotrait–monotrait ratio

Table 3

Structural model: results

			R2 JS = 0.514
			R2 WFC = 0.066
			R2 WFE = 0.069
					Bootstrap percentile
Relationships			β	t value	95% CI		f2	Variance accounted for	VIF	Supported
					Lowest	Highest
Hypothesis		Direct effects
H1	+	USOB → JS	0.140***	5.470	0.097	0.181	0.036	4.62%	1.128	Yes
		USOB → WFC	−0.235***	7.760	−0.284	−0.186	0.057	5.41%	1.042
		USOB → WFE	0.240***	8.155	0.192	0.287	0.059	6.00%	1.042
		WFC → JS	−0.132***	4.582	−0.179	−0.086	0.028	5.45%	1.270
		WFE → JS	0.588***	22.077	0.543	0.631	0.558	40.27%	1.274
Control variables
Gender → WFC			0.014	0.440	−0.039	0.068	0.0002	0.05%
Gender → WFE			0.023	0.750	−0.028	0.073	0.001	−0.02%
Gender → JS			−0.039	1.593	−0.079	0.00002	0.003	0.26%
Spouse or partner → WFC			−0.076**	2.344	−0.130	−0.023	0.006	0.62%
Spouse or partner → WFE			0.008	0.250	−0.046	0.060	0.0001	0.04%
Spouse or partner → JS			0.009	0.389	−0.031	0.050	0.0002	0.08%
Dependent children → WFC			0.105***	3.216	0.051	0.158	0.011	0.48%
Dependent children → WFE			0.059*	1.813	0.006	0.113	0.003	0.57%
Dependent children → JS			0.036	1.436	−0.006	0.076	0.002	0.42%
Dependent adults → WFC			−0.009	0.274	−0.060	0.043	0.0001	0.01%
Dependent adults → WFE			0.045	1.459	−0.006	0.096	0.002	0.28%
Dependent adults → JS			0.032	1.368	−0.007	0.070	0.002	0.26%

Note(s): USOB: Use of work–life benefits and policies (WLBPs); WFC: work-to-family conflict; WFE: work-to-family enrichment; and JS: job satisfaction. β: coefficient; and CI: confidence interval. f ²: effect size. Variance accounted for = β * correlation

t value for 5,000 bootstrapping subsamples: (based on a distribution t (4,999) of one-tailed student). t (0.05; 4,999) = 1.645; t (0.01; 4,999) = 2.327; and t (0.001; 4,999) = 3.092. *p < 0.05, **p < 0.01 and ***p < 0.001

Table 4

Structural model: evaluation indicators (prediction power)

Factor	R2	R2 Adjusted	Q2
Work-to-family conflict	0.066	0.061	0.032
Work-to-family enrichment	0.069	0.064	0.043
Job satisfaction	0.514	0.510	0.325

Note(s): WFC: work-to-family conflict; WFE: work-to-family enrichment; and JS: job satisfaction

Table 5

Summary of the mediating effect test

Total effect (c)		Direct effect (c′)			Indirect effects			Bootstrap percentile		VAF
β	t value		β	t value		Estimated point	t value	95% CI
								Lowest	Highest
0.313***	11.424	H2: c′	0.140***	5.470	Total = a1×b1 + a2×b2	0.172***	8.442	0.139	0.206	55.08%
					H4a = a1×b1 M1	0.031***	3.931	0.019	0.045	18.09%
					H4b = a2×b2 M2	0.141***	7.359	0.110	0.173	50.13%

Note(s): β: Coefficient; VAF: variance accounted for test. t value for 5,000 bootstrapping subsamples: (based on a distribution t (4,999) of one-tailed student). t (0.05; 4,999) = 1.645; t (0.01; 4,999) = 2.327; and t (0.001; 4,999) = 3.092. *p < 0.05, **p < 0.01 and ***p < 0.001

Table 6

Comparison of the mediation effects

Differential effect	β	Bootstrap percentile
		95% CI
		Lowest	Highest
M1 − M2 = (a1 × b1) − (a2 × b2)	− 0.110	− 0.152	− 0.069

Note(s): β: Coefficient