Work–life balance has grown to be a major concern in today’s organisations. The dynamism of the business environment continues to accentuate a greater difficulty in the balance between professional roles and personal responsibilities (Shankar & Bhatnagar 2010). Dating back to the 19th century, the construct of WLB was seen as more inclined to women as it became prominent at a time when women entered the labour market. However, Aruna Shantha (2019) notes the differences in the global communities, which reflect a different picture whereby all employees regardless of gender also want flexibility and control over their work and personal life. Today’s generation has many responsibilities, work obligations and family responsibilities owing to the complexity brought about by globalisation and technology, hence employers should craft effective and efficient family friendly policies and practices supportive of WLB.

According to Brough et al. (2020), WLB is a multifaceted concept that encompasses various elements that include equity across multiple roles, perceived control between multiple roles, a satisfaction between multiple roles and the fulfilment of role salience between multiple roles. There is a strong need for retailers to build pillars of support both at work and home. Every employee deserves work opportunities that strike a balance between their career and personal life in order to nurture their well-being (Sen & Hooja 2018).

Although an enormous amount of literature deals with the issue of work–family conflict and family–work conflict, a full review of this literature goes beyond the purview of this study. Work–family conflict occurs when an incompatibility between the pressures of work and family happens to an employee (Akkas, Hossain & Rhaman 2015). Work–family conflict can be experienced when an employee has to perform the multiple roles of spouse, parent and worker (Akkas et al. 2015). In order to be executed effectively, each of the aforementioned roles demand energy, time and commitment. However, demands from work and family roles are often mutually incompatible – participation in one role makes participation in the other role more difficult. Work–family conflict has become more ubiquitous than family–work conflict (where family interferes with work) although both can occur (Akkas et al. 2015). However, irrespective of the direction of causation, when one realm gets acrimonious with the other realm, the result is usually conflict and escalating stress on the employee.

A research study conducted by Murphy, Duxbury and Higgins (2007) revealed that high levels of work–family conflict were experienced by people with a lot of care-needing dependents who include the elderly and children. The research shows that employees with a lot of care-needing dependents fail to strike a balance between job demands and family responsibilities, thus work–family conflict results. Working mothers with children below the age of three were found to experience high levels of work–family conflict because of child care needs. Moreover, families with strong bonds and who were fully immersed in their family traditions and values experienced a lot of family–work interferences (Murphy, Duxbury & Higgins 2007). Most family-related factors are bound to a lot of care giving. This, however, has a negative bearing on the level of commitment, motivation and engagement to one’s work (Murphy, Duxbury & Higgins 2007).

People have different values and beliefs that may interfere with the balance between work and family. Individual needs and values can contribute to work–family conflict (Marolt, Zimmermann & Žnidaršič 2020). According to (Kossek & Lee 2017) importance, priority and centrality help to explain the individual factors that influence work–life conflict. The value that an individual places on a role determines its importance and how it is prioritised (Kossek & Lee 2017). Importance is the value expression that an individual uses to determine the gravity and urgency of a role (Gonçalves, Curado & Martsenyuk 2023).

According to Žnidaršič and Marič (2020), literature, identifies two types of conflicts, namely time and strain conflict. On one hand, time conflict occurs when the time allocated to work roles interferes with the fulfilment and functioning of duties of the family and vice versa. This type of conflict is a result of people failing to equally distribute their time among different roles (Cui & Li 2021). A good example of time-based conflict is when an individual chooses to work overtime while missing a family function or commitment (Cui & Li 2021). Higher levels of work–family conflict are likely to have a negative bearing on employee well-being and performance (Fay & Hüttges 2017).

On the other hand, strain conflict occurs when stress from work (or family) hinders family (or work) duties. Netemeyer, Boles and McMurrian (1996), refer to a case where work-related anxiety and irritability deter the accomplishment of family (or work) roles and vice versa. For instance, marital problems at home could influence one to behave poorly at work or miss deadlines at work. Employees in the retail sector are more likely to be vulnerable to strain-based conflict because of the excessive stress levels associated with the industry (Vutete & Vutete 2015).

The employees in the Bulawayo Urban retail sector were the target population of the research study (Cooper & Schindler 2006). The study used the convenience sampling technique, which allowed researchers to get more less costly samples and in a short period.

A structured questionnaire with closed items was used for this study because it allows for minimal researcher involvement and influence on respondents (Khothari & Garg 2014). The researchers adapted standardised items for all the three variables under study. Customised items from the Self-Assessment WLB scale by Hayman (2005), the job engagement scale by Schaufeli et al. (2002) and the JS Rating scale by Warr, Cook and Wall (1979) was used.

Section A focussed on demographic data. Sections B to D used a five point Likert scale. Participants were expected to choose their responses from: ‘strongly disagree’ – 1 point, ‘disagree’ – 2 points, ‘neutral’ – 3 points, ‘agree’ – 4 points and ‘strongly agree’ – 5 points.

The researchers sought for ethical clearance first before questionnaires were physically distributed with strict adherence to coronavirus disease 2019 (COVID-19) safety protocols. After seeking respondents’ informed consent and assuring them of anonymity, respondents were given 3 days to complete the questionnaire. Of the 120 delivered questionnaires, 108 were completed and returned.

The researchers employed structural equation modelling (SEM) because it offers clear statistical analysis that can be testable in conducting further analysis (Hair et al. 2014). Some evidence support that some SEM models could be tested meaningfully even with a small sample size of 100 to 150, which is regarded as the minimum size of a sample for conducting SEM (Tabachnick, Linda & Fidell 2001). In view of this, this study used a sample size of 108 respondents

Ethical clearance to conduct this study was obtained from the Great Zimbabwe University Julius Nyerere School of Social Sciences Research Ethics Committee.

To do SEM there are certain things such as discriminant validity and reliability tests that have to be met. The Cronbach’s alpha test (CA) that tests internal consistency, the composite reliability (CR) and even the average extracted variance (AVE) were calculated, and the results are summarised in Table 1. The considered variables were WLB, VI, JS, DE and AB.

The results shown in Table 1 indicate that the CA values do range from 0.757 to 0.878, suggesting internal consistency of the measurement constructs. All the CR values exceeded 0.7, a recommended threshold value by Hair et al. (2014). In addition, all the AVE values exceeded the 0.5 threshold value. These results imply that the measurement constructs were reliable, hence better results may be attained.

The structure and cross-loadings of the variables were calculated to determine convergent validity of the factors with the results being summarised in Table 2.

It can be noticed, from the Table 2, that all the cross-loading values exceeded 0.6, showing that convergent validity existed on the measurement constructs. All variables with factor values (loadings) below 0.6 were removed from the model. Inter-construct correlations helped to examine discriminant validity. Inter-constructs correlations are presented in Table 3.

According to the findings displayed in Table 3, discriminant validity existed in all the measurement items, as shown by the square root of the AVE values (diagonal values) for the factors, which exceeded the corresponding correlation coefficient values of other factors. As reliability tests and discriminant validity were confirmed, a structural equation model can be fitted (see Figure 1).

From the SEM displayed in Figure 1, the following hypotheses were derived and tested:

The partial least squares (PLS) approach was employed in analysing the data through the use of SmartPLS software. The SEM results from the analysis are presented in Table 2.

The first hypothesis examined the relationship between WLB and VI. A coefficient value of 0.704 shown on the table above, shows that WLB has a positive effect on VI. The association between WLB and VI is statistically significant at 1% level because the probability value is less than 0.001. This suggests the acceptance of H₁, which says WLB is positively associated with VI. This means that as WLB improves, VI will also improve.

The second hypothesis examined the relationship between VI and JS. It can be noticed, from the above table, that VI positively affects JS because of the coefficient value of 0.238. Furthermore, there is a statistically significant relationship at 5% level because the probability value of 0.028 is less than 0.05. The results suggest the acceptance of the H₂, which states that VI is positively associated with JS. This implies that improvements in VI will result in JS improvement.

The relationship between WLB and JS formed the third hypothesis of the study. According to the given results, WLB had a positive effect on JS as indicated by a positive coefficient value of 0.304. Work–life balance and JS have a statistically significant relationship at 5% level because of the probability value of 0.046, which is below 0.05. From these results, hypothesis H₃ is accepted. This means that the improvements in WLB will be reflected by the improvement of JS.

The fourth hypothesis attempted to examine the relationship between WLB and DE. The given results suggest that WLB had a positive effect on DE, as supported by a positive coefficient value of 0.912. A statistically significant relationship exists at 1% level because the probability value is less than 0.001. These results suggest the acceptance of H₄, which says WLB is positively associated with DE. These results suggest that as WLB improves, DE improves as well.

The fifth hypothesis investigated the association between DE and JS. The results displayed in the table suggest that DE positively affects JS, as indicated by the positive coefficient value of 0.077. However, the relationship lacks statistical significance, as evidenced by a probability value of 0.635, which lies above the 0.05 threshold value. These results suggest the rejection of H₅.

The sixth hypothesis was meant to examine the relationship between WLB and AB. According to the results displayed in the table above, WLB had a positive effect on AB because of the positive coefficient value of 0.709. Furthermore, the relationship becomes statistically significant at 1% level because the probability value is less than 0.01. These results suggest the acceptance of H₆, which states that WLB is positively associated with AB. The result implies that the improvements in AB will be influenced by improvements in WLB.

The seventh hypothesis examined the relationship between AB and JS. According to the results in Table 4, AB had a positive effect on JS because of the coefficient value of 0.246. The relationship between AB and JS is statistically significant at 5% level because the probability value of 0.025 is below the threshold value of 0.05. These results suggest the acceptance of H₇, which says AB is positively associated with JS. This means that as AB improves, JS improves as well. The results in Table 4 suggest that H₁, H₂, H₃, H₄ and H₆ are supported, while H₅ is not supported.

The next section presents a summary of the fitted model containing coefficients and factor loadings.

Brioness Penalver et al. (2017) observed that R² and Q² are helpful in deciding the goodness of fit of a model and showed that the values have to be above zero. Both the R² (R² = 0.609) and Q² (Q² = 0.290) exceed the zero threshold value. Hair et al. (2016) postulate that a Q² value above 0 for an endogenous latent variable depicts the path model’s predictive relevance for a certain dependent construct. According Figure 2, 60.9% (R² = 0.609) total variability in JS was being explained in the model by the independent variables (VI, DE, AB and WLB). Henseler, Ringle and Sinkovics (2009) and Hair, Ringle and Sarstedt (2011) suggest that the R² for the fitted model implies good levels of predictive accuracy of the model. Table 5 shows some of the goodness of fit results.

Table 5 indicates that, the standardised root mean square residual (SRMR) value for the fitted model is 0.077, which lies below the threshold value of 0.1, suggesting acceptance of the model. Furthermore, the normed fit index (NFI) value of 0.91 for the model, which is slightly above the recommended threshold value of 0.9, suggests that the fitted model is a good.

Structural equation modelling was employed to assess how WLB and job engagement relate. Job engagement was characterised by DE, VI and AB (Schaufeli et al. 2002). The first hypothesis presented by the SEM examined the relationship between WLB and VI. The SEM results revealed that WLB had a positive effect on VI, DE and AB, with coefficient values of 0.704, 0.912 and 0.709, respectively. These indicated a strong positive correlation between WLB and the three variables measuring job engagement. The p-values of 0.001 and less for the three variables at 1% level of significance suggest the acceptance of the alternative three hypotheses H₁, H₄ and H6, which imply a statistically significant relationship between WLB and the three variables measuring job engagement, viz VI, DE and AB

The aforementioned findings are supported by studies carried out by (Li, McCauley & Shaffer 2021) and Vance (2017), who observed that the resources offered by a job have significant and motivational influence in increasing job engagement. Furthermore, other studies (Dhas 2018; Presa 2018) indicated a positive correlation between WLB and job engagement. Ibrahim et al. (2020) argued that poor WLB practices such as long working hours, erratic work schedules, poor communication, a lack of autonomy among other factors have a negative impact on employee engagement levels. The issue of WLB poses some challenges in various degrees and with different outcomes to employees despite their occupation, gender and age. It is therefore the duty of management to ensure that effective and efficient WLB practices are in place (Presa 2018). The major concern is that organisations and regulatory boards do revise unfavourable laws, policies and practices to further support the concept of WLB thereby enhancing workforce engagement and corporate effectiveness (Dhas 2018). Job engagement and WLB are catalysts that have incredible potential in unleashing continuous corporate growth. The retail sector characterised by long working hours, busy schedules, fast movement of goods and services, requires that management do exercise effective WLB practices to enhance employee job engagement (Chikweche 2017).

The hypothesis examining the relationship between WLB and JS was again tested using SEM. The SEM results revealed that WLB had a positive but weak effect on JS. This was evidenced by a positive coefficient value of 0.304, which was indicative of a weak positive correlation between WLB and JS. In addition, a probability value of 0.046, which is below 0.05 proved that the relationship between WLB and JS was statistically significant at 5% level. The results suggest acceptance of the alternative hypothesis H₃, which implies a positive association between WLB and JS. The research findings support those from a related study by Aruna Shantha (2019) that found a statistically significant and positive correlation between WLB and JS. Žnidaršič et al. (2020) and Casper et al. (2018) also assert that striking a balance between work and family commitments enhances JS. Kashyap et al. (2016) however observes how challenging it is for organisations to achieve WLB in the retail sector, especially in a developing country such as Zimbabwe characterised by a high rate of inflation, political instability, unstable currency and high unemployment.

Interestingly, Adikaram (2018) seems to realise that managerial employees can tailor-make their own schedules to balance their family and work commitments, whereas lower-level employees have little to say on how their work schedules are crafted. This could explain why levels of JS could differ among employees in the same organisation. It therefore implies that job autonomy, being one of the significant elements of WLB shares a positive relationship with JS (Adikaram 2018). As a result, the greater autonomy one has, the more power they wield in deciding flexible working arrangements and the greater the likelihood of satisfaction outcomes (Adikaram 2018).

Hypotheses 2, 5 and 7 presented by the SEM examined the relationship between the three variables of job engagement (VI, DE and AB) and JS. A coefficient value of 0.238 was indicative of a weak positive correlation between VI and JS. The relationship was, however, statistically significant at 5% level, with a probability value of 0.028, which is less than 0.05. This led to the acceptance of the alternative hypothesis H₂, which proposes that VI is positively associated with JS. The fifth objective investigated the association between DE and JS. The results displayed in the table suggest that DE weakly but positively affects JS, as indicated by a positive coefficient value of 0.077. However, the relationship was not statistically significant, as evidenced by a probability value of 0.635, which is above the threshold value of 0.05. These results suggest the rejection of the alternative hypothesis, which suggests a statistically significant relationship between DE and JS. Finally, the seventh hypothesis presented on the SEM examined the relationship between AB and JS. A coefficient value of 0.246 indicated a weak positive correlation between the two. The relationship between AB and JS was statistically significant at 5% level because of a probability value of 0.025, which was below the threshold value of 0.05. The results therefore suggest the acceptance of the alternative hypothesis that suggests a positive association between AB and JS.

The aforementioned results suggest a weak positive correlation between job engagement (as measured by all the three variables) and JS, while statistically significant relationships existed between only two variables measuring job engagement (i.e. VI and AB) and JS. This means that as employee job engagement improves, their JS would weakly improve as well. Job engagement was therefore instrumental in influencing JS. These research findings concur with Harini, Luddin and Hamidah (2019) who established that the level of commitment and satisfaction one has towards his or her job is derived from the level of engagement they have towards the job. Individuals who are highly engaged to their work are very persistent in achieving their tasks even in the face of extreme challenges, all because of the satisfaction derived. Extant literature (Aziri 2019; Bellani et al. 2018; McIlveen et al. 2021) also confirm the positive relationship between job engagement and JS Effective employee engagement practices such as valuing the opinions of employees, getting them involved in crafting policies and practices, autonomy, among others, enhance outcomes of JS to a greater extent. Furthermore, Maryati and Astuti (2022) assert that effective job engagement practices create communication channels between the workforce and its management facilitating outcomes of JS, motivation and commitment.

It can however be highlighted that the lack of a statistically significant relationship between DE and JS, although confounding, has been explained by Vance (2017). Employees may be dedicated to their jobs and lack satisfaction from it or vice versa because of various reasons such as employment opportunity, ineffective collective bargaining in terms of the value of salaries, recognition, working conditions, leadership, style of management and the job itself (Vance 2017). Although previous studies (Abu Khalaf, Hmoud & Obeidat 2019; Aziri 2019; Bellani et al. 2018; Casper et al. 2018; Harini et al. 2019; Jaiswal et al. 2017; Lee 2017; McIlveen et al. 2021; Abu Khalaf, Hmoud, & Obeidat 2019) indicate a positive relationship between job engagement and JS, this study indicated that the DE of employees to their work had no influence on JS. Therefore, investing in employee engagement, specifically regarding DE, will not enhance outcomes of JS as previously anticipated. However, the contrasting results could be attributed to the unique external environment under which the study was conducted. This study was conducted in a country in the global south, characterised by hyperinflation, an unstable currency, regressive laws, political instability, dilapidating economy and a high rate of unemployment among many other depressive characteristics (Chikweche 2017).