Engineering work in South Africa is carried out against a backdrop of persistent skills scarcity and infrastructure delivery pressures that directly affect how engineering work is organised and experienced by professionals. At a national level, South Africa faces a longstanding deficit of engineering professionals and supporting occupations, affecting both capital projects and, especially, ongoing operations and maintenance activities that require sustained technical capacity over time (Academy of Science of South Africa [ASSAf], 2021). The country’s development needs demand greatly increased engineering skills to conceive, design, build and sustainably operate the services and infrastructure required for quality of life (ASSAf, 2021). Consistent with this macro-level picture, evidence from the built-environment sector confirms that an engineering skills shortage poses a major threat to South Africa’s built environment and that shortages of engineers hinder the ability of the government to sustain the economic growth needed for development and the eradication of poverty (Chipanga & Bowen, 2022; Gomes et al., 2022). The civil engineering labour market is similarly constrained, with estimates indicating approximately 15 000 civil engineering professionals and 3000 to 6000 vacancies; short-term organisational responses may include outsourcing and encouraging longer working hours (Visser & Joubert, 2022). National infrastructure programmes further intensify the demand for engineering capability by generating additional skills requirements that must be met both ‘in advance of’ and ‘through’ implementation, thereby placing pressure on education, training and workplace learning systems (Department of Higher Education and Training [DHET], 2014). The Strategic Integrated Projects (SIPs) Skills Plan was explicitly premised on an anticipated shortage of critical skills for infrastructure delivery while acknowledging that skills planning is complicated because project-specific skills needs cannot be separated from broader labour-market demand and the mobility of skilled personnel (DHET, 2014). Importantly, the SIPs Skills Plan articulated the principle that ‘building people is as critical as building physical assets’, underscoring that engineering capacity is inseparable from workforce development and sustainability (DHET, 2014). In practice, sustained shortages can lead to heavier individual workloads, as employees in occupations affected by skills shortages are ‘often burdened with additional duties and responsibilities’, which may increase their vulnerability to work-related stress and undermine optimal functioning (Thasi & Van der Walt, 2020). Under these conditions, occupational stress (OS) and work–life balance (WLB) are not peripheral concerns but central features of the sustainability and performance of the engineering workforce in South Africa (Bowen et al., 2014a; Rothmann & Malan, 2006). In South African engineering samples, WLB and work overload have been identified as the most important occupational stressors for engineers, indicating that time-based and energy-based pressures are salient in this professional group (Rothmann & Malan, 2006). Evidence further suggests that work overload, WLB and perceived control jointly function as key predictors of both physical and psychological ill-health, linking job design and boundary-related strain to well-being outcomes that matter for retention and capability in scarce-skills contexts (Rothmann & Malan, 2006). When South African engineers were compared with normative data, they reported higher psychological outcomes of stress, indicating that occupational stressors are associated with materially elevated psychological strain (Rothmann & Malan, 2006). Moreover, occupational stressors, alongside low organisational commitment, explained 18% and 24% of the variance in physical and psychological ill-health, respectively, in this population, reinforcing the relevance of stress processes to health and organisational outcomes (Rothmann & Malan, 2006). Sector-adjacent evidence from South Africa’s construction and built-environment professions underscores the salience of occupational stressors structurally embedded in project work. Among South African construction professionals, most respondents report high levels of work-related stress, and tight deadlines and long working hours ‘probably play a bigger role’ in their stress experiences than perceived control over job situations (Bowen et al., 2012). Job-demand factors such as critical project time constraints, having to work long hours and ‘having to skew the work–family life balance’ are identified among the most important project-related contributors to perceived stress, highlighting the centrality of time pressure and work–home boundary disruption in project-based professional work (Bowen et al. 2014c). In predictive modelling of workplace stress in South African construction contexts, work–life imbalance emerges as the strongest predictor of high workplace stress, and the perceived need to work long hours is described as overwhelmingly mediated by the imbalance between work and life or family commitments (Bowen et al., 2014a). These findings are particularly relevant in a developing-country context characterised by economic hardship and social problems that compound occupational pressures on built-environment professionals (Bowen et al., 2012). Work–life balance in this study is conceptualised as the ability of individuals to successfully pursue their work and non-work lives ‘without undue pressures from one undermining the satisfactory experience of the other’ (Greenhaus & Allen, 2011). Relatedly, work–life conflict can be understood as a form of inter-role conflict that occurs when the energy, time or behavioural demands of the work role conflict with those of family or personal life roles, linking balance to role pressures and resource allocation across domains (Byron, 2005). Work–life conflict is consequential because it affects work, non-work and personal outcomes, including health and stress, thereby establishing a clear theoretical and practical rationale for studying balance as a proximal determinant of strain outcomes in high-demand roles (Byron, 2005). Role Theory (Kahn et al., 1964) provides an early and enduring account of how incompatible or ambiguous role expectations generate the tension underlying both work–life conflict and OS. In engineering, these conflicting expectations involve delivering technical output, managing stakeholder relationships, ensuring compliance and coordinating team activities simultaneously; this constellation of role demands makes engineers particularly susceptible to boundary erosion and strain (Rothmann & Malan, 2006). Occupational stress in engineering-relevant contexts is evident not only in immediate strain reports but also in downstream well-being risks such as burnout, defined as a state of physical, emotional and mental exhaustion resulting from long-term involvement in emotionally demanding work situations (Rothmann & Malan, 2006). Burnout has been associated with health problems, depression, reduced productivity, absenteeism and job turnover, suggesting that chronic stress processes erode both employee well-being and organisational capacity in professions critical to infrastructure and economic development (Rothmann & Malan, 2006). In engineering work specifically, the work environment is described as demanding and stressful, and burnout arises from complex interactions between individual characteristics and work-environment issues rather than a single cause (Rothmann & Malan, 2006). Consistent with an interface perspective, evidence from practising civil engineers indicates that burnout can be predicted by combinations of work and family stressors and that prevention strategies must extend beyond the work environment to address issues at the work–family interface (Lingard, 2003). To explain why demands and resources should be studied together, the Job Demands-Resources (JD-R) model (Bakker & Demerouti, 2007) provides a flexible framework for understanding employee well-being by integrating negative and positive indicators and by considering how the structure of work shapes both strain and motivation processes. Job demands (JD) are defined as those physical, psychological, social or organisational aspects of the job that require sustained effort and are associated with physiological and/or psychological costs, while job resources are aspects of the job that are functional in achieving work goals, reduce JD and their associated costs, or stimulate personal growth and development (Bakker & Demerouti, 2007; Schaufeli & Taris, 2014). The model specifies two relatively independent processes: a health-impairment process in which high JD exhaust employees’ resources and may lead to energy depletion and burnout, and a motivational process in which job resources induce employees to meet goals and may lead to work engagement (Bakker et al., 2005). Empirical evidence further supports a buffering proposition in which the allocation of specific job resources can mitigate the effects of specific JD on core dimensions of burnout, especially when demands are high (Bakker et al., 2005). Perceived organisational support (POS) is treated in this study as a salient resource-related perception reflecting whether engineers feel valued and cared for in demanding work settings (Eisenberger et al., 1986). Evidence from engineering-related contexts indicates that perceptions of being valued and supported are meaningfully connected to stress-related mental health outcomes. A regression model identified feeling of appreciation (β = −0.301), ability to handle pressure (β = −0.253), and balance between professional and personal life (β = −0.182) as significant predictors of lower combined anxiety, depression and stress symptoms among engineers (Teles et al., 2020). In parallel, South African stress-modelling research suggests that support received from colleagues in difficult situations may function as a moderator in the stress process, supporting the conceptual role of support-related resources as protective factors even when demands remain high (Bowen et al., 2014a). Despite clear evidence that WLB and overload-related demands are central stressors for South African engineers, gaps persist in how these constructs have been examined in integrated ways within the South African engineering context. Earlier South African scholarship explicitly noted that no research had been conducted on burnout in the engineering profession in South Africa and that the broader burnout literature in South Africa suffered from a lack of systematic empirical research, small sample sizes and poorly controlled study designs (Rothmann & Malan, 2006). Given that engineers are positioned as key contributors to South Africa’s growing economy and must cope with increasing pressure to perform, the ‘work wellness’ of engineers is directly relevant to the profession’s potential to contribute to national success (Rothmann & Malan, 2006). This study, therefore, tests an integrated model linking WLB to OS and examines whether POS and JD mediate this relationship, with the envisaged findings serving both theoretical understanding and organisational practices that protect engineers’ well-being and support sustainable performance.

The research aimed to examine the relationship between WLB and OS among engineers in South Africa. It also investigated whether POS and JD mediated this relationship.

To fulfil this purpose, the following research objectives were formulated:

By achieving these objectives, the study sought to provide context-specific empirical evidence to inform practice and advance broader research on OS and WLB in high-demand professional roles.

Work–life balance is defined most directly as a perceived capability to balance work and non-work commitments, which already implies a stress-relevant appraisal of whether role demands are manageable (Haar et al., 2014). Complementary definitions specify WLB as the ability to manage work responsibilities and personal life so that neither overwhelms the other, foregrounding boundary manageability and the potential for role overload to become strain-producing (Greenhaus & Allen, 2011). Work–life balance is also framed as the extent to which individuals divide time, energy and focus between work and non-work domains, making resource allocation explicit and therefore relevant to stress processes that worsen when resources are depleted (Weerarathna et al., 2022). Work–life balance is defined as ‘the extent to which an individual successfully balances work, family, and personal life without experiencing persistent conflict between work and non-work domains’ (Buddhapriya, 2005). In addition, WLB is defined as ‘the ability of an individual to balance work and family responsibilities without negative spillover effects on either domain’ (Chaudhary, 2024). Modern studies on WLB recognise that it is a double-edged sword that may, at times, interfere with work or family life while also supporting both (Rothbard et al., 2021). There are multiple theoretical foundations for WLB. Role theory argues that people play various roles and experience strain when these roles are in conflict or ambiguous (Kahn et al., 1964). In engineering, these conflicting expectations concern the delivery of engineering output, the management of stakeholder relationships, the ensuring of compliance and the coordination of team activities. Conservation of Resources (COR) theory offers a framework for understanding imbalance as a resource-related process. Individuals are motivated to conserve and protect their resources, which include time, energy and social support (Hobfoll, 2001, 2011). As these resources are depleted by work and opportunities to recover are limited, strain will increase. Because WLB can be operationalised using multidimensional instruments that distinguish work interference with personal life from personal life interference with work (Hayman, 2005), a model-based literature review must acknowledge that ‘low WLB’ can reflect different interference patterns with potentially different implications for stress. Qualitative evidence shows that employees report significant difficulty attaining WLB and describe negative spillover – working long hours and being short-staffed preoccupy their thinking even at home (Maphumulo & Bhengu, 2019). In contrast, other occupational samples show moderate OS alongside high WLB (Singh et al., 2018), signalling that WLB and stress do not map one-to-one and reinforcing the need to test mediating mechanisms. This heterogeneity directly motivates the present parallel mediation model (H1–H3). Work–life balance remains a concern in engineering practice, where work overload strains it, thereby negatively impacting personal relationships and job satisfaction (Lingard & Sublet, 2002). Extensive research among software engineers in Sri Lanka identified WLB item means ranging from 2.92 to 3.50 on a five-point scale (Weerarathna et al., 2022), while a study of civil engineers in Chennai reported a mean of 3.74 on a seven-point scale, with more than half (55%) scoring in the low band (Anandh & Gunasekaran, 2016), indicating persistent WLB problems in engineering contexts.

Occupational stress relates to psychological or physical symptoms such as exhaustion, anxiety, depression or hypertension (Mucci et al., 2016; Quick & Henderson, 2016). It is associated with reduced organisational productivity, absenteeism and healthcare costs (Batista-Taran & Reio, 2013). Other studies note its multi-component nature, involving exposure, appraisal and strain (Gunasekara & Perera, 2023). The Job Demand-Control model holds that stress rises when demands are high and decision latitude is low (Karasek, 1979, 1990). The Effort–Reward Imbalance model posits that high effort combined with low rewards increases strain (Siegrist, 1996). The Transactional model highlights appraisal and coping processes in the stress response, explaining why employees react differently to similar demands. Critically, in the present model, appraisal-based stress logic implies that organisational resources, such as POS, can shape how work demands are interpreted – providing the theoretical mechanism underlying H2. Empirical evidence confirms that stress appraisal mediates downstream outcomes when POS is low but not when POS is high (Ng & Sorensen, 2008), directly supporting POS as an appraisal-altering mediator. In the engineering industry, the major stressors are work overload, difficulty maintaining WLB, time management issues and a lack of control (Rothmann & Malan, 2006; Saleh & Desai, 1986). In the Saleh and Desai (1986) study, macro-stressors had a mean of 13.38 (SD [standard deviation] = 4.30) and micro-stressors a mean of 11.53 (SD = 3.63). Rothmann and Malan (2006) found that South African engineers’ stress scores averaged 5.5 using the ASSET instrument, with work overload and WLB as the two most significant stressors. These findings establish the practical salience of OS in engineering and support H1. Methodological limitations must be acknowledged: the vast majority of JD-R studies use cross-sectional self-report data, making it difficult to draw strong causal inferences about directionality (Schaufeli & Bakker, 2004). This reinforces the need to present mediation as a theory-driven mechanism test rather than a definitive causal claim.

Perceived Organisational Support is defined as employees’ perception that the organisation supports them by valuing their contributions and investing in their well-being (Eisenberger et al., 1986). Perceived Organisational Support is an experience-based attribution of organisational policies, norms, procedures and actions (Eisenberger et al., 2001), thereby strengthening its suitability as a mechanism variable, as it is tied to interpretable organisational practices. Perceived Organisational Support was developed on the basis of fair treatment, management support, legitimate HR practices and employee recognition (Rhoades & Eisenberger, 2002; Sun, 2019). Meta-analytic findings confirm POS is positively related to job satisfaction, citizenship behaviour, performance and turnover intention (Kurtessis et al., 2017). Organisational Support Theory proposes that POS fulfils socio-emotional needs, thereby increasing the sense of obligation to reciprocate (Rhoades & Eisenberger, 2002). Social Exchange Theory explains how reciprocity operates between organisations and employees (Wayne et al., 1997). Psychological Contract Theory posits that unfulfilled promises undermine trust, as reflected in POS (Aselage & Eisenberger, 2003). These theories collectively position POS as a buffer against strain, particularly in high-demand situations – which is directly relevant to H2. Critically, POS is theorised within transactional stress and JD-R frameworks to buffer aversive relationships by fostering a favourable, less stressful appraisal and serving as a resource employees can rely on under excessive JD (Ng & Sorensen, 2008). Empirical findings confirm that organisational support shows a significant negative correlation with perceived stress (Maphumulo & Bhengu, 2019) and that it partially mediates the relationship between stressors and burnout (Bakker et al., 2005). Broader evidence confirms that workplace stress factors are mediated by POS in relation to employee well-being (Allen et al., 2003). These findings directly support H2. At the same time, evidence shows that POS has a marked positive effect on WLB (Aruldoss et al., 2022), and that WLB can mediate the effect of POS on well-being (Haar et al., 2014), implying that alternative orderings exist. This contrast justifies testing the specific ordering of POS mediation in the present model rather than assuming it. The study by Singh et al. (2018) found a mean POS of 4.41 (SD = 1.41) on a seven-point scale among 245 US engineers, indicating moderate perceived support. Jawahar et al. (2007) similarly found a mean POS of 4.51 (SD = 1.18) among 120 technology professionals. The large s.d.s indicate substantial variability in perceived support – consistent with the idea that POS moderates the experience of WLB-related strain.

Job demands include the amount of work, time pressure and task complexity (Van Veldhoven, 2013). Increased demands push employees to work faster and take on more responsibility (Kubicek & Korunka, 2015), and when demands exceed resources, performance and well-being suffer (Nauman et al., 2021).

The JD-R model explains two pathways: demands cause exhaustion via the health impairment process, while resources enhance engagement via the motivational process (Bakker & Demerouti, 2007; Demerouti et al., 2001). Critically, JD are boundary-eroding pressures – evidence shows that continuous learning requirements and extended working hours enabled by digital connectivity blur the boundary between work and personal domains (Weerarathna et al., 2022). In skills-shortage contexts, employees are burdened with additional duties and become more prone to work-related stress (Kleynhans & Linde, 2016), which is directly relevant to South African engineering labour markets. The JD-R model has been validated across work domains, including engineering (Bakker & Demerouti, 2007). Prior Malaysian studies have shown that high JD is associated with elevated stress among engineers (Rahim & Siti-Rohaida, 2015). Bowen et al. (2014a) measured JD among construction project consultants using the JD–CS framework, with means of 19.77 (SD = 3.49) for men and 19.21 (SD = 4.05) for women, reflecting high time pressure and workload. However, the JD-R model has been critiqued for limited explanatory specificity about underlying mechanisms (Schaufeli & Taris, 2014), and most evidence is cross-sectional. The present study therefore tests JD as a mediating mechanism in a specific South African engineering context rather than assuming generic causality. The high job demand levels observed in engineering work support H3: that JD carries part of the WLB–stress association through boundary erosion and workload intensification.

Work–life balance and OS are closely linked. When WLB is poor, particularly when work demands encroach on recovery time, stress tends to increase due to reduced coping capacity and heightened perceptions of unmanageable demands (World Health Organization [WHO], 2020). Poor WLB tends to manifest as after-work fatigue and difficulty maintaining a satisfactory personal life – a strain-based conflict experience (Fisher et al., 2009). Multiple empirical studies support a negative association between WLB and OS. A statistically significant negative correlation between OS and WLB was reported in Indian CRPF samples (Singh & Singh, 2018). A moderate negative correlation was found between WLB and OS in nursing populations (Maphumulo & Bhengu, 2019). Work–life balance is negatively correlated with job stress (r = −0.52, p < 0.01) in Filipino professionals (Saguin, 2019). Work interference with personal life and personal life interference with work each have significant negative relationships with work stress (Hayman, 2005). These convergent findings across diverse occupational contexts support H1. The WLB–stress relationship is also dynamic. Occupational stress can extend work hours, increase after-work rumination and deplete resources needed for personal life (Landsbergis et al., 2017; WHO, 2020). In turn, poor WLB can increase stress, creating a vicious cycle. However, the WLB–stress relationship is not uniformly significant. One study found no significant direct effect of WLB on job stress (significance value = 0.106) (Puteri et al., 2022). Other evidence reports non-significant WLB–well-being associations, attributed to low scale reliability (Govender & Parumasur, 2010). This heterogeneity supports the inclusion of mediators, POS and JD, that can explain when and how WLB becomes consequential for stress outcomes, directly motivating the present mediation model:

H₀₁: Work–life balance does not significantly influence OS among South African engineers.

H_a1: Work–life balance significantly influences OS among engineers in South Africa.

Perceived organisational support is theoretically justified as a mediator between WLB and OS when treated as an organisational resource that shapes stress appraisal under demanding conditions. In transactional stress and JD–R frameworks, POS mitigates aversive relationships by fostering a favourable, less stressful appraisal and serving as a resource in the face of excessive JD (Ng & Sorensen, 2008). Empirical evidence that stress appraisal operates when POS is low but not when POS is high (Ng & Sorensen, 2008) provides direct support for POS as an appraisal-conditioning mechanism. From a Social Exchange Theory perspective, POS involves employees’ belief that the organisation values their input and well-being. When employees experience good WLB, they perceive the organisation as caring and supportive, which reduces OS. When WLB is poor, perceived support declines, and stress increases (Aruldoss et al., 2022). Perceived organisational support during the coronavirus disease 2019 (COVID-19) period was shown to reduce the effects of stressors on well-being (Liu et al., 2024) and increase motivation to enhance WLB (Tan et al., 2024). Consistent with COR theory, POS acts as a resource that buffers stress appraisal when WLB-related resource depletion occurs. Empirical evidence further supports the mediating role of POS: POS is negatively correlated with perceived stress (Maphumulo & Bhengu, 2019), and it partially mediates the stressor-to-burnout relationship (Bakker et al., 2005). Workplace stress factors are mediated by POS for employee well-being (Allen et al., 2003). These findings collectively support H2. Several studies suggest that POS mediates the relationship between WLB and OS:

H₀₂: Organisational support does not significantly mediate the relationship between WLB and OS among engineers in South Africa.

H_a2: Organisational support significantly mediates the relationship between WLB and OS among engineers in South Africa.

Within the JD-R framework (Bakker & Demerouti, 2007), JD are the demand-side pathway through which WLB difficulties translate into OS: when WLB is compromised, employees typically face elevated demands that blur work–personal boundaries, deplete energy and increase stress appraisals. This boundary-erosion mechanism is empirically supported by evidence that extended working hours enabled by digital connectivity blur the work–personal boundary (Weerarathna et al., 2022), and that skills shortages burden employees with additional duties, making them more prone to work-related stress (Kleynhans & Linde, 2016). Gan and Kee (2024) found that psychosocial safety climate reduces JD, increasing work engagement and job satisfaction while reducing workload and role strain. Atiku and Van Wyk (2024) demonstrated that JD fully mediates the relationship between leadership practices and work engagement in higher education. Wood et al. (2020) showed that work–non-work supports affect well-being by modifying JD. These studies collectively show that JD can function as a mediating mechanism in well-being-related models. Hessels et al. (2017) found that self-employment increases stress due to high JD. Kloutsiniotis and Mihail (2020) found that high-performance work systems increase emotional exhaustion via high JD. These contrasting findings reinforce that demand levels shape stress outcomes through a boundary-erosion and resource-consumption pathway, supporting H3:

H₀₃: Job demands do not significantly mediate the relationship between WLB and OS among engineers in South Africa.

H_a3: Job demands significantly mediate the relationship between WLB and OS among engineers in South Africa.

A defensible gap claim requires grounding in evidence of what is known, what is missing and why this specific model is needed – no single-citation assertions of scarcity. What is known: Engineers experience significant OS. In construction professions, 61.9% of professionals reported workplace stress, with strain effects including physical fatigue, psychological frustration and strained personal relationships in high-workload, high-complexity environments (Bowen et al., 2014b). In South Africa, Rothmann and Malan (2006) established that work overload and WLB are the two primary stressors among engineers. Malaysian studies confirm that high JD is associated with elevated stress and adverse psychological health outcomes among engineers (Rahim & Siti-Rohaida, 2015). Skills shortages in South Africa burden employees with additional duties, increasing their proneness to stress (Kleynhans & Linde, 2016). What is missing: Most studies examine WLB, OS, POS and JD as separate constructs, limiting understanding of how they interact within a single system (Bakker & Demerouti, 2007; WHO, 2020). Empirical work explicitly notes that the link between POS, stress and engagement was not examined (Leiter & Maslach, 2004). There is a dearth of stress-focused research in the South African engineering context, despite the industry’s stressful nature (Schutte et al., 2012). No identified study has tested whether POS and JD simultaneously mediate the WLB–stress relationship among South African engineers using a parallel mediation design. Why this model: Heterogeneity in direct WLB–stress associations across studies (some significant, some not) suggests that mediating mechanisms explain when WLB becomes stress-consequential. The JD-R model and transactional stress theory together justify testing POS (resource-side, appraisal-buffering pathway) and JD (demand-side, boundary-erosion pathway) as simultaneous mediators.

An empirical framework was developed based on a comprehensive review of the literature (see Figure 1). Work–life balance was the independent variable, OS was the dependent variable, and POS and JD were the mediators. Work-life balance is considered to have a direct negative effect on stress. Role theory suggests that stress would increase when engineers experience conflicting or ambivalent expectations in their work and personal lives (Kahn et al., 1964). Conservation of Resources (COR) theory suggests that an imbalance compromises or depletes critical resources – such as time and energy – thereby increasing stress (Hobfoll, 1989, 2001). Perceived organisational support is the first mediator: Organisational Support Theory suggests that when engineers experience POS, they have resources to cope with stress and experience less of it (Eisenberger et al., 1986; Rhoades & Eisenberger, 2002). Job demands are the second mediator: the JD-R model and the Demand-Control-Support model suggest that when JD are high and lead to energy exhaustion, stress increases (Bakker & Demerouti, 2007; Karasek, 1990).

The study was conducted using a quantitative post-positivistic framework. Post-positivism assumes that researchers can systematically measure the social world. However, the study recognises that the measurement results are only probable. This paradigm is appropriate for the study because it allows testing of theory-based hypotheses and estimating the structural relationships among latent variables using multivariate analysis. A cross-sectional survey design was used, with engineers registered with professional bodies recognised by the Engineering Council of South Africa (ECSA) surveyed at a single point in time. This design was appropriate because it enabled the study to estimate levels of WLB, OS, POS and JD and to test the hypothesised structural relationships among these constructs within a single empirical model.

The target group was engineers registered with various associations recognised by the ECSA (2023) in South Africa. The respondents were accessed through these professional associations, which acted as gatekeepers. However, due to limitations in disclosing membership figures and due to varying levels of cooperation from these associations, probability sampling was not used. Therefore, non-probability cluster sampling was used (Chen et al., 2015). A total of eight professional associations agreed to forward the link to their members. In total, 148 questionnaire responses were received; 143 met the inclusion criteria (engineers) and were included in the analysis. Furthermore, the sample group was largely male (86%) and included a high proportion in supervisory positions (91%). This indicates the level of seniority among the engineers in the sample. The level of experience in the sample group was high, with a notable number in the 50–59 years and 60+ years age groups. The demographic data indicate the environment in which the study on WLB and workplace stress was conducted.

The questionnaire had five sections: demographics, WLB, OS, POS and JD. The demographic section comprises age group, gender, tenure in current position, supervisory role (yes/no), highest qualification and engineering discipline (with a ‘not an engineer’ option). Work–life balance was measured with the Work and Nonwork Interference and Enhancement Questionnaire (Fisher et al., 2009; Manivannan et al., 2022). It includes four subscales: Work Interference with Personal Life (WIPL) (5 items), Personal Life Interference with Work (PLIW) (6 items), Work Enhancement of Personal Life (WEPL) (3 items) and Personal Life Enhancement of Work (PLEW) (3 items). Responses were rated on a 5-point Likert scale from 1 (‘Not at all’) to 5 (‘Almost all the time’). Occupational stress was assessed using the Work–Stress Questionnaire (WSQ), which was grouped into influence at work (4 items), indistinct organisation and conflicts (7 items), individual commitment and demands (7 items) and work-to-leisure interference (3 items). Response formats vary by subscale (2-point, 3-point and 4-point options). Items in the organisation, conflict, demands and commitment domains also include a stress appraisal rating (‘Not stressful’ to ‘Very stressful’). Perceived organisational support was measured with the 8-item short form of the Survey of POS (Eisenberger et al., 1986). The scale is unidimensional and uses a 7-point response format from 0 (‘Strongly disagree’) to 6 (‘Strongly agree’). Prior studies report strong reliability for the short form (Eisenberger et al., 1990, 1997; Rhoades et al., 2001; Worley et al., 2009).

Job demands were measured using a 6-item scale drawn from the Brief Job Stress Questionnaire, covering both quantitative and qualitative demands. Responses were recorded on a 4-point Likert-type scale from 4 (‘Agree’) to 1 (‘Disagree’) (Kawada & Otsuka, 2011).

The data collection method used an online questionnaire created in Microsoft Forms. The link and the corresponding quick response (QR) code were shared via association mailing lists and professional communication networks. Participation in the study was voluntary, and participants could opt out at any time by stopping the survey.

The study received ethical clearance from the North-West University (NWU) Economic and Management Sciences Research Ethics Committee (EMS-REC) under the study number NWU-00618-24-A4. The study has followed the ethical guidelines of the NWU and the principles of the 1964 Declaration of Helsinki and its subsequent amendments. The research participants were informed of the study’s purpose and that the data collected would be handled appropriately. No personal information was requested from research participants, and they provided informed consent before participating in the study. These steps were taken to respect the rights of the research participants and the integrity of the research process (Bryman & Bell, 2019).

Data analysis was conducted using quantitative statistical methods appropriate for hypothesis testing and mediation analysis. Descriptive statistics were first calculated, including minimum and maximum values, means, and standard deviations for each construct. This helped evaluate the levels of WLB, OS, POS and JD among engineers. Next, Spearman’s rank correlation was conducted to examine the relationship between the constructs. This statistical analysis was used in the study because the data are ordinal and do not follow a normal distribution, making it the most appropriate, according to Field (2024). In structural equation modeling (SEM), Likert-type indicators were treated as continuous variables, consistent with common practice when the response categories exceed 4 (Hair et al., 2019). To examine the mediation model, the analysis used SEM. This statistical analysis model was used in the study because it allowed simultaneous estimation of the direct and indirect effects between the latent constructs and provided a comprehensive range of indices for evaluating the model’s overall fit (Hair et al., 2019). The analysis examined the direct effects of WLB on OS, as well as the indirect effects mediated by POS and JD. The overall model fit was assessed using established criteria, including the comparative fit index (CFI) and the root mean square error of approximation (RMSEA).

In Model 1, WLB did not significantly predict JD, B = −0.182, SE = 0.144, CR = −1.261, p = 0.207. However, WLB predicted OS IAW, B = −0.281, SE = 0.116, CR = −2.431, p = 0.015, and JD significantly predicted OS IAW, B = −0.079, SE = 0.040, CR = −1.989, p = 0.047 (see Table 3). The corresponding standardised coefficients indicated a small effect of WLB on JD (β = −0.133), a moderate effect of JD on OS IAW (β = −0.262), and a strong direct effect of WLB on OS IAW (β = −0.680). Notably, the JD → OS IAW coefficient was negative after controlling for WLB. This counterintuitive sign may reflect the measurement direction (e.g. scoring where higher JD values indicate lower demand), shared variance between WLB and JD that produces suppression in the structural model, or sample-specific artefacts. Accordingly, this path should be interpreted cautiously and verified in future studies using alternative specifications and disaggregated demand indicators. Because the WLB → JD path was not statistically significant, the indirect effect via JD was negligible, and mediation was not supported (see Table 4).

In Model 2, WLB significantly predicted JD (β = −0.384, p = 0.004) and OS WLTI (β = 0.811, p < 0.05). However, JD did not significantly predict OS WLTI (β = 0.013, p = 0.875). Accordingly, the indirect effect of WLB on OS WLTI via JD was trivial (β ≈ −0.005), and mediation was not supported (see Table 5).

The results clearly show the strong inverse relationship between WLB and OS. Engineers who experienced higher levels of WLB also experienced lower levels of OS, which is in line with the boundary theory and the recovery theory, along with the Conservation of Resources theory, where the preservation of time and energy is likely to reduce stress levels (Fisher et al., 2009; Hobfoll, 1989, 2001). From a practical perspective, interventions that preserve non-work time, reduce spillover into the after-work period and enhance recovery activities are likely to reduce stress levels (Allen et al., 2000; Fisher et al., 2009). While organisational factors are likely to affect WLB and OS, the results clearly show a direct relationship between them. This suggests that interventions focusing on WLB are likely to be effective through mechanisms such as reduced role conflict, recovery and reduced ruminating, irrespective of whether the organisation is viewed positively by the worker (Hobfoll, 2001; Kahn et al., 1964). It is also worth noting that the mediation models did not fit well, which could indicate issues with the models, including that other mechanisms may not have been considered. For example, it could be the case that organisational mechanisms are more complex than the one proposed, including aspects such as supervisor support rather than the more general POS, or other resources not considered in the model, which could impact OS levels (Hair et al., 2019; Rhoades & Eisenberger, 2002).

Perceived organisational support did not mediate the WLB–OS relation. Statistically, the WLB–POS pathway did not reach statistical significance, and the indirect effect is virtually zero. Substantively, this pattern suggests that WLB may reduce stress through recovery and boundary-management mechanisms, but not through perceptions of organisational support tied to feelings of reciprocity (Fisher et al., 2009; Rhoades & Eisenberger, 2002). This pattern may also not fit some accounts of the POS-mediated effect because of the nature of the engineering work environment, in which stressors are often organised around projects and work constraints (Bowen et al., 2014a; Rothmann & Malan, 2006). In such work settings, support may not act as a mediator but rather as a resource and/or moderator that buffers work demands. It is also possible that examining support as a moderator and using more immediate measures of support could be useful avenues for future research (Bakker & Demerouti, 2007; Rhoades & Eisenberger, 2002).

In all models examined in this evaluation, JD did not mediate the relationship between WLB and OS. In all models, however, at least one component of the mediation chain was non-significant, resulting in a negligible effect. This suggests that WLB may be related to OS without significantly altering structural work conditions (Bakker & Demerouti, 2007). This pattern best fits a parallel effect in which WLB is directly related to reduced OS, and JD is an independent contributor to stress. This interpretation is conceptually consistent with the JD–R model, which posits that demands and resources can have simultaneous, partially independent effects on strain outcomes (Bakker & Demerouti, 2007). Future studies should disaggregate demands (e.g. qualitative vs quantitative) and test additional forms of resources and support to refine explanatory pathways (Kurtessis et al., 2017; Van Veldhoven, 2013).

The study offers guidelines to improve engineers’ well-being and productivity. A poor WLB is associated with increased OS, whereas a favourable WLB is associated with reduced OS and improved work performance. Work–life balance can be promoted in the workplace by offering engineers flexible working hours and the option to work from home, reducing the risk of burnout. This can be done while creating a positive perceived organisational safety climate that clearly communicates the organisation’s expectations regarding working hours, regular breaks and the protection of personal time. This approach is supported by evidence on the benefits of recovery and boundary control for stress levels (Bakker & Demerouti, 2007; Fisher et al., 2009). Perceived Organisational Support remains a key driver. When engineers are treated with respect and supportive care by their organisation, they demonstrate their ability to meet high standards. Organisations should improve managers’ effectiveness in managing staff, particularly in communication and people management. Organisational Support Theory suggests that showing care and appreciation for employees affects their attitudes positively (Kurtessis et al., 2017; Rhoades & Eisenberger, 2002). Although the JD did not provide statistical evidence of mediationin the relationship between WLB and stress, they remain a source of stress for engineers with high workloads and time constraints. Organisations should manage JD by effectively planning workloads, providing clear role definitions and aligning schedules with project cycles. Given the skills shortage in South Africa, these findings can be used to improve retention strategies and reduce mid-career exits, in line with SDGs 3 and 8.

However, there are several limitations. The study was limited to only engineers in South Africa. The study used a cross-sectional design. The study focused only on the WLB, POS, JD and OS. The study did not examine the other potential predictors of the outcome. All constructs were self-reported, increasing the risk of common method variance. Some structural models showed a weak fit, suggesting possible misspecification or omitted paths; results should therefore be interpreted with caution. Sampling relied on professional associations as gatekeepers, increasing selection bias and likely underrepresenting small contractors, remote sites and niche roles. Measurement choices also constrained interpretation: job demands were not differentiated into quantitative or qualitative or challenge or hindrance categories, and POS was measured globally rather than distinguishing between instrumental and socio-emotional support. The sample size of 143 engineers was adequate for preliminary model testing but limited for complex SEM mediation modeling; therefore, the indirect-effect findings should be interpreted cautiously and replicated with a larger sample. Future studies should use stratified, multi-stage sampling designs. Longitudinal and time-lagged designs would also strengthen causal inference. Experience sampling may also provide a better understanding of fluctuations in demand and recovery. Refining measurement should also allow differentiation between types of demands and support, as well as between support forms. Multilevel analysis and sector comparisons would also provide a clearer understanding of contextual versus generalisable patterns, thereby supporting better organisational policy development.