Social exchange theory (SET) is one of the most influential theories for examining work behaviour anchored in normative social interactions (Cropanzano & Mitchell, 2005). The theory suggests that social behaviour results from an exchange process aimed at optimising outcomes while reducing costs, with reciprocity as a central principle (Davlembayeva & Alamanos, 2023). Transformational leaders’ behaviours (e.g. inspiration, empowerment) support a high-quality leader-employee relationship characterised by trust and mutual respect (Bani Ahmad et al., 2024). Cropanzano et al. (2017) also propose that it is through such social relationships that leaders offer valuable resources, and that employees reciprocate by enacting discretionary positive behaviours, reflected in elevated job satisfaction, affective commitment and OCB. This perspective draws on SET to effectively frame the proposed mediation through goal difficulty, empowerment and performance feedback. This study, therefore, utilised SET to explore the relationship between transformational leadership, OCB and JS, as mediated by goal difficulty, empowerment and performance feedback.

Transformational leadership is defined as a leadership style that can create significant, vision-driven change in employees and an organisation. Transformational leaders display individual consideration towards employees, motivating them to perform at their best and develop their own leadership skills (Bass & Riggio, 2006). They foster a compelling vision, establish group goals and empower individuals to reach their highest potential, thereby enhancing overall commitment and performance. Transformational leaders also influence how individuals perceive themselves, inspiring them and enabling them to exceed expectations. This leadership style brings out the best in employees and motivates them to achieve optimal outcomes for the organisation (Schedlitzki & Edwards, 2014).

Research indicates that empowerment is positively related to performance feedback (Baird et al., 2020; Lee & Kim, 2020) and goal difficulty (Coun et al., 2022; Joo & Nam, 2019). Research has also empirically confirmed that performance feedback is positively related to goal difficulty (Atmaja et al., 2023; De Kleijn, 2023). Evidence from previous studies confirms that goal difficulty is positively related to performance feedback (Atmaja et al., 2023; De Kleijn, 2023) and empowerment (Coun et al., 2022; Joo & Nam, 2019).

The value of researching transformational leadership and its consequences within a company has received significant attention in the literature. The importance stems from transformational leadership’s association with important employee outcomes. It is therefore important to examine transformational leadership, because research suggests it is positively related to JS (Chi et al., 2023; Hilton et al., 2023) and OCB (Davis et al., 2022; Dafe et al., 2026). It can therefore be hypothesised that:

Previous studies have shown the relevance of focusing on empowerment, performance feedback and goal difficulty and on the consequences of these practices and why they are essential. According to the literature, it is critical to focus on empowerment and its consequences. It is essential to examine empowerment because research suggests it is positively related to JS (Akhter et al., 2021; Cruz et al., 2022) and OCB (Nauly et al., 2022; Van der Hoven et al., 2021). It can therefore be hypothesised that:

It is crucial to examine performance feedback because research suggests that it is positively related to job satisfaction (Davidescu et al., 2020; Doctor et al., 2023) and OCB (Guo & Ling, 2020; Haider et al., 2019). It can therefore be hypothesised that:

The importance stems from the association between goal difficulty and important employee outcomes. It is vital to examine goal difficulty because research suggests that it is positively related to JS (Gogoi & Baruah, 2021; Manolache & Epuran, 2023) and OCB (Faria, 2022; Ha & Moon, 2023). It can therefore be hypothesised that:

Choi et al. (2016) investigated and confirmed the relationships among perceived transformational leadership, empowerment and job satisfaction in Malaysia. This means that employee empowerment not only is indispensable for enhancing job satisfaction but also mediates the relationship between transformational leadership and job satisfaction. Saira et al. (2021) empirically tested the mediating effect of psychological empowerment in the relationship between transformational leadership and OCB. The findings of the present study indicate that psychological empowerment did mediate the relationship between transformational leadership and OCB. This finding concurs with the results of Ibrahim et al. (2024). It can therefore be hypothesised that:

The mediating role of performance feedback has been confirmed by studies that sought to explain the relationship between its variables. Results from Lee et al. (2019) found that performance feedback mediated the relationship between transformational leadership and job performance. It can therefore be hypothesised that:

According to Steinmann et al. (2018), transformational leaders influence followers’ job satisfaction and proactive behaviour through the mediating role of goal difficulty. A study by Caillier (2016) established the mediating role of goal difficulty in the relationship between transformational leadership and OCB. It can therefore be hypothesised that:

In line with the positivist paradigm, this study used a quantitative approach, utilising a cross-sectional research survey design, with a non-probability convenience sample. The micro-franchising company where the research was conducted normally comprise of the following layers of staff: The franchisor (issues the licence to trade using their name), the business manager or distributorship and franchisee, team leaders responsible for sales (80 of them in the Republic of South Africa (RSA) with R500 000.00 sales per month), managers (those with R25 000.00 sales per month) responsible for direct-selling agents (sales units). The company (population) employs 2000 managers at any given time, and this study targeted 400 managers from five distributorships within the direct-selling micro-franchising company in South Africa. The sample size was determined using the Krejcie and Morgan (1970) formula. A population of 2000 should use a sample of 322. We targeted 400 to minimise the risk of poor internet coverage and unreliable email addresses.

Team leaders sent an information sheet (invitation, informed consent and survey link) to 400 managers via email to participate in the study. The five distributorships were targeted because Internet access was better in their areas, and they were also closer to urban areas. As a result of the nature of the product sold and the customers targeted, 89% of the company’s employees are female. The researchers collected the data via a SurveyMonkey questionnaire. Two hundred and twenty-two (222: 56% response rate) managers fully completed the questionnaire. Additional features of the participants are presented in Table 1.

Table 1 shows that most respondents were female (89.2%), aged 41–50 years (32.9%), and from the Eastern Cape (33.9%).

The research variables were assessed using validated measurements from existing literature. Creditable psychometric properties were obtained when we tests the structured questionnaire with participants from the research population (n = 70) in a pilot run. Constructs were measured on a five-point Likert scale, ranging from 1 ‘strongly disagree’ to 5 ‘strongly agree’:

The study employed SmartPLS 4.1.1.8, a partial least squares structural equation modelling and bootstrapping software to analyse the proposed parallel mediation model. Partial least squares structural equation modelling is an analytical technique that prioritises repetition; thus, it maximises the variance in endogenous latent variables (Hair et al., 2019). Partial least squares structural equation modelling is suitable for exploratory research, especially for explaining the relationships among variables in a model (Hair et al., 2019; Ringle et al., 2013). Prior to data analysis, the dataset was screened for missing values and survey quality controls. The numeric marker –99 was used to signify a missing value in the data set. There were no respondents with missing data exceeding 15% of the total indicators; therefore, no cases were removed from the dataset prior to further processing. For the remaining cases, missing values were handled using mean replacement. The overall missing data accounted for less than 2.8% of the total dataset, ensuring no substantial reduction in statistical power. No IP duplications were detected.

Assessing a measurement model helps researchers evaluate how well and appropriately the model fits the reality in the field (based on collected data). This evaluation considered internal consistency reliability, which was assessed using Cronbach’s alpha and composite reliability (CR). Convergent validity was assessed by evaluating the item’s outer loading value and the average variance extracted (AVE) value of the construct. Discriminant validity was assessed using the Fornell-Larcker criterion (Hair et al., 2019) and the Heterotrait-Monotrait ratio (HTMT). In assessing the structural model, several values were considered, including standardised root mean squared residual (SRMR), exact fit criteria were used to assess model fit, the construct’s tolerance (VIF), coefficient of determination (R²), effect size (f²) and predictive relevance (Q²) value. Bootstrap confidence intervals of the indirect effects were considered to assess mediation effects.

Ethical clearance to conduct this study was obtained from the People Management Scientific Committee of North-West University Business School and the Research Ethics Committee (FEEMS REC) Faculty of Economics and Management Sciences (Ref. No. NWU-00665-22-A4).

According to Hair et al. (2019), convergent reliability indicates how well a measure correlates positively with other measures of the same construct. It is evaluated using the average variance extracted (AVE), which should be greater than 0.5, which was the case in this study.

Discriminant validity assessment has become a generally accepted prerequisite for analysing relationships between reflectively measured constructs. In the context of variance-based structural equation modelling, such as PLS-SEM, the Fornell-Larcker criterion and cross-loadings analysis are considered outdated methods for assessing discriminant validity. Discriminant validity was initially assessed by comparing an item’s cross-loading in the construct and its cross-loading in other constructs (attached as annexures). Screening of the cross-loadings confirmed that one item had the highest cross-loading on its own construct.

Discriminant validity was also evaluated using the Fornell-Larcker criterion (Hair et al., 2019). This method involves comparing the square roots of the AVE values with the correlations between the latent variable and other constructs. For discriminant validity to hold, the square root of the AVE for each construct should exceed its correlation with any other construct, as demonstrated in Table 4. Henseler et al. (2015) demonstrated that the Fornell-Larcker criterion and cross-loading approaches do not reliably detect the lack of discriminant validity in common research situations. They therefore propose an alternative approach, based on the multitrait-multimethod matrix, to assess discriminant validity: The heterotrait-monotrait ratio of correlations (HTMT). In PLS-SEM, all constructs correlate with the heterotrait-monotrait ratio (HTMT) and do not include 1.0.

Testing for common method bias (CMB) ensures that relationships in survey data are not artificially inflated or deflated by the use of the same measurement method. Researchers typically use three main statistical techniques to detect and evaluate this bias: Harman’s single-factor test, full collinearity assessment and the confirmatory factor analysis (CFA) marker technique. This study employed Harman’s single-factor test and full collinearity assessment.

Harman’s single-factor test: An exploratory factor analysis (EFA) was conducted using SPSS, forcing all study items into a single, unrotated factor. If a single factor accounts for 50% or more of the total variance, it suggests CMB is a significant issue. In this study, the first factor explained 49.2% of the variance, indicating that CMB is not an issue. This method is widely used but has been criticised for lacking statistical rigour and for providing a false sense of security.

Full collinearity assessment: This test assesses for collinearity by building an alternative model with a randomly generated marker variable and is often used in variance-based structural equation modelling (like SmartPLS 4). The variance inflation factor (VIF) for the variables used was examined, and all VIF values were less than or equal to 3.3, indicating the model is generally free from CMB (see Table 5).

Table 6 shows that TL significantly impacted EM (β = 0.512; p < 0.001; [0.389 – 0.618]), FB (β = 0.327; p < 0.001; [0.178 – 0.441]) and GD (β = 0.267; p < 0.001; [0.127 – 0.396]). Hence, H1, H2 and H3 were accepted. Table 6 also suggests that the paths from transformational leadership to the dependent variables, namely job satisfaction (β = 0.213; p = 0.225; [−0.053 – 0.227]), and OCB (β = 0.009; p = 0.893; [−0.113 – 0.151]), were insignificant. It can therefore be inferred that TL does not directly impact job satisfaction and OCB. Consequently, H4 and H5 were rejected.

Furthermore, empowerment significantly impacted job satisfaction (β = 0.462; p < 0.001; [0.198 – 0.690]) and OCB (β = 0.407; p < 0.001; [0.181 – 0.584]), and feedback significantly impacted on job satisfaction (β = 0.228; p = 0.015; [0.052 – 0.415]) and OCB (β = 0.302; p < 0.001; [0.134 – 0.464]). Accordingly, hypotheses H6, H7, H8 and H9 were supported. However, H10 was rejected because of the negative, non-significant relationship between goal difficulty (β = −0.122; p = 0.074; [−0.251 – 0.013]) and job satisfaction. In addition, goal difficulty did not significantly impact OCB (β = −0.03; p = 0.644; [−0.190 – 0.115]). Hence, H11 was also rejected.

This study was also interested in determining the indirect (mediating) effects of performance management activities such as goal difficulty, performance feedback and empowerment in the relationship between transformational leadership and job satisfaction and OCB. Bootstrap confidence intervals of the indirect effects were considered to assess mediation effects. The exclusion of zero (0) in the confidence intervals of the indirect effects, as well as significant indirect effects, was then considered as suggestive that the mediation role was confirmed. Table 7 presents the results of the mediation role of performance management activities, such as goal difficulty, performance feedback and empowerment, in the relationship between transformational leadership and job satisfaction and OCB.

Table 7 confirms the mediating roles of empowerment (β = 0.236, p = 0.002 [0.097 – 0.394]) and performance feedback (β = 0.075, p = 0.046 [0.016 – 0.161]) in the relationship between transformational leadership and job satisfaction. Table 7 also suggests that only the indirect effects of empowerment (β = 0.208; p < 0.001; [0.092 – 0.323]) and performance feedback (β = 0.099; p = 0.005; [0.038 – 0.173]) in the relationship between transformational leadership and OCB were confirmed. Hence, the acceptance of H12, H13, H14 and H15. Table 7 also suggests that goal difficulty (β = −0.033, p = 0.106 [−0.082 – −0.001]) did not significantly affect the relationship between transformational leadership and job satisfaction. Goal difficulty (β = −0.010; p = 0.670; [−0.059 – 0.031]) also did not significantly affect the relationship between transformational leadership and OCB, hence, the rejection of H16 and H17. All accepted hypotheses represent full mediation models because the direct relationships between the independent variable and the model outcomes (employee attitudes) were found to be insignificant.

The significance of this study can be viewed from different perspectives. This study is based on and underpinned by the expectations and assumptions of the social exchange theory (SET), which is one of the most reputable psychological theories for clarifying exchanges and interactions between individuals and their organisations (Ng, 2023). This theory explains the development and demonstration of positive attitudes among employees toward their employers and organisations. The theory posits that the employer–employee relationship is determined by employers’ perceptions of their benevolence, as reflected in the provision of employee-centred transformational leadership and favourable job conditions (Ahmad et al., 2022), such as empowerment, performance feedback and goal difficulty. Hence, an employee will reciprocate positively when such favourable conditions are implemented in their workplace.

Additionally, South African studies on transformational leadership are primarily focused on large corporations. This study extends the body of evidence on transformational leadership to the entrepreneurship industry. Moreover, although transformational leadership research is well established, there remains a need for research exploring the contextual mechanisms by which transformational leadership influences employee attitudes (Dahleez et al., 2021).

This study introduced empowerment, performance feedback and goal difficulty (performance management activities) as mediators and confirmed their relationships with transformational leadership, job satisfaction and OCB.

Firstly, the results show that transformational leadership does not have a direct relationship with employee outcomes (job satisfaction and OCB). Instead, the relationship exists indirectly through the mediators (empowerment and performance feedback). This suggests that direct-selling organisations should focus less on promoting transformational leadership as a broad leadership style and instead on specific motivators that convert to employee outcomes. In practice, leaders need to empower their employees by involving them in decision-making, giving them autonomy and promoting task ownership (Bose et al., 2021). This is especially important in a direct-selling context, where individuals are self-motivated and function independently.

Secondly, the findings indicate that performance feedback is a notable mediator of both relationships: Between TL and job satisfaction and between TL and OCB. This underscores the importance of continuous, well-structured communication between employers and employees (Zheng et al., 2012). There is not much formal, structured supervision in the direct-selling environment, so regular feedback will help maintain engagement, clarify roles and strengthen desired behaviours. Organisations can enhance the quality of performance feedback by providing communication training courses, coaching programmes, job aids and learning projects to their managers and employees (Zheng et al., 2012).

Thirdly, this research suggests that goal difficulty can sometimes have counterproductive effects, especially on job satisfaction. This proposes that unreasonably high-performance targets that do not align with employees’ abilities may negatively affect employee well-being in direct-selling settings, where the pressure to perform is already high.

Fourthly, as a result of the specific nature of the sample, these practical implications are most relevant to direct-selling companies characterised by independent, results-driven frameworks. Care should be taken not to generalise these results to conventional organisational settings.

The researchers accept that the present study was not without its faults. The quantitative cross-sectional research approach examined transformational leadership, mediators and employee outcomes at a single point in time; thus, a cause-and-effect relationship could not be tested (Schofield, 2009). Although relationships were established, the cross-sectional design means that the sequence of events, reciprocal effects or other factors not included in the study can also not be ruled out. Future studies could overcome this limitation by using longitudinal or experimental research designs investigating the effect of transformational leadership on employee outcomes through specific mediators over time.

Moreover, as the study used a convenience sampling approach involving managers from a direct-selling company in South Africa, generalising the findings to other organisational settings and industries is unfortunately not possible. In addition, the selected company specialises in the direct sale of food storage containers and therefore has predominantly female sales agents and managers, further limiting the generalisability of the sample. The relationships and leadership dynamics in a direct-selling setup may differ from those in conventional corporate structures.

Additionally, the study’s relatively small sample size increased the risk of sampling and volunteer bias, as those who chose to complete the survey questionnaire may have had different views from those who chose not to participate. Future studies should consider larger and more diverse samples drawn from different organisations in South Africa. Probability-based sampling or a mixed-methods approach could also be used to enhance external validity and the study’s broader relevance.

As the study employed a self-report survey method, issues related to measurement validity and CMB may have influenced the findings. The positive association between transformational leadership, mediators and employee outcomes may be partly because of single-source bias rather than a true relationship. While incorporating individual variables such as age, gender and education as mediating or moderating factors in future studies could add more context, it will, unfortunately, not mitigate method bias. This study employed Harman’s single-factor test and a full collinearity assessment to rule out the presence of CMB. The researchers also suggest that this design-related shortcoming be tested for in all future studies of this nature.

Future research could consider other procedural and statistical approaches to mitigate CMB, such as removing common-scale properties, balancing positive and negative items, or including reversed-coded items. Researchers could also collect data from multiple sources, such as managers and sales agents; use temporal separation between predictor and employee outcome measures; include the unmeasured latent factor technique or possibly add a confirmatory factor analysis marker technique as suggested by Podsakoff et al. (2014). The measurement instruments also need to be applied to other direct-selling contexts, using larger sample sizes in South Africa, to ensure the results are generalisable in this sector of the country.

Further studies could also extend the conceptual framework of the research by examining other situational (e.g. trust, communication and tools of the trade) and dispositional factors (e.g. locus of control, self-efficacy and self-esteem) as well as other leadership styles, such as servant leadership, authentic leadership or transactional leadership, to investigate whether similar correlations develop across different leadership approaches.