The study’s target group consists of permanent personnel from state-owned firms (BUMN) situated in Makassar, South Sulawesi Province, specifically PT Telkom, PT Pelindo, PT PLN and PT Pertamina. A total of 209 employees engaged in completing the research questionnaire. Respondents were recruited by purposive and snowball sampling approaches to include persons with pertinent experiences in leadership and empowerment situations (Jada & Mukhopadhyay 2019; Kundu et al. 2019). This method facilitates access to a varied sample, reflecting different organisational contexts and improving the generalisability of the results. The determined sample size surpasses 200 respondents, deemed adequate to guarantee statistical power for analyses, especially in mediation and structural equation modelling studies. This criterion corresponds with other studies that have effectively utilised analogous sample sizes to investigate similar constructs and connections (Kundu et al. 2019; Pigeon et al. 2017). Respondents were requested to take an online survey disseminated through email and professional networks. To optimise response rates, periodic follow-up reminders were dispatched. Participation in the study was completely voluntary, with informed consent acquired from all respondents. Ethical considerations were rigorously maintained, guaranteeing the confidentiality and anonymity of respondents during the data collection procedure.

The study employed well-established instruments to measure the key constructs:

Reliability was assessed using Cronbach’s alpha (CA) to ensure internal consistency, with prior studies confirming its robustness (Kundu et al. 2019; Mokhtarian & Mohammadi 2011; Prasanthi, Kumar & Mohammed 2023). Construct validity was evaluated through confirmatory factor analysis (CFA) to ensure that the factor structures aligned with theoretical components from previous research (Kundu et al. 2019; Ravelonarivo et al. 2024; Pigeon et al. 2017). Data analysis was conducted using partial least squares structural equation modelling (PLS-SEM) in SmartPLS, which was chosen for its effectiveness in handling complex models with multiple components and indicators (Valdez et al. 2023). Given that this is a cross-sectional survey, PLS-SEM was selected as the most appropriate analytical method. Additionally, bootstrapping was conducted to assess the significance of both direct and indirect effects.

This study adhered to internationally accepted ethical research principles to safeguard respondents’ rights and well-being.Full ethical approval was obtained from the Universitas Muhammadiyah Makassar Ethics Committee (approval number 015/K-7/ST/UNISMUH/II/2025, granted 26 February 2025) before data collection commenced. Participation was entirely voluntary: all respondents provided informed consent after being clearly briefed on the study’s aims, procedures, potential risks, and their right to withdraw at any time without penalty.

To protect confidentiality and anonymity, coded identifiers were used instead of personal information, and all data were securely stored with access limited to authorised researchers. Respondents were assured that their answers would be used solely for academic purposes and would not influence their employment in state-owned enterprises (BUMN) in Makassar, Indonesia. The recruitment process was free from coercion, deception, or manipulation. Throughout the research, data were collected, analysed, and reported honestly and transparently, in full compliance with the approved protocol and with academic integrity standards.

The respondents in this study comprised 209 permanent employees from four state-owned enterprises in Makassar City, South Sulawesi: PT Telkom, PT Pelindo, PT PLN and PT Pertamina. According to the company distribution, 39 employees (18.66%) were from PT Telkom, 51 employees (24.40%) from PT Pelindo, 62 employees (29.67%) from PT PLN and 57 employees (27.27%) from PT Pertamina. Respondents exhibit varied characteristics regarding age and tenure. Respondents are categorised into three primary age groups: those aged 21–30 years, predominantly present in PT Telkom and PT PLN; those aged 31–40 years, primarily located in PT Pelindo and PT Pertamina and individuals over 41 years, who are evenly distributed, particularly in PT PLN and PT Pertamina. Respondents with less than 5 years of service were predominantly at PT Telkom and PT PLN, those with 5–10 years of service were primarily at PT Pelindo and individuals with over 10 years of service were generally at PT Pertamina.

Convergent validity is assessed by comparing the outer loading values of each measurement indicator with the corresponding variable it represents. Indicators with an outer loading value of less than 0.6 demonstrate weak efficacy in representing the variables within the established model (Hair et al. 2017). The external loading values derived from each indicator in this investigation support this approach, ensuring the robustness of the measurement model and are presented in Table 1.

Figure 2 presents the partial least squares structural equation modelling (PLS-SEM) output. The five reflective indicators of empowering leadership (EL1–EL5) all load strongly on their latent construct, with t-values ranging from about 4.6 to 11.1, confirming measurement reliability. Job autonomy (JA) is measured by three items (JA1–JA3), each showing substantial loadings (≈8.4–14.9). Psychological empowerment (PE) is indicated by four items (PE1–PE4), with high t-values (≈4.6–29.3). Voice behaviour (VB) is represented by two items (VB1–VB2), again demonstrating robust loadings (≈9.6–20.3).

The outer loading values for each measurement instrument corresponding to each variable in our study model, as presented in Table 1, indicate that each instrument effectively represents its respective variable. No instrument had an outer loading value below 0.7, thus demonstrating that each instrument effectively characterises the latent variable it signifies. Based on the calculations carried out by the PLS Algorithm for the indicators in the following table, the average variance extracted (AVE) value and AVE squared value are obtained as in Table 2.

Table 2 shows that all AVE values above the required 0.5 threshold, with the lowest AVE (EL) at 0.704. Based on the loading factor values in Table 1 and AVE values in Table 2, the study meets the criteria for convergent validity. Next, construct reliability is tested using composite reliability (CR) and CA. A construct is considered reliable if both CR and CA exceed 0.6 (Hair et al. 2017). The results show a CR value of 0.7 and a CA value above 0.6, confirming that all variables – EL, JA, PE and VB – are reliable.

The concluding phase of inferential statistical analysis in this study entails hypothesis testing. The method involves comparing the t-statistic values with the critical t-table values obtained from the relationships among variables in the proposed model. The theories were developed based on theoretical principles and previous studies. This study employs a 95% confidence level, establishing the significance threshold (α) at 0.05. This threshold minimises the probability of rejecting a correct null hypothesis (Type I error), hence augmenting the reliability and validity of the results (Hair et al. 2017). Hypotheses are deemed supported if the t-statistic surpasses the crucial t-value or if the p-value is below 0.05 (Hair et al. 2017).

The investigation illustrated in Table 3, indicated that EL does not exert a statistically significant direct influence on VB. The standardised coefficient (Std. Coeff) for this association was 0.081, accompanied by a t-value of 0.969 and a p-value of 0.333. Given that the t-value does not surpass the critical threshold and the p-value is above 0.05, the alternative hypothesis H1 is not supported. This suggests that although EL may create a supportive atmosphere, it does not directly influence VB.

The findings indicate a statistically significant and affirmative correlation between EL and JA, evidenced by a standard coefficient of 0.327, a t-value of 5.448 and a p-value of 0.000. These findings corroborate H2, demonstrating that EL significantly influences employees’ perceptions of autonomy in their work. Leaders who delegate authority and encourage independence are likely to enhance JA among their subordinates. The correlation between EL and PE is also large and favourable, with a standard coefficient for this connection of 0.515, accompanied by a t-value of 11.778 and a p-value of 0.000. The findings corroborate H3, indicating that EL is a crucial catalyst for PE, allowing employees to perceive themselves as more competent, influential and autonomous in their positions.

The mediating effect of JA on the connection between EL and VB was determined to be insignificant. The standardised coefficient for this indirect impact was 0.042, with a t-value of 1.288 and a p-value of 0.198. As the t-value and p-value do not satisfy the necessary criteria for significance, hypothesis H4.a is not supported. This indicates that whereas empowered leadership increases JA, the latter does not substantially moderate the relationship with VB in this scenario. The mediating effect of PE on the association between EL and VB was insignificant. The standardised coefficient for this indirect impact was 0.077, with a t-value of 1.626 and a p-value of 0.105. As these values do not achieve statistical significance, H4.b is not supported. This suggests that while empowered leadership enhances PE, this empowerment does not significantly result in greater vocal behaviour within the parameters of this study.

Although prior research has indicated that empowered leadership promotes VB by establishing a psychologically safe and supportive atmosphere (Jada & Mukhopadhyay 2018; Zhang et al. 2020), the present investigation revealed no significant direct effect. This gap may stem from contextual factors, such as cultural or organisational dynamics, which could attenuate the impact of leadership styles. Social exchange theory posits that EL fosters reciprocal interactions, prompting people to articulate their thoughts (Qian et al. 2018). Nonetheless, this reciprocity may necessitate further mediators, such as psychological safety or trust, to materialise as voice action. Moreover, the relevance of burden or tasks, as suggested by job features theory, may influence the promotion or suppression of speech behaviour (Tian & Chae 2023). If employees view their workload as onerous or their roles as undervalued, their likelihood of exhibiting VB may decrease, even with empowered leadership present.

The notable correlation between EL and JA corresponds with SDT, which highlights the importance of autonomy in satisfying psychological needs and augmenting intrinsic motivation (Ju et al. 2019; Li et al. 2023a). Empowering leadership provides employees with increased decision-making ability and control over their responsibilities, therefore promoting JA. This association underscores the dual function of empowered leadership as a motivational asset, as delineated in JD-R theory (Li et al. 2023b). Notably, the findings support the notion that JA serves as both a driver and an outcome of EL. By permitting employees to organise their tasks and utilise autonomy, leaders foster a work atmosphere that promotes innovation and proactive behaviour (Alam et al. 2023).

The strong positive correlation between EL and PE substantiates empowerment theory, which asserts that leadership styles prioritising autonomy, support and competence augment employees’ intrinsic motivation (Ravelonarivo et al. 2024). Empowering leadership significantly enhances employees’ self-efficacy, competence and agency in their roles. Social cognitive theory clarifies this relationship by highlighting cognitive processes such as self-efficacy and perceived control as mediators of leadership efficacy (Fareed et al. 2023). Moreover, social exchange theory emphasises the significance of trust and mutual respect in enhancing the relationship between EL and PE (Lorinkova & Perry 2017).

Unexpectedly, neither JA nor PE significantly mediated the association between EL and VB. The findings indicate that whereas EL fosters autonomy and PE, other elements may be required to convert these qualities into VB. JA, as articulated by role theory, affords employees the latitude to delineate their responsibilities. The lack of substantial mediation suggests that autonomy alone may be inadequate to promote VB in the absence of psychological safety or organisational support (Cheng et al. 2024). Psychological empowerment, while closely linked to EL, may necessitate additional factors, such as inclusive leadership or high-quality LMX, to effectively facilitate VB (Dennerlein & Kirkman 2023; Younas et al. 2023).

The findings of this study have important implications for organisations, particularly state-owned enterprises (BUMN) in Makassar, Indonesia, seeking to enhance employee VB through leadership practices. While EL significantly increases JA and PE, its direct effect on VB is not evident, suggesting that additional organisational factors, such as psychological safety and supportive workplace culture, may be necessary to encourage proactive employee participation. This highlights the need for organisations to not only focus on leadership development programmes that enhance autonomy and empowerment but also create an environment where employees feel safe and supported in expressing their ideas and concerns. Managers should be trained to provide constructive feedback, actively listen to employees and foster trust to bridge the gap between empowerment and VB. Additionally, organisations should implement policies that reinforce an open communication culture and encourage proactive employee engagement. By addressing these factors, companies can maximise the effectiveness of EL in fostering a workplace culture where employees feel valued and encouraged to contribute, ultimately leading to improved organisational performance and innovation.