The role of complexity science in advancing safety leadership and safety compliance in safety-critical environments

In recent years, the number of accidents in the oil and gas industry has risen drastically and various hazardous occurrences have resulted in severe disasters (Bensonch et al., 2022). Although major industrial incidents are caused by many factors (e.g. operational, behavioral and technological factors), some of these factors are more common than others (Bensonch et al., 2022). A study by Willamson and Feyer (1990) recognized that 91% of the occupational fatalities involve behavioral factors. Along the same line, according to annual reports from the International Association of Oil and Gas Producers on accident investigations, lack of safety compliance is recognized as the most common and primary cause of accidents in the oil and gas industry (IOGP, 2023). Safety compliance is the degree to which individuals or workers comply with safety standards, rules, terms and conditions and regulations at the workplace (Masia and Pienaar, 2011). As safety within the oil and gas industry is highly regulated and all work operations are virtually governed by rules and procedures, a high level of safety presupposes a high level of safety compliance (Kvalheim and Dahl, 2016). Therefore, various strategies for improving safety compliance in oil and gas companies need to be examined.

Companies’ leaders and project management teams in oil and gas organizations have a legal responsibility to ensure that their organization is compliant with an extensive system of legal safety requirements, safety standards, guidance on safety practices and specific codes of practices. Leaders establish safety vision, practices and procedures, model safe behaviors, provide resources and training for employees and promote open communication about safety issues. As a result, organizational safety compliance is built on individual safety compliance through the agency of the leader and the practices that they enact (Pilbeam et al., 2016). Many studies (e.g. Dahl and Olsen, 2013; Clarke, 2013) considered safety leadership as a prerequisite for safety compliance among employees and safety performance improvement, thereby attempting to examine the effects of various leadership styles and behaviors of the leaders on employees’ safety compliance. Accordingly, company leaders, managers and supervisors have traditionally adopted a transactional-transformational perspective of leadership stressing the “command and control” form of regulation to ensure safety compliance among employees, which can be evident in the production of standard operating procedures and accompanying checklists for audit purposes (Fernández-Muñiz et al., 2017).

Nonetheless, achieving compliance depends not only on the knowledge, understandings and skills of the individuals but also on individual motivations (Tyler, 2011). The effectiveness of the traditional safety leadership approach on safety compliance has been challenged by the inherently different levels of individuals’ motivations to conform to safety requirements in dynamic work contexts. A review by Baltaci and Balcı (2017) indicated that previous leadership studies and practices have been developed based on the “command and control” form, thereby limiting its effects and degrees of influence on individuals’ adherence to safety regulations and standards. As a result, workers may comply with organizational rules to keep their jobs, receive rewards or avoid negative consequences of noncompliance (Hu et al., 2020). Existing safety leadership research predominantly conceptualizes safety compliance as a linear outcome of leadership behaviors, providing limited insight into how compliance develops through dynamic interactions among workers, leaders and organizational conditions (Ojuola, 2020). Thus, the findings from previous studies provide a limited understanding of the underlying mechanisms that drive individuals’ adherence to safety requirements in complex and adaptive work environments. Consequently, despite substantial efforts in improving safety leadership (e.g. Clarke, 2013; Fernández-Muñiz et al., 2017), the traditional safety leadership approach and practices may be insufficient to achieve a sustained level of safety compliance at the workplace. There is a need for an alternative theoretical perspective to better foster safety compliance in workplaces.

Complexity science principles (CSPs) provide a more effective and innovative framework for leadership in complex adaptive systems (CAS) (Weberg, 2012). Conventional management and leadership principles emphasize planning, directing and controlling, treat organizations as machine-like systems and manage the components properly to achieve the desired outcomes. While such a mechanical leadership approach works well under the same work conditions, they fail to equip organizations to adapt to dynamic environment or respond effectively to emerging challenges (Zimmerman et al., 1998). In contrast, CSPs regard leadership as a process that shapes the conditions in which CAS can evolve, generating adaptive, innovative and creative outcomes (Burns, 2001). Accordingly, leadership is distributed across networks of individuals, emphasizing adaptability, flexibility and co-creation rather than residing solely in formal authority or predefined roles (Mitleton-Kelly, 2003). By focusing on interactions, feedback loops and emergent patterns, CSPs enable organizations to leverage uncertainty as a source of innovation rather than treating it as a threat to be eliminated. This approach addresses the critical limitations of conventional leadership, which struggles with diverse motivations, dynamic work environments and unpredictable safety challenges. Specifically, the five principles of complexity science are:

1. connectivity;

2. interdependence;

3. feedback;

4. exploration-of-the-space-of-possibilities (i.e. creative exploration); and

5. coevolution (i.e. mutual adaptation and development) (Mitleton-Kelly, 2003).

By applying CSPs, leaders can cultivate adaptive networks and collaborative problem-solving processes, making it distinctly effective in contexts where conventional, command-and-control approaches fail. Therefore, CSPs have been proposed as a promising approach to modern organizations facing technological, organizational, social and political complexity work for achieving sustained safety performance (Ojuola, 2020).

Several studies introduced complexity science and examined its applications in the health and safety domain (Mckeon et al., 2006; Ojuola et al., 2020; Smaggus, 2019). Smaggus (2019) discussed the notion of complexity science to create high levels of safety and quality services through adaption, improvision and dedication. In clinical nursing, Mckeon et al. (2006) suggested that complexity science forms the basis for high-reliability teams to recognize even the most minor variances in expected outcomes and take strong action to prevent serious errors from occurring. Ojuola et al. (2020) put forward a framework for safety leadership functions using complexity science, and found that complexity science supports the flexibility, innovation and dynamism of leadership. A recent study by Mohamed et al. (2025) examined the interactions of safety leadership and CSPs, and confirmed that more effective safety leadership behaviors via one or more CSPs entail higher levels of safety participation in the oil and gas companies. Unfortunately, while it is acknowledged that safety compliance and safety participation are two key components of employees’ safety behaviors, safety compliance has not yet been investigated within the context of the interplay between leadership and CSPs. It remains unclear whether CSPs can explain how leaders enhance safety compliance in a uniquely effective manner within safety-critical organizations. Therefore, this study moves to the next phase of scientific inquiry to test the interactive effects of safety leadership and CSPs on the level of safety compliance in the oil and gas industry. The research findings are expected to clarify the critical role of CSPs in advancing safety leadership and safety compliance. Following this, the prospect of achieving sustained improvements in safety compliance will be discussed.

Safety leadership generally refers to a set of leadership behaviors that influence subordinates’ behaviors to attain particular safety goals (Li et al., 2020). Specifically, safety leadership can affect subordinates’ behaviors in handling safety issues in both direct and indirect ways. The direct ways could relate to their reinforcement of employees’ safe behaviors through monitoring and control. The indirect ways could be the establishment of norms relating to safety practices and procedures, thus cultivating a particular safety culture. As a result, these leadership behaviors directly and indirectly influence subordinates’ expectations and motivations, thus influencing subordinates’ safe or unsafe behaviors (Li et al., 2020).

A recent review by Adra et al. (2024) indicated that transformational and transactional leadership styles have been extensively studied and established as solid theories within the realm of leadership literature. Transformational leadership style for safety involves using inspirational strategies such as expressing satisfaction when jobs are performed safely; rewarding achievement of safety targets; continuous encouragement for safe working; maintaining a safe working environment (Kelloway et al., 2006). In addition, transactional leadership style in relation to safety typically establishes appropriate safety goals, monitors performance toward these goals and rewards behaviors that sustain or improve safety practices (Kapp, 2012). Drawing upon transformational and transactional styles of leadership, safety leadership has been examined and recognized as a multidimensional concept in safety literature (Adra et al., 2024). These dimensions can include safety caring, safety controlling, safety motivation, safety inspiring, safety coaching, safety concern, safety policy, etc. (Oswald et al., 2022; Subramaniam et al., 2023).

A recent study by Mohamed et al. (2025) developed a scale for assessing safety leadership, and confirmed three safety leadership dimensions incorporating elements of both transformational and transactional leadership behaviors. They include empowering and engaging; modeling and reinforcing; and promoting and enabling. Empowering and engaging (EE) refers to a process where power is shared, and followers are given more autonomy and responsibilities through activities that enhance the meaningfulness of work, encourage participation in decision-making, express confidence in high performance and provide freedom from bureaucratic constraints (Amundsen and Martinsen, 2014). Modeling and reinforcing (MR) refers to setting an example for others to follow (Williams, 2010). In essence, if leaders expect respectful communication among team members, they must exhibit respectful communication themselves. Finally, Promoting and enabling (PE) represents leadership that actively fosters a safe work environment (Barling et al., 2002). When leaders engage in behaviors that promote and enable safety, employees perceive a positive safety climate, leading to increased engagement in safety practices and a reduction in injuries and accidents due to heightened awareness and focus on safety (Liu et al., 2025).

Since the introduction of the Health and Safety at Work Act, approaches to managing occupational health and safety have evolved through a series of ages, each emphasizing different aspects of the system: the technical, human factors, sociotechnical and organizational culture. The technical age attributed accidents mainly to mechanical failures; the human factors age focused on operator error; the sociotechnical age recognized the interaction between human and technical elements; and the age of organizational culture emphasized the influence of shared workplace culture and collaboration on safety (Harvey et al., 2016). Most recently, the adaptive age has emerged with growing evidence in safety research suggesting that accidents are caused by the interactive complexity of the system itself. Accordingly, uncertainty, complexity and contradictory requirements are inherent in CAS (Woods et al., 2017).

Complexity science theory seeks to foster CAS dynamics while simultaneously enabling control structures appropriate for coordinating formal organizations and producing outcomes appropriate to the vision and mission of the system (Weberg, 2012). Complexity science provides a more useful framework for safety management in CAS (Reiman et al., 2015). In complexity science-based view, safety management has its five components. First, complexity science suggests that organizations operate through multiple, sometimes competing, logics, requiring management to remain adaptive and balance these dynamics according to situational demands. Second, rather than relying solely on control and command over employees, management should concentrate on establishing the preconditions that enable safety and strengthen the organization’s capacity for its safety performance. Third, because CAS are inherently uncertain, effective safety management requires continuous adaptation to evolving constraints and emerging opportunities. Fourth, variations within the system can serve as drivers of learning and development, meaning that variability in CAS can stimulate innovation. Finally, leadership is distributed and leaders are participants within the system and must recognize that other individuals also shape the system’s direction and outcomes (Reiman et al., 2015).

A study by Mitleton-Kelly (2003) outlines five CSPs, which can be categorized into three areas:

1. relationships between agents (i.e. connectivity, interdependence and feedback);

2. patterns of behavior (i.e. creative exploration); and

3. enabling functions (i.e. mutual adaptation and development).

In terms of relationships between agents, leaders can foster connections, build trust and facilitate effective feedback to solve organizational issues based on the principles of connectivity, interdependence and feedback within complex systems. In terms of patterns of behavior, leaders can encourage team members to try new strategies based on the creative exploration principle. Rather than relying heavily on pin-point forecasting, top-down planning or elaborate controls, leaders acknowledge the value of exploring various strategies for organizational development. Finally, enabling functions illustrate how leaders act as agents of change within an organization guided by the principle of mutual adaptation and development. Leadership is influenced by unique factors stemming from specific tasks, professional affiliations and roles within the organization. Analyzing how such mutual adaptation and development change can provide insight into the most suitable fit and collective leadership capacity (Mitleton-Kelly, 2003).

As discussed earlier, safety leadership is demonstrated to predict staffs’ safe or unsafe behaviors through various factors such as cognition, personality, safety knowledge and motivation and well-being (Jung et al., 2020; Liu et al., 2025; Yang et al., 2024). Nonetheless, previous studies only focused on leaders’ influences, while overlooking the dynamic interactions and feedback loops among employees, leaders and organizational conditions associated with the increasing complexity of safety-critical environments. For instance, Jung et al. (2020) found that employees may experience negative interactions with colleagues, supervisors and managers, characterized by distrust, isolation and resistance resulting from psychological states such as anxiety and avoidance. As a result, these workplace dynamics can weaken the effectiveness of safety leadership and hinder improvements in safety compliance. Consequently, there is still limited understanding of how safety compliance can be effectively achieved in such work environments.

The above review of safety leadership theory also indicates the three dimensions of safety leadership (i.e. EE, MR and PE) (Mohamed et al., 2025). These leadership behaviors are expected to improve employees’ safety compliance. A comparison of safety leadership and complexity science reveals that both concepts deal with how people behave in organizations. In addition, complexity science theory enriches leadership theory by viewing safety-critical organizations as CAS where safety outcomes emerge from dynamic interactions, feedback loops and nonlinear change. Complexity science theory implies that leadership can arise from distributed interactions, not just from formal authority, therefore proposing five principles (i.e. connectivity, interdependence, feedback, creative exploration and mutual adaptation and development) aimed at enhancing leadership behaviors. It is implied that when leaders, managers and supervisors demonstrate stronger leadership behaviors through the adoption of CSPs, higher levels of safety compliance among employees are more likely to be achieved. Specifically, while safety leadership directly influences safety compliance, the adoption of one or more CSPs may help create psychological conditions and motivation that encourage employees to comply with safety requirements. In other words, the three dimensions of safety leadership are predictors of each of the CSPs, which, in turn predict levels of safety compliance. Based on this argument, this study proposes three dimensions of safety leadership as antecedents of safety compliance, where each of the five CSPs serves as the mediator of these relationships. Hence, it is reasonably hypothesized that five CSPs mediate the relationships between three dimensions of safety leadership and safety compliance. The following hypotheses are set out:

Figure 1 depicts the research hypotheses proposed in this study.

A survey research design and a quantitative approach were adopted in this study for two main reasons:

1. The research objective required examining the variation in key variables (i.e. safety leadership, CSPs and safety compliance) across multiple cases (e.g. oil and gas projects).

2. The hypotheses sought to identify the patterns of association among variables. Accordingly, data were collected primarily in quantitative form.

Based on the hypotheses, three main research variables are identified: safety leadership, CSPs and safety compliance. The survey was therefore conducted in the form of a questionnaire targeting professionals working in the oil and gas industry with safety responsibilities. The questionnaire consists of four primary sections. Section 1 aimed to collect general details on the respondents and the characteristics of their organization. Another three sections aimed to measure the three constructs incorporated in this research.

To measure the three dimensions of safety leadership, 13 measurement items developed by Mohamed et al. (2025) were used. Accordingly, respondents were required to indicate their level of agreement regarding the leadership in their organization using a five-point Likert scale (between 1 = strongly disagree and 5 = strongly agree).

To measure the five principles of complexity science, 28 measurement items developed by Mohamed et al. (2025) were used. Accordingly, respondents were required to rate the impact of leadership actions in their organization using a five-point Likert scale (between 1 = very low and 5 = very high).

To measure safety compliance, the questionnaire contained 6 measurement items. It was based on a review of previous studies (Clarke, 2013; Martínez-Córcoles et al., 2013). Respondents were therefore required to indicate their level of agreement regarding employees’ compliance using a five-point Likert scale (between 1 = strongly disagree and 5 = strongly agree).

Content validity and reliability of the data collection instrument were assured by a pilot study in which preliminary questionnaires were sent to four academics and industry experts working in the field:

• to identify whether the questions effectively measured the intended constructs; and

• to examine whether the questions were worded appropriately unambiguous and easy to understand.

Comments included clarifying several terminologies and modifying ambiguous questions. The questionnaire was then amended based on the experts’ feedback.

Sampling techniques are generally classified into two categories: random sampling and nonrandom sampling. This study required a project-level analysis, with the target population comprising oil and gas projects worldwide. However, random sampling was not feasible because the exact population could not be clearly identified. Therefore, judgmental sampling was used to obtain the research data as recommended by Sekaran and Bougie (2016). Judgmental sampling is a nonprobability sampling technique wherein the sample members are chosen only based on the researcher’s knowledge and professional judgment. Accordingly, two selection criteria were followed:

1. The main form of business must be oil and gas; and

2. The project must have at least 50 site staff members.

In addition, a “key informant” approach was adopted to select potential respondents in the survey. This method is appropriate when respondents can be identified based on their organizational roles and are able to provide informed opinions and insights that accurately reflect the perspectives of other key decision-makers within the organization (Phillips, 1981). In this study, respondents were professionals with safety-related responsibilities and were therefore specifically targeted for data collection. The minimum sample size of 70 was considered adequate for the modeling purposes in this study as suggested by Aibinu et al. (2008).

Ethical approval was granted by the Griffith University Human Research Ethics Committee. The data collection procedure involved three steps. In the first step, potential participants were contacted. Accordingly, an online survey administered via EnergyPeople was used as the platform of survey questionnaire delivery for this research (Energy People Group, 2023). The second step involved reaching out to potential respondents via email and inviting them to participate in this study. In the third and final step, respondents were instructed to perform the following tasks:

• Review the cover letter and the Participation Information Sheet and give their consent.

• Complete the survey based on the implemented safety practices and project records.

A total of 255 completed responses were obtained from the prospective participants. To ensure data validity, all received responses were carefully examined to eliminate any duplicates originating from the same project location or projects with fewer than 50 staff members. Through this process, 82 invalid responses were identified and removed from 255 responses received. As a result, 173 responses were considered valid for final data analysis, representing 67.84% of the total collected number. The data of this study were mostly collected from well-grounded organizations existing for more than 11 years (89%). In terms of the organization types, most organizations had a presence internationally (78%). Most projects were privately owned (95%) and had 50–100 staff (74%). The profile of respondents indicates that most respondents are project managers (73%) and have at least five years of experience (87%).

Structural equation modeling (SEM) is a model that encompasses various data analysis methods such as multiple regression, analysis of variance, confirmatory factor analysis (CFA) and path analysis (Bowen and Guo, 2011). In this study, SEM was adopted because it enables the simultaneous examination of relationships between latent constructs and their measurement items via CFA and the structural relationships among multiple constructs via path analysis. Specifically, CFA was used to confirm the reliability and fitness of the factor structure of the constructs involved in this study (i.e. safety leadership, CSPs and safety compliance) before performing the substantive analyses. It concerns the adequacy of individual sets of indicators in capturing their related constructs by evaluating the internal consistency reliability, convergent validity and discriminant validity of constructs specified (Bowen and Guo, 2011). To establish internal consistency reliability, convergent validity and discriminant validity, any inconsistent or insignificant indicators were considered to be removed according to a set of rules. For indicator reliability, indicators with low loadings (below 0.5) should be removed from the construct; for internal consistency, the composite reliability score should be above 0.70; for convergent validity, average variance extracted (AVE) scores should be higher than 0.5; for discriminant validity, the square root of the AVE score of each construct should be higher than its correlation with any other construct (Hair et al., 2006). In addition, the maximum likelihood method was used to measure the measurement model parameters and fit indices (Hair et al., 2006). Accordingly, a measurement model is considered acceptable when its fit indices meet the following criteria: (1) $χ2/df<3.0$⁠, (2) GFI, TLI, CFI and IFI > 0.9 and (3) RMSEA < 0.08 (Hair et al., 2006).

Path analysis of SEM concerns an examination of multiple structural relationships between constructs via assessment of explanatory power and path coefficients (Bowen and Guo, 2011). In this study, the hypotheses propose that the dependent variable (i.e. SC) is influenced by three independent variables (i.e. EE, MR and PE) through a set of five mediators (i.e. CON, INT, FBK, EXP and MAD). Therefore, several SEM models were developed to test the research hypotheses. For example, to examine the mediating effect of CON on the relationship between EE and SC (H1.1), two SEM models were developed. Accordingly, the total effect of EE on SC must be first examined via a nonmediated SEM model. If this effect was confirmed as statistically significant, another structural model was then developed to test whether the effect of EE on SC is mediated/transmitted by CON.

The results of validity and reliability are presented in Table 1. After a removal of inconsistent items, all factor loadings exceed 0.5, demonstrating the high level of convergent validity. All R² values of measurement items surpass 0.5, showing that the indicator reliability was acceptable. The results also show that all composite reliability scores of the constructs were higher than 0.7 and AVE scores exceeded 0.5. Thus, the measurement items were appropriate for their respective constructs. In addition, Table 2 presents the results of testing measurement model parameters and fit indices. The results confirm the reliability and validity of the measurement models for structural model evaluation.

The total effects of each dimension of safety leadership on safety compliance were tested using a nonmediated SEM model. The results show that the positive correlations between:

• EE and SC (β = 0.183*, t-value = 1.875, p = 0.061);

• MR and SC (β = 0.226**, t-value = 2.745, p = 0.006); and

• PE and SC (β = 0.205*, t-value = 1.928, p = 0.054) are statistically significant.

The results provide evidence that each dimension of safety leadership positively and directly predicts safety compliance. These results also provide a basis to examine whether and how these effects are mediated by various CSPs.

Mediating effects of CSPs on the relationships between EE and SC were examined by its corresponding SEM models and reported in Block 1 of Table 3. The results show that each of CSPs predicts SC (p < 0.001). Furthermore, EE has indirect relationships with SC, mediated by CON (β = 0.111*, p < 0.1), INT (β = 0.103*, p < 0.1), EXP (β = 0.116*, p < 0.1) and MAD (β = 0.115*, p < 0.1). EE has indirect relationships with SC, mediated by FBK (β = 0.085**, p < 0.05). The percentage of overall change (R²) in SC due to EE was observed via CON (33.2%), INT (37.5%), FBK (37.4%), EXP (37.8%) and MAD (31.9%), indicating a large effect size. These results indicate that each of CSPs partially mediates the relationships between EE and SC. This provides empirical evidence to support H1.1, H1.2, H1.4 and H1.5 at p < 0.1; and H1.3 at p < 0.05.

Mediating effects of CSPs on the relationships between MR and SC were investigated by its corresponding SEM models and reported in Block 2 of Table 3. The results show that each of CSPs predicts SC (p < 0.001). There are statistically significant and positive correlations between:

• MR and CON (β = 0.176**, p < 0.05);

• MR and INT (β = 0.126**, p < 0.05);

• MR and FBK (β = 0.208**, p < 0.05); and

• MR and EXP (β = 0.135*, p < 0.1).

No significant relationship was found between MR and MAD (p > 0.1). Furthermore, MR has indirect relationships with SC, mediated by CON (β = 0.136**, p < 0.05), INT (β = 0.158**, p < 0.05) and EXP (β = 0.152**, p < 0.05). MR has indirect relationships with SC, mediated by FBK (β = 0.110*, p < 0.1). The percentage of overall change (R²) in SC due to MR was observed via CON (33.8%), INT (38.9%), FBK (37.8%) and EXP (38.8%), indicating a large effect size. These results indicate that most CSPs (i.e. CON, INT, FBK and EXP) partially mediate the relationships between MR and SC. Therefore, H2.1, H2.2 and H2.4 are supported at p < 0.05; H2.3 is supported at p < 0.1; whereas H2.5 is not supported at p < 0.1.

Mediating effects of CSPs on the relationships between PE and SC were examined by its corresponding SEM models and shown in Block 3 of Table 3. The results show that each of CSPs predicts SC (p < 0.001). In addition, PE has indirect relationships with SC, mediated by CON (β = 0.078**, p < 0.05), INT (β = 0.048**, p < 0.05) and MAD (β = 0.128**, p  < 0.05). PE has indirect relationships with SC, mediated by FBK (β = 0.012*, p < 0.1) and EXP (β = 0.103*, p < 0.1). The percentage of overall change (R²) in SC due to PE was observed via CON (32.6%), INT (36.7%), FBK (36.7%), EXP (37.6%) and MAD (32.3%), indicating a large effect size. These results indicate that each of CSPs partially mediates the relationships between PE and SC. This provides empirical evidence to support H3.1, H3.2 and H3.5 at p < 0.05; H3.3 and H3.4 at p < 0.1. The statistical results are discussed in the next section.

The results show that empowering and engaging leadership behaviors are positively correlated to safety compliance (β = 0.183*, t-values = 1.875, p = 0.061). This result indicates that staff exhibit greater compliance with safety rules, requirements and procedures when their leaders more effectively encourage participatory management and empower team members to enhance workplace safety. This finding is consistent with previous studies (Bryden, 2002; O'Dea and Flin, 2001). The study of Bryden (2002) recognized engaging employees with relevant experience in decision-making as one of the critical senior managers’ behaviors for safety in an oil company. In another study, O'Dea and Flin (2001) suggested that safety performance can be improved when leaders promote workforce involvement and empowerment in safety activities.

The finding of this current study advanced the theories of leadership and safety management by giving further insights into the effects of leadership behaviors on safety compliance, where the results of mediated analysis confirmed that the effects of empowering and engaging leadership behaviors on safety compliance are partially transmitted by all CSPs (i.e. CON, INT, FBK, EXP and MAD) (Block 1 of Table 3). In particular, EE has significant indirect effects on SC via CON (β = 0.111*, p < 0.1), INT (β = 0.103*, p < 0.1), EXP (β = 0.116*, p < 0.1), FBK (β = 0.085**, p < 0.05) and MAD (β = 0.115*, p < 0.1). This result indicates that some of the effects of empowering and engaging leadership behaviors on safety compliance are direct, while some are indirect. In other words, some of these effects may not produce a direct effect on the level of safety compliance, while they contribute to the cultivation of Connectivity, Interdependence, Feedback, Creative exploration and Mutual adaptation and development in an organization, all of which then transmit the effects of such safety leadership behaviors to staff’s safety compliance.

The two direct and indirect effects of empowering and engaging leadership behaviors on staff’s safety compliance could be explained by Emotional Labor Theory (ELT), which suggests that individuals have different intentions when conforming to organizational requirements, and as a result of those different intentions, use two different approaches (i.e. surface vs deep) to address the demands imposed by the organization (Steinberg and Figart, 1999). While a surface approach involves the deployment of minimal effort directed to display the required behavior at a superficial level, a deep approach requires the deployment of cognitive and attentional effort toward the attainment of the organizational outcomes. Based on ELT, individuals may display surface compliance when they perceive that complying with organizational requests will lead to desirable outcomes or rewards or avoid negative consequences of noncompliance. This means an employee may outwardly perform and follow the rules without truly believing in their importance. In a different manner, deep compliance may occur when individuals find it personally meaningful and enjoyable to comply with rules even in the absence of consequences or rewards. In the context of this study, it is possible that traditional empowering and engaging leadership behaviors are more likely to direct its effect on staff’s surface safety compliance, while the inclusion of a complexity perspective and its principles set the stage for such empowering and engaging leadership behaviors to indirectly influence the staff’s deep safety compliance. More specifically, when complexity perspective and its principles are embedded in those leadership behaviors, staff experience a sense of purpose, connection and ownership in following safety rules. Instead of viewing safety compliance as an external demand, staff see it as an integral part of their contribution to a supportive and evolving workplace. This makes safety compliance more intrinsically rewarding, fulfilling and sustained.

The results show that modeling and reinforcing leadership behaviors are positively correlated to safety compliance (β = 0.226**, t-values = 2.745, p = 0.006). This result indicates that staff exhibit greater levels of safety compliance when their leaders and managers have a more visible and consistent commitment to safety and participate more in safety activities. This finding is consistent with previous studies (e.g. Wu et al. 2015), which suggested that leaders, through their behaviors, can lead by example and influence subordinates with charisma and high-level morality, ultimately convincing them that safety is the top priority and inspiring them to strive for safety excellence.

Furthermore, this study offers insight into the effects of modeling and reinforcing leadership behaviors on safety compliance, where the results of mediated analysis confirmed that most CSPs (i.e. CON, INT, FBK and EXP) partially mediate the relationships between MR and SC (Block 2 of Table 3). In particular, MR has significant indirect effects on SC via CON (β = 0.136**, p < 0.05), INT (β = 0.158**, p < 0.05), FBK (β = 0.110*, p < 0.1) and EXP (β = 0.152**, p < 0.05). This result indicates that modeling and reinforcing leadership behaviors influence safety compliance through both direct and indirect effects. The direct effects are supported by several researchers (Bandura, 1977; Casey and Krauss, 2013). Bandura’s (1977) social learning theory (SLT) proposes that individuals learn behaviors through observing others, imitating their actions and experiencing reinforcement. Based on SLT, it is implied that leadership behaviors directly influence staff’s safety behaviors through modeling, reinforcement and social interactions. Casey and Krauss’s (2013)’s study also found that when managers and supervisors emphasize the importance of safe work practices and consistently demonstrate them through their actions, these safe work behaviors are reinforced and modeled for employees.

The indirect effects of modeling and reinforcing leadership behaviors on safety compliance could be explained by normative social influence theory (NSIT) as proposed by Latané and Wolf (1981). While SLT suggests that observing others’ behaviors increases the likelihood of imitation, Normative Social Influence occurs when individuals conform to group expectations more indirectly. According to NSIT, individuals tend to align with the expectations, rules or norms of a group to gain approval, avoid disapproval and foster a sense of belonging. This theory is also supported by Borsari and Carey (2003), where it is found that students often overestimate the prevalence and approval of alcohol use among peers. As a result, students are less likely to perceive their alcohol use as problematic, and feel pressure to use alcohol to gain acceptance by peers. This evidence highlights the power of Normative Social Influence in shaping individuals’ normative beliefs and related behaviors. In the context of this study, it is possible that while traditional modeling and reinforcing leadership behaviors shape staff’s safety compliance through learning and imitating, incorporating a complexity perspective and its principles can further enhance a sustained level of safety compliance. This is because staff are not only influenced by direct learning but also by a desire to fit in and gain acceptance from their managers and leaders, fostering a more sustained commitment to workplace safety.

However, H2.5 failed the test, suggesting that MR is not an effective driver of staff’s safety compliance through MAD. One possible reason for this is that MR primarily relies on leaders leading by example, reinforcing employees’ commitment to safety by ensuring consistent adherence to safety protocols and procedures. In contrast, MAD positions leaders as change agents, emphasizing innovation and continuous improvement. The fundamental difference between these approaches, where MR prioritizes stability and compliance, while MAD promotes transformative change, likely undermines the integration of both strategies, ultimately limiting their ability to work together in fostering staff’s safety compliance.

The results show significant direct and indirect correlations between promoting and enabling leadership behaviors and safety compliance (Block 3 of Table 3). In particular, PE has significant indirect effects on SC via CON (β = 0.078**, p < 0.05), INT (β = 0.048**, p < 0.05), FBK (β = 0.012*, p < 0.1), EXP (β = 0.103*, p < 0.1) and MAD (β = 0.128**, p < 0.05). As indicated in the direct effect, when staff are more adequately provided with necessary skills, resources and support by the management, they exhibit greater levels of safety compliance. This finding is supported by social exchange theory (SET) (Ladd and Henry, 2000), which suggests that workplace relationships are built on a system of reciprocal exchanges, and employees’ perceptions of reciprocity within the organization affect how they respond to perceived organizational support. SET implies that when management provides more adequate support, staff are more likely to feel obligated to reciprocate by demonstrating compliance with organizational requirements. This study’s finding is also consistent with previous researchers (e.g. Subramaniam et al., 2023), who examined and confirmed the positive effect of facilitating and assisting leadership behaviors on employees’ safety performance.

The results of the mediation analysis in this study confirm that all CSPs (i.e. CON, INT, FBK, EXP and MAD) serve as partial mediators in the relationships between PE and SC (Block 3 of Table 3). This result shows that more promoting and enabling leadership behaviors via one or more CSPs entail higher levels of safety compliance. While organizational reciprocity, as suggested by SET, influences staff’s behaviors, individual actions are also driven by intrinsic motivations. As a result, such reciprocity encourages compliance but does not guarantee it. In other words, when staff receive essential safety training, resources and support from the management, they typically but not invariably deploy their effort to comply with safety rules and procedures. A sustained level of safety compliance therefore can only be achieved by an intrinsic motivation within individuals. This is supported by self-determination theory (SDT) (Deci and Ryan, 2004), which suggests that human motivation and performance are primarily driven by an innate need for autonomy, competence and relatedness. According to SDT, employees are more likely to exhibit sustained compliance when they have control over their work, opportunities to develop their skills and a strong sense of connection with their leaders, colleagues and organization. In this study, while traditional promoting and enabling leadership behaviors promote staff’s safety compliance through organizational reciprocity and a sense of obligation, integrating a complexity perspective and its principles foster autonomy, competence and relatedness among staff, thereby strengthening long-term safety compliance.

The contributions of the present study are two-fold.

By empirically examining the mediated effects of three selective dimensions of safety leadership on safety compliance, this study presented an approach to explain variations in safety compliance across different workplaces, particularly within increasing complexity of safety-critical environments. In addition, three dimensions of safety leadership can serve as reliable predictors of safety compliance, enabling organizations to identify and prioritize these behaviors in training programs and performance assessments. The proposed approach can be used for introducing and examining other targeted leadership behaviors and improving other safety outcomes at any workplace.

This study incorporated the five CSPs into safety leadership and offered a mechanism-based understanding of how consistently high safety compliance can be achieved and maintained in complex operational settings. The findings imply that traditional safety leadership approaches often compel employees to comply with rules and procedures superficially, relying on extrinsic motivation, which may fail to ensure a deep understanding of the protective effects of these measures. The integration of CSPs empowers leadership behaviors to foster intrinsic motivation such as cultivating a sense of responsiveness, meaningfulness and enjoyment, thereby encouraging staff to invest greater effort in making informed and long-term safety decisions. The findings therefore advance theories and concepts of complexity science and leadership in the safety-critical organizations by clarifying the value of CSPs for improvements of safety leadership and safety compliance, while also offering an interdisciplinary approach for more adaptive and effective safety interventions. In practice, organizations should encourage leaders to foster a shared social identity, support collective sense-making and create opportunities for reflection and dialogue among employees. Such approaches may help organizations better respond to dynamic and uncertain safety environments by strengthening adaptive interactions and shared understanding across teams.

This study examined the interactive effects of safety leadership and CSPs on safety compliance in the oil and gas industry. The results of hypotheses testing indicate that the selective safety leadership behaviors via one or more of CSPs entail higher levels of safety compliance. These findings imply that leadership behaviors should direct attention to making sure that staff clearly understand the intended implications of safety rules and requirements and facilitating their sense of joyfulness of adherence, in order to sustained improvement in safety compliance at the workplace. This can be achieved through the integration of proposed CSPs into leadership behaviors in safety production-related activities.

The findings of this study contribute to the knowledge of safety management by giving insights into the positive effects of selective safety leadership behaviors on safety compliance, and the mediating role played by CSPs. Theories of safety leadership and complexity science deserve further investigation to address the gaps in knowledge (e.g. how do leaders adopt CSPs addressing the safety challenges brought about by the fourth industrial revolution and secure the sustained levels of safety compliance?; how do leaders via CSPs influence distinctively on safety compliance of different demographic groups of employees such as gender, age, education, etc.?). More case studies on the adoption of CSPs on various oil and gas projects may bridge the gaps between theory and practice.

There are a few limitations in this study. First, this study relied on a self-reported survey completed by managerial-level respondents, which may introduce response bias. Nonetheless, the impact of response biases was minimized by the following precautions:

• A careful selection of appropriate respondents.

• The data collection procedure ensures voluntary nature of participation, anonymity and confidentiality of respondents’ responses.

• Assuring the comprehensiveness and clarity of the questionnaire to avoid unintended error made by respondents through a pilot study.

Another limitation was the generalizability of the findings. Although the research findings were based on the data collected globally, most of them are derived from privately owned projects (95%). Therefore, the research findings should only be interpreted in such a form of project. The final limitation concerns the statistical power of the research results. In addition to the conventional threshold, the results at p < 0.1 were reported as marginally significant. This approach helps identify meaningful theoretical relationships between leadership behaviors and safety compliance that might otherwise be overlooked due to the limited sample size and complex structural models. Nonetheless, the validity of the findings was assured by confirming the reliability and validity of constructs before performing any substantive analyses and interpreting the research results within an extensive review of the relevant literature.

This research was undertaken as part of John Ojuola’s PhD at Griffith University and was supported by a Griffith University International Postgraduate Research Scholarship (GUIPRS). The funder had no role in the study design, data collection and analysis, decision to publish or preparation of the manuscript.