Multidimensional analysis of supervisors’ safety leadership on safety violations of construction workers: An empirical investigation

2.1 Safety violations

Construction safety violations are the opposite of safety compliance, referring to behaviors intentionally carried out by construction workers during the operation process, which, although unintentionally causing any harm, violate the organizational safety system and procedures [7, 40]. Safety violations are usually categorized into two dimensions: individuals’ routine safety violations (IRSV) and individuals’ situational safety violations (ISSV). IRSV refer to the intentional circumvention of safety systems, procedures, or practices by workers to achieve organizational (e.g., timely completion of work tasks) or personal (e.g., saving time or physical effort) benefits. When committing routine safety violations, workers habitually choose the most physically efficient operation or a “shortcut” [41]. In contrast, IRSV and ISSV result from situational constraints in the workflow that render compliance with the safety system impractical [42]. For instance, an operator may violate a safety system when protective equipment is unavailable. The main difference between these types of safety violations lies in whether they are caused by situational constraints. Existing studies indicate that situational and routine safety violations are influenced by different antecedent variables. For example, Liang et al. [39] found that workers’ perceived social support and safety motivation affect ISSV, while perceived production stress and ambivalent attitudes affect IRSV. Another study by Liang and Zhang [40] concluded that workers with low education levels are more likely to violate situational safety regulations, while inexperienced workers are more inclined towards routine safety violations. On this basis, this study also classified construction workers’ safety violations into IRSV and ISSV. It explored the intrinsic mechanisms between the different dimensions of supervisor safety leadership and these two types of safety violations.

2.2 Supervisors’ safety leadership (SSL)

2.3 SSL and safety violations

SSL has been shown to be highly effective in improving human safety behavioral performance. For example, Dickerson et al. [44] argued that behavior-based safety coaching is a successful strategy for reducing errors in the context of industrial occupational safety. Leader coaching behavior can effectively guide subordinate safety behavior [45]. Kapp [28] and Petitta et al. [46] both pointed out that strong safety control can directly enhance workers’ safety performance and reduce the frequency of their mistakes. Liu et al. [47] found that safety support from superiors can directly and positively influence employees’ safety citizenship behavior. Based on these findings, we argue that SSL can attenuate worker safety violations and propose hypotheses 1 and 2.

H1: Three sub-dimensions of SSL, namely Safety coaching (H1a), Safety controlling (H1b), and Safety caring (H1c), have negative influences on ISSV.

H2: Three sub-dimensions of SSL, namely Safety coaching (H2a), Safety controlling (H2b), and Safety caring (H2c), have negative influences on IRSV.

2.4 Safety motivation as the mediating variable

Safety motivation (SM) refers to an individual’s willingness to make efforts to achieve safety and the corresponding actions, reflecting the importance employees attribute to safety [48–50]. Previous research conceptualized safety behavior as an interaction between proximal individual differences (e.g., safety motivation, etc) and distal situational factors (e.g., leadership or safety management, etc). Therefore, safety leadership has been recognized as an important antecedent of safety motivation. Current research has confirmed this relationship. For example, Kovjanic et al. [51] argued that when supervisors provide careful instruction and safety care, these high-quality behaviors resonate with workers and enhance their sense of identification with the organization, which helps workers foster safety motivation. Probst and Brubaker [52] noted that workers’ safety motivation relates to supervisors’ perceptions of safety policy enforcement, including the extent to which supervisors commend safety compliance and sanction non-compliance. Inadequate enforcement by supervisors correlates with low levels of worker safety motivation. Based on the above discussions, hypothesis 3 is proposed:

H3: Three sub-dimensions of SSL, namely: Safety coaching (H3a), Safety controlling (H3b), and Safety caring (H3c) have positive influences on SM

In addition, existing research generally recognizes safety motivation as a key determinant of whether an individual makes safety violations [39]. Safety motivation can significantly enhance employees’ level of safety compliance [53–55]. However, the extent to which safety motivation plays a role in safety violations may vary depending on the nature of the violations. For example, when workers still choose to adhere to safety rules without safety protective equipment, they may need more safety motivation. On the contrary, if they are adhered to just to comply with safety procedures, the effectiveness of safety motivation may be relatively low [48]. Therefore, hypothesis 4 is proposed to test the effect of safety motivation on different types of safety violations:

H4: SM have negative influences on two sub-dimensions of safety violations, namely: ISSV (H4a) and IRSV (H4b)

On the basis of hypotheses 3 and 4, we further combine the Self-Determination Theory (SDT) to explain the mediating role of safety motivation between SSL and safety violations: SDT recognizes human beings as positive organisms with innate potential for psychological growth and development [56]. However, these intrinsic potentials often require external stimuli for reinforcement and gradually develop into control over behavior. Therefore, key external environmental variables (e.g., leadership) help stimulate autonomous motivation in employees to further drive their growth and progress [51]. Therefore, hypothesis 5 is proposed:

H5: SM can mediate the relationship between SSL and safety violations: SM can mediate the relationship between the three sub-dimensions of SSL (Safety coaching (H5a), Safety controlling (H5b), and Safety caring (H5c)) and ISSV. SM can mediate the relationship between the three sub-dimensions of SSL (Safety coaching (H5d), Safety controlling (H5e), and Safety caring (H5f)) andIRSV.

2.5 Group safety norm as the mediating variable

Group safety norm (GSN) indicates the social pressure to perform safety behavior, namely, what others expect the individual should undertake to perform tasks safely. It is a collective cognition of a group [57]. The common belief of group is formed by the interaction between internal members. The interaction occurs not only between workers, but also between workers and their direct leaders [21, 58]. Previous studies have affirmed the pivotal role of supervisors and their leadership in the formation of safety-related group norms in construction teams. For example, Fang et al. [21] found that the training and prevention behavior performed by supervisors significantly affects the group-level safety climate, including the involvement and mutual influence of workers. McFadden et al. [59] argued that reasonable control over employees can effectively cultivate positive group safety norms. He et al. [60] stated that a high-quality leader-member exchange relationship directly and positively affects the safety climate of the team. Based on the above discussions, hypothesis 6 is proposed:

H6: Three sub-dimensions of SSL, namely: Safety coaching (H6a), Safety controlling (H6b), and Safety caring (H6c) have positive influences on GSN

Moreover, due to the complexity of the construction site, formal rules cannot regulate every aspect of workers’ behavior at all. Behavior changes driven by social norms are more sustainable and effective, especially when such social norms spread to the work groups [61]. Under such an environment, the safety behavior of workers can be improved unconsciously, and they will do their best to avoid violations. Therefore, hypothesis 7 is proposed:

H7: GSN have negative influences on two sub-dimensions of safety violations, namely: ISSV (H7a) and IRSV (H7b)

Based on hypotheses 6 and 7, we further combine the Social Cognition Theory to explain the mediating role of group safety norm between SSL and safety violations: according to social cognition theory, individual workers always try to belong to a social group and adjust their behavior to what they perceive as socially acceptable within the crew [62]. Despite the belief in the effectiveness of GSN in mitigating the relationship between SSL and worker safety violations, few studies have synthesized the multidimensionality of SSL and safety violations. Therefore, hypothesis 8 is proposed:

H8: GSN can mediate the relationship between SSL and safety violations: GSN can mediate the relationship between the three sub-dimensions of SSL (Safety coaching (H8a), Safety controlling (H8b), and Safety caring (H8c)) and ISSV. GSN can mediate the relationship between the three sub-dimensions of SSL (Safety coaching (H8d), Safety controlling (H8e), and Safety caring (H8f)) and IRSV.

2.6 Self-efficacy as the mediating variable

In the field of safety, workers’ self-efficacy refers to “workers’ estimates and judgments of their ability to accomplish safety-related goals and tasks” [63]. Evidence has shown that in safety critical industries, the involvement of supervisors in safety management can effectively enhance the self-efficacy of subordinates towards safety rules. Similarly, the positive relationship between SSL and employees’ self-efficacy has been widely demonstrated. Bono and Ilies [64] suggested that subordinates emulate the exemplary behavior of their supervisors to stimulate their own positive feelings and efficacy. Fugas et al. [65] argued that supervisors’ “injunctive norms” contribute to workers’ accurate perceptions of their own abilities and motivate them to surmount difficulties. In addition, care and support from supervisors have been identified as effective drivers of self-efficacy [66]. However, few studies have comprehensively compared the differences in the impact of different dimensions of SSL on self-efficacy. Based on the above discussions, hypothesis 9 is proposed:

H9: Three sub-dimensions of SSL, namely: Safety coaching (H9a), Safety controlling (H9b), and Safety caring (H9c) have positive influences on SE

Additionally, several studies have reported a positive relationship between workers’ self-efficacy on safety performance. For instance, Yen et al. [67] and Saleem et al. [68] both concluded that workers’ self-efficacy positively affects their safety compliance. Kim and Jang [69] revealed that safety self-efficacy is a key skill for individuals to inhibit human errors and thus avoid potential accidents. Therefore, we believe that the self-efficacy of workers can reduce the probability of safety violations and thus propose hypothesis 10:

H10: SE have negative influences on two sub-dimensions of safety violations, namely: ISSV (H10a) and IRSV (H10b)

Based on hypotheses 9 and 10, we further combine the Social Information Processing (SIP) theory to explain the mediating role of self-efficacy between SSL and safety violations: according to SIP, an individual processes information based on their understanding of how cognitive and behavioral responses come about through social interactions [70, 71]. Since behaviors, attitudes, and perceptions are affected by social information, all aspects of individual behaviors, attitudes, and perceptions are presented as a result of cognitive products of information processing [72, 73]. In construction workplaces, workers heavily rely on cues or signals from leaders to confirm their understanding of the environment in the organization and adjust their cognitions accordingly. Thus, in explaining the link between safety leadership and safety violations, we hypothesized a mechanism that could potentially mediate this relationship. From this perspective, workers’ self-efficacy was identified as another mediator at the individual level. Therefore, hypothesis 11 is proposed:

H11: SE can mediate the relationship between SSL and safety violations. SE can mediate the relationship between the three sub-dimensions of SSL (safety coaching (H11a), safety controlling (H11b), and safety caring (H11c)) and ISSV. SE can also mediate the relationship between the three sub-dimensions of SSL (Safety coaching (H11d), Safety controlling (H11e), and Safety caring (H11f)) and IRSV.

Based on the above analysis, the complete research model of this paper is shown in Fig. 1.

OPEN IN VIEWER

Fig. 1

The hypothesized model regarding the supervisors’ safety leadership-safety violations relationship.

3.1 Participants and procedures

The questionnaire survey was conducted between 01/05/2023 and 31/08/2023. Data were collected using both electronic and paper questionnaires. A valid survey link was provided through the questionnaire website (Question-star) for respondents to fill in the questionnaire. As per the research scope from the perspective of construction workers, 500 paper questionnaires were distributed to frontline construction workers in major Chinese cities, including Shanghai, Suzhou, Nanjing, Chongqing, Xi’an and Wuhan. We collected 255 paper questionnaires and 198 online questionnaires. After excluding those that did not meet the requirements, 195 valid paper questionnaires and 151 online questionnaires were finally recovered, with valid questionnaire recovery rates of 76.5% and 76.3%, respectively. A total of 346 valid questionnaires were recovered. The demographic information of the respondents is shown in Fig. 2. These sample characteristics are consistent with the nature and uniqueness of the construction industry.

OPEN IN VIEWER

Fig. 2

Demographic information of the respondents. Note: ^*represents p < 0.05, ^**represents p < 0.01.

We adapted previously validated measurement scales and developed specific scales for use in the construction context. Before the formal scale was finalized, the original English questions were translated into Chinese following standard translation procedures, followed by a panel discussion among five experts in the field of engineering safety to determine the final measurement scales (e.g., incorporating industry experience and knowledge). Each measurement question was rated on a Likert 5-point scale ranging from 1 (strongly disagree) to 5 (strongly agree). Table 1 shows the variables, measures of study variables, number of items, and supporting literature.

A 19-item scale was used in this study to assess each worker’s perceptions of safety leadership across three adapted subscales: Safety coaching, Safety controlling and Safety caring [33]. Sample question items include, “My supervisor helps workers realize the importance of safety” (SCoa), “My supervisor orders us to be determined to accomplish safety goals” (SCon), “My supervisor works hard to meet our safety needs” (SCar), etc.

The scale used to measure worker safety motivation was selected from the validated scale developed by Nykanen et al. [74], which measured safety motivation among vocational high school students with three items. We fine-tuned its content by incorporating the context of the construction safety industry (one of the sample items was “I feel that working to maintain or improve my personal safety is worthwhile”).

To measure worker group safety norms, we adapted a scale from the work of Su et al. [7]. This scale contains five items to assess workers’ perceptions of whether fellow team members recognize certain behaviors. One of the sample items is: “Members within our shift would strongly disapprove of behavior that violates safety rules and regulations”.

The scale used to measure workers’ self-efficacy was selected from the Psychological Capital Scale developed and validated by Luthans et al. [75], which measured employees’ self-efficacy using six items. We adjusted it to incorporate the context of the construction safety industry (one of the sample items was: “I believe that I can always analyze problems and find solutions independently”).

The questions for assessing both routine and situational safety violations were adapted from the self-report items validated by Chmiel et al. [35] and Su et al. [7]. Routine safety violations were measured with four items, with a high score indicating a greater frequency of violations. A sample item reads, “I’ll take shortcuts to get the job done if I can make it easy”. Situational safety violations were measured with six items and were reversely scored: a high score indicated a low level of violations. A sample item is “Even if it’s inconvenient sometimes, I always wear a safety helmet”.

The data analysis was conducted in three stages using SPSS 21.0 and AMOS 24.0 software. First, SPSS 21.0 was used for descriptive analysis. Second, reliability tests were conducted using SPSS 21.0, and AMOS 24.0 was used for confirmatory factor analysis (CFA) to test the dimensionality and discriminant validity of our multiple-item measures, and metrics were calculated to test convergent validity. Finally, all hypotheses (across the entire model) were tested (see Chapter 4), including tests for direct effects between variables as well as mediating effects (using AMOS 24.0 software).

4.1 Structural model fit

To clarify the direct influence relationship among the variables, we analyze the model composed of all variables using AMOS 24.0 software. The first step is to determine the validity of the structural equation model [83], mainly gauged through the measurement of some fit indicators [84, 85]. The model fit metrics and their corresponding criteria are shown in Table 5. All metrics meet the criteria except for the NFI, which is close to 0.9. This suggests an acceptable overall model fit.

Figure 3 demonstrates the final estimation results of the structural model together with the standardized path coefficients and their respective significance levels. On this basis, the standardized path coefficients (β), critical ratio (C.R.) values, p-values, and hypothesis testing results of the model are presented more visually in Table 6. The path coefficients (β) reflect the relationship between the variables and their respective degrees of influence, while the C.R. determines the significance of regression coefficients. The C.R. value equal to or greater than 1.96 indicates significance at a p-value less than 0.05 [84].

OPEN IN VIEWER

Fig. 3

The structural model and impact paths.

As seen in Fig. 3 and Table 6, among the three dimensions of SSL, only SCar exhibits a significant effect on ISSV (β= –0.159, p < 0.05). Therefore, only H1C holds for hypotheses H1a-c. Furthermore, only SCoa has an effect on IRSV (β= –0.161, p < 0.05). Therefore, only H2a holds true for hypotheses H2a-c. SCoa, SCon, and SCar all positively affect SM (β= 0.167, 0.232, 0.581, p < 0.001), confirming the validity of H3a-c. However, SM only influences ISSV (β= –0.469, p < 0.001), thus only supporting H4a in H4a-b. GSN is positively influenced by SCon and SCar (β= 0.383, 0.360, p < 0.001), thus validating H6b and H6c in H6a-c. However, GSN only influences ISSV (β= –0.230, p < 0.001), hence, only H7a is upheld within H7a-b. Finally, SE is only positively affected by SCoa (β= 0.199, p < 0.01). Thus, only H9a is supported within H9a-c. However, SE only influences IRSV (β= –0.428, p < 0.001), confirming only H10b in H10a-b.

To understand the mediating role of workers’ safety motivation, group safety norms, and self-efficacy, the bootstrap method was used [86]. The mediating effect was tested by 2000 iterations, calculating 95% confidence intervals.

We provide a more intuitive list in Table 7, detailing the mediation effect values, 95% confidence intervals, and hypothesis testing results for each mediation pathway. For example, considering the SCoa⟶SM⟶ISSV path, the mediation effect value is –0.089. The upper and lower 95% confidence intervals exclude 0, and the p-value is less than the significant level of 0.05, indicating a mediation effect in the path. Similarly, the standardized indirect effects of SCon and SCar on ISSV through SM are (–0.113, p < 0.01, 95% CI [–0.207, –0.048]) and (–0.263, p < 0.01, 95% CI [–0.418, –0.146]), respectively. Therefore, hypotheses H5a, b, and c are all hold. Conversely, for the SCoa⟶SM⟶IRSV path, the mediation effect value is – 0.041. The upper and lower 95% confidence intervals contain 0, and the p-value is greater than the significance level of 0.05, indicating the absence of a mediation effect in this path. Similarly, the standardized indirect effects of SCon and SCar on IRSV through SM are not statistically significant (95% confidence interval including 0). Therefore, hypotheses H5d, e, and f are not hold. Hypothesis H5 is partially supported. In addition, the standardized indirect effects of SCon and SCar on ISSV through GSN are (–0.091, p < 0.01, 95% CI [–0.177, –0.036]) and (– 0.08, p < 0.01, 95% CI [–0.164, –0.028]), respectively. Thus, hypotheses H8b, and c are all hold. The standardized indirect effect of SCoa on ISSV through GSN and the standardized indirect effects of SCoa, SCon, and SCar on IRSV through GSN are not statistically significant (95% confidence interval including 0). Therefore, hypotheses H8a, d, e, and f are not hold. Hypothesis H8 is partially supported. Furthermore, the standardized indirect effect of SCoa on IRSV through SE is (– 0.121, p < 0.05, 95% CI [–0.226, –0.024]). Therefore, hypothesis H11d holds. However, the standardized indirect effects of SCoa, SCon and SCar on ISSV through SE and the standardized indirect effects of SCon and SCar on IRSV through SE are not statistically significant (95% confidence interval including 0). Therefore, hypotheses H11a, b, c, e, and f are not hold. Hypothesis H11 is partially supported.

5.1 The direct relationship between SSL and safety violations

The tests conducted on hypotheses 1 and 2 show that only two paths, SCar⟶ISSV (H1c) and SCoa⟶IRSV (H2a), hold. Few studies have been able to account for this result. Therefore, we propose a possible explanation for the direct impact of safety caring on ISSV. As mentioned in Section 2.1., ISSV is caused by situational constraints during construction work (such as safety equipment not being easily accessible, etc.). We argue that resource support and emotional care from supervisors directly benefit workers. When supervisors provide care and support, workers’ safety needs are met and safeguarded, thereby removing situational constraints that impede their safety compliance [22]. As a result, their situational violations are diminished. In response to the direct influence of safety coaching on IRSV, we argue that within construction crews, workers tend to look to the supervisor as an opinion leader and seek their guidance to determine operational procedures [24]. Therefore, individual behavior is influenced by supervisors’ demonstration of safety behavior. It can be inferred that some of the individual’s undesirable behavior (e.g., failure to comply with safety procedures) also mostly comes from the imitation of supervisors’ erroneous safety coaching. Meanwhile, compared to other roles, the influence of supervisors is often more effective [43], which leaves a deeper impression on workers and leads to their daily and ongoing safety violations.

5.2 The mediating mechanisms between SSL and safety violations

5.3 Theoretical contributions

The findings of this study yield significant theoretical contributions to the study of leadership behavior and occupational health and safety in the construction industry. While Current research on construction safety violations still focuses on organizational or individual factors, there have been relatively few studies on SSL. To the best of our knowledge, this is the first study to explore the multiple mediating mechanisms of the relationship between SSL and individuals’ safety violations from the perspective of construction workers. Therefore, this study offers some meaningful empirical insights into construction safety management. For example, three dimensions of SSL and three mediating variables were established to relate to different organizational and individual levels. This multidimensional analysis has addressed a limitation in previous safety leadership research, which tends to relate to organizational-level mediating variables. On this basis, the direct and indirect effects of three leadership characteristics on two types of safety violations (routine and situational) have been clarified. The results of the study highlight safety caring from supervisors can better stimulate the willingness of workers to actively engage in safety related actions, thereby reducing the probability of situational safety violations. In addition, supervisors’ safety controlling can be more effective in creating a good group safety climate. In such an organizational environment, workers are constantly subjected to good conduct and professional safety terms, which make them less susceptible to breaches of safety rules. Furthermore, supervisors’ safety coaching enhances workers’ competence and confidence, thereby promoting sustained adherence to safety protocols. As a result, this evidence-based research can inform the development of targeted interventions designed to effectively reduce both types of workers’ safety violations, namely, ISSV and IRSV. Therefore, this study clarifies the complex relationships between supervisor behavior and safety outcomes, particularly from the perspective of construction workers and their perceptions of supervisors’ actions and leadership.

5.4 Practical implications

This study offers practical implications for construction safety management. Firstly, top management in the construction industry should prioritize education and training for frontline supervisors to enhance their skills, safety competence, and correct safety behavior [89]. In this way, these principles penetrate the organization from the top down, and the workers learn through emulation coupled with the supervisors’ correct safety coaching; workers can significantly improve their safety competence and confidence to ensure diligent performance of safety duties. Secondly, supervisors should create a positive safety climate at the workplace by communicating the principle of “safety first” to workers. This involves a pattern of high expectations, high standards of supervision, and rewards and penalties. Through the establishment of “prohibitive norms” within the organization, worker groups can demonstrate high standards of safety behavior, and workers will remind and help each other to adhere to these regulations. Moreover, supervisors should increase their caring for workers’ living conditions, take flexible measures to meet sensible and reasonable work requirements within the organizational framework, offer increased encouragement to employees and humbly accept workers’ safety-related suggestions. The integration of humanistic care and respect for individuality within institutional constraints greatly improves workers’ satisfaction and recognition of their supervisors [90]. This approach also better shapes workers’ safety values and creates stronger motivation among employees to prioritize safety in their work.