*2.2. Employee's Job Strain and Safety Behavior*

Safety behavior has been considered as one of the most important safe performances in an organization, being defined as employees' behaviors which seek to prevent mental and physical hazards [12,32–34]. Many previous works have reported that safety behavior is closely associated with occupational injuries and accidents in various industries [33,35–37].

Based on previous studies [12,18–21], we suggest that job strain would decrease the level of an employee's safety behavior. As job strain increases, employees may experience functional diminishment in their cognitive/emotional/physical areas [18–20]. The loss of such functions has a serious adverse effect on an employee's abilities pertinent to both attention and prevention for safety [12,18–20]. According to the explanation of the stress–thought model, stress not only increases psychological anxiety and physical fatigue, which cause deterioration of cognitive functions, but also makes normal thinking impossible [19,20]. Considering that adequate situational judgments and decision-making processes are essential for conducting proper safety behavior, stress can have a critical and harmful effect on safety behavior. In addition, according to extant works which have explored the relationship between emotions and decision-making processes [38,39], individuals are likely to experience negative emotions when they are under stress. Then, these negative experiences would lower the level of their logical thinking and judgment ability, eventually resulting in impulsive decision-making and unsafe behavior [38,39]. Based on the above studies, we can hypothesize that job strain will reduce the level of employees' safety behavior.

**Hypothesis 2.** *An employee's strain is negatively related to his or her safety behavior.*

#### *2.3. Mediating Role of Job Strain between Transformational Leadership and Safety Behavior*

As described above, we suggest that an employee's job strain would mediate the relationship between transformational leadership and safety behavior. Based on the above arguments, we believe that transformational leadership may enhance the level of an employee's safety behavior by diminishing his or her job strain at work.

To integrate our hypotheses, which are described above, based on a theoretical ground, we have relied on the context–attitude-behavior framework [14,15] that bolsters the mediation structure. This perspective suggests that a variety of contexts at work (e.g., organizational systems, rules, leadership, and environments) are important preceding factors which significantly affect the attitudes and behavior of employees. Grounded on it, we expect that transformational leadership, as one of the critical contexts, would create employees' behavior (i.e., safety behavior) by affecting their attitude (i.e., job strain). Previous works theoretically and empirically bolster our hypotheses by demonstrating the negative association between transformational leadership and job strain [7,16,17] and job strain and safety behavior [12,18–21]. Thus, we hypothesize as follows.

**Hypothesis 3.** *Employees' job strain mediates the relationship between transformational leadership and safety behavior.*

#### *2.4. Moderating E*ff*ect of Employees' Self-E*ffi*cacy regarding Safety between Job Strain and Safety Behavior*

We suggest that there may be contingent factors which moderate the relationship between job strain and safety behavior. Although job strain would decrease the quality of an employee's safety behavior, there may be some buffering factors which can diminish the negative impact of job strain on safety behavior. Among various buffering factors, in this paper, we have focused on employees' self-efficacy regarding safety because the concept of self-efficacy has been known as one of the most fundamental variables which explain individuals' perceptions, attitudes, and behavior [22,23].

According to Bandura, self-efficacy is defined as the "beliefs in one's capabilities to organize and execute courses of action required in managing prospective situations. Efficacy beliefs influence how people think, feel, motivate themselves, and act" [22]. Previous works have reported the significant role of self-efficacy in various organizational outcomes [22–24]. This concept is considered as a both task-specific and general variable, and it has been known as a dispositional trait that significantly explains individual behaviors across various situations [40]. In this research, we have applied the concept into safety-related contexts. Thus, based on Eden and Zuk's definition [24], we have defined self-efficacy regarding safety as an individual's overall estimate or expectation of his or her ability to effectively deal with safety-related situations.

We believe that an employee's self-efficacy regarding safety can function as a buffering factor that diminishes the negative effects of job strain on the employee's safety behavior. As described above, job strain would deteriorate the quality of an employee's safety behavior. The employee's anxiety and physical fatigue at work which originate in job strain would diminish his or her cognitive functions, directly destroying adequate safe-related decision making. However, if the employee has a high level of self-efficacy regarding safety, he or she can protect himself/herself from the harmful influence of job strain. By virtue of the high-level of self-efficacy, he or she may feel considerable competence in effectively dealing with the harmful effects of job strain. Then, the negative psychological and physical states from job strain may no longer significantly decrease the quality of his or her safety behavior.

In contrast, when an employee has a low level of self-efficacy regarding safety, he or she may feel that he or she cannot deal with various problems from unsafe situations at work (e.g., how to implement safe-related rules and procedures or how to decrease the possibility of safe accidents). In that situation, the negative impact of job strain would be facilitated and amplified, significantly damaging the employee's cognitive abilities. Then, the quality of safe-related decision-making would be substantially decreased, critically destroying his or her safe behavior. Thus, we propose that

an employee's self-efficacy regarding safety may moderate the relationship between job strain and the employee's safety behavior (Please see Figure 1).

**Figure 1.** Framework of research model.

**Hypothesis 4:** *An employee's self-e*ffi*cacy regarding safety may moderate the relationship between his or her job strain and safety behavior.*

#### **3. Method**

#### *3.1. Data Collection*

Considering the residence and industry of the respondents, we chose and contacted companies which had more than 15 employees. Then, considering the size of the company, roughly 3–8 employees were randomly selected. Survey-trained researchers conducted the survey using structured questionnaires. When the quality of the response was bad, the survey was conducted again. Through these processes, data from 997 employees from 103 firms who adequately responded to all the items were utilized in the analysis. The characteristics of the sample are described below (Please see Table 1).


**Table 1.** Descriptive characteristics of our sample.



#### *3.2. Measures*

We measured the research variables with a five-point Likert scale (1 = strongly disagree, 5 = strongly agree). Then, we computed internal consistency of the variables by using Cronbach alpha values.

#### 3.2.1. Transformational Leadership

We utilized 13 items that were adapted from the Multifactor Leadership Questionnaire (MLQ) to measure transformational leadership. The scale was developed by Bass and Avolio [30], consisting of four sub-dimensions: Idealized influence, inspirational motivation, intellectual stimulation, and individual consideration. The 13 items were selected by the suggestions of previous studies on transformational leadership [5–7]. Sample items were "the leader in my organization is a role model I want to be" and "my leader articulates a compelling vision of the future". The Cronbach alpha value was 0.92.

#### 3.2.2. Job Strain

To measure the level of job strain, we utilized 10 items of the job strain scale by adapting the scale of DeJoy and his colleagues [41]. The 10 items of job strain were selected by the authors because those adequately reflected the core components of the measure [41]. Sample items were "I feel nervous and strain because of work" and "I feel nervous when I work". The Cronbach alpha value was 0.85.

#### 3.2.3. Safety Behavior

We utilized 8 items of the Neal and Griffin's [36] scale to measure safety behavior. The scale consisted of two sub-dimensions: Safety participation (SP) and safety compliance (SC). Sample items included "I use all necessary safety equipment to do my job" (SC) and "I put in extra effort to improve the safety of workplace" (SP). The value of Cronbach alpha was 0.94.

#### 3.2.4. Self-Efficacy regarding Safety

To measure the level of employees' self-efficacy regarding safety, we utilized four items by adapting the self-efficacy scale of Bandura [42]. Sample items were "I am confident in reducing the risk of accidents" and "I am capable of maintaining and improving the safety of my workplace". The value of Cronbach alpha was 0.89.

#### 3.2.5. Control Variables

Considering that various factors can influence employees' safety behavior [33], we included employees' gender, tenure, position, and education level in our analysis to control for employees' safety behavior.

#### *3.3. Statistical Analysis*

Frequency analysis and correlation analysis were performed using the SPSS 21.0 program. Furthermore, we conducted a structural equation modeling (SEM) analysis by using the Amos 21.0 program. SEM, unlike the existing multiple regression analysis methodology, is capable of "simultaneously" analyzing the direct or indirect path between variables in an integrated model.

Considering the suggestion of Anderson and Gerbing [43], we took a two-step approach which includes the measurement model and the structural model. To evaluate the model fit of our hypothesized model, various fit indices such as the comparative fit index (CFI), the turker–lewis index (TLI), and the root mean square error of approximation (RMSEA) were utilized. According to previous studies [44,45], when the values of CFI and TLI of a certain model are greater than 0.90 and the value of RMSEA is less than 0.06, then the model can be considered as a good model. Based on this, bootstrapping analysis was conducted to confirm whether the indirect effect of our research model was significant.

#### **4. Results**

#### *4.1. Descriptive Statistics*

The descriptive statistics of this research are shown in Table 2. The main research variables including the independent variable, mediator, moderator, and dependent variable were highly correlated.


**Table 2.** Means, standard deviations, and correlations between variables.

Note: \* *p* < 0.01. With regard to gender, male is coded as 1, and female is coded as 2. With regard to position, general manager or higher are coded as 5, deputy general manager and department manager as 4, assistant manager as 3, clerk as 2, and others below clerk as 1. With regard to education, "below high school diploma" level is coded as 4, "community college" level as 3, "bachelor's" level as 2, and "master's degree or more" level is coded as 1.

#### *4.2. Measurement Model*

To check whether the level of discriminant validity was appropriate, we conducted a confirmatory factor analysis (CFA) for the research variables which were evaluated by the same employee (i.e., transformational leadership, job strain, safety behavior, and self-efficacy regarding safety). The four-factor model had a good fit to the observations (χ2 (df = 140) = 594.71; CFI = 0.965; TLI = 0.958; RMSEA= 0.057). Then, we conducted sequential chi-square (χ2) difference tests to compare the four-factor model with the three-factor, two-factor, and single-factor model, respectively. Specifically, considering that safety behavior and self-efficacy were very highly correlated, we made the two-factor model which loaded the two variables on the same factor The results of the test showed that the four-factor model had the best fit among all alternative models. Therefore, we believe that the research variables are distinctive (Please see Table 3).

**Table 3.** Chi-square difference tests among alternative measurement models.


Note: CFI means comparative fit index, TLI means turker–lewis index, and RMSEA means root mean square error of approximation. In addition, Tl means transformational leadership.

#### *4.3. Structural Model*

#### 4.3.1. Result of Mediation Analysis

We established a moderated mediation model by utilizing the SEM technique. In the analysis, the association between transformational leadership and safety behavior was mediated by employees' job strain.

The fit indices of our hypothetical model (Model 1) was good enough: χ2 = 323.22 (df = 77), CFI = 0.960, TLI = 0.945, and RMSEA = 0.057. In the model, all the control variables (i.e., gender, position, tenure, and education level) were not statistically significant. The model demonstrated that transformational leadership was significantly and negatively associated with job strain (β = −0.20, *p* < 0.001), and job strain was significantly and negatively related to safety behavior (β = −0.06, *p* < 0.05). The results suggest that Hypothesis 1 and 2 were supported.

#### 4.3.2. Result of Moderation Analysis

To test the moderating effect of employees' self-efficacy regarding safety, we built a moderated mediation model which simultaneously included both the mediation structure and moderation structure. The moderation effect of employees' self-efficacy regarding safety on the association between job strain and safety behavior was tested by the model (see Figure 2). As described above, job strain and safety behavior were transformed into mean-centered variables and the interaction term was calculated by multiplying the two transformed variables [46]. Please consider that centered variables are useful in (i) estimating the interaction terms without loss of correlations and (ii) decreasing and testing multicollinearities among research variables. In addition, we tested whether there was a multicollinearity bias between job strain and self-efficacy regarding safety by using the SPSS program. To test this, we calculated the variance inflation factors (VIF) and tolerances [47]. The VIF for job strain and self-efficacy regarding safety was 1.02 and 1.02, respectively, and the tolerance statistics were 0.99 and 0.99, respectively. Because the obtained VIF values were smaller than 10, as well as the tolerance statistics above 0.2, we can conclude that the two variables (job strain and self-efficacy regarding safety) were relatively free from the issue of multicollinearity.

The coefficient of the interaction term (β = 0.07, *p* < 0.01) was significant, implying that employees' self-efficacy regarding safety functions as a moderator in the association between job strain and safety behavior. In other words, when the level of an employee's self-efficacy regarding safety is high, the negative effect of job strain on safety behavior is decreased. Thus, the results support Hypothesis 4.

#### *4.4. Bootstrapping Analysis*

We conducted bootstrapping analysis with a sample of 5000 to test Hypothesis 3, which suggested that there is a mediating effect of job strain between transformational leadership and safety behavior. Note that the mediation effect is significant at a 5% level when the 95% bias-corrected confidence interval (CI) for the mediation effect does not include zero [48]. In the analysis, the bias-corrected CI for the effect on the pathway from transformational leadership to safety behavior via job strain excluded

zero (95% CI = (0.01, 0.04)). Thus, the result indicates that the mediation effect of job strain on the path was significant at the level of 5%, supporting Hypothesis 3.

**Figure 2.** The standardized estimate values of the final model. Notes: \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001.

#### **5. Discussion**

In the present paper, we examined the underlying mechanisms of the relationship between transformational leadership and safety behavior. To empirically test our hypotheses, data from employees in South Korea were utilized. By conducting a moderated mediation model analysis with the SEM technique, we found that employees' job strain mediated the association between transformational leadership and safety behavior. In addition, employees' self-efficacy regarding safety functioned as a moderator in the relationship between job strain and safety behavior. In this section, some theoretical and practical implications can be drawn from our results.

#### *5.1. Theoretical Implication*

We believe that this research may contribute to extending transformational leadership and safety behavior literature by providing these theoretical implications.

First, we delved into the mediating role of job strain to explain the influence of transformational leadership on employees' safe behavior. Previous works which examined the association between these variables had paid less attention to the significance of job strain in describing intermediating processes in the link. Considering that employees' job strain substantially influences employees' perceptions, attitudes, and behaviors, eventually decreasing the quality of various organizational outcomes [10–13], our attempt to reveal the important role of job strain as a mediator in the transformational leadership–safety behavior link would be meaningful. Through it, we expect that this research may contribute to elaborating transformational leadership and safety behavior literature.

Second, we found a contingent factor which moderated the relationship between transformational leadership and employees' safety behavior. Although some previous studies have reported mediators in the link, those have paid less attention to the contingent or contextual factors which moderate the relationship. Of course we acknowledge that investigating mediators in the link would be highly required. However, to elaborately explain the intermediating mechanisms in the transformational leadership–safety behavior link, it is highly required to find certain conditions or contexts under which the mediators work well, since the mediating variables cannot always intermediate the link in all situations. Therefore, we expect that our finding that employees' self-efficacy regarding safety functions as a buffering factor to decrease harmful effects of job strain on safety behavior would enrich previous works on transformational leadership and safety behavior.

#### *5.2. Practical Implications*

We expect that our findings may provide business leaders with practical implications. First, leaders or top management teams who want to enhance employees' safety behavior through implementing transformational leadership would get some insights from this paper. Our results demonstrated that transformational leadership can increase the quality of an employee's safety behavior through decreasing his or her job strain. Thus, to check whether their attempt to enhance employees' safety behavior through facilitating transformational leadership is successful, the top management teams or leaders should monitor the changes in the level of employees' job strain as an indicator. If the level of job strain has not changed or has even increased, this would indicate that their transformational leadership does not function effectively enough to boost employees' safety behavior. In addition, the leaders using transformational leadership should focus their leadership behavior on employees' job strain by implementing "job strain-specific" leadership.

Second, the findings demonstrated that employees' self-efficacy regarding safety functions as a buffering factor in the association between employees' job strain and safety behavior. We expect that the results may emphasize the importance of employees' individual characteristics in preventing the harmful effects of job strain on safety behavior. As the finding suggests, an employee's job strain is not always harmful to his or her safety behavior. Its negative impact would be minimized when the employee has a high-level of self-efficacy regarding how to implement safe-related procedures and how to decrease the possibility of safe accidents. Therefore, we suggest that top management teams or leaders should attempt to foster employees' self-efficacy regarding safety by providing them with safe education programs, safety guidance, and safety systems in an organization.

### *5.3. Limitations and Suggestions for Future Studies*

Although we believe that this research has valuable implications from the theoretical and empirical point of view, it has some limitations which need to be addressed. First, in this study, we found that transformational leadership affects safety behavior through "psychological factors" such as job strain. However, prior to such psychological factors, "physical environments" or "physical states of employees" may affect safety behavior. Further studies are needed to verify this. Second, although we conducted a SEM analysis to test our mediation hypothesis, we could not adequately reveal the causal relationships that our research hypotheses claim since this study only utilized cross-sectional data. This should be complemented and alleviated by not only utilizing a longitudinal research design but also by considering the influence of third variables or alternative explanations [49]. In addition, future studies should deal with the fundamental concern that cross-sectional data cannot adequately describe and explain the interaction effects between variables. Third, the data of this study were collected through participants' self-reports. Since the employee's behavior which is reflected in the self-report survey may be different from his or her actual behavior, there needs to be adequate supplementation. For example, a third party's observation or behavioral assessment can be good alternative ways to collect data. Fourth, because the same respondents responded to our survey at the same time, they cannot be free from the common method bias problem. The problem is likely to lead to an overestimation of the correlation between the variables. This limitation needs to be dealt with. Lastly, this paper could not fully utilize the entire items of each measure for our research variables (i.e., transformational leadership and job strain). Although we chose core and essential items from the original version of the measures, future studies should use the full items of the original measures. Fifth, in this study, we only focused on employees' positive behavior, such as safety behavior, when we investigated the impact of transformational leadership in an organization. However, considering that negative behavior of employees also critically influences organizational outcomes, future studies are needed to deal with negative behavior such as unsafe behavior or counterproductive work behavior. Lastly, this paper did not adequately deal with the issue of nesting of data. Considering that some respondents may be nested into the same organization, the respondents are likely to share the same culture, climate, and leadership style. Thus, the perceptions of the respondents on their transformational leadership tend to be more similar within the organization than between organizations. This research could not deal with this issue. To complement it, additional multi-level approaches are recommended.

### **6. Conclusions**

Although this study has various limitations, we believe that it contributes to deepening the transformational leadership and safety behavior literature by investigating a mediating factor between transformational leadership and safety behavior, as well as a contingent factor through which job strain influences safety behavior. Specifically, through this study, we have shown two important findings. First, transformational leadership reduces the level of employees' job strain, and then the reduced strain ultimately increases their level of safety behavior. Second, although job strain decreases the quality of employees' safety behavior, their self-efficacy regarding safety functions as a buffering factor by moderating the relationship. The findings show that an employee's job strain plays an intermediating role in connecting transformational leadership with safety behavior. We also demonstrated the importance of self-efficacy regarding safety in minimizing the negative effects of job strain on safety behavior.

**Author Contributions:** B.-J.K. and S.-Y.J. contributed to all the processes including writing of the original draft, conceptualization, data collection, formal analysis, and methodology of this paper.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

International Journal of *Environmental Research and Public Health*
