The Moderating/Mediating Effect of Relationships among Citizenship Behavior, Social Reflection, and Safety Behavior

Most extant research has focused on antecedent and consequence factors of citizenship behavior (CB) within organizations. The long-term lifetime effect of CB has rarely been researched. This study employed a questionnaire survey to collect respondents’ perspectives of individual CB at different life stages. EFA was applied to explore the primary components of CB, i.e., altruism, conscientiousness, sportsmanship, and civic virtue. ANOVA and SEM were performed to examine the hypotheses in this study. The results demonstrated that CBs were influenced by gender, work status, and work experience. CBs also exerted a positive effect on safety behavior, although this effect was mediated by social reflection. Research theory and implementations were discussed.

tional citizenship behavior (OCB), and it has been shown to increase productivity, efficiency and customer satisfaction, and reduce costs and rates of turnover and absenteeism [3]. Thus, CB is considered to be a critical issue in organizations.
However, it is well-known that behavior is cultivated through education and interaction with the environment from childhood onwards. When behavior has been formed and presents in daily life, it then tends to be resistant to change, as expressed in the Chinese proverb, "Change easily, nature is hard to move". Moreover, Hofstede [9] claims that cultural differences exert a profound influence on individuals' attitudinal tendencies. He highlights the cultural difference concept, and defines culture as dissimilarities in collective values of society [9].
Mole [12] argues that culture is a combination of various influences that encourage individuals to do things in a certain manner. Therefore, individuals' attitudinal patterns could be influenced by culture [25].
All managers want members within their organizations to achieve excellent performance. Human resource managers are particularly responsible to recruit the most appropriate people to join their organization, in order to form good OCB and to maintain superior performance. Consequently, this study focuses on the influence of CB on performance and how it affects performance. The remainder of this paper is organized into five sections. A literature review is presented in the second section, which discusses previous research regarding the theory of OCB, behavioral influence factors, and their results. The third section describes the research methodology, including the questionnaire designed, target sampling, analysis techniques, etc. The results of data analyses are shown in the fourth section. Finally, the study's findings are discussed in the final section.

Literature Review
CB was defined as an employee's sense of intentional involvement in organizational activities without expecting any type of benefits [26]. It was derived from the "innovative and spontaneous behavior" concept [27] [28], and modified Barnard's [1] conception by pointing out supra-role behavior in an organization's social relationships with its employees [20] [29]. These concepts suggested Podsakoff and colleagues [29] identified over 30 different forms of CB. Recognizing the large overlap and similarity among CBs, the researchers categorized the behaviors into seven themes, including helping behaviors, sportsmanship, organizational loyalty, organizational compliance, individual initiative, civic virtue, and self-development [28]. Organ [23] identifies five categories of discretionary behavior, and explains how each assists to improve efficiency in the organization. They are called OCB [23] [32] and comprise altruism, conscientiousness, sportsmanship, courtesy, and civic virtue.
OCB is considered to comprise a special role that members of an organization are expected to perform. Organizations could not survive or prosper without their members behaving as good citizens by engaging in many types of positive behaviors [23]. Moreover, previous literature found that culture differences have impacts on OCB [10] [11]. Lin and Ho [11] reported that people with a collectivist inclination engage in OCB more positively, since group harmony and unity are primary tenets of collectivism. Previous studies [33] [34] [35] [36] [37] also found that results of interactions with the environment will direct employees' emotions to enhance behavior within the organization. Tziner and Sharoni's [37] study demonstrated that perceived organizational justice affects employees' behavior, satisfaction, esteem, loyalty, and commitment to the organization and its leaders. The results imply that OCB is a by-product of organizational justice, which contributes to organizational success and expresses employees' loyalty and commitment to the organization's goals. These findings indicate that OCB is positively affected by culture and the environment.
Numerous investigations suggested that OCB comprises two factors based on the target of the behavior, including organizationally-targeted and interpersonally-targeted [24] [38] [39]. Organizationally-targeted behaviors, such as augmenting the reputation of the organization, are referred to as organizational; whereas, interpersonally-targeted OCB, such as assisting to acclimate a new employee, are referred to as interpersonal. This study aimed to determine precisely how to identify the optimal person with better CB before he or she enters an organization. The author applies the following four interpersonal targeted behaviors as CB: altruism, conscientiousness, civic virtue, and sportsmanship. Therefore, the hypothesis postulated as: H1: CB has a positive effect on safety behavior.
Exception considering the behavior to be predicted by CB, it may also be affected by certain social reflection factors, such as personal perspective and attitude. According to reflection theory, our knowledge reflects the real world. Previous researchers [40] [41] suggested that all ideas derive from experiences of the H3: CB has a positive effect on social reflection factors.
H4: Social reflection factors modulate the effect of CB on safety behavior.
The research model is shown in Figure 1.

Questionnaire Design
The study aimed to determine the role that CB plays in performance and its ex-   Table A1). The questions were measured on a five-point Likert scale, ranging from (1) "strongly disagree" to (5) "strongly agree".

Sampling
The study focused on exploring the effects of CB on safety performance among different life stages in Taiwan

Data Analysis Methodology
According to previous research [18] [19], an exploratory factor analysis can reduce a large set of variables to a smaller set of underlying dimensions, which assists to detect the presence of meaningful patterns among the original variables.
Therefore, to ensure that the data of CB are suitable for performing a factor analysis, the Bartlett Test of Sphericity must be significant and the Kaiser-Meyer-Olkin value must be above 0.8 [52]. In addition, a minimum of five subjects per variable or a sample of 100 is requisite for a factor analysis [53]. The criteria include eigenvalues over 1 and a minimum of 5% variance per factor.
Moreover, the scree plot is commonly utilized to extract factors. Factors with loadings of 0.40 or more are retained [54] [55].
Reliability is determined as the ratio of true score variance to observed score variance [52] [56], which could be assessed in three forms: test-retest, alternate-form, and internal consistency [57]. Generally, Cronbach's alpha is used to measure internal consistency reliability among a group of items combined to form a single scale; levels of 0.7 or more are considered reliable in basic research [55] [57] [58]. Since a moderating/mediating effect occurs when some variables are combined, structure equation modeling (SEM) was finally performed to examine the effects of relationships among CB, social reflection, and safety behavior in this study.

Respondents' Profiles
Respondents' characteristics in terms of gender, age, education level, work status, specialized skills, and sea age (work experience after graduating) were elicited to ascertain whether respondents' perceptions were influenced by these characteristics. Among the respondents, 457 were male and 109 were female (see Table 1). Regarding age, there were 64 respondents younger than 15,344 were aged between 16 and 20, 37 respondents were aged between 21 and 25, 58 were aged between 26 and 30, 29 were aged between 31 and 35, and 34 were older than 35. Concerning education level, 70 respondents indicated that they were in secondary school, 276 were in high school, 125 respondents were in college/university, and 95 respondents had completed university education. Regarding work status, 44 respondents stated that they were not working, 414 were students, and 108 respondents were working on shore or serving onboard. Concerning specialized skills, 123 responded that they had none, 220 were specialized in navigation, 215 were specialized in engine, and eight responded "others". Regarding sea age (work experience after graduating), most respondents (438, 77.4%) had work experience of less than one year, 63 between one and three years, 36 between four and six years, and 29 respondents had six or more years. Results presented in Table 1 indicate that four factors accounted for approximately 70.41% of the total variance, and thus represented the primary component of this study. Moreover, an examination of loading factors in Table 2 shows that all items on each of the factors were 0.5 or higher, indicating acceptable

Social Reflection Attributes
Factor analysis with Varimax rotation was conducted to identify the underlying dimensions of social reflection perceptive attributes. A Kaiser-Meyer-Olkin value of 0.814 indicated that the data were suitable for conducting factor analysis, and the Bartlett Test of Sphericity [χ 2 = 1739.82, P < 0.001] suggested that correlations existed among some of the response categories. Results, as shown in Table 3, revealed that two factors, namely, safety perspective and work attitude, accounted for approximately 63.69% of the total variance, and thus represented all of the social reflection attributes in this study.

ANOVA Test
One-way ANOVA was utilized to identify whether perceived differences in dimensions of CB existed between various groups based on demographic variables, such as age, education level, work status and experience, and specialized skills.
The results are presented in Tables 4-6. The results in Table 4 revealed that gender perceptions did not significantly differ regarding the dimensions of conscientiousness and civic virtue. However, significant differences were identified between gender perceptions towards the two remaining dimensions, i.e., altruism and sportsmanship, at the 5% significance level. Concerning male respondents, Table 4 shows that they have a lower mean score (3.89) on altruism, but have a higher mean score (3.54) on sportsmanship, compared to the mean scores of female respondents on these two dimensions, which were 4.03 and 3.34, respectively. These results indicated that the different genders perform different CBs.      The results show that the respondents with less work experience had a lower mean score, while the respondents who had more work experience had a higher mean score on the two dimensions of altruism and civic virtue. These results provided evidence that work experience has a significant influence on CB.

Test of the Proposed Structural Equation Modeling
After ANOVA testing, structure equation modelling (SEM) was applied to examine the path effects among different models with various independent and dependent variables, including CB, safety perspective, work attitude, and safety  Table 7. Because the chi-square (χ 2 ) value is sensitive to sample size [63] [64], model examinations were based on other fitness indices, including chi-square ratio (χ 2 /df), root mean square residual (RMR), root mean square error of approximation (RMSEA), goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI), comparative fit index (CFI), and the Tucker Lewis Index (TLI) [65] [66] [67] [68]. The results showed that the chi-square ratios (χ 2 /df) were all less than four, and the value of RMR and RMSEA were less than 0.03 and 0.07, which were all smaller than the threshold value of 0.05 and 0.08, respectively, and were acceptable. The other fitness indices values, i.e., GFI, AGFI, CFI and TLI, were all greater than the recommended level of 0.90 (see Table 7). In addition to these fitness indices, the items of each dimension were examined based on certain criteria, including standardized residuals in absolute terms greater than 2.58 [52] and completely standardized expected changes greater than 0.3 [66]. All of these models had an acceptable model-to-data fit. These fit measures indicated that the hypothesized model fit the data well in each model [69].

Results of the Hypothesized Relationships
This study aimed to discern the effects of CB on safety behavior and the roles that additional factors may play in the full influence model. Therefore, the author first examined the simplest model, which contains the two dimensions of CB and the explanation of safety behavior. This model has a good model fitness value: a chi-square ratio (χ 2 /df) = 3.11; RMR = 0.021; RMSEA = 0.061; GFI = 0.98; AGFI = 0.95; CFI = 0.98; and TLI = 0.97 (see Model 1 in Table 7). The coefficient (β = 0.48, P < 0.01) displayed in Model 1 of Table 8 presents the relationship between CB and safety behavior, which was found to exert a positive effect. Secondly, an additional factor (i.e., safety perspective) was added into the examining model, which also has a good model fitness value: chi-square ratio (χ 2 /df) = 2.80; RMR = 0.028; RMSEA = 0.056; GFI = 0.97; AGFI = 0.95; CFI = 0.97; and TLI = 0.96 (see Model 2 in Table 7). The coefficient values between CB and safety behavior, between CB and safety perspective, and between safety perspective and safety behavior were 0.22, 0.60, and 0.44, respectively, and all coefficients were significant at the 0.01 level (see Model 2 in Table 8). However, the coefficient of the relationship between CB and safety behavior decreased from 0.48 to 0.22 in this model.  The author then examined the model with three dimensions, including CB, work attitude, and safety behavior. The results are presented in Model 3 of Table   7 and Table 8. This model has a good model fitness value: chi-square ratio  Table 8). However, the standardized coefficient of the relationship between CB and safety behavior decreased from 0.48 to 0.30 in this model, as well. Accordingly, these results constituted evidence that social reflections regarding safety perspective and work attitude play a moderating role when they individually combine with CB to affect safety behavior.
Finally, to provide more evidence to support the proposed model, both the direct and indirect effects of the full model, which contains all variables, including CB, social reflection which comprises two dimensions of safety perspective and work attitude, and safety behavior, were examined using AMOS21.  The study also indicated that safety behavior was separately modulated by safety perspective and work attitude, and safety behavior was mediated by these two external factors when they were combined. The significant impact of safety perspective and work attitude on safety behavior implies that the environment plays a key role in augmenting CB. The study also revealed that safety perspective and work attitude, as social reflection factors, were significantly influenced by CB. This finding strongly suggests that CB can generate a safety perspective and work attitude in general life. self-reported behavior, but it did not verify the time point of greatest influence.

Study Limitations and Future Research
Since organizations aim to improve their performance, future research should consider relationships between CB, OCB, organization culture, management systems, and their impacts on performance. A more comprehensive understanding of factors influencing performance could help to recruit the most appropriate people and implement optimal management practices within the organization. Finally, the collected data obtained from self-reported safety behaviors and perceptions of CB may have been subject to bias due to respondents' reluctance to report actual behavior. A potential social desirability bias or response set effect should therefore be considered. To address this, further research might measure respondents' behaviors by actual observation.