Multilevel approach to organizational and group safety climate and safety performance: Co-workers as the missing link

Safety Science 50 (2012) 1847–1856

Contents lists available at SciVerse ScienceDirect

Safety Science journal homepage: www.elsevier.com/locate/ssci

Multilevel approach to organizational and group safety climate and safety performance: Co-workers as the missing link Margherita Brondino a,⇑, Silvia A. Silva b, Margherita Pasini a a b

Department of Philosophy, Pedagogy and Psychology, University of Verona, Lungadige Porta Vittoria, 17, 37129 Verona, Italy Instituto Universitário de Lisboa (ISCTE-IUL), Lisboa, Portugal

a r t i c l e

i n f o

Article history: Received 21 May 2011 Received in revised form 28 March 2012 Accepted 11 April 2012 Available online 29 May 2012 Keywords: Organizational safety climate Supervisor’s safety climate Co-workers’ safety climate Safety agents Safety performance Multilevel structural equation modelling

a b s t r a c t The aim of this study is to test a model on the relationships between organizational and group safety climate and safety performance, that highlights the importance of co-workers as a safety climate agent side by side supervisors at group level. The idea is to consider the co-workers’ safety climate as a necessary part of a multilevel model of safety climates’ framework associated to safety performance. Firstly, the assessment of the safety climates’ framework which consider organizational safety climate and at group level supervisor’s and co-workers’ safety climate was performed. Then, the mediating role of co-workers’ safety climate between organizational and supervisor’s safety climate, and worker’s safety behaviours was explored. From the literature, the importance to study safety climate in a multilevel perspective by a theoretical and methodological point of view is known. For these reasons the proposed models were tested with multilevel structural equation modelling. We used a two-level design which considered the individual level and the work-group level. Data collection involved 991 blue-collars, belonging to 91 work groups, from five Italian manufacturing companies. The research highlighted the importance of considering at group level not only climate referred to supervisor, but also climate referred to co-workers. Furthermore, results confirmed the mediating role of co-workers’ safety climate and revealed that co-workers’ safety climate had a stronger influence on safety behaviours, and in particular on safety participation, than supervisor’s safety climate, at individual level as well at group level. Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction Since the 1990s, research on safety at work has often centred on safety climate, as an antecedent of safety performance, defining safety climate as the shared perceptions of the employees on policies, procedures, and practices relating to safety. It can be investigated at two hierarchical levels: group level, and organizational level. At the group level, safety climate usually refers to the role of supervisor (e.g. Zohar, 2000; Zohar and Luria, 2005; Wallace and Chen, 2006; Meliá and Sesè, 2007). Although the psychosocial, organizational, and safety literature recognizes the importance of co-worker influence, research has not systematically included coworkers. Therefore, the present study adds co-workers as a safety climate agent side by side with top management and supervisors; and it analyses the safety climate at the organizational and the group level. This approach allows us to not only study the importance of co-workers in creating a safe climate and culture in an organization but also to explore the mediating role of co-workers’ safety climate in the relationship between safety climate (connected to top ⇑ Corresponding author. Tel.: +39 045 802 8558. E-mail address: [email protected] (M. Brondino). 0925-7535/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ssci.2012.04.010

management and supervisor) and safety performance. By using multilevel structural equation modelling, the present study represents one of the first, if not the first, attempt to analyse data on safety climate while keeping the complexity of its structure and permitting the proper study of relationships between relevant constructs at different organizational levels. 1.1. Safety climate Researchers now define safety climate as a multilevel construct (Zohar, 2000; Zohar and Luria, 2005; Zohar, 2008, 2010; Glendon, 2008; Meliá et al., 2008). Organizational processes take place simultaneously at several levels, and these processes are linked to one another (e.g. Kozlowski and Klein, 2000; Shannon and Norman, 2009). Hence, processes that take place at one hierarchical level have an influence on other levels. For safety, the safety climate may have different meanings at different levels of an organization, and it may have relationships across levels. Zohar and Luria (2005) suggested that the core meaning of safety climate relates to socially constructed indications of desired role behaviour, coming simultaneously from policy and procedural actions of top management and from practices of the supervisors. Thus, at the organizational level, safety climate refers to worker

1848

M. Brondino et al. / Safety Science 50 (2012) 1847–1856

Fig. 1. Zohar and Luria model (Zohar and Luria, 2005).

perceptions of the top management’s policies and procedures, while, at the group level, safety climate refers to worker perceptions of how the supervisors transform these policies and procedures into daily practice. In addition, the group (supervisor) safety climate mediates between the effect of organizational safety climate on safety behaviours in work groups (see Fig. 1). Yet, for the group safety climate, the research on safety climate tends to overlook the role of co-workers, and focus more on the leadership perspective. It considers the supervisor as ‘‘enough’’ to represent the group climate. However, strong evidence from social and organizational psychology highlights the need to consider the influence of coworkers on the group safety climate (e.g. Ashforth, 1985; Bandura, 1986; Deutsch and Gerard, 1955). Co-workers offer information, show behavioural support for desired practices while discouraging others and might shape their co-workers’ roles through offering lateral mentoring (e.g. Ashforth, 1985; Chiaburu and Harrison, 2008). Researchers have studied the role of co-workers for co-workers’ support (e.g. Chiaburu and Harrison, 2008; Burt et al., 2008); practices (e.g. Meliá and Becerril, 2006; Meliá et al., 2008; Jiang et al., 2009); social norms (e.g. Fugas et al., 2009; Kath et al., 2010); interaction (e.g. Cavazza and Serpe, 2009; Zohar and Tenne-Gazit, 2008; Zohar, 2010); and a more generalized content as co-worker safety (e.g. Gyekye and Salminen, 2009; Lingard et al., 2011; Morrow et al., 2010), and others have called for the inclusion of the study of group influences on the safety climate (e.g. Chiaburu and Harrison, 2008; Ehrhart and Naumann, 2004; Fugas et al., 2011; Westaby and Lowe, 2005). However, studies have only occasionally used items about co-workers as part of a safety climate scale. One study of safety climate from the perspective of the agents who perform or are responsible for each safety climate action or omission inside the company (safety response) identified co-workers as a safety agent as important as the organization and the supervisor (Meliá et al., 2008). The organizational safety response and supervisor’s safety response positively predict co-workers’ safety response at a statistically significant level. Notably, the authors emphasized safety climate as a diagnostic tool to explore specific ways to improve safety at work. From this premise, they identified four main agents who are responsible for every safety issue inside the organization: organization, supervisors, co-workers and workers. For these agents, the authors studied five safety climate variables: organizational safety response, supervisors’ safety response, co-workers’ safety response, worker safety response and perceived risk of accidents. Studying the statistically significant relationships between these five variables in four different samples (see Fig. 2) they found complex interrelationships. The organizational safety response predicted the safety response of the supervisors and the co-workers; and the safety response of the supervisors predicted the safety response of the co-workers. The safety response of the co-workers and of the organization predicted the safety response of workers. Only in two of the four samples, the safety response of the supervisor predicted worker safety response. In agreement with this, another study found that co-workers’ support and antagonism have a unique effect on employees’ outcomes beyond that of leader influences and that co-workers’ support has a strong positive relationship with task performance (Chiaburu and Harrison, 2008). In sum, there are four reasons for including co-workers at the group level:

Therefore, it is essential to define group safety climate to takes into account what a group stands for and distinguishes between the role of the supervisor and co-workers. While Zohar (2008) described the safety climate as a multilevel construct coming simultaneously from policy and procedural actions of top management and from practices of the supervisors, the evidence suggests a broader definition that includes the practices of co-workers. Consequently, the present study seeks to test a model that integrates approaches of Zohar and Luria (2005), and Meliá et al. (2008) and explores the role of co-workers as an agent of safety climate at the group level and as mediating role between organization and supervisor’s safety climate, and workers safety behaviours. The relevance of co-workers was also supported by Chiaburu and Harrison (2008) who showed that co-worker support and antagonism have a unique effect on employees’ outcomes beyond that of leader influences, and that co-workers’ support has a strong positive relationship with task performance. Based on the empirical evidence, we propose the multilevel conceptual model of the framework associated with safety outcomes shown in Fig. 3. Note that as workers are nested within work groups, the analysis cannot treat the data from individual workers as completely independent. Instead, it must decompose the variances into a between-group component (variability among work groups) and a within-group component (variability among individuals within each group). The model specifies effects of organizational, supervisor’s and co-workers’ safety climates at individual level (the within-group model, which concerns the within-group variance, below the dotted line in Fig. 3) and at group level (the between-group model, which concerns the betweengroup variance, above the dotted line in Fig. 3). The following hypotheses describe this model in detail.

1. The safety climate is constructed through interactions between people.

H3b. Co-workers’ safety climate mediates the relationship between supervisor’s safety climate and workers’ safety behaviours.

Fig. 2. Model of Meliá et al. (2008).

2. The group climate is important and groups include more than only a ‘‘formal leader’’. 3. Peers/co-workers influence one another. 4. Research has recognized the relevance of co-workers.

H1. Organizational safety climate positively predicts supervisor’s safety climate and co-workers’ safety climate. H2. Supervisor’s safety climate mediates, at least partially, the relationship between organizational safety climate and co-workers’ safety climate. H3a. Co-workers’ safety climate mediates, at least partially, the relationship between organizational safety climate and workers safety behaviours.

1849

M. Brondino et al. / Safety Science 50 (2012) 1847–1856

Fig. 3. Conceptual multilevel model of safety climates framework associated to safety performance. Note: Org_C = Organizational Safety Communication; Org_T = Organizational Safety Training; Org_S = Organizational Safety Systems; Org_V = Organizational Safety Values; Su_R = Supervisor’s Safety Reactions; Su_E = Supervisor’s Safety Efforts; Co_C = Co-workers’ Safety Communication; Co_M = Co-workers’ Safety Mentoring; Co_S = Co-workers’ Safety Systems; Co_V = Co-workers’ Safety Values; B_C = Safety Compliance; B_P = Safety Participation.

H4. In predicting workers’ safety behaviours, a model which considers not only the role of organizational safety climate and supervisor’s safety climate, but also the mediating role of co-workers’ safety climate is more explicative than a model that does not include the co-workers’ role. The present work represents the first attempt to model these relations with a multilevel analysis.

1.2. Safety performance Research classifies work behaviours that relate to safety into categories that parallel the categories for work behaviours that relate to work performance. Borman and Motowidlo (1993) proposed two main components of work performance: task performance and contextual performance. Task performance for workers is defined as ‘‘the activities that are formally recognized as part of their jobs, activities that contribute to the organization’s technical core either directly or indirectly’’ (p. 73). Contextual performance ‘‘supports the organizational, social and psychological environment in which the technical core must function’’ (p. 73). Griffin and Neal (2000) applied the same two categories to differentiate safe and unsafe behaviour in the workplace. Task performance became safety compliance. It refers to activities that are a formally recognized part of the job and that contribute to work safety. They include such things as obeying safety regulations, following the correct procedures and using appropriate equipment. Contextual performance became safety participation. It refers to behaviours that do not directly increase workplace safety, but that help create an atmosphere supportive of safety. Examining the relationship of the organizational climate to each, the study found a stronger relationship to safety participation than to safety compliance. Similarly Christian et al. (2009) found a stronger relationship of group safety climate to safety participation than to safety compliance. These findings suggest the following hypothesis

Table 1 Characteristics of the Participants. Variables

N

%

Gender Male Female

850 137

86 14

Age 18–25 26–35 36–45 46–55 >55

54 229 385 253 36

6 24 40 26 4

Nationality Italian Foreign

745 246

75 25

Educational level <5 y 5–8 y 9–13 y >13 y

56 366 433 118

6 38 44 12

Years of work experience in the company <1 y 47 1–5 y 235 >5 y 658

5 25 70

Injuries involvements in the company in the last 2 years None 614 One 221 More than one 141

63 23 14

Micro-accidents in the last 6 months None One More than one

83 8 9

812 75 90

H5. A model predicting the effect of safety climate explains more of the variance if it centres on safety participation rather than on safety compliance.

1850

M. Brondino et al. / Safety Science 50 (2012) 1847–1856

2. Method 2.1. Participants The sample had 991 blue collar workers (86% men, 14% women). As you can see in Table 1, most of participants were Italian (75%), had 5–13 years of education (82%), had worked for the same company for 5 years or more (70%), and had a permanent contract (66%). A small minority (5%) had worked for the same company for less than 1 year. To get the sample, we contacted the main branches of some metal and mechanical sector companies in the region of Veneto. Veneto is a developed industrial zone with a high rate of workplace accidents (INAIL, 2011), particularly in the metal-mechanical sector, a major sector in the region. The companies varied in size from small (from 0 to 50 employees), to medium (from 50 to 200), to large (200 and beyond). Five companies (one small, two medium and two large companies) agreed to participate to the study; 83% of the blue-collar workers at those companies took part in the interviews (Table 2 shows the characteristics of the five companies). Each response represented the individual level. A two-level design was used, considering the individual level and the work-group level. All data were collected at individual level. To get group level data, we recorded for each participant his or her work-group. The total number of work-groups in the five companies was 91.1 2.2. Measures The present study used the three safety climate measures (Organizational, Supervisor and Co-worker) developed and validated by Brondino et al. (in press) and shown in Appendix A. Each item on each measure has a 7-point scale that ranges from 1 = never to 7 = always. The present study confirmed the reliability of each scale. Organizational safety climate (OSC) is measured with a 12-item scale, in which the worker is asked to judge the safety climate of the entire organization. Through a validation process, Brondino et al. (in press) merged ten items from the Multilevel Safety Climate Scale of Zohar and Luria (2005) with two items from the Safety Climate Scale of Griffin and Neal (2000, personal communication). Each item of OSC scale relates to one of the four domains identified by Griffin and Neal (2000, personal communication): Management values, Safety systems, Safety communication, and Safety training (see Table 3). Management values describe the degree to which managers valued safety in the workplace, and it has items such as ‘‘Top management considers safety when setting production speed and schedules’’. Safety systems refer to the effectiveness of safety systems in the organization, and it has items such as, ‘‘Top management provides all the equipment needed to do the job safely’’. Safety communication refers to how well safety issues are communicated, and it has items such as, ‘‘Top management listens carefully to workers’ ideas about improving safety’’. Safety training refers to the quality and quantity of the employees’ opportunities for training, and it has items such as ‘‘Employees receive comprehensive training in workplace health and safety issues’’. Since the validation of the three safety climate scales is described in Brondino et al. (in press), the present study reports only the inter-item alpha reliability of this scale which was high (a = .93). Furthermore, Construct Reliability (CR) and Average Variance 1 Only groups composed of workers within the same department, working in the same shift and with the same supervisor were selected. Each group had at least four members, due to the assumption that shared perceptions about climate need the presence of at least three-four individuals (Zohar, personal communication)

Extracted (AVE) for each first-order factor had values above the fixed threshold suggested by Hair et al. (1998): values (CR = .80; AVE = .58), safety system (CR = .77; AVE = .53), safety communication (CR = .78; AVE = .54) and training (CR = .80; AVE = .58). Supervisor’s safety climate (SSC) is measured with a 10-item scale, in which the workers had to judge the real importance given to safety by their direct supervisor in the work-group. The SSC is an adjusted version of the Group-level Safety Climate scale by Zohar and Luria (2005). Each item of SSC scale refers to two domains identified as supervisor’s reaction to the workers’ safety behaviours (for example ‘‘My direct supervisor is strict about working safely when we are tired or stressed’’) and supervisor’s own safety behaviour and effort to improve safety (for example ‘‘My direct supervisor uses explanations (not just compliance) to get us to act safely’’) (Meliá and Sesé, 2007; Zohar, 2000) (see Table 3). As with the OSC scale, psychometric properties of SSC scale were assessed with multilevel confirmatory factor analysis in Brondino et al. (in press). This scale had high inter-item reliability (a = .95) and acceptable values for CR and AVE for each first-order factor: first factor (CR = .93; AVE = .69); second factor (CR = .91; AVE = .72). Co-workers’ safety climate (CSC) is measured with a 12-item scale in which the workers rate the degree to which safety is a real priority of their colleagues. Items of the CSC scale were derived from the adjustment to co-workers of the Group-level Safety Climate scale by Zohar and Luria (2005), and comparing the resulted items with items’ content of co-workers’ scales by co-workers’ safety climate literature (e.g. Fugas et al., 2009; Singer et al., 2007; Meliá, 1998; Meliá and Becerril, 2006; Meliá et al., 2008; Jiang et al., 2009). Every item of CSC scale connects to one of the four domains identified by Griffin and Neal (2000, personal communication): Safety mentoring, co-workers’ Values, Safety systems, and Safety communication. We changed the Griffin and Neal’s dimension of ‘training’ into ‘mentoring’, as it was a better fit to the co-workers’ role. The domain of Safety mentoring refers to co-worker activities aimed at helping their colleagues behave more safely, for example giving them suggestions and calling attention to safety (Ensher et al., 2001), for example, ‘‘If it is necessary, my team members use explanations to get other team members to act safely’’. The domain of co-workers’ Values concern the degree to which co-workers value safety in the workplace, represented by items such as ‘‘My team members are careful about working safely also when we are tired or stressed.’’ The domain of Safety systems refer to the attention co-workers pay to safety systems, for example ‘‘My team members are careful that the other members receive all the equipment needed to do the job safely.’’ Finally, the domain of Safety communication refers to the way in which team members discuss safety issues, for example ‘‘My team members talk about safety issues throughout the work week’’. As with the other scales, we assessed the psychometric properties of the scale of the individual perception with multilevel confirmatory factor analysis. The CSC scale had high inter-item reliability (a = .95) and acceptable values for CR and AVE for each first-order factor: Mentoring (CR = .87; AVE = .69), Values (CR = .84; AVE = .63), Safety System (CR = .90; AVE = .75), and Safety Communication (CR = .87; AVE = .68). Safety performance is measured with an 8-item scale which refers to workers safety behaviours. The scale is an adjusted version of Griffin and Neal scale about safety behaviour (2000, personal communication). It measures two components of safety performance: safety compliance and safety participation. Safety compliance is assessed by four items asking about the individual’s performance of safety compliance (for example ‘‘I use all the necessary safety equipment to do my job’’). Safety participation is assessed by four items about participation that support safety in the workplace, but do not necessarily involve a specific activity related to safety (for example ‘‘I put in extra effort to improve the safety of the workplace’’). A model with a second-order factor (safety

1851

M. Brondino et al. / Safety Science 50 (2012) 1847–1856 Table 2 Characteristics of the companies. Company

Products

Company size

Workgroups

Participants

% of participants on the total number of the blue-collars (%)

Micro-accidents in the last 6 months (% of one ore more, self-report) (%)

Injuries in the company (% of one ore more, selfreport) (%)

1

Electric and petrol driven chainsaws, brush cutters and hedge cutters Metal furniture for superand hyper-markets Cooling, conditioning and purifying systems Electrodes and metal wires Excavators and trucks

Large

49

540

55

17

31

Small

13

81

85

41

37

Medium

10

114

95

17

34

Small Medium

6 13

32 224

90 88

19 6

34 53

91

991

82.60

2 3 4 5 Total

Table 3 Results from analysis on between-group variability. Construct

F

Degree of freedom

p

ICC

Org. – safety communication Org. – safety training Org. – safety systems Org. – values Sup. – reaction to workers behaviours Sup. – effort to improve safety Co-w. – safety communication Co-w. – safety mentoring Co-w. – safety systems Co-w. – values Safety compliance Safety participation

3.21 4.74 3.91 3.97 4.17

63 63 63 63 63

<.001 <.001 <.001 <.001 <.001

.14 .22 .18 .18 .20

5.67 2.60 3.03 3.60 3.94 3.32 2.88

63 63 63 63 63 63 63

<.001 <.001 <.001 <.001 <.001 <.001 <.001

.26 .11 .14 .16 .18 .16 .14

behaviour) and two first-order factors (Safety Compliance and Safety Participation) was estimated. Psychometric properties of the scale are assessed with confirmatory factor analysis. Also in this case the estimated model provided a good fit indices, v2(18; N=964) = 47.38, p < .001; TLI = .98, CFI = .99; SRMR = .023. The inter-item reliability of this scale was high (a = .84), and the CR and AVE for each first-order factor were acceptable: Compliance (CR = .83; AV = .54) and Participation (CR = .73; AVE = .40).

2.2.1. Other questions in the questionnaire The survey also gathered background information on each participant: their gender, age, educational level, nationality, length of employment in the company, kind of job-contract, department, and work shift.

2.3. Procedures A few days before administering the questionnaire, either the top management organized an ad hoc meeting with unions, the Safety Commission and the safety officer, or a trade-union meeting was held. In each meeting, workers learned that they were part of a larger sample of workers involved in a research supported by INAIL,2 and received information about the research program. Participants were informed that the questionnaire was anonymous, that all data would be collected and protected by the research team, 2 INAIL (Istituto Nazionale per l’Assicurazione contro gli Infortuni sul Lavoro, the OSH national institution of Italy) is an Italian institution pursuing several objectives: the reduction of accidents at work, the insurance of workers involved in risky activities; the re-integration in the labor market and in social life of work accident victims.

and that only aggregate results would be given to the management of the company. All participants answered the questionnaire during working hours, at the end or at the beginning of their work shift, and were asked to answer as honestly as possible. They were told that questions referred to their own perception of the organizational management, their direct supervisor, and their work-group coworkers about safety at work. They were also told that, in case they found difficult to answer to an item, due to lack of knowledge about something such as organizational policy, they should choose the answer closest to their perception. At the end of the questionnaire participants were asked to answer questions about some socio-demographic data. Along with the Italian questionnaire, English and French translations were also provided for foreign workers. If needed, researchers were available to help participants. The procedure took about 20 min. 2.4. Data analysis To model relations among variables at multiple levels, we used multilevel structural equation modelling (ML-SEM) with full maximum likelihood estimation in Mplus 5.2 (Muthén and Muthén, 1998–2008). In ML-SEM the variability in variables is decomposed into two latent components, a within-group component (variability at individual level), and a between-group component (variability at group level) (Muthén and Asparouhov, 2010). ML-SEM permits to model the relationships among these variance components within each level through the specification of measurement and structural models. At the individual level, variables were specified as having intercepts (the means for each group) that vary across groups (random intercepts). Therefore, at the group level this random intercepts are modelled as latent variables (see Fig. 3). To perform the ML-SEM analysis, the present study followed the steps recommended by Preacher et al. (2010) and Muthén (1994). The analysis only considered two levels, work-group and individual level, though it could have also considered the organizational level, because of the complexity of the model and the limited number of organizations. To run the ML-SEM analysis, we first had to run some preliminary operations. First, we had to examine the between-group variability. We did this by computing the intraclass correlation (ICC) (which varies from 0 to 1) for each variable of the three climate scales (OSC, SSC, and CSC).3 If values are close to zero (e.g. .05) the multilevel modelling will be meaningless (Dyer et al., 2005). 3 Muthén (1994) suggested estimating a unique type of ICC to determine potential group influence. Muthen’s ICC index is conceptually similar to ICC(1). However, Muthen’s ICC is obtained by random effects ANOVA, while ICC(1) is obtained by fixed effects ANOVA.

1852

M. Brondino et al. / Safety Science 50 (2012) 1847–1856

Table 4 Descriptive statistics for study variables. Construct

Mean

SD

OSC. S. comm. OSC. S. train. OSC. S. system OSC. S. values SSC. reactions SSC. effort CSC. S. comm. CSC. S. Train. CSC. S. system CSC. S. values Compliance Participation

3.83 4.40 4.52 3.83 4.15 3.78 3.38 3.76 3.42 3.81 5.49 4.74

1.53 1.54 1.43 1.53 1.76 1.75 1.54 1.68 1.65 1.59 .99 1.16

(3.86) (4.42) (4.57) (3.88) (4.21) (3.81) (3.45) (3.83) (3.52) (3.89) (5.55) (4.80)

(.89) (1.01) (.86) (.91) (1.25) (1.14) (.74) (.93) (.94) (.94) (.44) (.53)

1

2

– .71 .67 .73 .58 .63 .37 .43 .36 .46 .27 .35

.83 – .70 .68 .56 .57 .37 .41 .36 .43 .27 .33

3 .97 .90 – .73 .62 .58 .34 .43 .35 .47 .34 .33

4

5

.94 .85 .98 – .62 .63 .37 .45 .39 .52 .33 .36

.86 .72 .86 .84 – .82 .47 .55 .46 .54 .36 .37

6 .87 .86 .90 .87 .94 – .41 .54 .47 .58 .33 .32

7

8

.81 .63 .84 .83 .76 .77 – .73 .76 .67 .26 .43

.90 .77 .95 .92 .80 .83 .93 – .74 .75 .32 .42

9 .89 .72 .91 .91 .84 .83 .88 .92 – .67 .31 .41

10

11

.89 .83 .93 .89 .85 .84 .93 .96 .93 – .34 .42

.74 .67 .80 .86 .77 .63 .79 .73 .84 .75 – .52

12 .79 .68 .83 .87 .81 .70 .84 .78 .88 .79 .99 –

Note: Means and standard deviations without parentheses are based on individual-level data (N = 895) and means and standard deviations in parentheses are based on grouplevel data (N = 64). Correlations below the diagonal are based on individual-level data and correlations above the diagonal are based on group-level data. All individual-level correlations and group level correlations are significant at . p < .05. p < .01. p < .001.

We also assessed the homogeneity of climate perceptions with the median value of rwg(j) (Bliese, 2000) for each work group.4 We used this method to ensure a sufficient level of within-group agreement (inside the groups) in the variables of substantive interest at the group level. Next, we performed an investigation of a properly specified within-group model, fitting the full model allowing the group-level constructs to freely co-vary. In the third step, we analysed the multilevel structural model illustrated in Fig. 3, to test hypotheses from H1 to H3. Then, we fitted the Zohar and Luria’s model (Zohar and Luria, 2005; see Fig. 1) to ML-SEM, to compare it with the new proposed model. Here we sought to find out if adding co-worker safety climate as mediator between supervisor safety climate and safety behaviours accounted for more variability than a model without the co-worker role (H4). Finally, to test whether the proposed model predicts safety participation better than safety compliance, we split safety behaviours into the two components (H5). We evaluated the goodness of fit of the models with the Tucker Lewis Index (TLI; Tucker and Lewis, 1973), the comparative fit index (CFI; Bentler, 1990), the root mean square error of approximation (RMSEA; Hu and Bentler, 1999), and the standardized root mean square residual (SRMR). To compare the models, we used the Akaike Information Criterion (AIC; Akaike, 1974), and the Bayesian Information Criterion (BIC; Schwarz, 1978). 2.5. Descriptive statistics and aggregation analysis We had to conduct three preliminary analyses before testing the model. First, we calculated the frequency of missing values for each variable in the sample, and removed all cases with more than 5% missing values (Chemolli and Pasini, 2007). This resulted in removing 28 cases (3% of the full sample). To make sure that these deletions did not invalidate our sample, we compared the socio-demographic characteristics of the samples with and without the removed cases. The v2 tests showed that the deletions were equally distributed among the various socio-demographic characteristics of the sample. Then, we analysed the composition of the work groups and homogeneity of the perceptions of the safety climate, and eliminated work groups that did not satisfy the critical conditions. These preliminary analyses resulted in a sample having 895 cases and 64 4 Agreement was evaluated using LeBreton and Senter’s (2008) revised standards for interpreting inter-rater agreement estimates. For each group-level construct, organizational, supervisor and co-worker safety climate, the level of agreement (median values P.70) supported their inclusion (LeBreton and Senter, 2008).

work groups. Table 3 reports the results for the variability between groups to support multilevel analyses. The analyses showed significant between-group variance for all variables, with ICCs ranging from .11 (safety communication between co-workers) to .26 (supervisor’s reaction to workers safety behaviours). The links of group membership to individual observations underlines the importance of conducting a ML-SEM. We also analysed the median values of rwg(j) across groups and found strong homogeneity within groups for organizational safety climate (.87), supervisor safety climate (.70) and co-worker safety climate (.85). Finally, for each indicator, we calculated the mean and standard deviation, and, to check for normal distribution, we calculated the skewness and kurtosis. All the items with values into the range 1/+1 were considered normally distributed. The responses approximated a normal distribution. Skewness ranged from .61 to .58 and kurtosis ranged from 1.17 to .62. For kurtosis, the unique value out of the range ( 1.17) came from the supervisor’s reactions to workers behaviours. It did not represent a problem, because the mean kurtosis (|M| = .74) was less than 1.0 (Muthén and Kaplan, 1985). Table 4 shows means, standard deviations, and bivariate correlations for the measures used in the present study. The means show that respondents perceive high levels of safety climate for each safety agent.

3. Results To test the hypothesized multilevel structural model shown in Fig. 3, we first had to estimate the measurement model. It used the mean values of the first order factors as the observed variables. The measurement model had a good fit to the data (see Table 5). All factor loadings were statistically significant, suggesting that all indicators adequately reflected the latent constructs. The fit indices for the within-group structural model were also good (see Table 5). Then, we simultaneously estimated the within-group and between-group structural models. The ML-SEM model again showed good fit indices (see Table 5). The accounted variance in supervisor’s safety climate, in co-workers’ safety climate, and in safety behaviours were, at individual level, 63%, 44% and 38% respectively, and at group-level 83%, 91% and 76% respectively. Finally, to test the mediation role of SSC and CSC (H2 and H3a), we estimated an alternative model which added a direct path between OSC and safety behaviours. Fig. 4 (final model) shows this model along with path estimates. As expected, the final model had a better fit than the previous model (Dv2(2, N=895) = 13.85, p < .001), and as a result, we chose it as the best one. Its other indices of fit were good (see Table 5), and the AIC and BIC indices were

1853

M. Brondino et al. / Safety Science 50 (2012) 1847–1856 Table 5 Fit indexes for measurement and structural models. Model

v2 (df)

Measurement model Within model Hypothesized multilevel model Final multilevel model Model OSC ? SSC ? Beh. Model OSC ? CSC ? Beh. Final mod. with safety compliance Final mod. with safety participation

364.62 378.24 380.83 366.98 144.77 226.19 349.72 343.14

(99) (100) (101) (99) (38) (68) (82) (82)

p

CFI

TLI

RMSEA

SRMRw

SRMRb

.001 .001 .001 .001 .001 .001 .001 .001

.96 .96 .96 .96 .97 .97 .96 .96

.95 .95 .95 .95 .96 .96 .94 .95

.06 .06 .06 .06 .06 .05 .06 .06

.04 .04 .04 .04 .02 .03 .03 .03

.05 .05 .05 .05 .05 .05 .11 .11

significant (indirect effect = .44, p < .001). This result, in combination with the presence of direct effect of OSC on CSC, indicates a partially mediated relationship between the two constructs. At the group level, SSC did not mediate the relation between OSC and CSC. H3a and H3b: That CSC mediated the relationship between OSC and safety behaviours, and between SSC and safety behaviours were supported at individual level. H3a: The relationship between OSC and safety behaviours was partially mediated by CSC (indirect effect = .25, p < .001). It was partially mediated, because of the statistically significant coefficient of the direct path between OSC and safety behaviours. H3b: The relationship between SSC and safety behaviours was fully mediated by CSC (indirect effect = .24, p < .001). As for the H2, at group level neither hypothesis (H3a and H3b) was supported. At the group level none of the relationships between safety climates and safety behaviours were statistically significant (see also Footnote 5). To better understand these results, we tested two models that separately analysed the mediating role of SSC (Fig. 5) and CSC (Fig. 6) in the relationship between OSC and safety behaviours. Fig. 4. Results for Final Integrated Model. Note: OSC = Organizational Safety Climate; SSC = Supervisor’s Safety Climate; CSC = Co-workers’ Safety Climate; BEH = Safety Behaviours. Beside latent variables accounted variability is shown.  p < .05. p < .01. p < .001.

nearly equal to those in the previous model (previous model AIC = 30279 vs final model AIC = 30289; previous model BIC = 30610 vs final model BIC = 30610). The direct path between OSC and safety behaviours was statistically significant at individual level (b = .25, p < .01) but not at group level5 (b = .42, n.s.). In the final model, the relationship between SSC and safety behaviours ceased to be statistically significant (b = .02, n.s., in the within-group level part of the SEM; in the previous model b = .23, p < .01). This result means that SSC did not mediate the relationship between OSC and safety behaviours. Now consider the results for each hypothesis within the framework of the final model. H1: That OSC positively predicts SSC and CSC, was confirmed: OSC had a strong positive and statistically significant relationship with SSC at both the individual level (b = .79, p < .001) and the group level (b = .91, p < .001). Also OSC had a positive and statistically significant relationship with CSC at the individual level (b =.14, p < .05) and at the group level (b = .87, p < .001). Note that the relationship between OSC and CSC was stronger at group level than at individual level. H2: That SSC mediated the relationship between OSC and CSC was supported but only at the individual level. At the individual level, the indirect effect of OSC on CSC was positive and statistically

5 Considering the between-group level or the ML-SEM, no path achieved significance except the link between OSC and the SSC and CSC climates. Because of the small sample size at the group level, the insignificant results should be treated with caution. It does not necessarily indicate the absence of the link.

Fig. 5. Results of the Model with Supervisor’s Mediating Role. Note: OSC = Organizational Safety Climate; SSC = Supervisor’s Safety Climate; CSC = Co-workers’ Safety Climate; BEH = Safety Behaviours. Beside latent variables accounted variability is shown. p < .05. p < .01. p < .001.

Fig. 6. Results of the Model with co-workers’ Mediating Role. Note: OSC = Organizational Safety Climate; SSC = Supervisor’s Safety Climate; CSC = Co-workers’ Safety Climate; BEH = Safety Behaviours. Beside latent variables accounted variability is shown. p < .05. p < .01. p < .001.

1854

M. Brondino et al. / Safety Science 50 (2012) 1847–1856

Both models had good fit indices (see Table 5), and both models supported the mediating role of SSC and of CSC: SSC partially mediated the relationship at individual level (indirect effect = .21, p < .001; direct effect: .32, p < .001) and fully mediated it at group level (indirect effect = .76, p < .001; direct effect: .02, n.s.); CSC partially mediated the relationship between OSC and safety behaviours at individual level (indirect effect = .25, p < .001; direct effect = .26, p < .001), and fully mediated it at group level (indirect effect = .83, p < .001; direct effect = .07, n.s.). H4: That adding CSC as a mediating variable explained more of the variance than models without it was fully supported. At the individual level, the model with CSC as a mediating variable explained more of the variance than did the model without it (39% vs 31%), and at the group level, the model with CSC as a mediating variable explained more of the variance than did the model without it (75% vs. 69%). Finally, for H5: That a model predicting effects of safety climate explains more of the variance if it centres on safety participation than on safety compliance, we estimated two new models. One replaced the latent construct ‘‘safety behaviours’’ with its safety compliance component; and the other replaced safety behaviours with its safety participation component. As hypothesized, the model predicting safety participation accounted for more of the variance than did the model predicting safety compliance at both the individual level (participation, 26% vs. compliance 17%) and the group level (participation, 81% vs., compliance, 77%).

4. Discussion and future directions The present study sought to determine if co-worker safety climate predicts safety behaviours and mediates between organizational safety climate and safety behaviours, and between supervisor safety climate and safety behaviours. Our unique approach of using a multilevel structural equation model maintained the complexity of the integrated structure of the safety climates and allowed a better understanding of the relationships between constructs at different organizational level. As predicted, improvements in the organizational safety climate yielded statistically significant improvements in co-worker and supervisor safety climate at the individual and group level; coworker safety climate had a stronger mediating role than did supervisor safety climate at the individual and group level. Neither one had a statistically significant mediation role at the group level and at the individual level, but at the individual level co-worker safety climate mediated at a statistically significant level the relationships between organizational safety climate and safety behaviours, and between supervisor’s safety climate and safety behaviours, but supervisor safety climate only partially mediated the relationship between organizational safety and co-worker safety climate. Perhaps, at the group level the association of supervisor’s safety climate and co-workers’ safety climate reduces or cancels the effects of both on safety behaviours. In particular, the models indicate that coworker safety climate may reduce the effect of supervisor’s safety climate. These findings agree with Chiaburu and Harrison (2008) findings that co-worker support was a better predictor than was leader support of many employee outcomes. These findings suggest that as Chiaburu and Harrison (2008) suggested future research should explore the lateral relationships of supervisor and co-worker safety climate as these relationships may well have reciprocal influences (e.g. additive, interactive, or compensatory) that previous theory and research has missed. The integrated model of safety climate predicted safety participation better than safety compliance. These results confirm previous findings (e.g. Griffin and Neal, 2000; Christian et al., 2009) that

safety climate has a greater influence on behaviours that are contextual, because workers must by definition comply with obligatory procedures and practices. If so, then when individuals believe that their work environment is safe, they will reciprocate by putting extra effort into discretionary activities for safety. Therefore, organizations, seeking to improve safety, might do well to focus more on improving the safety climate than on blaming and punishing individuals who fail to comply with standard work procedures. Our findings suggest key intervention points to improve workplace safety. Interventions that focus on improving safety communication among colleagues, or co-workers’ commitment to safety could improve safety performance. Safety-training programs should target teams and work groups. Such programs can both convey knowledge about safety and strengthen group norms for safety. They need have relevance to the work-group context. For example, programs could use grouptraining with ‘‘hands on’’ exercises, simulation and role-playing in work (safety relevant) situations and focus on safety responsibilities for safety initiatives/actions around other co-workers. It would also help to use short safety meetings (once a week or month) where co-workers play an active role in discussing safety issues (both good initiatives and problems), and propose ways to improve safety, following a continuous improvement model. Finally, it would help to publically reward or acknowledge work groups or co-workers for successful safety performance, particularly where workers work in teams close together. While the present study supports previous findings (e.g. Meliá et al., 2008) that stress the need to consider co-workers in studies of organizational and group safety climate, it expands on the previous findings by using a more complex operationalization of safety climate (Brondino et al., in press), considering contextual behaviours and using a multilevel approach. This study used self-report measures for all dimensions of safety climate. To avoid potential confounding by common method variance of the estimates of the relationships between the measures, future research could obtain independent measures of each dimension. It could also obtain objective measures of behaviours to find out how well the self-report measures reflect the relationship between the integrated safety climate and safety behaviours. It could use a larger sample to both increase the power and allow the specification of random slopes to assess cross-level interactions, and it could benefit from getting data from a larger number of organizations. To better understand the dynamics of the full integrated system, future research could expand the model in several ways. Recent work suggests the need to study not only climate level, as was done here, but also climate strength (e.g. Zohar and Luria, 2005). The relationships between climate and outcomes are generally greater within a strong climate. The present work centred on groups that had a strong climate, because it sought to understand relationships and as the presence of a weak climate might disturb the analysis of the relationships. Future research could consider the potential moderating role of climate strength. Finally, it could assess the mediating role of the safety performance determinants safety knowledge and safety motivation (Campbell et al., 1993; Neal et al., 2000). Research on the mediating role of these constructs found that they predict safety performance (Christian et al., 2009; Sinclair et al., 2008). Studying these relationships, integrated in a larger system of variables, with a multilevel approach, could help us to understand the mechanisms that influence safety behaviours at different organizational levels and therefore helping to have instruments to understand how to improve safety in an ever more effective way. In sum, although the safety climate literature has developed a solid knowledge base for use by organizations to improve safety,

M. Brondino et al. / Safety Science 50 (2012) 1847–1856

the present study both adds to that knowledge and reveals important ‘‘avenues’’ for future research.

1855

B.5. Co-workers’ safety climate scale B.5.1. Safety communication

Acknowledgments We gratefully acknowledge the financial support by Istituto Nazionale per l’Assicurazione contro gli Infortuni sul Lavoro (INAIL, the OSH national institution of Italy) of Vicenza and by INAIL of the Veneto Region, and by the three main Italian union federations of metal workers (Federazione Italiana Metalmeccanici (FIM), Federazione Impiegati e Operai Metallurgici (FIOM) and Unione Italiana Lavoratori Metalmeccanici (UILM). We gratefully acknowledge the financial support by INAIL Vicenza and FIM, FIOM and UILM Veneto and participation of the workers of companies which participated at the survey.

Team members’ speaking on safety on the week Team members’ discussing about incident prevention Team members’ discussion about safety hazard B.5.2. Safety mentoring Team members’ emphasis to peer on safety care when under pressure Team members’ care of peers’ safety awareness Team members’ mentoring to peer about working safely B.5.3. Safety values

Appendix A. The Integrated Organizational Safety Climate questionnaire with the short description of items and the specification of the dimensions B.3. Organizational safety climate scale B.3.1. Safety communication Space to discuss in meeting Management attention to workers ideas to improve safety Workers consultation on safety issues

Team members’ safety care at the shift end Team members’ safety care when tired Team members’ safety care when a delay in production schedule occurs B.5.4. Safety systems Team members care to other workers’ safety equipment Team members remind safety equipment use Team members care to other members’ safety compliance

B.3.2. Safety training Information supply on safety issues Investments on safety training Quality of safety training B.3.3. Safety values Management safety care in production schedule Management safety care in moving-promoting people Management safety care on a delay in production schedule B.3.4. Safety systems Management effort on safety improvement Management reaction to solve safety hazard Power given to safety officers B.4. Supervisor safety climate scale B.4.1. Supervisor’s effort to improve safety Supervisor’s care of safety rules when a delay in production schedule occurs Supervisor’s care to provide workers needed safety equipment Supervisor’s care to the use of safety equipment Supervisor’s care of safety rules when workers are tired Supervisor’s care to all safety rules Supervisor controls the compliance of all the workers B.4.2. Supervisor’s reactions to workers behaviours Supervisor discusses with workers on safety improvement Supervisor’s care to workers safety awareness Supervisor’s coaching about safety care Supervisor praise to very careful safety behaviours

References Ashforth, B.E., 1985. Climate formation: issues and extensions. Academy of Management Review 10, 837–847. Bandura, A., 1986. Social Foundations for Thought and Action: A Social Cognitive Theory. Prentice Hall, Englewood Cllifs, NJ. Bentler, P.M., 1990. Comparative fit indexes in structural models. Psychological Bulletin 107 (2), 238–246. Bliese, P.D., 2000. Within-group agreement, non-independence, and reliability. In: Kline, K.J., Kozlowski, S.W. (Eds.), Multilevel Theory, Research, and Methods in Organizations. Jossey-Bass, San Francisco, pp. 349–381. Borman, W.C., Motowidlo, S.J., 1993. Expanding the criterion domain to include elements of contextual performance. In: Schmitt, N., Borman, W.C., Associates (Eds.), Personnel Selection in Organizations. Jossey-Bass, San Francisco, pp. 71– 98. Brondino, M., Pasini, M., Silva, S.A., in press. Development and validation of an integrated organizational safety climate questionnaire with multilevel confirmatory factor analysis. Quality and Quantity. Burt, C.D.B., Sepie, B., McFadden, G., 2008. The development of a considerate and responsible safety attitude in work teams. Safety Science 46, 79–91. Campbell, J.P., McCloy, R.A., Oppler, S.H., Sager, C.E., 1993. A theory of performance. In: Schmitt, N., Borman, W.C., & Associates (Eds.), Personnel Selection in Organizations. Jossey-Bass, San Francisco, pp. 35–70. Cavazza, N., Serpe, A., 2009. Effects of safety climate on safety norm violations: exploring the mediating role of attitudinal ambivalence toward personal protective equipment. Journal of Safety Research 40, 277–283. Chemolli, E., Pasini, M., 2007. I dati mancanti. DIPAV 20, 51–56. Chiaburu, D.S., Harrison, D.A., 2008. Do peers make the place? Conceptual synthesis and meta-analysis of co-worker effects on perceptions, attitudes, OCBs, and performance. Journal of Applied Psychology 93, 1082–1103. Christian, M.S., Bradley, J.C., Wallace, J.C., Burke, M.J., 2009. Workplace safety: a meta-analysis of the roles of person and situation factors. Journal of Applied Psychology 94, 1103–1127. Deutsch, M., Gerard, H.B., 1955. A study of normative and informational social influences upon individual judgment. Journal of Abnormal and Social Psychology 51, 629–636. Dyer, N.G., Hanges, P.J., Hall, R.J., 2005. Applying multilevel confirmatory factor analysis techniques to the study of leadership. Leadership Quarterly 16, 149– 167. Ehrhart, M., Naumann, S.E., 2004. Organizational citizenship behavior in work groups: a group norms approach. Journal of Applied Psychology 89, 960–974. Ensher, E.A., Thomas, C., Murphy, S.E., 2001. Comparison of traditional, step-ahead, and peer mentoring on protégés’ support, satisfaction, and perceptions of career success: a social exchange perspective. Journal of Business and Psychology 15, 419–438.

1856

M. Brondino et al. / Safety Science 50 (2012) 1847–1856

Fugas, C., Silva, S.A., Meliá, J.L., 2009. Exploratory and confirmatory analysis of the relationship between social norms and safety behavior. In: Martorell, S., Soares, C.G., Barnett, J. (Eds.), Safety, Reliability and Risk Analysis: Theory, Methods and Applications. Taylor & Francis Group, London. Fugas, C., Melia, J.L., Silva, S.A., 2011. The ‘‘is’’ and ‘‘the ought’’: how perceived social norms influence safety behaviors at work? Journal of Occupational Health Psychology 16 (1), 67–79. Glendon, I., 2008. Safety culture: snapshot of a developing concept. Journal of Occupational Health Safety 24 (3), 179–189. Griffin, M.A., Neal, A., 2000. Perceptions of safety at work: a framework for linking safety climate to safety performance, knowledge, and motivation. Journal of Occupational Health Psychology 5, 347–358. Gyekye, S.A., Salminen, S., 2009. Educational status and organizational safety climate: does educational attainment influence workers’ perceptions of workplace safety? Safety Science 47, 20–28. Hair, J.F., Anderson, R.E., Tatham, R.L., Black, W., 1998. Multivariate Data Analysis, fifth ed. Prentice-Hall, Englewood Cliffs, NJ. Hu, L., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling 6, 1–55. INAIL, 2011. Rapporto annuale 2010. Milano, Italy. Jiang, L., Yu, G., Li, Y., Li, F., 2009. Perceived colleagues’ safety knowledge/behavior and safety performance: safety climate as a moderator in a multilevel study. Accident Analysis and Prevention 42 (5), 1468–1476. Kath, L.M., Marks, K.M., Ranney, J., 2010. Safety climate dimensions, leader-member exchange, and organizational support as predictors of upward safety communication in a sample of rail industry workers. Safety Science 48 (5), 643–650. Kozlowski, S.W., Klein, K.J., 2000. A multilevel approach to theory and research in organizations: contextual, temporal, and emergent processes. In: Kline, K.J., Kozlowski, S.W. (Eds.), Multilevel Theory, Research, and Methods in Organizations. Jossey-Bass, San Francisco, pp. 3–90. LeBreton, J.M., Senter, J.L., 2008. Answers to 20 questions about interrater reliability and interrater agreement. Organizational Research Methods 11, 815–852. Lingard, H., Cooke, T., Blismas, N., 2011. Coworkers’ response to occupational health and safety: an overlooked dimension of group-level safety climate in the construction industry? Engineering, Construction and Architectural Management 18, 159–175. Meliá, J.L., 1998. A psychosocial causal model of work accidents. Anuario de Psicologia 29, 25–43. Meliá, J.L., Becerril, M., 2006. Safety climate dimensions from the ‘‘agent’’ point of view. In: Mondelo, P., Mattila, M., Karwowski, W., Hale, A. (Eds.), Proceedings of the Fourth International Conference on Occupational Risk Prevention. Seville. Meliá, J.L., Sesè, A., 2007. Supervisor’s safety response: a multisample confirmatory factor analysis. Psicothema 19 (002), 231–238. Meliá, J.L., Mearns, K., Silva, S., Lima, M.L., 2008. Safety climate responses and the perceived risk of accidents in the construction industry. Safety Science 46, 949– 958.

Morrow, S.L., McGonagle, A.K., Dove-Steinkampa, M.L., Walker Jr., C.T., Marmeta, M., Barnes-Farrella, J.L., 2010. Relationships between psychological safety climate facets and safety behavior in the rail industry: a dominance analysis. Accident Analysis and Prevention 42 (5), 1460–1467. Muthén, B.O., 1994. Multilevel covariance structure analysis. Sociological Methods and Research 22, 376–398. Muthén, B., Asparouhov, T., 2010. Beyond multilevel regression modeling: multilevel analysis in a general latent variable framework. In: Hox, J., Roberts, J.K. (Eds.), The Handbook of Advanced Multilevel Analysis. Taylor and Francis. Muthén, B., Kaplan, D., 1985. A comparison of some methodologies for the factor analysis of non-normal Likert variables. British Journal of Mathematical and Statistical Psychology 38, 171–189. Muthén, L.K., Muthén, B.O., 1998–2008. Mplus User’s Guide: Statistical Analysis with Latent Variables, fifth ed. Muthén & Muthén, Los Angeles. Neal, A., Griffin, M.A., Hart, P.M., 2000. The impact of organizational climate on safety climate and individual behavior. Safety Science 34, 99–109. Preacher, K.J., Zyphur, M.J., Zhang, Z., 2010. A general multilevel SEM framework for assessing multilevel mediation. Psychological Methods 15 (3), 209–233. Shannon, H.S., Norman, G.R., 2009. Deriving the factor structure of safety climate scales. Safety Science 47, 327–329. Sinclair, R.R., Martin, J.E., Sears, L.E., 2008. Labor unions and safety climate: perceived union safety values and retail employee safety outcomes. Accident Analysis and Prevention 42 (5), 1477–1487. Singer, S., Meterko, M., Baker, L., Gaba, D., Falwell, A., Rosen, A., 2007. Workforce perceptions of hospital safety culture: development and validation of the patient safety climate in healthcare organizations survey. Health Services Research 42 (5), 1999–2021. Tucker, L., Lewis, C., 1973. A reliability coefficient for maximum likelihood factor analysis. Psychometrika 38 (1), 1–10. Wallace, C., Chen, G., 2006. A multilevel integration of personality, climate, regulatory focus, and performance. Personnel Psychology 59, 529–557. Westaby, J.D., Lowe, J.K., 2005. Risk-taking orientation and injury among youth workers: examining the social influence of supervisors, coworkers, and parents. Journal of Applied Psychology 90, 1027–1035. Zohar, D., 2000. A group-level model of safety climate: testing the effect of group climate on micro-accidents in manufacturing jobs. Journal of Applied Psychology 85, 587–596. Zohar, D., 2008. Safety climate and beyond: a multi-level multi-climate framework. Safety Science 46, 376–387. Zohar, D., 2010. Safety climate: conceptual and measurement issues. In: Quick, J.C., Tetrick, L.E. (Eds.), Handbook of Occupational Health Psychology. American Psychology Association, Washington, DC, pp. 123–142. Zohar, D., Luria, G., 2005. A multilevel model of safety climate: cross-level relationships between organization and group-level climates. Journal of Applied Psychology 90, 616–628. Zohar, D., Tenne-Gazit, O., 2008. Transformational leadership and group interaction as climate antecedents: a social network analysis. Journal of Applied Psychology 93, 744–757.