9
Empowering team leadership and safety performance in nuclear power plants: A multilevel approach Mario Martínez-Córcoles a,, Francisco J. Gracia a , Inés Tomás a , José M. Peiró a,b , Markus Schöbel c a Research Institute on Personnel Psychology, Organizational Development, and Quality of Working Life (IDOCAL), University of Valencia, Spain b Valencian Institute of Economic Research, Spain c Center for Economic Psychology, University of Basel, Spain article info Article history: Received 22 November 2011 Received in revised form 26 June 2012 Accepted 1 August 2012 Available online 1 September 2012 Keywords: Team leadership Safety performance Risky behavior Safety compliance Safety participation Nuclear power plant abstract Despite the large body of work on team leadership, hardly any literature has dealt with team leadership in safety performance settings. The goal of the present study is to analyze how team leader behaviors influence team members’ safety performance in nuclear power plants. For this purpose, an empowering leadership approach was assessed. We consider a multilevel model in which safety performance is divided into three types of behaviors. The sample was composed of 479 workers in 54 groups from two Spanish nuclear power plants. The results suggested that leaders’ empowering behaviors generated higher safety compliance behaviors and higher safety participation behaviors by team members, whereas risky behaviors were reduced. Empirical support was found for hierarchical linear modeling linking lead- ership and safety performance behaviors. Practical implications, study limitations and directions for future research are discussed. Ó 2012 Published by Elsevier Ltd. 1. Introduction The role of leadership is of prime importance for the safety of nuclear power plants (NPPs). In general, leadership is viewed as a shift lever for safety culture, and as an important antecedent of achieving high levels of safety (e.g., Fahlbruch, 2005; IAEA, 2002, 2007, 2008). Although safety leadership is promoted through sem- inars, simulator trainings or safety culture reviews in the nuclear industry, only a few studies have empirically analyzed and speci- fied the impact of leadership on safety performance (e.g., Martí- nez-Córcoles et al., 2011; Yule et al., 2007). The question of which leadership style might best fit within the highly regulated work context of nuclear power plants still remains unanswered. Whereas leadership theories are primarily concerned with enhanc- ing the effectiveness and efficiency of employees’ work perfor- mance, it is questionable whether these promoted leadership styles will obtain the same positive results within a work context where the trade-off between efficient and safe performance is sometimes crucial. The present study aims to examine the differ- ential impact of safety leadership on safety performance in that context. Using a multi-level approach, we tested the effects of an empowering team leadership style (ELQ, Arnold et al., 2000) on safety performance. As a result, we introduce three dimensions for the safety performance construct, namely, safety compliance, safety participation and risky behavior. In the next two sections, we introduce research on safety leadership and the leadership model utilized in our study. 1.1. Safety leadership Empirical findings in safety research have shown that leadership constitutes the strongest factor affecting organizational safety per- formance. Many of these studies focus on well-known leadership approaches, namely leader-member exchange (LMX) (Dansereau et al., 1975; Graen and Cashman, 1975) and transformational lead- ership (Bass, 1985, 1990), and they have been applied to several industrial sectors, such as manufacturing, metal processing, con- struction or food service (e.g., Barling et al., 2002; Clarke and Ward, 2006; Hofmann and Morgeson, 1999; Michael et al., 2006; Mullen and Kelloway, 2009; Simard and Marchand, 1994; Zohar, 2000, 2002). For instance, Hofmann et al. (2003), using a multilevel ap- proach, demonstrated with transportation members of the US Army that high-quality leader–member exchange relationships expanded safety citizenship role definitions when there was a positive safety climate, and that there was no such expansion in a less positive safety climate. Both leader–member exchange and safety citizen- ship role definitions were positively related to safety citizenship behavior. Research on transformational full-range leadership 0925-7535/$ - see front matter Ó 2012 Published by Elsevier Ltd. http://dx.doi.org/10.1016/j.ssci.2012.08.001 Corresponding author. Tel.: +34 665784288. E-mail addresses: [email protected] (M. Martínez-Córcoles), francisco.gra- [email protected] (F.J. Gracia), [email protected] (I. Tomás), [email protected] (J.M. Peiró), [email protected] (M. Schöbel). Safety Science 51 (2013) 293–301 Contents lists available at SciVerse ScienceDirect Safety Science journal homepage: www.elsevier.com/locate/ssci

Empowering team leadership and safety performance in nuclear power plants: A multilevel approach

Embed Size (px)

Citation preview

Safety Science 51 (2013) 293–301

Contents lists available at SciVerse ScienceDirect

Safety Science

journal homepage: www.elsevier .com/locate /ssc i

Empowering team leadership and safety performance in nuclear power plants:A multilevel approach

Mario Martínez-Córcoles a,⇑, Francisco J. Gracia a, Inés Tomás a, José M. Peiró a,b, Markus Schöbel c

a Research Institute on Personnel Psychology, Organizational Development, and Quality of Working Life (IDOCAL), University of Valencia, Spainb Valencian Institute of Economic Research, Spainc Center for Economic Psychology, University of Basel, Spain

a r t i c l e i n f o

Article history:Received 22 November 2011Received in revised form 26 June 2012Accepted 1 August 2012Available online 1 September 2012

Keywords:Team leadershipSafety performanceRisky behaviorSafety complianceSafety participationNuclear power plant

0925-7535/$ - see front matter � 2012 Published byhttp://dx.doi.org/10.1016/j.ssci.2012.08.001

⇑ Corresponding author. Tel.: +34 665784288.E-mail addresses: [email protected] (M. Mart

[email protected] (F.J. Gracia), [email protected] (I. Tomás), [email protected] (M. Schöbel).

a b s t r a c t

Despite the large body of work on team leadership, hardly any literature has dealt with team leadershipin safety performance settings. The goal of the present study is to analyze how team leader behaviorsinfluence team members’ safety performance in nuclear power plants. For this purpose, an empoweringleadership approach was assessed. We consider a multilevel model in which safety performance isdivided into three types of behaviors. The sample was composed of 479 workers in 54 groups fromtwo Spanish nuclear power plants. The results suggested that leaders’ empowering behaviors generatedhigher safety compliance behaviors and higher safety participation behaviors by team members, whereasrisky behaviors were reduced. Empirical support was found for hierarchical linear modeling linking lead-ership and safety performance behaviors. Practical implications, study limitations and directions forfuture research are discussed.

� 2012 Published by Elsevier Ltd.

1. Introduction

The role of leadership is of prime importance for the safety ofnuclear power plants (NPPs). In general, leadership is viewed as ashift lever for safety culture, and as an important antecedent ofachieving high levels of safety (e.g., Fahlbruch, 2005; IAEA, 2002,2007, 2008). Although safety leadership is promoted through sem-inars, simulator trainings or safety culture reviews in the nuclearindustry, only a few studies have empirically analyzed and speci-fied the impact of leadership on safety performance (e.g., Martí-nez-Córcoles et al., 2011; Yule et al., 2007). The question ofwhich leadership style might best fit within the highly regulatedwork context of nuclear power plants still remains unanswered.Whereas leadership theories are primarily concerned with enhanc-ing the effectiveness and efficiency of employees’ work perfor-mance, it is questionable whether these promoted leadershipstyles will obtain the same positive results within a work contextwhere the trade-off between efficient and safe performance issometimes crucial. The present study aims to examine the differ-ential impact of safety leadership on safety performance in thatcontext. Using a multi-level approach, we tested the effects of anempowering team leadership style (ELQ, Arnold et al., 2000) on

Elsevier Ltd.

ínez-Córcoles), [email protected] (J.M. Peiró),

safety performance. As a result, we introduce three dimensionsfor the safety performance construct, namely, safety compliance,safety participation and risky behavior. In the next two sections,we introduce research on safety leadership and the leadershipmodel utilized in our study.

1.1. Safety leadership

Empirical findings in safety research have shown that leadershipconstitutes the strongest factor affecting organizational safety per-formance. Many of these studies focus on well-known leadershipapproaches, namely leader-member exchange (LMX) (Dansereauet al., 1975; Graen and Cashman, 1975) and transformational lead-ership (Bass, 1985, 1990), and they have been applied to severalindustrial sectors, such as manufacturing, metal processing, con-struction or food service (e.g., Barling et al., 2002; Clarke and Ward,2006; Hofmann and Morgeson, 1999; Michael et al., 2006; Mullenand Kelloway, 2009; Simard and Marchand, 1994; Zohar, 2000,2002). For instance, Hofmann et al. (2003), using a multilevel ap-proach, demonstrated with transportation members of the US Armythat high-quality leader–member exchange relationships expandedsafety citizenship role definitions when there was a positive safetyclimate, and that there was no such expansion in a less positivesafety climate. Both leader–member exchange and safety citizen-ship role definitions were positively related to safety citizenshipbehavior. Research on transformational full-range leadership

Team Leadership

Safety Participation

Safety Compliance

Riskybehaviors

GROUP LEVEL

INDIVIDUAL LEVEL

Fig. 1. Hypothesized multilevel model.

294 M. Martínez-Córcoles et al. / Safety Science 51 (2013) 293–301

models (Bass, 1990; Bass and Avolio, 1997) reveals that group-levelleadership [transformational and constructive (contingent-reward)leadership] predicted injury rates in manufacturing jobs throughthe mediation of group-level safety climate (Zohar, 2000).

In nuclear power plants, research about leadership and its rela-tionship with safety performance is very scarce. For example,Kivimaki et al. (1995) found that participative management(providing communication and feedback to subordinates) was pos-itively associated with safety performance. According to Flin andYule (2004), stimulating, individually considerate, and rewardingstyles were found to be influential in leaders’ impact on workerssafety behaviors. Katsva and Condrey (2005) suggested that themost effective leadership style in nuclear power plants (at all lead-ership levels) would be characterized by flexibility, the develop-ment and implementation of new ideas, and the encouragementof new initiatives. Martínez-Córcoles et al. (2011) conducted astudy in a nuclear power plant that assessed the impact of anempowering leadership style1 (EL) on the perceived safety behav-iors of employees. Focusing on individual leadership, they found thatEL enhances perceived safety behaviors by means of safety climate.Moreover, the authors show that empowering leaders influenceemployees’ perceived safety behaviors positively in both strongand weak safety cultures. Thus, one can tentatively conclude thatleadership styles that encourage and empower workers should havea positive effect on safety in NPPs.

The present study aims to expand the findings of Martínez-Córcoles et al. (2011) in three distinct ways. Firstly, a multi-levelapproach is used to assess the impact of EL at the group-level;i.e. the group influence of team leadership is assessed. This ap-proach agrees with recent team and leadership literature callingfor the use of multilevel methodology in team and organizationalsettings (Burke et al., 2006). Secondly, safety performance is differ-entiated into three different types of safety performance behaviors(see Fig. 1) that are relevant for work performance in NPPs. Thirdly,we test the direct effects of leadership on safety performance. Inaddition to their climate mediated effects (see Martínez-Córcoleset al., 2011), supportive leadership styles are also likely to havedirect (i.e., unmediated) effects on safety records, due to theircharacteristics (i.e., open and egalitarian communication withinand across group boundaries) (Zohar, 2003). Next, we will firstintroduce the empowering leadership model, and then state ourpredictions about the impact of EL on safety compliance, safetyparticipation and risky behavior.

1 According to previous definitions offered by Carson et al. (2007, p. 1218),Srivastava et al. (2006, p. 1240) and Vecchio et al. (2010, p. 531), we define anempowering leadership style as leaders’ behaviors that share power with subordi-nates with the aim of creating self-directedness in them. Following this definition, werefer to empowering team leadership style as team leaders’ behaviors that sharepower with the team, with the aim of creating a self-managed team.

1.2. Empowering leadership

The empowering leadership model by Arnold et al. (2000)claims that the main function of a leader is to increase a team’spotential for self-management. The empowering leadership modeldistinguishes five dimensions corresponding to different behaviorsthat empowering leaders must show. The first one is ‘‘leading byexample’’, which refers to a set of behaviors that demonstratethe leader’s commitment to his or her own work as well as tothe work of his/her members. With regard to safety, one can as-sume that leaders who practice what they preach will obtain goodsafety results, since a sense of coherence between what is said andwhat is done is created. The second dimension is ‘‘participativedecision making’’, which refers to the leader’s use of members’ in-puts in decision-making, displaying behaviors such as encouragingmembers to express their ideas and opinions. Leaders who encour-age their employees to participate in decision-making about safety(e.g., with regard to potential safety improvements) should instill asense of group belonging as well as a greater commitment tosafety. Tjosvold (1990) found that members of a flight crew per-formed more effectively in dangerous situations because teammembers were motivated by their leader to contribute to teamfunctioning with their ideas. The third dimension is ‘‘coaching’’,the ability of leaders to encourage their team members to solveproblems in a self-managed way, thereby providing members withopportunities to share and increase their knowledge. Yule et al.(2007) found that as team knowledge increases, the propensityto engage in risk-taking behaviors decreases. The fourth dimensionis ‘‘informing’’, which refers to the dissemination of information byleaders, such as the organization’s mission, philosophy or otherimportant information. One can assume that good safety resultswill be achieved when information is transmitted by leaders on aregular basis, e.g., by means of regular group meetings where infor-mation between team members and leaders is shared. Moreover,organizations with better safety records were characterized by ahigh frequency of safety discussions between managers andemployees (Cohen, 1977; Smith et al., 1978). The last dimensionis ‘‘showing concern/interacting with employees’’, which includesbehaviors such as taking time to discuss members’ concerns orshowing concern for their welfare. Leaders who provide theiremployees with emotional support acknowledge and reward teamperformance. Consequently, their behavior will achieve high levelsof safety commitment among their group members. Katsva andCondrey (2005) highlight individual treatment and feedback ascrucial to obtaining good safety outcomes in nuclear power plants.

The EL is a team leadership model that is suitable for applicationusing a multilevel approach, in order to capture the specific influ-ence of team leaders on team performance. In the present paper, ateam2 is considered as a group of people who work in a setting char-acterized by a team task, with clear boundaries, a specified authorityto manage work processes, and some degree of membership stability(Hackman, 2002). To date, there have been relatively few multileveltheories and little empirical research on team leadership (see Avolioand Bass, 1995; Pearce and Conger, 2003; Pearce and Sims, 2002).Although some authors have taken multilevel approaches into con-sideration in assessing the impact of group constructs like safetyclimate (group-level safety climate) or leadership (team leadership)on safety results, these studies have predominantly been conductedin domains where personal safety is at stakes (e.g., Hofmann et al.,2003; Neal and Griffin, 2006; Zohar, 2002, 2008). However, therehas been too little multilevel research about leadership in NPPs, eventhough scholars and practitioners recognize that team leaders play a

2 Note that although other previous research differentiates between a group, ateam, and a unit, in the present paper these terms are treated as similar.

M. Martínez-Córcoles et al. / Safety Science 51 (2013) 293–301 295

pivotal function within these settings with regard to the optimizationof safety processes (e.g., Carvalho et al., 2005; SCART Guidelines,2008).

Moreover, the EL model embraces leadership behaviors thatmight be especially relevant for nuclear power plants, since it cov-ers task-focused behaviors (i.e., informing), such as facilitating theunderstanding of task requirements and operating proceduresand compliance with them. More importantly, is also deals withperson-focused behaviors (i.e., showing concern/interacting withemployees), which facilitate behavioral interactions and influencecognitive structures and attitudes that must be developed in orderto work effectively as a team (Salas et al., 1992). Although compli-ance with rules and procedures is an important leadership out-come in establishing safety in NPPs, person-focused leadershipbehavior has the potential to enhance employees’ safety perfor-mance by going beyond complying with safety standards, (e.g.,by reporting near-misses or minor events). With a clear orientationtoward person-focused behaviors, empowering leadership has pro-ven to be the most functional leadership in team performance(Burke et al., 2006). In the following paragraphs, we specify ourassumptions about the differential impact of team leadership onsafety performance.

1.3. Safety performance

Research on work performance has shown a strong consensusabout two constructs simultaneously, task performance and con-textual performance (Borman and Motowidlo, 1993; Motowidloand Van Scotter, 1994). Task performance is defined as work activ-ities that contribute to an organization’s primary task and areprescribed by formal job descriptions. Contextual performance isdefined as an activity that contributes to the social and psycholog-ical core of the organization; it is voluntary in nature and consid-ered as an informal activity (Borman and Motowidlo, 1997).Although synonymous or similar concepts can be found through-out the literature on these two concepts, especially in the case ofcontextual performance [e.g., organizational citizenship behavior(Smith et al., 1983), extra-role behaviors (Katz and Kahn, 1966),civic virtue (Organ, 1988) and pro-social organizational behavior(Brief and Motowidlo, 1986)], this duality continues to be the mostwidely-used explanation in organizational psychology for the workperformance construct. However, a review on job performancecarried out by Rotundo and Sackett (2002) suggests that thereare three distinct groups of behaviors that constitute the whole do-main of job performance. These behaviors include task, citizenshipand counterproductive performance. According to Robinson andBennett’s (1995) definition, counterproductive behaviors are thosevoluntary behaviors that harm the well being of the organization.

Safety research also started to apply the aforementioned dualityto the safety performance construct (e.g., Griffin and Neal, 2000;Neal and Griffin, 2004, 2006). Thus, safety performance has usuallybeen conceptualized as two types of safety behaviors: Safety com-pliance and safety participation.3 Safety compliance refers to thework activities that individuals need to carry out in order to estab-lish workplace safety. These behaviors include adhering to standardwork procedures and wearing personal protective equipment. Safetyparticipation describes behaviors that do not directly contribute toan individual‘s personal safety, but that help to develop a work envi-ronment that supports process safety. It includes activities such asparticipating in voluntary safety activities, helping coworkers with

3 Also, for safety compliance and safety participation synonymous or very similarconcepts have emerged. For safety compliance: Rule-related safety (Schöbel, 2005). Forsafety participation: Safety proactivity (Parker et al., 2003) or proactive safetybehaviors (Geller et al., 1996); Safety citizenship (Hofmann et al., 2003); Commit-ment-based safety (LePine et al., 2002), etc.

safety-related issues or attending safety meetings (Neal and Griffin,2006). Although recent safety research does not focus on empower-ing leadership per se, it can be shown that efficient leadership inhigh-risk settings is associated with increasing safety complianceand safety participation. For example, Christian et al. (2009) inte-grates safety literature related to antecedents of safety performancebehaviors (distinguishing between safety compliance and safety par-ticipation) and safety outcomes. In their meta-analysis, leadership(conceptualized as LMX and transformational leadership models) isan antecedent of both kinds of behavior, with the influence beingeven stronger on safety participation than on safety compliance(Mp = .24 for compliance, Mp = .35 for participation).

Following the threefold job performance structure found byRotundo and Sackett (2002), we think that research on safety per-formance should also integrate those behaviors (i.e. risky behav-iors) which have the potential to cause adverse consequences,thereby completing the safety performance construct, along withsafety compliance and participation. However, we contrast riskybehavior with counterproductive behavior in the Rotundo andSacket model (2002), since risky behavior does not necessarily pro-duce detrimental effects on productivity. Based on Ramanujam andGoodman‘s concept of latent errors (2003), we understand riskybehavior to be a deviation from standard organizational practices,procedures and expectations that does not always produce directand immediate adverse consequences and can lead to efficient(but not necessarily safe) outcomes. Therefore, the safety perfor-mance construct in our study is composed of safety compliance,safety participation and risky behaviors.

Safety compliance involves adherence to rules and proceduresdeveloped by the company and/or regulatory bodies. Researchfindings on safety compliance have shown that a leader’s empow-ering behavior predicts workers’ compliance with rules and proce-dures. Simard and Marchand (1997) demonstrated that aworkgroup’s safety compliance is positively influenced by a coop-erative workgroup–supervisor relationship and a participativeleadership style of the supervisor. Similarly, Hofmann and Morge-son (1999) found that employees are highly committed to follow-ing safety procedures when they feel comfortable discussingsafety-related issues with their supervisors. Assuming EL as aneffective leadership style that achieves team members’ compliancewith safety requirements and procedures, we hypothesize thatteam EL will predict safety compliance.

H1. Empowering team leadership will positively predict safetycompliance.

Even when the performance of a NPP is good and has been goodfor several years, there is a strong need to support, challenge andmaintain employees’ and managers’ alertness (Wahlström, 2005).Continual vigilance is an important pre-condition for identifyingand correcting latent failures or hidden issues in technological sys-tems that have the potential to incubate and ultimately producesystem accidents (e.g., Reason, 1990). One potential way to keepan organization alert is to promote learning processes by meansof meetings, exchanging viewpoints, discussions or self-practicalexperimentation (Richter, 2003). The goal is to confront organiza-tional members with different interpretations of safety issuesand make them aware of the normal, the necessary and the unex-pected. One important pre-condition of the success of such inter-ventions is that members actively participate and contribute.Therefore, safety participation should complement safety compli-ance, since procedures cannot cover all possible risks (Zohar,2008). According to Naevestad (2008), it is crucial for leaders toencourage their workers to participate in ongoing safety discus-sions and campaigns, in order to create frames of reference thatallow for cultural redundancy, i.e. the proactive integration of

296 M. Martínez-Córcoles et al. / Safety Science 51 (2013) 293–301

different views and perspectives on safety as a key value of anorganizational culture. For this purpose, it is essential that leadersprovide and foster safety participation, which results in continuouslearning processes. Due to the potential of EL to strengthen work-ers’ initiatives and self-management, it is hypothesized that EL hasthe potential to enhance safety participation.

H2. Empowering team leadership will positively predict safetyparticipation.

Most workplace accidents and incidents can be attributed tounsafe or risky behavior (Hollnagel, 1993; HSE, 2002). Throughoutthe safety literature on leadership, characteristics of the EL model(i.e. leaders’ commitment with safety, encouraging members toexpress their ideas and opinions, showing concern and interactwith employees rewarding their performance, and informing aboutsafety issues) have been directly connected with the reduction ofunsafe acts. For instance, Flin and Yule (2004) state that the mosteffective direct leader for safety had a more supportive style ofleadership, initiated discussions about safety and provided positivefeedback regarding safety issues. Regarding the nuclear industry,Yule et al. (2007) found in six power stations that both managers’commitment and supervisors‘ involvement reduced risk-takingbehavior by means of safety knowledge and training. The authorsconclude that workers are more likely to follow safety procedureswhen faced with the option of taking a risk if they understand therisks of not following the procedures. As aforementioned, empow-ering leadership provides members with opportunities to shareand increase their safety knowledge (i.e. due to coaching).

However, the link between the EL model and the reduction ofrisky behavior in nuclear power plants is not straightforward.One might argue that providing autonomy in such a highly stan-dardized setting will necessarily lead to risk-taking behavior or,in other words, may have opposite effects on safety performance.Contrary, we argue that these systems substitute leadership func-tions not only in the way that they prescribe special types ofbehavior (by procedures), but also that they align the goals of orga-nizational members to safety and reliability. It can be assumed thatfostering safety knowledge exchange and providing autonomy byempowering leadership motivates employees to achieve thesesafety goals. And this should become relevant in situations whereemployees have to evaluate the riskiness of different behavioraloptions (e.g., in situations where procedures do not match or arein conflict with each other). Therefore, we assume that empower-ing leadership style will have a potential effect on risky behavior:

H3. Empowering team leadership will negatively predict riskybehavior.

2. Method

2.1. Participants and procedure

Our original sample was composed of 495 workers from twonuclear power plants. Our analysis was based on 479 respondentsin 54 groups, as missing data was deleted (13 respondents). More-over, three groups were not considered because they consisted ofonly one person (three respondents). In our sample, the percentageof missing data was quite small (3%); thus it is unlikely to be a seri-ous problem (e.g., Graham and Hofer, 2000). The average team sizewas nine members, with a standard deviation of 4.6. Maximumworkers per team were 23, and minimum three. All responsibilitylevels and functional areas in the nuclear facilities were included.Data were collected in March 2011. The total size for both organi-zations was 760 employees. Thus, we obtained an overall responserate of 65.1%.

Within our sample, 3% of the respondents are younger than30 years old, 18% are between 30 and 45 years of age, and 79%are older than 45 years of age. In addition, 47.3% of the respondentshold a university degree.

The questionnaire was administered in the workplace as part ofa broader battery of questionnaires designed to evaluate safetyculture. The entire battery took about 30 min to be filled out. Theparticipation was voluntary and occurred during work time. Ano-nymity and confidentiality were guaranteed. Several researcherswent to the workplace and stayed there for about 3 days to collectdata in each facility. In all sessions, the researchers explained theobjective of the research. While the participants were answeringthe questionnaire, the researchers were available to solve anyproblems or answer any questions they might have.

2.2. Measures

2.2.1. Empowering leadershipAn adaptation of the ‘‘empowering leadership questionnaire

(ELQ)’’ (Arnold et al., 2000) was measured. A total of 17 items madeup the scale, after omitting some items from the original scale, as itwas too long (38 items) to adapt to time constraints. This scale hadbeen applied in a previous study (Martínez-Córcoles et al., 2011). A5-point Likert response scale ranging from 1 (never) to 5 (always)was used. Sample items assessing leadership include: My immedi-ate boss ‘‘sets high standards for safety performance by his/herown behavior’’, ‘‘encourages work group members to expressideas/suggestions’’ or ‘‘pays attention to my work group’s efforts’’.

2.2.2. Safety complianceIn order to measure compliance with safety rules and proce-

dures, the original scale by Neal and Griffin (2006) was used. Threeitems make up the scale, with a 5-point Likert response scale rang-ing from 1 (Completely disagree) to 5 (Completely agree). Scale itemswere: ‘‘I use all the necessary safety equipment to do my job’’,‘‘I use the correct safety procedures for carrying out my job’’, and‘‘I ensure the highest levels of safety when I carry out my job’’.

2.2.3. Safety participationAs in the case of safety compliance, we used the original safety

participation scale by Neal and Griffin (2006). The scale consistedof three items, with a 5-point Likert response scale ranging from1 (Completely disagree) to 5 (Completely agree). Scale items were:‘‘I promote the safety program within the organization’’, ‘‘I makeextra effort to improve safety in the workplace’’, and ‘‘I voluntarilycarry out tasks or activities that help to improve workplace safety’’.

2.2.4. Risky behaviorRisky behavior was recorded on a scale based on Mearns et al.

(2001). The original scale had 12 items, but two items were deletedbecause they were not appropriate for the nuclear sector. Thismodified scale had already been used in a previous study (Martí-nez-Córcoles et al., 2011). Therefore, our scale was composed of10 items. A 5-point Likert scale ranging from 1 (never) to 5(usually) was used, so that higher scores reflected risky behavior.Sample items include: ‘‘I ignore safety regulations to get the jobdone’’, ‘‘In order to be more effective in my job, I have to breakwork procedures’’, or ‘‘I take shortcuts that involve little or no risk’’.

2.3. Analyses

The first step in the data analysis was to test the factorial struc-ture of the scales through the present sample, in order to obtainevidence of the validity of the four scales presented above. For thispurpose, we performed confirmatory factor analyses (CFAs) usingLISREL 8.8 (Jöreskog and Sörbom, 2006). Considering the ordinal

M. Martínez-Córcoles et al. / Safety Science 51 (2013) 293–301 297

nature of the items, weighted least square (WLS) was used toestimate model parameters, and both the polychoric correlationsmatrix and the asymptotic covariances matrix were used as inputfor the analyses.

First of all, we conducted two confirmatory factor analyses,making use of the empowering leadership scale [adapted fromArnold et al. (2000)], in order to discover the best factorial structurein our sample. Arnold et al. (2000) confirmed a 5-dimensionalmodel in their empowering leadership questionnaire (leading byexample, participative decision making, coaching, informing, andshowing concern/interacting with the employees). In their analyses,these authors found that some leadership factors had high correla-tions with each other. Running confirmatory factor analyses (usinggeneralized least squares), they compared several factor models(eight, seven, six, five, four, one, and zero factors), with subsequentcross-validation results, and they found support for the five-factorstructure as the best solution. In contrast, a one-dimensional modelwas chosen as the best factor solution in recent studies (seeMartínez-Córcoles et al., 2011). Thus, we decided to compare thefive-factor structure with the one-dimensional structure.

Second, we also wanted to determine the factorial structure ofsafety performance for two reasons. First, although most of theprevious literature identified safety compliance and safety partici-pation as the two predominant factors of safety performance (seeChristian et al., 2009), we have taken risky behaviors into consider-ation as well. Since we wanted to make sure that risky behaviorsand safety compliance did not overlap in terms of correlations,we decided to test the factorial structure of safety performance,comparing three alternative models by means of confirmatoryanalyses: (1) a one-dimensional structure, (2) a two-factor struc-ture considering safety compliance and risky behavior as the firstfactor and safety participation as the second factor, and (3) athree-factor structure including safety compliance, safety partici-pation, and risky behaviors as three independent factors.

In order to assess the fit of these models, we examinedRMSEA (root mean square error of approximation), CFI (compara-tive fit index), NNFI (non-normed fit index), and AGFI (adjustedgoodness of fit index) goodness of fit statistics. The interpretationof these indexes is as follows: RMSEA < .08 = acceptable model;CFI > .90 = acceptable model, and >.95 = excellent model; NNFI >.90 = acceptable model, and >.95 = excellent model; AGFI > .90 =acceptable model. In order to test differences between the modelsand decide which one presents a better fit, several modeling ratio-nale criteria were considered. Thus, for example, differences notlarger than .01 between NNFI and CFI values (DNNFI and DCFI)are considered an indication of negligible practical differences(Cheung and Rensvold, 2002; Widaman, 1985). Chen (2007) sug-gests that when the RMSEA increases by less than .015, one canalso claim support for the more constrained (parsimonious) model.

Finally, with the purpose of providing support for our threehypotheses, a multi-level analysis with LISREL 8.8 (PRELIS) (Jöre-skog and Sörbom, 2006) was performed, in which leadership wasaggregated at the group-level, based on the results of the AverageDeviance Indexes (ADM(J)). Safety compliance, safety participationand risky behaviors were taken into account as individual-levelvariables. A sequence of hierarchical linear models was run: (1) a

Table 1Descriptive statistics, Cronbach’s a, and intercorrelations between study variables.

Factor M SD Cronbach’s

1. Leadership 3.52 .97 .982. Risky behaviors 1.47 .53 .913. Safety compliance 4.56 .60 .884. Safety participation 4.10 .75 .86

** p < .01.

one-way ANOVA model to determine the variability of safety per-formance variables among teams and within teams (inter-teamand intra-team variance), calculating at the same time intraclasscorrelation coefficients in safety performance indicators, and (2)an intercepts-as-outcomes model to test the cross-level effectsbetween leadership and safety performance indicators.

3. Results

Descriptive statistics and alpha coefficients (Cronbach, 1951)for all measures are presented in Table 1. In general, the partici-pants reported high scores for leadership (M = 3.52, SD = .97),safety compliance (M = 4.56, SD = .60) and safety participation(M = 4.10, SD = .75), and low scores for risky behaviors (M = 1.47,SD = .53). Internal consistencies were excellent (Cronbach’s a coef-ficients from .86 to .98). Pearson correlations revealed positiverelations between leadership, safety compliance and safety partic-ipation (r = .32, r = .37, r = .56, p < .01 respectively), and negativerelations between each of these three variables and risky behaviors(r = �.26, r = �.47, r = �.39, p < .01, respectively).

3.1. Confirmatory factor analyses

Regarding the leadership construct, we evaluated both factorstructures (the one-dimensional structure and the five-factorstructure) in our sample. Theoretically, the five-factor structureshould be the solution with the best fit, as we tried to maintainit throughout our research (the items selected corresponded to fivefactors). Actually, the five-factor model provided an excellent fit(v2 = 175.380, df = 109, p < .01; RMSEA = .036; CFI = .999;NNFI = .998; AGFI = .996), but so did the one-dimensional model(v2 = 249.300, df = 119, p < .01; RMSEA = .049; CFI = .998;NNFI = .997; AGFI = .995). As the incremental fit indices’ differ-ences indicated negligible practical differences between the twotested models, we chose the more parsimonious model, namely,the one-dimensional model (internal consistency reliability forthe scale was .97). Incidentally, the five-factor model presentedhigh correlations among the leadership factors in our sample (seeTable 2), another indication that the one-dimensional model wasthe best one.

With regard to safety performance, we tested three factor struc-tures in order to determine which was best for our study. The threemodels presented a satisfactory fit, although the three-factor mod-el was slightly better with regard to goodness of fit statistics. Whencomparing the one-dimensional (v2 = 382.559, df = 104, p < .01;RMSEA = .077; CFI = .975; NNFI = .971; AGFI = .973) and three-dimensional models (v2 = 205.928, df = 101, p < .01; RMSEA = .048;CFI = .990; NNFI = .989; AGFI = .985), differences in practical fit ap-peared for all practical goodness of fit statistics. In the comparisonbetween the two-factor (v2 = 295.966, df = 103, p < .01;RMSEA = .064; CFI = .983; NNFI = .980; AGFI = .979) and three-fac-tor models, the DRMSEA also showed a non-negligible differencein practical fit. Therefore, we considered the three-factor modelto be the best safety performance factorial structure. Results ofour confirmatory analyses are shown in Table 3.

a 1 2 3 4

–�.26** –

.32** �.47** –

.37** �.39** .56** –

Table 2Descriptive statistics and intercorrelations between leadership factors.

Factor M SD 1 2 3 4 5

1. Leading by example 3.64 1.03 –2. Participative decision-making 3.73 1.02 .82** –3. Coaching 3.45 1.00 .84** .86** –4. Informing 3.41 1.05 .73** .77** .82** –5. Showing concern 3.36 1.17 .77** .84** .86** .81** –

** p < .01.

Table 3Goodness of fit statistics of five confirmatory factor analyses performed.

Model v2 df p RMSEA CFI NNFI AGFI

Leadership (1 factor) 249.300 119 <.01 .049 .998 .997 .995Leadership (5 factors) 175.380 109 <.01 .036 .999 .998 .996Safety performance (1 factor) 382.559 104 <.01 .077 .975 .971 .973Safety performance (2 factors) 295.966 103 <.01 .064 .983 .980 .979Safety performance (3 factors) 205.928 101 <.01 .048 .990 .989 .985

Note: Leadership (5 factors) = leading by example, participative decision making, coaching, informing, and showing concern/interacting with the employees.Safety performance (2 factors) = (safety compliance + risky behavior) and safety participation.Safety performance (3 factors) = safety compliance, safety participation and risky behavior.

298 M. Martínez-Córcoles et al. / Safety Science 51 (2013) 293–301

3.2. Multilevel analyses

Before we tested our hypotheses with a multi-level analysisusing LISREL 8.8 (PRELIS), we tested within-group agreement onleadership from a consensus-based approach with the averagedeviation index (ADM(J)) by means of SPSS 17.0. ADM(J) was pro-posed by Burke et al. (1999). This index first calculates the averagedeviation for each scale item, and then the mean values of eachaverage deviation. According to Gonzalez-Romá et al. (2002), thereare some advantages of ADM(J) compared to the interrater agree-ment index (rwg) (James et al., 1984). First, ADM(J) does not needto model the random or null response distribution; it only requiresa priori specification of a null response range of interrater agree-ment. Second, the ADM(J) index provides estimates in the metricof the original response scale. Following Dunlap et al. (2003), apractical cut-off value for a Likert-type response scale with five op-tions was .83 or less because ADM(J) is a dispersion index. Thus, anexamination of ADIs showed them to be satisfactory for leadership(ADM(J) = .71, SD = .20), suggesting that shared perceptions of lead-ership were apparent among the team members. Therefore, weaggregated the leadership variable using the average, as in thecomposition model (Kozlowski and Klein, 2000).

Hierarchical linear modeling (HLM) was conducted to test howsafety performance indicators (safety compliance, safety participa-tion and risky behaviors) were influenced by group perceptions ofleadership (team-level factor). Three separate hierarchical linearmodels were estimated (one for each of the three indicators ofsafety performance), using the maximum likelihood method. Theone-way ANOVA model determined significant intra-variabilitywithin groups, but non-significant differences between groups (in-ter-variability). The intraclass correlation coefficient (ICC) was cal-culated in these three dimensions (ICC for risky behaviors was .03,for safety compliance .06, and for safety participation .05). An addi-tional analysis of variance with SPSS was run for each of the safetyperformance indicators, in order to examine whether there wasany variance among the teams on the indicators of safety perfor-mance. The results indicated non-significant differences betweengroups for risky behaviors (F(53,410) = 1.25; p > .05), but significantdifferences between groups for safety compliance (F(53,417) = 1.69;p < .01) and safety participation (F(53,410) = 1.53; p < .05). Taken to-gether, the results suggest that 6% and 5% of the variance in safetycompliance and safety participation, respectively, can be attributed

to differences between groups, making it appropriate to examinewhether a group level variable (in this case, the team level percep-tion of empowering leadership) accounts for the group level vari-ance found in these safety performance indicators. Although littlevariance can be attributed to differences between groups for riskybehavior, the intercepts-as-outcomes model was run to test thecross-level effects for the three safety performance indicators.

As in the hypothesized models there were no individual-levelpredictors, the individual-level models were only composed ofparameters b0j, which represent the average indicator of the safetyperformance factor for team j, and eij, which symbolizes the ran-dom error. Leadership was included in the team-level models(intercepts-as outcomes model) to determine whether variationsin each safety performance indicator could be explained by groupleaders’ behaviors. See models below.

Individual level model

Safety performance indicators ¼ b0j þ eij

Team level model

b0j ¼ c00 þ c01ðLeadershipÞj þ u0j

In the team-level model, the average indicator of a team’s safetyperformance, represented by b0j, was modeled as a function of theteam’s perception of leadership. In the model, c00 is the interceptfor the team level model, c01 represents the change in the indicatorof safety performance due to teams’ perception of leadership, andu0j symbolizes the random effects.

The results of the multilevel models are displayed in Table 4.Examination of team level effects revealed that safety complianceand safety participation were positively predicted by team-levelleadership. In contrast, as we expected, risky behaviors were neg-atively predicted by team leadership. Therefore, support was ob-tained for our three hypotheses.

4. Discussion and conclusions

The present study was designed to test the effect of team lead-ership on safety performance. Our goal was to test a model to ex-plain the differential impact of team empowering leadership onthree facets of safety performance (safety compliance, safety par-ticipation and risky behaviors) in nuclear power plants. Therefore,

Table 4Summary of the three hierarchical linear models (one for each safety performanceindicator).

Fixed effect Parameter Coefficients SE z-value

Risky behaviorsIntercept b0j 1.85 .18 10.47**

Leadership c01 �.11 .05 �2.15*

Safety complianceIntercept b0j 3.65 .20 18.57**

Leadership c01 .26 .05 4.69**

Safety participationIntercept b0j 3.14 .25 12.31**

Leadership c01 .27 .07 3.74**

Note: Models tested the cross-level effects between leadership and each safetyperformance indicator.** p < .01.* p < .05.

4 Critically questioning rules and procedures is seen as very important behavior innuclear industries, since nearly 70% of all human performance problems could betraced to bad procedures (Reason, 2008). That is, procedures which were inappro-priate and unworkable in the current situation, that gave the wrong information, orwere not found, or could not be understood, or that had not been written to cover thisparticular task (Reason, 2008). Moreover, safety of nuclear power plants is an additiveattribute of these systems, since safety solutions primarily rest on more rules andprocedures (Amalberti, 2001). Additive safety implies that probability of conflictingprocedures, inconsistent procedures or ‘‘not matching’’ procedures (due to newtechnologies) is very high even under normal operation.

M. Martínez-Córcoles et al. / Safety Science 51 (2013) 293–301 299

we expand the findings of a recent study by Martínez-Córcoleset al. (2011), in which it was shown that EL positively influencedsafety behaviors by means of safety climate. Apart from this study,no other study has investigated the influence of empowering lead-ership on safety performance within the nuclear field. Two direc-tions have been adopted in the present study. The first was tomore fully explore the direct relationships between EL and safetyperformance from a multilevel approach (team EL). The secondwas to conceptualize safety performance as three different typesof behaviors, with the novelty of adding risk-taking behavior tothe predominant duality of safety performance (safety compli-ance-safety participation) in safety research. The results of HLMshow clearly that leadership significantly predicts the three safetyperformance behaviors.

In light of the findings from our study, we can conclude thatteam EL is a potential antecedent of safety performance. Specifi-cally, when leaders behave like empowering leaders, they producecompliance with safety procedures and requirements (Hypothesis1), enhance safety participation behaviors (Hypothesis 2), and theyare able to reduce team members’ risky behaviors (Hypothesis 3).Thus, our study reveals that EL directly influences subordinates’safety performance behaviors. Our results are in line with Zohar’s(2003) conclusion that behavioral leadership can directly influenceemployees’ behaviors (without mediators) because of the effects ofsocial learning processes (Bandura, 1977). Regarding social learn-ing theory, team leaders’ behaviors (i.e. leading by example) areobserved by team members, who imitate these behaviors becausethey are rewarded by team leader or team colleagues (i.e. due torecognition of these behaviors). EL is a purely behavioral leader-ship style, and in our study we recorded the team members’ re-sponses regarding the direct interaction with their leader. Thus,our study is a pioneer in finding direct relationships between lead-ership and safety performance, shedding empirical light about anew way in which leaderships proxies to transformational or par-ticipative styles (e.g. empowering leadership) impact on safetyperformance, beyond the already known mediated effect of safetyclimate mentioned in previous empirical studies (e.g. Clarke andWard, 2006; Kelloway et al., 2006; Martínez-Córcoles et al.,2011; Zohar, 2002).

We believe that the results of our research could have impor-tant implications for safety research in two ways. First of all, a mul-tilevel approach was used, which allows us to detect the groupeffects and the influences (social interaction, team membership,etc.) to which individuals are subjected. Therefore, using hierarchi-cal linear modeling, we estimated three cross-level relationships,assuming dependent observations due to the hierarchical datastructure. That is, individual shared perceptions within units abouttheir team leader were considered. Second, measuring safety

performance in HROs implies not only focusing on safety compli-ance and safety participation (behaviors associated with positivesafety performance), but also considering behaviors such as devi-ance from rules and procedures that involve risks and could pro-duce fatal consequences. In that case, we found that empoweringleaders’ behaviors were related to a decrease in employees‘ riskybehaviors.

The latter result needs some more clarification, especially, whenit comes to the practical implications of our study. As stated above,we argue that highly standardized systems (in terms of proceduresand safety rules) like nuclear power plants substitute leadershipfunctions by aligning the goals of organizational members to safetyand reliability. Promoting an empowering leadership style in nucle-ar power plants implies that team members have more autonomy infulfilling their work tasks, which might foster their risk-takingbehavior. However, we think that the alignment of safety goalsdue to procedural systems paired with an empowering leadershipstyles counteracts this argument. Empowering leaders motivatetheir subordinates to critically question rules and procedures,4

which will be a much more harder endeavor for (e.g.) authoritarianleaders who will mainly focus on rule compliance without subordi-nate‘s critical questioning of the match between the rule and the sit-uation. Thus, by fostering safety knowledge exchange (i.e. more safetyparticipation) empowering leaders support their subordinates to bet-ter evaluate rule compliance situations, to make safety-optimizingdecisions and to cope with unknown (or not anticipated) situations,since they have internalized the goal of to be developed behavior(i.e. safety) and therefore have the knowledge to cope with these sit-uations in a safety-directed manner. Empowering leadership providesmeaning to workers’ rule compliance in responding to complex prob-lems. When people perceive that complying with rules makes sense,they get a better understanding of potential hazards and are moti-vated to develop their adaptive capacity further. The results of ourstudy clearly suggest that this might be the case (e.g. less risk-takingbehavior, more safety participation and more safety compliance).

Several limitations exist in the present study. First, we use self-reports to measure safety performance. This may lead to inflatedresults due to respondents’ tendencies to respond in a consistentand socially desirable manner. Thus, we measured perceived safetybehaviors and, therefore, do not know the real occurrence and out-come of these behaviors. Second, the cross-sectional character ofthe study reduces the nature of the variables to a ‘‘snapshot’’, in-stead of being dynamic processes over time. For instance, howleadership develops its impact on each type of safety performanceover time is missing, and causal relations cannot be tested. Third,we have not taken mediator variables into account. Most previousempirical research (although not particularly in HROs) agrees thatsafety climate is the mediator variable through which leadershipinfluences safety outcomes (e.g., Zohar, 2000). However, one rea-son why we explored the direct relationships between leadershipand safety performance is because of the recent appearance ofthe EL model (to date there is no empirical multilevel researchabout EL for safety). Therefore, first testing the direct cross-levelrelationships between team EL and safety performance types hasbeen our main goal in this study. Another one was to test if abehavioral leadership style could have direct impact on employees’

300 M. Martínez-Córcoles et al. / Safety Science 51 (2013) 293–301

safety performance as previously has been suggested (e.g., Zohar,2003). Finally, the reader should be careful regarding the general-izability of the results of the present study. According to a cross-cultural comparison of work in reactor control rooms, Rochlinand von Meier (1994) found several functional variations, whichcan be lastly attributed to the impact of national cultures. At least,four different cultural archetypes varying on the dimensions ‘‘hier-archical power distance’’ and ‘‘rule orientation’’ were identified(Latin, Germanic, Scandinavian-British, and Asia countries). There-fore, our results concerning the outcomes of empowering leader-ship should be seen in the light of this impact. They could bealtered depending on the cultural milieu embedded within thenational culture. Future studies, which generalize data susceptibleto be influenced by culture, should take this limitation intoconsideration.

The limitations detected also indicate possible directions for fu-ture research. Empirical studies should explore potential mediatorsand/or moderators (i.e. group-level safety climate) in the teamleadership-safety performance relationship from a multilevel ap-proach. Moreover, future research should consider a longitudinaldesign in order to capture the dynamic quality of the team leader-ship construct over time. At the same time, other methods shouldbe used in addition to self-report instruments, such as observationsor interviews, to control tendencies of consistent responses or so-cial desirability. The authors hope that the present study shedsnew light on leadership and safety performance in HROs, and pro-vides new orientations for practitioners.

Acknowledgements

The authors wish to thank CIEMAT and the Department of Psy-chology and Ergonomics at Technische Universität Berlin for thesupport rendered.

This investigation was supported by Research Grant CONSOLID-ER-C (SEJ2006-14086/PSIC) and FEDER from the Spanish Ministryof Education and Science.

References

Amalberti, R., 2001. The paradoxes of almost totally safe transportation systems.Safety Science 37, 109–126.

Arnold, J.A., Arad, S., Rhoades, J.A., Drasgow, F., 2000. The empowering leadershipquestionnaire: the construction and validation of a new scale for measuringleader behaviors. Journal of Organizational Behavior 21, 249–269.

Avolio, B.J., Bass, B.M., 1995. Individual consideration viewed at multiple levels ofanalysis: a multi-level framework for examining the diffusion oftransformational leadership. Leadership Quarterly 62, 199–218.

Bandura, A., 1977. Social Learning Theory. General Learning Press.Barling, J., Loughlin, C., Kelloway, E.K., 2002. Development and test of a model

linking safety-specific transformational leadership and occupational safety.Journal of Applied Psychology 87, 488–496.

Bass, B.M., 1985. Leadership and Performance Beyond Expectations. The Free Press,New York.

Bass, B.M., 1990. Bass & Stogdill’s Handbook of Leadership, 3rd ed. The Free Press,New York.

Bass, B.M., Avolio, B.J., 1997. Full Range Leadership Development: Manual for theMultifactor Leadership Questionnaire. Mindgarden, Palo Alto, CA.

Borman, W.C., Motowidlo, S.J., 1993. Expanding the criterion domain to includeelements of contextual performance. In: Schmidt, N., Borman, W.C., Howard, A.,Kraut, A., Ilgen, D., Schneider, B., Zedeck, S. (Eds.), Personnel Selection inOrganizations. Jossey-Bass, San Francisco, pp. 1–98.

Borman, W.C., Motowidlo, S.J., 1997. Introduction: organizational citizenshipbehavior and contextual performance. Special issue of human performance.In: Borman, W.C., Motowidlo, S.J., (Eds.), Human Performance 10, pp. 67–69.

Brief, A.P., Motowidlo, S.J., 1986. Prosocial organizational behaviors. Academy ofManagement Review 11, 710–725.

Burke, M.J., Finkelstein, L.M., Dusig, M.S., 1999. On average deviation indices forestimating interrater agreement. Organizational Research Methods 2, 49–68.

Burke, C.S., Stagl, K.C., Klein, C., Goodwin, G.F., Salas, E., Halpin, S.M., 2006. Whattype of leadership behaviors are functional in teams? A meta-analysis. TheLeadership Quarterly 17, 288–307.

Carson, J.B., Tesluk, P.E., Marrone, J.A., 2007. Shared leadership in teams: aninvestigation of antecedent conditions and performance. Academy ofManagement Journal 50, 1217–1234.

Carvalho, P.V.R., dos Santos, I.L., Vidal, M.C.R., 2005. Nuclear power plant shiftsupervisor’s decision making during microincidents. International Journal ofindustrial Ergonomics 35, 619–644.

Chen, F.F., 2007. Sensitivity of goodness of fit indexes to lack of measurementinvariance. Structural Equation Modelling 14, 464–504.

Cheung, G.W., Rensvold, R.B., 2002. Evaluating goodness-of-fit indexes for testingmeasurement invariance. Structural Equation Modelling 9, 233–255.

Christian, M.S., Bradley, J.C., Wallace, J.C., Burke, M.J., 2009. Workplace safety: ameta-analysis of the roles of person and situation factors. Journal of AppliedPsychology 94 (5), 1103–1127.

Clarke, S.G., Ward, W., 2006. The role of leader influence tactics and safety climatein engaging employees’ safety participation. Risk Analysis 26 (5), 1175–1185.

Cohen, A., 1977. Factors in successful occupational safety programs. Journal ofSafety Research 9, 168–178.

Cronbach, L.J., 1951. Coefficient alpha and the internal structure of tests.Psychometrika 16, 297–334.

Dansereau, F., Graen, G.B., Haga, W.J., 1975. A vertical dyad linkage approach toleadership within formal organizations. Organizational Behavior and HumanPerformance 13, 46–78.

Dunlap, W.P., Burke, M.J., Smith-Crowe, K., 2003. Accurate tests of statisticalsignificance for rwg and average deviation interrater agreement indices. Journalof Applied Psychology 88, 356–362.

Fahlbruch, B., 2005. Safety culture training for nuclear management and supervisors.In: Itoigawa, N., Wilpert, B., Fahlbruch, B. (Eds.), Emerging Demands for theSafety of Nuclear Power Operations. CRC Press LLC, Florida, pp. 109–117.

Flin, R., Yule, S., 2004. Leadership for safety: industrial experience. Quality andSafety in Health Care 13 (Suppl. II), 45–51.

Geller, E.S., Roberts, D.S., Gilmore, M.R., 1996. Predicting propensity to actively carefor occupational safety. Journal of Safety Research 27, 1–9.

Gonzalez-Romá, V., Peiró, J.M., Tordera, N., 2002. An examination of the antecedentsand moderator influences of climate strength. Journal of Applied Psychology 87,465–473.

Graen, G.B., Cashman, J.F., 1975. A role making model in formal organizations: adevelopmental approach. In: Hunt, J.G., Larson, L.L. (Eds.), Leadership Frontiers.Kent State University Press, Kent, OH, pp. 143–165.

Graham, J.E., Hofer, S.M., 2000. Multiple imputation in multivariate research. In:Little, T.D., Schnable, K.U., Baumert, J. (Eds.), Modeling Longitudinal andMultilevel Data: Practical Issues, Applied Approaches, and Specific Examples.Erlbaum, Mahwah, NJ, pp. 201–218.

Griffin, M.A., Neal, A., 2000. Perceptions of safety at work: a framework for linkingsafety climate to safety performance, knowledge, and motivation. Journal ofOccupational Health Psychology 5, 347–358.

Hackman, J.R., 2002. Leading Teams: Setting the Stage for Great Performances. HBSPress, Boston.

Health and Safety Executive, 2002. Strategies to Promote Safe Behaviors as Part of aHealth and Safety Management System. HSE Books, Suffol, England.

Hofmann, D.A., Morgeson, F., 1999. Safety-related behavior as a social exchange: therole of perceived organizational support and leader-member exchange. Journalof Applied Psychology 84, 286–296.

Hofmann, D.A., Morgeson, F., Gerras, S., 2003. Climate as a moderator of therelationship between leader–member exchange and content specificcitizenship: safety climate as an exemplar. Journal of Applied Psychology 88(1), 170–178.

Hollnagel, E., 1993. Human Reliability Analysis: Context and Control. HarcourtBrace, London.

International Nuclear Safety Advisory Group, Safety Culture, INSAG-15, IAEA,Vienna, Austria, 2002.

International Atomic Energy Agency, 2007. SCART Guidelines: Reference report forIAEA Safety Culture Assessment Review Team. IAEA Service Series No. IAEA,Vienna.

International Atomic Energy Agency, 2008. SCART Guidelines: Reference report forIAEA Safety Culture Assessment Review Team. IAEA Service Series No. 16. IAEA,Vienna.

James, R.L., Demaree, R.G., Wolf, G., 1984. Estimating within-group interraterreliability with and without response bias. Journal of Applied Psychology 69,85–98.

Jöreskog, K.G., Sörbom, D., 2006. LISREL 8.80 for Windows [Computer Software].Scientific Software International, Inc., Lincolnwood, IL.

Katsva, M., Condrey, S.E., 2005. Motivating Personnel at Russian nuclear powerplants: a case-study of motivation theory application. Public PersonnelManagement 34 (4), 343–356.

Katz, D., Kahn, R.L., 1966. The Social Psychology of Organizations. Wiley, New York.Kelloway, E.K., Mullen, J., Francis, L., 2006. Divergent effects of transformational and

passive leadership on employee safety. Journal of Occupational HealthPsychology 11 (1), 76–86.

Kivimaki, M., Kalimo, R., Salminen, S., 1995. Perceived nuclear risk, commitmentand appraisals of management: a study of nuclear power plant personnel. RiskAnalysis 15 (3), 391–395.

Kozlowski, S.W.J., Klein, K.J., 2000. A multilevel approach to theory and research inorganizations: contextual, temporal, and emergent processes. In: Klein, K.J.,Kozlowski, S.W.J. (Eds.), Multilevel Theory, Research and Methods inOrganizations: Foundations, Extensions, and New Directions. Jossey-Bass, SanFrancisco, CA, pp. 3–90.

LePine, J.A., Erez, A., Johnson, D., 2002. The nature and dimensionality oforganizational citizenship behavior: a critical review and meta-analysis.Journal of Applied Psychology 87, 53–65.

M. Martínez-Córcoles et al. / Safety Science 51 (2013) 293–301 301

Martínez-Córcoles, M., Gracia, F., Tomás, I., Peiró, J.M., 2011. Leadership andemployees’ safety behaviors in a nuclear power plant: a structural equationmodel. Safety Science 49, 1118–1129.

Mearns, K., Flin, R., Gordon, R., Fleming, M., 2001. Human and organizational factorsin offshore safety. Work & Stress 15, 144–160.

Michael, J.H., Guo, Z.G., Wiedenbeck, J.K., Ray, C.D., 2006. Production supervisorimpacts on subordinates safety outcomes: an investigation of leader–memberexchange and safety communication. Journal of Safety Research 3, 469–477.

Motowidlo, S.J., Van Scotter, J.R., 1994. Evidence that task performance should bedistinguished from contextual performance. Journal of Applied Psychology 79(4), 475–480.

Mullen, J.E., Kelloway, E.K., 2009. Safety leadership: a longitudinal study of theeffects of transformational leadership on safety outcomes. Journal ofOccupational and Organizational Psychology 82, 253–272.

Naevestad, T.O., 2008. Safety cultural preconditions for organizational learning inhigh-risk organizations. Journal of Contingencies and Crisis Management 16 (3),154–163.

Neal, A., Griffin, M.A., 2004. Safety climate and safety at work. In: Barling, J., Frone,M. (Eds.), The Psychology of Workplace Safety. American PsychologicalAssociation, Washington, DC.

Neal, A., Griffin, M.A., 2006. A study of the lagged relationships among safetyclimate, safety motivation, safety behavior, and accidents at the individual andgroups levels. Journal of Applied Psychology 91 (4), 946–953.

Organ, D.W., 1988. Organizational Citizenship Behavior: The Good SoldierSyndrome. Lexington Books, Lexington, MA.

Parker, S.K., Turner, N., Griffin, M.A., 2003. Designing healthy work. In: Hofmann, D.A.,Tetrick, L.E. (Eds.), Health and Safety in Organizations: A Multilevel Perspective.Society for Industrial and Organizational Psychology. Jossey-Bass, NJ, pp. 1–130.

Pearce, C.L., Conger, J.A., 2003. Shared Leadership: Reframing the Hows and Whys ofLeadership. Sage, Thousand Oaks, CA.

Pearce, C.L., Sims, H.P., 2002. Vertical versus shared leadership as predictors of theeffectiveness of change management team: an examination of aversive,directive, transactional transformational, and empowering leader behaviors.Group Dynamics: Theory Research, and Practice 6 (2), 172–197.

Reason, J., 1990. Human Error. Cambridge University Press, Cambridge, UK.Reason, J., 2008. The Human Contribution: Unsafe Acts, Accidents and Heroic

Recoveries. Ashgate, Farnham Surrey, UK.Richter, A., 2003. New ways of managing prevention: a cultural and participative

approach. Safety Science Monitor 7 (1), 1–10.Robinson, S., Bennett, R., 1995. A typology of deviant workplace behaviors: a multi-

dimensional scaling study. Academy of Management Journal 38 (2), 555–572.Rochlin, G.I., von Meier, A., 1994. Nuclear power operations: a cross-cultural

perspective. Annual Review of Energy and the Environment 19, 153–187.Rotundo, M., Sackett, P.R., 2002. The relative importance of task, citizenship, and

counterproductive performance to global ratings of job performance: a policycapturing approach. Journal of Applied Psychology 87 (1), 66–80.

Salas, E., Dickinson, T.L., Converse, S.A., Tannenbaum, S.I., 1992. Toward anunderstanding of team performance and training. In: Swezey, R.W., Salas, E.(Eds.), Teams: Their Training and Performance. ABLEX, Norwood, NJ, pp. 3–29.

Schöbel, M., 2005. Compliance with safety rules: the role of environmentalstructures. In: Itoigawa, N., Wilpert, B., Fahlbruch, B. (Eds.), EmergingDemands for the Safety of Nuclear Power Operations. CRC Press LLC, Florida,pp. 99–108.

Simard, M., Marchand, A., 1994. The behavior of first-line supervisors in accidentprevention and effectiveness in occupational safety. Safety Science 17,169–185.

Simard, M., Marchand, A., 1997. Workgroups’ propensity to comply with safetyrules: the influence of micro-macro organizational factors. Ergonomics 40 (2),172–188.

Smith, M.J., Cohen, H.H., Cohen, A., Cleveland, R.J., 1978. Characteristics of successfulsafety programs. Journal of Safety Research 10, 5–15.

Smith, C.A., Organ, D.W., Near, J.P., 1983. Organizational citizenship behavior: itsnature and antecedents. Journal of Applied Psychology 68, 653–663.

Srivastava, A., Bartol, K.M., Locke, E.A., 2006. Empowering leadership inmanagement teams: effects on knowledge sharing, efficacy, and performance.Academy of Management Journal 49, 1239–1251.

Tjosvold, D., 1990. Flight crew collaboration to manage safety risks. Group andOrganization Studies 15, 177–191.

Vecchio, R.P., Justin, J.E., Pearce, C.L., 2010. Empowering leadership: an examinationof mediating mechanisms within a hierarchical structure. The LeadershipQuarterly 21, 530–542.

Wahlström, B., 2005. Challenges in the nuclear industry: perspectives from seniormanagers and safety experts. In: Itoigawa, N., Wilpert, B., Fahlbruch, B. (Eds.),Emerging Demands for the Safety of Nuclear Power Operations. CRC Press LLC,Florida, pp. 17–29.

Widaman, K.F., 1985. Hierarchically nested covariance structure models formultitrait–multimethod data. Applied Psychological Measurement 9, 1–26.

Yule, S., Flin, R., Murdy, A., 2007. The role of management and safety climate inpreventing risk-taking at work. International Journal of Risk Assessment andManagement 7 (2), 127–151.

Zohar, D., 2000. A group-level model of safety climate: testing the effect of groupclimate on microaccidents in manufacturing jobs. Journal of Applied Psychology85 (4), 587–596.

Zohar, D., 2002. The effects of leadership dimensions, safety climate, and assignedpriorities on minor injuries in work groups. Journal of Organizational Behavior23, 75–92.

Zohar, D., 2003. The influence of leadership and climate on occupational health andsafety. In: Hofmann, D.A., Tetrick, L.E. (Eds.), Health and Safety in Organizations:A Multilevel Perspective. Society for Industrial and Organizational Psychology.Jossey-Bass, NJ, pp. 201–230.

Zohar, D., 2008. Safety climate and beyond: a multi-level multi-climate framework.Safety Science 46, 376–387.