See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/26762637

Workplace Safety: A Meta-Analysis of the Roles of Person and Situation

Factors

Article in Journal of Applied Psychology · September 2009

DOI: 10.1037/a0016172 · Source: PubMed

CITATIONS READS
1,448 29,501

4 authors:

Michael S. Christian Jill C. Bradley-Geist

University of North Carolina at Chapel Hill California State University, Fresno
47 PUBLICATIONS 7,144 CITATIONS 13 PUBLICATIONS 2,857 CITATIONS

SEE PROFILE SEE PROFILE

Craig Wallace Michael Burke

University of Denver Tulane University
46 PUBLICATIONS 5,676 CITATIONS 97 PUBLICATIONS 13,967 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Michael Burke on 04 June 2014.

The user has requested enhancement of the downloaded file.

Journal of Applied Psychology © 2009 American Psychological Association
2009, Vol. 94, No. 5, 1103–1127 0021-9010/09/$12.00 DOI: 10.1037/a0016172

Workplace Safety: A Meta-Analysis of the Roles of Person

and Situation Factors

Michael S. Christian Jill C. Bradley

University of Arizona California State University, Fresno

J. Craig Wallace Michael J. Burke

Oklahoma State University Tulane University

Recent conceptual and methodological advances in behavioral safety research afford an opportunity to
integrate past and recent research findings. Building on theoretical models of worker performance and
work climate, this study quantitatively integrates the safety literature by meta-analytically examining
person- and situation-based antecedents of safety performance behaviors and safety outcomes (i.e.,
accidents and injuries). As anticipated, safety knowledge and safety motivation were most strongly
related to safety performance behaviors, closely followed by psychological safety climate and group
safety climate. With regard to accidents and injuries, however, group safety climate had the strongest
association. In addition, tests of a meta-analytic path model provided support for the theoretical model
that guided this overall investigation. The implications of these findings for advancing the study and
management of workplace safety are discussed.

Keywords: occupational safety, safety performance, accidents, safety climate, Occupational Safety and
Health Administration

Thousands of deaths and disabilities occur because of occupa- Hence, we have four goals for the current study. First, we
tional accidents each year in the United States, including 5,804 illustrate the benefits of developing clear operationalizations of
work-related fatalities and 4.1 million nonfatal occupational inju- safety constructs. Second, we build on existing theory and
ries and illnesses in 2006 (Bureau of Labor Statistics, U.S. De- research (e.g., Campbell, McCloy, Oppler, & Sager, 1993; Neal
partment of Labor, 2007). Given these statistics, researchers have & Griffin, 2004) by detailing a conceptual framework with
devoted much effort to studying workplace safety. Although an which to organize and study relationships between antecedents
impressive quantity of information has resulted, much of the and safety criteria. To this end, we organize constructs to
behaviorally oriented occupational safety research is plagued by develop a parsimonious description of the person- and
lack of theory, weak methodology, and unclear conceptualizations situation-related antecedents of workplace safety. Third, using
of constructs. Moreover, studies of antecedents to safety have this conceptual framework, we meta-analytically estimate hy-
tended to focus on either individual differences or contextual pothesized relationships. Fourth, we use meta-analytic path
factors but rarely on both. Additionally, though previous studies modeling to test an exemplar model of the integrated conceptual
have summarized aspects of this literature (Clarke, 2006a; Clarke framework.
& Robertson, 2005), these efforts have not integrated the array of
situational and individual antecedents to safety nor have they
attended to levels-of-analysis issues that have implications for the Conceptualizing Workplace Safety
interpretation of findings.
One shortcoming in the safety literature is a lack of clear and
consistent construct definitions and conceptualizations, both on
Michael S. Christian, Eller College of Management, University of the predictor and criterion sides (cf. Clarke & Robertson, 2005).
Arizona; Jill C. Bradley, Craig School of Business, California State Uni- As a result, inconsistencies exist between studies, and empirical
versity, Fresno; J. Craig Wallace, William S. Spears School of Business, findings do not always align with theoretical predictions. Al-
Oklahoma State University; Michael J. Burke, A. B. Freeman School of though there have been efforts to overcome this situation in
Business, Tulane University. particular domains (e.g., safety climate; Flin, Mearns,
Michael S. Christian and Jill C. Bradley contributed equally to this O’Connor, & Bryden, 2000), no study has comprehensively
article. We thank Adela Garza for her invaluable assistance throughout the
addressed such deficiencies. Clear delineation of constructs is a
research process. Also, thanks to Kim Mathe for her help with article
critical step to facilitate not only the organization of accumu-
compilation.
Correspondence concerning this article should be addressed to Michael lated knowledge, but also the development of theory in the
S. Christian, Eller College of Management, University of Arizona, Depart- safety domain. Thus, we begin by clarifying conceptualizations
ment of Management and Organizations, McClelland Hall, P.O. Box of safety criteria before presenting a model for classifying and
210108, Tucson, AZ 85721-0108. E-mail: msc@email.arizona.edu understanding their antecedents.

1103
1104 CHRISTIAN, BRADLEY, WALLACE, AND BURKE

Problematically, the term safety performance may be used to safety performance behaviors, which then directly relate to safety
refer to two different concepts. At times, safety performance might outcomes, such as accidents and injuries. We used a modified
refer to an organizational metric for safety outcomes, such as version of Neal and Griffin’s framework for organizing the liter-
number of injuries per year. Conversely, safety performance may ature and studying construct relations (see Figure 1). Accordingly,
refer to a metric for safety-related behaviors of individuals (e.g., we posit that situational factors, individual differences, and atti-
Burke, Sarpy, Tesluk, & Smith-Crowe, 2002; Neal & Griffin, tudes are distal in their relationships with safety performance and
2004). Distinguishing safety-related behaviors from the outcomes are even more distally related to safety outcomes. These factors are
of those behaviors is important, because each might have differ- expected to impact more proximal states or self-regulatory pro-
ential relationships with antecedents. Thus, we consider safety cesses that directly affect safety performance behaviors. Of im-
performance behaviors and safety outcomes to be distinct. In portance, this theoretical framework informs not only the magni-
contrast to safety performance behaviors, safety outcomes are tudes of the relationships we expected to observe between various
tangible events or results, such as accidents, injuries, or fatalities. antecedents and safety criteria but also the processes through
Conceptualizing safety performance as individual behaviors which workplace accidents and injuries occur.
provides researchers with a measurable criterion, which is more
proximally related to psychological factors than accidents or inju- Antecedents of Safety Performance and Safety Outcomes
ries. Safety performance behaviors can be predicted with greater
accuracy than outcomes, which often have a low base rate and At the broadest level, we classified antecedents as person re-
skewed distributions (cf. Zohar, 2000). Similar to job performance lated or situation related; within each of these areas, we identified
in general, safety performance behaviors can be scaled by the more proximal and more distal antecedents to safety performance
frequency with which employees engage in the behaviors and are behaviors (see Figure 1). We considered safety knowledge and
distinguishable in terms of their antecedents and covariation with safety motivation proximal antecedents to safety performance be-
safety outcomes (Burke, Sarpy, et al., 2002). However, although haviors. In contrast, situation-related factors and individual dispo-
safety performance is conceptually similar to job performance in sitional characteristics and attitudes were considered to be more
general, it does not fit neatly into task, contextual, or adaptive distal. With regard to safety outcomes, all of the antecedents are
performance and thus should be treated as a separate domain of job indirect in that they operate through safety performance behaviors.
performance (Burke, Sarpy, et al., 2002; Parker & Turner, 2002). The only direct antecedent to safety outcomes in the theoretical
Several conceptual models of safety performance have been model is safety performance behavior. However, for consistency,
advanced. The model of safety performance outlined by Burke, we refer to the antecedents as either distal or proximal, indicating
Sarpy, et al. (2002)— defined as “actions or behaviors that indi- relative distance from either criterion.
viduals exhibit in almost all jobs to promote the health and safety As a general rule, proximal factors were anticipated to yield
of workers, clients, the public, and the environment” (p. 432)— larger relationships than distal factors. Further, where theoretically
includes four factors: (a) using personal protective equipment, (b) relevant, we distinguish between safety compliance and safety
engaging in work practice to reduce risk, (c) communicating participation (Neal & Griffin, 2004). Consistent with research on
hazards and accidents, and (d) exercising employee rights and task versus contextual performance in the general organizational
responsibilities. Although the factors are distinct (but correlated), literature, Griffin and Neal (2000) found that safety motivation
Burke, Sarpy, et al. suggested that under certain conditions, using was more strongly related to safety participation than safety
the aggregate of the four factors is appropriate. Other conceptual- knowledge, whereas the converse was true for safety compliance.
izations of safety performance distinguish between safety “com- We likewise anticipate that motivation should play a larger role in
pliance” and safety “participation,” with the former referring to discretionary safety participation behaviors, whereas knowledge
“generally mandated” safety behaviors and the latter referring to should be more related to compulsory safety compliance.
safety behaviors that are “frequently voluntary” (Neal, Griffin, &
Hart, 2000, p. 101). This distinction is similar to that between task Person Related: Proximal Antecedents
and contextual performance in the job performance literature (e.g.,
Borman & Motowidlo, 1993). Safety knowledge was the first proximal person-related factor. In
line with our conceptual model, we anticipated that knowledge
would have a strong positive relationship with safety performance
Predicting Safety Criteria: A Conceptual Model because knowledge is a direct determinant of performance behav-
To develop a model of the processes through which situations iors. In short, knowing how to perform safely (e.g., handling
and individual difference factors influence safety performance hazardous chemicals, emergency procedures) is a precondition to
behaviors and outcomes, we build upon Neal and Griffin’s (2004) enacting safe behaviors. Thus, safety knowledge should be
model of workplace safety. This model is grounded in Campbell et strongly related to safety performance behaviors. Furthermore,
al.’s (1993) theory of performance, which identifies three proximal
determinants of an individual’s performance— knowledge, skills, 1
Kanfer (1990, 1992) proposed that distal traits relate to performance
and motivation to perform—and suggests that distal antecedents of
and outcomes through proximal statelike individual differences and self-
performance (e.g., training, organizational climate, personality) regulatory processes. A great deal of empirical support has been gleaned
presumably influence performance through increases in these for this motivational process as well (e.g., Barrick et al., 1993; Bergman,
proximal determinants.1 Hence, Neal and Griffin posited that Donovan, Drasgow, Overton, & Henning, 2008; Chen, Casper, & Cortina,
antecedents like safety climate or personality directly influence 2001; Chen, Gully, Whiteman, & Kilcullen, 2000), even in the safety
safety motivation and knowledge, which in turn directly influence literature (e.g., Wallace & Chen, 2006).
 WORKPLACE SAFETY 1105

Figure 1. An integrative model of workplace safety.

safety knowledge should exhibit higher correlations with safety related to motivation (Furnham, Petrides, Jackson, & Cotter,
compliance than with safety motivation (cf. Griffin & Neal, 2000). 2002). Thus, because conscientiousness and safety motivation
For safety outcomes, we expected moderate negative relationships, should be positively related (e.g., Griffin, Burley, & Neal, 2000),
because knowledge effects should operate through safety perfor- we anticipated that conscientiousness would indirectly relate to
mance behaviors. safety performance behaviors, particularly those that are voluntary
Safety motivation was another direct person-related antecedent (i.e., safety participation). Because of the distal relationship be-
of safety performance behaviors. Safety motivation reflects “an tween conscientiousness and safety outcomes, we anticipated a
individual’s willingness to exert effort to enact safety behaviors weak negative or equivocal relationship between conscientious-
and the valence associated with those behaviors” (Neal & Griffin, ness and accidents and injuries, as indicated by Clarke and Rob-
2006, p. 947). For the same reasons identified for safety knowl- ertson’s (2005) meta-analysis.
edge, we expected safety motivation to be strongly related to safety People high in neuroticism may have difficulty coping with
performance and moderately related to fewer accidents and inju- threatening situations (cf. Hobfoll, 1989; Kanfer & Ackerman,
ries. Given the motivation-related conceptualization of safety par- 1989), in part because they may devote more resources to worry
ticipation, we expected safety motivation would be more strongly and anxiety as opposed to the task at hand. Additionally, neurot-
related to safety participation than safety compliance. icism is negatively related to intrinsic motivation (Furnham et al.,
2002) and the desire to take control over one’s environment
Person Related: Distal Antecedents (Judge, 1993), both of which likely hamper safety performance.
With a few exceptions, we used the Big Five framework (e.g., On the other hand, given their vigilance toward negative stimuli in
Costa & McCrae, 1985) to organize dispositional traits. As noted the environment, people high in neuroticism may be attuned to
by Clarke and Robertson (2005) in their meta-analysis of person- signs of danger in the workplace (Mathews & MacLeod, 1985).
ality and accidents, the Big Five is useful for providing order to a Further, in their meta-analysis, Clarke and Robertson (2005) found
disordered literature. Still, the occupational safety literature con- that neuroticism had a negligible relationship with accidents. Thus,
tains too few studies of openness to experience and agreeableness we expected a weak negative correlation with safety performance
to consider in a cumulative investigation. behaviors and an even weaker positive association with accidents
Conscientiousness comprises both achievement and responsibil- and injuries.
ity (dependability) components (Hough, 1992). Conscientious in- Extraversion was anticipated to be only weakly or equivocally
dividuals are more likely to set, commit to, and strive for personal related to safety performance behaviors or safety outcomes. On the
goals; they also are more dependable and responsible than less one hand, high extraversion could be detrimental because the
conscientious individuals. Further, conscientiousness is positively sensation-seeking aspect of the trait could lead people to engage in
1106 CHRISTIAN, BRADLEY, WALLACE, AND BURKE

risky behavior (e.g., Golimbet, Alfimova, Gritsenko, & Ebstein, James, James, & Ashe, 1990). Thus, we define group-level safety
2007). On the other hand, extraversion is closely aligned with climate as shared perceptions of work environment characteristics
positive affect (Eysenck, 1967; Iverson & Erwin, 1997), which is as they pertain to safety matters that affect a group of individuals
associated with high self-efficacy (Judge, 1993) and contextually (e.g., Neal & Griffin, 2004; Zohar & Luria, 2005).
oriented behaviors. Moreover, extraversion has been found to be Further, we view safety climate as having a hierarchical
unrelated to accidents (e.g., Clarke & Robertson, 2005). structure at both the psychological and group levels, on the
Locus of control is the extent to which people feel they person- basis of theoretical arguments and the confirmatory factor an-
ally control the events in their lives as opposed to those events alytic work of L. A. James and James (1989) and Burke,
being controlled by the external environment (cf. Judge, Erez, Borucki, and Hurley (1992). At the higher order factor level, we
Bono, & Thoresen, 2002). The Big Five trait most strongly related conceptualize psychological climate with respect to employees’
to locus of control appears to be neuroticism (Judge & Bono, perceptions of well-being. That is, at the higher order factor
2001). However, given that over half of the variance in locus of level, we view first-order climate factors as driven by an
control cannot be explained by neuroticism, we examined these
employee’s emotional evaluation of the degree to which the
two categories separately. People who believe they can control
work environment is perceived as personally beneficial or det-
events should be more motivated to learn about and engage in safe
rimental. In fact, Griffin and Neal’s (2000) factor analytic and
practices than people who do not believe they can control acci-
path modeling research, which involved a higher order safety
dents. Given its distal relationship with criteria, we anticipated a
climate factor, relied directly on L. A. James and James (1989)
moderate relationship between internal locus of control and safety
performance behaviors (particularly safety participation due to the and Burke et al.’s (1992) arguments in positing and confirming
motivational component inherent in locus of control) and a weaker that safety climate is driven by a singular, higher order factor
relationship with outcomes. reflecting assessments of well-being. When this perception is
Propensity for risk taking has been described as an amalgam- aggregated to the group level, or when the perception refers to
ation of several Big Five traits (Nicholson, Soane, Fenton- the degree to which the work environment is beneficial or
O’Creevy, & Willman, 2005). People high in risk taking tend to be detrimental to the group as a whole, this higher order factor is
impulsive sensation seekers (Zuckerman, Kuhlman, Thornquist, & conceptually a group-level factor (e.g., L. R. James et al.,
Kiers, 1991), who might be more apt than their coworkers to 2008).
engage in unsafe behaviors either because they underestimate the Safety climate was expected to positively influence safety
chances of accidents or because they are actually stimulated by performance behaviors (through safety knowledge and motiva-
risk. Thus, we expected risk taking to have a negative relationship tion) and to negatively influence outcomes. A positive safety
with safety performance and a positive relationship with safety climate should encourage safe action either through reward or
outcomes. Further, we predicted risk taking to be moderately through principles of social exchange (cf. Clarke, 2006a; Grif-
related to safety performance and weakly related to outcomes fin & Neal, 2000; Hofmann, Morgeson, & Gerras, 2003; Zohar,
because of its distal relationship with the criteria in the theoretical 2000). Further, positive safety climates should enhance safety
model. knowledge because they are reflective of environments where
Attitudes, unlike personality, are presumably fluid and suscep- safety information is communicated formally through training
tible to change depending on the situation (e.g., Petty & Cacioppo, and meetings and informally through on-the-job discussion.
1986). Here, we examine general job attitudes that people hold Thus, we anticipated that safety climate would be moderately
about their work (i.e., job satisfaction and organizational commit- related to safety performance behaviors and weakly related to
ment). In theory, more positive attitudes might lead to greater more distal safety outcomes. We further anticipated that safety
motivation to behave safely. However, given research suggesting climate would be more strongly related to safety participation
that attitudes are a distal and imperfect predictor of behavior (e.g., than safety compliance, because of the voluntary nature of
Fazio & Williams, 1986) and the equivocal findings linking job
participation and the motivational desire of employees to recip-
attitudes with performance (cf. Judge, Thoresen, Bono, & Patton,
rocate manager actions regarding safety (e.g., Clarke, 2006a;
2001), we do not offer firm predictions.
Hofmann et al., 2003). We also expected that group-level
climate would have stronger relationships with safety perfor-
Situation Related: Safety Climate mance and outcomes than psychological climate. Zohar (e.g.,
A recent meta-analysis by Clarke (2006a) demonstrated that 2000) has argued that group-level climate results from patterns
safety climate is a meaningful predictor of safety performance of behaviors and practices as opposed to isolated events or
behaviors (particularly safety participation) and is weakly related environmental circumstances. For perceptions to be shared
to accidents. In the current study, we build on Clarke’s findings by among individuals, an objective reality in the external environ-
further differentiating safety climate into psychological safety cli- ment must be concrete and influential enough that people can
mate and group safety climate. We define psychological safety agree in their perceptions.
climate as individual perceptions of safety-related policies, prac- To examine specific facets of safety climate, we utilized the
tices, and procedures pertaining to safety matters that affect per- taxonomy put forth by Neal and Griffin (2004). As shown in
sonal well-being at work (cf. L. A. James & James, 1989; L. R. Table 1, included within this taxonomy are the following: man-
James, Hater, Gent, & Bruni, 1978; L. R. James & Sells, 1981). agement commitment, human resources management practices,
When these perceptions are shared among individuals in a partic- safety systems, supervisory support, internal group processes,
ular work environment, a group-level climate emerges (L. R. boundary management (for which we found no relevant studies),
 WORKPLACE SAFETY 1107

Table 1
First-Order Factors of Safety Climate

Factor Element Example

Management commitment The extent to which people perceive that management values Perceived organizational support, management
safety and engages in communication and actions that safety practices and/or values, managerial
support safety communication of safety
Human resource management The extent to which people perceive that selection, training, Selection systems, safety training, perfor-
practices and reward systems contribute to safety mance management, reward systems
Safety systems Perceived quality of policies, procedures, or interventions Hazard management, incident investigations,
implemented by an organization with the intention of safety policies and procedures
improving safety outcomes
Supervisor support The extent to which people believe their supervisor values Supervisor safety consciousness, supervisory
safety as reflected in communication, encouragement, and safety values, supervisor safety
consequences communication, supervisory safety
orientation
Internal group processes Perceptions of communication and support for safety within Safety backup, safety communication, peer
work groups or the extent to which employees perceive safety orientation, trust in peers
that their coworkers provide them with safety-related
cooperation and encouragement
Boundary management The perceived quality of communication between the work N/Aa
group and other relevant stakeholders regarding safety
issues
Risk The extent to which workers perceive the work itself as Perceived job risk, perceived accident
dangerous potential, perceived physical hazards,
perceived job safety
Work pressure The extent to which the workload overwhelms one’s ability Production pressure, pressure to take
to perform safely shortcuts, workload, time pressure, role
overload

Note. Adapted from Neal and Griffin (2004).

a
No applicable studies were found that examined the boundary management factor.

risk, and work pressure.2 Because each first-order factor is distally outcomes. Included in our definition of safety outcomes were
related to safety behavior and outcomes, we anticipated that they accidents, injuries, and fatalities as well as safety performance
would generally have moderate relationships with safety perfor- behaviors. Keywords for the literature searches included combi-
mance behavior and weaker relationships with outcomes. nations of the following: safe(ty) climate; safe(ty) behaviors;
safe(ty) performance; (workplace, organizational, or occupa-
Situation Related: Leadership tional) (injuries, accidents, or fatalities). Through September
2008, we conducted electronic literature searches of databases,
Leadership refers to perceptions of how a manager behaves,
including PsycINFO, Social Science Citation Index, and
enacts, and achieves organizational or group objectives in general
MEDLINE. In addition, we conducted manual searches of major
(as compared with the supervisor support facet of safety climate,
journals relevant to industrial– organizational psychology and oc-
which refers to safety-specific supervisory behaviors; cf. Zohar,
cupational safety (e.g., Academy of Management Journal, Journal
2000). We included constructs such as leader–member exchange
(LMX) and transformational leadership in this category. Employ- of Applied Psychology, Personnel Psychology, Journal of Safety
ees who have positive feelings toward their leader are more likely Research, Journal of Occupational Health Psychology, Accident
to reciprocate when possible. As such, leadership quality has been Analysis and Prevention, Safety Science) to locate articles that did
found to be related to occupational safety and safety outcomes not surface in the database searches. We also consulted reference
(Hofmann et al., 2003; Hofmann & Morgeson, 1999; Zohar, sections of recent review articles to identify additional studies
2002a; Zohar & Luria, 2003). Further, Hofmann et al. (2003)
found that high-quality relationships with supervisors predicted 2
Many of our first-order climate factors (i.e., first-order climate factors
employees’ safety-related citizenship behaviors. Hence, we ex- coming from Neal & Griffin’s, 2004, work) were included in Griffin and
pected leadership to have a stronger relationship with safety par- Neal’s (2000) hierarchical model of safety climate (e.g., Internal Group
ticipation than with compliance. However, we expected leadership Processes in our study overlaps with the factor they labeled Safety Com-
to have a moderate relationship with performance and a weak munication; Management Commitment in our study was labeled as Man-
relationship with accidents. agement Values in their earlier work). Notably, our first-order climate
factors, which reflect more specific content related to these work environ-
ment characteristics, are highly consistent with common first-order factors
Method
(e.g., the focus on supervision, work group processes, human resource
Literature Search practices, role overload, etc.) identified in several reviews of the safety
climate literature (cf. Flin et al., 2000) and reviews of the general climate
A search was conducted to identify all peer-reviewed published literature (e.g., Burke, Borucki, & Kaufman, 2002; L. R. James et al.,
articles about predictors of occupational safety performance and 2008).
1108 CHRISTIAN, BRADLEY, WALLACE, AND BURKE

(e.g., Burke, Holman, & Birdi, 2006; Clarke & Robertson, 2005). ers in routine and nonroutine contexts, are requisite tasks for which
The initial searches yielded over 500 potential articles. individuals often receive extensive training.

Criteria for Inclusion Categorization of Predictor Variables

We first generated a list of all predictor constructs and specific
After collecting the articles, two researchers independently as-
operational measures utilized in each study. From this list, two
sessed each study to determine that (a) the study reported an effect
independent raters grouped together identical and conceptually
size between one or more antecedent and one or more safety
similar predictors based on logic, empirical evidence, and theoret-
outcome, (b) the outcome occurred on the job, and (c) the job or
ical considerations (e.g., combining neuroticism and negative af-
outcome was not driving related. We excluded driving outcomes
fectivity, combining conscientiousness and cognitive failures,
because many studies of driving safety confound work-related
which is reflective of poor dependability; cf. Wallace & Chen,
driving with personal-use driving. Moreover, evidence suggests
2005). At the broadest level, predictors were sorted into person-
differences in the antecedents of driving accidents versus other
related and situation-related factors. Within each factor, several
workplace accidents (Iversen & Rundmo, 2002; Lajunen, 2001;
domains were created to organize construct categories.
Wagenaar, 1992). After independent evaluation, agreement about
For person factors, the domains included proximal factors (i.e.,
study inclusion was near 100%, and the researchers resolved
safety knowledge, safety motivation) and distal factors (e.g., Big
several discrepancies through discussion. At the end of the pro-
Five personality traits, internal locus of control, risk-taking pro-
cess, 90 studies and 1,744 effect sizes had been identified for the
pensity, and job attitudes). For situation factors, categories were
meta-analysis, 477 of which were utilized in the predictor-criterion
based on whether the measure was individual level (e.g., psycho-
analyses.
logical safety climate or leadership) or group level (e.g., work
group or organizational safety climate).
Categorization of Criterion Variables In our categorizing of safety climate, we encountered two con-
ceptual decisions. First, we had to clarify our conceptualization of
To categorize criterion variables, two independent raters sorted climate with regard to its factor structure, and second we had to
them into the predetermined safety criterion categories described create a rule for how to operationalize climate at different levels of
next. After both raters categorized each effect, the results were analysis. First, we conceptualized climate as a higher order overall
compared to establish agreement, initially estimated at 94%. All factor in addition to a more specific set of first-order factors
discrepancies were resolved through discussion, resulting in 100% consistent with Neal and Griffin (2004). Thus, we created a cate-
agreement. gory for overall safety climate, which included studies reporting
Safety outcomes: Accidents and injuries. Accidents and inju- (a) an overall composite safety climate score or (b) at least two
ries are often treated interchangeably with regard to their predic- dimensions consistent with Neal and Griffin (2004), from which
tors (e.g., Cooper, Phillips, Sutherland, & Makin, 1994; Sulzer- we calculated a composite data point. Next, studies were grouped
Azaroff, Loafman, Merante, & Hlavacek, 1990; Tuncel, Lotlikar, into each dimension of safety climate.
Salem, & Daraiseh, 2006). Moreover, definitions of accidents are The second conceptual decision involved categorizing climate
often confounded with injuries, with some researchers labeling as measures at different levels of analysis. We coded climate at two
accidents only those occurrences that result in injuries needing levels: individual (psychological) and group (work group, team, and
medical attention (cf. Visser, Ysbrand, Stolk, Neeleman, & Ros- organizational levels). We defined psychological climate as individual
malen, 2007). Thus, we computed an overall composite of acci- perceptions of work environment characteristics as they pertain to
dents and injuries. safety matters that affect personal well-being (cf. L. R. James et al.,
Safety performance. Our conceptualization of safety perfor- 1978; L. R. James & Sells, 1981), whereas group safety climate was
mance refers to individual behaviors; either measured at the indi- defined as shared perceptions of the same (e.g., Neal & Griffin, 2004;
vidual level or aggregated (e.g., rated groups of workers). Safety Zohar & Luria, 2005) with regard to the work group or the organi-
performance was defined as safety compliance and safety partic- zation. To be defined as group-level safety climate, a study must have
ipation (cf. Neal & Griffin, 2004). Safety-related behaviors re- (a) collected individual-level data and aggregated on the basis of an
quired by the organization were classified as safety compliance. acceptable measure of agreement (e.g., rwg, .70 or higher; Glick,
Safety participation consisted of voluntary behaviors that did not 1985) or (b) collected data at the group level (e.g., supervisor ratings
contribute to personal safety but supported safety in the larger of safety climate). Consistent with recommendations by Ostroff and
organizational context. We augmented this definition with the six Harrison (1999), we did not include cross-level effects or nested data.
safety citizenship factors provided by Hofmann et al. (2003): Rather, we included only correlations obtained from primary studies
communication and voice, helping, stewardship, whistle-blowing, that measured the predictor and criterion at the same levels. Table 2
civic virtue, and initiating safety-related change. Finally, we in- illustrates the levels of analysis at which we coded constructs for each
cluded a category representing a higher order safety performance primary study.
factor, consistent with Burke, Sarpy, et al.’s (2002) arguments.
This composite included broad, overall measures of safety-related Coding of Studies
behaviors in addition to specific task (safety compliance) and
contextual (safety participation) behaviors. This composite vari- The coding process was conducted by three researchers. Ini-
able is conceptually meaningful because in many types of safety- tially, the studies were divided among the researchers such that
related work, contextual aspects of work, such as helping cowork- (text continues on page 1116)
Table 2
Studies Included in the Meta-Analysis (N ⫽ 90 Studies)

Level of Coded criterion Source

Study N Design Industrial sector Sample description analysis Coded predictor construct Source (predictor) construct (criterion)

Barling et al. 16,446 Concurrent Misc. industries Australian employees Individual Job attitudes, safety Self Injuries Self
(2003) systems, and work
pressure
Barling et al. 174 Concurrent Food service Misc. employees Individual Safety climate and Self Accidents/injuries Self
(2002) leadership and safety
participation
164 Concurrent Misc. occupations, Employees younger Individual Safety climate, leadership, Self Accidents/injuries Self
88% service than 25 years old and work pressure and safety
participation
Borofsky & Smith 53 Longitudinal Manufacturing Misc. employees Individual Safety systems Self Accidents Archival
(1993)
Brown et al. (2000) 551 Concurrent Steel mills Equipment and Individual Locus of control, risk Self Safety compliance Self
maintenance taking behavior, safety
employees climate, and work
pressure
Burke, Sarpy, et al. 127–133 Concurrent Hazardous waste Operators, technicians, Individual Safety knowledge Archival Safety performance Supervisor
(2002) electricians, (overall), safety
engineers, handlers, compliance, and
plumbers, misc. safety
participation
Burke et al. (2008) 31 Concurrent Health care Misc. employees Organizational Safety climate Archival Accidents Archival
67 Concurrent Health care Misc. employees Organizational Safety climate Archival Safety performance Archival
(overall)
Burt et al. (2008) 104 Concurrent Power generation N/A Individual Job attitudes Self Safety participation Self
and
construction
WORKPLACE SAFETY

Cellar et al. (2001) 202 Concurrent Varied College students Individual Conscientiousness, Self Accidents Self
extraversion, and
neuroticism
Clarke & Ward 83 Concurrent Glassware Misc. employees Individual Safety climate and Self Safety participation Self
(2006) manufacturing leadership
22 Concurrent Glassware Supervisors Individual Safety climate and Self Safety participation Self
manufacturing leadership
Clarke (2006b) 109 Concurrent UK car N/A Individual Safety climate, leadership, Self Accidents and Self
manufacturing and work pressure safety
compliance
Cooper & Phillips 540 Longitudinal Packaging Production employees Individual Safety climate Self Safety performance Other
(2004) (overall)
Davids & Mahoney 34 Concurrent Industrial plants N/A Individual Extraversion and Self Accidents Archival
(1957) neuroticism
DeJoy et al. (2004) 2,120–2,182 Concurrent Retail stores Retail employees Individual Safety climate, management Self Safety performance Self
commitment, leadership, (overall)
internal group processes,
risk, and work pressure
Donald & Canter 10 Concurrent Chemical industry Misc. employees Organizational Safety climate Self Accidents Self
(1994)
Dunbar (1993) 44 Concurrent Chemical Emergency response Individual HRM practices Self Safety compliance Observer
manufacturing
(table continues)
1109
 1110

Table 2 (continued )

Level of Coded criterion Source

Study N Design Industrial sector Sample description analysis Coded predictor construct Source (predictor) construct (criterion)

Eklöf (2002) 92 Concurrent Swedish fishermen Swedish fishermen Individual Safety knowledge, locus of Self Safety compliance Self
control, and risk
Eklöf & Törner 92 Concurrent Fishermen Swedish fishermen Individual Safety knowledge, locus of Self Safety participation Self
(2005) control, risk-taking
behavior, and risk
Fallon et al. (2000) 359 Concurrent Home Sales associates Individual Conscientiousness Self Safety performance Self
improvement (overall)
retail
organization
Fellner & Sulzer-
Azaroff (1984) 6 Longitudinal Paper mill N/A Group Management commitment Self Injuries Archival
17 Longitudinal Paper mill N/A Group Management commitment Archival/manipulated Safety compliance Observer
Frone (1998) 319 Concurrent Part-time Misc. employees Individual Neuroticism, risk-taking Self Injuries Self
employees who behavior, job attitudes,
are students leadership, internal group
processes, risk, and work
pressure
Fullarton & Stokes 868 Concurrent Industrial N/A Individual Safety climate Self Injuries Archival
(2007) organizations
Geller et al. (1996) 328 Concurrent Plastics Maintenance/operations Individual Conscientiousness, Self Safety participation Self
manufacturing extraversion, locus of
control, risk-taking
behavior, and internal
group processes
202 Concurrent Textile Maintenance/operations Individual Conscientiousness, Self Safety participation Self
manufacturing extraversion, locus of
control, risk-taking
behavior, and internal
group processes
Goldenhar et al. 408 Concurrent Construction Laborers in the Pacific Individual Neuroticism, safety climate, Self Injuries and safety Self
(2003) Northwest HRM practices, compliance
leadership, and work
CHRISTIAN, BRADLEY, WALLACE, AND BURKE

pressure
Griffin & Neal 326 Concurrent Manufacturing and Australian employees Individual Safety knowledge, safety Self Safety compliance Self
(2000) mining motivation, safety and safety
climate, and HRM participation
practices
1,264 Concurrent Manufacturing and Australian employees Individual Safety knowledge and Self Safety compliance Self
mining safety climate and safety
participation
Hansen (1989) 362 Concurrent Chemical industry Production and Individual Neuroticism and risk Self Archival Accidents Archival
maintenance
Accidents Archival
Harrell (1990) 244 Concurrent Professional and Canadians Individual Risk and work pressure Self Accidents Self
blue-collar jobs
(table continues)
Table 2 (continued )

Level of Coded criterion Source

Study N Design Industrial sector Sample description analysis Coded predictor construct Source (predictor) construct (criterion)

Hayes et al. (1998) 297 Concurrent N/A Industrial accident Individual Job attitudes, management Self Accidents and Self
victims seen by commitment, safety safety
medical consulting systems, supervisor compliance
firm support, internal group
processes, risk, and work
pressure
301 Concurrent N/A Industrial accident Individual Management commitment, Self Accidents Self
victims seen by safety systems,
medical consulting supervisor support,
firm internal group processes,
and risk
156 Concurrent Utilities Telephone line Individual Management commitment, Self Accidents and Self
employees HRM practices, safety safety
systems, supervisor compliance
support, internal group
processes, and risk
Hemingway & 252 Concurrent Health care Nurses Individual Supervisor support, internal Self Injuries Self
Smith (1999) group processes, risk,
and work pressure
Hofmann & 49 Concurrent Manufacturing Group leaders Individual Management commitment Self Safety participation Self and
Morgeson and leadership and accidents archival
(1999)
Hofmann & Stetzer 21 Concurrent Chemical Management, Group Safety climate, internal Self Accidents, safety Archival
(1996) processing administrative, group processes, and compliance, and and self
operating core work pressure safety
participation
Hofmann & Stetzer 159 Concurrent Utility worker Outdoor utility Group Safety climate Self Safety participation Self
(1998)
WORKPLACE SAFETY

1,359 Concurrent Utility worker Outdoor utility Individual Safety climate Self Safety participation Self
Hsu et al. (2008) 256 Concurrent N/A Japanese employees Individual Locus of control, Self Safety performance Self
management (overall)
commitment, safety
systems, supervisor
support, and internal
group processes
295 Concurrent N/A Taiwanese employees Individual Locus of control, Self Safety performance Self
management (overall)
commitment, safety
systems, supervisor
support, and internal
group processes
Huang et al. (2006) 1,856 Concurrent Manufacturing, N/A Individual Locus of control, Self Injuries Self
construction, management
service, and commitment, and HRM
transportation practices
Iverson & Erwin 361–362 Concurrent Unionized blue- Production and Individual Neuroticism, extraversion, Self Injuries Archival
(1997) collar assembly leadership, internal group
employees processes, risk, and work
pressure
(table continues)
1111
 1112

Table 2 (continued )

Level of Coded criterion Source

Study N Design Industrial sector Sample description analysis Coded predictor construct Source (predictor) construct (criterion)

Johnson (2007) 17 Longitudinal Heavy N/A Group Safety climate Self Injuries and safety Archival
manufacturing performance observer
Jones & Wuebker 280 Concurrent Health care Hospital employees Individual Locus of control Self Injuries Self
(1993) excluding physicians
Krause et al. 73 Longitudinal Misc. plants N/A Organizational HRM practices Other Injuries Archival
(1999)
Liao et al. (2001) 1,286 Concurrent Emergency Firefighters, officers Individual Neuroticism and Self Injuries Archival
services extraversion
Lilley et al. (2002) 367 Concurrent Logging or Machine operators, Individual Work pressure Self Injuries Self
silviculture pruners, skid/log
Lingard (2002) 14 Longitudinal N/A N/A Individual HRM practices Other Safety compliance Other
Maierhofer et al. 218 Concurrent Beauty Hairstylists Individual Safety motivation and work Self Safety compliance Self
(2000) pressure
Supervisor support Supervisor
McLain & Jarrell 234 Concurrent Industrial Misc. jobs Individual Safety climate, management Self Safety performance Self
(2007) commitment, and work (overall)
pressure
Mearns et al. 722 Concurrent Offshore oil and Maintenance, Individual Safety motivation and risk Self Accidents Self
(2001) gas installations production, catering,
administration
Mearns et al. 55 Concurrent Offshore oil and N/A Group Job attitudes and Self Accidents and Archival
(2003) gas installations management commitment safety
compliance
13 Concurrent Offshore oil and N/A Group Job attitudes and Self Accidents Self
gas installations management commitment
40 Concurrent Offshore oil and N/A Organizational Management commitment Self Accidents and Archival
gas installations and work pressure safety
compliance
Melamed & 532 Concurrent 8 industrial plants Nonshift daytime Group Risk Archival Injuries Archival
Oksenberg (power, metal employees
(2002) fabrication,
composite
material, heavy
CHRISTIAN, BRADLEY, WALLACE, AND BURKE

machinery,
repair)
Michael et al. 641 Concurrent Wood products Production employees Individual Job attitudes, supervisor Self Injuries Archival
(2005) manufacturing support, and risk
Michael et al. 598 Concurrent Wood products Hourly employees Individual Job attitudes and supervisor Self Accidents and Self and
(2006) manufacturing support injuries archival
Mohamed (1999) 36 Concurrent Construction Managers Organizational Safety climate Self Safety performance Self
(overall)
Neal & Griffin 33 Longitudinal Health care Hospital staff Group Safety climate Self Accidents Archival
(2006) Safety performance Self
(overall)
135 Longitudinal Health care Hospital staff Individual Safety motivation, Self Safety compliance Self
neuroticism, and safety and safety
climate participation
(table continues)
Table 2 (continued )

Level of Coded criterion Source

Study N Design Industrial sector Sample description analysis Coded predictor construct Source (predictor) construct (criterion)

Neal et al. (2000) 525 Concurrent Health care Australian hospital Individual Safety knowledge, safety Self Safety performance Self
motivation, safety (overall) and
climate, and management safety
commitment compliance
Paul & Maiti 300 Concurrent Mining Coal miner Individual Neuroticism, risk taking Self Injuries and safety Archival
(2007) behavior, and job performance and
attitudes (overall) supervisor
Probst (2004) 136 Concurrent Manufacturing Production employees Individual Safety knowledge and Self Accidents/injuries Self
safety climate and safety
compliance
Probst & Brubaker 134–138 Longitudinal Food processing Misc. Individual Safety knowledge, job Self Accidents/injuries Self
(2001) attitudes, and supervisor and safety
support compliance
Prussia et al. 121 Concurrent Steel company Managers/supervisors Group Management commitment, Supervisor Safety performance Supervisor
(2003) risk, and work pressure (overall) and
safety
compliance
Real (2008) 645 Concurrent N/A Production employees Individual Supervisor support and risk Self Injuries and safety Self
performance
(overall)
Ringenbach & 209 Concurrent Nuclear power Clerical, operational, Individual Conscientiousness, safety Self Injuries and safety Self
Jacobs (1995) plant supervisors climate, and work compliance
pressure
Rundmo (1994) 857–863 Concurrent Offshore Norwegian Individual HRM practices and safety Self Accidents Self
petroleum climate
Rundmo (1996) 993–1,001 Concurrent Petroleum Sea platform Individual Risk Self Safety compliance Self
employees
WORKPLACE SAFETY

Rundmo (2000) 730 Concurrent Fertilizer, energy, Misc. employees Individual Neuroticism Self Accidents Self
oil
Rundmo (2001) 814 Concurrent Agricultural, Varied Individual Locus of control, risk, and Self Safety compliance Self
energy and work pressure
petrochemical
producing
Saari & Lahtela 246 Concurrent Light metal Production employees Individual Work pressure Other (observed by Accidents Archival
(1979) working experimenter)
Salminen & Klen 228 Concurrent Forestry Forestry employees Individual Locus of control Self Safety compliance Self
(1994)
Salminen et al. 185–203 Concurrent Forestry Forestry employees Individual Neuroticism, extraversion, Self Accidents Archival
(1999) locus of control, and
risk-taking behavior
Seo et al. (2004) 620 Concurrent Grain elevators Nonclerical and clerical Individual Safety climate Self Accidents, safety Self
performance
(overall), and
safety
compliance
Simard & 68 Concurrent Manufacturing Supervisors, managers Organizational Safety systems and Self Safety compliance Archival
Marchand plants supervisor support and safety and self
(1994) participation
(table continues)
1113
 1114

Table 2 (continued )

Level of Coded criterion Source

Study N Design Industrial sector Sample description analysis Coded predictor construct Source (predictor) construct (criterion)

Siu et al. (2003) 374 Concurrent Construction Construction work, Individual Management commitment, Self Accidents/injuries, Self
general HRM practices, safety
supervisor support, compliance, and
internal group processes, safety
and work pressure participation
Siu et al. (2004) 374 Concurrent Construction Construction Individual Neuroticism, job attitudes, Self Accidents Self
employees: Hong and safety climate
Kong
Smith et al. (2006) 33 Concurrent Companies Misc. Organizational Safety climate Self Injuries Archival
(unspecified)
Snyder et al. 253 Concurrent Facilities Misc. Individual Safety climate and risk Self Injuries Self
(2008) department:
food service,
maintenance,
distribution,
moving,
catering
73 Concurrent Pipefitting Pipefitters Individual Safety climate and risk Self Injuries Self
Stephenson et al. 324 Concurrent Mining Miners Individual Safety systems Assigned Safety compliance Self
(2005)
Storeth (2007) 1,442 Concurrent Transportation Norwegian: Individual Supervisor support, risk, Self Safety performance Self
administration, sales, and work pressure (overall)
transport
Sutherland & 554 Longitudinal Offshore drilling Misc. Individual Neuroticism Self Accidents Self
Cooper (1991) and production
Thoms & 23 Concurrent Manufacturing Department managers Group HRM practices and risk Self Accidents Archival
Venkataraman Other
(2002)
Torp & Moen 436 Longitudinal Garage Managers and Individual Safety systems Self Safety compliance Self
(2006) employees
Trimpop et al. 778 Concurrent Veterinary Vet surgeons, Individual Job attitudes, internal group Self Accidents Self
(2000) administration, lab processes, and work
technicians pressure
CHRISTIAN, BRADLEY, WALLACE, AND BURKE

Turner et al. (2005) 334 Concurrent Railway Trackside Individual HRM practices and work Self Safety participation Self
pressure
Varonen & Mattila 8 Concurrent Sawmills, Finnish Organizational Safety motivation, safety Self Accidents Archival
(2000) plywood, climate, and supervisor
parquet plants support
Varonen & Mattila 13 Concurrent Wood processing Misc. Group Safety systems Self Accidents Archival
(2002)
Wallace & 222 Concurrent Military Electrical work Individual Conscientiousness Self Accidents Self
Vodanovich
(2003a)
Wallace & 219 Concurrent Production Maintenance/assembly Individual Conscientiousness Self Accidents and Self
Vodanovich line safety
(2003b) performance
(overall)
(table continues)
Table 2 (continued )

Level of Coded criterion Source

Study N Design Industrial sector Sample description analysis Coded predictor construct Source (predictor) construct (criterion)

Watson et al. 408 Concurrent Steel manufacturing Sheet rolling Individual Management commitment, Self Safety compliance Self
(2005) leadership, internal group
processes, and risk
Wogalter et al. 79 Concurrent Misc. Introduction to Individual Safety systems Manipulated Safety compliance Self
(1999) psychology students
Wu et al. (2008) 920 Concurrent University Faculty, staff, Individual Safety climate and Self Safety participation Self
laboratories technicians, management commitment
janitorial
Wuebker (1986) 120 Concurrent Hospitality Hotel employees Individual Locus of control Self Accidents Archival
Zacharatos et al. 138 Concurrent Chemical, automotive, Human resources and Organizational HRM practices Archival Injuries Archival
(2005) construction safety directors
189 Concurrent Petroleum and Plant/field operators, Individual Safety knowledge, safety Self Injuries and safety Self
telecommunications technicians motivation, job attitudes, compliance
safety climate,
management
commitment, and HRM
practices
Zohar (2000) 53 Concurrent Metal-processing Fixing heavy Group Supervisor support and risk Self Injuries and Self
production equipment accidents
(microaccidents)
WORKPLACE SAFETY

Other
534 Concurrent Metal-processing Fixing heavy Individual Supervisor support, risk, Self Accidents Self
production equipment and work pressure (microaccidents)
Zohar (2002a) 42 Concurrent Metal processing Production employees Group Safety climate and Self Injuries Self
supervisor support
Zohar (2002b) 33 Concurrent Equipment Line workers and Group HRM practices Archival/manipulated Accidents Other
maintenance supervisors (microaccidents)
Work pressure Supervisor
Zohar & Luria 42 Concurrent Military Infantry soldiers Group Safety motivation, safety Self Injuries Archival
(2004) climate, supervisor
support, and risk
Supervisor
Zohar & Luria 36–401 Concurrent Metal, food, plastics, Manufacturing plant Group and Safety climate Self Safety performance Other
(2005) chemical industry employees organizational (overall)

Note. N ⫽ number of participants in each independent sample. Misc. ⫽ miscellaneous; N/A ⫽ study did not provide this information; HRM ⫽ human resource management; Other ⫽ rated by external observer(s).
1115
1116 CHRISTIAN, BRADLEY, WALLACE, AND BURKE

two individuals were responsible for coding and double coding, or Table 3
verifying the initial coding of each article. In instances of disagree- Mean Sample-Based Reliability Estimates Used for Analyses
ment between the first and second coding, a third researcher also
coded the study and either resolved the issue or the researchers met Mean reliability
Construct k N estimate
to arrive at a consensus through discussion. Following this process,
initial agreement was 89%, and through discussion, 100% consen- Predictor measure
sus was achieved. Safety knowledge 8 2,758 .784
Safety motivation 5 1,393 .845
Conscientiousness 7 1,601 .896
Meta-Analytic Calculations Neuroticism 9 3,255 .740
Extraversion 6 1,461 .741
We used the meta-analytic procedures proposed by Raju, Burke, Locus of control 7 2,307 .755
Normand, and Langlois (1991). Raju et al.’s procedure yields Risk taking 4 1,173 .773
estimates of construct-level effect sizes by correcting for artifac- Job attitudes 10 19,780 .766
tual error (i.e., sampling error, unreliability of measures), using Psychological safety climate 48 33,739 .794
Management commitment 9 4,352 .889
sample-based artifact data (i.e., reliability estimates from the pri-
HRM practices 4 2,964 .772
mary studies) as opposed to using artifact distributions. Raju et Safety systems 6 17,442 .769
al.’s procedures were optimally designed in the sense of estimating Supervisor support 16 8,091 .807
appropriately defined standard errors for corrected correlations Internal group processes 10 4,867 .880
when sample-based artifact values, such as a sample-based crite- Perceived job risk 14 7,986 .850
Work pressure 13 4,816 .755
rion reliability estimates (or assumed-fixed population reliability Group-level safety climate 14 794 .851
estimates), are incorporated into the corrections. The reader is Management commitment 1 121 .880
referred to more recent discussions by Raju and Brand (2003) and HRM practices 3 129 1.000a
Raju, Lezotte, Fearing, and Oshima (2006) on the estimation of the Safety systems 3 219 .851
Supervisor support 8 358 .805
standard errors for individually corrected correlations with sample-
Perceived job risk 1 42 .870
based and assumed (fixed) artifact values within Raju et al.’s Work pressure 2 63 .803
meta-analytic procedures. In addition, the reader is referred to Leadership 10 4,207 .796
Burke and Landis (2003) for the equation used to estimate the
standard error of the mean corrected correlation (assuming a Criterion measure
random effects model) used in this meta-analysis. As noted by Accidents and injuries 5 1,800 .784
Safety performance (overall) 18 6,076 .858
several authors, the utilization of a random effects model results in Safety compliance 24 7,348 .734
more accurate Type I error rates and more realistic confidence Safety participation 21 5,620 .790
intervals than does a fixed effect model (e.g., Erez, Bloom, &
a
Wells, 1996; Overton, 1998). Measures of group-level HRM practices included in the analysis were all
When reliability information was not provided for a particular scored dichotomously (i.e., either the organization had or did not have
practices in place), so predictor measures were assumed to contain no error.
effect, we substituted the best estimate of reliability, based on the All estimates were conducted within level (i.e., group or individual).
population of studies. To arrive at this estimate, sample size HRM ⫽ human resource management.
weighted mean reliabilities were calculated from all reported reli-
abilities for each construct measure within the study population.
Because scale reliability estimates may vary at multiple levels tion matrix containing corrected correlations between each vari-
(Zyphur, Kaplan, & Christian, 2008), reliabilities were estimated able in the model. Three decision criteria were applied in the
within level. For archival criterion data (e.g., number of accidents generation of this matrix: (a) the variables must enable a strong
during the year), no corrections for unreliability were made. The exemplar test of the model we present in Figure 1 (i.e., they must
estimates of reliability calculated in the current study can be found represent some theoretically derived combination of indirect
in Table 3. Across our analyses, 63% of the reliability data were situation-related factors and person-related factors, direct factors,
sample based. safety performance, and safety outcomes); (b) the variables should
A number of the studies measured a particular construct cate- represent the largest possible combination of sample sizes in each
gory in multiple ways (e.g., with different safety climate mea- cell of the matrix; and (c) the variables should all be measured at
sures); in these cases, composite correlations were derived with the the individual level.
Spearman–Brown formula (Hunter & Schmidt, 2004, pp. 454 – Once we had applied these decision rules, we settled on a model
463). Composite correlations, in comparison to a simple averaging integrating conscientiousness, safety climate, safety knowledge,
of correlations, are advantageous in that they provide a higher level safety motivation, safety performance, and safety outcomes. As we
of construct validity and limit downward biasing. have argued, conscientiousness, although presumably not related
to safety knowledge, should have a direct relationship with safety
Meta-Analytic Path Analysis motivation. Safety climate should have a direct effect on both
safety motivation and safety knowledge (e.g., Griffin & Neal,
To provide a more comprehensive analysis of the theoretical 2000). Knowledge and motivation should be directly related to
relationships among the factors in our conceptual framework (see performance, which is directly related to safety outcomes.
Figure 1), we applied path analysis techniques to our meta-analytic The resulting input matrix consisted of 15 cells. Two cells in our
data to test an exemplar path model. As input, we used a correla- matrix were empty, between (a) conscientiousness and safety
 WORKPLACE SAFETY 1117

motivation and (b) conscientiousness and safety knowledge. Thus, In the next sections, we focus on our expectations for magnitude
following recommendations by Viswesvaran and Ones (1995), we and direction regarding each antecedent’s correlation with perfor-
used assumed corrected population values as estimates of these mance and outcomes, discussing correlation magnitudes according
relationships. For the relationship between conscientiousness and to the guidelines provided by Cohen (1988), which suggest that an
knowledge, we used the sample-weighted average of the values effect size between .1 and .3 be considered weak, an effect size
from Colquitt, LePine, and Noe, (2000) and Mauer, Lippstreu, and between .3 and .5 be considered moderate, and an effect size of .5
Judge (2008). This value (rc ⫽ .00, N ⫽ 1,908) represented a or higher be considered strong. Further, we interpret our findings
combination of declarative and procedural knowledge, which are with respect to the magnitudes of each relationship with overall
both components of safety knowledge (Burke, Sarpy, et al., 2002). safety performance and outcomes. We report the findings with
The assumed value between conscientiousness and motivation respect to safety participation and safety compliance only for those
(rc ⫽ .20, N ⫽ 574) represented a combination of goal commit- variables expected to exhibit differential magnitudes. Also, we
ment (Barrick, Mount, & Strauss, 1993), prior participation in omit reporting expected relationships for which we had insuffi-
development activities (Mauer et al., 2008), and self-efficacy cient data to generate estimates. The remaining estimates can be
(Colquitt et al., 2000). Additionally, we used the harmonic mean of found in Tables 4 and 5.
all the sample sizes contained in the matrix because the harmonic Proximal person-related factors. Consistent with expecta-
mean gives much less weight to large sample sizes than the tions, safety performance was strongly related to safety knowledge
arithmetic mean and is therefore a more conservative parameter (M␳ ⫽ .61) and safety motivation (M␳ ⫽ .57). Also, we expected
estimate (Viswesvaran & Ones, 1995). Although we report overall that safety knowledge would be more strongly related to compli-
fit statistics, we emphasize the magnitudes of direct and indirect ance than participation, which was not supported (M␳ ⫽ .60 for
effects when assessing model fit. compliance, M␳ ⫽ .61 for participation). Further, we expected
safety motivation to be more strongly related to participation than
compliance; however, we obtained sufficient data only for com-
Results
pliance (M␳ ⫽.44). Finally, although we expected a moderate
Descriptive Information relationship, safety knowledge was not significantly related to
safety outcomes (M␳ ⫽ ⫺.11).
For a complete description of the constructs coded from each Distal person-related factors. We expected that conscien-
study, the levels of analysis for each construct, and the source of tiousness, locus of control, and risk taking would be moderately
the ratings of each construct, please refer to Table 2. Also, Table correlated with safety performance and weakly correlated with
3 presents sample-weighted mean reliability coefficients computed safety outcomes. Expectations were partially supported for safety
at the construct level, using estimates of internal consistency performance, because safety performance was moderately related
provided by studies. to locus of control (M␳ ⫽ .35) but was weakly related to consci-
entiousness (M␳ ⫽ .18) and risk taking (M␳ ⫽ ⫺.28). We also
Predictor–Criterion Relationships expected that job attitudes would be weakly (or equivocally)
related to safety performance, which was supported (M␳ ⫽ .25).
A corrected mean correlation (i.e., M␳) is statistically significant Finally, we expected locus of control to have a stronger relation-
at the p ⬍ .05 level when its 95% confidence interval does not ship with safety participation than compliance. This was supported
include zero. Unless reported otherwise, for all mean effects re- (M␳ ⫽ .25 for compliance; M␳ ⫽ .43 for participation).
ported here, the confidence interval did not overlap zero. In addi- With regard to our expectations for safety outcomes, we again
tion, we report credibility intervals, which indicate the extent to found partial support. Conscientiousness (M␳ ⫽ ⫺.26), neuroti-
which individual correlations varied across studies for a particular cism (M␳ ⫽ .19), locus of control (M␳ ⫽ ⫺.26), and job attitudes
analysis distribution (Hunter & Schmitt, 2004). Specific informa- (M␳ ⫽ ⫺.17) were each weakly related to safety outcomes. How-
tion on these intervals and other meta-analytic findings are re- ever, extraversion (M␳ ⫽ ⫺.07) and risk taking (M␳ ⫽. 20) were
ported in Tables 4 and 5. In addition, we do not report corrected not significantly related.
predictor– criterion correlations for analyses of fewer than three Distal situation-related factors. We expected that both safety
studies. climate and leadership would have moderate relationships with
General expectations. We expected to find magnitudes of safety performance and weak relationships with safety outcomes.
relationships consistent with the conceptual model in Figure 1, These expectations were supported for overall safety climate and
which posits stronger effects for proximal factors and weaker safety performance, because safety climate was moderately related
effects for distal person-related and situation-related factors. As to safety performance at the individual level (M␳ ⫽ .49) and at the
shown in Tables 4 and 5, in general these expectations were group level (M␳ ⫽ .51). However, our expectation to find a
supported with regard to safety performance, because the two stronger relationship with safety performance between group-level
proximal factors—safety knowledge and safety motivation— climate and individual-level safety climate was not supported,
exhibited stronger effects for the safety performance composite although the effects were in the right direction. Leadership’s
(M␳ ⫽ .61 and M␳ ⫽ .57, respectively) than any of the distal relationship with safety performance (M␳ ⫽ .31) was in line with
factors (range: M␳ ⫽ .18 –.51). Conversely, our expectations were expectations.
not supported with regard to safety outcomes, because 20 of the 22 Regarding the first-order safety climate factors, we found sup-
distal factors we examined had stronger magnitudes than the M␳ ⫽ port for our expectation of moderate relationships with safety
.11 (ns) estimate we obtained for the proximal factor safety knowl- performance with the exception of individual-level perceived job
edge. risk (M␳ ⫽ ⫺.29) and individual-level work pressure (M␳ ⫽
1118 CHRISTIAN, BRADLEY, WALLACE, AND BURKE

Table 4
Results for Meta-Analysis of Person- and Situation-Related Factors With Safety Performance Composite, Safety Compliance, and
Safety Participation

95% conf. int. 80% cred. int.

Construct k N Mr SDr M␳ SEM␳ L U SD␳ L U

Person-related factors
Proximal
Safety knowledge 9 2,893 .47 .16 .61 .06 .50 .72 .16 .41 .81
Compliance 8 2,803 .46 .17 .60 .05 .50 .71 .14 .42 .79
Participation 4 1,815 .45 .11 .61 .08 .46 .76 .14 .42 .79
Safety motivation 5 1,393 .50 .24 .57 .11 .36 .78 .23 .27 .87
Compliance 4 868 .47 .15 .44 .12 .20 .68 .14 .14 .74
Distal
Conscientiousness 5 1,317 .15 .11 .18 .06 .06 .28 .10 .04 .31
Locus of control 9 2,858 .28 .16 .35 .07 .22 .48 .19 .11 .60
Compliance 4 1,685 .19 .11 .25 .08 .10 .41 .15 .06 .44
Participation 3 622 .33 .06 .43 .04 .34 .51 — — —
Risk taking 4 1,173 ⫺.23 .07 ⴚ.28 .04 ⫺.37 ⫺.19 .05 ⫺.35 ⫺.21
Participation 3 622 ⫺.19 .08 ⫺.24 .06 ⫺.36 ⫺.12 .06 ⫺.31 ⫺.16
Job attitudes 4 924 .20 .07 .25 .04 .16 .33 .04 .19 .30
Compliance 3 624 .24 .04 .30 .03 .25 .35 .00 — —

Situation-related factors
Psychological safety climate 31 15,327 .39 .18 .49 .05 .40 .58 .17 .24 .80
Compliance 18 6,783 .36 .19 .48 .07 .35 .61 .11 .28 .85
Participation 9 2,971 .45 .13 .59 .06 .47 .70 .17 .36 .81
Management commitment 12 5,823 .34 .18 .40 .05 .30 .49 .22 .19 .61
Compliance 6 1,949 .33 .13 .41 .08 .25 .57 .19 .17 .66
HRM practices 7 1,656 .31 .16 .42 .09 .24 .60 .23 .13 .71
Compliance 3 544 .40 .17 .57 .14 .29 .85 .24 .27 .88
Participation 3 1,034 .44 .18 .58 .16 .28 .89 .27 .24 .93
Safety systems 8 2,032 .31 .16 .38 .07 .25 .51 .18 .15 .60
Compliance 5 1,292 .22 .11 .27 .05 .16 .38 .10 .14 .40
Supervisor support 9 3,821 .30 .11 .38 .06 .28 .49 .16 .18 .59
Compliance 6 1,591 .32 .14 .43 .09 .26 .60 .20 .17 .68
Internal group processes 9 4,497 .32 .12 .40 .05 .31 .49 .13 .24 .56
Compliance 4 1,235 .38 .07 .48 .05 .38 .59 .09 .36 .60
Participation 3 904 .42 .15 .52 .09 .34 .69 .16 .34 .69
Perceived job risk 10 7,063 ⫺.24 .16 ⴚ.29 .06 ⫺.40 ⫺.17 .18 ⫺.52 ⫺.05
Compliance 6 2,764 ⫺.13 .07 ⫺.16 .04 ⫺.24 ⫺.08 .08 ⫺.26 ⫺.06
Work pressure 12 7,065 ⫺.11 .10 ⴚ.14 .04 ⫺.21 ⫺.07 .11 ⫺.28 .01
Compliance 7 2,771 ⫺.15 .08 ⫺.20 .04 ⫺.28 ⫺.12 .09 ⫺.31 ⫺.08
Participation 3 872 ⫺.17 .06 ⫺.22 .05 ⫺.32 ⫺.12 .04 ⫺.27 ⫺.16
Group-level safety climate 10 598 .43 .22 .51 .08 .36 .66 .24 .23 .79
Compliance 4 250 .33 .22 .40 .12 .17 .64 .21 .13 .68
Participation 3 248 .47 .25 .59 .16 .28 .90 .27 .27 .91
Management commitment 4 233 .45 .22 .51 .12 .27 .75 .22 .23 .79
Compliance 4 233 .45 .22 .52 .12 .28 .76 .22 .23 .81
Work pressure 3 182 ⫺.30 .17 ⴚ.35 .10 ⫺.55 ⫺.16 .11 ⫺.50 ⫺.21
Compliance 3 182 ⫺.32 .15 ⫺.38 .09 ⫺.56 ⫺.21 .07 ⫺.48 ⫺.29
Leadership 9 3,537 .25 .09 .31 .04 .24 .38 .11 .19 .43
Compliance 3 925 .19 .05 .24 .03 .18 .30 .00 — —
Participation 3 154 .30 .05 .35 .04 .27 .43 .00 — —

Note. k ⫽ the number of independent effect sizes included in each analysis; N ⫽ sample size (for individual-level estimates, N ⫽ number of individuals;
for group-level estimates, N ⫽ number of groups); Mr ⫽ mean uncorrected correlation; SDr ⫽ standard deviation of uncorrected correlations; M␳ ⫽ mean
corrected correlation (corrected for unreliability in the predictor and criterion); SEM␳ ⫽ standard error of M␳; 95% conf. int. ⫽ 95% confidence interval
for M␳; SD␳ ⫽ standard deviation of estimated ␳s; 80% cred. int. ⫽ 80% credibility interval; L ⫽ lower; U ⫽ upper; HRM ⫽ human resource management.
Values in bold indicate mean corrected correlations for safety performance composite.

⫺.14). Also, we expected to find stronger effects for safety climate participation, M␳ ⫽.40 for compliance). This expectation was also
and leadership with safety participation than with safety compli- supported for leadership (M␳ ⫽ .24 for compliance, M␳ ⫽ .35 for
ance. As predicted, psychological safety climate was more participation). We caution that although the magnitudes of these
strongly related to participation (M␳ ⫽ .59) than compliance relationships are in the expected directions, the respective mean
(M␳ ⫽ .48), as was group-level safety climate (M␳ ⫽ .59 for correlations had overlapping confidence intervals.
 WORKPLACE SAFETY 1119

Table 5
Results for Meta-Analysis of Person- and Situation-Related Factors With Accidents and Injuries Composite

95% conf. int. 80% cred. int.

Construct k N Mr SDr M␳ SEM␳ L U SD␳ L U

Person-related factors
Proximal
Safety knowledge 3 461 ⫺.07 .14 ⴚ.11 .11 ⫺.33 .12 .17 ⫺.32 .11
Distal
Conscientiousness 4 852 ⫺.22 .13 ⴚ.26 .07 ⫺.40 ⫺.11 .13 ⫺.42 ⫺.10
Neuroticism 12 5,129 .15 .16 .19 .06 .08 .31 .19 ⫺.06 .44
Extraversion 5 2,083 ⫺.06 .10 ⴚ.07 .05 ⫺.17 .04 .11 ⫺.20 .07
Locus of control 4 2,446 ⫺.20 .04 ⴚ.26 .03 ⫺.32 ⫺.21 .03 ⫺.30 ⫺.22
Risk taking 3 820 .16 .16 .20 .11 ⫺.02 .41 .18 ⫺.04 .43
Job attitudes 9 20,078 ⫺.13 .04 ⴚ.17 .02 ⫺.20 ⫺.13 .05 ⫺.23 ⫺.11

Situation-related factors
Psychological safety climate overall 27 27,639 ⫺.11 .07 ⴚ.14 .02 ⫺.17 ⫺.11 .07 ⫺.23 ⫺.04
Management commitment 7 3,222 ⫺.17 .05 ⴚ.21 .03 ⫺.26 ⫺.16 .04 ⫺.26 ⫺.16
HRM practices 5 3,657 ⫺.15 .04 ⴚ.19 .02 ⫺.24 ⫺.14 .03 .02 ⫺.16
Safety systems 6 17,439 ⫺.12 .03 ⴚ.16 .01 ⫺.19 ⫺.13 .03 ⫺.19 ⫺.12
Supervisor support 12 4,615 ⫺.12 .07 ⴚ.15 .03 ⫺.20 ⫺.10 .07 ⫺.24 ⫺.06
Internal group processes 8 2,839 ⫺.16 .08 ⴚ.19 .03 ⫺.25 ⫺.12 .07 ⫺.28 ⫺.10
Perceived job risk 15 5,693 .15 .13 .18 .04 .10 .26 .15 ⫺.02 .38
Work pressure 15 21,109 .06 .09 .07 .03 .01 .14 .12 ⫺.07 .22
Group-level safety climate overall 13 421 ⫺.34 .14 ⴚ.39 .05 ⫺.48 ⫺.29 ⫺.44 ⫺.33 .04
Management commitment 3 80 ⫺.33 .07 ⴚ.36 .04 ⫺.44 ⫺.27 .00 — —
HRM practices 3 129 ⫺.44 .14 ⴚ.46 .08 ⫺.62 ⫺.30 .06 ⫺.54 ⫺.38
Safety systems 3 219 ⫺.34 .12 ⴚ.38 .08 ⫺.53 ⫺.22 .10 ⫺.50 ⫺.25
Supervisor support 3 129 ⫺.21 .06 ⴚ.24 .10 ⫺.43 ⫺.05 .03 ⫺.28 ⫺.20
Perceived job risk 3 118 .12 .10 .13 .06 .02 .25 .00 — —
Work pressure 3 103 ⫺.27 .15 .33 .11 ⫺.54 ⫺.11 .05 ⫺.40 ⫺.26
Leadership 7 1,585 ⫺.14 .07 ⴚ.16 .03 ⫺.22 ⫺.10 .00 — —

Note. k ⫽ the number of independent effect sizes included in each analysis; N ⫽ sample size (for individual-level estimates, N ⫽ number of individuals;
for group-level estimates, N ⫽ number of groups); Mr ⫽ mean uncorrected correlation; SDr ⫽ standard deviation of uncorrected correlations; M␳ ⫽ mean
corrected correlation (corrected for unreliability in the predictor and criterion); SEM␳ ⫽ standard error of M␳; 95% conf. int. ⫽ 95% confidence interval
for M␳; SD␳ ⫽ standard deviation of estimated ␳s; 80% cred. int. ⫽ 80% credibility interval; L ⫽ lower; U ⫽ upper; HRM ⫽ human resource management.
Values in bold indicate mean corrected correlations.

Our expectations for safety climate and safety outcomes were mance and outcomes within our two largest predictor distributions
partially supported, because overall psychological safety climate was (i.e., individual- and group-level overall safety climate). Readers
weakly related to outcomes (M␳ ⫽ ⫺.14), as was each first-order may refer to Table 2 for a complete list of the sources of ratings for
climate factor. At the group level, we found moderate relationships for each primary study and to Table 6 for the results of our moderator
overall safety climate (M␳ ⫽ ⫺.39) and for four of the six first-order analyses. At the individual level, 92% of the safety criterion
climate factors: management commitment (M␳ ⫽ ⫺.36), human re- measures were self-reported, and 8% were archival or observer
source management practices (M␳ ⫽ ⫺.46), safety systems (M␳ ⫽ ratings. At the group level, 32% of the criterion measures were
⫺.38), and work pressure (M␳ ⫽ .33). Group-level supervisor support self-reported (aggregated to the group level), 5% were supervisor
(M␳ ⫽ ⫺.24) and perceived job risk (M␳ ⫽ .13) were weaker than rated, and 64% were archival or rated by outside observers or
expected. Finally, leadership, as expected, was weakly related to authorities. For safety performance, at the individual level we were
safety outcomes (M␳ ⫽ ⫺.16). unable to find any conclusive evidence of differential correlations
across criterion source, because the majority of effects were self-
Moderator Analyses reported. At the group level, we again did not find conclusive
To further explore our data, we conducted two sets of moderator evidence for moderation (i.e., no significant differences), although
analyses of the calculated relationships between predictors and crite- archival safety performance had a stronger relationship with safety
ria. For the sake of comprehensiveness, we present all available data climate (M␳ ⫽ .69) than did self-reported safety performance
regardless of k in Tables 6 and 7. However, here we present meta- (M␳ ⫽ .59). For safety outcomes at the individual level, psycho-
analytic results only for analyses with a k of 3 or more. logical safety climate was more strongly related to medical and
Criterion source. To examine the potential effects of common Occupational Safety and Health Administration (OSHA) records
method biases and other potential sources of error for which we of accidents and injuries (M␳ ⫽ ⫺.20) than to self-reported acci-
could not correct (i.e., reporting biases), we considered criterion dents and injuries (M␳ ⫽ ⫺.13), with the difference approaching
source as a moderator. We calculated these estimates for perfor- significance. At the group level, the same pattern emerged, with
1120 CHRISTIAN, BRADLEY, WALLACE, AND BURKE

Table 6
Results for Moderator Analyses for Safety Climate by Criterion Source

95% conf. 80% cred.

int. int.

Criterion and construct k N Mr SDr M␳ SEM␳ or SE␳ L U SD␳ L U

Safety performance
Psychological safety climate
Self-reported safety behaviors 30 14,787 .38 .17 .47 .01 .45 .49 .43 .24 .51
Archival/observer ratings 1 540 .79 — .88 .02 .85 .91 — — —
Group-level safety climate
Self-reported safety behaviors 5 317 .48 .19 .59 .05 .49 .69 .03 .55 .63
Supervisor-rated safety behavior 1 121 .27 — .35 .11 .14 .56 — — —
Archival/observer ratings 3 93 .63 .15 .69 .06 .57 .81 .65 .03 .73
Accidents/injuries
Psychological safety climate
Self-reported accidents/injuries 24 25,768 ⫺.10 .06 ⴚ.13 .01 ⫺.15 ⫺.11 ⫺.17 ⫺.09 .03
Medical records/OSHA 4 1,920 ⫺.16 .09 ⴚ.20 .03 ⫺.26 ⫺.14 ⫺.24 ⫺.16 .03
Group-level safety climate
Self-reported accidents/injuries 2 63 ⫺.19 .14 ⴚ.21 .12 ⫺.45 .03 — — —
Medical records/OSHA 11 360 ⫺.37 .14 ⴚ.42 .05 ⫺.52 ⫺.33 ⫺.46 ⫺.38 .03

Note. k ⫽ the number of independent effect sizes included in each analysis; N ⫽ sample size (for individual-level estimates, N ⫽ number of individuals;
for group-level estimates, N ⫽ number of groups); Mr ⫽ mean uncorrected correlation; SDr ⫽ standard deviation of uncorrected correlations; M␳ ⫽ mean
corrected correlation (corrected for unreliability in the predictor and criterion); SEM␳ ⫽ standard error of M␳; 95% conf. int. ⫽ 95% confidence interval
for ␳ or M␳; SD␳ ⫽ standard deviation of estimated ␳s; 80% cred. int. ⫽ 80% credibility interval; L ⫽ lower; U ⫽ upper; OSHA ⫽ Occupational Safety
and Health Administration. Values in bold indicate mean corrected correlations. Estimates of individual, disattenuated correlations were estimated with
Equation 2 of Raju, Burke, Normand, and Langlois (1991), and the standard errors for these disattenuated correlations were estimated with either Equation
5 or Equation 9 of Raju and Brand (2003), depending on the availability of sample-based reliability on the predictor or criterion measure. For use of Raju
et al.’s Equation 2 and Raju and Brand’s Equations 5 and 9, the range restriction factor was fixed at 1.0.

medical records and OSHA records of accidents and injuries significant. For safety outcomes, we found a significant difference
moderately related to climate (M␳ ⫽ ⫺.42), whereas self-reported between individual-level measures and higher levels of analysis.
outcomes were weak (M␳ ⫽ ⫺.21). Specifically, psychological safety climate (M␳ ⫽ ⫺.14) was
Level of analysis. Next, we investigated the extent to which weaker than work group safety climate (M␳ ⫽ ⫺.38) and organi-
operationalizing constructs at different levels of analysis (i.e., zational safety climate (M␳ ⫽ ⫺.39).
individual, work group, or organization level) moderates the rela-
tionship between climate and criteria. As depicted in Table 7, for Relationships Among Criteria
safety performance, individual-level measures and group-level
measures had similar magnitudes (M␳ ⫽ .49). Organizational-level We also conducted meta-analyses to determine the extent to
measures were weaker (M␳ ⫽ .38), although the difference was not which each of the criterion types included in our analyses were

Table 7
Results for Moderator Analyses for Safety Climate by Level of Analysis

95% conf. int. 80% cred. int.

Criterion and construct k N Mr SDr M␳ SEM␳ L U SD␳ L U

Safety performance
Psychological safety climate 31 15,327 .39 .18 .49 .05 .40 .58 .17 .24 .80
Work group-level safety climate 6 752 .43 .03 .49 .03 .43 .55 .18 .45 .53
Organization-level safety climate 5 247 .34 .21 .38 .06 .26 .50 .22 .34 .42
Accidents and injuries
Psychological safety climate 27 27,639 ⫺.11 .07 ⴚ.14 .02 ⫺.17 ⫺.11 .07 ⫺.23 ⫺.04
Work group-level safety climate 7 231 ⫺.33 .16 ⴚ.38 .06 ⫺.51 ⫺.26 ⫺.43 ⫺.34 .03
Organization-level safety climate 6 190 ⫺.34 .12 ⴚ.39 .07 ⫺.52 ⫺.26 ⫺.43 ⫺.35 .03

Note. k ⫽ the number of independent effect sizes included in each analysis; N ⫽ sample size (for individual-level estimates, N ⫽ number of individuals;
for group-level estimates, N ⫽ number of groups); Mr ⫽ mean uncorrected correlation; SDr ⫽ standard deviation of uncorrected correlations; M␳ ⫽ mean
corrected correlation (corrected for unreliability in the predictor and criterion); SEM␳ ⫽ standard error of M␳; 95% conf. int. ⫽ 95% confidence interval
for M␳; SD␳ ⫽ standard deviation of estimated ␳s; 80% cred. int. ⫽ 80% credibility interval; L ⫽ lower; U ⫽ upper. Values in bold indicate mean
corrected correlations. Because of nonindependence of group- and organization-level effects within some primary studies, the values in this table
may not add up to the totals in Tables 3 and 4.
 WORKPLACE SAFETY 1121

correlated. As shown in Table 8, the safety performance composite lated than more distally related variables. Together, the overall
was strongly related to safety compliance (M␳ ⫽ .63) and safety pattern of meta-analytic correlations and path-modeling results
participation (M␳ ⫽ .80). Safety participation was moderately demonstrated support for the veracity of this theoretical frame-
related to safety compliance (M␳ ⫽ .46). The accident and injuries work. In the best fitting path model, safety climate was positively
composite was more strongly correlated with the safety perfor- related to both safety knowledge and safety motivation, whereas
mance composite (M␳ ⫽ ⫺.31) than with safety participation conscientiousness was positively associated with just safety moti-
(M␳ ⫽ ⫺.15) and safety compliance (M␳ ⫽ ⫺.14). vation. Safety motivation was related to safety knowledge, and
both of these variables were positively related with safety perfor-
Exemplar Path Model mance. In turn, safety performance was correlated with accidents
Table 9 presents the meta-analyzed individual-level correlations and injuries. Although we were limited in our ability to run
among the variables in the exemplar path model. We sequentially multiple iterations of the path model or to test all variables, we
tested two nested models, inputting the harmonic mean sample size believe the model should hold for the other distal antecedents,
of 1,092. We first tested a full-mediation model in which consci- because the magnitudes of the meta-analytic correlations observed
entiousness and safety climate were exogenous and safety knowl- exhibit the patterns expected from the theoretical model. Given the
edge and safety motivation were endogenous mediators, which support for our exemplar path model and the overall pattern of
were directly related to safety performance. Safety performance meta-analytic correlations, we expect that situation-based factors
was, in turn, directly related to accidents and injuries. Although the (e.g., safety climate and leadership) and indirect person-based
path coefficients were significant, this model (Model 1) fit the data factors (e.g., job attitudes and personality) should influence safety
only moderately well, ␹2(9) ⫽ 622.5, p ⬍ .001; CFI ⫽ .68; GFI ⫽ performance behaviors indirectly by way of safety knowledge and
.86. On closer inspection of our data and modification indices, we safety motivation and that safety performance behaviors, in turn,
determined that a more accurate theoretical model would include a influence accidents and injuries. Later, we discuss more specific
path between safety motivation and safety knowledge. Theoreti- findings in relation to safety outcomes, followed by discussions of
cally, safety motivation should lead to safety knowledge acquisi- future research and practice directions.
tion. Indeed, motivation has been linked to learning outcomes and In terms of safety compliance versus safety participation, we
knowledge in many domains (see Colquitt et al., 2000). Thus, we observed that safety climate tended to be more highly related to
tested a second full mediation model (Model 2), depicted in Figure 2, safety participation than safety compliance. Because workers must
by freeing the path between safety motivation and safety knowl- by definition comply with obligatory or mandatory practices and
edge. This model showed an acceptable fit to the data, ␹2(8) ⫽ procedures, safety climate should not matter as much as for be-
313.5, p ⬍ .001; CFI ⫽ .90; GFI ⫽ .94. In addition, as shown in haviors that are compulsory. Consistent with this point, leaders are
Figure 2, all direct paths were significant ( p ⬍ .001), providing likely to have a stronger influence on workers’ safety participation
further support for the full mediation model. Although modifica- than safety compliance, which was supported in this meta-analysis.
tion indices indicated that freeing additional paths could improve In effect, the importance that leaders place on safety likely under-
overall model fit, we retained the fully mediated model because of girds the climate for safety and has a critical influence on discre-
its good fit and our desire for parsimony. Also, in Table 10, we tionary safety behaviors.
report the direct, indirect, and total effects for the relationships in Turning to safety climate and levels of analysis, we found that
Model 2. Notably, a majority of the indirect effects are moderate group and organizational safety climate generally had stronger
to large, adding further support. relationships with safety performance than psychological safety
climate. Psychological safety climate, by nature of its assessment
Discussion from the individual person’s perspective, is influenced by unique
Consistent with the theoretical framework in Figure 1, variables nuances of the person. In contrast, because group and organiza-
that are more proximally related tended to be more highly corre- tional safety climate are shared perceptions of individuals, climate

Table 8
Meta-Analysis of Relationships Between Safety Criteria

Safety compliance Safety participation Safety performance composite

SD␳ SD␳ SD␳

Variable Mr, M␳ (95% CI) (SEM␳) Mr, M␳ (95% CI) (SEM␳) Mr, M␳ (95% CI) (SEM␳)

Safety compliance — (—) — (—)

k, N — —
Safety participation .32, .46 (.26, .66) .17 (.10) — (—) — (—)
k, N 5 2,909 — —
Safety performance composite .38, .63 (.60, .66) .00 (.00) .57, .80 (.70, .88) .05 (.08) — (—) — (—)
k, N 1 1,264 3 1,448 — —
Accidents and injuries ⫺.11, ⴚ.14 (⫺.20, ⫺.08) .07 (.03) ⫺.12, ⴚ.15 (⫺.18, ⫺.13) .03 (.01) ⫺.25, ⴚ.31 (⫺.54, ⫺.31) .28 (.12)
k, N 8 1,905 4 2,004 6 1,876

Note. k ⫽ the number of independent effect sizes included in each analysis; N ⫽ sample size; Mr ⫽ mean uncorrected correlation; M␳ ⫽ mean corrected
correlation (corrected for unreliability in the predictor and criterion); 95% CI ⫽ 95% confidence interval for M␳; SD␳ ⫽ standard deviation of estimated
␳s; SEM␳ ⫽ standard error of M␳. Values in bold indicate mean corrected correlations.
1122 CHRISTIAN, BRADLEY, WALLACE, AND BURKE

at this level is likely to be more potent in that people perceive the

Note. k ⫽ the number of independent effect sizes included in each analysis; N ⫽ sample size; Mr ⫽ mean uncorrected correlation; M␳ ⫽ mean corrected correlation (corrected for unreliability in the
.13 (.07) ⫺.11, ⴚ.14 (⫺.17, ⫺.11) .07 (.02) ⫺.07, ⴚ.11 (⫺.33, .12) .17 (.11) ⫺.16, ⴚ.20 (⫺.29, ⫺.11) .04 (.05) ⫺.25, ⴚ.31 (⫺.54, ⫺.31) .28 (.12)
— (—)
(SEM␳)

predictor and criterion); 95% CI ⫽ 95% confidence interval for M␳; SD␳ ⫽ standard deviation of estimated ␳s; SEM␳ ⫽ standard error of M␳. Values in bold indicate mean corrected correlations.
1,876
SD␳

—
environment similarly and should thus be more influenced by it
than if they had divergent perceptions. These findings are consis-
Safety performance

tent with those in other domains regarding the relative influence of

isomorphic (similar) constructs at the group and individual levels
Mr, M␳ (95% CI)

of analyses (e.g., see Gully, Incalcaterra, Joshi, & Beaubien,

— (—)
2002).

6
Further, for accidents and injuries, correlations with group and
organizational safety climate were significantly larger than for
psychological climate. In comparing the current results with those
of Clarke (2006a), we found that climate (regardless of level of

Assumed values, calculated as corrected sample-weighted mean correlations derived from Barrick et al. (1993); Colquitt et al. (2000), and Maurer et al. (2008).
.23 (.11)
— (—)
(SEM␳)

analysis) was significantly correlated with safety outcomes,

1,393
SD␳

911
—
whereas Clarke’s estimate of ⫺.22 across all levels was not sig-
nificant, a difference that is likely due to the larger sample size of
Safety motivation

the current study. Also, our estimate between psychological safety

Mr, M␳ (95% CI)

.50, .57 (.36, .78)

climate and safety outcomes (⫺.14) is lower than that observed by

Clarke (⫺.22), but our estimates for group and organizational
— (—)

safety climate were much higher than Clarke’s (⫺.38 and ⫺.39,
—

respectively).
Notably, our findings did not support a possible inflationary
(common methods) bias for correlations with self-reported climate
measures and self or supervisory safety performance ratings in
.05 (.04)

.15 (.06)
— (—)
(SEM␳)

2,893

relation to correlations based on self-reported climate measures

SD␳

714

461
—

and archival and medical recordings criteria. Although the number

Safety knowledge

of primary studies prevented conclusive findings, our results re-

.60, .66 (.58, .74)

.47, .61 (.50, .72)

vealed that common methods bias may not be a major concern in

Mr, M␳ (95% CI)

the safety domain. If bias does exist, the pattern of our findings
— (—)

would suggest that self-reports for certain types of safety criteria

—

(e.g., safety performance or accidents/injury reports) may yield

slightly downward biased underestimates of relationships involv-
ing climate as a predictor. With regard to self-reports of accidents
and injuries, the slightly lower relationships may reflect artifactual
.08 (.05)

.05 (.04)

.17 (.05)
— (—)
(SEM␳)

15,327

27,639

or method-induced restriction in range of scores due to underre-

SD␳

987

849
—

porting. However, we cautiously allude to this possibility given

Meta-Analysis of Relationships Between Variables in Exemplar Path Model

that the latter relationships, as evidenced by the overlapping con-

Safety climate

fidence intervals, were not significantly different from each other.

Mr, M␳ (95% CI)

.43, .49 (.38, .59)

.42, .46 (.38, .54)

.39, .49 (.40, .58)

In general, the manner in which many variables are measured in

— (—)

the domain of workplace safety may preclude concerns that the

—

31

27
3

measurement method greatly distorts relationships between con-

structs. This point is reinforced by the work of Burke, Sarpy, et al.
(2002) who found minimal differences among self-, coworker, and
supervisory ratings of safety performance behaviors and by the
fact that the present safety performance reliabilities are uniformly
.00 (.03)

.12 (.06)
— (—)

— (—)

— (—)
(SEM␳)

1,908

1,421
SD␳

971

574

852
—

higher than average criterion reliabilities reported in the applied

psychology literature for coworker or supervisory ratings.
Conscientiousness

⫺.22, ⴚ.26 (⫺.40, ⫺.11)

Mr, M␳ (95% CI)

.10, .11 (.06, .16)

.17, .19 (.08, .31)

Practice Implications, Future Research Directions, and

.00a (—, —)

.21ⴱ (—, —)

Potential Limitations
— (—)
—

10
5

The results of this meta-analysis suggest that both the person

and the situation are important factors related to workplace safety.
Workers can be selected, trained, and supported through positive
safety climate to maximize safety motivation and safety knowl-
Safety performance

edge, which in turn leads to safe behaviors and fewer accidents and
Conscientiousness

Safety knowledge

Safety motivation

injuries. Also, our findings regarding particular individual differ-

Safety climate

Accidents and
Variable

injuries

ences suggest where to focus personnel assessments (e.g., on

conscientiousness in personnel selection contexts). In addition, our
Table 9

k, N

k, N

k, N

k, N

k, N

k, N

findings with regard to specific climate dimensions suggest key

intervention points related to enhancing workplace safety. For
a
 WORKPLACE SAFETY 1123

Figure 2. Maximum-likelihood parameter estimates for the hypothesized model. Statistics are standardized
path coefficients. ⴱp ⬍ .001.

instance, interventions focused on improving management com- uninjured. Thus, by recognizing that injuries are less common than
mitment to safety may meaningfully enhance safety performance accidents, future research could investigate how situational factors
and reduce accidents. might moderate individual difference (predictor) relationships with
A possible future research avenue might be to examine how accidents and injury criteria. For example, workers low (rather
person and situation factors interact to influence safety. One way than high) in conscientiousness might be more likely to acciden-
to approach the issue of person–situation interactions is with tally spill a noxious chemical (i.e., accident) but may be no more
Schneider’s (1987) attraction–selection–attrition model, which likely to be injured by the spill if the organization requires the use
suggests that individuals are differentially attracted to, selected to, of protective clothing. Along this line, we encourage future re-
and retained within different work environments on the basis of search that examines microaccidents, or accidents requiring only
their values, personality, and other individual differences. For basic first aid treatment (Zohar, 2000, 2002b). We refer the reader
example, thrill-seeking people may be more likely to seek out to Wallace and Chen (2006) and Zohar (2000, 2002b), who have
high-risk jobs. To the extent that risk-seeking individuals congre- highlighted methodological advantages of studying microaccidents
gate in riskier environments, the organizational climate may be- relative to accidents.
come socially constructed to lead to riskier decisions and actions. As with any meta-analysis, the current findings are limited by
Furthermore, although researchers have made progress in defin- the primary studies used. In general, the early literature was often
ing and capturing safety performance behaviors (Burke, Sarpy, et difficult to integrate, in large part because many early studies
al., 2002; Marchand, Simard, Carpentier-Roy, & Ouellet, 1998), failed to provide statistical information (e.g., effect sizes, sample
they have not similarly progressed on outcome criteria. Most sizes) required for a meta-analysis. Our inability to analyze such
studies examining workplace accidents have operationalized acci- studies represents a research opportunity in that meta-analytic
dents in terms of the number of recordable accidents as defined by distributions with small numbers of effects are ones where more
OSHA, meaning those that require more than simple first aid primary research is clearly needed. Obtaining more stable param-
treatment (e.g., Hofmann & Stetzer, 1996), or as lost work days eter estimates for some meta-analytic distributions will greatly
resulting from an injury (e.g., Zohar, 2000). Clearly, when an contribute to researchers’ ability to overcome issues with the use
injury has occurred, an accident has also taken place. However, the of mean corrected effects in path analyses. Additionally, the fact
converse is not true; for example, a worker could fail to stabilize that few of the primary studies (12 out of 90) were longitudinal
a ladder and suffer a fall (accident) but be fortunate enough to go field studies limits our ability to make causal statements. For
example, although reverse causality can be ruled out for distal
traits’ relationships with outcomes, the possibility remains that
Table 10 more proximal states, like safety motivation, have a reciprocal or
Direct, Indirect, and Total Effects of Person and Situation- reverse-causal relationship with safety performance. Furthermore,
Related Factors for Safety Performance and Safety Outcomes the safety literature on the whole could do better to develop
Direct Indirect Total
stronger theoretical rationales and more rigorous research designs
Model effects effects effects to control potentially spurious or third variable effects that could
explain some of the relationships presented herein. Hence, we
Safety performance suggest future research is needed to further the understanding of
Safety climate — .33 .33
occupational safety, particularly with an emphasis on theoretically
Conscientiousness — .09 .09
Safety motivation .30 .23 .52 driven longitudinal research designs.
Safety knowledge .41 — .41
Accidents and injuries
Safety climate — ⫺.10 ⫺.10 References
Conscientiousness — ⫺.03 ⫺.03
Safety motivation — ⫺.16 ⫺.16 References marked with an asterisk indicate studies included in the
Safety knowledge — ⫺.13 ⫺.13 meta-analysis.
Safety performance ⫺.31 — ⫺.31
*Barling, J., Kelloway, E. K., & Iverson, R. D. (2003). High-quality work,
Note. All computations were conducted by inputting the harmonic mean job satisfaction, and occupational injuries. Journal of Applied Psychol-
for the sample size (Nh ⫽ 1,092). ogy, 88, 276 –283.
1124 CHRISTIAN, BRADLEY, WALLACE, AND BURKE

*Barling, J., Loughlin, C., & Kelloway, E. K. (2002). Development and performance: A meta-analytic review. Journal of Occupational Health
test of a model linking safety-specific transformational leadership and Psychology, 11, 315–327.
occupational safety. Journal of Applied Psychology, 87, 488 – 496. *Clarke, S. (2006b). Safety climate in an automobile manufacturing plant:
Barrick, M. R., Mount, M. K., & Strauss, J. P. (1993). Conscientiousness The effects of work environment, job communication and safety atti-
and performance of sales representatives: Test of the mediating effects tudes on accidents and unsafe behavior. Personnel Review, 35, 413– 430.
of goal setting. Journal of Applied Psychology, 78, 715–722. Clarke, S., & Robertson, I. T. (2005). A meta-analytic review of the Big
Bergman, M. E., Donovan, M. A., Drasgow, F., Overton, R. C., & Hen- Five personality factors and accident involvement in occupational and
ning, J. B. (2008). Test of Motowidlo et al.’s (1997) theory of individual non-occupational settings. Journal of Occupational and Organizational
differences in task and contextual performance. Human Performance, Psychology, 78, 355–376.
21, 227–253. *Clarke, S., & Ward, K. (2006). The role of leader influence tactics and
Borman, W. C., & Motowidlo, S. J. (1993). Expanding the criterion safety climate in engaging employee’s safety participation. Risk Analy-
domain to include elements of contextual performance. In N. Schmitt & sis, 26, 1175–1185.
W. C. Borman (Eds.), Personnel selection in organizations (pp. 71–98). Cohen, J. (1988). Statistical power analysis for the behavioral sciences
San Francisco: Jossey-Bass. (2nd ed.). Hillsdale, NJ: Erlbaum.
*Borofsky, G. L., & Smith, M. (1993). Reductions in turnover, accidents, Colquitt, J. A., LePine, J. A., & Noe, R. A. (2000). Toward an integrative
and absenteeism: The contribution of a pre-employment screening in- theory of training motivation: A meta-analytic path analysis of 20 years
ventory. Journal of Clinical Psychology, 49, 109 –116. of research. Journal of Applied Psychology, 85, 678 –707.
*Brown, K. A., Willis, P. G., & Prussia, G. E. (2000). Predicting safe *Cooper, M. D., & Phillips, R. A. (2004). Exploratory analysis of the
employee behavior in the steel industry: Development and test of a safety climate and safety behavior relationship. Journal of Safety Re-
sociotechnical model. Journal of Operations Management, 18, 445– 465. search, 35, 497–512.
Bureau of Labor Statistics, U.S. Department of Labor. (2007). Industry Cooper, M. D., Phillips, R. A., Sutherland, V. J., & Makin, P. J. (1994).
injury and illness data. Retrieved April 24, 2009, from http:// Reducing accidents using goal setting and feedback: A field study.
www.bls.gov/iif/oshsum.htm#06Summary%20News%20Release Journal of Occupational and Organizational Psychology, 67, 219 –240.
Burke, M. J., Borucki, C., & Hurley, A. E. (1992). Reconceptualizing Costa, P. T., Jr., & McCrae, R. R. (1985). The NEO Personality Inventory
psychological climate in a retail service environment: A multiple stake- manual. Odessa, FL: Psychological Assessment Resources.
holder perspective. Journal of Applied Psychology, 77, 717–729. *Davids, A., & Mahoney, J. T. (1957). Personality dynamics and accident
Burke, M. J., Borucki, C., & Kaufman, J. D. (2002). Contemporary proneness in an industrial setting. Journal of Applied Psychology, 41,
perspectives on the study of psychological climate. European Journal of 303–309.
Work and Organizational Psychology, 11, 325–340. *DeJoy, D. M., Schaffer, B. S., Wilson, M. G., Vandenberg, R. J., & Butts,
*Burke, M. J., Chan-Serafin, S., Salvador, R., Smith, A., & Sarpy, S. M. M. (2004). Creating safer workplaces: Assessing the determinants
(2008). The role of national culture and organizational climate in safety and role of safety climate. Journal of Safety Research, 35, 81–90.
training effectiveness. European Journal of Work and Organizational *Donald, I., & Canter, D. (1994). Employee attitudes and safety in the
Psychology, 17, 133–152. chemical industry. Journal of Loss Prevention in the Process Industries,
Burke, M. J., Holman, D., & Birdi, K. (2006). A walk on the safe side: The 7, 203–208.
implications of learning theory for developing effective safety and health *Dunbar, E. (1993). The role of psychological stress and prior experience
training. In G. P. Hodgkinson & J. K. Ford (Eds.), International review in the use of personal protective equipment. Journal of Safety Research,
of industrial and organizational psychology (pp. 1– 44). Chichester, 24, 181–187.
England: Wiley. *Eklöf, M. (2002). Perception and control of occupational injury risks in
Burke, M. J., & Landis, R. S. (2003). Methodological and conceptual fishery—A pilot study. Work & Stress, 16, 58 – 69.
challenges in conducting and interpreting meta-analyses. In K. Murphy *Eklöf, M., & Törner, M. (2005). Participatory analysis of accidents and
(Ed.), Validity generalization: A critical review (pp. 287–309). Mahwah, incidents as a tool for increasing safety behaviour in fishermen—A pilot
NJ: Erlbaum. intervention study. Work & Stress, 19, 360 –369.
*Burke, M. J., Sarpy, S. A., Tesluk, P. E., & Smith-Crowe, K. (2002). Erez, A., Bloom, M. C., & Wells, M. T. (1996). Using random rather than
General safety performance: A test of a grounded theoretical model. fixed effects models in meta-analysis: Implications for situational spec-
Personnel Psychology, 55, 429 – 457. ificity and validity generalization. Personnel Psychology, 49, 275–306.
*Burt, C. D., Sepie, B., & McFadden, G. (2008). The development of Eysenck, H. J. (1967). The biological basis of personality. Springfield, IL:
considerate and responsible safety attitudes in work teams. Safety Sci- Charles C Thomas.
ence, 46, 79 –91. *Fallon, J. D., Avis, J. M., Kudisch, J. D., Gornet, T. P., & Frost, A. (2000).
Campbell, J. P., McCloy, R. A., Oppler, S. H., & Sager, C. E. (1993). A Conscientiousness as a predictor of productive and counterproductive
theory of performance. In N. Schmitt & W. C. Borman (Eds.), Personnel behaviors. Journal of Business and Psychology, 15, 339 –349.
selection in organizations (pp. 35–70). San Francisco: Jossey-Bass. Fazio, R. H., & Williams, C. J. (1986). Attitude accessibility as a moder-
*Cellar, D. F., Nelson, Z. C., Yorke, C. M., & Bauer, C. (2001). The ator of the attitude–perception and attitude– behavior relations: An in-
five-factor model and safety in the workplace: Investigating the rela- vestigation of the 1984 presidential election. Journal of Personality and
tionships between personality and accident involvement. Journal of Social Psychology, 51, 505–514.
Prevention & Intervention in the Community, 22, 43–52. *Fellner, D. J., & Sulzer-Azaroff, B. (1984). Increasing industrial safety
Chen, G., Casper, W. J., & Cortina, J. M. (2001). The roles of self-efficacy practices and conditions through posted feedback. Journal of Safety
and task complexity in the relationships among cognitive ability, con- Research, 15, 7–21.
scientiousness, and work-related performance: A meta-analytic exami- Flin, R., Mearns, K., O’Connor, P., & Bryden, R. (2000). Measuring safety
nation. Human Performance, 14, 209 –230. climate: Identifying the common features. Safety Science, 34, 177–192.
Chen, G., Gully, S. M., Whiteman, J. A., & Kilcullen, R. N. (2000). *Frone, M. R. (1998). Predictors of work injuries among employed ado-
Examination of relationships among trait-like individual differences, lescents. Journal of Applied Psychology, 83, 565–576.
state-like individual differences, and learning performance. Journal of *Fullarton, C., & Stokes, M. (2007). The utility of a workplace instrument
Applied Psychology, 85, 835– 847. in prediction of workplace injury. Accident Analysis and Prevention, 39,
Clarke, S. (2006a). The relationship between safety climate and safety 28 –37.
 WORKPLACE SAFETY 1125

Furnham, A., Petrides, K. V., Jackson, C. J., & Cotter, T. (2002). Do Hunter, J. E., & Schmidt, F. L. (2004). Methods of meta-analysis: Cor-
personality factors predict job satisfaction? Personality and Individual recting for error and bias in research findings. Newbury Park, CA:
Differences, 33, 1325–1342. Sage.
*Geller, E. S., Roberts, D. S., & Gilmore, M. R. (1996). Predicting Iversen, H., & Rundmo, T. (2002). Personality, risky driving and accident
propensity to actively care for occupational safety. Journal of Safety involvement among Norwegian drivers. Personality and Individual Dif-
Research, 27, 1– 8. ferences, 33, 1251–1263.
Glick, W. (1985). Conceptualizing and measuring dimensions of organi- *Iverson, R. D., & Erwin, P. J. (1997). Predicting occupational injury: The
zational and psychological climate: Pitfalls in multi-level research. role of affectivity. Journal of Occupational and Organizational Psychol-
Academy of Management Review, 10, 601– 616. ogy, 70, 113–128.
*Goldenhar, L. M., Williams, L. J., & Swanson, N. G. (2003). Modeling James, L. A., & James, L. R. (1989). Integrating work environment
relationships between job stressors and injury and near-miss outcomes perceptions: Explorations into the measurement of meaning. Journal of
for construction labourers. Work & Stress, 17, 218 –240. Applied Psychology, 74, 739 –751.
Golimbet, V., Alfimova, M., Gritsenko, I., & Ebstein, R. (2007). Relation- James, L. R., Choi, C. C., Ko, C. E., McNeil, P. K., Minton, M. K., &
ship between dopamine system genes and extraversion and novelty Wright, M. A. (2008). Organizational and psychological climate: A
seeking. Neuroscience and Behavioral Physiology, 37, 601– 606. review of theory and research. European Journal of Work and Organi-
Griffin, M. A., Burley, N., & Neal, A. (2000, August). The impact of zational Psychology, 17, 5–32.
supportive leadership and conscientiousness on safety at work. Paper James, L. R., Hater, J. J., Gent, M. J., & Bruni, J. R. (1978). Psychological
presented at the 25th annual meeting of the Academy of Management, climate: Implications from cognitive social learning theory and interac-
Toronto. tional psychology. Personnel Psychology, 31, 781– 813.
*Griffin, M. A., & Neal, A. (2000). Perceptions of safety at work: A James, L. R., James, L. A., & Ashe, D. K. (1990). The meaning of
framework for linking safety climate to safety performance, knowledge, organizations: The role of cognitions and values. In B. Schneider (Ed.),
and motivation. Journal of Occupational Health Psychology, 3, 347– Organizational climate and culture (pp. 40 – 84). San Francisco: Jossey-
358. Bass.
Gully, S., Incalcaterra, K., Joshi, A., & Beaubien, J. (2002). A meta- James, L. R., & Sells, S. B. (1981). Psychological climate: Theoretical
analysis of team-efficacy, potency, and performance: Interdependence perspectives and empirical research. In D. Magnusson (Ed.), Toward a
and level of analysis as moderators of observed relationships. Journal of psychology of situations. An international perspective (pp. 275–295).
Applied Psychology, 87, 819 – 832. Hillsdale, NJ: Erlbaum.
*Hansen, C. P. (1989). A causal model of the relationship among accidents, *Johnson, S. E. (2007). The predictive validity of safety climate. Journal
biodata, personality, and cognitive factors. Journal of Applied Psychol- of Safety Research, 38, 511–521.
ogy, 74, 81–90. *Jones, J. W., & Wuebker, L. J. (1993). Safety locus of control and
*Harrell, W. A. (1990). Perceived risk of occupational injury: Control over employees’ accidents. Journal of Business and Psychology, 7, 449 – 457.
pace of work and blue-collar versus white-collar work. Perceptual and Judge, T. A. (1993). Does affective disposition moderate the relationship
Motor Skills, 70, 1351–1359. between job satisfaction and voluntary turnover? Journal of Applied
*Hayes, B. E., Perander, J., Smecko, T., & Trask, J. (1998). Measuring Psychology, 78, 395– 401.
perceptions of workplace safety: Development and validation of the Judge, T. A., & Bono, J. E. (2001). Relationship of core self-evaluations
work safety scale. Journal of Safety Research, 29, 145–161. traits—self-esteem, generalized self-efficacy, locus of control, and emo-
*Hemingway, M. A., & Smith, C. S. (1999). Organizational climate and tional stability—with job satisfaction and job performance: A meta-
occupational stressors as predictors of withdrawal behaviors and injuries analysis. Journal of Applied Psychology, 86, 80 –92.
in nurses. Journal of Occupational and Organizational Psychology, 72, Judge, T. A., Erez, A., Bono, J. E., & Thoresen, C. J. (2002). Are measures
285–299. of self-esteem, neuroticism, locus of control, and generalized self-
Hobfoll, S. E. (1989). Conservation of resources: A new attempt at con- efficacy indicators of a common core construct? Journal of Personality
ceptualizing stress. American Psychologist, 44, 513–524. and Social Psychology, 83, 693–710.
*Hofmann, D. A., & Morgeson, F. P. (1999). Safety-related behavior as a Judge, T. A., Thoresen, C. J., Bono, J. E., & Patton, G. K. (2001). The job
social exchange: The role of perceived organizational support and satisfaction–job performance relationship: A qualitative and quantitative
leader–member exchange. Journal of Applied Psychology, 84, 286 –296. review. Psychological Bulletin, 127, 376 – 407.
Hofmann, D. A., Morgeson, F. P., & Gerras, S. J. (2003). Climate as a Kanfer, R. (1990). Motivation theory and industrial/organizational psy-
moderator of the relationship between leader–member exchange and chology. In M. D. Dunnette & L. M. Hough (Eds.), Handbook of
content specific citizenship: Safety climate as an exemplar. Journal of industrial and organizational psychology (2nd ed., Vol. 1, pp. 75–170).
Applied Psychology, 88, 170 –178. Palo Alto, CA: Consulting Psychologists Press.
*Hofmann, D. A., & Stetzer, A. (1996). A cross-level investigation of Kanfer, R. (1992). Work motivation: New directions in theory and re-
factors influencing unsafe behaviors and accidents. Personnel Psychol- search. In C. L. Cooper & I. T. Robertson (Eds.), International review of
ogy, 49, 307–339. industrial and organizational psychology (Vol. 7, pp. 1–53). Chichester,
*Hofmann, D. A., & Stetzer, A. (1998). The role of safety climate and England: Wiley.
communication in accident interpretation: Implications for learning from Kanfer, R., & Ackerman, P. L. (1989). Motivation and cognitive abilities:
negative events. Academy of Management Journal, 41, 644 – 657. An integrative/aptitude–treatment interaction approach to skill acquisi-
Hough, L. M. (1992). The “Big Five” personality variables– construct tion. Journal of Applied Psychology, 74, 657– 689.
confusion: Description versus prediction. Human Performance, 5, 139 – *Krause, T. R., Seymour, K. J., & Sloat, K. C. M. (1999). Long-term
155. evaluation of a behavior-based method for improving safety perfor-
*Hsu, S. H., Lee, C. C., Wu, M. C., & Takano, K. (2008). A cross-cultural mance: A meta-analysis of 73 interrupted time-series replications. Safety
study of organizational factors on safety: Japanese v. Taiwanese oil Science, 32, 1–18.
refinery plants. Accident Analysis and Prevention, 40, 24 –34. Lajunen, T. (2001). Personality and accident liability: Are extraversion,
*Huang, Y. H., Ho, M., Smith, G. S., & Chen, P. Y. (2006). Safety climate neuroticism and psychoticism related to traffic and occupational fatali-
and self-reported injury: Assessing the mediating role of employee ties? Personality and Individual Differences, 31, 1365–1373.
safety control. Accident Analysis and Prevention, 38, 425– 433. *Liao, H., Arvey, R. D., Butler, R. J., & Nutting, M. (2001). Correlates of
1126 CHRISTIAN, BRADLEY, WALLACE, AND BURKE

work injury frequency and duration among firefighters. Journal of Sonnentag (Ed.), Psychological management of individual performance
Occupational Health Psychology, 6, 229 –242. (pp. 69 –93). Chichester, England: Wiley.
*Lilley, R., Feyer, A. M., Kirk, P., & Gander, P. (2002). A survey of forest *Paul, P. S., & Maiti, J. (2007). The role of behavioral factors on safety
workers in New Zealand: Do hours of work, rest, and recovery play a management in underground mines. Safety Science, 45, 449 – 471.
role in accidents and injury? Journal of Safety Research, 33, 53–71. Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of
*Lingard, H. (2002). The effect of first aid training on Australian construc- persuasion. Advances in Experimental Social Psychology, 19, 123–205.
tion workers’ occupational health and safety motivation and risk control *Probst, T. M. (2004). Safety and insecurity: Exploring the moderating
behavior. Journal of Safety Research, 33, 209 –230. effect of organizational safety climate. Journal of Occupational Health
*Maierhofer, N. I., Griffin, M. A., & Sheehan, M. (2000). Linking manager Psychology, 9, 3–10.
values and behavior with employee values and behavior: A study of *Probst, T. M., & Brubaker, T. L. (2001). The effects of job insecurity on
values and safety in the hairdressing industry. Journal of Occupational employee safety outcomes: Cross-sectional and longitudinal explora-
Health Psychology, 5, 417– 427. tions. Journal of Occupational Health Psychology, 6, 139 –159.
Marchand, A., Simard, M., Carpentier-Roy, M. C., & Ouellet, F. (1998). *Prussia, G. E., Brown, K. A., & Willis, P. G. (2003). Mental models of
From a unidimensional to a bi-dimensional concept and measurement of safety: Do managers and employees see eye to eye? Journal of Safety
workers’ safety behavior. Scandinavian Journal of Work, Environment Research, 34, 143–156.
& Health, 24, 293–299. Raju, N. S., & Brand, P. A. (2003). Determining the significance of
Mathews, A., & MacLeod, C. (1985). Selective processing of threat cues in correlations corrected for unreliability and range restriction. Applied
anxiety states. Behavioural Research and Therapy, 23, 563–569. Psychological Measurement, 27, 52–71.
Mauer, T. J., Lippstreu, M., & Judge, T. A. (2008). Structural model of Raju, N. S., Burke, M. J., Normand, J., & Langlois, G. M. (1991). A new
employee involvement in skill development activity: The role of indi- meta-analytic approach. Journal of Applied Psychology, 76, 432– 446.
vidual differences. Journal of Vocational Behavior, 72, 336 –350. Raju, N. S., Lezotte, D. V., Fearing, B. K., & Oshima, T. C. (2006). A note
*McLain, D. L., & Jarrell, K. A. (2007). The perceived compatibility of on correlations corrected for unreliability and range restriction. Applied
safety and production expectations in hazardous occupations. Journal of Psychological Measurement, 30, 145–149.
Safety Research, 38, 299 –309. *Real, K. (2008). Information seeking and workplace safety: A field
*Mearns, K., Flin, R., Gordon, R., & Fleming, M. (2001). Human and application of the risk perception attitude framework. Journal of Applied
organizational factors in offshore safety. Work & Stress, 15, 144 –160. Communication Research, 36, 339 –359.
*Mearns, K., Whitaker, S. M., & Flin, R. (2003). Safety climate, safety *Ringenbach, K. L., & Jacobs, R. R. (1995). Injuries and aging workers.
management practice and safety performance in offshore environments. Journal of Safety Research, 26, 169 –176.
Safety Science, 41, 641– 680. *Rundmo, T. (1994). Associations between safety and contingency mea-
*Melamed, S., & Oksenberg, A. (2002). Excessive daytime sleepiness and sures and occupational accidents on offshore petroleum platforms. Scan-
risk of occupational injuries in non-shift daytime workers. Sleep, 25, dinavian Journal of Work, Environment & Health, 20, 128 –131.
315–322. *Rundmo, T. (1996). Associations between risk perception and safety.
*Michael, J. H., Evans, D. D., Jansen, K. J., & Haight, J. M. (2005). Safety Science, 24, 197–209.
Management commitment to safety as organizational support: Relation- *Rundmo, T. (2000). Safety climate, attitudes and risk perception in Norsk
ships with non-safety outcomes in wood manufacturing employees. Hydro. Safety Science, 34, 47–59.
Journal of Safety Research, 36, 171–179. *Rundmo, T. (2001). Employee images of risk. Journal of Risk Research,
*Michael, J. H., Guo, Z. G., Wiedenbeck, J. K., & Ray, C. D. (2006). 4, 393– 404.
Production supervisor impacts on subordinates’ safety outcomes: An *Saari, J. T., & Lahtela, J. (1979). Characteristics of jobs in high and low
investigation of leader–member exchange and safety communication. accident frequency companies in the light metal working industry.
Journal of Safety Research, 37, 469 – 477. Accident Analysis and Prevention, 11, 51– 60.
*Mohamed, S. (1999). Empirical investigation of construction safety man- *Salminen, S., & Klen, T. (1994). Accident locus of control and risk taking
agement activities and performance in Australia. Safety Science, 33, among forestry and construction workers. Perceptual and Motor Skills,
129 –142. 78, 852– 854.
Neal, A., & Griffin, M. A. (2004). Safety climate and safety at work. In J. *Salminen, S., Klen, T., & Ojanen, K. (1999). Risk taking and accident
Barling & M. R. Frone (Eds.), The psychology of workplace safety (pp. frequency among Finnish forestry workers. Safety Science, 33, 143–153.
15–34). Washington, DC: American Psychological Association. *Salminen, S., & Tallberg, T. (1996). Human errors in fatal and serious
*Neal, A., & Griffin, M. A. (2006). A study of the lagged relationships occupational accidents in Finland. Ergonomics, 39, 980 –988.
among safety climate, safety motivation, and accidents at the individual Schneider, B. (1987). The people make the place. Personnel Psychology,
and group levels. Journal of Applied Psychology, 91, 946 –953. 40, 437– 453.
*Neal, A., Griffin, M. A., & Hart, P. M. (2000). The impact of organiza- *Seo, D., Torabi, M. R., Blair, E. H., & Ellis, N. T. (2004). A cross-
tional climate on safety climate and individual behavior. Safety Science, validation of safety climate scale using confirmatory factor analytic
34, 99 –109. approach. Journal of Safety Research, 35, 427– 445.
Nicholson, N., Soane, E., Fenton-O’Creevy, M., & Willman, P. (2005). *Simard, M., & Marchand, A. (1994). The behaviour of first-line super-
Personality and domain-specific risk taking. Journal of Risk Research, 8, visors in accident prevention and effectiveness in occupational safety.
157–176. Safety Science, 17, 169 –185.
Ostroff, C., & Harrison, D. A. (1999). Meta-analysis, level of analysis, and *Siu, O., Phillips, D. R., & Leung, T. W. (2003). Age differences in safety
best estimates of population correlations: Cautions for interpreting meta- attitudes and safety performance in Hong Kong construction workers.
analytic results in organizational behavior. Journal of Applied Psychol- Journal of Safety Research, 34, 199 –205.
ogy, 84, 260 –270. *Siu, O., Phillips, D. R., & Leung, T. W. (2004). Safety climate and safety
Overton, R. C. (1998). A comparison of fixed-effects and mixed (random- performance among construction workers in Hong Kong: The role of
effects) models for meta-analysis tests of moderator variable effects. psychological strains as mediators. Accident Analysis and Prevention,
Psychological Methods, 3, 354 –379. 36, 359 –366.
Parker, S. K., & Turner, N. (2002). Work design and individual work *Smith, G. S., Huang, Y. H., Ho, M., & Chen, P. Y. (2006). The relation-
performance: Research findings and an agenda for future inquiry. In S. ship between safety climate and injury rates across industries: The need
 WORKPLACE SAFETY 1127

to adjust for injury hazards. Accident Analysis and Prevention, 38, climate, self-regulation, and performance. Personnel Psychology, 59,
556 –562. 529 –557.
*Snyder, L. A., Krauss, A. D., Chen, P. Y., Finlinson, S., & Huang, Y. H. *Wallace, J. C., & Vodanovich, S. J. (2003a). Can accidents and industrial
(2008). Occupational safety: Application of the job demand– control– mishaps be predicted? Further investigation into the relationship be-
support model. Accident Analysis and Prevention, 40, 1713–1723. tween cognitive failure and reports of accidents. Journal of Business and
*Stephenson, M. T., Witte, K., Vaught, C., Quick, B. L., Patel, D., & Psychology, 17, 503–514.
Zuckerman, C. (2005). Using persuasive messages to encourage volun- *Wallace, J. C., & Vodanovich, S. J. (2003b). Workplace safety perfor-
tary hearing protection among coal miners. Journal of Safety Research, mance: Conscientiousness, cognitive failure, and their interaction. Jour-
36, 9 –17. nal of Occupational Health Psychology, 8, 316 –327.
*Storeth, F. (2007). Affective job insecurity and risk taking at work. *Watson, G. W., Scott, D., Bishop, J., & Turnbeaugh, T. (2005). Dimen-
International Journal of Risk Assessment and Management, 7, 189 –204. sions of interpersonal relationships and safety in the steel industry.
Sulzer-Azaroff, B., Loafman, B., Merante, R. J., & Hlavacek, A. C. (1990). Journal of Business and Psychology, 19, 303–318.
Improving occupational safety in a large industrial plant: A systematic *Wogalter, M. S., Young, S. L., Brelsford, J. W., & Barlow, T. (1999). The
replication. Journal of Organizational Behavior Management, 11, 99 – relative contributions of injury severity and likelihood information on
120. hazard-risk judgments and warning compliance. Journal of Safety Re-
*Sutherland, V. J., & Cooper, C. L. (1991). Personality, stress and accident search, 30, 151–162.
involvement in the offshore oil and gas industry. Personality and Indi- *Wu, T. C., Chen, C. H., & Li, C. C. (2008). A correlation among safety
vidual Differences, 12, 195–204. leadership, safety climate, and safety performance. Journal of Loss
*Thoms, P., & Venkataraman, R. R. (2002). Relation of mangers’ person- Prevention in the Process Industries, 21, 307–318.
ality to accident and injury rates. Psychological Reports, 91, 1107–1115. *Wuebker, L. J. (1986). Safety locus of control as a predictor of industrial
*Torp, S., & Moen, B. E. (2006). The effects of occupational health and accidents and injuries. Journal of Business and Psychology, 1, 19 –30.
safety management on work environment and health: A prospective *Zacharatos, A., Barling, J., & Iverson, D. I. (2005). High performance
study. Applied Ergonomics, 37, 775–783. work systems and occupational safety. Journal of Applied Psychology,
*Trimpop, R., Kirkcaldy, B., Athanasou, J., & Cooper, C. (2000). Individ- 90, 77–93.
ual differences in working hours, work perceptions and accident rates in *Zohar, D. (2000). A group-level model of safety climate: Testing the
veterinary surgeries. Work & Stress, 14, 181–188. effect of group climate on microaccidents in manufacturing jobs. Jour-
Tuncel, S., Lotlikar, H., Salem, S., & Daraiseh, N. (2006). Effectiveness of nal of Applied Psychology, 85, 587–596.
behaviour based safety interventions to reduce accidents and injuries in *Zohar, D. (2002a). The effects of leadership dimensions, safety climate,
workplaces: Critical appraisal and meta-analysis. Theoretical Issues in and assigned priorities on minor injuries in work groups. Journal of
Ergonomics Science, 7, 191–209. Organizational Behavior, 23, 75–92.
*Turner, N., Chmiel, N., & Walls, M. (2005). Railing for safety: Job *Zohar, D. (2002b). Modifying supervisory practices to improve subunit
demands, job control, and safety citizenship role definition. Journal of safety: A leadership-based intervention model. Journal of Applied Psy-
Occupational Health Psychology, 10, 504 –512. chology, 87, 156 –163.
*Varonen, U., & Mattila, M. (2000). The safety climate and its relationship Zohar, D., & Luria, G. (2003). The use of supervisory practices as leverage
to safety practices, safety of the work environment and occupational to improve safety behavior: A cross-level intervention model. Journal of
accidents in eight wood-processing companies. Accident Analysis and Safety Research, 34, 567–577.
Prevention, 32, 761–769. *Zohar, D., & Luria, G. (2004). Climate as a social– cognitive construction
*Varonen, U., & Mattila, M. (2002). Effects of the work environment and of supervisory safety practices: Scripts as proxy of behavior patterns.
safety activities on occupational accidents in eight wood-processing Journal of Applied Psychology, 89, 322–333.
companies. Human Factors and Ergonomics in Manufacturing, 12, 1–5. *Zohar, D., & Luria, G. (2005). A multilevel model of safety climate:
Visser, E., Ysbrand, J. P., Stolk, R. P., Neeleman, J., & Rosmalen, J. G. M. Cross level relationships between organization and group-level climates.
(2007). Accident proneness, does it exist? A review and meta-analysis. Journal of Applied Psychology, 90, 616 – 628.
Accident Analysis and Prevention, 39, 556 –564. Zuckerman, M., Kuhlman, D. M., Thornquist, M., & Kiers, H. (1991). Five
Viswesvaran, C., & Ones, D. S. (1995). Theory testing: Combining psy- (or three) robust questionnaire scale factors of personality without cul-
chometric meta-analysis and structural equations modeling. Personnel ture. Personality and Individual Differences, 12, 929 –941.
Psychology, 48, 865– 885. Zyphur, M. J., Kaplan, S. A., & Christian, M. S. (2008). Assumptions of
Wagenaar, W. A. (1992). Risk taking and accident causation. In J. F. Yates cross-level measurement and structural invariance in the analysis of
(Ed.), Risk-taking behavior (pp. 257–281). Chichester, England: Wiley. multilevel data: Problems and solutions. Group Dynamics: Theory,
Wallace, J., & Chen, G. (2005). Development and validation of a work- Research, and Practice, 12, 127–140.
specific measure of cognitive failure: Implications for occupational
safety. Journal of Occupational and Organizational Psychology, 78, Received June 9, 2008
615– 632. Revision received March 30, 2009
Wallace, J. C., & Chen, G. (2006). A multilevel integration of personality, Accepted April 2, 2009 䡲

View publication stats