Safety Climate and Injuries: An Examination of Theoretical and Empirical Relationships

Abstract

Our purpose in this study was to meta-analytically address several theoretical and empirical issues regarding the relationships between safety climate and injuries. First, we distinguished between extant safety climate-->injury and injury-->safety climate relationships for both organizational and psychological safety climates. Second, we examined several potential moderators of these relationships. Meta-analyses revealed that injuries were more predictive of organizational safety climate than safety climate was predictive of injuries. Additionally, the injury-->safety climate relationship was stronger for organizational climate than for psychological climate. Moderator analyses revealed that the degree of content contamination in safety climate measures inflated effects, whereas measurement deficiency attenuated effects. Additionally, moderator analyses showed that as the time period over which injuries were assessed lengthened, the safety climate-->injury relationship was attenuated. Supplemental meta-analyses of specific safety climate dimensions also revealed that perceived management commitment to safety is the most robust predictor of occupational injuries. Contrary to expectations, the operationalization of injuries did not meaningfully moderate safety climate-injury relationships. Implications and recommendations for future research and practice are discussed.

Full text

Safety Climate and Injuries:
An Examination of Theoretical and Empirical Relationships
Jeremy M. Beus, Stephanie C. Payne, Mindy E. Bergman, and Winfred Arthur Jr.
Texas A&M University
Our purpose in this study was to meta-analytically address several theoretical and empirical issues
regarding the relationships between safety climate and injuries. First, we distinguished between extant
safety climate3injury and injury3safety climate relationships for both organizational and psychological
safety climates. Second, we examined several potential moderators of these relationships. Meta-analyses
revealed that injuries were more predictive of organizational safety climate than safety climate was
predictive of injuries. Additionally, the injury3safety climate relationship was stronger for organiza-
tional climate than for psychological climate. Moderator analyses revealed that the degree of content
contamination in safety climate measures inflated effects, whereas measurement deficiency attenuated
effects. Additionally, moderator analyses showed that as the time period over which injuries were
assessed lengthened, the safety climate3injury relationship was attenuated. Supplemental meta-analyses
of specific safety climate dimensions also revealed that perceived management commitment to safety is
the most robust predictor of occupational injuries. Contrary to expectations, the operationalization of
injuries did not meaningfully moderate safety climate–injury relationships. Implications and recommen-
dations for future research and practice are discussed.
Keywords: organizational safety climate, psychological safety climate, injury, meta-analysis
The study of safety is of obvious organizational importance. In
2007, there were over four million nonfatal work injuries and more
than 5,600 work fatalities reported in the United States (Bureau of
Labor Statistics, 2008). However, a recent audit of workplace
injuries reported to the Occupational Safety and Health Adminis-
tration (OSHA) suggested that up to two thirds of all workplace
injuries and illnesses go unreported by employees due to factors
such as fear of disciplinary action or the loss of valued incentives
(U.S. Government Accountability Office, 2009). Consequently,
actual workplace injury rates may be far greater than indicated by
existing OSHA injury records. Safety climate, or employees’ per-
ceptions of organizational safety policies, procedures, and prac-
tices (Zohar, 2003), plays a critical role in workplace safety. Past
meta-analytic research has demonstrated that safety climates that
are supportive of safety are associated with fewer occupational
injuries than safety climates that are not supportive of safety
(Christian, Bradley, Wallace, & Burke, 2009; Clarke, 2006a).
However, it is noteworthy that the presumed effect of safety
climate on workplace injuries has often been examined in studies
that related an assessment of safety climate to injuries that oc-
curred prior to that assessment time. In her meta-analysis, Clarke
(2006a) distinguished between retrospective designs (i.e., injuries
assessed prior to safety climate assessment) and prospective de-
signs (i.e., injuries assessed after safety climate assessment) and
found this distinction to moderate the safety climate–injury rela-
tionship. However, because Clarke did not disaggregate
individual-level (psychological climate) and group-level (organi-
zational climate) primary studies in her meta-analysis, important
differences in the safety climate–injury relationship that result
from the conceptual levels of safety climate (i.e., psychological vs.
organizational; Christian et al., 2009; Ostroff, Kinicki, & Tamkins,
2003) were obscured. In addition, Clarke conceptualized retro-
spective and prospective designs as different means of assessing
the influence of safety climate on workplace injuries (i.e., safety
climate3injury) and not as the assessment of theoretically distinct
relationships. We contend, in contrast, that previous retrospective
studies purporting to measure the effect of safety climate on injuries
have instead assessed the influence of injuries on safety climate (i.e.,
injury3safety climate). Thus, we argue that this is not a simple design
distinction but that there are instead important theoretical differences
between the safety climate3injury and injury3safety climate con-
ceptualizations. Consequently, one objective in the present study was
to use meta-analytic procedures to empirically investigate this distinc-
tion of safety climate3injury and injury3safety climate relationships
while the distinction between psychological and organizational safety
climate was maintained. These distinctions thus resulted in the poten-
tial for four meta-analyses of the relationships between safety climate
and injuries.
An additional purpose in this study was to examine a number of
moderators posited to affect the relationships between safety cli-
Jeremy M. Beus, Stephanie C. Payne, Mindy E. Bergman, and Winfred
Arthur Jr., Department of Psychology, Texas A&M University.
This research was supported by the Mary Kay O’Connor Process Safety
Center at Texas A&M University. We thank Jennifer Rodriguez and Ian
Wilson for their assistance with coding measures and articles and Richard
Woodman for his feedback on an earlier draft of this paper. This article is
based on Jeremy M. Beus’s master’s thesis. An earlier version of this paper
was presented at the annual conference of the Society for Industrial and
Organizational Psychology, New Orleans, Louisiana, April 2009.
Correspondence concerning this article should be addressed to Jeremy
M. Beus, Department of Psychology, Texas A&M University, 4235
TAMU, College Station, TX 77843-4235. E-mail: jeremybeus@gmail.com
Journal of Applied Psychology © 2010 American Psychological Association
2010, Vol. 95, No. 4, 713–727 0021-9010/10/$12.00 DOI: 10.1037/a0019164
713

mate and injuries. Injuries are defined here as workplace incidents
that result in personal harm to an individual, ranging from slips,
trips, and other minor occurrences (Evans, Michael, Wiedenbeck,
& Ray, 2004; Oliver, Cheyne, Tomas, & Cox, 2002) to those that
require first aid treatment (Hofmann & Stetzer, 1996; Michael,
Evans, Jansen, & Haight, 2005) or time off work (Neal & Griffin,
2006). OSHA defined reportable injuries as those that at a mini-
mum require more than basic first aid treatment. For analyses
discussed subsequently, we incorporated this definition of work-
place injuries (i.e., OSHA injuries), as well as a more expansive
definition (i.e., more than OSHA injuries). The moderators exam-
ined in this study pertain to (a) how safety climate and injuries
have been associated temporally and (b) how they have been
operationalized. These moderators represent meaningful theoreti-
cal and conceptual variations across studies that pose important
implications for both research and practice. The specific modera-
tors examined were the length of time over which workplace
injuries are assessed, measures of content contamination and de-
ficiency of safety climate, and the operationalization of injuries. A
brief discussion of each of these factors follows.
Safety Climate–Injury Relationships
Zohar (2003) conceptualized safety climate as a distal anteced-
ent of workplace injuries. That is, the influence of perceptions of
workplace safety policies, procedures, and practices on injuries is
theoretically mediated via their more direct effects on behavior–
outcome expectancies, which subsequently affect safety behavior
and performance (Christian et al., 2009; Clarke, 2006a; Gulden-
mund, 2000; Neal & Griffin, 2004; Zohar, 2003). Because safety
climate informs behavior– outcome expectancies, a supportive
safety climate, in which safe behavior is reinforced, is expected to
be associated with fewer injuries, whereas an unsupportive cli-
mate, in which safe behavior is not reinforced— or is possibly even
punished—is expected to be associated with more frequent injuries
(Zohar, 2003). Conversely, although safety climate is typically
conceptualized as a predictor of workplace injuries (i.e., safety
climate3injuries) regardless of the relative sequencing of the mea-
surement of these two constructs (e.g., Christian et al., 2009; Clarke,
2006a; Hofmann & Stetzer, 1996; Johnson, 2007; Mearns, Whitaker,
& Flin, 2003; Zohar, 2000; Zohar & Luria, 2004), injuries are also
considered to be predictive of safety climate (i.e., injuries3safety
climate; Zohar, 2003) because injuries provide information about the
safety of the workplace. That is, when injuries occur, they are signals
about the underlying safety climate in the organization (Spence, 1973)
such that employees’ reflection on past injury-related events and
experiences will influence employees’ perceptions of safety policies,
procedures, and practices (Schneider & Reichers, 1983).
As previously noted, the safety climate3injury relationship has
been investigated in studies using either prospective or “retrospec-
tive” designs. The methodological threats associated with retro-
spective designs are widely recognized (Bitektine, 2008; Denrell &
Kovacs, 2008). However, in the safety climate–injuries domain,
they are exacerbated because of an alternative theoretical framing
that posits an injury3safety climate relationship. If this is the case,
what is designated as a retrospective design under the safety
climate3injury conceptualization is in fact a prospective design
under the injury3safety climate conceptualization. Thus, although
previous research may have labeled these as tests of prospective
versus retrospective designs (e.g., Clarke, 2006a), we contend that
these really represent tests of two competing theoretical frame-
works: safety climate3injury versus injury3safety climate con-
ceptualizations. Accordingly, in the present study, we considered
prospective designs to represent tests of a safety climate3injury
relationship and so-called retrospective designs to represent tests
of an injury3safety climate relationship. Although we consider
the retrospective studies included in this meta-analysis to be tests
of the injury3safety climate relationship, it should be clarified
that, in most cases, the authors considered injuries not to be a
predictor of safety climate but rather a postdictive criterion.
Although there is a sound theoretical basis for both conceptu-
alizations of the safety climate-injury relationship (i.e., safety
climate3injury and injury3safety climate), there is little theoret-
ical guidance as to whether one should result in a stronger rela-
tionship than the other. Clarke’s (2006a) meta-analysis of safety
climate and injuries revealed a weaker relationship for studies
using retrospective designs (i.e., the injury3safety climate con-
ceptualization; ␳⫽⫺.22) than for studies that used prospective
designs (i.e., the safety climate3injury conceptualization; ␳⫽
⫺.33). However, as previously noted, Clarke did not disaggregate
the organizational and psychological climate studies in her meta-
analysis, so it is unclear how much these different conceptualiza-
tions contributed to differences in her results. Hence, in the ab-
sence of a compelling theoretical or empirical basis for why one
should result in a stronger relationship than the other, we pose an
exploratory research question: Does the magnitude of the safety
climate3injury relationship differ from that of the injury3safety
climate relationship?
Organizational and Psychological Safety
Climate–Injury Relationships
Ostroff et al. (2003) argued that climate can be conceptualized
at the psychological and organizational levels.
1
Psychological cli-
mates constitute individuals’ perceptions about a coherent set of
policies, procedures, and practices and are born of both direct and
indirect exposure to these (Ostroff et al., 2003). On the other hand,
organizational climate is the collective’s perceptions of those
policies, procedures, and practices; it is an emergent group-level
phenomenon that constitutes an aggregate of the climate percep-
tions within a group (Kozlowski & Klein, 2000). Because similar
phenomena can have very different effects across organizational
levels (Kozlowski & Klein, 2000), it is essential that organiza-
tional and psychological climate be distinguished from each other
(Ostroff et al., 2003). Thus, psychological safety climate reflects
individual perceptions of safety policies, procedures, and practices
in the workplace, whereas organizational safety climate is the
collective of perceptions regarding the same (Christian et al.,
2009).
The safety climate3injury and injury3safety climate relation-
ships are likely to be influenced by unique processes for organi-
zational and psychological climate. In particular, the safety climate
perceptions of individuals will be affected by their own idiosyn-
cratic worldviews, perceptual biases, and experiences (Ostroff &
1
In this paper, we refer to “organizational” safety climate as any
group-level safety climate (e.g., work group, worksite, organization).
714 BEUS, PAYNE, BERGMAN, AND ARTHUR

Bowen, 2000). These idiosyncrasies can lead to different employee
interpretations of the same organizational phenomena (Guzzo &
Noonan, 1994). Varying organizational interpretations can differ-
entially affect individual behavior– outcome expectancies (Rous-
seau & Wade-Benzoni, 1994), subsequent safety behavior, and
ultimate injury occurrences (Zohar, 2003) and will consequently
attenuate relationships with injuries. Likewise, individual idiosyn-
crasies should also affect employee interpretations of workplace
injuries and their resulting safety climate perceptions. Thus, the
safety climate3injury and injury3safety climate relationships
may be attenuated when considered at the psychological level as
compared to the organizational level.
Conversely, organizational safety climate represents the higher
level emergence of individual interpretations of and expectations
regarding workplace safety. This higher level emergence corre-
spondingly represents the amplification of a psychological phe-
nomenon to a more pervasive group-level phenomenon (Kozlow-
ski & Klein, 2000). Consequently, when the organizational
environment leads to the group-level convergence of workplace
safety perceptions, the safety-related behaviors within the group
are likely to be more similar, leading to stronger associations with
both past and subsequent workplace injuries within the group. That
is, when considered at the group level, safety climate should have
a stronger influence on future injuries and injuries should have a
stronger influence on subsequent group safety climate perceptions.
In line with this reasoning, Christian et al. (2009) found that
organizational safety climate had a stronger association with inju-
ries than did psychological safety climate, though they did not
separate safety climate3injury and injury3safety climate rela-
tionships. We expected, consistent with Christian et al.’s (2009)
results, that the safety climate3injury and injury3safety climate
relationships would be stronger for organizational safety climates
than for psychological safety climates.
Hypothesis 1: Safety climate3injury and injury3safety cli-
mate relationships will be stronger for organizational climates
than for psychological climates.
Proposed Moderators of Safety Climate–Injury
Relationships
Because time has been acknowledged to be an important factor
in the relationships between predictors and criteria (Ancona,
Goodman, Lawrence, & Tushman, 2001; George & Jones, 2000;
Mitchell & James, 2001), we sought to examine its influence on
safety climate–injury relationships as a means of informing both
theory and practice. We examined the length of time over which
injuries are assessed as a potential moderator of the relationships
between safety climate and injuries.
In addition, specification of the content domain of any two
constructs necessarily affects the extent to which they will share
consistent associations. For safety climate, the absence of a com-
monly accepted conceptualization of the construct space has re-
sulted in tremendous variation regarding safety climate at both the
theoretical and the empirical level. Further, given that injuries can
range from minor to severe and, at times, have vastly different
explanations for their occurrences, the specification of what con-
stitutes an injury can greatly affect its association with safety
climate. Thus, we examined safety climate content contamination
and deficiency as well as the operationalization of injuries as
potential moderators of safety climate–injury relationships.
Length of Time Over Which Injuries Are Assessed
The length of time over which injuries are assessed is a potential
source of variance in safety climate–injury relationships. In the
safety climate literature, the time frames for injury assessment
have ranged from as little as three months (Hofmann & Mark,
2006) to more than two years (Garavan & O’Brien, 2001; Lyon,
2007; Varonen & Mattila, 2000) to employees’ total tenure with
the organization (Clarke, 2006b; Huang, Ho, Smith, & Chen, 2006;
Williamson, Feyer, Cairns, & Biancotti, 1997). These vastly dif-
ferent time frames for injury occurrences could moderate the
relationships between safety climate and injuries. However,
whether longer time frames increase or decrease safety climate–
injury relationships is an open question, as there are competing
explanations regarding how the length of time over which injuries
are assessed affects these relationships.
On one hand, Harrison and Hulin (1989) argued that the aggre-
gation of criterion data over long time periods can constrain causal
inferences because of the increased temporal distance between the
measurement of the predictor and the included criterion. That is,
for causation to be inferred accurately, the assumed cause must be
contiguous to the presumed effect (Cook & Campbell, 1979). As
the time frame expands to include injuries that are more temporally
distal from the measurement of safety climate, more factors may
intervene in the process and influence relationships, thereby dis-
rupting the contiguity of the variables examined. For example, the
acquisition of new technology or a change in existing work pro-
cedures could alter the relationships between safety climate and
injuries, and such events are more likely to occur during longer
time frames. Likewise, over a longer time period, safety climate
itself could change from what it was when it was originally
assessed, altering its associations with previous and subsequent
organizational injuries. Thus, use of a long time frame could lead
researchers to overlook the incremental and nonrecursive adjust-
ments in the relationships between safety climate and injuries. This
could occur following major organizational events, such as merg-
ers, acquisitions, unionization, or top management team changes,
but could also occur due to smaller events, such as a series of
minor injuries (e.g., trips and falls, near injuries) or changes in
personnel. Thus, this perspective argues that there is greater con-
tiguity between safety climate and injuries over shorter time peri-
ods than over longer time periods, and this should hold true for
both injury3safety climate and safety climate3injury relation-
ships. Therefore, one possible effect of the length of time over
which injuries are assessed on the safety climate–injury relation-
ship could be that shorter time frames are associated with larger
injury3safety climate and safety climate3injury associations due
to the likelihood that fewer factors intervene between causal pro-
cesses and outcomes.
Alternatively, because workplace injuries are low base-rate phe-
nomena (Harrison & Hulin, 1989; Jacobs, 1970), when there are no
injuries in a workplace it is unclear if the lack of injuries is due to
an organizational propensity for few injuries to occur or to an
insufficient amount of time for injuries to appear when there is a
propensity for injuries to occur. Hulin and Rousseau (1980) re-
ported that a common means of studying infrequent events is to
715
SAFETY CLIMATE–INJURY META-ANALYSIS

gather criterion data over longer time intervals. That is, because
injuries are low base-rate events, longer time periods for gathering
injury data are often necessary for amounts of variance to be
sufficient for detection of a relationship between safety climate
and injuries. Consequently, it could be that longer time frames
over which injuries are collected are associated with larger
injury3safety climate and safety climate3injury effects. Due to
the presence of competing rationales, we sought to investigate
whether injury3safety climate and safety climate3injury rela-
tionships are moderated by the length of time over which injuries
are assessed and, if so, to determine whether shorter or longer time
frames are associated with stronger relationships.
Safety Climate Measure Content Contamination and
Deficiency
There is little consensus as to which factors adequately consti-
tute the construct of safety climate (Guldenmund, 2000). One
reason for this is that safety climate is often conceptualized induc-
tively by examining the safety literature and conducting interviews
and focus groups to create industry- and situation-specific mea-
sures (Flin, Mearns, O’Connor, & Bryden, 2000). This approach
has led to measures with vastly different numbers and types of
factors. In the extant literature, measures of safety climate have
factor structures ranging from a low of one factor (Barling, Lough-
lin, & Kelloway, 2002; Evans et al., 2004; Neal & Griffin, 2006)
to a high of 12 factors (Krispin, 1997). Flin et al. (2000) reviewed
the safety climate literature and concluded that the most common
safety climate factors were management commitment to safety,
safety systems, risk, work pressure, and competence. However,
although these factors are the most common, there is no evidence
that they constitute the best conceptualizations of the safety cli-
mate construct; in fact, there is reason to suggest that some of these
dimensions may not represent safety climate at all. For example,
“inherent risk” (i.e., hazards present in the job) is arguably not a
theoretically appropriate aspect of the safety climate content do-
main; some jobs and their task elements are dangerous or risky
independent of the prevailing safety climates. For example, off-
shore oil drilling, high-rise construction projects, firefighting, and
police work are inherently risky because of the nature of the work
itself and not necessarily the existing safety climates. Accordingly,
it is informative to determine the degree to which safety climate
measures correspond to a theoretical conceptualization of safety
climate as a means of better understanding the true relationships
between safety climate and injuries.
For the purposes of this study, we adopted Zohar’s (2003)
theoretical definition of safety climate as our exemplar.
2
Zohar’s
definition (employee’s perceptions of the policies, procedures, and
practices concerning safety) is consistent with the broader organi-
zational climate literature (Hellreigel & Slocum, 1974; James &
Jones, 1974) and aligns with Reichers and Scheider’s (1990)
definition of organizational climate as employees’ perceptions of
organizational policies, procedures, and practices as they pertain to
specific organizational phenomena (e.g., safety, service; Schneider
& Reichers, 1983). Additionally, Zohar’s definition does not in-
clude some of the contaminants identified by other safety climate
researchers (e.g., individual differences such as fatalism and safety
attitudes; Neal & Griffin, 2004).
Messick (1980, 1995) argued that there are two major threats to
construct validity. The first is construct-irrelevant variance (i.e.,
contamination). A measure is contaminated if it includes content
that is not associated with the conceptual content domain (Mes-
sick, 1995). The presence of content contamination can distort the
predictor– criterion relationship of interest (Messick, 1995), such
that contamination in a safety climate measure can create “noise”
in safety climate–injury relationships and subsequently attenuate
effect sizes. For the present meta-analysis, content contamination
was operationalized as the proportion of items within a given
study’s safety climate measure that assessed something other than
safety climate, as defined by Zohar (2003).
Hypothesis 2: Higher levels of safety climate content con-
tamination will be associated with weaker safety climate–
injury relationships.
Messick (1995) argued that the second major threat to construct
validity is construct underrepresentation (i.e., deficiency). In this
instance, it is the degree to which a measure of safety climate fails
to represent the complete specified content domain. Accordingly,
when a safety climate measure is deficient, empirical results using
that measure should be attenuated relative to studies that use
nondeficient measures. This is because the underrepresentation of
a domain in a measure should result in a failure to capture the
breadth of indicators of the phenomenon and overlook relevant
variance components. Therefore, deficient measurement is also
expected to attenuate safety climate–injury relationships. In this
study, content deficiency was operationalized as the degree to
which, relative to Zohar’s (2003) theoretical definition, a study’s
safety climate measure did not tap the full safety climate domain.
That is, deficiency represented the extent to which safety climate
measures failed to assess employee perceptions of organizational
safety policies, procedures, and practices.
Hypothesis 3: Higher levels of safety climate content defi-
ciency will be associated with weaker safety climate–injury
relationships.
We conducted a series of dimension-level meta-analyses as a
supplement to our scale-level analyses of safety climate content
contamination and deficiency to provide a richer illustration of
how common dimensions, whether relevant to the safety climate
content domain or not, differentially relate to injury occurrences.
3
We note that these dimension-level analyses constitute a more
fine-grained examination of safety climate dimensions than was
recently reported by Christian et al. (2009). That is, we did not
aggregate several distinct safety climate dimensions that Christian
et al. chose to combine into single dimensions (e.g., management
safety practices, safety values, and safety communication into
“management commitment”). These exploratory dimension-level
analyses complement our general tests of contamination and defi-
2
For reviews of the development of the safety climate construct, see
Guldenmund (2000); Neal and Griffin (2004); and Zohar (2003). For
alternative definitions of safety climate, see Guldenmund (2000, Table 3).
3
We thank an anonymous reviewer for encouraging these analyses.
716 BEUS, PAYNE, BERGMAN, AND ARTHUR

ciency by revealing where common contaminants and deficiencies
lie.
Injury Operationalization
Injury operationalizations often vary in terms of the degree of
severity required for an injury to be reported and, subsequently,
counted or measured. Minimum inclusion criteria for injury op-
erationalizations determine the range of injuries that are included
in a given study. That is, less severe minimum inclusion criteria
lead to broader ranges of injury data, whereas more severe mini-
mum criteria limit the types of injuries that can be included
because injuries must be of a higher level of severity to be counted.
In studies of safety climate–injury relationships, the minimum
criteria for injuries have ranged from including nearly all injuries,
such as slips and trips (Evans et al., 2004; Oliver et al., 2002), to
only those that required at least first aid treatment (Hofmann &
Stetzer, 1996; Michael et al., 2005) or time off work (Neal &
Griffin, 2006).
It is common practice for organizations to use OSHA criteria for
recording injuries. OSHA’s specific criteria for reportable injuries
include only those injuries that result in death, days away from
work, restricted work or transfer to another job, medical treatment
beyond first aid, loss of consciousness, or diagnosis by a physician
or licensed health care professional (OSHA, 2008). Thus, OSHA
criteria include only those injuries that are comparatively severe
and exclude injuries that are of a less serious nature (Komaki,
Barwick, & Scott, 1978). As previously noted, the nature of the
reporting criteria can lead to situations in which injuries are
underreported. The result is deficient data and reports of workplace
injury rates (U.S. Government Accountability Office, 2009). Al-
ternatively, employees can be asked to report all injuries regardless
of their severity (Clarke, 2006b; Cree & Kelloway, 1997; Donald
& Canter, 1994). The minimum severity of injury operationaliza-
tions determines the available sample of injury reports. Further,
given that severe injuries are less likely to occur than more minor
injuries (Crowl & Louvar, 2002; Heinrich, 1931), the use of more
inclusive injury criteria should result in the majority of included
injuries being less serious in nature.
Safety climate—whether organizational or psychological—
should influence future injuries, regardless of their severity. Both
minor and major injuries can result from safety lapses that are
indicative of poor safety climates. Additionally, a pattern of minor
injuries is often a precursor to more severe injuries (Crowl &
Louvar, 2002). Consequently, given that minor injuries often pre-
cede more serious ones, they are more proximal to safety climate
than major injuries. Thus, safety climate should be more effective
in predicting injuries of a less serious nature.
On the basis of a review of the safety climate⫺injury literature,
we dichotomized injury operationalizations as studies that used
only OSHA-reportable injuries and studies that included both
OSHA-reportable injuries and less severe injuries. Thus, the range
of injury severity was much broader (as it was more inclusive) for
this latter group of studies than for the OSHA-criteria-only studies.
Therefore, we made the following hypothesis, based on this di-
chotomization and the theoretical arguments above.
Hypothesis 4a: The safety climate3injury relationship will
be moderated by injury operationalization such that the in-
clusion of more than OSHA-reportable injuries will be asso-
ciated with stronger relationships than will the inclusion of
only OSHA-reportable injuries.
With regard to the injury3safety climate relationship, it is
likely that the severity of injuries affects subsequent safety climate
perceptions to differing degrees. That is, minor injuries would be
expected to have a lesser impact on safety climate perceptions than
would more major injuries (e.g., OSHA-reportable injuries).
Whereas less severe injuries can often be discounted on an indi-
vidual basis, severe injuries are more memorable and would likely
have a greater influence on subsequent perceptions of workplace
safety. This is not to say that minor injuries are not meaningful
indicators of safety climate; rather, observing or personally expe-
riencing a minor injury should have a smaller effect on consequent
safety perceptions than should observing or experiencing a severe
injury. For example, witnessing a group member receive a minor
cut would likely have less of an effect on safety climate percep-
tions than witnessing a group member break a limb. Thus, more
severe injuries should have a stronger influence on safety climate
than should less severe injuries.
Hypothesis 4b: The injury3safety climate relationship will
be moderated by injury operationalization such that OSHA-
reportable injuries will have a stronger effect on safety cli-
mate than the inclusion of more than OSHA-reportable inju-
ries.
Method
Literature Search
To locate studies, we conducted an online literature search of the
PsycINFO, PubMed, and dissertation databases using the key-
words safety climate and injury, injuries, accident, or accidents. A
search of Society for Industrial and Organizational Psychology,
Academy of Management, and Human Factors and Ergonomics
Society conference programs from the past 7 years (2003⫺2009)
was conducted to locate unpublished studies. Additionally, re-
quests for published and unpublished safety climate studies were
posted on three Listservs (i.e., HRDivNet, RMNet, OBlistserv).
Further, researchers in the fields of safety climate and injuries were
contacted directly to seek unpublished studies.
Inclusion Criteria
Studies were initially eligible for inclusion if they reported the
relationship between a measure of safety climate and a measure of
workplace injuries and included either an appropriate effect size or
sufficient information to permit the computation of one. Further,
given that the terms safety culture and safety climate are often used
interchangeably despite their theoretical differences (Denison,
1996), we examined studies that purported to measure safety
culture to determine if the substance of what was being measured
was in fact safety climate. We also examined studies that assessed
related constructs such as “safety attitudes.” Additionally, studies
had to provide sufficient information for us to determine if the
reported injuries occurred before or after the assessment of safety
climate. We excluded studies in which this information was not
717
SAFETY CLIMATE–INJURY META-ANALYSIS

provided and those in which injuries and the assessment of safety
climate could have occurred concurrently.
In cases where an overall safety climate–injury effect size was
not reported and factor-specific effect sizes were present, a com-
posite effect size was computed (Nunnally & Bernstein, 1994).
Multiple effect sizes within a given study were each eligible for
inclusion as long as they were associated with independent sam-
ples or offered unique contributions to the meta-analyses. For
example, safety climate3injury and injury3safety climate con-
ceptualizations associated with the same sample (e.g., Neal &
Griffin, 2006) were eligible for inclusion, given that these rela-
tionships were meta-analyzed separately.
Information pertinent to moderator analyses (e.g., length of time
over which injuries were assessed or safety climate measure used)
was required for inclusion in the relevant analyses. For example, a
study lacking information on the operationalization of injuries
would not be eligible for inclusion in the analysis of injury opera-
tionalization as a moderator. These inclusion criteria resulted in an
overall sample of 32 injury3psychological safety climate effect
sizes (N⫽16,011), 10 injury3organizational safety climate
effect sizes (N⫽251), and 11 organizational safety
climate3injury effect sizes (N⫽458). Because we located only
one study that examined the psychological safety climate3injury
relationship (i.e., Hofmann & Morgeson, 1999), we could not
conduct a meta-analysis of this relationship. As a result, there were
three rather than four highest level relationships for the present
study. Studies that were included in the meta-analyses are marked
by asterisks in the reference list.
Data Coding
Each study was coded for pertinent sample information, aspects
of the study design, and information on the predictor and criterion
measures. It should be noted that some effect sizes were transposed
to ensure that, for all effect sizes, a negative sign indicated that
more favorable safety climates are related to fewer injuries and
vice versa. Jeremy M. Beus and Stephanie C. Payne coded each
study independently. This exercise resulted in 91% agreement
across all coded elements between the two coders. Discrepancies
were resolved by discussion and by revisiting the articles in
question. For our supplemental analyses, safety climate dimen-
sions were grouped primarily on the basis of their author-assigned
labels, although for labels that were ambiguous or unclear, items
were examined where possible to facilitate more precise categori-
zations.
Meta-Analytic Procedures
This meta-analysis was conducted with Hunter and Schmidt’s
(2004) meta-analytic approach. Corrections were made for sam-
pling error and unreliability in the safety climate measures using
an artifact distribution; separate distributions were computed for
each subgroup analysis. Effect sizes were not corrected for unre-
liability in the measurement of injuries because reliabilities are
typically not reported for injury measures, in large part because
injury measures are often counts rather than psychological scales.
Analysis of Moderators
Because organizational and psychological safety climate3injury
and injury3safety climate relationships are theoretically distinct,
there are four separate theoretical associations. We were able to
examine three of them meta-analytically: (a) organizational safety
climate3injury; (b) injury3organizational safety climate; and (c)
injury3psychological safety climate. To evaluate research questions
and hypotheses relating to these highest level conditions, we com-
pared the absolute values of the population estimates obtained and
interpreted them as absolute differences that reflect true population
differences (Hunter & Schmidt, 2004).
To determine whether moderators were likely to be operating,
we calculated the percentage of variance explained by statistical
artifacts and constructed 95% credibility intervals. If after correc-
tion for statistical artifacts much of the variance in effect sizes is
unaccounted for, it is likely that moderators are present. Likewise,
credibility intervals provide an estimate of the amount of variabil-
ity across studies and suggest moderation when the interval in-
cludes zero (Hunter & Schmidt, 2004). However, considerable
variance may still be indicated by a large standard deviation value
and thus still suggest moderation even when the credibility interval
does not include zero. Thus, moderation was determined to exist if
credibility intervals included zero and if the standard deviations of
the different moderator conditions were not smaller relative to the
overall population estimate’s standard deviation (i.e., variance did
not decrease after accounting for the proposed moderators).
All proposed moderators were examined hierarchically within
organizational and psychological safety climate3injury and
injury3safety climate relationships. Categorical moderators were
assessed with Hunter and Schmidt’s (2004) subgroup analysis.
That is, separate meta-analyses were performed for each proposed
moderator condition (i.e., OSHA injuries vs. more than OSHA
injuries) to allow for comparisons between conditions. As an a
priori decision rule, proposed moderator conditions were deter-
mined to be meaningfully different by comparing confidence in-
tervals around the sample-weighted mean observed effect size for
each condition; overlapping confidence intervals suggest that mod-
erator conditions are not meaningfully different.
For continuous moderators, weighted least squares multiple
regression (WLS) was used to test for moderation. Monte Carlo
simulations demonstrate that this method provides the most accu-
rate estimates and is least affected by multicollinearity, even with
small sample sizes (Steel & Kammeyer-Mueller, 2002). The
weighting factor used was the inverse square root of the sampling
error (Steel & Kammeyer-Mueller, 2002).
Contamination and deficiency in safety climate measures.
Twenty-nine of a possible 35 measures, which accounted for 47 of
the 53 total effect sizes in this meta-analysis, were evaluated for
content contamination and deficiency. Items were gathered by
reviewing published sources or by contacting the authors; we were
unable to obtain the six remaining measures. Of the 29 measures
obtained, 15 were used in psychological climate studies only, 10
were used in organizational climate studies only, and four mea-
sures were used at least once for each safety climate conceptual-
ization. The number of items in these measures ranged from three
to 69.
Before the measures were evaluated, all identifying information
(e.g., names of the authors for the studies utilizing the measure;
year of publication; source of publication) was removed, leaving
only the set of items. Four subject matter experts (SMEs; two
industrial– organizational psychology professors and two
industrial– organizational psychology doctoral students) evaluated
718 BEUS, PAYNE, BERGMAN, AND ARTHUR

contamination and deficiency on the basis of divergence from or
convergence with Zohar’s (2003) definition of safety climate and
his delineation of policies, procedures, and practices (see Appen-
dix for rating cover sheet). Contamination was operationalized as
the proportion of contaminated items in a measure. Deficiency was
rated on a 7-point Likert scale ranging from 1 (not at all deficient)
to7(completely deficient). The SMEs rated each of the 29 mea-
sures individually and then met to resolve discrepancies and reach
consensus. Preconsensus levels of interrater agreement for con-
tamination (r
wg
⫽.89) and deficiency ratings (r
wg
⫽.81) were
satisfactory.
Results
Safety Climate3Injury Versus Injury3Safety
Climate Relationships
Results for the categorical analyses are provided in Table 1. With
regard to our first research question, the injury3organizational safety
climate relationship (␳⫽⫺.29, k⫽10, N⫽251) was stronger than
the organizational safety climate3injury relationship (␳⫽⫺.24, k⫽
11, N⫽448). This suggests that for organizational climate, injuries
are more predictive of safety climate than safety climate is of injuries.
With regard to Hypothesis 1, the injury3organizational safety cli-
mate relationship (␳⫽⫺.29) was stronger than the
injury3psychological safety climate relationship (␳⫽⫺.16, k⫽32,
N⫽16,011). Comparisons of organizational and psychological safety
climate3injury relationships could not be made due to the unavail-
ability of psychological safety climate3injury studies. Thus, for the
available data, Hypothesis 1 was supported.
With regard to the likelihood of moderation, the credibility
intervals for injury3psychological safety climate (CV ⫽⫺.39 to
.07), injury3organizational safety climate (CV ⫽⫺.66 to .08),
and organizational safety climate3injury (CV ⫽⫺.65 to .18) all
included zero, suggesting that moderators may be operating. These
are addressed next.
Length of Time Over Which Injuries Were Assessed
Results for the analyses of continuous moderators are provided
in Table 2. The effects of the length of time over which injuries are
assessed were tested with WLS multiple regression. Results
showed that the length of the time interval over which injuries
were assessed moderated only the organizational safety
climate3injury relationship, accounting for 39% of the variance in
those effect sizes with a positive standardized regression coeffi-
cient (␤⫽.62, p⬍.05). Because the corrected correlation be-
tween safety climate and injuries was negative, a positive beta
indicates that effect sizes tended to approach zero (i.e., become
smaller) as the length of time over which injuries were assessed
increased. Thus, this suggests that the ability of a given assessment
of safety climate to predict future injuries is lessened as the time
period over which injuries are assessed increases.
The injury3safety climate relationships for organizational and
psychological climate were not moderated by the length of time for
assessing injuries. For both, no variance (R
2
⫽.00) was accounted
for by the time frame over which injuries were assessed. These
results consequently suggest that the influence of injuries on safety
climate perceptions is largely unaffected by the injury time frame.
Safety Climate Content Contamination and Deficiency
The effects of safety climate measure content contamination and
deficiency on safety climate–injury relationships were tested with
WLS multiple regression. Both proposed moderators were entered
simultaneously into regression equations to account for shared
Table 1
Results for Categorical Moderators of Safety Climate–Injury Relationships
Variable kN r៮
% var.
sampling
error
95% CI
␳SD␳
% var.
accounted
for
95% CV
LL UL LL UL
Injury3psychological safety climate 32 1,6011 ⫺.15 14.29 ⫺.19 ⫺.11 ⫺.16 .12 14.82 ⫺.39 .07
OSHA injury criteria 3 961 ⫺.23 4.20 ⫺.52 ⫺.08 ⫺.25 .28 4.30 ⫺.79 .30
OSHA and more 29 1,5050 ⫺.14 19.42 ⫺.18 ⫺.11 ⫺.15 .10 20.16 ⫺.34 .03
Injury3organizational safety climate 10 251 ⫺.26 55.37 ⫺.41 ⫺.10 ⫺.29 .19 55.76 ⫺.66 .08
OSHA injury criteria 6 122 ⫺.35 38.08 ⫺.61 ⫺.09 ⫺.39 .28 38.62 ⫺.94 .16
OSHA and more 4 129 ⫺.17 100 ⫺.28 ⫺.06 ⫺.19 0
a
100 — —
Psychological safety climate3injury
Psychological safety climate
b
—— — — —— —— — ——
Organizational safety climate3injury 11 448 ⫺.22 36.75 ⫺.37 ⫺.07 ⫺.24 .21 36.84 ⫺.65 .18
OSHA injury criteria 2 50 ⫺.08 100 ⫺.35 .19 ⫺.08 0
a
100 — —
OSHA and more 9 398 ⫺.24 33.05 ⫺.40 ⫺.07 ⫺.26 .22 33.11 ⫺.70 .18
Note. Dashes indicate analyses that could not be conducted. k⫽number of safety climate–injury effect sizes; N⫽total number of participants across
studies; r៮⫽sample weighted mean observed r; % var. sampling error ⫽percentage of variance attributed to sampling error; 95% CI ⫽confidence interval,
lower (LL) and upper (UL) bounds; ␳⫽corrected mean r;SD␳⫽standard deviation of corrected effect size; % var. accounted for ⫽percentage of variance
attributed to corrected statistical artifacts; 95% CV ⫽credibility interval, lower (LL) and upper (UL) bounds; OSHA ⫽Occupational Safety and Health
Administration.
a
For cells with a SD␳of 0, the variance of ␳was less than the average sampling error that was subtracted from it. This leads to a negative SD␳, which
Hunter and Schmidt (2004) argued should be interpreted as zero variance.
b
Only one study was found that investigated the safety climate3injury
relationship at the psychological level (i.e., Hofmann & Morgeson, 1999).
719
SAFETY CLIMATE–INJURY META-ANALYSIS

variance.
4
The combined effects of safety climate content contam-
ination and deficiency accounted for 13%, 79%, and 5% of the
variance in safety climate–injury effect sizes for
injury3psychological safety climate, injury3organizational
safety climate, and organizational safety climate3injury relation-
ships, respectively.
Safety climate content contamination. The results for Hy-
pothesis 2, which posited that safety climate content contamination
would result in weaker safety climate–injury relationships, were
mixed. Contamination had a meaningful effect on injury3safety
climate relationships for psychological climate (␤⫽⫺.16, p⬍
.05) and organizational climate (␤⫽⫺.71, p⬍.05) but not for the
organizational safety climate3injury relationship (␤⫽⫺.07, p⬎
.05). However, contrary to expectation, greater contamination was
associated with stronger injury3safety climate relationships for
organizational and psychological climate. Although contamination
was found to be a meaningful moderator in two out of three
analyses, because effects were contrary to expectation, Hypothesis
2 was not supported.
Safety climate content deficiency. Hypothesis 3 posited that
deficiency in safety climate measures would moderate the safety
climate–injury relationship such that deficiency would bias effect
sizes downward (i.e., make effect sizes smaller). The results re-
vealed that deficiency in safety climate measures was a moderator
for injury3safety climate relationships both for psychological
climate (␤⫽.38, p⬍.05) and for organizational climate (␤⫽.77,
p⬍.05) but not for the organizational safety climate3injury
relationship (␤⫽⫺.09, p⬎.05). As hypothesized, greater defi-
ciency was associated with weaker safety climate-injury relation-
ships. However, because deficiency was a moderator only for
injury3safety climate relationships, there is only partial support
for Hypothesis 3, with two of three possible meta-analyses con-
sistent with our expectations.
Safety climate dimensions. We conducted dimension-level
meta-analyses of safety climate and injury relationships to further
examine how common safety climate dimensions and contami-
nants variously relate to injuries and consequently contribute to
observed effects. Table 3 lists the meta-analytic results of all the
reported dimension-level effect sizes that could be gleaned from
the included studies. We reported the meta-analytic results for any
dimensions for which there were two or more reported effect sizes.
This resulted in meta-analyses of 10, nine, and seven separate
dimensions for injury3psychological safety climate, organiza-
tional safety climate3injury, and injury3organizational safety
climate relationships, respectively.
Of the noncontaminated safety climate dimensions meta-
analyzed, perceived management commitment to safety was the
most common. The prevalence of management safety commitment
in safety climate measures is not surprising, given that it is one of
the most widely agreed-upon dimensions in the safety climate
literature (Flin et al., 2000) and a salient indicator for employees
regarding organizational safety policies and practices (Zohar,
2003). Consequently, it is noteworthy that perceived management
commitment to safety had stronger effects in its prediction of
future injuries (␳⫽⫺.30, k⫽10, N⫽431) than injuries did in
predicting management safety commitment both for organizational
safety climate (␳⫽⫺.22, k⫽6, N⫽120) and for psychological
safety climate (␳⫽⫺.12, k⫽10, N⫽5,903). Furthermore,
perceived management commitment to safety demonstrated a
stronger association with subsequent injuries than any other safety
climate dimension (with the exception of safety reporting, which
had the same effect), and it is also the only assessed dimension in
which validity generalized (i.e., the credibility interval did not
include zero). Consequently, for organizational safety climate,
management commitment to safety is the dimension that consti-
tutes the most robust predictor of injuries. Table 3 provides a
listing of the other safety climate dimensions assessed and their
respective effects.
Contaminated dimensions were job safety/risk, personal safety
attitudes, and supervisor competence. Note that these were not the
only contaminated dimensions identified in the included studies
but rather were the most commonly reported dimension-level
content contaminants. Other content contaminants that could not
be meta-analyzed included fatalism, optimism, job security, job
commitment, and employee appreciation.
4
Contamination and deficiency were correlated .17, ⫺.18, and .01 for
psychological injury3safety climate, organizational injury3safety cli-
mate, and organizational safety climate3injury relationships, respectively.
Table 2
Results for Continuous Moderators of Safety Climate–Injury
Relationships
Variable kMSD␤R
2
Injury3psychological
safety climate 32
Time frame for gathering
injury data
b
25 11.24
a
6.65 ⫺.05 .00
Safety climate content
analyses
Overall model 28 .13
Contamination 28 0.21 0.21 ⫺.16
ⴱ
Deficiency 28 2.32 1.02 .38
ⴱ
Injury3organizational
safety climate 10
Time frame for gathering
injury data 10 16.20
a
8.97 ⫺.03 .00
Safety climate content
analyses
Overall model 9 .79
Contamination 9 0.28 0.18 ⫺.71
ⴱ
Deficiency 9 3.11 1.69 .77
ⴱ
Organizational safety
climate3injury 11
Time frame for gathering
injury data 11 9.45
a
3.62 .62
ⴱ
.39
Safety climate content
analyses
Overall model 10 .05
Contamination 10 0.04 0.07 ⫺.19
Deficiency 10 2.80 0.92 ⫺.09
Note. Moderation was examined with weighted least squares multiple
regression. k⫽number of effect sizes; M⫽mean; SD ⫽standard
deviation; ␤⫽beta weight; R
2
⫽proportion of variance attributed to
moderators.
a
Number of months.
b
Seven injury3psychological safety climate stud-
ies were excluded from this analysis because injuries were gathered over
employees’ organizational tenure and not fixed time frames.
ⴱ
p⬍.05.
720 BEUS, PAYNE, BERGMAN, AND ARTHUR

As previously stated, because the risk level for a given job can
often be attributed to the nature of the work itself and not neces-
sarily to the prevailing safety climate, job safety/risk represents a
contaminated assessment of safety climate. Job safety/risk is per-
haps the most pervasive contaminant of safety climate measures.
That is, in addition to constituting the most prevalent dimension-
level contaminant, individual job safety/risk items were found to
be present in 50% of all the safety climate measures that were rated
by SMEs to contain any degree of contamination. The injury3job
safety/risk effects at the psychological level (␳⫽⫺.23, k⫽7, N⫽
3,979) and organizational level (␳⫽⫺.47, k⫽3, N⫽28) were
greater than the overall meta-analytic injury3safety climate rela-
tionships obtained for both respective climates (␳⫽⫺.16, ␳⫽
⫺.29). Thus, consistent with the general effects of content con-
tamination, the assessment of job safety/risk in safety climate
measures would have spuriously inflated safety climate and injury
relationships.
Personal safety attitudes represent another contaminant given
that they are not descriptive of an organization’s safety policies,
procedures, or practices, to which safety climate refers even when
considered at the psychological level. At the psychological level,
the injury3personal safety attitudes effect (␳⫽⫺.09, k⫽5, N⫽
2,943) was weaker than the overall injury3psychological safety
climate effect (␳⫽⫺.16). For organizational climate, the
injury3personal safety attitude effect (␳⫽⫺.29, k⫽4, N⫽42)
did not differ from the overall injury3organizational safety cli-
mate effect (␳⫽⫺.29).
Finally, because the perceived competence of supervisors does
not necessarily bear a direct relationship to their safety behaviors,
supervisor competence is also a safety climate contaminant. For
organizational climate, the injury3supervisor competence associ-
ation (␳⫽.19, k⫽2, N⫽26) was positive in direction and
weaker in magnitude relative to the overall injury3organizational
safety climate relationship (␳⫽⫺.29).
Injury Operationalization
Hypothesis 4a posited that the use of more than OSHA-
reportable injuries (i.e., including more minor injuries) would be
more associated with stronger safety climate3injury relationships
than would the use of only OSHA-reportable injuries (i.e., inclu-
sion restricted to severe injuries). Consistent with expectation, the
Table 3
Dimension-Level Meta-Analyses of Safety Climate–Injury Relationships
Variable kN r៮
% var.
sampling
error
95% CI
␳SD␳
% var.
accounted
for
95% CV
LL UL LL UL
Injury3psychological safety climate 32 1,6011 ⫺.15 14.29 ⫺.19 ⫺.11 ⫺.16 .12 14.82 ⫺.39 .07
Management commitment to safety 10 5,903 ⫺.11 34.16 ⫺.15 ⫺.06 ⫺.12 .06 35.11 ⫺.24 .00
Management safety attitudes 4 992 ⫺.19 11.40 ⫺.37 ⫺.01 ⫺.20 .18 11.52 ⫺.55 .15
Management safety practices 6 2,732 ⫺.15 3.86 ⫺.34 .04 ⫺.16 .24 4.14 ⫺.64 .32
Specific safety policies 4 3,614 ⫺.08 2.19 ⫺.30 .14 ⫺.09 .24 2.24 ⫺.56 .39
Coworker safety 7 2,760 ⫺.20 5.63 ⫺.35 ⫺.05 ⫺.22 .21 5.73 ⫺.63 .20
Safety communication 2 1,096 ⫺.15 100 ⫺.17 ⫺.14 ⫺.18 0
a
100 — —
Safety training 3 2,407 ⫺.17 71.06 ⫺.21 ⫺.12 ⫺.19 .02 76.04 ⫺.24 ⫺.15
Housekeeping 2 450 ⫺.15 100 ⫺.17 ⫺.12 ⫺.24 0
a
100 — —
Personal safety attitudes 5 2,943 ⫺.08 38.94 ⫺.14 ⫺.03 ⫺.09 .06 39.45 ⫺.20 .02
Job safety/risk 7 3,979 ⫺.21 5.11 ⫺.34 ⫺.08 ⫺.23 .19 5.47 ⫺.59 .14
Injury3organizational safety climate 10 251 ⫺.26 55.37 ⫺.41 ⫺.10 ⫺.29 .19 55.76 ⫺.66 .08
Management commitment to safety 6 120 ⫺.21 66.15 ⫺.42 .01 ⫺.22 .17 66.36 ⫺.56 .11
General safety policy 2 21 ⫺.53 100 ⫺.82 ⫺.23 ⫺.62 0
a
100 — —
Safety procedures 2 20 ⫺.39 100 ⫺.56 ⫺.22 ⫺.43 0
a
100 — —
Safety communication 2 26 ⫺.19 56.56 ⫺.70 .32 ⫺.21 .27 56.56 ⫺.73 .32
Safety reporting 3 72 ⫺.11 61.22 ⫺.40 .19 ⫺.12 .18 61.23 ⫺.47 .24
Safety behavior 4 41 ⫺.26 38.17 ⫺.75 .22 ⫺.30 .44 38.19 ⫺1.0 .57
Personal safety attitudes 442⫺.24 100 ⫺.53 .05 ⫺.29 0
a
100 — —
Job safety/risk 328⫺.40 74.91 ⫺.78 ⫺.03 ⫺.47 .19 76.41 ⫺.85 ⫺.10
Supervisor competence 2 26 .17 100 .11 .23 .19 0
a
100 — —
Psychological safety climate3injury — — — — — — — — —
Organizational safety climate3injury 11 448 ⫺.22 36.75 ⫺.37 ⫺.07 ⫺.24 .21 36.84 ⫺.65 .18
Management commitment to safety 10 431 ⫺.27 87.44 ⫺.37 ⫺.18 ⫺.30 .06 87.87 ⫺.41 ⫺.18
Management safety practices 6 195 ⫺.08 27.65 ⫺.35 .19 ⫺.09 .31 27.65 ⫺.69 .52
Safety procedures 4 96 .17 56.15 ⫺.09 .44 .19 .19 56.15 ⫺.19 .56
Safety communication 4 96 .18 52.52 ⫺.09 .45 .19 .21 52.52 ⫺.21 .60
Safety reporting 3 139 ⫺.27 100 ⫺.33 ⫺.20 ⫺.30 0
a
100 — —
Safety behavior 2 30 ⫺.06 82.87 ⫺.47 .34 ⫺.07 .13 82.89 ⫺.32 .18
Note. Italicized dimension labels signify contaminated dimensions; dashes indicate analyses that could not be conducted. k⫽number of safety
climate-injury effect sizes; N⫽total number of participants across studies; r៮⫽sample weighted mean observed r; % var. sampling error ⫽percentage
of variance attributed to sampling error; 95% CI ⫽confidence interval, lower (LL) and upper (UL) bounds; ␳⫽corrected mean r;SD␳⫽standard deviation
of corrected effect size; % var. accounted for ⫽percentage of variance attributed to corrected statistical artifacts; 95% CV ⫽credibility interval, lower
(LL) and upper (UL) bounds.
a
For cells with a SD␳of 0, the variance of ␳was less than the average sampling error that was subtracted from it. This leads to a negative SD␳, which
Hunter and Schmidt (2004) argued should be interpreted as zero variance.
721
SAFETY CLIMATE–INJURY META-ANALYSIS

use of more than OSHA injuries (r៮⫽⫺.24) resulted in a stronger
sample-weighted observed effect size for the organizational safety
climate3injury relationship than did the use of only OSHA inju-
ries (r៮⫽⫺.08). However, overlapping confidence intervals sug-
gest that the magnitude of the safety climate3injury relationship
does not differ substantially on the basis of the severity of injury
operationalizations. Thus, Hypothesis 4a was not supported.
Hypothesis 4b posited that the inclusion of only OSHA-
reportable injuries would result in a stronger injury3safety cli-
mate relationship than would the inclusion of more than OSHA-
reportable injuries. Results revealed, consistent with the direction
of our hypothesis, that the use of OSHA-reportable injuries dem-
onstrated stronger sample-weighted mean effect sizes than did the
inclusion of more than OSHA-reportable injuries for psychological
climate (r៮⫽⫺.23 vs. r៮⫽⫺.14) and organizational climate (r៮⫽
⫺.35 vs. r៮⫽⫺.17). However, overlapping confidence intervals
for both comparisons suggest that this is not a meaningful source
of moderation. Consequently, Hypothesis 4b was not supported.
Discussion
Our purpose in this study was to distinguish safety
climate3injury and injury3safety climate relationships for orga-
nizational and psychological safety climates and to investigate the
effects of several potential moderators of these relationships. Re-
sults revealed that the predictive effects of injuries on organiza-
tional safety climate (␳⫽⫺.29) are slightly stronger than those of
organizational safety climate on injuries (␳⫽⫺.24). Further, the
injury3safety climate relationship was stronger for organizational
climate (␳⫽⫺.29) than for psychological climate (␳⫽⫺.16).
Credibility intervals that included zero and large proportions of
unexplained variance after correcting for statistical artifacts sup-
ported a priori expectations concerning the presence of moderators
for each safety climate–injury relationship. Accordingly, the
length of time over which injuries were assessed was found to be
a significant moderator of the organizational safety
climate3injury relationship, with longer time frames yielding
weaker associations. Concerning the operationalization of safety
climate, safety climate content contamination and deficiency mod-
erated the injury3safety climate relationships for organizational
and psychological climate but not the organizational safety
climate3injury relationship. However, the direction of contami-
nation’s effect was contrary to expectations, such that greater
contamination led to stronger safety climate–injury effects. In
addition, supplemental analyses revealed that perceived manage-
ment commitment to safety is the safety climate dimension with
the most robust association with future injuries. Finally, injury
operationalization was not found to be a moderator for any of the
safety climate–injury relationships. We elaborate on these findings
and discuss their implications below.
Conceptual Framing of the Safety Climate–Injury
Relationship for Psychological and Organizational
Climates
Our findings suggest that injuries have a greater predictive
effect on safety climate than safety climate has on injuries, but the
magnitude of this difference is very small. Safety climate’s effect
on workplace injuries does not appear to be substantively different
from the effect of injuries on safety climate. Thus, consistent with
theory, not only is safety climate associated with future injuries but
the converse is also true, such that individuals appear to recalibrate
their perceptions of organizational safety following injuries, re-
sulting in changes to both psychological and organizational cli-
mate (Zohar, 2003). The contention that safety climates shift
following workplace injuries supports the notion that climates are
dynamic, as opposed to static, organizational phenomena (Schnei-
der & Reichers, 1983). Consequently, although safety climate is
most frequently hypothesized to affect injuries, these results sug-
gest that injuries have a very similar—and even a slightly stron-
ger— effect on organizational safety climate. The slightly greater
magnitude for the effect of injuries on safety climate might be due
to injuries’ more proximal influence on safety climate relative to
safety climate’s influence on injuries, as injuries directly inform
safety climate perceptions but safety climate affects injuries via
workplace behaviors. Consistent with this, our safety
climate3injury effect sizes were smaller than the safety
climate3safety behavior effect sizes reported by Christian et al.
(2009).
The evidence of a stronger injury3safety climate relationship
for organizational climate than for psychological climate suggests
that injuries that occur within a group have a greater impact on the
group’s safety climate than individual injuries have on the injured
person’s psychological safety climate. For safety climate research-
ers, this further underscores the need to differentiate between
organizational and psychological safety climates, as the use of one
conceptualization over another affects empirical relationships and
subsequent research conclusions due to unique processes that
occur within levels.
An alternative explanation for stronger effects at the group level
is that injuries within a group may predict organizational safety
climate better than individual injuries predict psychological safety
climate, due to the fact that workplace injuries are generally rare
events (Jacobs, 1970). There normally is a greater probability of
injuries being incurred within a group of people than by any
specific individual. Greater variance in injury data in turn allows
for greater statistical power to detect effects at the group level.
Although we cannot rule out this alternative explanation, we posit
that emergent group-level processes, not the mere probability of
injury occurrences, are primarily responsible for the stronger
injury3safety climate effects demonstrated for organizational
safety climate.
Moderators of the Safety Climate–Injury Relationship
Our finding that safety climate assessments appear to lose their
ability to predict injuries over time indicates that time is important
when examining safety climate’s influence on injuries. However,
we also found that the influence of injuries on psychological and
organizational safety climate remains stable regardless of the time
period over which injuries are assessed. This is perhaps not sur-
prising, given that workplace injuries are rare events that are likely
stored in employees’ long-term memories (Jacobs, 1970). These
results point to a number of research questions that need to be
addressed about the role of time in safety climate–injury relation-
ships. For example, at what point in time (e.g., after 6 months? 1
year?) does a given snapshot of safety climate no longer affect
injury occurrences? Further, to what degree does safety climate
722 BEUS, PAYNE, BERGMAN, AND ARTHUR

change when workplace injuries occur? Answers to these ques-
tions concerning the relationship between safety climate and inju-
ries over time would aid safety researchers and practitioners in
future safety climate research and the development of safety in-
terventions.
Although our analyses revealed that content contamination and
deficiency are meaningful moderators of injury3safety climate
relationships, it is noteworthy that they did not moderate the
organizational safety climate3injury relationship. There is no
conceptual reason to expect the effects of content contamination
and deficiency to differ for the various safety climate–injury
relationships, so it was unexpected to find that this particular
relationship was unaffected by contamination and deficiency when
the other relationships were. A closer examination of organiza-
tional safety climate3injury studies reveals that, on average, the
safety climate measures used had very small proportions of con-
tamination (M⫽0.04). This suggests that the safety climate
measures used in these studies aligned more closely with Zohar’s
definition. It is probable that, if there had been contamination
present in these measures, the effect of contamination on the
organizational safety climate3injury relationship would have
been similar to that found in the other safety climate–injury anal-
yses. With regard to the null effects of content deficiency on the
organizational safety climate3injury relationship, although the
average deficiency rating for these studies (M⫽2.80) was not
noticeably different from those for the other study subgroups,
the variability was considerably lower (SD ⫽0.92). This re-
duced the ability to detect the effect of this potential moderator.
Again, if the variability had been more similar to that for
injury3safety climate relationships, it is likely that deficiency
would have had a similar effect on this relationship.
Safety Climate Dimensions
Our examination of safety climate dimensions sheds light on
why content contamination resulted in larger safety climate–injury
relationships. A number of the dimensions (e.g., job safety/risk;
individual safety attitudes) contaminating safety climate measures
are important aspects of the context surrounding safety in organi-
zations, even if they are not indicators of safety climate them-
selves. Inherent job risk is a useful variable to consider, especially
given its meaningful association with both past and subsequent
workplace injuries as conveyed here. Although job risk, as it is
commonly assessed in safety climate measures, refers to the level
of risk inherent in the job being performed, it is important to note
that perceptions of job risk can also be influenced by coworker or
supervisor actions. For example, the actions of an employee who
deliberately disregards safety practices and endangers his or her
fellow coworkers would certainly be expected to affect fellow
employees’ perceptions regarding workplace risk. Whereas we
contend that inherent job risk is not an aspect of safety climate, the
safety-related behaviors of coworkers and supervisors arguably
are a part of safety climate.
5
In extant safety climate measures, the
safety-related behaviors of coworkers are often assessed within a
different dimension (e.g., coworker safety practices or safety be-
havior). Given the influence that coworker safety behaviors should
have on perceptions of job risk, future research should determine
the extent to which this safety climate dimension affects subse-
quent risk perceptions. However, because inherent job risk is not
an aspect of safety climate, we reiterate that it should be treated
separately.
Despite our finding that contaminants tended to strengthen the
relationships between safety climate and workplace injuries, it is
important to note that without clean (i.e., noncontaminated) mea-
sures of focal constructs, it is difficult to draw conclusions about
the factors that influence organizational events and therefore dif-
ficult for organizational stakeholders to make decisions about
interventions to improve organizational safety. Conversely, al-
though the content of safety climate measures can readily be
identified as contaminated or not on the basis of Zohar’s (2003)
theoretical definition, it is difficult to provide a succinct recom-
mendation regarding the complete, nondeficient safety climate
content domain. As has been evidenced in the extant safety climate
literature, there are numerous safety-related policies, procedures,
and practices that likely contribute to employees’ safety climate
perceptions. Thus, a value of Zohar’s (2003) definition of safety
climate, in addition to its correspondence with the extant organi-
zational climate literature, is its applicability across organizational
settings.
Our findings regarding content deficiency are informative for
future safety climate research. For example, of the safety climate
measures that were rated by SMEs as neither deficient nor con-
taminated, the reported dimensions fell under the categories of
management commitment to safety; the priority of safety; general
safety policies, procedures, and practices; safety training; safety
communication; safety reporting; and employee safety involve-
ment. Although we do not posit that the above-listed dimensions
necessarily represent the complete safety climate content domain,
we would argue that they do represent substantial breadth within
the domain. Unfortunately, we were unable to conduct dimension-
level meta-analyses for all of these dimensions, and many that
could be meta-analyzed were based on a limited number of studies.
Consequently, we encourage future safety climate researchers to
consider these dimensions in their assessments of safety climate to
ensure more complete representation of the construct space.
Limitations and Future Directions
A limitation of this meta-analysis was the unavailability of data
to test all of the proposed safety climate and injury relationships.
Only one psychological safety climate3injury study was found, so
the meta-analytic comparison of psychological safety
climate3injury and injury3psychological safety climate relation-
ships was not possible. This study (Hofmann & Morgeson, 1999)
reported a ⫺.28 relationship between psychological safety climate
and injuries for a sample of 49 workers. Clearly, additional re-
search on this relationship is warranted. Further, the present meta-
analyses were based on a relatively small number of studies. In
spite of this, our study represents a more extensive quantitative
examination of the safety climate and injuries literature than past
safety climate meta-analyses and addresses important theoretical
and conceptual issues not previously considered (e.g., safety
climate3injury vs. injury3safety climate relationships).
Our dichotomization of workplace injuries was also a limitation,
given that the more inclusive injury category included severe
5
We thank an anonymous reviewer for bringing this to our attention.
723
SAFETY CLIMATE–INJURY META-ANALYSIS

injuries in addition to the more minor injuries. Thus, although the
majority of the injuries included in the “more than OSHA” cate-
gory would likely have been minor, severe injuries would have
been included as well. Unfortunately, the effect of injury severity
could not be teased apart meta-analytically. Hence, there is need
for research that separates injury occurrences by levels of severity
to better delineate the influence of safety climate on the severity of
workplace injuries and vice versa.
A third potential limitation of this study was the reliance on
SME ratings to assess safety climate content contamination and
deficiency. Because these proposed moderators could be assessed
only by using rater judgments, attempts were made to minimize the
possible influence of rater effects by using a common theoretical
framework, preserving scale anonymity, and holding consensus
meetings to discuss ratings. Additionally, it was not always pos-
sible to verify dimension labels, based on corresponding item
content, for our dimension-level analyses; thus, some of the safety
climate dimensions could only be coded according to the labels
provided in the primary studies.
There are additional theoretical moderators that warrant further
investigation. For example, groups vary in terms of climate
strength, or how much group members agree about safety climate
(Schneider, Salvaggio, & Subirats, 2002). Stronger climates are
more indicative of group behavior than are weaker climates, so
strong safety climates should have larger associations with safety-
related outcomes than do weak safety climates. However, such
analyses could not be performed here, as a consistent measure of
the variability of group safety climate perceptions across studies is
required; for group-level safety climate studies, this necessi-
tates information regarding the variability of safety climate
perceptions within each examined group. A related moderator is
that groups also vary in terms of the interdependence of their
work. Safety climates in highly interdependent groups would be
expected to relate more strongly to subsequent safety outcomes
than would safety climates in groups with low interdependence.
A favorable safety climate would be critical for groups in which
individual safety is dependent on the actions and perceptions of
others. Thus, future research is needed to examine group inter-
dependence and climate strength as moderators of safety
climate–injury relationships.
Practical Implications
There are a number of notable implications for managers and
safety practitioners that can be gleaned from this study’s findings.
The role of perceived management commitment to safety in re-
ducing future occupational injuries is of particular importance.
Employee perceptions of the extent to which managers and super-
visors are committed to workplace safety likely influence em-
ployee safety behavior and, subsequently, injuries. This has spe-
cific implications for manager and supervisor safety training.
Managerial safety training that emphasizes the manager’s or su-
pervisor’s role as a safety referent could increase awareness of the
extent to which employees scrutinize managerial actions regarding
safety. Commitment to safety can be communicated through words
or actions. Employees’ perceptions of management’s commitment
to safety, regardless of how it is conveyed, clearly play an impor-
tant role in safety climate perceptions.
Our finding that the length of time over which injuries were
assessed did not moderate the effect of injuries on subsequent
safety climate perceptions is also meaningful. This suggests that
past injuries—arguably those of greater severity—tend to influ-
ence employee interpretations of organizational safety even well
after actual injury occurrences. For managers and safety practitio-
ners, this suggests the need to investigate and address the factors
that have contributed to past injuries and then to clearly commu-
nicate to employees what has been done to decrease the likelihood
of their recurrence. This approach could potentially alleviate the
negative influence of past safety incidents on subsequent safety
climate perceptions. However, given that managers can only ad-
dress the problems of which they are aware, it is critical to
encourage timely and open reporting of workplace injuries and
accidents by employees. The meaningful negative relationship that
safety reporting demonstrated with consequent injuries in this
meta-analysis underscores this need.
Furthermore, the stronger negative relationship that injuries
demonstrated with subsequent organizational safety climates rel-
ative to psychological safety climates suggests that injuries have a
stronger effect on safety perceptions at the group level. Conse-
quently, organizational efforts to increase the extent to which
employees share safety climate perceptions may be particularly
effective in decreasing future injuries. For example, organizational
socialization tactics (e.g., new employee orientations, formal men-
toring, ongoing training) focused specifically on communicating
safety expectations and norms may be an effective means for
organizations and work groups to create greater homogeneity in
safety climate perceptions and, thus, to improve occupational
safety.
Many of the moderators investigated in this study are associated
with methodological decisions for those studying safety climate–
injury relationships; therefore, we offer a number of recommen-
dations for future primary studies and safety climate assessments
in the field. First, in cases where researchers and practitioners are
interested in how safety climate affects the occurrence of injuries,
prospective safety climate3injury relationships should be exam-
ined along with theoretical mediators (e.g., safety behavior), as this
allows more accurate causal inferences regarding the influence of
safety climate on injuries. Further, given that injuries were dem-
onstrated to affect safety climate, a given safety climate cannot be
presumed to have been the same phenomenon prior to the injuries
that preceded its assessment. Second, future research on safety
climate–injury relationships should specify theoretical reasons for
studying organizational or psychological safety climates and offer
conclusions that are consistent with the specified theoretical level
(Ostroff et al., 2003; Zohar, 2003). Third, there is a need for
developmental studies (e.g., Beck & Wilson, 2001) to better de-
termine how both safety climate and workplace injuries change,
how their relationships change over time, and what influences
those changes. Developmental designs would also enable research-
ers to examine both safety climate3injury and injury3safety
climate relationships, as well as to better understand what is
happening within time frames in which injuries are typically
assessed. Fourth, safety researchers and practitioners may want to
consider the moderators we tested when examining other outcomes
besides injuries. Examples include safety behavior, both compli-
ance and participation, and errors. Finally, we encourage critical
examination by safety researchers of safety climate assessment
724 BEUS, PAYNE, BERGMAN, AND ARTHUR

tools for deficiency (e.g., a lack of management commitment to
safety) and potential contaminants (e.g., inherent risk).
Conclusion
Our purpose in the present study was to meta-analytically ex-
amine a number of factors posited to influence the relationships
between safety climate and injuries. This study contributes to the
extant safety climate literature in several ways. First, this is the
only known attempt to differentiate meta-analytically between
organizational and psychological safety climates and to distinguish
between safety climate3injury and injury3safety climate rela-
tionships. Our study contributes to the safety climate literature by
elucidating the relationships that exist between safety climate and
injuries at the theoretically appropriate levels of analysis. It re-
vealed that injuries are more predictive of safety climate than
safety climate is of injuries and that the injury3safety climate
relationship is stronger for organizational climates than for psy-
chological climates. Second, this study represents the most exten-
sive examination to date of potential moderators of these relation-
ships. Of note, we found that organizational safety climate’s
influence on subsequent injuries is weakened as the time frames
over which injuries are assessed increase. Further, this study
represents the first attempt to quantitatively assess the effect of
safety climate content contamination and deficiency on safety
climate–injury relationships. Whereas contamination tended to
inflate the relationships between safety climate and injuries, defi-
ciency tended to attenuate them. Given the number of safety
climate measures in existence and the extensive content variability
among them, these findings suggest that greater attention should be
paid to the operationalization of safety climate. That is, we rec-
ommend that safety climate researchers and practitioners select or
create measures that are consistent with the extant theoretical
conceptualization of safety climate. We hope the findings of this
study will facilitate greater convergence in future safety climate
assessments as well as inspire additional research that will enrich
our understanding of the relationships between safety climate and
injuries, so that organizations can ultimately become safer places
for all workers.
References
References marked with an asterisk indicate studies included in the
meta-analysis.
*Alvarado, C. J. (2003). Safety climate in a university physical plant and
its relationship to self-reported injury (Unpublished doctoral disserta-
tion). University of Wisconsin—Madison.
Ancona, D. G., Goodman, P. S., Lawrence, B. S., & Tushman, M. L.
(2001). Time: A new research lens. Academy of Management Review,
26, 645– 663.
*Baas, J. (2002). An exploratory study of the role of trust in safety climates
and overall safety (Unpublished doctoral dissertation). Alliant Interna-
tional University, Los Angeles.
*Barling, J., Loughlin, C., & Kelloway, K. E. (2002). Development and
test of a model linking safety-specific transformational leadership and
occupational safety. Journal of Applied Psychology, 87, 488 – 496.
Beck, K., & Wilson, C. (2001). Have we studied, should we study, and can
we study the development of commitment? Methodological issues and
the developmental study of work-related commitment. Human Resource
Management Review, 11, 257–278.
Bitektine, A. (2008). Prospective case study design: Qualitative method for
deductive theory testing. Organizational Research Methods, 11, 160 –
180.
Bureau of Labor Statistics. (2008). Workplace injuries and illnesses in
2007. Retrieved from http://www.bls.gov/iif//oshwc/osh/os/
osnr0030.pdf
Christian, M. S., Bradley, J. C., Wallace, J. C., & Burke, M. J. (2009).
Workplace safety: A meta-analysis of the roles of person and situation
factors. Journal of Applied Psychology, 94, 1103–1127.
Clarke, S. (2006a). The relationship between safety climate and safety
performance: A meta-analytic review. Journal of Occupational Health
Psychology, 11, 315–327.
*Clarke, S. (2006b). Safety climate in an automobile manufacturing plant:
The effects of work environment, job communication, and safety atti-
tudes on injuries and unsafe behavior. Personnel Review, 35, 413– 430.
Cook, T. D., & Campbell, D. T. (1979). Quasi-experimentation: Design
and analysis for field settings. Chicago, IL: Rand McNally.
*Cree, T., & Kelloway, E. K. (1997). Responses to occupational hazards:
Exit and participation. Journal of Occupational Health Psychology, 2,
304 –311.
Crowl, D. A., & Louvar, J. F. (2002). Chemical process safety: Funda-
mentals with applications (2nd ed.). Upper Saddle River, NJ: Prentice
Hall.
Denison, D. R. (1996). What is the difference between organizational
culture and organizational climate? A native’s point of view on a decade
of paradigm wars. Academy of Management Review, 21, 619 – 654.
Denrell, J., & Kovacs, B. (2008). Selective sampling of empirical settings
in organizational studies. Administrative Science Quarterly, 53, 109 –
144.
Donald, I., & Canter, D. (1994). Employee attitudes and safety in the
chemical industry. Journal of Loss Prevention in the Process Industries,
7, 203–208.
*Evans, D. D., Michael, J. H., Wiedenbeck, J. K., & Ray, C. D. (2004).
Relationships between organizational climates and safety-related events
at four wood manufacturers. Forest Products Journal, 55, 23–28.
*Felknor, S. A., Aday, L. A., Burau, K. D., Delclos, G. L., & Kapadia,
A. S. (2000). Safety climate and its association with injuries and safety
practices in public hospitals in Costa Rica. International Journal of
Occupational and Environmental Health, 6, 18 –25.
Flin, R., Mearns, K., O’Connor, P., & Bryden, R. (2000). Measuring safety
climate: Identifying the common features. Safety Science, 34, 177–192.
Garavan, T. N., & O’Brien, F. (2001). An investigation into the relation-
ship between safety climate and safety behaviors in Irish organizations.
Irish Journal of Management, 22, 141–170.
George, J. M., & Jones, G. R. (2000). The role of time in theory and theory
building. Journal of Management, 26, 657– 684.
*Goldenhar, L. M., Williams, L. J., & Swanson, N. G. (2003). Modelling
relationships between job stressors and injury and near-miss outcomes
for construction labourers. Work & Stress, 17, 218 –240.
*Grosch, J. W., Gershon, R. R. M., Murphy, L. R., & DeJoy, D. M. (1999).
Safety climate dimensions associated with occupational exposure to
blood-borne pathogens in nurses. American Journal of Industrial Med-
icine, 36(Suppl. 1), 122–124.
Guldenmund, F. W. (2000). The nature of safety culture: A review of
theory and research. Safety Science, 34, 215–257.
Guzzo, R. A., & Noonan, K. A. (1994). Human resource practices as
communications and the psychological contract. Human Resource Man-
agement, 33, 447– 462.
*Gyekye, S. A. (2005). Workers’ perceptions of workplace safety and job
satisfaction. International Journal of Occupational Safety and Ergonom-
ics, 11, 291–302.
*Hahn, S. E., & Murphy, L. R. (2008). A short scale for measuring safety
climate. Safety Science, 46, 1047–1066.
Harrison, D. A., & Hulin, C. L. (1989). Investigations of absenteeism:
725
SAFETY CLIMATE–INJURY META-ANALYSIS

Using event history models to study the absence-taking process. Journal
of Applied Psychology, 74, 300 –316.
*Hayes, B. E., Perander, J., Smecko, T., & Trask, T. (1998). Measuring
perceptions of workplace safety: Development and validation of the
Workplace Safety Scale. Journal of Safety Research, 29, 145–161.
Heinrich, H. W. (1931). Industrial accident prevention: A scientific ap-
proach. New York, NY: McGraw-Hill.
Hellreigel, D., & Slocum, J. W., Jr. (1974). Organizational climate: Mea-
sures, research, and contingencies. Academy of Management Journal,
19, 255–280.
*Hofmann, D. A., & Mark, B. (2006). An investigation of the relationship
between safety climate and medication error as well as other nurse and
patient outcomes. Personnel Psychology, 59, 847– 869.
Hofmann, D. A., & Morgeson, F. P. (1999). Safety-related behavior as a
social exchange: The role of perceived organizational support and
leader–member exchange. Journal of Applied Psychology, 84, 286 –296.
*Hofmann, D. A., & Stetzer, A. (1996). A cross-level investigation of
factors influencing unsafe behaviors and injuries. Personnel Psychology,
49, 307–339.
*Huang, Y., Ho, M., Smith, G. S., & Chen, P. Y. (2006). Safety climate
and self-reported injury: Assessing the mediating role of employee
safety control. Injury Analysis and Prevention, 38, 425– 433.
Hulin, C. L., & Rousseau, D. M. (1980). Analyzing infrequent events:
Once you find them, your troubles begin. In K. H. Roberts & L. Burstein
(Ed.), Issues in aggregation: New directions for methodology and be-
havioral science (Vol. 6, pp. 65–75). San Francisco, CA: Jossey-Bass.
Hunter, J. E., & Schmidt, F. L. (2004). Methods of meta-analysis: Cor-
recting error and bias in research findings (2nd ed.). Thousand Oaks,
CA: Sage.
Jacobs, H. H. (1970). Towards more effective safety measurement. Journal
of Safety Research, 2, 160 –175.
James, L., & Jones, A. (1974). Organizational climate: A review of theory
and research. Psychological Bulletin, 81, 1096 –1112.
*Johnson, S. E. (2007). The predictive validity of safety climate. Journal
of Safety Research, 38, 511–521.
*Katz-Navon, T., Naveh, E., & Stern, Z. (2005). Safety climate in health
care organizations: A multidimensional approach. Academy of Manage-
ment Journal, 48, 1075–1089.
*Kelloway, E. K., Mullen, J., & Francis, L. (2006). Divergent effects of
transformational and passive leadership on employee safety. Journal of
Occupational Health Psychology, 11, 76 – 86.
Komaki, J., Barwick, K. D., & Scott, L. R. (1978). A behavioral approach
to occupational safety: Pinpointing and reinforcing safe performance in
a food manufacturing plant. Journal of Applied Psychology, 63, 434 –
445.
Kozlowski, S. W. J., & Klein, K. J. (2000). A multilevel approach to theory
and research in organizations: Contextual, temporal and emergent pro-
cesses. In K. J. Klein & S. W. Kozlowski (Eds.), Multilevel theory,
research, and methods in organizations (pp. 3–90). San Francisco, CA:
Jossey-Bass.
*Krispin, J. (1997). The construction and validation of a measure of safety
climate: Exploring the link between attitudes and perceptions around
safety and safe performance (Unpublished doctoral dissertation). Tem-
ple University.
*Lyon, J. S. (2007). The missing link: An examination of safety climate and
patient clinical outcomes in a national sample of hospitals (Unpublished
doctoral dissertation). University of Maryland.
*Mearns, K., Flin, R., Gordon, R., & Fleming, M. (1998). Measuring safety
climate on offshore installations. Work & Stress, 12, 238 –254.
*Mearns, K., Whitaker, S. M., & Flin, R. (2003). Safety climate, safety
management practice and safety performance in offshore environments.
Safety Science, 41, 641– 680.
Messick, S. (1980). Test validity and the ethics of assessment. American
Psychologist, 35, 1012–1027.
Messick, S. (1995). Validity of psychological assessment: Validation of
inferences from persons’ responses and performances as scientific in-
quiry into score meaning. American Psychologist, 50, 741–749.
*Michael, J. H., Evans, D. D., Jansen, K. J., & Haight, J. M. (2005).
Management commitment to safety as organizational support: Relation-
ships with non-safety outcomes in wood manufacturing employees.
Journal of Safety Research, 36, 171–179.
Mitchell, T. R., & James, L. R. (2001). Building better theory: Time and
the specification of when things happen. Academy of Management
Review, 26, 530 –547.
*Naveh, E., Katz-Navon, T., & Stern, Z. (2005). Treatment errors in health-
care: A safety climate approach. Management Science, 51, 948 –960.
Neal, A., & Griffin, M. A. (2004). Safety climate and safety at work. In J.
Barling & M. R. Frone (Ed.), The psychology of workplace safety (pp.
15–34). Washington, DC: American Psychological Association.
*Neal, A., & Griffin, M. A. (2006). A study of the lagged relationships
among safety climate, safety motivation, safety behavior, and injuries at
the individual and group levels. Journal of Applied Psychology, 91,
946 –953.
Nunnally, J. C., & Bernstein, I. H. (1994). Psychometric theory (3rd ed.).
New York, NY: McGraw-Hill.
Occupational Safety and Health Administration. (2008). Recording and re-
porting occupational injuries and illness (Part No. 1904). Retrieved from
http://www.osha.gov/pls/oshaweb/owadisp.show_document?p_id⫽
16312&p_table⫽FEDERAL_REGISTER
Oliver, A., Cheyne, A., Tomas, J. M., & Cox, S. (2002). The effects of
organizational and individual factors on occupational injuries. Journal of
Occupational and Organizational Psychology, 75, 473– 488.
Ostroff, C., & Bowen, D. E. (2000). Moving HR to a higher level: HR
practices and organizational effectiveness. In K. J. Klein & S. W.
Kozlowski (Eds.), Multilevel theory, research, and methods in organi-
zations (pp. 211–266). San Francisco, CA: Jossey-Bass.
Ostroff, C., Kinicki, A. J., & Tamkins, M. M. (2003). Organizational
culture and climate. In W. C. Borman & D. R. Ilgen (Eds.), Handbook
of psychology: Industrial and organizational psychology (Vol. 12, pp.
565–593). New York, NY: Wiley.
*Payne, S. C., & Bergman, M. E. (2008, September). Leading and lagging:
The safety climate–injury relationship. Paper presented at the annual
Power Switching Safety and Reliability Conference of the Electric
Power Research Institute, San Antonio, TX.
*Probst, T. M. (2004). Safety and insecurity: Exploring the moderating
effect of organizational safety climate. Journal of Occupational Health
Psychology, 9, 3–10.
*Probst, T. M., & Estrada, A. X. (2008, April). Injury under-reporting: The
moderating effects of organizational safety climate. In S. C. Payne & J. M.
Rodriguez (Chairs), Safety in organizations: Moderators and mediators of
safety climate. Symposium conducted at the meeting of the Society for
Industrial and Organizational Psychology, San Francisco, CA.
Reichers, A. E., & Schneider, B. (1990). Climate and culture: An evolution
of constructs. In B. Schneider (Ed.), Organizational climate and culture
(pp. 5–39). San Francisco, CA: Jossey-Bass.
*Ringenbach, K. L., & Jacobs, R. R. (1995). Injuries and aging workers.
Journal of Safety Research, 26, 169 –176.
Rousseau, D. M., & Wade-Benzoni, K. A. (1994). Linking strategy and
human resource practices: How employee and customer contracts are
created. Human Resource Management, 33, 463– 489.
Schneider, B., & Reichers, A. E. (1983). On the etiology of climates.
Personnel Psychology, 36, 19 –39.
Schneider, B., Salvaggio, A. N., & Subirats, M. (2002). Climate strength:
A new direction for climate research. Journal of Applied Psychology, 87,
220 –229.
*Seo, D., Torabi, M. R., Blair, E. H., & Ellis, N. T. (2004). A cross-
validation of safety climate scale using confirmatory factor analytic
approach. Journal of Safety Research, 35, 427– 445.
726 BEUS, PAYNE, BERGMAN, AND ARTHUR

*Sinclair, R. R., Martin, J. E., & Sears, L. E. (2008, April). Retail
employees’ perceived safety climate and hazard exposure outcomes. In
J. W. Grosch (Chair), Occupation/industry focused studies of safety
climate. Symposium conducted at the meeting of the Society for Indus-
trial and Organizational Psychology, San Francisco, CA.
*Siu, O., Phillips, D. R., & Leung, T. (2003). Age differences in safety
attitudes and safety performance in Hong Kong construction workers.
Journal of Safety Research, 34, 199 –205.
*Smith, G. S., Huang, Y., Ho, M., & Chen, P. Y. (2006). The relationship
between safety climate and injury rates across industries: The need to
adjust for injury hazards. Injury Analysis & Prevention, 38, 556 –562.
*Snyder, L. A., Krauss, A. D., Chen, P. Y., Finlinson, S., & Huang, Y.
(2008). Occupational safety: Application of the job demand– control–
support model. Accident Analysis and Prevention, 40, 1713–1723.
Spence, A. (1973). Job market signaling. Quarterly Journal of Economics,
87, 355–379.
Steel, P. D., & Kammeyer-Mueller, J. D. (2002). Comparing meta-analytic
moderator estimation techniques under realistic conditions. Journal of
Applied Psychology, 87, 96 –111.
U.S. Government Accountability Office. (2009). Workplace safety and
health (GAO-10 –10). Washington, DC: Author.
Varonen, U., & Mattila, M. (2000). The safety climate and its relationship
to safety practices, safety of the work environment and occupational
injuries in eight wood-processing companies. Injury Analysis & Preven-
tion, 32, 761–769.
*Vinodkumar, M. N., & Bhasi, M. (2009). Safety climate factors and its
relationship with accidents and personal attributes in the chemical in-
dustry. Safety Science, 47, 659 – 667.
*Vitro, T. M. (1991). Organizational safety climate: Its variation within a
single organization and its relationship to traditional indices of safety
(Unpublished doctoral dissertation). University of Tulsa.
*Williamson, A. M., Feyer, A., Cairns, D., & Biancotti, D. (1997). The
development of a measure of safety climate: The role of safety percep-
tions and attitudes. Safety Science, 25, 15–27.
*Wu, T., Liu, C., & Lu, M. (2007). Safety climate in university and college
laboratories: Impact of organizational and individual factors. Journal of
Safety Research, 38, 91–102.
*Zacharatos, A., Barling, J., & Iverson, R. D. (2005). High-performance
work systems and occupational safety. Journal of Applied Psychology,
90, 77–93.
*Zohar, D. (2000). A group-level model of safety climate: Testing the
effect of group climate on microinjuries in manufacturing jobs. Journal
of Applied Psychology, 85, 587–596.
*Zohar, D. (2002). The effects of leadership dimensions, safety climate,
and assigned priorities on minor injuries in work groups. Journal of
Organizational Behavior, 23, 75–92.
Zohar, D. (2003). Safety climate: Conceptual and measurement issues. In
J. C. Quick & L. E. Tetrick (Eds.), Handbook of occupational health
psychology (pp. 123–142). Washington, DC: American Psychological
Association.
*Zohar, D., & Luria, G. (2004). Climate as a social– cognitive construction
of safety practices: Scripts as proxy of behavior patterns. Journal of
Applied Psychology, 89, 322–333.
Zohar, D., & Luria, G. (2005). A multilevel model of safety climate:
Cross-level relationships between organization and group-level climates.
Journal of Applied Psychology, 90, 616 – 628.
Appendix
Safety Climate Scale Rating Questionnaire
The purpose of this questionnaire is to rate a set of safety climate
measures according to the degree to which they represent a commonly
held definition of safety climate. Specifically, safety climate measures
will be rated according to their level of deficiency and contamination
in relation to the specified safety climate construct. Deficiency is the
degree to which a given scale fails to represent the specified content
domain (Messick, 1980) and will be rated on a seven-point Likert
scale. Contamination is the degree to which a scale measures con-
struct irrelevant content (Messick, 1980) and will be operationalized
as the proportion of contaminated items within a measure. The defi-
nition of safety climate for the purpose of gauging deficiency and
contamination appears below.
Definition of Safety Climate
Zohar (2003) defined safety climate as the perception of the
policies, practices, and procedures pertaining to safety. Accord-
ing to Zohar and Luria (2005), safety policies define strategic
safety goals and the means for their achievement while safety
procedures provide planned courses of action relating to those
goals. Safety policies and procedures both exist at the organi-
zational level and are maintained by upper management (Zohar
& Luria, 2005). Safety practices refers to the implementation of
policies and procedures at the work group level (Zohar & Luria,
2005).
Received May 20, 2009
Revision received January 22, 2010
Accepted January 28, 2010 䡲
727
SAFETY CLIMATE–INJURY META-ANALYSIS