Accident Analysis and Prevention 134 (2020) 105350

Contents lists available at ScienceDirect

Accident Analysis and Prevention

journal homepage: www.elsevier.com/locate/aap

A dual perspective on risk perception and its effect on safety behavior: A T

moderated mediation model of safety motivation, and supervisor’s and
coworkers’ safety climate
Nini Xiaa,*, Qiuhao Xieb, Xiaowen Huc, Xueqing Wangd, Hao Menge
a
Department of Construction and Real Estate, School of Civil Engineering, Southeast University, No. 2 Southeast Road, Jiangning District, Nanjing 211189, China
b
Business School, Hohai University, No. 8 West Focheng Road, Jiangning District, Nanjing 211100, China
c
Queensland University of Technology, GPO Box 2434, Brisbane, QLD 4001, Queensland, Australia
d
College of Management and Economics, Tianjin University, No. 92 West Weijin Road, Nankai District, Tianjin 300072, China
e
Bank of Tianjin Co., Ltd., No.15 Youyi Road, Hexi District, Tianjin 300202, China

A R T I C LE I N FO A B S T R A C T

Keywords: Previous studies have acknowledged the impact of risk perception on safety behavior, but were largely con-
Risk perception troversial. This study aims to clarify this conflict and the mechanism through which risk perception can have an
Safety behavior impact on safety behavior. From the perspective of the dual attribute of the job demand concept in job de-
Job demands mands–resources theory, we posit that risk perception can be considered as a job hindrance or a job challenge
Safety motivation
depending on the context, thereby resulting in a negative or positive impact on safety behavior, respectively. The
Safety climate
current research context is the construction industry and the hypotheses were tested using hierarchically nested
Construction industry
data collected from 311 workers in 35 workgroups. Risk perception was demonstrated to be a job hindrance
exerting a negative impact on safety behavior and safety motivation mediated this effect. In addition, two di-
mensions of group-level safety climate––supervisor’s and coworkers’––were expected to alleviate or even reverse
the detrimental effects of hindrance risk perception on safety motivation and on safety behavior via motivation.
A moderation model and a first-stage moderated mediation model were established, respectively, for testing the
moderating roles of safety climate in the relationship between risk perception and safety motivation, and in the
indirect relationship of risk perception with safety behavior via motivation. Surprisingly, contrary to the hy-
potheses, when supervisor’s safety climate changed from a low level to a high level, the impact of risk perception
on safety motivation changed from positive to negative, and the negative effect of risk perception on safety
behavior via safety motivation was not alleviated but worsened. As expected, for workers in a positive cow-
orkers’ safety climate, the negative effect of risk perception on motivation and the indirect negative effect of risk
perception on behavior were both reversed to the positive. This indicates that coworkers’ safety climate helped
to change perceived risk from a job hindrance to a challenge. This research contributes to workplace risk per-
ception and safety behavior research by theoretically viewing risk perception as a dual job hindrance–challenge
concept and proposing two competing hypotheses concerning the impact of risk perception on safety behavior.
The empirical investigation confirmed the hindrance attribute of risk perception in the construction context. It
provides a theoretical framework and empirical evidence for future research to synthesize the conflict risk
perception–safety behavior relationship. We also contribute to the literature by pointing out the potential ne-
gative role of certain supervisor safety activities such as paternalistic leadership in influencing employee safety.

1. Introduction occupational accident or disease (International Labour Organization,

2018). Safety accidents in various industries not only cause personal
Injuries and accidents in the workplace remain a serious safety injuries, but also great economic losses (Beus et al., 2015; Hofmann
concern worldwide (Christian et al., 2009; Li et al., 2013; Nahrgang et al., 2003; Zou et al., 2014). Unsafe acts on the part of employees are
et al., 2011). Every day, approximately 7600 people die from an the main cause of industrial accidents (Heinrich et al., 1980). Haslam

⁎
Corresponding author.
E-mail addresses: ninixia@seu.edu.cn (N. Xia), xieqiuhao@hhu.edu.cn (Q. Xie), xiaowen.hu@qut.edu.au (X. Hu), wxqlab@126.com (X. Wang),
menghao@bankoftianjin.com (H. Meng).

https://doi.org/10.1016/j.aap.2019.105350
Received 3 February 2019; Received in revised form 28 August 2019; Accepted 25 October 2019
0001-4575/ © 2019 Elsevier Ltd. All rights reserved.
N. Xia, et al.
et al. (2005) analyzed 100 accidents in the construction industry and
found that 70 % of them were related to human factors, especially the
unsafe behavior of employees. Thus, reduction in the occurrence of
occupational injuries and accidents largely depends on improvement in
employees’ safety behavior.
Employees, as front-line personnel in the organization, are directly
exposed to workplace risks, injuries, and accidents. If they perceive
their work to be at a high level of risk, they are likely to behave safely to
avoid risk to their own safety (Didla et al., 2009). If they do not behave
safely, they increase their likelihood of suffering workplace injuries and
accidents (Christian et al., 2009). Some studies confirmed the positive
relationship between risk perception and safety behavior (e.g.,
Kouabenan et al., 2015; Seo, 2005; Tomás et al., 1999; Xia et al., 2017).
However, the current findings concerning the impact of risk perception
on safety behavior are still contradictory. For example, Nahrgang et al.
(2011) considered perceived level of workplace risks and hazards as
one type of job hindrance demands (i.e., stresses), which requires sus-
tained physical and/or mental effort and thereby leaves limited per-
sonal resources for handling safety behavior (Bakker and Demerouti,
2017; Hobfoll, 2001). Thus, additional research is warranted for clar-
ifying the impact of risk perception on safety behavior.
We argue that the conflict between risk perception and safety be-
havior can be explained by drawing on the dual attribute of the job
demand concept in job demands–resources (JD–R) theory (Bakker and
Demerouti, 2017). Nahrgang et al. (2011) considered perceived level of
workplace risks as a job hindrance demand; however, recent develop-
ments in JD–R theory have found that job hindrances can also be ex-
perienced as job challenges. Job challenges can cost effort but si-
multaneously can contribute to goal achievement or learning
(Breevaart and Bakker, 2017). Studies have found that a job hindrance,
such as work pressure experienced in one occupation, can be considered
as a challenge in other occupations (Bakker and Demerouti, 2017;
Bakker and Sanz-Vergel, 2013). Whether a job demand can be experi-
enced as a job hindrance or a job challenge depends on the context.
Using this lens, we posit that perceived risk may be considered as a job
hindrance and one that inhibits employees’ safety behavior, while in
other contexts it may serve as a job challenge and thus elicits em-
ployees’ safety behavior for the achievement of safety needs. Basically,
the hindrance or challenge attribute of perceived risk, as well as the
negative or positive relationship between risk perception and safety
behavior are empirical questions that depend on the context.
This dual hindrance and challenge job demand perspective on the
concept of risk perception can provide a theoretical framework for
synthesis of the current conflict in the risk perception–safety behavior
relationship. The construction industry is characterized by high risk
(Barbaranelli et al., 2015; Fang and Wu, 2013; Murie, 2007; Yu et al.,
2014), and the unsafe behavior of workers appears to be a critical factor
in a large proportion of construction accidents (Fogarty and Shaw,
2010; Xia et al., 2017). Thus, analysis of a sample of construction
workers should add empirical evidence to the existing risk percep-
tion–safety behavior relationship research. Furthermore, this study will
explore the moderated mediation effect of individual safety motivation
and group-level supervisor’s and coworkers’ safety climate in the risk
perception–safety behavior relationship. This will contribute to re-
vealing the underlying mechanism through which risk perception can
impact safety behavior, as most previous studies have merely in-
vestigated the direct relationship between them (e.g., Seo, 2005; Tomás
et al., 1999; Xia et al., 2017). Furthermore, we distinguish the two
safety climate dimensions, which will help to examine whether they
play the same role in the risk perception–safety motivation–behavior
relationship. The entire conceptual model is shown in Fig. 1. Details of
the development of the hypotheses are presented in the following sec-
tion.
2
Accident Analysis and Prevention 134 (2020) 105350
2. Theoretical background and hypotheses
2.1. Job demands–resources theory and the link between risk perception and
safety behavior
Workplace risks can be defined as the extent to which individuals
are exposed to danger when working (Frone, 1998), and risk perception
refers to “individuals’ subjective judgment of the risk” (Aven and Renn,
2009, p. 1). Safety behaviors are those activities ensuring that in-
dividuals or the workplace are free from physical threat or harm (Beus
et al., 2015). Following job performance theory, Griffin and Neal
(2000) defined and differentiated two components of safety behavior:
safety compliance and safety participation. Safety compliance corre-
sponds to task performance and refers to “the core activities that in-
dividuals need to carry out to maintain workplace safety” (Griffin and
Neal, 2000, p. 349), such as adhering to safety norms. Safety partici-
pation corresponds to contextual performance and refers to “behaviors
that do not directly contribute to an individual’s personal safety but
that do help to develop an environment that supports safety” (Griffin
and Neal, 2000, p. 349), such as participating in voluntary programs for
improving workplace safety.
We propose that the conflict findings concerning the impact of risk
perception on safety behavior can be clarified by drawing on JD–R
theory, especially its perspective of hindrance and challenge demands.
Put forward in 2001, JD–R theory was frequently used for explaining
job stress (Demerouti et al., 2001). The first proposition of JD–R theory
was that all types of job characteristics can be classified into one of two
categories: job demands and job resources. Job demands are defined as
“those physical, psychological, social, or organizational aspects of the
job that require sustained physical and/or psychological effort and are
therefore associated with certain physiological and/or psychological
costs” (Demerouti et al., 2001, p. 501). They are the specific stresses
related to the work context (Ten Brummelhuis and Bakker, 2012). In
high-risk settings, examples of job demands include exposure to risks
and hazards, physical demands, and complexity of the work (Nahrgang
et al., 2011). Job demands as work stresses will reduce individuals’ job
performance (e.g., safety behavior) by exhausting personal resources or
energy for displaying organization-expected behaviors (Bakker and
Demerouti, 2017; Hobfoll, 2001).
JD–R theory has been developed for approximately 20 years, and
two further types of job demands have been distinguished: hindrance
job demands and challenge job demands. Hindrance job demands are
defined as “job demands or work circumstances that involve excessive
or undesirable constraints that interfere with or inhibit an individual’s
ability to achieve valued goals,” thereby exerting a negative impact on
work engagement and performance (Bakker and Demerouti, 2017, p.
277). Challenge job demands are defined as “demands that cost effort
but that potentially promote personal growth and achievement of the
employee,” exerting a positive impact on work engagement and per-
formance similar to job resources (Bakker and Demerouti, 2017, p.
277). Certain job demands are job hindrances such as role conflict while
others are job challenges such as high levels of responsibility. Fur-
thermore, studies have found that a job hindrance such as work pres-
sure experienced in one occupation can be considered as a challenge in
other occupations (Bakker and Demerouti, 2017; Bakker and Sanz-
Vergel, 2013). That is, job demands can be experienced as either a job
hindrance demand or job challenge demand depending on the context.
Drawing on the dual hindrance and challenge attribute of job de-
mands, we posit that employees’ risk perception varies as a job hin-
drance or challenge in different contexts, which then decreases or in-
creases safety behavior in different population samples from different
occupations, respectively. The hindrance or challenge attribute of risk
perception and its subsequent negative or positive relationship with
safety behavior is basically an empirical question. For the current re-
search sample, construction workers, we propose two competing hy-
potheses.
N. Xia, et al.
Fig. 1. Conceptual model.
Hypothesis 1. Risk perception will have a negative effect on safety
compliance (H1a) and safety participation (H1b).
Hypothesis 1’. Risk perception will have a positive effect on safety
compliance (H1a’) and safety participation (H1b’).
Safety motivation refers to an individual’s willingness to exert effort
to enact safety behaviors and the valence associated with those beha-
viors (Chen and Chen, 2014; Neal and Griffin, 2006; Vinodkumar and
Bhasi, 2010). Perceptions of workplace risks likely influence workers’
motivation to act safely first and then promote their safety behavior. It
can also be inferred that safety motivation plays a mediating role in the
impact of risk perception on safety behavior, given the confirmed job
demands–work engagement–job performance relationship in the JD–R
theory (Bakker and Demerouti, 2017, p. 277). While work engagement
is often defined as a generic positive motivation state of vigor, ded-
ication, and absorption toward work (Bakker et al., 2014), safety mo-
tivation refers to safety-specific motivation toward work (Neal and
Griffin, 2006). In the workplace safety research, safety motivation and
knowledge have been demonstrated to be the two determinants of
safety behavior, while other factors influence safety behavior via these
two determinants (Christian et al., 2009; Griffin and Neal, 2000).
Therefore, we propose the following hypotheses.
Hypothesis 2. Risk perception is negatively associated with safety
motivation.
Hypothesis 3. Safety motivation mediates the impacts of risk
perception on safety compliance (H3a) and safety participation (H3b).
2.2. The moderating role of safety climate
In high-risk settings, the safety climate is an important environ-
mental factor and has been widely confirmed to affect safety behavior
(Brondino et al., 2012; Christian et al., 2009; Zohar and Polachek,
2014). The safety climate can be defined as shared perceptions among
members of a group or organization concerning policies, procedures,
and practices related to safety in the workplace (Dov, 2008; Morrow
et al., 2010; Zohar, 1980). Brondino et al. (2012) proposed three di-
mensions of safety climate, namely, top management’s safety climate,
supervisor’s safety climate, and coworkers’ safety climate, reflecting the
actual priority given to safety by top management, supervisors, and
coworkers, respectively (Zohar, 2003). Owing to the short tenure
characteristic of workers on construction sites, complex subcontracting,
and the temporality of projects in the construction industry, construc-
tion workers are prone to disconnect with top management (Schwatka
and Rosecrance, 2016). In reality, construction workers tend to work in
groups and interact frequently with their supervisors and coworkers
within the group. Research has found significant effects of both su-
pervisor’s and coworkers’ safety climate on construction workers’ safety
behavior (Fang et al., 2015; Liang et al., 2018). To capture the parti-
cularity of the setting, we consider supervisor’s and coworkers’ safety
climate as a group-level moderator in the relationship between risk
perception and safety behavior.
3
Accident Analysis and Prevention 134 (2020) 105350
The moderating role of supervisor’s and coworkers’ safety climate
can be explained by JD–R theory. According to this theory, job re-
sources refer to “those physical, psychological, social, or organizational
aspects of the job that are functional in achieving work goals, reduce
job demands and the associated physiological and psychological costs,
or stimulate personal growth, learning, and development” (Bakker and
Demerouti, 2017, p. 274). One typical example of job resources in high-
risk settings is a positive safety climate, because it reduces job demands
and their associated physiological and psychological costs and moti-
vates the development of safety behavior expected, supported, and re-
warded by the organization (Nahrgang et al., 2011; Schaufeli et al.,
2009). JD–R theory proposes that job resources can buffer the negative
impact of job hindrance demands on work engagement and perfor-
mance (Bakker and Demerouti, 2017). That is, employees who have
access to sufficient resources can better cope with demanding situa-
tions. This is because job resources refill energetic resources that are
consumed when meeting job hindrances, by providing employees with
emotional support or new tools to cope with stresses at work (Bakker
and Demerouti, 2014). In the context of workplace safety, Bronkhorst
(2015) focused on health care employees and revealed that the safety
climate buffers the negative impact of a job hindrance demand (i.e., job
insecurity) on safety behavior. Similarly, Kao et al. (2016) con-
ceptualized supervisor safety priority as a job resource and found that it
can alleviate the deleterious impact of insomnia on safety behavior and
injuries.
Furthermore, according to the standpoint of conservation of re-
sources theory (COR) (Hobfoll, 1989, 2002), both supervisor’s and
coworkers’ safety climate are a positive interpersonal activity related to
safety, which is likely to generate psychological resources and positive
affect by fulfilling human needs such as relatedness and competence
(Koopman et al., 2016; Weinstein and Ryan, 2010). As a positive event,
supervisor’s and coworkers’ safety climate can also reinforce self-eva-
luations (Williamson and Clark, 1989). These increased relatedness,
competence, self-evaluations, and associated psychological resources
and positive affect may also help workers to better cope with workplace
risks acting as job hindrances, or to change their perceptions of work-
place risks from negative (a job hindrance) to positive (a job challenge),
thereby motivating them to demonstrate safety behavior. Empirical
research has found that demands are sometimes considered as both
hindering and challenging (Searle and Auton, 2015).
For simplicity and clarity, we assume that there is a negative re-
lationship between risk perception and safety behavior for construction
workers. In line with the proposition in JD–R theory and the empirical
findings, we conceptualize supervisor’s and coworkers’ safety climate as
a job resource that will moderate the impact of risk perception as a job
hindrance demand on safety behavior. When the safety climate is po-
sitive, employees are in a workplace where safety policies, procedures,
and practices are clearly established and supervisors and coworkers
show concern for safety. In this situation, workers are likely to be better
at dealing with contextual hindrance demands, as the emotional and/or
behavioral safety support provided by supervisors and coworkers help
to replenish effort lost when meeting job hindrances (Ten Brummelhuis
N. Xia, et al. Accident Analysis and Prevention 134 (2020) 105350

and Bakker, 2012). That is, job resources relating to a positive safety completing questionnaires and to respond to questions. At the begin-
climate can counter the tendencies of workers to demonstrate unsafe ning of the survey, the participants were informed that their responses
behavior by perceiving workplace risks as a job hindrance. By further were anonymous, that there were no right or wrong answers, and that
considering the standpoint of COR and the finding in Searle and Auton they should answer questions as honestly as possible. All of the parti-
(2015), we expect that the negative effect of risk perception on safety cipants were assured that their responses would remain confidential
motivation and behavior can be alleviated or even reversed by a posi- and be kept by the authors for research purposes only.
tive safety climate. On the other hand, when the safety climate in the A total of 490 copies of the questionnaire were distributed. Because
workplace is poor, there is a lack of job resources supporting safety the survey was held on site, all questionnaires were returned. After
behavior, probably manifested by ambiguous safety policies, proce- checking the answers, 179 questionnaires were deemed unusable or
dures, and practices, supervisors’ insufficient concern for safety, and careless and thus removed from future use. To ensure the quality of our
coworkers’ demonstration of unsafe conduct. Lacking job resources, data, we applied three criteria to remove unusable or careless response
workers are likely to suffer from resource loss spirals, becoming more data: (a) multiple occurrence of two options were chosen for one item
sensitive to resource loss and tending to perceive workplace risks as (N = 50); (b) questionnaires were finished with more than 5 % missing
hindrances consuming their personal resources (Halbesleben et al., items, for example, one respondent gave no response to items on one
2014), which consequently reduces their safety motivation and then entire page (N = 36; this criterion was used by Seo (2005)); and (c)
behavior (Bakker and Demerouti, 2014). Therefore, we expect that with no variations across negatively and positively worded items on a
supervisor’s and coworkers’ safety climate will moderate the direct personality measure, for example, one respondent provided the same
relationship of risk perception to safety motivation and the indirect response (e.g., “strongly agree”) to the item “extraverted, enthusiastic”
relationship of risk perception to safety behavior. and the item “reserved, quiet” (N = 93; this criterion was used by
Griffin and Hu (2013)).
Hypothesis 4a. Supervisor’s safety climate moderates the relationship
Our final sample for the analyses comprised 311 workers nested in
between risk perception and safety motivation. Specifically, when
35 workgroups, producing a valid response rate of 63.47 %. This rate
supervisor’s safety climate is positive, the negative effect of risk
was considered high as most questionnaire surveys in the construction
perception on safety motivation is weaker or may change to be
industry normally range from 20 % to 30 % (Akintoye, 2000). Such a
positive, compared to when supervisor’s safety climate is negative.
relative lower rate compared to other industries may be due to the fact
Hypothesis 4b. Coworkers’ safety climate moderates the relationship that our subjects are construction workers, the majority of whom were
between risk perception and safety motivation. Specifically, when of low educational status and might struggle to answer the questions
coworkers’ safety climate is positive, the negative effect of risk (Xia et al., 2017).
perception on safety motivation is weaker or may change to be Nonetheless, deleting data may create sampling bias. To reduce this
positive, compared to when coworkers’ safety climate is negative. concern, we further implemented the following strategies. First, we
conducted t-tests for comparing demographic variables of the 179 de-
Hypothesis 5. Supervisor’s safety climate moderates the indirect
leted respondents with those of the 311 final respondents. Results show
negative effects of risk perception on safety compliance (H5a) and
that demographics between the two groups, on the whole, did not show
participation (H5b) (via safety motivation). Specifically, supervisor’s
any significant difference: gender (t = .528, p = .598, n.s.), age (t =
safety climate is positive, the indirect negative relationship of risk
−1.303, p = .194, n.s.), work experience (t = −.752, p = .232, n.s.),
perception with safety behavior via safety motivation is weaker or may
and work position (t = .971, p = .332, n.s.). Second, we calculated the
change to be positive, compared to when supervisor’s safety climate is
binary correlations between the response rates at workgroups and all
negative.
the study variables and found none of the relationships was statistically
Hypothesis 6. Coworkers’ safety climate moderates the indirect significant at the .05 level. Based on these tests, it can be concluded that
negative effects of risk perception on safety compliance (H6a) and sampling bias should not be a problem in contaminating the results. The
participation (H6b) (via safety motivation). Specifically, when retained sample was eligible for further analyses. Details of the final
coworkers’ safety climate is positive, the indirect negative participants’ demographics are presented in Table 1.
relationship of risk perception with safety behavior via safety
motivation is weaker or may change to be positive, compared to
when coworkers’ safety climate is negative. Table 1
Final participant demographics (N = 311).
Number Percent(%)
3. Methods
Gender
Male 286 6.7
3.1. Procedures and participants
Female 21 92.0
Missing 4 1.3
This study was supervised by an independent board at Southeast Age
University for the ethical review. Questionnaires were distributed on 15∼29 years 41 13.2
site in 35 construction projects in China, including housing projects, 30∼39 years 102 32.8
40∼49 years 121 38.9
subway projects, waterway projects, highway projects, etc. In one
50 years and over 43 13.8
project, we randomly chose one workgroup as a survey unit, and within Missing 4 1.3
each group 8–15 workers were randomly selected as participants. From Work experience
August 2017 to October 2017, 15 workgroups in 15 projects were 0∼5 years 56 18.0
6∼10 years 96 30.9
surveyed; from June 2019 to July 2019, another 20 workgroups in 20
11∼15 years 46 14.8
projects were surveyed. We approached the participants by contacting 16∼20 years 52 16.7
and getting permission from the project manager. Participation was 21 years and over 55 17.7
voluntary and the participants had the right to withdraw from the study Missing 6 1.9
at any time. Work position
Foreman 49 15.8
The participants completed the questionnaires during the lunch
Ordinary worker 256 82.3
break in the project meeting room. This process lasted for about 20 min, Missing 6 1.9
during which the authors participated to provide guidance on

4
N. Xia, et al.
3.2. Measures
3.2.1. Risk perception
Risk perception was assessed using nine items based on the scale
developed by Gyekye and Salminen (2005). One sample item was “I feel
that my work is very dangerous.” All items were rated on a 5-point
Likert scale, with 1 indicating “strongly disagree” and 5 indicating
“strongly agree.” The Cronbach’s alpha of the current sample was .931.
As indicated by Nunnally (1978), when Cronbach’s alpha is greater
than .70, the items for measuring a variable show an adequate relia-
bility.
3.2.2. Safety behavior
The two types of safety behavior, namely, safety compliance and
safety participation, were assessed using the scale developed by Neal
and Griffin (2006). Safety compliance was measured by four items; a
sample item was “I use all the necessary safety equipment to do my
job.” Safety participation was measured by three items; a sample item
was “I put in extra effort to improve the safety of the workplace.” Re-
sponses were recorded on a 5-point Likert scale, ranging from 1
(strongly disagree) to 5 (strongly agree). The Cronbach’s alphas of the
current sample were .869, and .809 for safety participation and safety
compliance, respectively.
3.2.3. Safety motivation
Safety motivation was assessed using five items based on the scale
developed by Vinodkumar and Bhasi (2010). A sample item was “I
think the safety of the workplace is very important.” All items were
rated on a 5-point Likert scale, with responses ranging from 1 (strongly
disagree) to 5 (strongly agree). The Cronbach’s alpha of the current
sample was .864.
3.2.4. Safety climate
We adopted the safety climate subscales developed by Brondino
et al. (2012) to measure the supervisor’s safety climate and coworkers’
safety climate. In each surveyed workgroup, one supervisor was as-
signed for the worker participants to assess the degree to which safety is
a priority of the workgroup’s supervisor. The supervisor’s safety climate
was measured by ten items; a sample item was “The supervisor cares for
safety rules when a delay in production occurs.” To assess the im-
portance given to safety by coworkers, coworkers’ safety climate was
measured by four dimensions, including safety communication, safety
mentoring, safety values, and safety systems. The four dimensions
contained 12 items; sample items were “Team members discuss incident
prevention” for safety communication, “Team members emphasize
safety care to peers when tired” for safety mentoring, “Team members
engage in safety care when tired” for safety values, and “Team members
care for other workers’ safety equipment” for safety systems. The re-
sponse coding used was a 5-point Likert scale, ranging from 1 (strongly
disagree) to 5 (strongly agree).
We operationalized group-level supervisor’s safety climate and
coworkers’ safety climate by aggregating the individual worker scores
to the group level and testing the within-group agreement. As suggested
by Sirotnik (1980), the internal consistency reliability was computed at
the group level. The Cronbach’s alphas of the current sample were .968
and .965 for supervisor’s safety climate and coworkers’ safety climate,
respectively.
3.2.5. Control variables
To maximize internal validity and rule out alternative explanations,
we controlled variables. We included individual characteristics of
workers: gender, age, work experience, and work position. Personality
traits are also important factors influencing employees’ safety motiva-
tion and behavior and the “big five” have been the most acknowledged
measures of personal traits (Beus et al., 2015). Therefore, a ten-term
personality inventory that was suitable for Chinese employees was
5
Accident Analysis and Prevention 134 (2020) 105350
adopted (Li, 2013). This scale comprised measures on five dimensions:
extraversion, agreeableness, conscientiousness, emotional stability, and
openness to experience. All personality items were rated on a 5-point
Likert scale, with responses ranging from 1 (very inaccurate) to 5 (very
accurate).
3.3. Analytic strategy
SPSS 19.0 was used to perform descriptive analyses, correlation
analyses, and reliability tests. Because the study variables were mea-
sured by workers’ self-reports, we assessed the construct validity of our
measures by examining structure validity, criterion-related validity,
convergent validity, and discriminant validity. Individual perceptions
of safety climate were aggregated to the group level; thus, we checked
the viability of the group-level variables by examining within-group
interrater reliability, intra-class correlation (ICC[1]), and reliability of
the mean (ICC[2]).
Because of the nested nature of the data, we estimated a null model
that had no predictors at either level 1 (the individual level) or level 2
(the group level) to partition the safety motivation variance into within-
and between-group components. The level 2 residual variance of the
intercept was .259 (p < .01). The ICC(1) was .59, indicating that 59%
of the variance in safety motivation resided between groups, and 41%
of the variance resided within groups. For safety compliance, the level 2
residual variance of the intercept was .273 (p < .01). The ICC(1) was
.58, indicating that 58% of the variance in safety motivation resided
between groups, and 42% of the variance resided within groups. For
safety participation, the level 2 residual variance of the intercept was
.319 (p < .01). The ICC(1) was .55, indicating that 55% of the var-
iance in safety motivation resided between groups, and 45% of the
variance resided within groups.
These results consistently indicated that ICC values were at accep-
table levels (Bliese, 2000; Cheung and Au, 2005) and that between-
groups’ variances in safety motivation, safety compliance, and safety
participation were significantly different from zero. Thus, hypotheses
were tested by means of hierarchical linear modeling (Raudenbush and
Bryk, 2002) using Mplus 7.11 (Muthén and Muthén, 2013). Tests of the
main, mediating, moderating, and moderated mediating effects are
described below.
3.3.1. Tests of main and mediation effects
Hierarchical linear models were established in Mplus 7.11 to test
the main effects, namely, the effects of risk perception on safety com-
pliance (H1a), safety participation (H1b), and safety motivation (H2),
as well as the mediation effects, namely, the mediating roles of safety
motivation in the effects of risk perception on safety compliance (H3a)
and safety participation (H3b). Moreover, to test the mediation effects,
we used a percentile bootstrap approach to obtain bias-corrected con-
fidence interval estimates (MacKinnon et al., 2004). The bootstrap
method can overcome the problems of non-normal sampling distribu-
tions of mediating effects estimates (Preacher and Hayes, 2004; Zhao
et al., 2010).
3.3.2. Tests of moderation effects
We tested the moderating effects of group-level supervisor’s safety
climate (H4a) and coworkers’ safety climate (H4b) in the relationship
between risk perception and safety motivation by establishing hier-
archical linear models in Mplus 7.11. For the current sample, the two
proposed moderators were correlated (r = .807, p < .01); thus, these
two moderators were included simultaneously in one moderation
model.
3.3.3. Tests of moderated mediation
To test H5 and H6 (i.e., the moderated mediation effects), we es-
tablished a two-level first-stage moderated mediation model (Bauer
et al., 2006) in Mplus 7.11 and used a bootstrap method to estimate the
N. Xia, et al. Accident Analysis and Prevention 134 (2020) 105350

Table 2
Descriptive statistics and correlations among variables.
Variables M SD 1 2 3 4 5 6 7 8 9 10 11 12

Individual level, construction worker

1. Safety compliance 4.027 .672 (.869)
2. Safety participation 3.841 .750 .667** (.809)
3. Risk perception 2.592 .859 −.234** −.234** (.931)
4. Safety motivation 4.053 .648 .649** .530** −.203** (.864)
5. Extraversion 3.471 .817 .045 .191** .025 .089 –
6. Agreeableness 2.500 .825 −.111 −.143* .091 −.075 .350** –
7. Conscientiousness 3.722 .724 .157** .256** −.128* .189** .071 −.308** –
8. Emotional stability 2.564 .819 −.116* −.159** .058 −.162** .219** .591** −.359** –
9. Openness to experience 3.259 .685 .182** .271** −.009 .234** .201** −.287** .458** −.263** –
10. Gender .930 .253 .046 .007 .070 .021 .142* .022 .022 .043 .167** –
11. Age 2.541 .893 −.046 −.040 −.031 −.086 −.025 −.041 .033 −.027 −.072 .033 –
12. Work experience 2.849 1.387 −.049 −.035 .019 −.058 .024 −.081 .053 −.086 .021 .158** .581** –
13. Work position 1.840 .368 −.049 −.125* .099 −.111 −.079 .113* −.198** .085 −.182** −.085 −.087 −.139*
Group level, workgroup
1. Supervisor’s safety climate 3.887 .609 (.968)
2. Coworkers’ safety climate 3.722 .588 .807** (.965)

Notes: N = 311 at the individual level; N = 35 at the group level. Gender: 0 = female, 1 = male. Age: 1 = 15∼29 years, 2 = 30∼39 years, 3 = 40∼49 years,
4 = 50 years and over. Work experience: 1 = 0∼5 years, 2 = 6∼10 years, 3 = 11∼15 years, 4 = 16∼20 years, 5 = 21 years and over. Work position: 1 = foreman,
2 = ordinary worker. Cronbach’s alphas are in the parentheses on the diagonal. *p < .05, **p < .01, two-tailed.

conditional indirect relationship of risk perception to safety behavior
via safety motivation at higher (+1 standard deviation) and lower le-
vels (-1 standard deviation) of safety climate as a moderator. Similar to
the model for testing H4, the two moderators were included simulta-
neously in one first-stage moderated mediation model.
4. Results
4.1. Descriptive analyses and correlations
Table 2 shows the means, standard deviations, and correlations
among all the variables. For the controls, safety compliance was related
to conscientiousness (r = .157, p < .01), emotional stability (r =
−.116, p < .05), and openness to experience (r = .182, p < .01);
safety participation was related to extraversion (r = .191, p < .01),
agreeableness (r = −.143, p < .05), conscientiousness (r = .256,
p < .01), emotional stability (r = −.159, p < .01), openness to
experience (r = .271, p < .01), and work position (r = −.125, p <
.05); safety motivation was related to conscientiousness (r = .189, p <
.01), emotional stability (r = −.162, p < .01), and openness to ex-
perience (r = .234, p < .01). In the tests of related hypotheses, ex-
traversion, agreeableness, conscientiousness, emotional stability,
openness to experience, and work position were included. Other con-
trols were not included in our analyses as their inclusion had no sub-
stantive effect on the results (Becker, 2005).
4.2. Validity of measures
We conducted the following analyses to assess the validity of our
measures in this data.
First, we conducted confirmatory factor analysis (CFA) to examine
the structure validity of the measures (DiStefano and Hess, 2005). As
shown in Table 3, our hypothesized six-factor model had a reasonable
fit to the data (Browne and Cudeck, 1992; Hu and Bentler, 1999).
Compared to alternative models, our baseline model provided sub-
stantial improvement in fit indexes. Thus, the structure validity of our
measures was demonstrated.
Second, we examined the criterion-related validity. As reported in
Table 2, at the individual level, safety compliance was significantly
correlated with safety participation (r =.667, p < .01), and safety
motivation was significantly correlated with safety compliance (r =
.649, p < .01) and participation (r =.530, p < .01); at the group
level, supervisors’ safety climate was significantly correlated with
coworkers’ safety climate (r =.807, p < .01). As these correlations
were consistent with theories about and empirical evidence on safety
behavior and safety climate (e.g., Brondino et al., 2012; Neal and
Griffin, 2006), the results provided criterion-related validity evidence
for the measures.
Third, the convergent validity of the measures was examined by
three indices suggested by Hair et al. (2009), including item reliability
(standardized factor loading, SFL), composite reliability (CR), and
average variance extracted (AVE). Table 4 shows the result of con-
vergent validity examination. The SFLs range (.528–.897) exceeded the
critical value of .50 suggested by Hair et al. (2009). The CRs range
(.752–.936) exceeded the critical value of .60 suggested by Fornell and
Larcker (1981). The AVEs range (.504–.617) was above the value of .50
suggested by Fornell and Larcker (1981). Thus, the convergent validity
of our measures was demonstrated. The values of CRs and AVEs also
indicated good reliabilities of the measures (Fornell and Larcker, 1981).
Fourth, the discriminant validity of the measures was assessed by
comparing the square root of a variable’s average variance extracted
(AVE) and correlations involving the variable (Fornell and Larcker,
1981). Results in Table 5 indicated that for all the individual variables,
the square root of AVE for safety compliance, safety participation, risk
perception, and safety motivation was .785, .710, .782, and .761,
greater than the correlations involving the variable. The correlation
between supervisor’s safety climate and coworkers’ safety climate
(.807), however, was greater than the square root of AVE for coworkers’
safety climate (.773) and supervisor’s safety climate (.744). This was
acceptable as the two dimensions of safety climate are distinct but
highly correlated (Brondino et al., 2012). Thus, discriminant validity of
measures was largely demonstrated.
Collectively, the above results demonstrated that the current mea-
sures had a six-factor structure, good criterion-related validity, and
reasonable convergent and discriminant validity, thus demonstrating
the validity of our measures.
4.3. Aggregation of group-level variables
For supervisor’s safety climate, the median value of within-group
interrater reliability (Rwg) was .98 (Rwg > .70; George and
Bettenhausen, 1990), intra-class correlation (ICC(1)) was .69 (ICC(1) ≥
.12; James et al., 1993), and ICC(2) was .95 (ICC(2) ≥ .60; Bartko,
1976). There was also significant between-group variance in super-
visor’s safety climate, F(34, 270) = 20.768, p < .001. For coworkers’
safety climate, the median value of Rwg was .98, ICC(1) was .68, and

6
N. Xia, et al. Accident Analysis and Prevention 134 (2020) 105350

Table 3
Confirmatory factor analysis for testing structure validity.
Model χ2 (df, p) Δχ2 CFI TLI SRMR RMSEA

Six factors (baseline model): safety compliance, safety participation, risk perception, safety motivation, 1798.016 .912 .908 .068 .065
supervisor’s safety climate, coworkers’ safety climate (803, p < .01)
Five factors: collapsing safety compliance and safety participation 2883.255 1085.239** .786 .773 .080 .091
(850, p < .01)
Five factors: collapsing supervisor’s safety climate and coworkers’ safety climate 3073.997 1275.981** .766 .752 .080 .095
(850, p < .01)
One factor: collapsing all the variables studied 5636.505 3838.489** .598 .473 .141 .138
(860, p < .01)

** p < .01, two-tailed.

Table 4 Table 5
Results of convergent validity examination. Results of discriminant validity examination.
Variable Indicator Standardized Composite Average Variables 1 2 3 4
factor loading reliability variance
(SFL) (CR) extracted Individual level, construction worker
(AVE) 1. Safety compliance (.785)
2. Safety participation .667** (.710)
Safety compliance SC1 .730 .866 .617 3. Risk perception −.234** −.234** (.782)
SC2 .820 4. Safety motivation .649** .530** −.203** (.761)
SC3 .790 Group level, workgroup
SC4 .800 1. Supervisor’s safety climate (.733)
Safety SP1 .791 .752 .504 2. Coworkers’ safety climate .807** (.744)
participation SP2 .653
SP3 .678 Notes: **p < .01, two-tailed. The square roots of AVEs are reported in par-
Risk perception PER1 .599 .933 .611 entheses along the diagonal of the correlation of variables.
PER2 .755
PER3 .883
PER4 .773 p < .05; see Model 1 (H1a)) and participation (B = −.214, SD = .061,
PER5 .727 p < .01; see Model 1 (H1b)), thereby supporting H1a and H1b, re-
PER6 .868 spectively. Accordingly, the competing hypothesis H1’ was rejected.
PER7 .779 Risk perception was significantly and negatively related to safety mo-
PER8 .812
tivation (B = −.187, SD = .080, p < .01; see Model 1 (H2)), thereby
PER9 .802
Safety motivation SM1 .770 .872 .579 supporting H2.
SM2 .842 For the mediation effects, after controlling for safety motivation as a
SM3 .710 mediator, the relationship between risk perception and safety com-
SM4 .686
pliance was reduced (B = −.115, SD = .071, p < .05; see Model 2
SM5 .785
Supervisor’s safety SSC1 .773 .920 .537 (H3a) in Table 6) and the relationship between risk perception and
climate SSC2 .735 safety participation was reduced (B = −.130, SD = .086, p < .05; see
SSC3 .780 Model 2 (H3b) in Table 6), providing evidence of a partial mediation. In
SSC4 .528 addition, if the 95% bias-corrected confidence interval of a mediation
SSC5 .805
effect excludes zero, the mediation effect is significant (Preacher et al.,
SSC6 .762
SSC7 .679 2004). The 95% bias-corrected confidence interval for the mediation
SSC8 .793 effect in H3a was (-.121, -.032) and was (-.134, -.029) for the mediation
SSC9 .713 effect in H3b. Thus, H3a and H3b were supported.
SSC10 .722
According to Table 7, the interaction effects of risk perception with
Coworkers’ safety CSC1 .732 .936 .554
climate CSC2 .748
supervisor’s safety climate (B = −1.851, SD = .658, p < .05) and
CSC3 .735 coworkers’ safety climate (B = 1.930, SD = .356, p < .05) on safety
CSC4 .580 motivation were both significant.
CSC5 .731 To further identify how each moderator moderates the relationship
CSC6 .830
between risk perception and safety motivation, we drew simple slopes
CSC7 .667
CSC8 .555 for the two moderating effects in Figs. 2 and 3. According to Fig. 2,
CSC9 .764 under a positive supervisor’s safety climate, risk perception was nega-
CSC10 .817 tively related to safety motivation; however, under a negative super-
CSC11 .897
visor’s safety climate, risk perception was positively related to safety
CSC12 .801
motivation. That is, although the moderating effect of supervisor’s
safety climate was significant, this effect was contrary to H4a (Super-
ICC(2) was .95. There was also significant between-group variance in visor’s safety climate moderates the relationship between risk percep-
coworkers’ safety climate, F(34, 276) = 9.551, p < .001. Hence, it tion and safety motivation. Specifically, when supervisor’s safety cli-
was appropriate to aggregate individual perceptions of supervisor’s mate is positive, the negative effect of risk perception on safety
safety climate and coworkers’ safety climate to the group level. motivation is weaker or may change to be positive, compared to when
supervisor’s safety climate is negative). Taken together, H4a was not
supported. According to Fig. 3, under a positive coworkers’ safety cli-
4.4. Hypotheses testing
mate, risk perception was positively related to safety motivation;
however, under a negative coworkers’ safety climate, risk perception
For the main effects, results in Table 6 show that risk perception was
was negatively related to safety motivation. Hence, this study
negatively related to safety compliance (B = −.167, SD = .042,

7
N. Xia, et al. Accident Analysis and Prevention 134 (2020) 105350

Table 6
Hierarchical linear modeling estimates for the relationships among risk perception, safety motivation, and safety behaviors (main and mediation effects).
Predictor Safety motivation Safety compliance Safety participation

Model 1 (H2) Model 1 (H1a) Model 2 (H3a) Model 1 (H1b) Model 2 (H3b)

Intercept 3.705*** (.769) 3.920*** (.368) 2.466** (1.170) 3.651*** (.425) 2.490*** (1.814)
Control variable
Extraversion .143 (.101) .140 (.107)
Agreeableness −.124 (.052) −.195 (.164)
Conscientiousness .042 (.104) .025 (.041) .044 (.166) .063 (.057) .055 (.211)
Emotional stability −.063 (.098) −.048 (.038) −.042 (.092) −.048 (.102) −.007 (.169)
Openness to experience .110* (.068) .060 (.075) −.017 (.296) .065 (.079) .076 (.083)
Work position .039 (.215) .047 (.137)
Direct effect
Risk perception −.187** (.080) −.167* (.042) −.115* (.071) −.214** (.061) −.130* (.086)
Safety motivation .283** (.034) .243** (.143)

Notes: N = 311 at the individual level; N = 35 at the group level. Entries are standardized estimates. Standard deviations are in parentheses. *p < .05, **p < .01,
***p < .001, two-tailed.

Table 7
Hierarchical linear modeling estimates for the moderating effects of super-
visor’s safety climate and coworkers’ safety climate in the relationship between
risk perception and safety motivation (moderation effects).
Predictor Safety motivation

B SD

Intercept 4.153* 1.754

Level 1 direct effect
Risk perception −.806 .527
Level 2 direct effect
Supervisor’s safety climate (clim_supervisor) 1.738** 1.237
Coworkers’ safety climate −1.968** 1.336
(clim_cowo)
Interaction effect
Risk perception × clim_supervisor −1.851* .658
Risk perception × clim_cowo 1.930* .356

Notes: N = 311 at the individual level; N = 35 at the group level. B represents Fig. 3. Moderating effect of coworkers’ safety climate on the relationship be-
standardized estimates. SD represents standard deviations. Controls were not tween risk perception and safety motivation. clim_cowo = coworkers’ safety
reported in this table. *p < .05, **p < .01, ***p < .001, two-tailed. climate.

perception to safety compliance via safety motivation at higher levels

Fig. 2. Moderating effect of supervisor’s safety climate on the relationship be-
tween risk perception and safety motivation. clim_supervisor = supervisor’s
safety climate.
supported H4b (Coworkers’ safety climate moderates the relationship
between risk perception and safety motivation. Specifically, when
coworkers’ safety climate is positive, the negative effect of risk per-
ception on safety motivation is weaker or may change to be positive,
compared to when the coworker safety climate is negative).
The estimate of the conditional indirect relationship of risk
(+1 standard deviation) was −1.044 (SD = .521, p < .05) and was
−.647 (SD = .336, p = .063) at lower levels (-1 standard deviation) of
supervisor’s safety climate as a group-level moderator. Moreover,
bootstrapping results revealed that the indirect effect from risk per-
ception to safety compliance (through safety motivation) was sig-
nificant at high levels (95% CI = [−2.193, −0.064]) but not sig-
nificant at low levels (95% CI [−1.212, 0.004]) of supervisor’s safety
climate. This means that at both lower and higher levels of supervisor’s
safety climate, the effect of risk perception on safety compliance via
safety motivation was negative, and the negative effect was significant
and greater at higher levels of supervisor’s safety climate. To sum-
marize, contrary to our hypothesis (H5a), supervisor’s safety climate
did not alleviate but worsened the negative effect of risk perception on
safety compliance.
Similar results were also obtained for the test of H5b: The estimate
of the conditional indirect relationship of risk perception to safety
participation via safety motivation at higher levels (+1 standard de-
viation) was −.563 (SD = .312, p < .05) and was −.321 (SD = .187,
p < .05) at lower levels (-1 standard deviation) of supervisor’s safety
climate as a group-level moderator. Moreover, bootstrapping results
revealed that the indirect effect from risk perception to participation
(through safety motivation) was significant at both high levels (95% CI
= [−1.341, −0.036]) and low levels (95% CI = [−0.876, −0.005])
of supervisor’s safety climate. This means that at both lower and higher
levels of supervisor’s safety climate, the effect of risk perception on

8
N. Xia, et al.
safety participation via safety motivation was significant and negative,
and the negative effect was greater at higher levels of supervisor’s
safety climate. That is, contrary to our hypothesis (H5b), supervisor’s
safety climate did not alleviate but worsened the negative effect of risk
perception on safety participation.
The estimate of the conditional indirect relationship of risk per-
ception to safety compliance via safety motivation at higher levels (+1
standard deviation) was .932 (SD = .516, p < .05) and was −.533
(SD = .321, p = .115) at lower levels (-1 standard deviation) of cow-
orkers’ safety climate as a group-level moderator. Moreover, boot-
strapping results revealed that the indirect effect from risk perception to
safety compliance (through safety motivation) was significant at high
levels (95% CI = [0.015, 2.155]) but not significant at low levels (95%
CI = [−0.008, 1.379]) of coworkers’ safety climate. This means that at
lower levels of a coworkers’ safety climate, the effect of risk perception
on safety compliance via safety motivation was negative; however, at
higher levels of coworkers’ safety climate, the effect of risk perception
on safety behavior via safety motivation was significantly positive.
Hence, as we expected in H6a, coworkers’ safety climate alleviated the
negative effect of risk perception on safety compliance via safety mo-
tivation: when coworkers’ safety climate was at higher levels, the
moderation effect was strong and reversed this indirect impact from
negative to positive.
Similar results were obtained for the test of H6b: The estimate of the
conditional indirect relationship of risk perception to safety participa-
tion via safety motivation at higher levels (+1 standard deviation) was
.497 (SD = .314, p < .05) and was −.302 (SD = .170, p = .108) at
lower levels (-1 standard deviation) of coworkers’ safety climate as a
group-level moderator. Moreover, bootstrapping results revealed that
the indirect effect from risk perception to safety participation (through
safety motivation) was significant at high levels (95% CI = [0.032,
1.289]) but not significant at low levels (95% CI = [−0.005, 0.785]) of
coworkers’ safety climate. This means that at lower levels of a cow-
orkers’ safety climate, the effect of risk perception on safety participa-
tion via safety motivation was negative; however, at higher levels of
coworkers’ safety climate, the effect of risk perception on safety beha-
vior via safety motivation was significantly positive. Thus, as we ex-
pected in H6b, coworkers’ safety climate indeed alleviated the negative
effect of risk perception on safety participation via safety motivation:
when coworkers’ safety climate was at higher levels, the moderating
effect of coworkers’ safety climate was strong and reversed this indirect
impact from negative to positive.
5. Discussion
In targeting construction workers, this study investigated the direct
effect of risk perception on safety behavior and the indirect effect from
risk perception to safety behavior via safety motivation. Moreover, we
explored whether two salient safety climate dimensions––supervisor’s
and coworkers’––in the construction industry moderated the relation-
ship between risk perception and safety motivation and the mediated
relationship. Our study is unique in that we drew on the dual nature of
the job demand concept, and proposed that risk perception can be
viewed as a job hindrance demand or a job challenge demand de-
pending on the context, which will have either a negative or positive
impact on safety behavior. Accordingly, two competing hypotheses
concerning the impact of risk perception on safety behavior were pro-
posed. For the construction workers studied, it was found that risk
perception was a job hindrance that reduced safety motivation and,
consequently, both safety compliance and participation behavior. We
expected that both supervisor’s and coworkers’ safety climate would
alleviate the negative impact of risk perception on safety motivation
and then on behavior, or may change the impact from negative to po-
sitive. Our results demonstrated this function of coworkers’ safety cli-
mate: for workers in a positive coworkers’ safety climate, the negative
effect of risk perception on motivation and the indirect negative effect
9
Accident Analysis and Prevention 134 (2020) 105350
of risk perception on behavior were both reversed to be positive. This
implies that a positive coworkers’ safety climate helped to change
workers’ perceptions of workplace risks from negative (hindrances) to
positive (challenges). However, an adverse moderation effect of su-
pervisor’s safety climate was identified.
5.1. Theoretical contributions
This study offers a new theoretical perspective on risk perception
and its impact on safety behavior. The previous findings concerning this
impact are largely contradictory. Some research has supported a posi-
tive impact (Kouabenan et al., 2015; Seo, 2005; Tomás et al., 1999; Xia
et al., 2017), while other studies have supported a negative impact
(Goldenhar et al., 2003; Nahrgang et al., 2011). It can be concluded
that the former studies mainly consider safety behavior as risk-avoiding
behavior that employees implement to reduce their risk. The latter re-
search mainly reviews perceived risk as job hindrance demands (i.e.,
stresses) that exhaust the personal resources required for demonstrating
safety behavior. We propose that this conflict can be addressed by
drawing on the recent development of the job demand concept, namely,
the hindrance demand–challenge demand distinction (Bakker and
Demerouti, 2017; Searle and Auton, 2015). Risk perception is a dual
concept by nature, and whether perceived risk is viewed as a job hin-
drance or challenge and, subsequently, whether it negatively or posi-
tively impacts safety behavior is basically an empirical question. Our
findings showed that perceived risk was viewed as a job hindrance
demand by construction workers, and thus high-risk perceptions led to
low levels of safety behavior. The dual hindrance and challenge de-
mand perspective on the concept of risk perception provides a new
theoretical reference for conceptualizing risk perception and synthe-
sizing the conflict relationship between risk perception and safety be-
havior examined in different occupations.
This study broadens our understanding of the role of supervisor
safety supervision and leadership in protecting and boosting employee
safety. Most previous studies have underscored the positive effect of
safety-oriented supervisors in preventing unsafe behavior and pro-
moting safety behavior (e.g., Clarke, 2013; Fang et al., 2015; Zohar and
Polachek, 2014). It was also found that the safety climate, including
supervisor’s and coworkers’ climate, buffered the deleterious impact of
job demands (i.e., work–family conflict and job insecurity) on safety
behavior (Bronkhorst, 2015). Our findings further supported the ex-
isting literature by confirming that coworkers’ safety climate reversed
the negative impacts of the risk perception on safety motivation and
behavior to be positive. However, contrary to the safety literature and
our expectations, when supervisors’ safety climate changed from a low
to a high level, the impact of risk perception on safety motivation
changed from positive to negative, and the negative effect of risk per-
ception on safety behavior via safety motivation was not alleviated but
worsened.
The different effects of supervisor’s safety climate and coworkers’
safety climate may be explained from the perspective of work stress.
Work stress can be derived from work requirements (Nahrgang et al.,
2011) and interpersonal relations (Weiss et al., 1976). In the present
study, most of the items for measuring supervisors’ safety climate were
imperatives, such as “The supervisor will ask us to abide by all safety
rules” and “The supervisor will still strictly comply with the safety rules
when we are tired.” These items reflect the characteristics of a pa-
ternalistic leader defined as “a style that combines strong discipline and
authority with fatherly benevolence” (Farh and Cheng, 2000, p. 91).
Paternalistic leadership imposes work stress on employees through
strict work requirements and/or interpersonal tensions, thus worsening
the negative effects of risk perception on safety motivation and beha-
vior (Chao and Kao, 2005; Chen and Kao, 2009). On the contrary,
coworkers’ safety climate items such as “Team members will actively
discuss how to prevent accidents” and “Team members often remind me
to stay safe” show that workers maintain a good relationship with each
N. Xia, et al.
other. These activities helped to ease the stress (risk perception as job
hindrance demands) placed on workers and subsequent consequences
produced by the stress, or even changed their negative perceptions of
workplace risks as hindrances (followed by a negative reaction, namely,
a reduction in safety motivation and behavior) to positive perceptions
as challenges (followed by a positive reaction, namely, an improvement
in safety motivation and behavior). In addition, using the lens of the
JD–R theory, paternalistic supervisor’s safety climate can be viewed as
a job hindrance demand while friendly coworkers’ safety climate tends
to be a job resource, as we hypothesized. A job hindrance demand will
worsen the negative effects of another job hindrance on work engage-
ment and performance, while a job resource will alleviate these effects
(Bakker and Demerouti, 2017). We contribute to the literature by
pointing out the potential negative role of certain supervisor safety
activities in influencing employee safety.
Last, our study also contributes to the hindrance–challenge job de-
mand framework. The concept of job demands was first proposed as job
hindrances that decreased employees’ work engagement and perfor-
mance (Demerouti et al., 2001). However, recent empirical studies have
found that some job demands, such as job responsibility, can promote
work engagement and performance (Crawford et al., 2010). Such job
demands are redefined as challenge demands that cost effort but si-
multaneously contribute to learning and goal achievement (Bakker and
Demerouti, 2017; Breevaart and Bakker, 2017). Furthermore, a demand
such as work pressure can be experienced as a job hindrance or chal-
lenge in different contexts (Searle and Auton, 2015). Research has also
found that demands are sometimes considered as both hindering and
challenging (Searle and Auton, 2015). Thus, our research fine-tunes the
hindrance–challenge job demand distinction in JD–R theory, by con-
ceptualizing risk perception as a concept that can be a job hindrance or
challenge, as well as by demonstrating that job demands (risk percep-
tion) can be both hindrances and challenges with the moderating roles
of coworkers’ safety climate.
5.2. Implications for practice
Given the hindrance attribute of risk perception for construction
workers and its negative impact on workers’ safety behavior, the
management should assess and take measures to reduce risk perception
levels in construction sites. Possible measures include providing suffi-
cient individual protection equipment, removing obstacles in con-
struction sites, and revealing safety management systems for removing
workers’ misperceptions of workplace risk. Considering the potential
dual attribute of risk perception, it is suggested that the management in
other occupations should first investigate the relationship between
perceived risk and safety behavior, and then decide to take measures to
increase or decrease employees’ risk perception levels.
Regarding the buffer effect of coworkers’ safety climate against the
deleterious impacts risk perception has on safety motivation and then
on safety behavior, management should commit itself to creating po-
sitive cooperation and communication concerning safety among
workers. More importantly, the management should realize that safety-
oriented supervisors will not always promote workers’ safety behavior;
for example, paternalistic safety leadership may produce negative im-
pacts on safety motivation and behavior.
5.3. Limitations and future directions
One limitation in our research design is that our data were workers’
self-reports. Thus, common method variance may exist. As suggested by
Podsakoff et al. (2003), we used statistical remedies to test common
method variance. First, Harman’s single-factor test showed that the one-
factor model fit the data poorly (χ2 = 5636.505, df = 860, p < .01;
CFI = .598; TLI = .473; SRMR = .141; RMSEA = .138). We also used
the technique of controlling for the effects of a single unmeasured latent
method factor. The CFA results indicated that the revised model (6
10
Accident Analysis and Prevention 134 (2020) 105350
substantive factors + 1 method factor) has poorer fit indexes (χ2=
2255.955, df = 802, p < .01; CFI = .847; TLI = .828; SRMR = .064;
RMSEA = .079) than the baseline six-factor model (χ2 = 1798.016,
df = 803, p < .01; CFI = .912; TLI = .908; SRMR = .068; RMSEA =
.065). Furthermore, Evans (1985) conducted a Monte Carlo study and
concluded that common method variance can only alleviate true in-
teractions rather than generating artificial interactions. Spector (1987)
also argued that interactions were unlikely to be a consequence of
method variance. As moderation effects were found in our research,
common method variance should not pose a serious threat to the study’s
validity. Nevertheless, to mitigate common method variance, we sug-
gest that future research should try to collect data from different
sources (objective observation or supervisor-assessment of worker
safety behavior) to examine and validate the relationships found in the
present study.
We suggest two future research directions. To the best of our
knowledge, the dual job hindrance and challenge perspective on the
concept of risk perception was first proposed here, hence, more work is
warranted on this issue. For example, the perception of workplace risks
and its impact on safety behavior should be widely examined in various
high-risk contexts. These empirical studies may help to further reveal
the conditions under which perceived risk serves as a job hindrance
demand versus a job challenge demand; eventually, the synthesis of the
conflict risk perception–safety behavior relationship from the dual job
hindrance and challenge perspective can be achieved. The role of su-
pervisor’s safety climate in the relationship between risk perception and
safety behavior should also be further clarified. As we focused mainly
on paternalistic safety leadership, other leadership styles, such as
transformational leadership, should be investigated. In so doing, we can
develop a holistic picture of the role of supervisors in influencing em-
ployee safety.
6. Conclusions
This study draws on the hindrance–challenge job demand concept,
proposing that perceived risk is by nature a dual concept. Whether
perceived risk is viewed as a job hindrance or challenge and, subse-
quently, whether it negatively or positively impacts safety behavior is
basically an empirical question that depends on context. Empirical data
collected from construction workers demonstrated that risk perception
served as a job hindrance, which had a further negative impact on
safety motivation and behavior. This study also confirmed that cow-
orkers’ safety climate helped workers to change their perceptions of
workplace risks from negative (a job hindrance) to positive (a job
challenge), thereby motivating them to demonstrate safety behavior.
However, when supervisor’s safety climate changed from a low level to
a high level, the impact of risk perception on safety motivation changed
from positive to negative, and the negative effect of risk perception on
safety behavior via safety motivation was not alleviated but worsened.
Overall, this study provides a theoretical framework and empirical
evidence for the risk perception concept and its relationship to safety
behavior, and indicates the potentially different role of supervisor’s and
coworker’s safety climate in the formulation of employee safety.
Declaration of Competing Interest
None. The authors claim that this paper is only being considered for
Accident Analysis & Prevention. All of the authors and their responsible
authorities have approved the manuscript and agreed its final submis-
sion.
Acknowledgements
This work was supported by the Fundamental Research Funds for
the Central Universities (Grant Nos. 2242019R10, 2019B69014),
Research Foundation of Humanities and Social Sciences of Ministry of
N. Xia, et al.
Education of China (18YJC630121), the National Natural Science
Foundation of China (Grant Nos. 71772736, 71722004), and the
Postgraduate Research & Practice Innovation Program of Jiangsu
Province (Grant No. SJKY19_0399). We would like to thank the parti-
cipants who provided us with valuable empirical data. Particular thanks
also go to the editor and anonymous reviewers for their work and
contributions.
References
Akintoye, A., 2000. Analysis of factors influencing project cost estimating practice.
Constr. Manage. Econ. 18, 77–89.
Aven, T., Renn, O., 2009. On risk defined as an event where the outcome is uncertain. J.
Risk Res. 12 (1), 1–11.
Bakker, A.B., Demerouti, E., 2014. Job demands-resources theory. In: Cooper, C., Chen, P.
(Eds.), Wellbeing: A Complete Reference Guide. Wiley-Blackwell, Chichester, UK.
Bakker, A.B., Demerouti, E., 2017. Job demands–resources theory: taking stock and
looking forward. J. Occup. Health Psychol. 22 (3), 273–285.
Bakker, A.B., Demerouti, E., Sanz-Vergel, A.I., 2014. Burnout and work engagement: the
JD-R approach. Annu. Rev. Organ. Psychol. Organ. Behav. 1, 389–411.
Bakker, A.B., Sanz-Vergel, A.I., 2013. Weekly work engagement and flourishing: the role
of hindrance and challenge job demands. J. Vocat. Behav. 83 (3), 397–409.
Barbaranelli, C., Petitta, L., Probst, T.M., 2015. Does safety climate predict safety per-
formance in Italy and the USA? Cross-cultural validation of a theoretical model of
safety climate. Accid. Anal. Prev. 77, 35–44.
Bartko, J.J., 1976. On various intraclass correlation reliability coefficients. Psychol. Bull.
83, 762–765.
Bauer, D.J., Preacher, K.J., Gil, K.M., 2006. Conceptualizing and testing random indirect
effects and moderated mediation in multilevel models: new procedures and re-
commendations. Psychol. Methods 11 (2), 142.
Becker, T.E., 2005. Potential problems in the statistical control of variables in organiza-
tional research: a qualitative analysis with recommendations. Organ. Res. Methods 8
(3), 274–289.
Beus, J.M., Dhanani, L.Y., McCord, M.A., 2015. A meta-analysis of personality and
workplace safety: addressing unanswered questions. J. Appl. Psychol. 100 (2),
481–498.
Bliese, P.D., 2000. An introduction to multilevel modeling techniques. Pers. Psychol. 53
(4), 1062–1065.
Breevaart, K., Bakker, A.B., 2017. Daily job demands and employee work engagement: the
role of daily transformational leadership behavior. J. Occup. Health Psychol. 23 (3),
338–349.
Brondino, M., Silva, S.A., Pasini, M., 2012. Multilevel approach to organizational and
group safety climate and safety performance: co-workers as the missing link. Saf. Sci.
50 (9), 1847–1856.
Bronkhorst, B., 2015. Behaving safely under pressure: the effects of job demands, re-
sources, and safety climate on employee physical and psychosocial safety behavior. J.
Safety Res. 55, 63–72.
Browne, M.W., Cudeck, R., 1992. Alternative ways of assessing model fit. Sociol. Methods
Res. 21 (2), 230–258.
Chao, A., Kao, S., 2005. Paternalistic Leadership and subordinated stress in Taiwanese
enterprises. Res. Appl. Psychol. 27 (1), 111–131.
Chen, C.F., Chen, S.C., 2014. Measuring the effects of safety management system prac-
tices, morality leadership and self-efficacy on pilots’ safety behaviors: safety moti-
vation as a mediator. Saf. Sci. 62, 376–385.
Chen, H.Y., Kao, H.S.R., 2009. Chinese paternalistic leadership and non-Chinese sub-
ordinates’ psychological health. Int. J. Hum. Resour. Manage. 20 (12), 2533–2546.
Cheung, M.L., Au, K., 2005. Applications of multilevel structural equation modeling to
cross-cultural research. Struct. Equ. Model. 12 (4), 598–619.
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. J. Appl. Psychol. 94 (5),
1103–1127.
Clarke, S., 2013. Safety leadership: a meta-analytic review of transformational and
transactional leadership styles as antecedents of safety behaviours. J. Occup. Organ.
Psychol. 86 (1), 22–49.
Crawford, E.R., LePine, J.A., Rich, B.L., 2010. Linking job demands and resources to
employee engagement and burnout: a theoretical extension and meta-analytic test. J.
Appl. Psychol. 95 (5), 834.
Demerouti, E., Bakker, A.B., Nachreiner, F., Schaufeli, W.B., 2001. The job demands-
resources model of burnout. J. Appl. Psychol. 86 (3), 499–512.
Didla, S., Mearns, K., Flin, R., 2009. Safety citizenship behaviour: a proactive approach to
risk management. J. Risk Res. 12 (3–4), 475–483.
DiStefano, C., Hess, B., 2005. Using confirmatory factor analysis for construct validation:
an empirical review. J. Psychoeduc. Assess. 23 (3), 225–241.
Dov, Z., 2008. Safety climate and beyond: a multi-level multi-climate framework. Saf. Sci.
46 (3), 376–387.
Evans, M.G., 1985. A Monte Carlo study of the effects of correlated method variance in
moderated multiple regression analysis. Organ. Behav. Hum. Decis. Process. 36 (3),
305–323.
Fang, D.P., Wu, C.L., Wu, H.J., 2015. Impact of the supervisor on worker safety behavior
in construction projects. J. Manag. Eng. 31 (6), 04015001.
Fang, D.P., Wu, H.J., 2013. Development of a Safety Culture Interaction (SCI) model for
construction projects. Saf. Sci. 57, 138–149.
Farh, J.L., Cheng, B.S., 2000. A Cultural Analysis of Paternalistic Leadership in Chinese
11
Accident Analysis and Prevention 134 (2020) 105350
Organizations. Management and Organizations in the Chinese Context. Palgrave
Macmillan, London, pp. 84–127.
Fogarty, G.J., Shaw, A., 2010. Safety climate and the theory of planned behavior: towards
the prediction of unsafe behavior. Accid. Anal. Prev. 42 (5), 1455–1459.
Fornell, C., Larcker, D.F., 1981. Structural equation models with unobservable variable
and measurement error: algebra and statistics. J. Mark. Res. 18 (3), 382–388.
Frone, M.R., 1998. Predictors of work injuries among employed adolescents. J. Appl.
Psychol. 83 (4), 565–576.
George, J.M., Bettenhausen, K., 1990. Understanding prosocial behavior, sales perfor-
mance, and turnover: a group–level analysis in a service context. J. Appl. Psychol. 75
(6), 698–709.
Goldenhar, M.L., Williams, L.J., Swanson, N.G., 2003. Modelling relationships between
job stressors and injury and near-miss outcomes for construction labourers. Work
Stress 17 (3), 218–240.
Griffin, M.A., Hu, X., 2013. How leaders differentially motivate safety compliance and
safety participation: the role of monitoring, inspiring, and learning. Saf. Sci. 60,
196–202.
Griffin, M.A., Neal, A., 2000. Perceptions of safety at work: a framework for linking safety
climate to safety performance, knowledge, and motivation. J. Occup. Health Psychol.
5 (3), 347–358.
Gyekye, S.A., Salminen, S., 2005. Are “good soldiers” safety conscious? An examination of
the relationship between organizational citizenship behaviors and perception of
workplace safety. Soc. Behav. Personal. Int. J. 33 (8), 805–820.
Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E., 2009. Multivariate Data Analysis.
Prentice-Hall, Upper Saddle River, NJ.
Halbesleben, J.R.B., Neveu, J.P., Paustian-underdahl, S.C., Westman, M., 2014. Getting to
the “COR”: understanding the role of resources in conservation of resources theory. J.
Manage. 40 (5), 1334–1364.
Haslam, R.A., Hide, S.A., Gibb, A.G.F., Gyi, D.E., Pavitt, T., Atkinson, S., Duff, A.R., 2005.
Contributing factors in construction accidents. Appl. Ergon. 36 (4), 401–415.
Heinrich, H.W., Petersen, D.C., Roos, N.R., Hazlett, S., 1980. Industrial Accident
Prevention: A Safety Management Approach. McGraw-Hill, New York, NY.
Hobfoll, S.E., 1989. Conservation of resources: a new attempt at conceptualizing stress.
Am. Psychol. 44, 513–524.
Hobfoll, S.E., 2001. The influence of culture, community, and the nested-self in the stress
process: advancing conservation of resources theory. Appl. Psychol. 50 (3), 337–421.
Hobfoll, S.E., 2002. Social and psychological resources and adaptation. Rev. Gen. Psychol.
6, 307–324.
Hofmann, D.A., Morgeson, F.P., Gerras, S.J., 2003. Climate as a moderator of the re-
lationship between leader-member exchange and content specific citizenship: safety
climate as an exemplar. J. Appl. Psychol. 88 (1), 170–178.
Hu, L.T., Bentler, P.M., 1999. Cutoff criteria for fit indexes in covariance structure ana-
lysis: conventional criteria versus new alternatives. Struct. Eq. Model.- Multidiscip. J.
6 (1), 1–55.
International Labour Organization, 2018. Safety and Health at Work. Retrieved from:.
https://www.ilo.org/global/topics/safety-and-health-at-work/lang–en/index.htm.
James, L.R., Demaree, R.G., Wolf, G., 1993. Rwg: an assessment of within-group inter-
rater agreement. J. Appl. Psychol. 78, 306–309.
Kao, K.Y., Spitzmueller, C., Cigularov, K., Wu, H., 2016. Linking insomnia to workplace
injuries: a moderated mediation model of supervisor safety priority and safety be-
havior. J. Occup. Health Psychol. 21 (1), 91–104.
Koopman, J., Lanaj, K., Scott, B.A., 2016. Integrating the bright and dark sides of OCB: a
daily investigation of the benefits and costs of helping others. Acad. Manage. J. 59
(2), 414–435.
Kouabenan, D.R., Ngueutsa, R., Mbaye, S., 2015. Safety climate, perceived risk, and in-
volvement in safety management. Saf. Sci. 77 (3), 72–79.
Li, F., Jiang, L., Yao, X., Li, Y.J., 2013. Job demands, job resources and safety outcomes:
the roles of emotional exhaustion and safety compliance. Accid. Anal. Prev. 51,
243–251.
Li, J.D., 2013. Psychometric properties of ten-item personality inventory in China. China
J. Health Psychol. 21, 1688–1692. https://doi.org/10.13342/j.cnki.cjhp.2013.11.
008. (in Chinese).
Liang, H.K., Lin, K.Y., Zhang, S.J., Su, Y.K., 2018. The impact of coworkers’ safety vio-
lations on an individual worker: a social contagion effect within the construction
crew. Int. J. Environ. Res. Public Health 15 (4), 773.
MacKinnon, D.P., Lockwood, C.M., Williams, J., 2004. Confidence limits for the indirect
effect: distribution of the product and resampling methods. Multivariate Behav. Res.
39 (1), 99–128.
Morrow, S.L., McGonagle, A.K., Dove-Steinkamp, M.L., Walker Jr, C.T., Marmet, M.,
Barnes-Farrell, J.L., 2010. Relationships between psychological safety climate facets
and safety behavior in the rail industry: a dominance analysis. Accid. Anal. Prev. 42
(5), 1460–1467.
Murie, F., 2007. Building safety: an international perspective. Int. J. Occup. Environ.
Health 13 (1), 5–11.
Muthén, L.K., Muthén, B.O., 2013. Mplus 7.11. Los Angeles, CA.
Nahrgang, J.D., Morgeson, F.P., Hofmann, D.A., 2011. Safety at work: a meta-analytic
investigation of the link between job demands, job resources, burnout, engagement,
and safety outcomes. J. Appl. Psychol. 96 (1), 71–94.
Neal, A., Griffin, M.A., 2006. A study of the lagged relationships among safety climate,
safety motivation, safety behavior, and accidents at the individual and group levels.
J. Appl. Psychol. 91 (4), 946–953.
Nunnally, J.O., 1978. Psychometric Theory. McGraw-Hill, New York, NY.
Podsakoff, P.M., MacKenzie, S.B., Lee, J.-Y., Podsakoff, N.P., 2003. Common method
biases in behavioral research: a critical review of the literature and recommended
remedies. J. Appl. Psychol. 88, 879–903.
Preacher, K.J., Hayes, A.F., 2004. SPSS and SAS procedures for estimating indirect effects
N. Xia, et al.
in simple mediation models. Behav. Res. Methods Instrum. Comput. 36 (4), 717–731.
Raudenbush, S.W., Bryk, A.S., 2002. Hierarchical Linear Models: Applications and Data
Analysis Methods. Sage, Thousand Oaks, CA.
Schaufeli, W.B., Bakker, A.B., van Rhenen, W., 2009. How changes in job demands and
resources predict burnout, work engagement, and sickness absenteeism. J. Organ.
Behav. 30, 893–917.
Schwatka, N.V., Rosecrance, J.C., 2016. Safety climate and safety behaviors in the con-
struction industry: the importance of co-workers commitment to safety. Work 54 (2),
401–413.
Searle, B.J., Auton, J.C., 2015. The merits of measuring challenge and hindrance ap-
praisals. Anxiety Stress Coping 28 (2), 121–143.
Seo, D.C., 2005. An explicative model of unsafe work behavior. Saf. Sci. 43 (3), 187–211.
Sirotnik, K.A., 1980. Psychometric implications of the unit of analysis problem (with
examples from the measurement of organizational climate). J. Educ. Meas. 17 (4),
245–282.
Spector, P.E., 1987. Method variance as an artifact in self-reported affect and perceptions
at work: myth or significant problem? J. Appl. Psychol. 72 (3), 438–443.
Ten Brummelhuis, L.L., Bakker, A.B., 2012. A resource perspective on the work–home
interface: the work–home resources model. Am. Psychol. 67 (7), 545–556.
Tomás, J.M., Melia, J.L., Oliver, A., 1999. A cross-validation of a structural equation
model of accidents: organizational and psychological variables as predictors of work
safety. Work Stress 13 (1), 49–58.
Vinodkumar, M., Bhasi, M., 2010. Safety management practices and safety behaviour:
assessing the mediating role of safety knowledge and motivation. Accid. Anal. Prev.
42 (6), 2082–2093.
Weinstein, N., Ryan, R.M., 2010. When helping helps: autonomous motivation for
12
Accident Analysis and Prevention 134 (2020) 105350
prosocial behavior and its influence on well-being for the helper and recipient. J.
Pers. Soc. Psychol. 98, 222–244.
Weiss, S., Baker, G., Gupta, R., 1976. Vibrational residual stress relief in a plain carbon
steel weldment. Welding Res. Suppl. 55 (2) 47s–51s.
Williamson, G.M., Clark, M.S., 1989. Providing help and desired relationship type as
determinants of changes in moods and self-evaluations. J. Pers. Soc. Psychol. 56,
722–734.
Xia, N., Wang, X., Griffin, M.A., Wu, C., Liu, B., 2017. Do we see how they perceive risk?
An integrated analysis of risk perception and its effect on workplace safety behavior.
Accid. Anal. Prev. 106, 234–242.
Yu, Q., Ding, L., Zhou, C., Luo, H., 2014. Analysis of factors influencing safety manage-
ment for metro construction in China. Accid. Anal. Prev. 68, 131–138.
Zhao, X., Lynch Jr, J.G., Chen, Q., 2010. Reconsidering Baron and Kenny: myths and
truths about mediation analysis. J. Consum. Res. 37 (2), 197–206.
Zohar, D., 1980. Safety climate in industrial organizations: theoretical and applied im-
plications. J. Appl. Psychol. 65 (1), 96–102.
Zohar, D., 2003. Safety climate: conceptual and measurement issues. In: Quick, J.C.,
Tetrick, L. (Eds.), Handbook of Occupational Health Psychology. American
Psychological Association, Washington, DC.
Zohar, D., Polachek, T., 2014. Discourse-based intervention for modifying supervisory
communication as leverage for safety climate and performance improvement: a
randomized field study. J. Appl. Psychol. 99 (1), 113–124.
Zou, P.X.W., Sunindijo, R.Y., Dainty, A.R.J., 2014. A mixed methods research design for
bridging the gap between research and practice in construction safety. Saf. Sci. 70,
316–326.