Content uploaded by Tao Wang
Author content
All content in this area was uploaded by Tao Wang on Feb 20, 2016
Content may be subject to copyright.
Core Dimensions of the Construction Safety Climate for a
Standardized Safety-Climate Measurement
Chunlin Wu1; Xinyi Song2; Tao Wang3; and Dongping Fang4
Abstract: The prevalent disparity and divergence in the identification of safety-climate dimensions in the academia cause general confusion
and inconvenience to both construction researchers and practitioners in terms of safety-climate measurement. Existing review studies
identified several key dimensions or common features of safety climate, but only in a qualitative way. Whether these common features
fit the reality and reflect the essence of construction safety climate is still to be verified by empirical studies. This research defined the
core dimension and specific dimension of safety climate, identified the four most commonly used dimensions, and built a core dimension
structure of safety climate accordingly. Empirical data collected from 21 Chinese construction enterprises were analyzed by means of struc-
tural equation modeling. The proposed core dimension structure and the corresponding measurement scale were validated rigorously by
structural equation modeling approaches. Furthermore, two specific subgroups of the enterprises were analyzed to prove that the core di-
mension structure also applies to specific construction enterprise types, such as building contractors and specialty trade contractors. It is
concluded that the proposed core dimension structure of safety climate is applicable in construction practices, especially to building enter-
prises. Practical implications of the safety-climate core dimension research are discussed in detail. This study contributes to the construction
safety-climate study primarily by depicting relationships among the common dimensions as well as relationships between common dimen-
sions and specific dimensions, which have rarely been involved or interpreted deeply in the past research. This can in turn facilitate the
standardization of construction safety-climate measurement by providing a unified criterion (core dimension structure of safety climate) for
both researchers and practitioners. It is a valid starting point to design specific measurement scales in different settings. DOI: 10.1061/
(ASCE)CO.1943-7862.0000996.© 2015 American Society of Civil Engineers.
Author keywords: Construction safety; Safety climate; Core dimensions; Structural equation modeling; Labor and personnel issues.
Introduction
Construction is one of the most hazardous industries, in which large
numbers of accidents result in workers deaths, injuries, and work-
related illness as well as other direct and indirect heavy losses
(Fang and Wu 2013). Recent statistics from the United States,
the United Kingdom, Hong Kong, and the Chinese mainland reveal
no significant reduction in the number of fatalities in the con-
struction industry while an overall safety improvement has been
achieved within all industrial sectors (U.S. Department of Labor
2010;U.K. Health and Safety Executive 2011;Hong Kong Labor
Department 2011;MOHURD 2013). In recent years, to further
improve safety conditions on the worksite, safety researchers and
practitioners have both shifted their focus from technical ap-
proaches to psychological, organizational, and other nonphysical
safety factors (Zohar 1980;Reason 1990;Fang and Wu 2013).
As a significant nonphysical factor, organizational climate is a
valuable resource for organizational effectiveness because it influ-
ences employees’loyalty and commitment to the organization
(Van Vianen et al. 2011). Safety climate is a subset of organiza-
tional climate, which is related to safety performance improvement,
and has been proved to be one of the most important antecedents for
on-site safety performance (Zohar 2010). Safety climate can be
regarded as a set of common perceptions of employees in terms
of safety policies, safety codes (including regulations and proce-
dures), safety conventions, and safety behavior, all of which are
attributes reflecting the organization’s safety priorities (Zohar
1980;Guldenmund 2000).
Zohar (1980) initially proposed the construct of safety climate
and its dimensions. Later on, safety and psychosocial researchers
have carried out extensive studies on safety-climate conceptions,
dimensions, indicators, factor structures, applicative organizational
levels, and roles in the causal chains of accidents. As is widely ac-
cepted, safety climate is the “snapshot”of safety culture (Fang et al.
2006). Safety culture is a macroscopic concept, being stable and
consistent (Glendon and Stanton 2000). However, from the micro-
cosmic perspective, safety culture may change slightly with the
variations in organizational structure, employees, and contracted
projects. Safety climate, as the snapshot of safety culture, is sensi-
tive to these variations, so it can reflect the safety conditions in a
specific status of the organization as well as its employees. As a
matter of fact, currently safety climate survey is almost the only
way to represent safety culture through quantitative measurement
(Fang and Wu 2013). In particular, safety climate is highly appli-
cable to construction projects because of their characteristics, such
as temporality, decentralization, mobility, high turn-over rate, and
environmental complexity (Fang and Wu 2013).
1Ph.D. Candidate, Dept. of Construction Management, Tsinghua Univ.,
Beijing 100084, China. E-mail: wucl11@mails.tsinghua.edu.cn
2Assistant Professor, School of Building Construction, Georgia Institute
of Technology, 280 Ferst Dr., Atlanta, GA 30308. E-mail: xinyi.song@coa
.gatech.edu
3Lecturer, School of Management Science and Engineering, Central
Univ. of Finance and Economics, Beijing 100081, China (corresponding
author). E-mail: wangtaothu@163.com
4Professor, Dept. of Construction Management, Tsinghua Univ.,
Beijing 100084, China. E-mail: fangdp@tsinghua.edu.cn
Note. This manuscript was submitted on May 1, 2014; approved on
February 9, 2015; published online on March 18, 2015. Discussion period
open until August 18, 2015; separate discussions must be submitted for
individual papers. This paper is part of the Journal of Construction En-
gineering and Management, © ASCE, ISSN 0733-9364/04015018(12)/
$25.00.
© ASCE 04015018-1 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
Safety-Climate Dimensions
In order to measure safety climate quantitatively, the main task is
to construct a multidimensional factor structure. Based on the prin-
ciple of structural equation modeling (SEM), safety climate can be
seen as a high-order latent variable, which is described and mea-
sured by several low-order latent variables (Mohamed 2002;Zhou
et al. 2011). These low-order variables are defined as safety-climate
dimensions. Each dimension, or latent variable, uses several mani-
fest variables as observable indicators. Mohamed (2002) proposed
an evaluative model for safety climate on construction sites and
carried out questionnaire surveys to validate it. Results corrobo-
rated the importance of the indicators in the model in achieving
a positive safety climate. These indicators include management
commitment, communication, worker’s involvement, attitudes,
and competence, as well as supportive and supervisory environ-
ments. Glendon and Litherland (2001) tested the safety climate
of a road construction contractor, and identified several elements
which constituted the factor structure of safety climate, including
communication and support, adequacy of procedures, work pres-
sure, personal protective equipment, relationships, and safety rules.
Fang et al. (2006) undertook a case study in a Hong Kong construc-
tion enterprise and identified 10 critical dimensions of safety cli-
mate, such as safety attitude and management commitment, safety
consultation and safety training, supervisor’s role and workmate’s
role, risk-taking behavior, appraisal of safety procedure and work
risk, improper safety procedure, worker’s involvement, and com-
petence. Zhou et al. (2011) adopted the case study approach and
used one safety-climate instrument to carry out two rounds of in-
vestigation with an interval of three years. The purpose was to
verify the existence of a standard first-order factor structure,
i.e., safety-climate dimensions, and second-order construct,
i.e., safety climate. The dimensions applied in their study are safety
regulations; safety supervision, safety training and workmate’s sup-
port; management commitment; and safety attitude.
According to the above literature review, researchers have iden-
tified some common safety-climate dimensions, but there still ex-
ists inconsistency. This may be attributed to the nature of safety
climate. It is a high-order construct, which can be interpreted in
various ways. Most studies in safety climate dimensions are explor-
atory in essence, with minor exceptions like DeDobbeleer and
Béland (1991) and Huang et al. (2006) which applied confirmatory
approaches. Research works carried out in different companies
and industries may have different results. In addition, every re-
searcher of safety climate has considerable freedom to label his
or her dimensions, and there is not much correspondence among
them. Obviously, different researchers did not have the need (both
methodological and terminological) to connect to previous research
in terms of their dimensions (Guldenmund 2000). The current dis-
parity in the identification of safety-climate dimensions causes con-
fusion and inconvenience to both researchers and practitioners to
evaluate safety climate. Existing review studies on safety climate or
safety culture did identify several key dimensions or common fea-
tures of safety climate across industries (e.g., Guldenmund 2000;
Flin et al. 2000), but most of these papers were summaries of the
results from previous studies and obtained common dimensions
qualitatively. They rarely probed into the internal structure of safety
climate and studied the interrelationships among different common
dimensions based on empirical data. It is still to be verified by em-
pirical studies whether the identification of common features has
enough reliability and validity, and whether these common features
fit the reality and reflect the essence of construction safety climate.
Moreover, previous research did not cover one critical issue,
that is, the relationship between common dimensions and specific
dimensions in specific settings. Once the common dimensions or
features of safety climate are generated, another important question
arises: how can those common dimensions guide future safety cli-
mate measurement in various contexts? This question can be an-
swered only when the relationships among common dimensions
as well as the relationships between common dimensions and spe-
cific dimensions are discovered. While some previous studies men-
tioned this question as a direction for future study, few looked into
it in detail.
This study contributes to the safety-climate study by depicting
relationships among the common dimensions as well as relation-
ships between common dimensions and specific dimensions. Es-
sential attributes of safety climate and its dimensions are explored,
and the internal structure of safety climate will be validated quan-
titatively based on an empirical study.
Core Dimensions and Specific Dimensions of
Safety Climate
Jones and James (1979) analyzed the multidimensionality of
organizational culture and climate, and pointed out that there exist
“the core of dimensions”and “specific dimensions”applying to
particular situations. Guldenmund (2000) reviewed safety culture
and safety climate and concluded that the safety climate factor
structure contains multiple layers. Hofstede (1991) defined the first
layer, or the central core, of safety culture as norms and values, the
second layer as rituals, the following as heroes, and the outer layer
as symbols. The divergence of safety-climate dimensions men-
tioned above is the reflection of the existence of “specific dimen-
sions.”According to Jones and James (1979), specific dimensions
may just be the variants of the dimension core in specific environ-
ments. Researchers can classify different dimensions by extracting
the same characteristics, so as to achieve dimension reduction and
divergence elimination. When a couple of higher-order latent var-
iables are extracted from these specific dimensions which can be
seen as low-order latent variables, different studies may take on
more similar results (Guldenmund 2000). The idea also conforms
to the parsimony principle of SEM, which requires the structural
model to be as simple and clear as possible. This serves as the
methodological basis of this study. The definitions of both core di-
mensions and specific dimensions in this study are (Fig. 1for the
illustration of the definitions) as follows.
Safety climate, as a core construct or high-order latent variable
itself, is multihierarchical and multidimensional. It has several
layers of dimensions, and each layer has a number of dimensions.
From the inside out, these dimensions can be labeled from high-
order to low-order latent variables. The outermost layer is com-
posed of the manifest variables, or the observable indicators. In this
hierarchy, the core construct as well as some inner layers can be
applied universally, so the dimensions within can be called core
dimensions; dimensions in the other outward layers are variants
of the core dimensions in specific environments, and are so-called
specific dimensions.
According to the definition, core dimensions and specific di-
mensions are both structures; that is, they may be either single or
multiple, and have either single or multiple layers. Correlative links
can exist within the structure. As the existence of a single “dimen-
sion core”has not been sufficiently verified (Guldenmund 2000),
this definition is therefore more precise and inclusive.
The development of core dimensions has significant practical
importance. The determination of core dimensions is a crucial pre-
condition for safety-climate evaluations. Core businesses or internal/
external environments are different in different construction
© ASCE 04015018-2 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
enterprises, leading to different kinds of safety management issues.
However, due to the similar processes of construction, construction
enterprises may share the common safety climate factors, or the core
dimensions. The goal of this research is to discover those core
dimensions of construction safety climate, which serve as the foun-
dation for developing comprehensive safety-climate scales. Core
dimensions and specific dimensions can serve as important “organi-
zational process assets”[Project Management Institute (PMI) 2008],
namely the organizational guideline to design safety-climate mea-
sures and achieve safety performance improvement. To achieve
the objective of establishing core dimensions of safety climate,
(1) core dimensions in the safety climate factor stru ctures of different
construction enterprises should be identified, and (2) a core dimen-
sion structure embodying interrelationships among different core
dimensions should be built and validated.
Literature Review: Developing the Safety-Climate
Core Dimension Structure
The identification of construction safety climate core dimensions
started from summarizing the existing related research results and
finding out the common characteristics for extraction. According to
Jones and James (1979) and Guldenmund (2000), every researcher
of safety climate has considerable freedom to label his or her
dimensions, and there is not much correspondence among them.
To solve this problem, a renaming and grouping exercise is con-
ducted (Guldenmund 2000); that is, one could define a small set of
common denominators (i.e., core dimensions) to classify compa-
rable dimensions.
Identification of core dimensions may naturally depend on the
importance and frequencies of occurrences of various dimensions
in the existing literature. Just as Guldenmund (2000) argued, when
the number of times a dimension is found is taken into account, it
will become obvious that certain dimensions are mentioned more
often than others. However, it is not reliable or convincing enough
to simply calculate frequencies of occurrences while ignoring the
existing theoretical evidence. The identification of core dimensions
can get great benefit from references to the influential research re-
sults in safety science, especially literature review papers.
Flin et al. (2000) and Gadd (2002), respectively, made compre-
hensive literature reviews on safety climate, and discovered five
dominant dimensions, namely, management, safety system, risk,
working pressure, and competence. Management refers to the
organizational managers’(especially top managers) commitment
and priorities regarding safety issues. Safety system reflects various
aspects of the organizational safety management system, including
safety supervision, safety committee, work permits, and safety
training. Risk covers a series of dimensions related to safety risk
issues, like the self-reported level of risk tolerance, worksite hazard
identification, and risk attitudes. Working pressure consists of
working paces and working load. Competence is the employees’
perceptions of their own aptitudes, workmanship, and capabilities
(Gadd 2002). Flin et al. (2000) further pointed out that regulations
and rules are of equal importance to the above five dimensions. It is
also one of the most frequently occurring dimensions discovered in
the reviewing study of Guldenmund (2000). It refers to the employ-
ees’perceptions of safety rules as well as observance/violation of
safety regulations, and also is relevant to risk-taking behavior, since
taking risks, in most cases, means safety regulation violation.
Zohar (2010) provided reflections and possible future directions
for safety-climate research after 30 years since this construct was
firstly defined and measured. He proposed that safety-climate per-
ceptions should focus on the nature of relationships among safety
policies, procedures, and practices. He further pointed out that the
overall level of safety climate represents the shared perceptions of
the priority of safety compared to other competing priorities. Safety
policies embody and convey safety priorities of the management,
and safety practices involve such aspects of the safety system as
safety training, safety supervision, and safety involvement.
Based on the above review studies of safety climate, more
than 30 technical papers which studied safety-climate dimensions
were further reviewed to generate safety-climate core dimensions,
including those studies reviewed above in the “Safety-Climate
Dimensions”subsection. Other typical dimensions include impor-
tance of safety training programs, management attitudes toward
safety, effects of safety conduct on promotion, status of safety of-
ficer, effects of safety conduct on social status, status of safety
committee (Zohar 1980); communication and support, adequacy
and procedures, personal protective equipment, relationships, work
pressure, safety rules (Glendon and Litherland 2001); safety atti-
tude and management commitment, safety consultation and safety
training, risk-taking behavior, improper safety procedure, workers’
involvement (Fang et al. 2006).
Fig. 2presents those dimensions which have an occurrence fre-
quency of more than 30%. Based on the classifications in the liter-
ature review, those dimensions which have different names but bear
the same or similar attributes with one another were categorized
into one core dimension.
The following are preliminarily identified core dimensions of
safety climate of construction enterprises. The percentage behind
the dimension names is the frequency of occurrence as it appears
in previous studies. The four core dimensions listed below have
much higher occurrence frequencies than others, which to some
extent validate the dimensions chosen in this research. Relation-
ships between the four core dimensions and those dimensions iden-
tified by major review studies are also discussed.
Safety priority (85%): It means the importance of safety com-
pared with other organizational goals (such as production, sched-
ule, and cost) perceived by both employers and employees. In
the existing literature, there is close correlation between safety pri-
ority and management commitment, but here this dimension also
Core
dimension
Specific
dimension
Observable
indicator
Core dimension
layer (s)
Specific dimension layer (s)
Observable indicator layer (s)
Fig. 1. Schematic diagram for the definitions of core dimensions and
specific dimensions of safety climate
© ASCE 04015018-3 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
depends on the working pressure of the staff. Therefore, it involves
both management and working pressure.
Safety rules and procedures (70%): This dimension is related
to observance/violation of safety rules,perceptions of safety pro-
cedures, and risk-taking behavior.
Safety supervision, training, and communication (90%): This
dimension mainly explains the safety system, and also covers the
most common indicators related to safety system in the existing lit-
erature. Safety system is a professional term and an indirect ap-
proach that the managers use to show their safety commitment, or
the exchange of ideas between the managers and workers on the
issue of safety.
Safety involvement (85%): The idea of “total safety manage-
ment (TSM)”(Herrero et al. 2002) has popular support and
endorsement in the academia. Favorable safety culture and climate
need the involvement of the whole staff. Safety involvement is re-
lated to management,safety system, and competence. It puts more
emphasis on subjective initiatives of human beings, other than ob-
jective conducive or adverse factors.
Based on the previous research, correlative links among the core
dimensions are also identified. The dimensions and their links con-
stitute the core dimension structure of safety climate, as shown
in Fig. 3.
SEM is then adopted to validate the preliminary core dimension
structure of safety climate of construction enterprises (Fig. 3). Since
SEM requires a concise factor structure of the construct, the pre-
liminary theoretical framework is suitable for further analysis.
SEM Research Methodology
Based on the above theoretical analysis and previous literature, a
SEM research model was built, and a questionnaire-based survey in
the Chinese construction industry was designed to collect data to
test the validity and reliability of the research model. Various ap-
proaches and indices were used to guarantee the objectivity and
high model construction quality of the research.
The structural equation model consists of the measurement
model and the structural model (Byrne 2013). The former one links
the model and the reality, and addresses the issue of measuring the
latent variable (dimension). The latter one aims to describe the re-
lationships among different dimensions.
Research Model
Structural Model
The structural model is equivalent to what Fig. 3shows. As the
paths between exogenous dimensions (nodes where arrows start,
like safety priority) and endogenous dimensions (nodes where
arrows end, like safety involvement) have residuals, each of the
endogenous dimensions in the structural model should have one
residual variable.
The hypothesized links among core dimensions include
(1) safety priority predicts safety supervision, training, and commu-
nication positively and directly; (2) safety priority predicts safety
rules and procedures positively and directly; (3) safety priority pre-
dicts safety involvement positively and directly; (4) safety super-
vision, training, and communication predict safety involvement
positively and directly; and (5) safety rules and procedures predict
safety involvement positively and directly. All of these relation-
ships will be validated later in this study.
Measurement Model
The measurement model consists of the dimensions and their re-
spective manifest variables (observable indicators) which are also
the measurement items in the questionnaire. The details of scale
items are shown in Table 1. This questionnaire was developed
based on the questionnaire adopted by Zohar and Luria (2005), and
also referred to Mohamed (2002) and Fang et al. (2006), both being
exploratory studies of safety climate in the construction industry. It
was further modified based on the features of the Chinese construc-
tion industry, especially the perception modes and language habits
Fig. 2. Frequencies of main safety-climate dimensions’occurrence in literature review
Fig. 3. Preliminary core dimension structure of construction safety
climate based on literature review
© ASCE 04015018-4 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
of the Chinese construction personnel. All items describe percep-
tions or behaviors of organizational top management.
Whole Research Model
Based on Fig. 3and Table 1, the whole structure of safety-climate
core dimensions of construction enterprises was developed, as
shown in Fig. 4, which was generated by AMOS 7.0. AMOS, short
for Analysis of Moment Structures, is one of the most widely used
software tools for SEM (Arbuckle 1995). F1, F2, F3, and F4 stand
for safety priority;safety supervision, training and communication;
safety rules and procedures; and safety involvement, respectively.
SC1 to SC18 stand for the observable indicators (questionnaire
items) to measure the four latent variables (dimensions). e1 to e18
are the residual variables of observable indicators, and can be
regarded as the measurement errors in the measurement model.
e19 to e21 are the residual variables of endogenous dimensions
in the structural model. w1 to w19 are the path coefficients (AMOS
automatically sets one of the path coefficients between every latent
variable and its manifest variables as one). v1 to v19 are the
variances to their corresponding variables. The one-way straight
arrows indicate one-way direct effects. Nodes where arrows start
are antecedents. Nodes where arrows end are consequences. For
example, SC1, as an observable indicator of F1, is affected by both
F1 and residual variable e4. In other words, in the measurement, the
score of SC1 is decided by its high-order variable F1 and e4
(which represents other unmeasurable factors). Thus, there is an
arrow going from F1 to SC1, and another one going from e4
to SC1.
Table 1. Measurement Scale of Safety-Climate Core Dimensions of Chinese Construction Enterprises
Core dimensions
of safety climate
Codes of
measurement
items Measurement items
Safety priority SC1 Always referring to safety when talking about your company, especially to the public
SC4 Keeping regular and thorough on-site safety inspection
SC5 Endeavoring to improve safety performance of all projects in your company
SC6 Being able to provide the complete personal protective equipment for workers of all professions
Safety supervision,
training, and
communication
SC11 Constantly urging all project managers to improve safety performance
SC12 Inputting sufficient time and funds to safety training
SC13 Regularly organizing safety managers to participate in seminars for the constant improvement of safety rules and
procedures
SC14 Frequently discussing with employees at all levels in your company about safety issues
SC15 Taking safety into account when setting long-term and short-term goals
SC16 Offering to workers as much safety instruction and training as possible
SC17 Organizing regular safety lectures and training to the management (not merely safety managers)
Safety rules and
procedures
SC2 Probably cutting down safety investment when funding of your company or project is insufficient
SC3 Not always coming to the site immediately for investigation when informed of significant safety hazards
SC10 Generally not considering their previous safety performance when determining subordinates
SC18 The power and authority given to site safety supervisors cannot fully meet the demands of site safety management
Safety involvement SC7 Probably lowering safety requirements to workers for sake of schedule
SC8 Still rewarding those who report accidents and hazards when funding of your company or project is limited
SC9 Generally not informing all staff of safety investigation results
V1
F1
F2
F4
F3
W1
W2 W3
SC6
V2
e1
1
1
SC5
V3
e2
W6
1
SC4
V4
e3 W7
1
SC1
V5
e4
W8
1
SC11
V6
e5
1
1
SC12
V7
e6
W9
1
SC13
V8
e7
W10
1
SC14
V9
e8
W11
1
SC15
V10
e9
W12
1
SC16
V11
e10
W13
1
SC17
V12
e11
W14
1
SC18
V13
e12
1
1
SC10
V14
e13
W15
1
SC3
V15
e14
W16
1
SC2
V16
e15
W17
1
SC8
V17 e16
11
SC9
V18
e17
W18 1
SC7
V19 e18
W19
1
V20
e19
1
V21
e20
1
V22
e21
1
W4
W5
Fig. 4. Preliminary structural equation model of safety-climate core dimensions constructed by AMOS 7.0
© ASCE 04015018-5 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
Empirical Survey
Questionnaire Design
The questionnaire includes the 18 items shown in Table 1. Re-
spondents were asked to indicate the degree to which they agree or
disagree with the practices of their top management as described in
the items (Zohar 2008). A 5-point Likert scale was adopted from 1
(completely disagree) to 5 (completely agree). Note that higher
scores could indicate worse safety conditions in some items.
Participants and Data Collection
Data were collected from Chinese construction enterprises with dif-
ferent core businesses and operating areas. The potential participant
enterprises should have stable business operations and healthy
project contracting modes; i.e., they should undertake business ac-
tivities in the industry for more than 10 years and have no commer-
cial litigation record. Invitation letters were distributed to more than
40 qualified construction enterprises, and 21 responded. Table 2
shows the basic information of these enterprises. One important
attribute of the samples is type of enterprise, e.g., building contrac-
tors (i.e., contractors for the construction of buildings), specialty
trade contractors, and labor service contractors. This attribute will
be used as the grouping variable in the specific-group analysis later.
The main project management personnel with over five years’
working experience in the construction industry were invited as
respondents. In particular, project safety managers and worksite
supervisors were emphasized because they are directly responsible
for the safety performance of projects, have frequent contact with
both workers and top managers, and also tend to have good educa-
tional backgrounds (which qualify them to give accurate and high-
quality questionnaire responses).
There were a total of 623 respondents. All the questionnaire sur-
veys were undertaken face to face to minimize potential bias. Out of
623 answered questionnaires, 613 were considered valid. The other
10 are either incompletely filled out or duplicated from others. The
majority of the respondents are on-site safety officers and worksite
supervisors, both of whom are middle- and basic-level project man-
agers in charge of safety. The safety management experience of the
respondents averages 8.62 years. Table 2shows the summary of
the respondents’affiliations.
Statistical Analysis
The software of AMOS 7.0 was used to process the empirical data
and make the statistical analysis. Reliability and validity of the re-
search model were tested, so as to validate the existence of the core
dimension structure and the significance of the interrelationships
among different core dimensions.
In AMOS 7.0, the featured function of “variable grouping”can
make a specific group analysis; that is, it can group respondents by
their “types of enterprise,”which can help verify whether the
safety-climate core dimension structure is applicable to specific
groups. Details of the statistical analysis will be further illustrated
in the “Results”section.
Results
Reliability of the Questionnaire
Reliability of the questionnaire was evaluated. In the reliability test,
special attention should be paid to internal consistency reliability
which reflects correlations among questionnaire items belonging
to one dimension (Flynn et al. 1994). When one questionnaire con-
tains more than one dimension, the internal consistency of each
dimension should be tested individually.
Cronbach’sα, as a measure of internal consistency reliability, is
0.880 for the whole questionnaire. The αvalue is 0.727 for the
items belonging to the core dimension of safety priority. The α
values of safety supervision, training, and communication;safety
rules and procedures; and safety involvement are 0.809, 0.685, and
0.634, respectively. Since 0.6 is generally accepted as the bottom
line of the desired value of internal consistency, i.e., αvalues lower
than 0.6 indicate unreliable questionnaire designs (Flynn et al.
1994), the questionnaire can be seen as reliable, both as a whole
and from a single dimension’s perspective.
SEM Analysis
The software of AMOS 7.0 is applied to undertake SEM analysis.
The significance level is set as 0.05. Various goodness-of-fit indices
are used to test the fitness of the a priori research model deriving
from literature review (Hooper et al. 2008). Goodness-of-fit indices
evaluate whether the assumed model fits the empirical data, so they
are significant indicators of the validity and reliability of the model.
The system of goodness-of-fit indices consists of three parts,
including basic fit indices, overall model fit indices, and internal
structural model fit indices. The basic goodness-of-fit is the precon-
dition for the other two kinds of goodness-of-fit as well as the con-
structing quality. Overall model indices validate the external quality
of the model. It can be further divided into three parts, including
absolute fit indices, incremental fit indices, and parsimonious fit
indices (Hair et al. 1998). Internal structural model fit indices val-
idate the internal quality of the model. Goodness-of-fit tests should
comprehensively cover the whole index system, rather than part
of it. All the criteria in the following tests are based on the work
of Bagozzi and Yi (1988), Hair et al. (1998), and Bollen (1989),
unless specified otherwise.
Table 2. Participant Construction Enterprises
Codes of
participant
enterprises
Number of
respondents Types of enterprises
Sizes of
enterprisesa
1 64 Specialty trade contractors Medium
2 7 House building contractors Medium
3 28 House building contractors Large
4 30 House building contractors Medium
5 33 House building contractors Large
6 7 Specialty trade contractors Small
7 13 Labor service contractors Medium
8 57 Specialty trade contractors Medium
9 9 House building contractors Medium
10 14 House building contractors Medium
11 50 House building contractors Large
12 5 Specialty trade contractors Medium
13 11 Labor service contractors Small
14 28 House building contractors Large
15 18 House building contractors Medium
16 29 House building contractors Large
17 15 Labor service contractors Small
18 75 House building contractors Medium
19 15 House building contractors Small
20 21 Specialty trade contractors Medium
21 41 Labor service contractors Small
aSizes of enterprises are based on the Chinese classification standard for
construction enterprises: large enterprises have annual revenues of over
300 million RMB, medium enterprises have annual revenues of 30–
300 million RMB, and small enterprises have annual revenues of below
30 million RMB.
© ASCE 04015018-6 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
Basic Fit Test
The basic fit test assesses some fundamental attributes of the re-
search model, like error rates, variance level and significance of
major relationships among different variables, which are all impor-
tant prerequisites for further test and analysis in SEM. If the fitness
criteria of some indicators cannot be satisfied, the research model
should be modified accordingly. Five aspects of the model are re-
ferred to in the basic fit test (Bagozzi and Yi 1988).
Firstly, the estimated parameters should have few negative error
variances. In the model, except for the dimension of safety involve-
ment, all error variances (v2 to v19) are positive, which satisfies the
requirement of further tests.
Secondly, all error variances should be significant. In the model,
all 22 exogenous variables, including one dimension variable F1
and 21 error variables (i.e., e1 to e21 in Fig. 4), have significant
variances, which meets the above criteria.
Thirdly, the absolute values of correlation coefficients among
estimated parameters should not be too close to 1. Through calcu-
lation, the average of the absolute values is 0.303, and the maxi-
mum is 0.589, so it is convincing that no outliers exist in the model.
Fourthly, no extremely high or extremely low standard errors
should exist. Analysis results show that the level of standard errors
in the model is favorable.
Lastly, standardized regression weights of all links in the model
are suggested to lie in the range between 0.50 and 0.95. SEM analy-
sis results show that two links, F2→F4and F1→F4, do not sat-
isfy this criterion, with standardized regression weights of −0.234
and 0.440, respectively. A basic fit test of the model invalidates the
hypothesized links F2→F4and F1→F4, and thus the two links
are removed from the structural model shown in Fig. 4.
The following analysis will be based on the model without the
two invalid links. Table 3shows the regression analysis results by
SEM after the removal of two invalid links.
Overall Fit Test
As mentioned above, overall goodness of fit includes absolute
goodness of fit, incremental goodness of fit, and parsimonious
goodness of fit. Indicators within this test are diversified and also
subject to other factors such as sample sizes. In practice, these
indicators should be synthetically and comprehensively applied
to reach valid conclusions. Overall goodness of fit can generally
reflect the external quality of the model. The indicators showed
in Table 4are important and popularly used, and when combined,
can generally reflect the overall goodness of fit of the model.
According to Table 4, there are 14 out of 17 indices that satisfy
the fitness criteria, indicating that the external quality of the model
is fairly high. Two indices that are related to chi-square are below
the fitness criteria, most likely because of the large sample size.
The value of chi-square is easily affected by the sample size; i.e., the
larger the sample size, the higher the value of chi-square and the
lower the p-value (Bearden et al. 1982). However, in this study,
based on the computation of the overall index system, the model
shown in Fig. 4has a high external quality.
Internal Structure Fit Test
The internal structure fit test is essentially the significance test of the
paths between one dimension and its observable indicators (internal
quality of the model). The column of p-value in Table 3shows that
all factor loadings are significant in the significance level of 0.05,
indicating a good internal structure fitness of the model.
Preliminary Test of the Construction Quality of the Core
Dimension Structure
To some extent, a model with high goodness of fit can be deemed as
being successfully constructed as it fits the specific environment.
However, goodness of fit is essentially the consistency between the
assumed model and the empirical data, and thus can be largely af-
fected by the construction quality of the measurement model.
Therefore, only the conclusions of the above sections may not com-
pletely validate the high quality of the structural model, i.e., internal
validity of core dimension structure. This section was aimed at the
preliminary test of the construction quality of the whole core di-
mension structure. A high construction quality indicates that the
structural model is successfully constructed. In other words, the
core dimensions and their interrelationships are proven to be valid
and well founded (Bollen 1989).
According to the first three rows (except the title) of Table 3,
all regression weights among core dimensions are significant (all
p-values are under 0.05), so there exists a significant correlation
among different dimensions. The total impacting effects (standard-
ized values) between dimensions are as follows: F1→F30.653,
F1→F20.893, F1→F40.581, F3→F40.889. These coeffi-
cients can prove the strong correlation among core dimensions.
The interpretation of the above total impacting effects is as follows:
one unit variation in safety priority can lead to 0.653 unit variation
in safety rules and procedures, 0.893 unit in safety supervision,
training, and communication, and 0.581 unit in safety involvement;
one unit variation in safety rules and procedures can cause 0.889
unit variation in safety involvement. The high and significant path
coefficients are the evidence of high construction quality of the
model (Bollen 1989).
Revalidation of the Safety-Climate Core Dimensions
The above analysis results confirm the core dimension structure of
construction safety climate. The empirical data in the study were
obtained from 21 Chinese construction enterprises as a whole but
the differences among different enterprises were not distinguished.
The core dimensions were identified and validated using all the data
from 21 enterprises due to the requirement of sample size, but
whether these dimensions are applicable to one specific enterprise
type remains to be validated.
Table 3. Results of the Regression Analysis by SEM
Links
Unstandardized
regression
weight
Standardized
regression
weight
Standard
error
Critical
ratio pLabel
F1→F30.297 0.653 0.056 5.311 aW2
F1→F20.408 0.893 0.05 8.12 aW1
F3→F42.305 0.889 0.416 5.535 aW3
F1→SC61 0.674 ————
F1→SC50.887 0.710 0.06 14.857 aW6
F1→SC40.752 0.598 0.059 12.841 aW7
F1→SC10.673 0.604 0.052 12.848 aW8
F2→SC11 1 0.367 ————
F2→SC12 2.158 0.719 0.252 8.56 aW9
F2→SC13 1.949 0.712 0.227 8.566 aW10
F2→SC14 2.197 0.751 0.254 8.66 aW11
F2→SC15 1.592 0.678 0.188 8.482 aW12
F2→SC16 2.097 0.765 0.241 8.704 aW13
F2→SC17 1.726 0.665 0.206 8.366 aW14
F3→SC18 1 0.254 ————
F3→SC10 2 0.601 0.371 5.397 aW15
F3→SC32.137 0.596 0.396 5.392 aW16
F3→SC22.151 0.638 0.393 5.476 aW17
F4→SC81 0.746 ————
F4→SC90.738 0.582 0.058 12.672 aW18
F4→SC70.993 0.769 0.061 16.413 aW19
aIf the value of p<0.001, it is denoted by footnote “a”; otherwise, it will
show the exact numeral.
© ASCE 04015018-7 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
It has been mentioned above that in AMOS 7.0, specific-group
analysis can be taken by means of the variable grouping function.
Those enterprises with the same attribute were grouped into one
subgroup, and SEM was used to test the applicability of the core
dimension structure to each specific subgroup. The attribute types
of enterprise serves as the grouping variable in the specific-group
analysis.
Many researchers proposed the sample size requirement for
SEM. Schumacker and Lomax (1996) discovered that most SEM
sample sizes are above 200. Kline (1998) pointed out that if the
SEM sample size is lower than 100, researchers cannot get reliable
results. Huang (2004) suggested that if there are 15 observable
indicators, the sample size should be at least 75. Rigdon (1995)
thought that the smallest sample size is 150. In this study, consid-
ering the number of observable indicators and over 600 sample
individuals, the smallest specific-group sample size should be 100.
Twelve building contractors were categorized into one spe-
cific group, and five specialty trade contractors were categorized
into another specific group (Table 2). By means of SEM, the con-
structed safety-climate core dimension structure was revalidated.
The statistical results of the specific-group analysis are shown in
Tables 5and 6.
Table 5indicates that the two specific subgroups show a favor-
able goodness of fit, though some indices become inferior to those
Table 4. Summary of Overall Fit Test Results by SEM
Statistics Fitness criteria Values
Fitness judgment
(yes or no)
Absolute fit indices ———
χ2p>0.05 293.913 (p¼0.000 <0.05)No
RMR <0.05 0.036 Yes
RMSEA <0.08 (<0.05 is excellent, and <0.08 is good) 0.045 Yes (excellent)
GFI >0.90 0.950 Yes
AGFI >0.90 0.936 Yes
Incremental fit indices ———
NFI >0.90 0.925 Yes
IFI >0.90 0.957 Yes
RFI >0.90 0.913 Yes
TLI(NNFI) >0.90 0.950 Yes
CFI >0.90 0.957 Yes
Parsimonious goodness of fit ———
PGFI >0.50 0.734 Yes
PNFI >0.50 0.798 Yes
PCFI >0.50 0.826 Yes
CN >200 333 Yes
χ2=DOF <2.00 2.227 No
AIC The theoretical model should be lower than the
independent model and the saturated model
371.913 >342.000,371.913 <3945.452 No
CAIC The theoretical model should be lower than the
independent model and the saturated model
583.229 <1268.540,583.229 <4062.983 Yes
Table 5. Summary of Overall Goodness of Fit for Specific Groups by SEM
Statistics
House building contractors Specialty trade contractors
Values
Fitness
judgment
(yes or no) Values
Fitness
judgment
(yes or no)
Absolute fit indices —— — —
χ2213.461 (p¼0.000 <0.05) No 161.419 (p¼0.021 <0.05)No
RMR 0.056 No 0.067 No
RMSEA 0.054 Yes 0.044 Yes (excellent)
GFI 0.910 Yes 0.898 No
AGFI 0.879 No 0.862 No
Incremental fit indices —— — —
NFI 0.864 No 0.866 No
IFI 0.940 Yes 0.968 Yes
RFI 0.836 No 0.838 No
TLI(NNFI) 0.927 Yes 0.960 Yes
CFI 0.939 Yes 0.967 Yes
Parsimonious goodness of fit —— — —
PGFI 0.676 Yes 0.667 Yes
PNFI 0.717 Yes 0.718 Yes
PCFI 0.779 Yes 0.803 Yes
CN 168 No 134 No
χ2=DOF 1.681 Yes 1.271 Yes
AIC 301.461 <342.000301:461 <1607.660 Yes 497.306 <1103.128,497.306 <1687.779 Ye s
CAIC 249.419 <342.000249:419 <1236.194 Yes 423.165 <1017.238,423.165 <1307.272 Ye s
© ASCE 04015018-8 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
of the whole group in Table 4. About 64.7% of indicators in the
building contractors group and 58.82% of indices in the specialty
trade contractors group can well satisfy the fit criteria. Moreover,
two key indices, namely, the chi-square and chi-square divided by
DOF, improve much and are very close to 1. It should be noted that
the decline of goodness of fit in specific-group analysis can be
attributed to the significant shrink of the sample size.
Table 6shows that the standardized regression weights of the
two subgroups mostly lie between 0.50 and 0.95. For building con-
tractors and specialty trade contractors, 100 and 88.9% of weight
coefficients are significant, respectively. Thus, the observable
indicators can be explained by core dimensions, and the relation-
ships among core dimensions are strong. However, it should be
noted that for the specialty trade contractors, the relationships
among a few variables are insignificant. According to Table 6,it
is concluded that goodness of fit of specialty trade contractors is
inferior to that of building contractors.
The internal consistency reliability indices of safety-climate
core dimensions of specific subgroups are also examined. As to the
building contractors group, the internal consistency reliability in-
dices of the four dimensions from F1 to F4 are 0.7781, 0.8899,
0.5976, and 0.7257, respectively. As to the specialty trade contrac-
tors group, the indices of the four dimensions are 0.7219, 0.8287,
0.6202, and 0.7834, respectively.
In conclusion, the model has a favorable goodness of fit, high
internal quality, and acceptable reliability in the specific-group
analysis. Therefore, it was constructed successfully.
Discussion of the Results
By means of SEM, this research established a safety-climate core
dimension structure for construction enterprises. Based on the
findings of previous studies in this area, the model contains four
dimensions and three correlative links in between. This simplicity
and conciseness fit the model requirement of the SEM approach
well, and also leave enough room for researchers and practitioners
to design their specific dimensions out of the core dimensions.
Statistical analysis based on empirical data shows that the safety-
climate core dimension model has high validity and reliability, as
the basic fit test, overall fit test, internal structure fit test, and con-
struction quality test all have favorable results. Most significantly,
the high path coefficient values in the structural model validate
the empirical modeling work. Analysis of empirical data of specific
subgroups also support the model’s validity and reliability, and it
can be concluded that the core dimension structure of safety climate
is more suitable for building contractors than the specialty trade
contractors, because the former has a higher goodness of fit than
the latter. One reason lies in the generation of the a priori theoretical
framework. The existing research in the literature review is mostly
undertaken in building projects or enterprise. Dimensions and their
indicators are mostly derived from building environments, so the
core dimensions structure has a more favorable goodness of fit in
building contractors. The other reason is the homogeneity of the
selected specific subgroups. The 12 building contractors have sim-
ilar core businesses, management systems, organizational process
assets, and environmental factors. In other words, sample data in
the building contractors group have high homogeneity, and thus
have high convergence in the development of safety climate. In
contrast, the five specialty trade contractors have different specialty
trades such as installment, decoration and machinery manufactur-
ing, etc. They have different core businesses, which decreases the
convergence of sample data.
The purpose of the study is to assist in more valid and effective
designing of construction safety-climate scales by providing a
standardized basis and criterion. Core dimensions and their indica-
tors in the study can be directly applied in practice and also trans-
formed into specific dimensions based on the specific contexts.
The generation of specific dimensions is not only the flexible ap-
plication of the theoretical core dimension structure, but also the
enrichment and expansion of its content and scope, as well as the
improvement of its feasibility.
This study established correlative links among different core di-
mensions of safety climate. Safety priority, which also can also be
partly regarded as management commitment to safety, has direct
predictive relationships to both safety supervision, training, and
communication and safety rules and procedures. It also has indirect
predictive effects on safety involvement, which is mediated by
safety rules and procedures. Management commitment is proven
to be closely associated with safety leadership, and of paramount
importance for the development of safety policy as well as the al-
location of resources to safety (Mohamed 2002;Lu and Yang 2010;
Fang and Wu 2013). With the higher priority of safety over other
organizational and project goals, safety supervision will be rein-
forced to reduce safety risk levels as low as possible. Constant and
effective safety training will be provided to the whole staff, and
safety communication will be facilitated among the whole organi-
zation, driven by strong safety leadership (Dingsdag et al. 2006).
Safety commitment of the management can also enhance the ob-
servance of safety rules and procedures, and reduce risk-taking
behaviors of the employees as much as possible. In turn, voluntary
safety involvement of the staff can be promoted after violations of
safety rules are minimized and the whole organization benefits
from the safe environment (Fang et al. 2006). As follows, more
details of the essence and practical implications of the four core
dimensions are discussed and interpreted in order to offer guidance
for the application of the core dimension structure, including the
generation of observable indicators and specific dimensions.
Table 6. Results of Regression Analysis for Specific Groups by SEM
Links
House building
contractors
Specialty trade
contractors
Standardized
regression
weights p
Standardized
regression
weights p
F1→F30.663 a0.560 0.054
F1→F20.917 a0.843 a
F3→F40.851 a0.957 0.045
F1→SC60.677 —0.744 —
F1→SC50.607 a0.847 a
F1→SC40.589 a0.540 a
F1→SC10.635 a0.582 a
F2→SC11 0.252 —0.520 —
F2→SC12 0.783 a0.754 a
F2→SC13 0.702 a0.712 a
F2→SC14 0.696 a0.830 a
F2→SC15 0.621 a0.780 a
F2→SC16 0.736 a0.831 a
F2→SC17 0.625 a0.671 a
F3→SC18 0.303 —0.186 —
F3→SC10 0.611 a0.567 0.052
F3→SC30.577 a0.698 0.049
F3→SC20.643 a0.591 0.048
F4→SC80.742 —0.792 —
F4→SC90.674 a0.478 a
F4→SC70.799 a0.763 a
aIf the value of p<0.001, it is denoted by footnote “a”; otherwise, it will
show the exact numeral.
© ASCE 04015018-9 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
Safety priority is the center of core dimension structure of safety
climate; that is, it exerts direct and indirect significant influences
on the other three dimensions. It can be manifested in practice
by the commitment and endeavor of the management, especially
the top management. Hofmann et al. (2003) said that only when
management commitment is translated into perceivable actions
or utterance can it really improve safety climate. Measures showing
high safety priorities by the top management include constantly
communicating good safety management ideas and policies within
the organization and its stakeholders, keeping regular visibility
on the construction sites, actively organizing and participating in
safety meetings, and ranking safety over cost and schedule in de-
cision making, etc.
The dimension of safety supervision, training, and communica-
tion is one of the major measures of the management to achieve its
commitment, and also the dominant part of organizational routine
safety-promoting activities. Supervision, training, and communica-
tion are the basic ways of information exchange and mutual influ-
ence between employers and employees, between the high-level
management and the low-level management, and among employ-
ees. Safety supervision is the restraint, induction, and guidance of
the workers by supervisors in accordance with mandatory rules
and regulations. Safety training is the formal and informal safety
education and instruction to the inexperienced apprentices by the
experienced mentors. Safety communication is the exchange and
sharing of information related to safety issues among the whole
organizational staff. Management actions related to this core di-
mension include regular internal safety auditing; regular on-site
safety inspection; safety orientation for new employees and work-
force; safety training for the whole staff, such as those on funda-
mental safety knowledge and skills, safety risk assessment, and
safety emergency response; and regular safety meetings for in-
depth safety discussions.
Safety rules and procedures contains two aspects, i.e., safety
rules and safety procedures. The former is the self-oriented man-
datory codes and standards formulated by the organization based
on governmental laws and regulations, and the latter is the in-
structions designed for the purpose of enhancing work processes
and improving work safety conditions. In practice, only when a
thorough system of safety rules and procedures has been estab-
lished can the organization undertake safety supervision, training,
and communication successfully, and eventually reduce the viola-
tion of safety rules and procedures. Specific measures related to this
dimension include establishing skilled and experienced safety man-
agement teams, enacting effective accident prevention regulations
and procedures, and guaranteeing full implementation of on-site
regulations and procedures.
Safety involvement is a managerial method targeting the average
employees, with the major aim of encouraging the involvement and
participation of the workforce (Mohamed 2002). Its basic manifes-
tation is placing employees in the everyday information flows
and decision-making processes. With sufficient safety involvement,
employees will pay more attention to their behavioral safety be-
cause they are essentially involved in the decision-making proc-
esses and therefore responsible for the consequences of their
unsafe acts. With these management tactics, managers can build
up the employees’awareness of ownership, so as to lead them to
behavioral safety improvement. Workers’involvement includes
such issues as procedures for reporting injuries and potentially haz-
ardous situations.
Last but not least, it should be noted that the four aspects of
safety climate mentioned above are not at all isolated, but highly
interrelated with each other, and they together are the actual kernel
of a safety management system, which is the main embodiment of
organizational safety culture (Fernández-Mu˜niz et al. 2007). The
work of this study provides significant practical implications for
organizational safety management; that is, in order to possess a
favorable safety climate and in turn achieve a high safety perfor-
mance level, the organization, especially the top management,
should stick to high safety priority; develop and update safety rules
and procedures consisting of both mandatory codes and suggestive
instructions; guarantee the full implementation of rules and pro-
cedures by safety supervision, training, and communication; and
promote safety involvement to improve employees’loyalty to the
organization.
Limitations and Future Research
Potential limitations of the research are discussed as follows. First,
in this empirical research, due to the limited number of participants,
it was not possible to make specific-group analysis to cover every
single enterprise, but rather just selected building contractors and
some specialty trade contractors. Ideally, every enterprise should be
analyzed, and when all these specific subgroups satisfy the criteria
of goodness of fit and reliability, the revalidation of the safety-
climate core dimensions can be achieved. In addition, more field
research works are encouraged to test the identified links among
different core dimensions.
Second, the questionnaire used in this study is based on the pre-
vious research of safety climate, and may not involve all observable
indicators of those core dimensions. The study by no means limits
researchers and practitioners to only one safety-climate measure-
ment scale, but rather encourages them to apply more concrete cli-
mate indicators in different settings.
Third, in the current safety-climate factor research, one of the
common methodological flaws is the difficulty in the naming of
factors (especially when the method is exploratory in essence).
To some extent, some names cannot embrace all items contained
in the factor. The core dimensions are highly concentrated and have
abundant implications, so the temporary names in this research may
not be perfect. To practitioners, especially those who have limited
theoretical knowledge, these names may lead to misunderstand-
ings. The improvement of naming approaches will be focused on
in future research.
Conclusions
The prevalent disparity and divergence in safety-climate dimension
identification in academia cause confusion and inconvenience to
both construction researchers and practitioners in terms of safety-
climate measurement. Previous research has various limitations,
and thus is not able to guide practical safety-climate measurement
effectively. This study aims to contribute to the standardization of
construction safety-climate measurement by providing a unified set
of criteria, namely, a core dimension structure of safety climate, for
both researchers and practitioners:
1. It defined the core dimensions and specific dimensions of
safety climate based on previous studies. Distinguishing core
dimensions and specific dimensions facilitates the work of
discovering common and shared characteristics of safety-
climate factor structures developed by various studies, so as
to find out the true and inherent attributes of the safety-climate
construct. Relationships between core dimensions and specific
dimensions are interpreted and illustrated, which has not yet
been covered in previous studies. This can serve as the most
crucial step in resolving the disparity and divergence in safety-
climate dimension identification. Identification of these core
© ASCE 04015018-10 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
dimensions will not only lead to a better understanding of
safety climate, but also facilitate safety-climate measurement
in the whole construction industry.
2. Based on a thorough literature review, this study identified
core dimensions of safety climate and their internal relation-
ships, so as to build a core dimension structure for construc-
tion enterprises. The four core dimensions are safety priority;
safety supervision, training, and communication;safety rules
and procedures; and safety involvement. SEM analysis empiri-
cally validated the core dimension structure from different
aspects and perspectives. The core dimension structure is mul-
tilayered, which conforms to the properties of the safety-
climate construct. It is flexible and can be transformed or give
rise to specific dimensions in specific contexts. In other words,
core dimensions and their indicators in the study can be not
only directly applied in practice and other studies, but also
transformed into specific dimensions based on the specific
contexts.
3. Core dimensions and their mutual relationships also have
abundant practical implications, which are specified in detail
in the paper, so as to offer guidance for the application of the
core dimension structure, including the generation of observa-
ble indicators and specific dimensions. Core dimensions and
their relationships together are the actual kernel of safety man-
agement systems, and can offer guidance to effective construc-
tion safety management.
This study can in no way be regarded as the final step on safety-
climate study. In contrast, it is just the starting point. More research
and practical efforts should be made to test the identified links
among different core dimensions and develop effective specific di-
mensions and concrete indicators. A generic dimension structure of
safety dimensions exists (Zohar 2010), but measurement scales can
be diversified. Variations in the factor structure of safety climate
can exist in different contexts, but they should be based on the core
dimensions of safety climate, or else they may deviate from the
essence of safety climate. In this respect, the core dimension struc-
ture of this study can provide instructions for a better and more
valid development of safety-climate factor structure in different
contexts. It is a starting point for future studies exploring the unified
criteria in safety-climate application more profoundly. Construc-
tion safety-climate core dimension structure, together with its rela-
tionship with specific dimensions and observable indicators, can
guide more effective measurement scale design in different settings,
rather than limit the scale to only one format.
Acknowledgments
Acknowledgments are addressed to the National Natural Science
Foundation of China (Grant Nos. 71172013 and 71401191) and
121 Youth Ph.D. Development Foundation of Central University
of Finance and Economics (Grant No. QBJ1411). The authors
would also like to thank the anonymous reviewers for their invalu-
able and constructive comments on an earlier draft of this paper,
which contributed to the substantial revisions made since that
time.
References
Arbuckle, J. L. (1995). Amos™7.0 user’s guide, Amos Development
Corporation, Crawfordville, FL.
Bagozzi, R. P., and Yi, Y. (1988). “On the evaluation of structural equation
models.”J. Acad. Mark. Sci., 16(1), 74–94.
Bearden, W. O., Sharma, S., and Teel, J. E. (1982). “Sample size effects
on chi square and other statistics used in evaluating causal models.”
J. Mark. Res., 19(4), 425–430.
Bollen, K. A. (1989). Structural equations with latent variables, Wiley,
New York.
Byrne, B. M. (2013). Structural equation modeling with AMOS: Basic
concepts, applications, and programming, Routledge, New York.
DeDobbeleer, N., and Béland, F. (1991). “A safety climate measure for
construction sites.”J. Saf. Res., 22(2), 97–103.
Dingsdag, D. P., Biggs, H. C., Sheahan, V. L., and Cipolla, D. J. (2006).
A construction safety competency framework: Improving OH&S
performance by creating and maintaining a safety culture, CRC for
Construction Innovation, Brisbane, QLD, Australia.
Fang, D., and Wu, H. (2013). “Development of a safety culture interaction
(SCI) model for construction projects.”Saf. Sci., 57, 138–149.
Fang, D., Yang, C., and Wong, L. (2006). “Safety climate in construction
industry: A case study in Hong Kong.”J. Constr. Eng. Manage.,
10.1061/(ASCE)0733-9364(2006)132:6(573), 573–584.
Fernández-Mu˜niz, B., Montes-Pe´on, J. M., Vázquez-Ordás, C. J. (2007).
“Safety management system: Development and validation of a multi-
dimensional scale.”J. Loss Prev. Process Ind., 20(1), 52–68.
Flin, R., Mearns, K., O’Connor, P., and Bryden, R. (2000). “Measuring
safety climate: Identifying the common features.”Saf. Sci., 34(1–3),
177–192.
Flynn, B. B., Schroeder, R. G., and Sakakibara, S. (1994). “
A framework
for quality management research and an associated measurement instru-
ment.”J. Oper. Manage., 11(4), 339–366.
Gadd, S. (2002). Safety culture: A review of the literature, Human Factors
Group, Health and Safety Laboratory, Sheffield, U.K.
Glendon, A. I., and Litherland, D. K. (2001). “Safety climate factors, group
differences and safety behaviour in road construction.”Saf. Sci., 39(3),
157–188.
Glendon, A. I., and Stanton, N. A. (2000). “Perspectives on safety culture.”
Saf. Sci., 34(1–3), 193–214.
Guldenmund, F. W. (2000). “The nature of safety culture: A review of
theory and research.”Saf. Sci., 34(1), 215–257.
Hair, J. F., Anderson, R. E., Tatham, R. L., and Black, W. C. (1998).
Multivariate data analysis, 5th Ed., Prentice Hall, Upper Saddle River,
NJ.
Herrero, S. G., Salda˜na, M. A. M., del Campo, M. A. M., and Ritzel, D. O.
(2002). “From the traditional concept of safety management to safety
integrated with quality.”J. Saf. Res., 33(1), 1–20.
Hofmann, D. A., Morgeson, F. P., and Gerras, S. J. (2003). “Climate as a
moderator of the relationship between leader-member exchange and
content specific citizenship: Safety climate as an exemplar.”J. Appl.
Psychol., 88(1), 170–178.
Hofstede, G. R. (1991). Cultures and organisations: Software of the mind,
McGraw-Hill, London.
Hong Kong Labour Department. (2011). “Occupational safety and
health statistics 2010.”〈http:www.labour.gov.hk/eng/osh/pdf/OSH
_Statistics2010.pdf〉(Feb. 22, 2014).
Hooper, D., Coughlan, J., and Mullen, M. R. (2008). “Structural equation
modelling: Guidelines for determining model fit.”Electron. J. Bus. Res.
Methods, 6(1), 53–60.
Huang, F. M. (2004). Theory and application of structural equation model,
Wunan, Taipei.
Huang, Y. H., Ho, M., Smith, G. S., and Chen, P. Y. (2006). “Safety climate
and self-reported injury: Assessing the mediating role of employee
safety control.”Accid. Anal. Prev., 38(3), 425–433.
Jones, A. P., and James, L. R. (1979). “Psychological climate: Dimensions
and relationships of individual and aggregated work environment per-
ceptions.”Organ. Behav. Hum. Perform., 23(2), 201–250.
Kline, R. B. (1998). Principles and practice of structural equation model-
ing, Guilford Press, New York.
Lu, C. S., and Yang, C. S. (2010). “Safety leadership and safety behavior in
container terminal operations.”Saf. Sci., 48(2), 123–134.
Mohamed, S. (2002). “Safety climate in construction site environments.”
J. Constr. Eng. Manage.,10.1061/(ASCE)0733-9364(2002)128:5(375),
375–384.
© ASCE 04015018-11 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
MOHURD. (2013). “Safety accidents report in house building and munici-
pal engineering projects in December 2013.”〈http://www.mohurd.gov
.cn/zcfg/jsbwj_0/jsbwjgczl/201401/t20140117_216900.html〉(Jan. 31,
2014).
PMI (Project Management Institute). (2008). A guide to the project
management body of knowledge (PMBOK guide), 4th Ed., Project
Management Institute, Upper Darby, PA.
Reason, J. (1990). Human error, Cambridge University Press, Cambridge,
U.K.
Rigdon, E. (1995). “A necessary and sufficient identification rule for
structural equation models estimated.”Multivar. Behav. Res., 30(3),
359–383.
Schumacker, R. E., and Lomax, R. G. (1996). A beginner’s guide to
structural equation modeling, Lawrence Erlbaum Associates, Mahwah,
NJ.
U.K. Health and Safety Executive. (2011). “Construction: Work related
injuries and ill health.”〈http://www.hse.gov.uk/statistics/industry/
construction/construction.pdf〉(Feb. 24, 2014).
U.S. Department of Labor. (2010). “Census of fatal occupational injuries
charts, 1992–2010.”〈http://www.bls.gov/iif/oshwc/cfoi/cfch0009.pdf〉
(Feb. 24, 2014).
Van Vianen, A. E., De Pater, I. E., Bechtoldt, M. N., and Evers, A. (2011).
“The strength and quality of climate perceptions.”J. Manage. Psychol.,
26(1), 77–92.
Zhou, Q., Fang, D., and Mohamed, S. (2011). “Safety climate improve-
ment: Case study in a Chinese construction company.”J. Constr.
Eng. Manage.,10.1061/(ASCE)CO.1943-7862.0000241,86–95.
Zohar, D. (1980). “Safety climate in industrial organizations: Theoretical
and applied implications.”J. Appl. Psychol., 65(1), 96–102.
Zohar, D. (2008). “Safety climate and beyond: A multi-level multi-climate
framework.”Saf. Sci., 46(3), 376–387.
Zohar, D. (2010). “Thirty years of safety climate research: Reflections and
future directions.”Accid. Anal. Prev., 42(5), 1517–1522.
Zohar, D., and Luria, G. (2005). “A multilevel model of safety climate:
Cross-level relationships between organization and group-level
climates.”J. Appl. Psychol., 90(4), 616–628.
© ASCE 04015018-12 J. Constr. Eng. Manage.
J. Constr. Eng. Manage. 2015.141.
Downloaded from ascelibrary.org by TSINGHUA UNIVERSITY on 09/02/15. Copyright ASCE. For personal use only; all rights reserved.
