Available online at www.sciencedirect.

com

ScienceDirect
International Journal of Project Management 34 (2016) 789 – 805
www.elsevier.com/locate/ijproman

How safety leadership works among owners, contractors and

subcontractors in construction projects
Chunlin Wu a , Feng Wang b , Patrick X.W. Zou c , Dongping Fang a,⁎
a
(Tsinghua-Gammon) Construction Safety Research Center, Department of Construction Management, School of Civil Engineering, Tsinghua University,
Beijing 100084, China
b
Lanzhou Railway Bureau, 72 Hezheng Road, Lanzhou 730099, Gansu Province, China
c
Department of Civil and Construction Engineering & Centre for Sustainable Infrastructure, Swinburne University of Technology,
Hawthorn, Victoria 3122, Australia

Received 10 November 2015; received in revised form 18 February 2016; accepted 25 February 2016
Available online 2 April 2016

Abstract

Leadership is a key factor impacting construction safety, but previous research merely investigated the single-level relationship between safety
leadership and safety performance and ignored the leadership interaction between different project stakeholders. To fill this gap, this paper aims to
examine the relationships between safety leaderships of project owners, contractors and subcontractors and discover leadership dimensions which
significantly impact construction safety performance. An impacting mechanism involving owner safety leadership, contractor safety leadership and
subcontractor safety leadership are hypothetically proposed and empirically tested. The results show that significant relationships exist between
safety leaderships of the three key stakeholders. Project safety culture acts as a significant mediator in these relationships. In addition, among all
leadership dimensions of owners and contractors, safety influence and role modeling has the widest range of influence on project safety culture and
other stakeholders' safety leadership. As such, it is suggested that owners and contractors should cultivate charisma and the ability of being
influential about ideals in project managers and require them to behave as role models for others. The results also show that the caring dimension of
leadership is more required in the frontline environment. Supervisors need more attention and support from the contractor in their well-being and
organizational identification for the project. In conclusion, this paper establishes clear leadership impacting paths from owners to site supervisors of
subcontractors in construction projects, which provides insights into effective ways to implement managerial measures and publicize policies and
values to construction sites.
© 2016 Elsevier Ltd. APM and IPMA. All rights reserved.

Keywords: Safety leadership; Safety culture; Construction project; Owner; Contractor; Subcontractor

1. Introduction like high-speed railways (HSR) in China increased the

occupational fatalities in construction. In recent years, increasing
The construction industry is reported worldwide as having emphasis on accident prevention has been exemplified by the
the highest occupational injury rates (Abudayyeh et al., 2006; introduction of more and more occupational health and safety
Fang and Wu, 2013). In 2012, 2437 people died of (OHS) provisions. However, the effects of these provisions have
occupational accidents in Chinese construction industry, been far from satisfactory. Despite the existence of OHS laws,
which makes construction the most dangerous industry, accident frequency in construction still remains at a high level. As
exceeding any other industry including coal-mining (Chinese many researchers reported, the continuous unsafe conditions are
State Administration of Work Safety, 2013). Particularly, the mainly due to a misalignment of management commitment and
rapid development of huge and complex infrastructure projects subordinates' actions (Suraji et al., 2001; Arquillos et al., 2012;
Sunindijo and Zou, 2012; Martin and Lewis, 2014). More
⁎ Corresponding author. Tel.: +86 10 62770380; fax: +86 10 62773661. specifically, safety has remained a concept held by managers and
E-mail address: [email protected] (D. Fang). it is not fully disseminated to its subordinates, and managerial

http://dx.doi.org/10.1016/j.ijproman.2016.02.013
0263-7863/00/© 2016 Elsevier Ltd. APM and IPMA. All rights reserved.
790 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805

measures cannot be fully implemented to take effect on (Fang and Wu, 2013). Project managers from owners, construction
construction sites. These problems are in turn attributed to managers from contractors and supervisors from subcontractors
construction managers' lack of safety leadership (Tam et al., are selected as objects of empirical study when analyzing the
2004; Martin and Lewis, 2014), which improves safety perfor- three stakeholders' safety leaderships because they are represen-
mance in highly hazardous and complex working environments tative of the three stakeholders' staff, respectively. Safety
(Flin and Yule, 2004). Effective leadership plays an important role leaderships of the above three types of personnel are hereafter
in ensuring the success of temporary organizations facing a high called owner project managers' safety leadership (OPSL),
degree of uncertainty, which corresponds to the characteristics of contractor construction managers' safety leadership (CCSL) and
construction projects (Tyssen et al., 2014). A project team will be subcontractor supervisors' safety leadership (SSSL).
either empowered to succeed or condemned to fail largely by the Based on the previous studies which validated the impact of
quality of the leadership skills of project managers. Therefore, SSSL on safety performance of construction projects (Hoffmeister
strong safety leadership should be the key for improvement, et al., 2014; Fang et al., 2015), this study aims to explore
especially for those countries whose construction industries are relationships between OPSL, CCSL and SSSL. Three research
facing significant safety challenges and which need transforma- objectives are specified: (1) develop and validate measurement
tional development (Construction Users Roundtable, 2012). scales of safety leadership and safety culture of construction
Many empirical studies have demonstrated the importance of projects; (2) depict and validate the exact impacting paths between
leadership to safety (Barling et al., 2002; Zohar, 2002; Griffin and OPSL, CCSL and SSSL; and (3) discover leadership dimensions
Hu, 2013). Safety leadership is a significant antecedent of safety which can significantly impact safety leadership of other
culture and worker safety behavior (O'Dea and Flin, 2001; Wu stakeholders. The above analysis is able to contribute to leadership
et al., 2011; Griffin and Hu, 2013). However, safety leadership research by interpreting the role of safety leadership in
research within the construction industry is scarce compared with construction safety improvement in a holistic manner, and by
other industries (Keegan and Den Hartog, 2004; Chan and Chan, providing a framework for future theory development. From a
2005; Ofori and Toor, 2012). From a research method pragmatic and practical perspective, determining the relative
perspective, causality between safety leadership, safety culture contributions of leadership dimensions to safety can aid
and safety performance has not been fully validated. The main practitioners in developing better interventions for safety leader-
reason is that conclusions rely largely on cross-sectional studies, ship improvement. If some dimensions are much more important
and longitudinal studies of safety leadership are few. Cross- than the others in predicting outcomes, it is logical to allocate
sectional studies obtain empirical links or correlations rather resources to the more important dimensions. Empirical data are
than causal relations between variables. Causalities should be collected from high-speed railway construction projects in China
validated with time-series data and panel data, both of which are because they are highly complex construction projects with huge
obtained by longitudinal studies (Wooldridge, 2010). numbers of safety risks. The fulfillment of safety leadership and
Furthermore, although many researchers argued that senior safety management in HSR construction projects are typical and
leadership exerts the most significant influence on organizational representative. In addition, owners tend to play a crucial role in
performance (Kimmel, 1981; Smallman and John, 2001; Reid safety leadership and safety management of HSR construction
et al., 2008), safety leadership research has generally focused on projects, which correspond to the research objectives pertaining
the supervisory level (Lu and Yang, 2010; Martínez-Córcoles to exploring how owners' safety leadership works on other
et al., 2013; Hoffmeister et al., 2014). The higher the position stakeholders.
individuals are, the greater their potential to influence organiza-
tional outcomes (Flin and Yule, 2004). In construction projects, 2. Literature review and hypothesis development
owners' commitment and policies are transmitted by contractors
and subcontractors to affect the workers' priorities, attitudes and This section theoretically determines the dimensions of safety
behaviors. However, previous empirical studies tended to solely leadership and relationships between the three stakeholders'
address the single-level effects of safety leadership (mostly safety leaderships in construction projects in order to establish the
supervisors' leadership) on safety performance (DeChurch et al., research model.
2010; Wu et al., 2011; Martínez-Córcoles et al., 2013), but
somewhat ignored the relationships between safety leaderships 2.1. Safety leadership and its dimensions
of different stakeholders in construction projects, like owners
and contractors. Details of the underlying mechanisms by which Several previous studies have examined the effects of safety
leadership may influence safety are not yet well depicted and leadership on safety performance (O'Dea and Flin, 2001;
understood (Hoffmeister et al., 2014). The interactions and Barling et al., 2002; Zohar, 2002; Neal and Griffin, 2006;
combined effects on safety performance of the above three Griffin and Hu, 2013). Safety leadership is a sub-system of
stakeholders' leaderships have to be considered and explored leadership (Pater, 2001), and can be defined as “the process of
holistically. interaction between leaders and followers, through which leaders
This study focuses on three project stakeholders' safety can exert their influence on followers to achieve organizational
leaderships during the construction phase of projects, i.e. project safety goals under the circumstances of organizational and
owners, contractors and subcontractors. These are the three main individual factors” (Wu et al., 2008). Leadership is fully implicated
parties constituting the construction project management team in safety, and a majority of previous studies focused on the Full
 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805 791

Range Leadership (FRL) model, which mainly contains transfor- ultimately improve their motivation and knowledge in safety
mational and transactional leaderships (Barling et al., 2002; Lu and management. Safety caring and individual respect indicates the
Yang, 2010). Transactional leadership is related to monitoring and degree to which the leader shows interest in subordinates'
rewarding whereas transformational leadership is directed towards well-being, assigns projects individually, and pays attention to
inspiring and genuinely motivating the workforce (Reid et al., those who seem less involved in the group. Safety controlling and
2008). Transformational leadership has four dimensions: idealized performance management focuses on monitoring task execution
influence, inspirational motivation, intellectual stimulation and for any problems that might arise and correcting those problems
individualized consideration. Transactional leadership contains to maintain current performance. It also shows the degree to
two dimensions: contingent reward and management-by- which the leader tells others what to do in order to be rewarded,
exception (Avolio et al., 1999). The FRL model containing emphasizes what you expect from them and recognizes their
transformational leadership and transactional leadership can be accomplishments. It is the institutional basis for the realization of
regarded as the foundation of the safety leadership factor structure the owner's safety philosophy, and also the foundation for the
(Wu et al., 2015). Many other studies constructed their specific other three safety leadership dimensions in our paper because
dimensions of safety leadership in order to better measure it. leaders have to be able to control and remove any organizational
Dimensions of safety leadership reported in the literature include obstacles against goal realization before caring, motivating,
safety motivation, safety inspiring, safety policy, safety concern, influencing and conveying the right philosophy (Cooper, 1998;
safety monitoring, safety learning, safety coaching, safety caring, Wu et al., 2015).
safety controlling, etc. (Wu, 2005; Wu et al., 2008; Lu and Yang, For SSSL, i.e. subcontractor supervisor safety leadership,
2010; Griffin and Hu, 2013). Almost all these studies took many previous studies applied the empowering leadership scale
into account different aspects of transformational/transactional for its measurement (Martínez-Córcoles et al., 2011; Martínez-
leadership. As a result, the identified dimensions have close Córcoles et al., 2012; Bobbio et al., 2012). Empowering
relationships with those of transformational/transactional leader- leadership is a group (team) leadership model that is suitable for
ship. For example, safety motivation and safety inspiring have capturing the specific influence of group leaders (which are
similar meanings with idealized influence and inspirational supervisors in construction projects) on group performance
motivation. Safety learning and safety coaching overlap with (Arnold et al., 2000). Five dimensions are involved, including
intellectual stimulation. Safety caring is related to individual leading by example, participative decision making, coaching,
consideration. Safety policy, safety monitoring and safety informing and showing concern/interacting with employees.
controlling are closely linked to contingent reward and
management-by-exception (Wu et al., 2015). 2.2. Relationships between project stakeholders' safety leaderships
Based on the FRL model, Wu et al. (2015) made an
ethnographic study on construction projects and identified four Among the existing studies on leadership's impacting
categories of safety leadership practices, which interpreted the mechanisms, the majority merely examines single-level rela-
types of safety leaders required in the construction industry. They tionships, linking leadership to organizational outcomes. A
are safety influence and role modeling, safety motivation and valuable direction in this area is to develop theories on the
coaching, safety caring and individual respect, safety controlling top-down mechanisms through which senior leaders impact
and performance management. This paper selects the above middle leadership and in turn frontline leaders, and ultimately
four categories as the four dimensions of safety leadership for impact their organizations (DeChurch et al., 2010). When it
owner project managers and contractor construction managers comes to construction safety leadership, to achieve the above
(i.e. OPSL and CCSL), and further develops safety leadership objectives, relationships between OPSL and CCSL as well as
measurement scales based on these dimensions. The first three between CCSL and SSSL have to be depicted.
dimensions are associated with transformational leadership, and Leadership is essentially behaviors related to leading
the last one relates to transactional leadership. Safety influence subordinates to achieve organizational goals. Safety behavior
and role modeling indicates whether the leader holds subordi- of a given management level can be seen as safety leadership
nates' trust, maintains their faith and respect, shows dedication to for their lower-level management. Owner project managers'
them, appeals to their hopes and dreams and acts as their role purpose to influence contractor construction managers is to
model in their own safety management and leadership. It is based enhance their positive influence to supervisors. By influencing
on the leaders' charisma and the capability of being influential supervisors, contractor construction managers' ultimate goal is
about ideals, which in turn requires the highest level of morality improving workers' safety behavior. Moreover, there is also a
of leaders (Bass, 1999). Favorable role models built by leaders high consistency between dimensions of safety management
are able to predict better performance in other dimensions of behavior and safety leadership in the current literature. Thus,
subordinates' safety behaviors, including motivation, perception, although the relationship between leaderships tended to be
knowledge and managerial capabilities, so it is the first and ignored by the academia, studies on relationships between safety
foremost dimension of transformational leadership (Bass, 1999). leadership and safety behavior can be instructive and give
Safety motivation and coaching measures the degree to which insights to this study.
the leader provides a vision, uses appropriate symbols and images Relationships between safety leadership and safety behavior
to help subordinates focus on their safety work, encourages them were explored in the literature. In these relationships, safety
to be creative in looking at old problems in new ways, and culture (or its snapshot, safety climate) is widely proven to be
792 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805

the mediating variable (O'Dea and Flin, 2001; Wu et al., 2008; leadership (i.e. inspiring and motivating subordinates by full
Wu et al., 2011; Martínez-Córcoles et al., 2011). As follows, involvement and participation in work activities) and safety
based on the existing literatures, several hypotheses are participation. Safety compliance and safety participation are two
generated to establish the research model and will be validated fundamental safety behavior dimensions. Martínez-Córcoles et al.
in subsequent sections by empirical data. (2012) showed that those leadership behaviors as participative
The research model consists of two levels: one is the decision-making and interaction with subordinates are associated
impact of OPSL on CCSL (Fig. 1) and the other is the impact of with safety participation of all group members.
CCSL on SSSL (Fig. 2). The impact of OPSL on SSSL is not Thus, it is reasonably hypothesized that each OPSL dimension
considered because of the absence of cross-level distant influences the same and corresponding CCSL dimension. Besides,
leadership (Chun et al., 2009) in construction projects. Without as mentioned above, safety influence and role modeling (as the
the explicit contracting relationship, project owners and subcon- principal dimension) of OPSL also influences the other three
tractors tend to have little interaction with each other. Most of their dimensions of CCSL because it, based on charisma and the
communications take place through contractors. A pilot investi- capability of ideal role modeling, can be regarded as an all-
gation was undertaken on construction projects by the authors and inclusive transformational leadership dimension in influencing
found that most supervisors could not effectively appraise distant subordinates (Bass, 1999). Safety motivation and coaching (OPSL)
leadership of owners because they rarely interacted with owners predicts safety controlling and performance management (CCSL)
during construction. In Figs. 1 and 2, codes of all hypotheses are because it can improve subordinates' motivation and knowledge to
exhibited on their corresponding arrows. enforce effective safety monitoring and controlling, and thus
enhances their performance management abilities (Lu and Yang,
2.2.1. Relationships between OPSL and CCSL 2010). Thus, Hypothesis 1, Hypothesis 2 and Hypothesis 3 (H1,
Based on literature review, it is hypothesized that there are H2 and H3) are developed as follows:
significant relationships between OPSL and CCSL. It has been
H1. Each OPSL dimension positively and directly predicts its
widely proven that safety leadership predicts subordinates'
corresponding dimension of CCSL (H1a to H1d in Fig. 1).
safety behaviors which have the same or similar attributes with it.
For example, Hoffmeister et al. (2014) validated the positive links H2. Safety influence and role modeling (OPSL) positively and
between transactional leadership (i.e. monitoring and controlling directly predicts three CCSL dimensions, i.e. safety motivation
subordinates' compliance with organizational rules and proce- and coaching (H2a), safety caring and individual respect (H2b)
dures) and safety compliance, as well as between transformational and safety controlling and performance management (H2c).

OPSL Project safety culture CCSL

H4e
H4a

Safety influence and H1a Safety influence and

role modeling (OPSL) role modeling (CCSL)

H2a
H4f
H4b

Safety motivation and H1b Safety motivation and

coaching (OPSL) coaching (CCSL)

H2b
H4c H4g
H3 H2c

Safety caring and H1c Safety caring and

individual respect individual respect
(OPSL) (CCSL)

H4d H4h

Safety controlling and H1d Safety controlling and

performance performance
management (OPSL) management (CCSL)

Fig. 1. Hypothesized relationships between OPSL and CCSL.

 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805 793

Project safety culture

SSSL
H9e

H9a
CCSL
Leading by example
(SSSL)
H5a

Safety influence and

role modeling (CCSL) H9f
H5b

Participative decision-
H9b H5c
H6a making (SSSL)

Safety motivation and

H9g
coaching (CCSL)

H9c H6b Coaching (SSSL)

H5d
Safety caring and H7a H6c
individual respect H9h
(CCSL)
H5e
H7b
H9d Informing (SSSL)

Safety controlling and

performance H8 H9i
management (CCSL)
showing concern/
interacting with
workers (SSSL)

Fig. 2. Hypothesized relationships between CCSL and SSSL.

H3. Safety motivation and coaching (OPSL) positively and safety culture, and H4e to H4h represent links between project
directly predicts safety controlling and performance management safety culture and CCSL dimensions).
(CCSL).
2.2.2. Relationships between CCSL and SSSL
Project safety culture (Fang and Wu, 2013) plays as the Likewise, significant relationships exist between CCSL and
mediator in the relationship between OPSL and CCSL. OPSL SSSL. Safety influence and role modeling (as the principal
positively influences project safety culture, whether by role dimension) of CCSL influences all SSSL dimensions, whether
modeling, coaching, caring or controlling (Wu et al., 2010; in transformational or transactional styles.
Hoffmeister et al., 2014). Safety culture is also able to deeply and
holistically enhance safety behavior of all personnel (Cooper, H5. Safety influence and role modeling (CCSL) positively and
2000; Choudhry et al., 2007). Although owner leadership directly predicts all SSSL dimensions, i.e. leading by example
dimensions can directly predict contractor leadership dimensions, (H5a), participative decision-making (H5b), coaching (H5c),
their indirect impact through the mediation of safety culture can informing (H5d) and showing concern/interacting with workers
enhance middle leadership in a more profound and comprehen- (H5e).
sive manner (Fang et al., 2015). Hypothesis 4 is then proposed as
follows: Safety motivation and coaching of contractor construction
managers promotes “coaching” in SSSL because they are
H4. Project safety culture mediates the relationships between homogeneous leadership dimensions. This dimension also
OPSL and CCSL. In other words, each OPSL dimension predicts predicts the dimension “informing” because it provides a clear
project safety culture, which in turn predicts all CCSL dimensions safety vision and sufficient knowledge for supervisors to
(H4a to H4d represent links between OPSL dimensions and project disseminate safety mission, philosophy, fundamental skills and
794 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805

other important information to workers (Arnold et al., 2000). 3. Methodology

Furthermore, by encouraging supervisors to be creative in
looking at old problems in new ways and looking for innovative A short-term longitudinal study was carried out in HSR
viewpoints on safety, contractors are able to motivate supervisors construction projects in China in order to test the above hypotheses.
to actively use workers' inputs in decision-making, and to display Invitation letters explaining the purpose of the survey and seeking
behaviors such as encouraging workers to express their own permission to conduct site visits were distributed to more than 20
ideas and opinions (Martínez-Córcoles et al., 2011). Thus, this projects, covering almost the entire HSR construction scale in
dimension can also facilitate participative decision-making of Eastern China. Seven projects responded with positive attitudes to
SSSL. participate in the survey. Four of them have a designed operating
speed of 300 km/h and the other three have a designed operating
H6. Safety motivation and coaching (CCSL) positively and speed of 250 km/h. The average length of the seven HSRs is about
directly predicts three SSSL dimensions, i.e. participative 183 km and the average investment is 2.56 billion yuan (Chinese
decision-making (H6a), coaching (H6b) and informing (H6c). RMB). Most of these projects were in the middle of construction
stage, and their safety management systems worked well and their
Safety caring and individual respect directly predicts the project safety cultures were stable, which enhanced the quality and
dimension “showing concern/interacting with employees” of reliability of empirical data. The projects that declined the
SSSL because they share with each other the same attribute invitation felt that they could not allow researchers to visit their
which is individual concern and humanistic care. Moreover, sites because of the special site safety requirements or the difficulty
this dimension also embodies paying attention to those who in gathering enough number of respondents in a fixed place.
seem less involved in safety, and thus predicts the “participative The study had a non-experimental design, using self-reported
decision-making” dimension of SSSL. questionnaire data from three measurement rounds, R1–R3,
performed with an interval of four months from May 2014 to
H7. Safety caring and individual respect (CCSL) positively January 2015. The interval was chosen to counteract recall bias but
and directly predicts two SSSL dimensions, i.e. participative still can generate registration of significant fluctuations in the
decision-making (H7a) and showing concern/interacting with measures (Tholén et al., 2013). Based on the typical longitudinal
workers (H7b). study designs (Tucker et al., 2010; Tholén et al., 2013),
this time interval is suitable for a longitudinal questionnaire survey,
As a transactional leadership dimension, safety controlling especially in the Chinese construction industry. Chinese construc-
and performance management have close relations with the tion projects have rapidly changing internal and external
“informing” dimension, which is also a transactional leadership environments, and longer time intervals would not guarantee that
style. Informing is essentially task-oriented and obviously all respondents participate in all the three rounds of measurement.
facilitated by monitoring and controlling task execution of the The research team monitored and provided necessary help for
upper-level management. A rigorous safety performance man- the respondents before, during and after the questionnaire surveys.
agement of contractors impels supervisors to promote workers' Respondents were firstly informed of the research aims, research
understanding of task requirements, operating procedures, safety procedures and ethical issues such as voluntary participation,
compliance regulations and reward and punishment provisions anonymity and confidentiality regarding individual feedbacks.
(Martínez-Córcoles et al., 2011). They were then asked to indicate the degree to which they agreed
or disagreed with the phenomena or practices described in the
H8. Safety controlling and performance management (CCSL) questionnaire. The longitudinal study yielded three rounds of data
positively and directly predicts informing (SSSL). regarding OPSL, CCSL, SSSL and project safety culture.

Project safety culture acts as the mediator in the relationship 3.1. Measures and instruments
between CCSL and SSSL. CCSL positively influences project
safety culture, which in turn predicts SSSL dimensions. It is the The questionnaires include OPSL, CCSL, SSSL and project
similar case with the relationship between OPSL and CCSL. safety culture. A 5-point Likert scale was adopted from 1
(completely disagree) to 5 (completely agree).
H9. Project safety culture mediates the relationship between
CCSL and SSSL. In other words, each CCSL dimension 3.1.1. Safety leadership
predicts project safety culture, which in turn predicts all SSSL The questionnaire items assessing OPSL and CCSL were
dimensions (H9a to H9d represent links between CCSL developed specially for the study based on the existing related
dimensions and project safety culture, and H9e to H9i represent constructs of the four dimensions, i.e. safety influence and
links between project safety culture and SSSL dimensions). role modeling, safety motivation and coaching, safety caring
and individual respect, safety controlling and performance
Hypotheses H1–H9 are the assumed relationships among management. Measurement items in the FRL model as well as
the research variables (dimensions of OPSL, CCSL, SSSL and safety leadership scales developed by other studies (as
project safety culture). They will be validated by empirical mentioned in the literature review section) were referenced,
studies as illustrated in the next section. adapted and refined in order to be applicable in this study. The
 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805 795

safety leadership practices identified by Wu et al. (2015) were and the respondents are contractor construction managers.
also used for reference. OPSL was rated by contractor construction managers using the
When determining questionnaire items, it is crucial to OPSL questionnaire. Contractor construction managers also
ensure the validity of their content, which is an important measure rated their own safety leadership using the CCSL questionnaire.
of an instrument's accuracy. Content validity should be confirmed The second set is the relationships between CCSL and SSSL
carefully when questionnaire items were borrowed from one (Fig. 2), and the respondents are supervisors. CCSL was rated
industry to other industries. Its primary concern is “the degree to by supervisors using the CCSL questionnaire. Supervisors also
which a measure covers the range of meanings included within rated their own safety leadership using the SSSL questionnaire.
the concept” (Babbie, 1990). The content validity of question- To ensure consistency of the longitudinal study, valid
naires utilized in this study was confirmed through reviews of respondents must participate in all three rounds of measurement.
relevant literatures and a pilot investigation, i.e. structured Respondents who participated in one or two rounds were
interviews with practitioners including 10 contractor construction eliminated. This resulted in 221 valid respondents for the first
managers and 15 supervisors from subcontractors. Potential set (OPSL–CCSL) and 158 valid respondents for the second set
questionnaire survey respondents were asked whether they (CCSL–SSSL), and the average valid feedback rates were 71.5%
understood every item clearly and accurately. The items with and 60.8% for the two sets. Details of the distribution of the
vague or ambiguous expressions or did not conform to sample are presented in Table 1. Valid data were obtained from
construction management terminology were modified based on valid respondents' questionnaires which were completely filled
the respondents' comments. and not duplicated from others. Those respondents were
As a result, three sets of questionnaires were generated, consistent throughout three measurement rounds.
i.e. the OPSL questionnaires evaluated by contractor construction The ages of the valid respondents for the OPSL–CCSL
managers, the CCSL questionnaires evaluated by contractor relationship was from 30 to 65 years, with an average of
construction managers (i.e. self-evaluation), and the CCSL 42.49 years. Their working experiences were from 7.5 to
questionnaires evaluated by supervisors from subcontractors. 26 years, with an average of 17.10 years. All of them were at
Though sharing the common factor structure of safety leadership, least senior high school graduates and 92.8% were university
they have notable differences in the wording. Wording of each graduates. The ages of the valid respondents for the CCSL–
questionnaire type conforms to the reading and speaking norms SSSL relationship were from 25 to 57 years, with an average of
of the corresponding evaluators (respondents). See Appendix I 38.99 years. Their working experiences were from 5.5 to
for the details of the questionnaire. 23 years, with an average of 13.29 years. A total of 88.6% of
The empowering safety leadership questionnaire of Martínez- the respondents were at least senior high school graduates and
Córcoles et al. (2011) was used to assess SSSL by supervisors 53.8% were university graduates.
themselves. This questionnaire was adapted from the “empowering
leadership questionnaire” which was developed by Arnold et al. 3.3. Statistical analysis
(2000).
Confirmatory factor analysis and structural equation modeling
3.1.2. Project safety culture (SEM) were applied to test the reliability and validity of this
Project safety culture was measured by its snapshot, i.e. project empirical study, so as to validate the hypotheses mentioned
safety climate. Fang and Wu (2013) proposed six dimensions of previously. Confirmatory factor analysis was used to confirm the
project safety climate, including management commitment, safety reliability and fitness of the factor structure of owner and
management system, communication, safety involvement, safety contractor safety leadership and project safety climate, including
training and supervisory and supportive environment. The project their measurement items. SEM was to provide support (or
safety climate questionnaire used by this study contains 26 items. otherwise) for the nine hypotheses. The software SPSS 16.0 and
These items were adapted or originated from Mohamed (2002); Analysis of Moment Structures (AMOS) 22.0 were used to
Fang et al. (2006) and Zhou et al. (2011), which were all typical process the empirical data and make statistical analysis. The
studies measuring safety climate of construction projects. Their significance level was set as 0.05.
content validity was confirmed by the same method utilized by the The most crucial issue in the statistical analysis was to
safety leadership questionnaires (see Section 3.1.1). Two versions substantially validate the causal relations between safety
of project safety climate questionnaire with the same factor leaderships based on the three rounds of data. As mentioned
structure were generated, one for contractor construction man- above, most previous studies collected cross-sectional data and
agers and the other for supervisors. Similarly, they have notable thus could merely obtain empirical links between variables.
differences in the wording. See Appendix I for details. Granger (1969) indicated that a causal effect could be tested by
relating an independent variable measured at R1 to a dependent
3.2. Participants variable measured at R2, while controlling the effect of the
dependent variable at R1. Thus, utilizing the three-round panel
The contractor construction managers and supervisors of the data, this study made a statistical analysis based on the model
seven participating projects were involved in the empirical depicted in Fig. 3. The model shows that safety leadership as
study, which includes two sets of relationships. The first set the independent variable predicts future project safety culture and
corresponds to relationships between OPSL and CCSL (Fig. 1), safety leadership as the dependent variable. Safety culture also
796 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805

Table 1
Summary of distribution and feedback of the questionnaire survey.
Tested hypotheses The first set (OPSL–CCSL) The second set (CCSL–SSSL)

Respondents Contractor construction managers Supervisors from subcontractors

Number of distributed Number of valid Valid feedback Number of distributed Number of valid Valid feedback
copies feedback rate copies feedback rate
Project A 35 23 65.7% 24 14 58.3%
Project B 34 16 47.1% 25 15 60.0%
Project C 29 23 79.3% 20 9 45.0%
Project D 33 27 81.8% 26 13 50.0%
Project E 24 20 83.3% 17 14 82.4%
Project F 31 25 80.6% 23 11 47.8%
Project G 35 24 68.6% 23 20 87.0%
Total 221 158 71.5% 158 96 60.8%

predicts future safety leadership as the dependent variable, and thus safety motivation and coaching, safety caring and individual
mediates the lagging effect of safety leadership as the independent respect and safety controlling and performance management)
variable on safety leadership as the dependent variable. were 0.958, 0.886, 0.870, 0.814 and 0.868, respectively. The
The links D12 and D23 would be firstly focused on as they α values of project safety climate and its six dimensions
represent the impacting path between different stakeholders' (management commitment, safety management system, commu-
safety leadership during R1–R2 and R2–R3. For validating the nication, safety involvement, safety training and supervisory and
mediating role of project safety culture, the links I12 and I23 supportive environment) were 0.945, 0.855, 0.840, 0.824, 0.801,
would be emphasized. 0.778 and 0.832, respectively. Since 0.6 is generally accepted as
the bottom line of the desired value of internal consistency, i.e. α
4. Results values lower than 0.6 indicate unreliable questionnaire designs
(Flynn et al., 1994), both the supervisory behavior questionnaire
4.1. Confirmatory factor analysis and the safety climate questionnaire are proven as reliable.
The overall goodness of fit (GOF) test (Hair et al., 1998) was
The reliability of the questionnaires in this study was first performed to evaluate the quality of the four-dimension structure
evaluated. In the reliability test, special attention should be paid of safety leadership and the six-dimension structure of safety
to internal consistency reliability, which reflects correlations climate. If the overall GOF test has favorable results, the factor
between questionnaire items belonging to one dimension (Flynn structure of the two constructs can be confirmed. Overall goodness
et al., 1994). When one questionnaire contains more than one of fit includes absolute goodness of fit, incremental goodness of fit
dimension, the internal consistency of each dimension should be and parsimonious goodness of fit. Table 2 shows the overall GOF
respectively tested. Cronbach's α, as a measure of internal statistics of confirmatory factor analyses performed.
consistency reliability, is used. The α values of safety leadership Table 2 shows that the three aspects and 16 statistics of
and its four dimensions (safety influence and role modeling, overall GOF all indicate very favorable fitness judgments for

Safety leadership Safety leadership

R1 Project safety culture
(independent variable) (dependent variable)

D12

I12

Safety leadership Safety leadership

R2 Project safety culture
(independent variable) (dependent variable)

D23
I23

Safety leadership Safety leadership

R3 Project safety culture
(independent variable) (dependent variable)

Fig. 3. Statistical analysis model. Note: In this simplified form, dimensions of different stakeholders' safety leadership and their links are hidden for clarity. Only the
general lagging effects of safety leadership (independent variable) on safety leadership (dependent variable) are depicted to show the analytical approach in SEM.
 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805 797

both safety leadership and project safety climate. It not only RMSEA = 0.038 b 0.05; PGFI =0.754 N 0.5; PCFI = 0.869 N
confirms the factorial structures of the two constructs, but also 0.05; χ2/DOF = 1.762 b 2.0. In Fig. 4 to Fig. 9, the asterisks
validates the measurement scales designed for them. accompanying the coefficients (β) indicate the significance level.
Three asterisks (***) means p b 0.001, two asterisks (**) means
4.2. Structural equation modeling and hypotheses testing p b 0.01 and one asterisk (*) means p b 0.05.
Hypotheses testing results are shown in Figs. 4 to 9 and are
The software AMOS 22.0 was employed to undertake SEM interpreted in the following sections.
analysis for hypotheses testing by means of Granger's (1969) H1 proposed that each dimension of OPSL positively and
method in handling longitudinal data to validate causality. The directly predicts the same dimension of CCSL, which is validated
analysis was focused on the links D12, D23, I12 and I23 shown in by this study during both R1–R2 and R2–R3 (Figs. 4 and 5). H2
Fig. 3, which contains quite a number of detailed impacting paths and H3 proposed that safety influence and role modeling (OPSL)
between dimensions. This resulted in six sets of model testing and safety motivation and coaching (OPSL) positively and
results shown from Fig. 4 to Fig. 9. Figs. 4, 5 and 6 depict direct directly predicts some dimensions of CCSL besides their
and indirect relationships between OPSL and CCSL from corresponding dimensions. No evidence was found in this study
measurement rounds R1 to R3, and the other three figures depict for the two hypotheses (see Figs. 4 and 5). Therefore, the
relationships between CCSL and SSSL from R1 to R3. In the impacting path from OPSL to CCSL are one-to-one and
figures, R1, R2 and R3 above the constructs indicate the rounds in biunique. There is a distinct and simple correspondence between
which these constructs were measured. The results of hypotheses safety leaderships of the two main stakeholders. Each leadership
testing were divided into six figures for expression clearness. The dimension of contractor construction managers is influenced by
solid lines with estimated standardized effect coefficients (β) its homogenous leadership dimension of owner project managers.
represent significant links which passed the hypotheses tests, H4 is related to the mediating role of project safety culture in
while dotted lines represent non-significant links which failed the the relationships between OPSL and CCSL. Fig. 6 shows that
hypotheses tests. After these non-significant links had been this hypothesis was generally validated, although no significant
eliminated, the models showed a fairly good fit. Fit test results impacts of safety motivation and coaching (OPSL) and safety
for relationships between OPSL and CCSL (Figs. 4 to 6) were: caring and individual respect (OPSL) on project safety culture
p b 0.05, RMR = 0.041 b 0.05; RMSEA = 0.052 b 0.08; were identified. Notably, project safety culture predicts all
PGFI = 0.681 N 0.5; PCFI = 0.808 N0.05; χ2/DOF = 1.808 b safety leadership facets of contractor construction managers,
2.0; fit test results for relationships between CCSL and and their effect coefficients (β) on CCSL are higher than those
SSSL (Fig. 7 to Fig. 9) were: p b 0.05, RMR = 0.027 b 0.05; of OPSL. This finding well confirms the idea raised above that

Table 2
Overall GOF statistics of confirmatory factor analyses performed.
Safety leadership (OPSL and CCSL) Project safety climate
Statistics Fitness criteria
Values Fitness judgment Values Fitness judgment
Absolute fit indices
χ2 p b 0.05 276.649 (p = 0.000 b 0.05) Yes 513.967 (p = 0.000 b 0.05) Yes
RMR b 0.05 0.015 Yes 0.028 Yes
Yes
RMSEA b 0.08 (b 0.05 is excellent, and b0.08 is good) 0.032 0.044 Yes (excellent)
(excellent)
GFI N 0.90 0.968 Yes 0.918 Yes
AGFI N 0.90 0.958 Yes 0.901 Yes

Incremental fit indices

NFI N 0.90 0.975 Yes 0.920 Yes
IFI N 0.90 0.988 Yes 0.962 Yes
RFI N 0.90 0.970 0.906
TLI (NNFI) N 0.90 0.985 Yes 0.955 Yes
CFI N 0.90 0.988 Yes 0.961 Yes

Parsimonious fit indices

PGFI N 0.50 0.842 Yes 0.852 Yes
PNFI N 0.50 0.821 Yes 0.784 Yes
PCFI N 0.50 0.832 Yes 0.818 Yes
CN N 200 550 Yes 269 Yes
χ2/DOF b 2.00 1.921 Yes 1.855 Yes
Values of default model should be lower than those 368.649 b 380.000; 611.360 b 650.000
AIC Yes Yes
of independent model and saturated model 368.649 b 10,965.315 611.360 b 6223.561
Values of default model should be lower than those 634.473 b 1477.969; 976.950 b 2300.234
CAIC Yes Yes
of independent model and saturated model 634.473 b 11,075.112 976.950 b 6350.502
798 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805

R1 R2 R1

Safety influence and H1a Safety influence and Contractor

role modeling (OPSL) β =0.35*** role modeling (CCSL) construction managers’
β =0.79*** safety leadership

H2a
β =0.76***

Safety motivation and H1b Safety motivation and

coaching (OPSL) β =0.45*** coaching (CCSL)

H2c
H2b β =0.64***
β =0.59***

Safety caring and H1c Safety caring and

individual respect individual respect
β =0.43***
(OPSL) (CCSL)
H3

Safety controlling and H1d Safety controlling and

performance performance
management (OPSL) β =0.42*** management (CCSL)

Fig. 4. Relationships between OPSL and CCSL (R1–R2).

impacts of OPSL through the mediation of safety culture can found between safety influence and role model (CCSL) and three
enhance contractor safety leadership in a more profound and SSSL dimensions, i.e. participative decision-making, informing
comprehensive manner. and showing concern/interacting with workers. H6 is about
The above results support that significant impacts of OPSL on positive and direct impacts of safety motivation and coaching
CCSL exist in construction projects, and project safety culture (CCSL) on three SSSL dimensions, and this study successfully
can play as the mediator, which is able to enhance CCSL more validated one impacting path, i.e. the impact of safety motivation
holistically. and coaching (CCSL) on coaching (SSSL). H7 proposed that
H5 proposed that safety influence and role modeling (CCSL) safety caring and individual respect (CCSL) positively and directly
positively and directly predicts all SSSL dimensions, which was predicts two SSSL dimensions. This study only validated its link
not wholly validated by this study because no significant links were with showing concern/interacting with workers (SSSL), but not

R1 R2 R3 R2

Owner project β =0.74*** H1a β =0.71***

Contractor
Safety influence and Safety influence and
managers’ Construction managers’
role modeling (OPSL) β =0.33*** role modeling (CCSL)
safety leadership safety leadership

H2a β =0.69***
β =0.63***

H1b
Safety motivation and Safety motivation and
coaching (OPSL) β =0.39*** coaching (CCSL)
H2c
β =0.60*** H2b β =0.63***

Safety caring and H1c Safety caring and

individual respect individual respect β =0.52***
β =0.40***
(OPSL) (CCSL)
H3

β =0.59***
Safety controlling and H1d Safety controlling and
performance performance
β =0.38***
management (OPSL) management (CCSL)

Fig. 5. Relationships between OPSL and CCSL (R2–R3).

 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805 799

R1 R3 R2

Safety influence and Safety influence and β =0.71*** Contractor

role modeling (OPSL) role modeling (CCSL) construction managers’
R1 safety leadership
H4a
β =0.28*
H4e β =0.50*** β =0.69***
Project safety culture

Safety motivation and Safety motivation and

H4b β =0.60***
coaching (OPSL) R2 coaching (CCSL)
H4f

Project safety culture β =0.46*** β =0.63***

H4c β =0.44***
Safety caring and Safety caring and
H4g
individual respect individual respect β =0.52***
(OPSL) (CCSL)
H4d
β =0.42***
H4h
β =0.47***

Safety controlling and Safety controlling and

performance performance
management (OPSL) management (CCSL)

Fig. 6. Mediating role of project safety culture in relationships between OPSL and CCSL.

with participative decision-making (SSSL) (see Figs. 7 and 8). dimension of SSSL was directly predicted by none of the four
Thus, H6 and H7 were partially accepted by the empirical study. leadership dimensions of contractor construction managers.
Note that all the four dimensions of CCSL have direct predicting H9 concerns the mediating role of project safety culture in
effects on SSSL, but the “participative decision-making” the relationships between CCSL and SSSL. Fig. 9 shows that

R1 R2 R1

β =0.73*** Subcontractor
Leading by example
supervisors’ safety
H5a (SSSL)
leadership
β =0.35***
Safety influence and
β =0.78***
role modeling (CCSL)
H5b

H5c
β =0.34*** Participative decision-
H6a making (SSSL)

Safety motivation and

β =0.53**
coaching (CCSL)
H6b

β =0.23* Coaching (SSSL)

H7a
H5d
Safety caring and
individual respect β =0.64***
(CCSL) H5e
H7b H6c
β =0.32***
Informing (SSSL)

H8
Safety controlling and
performance β =0.25** β =0.59***
management (CCSL)
showing concern/
interacting with
workers (SSSL)

Fig. 7. Relationships between CCSL and SSSL (R1–R2).

800 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805

R1 R2 R3 R2

Subcontractor
Leading by example β =0.69***
supervisors’ safety
(SSSL)
H5a leadership
Contractor β =0.31***
β =0.66*** Safety influence and
construction managers’ β =0.75***
role modeling (CCSL)
safety leadership H5b

H5c
β =0.30*** Participativ edecision-
β =0.57*** H6a making (SSSL)

Safety motivation and β =0.52**

coaching (CCSL)
H6b
β =0.35***
β =0.89*** β =0.18* Coaching (SSSL)

Safety caring and H7a

individual respect H5d β =0.66***
(CCSL) H7b
β =0.33*** H6c

β =0.44** Informing (SSSL)

H8
Safety controlling and
performance β =0.22* H5e β =0.44**
management (CCSL)
showing concern/
interacting with
workers (SSSL)

Fig. 8. Relationships between CCSL and SSSL (R2–R3).

this hypothesis was generally validated, although no significant predicts all leadership dimensions of supervisors, and their
impacts of safety motivation and coaching (CCSL) and safety effect coefficients (β) on SSSL are higher than those of CCSL.
controlling and performance management (CCSL) on project The above results support that significant impacts of CCSL
safety culture were identified. Likewise, project safety culture on SSSL exist in construction projects, and project safety

R1 R3 R2

Leading by example Subcontractor

β =0.69***
(SSSL) supervisors’ safety
leadership
Safety influence and H9e
β =0.42*** β =0.75***
role modeling (CCSL) R1

H9a Participative decision-

Project safety climate
β =0.36* making (SSSL)
H9f

Safety motivation and β =0.57***

β =0.50***
coaching (CCSL) H9b R2 β =0.52***

Project safety climate H9g

Coaching (SSSL)
β =0.49***
β =0.41**

Safety caring and H9c

H9h
individual respect β =0.66***
β =0.50***
(CCSL)
H9d
Informing (SSSL)

Safety controlling and

performance H9i β =0.38*** β =0.44***
management (CCSL)
showing concern/
interacting with
workers (SSSL)

Fig. 9. Mediating role of project safety culture in relationships between CCSL and SSSL.
 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805 801

culture can play as the mediator, which is able to enhance SSSL organizational culture in the relationship between leadership and
more holistically. subordinates' behavior/organizational performance (Wu et al.,
2011; Wilderom et al., 2012), and this study discovered that
5. Discussions cultural factors also play important roles in the relationship
between leaderships of different project stakeholders. Compared
5.1. Findings and implications with individual factors, cultural factors tend to have greater power
to spread and penetrate throughout the organization, especially
This study validated the relationship between safety leaderships construction projects, which are temporary organizations with
of the three main stakeholders in construction projects, i.e. owners, multiple stakeholders. Thus, safety culture development should
contractors and subcontractors, and interpreted the role of safety be a very important approach for project managers to improve
leadership in construction safety improvement in a holistic manner. safety performance. Most notably, for some leadership dimen-
The nine hypotheses of this study were generally validated as there sions like participative decision-making (SSSL), it is project
are significant relationships between OPSL and CCSL, as well safety culture, but not safety leadership of other stakeholders, that
as CCSL and SSSL, although some links were proven to be non- has significant effects on them.
significant. The results show that the three safety leaderships, Safety influence and role modeling in both OPSL and CCSL
though belong to three different organizations, have close has the widest range of influence on safety leadership of other
interactions with each other. Thus, safety leadership can go beyond stakeholders, although some relevant hypothesized links are
traditional organizational boundaries within construction projects. non-significant in this study. This finding aligns with the
Owner safety leadership has distinct and one-to-one direct findings of previous studies that it is the first and foremost
effects on contractor safety leadership; that is, each OPSL dimension of transformational leadership. Safety influence and
dimension influences its corresponding and homogenous dimen- role modeling also has significant effects on project safety
sion of CCSL. This implies that when impacting contractor culture, which in turn influences all behavioral dimensions in
managers' safety leadership, the owner should implement specific safety leaders of other stakeholders.
interventions. For example, if the contractor managers have low Safety motivation and coaching in OPSL and CCSL has no
performance in the caring dimension of leadership, the owner significant effects on project safety culture. Actually, this
should focus on the caring dimension when fulfilling its safety leadership dimension only influences its homogenous leadership
leadership, instead of depending on other dimensions (like strict dimensions of another stakeholder, i.e. safety motivation and
controlling and monitoring). There is no owner leadership coaching (CCSL) and coaching (SSSL). Previous research (Blair,
dimension (even safety influence and role modeling) that directly 2002) argued that this dimension mainly makes subordinates
impacts its heterogeneous dimensions in contractor safety “know-how” to improve work performance while role modeling
leadership. According to the leader–member exchange (LMX) and charisma make subordinates “want” to achieve high
theory (Ilies et al., 2007), leadership is the dyadic relationship performance. “Know-how” means leaders must motivate the
between a leader and a follower. The fulfillment of leadership right person to take the right actions. It should be noted that
should be based on leaders' perceptions of the exchange process “know-how” must be based on “want”. Therefore, purely
between them and their followers and build unique relationship cramming safety regulations and procedures into subordinates'
with each follower. At least, the owner should know clearly which brains may improve their coaching ability, but cannot influence
aspect of the contractor's behavior they want to change and fulfill other leadership dimensions and safety culture, because they
their leadership according to this contractor's needs and features. probably lack adequate faith and safety awareness. Likewise,
Likewise, the effects of contractor safety leadership on preaching safety visions blindly without considering subordi-
supervisor safety leadership are also targeted. For example, safety nates' receptivity and individualized demands is also ineffective.
motivation and coaching (CCSL) influences coaching (SSSL); In the relationship from OPSL to CCSL, safety caring and
safety caring and individual respect (CCSL) predicts showing individual respect have no significant effects on project safety
concern/interacting with workers (SSSL); safety controlling and culture, while in the relationship from CCSL to SSSL, its
performance management (CCSL) is associated with informing significant effects on project safety culture were validated. This
(SSSL), and both belong to transactional leadership. Although the finding implies that the caring dimension of safety leadership is
correspondence between CCSL and SSSL dimensions is not as more effective and thus demanded in the lower-level and
clear-cut as that between OPSL and CCSL, the dyadic relationship frontline project staff. Supervisors need more attention and
between contractor managers and supervisors still apply, which support from contractors in their well-being, self-esteem
means contractor managers should also be specific and targeted and organizational identification. Contractor managers should
when improving supervisors' safety leadership. pay attention to maintain effective communication channels with
The mediating role of project safety culture is significant in the supervisors and respond positively to their needs and demands.
relationships between different stakeholders' safety leadership. In the relationship from OPSL to CCSL, the effects of safety
Without project safety culture, the safety leadership of one controlling and performance management on project safety
stakeholder can only influence part of the safety leadership of culture was validated, but this study failed to validate the
another stakeholder. In contrast, mediated by project safety significant influence of safety controlling and performance
culture, it can influence all safety leadership dimensions of management on safety culture in the relationship from CCSL to
another stakeholder. Previous studies proposed the importance of SSSL. It implies that transactional contingent reinforcement
802 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805

takes effect when the owner develops project safety culture 5.2. Limitations and future research
within the contractor management, but may not be effective
when the contractor enhances the lower-level safety culture Two limitations of this study need to be mentioned and point to
pertaining to the frontline worksite environment. In reality, future research directions. First, this study was undertaken in
what really matter in this situation are charisma, role modeling Chinese high-speed railway construction sites. In order to confirm
and caring behavior of the contractor managers (see Fig. 9). the generalizability of the results, similar studies can be designed in
Notably, contractor safety leadership is on one hand predicted other types of construction projects and in other areas outside
by project safety culture in the OPSL–CCSL relationship, and on China. Second, the owner safety leadership was treated as an
the other hand predicts project safety culture in the CCSL–SSSL independent variable in this study, i.e. owner safety leadership
relationship. Therefore, the interaction between leadership and cannot be impacted by other constructs. It should be acknowledged
culture was also found by this study. Leadership and culture are that owner leaders can adapt and change in response to the orga-
related constructs in that they both describe social interactions nizational change and be affected by cultural factors. Thus, owner
within organizations; however, their effects emanate clearly from safety leadership may be predicted by safety culture to some extent.
different levels-of-analyses (Yammarino et al., 2005). Whereas Besides, this study focuses on one-way top-down influence, i.e.
leadership denotes the behavior of one or more individuals who from owner to contractor and then to subcontractors. There are
act in a formal or informal role, culture is a byproduct of group several questions that remain unanswered. For example, is owner
processes. While it is widely accepted that leaders influence safety leadership influenced by contractor safety leadership? Is
organizational culture, culture influences leadership and the contractor safety leadership influenced by supervisor safety
development of good leaders as well. Leaders who appear to leadership? It is suggested that future research look into these
best represent the existing or desired culture are often the first to be questions to depict the interaction between all leadership agents as
recruited within the organization (Jaskyte, 2010). Moreover, well as leadership and culture, and address their combined effects
culture shapes managerial perceptions of events that, in turn, on safety.
affect how leaders manage (Jaskyte, 2010). For example,
managers tend to use participative leadership styles when the
6. Conclusion
culture endorses employee empowerment. Contractor leaders are
an important link in the safety leadership impacting chain because
This study investigated the safety leadership relationships
they not only pass the effects of owner leadership on to
between different stakeholders of construction projects, and
supervisors and workers but also transmit the upper-level project
identified safety leadership dimensions which have significant
safety cultural attributes to the frontline operating groups.
impacts on other stakeholders. Research on leadership's effects
Moreover, leadership is proven as very important in safety culture
on safety is transformed by this study from focusing on a single
development and maturity. The interaction between different
stakeholder to involving multiple stakeholders. Major conclu-
stakeholders is just the process in which different safety leadership
sions and their implications include:
levels influence the formation and development of construction
project safety culture.
The results also show that safety influence and role • Significant relationships exist between different project stake-
modeling (charismatic leadership) has wider range of effects holders' safety leaderships, and project safety culture acts as a
on project safety culture and other stakeholders' safety very significant mediator. There are close interactions between
leadership in both the OPSL–CCSL and CCSL–SSSL safety leadership and safety culture. Safety leadership can go
relationship. Thus, one important practical implication of this beyond traditional organizational boundaries within construc-
study is that construction organizations should select managers tion projects.
who have charisma and can behave as role models for others. • Relationships among different project stakeholders' safety
Construction project management training should emphasize leaderships are distinct and targeted. All dimensions of owner
more on socialized charismatic behaviors such as being safety leadership and contractor safety leadership make specific
altruistic, articulating safety core values, expressing confidence contributions to safety leadership of other stakeholders. The
in followers, shouldering major responsibilities in safety dyadic relationship between different stakeholders' safety
incidents and building identification within the group. leaderships means project leaders should be very specific
However, despite this, it should always be kept in mind that and targeted when improving safety leadership of other
all dimensions of owner and contractor safety leaderships make stakeholders.
specific contributions to safety, as all the dimensions have effects • Safety influence and role modeling has the widest range of
on project safety culture or other stakeholders' safety leadership, influence on project safety culture and safety leadership
although there is a difference in their influencing ways and among all dimensions. Thus, project owners and contractors
scopes. Thus, project leaders cannot neglect any leadership should cultivate charisma and idealized influence in project
domains when pursuing safety excellence, and abandoning any managers and require them to behave as role models for
leadership domain will ultimately lead to leadership deficiency of others. Project management training should also emphasize
other stakeholders. That is one of the most significant implications more on socialized charismatic behaviors such as being
of the biunique and targeted relationships of different leaderships altruistic, articulating safety core values, expressing confi-
discovered by this study. dence in followers, and building identification with the group.
 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805 803

• Charisma, role modeling and caring behaviors of safety values to frontline sites to contribute to substantial safety
leaders are particularly effective in the frontline environment. improvement in the construction industry.
Contractor managers should pay special attention to maintain
effective communication channels with supervisors and
Conflict of interest
respond positively to their needs and demands.
• Safety motivation and coaching is not as widely effective as
The authors declare that there are no conflicts of interest.
hypothesized. It is ineffective to purely cram safety regulations
and procedures into subordinates' brains and preach safety
visions blindly without considering subordinates' receptivity Acknowledgments
and individualized demands.
The authors would like to thank the Shanghai Railway
As mentioned above, the top-down diminishing effect of Bureau of China as well as its contractors and subcontractors
management commitment to safety is a major barrier against for their participation in the longitudinal study. Special
safety improvement in construction. Theoretical and practical appreciation goes to Jie Li, Xing Yu, Jun. Zhang, Wei Yuan,
implications, such as the emphasis on social relational Zhenguang Chen, Jianfeng Fang for their dedicated support to
perspective of leadership, will instruct researchers and practi- this research. Acknowledgments are also addressed to the
tioners to reinforce safety commitment, implement effective National Natural Science Foundation of China (grant numbers:
managerial measures and publicize advanced policies and 71572088 and 51378296).

Appendix 1. Measurement items of questionnaires

Dimensions Questionnaire items

1
OPSL and CCSL questionniares
Safety influence and role modeling Make subordinates feel proud of their work and confident about safety improvement
Talk about their safety values to subordinates frequently
Never sacrifice safety to meet other needs and requirements
Show a good model to obey safety rules and regulations
Always take major responsibilities when safety problems arise
Safety motivation and coaching Talk about safety visions clearly and enthusiastically
Seek different viewpoints and perspectives on safety to avoid arbitrary decisions
Suggest innovative ways and procedures for safety management
Facilitating safety coaching and learning all through the project
Encourage subordinates' participation in safety decision-making
Safety caring and individual respect Actively care about subordinates' everyday life and try to satisfy their need for safety and well-being
Are confident of subordinates' safety competence and provide sufficient resources for them
Pay special attention to onsite safety
Are impartial and maintain harmony between different departments when dealing with safety business
Safety controlling and performance Order subordinates to accomplish safety goals and work with them for the goals firmly
management Establish the safety responsibility system of the whole project personnel and review its implementation regularly
Amend and update safety regulations regularly and timely
Deal with near-misses and safety unconformities proactively and thoroughly
Reward and punish moderately and legitimately to consolidate safety controlling
Project safety climate questionnaire
Management commitment The upper management encourages safety improvement suggestions and provide corresponding incentives
The upper management actively and decisively facilitate the implementation of safety rules and procedures,
and acts quickly to correct safety problems
The upper management constantly and decisively require and motivate all onsite staff to work safely
The upper management seek advice of the onsite staff when safety procedures and rules are developed or reviewed
The project prepares sufficiently for safety emergencies
Safety management system The safety rules and procedures are able to prevent or mitigate all kinds of accidents
The safety rules and procedures are reviewed and updated timely and regularly
The safety rules and procedures have high feasibility and are easy to follow
The safety rules and procedures are strictly adhered to
The safety rules and procedures can guarantee our safety absolutely
Communication The upper management clearly communicates safety issues to all levels within the project
The upper management listens to and acts upon feedbacks from the onsite staff
Everyone can make comments and present opinions on safety issues based on their perception of safety policies
Safety involvement Everyone aims to achieve high levels of safety performance
Everyone actively reports safety accidents and potentially hazardous situations
Everyone is willing to participate in safety planning if being asked
Everyone contributes to job safety analysis
(continued on next page)
804 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805

Appendix
(continued)
I (continued)
Dimensions Questionnaire items
Safety training I fully understand current and relevant safety legislation
I am capable of identifying potential hazardous situations
I am capable of using relevant protective equipment and tools
I receive adequate training to perform my job safely and coach others
Supervisory and supportive My colleagues and I can help and support each other on safety issues
environment I am able to effectively communicate with others on safety issues
I regard my supervisor as a role model in behavioral safety and safety management
People can always get the equipment needed to work to safety rules and procedures
Safety inspection in the project is very helpful to improve workers' safety
Note: 1. Subjects of the questionnaire items are leaders whom the questionnaires evaluate.

References Griffin, M.A., Hu, X., 2013. How leaders differentially motivate safety
compliance and safety participation: the role of monitoring, inspiring, and
Abudayyeh, O., Fredericks, T.K., Butt, S.E., Shaar, A., 2006. An investigation learning. Saf. Sci. 60, 196–202.
of management's commitment to construction safety. Int. J. Proj. Manag. Hair, J.F., Anderson, R.E., Tatham, R.L., Black, W.C., 1998. Multivariate Data
24, 167–174. Analysis. fifth ed. Prentice Hall, Upper Saddle River, NJ, USA.
Arnold, J.A., Arad, S., Rhoades, J.A., Drasgow, F., 2000. The empowering Hoffmeister, K., Gibbons, A.M., Johnson, S.K., Cigularov, K.P., Chen, P.Y.,
leadership questionnaire: the construction and validation of a new scale for Rosecrance, J.C., 2014. The differential effects of transformational
measuring leader behaviors. J. Organ. Behav. 21 (3), 249–269. leadership dimensions on employee safety. Saf. Sci. 62, 68–78.
Arquillos, A.L., Romero, J.C.R., Gibb, A., 2012. Analysis of construction Ilies, R., Nahrgang, J.D., Morgeson, F.P., 2007. Leader–member exchange
accidents in Spain, 2003–2008. J. Saf. Res. 43 (5), 381–388. and citizenship behaviors: a meta-analysis. J. Appl. Psychol. 92 (1),
Avolio, B.J., Bass, B.M., Jung, D.I., 1999. Re-examining the components of 269–277.
transformational and transactional leadership using the multifactor leader- Jaskyte, K., 2010. An exploratory examination of correlates of organizational
ship. J. Occup. Organ. Psychol. 72 (4), 441–462. culture. Adm. Soc. Work. 34, 423–441.
Babbie, E., 1990. Survey Research Methods, Wadsworth, Belmont, CA. http:// Keegan, A.E., Den Hartog, D.N., 2004. Transformational leadership in a
www.cengage.com/wadsworth. project-based environment: a comparative study of the leadership styles
Barling, J., Loughlin, C., Kelloway, E.K., 2002. Development and test of a of project managers and line managers. Int. J. Proj. Manag. 22 (8),
model linking safety-specific transformational leadership and occupational 609–617.
safety. J. Appl. Psychol. 87 (3), 488–496. Kimmel, M.J., 1981. Senior Leadership: An Annotated Bibliography of the
Bass, B.M., 1999. Two decades of research and development in transforma- Military and Nonmilitary Literature. US Army Research Institute for the
tional leadership. Eur. J. Work Organ. Psychol. 8 (1), 9–32. Behavioral & Social Science, Alexandria, VA, USA.
Blair, M., 2002. Effective school leadership: the multi-ethnic context. Br. J. Sociol. Lu, C.S., Yang, C.S., 2010. Safety leadership and safety behavior in container
Educ. 23 (2), 179–191. terminal operations. Saf. Sci. 48 (2), 123–134.
Bobbio, A., Bellan, M., Manganelli, A.M., 2012. Empowering leadership, Martin, H., Lewis, T.M., 2014. Pinpointing safety leadership factors for safe
perceived organizational support, trust, and job burnout for nurses: a study construction sites in Trinidad and Tobago. J. Constr. Eng. Manag. 140 (2),
in an Italian general hospital. Health Care Manag. Rev. 37 (1), 77–87. 04013046.
Chan, A.T., Chan, E.H., 2005. Impact of perceived leadership styles on work Martínez-Córcoles, M., Gracia, F., Tomás, I., Peiró, J.M., 2011. Leadership
outcomes: case of building professionals. J. Constr. Eng. Manag. 131 (4), and employees' perceived safety behaviours in a nuclear power plant: a
413–422. structural equation model. Saf. Sci. 49 (8), 1118–1129.
Choudhry, R.M., Fang, D., Mohamed, S., 2007. The nature of safety climate: a Martínez-Córcoles, M., Schöbel, M., Gracia, F.J., Tomás, I., Peiró, J.M., 2012.
survey of the state-of-the-art. Saf. Sci. 45 (10), 993–1012. Linking empowering leadership to safety participation in nuclear power
Chun, J.U., Yammarino, F.J., Dionne, S.D., Sosik, J.J., Moon, H.K., 2009. plants: a structural equation model. J. Saf. Res. 43 (3), 215–221.
Leadership across hierarchical levels: multiple levels of management and Martínez-Córcoles, M., Gracia, F.J., Tomás, I., Peiró, J.M., Schöbel, M., 2013.
multiple levels of analysis. Leadersh. Q. 20 (5), 689–707. Empowering team leadership and safety performance in nuclear power
Cooper, D., 1998. Improving Safety Culture: A Practical Guide. Wiley, plants: a multilevel approach. Saf. Sci. 51 (1), 293–301.
Chichester, UK. Mohamed, S., 2002. Safety climate in construction site environments. J. Constr.
Cooper, M.D., 2000. Towards a model of safety climate. Saf. Sci. 36 (2), 111–136. Eng. Manag. 375-384.
DeChurch, L.A., Hiller, N.J., Murase, T., Doty, D., Salas, E., 2010. Leadership Neal, A., Griffin, M.A., 2006. A study of the lagged relationships among safety
across levels: levels of leaders and their levels of impact. Leadersh. Q. 21 climate, safety motivation, safety behavior, and accidents at the individual
(6), 1069–1085. and group levels. J. Appl. Psychol. 91 (4), 946–953.
Fang, D., Wu, H., 2013. Development of a safety climate interaction (SCI) O'Dea, A., Flin, R., 2001. Site managers and safety leadership in the offshore
model for construction projects. Saf. Sci. 57, 138–149. oil and gas industry. Saf. Sci. 37 (1), 39–57.
Fang, D., Chen, Y., Wong, L., 2006. Safety climate in construction industry: a Ofori, G., Toor, S.R., 2012. Leadership and construction industry development
case study in Hong Kong. J. Constr. Eng. Manag. 132 (6), 573–584. in developing countries. J. Constr. Dev. Ctries. 17, 1–21.
Fang, D., Wu, C., Wu, H., 2015. Impact of the supervisor on worker safety Pater, R., 2001. Leadership skills for the 21st century. Proceedings of 2001
behavior in construction projects. J. Manag. Eng. 31 (6), 04015001. ASSE Professional Development Conference and Exposition (CD-ROM).
Flin, R., Yule, S., 2004. Leadership for safety: industrial experience. Qual. Saf. Anaheim, CA, USA, Session 631.
Health Care 13 (Suppl. 2), ii45–ii51. Reid, H., Flin, R., Mearns, K., 2008. Influence from the Top: Senior Managers
Flynn, B.B., Schroeder, R.G., Sakakibara, S., 1994. A framework for quality and Safety Leadership. 2008 SPE International Conference on Health,
management research and an associated measurement instrument. J. Oper. Safety, and Environment in Oil and Gas Exploration and Production. SPE,
Manage. 11 (4), 339–366. Huston, USA, pp. 1–5.
Granger, C.J.W., 1969. Investigating causal relations by econometric models Construction Users Roundtable, 2012. Construction Owners' Safety Blueprint
and cross-spectral methods. Econometrica 37, 424–438. (Cincinnati, Ohio, USA).
 C. Wu et al. / International Journal of Project Management 34 (2016) 789–805 805

Chinese State Administration of Work Safety, 2013. Significant improvement Wooldridge, J.M., 2010. Econometric Analysis of Cross Section and Panel
was achieved for national work safety in 2012. http://www.chinasafety.gov. Data. MIT Press, MA, USA.
cn/newpage/kzzb/kzzb.htm. Wu, T.C., 2005. The validity and reliability of safety leadership scale in
Smallman, C., John, G., 2001. British directors' perspectives on the impact of universities of Taiwan. Int. J. Technol. Eng. Educ. 2 (1), 27–42.
health and safety on corporate performance. Saf. Sci. 38 (3), 227–239. Wu, T.C., Chen, C.H., Li, C.C., 2008. A correlation among safety leadership,
Sunindijo, R.Y., Zou, P.X.W., 2012. How project manager's skills may influence safety climate and safety performance. J. Loss Prev. Process Ind. 21 (3),
the development of safety climate in construction projects. Int. J. Project 307–318.
Organ. Manage. 4 (3), 286–301. Wu, T.C., Lin, C.H., Shiau, S.Y., 2010. Predicting safety climate: the roles of
Suraji, A., Duff, A.R., Peckitt, S.J., 2001. Development of causal model of employer, operations manager and safety professional. J. Saf. Res. 41 (5),
construction accident causation. J. Constr. Eng. Manag. 127 (4), 337–344. 423–431.
Tam, C.M., Zeng, S.X., Deng, Z.M., 2004. Identifying elements of poor Wu, T.C., Chang, S.H., Shu, C.M., Chen, C.T., Wang, C.P., 2011. Safety
construction safety management in China. Saf. Sci. 42, 569–586. leadership and safety performance in petrochemical industries: the
Tholén, S.L., Pousette, A., Törner, M., 2013. Causal relations between mediating role of safety climate. J. Loss Prev. Process Ind. 24 (6), 716–721.
psychosocial conditions, safety climate and safety behavior: a multi-level Wu, C., Fang, D., Li, N., 2015. Roles of owners' leadership in construction
investigation. Saf. Sci. 55, 62–69. safety: the case of high-speed railway construction projects in China. Int.
Tucker, S., Turner, N., Barling, J., McEvoy, M., 2010. Transformational J. Proj. Manag. 107, 185–194.
leadership and childrens' aggression in team settings: a short-term Yammarino, F.J., Dionne, S.D., Chun, J.U., Dansereau, F., 2005. Leadership and
longitudinal study. Leadersh. Q. 21 (3), 389–399. levels of analysis: a state-of-the-science review. Leadersh. Q. 16, 879–919.
Tyssen, A.K., Wald, A., Spieth, P., 2014. The challenge of transactional and Zhou, Q., Fang, D., Mohamed, S., 2011. Safety climate improvement: case
transformational leadership in projects. Int. J. Proj. Manag. 32 (3), 365–375. study in a Chinese construction company. J. Constr. Eng. Manag. 137 (1), 86–95.
Wilderom, C.P., van den Berg, P.T., Wiersma, U.J., 2012. A longitudinal study of Zohar, D., 2002. The effects of leadership dimensions, safety climate, and
the effects of charismatic leadership and organizational culture on objective assigned priorities on minor injuries in work groups. J. Organ. Behav. 23
and perceived corporate performance. Leadersh. Q. 23 (5), 835–848. (1), 75–92.