1
PROJECT MANAGER LEADERSHIP AND DIGITAL TRANSFORMATION
SUCCESS: INSIGHTS FROM CONTINGENCY THEORY
Section 1: Foundation of the Study
Digital transformation (DT) involves leveraging modern digital technologies such
as social media connectivity, the Internet of Things (IoT), artificial intelligence (AI),
cloud computing, machine learning, and intelligent manufacturing to transform
organizations for improved business performance, increased sales, increased profits, and
boosted customer satisfaction and retention (Doukidis et al., 2020; Kraus et al., 2021a;
C.-H. Lee et al., 2021). Through the successful implementation of DT projects,
organizations can achieve novel product and service offerings, advance the structure of
their supply chains, and vastly improve processing power through reduced costs and
improved product development, thereby reducing industry competition (Doukidis et al.,
2020; Kretschmer & Khashabi, 2020; C.-H. Lee et al., 2021; Nasiri et al., 2020; Verhoef
et al., 2021). In 2019, 40% of all technology investments by large industrial companies
(LICs) globally were in DT and growing at a compound annual rate of 17.5% (Appio et
al., 2021; Govindarajan & Immelt, 2019). According to Appio et al. (2021), the World
Economic Forum predicted that AI or other digital technology would drive the processes
and products of 90% of new enterprise applications by 2025, with only 21% of
companies having completed their DT processes, an estimated 67% of the $100 trillion
investment at stake from DT by 2025. Despite the significant investments, DT projects'
massive failure is causing an extensive social and economic crisis, mainly affecting LICs.
According to Brunner et al. (2023) and Müller et al. (2024), leadership behaviors (LB) of
project managers (PMs) are an essential driver of employee performance, task planning,
and productivity, leading to successful DT project implementation. With limited
literature
2
on the LB of DT PMs (Ahmad et al., 2022), this study investigated the relationship
between the LB of PMs of DT projects and successful DT project implementation in
LICs in the United States.
Background of the Problem
In the current digital age, the organizational survival of LICs depends mainly on
their DT capabilities. According to Ghosh et al. (2022), LICs operate in the industrial
environment where product differentiation is a competitive necessity, and digital
technologies such as software, AI, cloud computing, IoT, big data, and intelligent
manufacturing are major driving forces for growth and innovation. According to Cooney
et al. (2021), 93% of chief executive officers (CEOs) of industrial companies globally see
disruptive emerging technologies as driving competition in their industry and DT as a
critical change program that is required to operate and survive. DT is a significant
organizational change induced by digital technologies that have disrupted all industries
during the past decade (Cooney et al., 2021; Ghosh et al., 2022).
According to a recent trend, LICs are spending massive amounts of money on
transforming digitally for improved growth and innovation to enhance their business
models through reduced repair and delivery time, reduced cost, and enhanced quality
using digital tools. However, over 75% of the DT initiatives failed in LICs (Correani et
al., 2020; Datta & Nwankpa, 2021; Reeves et al., 2018), demonstrating that most large
companies from the industrial sector lack DT competency. Failed DT projects negatively
impact profitability, competitive advantage, and companies' sustainability (C.-H. Lee et
3
al., 2021). It is necessary to examine the reasons for the failure and identify the
contributing factors for successful DT project implementation in LICs.
In many LICs, the PMs' LB positively and significantly influenced the success of
DT project implementation (Ahmad et al., 2022; Brunner et al., 2023; Dubey et al., 2020;
Müller et al., 2024; Shao, 2019). PMs must understand the complexity of DT projects
and design their task structures effectively; they must also set clear project goals,
maintain good relationships with team members, motivate, coach, and control the team
effectively to accomplish project tasks and implement the DT projects successfully
(Brunner et al., 2023; Müller et al., 2024). There is a considerable gap in the research on
the impacts of LB of DT PMs on DT project implementation, resulting in over 75%
failure of DT projects in LICs in the United States.
Problem and Purpose
The specific business problem that triggered this study is that some PMs in LICs
do not know the relationship between PM's LB and DT project completion status.
Therefore, the purpose of this quantitative correlational research study was to investigate
the relationship between PM's LB and DT project completion status. The independent
variables were (a) PM's LB, (b) PM's member relationship, (c) PM's project task
structure, and (d) PM's position power. The dependent variable was DT project
completion status. The target population consisted of PMs of LICs in the United States,
focused on digitally transforming their businesses.
4
Population and Sampling
Population
The population for this doctoral research consisted of PMs who managed or were
managing DT projects during this study from large companies (# of employees ≥ 500 as
defined by the federal government) from the industrial sector in the United States and
who focused on transforming their businesses digitally. I used the third-party Centiment
Survey Panel (Centiment.co) to panel the population for this study. A large pool of PMs
who met the eligibility criteria based on the screening criteria provided in Appendix A
paneled through the Centiment.co company (hereafter Centiment) served as the
population for this study. This population was appropriate for answering the research
questions because this study's scope was to investigate the relationship between DT
project completion status and the PMs' LB measured at the PM level in three contingency
situations of the PM (PM's TS, PM's LMR, and PM's PP).
Unit of Analysis and Dependent Variable
The unit of analysis in a research study is the object of inquiry, which defines the
structure of research data and is a central consideration in any methodology (W. Li et al.,
2017). All persons conducting research studies must be clear about the unit of analysis
used in their research study. The unit of analysis used in this research was the PM. The
dependent variable of this study was whether the PM had completed a DT project or not
completed it within the allocated budget, within the scheduled time frame, and per
quality expectations by stakeholders but focused on transforming their businesses
digitally. I
5
have listed in Appendix C the screening questions used to collect data for the dependent
variable of this study.
Independent Variables
The independent variables of this study were (a) PM's LB, (b) PM's LMR, (c)
PM's TS, and (d) PM's PP. I collected primary data for the independent variables using
the following four scales developed and used by Fiedler (1967): (a) the Least Preferred
Coworker (LPC) Scale with 18 items, (b) the Leader–Member Relations (formerly
referred to as Group Atmosphere) Scale with eight questions, (c) the Task Structure
Rating Scale with 10 questions, and the (d) Leader's Position Power Rating Scale with
five questions (Fiedler & Chemers, 1984) to collect data for PMs LB, PM's LMR, PM's
TS, and PMs' PP, respectively. I administered the scales to the eligible participants
through the Centiment survey administration tool and collected the data.
Sampling
This research study involved a quantitative method and correlation design to
examine the relationship between the PM’s LB (primary independent variable) and DT
project completion status (dependent variable) at three contingency variable levels (PM’s
LMR, PM’s TS, and PM’s PP [independent variables]). The appropriate sampling
method for this study, therefore, was probabilistic. I chose the stratified random sampling
technique because there was variation in the targeted population, and the population size
was not fully known. The population size came from the paneling statistics provided by
Centiment, and the population size was only approximately known to me. The simple
6
random sampling (SRS) method is appropriate when the population is homogeneous, and
the research study involves all cases in the entire population (Bhardwaj, 2019).
The statistical testing technique used in this study was binary logistic regression
(BLR) because the dependent variable was dichotomous. There is no common acceptance
in determining sample size for BLR. A smaller than required sample size causes a loss of
information and misjudgments, and a larger than required sample wastes resources. I
conducted an a priori sample size analysis using the G*Power software (version 3.1.9.7)
with Z-tests and logistic regression options. G*Power is an effective statistical software
package that social scientists and business personnel use to conduct statistical power
analyses and sample size requirement determinations (Kang, 2021). A priori sample size
analysis conducted prior to the beginning of the study is ideal as a powerful analysis tool
for its ability to enable controlling Type I errors (the probability of rejecting a true null
hypothesis) and Type II errors (the probability of accepting a false null hypothesis;
Lakens, 2022).
G*Power covers statistical power analyses for many different statistical tests in
the families of F-test, t test, χ2-test, Z-test, and some exact tests (Kang, 2021). The
probability p1 = Pr(Y = 1|X = 1) = H0, the probability of occurrence of the negative (–)
effect is an input field in G*Power. The probability p2 = Pr(Y = 1| X = 1) = H1 is the
probability of occurrence of the positive (+) effect is another input field in G*Power.
Specifying the effect size is achievable in two ways: (a) directly by inputting the two
probabilities (p1 and p2) or (b) by calculating the odds ratio (OR) using the two
probabilities and inputting the OR in the analysis (Kang, 2021).
7
In this study, I tested whether a statistically significant effect existed; therefore, I
wanted to ensure that the sample size was large enough to prevent erroneous conclusions
about the desired effect size. Assuming a medium effect size (p1 = .24), a standard
significance alpha level of .05 (α = .05), and four predictor variables, the a priori power
analysis indicated a minimum sample size required of 143 to achieve a statistical power
of .80. A larger sample size lowers the likelihood of error in generalizing the findings to
a target population (Lakens, 2022). Therefore, this study's sample size of 214 PMs was
adequate and appropriate.
Nature of the Study
The chosen research method and the research design in this doctoral study to
address its research questions were quantitative and correlation. The approach was to
examine the relationship between PM’s LB (primary independent variable) and DT
project completion status. In this study, I analyzed numerical data using inferential
statistical tests to investigate the relationship between multiple independent variables and
a dependent variable and inferred the results from a larger population, for which the
quantitative research method was appropriate (Johnston et al., 2019).
Correlation research design enables prediction and explanation of the relationship
between two or more variables without varying, manipulating, or controlling any of the
variables. Unlike experimental research designs, correlational research designs analyze
the extent to which the variables are related and the nature of the relationship among two
or more variables without manipulating the variables (Seeram, 2019). I compared two
categories of LB of PMs (task-oriented [TO] vs. relationship-oriented [RO]) who
8
managed or were managing (during the study) DT projects on the two possible values of
the dependent variable (yes/no) in three variable contingency situations (PM's LMR,
PM's TS, and PMs' PP). The data collection for the dependent and independent variables
occurred without control, interference, or manipulation. I tested hypotheses about the
relationships between dependent and multiple independent variables without
manipulating any variables, which made the correlation the best design (Seeram, 2019)
for this study.
Research Questions
RQ1:
What is the relationship between PM’s leadership behaviors and DT
project completion status?
RQ2:
What homogenous clusters of PM’s leadership behaviors emerge based on
DT project completion status?
Hypotheses
Null Hypothesis H011: There is no significant relationship between PM’s
leadership behaviors and DT project completion status.
Alternative Hypothesis H111: There is a significant relationship between PM’s
leadership behaviors and DT project completion status.
Null Hypothesis H021: There are no homogenous clusters of PM’s leadership
behaviors that emerge based on DT project completion status.
Alternative Hypothesis H121: There are homogenous clusters of PM’s leadership
behaviors that emerge based on DT project completion status.
9
Theoretical Framework
The theory that grounded this study was the contingency theory of leadership
(CTL) introduced by Fiedler (1958). The key constructs underlying the CTL are (a) good
leader–member relations, (b) tasks with clear goals and procedures, and (c) the leader's
position power to reward and punish per members' performance. According to Fiedler
(1958, 1964, 1967), how the group receives the leader, the structure of tasks involved,
and whether the leader can control the group depends on the leader's LB. The theory has
evolved to address the adaptability of LB to manage external and internal business
environments. In situations where the group members are inadequately skilled or rely on
the structure of the task, such that tasks are complex or not well defined and structured, a
leader must rely more on task-oriented LB (TOLB) to accomplish goals (Fiedler, 1967).
On the contrary, when the members are skilled and independent, and the nature of the
task is less complicated and well defined, human relations are vital, and the leader must
rely on relationship-oriented LB (ROLB) to accomplish goals (Fiedler, 1967).
As applied to this doctoral study, the CTL was appropriate. CTL has directly
moderated the relationship between the PMs’ LB (TO vs. RO) and DT project
completion status. Per several past studies (Henkel et al., 2019; Shao, 2019; Warner &
Wager, 2019), the LB of leaders fell into a dichotomy of two metacategories: (a) TO and
(b) RO; this dichotomy of LB of PMs influenced the DT projects’ successful
implementation.
Operational Definitions
Digital transformation: The process by which organizations adapt themselves to
modern digital technologies such as the Internet of Things, social media connectivity,
10
cloud computing, artificial intelligence, big data, and predictive analytics, among others
(Albukhitan, 2020; Doukidis et al., 2020; Kretschmer & Khashabi, 2020).
Failed project: An abandoned, canceled, or unsuccessful project that the company
did not complete or completed without adhering to its requirements (Correani et al.,
2020).
Leadership behavior: The leader’s ability to induce subordinates to work with
enthusiasm and confidence is considered a central axis of the relationship between
superiors and subordinates and one of the aspects of mutual influence between
individuals and the group (Fiedler, 1967; Henkel et al., 2019).
Leader–member relationship: Fiedler (1967) defined the leader–member
relationship as the interpersonal relationship the leader establishes with his team.
Leader’s position power: The potential power that the organization provides for
the leader’s use to influence the leader’s group members to get them to comply with and
accept the leader’s direction and leadership (Fiedler, 1967; Kovach, 2020).
Project completion: When the project manager has delivered the project within
the agreed scope of the project, within the agreed cost, schedule, and quality, ensuring
the fulfillment of all acceptance criteria, the satisfaction of stakeholders, and meeting all
business objectives (L. H. Nguyen, 2021; Zid et al., 2020).
Project management: The planning, organization, monitoring, and control of all
aspects of the project, with the motivation of all people, included to achieve project goals
safely within the agreed schedule, budget, and performance criteria (Project Management
Institute, 2023a; Venczel et al., 2021).
11
Project manager: The professional who plans, organizes, and executes projects
within restraints such as allocated budgets and schedules, defines project goals, leads
entire project teams, liaises with stakeholders, and ensures the successful closure of the
project (Project Management Institute, 2023b).
Successful project: A project is successful if it meets cost objectives, completes
within schedule, and meets expected quality (L. H. Nguyen, 2021).
Task-structure: The dimensions and characteristics that classify and describe the
task or project that can be modified to enhance the dynamics of the task group or project
team, including the amount of freedom and discretion that a team member has in
performing assigned tasks, such as scheduling work, determining work methods, and the
extent to which an employee requires materials, information, and expertise from
colleagues to accomplish the task (Fiedler, 1967).
12
Assumptions, Limitations, and Delimitations
Assumptions
An assumption in research is an unwarrantable claim or supposition accepted as
accurate without proof about any research component, such as the phenomenon studied,
resources, cost, schedule, technology, or location (Nkwake, 2020). One of the central
assumptions of this study was that PMs could successfully complete every DT project if
they managed and controlled it correctly. Another critical assumption that drove this
study was that the success of every DT project is directly linked to the correct
application of contingency leadership behaviors (TO or RO) by project managers (PMs)
to manage projects and control contingencies. Other assumptions included participants
answered survey questions honestly and factually. A further assumption was that a DT
project would be successful and considered completed if it were completed within an
agreed- upon schedule and met the objectives of cost and quality criteria assigned during
initiation. For this study, I assumed the research findings would improve business
practices and people's lives in the United States.
Limitations
Limitations in research represent potential weaknesses, constraints, and
unanticipated challenges associated with any research component that may limit the
ability to generalize from the research’s findings, describe applications to practice, and
the usability of the findings (Ross & Zaidi, 2019). I used existing standard survey
instruments to collect primary data for the independent variables. This study had the
following survey instrument limitations: (a) responses to the survey were subjected to
the
13
limitation of predefined response categories, thereby limiting the range of responses, and
(b) respondents were limited to the text direction in the survey about how to complete it.
The information relevant to this study that participants were unwilling to provide further
limited the study. The assumption of linearity between the logit of the dependent and the
independent variables (van Smeden et al., 2019) and the subsequent requirement of linear
boundary constructs did limit the BLR procedure.
Delimitations
Delimitations in research reflect the choices the people conducting the research
make regarding what they aim to achieve by conducting the research and what they will
exclude studying (Akanle et al., 2020; Mengist et al., 2020). In this quantitative
correlational study, I aimed to investigate the relationship between PM’s LB and DT
project completion status. The study’s scope was limited to large (# employees ≥ 500)
industrial companies. PMs from small and medium-sized companies (# employees <
500) were excluded from this study. The intended population included the PMs from
U.S.- based LICs. PMs in LICs based in other countries were not considered and were
out of scope. This study investigated the relationship between the LB of PMs and the
completion status of DT projects managed in the past or managed during the data
collection for this study; thus, DT projects not managed by PMs were outside the scope.
Significance of the Study
This study is significant in that the results may provide new insights regarding the
identification and introduction of predictive models that can enable managers to make
predictions about future outcomes using historical data that enhance the managers’
14
capability and efficiency of making better decisions to complete DT projects and
competitively deliver advanced digital products and services at large U.S.-based
industrial companies. The results from the study may contribute to positive social
change by promoting and facilitating the systematization of key digital processes within
the practice of business. By adopting strategies related to DT, such as AI, big data
analytics, machine learning, IoT, smart manufacturing, and more online customer
services, LICs may lower costs and create tangible improvements regarding the
availability of digital services within underserved communities, improving their living
standards. They may enhance employment opportunities, increase spending, and boost
the economy of the local communities and the U.S.
A Review of the Professional and Academic Literature
A doctoral study becomes credible when prior studies' findings are reviewed for
conceptual, methodological, and thematic development, critically analyzed, and
synthesized logically, and when gaps are identified (Paul & Criado, 2020). I conducted
this quantitative correlational study to investigate the relationship between two categories
of LB (TO vs. RO) of PMs who managed or were still managing (during data collection
for this study) DT projects on DT project completion status in three contingency
situations (PM’s PP, PM's LMR, PM's TS). The following two null hypotheses guided
this study: (a) there is no significant relationship between PM's LB and DT project
completion status, and (b) no homogenous clusters of PM's LB emerge based on DT
project completion status. In this literature review, I discuss the theoretical framework
for this study, the CTL introduced by Fiedler (1958), who stated that leaders acquire
15
leadership style through their life experiences and that their leadership styles are
impossible to change and must fit the situation.
Information contained in some leadership theories from literature provide
contrasting evidence to Fiedler’s (1958) CTL, including (a) situational leadership theory
(SLT), introduced by Hersey and Blanchard (1970, as cited in Benmira & Agboola,
2021), who stated that leaders must compete with and adapt to the situation and
transform their leadership styles to match the maturity (readiness) of their subordinates;
(b) the behavioral theory of leadership (BTL) introduced by Blake and Moutan (1945, as
cited in Benmira & Agboola, 2021), who believed that great leaders are not born and
people can become leaders through proper training and observation and focused on the
actions of leaders, and not their mental qualities or internal traits; and (c) the
transformational leadership theory (TRFLT) founded by Burns (1978), who stated that
leaders inspire followers using the strength of their vision, personality, and charismatic
behavior, which alter the followers’ preconceptions and perspectives and motivate them
to work toward similar organizational objectives to improve organizational efficiency.
In contrast, several theories support CTL and help in examining its relevance and
application, including (a) the trait theory of leadership (TTL), introduced by Allport
(1936, as cited in Jayawickreme et al., 2019), who stated that successful leadership
characteristics could be either inherited or acquired through training and practice and that
successful leaders with the right combination of characteristics for a situation must be
identified; (b) the great man theory (GMT) introduced by Carlyle (1840, as cited in
Benmira & Agboola, 2021), who stated that leadership capacity is inherent that great
16
leaders are born, not made; and (c) the transactional leadership theory (TRALT),
developed by Burns (1978), who stated that leaders rely on authority such as reward and
punishment for followers’ efforts to motivate them to accomplish organizational goals. In
other words, transactional leaders influence followers through contingent rewards and
negative feedback or corrective coaching toward achieving established goals, completing
required tasks, avoiding unnecessary risks, and maintaining the current organizational
situation (Young et al., 2021).
In this academic literature review, I systematically review existing trends and
literature, summarize and synthesize information from the literature on the independent
and dependent variables, and discuss the primary independent variable of this study (LB
of DT PMs) and the three contingency independent variables, PMs LMR, PM's TS, and
PM's PP, through the theoretical framework lens of Fiedler's CTL (Fiedler, 1967; Henkel
et al., 2019). I also discuss the literature related to the dependent variable of this study,
the successful DT project completion status, critically analyzing the impacts of DT on
businesses, factors affecting successful DT project implementation, successful DT
project completion criteria, and significant reasons and impacts of DT project failures in
organizations. I further discuss the measurement of this study's independent and
dependent variables and describe how this study addressed a gap in the literature.
The scope for the literature search was around DT at large industrial companies
(LICs) in the private sector in the United States. I searched Walden University library
databases through (a) ScienceDirect, (b) ProQuest Central, (c) Emerald Management, (d)
EBSCOhost, (e) ERIC, (f) ABI/INFORM Complete, (g) Thoreau, (h) Sage Premier
17
Annual Reviews, and (i) Google Scholar. The key terms I used in the literature search for
this review included (a) digital transformation, (b) contingency theory, (c) leadership
behaviors, (d) digital leadership competencies, (e) project management, (f) project
completion criteria, (g) project success criteria, (h) project failure reasons, (i) large
industrial companies, and (j) skilled employees.
The literature review comprised 250 research articles, books, and web sources
published between 1958 and 2024, of which 222 (88.80%, Table 1) fell within the last 5
years (2019 and after). Twenty-eight articles (11.20%) had publication years between
1958 and 2018 (Table 1) and addressed the theoretical framework and data analysis
techniques. This review included 235 of the 250 sources (94.00%, Table 1) published in
high-impact, active, and peer-reviewed scholarly journals. I used Ulrich’s Web Global
Serials Directory to confirm that the references cited were peer reviewed and active.
Table 1
A Summary of Literature Review Sources
Type of source Current source
(2019–2024)
Older source
(before 2019) Total Percent
Peer-reviewed
source 215.00 20.00 235.00 94.00
Other source 7.00 8.00 15.00 6.00
Total 222.00 28.00 250.00 100.00
Percent 88.80 11.20 100.00
Contingency Theory of Leadership
Contingency theory (CT) is a class of behavioral theory that concerns the context
of leadership whose proponents claim that there is no best way to organize a
corporation, lead a company, or make decisions; the optimal course of action is
contingent upon the
18
internal and external situations (Fiedler, 1967; Shala et al., 2021). A group of researchers
from Ohio State University (OSU researchers) first developed CT in 1950 (Anderson &
Sun, 2017). According to the OSU researchers, effective leadership involves building
good interpersonal relationships and initiating a structure ensuring task completion and
goal attainment (Anderson & Sun, 2017). During the same time, another group of
researchers from the University of Michigan's Survey Research Center (UMSRC
researchers) investigated the relationship between group productivity and effective LB
and found similar structural behaviors identified by the OSU researchers (Anderson &
Sun, 2017). UMSRC researchers grouped the LB into two meta categories termed (a)
ROLB and (b) TOLB.
Fred Fiedler, considered the pioneer of contingency theories in 1958, extended
the research by UMSRC researchers and founded the contingency theory of leadership
(CTL, the framework of this study), emphasizing LB as taking control over situations
(Fiedler, 1967). Fiedler believed that leadership effectiveness significantly depended on
leaders' ability to control the situation and postulated the two primary behaviors (TOLB
and ROLB) of leaders in CTL. Fiedler (1967) developed the Least Preferred Coworker
(LPC) scale to determine the effectiveness of LB. This scale suggests that the situation is
highly favorable and fit when the job is clearly defined, the leader has the authority or
position power, and a healthy relationship exists between leader and followers (Fiedler,
1967; Rehman et al., 2020). Fiedler (1967) suggested that the LB one adopts is fixed and
challenging to change.
19
According to Fiedler (1967), in CTL, a leader's effectiveness depends on a
combination of two forces: (a) the leader's managerial style displayed as the LB and (b)
the favorableness of the situation. When the company aims to increase production and
where the work is technically complex, a leader must focus on completing the tasks
without necessarily considering employees' relationships or wishes. In contrast, if the
company is working to increase teamwork through collaboration for project success, a
leader must focus on RO management and getting the workers' consent. Thus, managers'
LB must fit the context of the situation for the projects to be completed (Fiedler, 1967;
Rehman et al., 2020). Once the LB is determined using the LPC questionnaire, a
determination of the contingency situation that describes the leader's situational control
is required, which, according to Fiedler (1964, 1967), includes the (a) leader's task
structure, (b) the leader's position power, and (c) the relationship the leader maintains
with their members.
Fiedler (1964), using the three contingency variables (leader's task structure,
leader's position power, and leader's member relationship), divided the leader's situation
into eight octants (I, II, III, IV, V, VI, VII, VIII) based on the favorability to the leader
(Table E1 [Appendix E]). Fiedler (1964) conducted 12 studies that yielded 63
relationship combinations between leaders' LB and group performance in several
industrial organizational situations and used the median Spearman's correlation for each
octant from the results of the studies to derive the standard CTL model. I have presented
Fiedler's (1964) results in Table E1 (Appendix E), representing the standard value of
Fiedler's CTL model. According to Fiedler (1964), Spearman's rho correlations between
20
the leaders' LPC and group performance measures were consistent across studies and the
relations within eight octants were highly nonrandom in distribution. Ayman et al.
(1995), Chemers and Skrzypek (1971, 1972), Fiedler and Chemers (1984), and Graen et
al. (1970) confirmed Fiedler's (1964) results in their studies. I have displayed the results
from Chemers and Skrzypek (1971, 1972) in Figure E3 (Appendix E).
Fiedler (1964), in CTL, further classified the eight octants into three categories of
favorableness to the leader: (a) favorable situations (octant I, octant II, and octant III), (b)
moderately favorable situations (octants IV, V, VI, and VII), and (c) unfavorable
situations (octant VIII). Later, Fiedler (1967) and several other investigators, including
Ayman et al. (1995) and Fiedler and Chemers (1984), classified octant VII as
unfavorable to the leader based on their study's results. Chemers and Skrzypek (1971,
1972) classified octant III as a moderately favorable situation for the leader (Figure E3 in
Appendix E).
According to Ayman et al. (1995), Fiedler (1964, 1967), and Fiedler and Chemers
(1984), TO leaders (with low LPC, < 73) perform best in favorable situations (octants I,
II, and III) and unfavorable situations (octant VII and VIII); they perform the least in
moderately favorable situations (octants IV, V, and VI), while RO leaders (with high
LPC, ≥ 73) perform best in moderately favorable situations and least in favorable and
unfavorable situations.
Constructs Underlying CTL
The vital constructs underlying the CTL are (a) good leader–member relations,
(b) tasks with clear goals and procedures, and (c) the leader's position power for rewards
and punishments per members' performance. How the group receives the leader, the
success
21
of tasks involved, and whether the leader can control the group depends on the leader's
LB and the situation. Per CTL, in situations where the group member is not skilled or
relies on the nature of the task such that tasks are not well defined, a leader must rely
more on a TOLB to accomplish goals (Fiedler, 1967). On the contrary, when the
members are skilled and independent, and the nature of the task is less complicated,
human relations are vital, and the leader must rely on the ROLB to accomplish goals
(Fiedler, 1967).
Fiedler (1967) defined favorableness as how the situation enables the leader to
exert influence over his group and introduced the Least Preferred Coworker Scale to
measure LB. The scale comprises 18 questions about how leaders handled their least
preferred coworkers. Also, when the leader and the group members have the requisite
physical resources, skills, and abilities, then the ability of the leader to motivate members
and to direct and coordinate their efforts depends on three major contingency factors: (a)
the leaders' position power, (b) the structure of the task, and (c) the interpersonal
relationship between leader and members (Fiedler, 1967). The CTL has evolved to
address the adaptability of LB to manage external and internal business environments for
successful project implementation (Farhan et al., 2024; Henkel et al., 2019; Nicolás-
Agustín et al., 2022; Shao, 2019; Warner & Wager, 2019).
Relevance of CTL to This Study
As applied to this doctoral study, CTL is appropriate because CTL has directly
moderated the relationship between the PMs' LB (TOLB vs. ROLB) and DT project
completion status. Henkel et al. (2019), in a quantitative descriptive research design
22
study, used the Fred Fiedler leadership behavioral style self-assessment survey tool
(FLBSSST) to measure 129 managers' LB (TOLB versus ROLB) related to successful
project completion and showed that the distributed combination of the managers' TOLB
and ROLB positively correlated to successful project completion. Popp and Hadwich
(2018), in a quantitative correlational research study, used Fiedler's CTL on 315
industrial service participants and found a significant positive correlation between ROLB
and employees' overall successful performance regardless of the three situations studied
(employee–customer relationship, task structure, and personal power of the employees).
Warner and Wager (2019), in a qualitative multiple case study design using CTL,
found cross-functional teams and leadership support as essential factors in building DT
capabilities. Brown et al. (2021), in their meta-analytic investigation, reported that the
TOLB and ROLB were critical regarding leader influence on virtual team collaboration.
When task interdependence was high and team size was large, leaders' ROLB was
necessary to enhance both the team and individual processes and outcomes; in contrast,
when the task complexity was high, leaders' TOLB was essential to strengthen both the
team and individual processes and projects (Brown et al., 2021). Research using CTL
theory indicated that no single task structure equally applied to DT project success in all
organizations. Still, organizational effectiveness depends on a fit or match between the
technology, people, environmental volatility, the organization's size, the organizational
structure's features, and its information system (Makhlouf & Allal-Chérif, 2019). Thus,
the CTL upholds the approach to the study of DT, where LB is contingent on factors such
23
as technology, culture, task complexity or organizational structure, and the external
environmental influence that impacts the design and implementation of DT projects.
Leader’s Position Power
French and Raven (1959) analyzed the complexities of leaders’ position power
(PP) and classified it into five categories: (a) referent, (b) expert, (c) legitimate, (d)
reward, and (e) coercive. Referent and expert power are considered informal because
they do not have a direct managerial span of control; they exist without any recognized
formal official authority of the leader of a company (Kovach, 2020). The other three
types of power (reward, coercive, and legitimate) are formal because these depend on the
leader’s formal position of authority in the organization (Kovach, 2020). According to
French and Raven (1959), these three formal leaders’ PP categories enable a person or
group in the dominant position to influence another person or group in a submissive role.
Fiedler (1967), in his CTL, stated that the leaders’ PP is highly related to French and
Raven’s (1959) concepts of legitimate power and reward power.
According to Kovach (2020), leaders’ PP allows the leader to influence and
modify the behaviors and attitudes of individuals and groups and is the primary source of
power for managers in achieving results or compliance from subordinates. Almazrouei et
al. (2020) and Gregory and Osmonbekov (2019) found that employees who received
rewards from their leaders developed a positive attitude toward their jobs and performed
effectively. According to Fiedler's (1967) CTL, leaders' relationship with their group
members depends to a significant extent on the power the leaders wield over their
members under their position. Leaders with high PP get their group members to comply
24
with and accept directions and leadership, making their job easier. In contrast, a leader
with low PP must first convince his group members that they follow the leader, which
hurts the leader–member relationship in the group. In CTL, Fiedler (1967) stated that the
leader's reward power (power based on the leader granting valuable rewards to followers
to carry out the leader's instructions) is a potent incentive to motivate followers to act.
Adequate reward power creates a good relationship between leader and follower and
influences how the followers perceive the leader. DT managers could enforce leaders’ PP
in the three formal French and Raven (1959) power categories (reward, coercive, and
legitimate) for projects' successful implementation.
The published literature has little evidence addressing the relationship between
the leaders’ PP and successful project implementation. The least understood is through
which processes some managers are acknowledged as good leaders by their subordinates
while other managers are not, even when both have the same authority to reward and
punish their staff, making it essential to study the LB of managers contingent on their
PP. A leader's PP is the potential to influence, while leadership is the leader's LB that is
conducive to exercising their influencing power. DT PMs' practice of the influence of
power can enhance employee performance leading to successful DT project
implementation in organizations because the leader's legitimate power of responsibility
can help guide the powerless group members to overcome any fear of change and
complexities of tasks.
Importance of Leaders’ PP in DT Projects. DT necessitates significant changes
in organizational structures, strategy, culture, and other properties to remain competitive,
25
which calls for strong leadership to manage the large-scale transformative projects to
align strategy with organizational culture and provide employees with the necessary
training, knowledge, support, and guidance to embrace the change effectively (Gilli et al.,
2023; Singh et al., 2020). The DT managers may need high PP to make the managers' job
easier and interact with their members in terms of the roles and mutual expectancies.
However, what leadership competencies are required to lead DT impactfully is unclear,
although DT is at the top of several organizations' management agendas aiming to
transform their organizations digitally (Firk et al., 2021; Gilli et al., 2023).
Leader–Member Relationship
Almazrouei et al. (2020) and Gregory and Osmonbekov (2019) showed that the
leader-member relationship (LMR) determines the quality of work the followers perform
in a workgroup. A high-quality LMR favors several benefits in projects, such as efficient
resource allocation, challenging task assignments, and professional mentoring and
guidance (Zhou et al., 2021). Good LMR attests to the positive relationship between
leader and follower, leading to followers' high-performance behavior and job satisfaction
and enhancing work performance, increasing operational efficiency and organizational
productivity (L. H. Nguyen, 2021; Zhou et al., 2021). Follow-up followers reciprocate the
leader's favorable treatment by engaging in discretionary behaviors to promote
organizational productivity. A high-quality relationship goes beyond job-related
contractual obligations and motivates followers to return with high productivity and in-
role performance. Good leaders use their power to establish a good relationship with their
subordinates, providing them with courage, rewards, and opportunities to motivate and
26
develop themselves to improve their performance at work to the leaders' expectations
(Fiedler, 1967; L. H. Nguyen, 2021). The above findings indicate that managers must
constantly care for and improve their relationship with their followers and maintain an
excellent relationship to enhance employees' job satisfaction and innovative capability
for successful DT project implementation and organizational performance.
According to Fiedler (1967), the interpersonal relationship is something that the
leader establishes with his team and depends on (a) leader's personality and (b) the
nature of the organization. The leader's effective relations with group members and the
acceptance and loyalty the leader can receive from his group members relate to the type
of person and how the leader behaves in critical situations during group activities.
Generally, the leader comes into the group with the required expert knowledge and the
organization's approval, and the followers are supposed to follow the leader. It will
require considerable artlessness on the leader's part to be rejected by his group (Fiedler,
1967). Some leaders who significantly overestimate their position power make over-
confident and over-ambitious judgments, decisions, and biased evaluations and destroy
the LMR. LMR is critical to enhancing positive outcomes for employees and the
organization, considering that positive working environments increase employee effort
and productivity (Fiedler, 1967; Kovach, 2020). According to Fiedler's (1967) CTL, a
leader can change the subordinates' perceptions of the leader through his behaviors; thus,
the most critical aspect of a good LMR is the leader (Fiedler, 1967).
27
Task-Structure
A task is an assignment the project group undertakes on behalf of the
organization (San Cristóbal et al., 2018). The organization has a stake in seeing the task
accomplished according to specifications, including time, cost, quality/scope, and
customer satisfaction, and the PM is responsible for achieving the task (San Cristóbal et
al., 2018). Fiedler (1967) in CTL classified the project task into two types: (a) structured
or (b) unstructured, and stated that the task structure (TS) significantly determines the
leader's influence on group members. When structured, tasks become clear and specific,
and employees can efficiently complete them in a particular order at an exact time. There
are never questions about a person's work assignment, and there is much less uncertainty
and more oversight because the management clearly defines goals, policies, and
procedures and expects all to follow them in a structured task system (Fiedler, 1967).
Employees easily understand tasks when structured, monitor their progress, and get
consistent feedback from management. Through consistent feedback, leaders influence
their members by efficiently applying the organizational rules and policies, reinforcing
their PP (Fiedler, 1967).
The unstructured tasks involve the day-to-day tasks that keep the organization
running and are much more flexible, allowing everyone in the team to learn on the job,
plan as they go, and complete tasks however they see fit without getting feedback from
management. Per CTL, when the project group engages in a highly unstructured task
system, the leader has a much more difficult job leading his group because the leader
cannot use his PP (Fiedler, 1967). To operationally measure task structure in CTL,
Fiedler
28
(1967) used the following four scales: (a) decision verifiability (the degree to which the
solution to the tasks is satisfactory to the authority); (b) goal clarity (the degree to which
the leader clearly states the requirements of the tasks for group members to understand);
(c) goal path multiplicity (the degree to which multiple procedures or methods can solve
the task); and (d) solution specificity (the degree to which there is more than one correct
solution).
Importance of TS in DT Projects. Projects involving DT, software integration,
and innovation continuously introduce changes and new product offerings (Guinan et al.,
2019). In this case, the tasks may be too complex to outline in a step-by-step sequence
making team members share knowledge, collaborate to figure out the best approach,
overcome obstacles, determine their progress, and use their experience and best practices
to help achieve the objectives. However, the PM’s competency, competent team, clear
project goals and objectives, adequate project planning, clear task structure, and usage of
PMT methodologies, tools, and techniques led to highly successful PMT and enabled DT
project success (Guinan et al., 2019; Singh et al., 2020). Project managers’ knowledge of
selecting the tools, methods, and procedures and how to use them to structure tasks are
essential for DT project success. To structure or not to structure and how much to
structure the DT tasks is an area that needs further research.
Supporting and Contrasting Theories
Because of the complexity and multidimensionality of the subject, leadership has
become more critical than ever in today’s fast-paced and increasingly globalized world.
Leadership continues to generate captivating and confusing debate, and many different
29
leadership theories exist in the literature. In this section, I discussed the following
contrasting theories of CTL: (a) SLT introduced by Hersey and Blanchard (1970, as cited
in Benmira & Agboola, 2021); (b) BTL introduced by Blake and Moutan (1945, as cited
in Benmira & Agboola, 2021); and (c) TRFLT developed by Burns (1978). I also
discussed in this section the following theories that support CTL: (a) GMT introduced by
Carlyle (1840, as cited in Benmira & Agboola, 2021); (b) TTL founded by Allport (1936,
as cited in Jayawickreme et al., 2019); and the (c) TRALT introduced by Burns (1978).
Supporting Theories
Great Man Theory (GMT). The Scottish-born Thomas Carlyle (1840, as cited
in Benmira & Agboola, 2021) introduced the GMT. Like CTL, GMT claims that
leadership traits of leaders are inherent and great leaders are born and not made. Both
GMT and CTL focus on the innate characteristics of leaders and insist on identifying the
personality traits and other effective leadership qualities in leaders. Fiedler, in CTL,
states that leadership styles of leaders are inherent and cannot be changed; two factors
determine a leader's success, (a) the leader's inherent leadership style and (b) the leader's
situational control (Fiedler, 1967). According to GMT, the ablest man or the leader is
truest-hearted and the most just and tells people to do the precise, wisest, and fittest, and
the subordinates loyally surrender to the great man's command and may find their
welfare in doing so. In CTL, Fiedler also suggests measuring a leader's leadership style
using measurement scales he developed, identifying the situation of leadership control,
and then determining if the leader's leadership style is suitable for the situation. In GMT,
Carlyle identified an eclectic group of great men as prophets, poets, priests, writers, and
30
kings whom he considered gifts from God and stated that the task for the rest of the
world is to recognize the gifted and to follow them (Benmira & Agboola, 2021).
Trait Theory of Leadership (TTL). According to the founders of the TTL,
people are born with inherited traits, and some of these traits or attributes make them
great leaders (Hunt & Fedynich, 2019). Several studies analyzed leaders' mental,
physical, and social characteristics to identify traits or the combination of attributes
common among leaders; however, they have yet to yield any conclusive results due to a
lack of psychometric evaluations (Hunt & Fedynich, 2019). However, the authors of TTL
suggested four primary traits that can lead to successful leadership: (a) emotional
stability, (b) admitting mistakes, (c) excellent interpersonal skills, and (d) intellectual
ability. Like CTL, TTL claims that people inherit traits, and some inherited traits or
attributes help them succeed as great leaders. In CTL, Fiedler (1967) used 18 traits or
qualities of leaders to measure their leadership style or behavior in his LPC scale.
Transactional Leadership Theory (TRALT). According to Burns (1978), the
founder of TRALT, a positive transaction that creates a mutually beneficial relationship
between the leader and the follower makes the leadership. Like CTL, TRALT claims that
the effectiveness of leadership depends on the leader's means to adequately reward (or
punish) his followers for performing the leader-assigned tasks (Changar & Atan, 2021).
The leader motivates followers through punishment and reward, and the anticipation of
the reward keeps the followers obedient. Leaders share a highly valued relationship with
their followers and depend on the followers' perception concerning the leader's fairness
and equity of the relationship (Young et al., 2021). Like Fiedler's claim in CTL, authors
31
of TRLT state that TRLT fits in organizations with a well-defined hierarchy that specifies
the roles of leaders and followers, where people agree about the need within the
organizational structure to complete tasks to accomplish goals which enhances leadership
efficiency. In this hierarchy, everyone should know who the leader is and who is
following, and the leader's effectiveness depends on the subordinates obeying the leader.
Contrasting Theories
Behavioral Leadership Theory (BLT). BLT was introduced by Blake and
Moutan (1945, as cited in Benmira & Agboola, 2021). BLT focuses on the behaviors of
the leaders and not considers their mental, physical, or any other inherent characteristics.
Behavioral studies using the cause-and-effect relationship of specific human behaviors of
leaders with psychometric measurements divided leaders into the following two
categories: (a) those concerned with the tasks and (b) those involved with the people. The
task concerned leaders focus on the organizational structure, the operating procedures,
and management strategies. The people-oriented leaders mainly focus on satisfying the
needs of the people and seeking to motivate staff through effective human relationships
(Benmira & Agboola, 2021).
Situational Leadership Theory (SLT). The SLT introduced by Hersey and
Blanchard (1970, as cited in Benmira & Agboola, 2021) indicates that leaders must adapt
their leadership styles based on their team members' maturity and readiness levels to
perform the specific tasks for projects under consideration. According to founders of
SLT, employees or team members are different in (a) being knowledgeable, (b) willing,
and (c) motivated to complete the assigned task or project work. Therefore, a manager
32
must analyze the maturity (readiness) level of their employees or team members before
applying a leadership style and then apply the appropriate leadership style for the
situation. Founders of SLT define the employees' maturity or readiness as a combination
of (a) competence (task-relevant knowledge and skills, and transferable skills) and (b)
confidence/commitment (motivation, self-confidence, and attitude toward work and
others).
According to the founders of SLT, there are four levels of maturity/readiness of
employees or teams based on their competence and confidence/commitment: (a) low
level of maturity/readiness (low competence and low confidence/commitment), (b)
moderate level of maturity/readiness (low competence and high
confidence/commitment), (c) moderately high level of maturity/readiness (high
competence and low confidence/commitment), and (d) high level of maturity/readiness
(high competence and high confidence/commitment) (Rodić & Marić, 2021). The
authors of SLT outlined four leadership styles for managers based on the four levels of
maturity/readiness of employees: (a) "Telling" leadership style, where managers must be
very specific, provide clear directions to employees of what is needed to accomplish the
tasks, answer all questions of all the aspects of the task assignment for the project; (b)
"Selling" leadership style, where managers must seek input from their employees/teams
to clarify the tasks for the project rather than just providing instructions on how to
accomplish the tasks for the project; (c) "Participating" leadership style, where managers
clarify the organization's goals and objectives with the teams and enhance
employees'/team's willingness and sense of security related to the tasks, and (d)
33
"Delegating" leadership style, where managers empower their employees/teams to work
independently, and only offer assistance when needed.
The theorists of SLT describe two behavioral categories appropriate for leaders:
(a) TO behavior and (b) RO behavior (Henkel et al., 2019). During highly TO behaviors,
managers should know all the details associated with all the tasks of the projects before
acting or directing the team members to accomplish the tasks and apply a 'Telling"
leadership style. Conversely, managers should develop trust and respect with the
employees or their teams during RO behaviors to build action plans by applying the
"Delegating" leadership style. The flexibility defined in STL that allows leaders to
change and adapt their leadership style to meet the needs of employees makes it different
from Fiedler's CTL, which states that a leader's LB is inherent and cannot be changed.
Transformational Leadership Theory (TRFLT). The TRFLT, one of the most
studied and popular leadership theories, indicates that leaders inspire followers to alter
preconceptions and perspectives and motivate them to work toward similar objectives
because of the strength of their vision, personality, and charismatic behavior
(Aboramadan & Kundi, 2020; Y. Lee et al., 2020; Yin et al., 2020). In transformational
leadership, charisma, intellectual stimulation, and individualized consideration play
significant roles as components of leadership, making the followers behave in a more
mature and idealistic way, showing more significant concern for achieving goals, self-
actualization, and the needs of their fellow workers, their organization, and society
(Aboramadan & Kundi, 2020). Transformational leaders demonstrate a vision of a
positive future, explain how this can be achieved, and lead by example through high-
34
performance standards, confidence, and determination, influencing followers to think
beyond their self-interest and work for the group with collective interest (Luu et al.,
2019).
The Unit of Analysis and Measurement of Variables
Unit of Analysis
This unit of analysis defines the structure of the research data. It provides the
correct contextual participant entities, which results in excellent and valid meanings that
ensure the integrity of research outcomes (W. Li et al., 2017). I investigated the
relationship between the LB of the PMs in charge of DT projects and DT project
completion status in three contingency situations, namely (a) the PM’s PP, (b) the PM’s
TS, and (c) the PM’s LMR (Fiedler, 1967). Data for all the independent
variables/constructs and the dependent variable I chose to study were available at the PM
level. Therefore, the unit of analysis for this study was the PM, which was appropriate.
Selecting the proper unit of analysis is an integral component of any research
study. Improper selection of the unit of analysis, such as choosing a unit at a micro-level
rather than what is needed, will make the research more time-consuming and costly (W.
Li et al., 2017). Suppose the chosen unit of analysis is at a much higher macro-level than
required. In that case, meaningful connections and contextual meanings at smaller units
may be missed in the investigation, leading to erroneous categorization and interpretation
of the data (W. Li et al., 2017). Also, successful DT project completion requires effective
resource governance, clear project goals, good LMR, and visibility into the DT journey
for improved products, profitability, and customer experience (Pulkkinen et al., 2019).
35
The PMs who managed or are currently managing DT projects’ successful
implementation in LICs in the United States must possess knowledge of the above.
The selection of the PM responsible for managing the DT project for successful
completion (dependent variable of this study) as the unit of analysis for this study aligns
with the purpose of this doctoral research. The PM demonstrates LB (an independent
variable in this study) and possesses the PP (independent variable) needed to influence
and guide the employees (Gilli et al., 2023). Further, the PM cultivates good LMR (an
independent variable in this study) to motivate group members and defines the project’s
task structure (an independent variable in this study) to enhance task completion (Appio
et al., 2021). Therefore, the PM was an appropriate object for this doctoral study's unit of
analysis.
Measurement of Variables
The dependent variable of this study was the DT project completion status
achieved by the PM in the selected companies. The variable could take only two possible
values (yes/no), thus binary and categorical. Besides, the primary independent variable
was the LB of the PMs managing during the data collection for this study or managed DT
projects in LICs in the United States. Per Fiedler's CTL model (Fiedler, 1967), PM's LB
(the primary independent variable) can take only two values (TOLB or ROLB) and is
categorical. Per Fiedler's CTL model (Fiedler, 1967), the three contingency independent
variables can take only two possible values, which are categorical and dichotomous. PM's
PP can be either weak or strong, the PM's LMR can be either good or bad, and the PM's
TS can be structured or unstructured (Fiedler, 1967).
36
Primary Independent Variable: Leadership Behaviors of Managers
Over the past one hundred years, leadership research yielded convincing evidence
that an organization's success depends mainly on its leaders' LB (Brunner et al., 2023;
Halliwell et al., 2022; Kapucu, 2020; Klein, 2020; Müller et al., 2024; Porfírio et al.,
2021; Yukl et al., 2019; Zeike et al., 2019). Three entities constitute the leadership: (a)
the leader, (b) the tasks that require completion, and (c) the followers who strive to
accomplish the tasks. According to several previous investigators, including Bartsch et
al. (2021), Halliwell et al. (2022), Klein (2020), and Zeike et al. (2019), the LB of leaders
falls into a dichotomy of the following two meta-categories: (a) TOLB and (b) ROLB.
This dichotomy of two meta-categories of LB was the root of many leadership theories
and taxonomies for more than 60 years (Bartsch et al., 2021; Halliwell et al., 2022; Klein,
2020; Zeike et al., 2019).
TOLB and ROLB categories are relevant to all work environments and project
structures (Bartsch et al., 2021; Solberg et al., 2020; Yukl et al., 2019). According to
Yukl et al. (2019), the TOLB category covers the following leadership competencies: (a)
enhancing understanding of the current situation, (b) mitigating risks, (c) strengthening
motivation, (d) encouraging innovation and collective learning, and (e) enhancing all
change-oriented activities and is best applied when facilitating implementation in the
planning and action phases of projects. The ROLB category covers effective leader-
follower engagement in accomplishing objectives such as (a) increasing coordination to
synchronize collective efforts, (b) promoting cooperation to encourage more outstanding
individual contributions, and (c) motivating and activating resources to expand valuable
37
contributions (Yukl et al., 2019). The two LB categories become important in contingent
situations. For example, leaders encouraging numerous contributions without an
established task structure cause risk. Also, leaders risk losing credibility and member
support if they do not recognize individuals' most valuable contributions or discourage
them from inappropriately contributing.
The Impact of LB on DT Project Implementation
The highly transformational situations in DT projects and the urgent requirement
for sourcing employees with digital skills from multiple areas around the globe pose
several challenges to companies. R. Ahmed et al. (2024), Brown et al. (2021),
Cetindamar et al. (2024), and Ivanenko and Artamonov (2020) showed that managers'
TOLB and ROLB are vital in overcoming these DT challenges. Managers with TOLB
focused on attaining organizational objectives by clarifying each task's goals and
monitoring work processes, while managers with ROLB focused on enhancing
collaborative interaction among corporate members and establishing a supportive climate
(Bartsch et al., 2021; Solberg et al., 2020). In their research, R. Ahmed et al. (2024)
showed that supervisors' TOLB (p < .001) and ROLB (p < .001) had significant positive
correlations with project success. In another study by Correani et al. (2020), DT
managers who clearly defined project goals, job roles, processes, and procedures
enhanced the organizational knowledge base through DT data generation and
successfully implemented DT projects.
According to Dubey et al. (2020), aligning DT leaders' LB with digital technology
adaption under different situations leads to enhanced operational performance. Research
38
by Shao (2019) indicated a significant positive correlation between DT managers' LB
and business strategy alignment and enterprise system assimilation mediated by
organizational culture. Bartsch et al. (2021), using a quantitative correlational research
design, investigating the relationship between leaders' TOLB and ROLB and the service
employees' work performance found significant positive correlations between ROLB of
leaders and individual employee's job autonomy (β = .79, p < .01) and team cohesiveness
(β = .33, p < .01). Their study involved digital maturity in an unaccustomed virtual work
environment (first-time exposure of employees to the virtual work environment) caused
by the COVID-19 crisis (Bartsch et al., 2021).
In a quantitative correlational study using 129 experienced managers worldwide,
Henkel et al. (2019) found a significant positive correlation between a distributed
combination of the managers' TOLB and ROLB and successful project completion. In
another quantitative correlational study conducted under multiple favorable and
unfavorable project situations, Popp and Hadwich (2018) reported a significant positive
correlation between managers' ROLB and employees' overall performance success
regardless of the situation. The above evidence indicated that effective LB of PMs could
lead to successful project implementation. The managers' TOLB and ROLB are the most
effective project success factors based on project contexts.
Despite extensive studies indicating its importance, there is a massive gap in
various dimensions of this significant subject. The effectiveness of LB (TOLB vs.
ROLB) may contribute positively or negatively according to the context in which the
leaders operate; leaders’ organizational power relations may influence leaders’ leadership
values
39
and challenge their approaches to leadership (Fiedler, 1967; Willis, 2019). A leader who
is effective in one organizational setup may not remain as compelling in another,
indicating that leadership studies related to DT project implementation must consider the
context in which the leaders operate (Fiedler, 1967; Willis, 2019). Therefore, the
relationship between leadership and successful DT project implementation necessitates
more comprehensive investigations in organizations.
Dependent Variable: Successful DT Project Completion Status
Successful DT project completion is a topical issue of significant importance for
all companies in all sectors worldwide as it changes customer relationships, internal
processes, and value creation (Albukhitan, 2020). DT project completion is the process
by which organizations adapt themselves to modern technology by leveraging digital
technologies such as the IoT, AI, social media connectivity, cloud computing, big data
analytics, and intelligent manufacturing technologies (I4.0) to transform organizations
for improved performance (Albukhitan, 2020; Doukidis et al., 2020; Kretschmer &
Khashabi, 2020; Warner & Wager, 2019). Successfully implementing and completing
DT projects requires changes to the company’s business model, products, and
organizational structures to adapt to digital technologies, the most pervasive managerial
challenge for companies currently and in the future.
According to Warner and Wager (2019), LICs must act now and invest in
digitized process platforms that facilitate operational excellence by choosing solid digital
technologies and strategies to enhance their value proposition, innovation, and
responsiveness to new market opportunities to stay competitive. However, successful
40
implementation of DT projects needs technology and people, as digital project tasks need
skilled employees and executives with digital management competencies to achieve
complex transformative task completion (Albukhitan, 2020; Kane, 2019; Ra et al., 2019).
In recent years, scholarly attention in the DT literature has steadily increased, leading to a
significant increase in articles addressing DT's different technological and organizational
aspects and leaders' role in driving positive results from investments in digital initiatives.
However, an underrepresentation of the leadership's role in DT project success results in
incomplete DT project completion, negatively affecting business performance (Ahmad et
al., 2022; Warner & Wager, 2019), and needs improvement in the literature. Considering
this development, I provided a descriptive literature review reflecting on the current state
of knowledge and a critical analysis of the field, assessing where, how, and who
researched DT project implementation and leadership's role in its completion.
Impacts of Successful DT Implementation to Businesses
According to several previous researchers, including Albukhitan (2020),
Doukidis et al. (2020), Govindarajan and Immelt (2019), Kretschmer and Khashabi
(2020); Mustafa et al. (2020), Pacchini et al. (2019), Schumacher et al. (2019), and Singh
et al. (2020) DT dramatically improve business processes, business strategies, customer
experience, innovation, and operational execution across various levels, including
individual, organizational, environmental, and societal. Cooney et al. (2021) reported
that ninety-three percent (93%) of CEOs of global industrial companies out of the 3000
they surveyed globally saw disruptive emerging technologies as driving competition in
their industry and DT as a critical change program and the solution to compete and
survive.
41
Therefore, successful DT (implies DT projects) implementation is a crucial driver of
competitive advantage that enables LICs to compete and survive; it is no longer optional
but the only way they can survive (Govindarajan & Immelt, 2019).
The successful implementation of DT offered many opportunities to enhance
customers, revenues, and performance (Ferreira et al., 2019; Hai et al., 2021). By
embracing new digital opportunities into their strategies, innovative and agile businesses
maintain their positions in competitive markets (Schumacher et al., 2019); by responding
to new digital opportunities, they become resilient against risk (Björkdahl, 2020).
Through the enhancement of resources and capabilities and the reconfiguration of
processes and structures per modern technologies (Chirumalla, 2021), adjustments in
leadership (Brunner et al., 2023; Klein, 2020; Ko et al., 2022; Porfírio et al., 2021), and
the implementation of digital culture (Kretschmer & Khashabi, 2020) organizations
enhanced their customer experience, productivity, competitive advantage, and
sustainability. Successful implementation of DT projects enabled novel product and
service offerings (Sony & Naik, 2019), transformed the structure of supply chains
(Ishfaq et al., 2021), and vastly improved processing power through reduced costs,
increased productivity, enhanced processing efficiency, and value-added through
dedicated services, enabling faster time-to-market (Kraus et al., 2021a; Verhoef et al.,
2021). The literature indicated that successful DT implementation enhanced customer
relationships (Doukidis et al., 2020; Kretschmer & Khashabi, 2020), reshaped industry
competition (Sony & Naik, 2019; Verhoef et al., 2021), and generated new value through
42
interconnecting physical and digital assets with data and ecosystems (Schumacher et al.,
2019; Verhoef et al., 2021).
Although success received much attention in DT literature, failure received much
lesser attention (Mustafa et al., 2020). Plenty of cash flows into digital initiatives at
industrial companies; despite the massive investment, the expected results often fail to
materialize. In eighty-four percent (84%) of LICs, including GE, Ford, P&G, and many
more, DT initiatives were a wasted opportunity that led to company failure (F. Li, 2020;
Reeves et al., 2018). According to Datta and Nwankpa (2021), McCarthy et al. (2024),
and Reeves et al. (2018), 70% of the U.S. enterprises' DT initiatives in 2018 failed to
achieve their stated goals as forecasted, equating to over $900 billion out of the 1.3
trillion worth of spending that went to waste. Based on another McKinsey global survey
of LICs, two-thirds of organizations generated only ten (10) to fifteen (15) percent of
revenue through digital amidst a large amount of money invested in DT (1.3 trillion in
the United States alone in 2018). According to Correani et al. (2020), recent estimates
indicated 66% to 84% (an average of 75%) of DT projects' failures, a sizable proportion
considering the monetary and other costs of putting these projects in place. The potential
impact and scale are so significant that flawed and imperfect DT can significantly hurt
companies' competition and survival. The uncertainties and challenges heighten the need
to examine mechanisms through which DT can deliver the desired value.
Factors Influencing Successful DT Project Implementation
DT project implementation involves adopting complex digital technologies,
acquiring digital talents, upgrading internal structures per selected digital technologies,
43
enhancing digital platforms and innovation to improve customer experience, creating a
digital culture, and achieving operational excellence by integrating processes and people
with digital technologies. The literature contains limited representative examples of
companies demonstrating successful DT project implementation (Ivančić et al., 2019).
Also, the implementation of DT projects involves several dimensions: (a) technology, (b)
organization and business processes, (c) people, and (d) environment (Müller et al.,
2024) and many success factors, including strategies, leadership, culture, employee
training and motivation, and customer and stakeholder relations (Schumacher et al.,
2019; Verhoef et al., 2021).
A significant trend discovered from the literature review was a shift of attention
from technological factors to managerial and organizational issues in the years, a theory
supported by several past researchers including Brunner et al. (2023), Facchini et al.
(2022), Ko et al. (2022), Metwally et al. (2019), Müller et al. (2024), Mustafa et al.
(2020), Schumacher et al. (2019), and Singh et al. (2020). Incompatible leadership
behaviors with the organizational culture and capabilities potentially pose significant
challenges, causing cultural incongruence, transformation obstacles, and managerial
challenges leading to the failures of DT projects (Brunner et al., 2023; Müller et al.,
2024). Also, no standard methods and best management practices are currently available
for organizations to successfully implement DT projects (Cichosz et al., 2020; Fischer et
al., 2020).
44
Success Dimensions of DT Project Implementation
When considering projects, there are two main success concepts: (a) project
success and (b) project management (PMT) success. According to the Project
Management Institute (2023a), PMT is the "application of specific knowledge, skills,
tools, and techniques to deliver value to people." According to Mishra (2020) and
Venczel et al. (2021), PMT is "planning, organization, monitoring, and control of all
aspects of the project," with the motivation of all people to achieve project goals safely
within the agreed schedule, budget, and performance criteria. Thus, PMT aims to
complete projects as intended, most efficiently, by minimizing cost and time and
achieving external goals related to customer needs.
Although there are several similarities and differences between the two success
dimensions, one significant difference is that project success is the overall project goals
achievement, while PMT success is successful project implementation within the
traditional measurements of time, cost, and quality; the two success criteria are mutually
related, and therefore it is hard to differentiate between them strongly. Factors such as
scope, time, budget, resources, risk, and performance specifications designed to meet a
specific customer need limit the project's success (San Cristóbal et al., 2018). PMT and
project manager's leadership competencies played an essential role in achieving project
success and positively influenced it (Alvarenga et al., 2020; Oh et al., 2021). Therefore,
PMT practices and project success are significantly positively related, and PMT success
is an essential element of project success because the latter is hardly achievable without
it.
45
Despite the considerable efforts to meet success goals, many projects continue to
run late, exceed their budgets, or fail to meet customer satisfaction (Correani et al., 2020;
Ivančić et al., 2019; Mustafa et al., 2020; Reeves et al., 2018). No standard methods and
best management practices are available for organizations to successfully implement DT
projects (Cichosz et al., 2020; Fischer et al., 2020). Lack of employee engagement and
ineffective strategy planning (Ko et al., 2022), weak or nonexistent cross-functional
collaboration, poor organizational design (Singh et al., 2020; Verhoef et al., 2021), lack
of knowledge among PMs regarding modern technology and how to structure project
tasks, lack of strategic guidance from top management towards realization (Brunner et
al., 2023; Müller et al., 2024; Pacchini et al., 2019; Schumacher et al., 2019), and poor
quality management practices (Brunner et al., 2023; Müller et al., 2024) are causes of the
failure of DT projects. There is a huge need for scholars to investigate how organizations
could manage a DT project to succeed.
Successful DT Project Completion Criteria and Models
According to the PMT Institute (San Cristóbal et al., 2018), a project involves a
group of activities that begin and end at specific points in time based on a pre-defined
schedule and within a consistent budget that specified team members execute to meet a
set of objectives to achieve a desired goal. A DT project provides the necessary structure
through which an organization implements its DT initiative with DT success as the goal
and embodies the processes necessary to reach that goal. PMs combine the project's
different parts, the team, technology, processes, governance structures, and other
elements via complex interactions to deliver a common objective. Hence, DT projects
are
46
complex systems with a high risk of facing uncertainties that may benefit from effective
management practices (Fischer et al., 2020). Although there is no consensus definition of
project success, there is an agreement in the literature that good PM actions, such as good
PMT, lead to project success.
The iron triangle model (ITM) was the first applied to measure PMT and project
success (Goldsmith & Boeuf, 2019). According to the concept of ITM, a project is
successful if it meets the objectives of ITM (within budget, on schedule, and per agreed-
upon quality [per stakeholder satisfaction]). ITM has been the standard measure of
project success for decades in the business community; businesses commonly measure
the successful completion of projects through the ITM components (Goldsmith & Boeuf,
2019; Zid et al., 2020). According to Goldsmith and Boeuf (2019) and L. H. Nguyen
(2021), ITM's three components contribute significantly, are strongly linked to
successful project implementation, and are reliable. An increase in scope without a
corresponding increase in time and cost can result in poor quality of work, or a decrease
in time without a decrease in scope can lead to poor quality if cost remains constant (Zid
et al., 2020).
Organizations could define project success using the ITM success criteria (cost, time, and
quality) to meet the stakeholders’ satisfaction and expectations; the inability to complete
projects on time, within budget, or per quality expectations poses challenges in executing
projects (Ika & Pinto, 2022; Zid et al., 2020). Despite the massive investments, many
projects face many challenges, such as expenditures exceeding the budget, which can
cause project failures, delays in completing the project on time, cost overruns, and low-
47
quality products that build defects, leading to disappointment and lost customers (Ika &
Pinto, 2022; X. Wang et al., 2023; Zid et al., 2020).
Recent research by L. H. Nguyen (2021) indicated that the ITM measure of
success implies that a project that is a day late or above budget by a dollar might be
unsuccessful even if it meets quality expectations; hence, it is not a valid measure of
project success. According to Goldsmith and Boeuf (2019) and Yan et al. (2019), project
success depends not only on time, cost, and quality but also on integration, human
resources, technology, communication, risk, and procurement management. According to
Al Dabbas and Alkshali (2021), businesses must include customer satisfaction in project
success assessments using the ITM objectives. Thus, the ITM measure needs broadening
to include these factors for stakeholders' satisfaction, benefits to the organization that
owns the project, and long-term impacts on the project environment. Currently, only
conflicting evidence exists regarding the success criteria of projects.
Integration and Capability Building
DT is an ongoing process and depends on the capabilities in multiple dimensions
of the companies to balance internal and external collaboration, redesign manageable
governance structures, and improve and promote the workforce's productivity (Rehman
et al., 2020; Warner & Wager, 2019; Zhou et al., 2021). Managing the multi-product
customer experiences and changes to business processes requires integration across
business units and managing customer data, global operations, work silos, organizational
politics, and path-breaking exogenous and endogenous mechanisms (Rehman et al.,
2020; Warner & Wager, 2019; Zhou et al., 2021). The capability-building process
48
depends mainly on trust, collaboration, knowledge sharing, and good interpersonal
relationships among all the people involved (Klein, 2020; Porfírio et al., 2021; Warner &
Wager, 2019). Superior leadership with digitalization experience and knowledge of
managing conflicting demands are significant to successful DT project implementation.
In the literature, the following two capabilities currently exist as core to DT
project success: (a) digital capability (DC) and (b) leadership capability (LC); Companies
can transform digital technology into a business advantage with these two capabilities
(Klein, 2020; Porfírio et al., 2021). DC enables companies to use innovative digital
technologies to improve elements of their business (Porfírio et al., 2021). LC enables
companies to envision and drive organizational change, fostering strategic technology
initiatives systematically and profitably in the digital era (Klein, 2020; Porfírio et al.,
2021). According to Bonnet and Westerman (2021), Ivančić et al. (2019), and Mager and
Katzenbach (2021), successful change management and talent development are essential
requirements of DT in LICs and supported by strategic DT initiatives from the top
leaders necessitating mutually respectful relationship among all stakeholders.
Collaboration among all stakeholders, creating cultural change, enabling agile
work environments, and choosing appropriate DT technology led to DT projects' success
in LICs (Gurbaxani & Dunkle, 2019; Ko et al., 2022). Suitable alignment of a company's
DT strategy to the leaders' task focus (Singh et al., 2020), adequate planning during the
DT implementation (Correani et al., 2020; Fischer et al., 2020; Gurbaxani & Dunkle,
2019; Reeves et al., 2018), project managers adopting quality leadership behaviors
(Klein, 2020; Müller et al., 2024; Porfírio et al., 2021), and employees' firsthand insights
49
into where processes needed improvement (Bonnet & Westerman, 2021; Sousa-Zomer et
al., 2020) had strong potential to support organizations' successful DT implementation.
How This Research Addressed a Literature Gap
In the literature, research investigating the role of LB in the context of planned
organizational change existed. However, no studies exist related to the role of effective
LB in a rapid and volatile organizational change like DT. This study offered practical
contributions to the meager literature on LB, which could support businesses in managing
the challenging digital task fulfillment and employee relationship for successful DT
project implementation and completion. Because of the scant literature on DT and LB,
this study aimed to reveal information on enabling and managing LB for successful DT
project implementation and completion.
Transition
In section 1, I introduced this doctoral study with a brief background related to its
business problem. I included a few critical elements of the study, such as a problem
statement, a statement of the study’s purpose, the nature of the study, research questions
addressed, hypotheses tested, theoretical framework used, operational definitions,
assumptions, limitations, and the study’s delimitations, the significance of the study, and
a review of the professional and academic literature. In section 2, I described this study’s
quantitative research method and design approach, including the participants, population
and sampling, instrumentation, data collection techniques, data analysis, and study
validity. In section 2, I also included my role as a researcher in this study and ethical
research conduct. In section 3, I presented this research’s findings, including the
50
application to professional practice, implications for social change, recommendations for
action, suggestions for further research, reflection, and the study’s conclusion.
51
Section 2: The Project
In Section 2, I discuss the project procedures related to studying the relationship
between project managers' (PMs') leadership behaviors (LB) and successful digital
transformation (DT) project completion in large industrial companies (LICs) that
employed over 500 people in the United States. Section 2 also includes a purpose
statement and addresses my role as a researcher in conducting this research, the study
participants, the sampling strategy, and the data collection process, including the chosen
population for this study and the total sample size, participant recruitment procedure, and
instruments used to collect data. Section 2 further includes the research method and
design chosen to support the study’s relevancy, an overview of the data analysis process,
the study's validity, and ethical conduct during the entire study.
Purpose Statement
The specific business problem was that some PMs in LICs do not know the
relationship between a PM's LB and DT project completion status. Therefore, the
purpose of this quantitative correlational study was to investigate the relationship
between PM's LB and DT project completion status. The independent variables were (a)
PM's LB, (b) PM’s LMR, (c) PM's TS, and (d) PM's PP. The dependent variable was DT
project completion status. The target population consisted of PMs of LICs located in the
United States, with a focus on digitally transforming their businesses.
Role of the Researcher
In quantitative (QUAN) research studies, the primary roles of the person
conducting the study include generating practical research questions related to the study’s
52
business problem, collecting and organizing data, conducting hypothesis tests, and
combining and documenting the findings (Mohajan, 2020). I identified and reviewed
exhaustively the most relevant scholarly literature published by previous researchers
related to this study’s area within the past 5 years, carefully selected and reported the test
variables, designed relevant research questions to explain the research problem, and
selected appropriate research design(s) to fit the research questions addressed. I identified
and used psychometrically robust and contextually sensitive measurement instruments
and appropriate data collection tools, appropriately selected participants, generated
sufficient data, and performed accurate analysis. I further chose and applied appropriate
statistical methods to process the data, whether the prediction was confirmed or not,
verified the results, drew conclusions, presented the findings from the hypotheses testing,
and shared the findings with participants, the public, and other relevant authorities.
Completing the Collaborative Institutional Training Initiative (CITI Program)
certification prepared me to follow ethical research principles. I followed the Belmont
Report's protocols (Office of Human Research Protections [OHRP], 2022) and protected
participants' rights, including confidentiality, privacy, respect for persons, beneficence,
and justice, which captured the ethical values inherent in quality research. I ensured
participants' privacy by conducting an anonymous survey and not collecting participants'
names or business organizations. In QUAN studies, data collection happens
independently of the person collecting the data, thereby minimizing the chances of bias
and undue influence on the participants. I ensured respondents' confidentiality strictly
53
through a robust informed consent process. I ensured participants' confidentiality through
(a) explaining to the participants that their participation in the study was strictly
voluntary without compensation or incentive and (b) ensuring that research participants
understood that they could withdraw from participation at any time without penalty or
loss of any benefits that they might be entitled to if they intended to.
Per the Belmont Report's basic ethical principle of beneficence (OHRP, 2022), I
chose the appropriate research design and situation to maximize benefits and reduce
risks to participants as much as possible and ensured that this research's findings would
improve their lives and circumstances (Head, 2020). Per the principle of justice of the
Belmont Report protocol (OHRP, 2022), I ensured justice during research by (a)
considering that the potential societal benefit from the research justifies the cost to
subjects and (b) not using the data collected for purposes outside the scope of this
research study. The Belmont Report further insists on informing the participants of the
study's purpose and their eligibility to access the results and a copy of the research
findings. The informed ICF included information about this study's purpose and
participants' rights and access to the study's findings (Appendix B). No termination of
any participants from the study occurred.
In quantitative studies, one of the main goals is to generalize the findings from a
sample to a population (Shieh, 2020). When the sample size in the hypothesis test is
large, minor effects become detectable through hypothesis testing, sampling error
becomes less of a problem, and accurate inference is guaranteed, which is a closer
54
approximation of the population (Bhardwaj, 2019; Fricker et al., 2019). However, any
research study that involves data collection and analysis requires resources, including
subjects, time, and money, that need consideration when selecting a sample size
(Bhardwaj, 2019; Lakens, 2022; Park et al., 2020). I conducted an a priori power analysis
and identified the required minimum sample size that satisfied the following: (a)
balanced the research's statistical validity and resource constraints, (b) avoided an
underpowered study with a low probability of detecting a significant effect with a smaller
than required sample size (Lakens, 2022), (c) was large enough to have sufficient power
to detect meaningful effects (Bhardwaj, 2019; Lakens, 2022), (d) met the requirements of
the chosen analysis methods and tools, (d) minimized waste, and (e) was small enough to
avoid exposing participants to unnecessary risks (Kang, 2021).
Participants
This doctoral study’s participants consisted of PMs who managed or were
managing (during data collection of this study) DT projects sampled from large
companies (# of employees ≥ 500 as defined by the federal government) from the
industrial sector (LIC) in the United States, with a focus on digitally transforming their
businesses. The participants were from the LICs because LICs possess tremendous
potential and the necessity for DT (Ghosh et al., 2022; Reeves et al., 2018). The
eligibility criteria that I considered for the participants of this study included (a)
participants were PMs who managed or were managing one or more projects that
involved DT and were initiated within the past 5 years from the beginning of data
55
collection for this study; (b) the sample of participants came from the target population;
(c) the participants were between 18 and 80 years of age, male or female, of any race and
nationality, located in the United States, and had worked or were working (during data
collection for this study) for a U.S.-based large (companies having ≥ 500 employees)
industrial company (LIC); and (d) the participants did or did not complete the DT project
they managed or were managing. In addition, participants were willing to participate in
the study and had access to the internet or electronic mail to complete the survey.
This study included participant PMs with one of the following project completion
statuses: (a) completed at least one DT project successfully (on time, within budget, and
per quality standards), or (b) did not complete any of the DT projects that they initiated
successfully (did not complete the DT projects or completed not on time, not within
budget or not per agreed-on quality standards or a combination of the three). A project
completed successfully in this study meant the project was completed within schedule
(within 3 years or as defined by the company), within the allocated budget, and per the
quality criteria intended. A project not completed successfully meant one of the
following: (a) the project was either completed but not within the schedule or not within
the allocated budget, or not per agreed quality criteria and any combination of these; (b)
the project was never completed and in progress after the scheduled completion date; or
(c) the project was abandoned due to causes such as inadequate resources, a flaw in the
design, or the like.
56
I conducted an anonymous survey using Walden's IRB's Anonymous Survey
Consent Form for DBA Survey (ASCFDS). As the first step, Centiment obtained a panel
of participants from the LICs who met the participant eligibility criteria (Appendix A)
and screening questions on project completion status (Appendix C) I provided. Through
Centiment's online survey tool, the eligible, willing, and available participants accessed
the ASCFDS form (Appendix B). I transferred the survey of about 50 closed-ended
questions and the ASCFDS form to Centiment for administration through Centiment's
free online survey tool to eligible and willing participants. All eligible participants who
read and agreed to ASCFDS's terms and clicked a continuing link through Centiment's
online survey tool and completed the survey indicated their consent to participate. The
survey remained anonymous, and no personal information was collected. I monitored the
website periodically for any assistance needed by any participants. I exercised caution
concerning the legitimacy of the internet source used to collect data and cross-checked
the accuracy of the data collected with other legitimate sources and documents.
Centiment is highly reliable, profiles its respondents extensively, and collects high-
quality data (https://www.linkedin.com).
Effective and continuous communication stimulates active participation and
safety of participants (Y. Wang et al., 2019). The anonymous online survey conducted
in this study facilitated easy access, speed of data collection, and lower cost. I ensured
that Centiment notified the participants via statements the week before posting the
survey questionnaire, provided reminders periodically after posting the survey, and gave
57
adequate time to prepare and answer the questions through Centiment's online survey
tool. I provided the participants with a brief overview of the study in ASCFDS without
discussing the survey questions or directly discussing the study's aim, participant
selection criteria, and how I planned to use the results. Through the ASCFDS, I
emphasized that the information gathered would help formulate better strategies and
models to improve DT initiatives and enhance the company's profitability and success.
None of the participants needed special aids, such as translation tools or training, to
answer the questions.
Research Method and Design
Research Method
Quantitative (QUAN), qualitative (QUAL), or mixed-method research methods
are possible choices for doctoral studies (Blair et al., 2019). QUAN was the chosen
research method for answering the research questions for this doctoral study. QUAN
methods yield better results and allow for informed decisions in doctoral research studies
aimed at collecting numerical data that involve technology (Hosseini et al., 2019), are
appropriate when testing hypotheses for verifying existing theories (Yue & Xu, 2019),
and are essential in studying project success in dynamic and complex environments with
disruption risks (Hosseini et al., 2019). In this research, I used a single theory and
correlational research design to find answers to research questions by testing hypotheses
using objective and impartial statistical methods.
58
I used a theoretical framework to obtain a reliable estimate of a generalized
relationship between the outcome and multiple predictor variables, used structured
research instruments such as survey questionnaires to collect numerical data for the
selected variables, developed and tested hypotheses guided by the framework, and
concluded and inferred the results to a larger population. Therefore, the QUAN method
was appropriate for this study. This study constituted a large sample size (N = 214) to
represent the population and aimed at inferring the results to a larger population. QUAN
methods use large samples with results generalized to an entire population,
subpopulation, and other legitimate studies (Lo et al., 2020; Mohajan, 2020; Nzekwe-
Excel, 2022) and were, therefore, appropriate for this study. QUAN methods also adopt
an objective perspective and help minimize researcher bias (Bloomfield & Fisher, 2019).
In doctoral studies, QUAL methods do not allow the measurement of facts or
objects or report numerical data (Köhler et al., 2022; Nzekwe-Excel, 2022) and use data
collected from a few participants' experiences and behaviors through interviews,
conversations, observations, and or field notes aligned with subjectivist epistemology,
which reduces the possibility of generalizability to a larger population (Ames et al.,
2019; Köhler et al., 2022; Nzekwe-Excel, 2022), making QUAL methods inappropriate
for this study, as this study used numerical data and aimed to generalize the findings to a
larger population or subpopulation. An MM approach that combines QUAL and QUAN
methods into a single study is appropriate only when neither a QUAN nor QUAL
approach alone addresses the research topic or when a study requires one method to
59
inform or clarify another (Fàbregues et al., 2021; Halcomb, 2019; Şahin & Öztürk,
2022). The QUAN method was appropriate and adequate, but a QUAL or an MM
approach was not proper or necessary for this doctoral study. Also, according to the
theory of incompatibility of paradigms, the assumptions for QUAN and QUAL methods
are different, and ad hoc mixing of the QUAN and QUAL methods can severely threaten
research validity (Denzin, 2012; Şahin & Öztürk, 2022; Yousefi Nooraie et al., 2020).
Research Design
A research design is a plan that helps the person conducting the research answer
specific research questions by determining the hypothesis, conducting the study, and
analyzing and interpreting the data (Bloomfield & Fisher, 2019). I chose a correlation
research design for this research to systematically investigate the nature of the
relationship between multiple dichotomous independent (predictor) variables and a
dichotomous dependent variable without manipulating the variables but occurring
naturally. For the dependent and independent variables, data collection occurred without
interference or manipulation by the person collecting the data (in this case, me), for
which the nonexperimental design, such as correlational research design, was
appropriate (Seeram, 2019). Further, I used statistical analysis, including calculating
correlation coefficients, conducting regression analysis, and conducting other statistical
tests to determine the strength and direction of the relationship between the dependent
and predictor variables and to determine whether the data supported a theory for which
correlational research design was the best (Seeram, 2019).
60
Descriptive research designs enable the description of the status of a variable or
phenomenon, such as the frequency of something that exists, particularly a new
phenomenon, or about which very little information is available (Bloomfield & Fisher,
2019). Although the findings from descriptive research studies may help develop
hypotheses, they do not help establish relationships or study the correlation between
variables; therefore, they were inadequate and inappropriate for this study. Experimental
and quasi-experimental research designs help establish cause-effect relationships among
multiple variables and test interventions' effectiveness. Therefore, they require the
manipulation of one or more independent variables while controlling other variables to
study the cause–effect relationship (Bloomfield & Fisher, 2019). Experimental or quasi-
experimental designs did not apply to this study because this study predicted the
relationship between multiple binary independent variables and a binary dependent
variable without manipulating the variables.
Population and Sampling
Population
The population for this doctoral research consisted of project managers (PMs)
who managed or were managing during data collection for this study DT projects from
large companies (# of employees ≥ 500 as defined by the federal government) from the
industrial sector in the United States who focused on digitally transforming their
businesses. This population aligned with the research question and the hypotheses for
this study because this study investigated the relationship between PM's LB and DT
project
61
completion status. The population consisted of PMs paneled through the Centiment
Survey Panel (Centiment). Centiment paneled a large pool of PMs who met the eligibility
criteria based on the screening criteria provided in Appendix A and Appendix C, who
were the population for this study.
Sampling
Sampling involves selecting a subset from the target population. I did not have
adequate resources or time to use the entire population of all PM participants managed or
managing DT projects from all the LICs in the United States in this study; therefore, I
selected a sample from the large target population using an appropriate sampling method
to answer the RQs of this study. Sampling has two primary methods: (a) probability or
random sampling and (b) nonprobability or nonrandom sampling. Selecting an
appropriate sampling method (probability or nonprobability) in a research study ensures
the study’s quality and validity (Matthes & Ball, 2019).
Probability Sampling
Probability sampling, which is selected based on probability theories from
mathematical statistics, is the primary sampling method appropriate in quantitative
studies (Berndt, 2020; Bhardwaj, 2019). Probability or random sampling has the greatest
freedom from bias and is an efficient way to reduce bias (Berndt, 2020; Bhardwaj,
2019); it ensures that every item in the target population possesses an equal chance of
being selected, making the sample more representative and findings effectively
generalized to a
62
larger population (Lamm & Lamm, 2019; Rahman et al., 2022), therefore was best suited
to this QUAN study.
Nonprobability Sampling
The nonprobability sampling method uses nonrandomized methods such as the
judgment of the person conducting the research, convenience, and ease of access to the
population (ex., selecting friends and classmates) in place of randomization (Andrade,
2021; Rahman, 2023). The purposive nonprobability sampling selects participants based
on characteristics defined for a purpose relevant to the study, and the convenience
nonprobability sampling method involves selecting the study participants based on their
availability (Andrade, 2021; Rahman, 2023). Although nonprobability sampling costs
less and consumes less time, they do not adequately represent the population (Andrade,
2021; Rahman, 2023), and results using data from this sampling technique lack
generality and generalizability to a larger population (Lamm & Lamm, 2019), therefore
was inappropriate for this QUAN doctoral study.
Sampling Subcategory
Probability sampling has two subcategories: (a) stratified random sampling
(STRS) and (b) simple random sampling (SRS). The chosen sampling subcategory for
this doctoral study was the STRS. I could not use the SRS because SRS is appropriate
when the population is homogeneous and a complete sampling frame, which is all the
units in the entire population, is known (Bhardwaj, 2019; Fowler & Lapp, 2019). This
study’s population size consisted of paneling statistics provided by Centiment, and the
63
population size was only approximately known to me. Centiment divided the target
population into subgroups called strata and drew samples from each stratum. The strata
were more homogeneous than the total population, with each element within a stratum
identified, the situation was well-informed, and the samples were more reliable and
contained detailed participant information. This sampling approach of participants
enabled effective generalization of the research findings to the larger population.
Sample Size Selection
I performed an a priori power analysis using G*Power 3.1.9.7, a standard method
to identify the minimum required sample size for BLR (Darling, 2021; Kang, 2021). I
have displayed the G*Power sample size interphase with the settings used and calculation
in Figure 1. The G*Power estimates statistical power using the Wald test and calculates
effect size directly from user inputs (Figure 1) of the two probabilities, p1 = (Pr(Y = 1|X
= 1) H0)) and the p2 = (Pr(Y = 1|X = 1) H1) or by calculating the odds ratio (OR) and
inputting the OR (Darling, 2021). The p2 is the probability of occurrence of the positive
factor; the group coded 1 in the binary outcome variable; the p1 is the probability of
occurrence of the negative factor; the group coded 0 in the binary outcome variable.
I chose the 'Z tests' as the test family and 'logistic regression' as the statistical test
in the power analysis because the dependent variable was binary (had only two responses
[yes = 1; no = 0]). I chose a power value of .80 in the analysis because power values
between .80 and .90 are acceptable in research studies (Giner-Sorolla et al., 2024; Kang,
2021). I assumed a moderate correlation of the three contingency predictor variables
64
(PMs' PP, PM's LMR, and PM's TS) with the primary predictor variable (PM's LB) and
used an R2 value of .09 in the power analysis (Figure 1).
Figure 1
Sample Size Determination Using G*Power Analysis
I further based the power analysis on a two-tailed distribution because
symmetrical distributions like the Z distributions (ex., binomial distribution) have two
tails (A. Ali et al., 2019; Costello et al., 2022) and chose binomial for the X distribution a
standard when the primary predictor variable is dichotomous (Darling, 2021; Riley et al.,
2019). I assumed a value of .24 for the p2 (Figure 1) because, on average, there is a 76%
failure rate of DT projects' completion by the PMs in U.S.-based LICs (Correani et al.,
65
2020; Datta & Nwankpa, 2021; Reeves et al., 2018). I assumed an alpha value for error
probability (α) as .05 for the power analysis to have a less than 5% chance that the data
tested occurring under the null hypothesis; the α value measures the degree of data
compatibility with the null hypothesis (Di Leo & Sardanelli, 2020). With the above
settings and information, the a priori power analysis with G*Power 3.1.9.7 yielded a
minimum sample size of 143. The actual sample size of this study was 214, which was
larger than the minimum sample size determined through power analysis. A larger
sample size lowers the likelihood of error in generalizing the findings to the target
population and a larger sample size than 100 increases the accuracy of the estimates in
BLR analysis (Lakens, 2022; Riley et al., 2019). Therefore, the sample size of 214 was
adequate and appropriate for this study.
Ethical Research
Ethical conduct in research promotes achieving the research goals, such as
knowledge of the truth, minimizes error by prohibiting fabricating, falsifying, or
misrepresenting research data, and promotes the values essential to collaboratively
conduct research, ensure trust, accountability, mutual respect, and fairness (Cumyn et al.,
2019; Head, 2020), and attracts funding and employment opportunities (Hutchings &
Michailova, 2022). I followed the appropriate ethical public, federal, university, and
other institutional guidelines for proper research conduct, including authorship
guidelines, copyright policies, patenting, data sharing, confidentiality rules, protection of
participant rights, and intellectual property rights during this research. I also ensured
appropriate
66
research conduct to not jeopardize the Walden school staff's and employees' safety in
business.
Informed Consent Process
Walden University's Institutional Review Board (IRB) approved this study's
research protocol. It provided an informed consent form (ICF), the ASCFDS, for use with
the participants during data collection (displayed in Appendix B). Walden University's
ethics IRB approval number for this study was 10-13-23-1017545. The ASCFDS form
included the ethics approval number, information on the study's purpose, implications to
business, the participants' roles and rights in the study, the requirement for participants to
give their voluntary consent to participate, the participant's right to withdraw from the
study at any time without hindrance and encumbrances, and the confidentiality of
participants' data and information collected. I provided participants with a link to the
ASCFDS during survey administration. I ensured they read it and agreed to its terms
before involving them in all data collection processes; participants who agreed to the
ASCFDS's terms clicked a continuing link voluntarily and completed the survey. Before
administering the survey to participants, I purchased a license from the publisher
(Appendix F) to use and reproduce the survey instruments.
Participants’ Withdrawal Process
The ASCFDS contained the necessary information on participants' right to
withdraw from the study at any time without penalty or loss of any benefits they may be
entitled to. I provided the following options for the participants who intended to withdraw
67
from the study: (a) stop participating in the data collection without any notice or (b) send
an email or text message informing me of their withdrawal. I also provided the necessary
instructions in the ASCFDS (Appendix B) on whom to contact if they decided to
discontinue participation or had any questions or concerns during or after completing the
study. I also informed the participants within the survey that during the participant's
voluntary withdrawal from the data collection process at any time before completion, the
data already collected before withdrawal may be retained and used by the study
investigators consistent with the study's purpose.
Ethical Protection of Participants
The data collection process was entirely voluntary, and others did not learn about
or influence the volunteers' participation in the study. I further protected participants'
rights and privacy by (a) taking good care of the data collected, (b) not using the data
collected for purposes outside the scope of the research study, (c) ensuring participants'
information was kept very private during data gathering, data storage, and data analysis
(Hutchings & Michailova, 2022), and (d) making sure the participants and their
organizations were unidentifiable directly or unintentionally through conducting
anonymous surveys. I carefully designed the data collection process (including
participant screening questions) to avoid recruiting vulnerable adults (elderly [>80
years]) and never even unknowingly recruited minors (<18 years old); as per the Belmont
Report, adult recruitment procedures must deliberately avoid recruiting minors.
68
I ensured that only the study's primary researcher (me) and the Walden
faculty/staff (ex. committee chair, second committee member) viewed the raw data. I
have stored all the collected data as follows: (a) paper documents, including any notes
and permissions letters sent via mail to me in locked file cabinets at my home; (b) all
electronic surveys and other electronic documents on a password-protected personal
computer backed up on a personal external password-protected hard drive, and a
password-protected cloud drive, and (c) the collected data remained stored for five years
and disposed of securely after 5-years. This study did not require organizational
masking and exceptions to organization-masking regulations and did not incentivize
participants because the data collection was anonymous.
Data Collection Instruments
In this study, I used existing survey questionnaires as the primary data collection
instrument to collect primary data for the independent variables. The survey instruments
used to collect data for this study were published and available in the book by Fiedler
(1967) and the book chapter by Ayman et al. (1998). I legally purchased copyright
permission (license #: 1412887-1, Appendix F) to reprint, use, and reproduce the scales
in this study from the publisher, Emerald Group Publishing Limited. I have included the
license's number (license #: 1412887) granted to me by the publisher in all scales
reproduced in Appendix D. I used or reproduced these survey scales in this study with
the copyright holder's permission, thereby eliminating infringing on the copyright. Raw
data
69
will be available upon request from the researcher. I have listed below the details of each
questionnaire.
Least Preferred Coworker Scale
I used the LPC survey questionnaire (Table D4) developed by Fred Edward
Fiedler (Fiedler, 1967) to collect data on participating PMs' LB (primary independent
variable). Several previous researchers, including Henkel et al. (2019), Kundu and
Mondal (2019), Shala et al. (2021), and Yammarino et al. (2020), have confirmed the
validity of the applicability of the LPC questionnaire to measure LB of PMs. On the LPC
scale, the leader rates the one person he or she has worked with in the past or is currently
working with whom he or she can work the least well. The LPC scale consists of 18
bipolar (positive/negative) adjective items (ex. pleasant-unpleasant, friendly-unfriendly)
with ratings from 1 to 8. The LPC score for a PM is the sum of the individual item scores
on the 18 bipolar adjective items with a high LPC score (≥ 73), indicating ROLB and
TOLB otherwise. Raters can complete the LPC scale in 5 minutes.
According to Fiedler (1967), the LPC score is an indication of the relative
strength of two discrete leadership orientations: (a) the leader is TO and (b) the leader is
RO. Leaders who score low on the LPC scale (< 73) are more TO than RO, find
gratification and self-esteem through task achievement, and feel confident and
comfortable when a task is highly structured. In contrast, the leaders who score high on
the LPC scale (≥ 73) are more strongly RO than TO, value how others regard them, and
find satisfaction, self-esteem, and confidence in maintaining good interpersonal
relations
70
(Fiedler, 1967; Fiedler & Chemers, 1984). Ayman et al. (1995) reported the LPC scale to
have high internal consistency (IC) reliability (Cronbach's Alpha = .88) and
comparatively high test-retest reliability (.67). LPC scale yields uniformly high split-half
reliability coefficients of about .90 (Fiedler, 1967, p. 44). Based on the above results of
psychometric properties, LPC can be considered a stable instrument. Further, the LPC
scale is reliable, effective in describing the LB of leaders, and identifies the hierarchy
(the psychological distance) the leader maintains between the leader and his subordinates
(Bartsch et al., 2021; Fiedler, 1967; Henkel et al., 2019; Kundu & Mondal, 2019;
Yammarino et al., 2020).
Leader’s Contingency Situation
Fiedler (1964) considered three contingency variables that determine the
situational favorability for the leader: (a) the leader’s leader-member relationship (LMR),
(b) the leader's position power (PP), and (c) the leader's task structure (TS), that together
with leader's LB determine whether TO or RO leadership is the best for the leader's
project environment (Table 2; Fiedler, 1964). I used the three contingency variables from
Fiedler's (1967) CTL as independent variables to describe the PMs’ situation in this
study. Fiedler (1967), in CTL, based on the three contingency variables, grouped the
leader’s situation into eight octants (Table E1) of favorableness (favorable to
unfavorable). According to Fiedler (1967), in real-life business and individual
organizations, in favorable and unfavorable situations, low LPC leadership (TO
leadership) is appropriate,
71
and in moderately favorable situations, high LPC leadership (RO leadership) is
appropriate (Table 2, Table E1).
Table 2
Contingency Model of Leadership Effectiveness
Leader-
member
relations Task structure
Leader’s
position
power
Favorability to
leader Most effective leader
Good Structured Strong Favorable TO
Good Structured Weak Favorable TO
Good Unstructured Strong Favorable TO
Good Unstructured Weak Moderately
favorable RO
Poor Structured Strong Moderately
favorable RO
Poor Structured Weak Moderately
favorable RO
Poor Unstructured Strong Moderately
favorable RO
Poor Unstructured Weak Unfavorable TO
Note. TO = task-oriented, RO = relationship-oriented. Adapted from “Contingency Model of Leadership
Effectiveness: Antecedent and Evidential Results,” by G. Graen, K. Alvares, J. B. Orris, and J. A. Martella,
1970, Psychological Bulletin, 74(4), 285–296 (https://doi.org/10.1037/h0029775).
Leader Member Relations (LMR) Scale
Fiedler used a self-report instrument, which the leader completed, called the LMR
scale, with eight LMR questions, to collect data for the LMR variable. Each item in the
LMR scale scored on a 1 to 5 Likert scale ('strongly agree,' 'agree,' neither agree nor
disagree,' disagree,' or 'strongly disagree'). The LMR scale is a powerful indicator of the
leader's LMR (Fiedler, 1967). The LMR variable indicates how well the leader gets along
with individual group members, which reflects the level of cohesiveness in the work team
and the degree of support the leader gets from his team members (Ayman et al., 1995;
72
Fiedler, 1964, 1967). Ayman et al. (1995) reported high IC reliability (Cronbach's Alpha
= .80) and high construct validity for the LMR scale. The highest score possible on the
scale is 40. A combined score of 20 or above indicates a good to moderately good
leader's LMR, and a score below 20 indicates a poor leader's LMR (Fiedler, 1967; Fiedler
& Chemers, 1984). In this study, I used the leader's rating of the LMR scale used by
Fiedler (1967) with eight questions to collect primary data for PM's LMR, a binary
independent variable (good PM's LMR vs. poor PM's LMR).
Task Structure (TS) Rating Scale
TS, defined as the clarity of tasks designed by the leader and perceived by the
project team members, is another contingency situational variable that Fiedler (1967)
used. According to Scheiter et al. (2020), employee performance is affected by PM's TS,
which moderates the employees' effort and task performance. Fiedler (1967) and Fiedler
and Chemers (1984) used a 'TS Rating Scale,' which includes ten questions and can
measure TS in the following four dimensions: (a) goal clarity, the extent to which
members clearly understand the task's requirements (Fiedler, 1967; Fiedler & Chemers,
1984; Shaw, 1963); (b) goal-path multiplicity, the extent to which a variety of different
procedures or paths exists for the team members to perform the task, (c) decision
verifiability, the extent to which the leader uses logic, mathematics, or feedback to
demonstrate the correct solution to team members (Fiedler, 1967; Laughlin & Ellis,
1986), and (d) solution specificity, the extent to which there is more than one correct
solution.
73
In the TS rating scale, the leader rates each of the ten questions on a three-point
scale scored as 'usually true = 2', 'sometimes true = 1', and 'seldom true = 0'. Scores on all
ten questions summed give the leader's TS score. The maximum possible score on the TS
scale is 20, and a combined score of 14 or above indicates the tasks are structured and
unstructured otherwise (Fiedler, 1967; Fiedler & Chemers, 1984). Ayman and Chemers
(1991) reported high IC reliability (Cronbach's Alpha = .81), and Fiedler (1967) reported
high interrater reliability (between .80 and .88) for the TS Rating Scale. In this
study, I used this 'TS Rating Scale' to collect primary data for the PM's TS, a binary
independent variable.
Position Power (PP) Scale
The third and final contingency situational variable Fiedler (1967) considered in
the CTL model was the leader's position power (PP), defined as the legitimate power
inherent in the leadership position, such as rights, duties, and obligations. PP measures
whether the leader has the authority to hire or fire, give rewards and punishments, and
approve raises in rank and compensation to followers, which is the right of the leader to
exercise power to persuade the followers to support the leader's efforts and establish good
relationships with subordinates while maintaining authority. Fiedler (1967) originally
developed the 'PP Scale,' with 13 questions. Fiedler and Chemers (1984) modified the 13-
question 'PP Scale' into five questions answered on a three-point scale. According to
Fiedler and Chemers (1984), the sum of the scores to the five questions is a highly
reliable scale for measuring a leader's PP. Fiedler (1967) reported a high interpreter
74
reliability of .95 for the PP Scale. The highest score possible on the PP scale is ten, and a
value of seven or above denotes a strong leader's PP (Fiedler, 1967; Fiedler & Chemers,
1984). In this study, I used the modified five-question PP Scale by Fiedler and Chemers
(1984) to collect primary data for the PM's PP, a binary independent variable.
Instrument Administration
I used the free online survey tool that Centiment provided to administer surveys
online. All the respondents accessed the online survey tool, which was the primary and
sole method of survey administration in this study. With online survey administration, the
return rate was higher than by post or email because it was easier and faster for the
respondents to complete the survey online. Online survey administration compels the
respondents to answer the questions, enhancing the return rate and clarifying if the survey
contains ambiguous questions (Einola & Alvesson, 2021).
Strategies Used to Assess Instrument Validity and Reliability
Survey instruments used to collect data in research studies must be valid and
reliable (Elangovan & Sundaravel, 2021; Story & Tait, 2019). Validity is how well the
survey instrument measures the intended attribute (Almanasreh et al., 2019; Knekta et
al., 2019). Reliability is the values obtained of the same attribute from the same
instrument in multiple experiments under the same conditions are the same (Elangovan
& Sundaravel, 2021). I performed a reliability and validity analyses of the four survey
instruments I used to collect primary data for the study’s independent variables. I have
presented and discussed the results of the reliability and validity analyses in section 3 of
this document.
75
Content Validity and Criterion Validity
A content validity analysis of items in the questionnaire ensures that the
questionnaire includes all essential items and eliminates undesirable items to a particular
construct (Almanasreh et al., 2019). A criterion validity analysis evaluates if the measure
agrees with a gold standard and to what extent. I performed item-to-total (ITT) score
correlation analyses and quantitative correlation analyses on the data collected from the
four survey scales to evaluate their content and criterion validity.
Construct Validity
Construct validity evaluates how well the constructs in a study translate into
functioning and operating measures. The construct validity has two components: (a)
convergent and (b) discriminant (divergent) validity. Discriminant validity allows the
evaluation of whether the individual indicators of the construct are significant (account
for acceptable variance in the observed variables). Convergent validity allows the
estimation of when two measures of constructs are related and when they should be
related. I evaluated the construct validity (discriminant and convergent validity) of the
four survey instruments by conducting factor analysis (FA) with the principal component
analysis (PCA) extraction method and the varimax rotation and verified the significance
of the factor loadings (if > .40) using the 'factor ' procedure in SPSS. The FA results
satisfy the construct validity criteria if the loading is at least .40 (Knekta et al., 2019). I
used the Kaiser-Meyer-Olkin (KMO) and Bartlett's sphericity tests to assess the data's
factorability, as Shrestha (2021) suggested.
76
Reliability
The reliability of existing survey questionnaires relates to consistency across the
parts of the measuring instrument (internal consistency [IC]) (Knekta et al., 2019;
Schrepp, 2020). The most used IC reliability measure to assess instruments is Cronbach’s
Alpha coefficient, and a minimum value of .70 for Cronbach’s Alpha is agreeable
(Schrepp, 2020). I evaluated the IC reliability of the four survey questionnaires I used to
collect primary data for the independent variables through Cronbach’s Alpha coefficient,
ensuring a minimum Cronbach’s Alpha value of .70.
Data Collection Technique
Online Survey
The survey is the best and the most common method to collect quantitative data,
among several primary data collection methods often used by psychologists and
sociologists to analyze leadership behaviors (Knekta et al., 2019; Story & Tait, 2019).
With the wide availability of computer systems and internet connectivity, online surveys
such as computer-administered surveys, electronic mail surveys, and web surveys allow a
global reach and much easier, faster, and more flexible administration (M. -J. Wu et al.,
2022). To address the research questions of this study, I collected quantitative numerical
data for the dependent and independent variables using questionnaires administered
through an online survey. I prepared the survey questionnaires using the free online tool
Centiment provided, included the link to the ICF I obtained from IRB, and then passed it
on to Centiment for online administration. Centiment successfully administered the
77
survey questionnaires to the eligible participants online. Eligible participants who read
and agreed to the terms of the ICF could click on a continuing link provided along with
the ICF, access the survey, and complete it.
In this study, online survey services enabled faster data collection, convenience,
ease of data entry and analysis, large samples easier to manage, and prevented the
inefficiency and expense of transforming the paper data into an electronic form for
processing and analysis, allowed question format diversity and ease of providing
instructions and reminders through alerts to enhance response rates resulting in sufficient
data. Survey research errors such as an unrepresentative sample, a low response rate
(<30%), and nonresponder bias could reduce the validity and reliability of the data
collected through online surveys (M. -J. Wu et al., 2022). I avoided or minimized these
errors by (a) providing adequate information and clarity to participants on survey
questions, (b) evaluating the validity and reliability of survey questionnaires and data
collected through appropriate statistical tests, and (c) using adequate sample size
identified through an a priori power analysis which improved the response rate and
minimized nonresponder bias.
Data Analysis
The research questions addressed in this study were: (1) what is the relationship
between PM’s leadership behaviors and DT project completion status (RQ1)? and (2)
what homogenous clusters of PM’s leadership behaviors emerge based on DT project
completion status (RQ2)? The RQ1 involved a dependent variable, one primary
78
independent variable (PM’s LB), and three contingency independent variables: (a) PM’s
PP, (b) PM’s LMR, and (c) PM’s TS. I used the BLR statistical method to test the
hypotheses and answer RQ1. I used a two-step cluster analysis with pre-clustering
performed with the hierarchical cluster (HC) analysis followed by the k-means method to
answer RQ2.
Logistic Regression
Logistic regression (LR), which describes the relationship between a predictor
variable Xi (or a series of predictor variables) and the conditional probability that an
outcome variable Yi equals one (success event), is appropriate when the research
involves a categorical dependent variable given one or more independent variables (Dao
et al., 2022; Sommet & Morselli, 2017; Zou et al., 2019). The equation of the predictor
variable in the LR model, β0 + βi * Xi, is the same as in the linear regression; however, in
the LR model, the exponent of the equation of the predictor variable, exp(β0 + β1 * Xi) is
applied to obtain an odds ratio (OR). The OR is the factor by which the probability of an
event occurring rather than not occurring ((P(Yi = 1))/(1 – P(Yi = 1))) changes (increases
or decreases depending on the sign of βi) when the predictor variable Xi increases by one
unit (Šinkovec et al., 2019; Sommet & Morselli, 2017). The sign of βi can be positive or
negative. When OR is not significantly different from 1, the odds of an event occurring
remain the same as Xi changes. This case indicates that the null hypothesis (H0) is true
and must be accepted; a value of OR significantly different from 1 and greater than 1
indicates a significant positive effect (Šinkovec et al., 2019; Sommet & Morselli, 2017).
79
The higher the predictor variable effect value, the higher the odds of an event occurring.
Suppose OR is < 1 and significantly different from 1, the lower the odds of the event
occurring (a negative effect); in these two situations, the H0 must be rejected (Šinkovec et
al., 2019; Sommet & Morselli, 2017).
Binary Logistic Regression
There are three types of LR: (a) binary LR (BLR), (b) multinomial LR (MLR),
and (c) ordinal LR (OLR). In BLR, the dependent variable can take only two possible
categories (ex., yes/no, male/female, true/false); in MLR, the response of the dependent
variable has three or more categories without any ordering within the categories (ex., four
different presidential candidates); in OLR, the dependent variable has three or more
categories with an ordering among the categories (ex., rating variables,
bad/good/excellent) (Dao et al., 2022; Harris, 2021). The BLR allows modeling the
relationship between a dichotomous dependent variable and multiple independent
variables, either continuous or categorical (Dao et al., 2022; Harris, 2021). The
dependent variable of this study, the DT project completion status achieved by the PMs
in their companies, could take only two possible values (yes/no), thus categorical and
binary.
Therefore, BLR was the appropriate analysis method for this research. Linear regression
methods such as multiple linear regression and analysis of variance (ANOVA) require
that the dependent variable be continuous; hence, it is not appropriate to model this
study’s data with a dichotomous dependent variable.
80
Cluster Analysis
Cluster analysis (CA), a technique to cluster similar observations based on the
observed values of several variables, can create groups of observations in data, like
customers, products, employees, projects, and the like, based on similar properties
(Karim et al., 2021). When the outcome variable is dichotomous, variation and
heterogeneity in outcomes leading to clusters occur (Austin & Leckie, 2020). Clustering
methods include HC, centroid-based (e.g., k-means and k-median clustering),
distribution-based, density-based, and self-organizing maps (Karim et al., 2021). HC and
centroid-based (k-means and k-median) methods are the simplest yet most effective
ways of creating clusters, frequently in research studies, and are the commonly used
approaches for clustering dichotomous data (Adolfsson et al., 2019; M. Ahmed et al.,
2020; Torrente & Romo, 2021). The HC method in SPSS allows the calculation of
distances and linking clusters by the calculated distances to accurately identify the
number of clusters; therefore, it helps identify how many clusters the data has (Galak,
2020a). K-means partitioning, the most prominent clustering approach in scientific
research has proved effective for partitioning dichotomous data (Galak, 2020a). The k-
means cluster analysis method allows effective data partitioning by the number of
clusters identified from the HC method (Galak, 2020b; Torrente & Romo, 2021).
In CA, after performing a cluster analysis using the k-means or HC method, a
Silhouette analysis used to obtain a mean Silhouette score helps assess whether the
number of clusters chosen was correct, whether the clusters formed were valid, and if the
81
clustering solution was suitable. The Silhouette analysis in SPSS uses the cluster number
assigned for each observation by HC or k-means clustering. It calculates a Silhouette
score for each observation, which varies between -1 and 1, with higher values > .50,
indicating that the observation fits its cluster well and adequately differs from members
in other clusters (Pedersen et al., 2023). The total mean Silhouette score of all
observations provides valuable insights into cluster cohesion and separation of the data,
allowing to validate data groupings; mean Silhouette score > .50 confirms the correct
number of clusters chosen and the data points were grouped meaningfully (Shutaywi &
Kachouie, 2021).
I initially clustered the data using the hierarchical clustering (HC) technique,
Ward’s cluster separation method, and the squared-Euclidean distance measure with all
variables included in SPSS software. The HC method in SPSS allows the calculation of
distances and linking clusters by the calculated distances for accurately identifying the
number of clusters; therefore, it helps identify how many clusters the data has (Galak,
2020a). Ward’s method, a popular clustering partitioning method for binary and
continuous data, allows calculating the distance of all clusters to the grand average of the
clusters, creating evenly sized clusters based on the distances and predicting the
significance of differences between the clusters using the F statistic value (Galak, 2020a;
Govender & Sivakumar, 2020). The squared Euclidean distance measure is a popular
dissimilarity distance measure for binary data, computed as the number of discordant
82
cases with a minimum value of 0 and no upper limit, and quickly processed using
statistical software like SPSS (Albuquerque et al., 2022).
Next, I used the k-means clustering method and partitioned the data into the
number of clusters already identified by the HC method. This two-step approach to
cluster analysis allowed me to identify significant clusters on dependent and independent
variable groups with binary variables (Galak, 2020b; Karim et al., 2021). I then
performed a Silhouette analysis using the cluster number for cases obtained from the k-
means method using the absolute difference dissimilarity measure. I used the Silhouette
score obtained from Silhouette analysis for each observation and calculated the mean
overall Silhouette score.
Bootstrapping
Bootstrapping is a statistical procedure to model uncertainty and variability
during statistical estimation (Austin & Leckie, 2020). According to the American
Statistical Association (ASA), conclusions from research studies and anything scientific
or practical importance must not solely rely on the statistical significance of associations.
Using the bootstrap method in research studies enables the estimation of the sampling
variability of measures of variance and heterogeneity, which allows for precise
assessment of statistics in the analysis (Austin & Leckie, 2020). I used bootstrapping to
estimate the sampling variability in the BLR models, keeping with ASA’s
recommendations. As stated by Austin and Leckie (2020), the bootstrapping increased
the sample size and permitted the construction of confidence intervals around BLR and
83
cluster-specific predicted random effects, which provided a richer interpretation of the
data than a simple reliance on statistical significance testing, allowing valid conclusions
from the statistical analysis.
Data Cleanup Strategies
Study results can be unreliable or misleading when data collection and analysis
include incorrect or poor-quality data and cleaning the data before any statistical analysis
can improve data quality and results. Data quality is usually considered in the following
four dimensions: (a) accuracy, (b) completeness, (c) consistency, and (d) timeliness (Xu
et al., 2020). Data is accurate when the collected data is appropriate for the phenomenon
studied, and data is complete when it covers the entire study objectives. Consistency is
when the data is in the correct format and structure. In this study, I cleaned the data for
outliers, missing values, and inaccuracy (when data did not meet the statistical
assumptions) using SPSS. Outliers in data indicate incorrect data. Missing values in the
data collected break the continuity of data. Missing data is a severe issue in quantitative
research with questionnaires because most statistical analyses assume no missing data
and need complete observations in the calculations (Xu et al., 2020). Inaccurate data
makes it challenging to apply valid statistical analysis methods and make intelligent
conclusions or extract valuable information (Xu et al., 2020).
Handling Missing Values
Handling missing data in regression models includes listwise and pairwise
deletion techniques. The listwise deletion technique, also known as the complete-case
84
analysis (CCA), removes all data for the case or observation with one or more missing
values, a commonly used technique when conducting empirical studies (Osman et al.,
2018; Stavseth et al., 2019). However, the listwise and pairwise deletion techniques
require an essential assumption that data are missing completely at random (MCAR); in
other words, CCA relies on the assumption that the probability of data missing in the
dependent variable is unrelated to the independent and dependent variables (Osman et al.,
2018).
CCA may result in reduced power, significant bias, and too wide confidence
intervals because of reduced sample size; however, CCA is still the most used approach
to handling missing data (Osman et al., 2018). Most statistical analyses and software
packages, including SPSS, assume that all variables in the model are measured and, by
default procedure, usually delete cases with missing data on the variables of interest,
which is CCA (Osman et al., 2018). When using a data set with missing observations,
the major disadvantage is that the software package will remove a large proportion of the
sample, leading to a severe loss of statistical observations and power (Osman et al.,
2018).
According to Fiedler (1967), the CTL framework requires nonmissing data for all
variables to adequately describe the PM's LB based on the projects' situations the PMs
manage to make valid conclusions. I obtained 214 complete survey responses without
missing data via Centiment. PM's LB was contingent on the three situational independent
variables: (a) PM's PP, (b) PM's LMR, and (c) PM's TS. Therefore, I carefully
85
designed the data collection process to prevent and minimize missing data by (a)
providing adequate information and clarity to participants on survey questions and (b)
using an adequate sample size to increase the response rate and minimize nonresponder
bias, which resulted in 214 complete responses.
Assumptions of Statistical Analysis
The inferences drawn from statistical test results are valid only when the data
meets assumptions associated with the statistical tests (Knief & Forstmeier, 2021; Nyitrai
& Virág, 2019; Turner & Deng, 2020). Violations of statistical assumptions can lead to
the inaccurate probability of the test statistic, distorting Type I and Type II error rates
(Knief & Forstmeier, 2021; Turner & Deng, 2020). In this study, I tested the data for the
assumptions associated with BLR statistical analysis and the associated factor analysis. I
identified and used appropriate statistical techniques to check for assumptions and
identified and applied ways to find remedies when data did not meet the necessary
assumptions. I used SPSS software to test and apply resolutions for the violation of
assumptions.
Assumptions of BLR
An excellent fit of the BLR model to the sample data occurs when the difference
between the model-predicted and observed values is statistically insignificant (Nyitrai &
Virág, 2019). The various goodness-of-fit diagnostic statistical measures available to
estimate a BLR model's adequacy must meet a few assumptions (Boateng & Abaye,
2019). BLR analysis must meet the following assumptions: (a) the dependent variable in
86
BLR must be dichotomous; (b) independence of observations, which means the
observations must be independent and not from repeated measurements or matched data;
(c) linearity in the logit for continuous independent variables, (d) absence of or little
multicollinearity among the independent variables, which means little or no correlation
among the independent variables, and (d) lack of powerful outliers, influential
observations, or high leverage points (Nyitrai & Virág, 2019). Additionally, the number
of events or observations per independent variable must be adequate to avoid model
overfitting. I tested for the above assumptions of BLR using standard tests in SPSS. I
have explained in sections below the above assumptions and the processes I used to test
data for the assumptions. I have discussed the test results of the BLR assumptions
analyses in section 3.
Outliers, Influential Observations, and High Leverage Points. Outliers,
influential observations, and high leverage points in data may result from human errors,
instrument errors, sampling errors, incorrect or corrupted data, or usage of missing
values coded as actual data (Costa e Silva et al., 2020; H. Wang et al., 2019). Outliers are
unusual observations with exceptionally large outcome values that generate large
residuals (Costa e Silva et al., 2020). High leverage points are observations with extreme
predictor values for one or more predictors and influential observations are unusual
observations that unduly influence one or more areas of the regression analysis,
including the predicted responses, the estimated coefficients, and the hypothesis test
results (Costa e Silva et al., 2020). Outliers, high leverage points, and influential
observations can lead
87
to inaccurate parameter estimates and incorrect fit of the BLR model (Costa e Silva et al.,
2020).
Outlier Detection and Mitigation. The sigmoid function tapers with outliers in
the BLR models. Presence of extreme outliers affect the covariate pattern, resulting in
misleading interpretations, and lowering the performance of the BLR model; therefore,
detecting outliers in BLR modeling and taking appropriate mitigation measures to obtain
a good fit is necessary. Residual measures are the generally used techniques to identify
outliers in BLR models. Several residuals calculated from a fitted BLR model, including
Pearson residuals and studentized or normalized Pearson residuals, and deviance
residuals allow outlier detection (Hickey et al., 2019; Nyitrai & Virág, 2019; Sarkar et
al., 2011).
Plotting BLR model residuals against the predicted probabilities that displays a
linear trend, are accurate and reliable ways of detecting outliers (Sarkar et al., 2011).
Generally, absolute values of standardized or deviance residuals for good observations in
BLR are within ±2. Standardized, normalized, or deviance residuals outside the range of
± 2 or ≥ 3 in extreme cases are potential outliers and require closer attention, and may
require exclusion from the analysis (Costa e Silva et al., 2020; Hickey et al., 2019; Sarkar
et al., 2011). Deleting outlying cases with the most significant residuals almost always
improves the fit of the BLR model (Hickey et al., 2019).
Influential Observations Detection and Mitigation. Diagnostic plots, such as
influence plots (plotting the derived diagnostic statistics against the estimated logistic
88
probability and observed cases), can reveal the presence of influential observations
(Hickey et al., 2019). For a BLR model, Cook’s distance (Cook’s D) can measure the
influence of each observation on the regression parameter estimates. Cook’s D of the
fitted model greater than or equal to .50 indicates the observation is influential (Costa e
Silva et al., 2020).
High Leverage Points Detection and Mitigation. High leverage points are
observations with extreme predictor values. In any BLR model, the leverage values vary
between 0 and 1 inclusive. An observation is a high leverage point if its leverage value is
larger than 2 times the mean leverage value (MLV) or, in extreme cases, larger than 3
times the MLV (Costa e Silva et al., 2020). Plotting the BLR model predicted leverage
values against the estimated logistic probabilities or observed cases can reveal the
presence of high leverage points.
Multicollinearity. Multicollinearity in the BLR model indicates the existence of
highly correlated independent variables. These variables can reduce the accuracy of the
BLR model fit when included together in a single model as separate independent
predictors, especially when they share 49% or more variance (Harris, 2021; Ozgur &
Franklin, 2021; Zeng & Zeng, 2021). Predictors with multicollinearity cause unstable
estimates and inflate variances of the parameter estimates, leading to incorrect
predictions of the relationships between independent and dependent variables.
Multicollinearity affects confidence intervals and hypothesis tests and causes type II
errors (Ozgur & Franklin, 2021).
89
Multicollinearity Detection. The Spearman's rho correlation matrix obtained with
the pairwise correlation coefficients for the binary independent variables from the SPSS
output can allow the identification of the presence of multicollinearity in binary
independent variables. Alternately, Phi and Cramer's V coefficients measure the strength
of an association between two categorical variables; Phi and Cramer's V accurately
predict the correlation between categorical independent variables (Akoglu, 2018). Phi
and Cramer's V coefficients measure the strength of association between two nominal
variables, are nonparametric tests that utilize a contingency table (also known as a cross-
tabulation, crosstab, or two-way table) in which data classification is according to two
categorical variables (Akoglu, 2018). Each categorical variable must have two or more
categories. The categories for one variable appear in the rows, and the categories for the
other variable appear in columns. Each cell reflects the total count of cases for a specific
pair of categories. Statistical software like SPSS allows the calculation of Cramer's V and
Phi. Cramer's V values vary between 0 and 1 without any negative values, and a value
close to 0 means no association, and a value bigger than .25 of Cramer's V may indicate a
very strong association (Akoglu, 2018). Phi values vary between -1 and 1, and a value
close to 0 means no association, and a value bigger than .25 or smaller than -.25 of Phi
may indicate a very strong association (Akoglu, 2018).
Examining the correlation matrix may help detect multicollinearity, but it is
insufficient; it is possible to have data in which no pair of variables has a high
correlation, but several variables may be highly interdependent (Harris, 2021). The
90
tolerance and variance inflation factor (VIF) allows a better multicollinearity diagnosis.
The tolerance of any specific independent variable is 1 - R2, where R2 is the coefficient of
determination for the regression of the independent variable involved in all remaining
independent variables. In linear regression models, tolerance values <.10 and <.20 in
extreme cases often indicate multicollinearity, but in nonlinear models such as BLR,
variables with tolerance values <.40 may be problematic (Harris, 2021). The VIF is the
reciprocal of tolerance, estimated as 1/ (1 - R2) (Harris, 2021). The VIF indicates how
much the multicollinearity inflates the variance of the coefficient estimate; in linear
regression models, VIF values exceeding ten (10) indicate multicollinearity, but in
nonlinear models such as BLR, variables with VIF values above 2.50 may be problematic
(Harris, 2021).
Multicollinearity Mitigation. The Spearman's rho correlation matrix, Cramer's V,
and Phi coefficient scores obtained for the independent variables indicated the existence
of multicollinearity in this study's data. I have discussed the results in section 3. Because
multicollinearity existed, I performed dimension reduction through exploratory FA with
the PCA method and varimax rotation. I used the factors to fit the BLR model with
bootstrapping to solve for multicollinearity. The uncorrelated factors created through FA
with PCA minimized information loss and improved the BLR model's predictability. I
also obtained the factors' correlation matrix to confirm the multicollinearity resolution in
the data. I resampled the data through bootstrapping, performed the hypotheses testing
and statistical analysis, and estimated confidence intervals on bootstrapped samples to
91
eliminate the influence of multicollinearity and violations of other possible assumptions,
and improved inferences about the study population.
Independent Observations. The BLR model requires observations that are
independent of each other and that they do not come from repeated measurements or
matched data. In this study, the 214 participants who completed the survey to obtain
primary data for the study's variables were unrelated, and the observations were
independent and not from repeated measurements. I used bias-corrected estimates of the
parameters for the BLR model to account for any unknown dependence of observations
that arose from the surveys with potential stratified and clustered designs.
Large Sample Size. BLR requires a relatively large sample size. The
recommended minimum value ranges from 10 to 20 events per covariate (Nyitrai &
Virág, 2019). I determined this study's minimum sample size requirement through an a
priori power analysis using G*Power 3.1.9.7 to achieve a power level of .80, a
recommended standard (Giner-Sorolla et al., 2024; Kang, 2021; Riley et al., 2019). An
experiment with a power level of .80 has an 80% chance of predicting the effect present
at the population level. The sample size used in this study (214) was larger than the
minimum required sample size of 143 obtained from the power analysis for BLR analysis
and was large and adequate.
Assumptions of Factor Analysis
When performing FA, unlike the maximum likelihood extraction method, the
PCA extraction method does not assume the sample is from a multivariate normal
92
distribution but focuses on maximizing the variance of the components. However, FA
with PCA assumes the following: (a) multiple variables on a numeric scale, preferably
continuous; (b) sampling adequacy; PCA extraction method depends on large enough
sample sizes to produce reliable results. I used the Kaiser-Meyer-Olkin (KMO) measure
of sampling adequacy in SPSS in this study, a generally used procedure to evaluate data
for this assumption (Schreiber, 2021; Shrestha, 2021); (c) suitability of data for reduction,
PCA requires significant or adequate correlations between the variables for reduction into
a smaller number of components. I used Bartlett's test of sphericity in SPSS, a generally
used procedure to evaluate this assumption (Schreiber, 2021; Shrestha, 2021); and (d) no
significant outliers; I used residual measures including Pearson residuals and studentized
or normalized Pearson residuals, and deviance residuals to identify outliers in the data, a
recommended procedure (Nyitrai & Virág, 2019).
I encoded the survey scores of the independent variables as numeric binary
variables (0s and 1s) and standardized them into scale variables through z-score
normalization before performing the FA with PCA. In the normalization step, I computed
the mean values of the variables subtracted from the original variable and divided by the
standard deviation for each variable so that all variables used the same scale and
contributed equally to the analysis. This procedure is recommended when performing FA
with PCA extraction with binary variables (Greenacre et al., 2022; IBM, 2023; R. M. Wu
et al., 2023). I have displayed the descriptive statistics of the z-normalized independent
variables in Table E2 (Appendix E).
93
PCA is a linear technique better suited for variables with linear relationships
(Schreiber, 2021). PLC coefficient is a statistical measure that indicates the degree to
which variables change their values about each other. PLC is commonly the most used
linear correlation coefficient, expressing the level to which two variables are linearly
related (Šverko et al., 2022). I validated the linearity among the z-normalized
independent variables through correlation analysis; significant bivariate Pearson linear
correlations (PLC) of independent variables indicated linearity as stated by Šverko et al.
(2022).
Assumptions of Cluster Analysis
There are no assumptions associated with cluster analysis (CA). However, CA
requires significant care when choosing the group of variables or the construct of interest
that can group similar observations into clusters. Besides, CA requires selecting an
appropriate clustering method, testing its robustness before application, and performing
sensitivity analyses using various cluster solutions and different sets of clustering
variables with the chosen methods to determine their appropriateness (Karim et al.,
2021). Also, confirming the validity of CA results by theory and cluster descriptions is
helpful (Karim et al., 2021).
Study Validity
Statistical Conclusion Validity
Threats to internal validity do not apply to the nonexperimental
correlation research design, which was the research design of this study.
94
However, threats to the validity of the statistical conclusion related to statistical
power are of concern. Statistical power equal to 1- β of a hypothesis test is the
probability of correctly predicting a sample effect in the population (Kang, 2021).
The statistical power is inversely related to a Type II error. A test with high
statistical power has a more significant probability of correctly rejecting a false
null hypothesis (Kang, 2021).
In statistical analysis, the statistical power of the test (1- β) depends on
three factors: (a) the significance alpha level, (b) the study sample size, and (c)
the effect size (Darling, 2021; Kang, 2021). The more significant the significance
alpha level, the larger the sample size, and the greater the effect size, the greater
the power of the test (Darling, 2021; Kang, 2021). I warranted a balance between
the statistical validity of the research and the practical resource constraints by
selecting the right sample size to make valid conclusions. I performed an a priori
power analysis with G*Power 3.1.9.7 software, input appropriate values for the
significance alpha level (.05, a standard) and the effect size (Figure 1), and
calculated the required sample size with a power of .80 (an accepted standard);
the power analysis resulted in a minimum sample size of 143. By selecting a
larger than the minimum required sample size of 214, I ensured a larger actual
power (.92) in this study, which enabled the assurance of accurate inference. I
minimized bias though ensuring the reliability of the data collected from the
survey questionnaires through Cronbach’s Alpha coefficient, ensuring a
minimum
95
Cronbach’s Alpha value of .70 for all independent variables; reliable data are
relatively independent of the skills, moods, and honesty of the person collecting it
hence are less biased, leading to accurate inference (Duckett, 2021; Moon, 2019).
External Validity (Generalizability)
Large random samples in the hypotheses test increase the statistical power and
more closely approximate the population (Darling, 2021; Kang, 2021). I selected a larger
than required sample size (214), ensuring the study’s generalizability. I used
bootstrapping to increase the sample size further, which provided a richer interpretation
of the data, allowing valid conclusions from the statistical analysis and leading to a
superior generalizability of the findings. I also ensured the validity and reliability of the
instruments used to collect the study’s data. I also assessed and mitigated the study’s
data for all violations of statistical assumptions through appropriate statistical tests and
methods using SPSS software. Given the validity and reliability of the instruments and
accurate statistical analysis, the inferential results based on the data were valid and
reliable, leading to better generalizability.
Objective Validity
Objective validity evaluates using no subjective judgment by the person
collecting the data when recording or interpreting data. Objectivity is of concern if the
survey questionnaires measure attitudes, emotional characteristics, and the like. I
collected primary data using survey questionnaires for PM's LB, PM's LMR, PM's PP,
and PM's TS, which did not measure the attitudes or emotional characteristics but
measured PMs’
96
LB. I ensured the objective validity by collecting adequate samples, applying reliable
data collection procedures, using appropriate statistical analysis methods, and testing and
mitigating data for statistical assumptions. I also performed a validity analysis in SPSS
for the data collected using the instruments by obtaining item-to-total score correlations,
the correlation between each item's score and the total score from all items on the survey
scale, which is a reliable assessment method for the validity of survey scales (Rossell et
al., 2019).
Transition and Summary
In section 2 of this research document, I described the research method and
design, the population, sampling, participants, data collection instruments, data
collection techniques, and data analysis methods used for the study's reliability and
validity analysis of the instruments and inferential data analysis. In section 2, I also
discussed my role as a researcher and ethical research conduct. In section 3, I presented
the results of the reliability and validity analysis of the instruments used to collect
primary data for the independent variables, the research findings from the inferential data
analysis, and an evaluation of the statistical assumptions used in the analysis with
appropriate tables, figures, and illustrations. In section 3, I also presented a detailed
description of the applicability of the findings to professional practices, the tangible
social change implications, recommendations for action, reflection, and suggestions for
further research. Finally, I presented the study's conclusion.
97
Section 3: Application to Professional Practice and Implications for Change
Introduction
This quantitative correlational study uncovered the intricate relationship between
PMs’ LB and DT project completion status in three project situations of the PM. I
examined four independent variables: (a) PM's LB, (b) PM's LMR, (c) PM's TS, and (d)
PM's PP. The dependent variable I focused on was DT project completion status. The
good psychometric properties of the instruments used in this study led to valid and
reliable data collection and inferential results. Factor analysis with adequate of KMO test
measure (> .50), highly significant Bartlett's sphericity test measure (χ2 [6, N = 214] =
39.14, p < .001), significant factor loadings (> .40) for all the independent variables,
significant communalities (> .50) of all variables in the factors, highly nonsignificant
Pearson correlation coefficient (.000, p = 1.00) between the factors, anti-image
correlation coefficients ≥ .57 for all variables in the factors allowed the extraction of two
valuable factors from the independent variables. BLR analysis with the extracted factors
indicated a significant (p < .05) relationship between PM's LB and DT project
completion status in favorable and unfavorable situations, leading to accepting the
alternate hypothesis for Research Question 1 with 95% confidence. In favorable and
unfavorable contingency situations, the PMs with ROLB could not complete their DT
projects successfully, creating a need for PMs with TOLB. Cluster analysis with
significant contribution of all variables (p < .05) indicated the existence of three distinct
clusters in the data, leading to accepting the alternate hypothesis for Research Question 2
with 95%
98
confidence. The results supported and confirmed the application of CTL for DT projects
in LICs in the United States.
Presentation of the Findings
In this section, I present descriptive statistics of data, results of missing value
analysis, analysis of violations of statistical assumptions, the reliability and validity
analysis of the survey scales used to collect primary data for the independent variables,
and the results of the inferential statistical tests of BLR and cluster analyses. I have
discussed how the analyses I performed addressed the hypotheses of this study and
answered the research questions and the theoretical perspectives of the findings.
Appropriate APA tables and figures accompany the analysis results. The research
questions addressed in this study were as follows:
RQ1. What is the relationship between PM's leadership behaviors and DT
project completion status?
RQ2. What homogenous clusters of PM's leadership behaviors emerge based
on DT project completion status?
I performed a BLR analysis procedure to address RQ1 and cluster analysis to address
RQ2 in SPSS software (version 28).
BLR is the correct statistical analysis method when predicting a binary outcome
(dependent) variable from one or more independent (predicting) variables (Boateng &
Abaye, 2019; Dao et al., 2022; Harris, 2021). A preliminary Kolmogorov-Smirnov
normality test of data indicated that the independent variables belonged to nonnormal
99
distributions (Table E3 [Appendix E]), which further confirmed the appropriateness of
BLR analysis of the data. I also performed an exploratory FA using the PCA method and
varimax rotation to correct for significant associations among independent variables and
ensure the independence of observations. I used the factors to fit the BLR model and
enhanced its accuracy. The answer to RQ2 included a cluster analysis (CA, also known
as clustering) to investigate the underlying structure of the data, described as the
grouping of objects that shared similar characteristics. I performed a two-step cluster
analysis in SPSS: (a) initially identified the number of clusters by the HC method, and (b)
applied the k-means clustering method and partitioned the data into three clusters already
identified by the HC method.
I used bootstrapping (≥ 1,000 samples) to increase the sample size, calculating
bootstrapped 95% confidence intervals (CI) for all estimates, including descriptive
statistics, correlations, model coefficients, and estimates necessary to solve potential
assumption violations and to improve the BLR model fit. Bootstrapping allows
resampling the data and analyzing the resampled data, improving BLR model fit and
making more accurate inferences about a study's population (Noma et al., 2021;
Ogundunmade & Adepoju, 2019). Through bootstrapping, I increased the sample size,
made valid conclusions from the statistical analysis, and enhanced the generalizability of
the findings.
100
Descriptive Statistics
In this subsection, I present descriptive statistics for the study's categorical
dependent (project completion status) and categorical independent (PM's LB, PM's PP,
PM's TS, and PM's LMR) variables for the study's data. I collected 214 complete
responses from participants through anonymous surveys. A missing value analysis of the
data in SPSS indicated no missing values in any of the variables for any observations
(Table 3).
Table 3
Missing Value Analysis Results
Missing
Variable NCount Percent
Project completion status 214 0 0
PM's LB 214 0 0
PM's LMR 214 0 0
PM's PP 214 0 0
PM's TS 214 0 0
Note: PM = project manager, LB = leadership behavior, LMR = leader member relations, PP = position
power, TS = task structure.
I have presented the frequencies and percentages of the study's dependent and
independent variables with a bootstrapped (1,000 samples) 95% confidence interval (CI)
for the percentages in Table 4. Results indicated that 56.1% of the 214 participants
(hereafter PMs) completed the DT projects successfully, and 43.9% did not (Table 4).
Most of the PMs (91.1%) displayed high LPC (≥ 73), indicating ROLB, while 8.9%
displayed low LPC (< 73), indicating TOLB (Table 4). Ninety percent of PMs who
completed their DT projects and 93% who did not complete their DT projects
101
successfully displayed ROLB. Ten percent of PMs who completed and 7% who did not
complete their DT projects successfully displayed TOLB.
Table 4
Frequencies and Percentages of Variables
Variable Type Category Frequency Percent Bootstrapped 95%
percentaCI of
LL UL
Project Dependent Yes 120 56.1 49.5 62.6
completion No 94 43.9 37.4 50.5
status Total 214 100.0
PM's LB Major RO 195 91.1 86.9 94.9
independent TO 19 8.9 5.1 13.1
Total 214 100.0
PM's LMR Independent Good 51 23.8 18.2 29.9
Poor 163 76.2 70.1 81.8
Total 214 100.0
PM's PP Independent Strong 66 30.8 24.8 37.4
Weak 148 69.2 62.6 75.2
Total 214 100.0
PM's TS Independent Structured 140 65.4 58.9 71.5
Unstructured 74 34.6 28.5 41.1
Total 214 100.0
Note: PM = project manager, LB = leadership behavior, LMR = leader–member relations, PP = position
power, TS= task structure, LPC = least preferred coworker, TO = task-oriented (LPC score < 73), RO =
relationship-oriented (LPC score ≥ 73), CI = confidence interval, LL = lower level, UL = upper level.
Sample size (N) =214.
a Bootstrap results are based on 1,000 bootstrap samples.
Correlation Between PM’s Leadership Behavior and Digital Transformation
Project Completion Status
I divided this study's data into eight octants, grouping them into the three
categories of situational favorableness for the leader: (a) favorable (octants I, II, and III),
(b) moderately favorable (octants IV, V, and VI), and (c) unfavorable (octants VII and
102
VIII), as indicated in the CTL model (Fiedler, 1967; Fiedler & Chemers, 1984). I
obtained descriptive statistics of PM's LB and the dependent variable (DT project
completion status) by the three situational favorableness categories. I also obtained
nonparametric Spearman's rho correlations between PM's LB and the dependent variable
in each octant. I have presented the results in the next two subsections.
Descriptive Statistics by Situational Favorability
Of the 214 PMs, 77.10% experienced favorable situational control, 12.62%
experienced moderately favorable situational control, and 10.28% expressed unfavorable
situational control. Most PMs under favorable situations displayed ROLB (94.50%),
while 77.80% under moderately favorable situations and 81.80% under unfavorable
situations displayed ROLB. PMs with TOLB were 5.50% under favorable situations,
22.20% under moderately favorable situations, and 18.20% under unfavorable situations.
Also, 57.60% of PMs under the favorable situation, 50.00% under the moderately
favorable situation, and 51.90% under the unfavorable situation completed their DT
projects successfully.
Correlations of PM's LB and DT Project Completion Status by Octants
I have displayed the Spearman's rho correlations between the dependent variable
and PM's LB in Figure 2, with Fiedler's standard CTL curve superimposed for
comparison. Results indicated that the relationship between PM's LB and DT project
completion status was nonrandom in this study, as shown by Fiedler (1964) in the CTL
model and other investigators, including Ayman et al. (1995), Chemers and Skrzypek
103
(1971, 1972), and Fiedler and Chemers (1984) in their studies. Negative correlation
existed between the successful DT project completion status and the PM's LB under
favorable situation: (a) octant I (rho = -.13, [-.24, .03]) and (b) octant II (rho = -.47, [-
1.00, -.26]) and under unfavorable situation: (a) octant VII (rho = -.02, [-.47, .41]) and
(b) octant VIII (rho = -.67, [-1.00, -.25]). The successful DT project completion status
increased with PM's LB from favorable situation (octant I, octant II, and octant III)
towards moderately favorable situation (octant IV, octant V, and octant VI), remained
positive and high in moderately favorable situation, and became negative and decreased
as the situational favorability decreased towards unfavorable (octant VII and octant VIII)
situation (Figure 2).
The negative correlation coefficients in the favorable and unfavorable situations
indicated that the PMs of ROLB (with high LPC) performed the least, and the PMs of
TOLB (with low LPC) performed the best in these two situations. In moderately
favorable situation, the PMs of ROLB showed better performance (more successful
completion of the DT projects they managed) as indicated by the positive correlations
between DT project completion status and PM's LB in this situation: (a) octant III (rho =
.18, [.14, .37]), (b) octant IV (rho = .15, [-.48, .90]), (c) octant V (rho = .35, [.18, .65]),
and octant VI (rho = .17, [.49, -.67]).
104
Figure 2
Spearman rho Correlations Between PM's LB and DT Project Completion Status
Note: Struc. = structured, Unstruc. = unstructured, PM = project manager, LB = leadership behaviors. The
correlation coefficients on the Y-axis are Spearman's rho coefficients. Source of Fiedler's CTL model:
“Contingency Model of Leadership Effectiveness: Antecedent and Evidential Results,” by G. Graen, K.
Alvares, J. B. Orris, and J. A. Martella, 1970, Psychological Bulletin, 74(4), 285–296
(https://doi.org/10.1037/h0029775).
The sign and correlation coefficient numbers remained in the hypothesized
direction of Fiedler's (1964) CTL model in all eight octants. However, some of the
correlations were low and nonsignificant at the .05 level because of the transition from
positive to negative values and vice versa of correlations, a criterion of statistical
instability. Overall, the study's results supported the central hypothesis of Fiedler's (1964,
1967) CTL model and its application to DT projects in LICs in the United States. The
105
results were compatible with the CTL theory's assumption that the leader's contribution to
his group's performance depends upon the leader's LB and the nature of the situation
(favorableness for the leader).
Test of Statistical Assumptions
The BLR model assumes (a) a binary dependent variable; (b) a large sample; (c)
no multicollinearity in data; (d) no outliers, influential observations, or high leverage
points existing in the data; and (e) the observations being mutually exclusive and
exhaustive. I evaluated the data for the above statistical assumptions of the BLR model. I
used the standard tests in SPSS to ascertain violations.
Binary Dependent Variable
In this study, the dependent variable (project completion status) took only one of
two values (yes or no). I coded the values as yes = 1 and no = 0; thus, the dependent
variable was binary. Frequency analysis of the survey data (Table 4) indicated that the
response of all 214 participants to the dependent variable questions fell into one of these
two categories. The survey results indicated no redundant, incorrect, or invalid
participant responses, and I used all 214 observations in the analysis. The data met the
assumption of binary dependent variable for BLR analysis.
Large Sample
Small to moderate sample sizes overestimate the effect they measure because,
with small sample sizes, the Hosmer-Lemeshow goodness of fit test has low power and
cannot detect minor deviations from the BLR model (Boateng & Abaye, 2019). I
106
performed an a priori power analysis with G*Power 3.1.9.7, which yielded a minimum
required sample size of 143 with an actual power of .80 and a critical Z value of 1.96 for
BLR analysis (Figure 1). The actual study sample size (N = 214) was larger than 143,
large, and adequate for BLR analysis with four independent variables. There were no
missing data in the sample (Table 3), which could reduce the sample size when removing
the participant cases with missing data. I used all the 214 cases in the BLR analysis.
Besides, I used bootstrapping to increase the sample size (≥ 1,000) in all tests to obtain
more accurate estimates with the 95% CI. The data met the assumption of the large
sample.
Multicollinearity
I computed the nonparametric Spearman's rho pairwise correlation coefficients for
the study's independent variables in SPSS to evaluate the existence of multicollinearity.
Unlike the Pearson correlation, the Spearman correlation does not assume that the
variables are normally distributed and computes bivariate correlations expressed as the
strength of association between two variables in a single value between -1 and +1, hence
could be effectively used to estimate the correlation between categorical variables;
however, for categorical variables, Spearman's rho values as small as ±.25, may indicate
a strong correlation (Mesfioui et al., 2022). I also used the chi-square correlation test for
categorical variables (Phi and Cramer's V tests) in SPSS to determine pairwise
correlations of the study's categorical independent variables. Phi and Cramer's V are
nonparametric tests that measure the strength of association between two categorical
107
variables, (Akoglu, 2018). A value greater than .25 or less than -.25 of Phi or greater than
.25 of Cramer's V indicates a very strong correlation between the categorical variables
(Akoglu, 2018). Using a linear regression procedure, I estimated the tolerance and VIF
multicollinearity statistics for the study's independent variables in SPSS. The variance
inflation factor (VIF) and tolerance scores obtained in SPSS for a variable quantifies how
well the other variables explain that variable in the model. For BLR, the VIF score value
of two or larger (≥ 2.0) contain multicollinearity and can be problematic (Harris, 2021).
The nonparametric Spearman's rho (rho) correlation matrix for the study's
independent variables indicated a significant positive correlation between (a) PM's LMR
and PM's LB (rho = .18, p = .008), (b) PM's LMR and PM's PP (rho = .19, p = .005), (c)
PM's LMR and PM's TS (rho = .22, p = .001), (d) PM's LB and PM's PP (rho = .25, p <
.001), and (e) PM's PP and PM's TS (rho = .15, p = .025) (Table 5), indicating the
possible existence of multicollinearity in the data. No evidence of high correlations
existed between PM's LB and PM's TS (rho = .03, p = .653). The Phi and Cramer's V
correlation coefficients for the pairwise binary independent variables also indicated
similar significant positive correlations between (a) PM's LMR and PM's LB, (b) PM's
LMR and PM's PP, (c) PM's LMR and PM's TS, (d) PM's LB and PM's PP, and PM's PP
and PM's TS (Table 6).
108
Table 5
Spearman's rho Coefficients, Significance, and CI of Independent Variables
Variable Pair N
Spearman's
rho Significance
Bootstrapped 95% CI for
Spearman's rhoa
LL UL
1. PM's LMR vs. PM's LB 214 .18** .008 .01 .35
2. PM's LMR vs. PM's PP 214 .19** .005 .02 .36
3. PM's LMR vs. PM's TS 214 .22** .001 .08 .36
4. PM's LB vs. PM's PP 214 .25** <.001 .04 .45
5. PM's LB vs. PM's TS 214 .03 .655 -.11 .18
6. PM's PP vs. PM's TS 214 .15*.025 .01 .30
Note. CI = Confidence interval, LL = Lower level, UL = Upper level, PM = Project manager,
LB = Leadership behaviors, LMR = Leader member relations, PP = Position power, TS = Task structure.
a Bootstrap result are based on 1000 bootstrap samples.
** Correlation is significant at the .01 level (two-tailed).
* Correlation is significant at .05 level (two-tailed).
Table 6
Phi and Cramer's V Coefficients and Significance of Independent Variables
Variable Pair NPhi Cramer's VApproximate
significance
Exact
significance
1. PM's LMR vs. PM's LB 214 .18** .18** .008 .014
2. PM's LMR vs. PM's PP 214 .19** .19** .005 .008
3. PM's LMR vs. PM's TS 214 .22** .22** .001 .002
4. PM's LB vs. PM's PP 214 .25** .25** <.001 .002
5. PM's LB vs. PM's TS 214 .03 .03 .653 .798
6. PM's PP vs. PM's TS 214 .15*.15*.025 .038
Note. PM = Project manager, LB = Leadership behaviors, LMR = Leader member relations, PP = Position power,
TS = Task structure.
109
** Correlation is significant at the .01 level (two-tailed).
* Correlation is significant at .05 level (two-tailed).
None of the independent variables indicated VIF > 2.50 or tolerance < .40 (Table
7). All the variables indicated VIF values below 1.20 and high tolerance values above or
equal to .90 (Table 7), indicating no severe multicollinearity of variables in the study
data. However, the VIF and tolerance value computation assumed a linear regression
model and may not apply for the nonlinear BLR model. Based on the results from the
nonparametric bivariate correlations, this study's data indicated the possible existence of
severe multicollinearity. Five of the six bivariate independent variable pairs indicated
significant correlations with p < .05 (Table 5 and Table 6).
Table 7
Collinearity Statistics of Independent Variables
Collinearity statistics
Variable NTolerance VIF
1. PM's LMR 214 .91 1.11
2. PM's LB 214 .92 1.09
3. PM's PP 214 .90 1.11
4. PM's TS 214 .94 1.07
Note. VIF = Variance inflation factor, PM = Project manager, LMR = Leader member relations, LB =
Leadership behaviors, PP = Position power, TS = Task structure.
Outliers, Influential Observations, and High Leverage Points
I obtained and analyzed the residuals and diagnostic statistics from an initial BLR
model fitted for the binary dependent and the four binary independent variables (PM's
LB, PM's PP, PM's LMR, and PM's TS). The Hosmer–Lemeshow goodness of fit test
results (χ2 [4, N = 214] = 1.98, p = .74) indicated an adequate fit of the initial BLR
110
model. I plotted the model-generated residuals, Cook's D values, and leverage points
against the observation indices and the model-predicted probabilities to evaluate data for
the existence of outliers, influential observations, and high leverage points.
Outliers. I have presented the plots of deviance (Figure 3) and standardized
(Figure E1 in Appendix E ) residuals from the initial BLR model against the model-
predicted (predicted) probabilities. Model residuals plotted against the predicted
probabilities indicated two almost linear trends, with slope -1 indicating decreasing
residuals, as expected. The deviance residual values ranged from -1.50 to 1.25, and none
of the observations in the plot indicated deviance residual values > 2 or < -2 to be an
outlier. Both residual plots indicated no residuals with values > +2 or < -2. The plot of
deviance residuals from the initial BLR model against the observed case indices (Figure
4) indicated no observations with deviance residual values outside of ±2. Thus, there
were no outliers in the observed data, as all the residual values remained within ±2. The
frequency histograms plotted (with trend lines added) of the deviance (Figure 5) and
standardized (Figure E2 in Appendix E) residuals from the initial BLR model displayed
asymmetric distribution and did not indicate a long tail in one direction or a bar outside
the ±2 values, indicating no outliers present in the data.
111
Figure 3
Deviance Residuals Against Predicted Probabilities From Initial BLR Model
2.00
D
e1.50
vi
a1.00
n0.50
c
e0.00
R-0.50
e
si -1.00
d
u-1.50
al
-2.00
0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80
Predicted Logistic Probabilities
Note. Dependent variable is project completion status. Independent variables are PM's LMR, PM's PP,
PM's TS, and PM's LB. N = 214.
112
Figure 4
Deviance Residuals From Initial BLR Model Against Observation Indices
D2.00
e
vi 1.50
a1.00
n
c0.50
e
0.00
R-0.50
e
si -1.00
d
u-1.50
al
-2.00
1 19 37 55 73 91 109 127 145 163 181 199 217
Observation Index
Note. Dependent variable is project completion status. Independent variables are PM's LMR, PM's PP,
PM's TS, and PM's LB. N = 214.
113
Figure 5
Frequency Histogram of Deviance Residuals From Initial BLR Model
120
100
80
60
40
20
0
-2.00 -1.50 -1.00 0.00 1.00 1.50 2.00
Deviance Residuals
Note. Dependent variable is project completion status. Independent variables are PM's LMR, PM's PP,
PM's TS, and PM's LB. N=214.
Influential Observations. Figure 6 displays the Cook's distance (Cook's D) from
the initial BLR model for the study data against the observed case indices. Cook's D
measures the influence of each observation on the BLR model parameter estimates; an
influential observation is that it has Cook's D value > .50 (Costa e Silva et al., 2020). In
the BLR model of study data, Cook's D values ranged from .005 to .214, and none of the
observed cases indicated Cook's D > .50 to be influential. Thus, no influential
observations existed in the study data.
Frequency
114
Figure 6
Cook's Distances From Initial BLR Model Against Observation Indices
0.50
C0.45
o
o0.40
k' 0.35
s
0.30
D0.25
is 0.20
t
a 0.15
n0.10
c
e0.05
0.00
1 19 37 55 73 91 109 127 145 163 181 199 217
Observation Index
Note. Dependent variable is project completion status. Independent variables are PM's LMR, PM's PP,
PM's TS, and PM's LB. N=214.
High Leverage Points. Figure 7 displays the leverage values from the initial BLR
model for the study data against the observed case indices. In the BLR model of study
data, the leverage values ranged between .006 and .196, with a mean leverage value
(MLV) of .019; 2 times the MLV was .037 and 3 times the MLV was .056. Twenty-one
(21) observations had leverage values larger than 2 times the MLV or > .037, and 17
observations had leverage values larger than 3 times the MLV or > .056 indicating
existence of several high-leverage points in the data (Figure 7).
115
Figure 7
Leverage Values From Initial BLR Model Against Observation Indices
1.00
L 0.50
e
v0.25
e
r0.13
a
g0.06
e
0.03
0.02
0.01
1 19 37 55 73 91 109 127 145 163 181 199 217
Observation Index
Note. MLV = Mean leverage value. MLV = .019, MLV * 2 = .037, MLV * 3 = .056.
Independence of Observations
The data for each observation was independent or came from separate
participants. The data did not come from repeated measurements for any participants
entered twice. Further, the deviance residuals plot against the observation indices (Figure
4) indicated random patterns around the two dependent categories (0 and 1), indicating
that the observations were independent. Thus, the data met the assumption of
independent observations.
Leverage Points MLV * 2
MLV * 3
116
Reliability and Validity Analyses of Survey Instruments
I performed reliability and validity analyses on the four survey scales (LPC rating
scale [18 items], TS rating scale [10 items], LMR rating scale [8 items], and PP rating
scale [5 items]) that I used to collect primary data for the study's independent variables. I
assessed the reliability of the scales in terms of split-half reliability and IC reliability,
which are acceptable tests for the reliability assessment of survey scales (Kishore et al.,
2021; Rossell et al., 2019). I used the 'Alpha' and 'Split-Half ' options under reliability
analysis in SPSS separately to assess the survey scales for these two reliabilities. I
obtained the Cronbach's Alpha values for the four scales in SPSS to assess their IC
reliability. Analysis of the split-half reliability in SPSS allows to divide the
questionnaire's items into two halves and calculate correlations between the two halves
with high correlations (> .50) between the two halves, indicating high split-half reliability
of the scales (Kishore et al., 2021; Rossell et al., 2019). I also obtained the Spearman-
Brown (SB) coefficient and the Guttman Split-Half coefficient of the four scales to
further assess their split-half reliability. I assessed instrument validity by obtaining item-
to-total score correlations. According to Rossell et al. (2019), the correlation between
each item's score on the survey scale and the total score from all items is a reliable
assessment method for the validity of survey scales.
Reliability Analysis Results
The Cronbach's Alpha values ≥ .70 for the LPC rating scale (.97) and TS rating
scale (.70) indicated adequate IC reliability of these survey scales. The < .70 Cronbach's
117
Alpha value of the LMR rating scale (.67) indicated questionable IC reliability of this
scale. The Cronbach's Alpha value of the PP rating scale (.40) indicated inadequate IC
reliability of this scale. Ayman and Chemers (1991) reported good IC reliability for the
LPC scale (Cronbach's Alpha = .90) and the LMR scale (Cronbach's Alpha = .80). In a
recent study, Arjanto et al. (2022) reported high IC reliability with a Cronbach's Alpha
value = .75 for the LPC scale. Previous IC reliability data for the PP rating scale is
available from only one study by Ayman and Chemers (1991), where it had a low
Cronbach’s Alpha of .31. According to Ayman and Chemers (1991), this low IC
reliability value of PP rating scale is the result of the multidimensional nature of the
scale, which measures several bases of power reported by the leader. This five-item scale
measures the leader's discretionary power to reward and punish, job-relevant expertise,
and official status. According to Fiedler and Chemers (1984), like the TS, PP is defined at
the individual level for the leader and contributes to the overall level of control in the
leader's situation; several field studies indicated that the leader's self-report was reliable.
The low IC reliability scores < .70 found in this study for the PP rating scale and
the LMR scale made understanding and applying the data per CTL model difficult. The
reliability statistics of the LMR scale indicated two items (Question 1 and Question 5,
Table D1) on the LMR scale having low IC scores; deleting these items would improve
the overall reliability of the data collected. Similarly, the reliability statistics of the PP
rating scale indicated one item (Question 5, Table D3) on the PP rating scale having a
low IC score; deleting this item would improve the overall reliability of the data
118
collected. These items were irrelevant and scored on different scales than the others,
creating nonreliable data. I removed these questionable items and reevaluated the
reliability of the LMR and the PP questionnaires. I have displayed in Table 8 the
Cronbach's Alpha values for all scales after removing the nonreliable items from the
LMR and PP rating scales. The Cronbach's Alpha values obtained for the LMR scale
(.85) and the PP rating scale (.48) in this study were larger (Table 8) than the values
reported by Ayman and Chemers (1991) for the LMR scale (.80) and the PP rating scale
(.31).
According to Fiedler and Chemers (1984), the cutoff scores for the high and low
levels of control need adjustment for the prescribed categories by the maximum number
of points possible on each scale. Previous researchers in their studies, in some cases,
have used extreme scores (cutoffs based on a standard deviation on each side of the mean
or the top and bottom 10% or thirds of the distribution [e.g., Ayman & Chemers, 1991]).
Past researchers used median or mean split in other studies to categorize high and low
scores (e.g., Chemers et al., 1985). Fiedler (1967) and Fiedler and Chemers (1984) used
the 75th percentile score of the distribution as the cutoff point for the high and low levels
of the scales; I used the 75th percentile score of the new total as the cutoff point for the
LMR and the PP scales.
The high (> .70) split-half SB reliability coefficients for the LMR rating scale
(.84), LPC rating scale (.95), and TS rating scale (.74) indicated high split-half reliability
of these scales (Table 8). Correlation coefficients between the split halves were also high
119
(> .50) for these scales, further indicating adequate split-half reliability (Table 8). The
low correlation (.18) coefficient and low SB coefficient (.30) of the PP rating scale
indicated that the reliability of this scale needed to be revised.
Table 8
Reliability Statistics of Survey Instruments
Reliability statistics
IC reliability
statistics Split-half reliability statistics
Instrument Cronbach’s
Alpha
Correlation
between halves
Spearman Brown
coefficient
Guttman split-
half coefficient
LMR rating scale .84 .74 .85 .85
TS rating scale .70 .53 .70 .70
PP rating scale .48 .18 .30 .28
LPC rating scale .97 .90 .95 .95
Note: IC = Internal consistency, LMR =Leader member relations, TS = Task structure, PP = Position
power, LPC = Least-preferred coworker.
Validity Analysis Results
Highly significant (p < .001) item-to-total score correlations for all items in all
four scales confirmed the high validity of the four scales (Table 9). All item-to-total
score correlations were positive, further verifying the four scales' validity. Negative
correlations between item and total scores indicate errors in the scale, thus invalidity of
scales (Rossell et al., 2019), which was not the case in this study. The high factor
loadings above .40 for all variables in the factors (Table 10) indicated that the data
satisfied the construct validity, including the convergent validity and the discriminant
validity (Knekta et al., 2019). Fiedler (1967) reported strong construct validity for the
four scales.
120
Table 9
Correlation Between Item and Total Scores of Survey Scales
Instrument Item # Pearson correlation
coefficient Significance (2-tailed) Bootstrapa 95 % CI
LL UL
LMR rating
scale
1 .64** <.001 .59 .72
2 .59** <.001 .48 .68
3 .62** <.001 .52 .71
4 .53** <.001 .42 .62
5 .57** <.001 .49 .66
6 .57** <.001 .47 .67
7 .68** <.001 .59 .76
8 .54** <.001 .43 .64
PP rating
scale
1 .69** <.001 .61 .76
2 .68** <.001 .59 .76
3 .37** <.001 .26 .47
4 .41** <.001 .31 .51
5 .36** <.001 .25 .45
TS rating
scale
1 .56** <.001 .46 .66
2 .51** <.001 .40 .60
3 .56** <.001 .45 .65
4 .45** <.001 .34 .56
5 .50** <.001 .38 .60
6 .52** <.001 .42 .61
7 .57** <.001 .46 .66
8 .53** <.001 .42 .62
9 .49** <.001 .38 .59
10 .42** <.001 .31 .53
LPC rating
scale
1 .75** <.001 .63 .84
2 .81** <.001 .72 .87
3 .82** <.001 .75 .87
4 .73** <.001 .68 .80
5 .71** <.001 .62 .79
6 .78** <.001 .70 .85
7 .79** <.001 .70 .87
8 .75** <.001 .67 .82
9 .80** <.001 .59 .80
10 .75** <.001 .77 .90
11 .72** <.001 .59 .80
12 .82** <.001 .73 .88
13 .83** <.001 .76 .88
14 .85** <.001 .77 .90
15 .83** <.001 .77 .88
16 .85** <.001 .81 .89
17 .88** <.001 .83 .91
18 .83** <.001 .75 .89
Note. Details of the items by item # in each scale are presented in Appendix D in Tables D1, D2, D3 and D4. CI= Confidence interval,
LL = Lower level, UL = Upper level.
a Bootstrap result are based on 1000 bootstrap samples.
** Correlation is significant at .01 level (2-tailed).
121
The reliability and validity analysis results confirmed that the instruments used in
this study had good psychometric properties, and the data collected based on these
instruments were of adequate quality. Given the validity and reliability of the
instruments, the inferential results based on the collected data were valid and reliable.
Based on the results from this study, further validations through test-retest reliability and
inter-rater reliability analysis with additional data would benefit future research.
Inferential Results
Factor Analysis Results
Although BLR provides a parsimonious combination of the best predictor
variables, the significant association of any two independent variables adversely impacts
the results of a BLR model by reducing the association of the independent variables with
the outcome variables (Akoglu, 2018). The existence of multicollinearity in the data
results in incorrect or underpredicted overall levels of significance from the BLR model
and leads to individual predictors not predicting the outcome correctly or the degree of
relationship between a predictor and the outcome incorrectly established (Akoglu, 2018).
Combining the correlated predictor variables or removing one of the variables could
significantly improve the accuracy of the BLR model (Akoglu, 2018).
Significant nonparametric correlations existed in this study's data between the
following binary independent variable pairs: (a) PM's LMR and PM's TS (p = .001), (b)
PM's LB and PM's PP (p < .001), (c) PM's PP and PM's TS (p = .025), (d) PM's LMR
and PM's LB (p = .008), and (e) PM's LMR and PM's PP (p = .005) (Table 5 and Table
122
6). Besides, the study data contained several high-leverage points (Figure 6), causing
significant problems, including erroneous goodness-of-fit statistics, wrong OR, and
wrong Wald statistics (Costa e Silva et al., 2020). Significant (p < .05) linear Pearson
bivariate correlations existed among the Z-normalized independent variables (Table E4
[Appendix E]), indicating linear relationships among variables. I performed an
exploratory FA using the PCA extraction method and varimax rotation to extract
significant factors of the study's Z-normalized independent variables using the 'factor'
procedure in SPSS. The goal in performing FA was to organize the four independent
variables obtained from the survey questionnaire scores into a valid set of uncorrelated
composite factors representing the original predictor variables.
In FA, the PCA extraction method is a valuable analysis tool to identify the
factors underlying the predictor variables of a data set to measure the outcome (Shrestha,
2021). PCA method transforms the data into orthogonal uncorrelated variables known as
principal components, preserves the total variance in the original data, and rejects the
combination of variables that do not explain much of the variance in the data; PCA
minimizes or eliminates common issues such as multicollinearity and overfitting
(Shrestha, 2021). The orthogonal varimax rotation procedure forces factors to be
independent (Nájera et al., 2023). It maximizes the variances of factor loadings across
variables of each factor, enhancing the interpretability of the result (Muhamad et al.,
2024). BLR with extracted principal components can result in a more accurate model
123
with a superior fit than the one fitted with the original correlated independent variables
(Costa e Silva et al., 2020; Muhamad et al., 2024; Shrestha, 2021).
I used the Kaiser-Meyer-Olkin (KMO) test to measure sampling adequacy,
Bartlett’s sphericity test to assess the factorability of the data and calculated the
determinant score of the correlation matrix to examine the multicollinearity among the
variables during FA. The KMO test measures the suitability of data for factor analysis
and measures the sample adequacy for each variable in the model; A KMO value > .60 is
acceptable for a sample size < 100, and a KMO value between .50 and .60 is acceptable
for sample sizes between 100 and 200 (Galak, 2020c; Shrestha, 2021). Because of the
larger than 200 (N = 214) sample size of the study’s data, a KMO value >.50 was
acceptable. Bartlett’s sphericity test tests the null hypothesis, Ho: When accepted, the
variables have an original correlation matrix (an identity matrix), indicating that the
variables are unsuitable for factor detection (Galak, 2020c; Muhamad et al., 2024). The
KMO test and the Bartlett sphericity test together measure the construct validity of the
factors through calculating sample adequacy (Shrestha, 2021).
KMO test measure (.59) for sampling adequacy and the significance of Bartlett’s
sphericity test (χ2 [6, N = 214] = 39.14, p < .001) results indicated that the correlation
matrix of the data was appropriate for FA (Shrestha, 2021). The factors indicated
convergent validity. The diagonal Anti-Image correlation coefficients larger than .50,
which must be larger than .50 (Shrestha, 2021) for all the independent variables (a) PM's
124
LB (.57), (b) PM's LMR (.61), (c) PM's PP (.60), and (d) PM's TS (.57) further indicated
adequacy of FA for the data (Table 10).
Table 10
Anti-Image Correlation Matrix of Independent Variables
Variable PM's LMR PM's PP PM's TS PM's LB
PM's LMR .61a-.13 -.20 -.15
PM's PP -.13 .60a-.12 -.22
PM's TS -.20 -.12 .57a.04
PM's LB -.15 -.22 .04 .57a
Note. PM = Project manager, LMR = Leader member relations, PP = Position power, TS = Task structure,
LB= Leadership behaviors.
a Measures of sampling adequacy (MSA).
The Kaiser's eigenvalue of a factor represents the amount of the total variance
explained by that factor. In FA, the factors having eigenvalue ≥ one explains more
common variance and hence are retained, which is an acceptable rule in FA (Shrestha,
2021). Further, the scree plot (plot of eigenvalue against the component number)
obtained in FA helps to identify the optimum number of factors; the factor before the plot
starts to flatten out should be extracted (Shrestha, 2021). I examined Kaiser's eigenvalue
criterion and Scree plot from the FA to determine the number of factors to retain and
apply and the varimax orthogonal factor rotation method to optimize the number of
variables with high loadings (> .40) on each factor. The FA with PCA method and
varimax rotation resulted in four components, with two having acceptable eigenvalues
(Table 11). I retained the two factors with rotated eigenvalues equal to 1.29 (Factor 1)
and 1.23 (Factor 2). FA's scree plot (Figure 8) started to flatten after component 3,
indicating that the amount of
125
unique variance began to dominate the common variance starting at component 3. The
eigenvalues and the scree plot confirmed the retention of Factor 1 and Factor 2 for further
analysis.
Table 11
Total Variance Explained by Extracted Factors
Rotation sums of squares loadings Rotated component matrix
Factor Percent
of
Cumulative
percent of PM’s LB PM’s PP PM’s TS PM’s LMR
Total variance variance
1 1.29 32.22 32.22 85 .67
2 1.23 30.66 62.88 .88 .62
Note. Extraction method is principal component analysis. PM = Project manager, LMR = Leader member
relations, PP = Position power, TS = Task structure, LB= Leadership behaviors.
Factor 1 explained 32.21% of the rotated total variance, and Factor 2 explained
30.66% with a cumulative variance of 62.88% (Table 11). Principal components with a
percentage of more than 10% variance and a cumulative variance of at least 60% are
considered acceptable (Muhamad et al., 2024). Factor 1 indicated high positive loading
values for PM's LB (.85) and PM's PP (.67) and adequately represented these two
independent variables (Table 12). Factor 2 indicated high positive loading values for
PM's TS (.88) and PM's LMR (.62) and represented them adequately (Table 12). The
correlation matrix's determinant score (.83), closer to 1, indicated the absence of
multicollinearity (Shrestha, 2021) in the extracted factors.
126
Figure 8
Scree Plot of Extracted Components
1.6
1.4
1.2
1
0.8
0.6
0.4
0.2
0
1234
Component Number
Table 12
Factor Loadings and Communalities of Independent Variables
Factor loading
Variable Factor 1 Factor 2 Extraction
communality
1. PM's LB .85 .72
2. PM's PP .67 .52
3. PM's LMR .62 .51
4. PM's TS .88 .77
Note. Extraction method: Principal component analysis. Rotation method: Varimax with kaiser
normalization. PM = Project manager, LB = Leadership behaviors, LMR = Leader member relations, PP =
Position power, TS = Task structure.
The communality of a variable in FA, which varies between 0 and 1, reveals the
proportion of variation in that variable explained by the extracted factors. The
Eigenvalue
127
communality of a variable will be equal to one when that variable does not have any
unique variance (its explained variance is 100% a result of other variables).
Communalities of variables > .50 in extracted factors indicate that the factors explain
most of the variation in those variables and are preferred (Shrestha, 2021). When a factor
has loadings greater than .60, the factor is stable regardless of sample size, and when the
communalities are ≥.50, the sample size needed is around 200 and below .50, 500, or
more (Schreiber, 2021). All the independent variables of the study used in the FA had
communalities >.50 (PM's LB [.72], PM's TS [.77], PM's PP [.52], and PM's LMR [.51]);
thus, extracted factors explained 72% of variation in PM's LB, 77% of variation in PM's
TS, 52% of variation in PM's PP, and 51% of variation in PM's LMR (Table 12). The
significant factor loadings (> .60) of all the independent variables in the two factors,
significant communalities (> .50) of all the variables in the factors, and the highly
nonsignificant Pearson correlation coefficient (.000, p = 1.00) between the two factors
indicated the factor model was superior. The results indicated that the FA allowed the
detection of relevant combinations of variables and the extraction of valuable factors
from the data set. Applying FA for component extraction allowed me to focus on a few
essential factors valuable in BLR model fitting.
BLR Analysis Results
To answer RQ1, I investigated the relationship between the four independent
variables (PM's LB, PM's LMR, PM's TS, and PM's PP) grouped into two factors, Factor
1 (PM's LB and PM's PP grouped) and Factor 2 (PM's LMR and PM's TS grouped) on
128
the probability of completing the DT projects successfully by the PMs through the BLR
analysis method with bootstrapped (> 1000) samples. I performed separate BLR analyses
by the three PMs' situational favorability: (a) favorable, (b) moderately favorable, and (c)
unfavorable. Table 13 illustrates the results from the BLR analysis in the three situational
favorability. I included the interaction effect of the relationship between the two factors
(Factor 1 * Factor 2) on the probability of successfully completing the DT projects by the
PMs as a predictor in the analysis. In a BLR analysis, an interaction effect (X * Z)
between two predictors (ex., X and Z) occurs when the relationship between one
predictor, X, and the outcome (response) variable, Y, depends on the value of the other
predictor variable, Z; an X * Z interaction term means that the X moderates the effect of
Z on Y and Z moderates the effect of X on Y(Fisher, 1992). A significant (p < .05)
interaction coefficient indicates that the association between X and the probability that Y
= 1 depends on the values of Z and vice versa, where X and Z may be binary or
continuous; in this case, the individual effects of the two predictors become less critical
(Fisher, 1992).
The nonsignificant p-value, almost close to 1 of the Hosmer and Lemeshow test,
which assesses the goodness of fit of the BLR model (Boateng & Abaye, 2019), for the
three situations: (a) favorable (χ2 = .01, p = .94), (b) moderately favorable (χ2 = .00, p =
1.00), and (c) unfavorable (χ2 = .13, p = .99) indicated, that the model predicted the data
exceptionally well in all three situations. The negative and significant value of β for the
interaction effect between Factor 1 and Factor 2 (Factor 1 * Factor 2) in the favorable (β
129
= -.99, p = .025) and unfavorable (β = -90.73, p = .040) situations (Table 13) indicated
that the two factors were dependent on each other, moderated the effect of each other,
and together interactively significantly decreased the chances of PMs successfully
completing the DT project in these two situations (favorable and unfavorable). Most of
PMs in the favorable (94.50% ) and unfavorable (81.80%) situations scored high on the
LPC scale displaying ROLB. In other words, when the situation was favorable (octant I,
octant II, and octant III) or unfavorable (octant VII and octant VIII) to the PMs with
ROLB, the odds of the PMs successfully completing the DT projects became less likely.
As indicated by Fiedler (1967) and Fiedler and Chemers (1984) in the CTL model these
two situations are best for the PMs with TOLB.
The odds ratio values less than 1 for the Factor 1 * Factor 2 interaction term (.37
[.14, .99]) under favorable conditions (Table 13) indicated that Factor 1 and Factor 2
interactively decreased the likelihood of the PMs successfully completing the DT projects
by 63 percent ([.37 - 1.00] * 100) when the factors' values increased from 0 to 1. The
almost zero odds ratio of the interaction term under unfavorable conditions (Table 13)
indicated that the factors interactively decreased the likelihood of the PMs successfully
completing the DT projects by almost 100 percent ([.00 - 1.00] * 100) when the factors'
values increased from 0 to 1.
The positive significant (p < .05) coefficient estimates (β) of the individual effects
of Factor 1 (.90, p = .047) and Factor 2 (.91, p = .032), and the larger than one odds ratio
for Factor 1 (2.47 [.78, 7.81] ) and Factor 2 (2 .49 [.98, 6.35]) in the favorable situation
130
(Table 13) indicated that the factors individually increased the successful completion of
the DT projects by the PMs. However, the significant and negative β for the interaction
effect of the two factors (β = -.99, p = .025) in this situation indicated that the two factors
moderated the effect of each other and significantly and jointly decreased the chances of
PMs completing the DT projects successfully in the favorable situation. The negative
significant (p < .05) of β and the almost zero odds ratio for the individual effect of Factor
1 and Factor 2 in the unfavorable situation (Table 13) indicated that the factors
individually decreased the successful completion of the DT projects by the PMs in the
unfavorable situation. Also, the significant and negative β for the interaction effect of the
two factors (β = -90.73, p = .040) in this situation indicated that the two factors
moderated the effect of each other and significantly and jointly decreased the chances of
PMs completing the DT projects successfully. It is important to note that when the
interaction effect of two predictors is significant, the individual effects of the two
predictors become less critical (Fisher, 1992). However, the significance of the individual
effects of predictors in these situations provides additional insights regarding the effects
of these factors.
The nonsignificant β for the interaction effect of Factor 1 and Factor 2 (Factor 1 *
Factor 2) in the moderately favorable (β = .58, p = .532) situation (Table 13) indicated
that the two factors did not moderate the effect of each other or significantly impact the
successful completion of DT projects by the PMs in the moderately favorable situation.
The nonsignificant (p > .05) coefficient estimates (β) of the individual effects of Factor 1
131
(.15, p = .639) and Factor 2 (-.67, p = .585) in the moderately favorable situation (Table
13) also indicated that changing values of Factor 1 (PM’s LB and PM’s PP grouped) or
Factor 2 (PM’s LMR and PM’s TS grouped) individually from 0 to 1 had no impact on
the chances of PMs successfully completing the DT project in the moderately favorable
situation. The PMs from the moderately favorable situation indicated a mix of ROLB and
TOLB.
Overall, results indicated that the PMs' LB and the PMs' contingency situation
(represented by PM's LMR, PM's PP, and PM's TS) moderated each other's effect and
together determined the successful completion of the DT project by the PMs in the
favorable and the unfavorable situation. Most PMs in the favorable situation (95.50%)
and unfavorable situation (81.80%) were RO leaders who, according to this study’s
results, were unable to complete the DT projects successfully, as indicated by the
negative and significant interaction effects of the four independent variables grouped into
the two factors (Table 13). The negative and significant interaction effect in the
favorable and unfavorable situations further indicated that the two factors (the four
independent variables) together interactively significantly decreased the chances of PMs
completing the DT project successfully in these two situations when the values of factors
(the values of variables associated with the factors) changed from 0 to 1, which indicated
that PMs with ROLB (high LPC) could not complete the DT projects successfully in the
favorable and unfavorable situations. These results confirmed the CTL theory that RO
leaders perform the least in favorable and unfavorable situations (Fiedler, 1967).
132
Managers who display a TOLB focus on details, give directions, and prescribe the
work to team members. Conversely, managers who display a ROLB create trust and
respect for the team members, allowing them to be part of project decisions (Fiedler,
1967; Henkel et al., 2019). According to Fiedler and Chemers (1984), the socio
independent leaders (middle LPC score leaders) also performed well in the favorable and
the moderately favorable situations. This study's findings and the above results by Fiedler
and Chemers (1984) indicate that a balance between a PM's TOLB and ROLB based on
the team members' readiness and skill level for specific tasks, experience, and maturity
might be appropriate in moderately favorable situations. Any concrete conclusions in the
moderately favorable situation would need more investigation with additional data and
predictors.
Table 13
BLR Analysis Results
Situational
favorability
Bootstrap
significance of
β (2-tailed)a
95% CI of exp(β)
Predictor β Exp(β) LL UL
Factor 1 .90 .047b2.47 .78 7.81
Favorable Factor 2 .91 .032b2.49 .98 6.35
Factor 1 * Factor 2 -.99 .025b.37 .14 .99
Constant -.40 .187b.67 -e-e
Factor 1 -177.81 .044c.00 .00 .00
Unfavorable Factor 2 -154.52 .040c.00 .00 .00
Factor 1 * Factor 2 -90.73 .040c.00 .00 .00
Constant -325.59 .043c.00 -e-e
Factor 1 .15 .639d1.16 .42 3.24
Moderately
favorable
Factor 2 -.67 .585d.51 .01 27.10
Factor 1 * Factor 2 .58 .532d1.79 .13 23.92
Constant -.43 .597d.65 -e-e
Note. CI = Confidence interval. LL = Lower level, UL = Upper level, Exp = Exponential, β = coefficient
estimate. Exp(β) is the odds ratio.
133
a Bootstrap result are based on 4000 bootstrap samples, b based on 3203 samples, c based on 846 samples, d
based on 3206 samples, e – indicates no CI can be estimated for a constant.
Cluster Analysis Results
The dendrogram (the graphical display of the merging of the clusters) obtained
from the HC analysis (Figure 9) indicated the existence of three distinct clusters in the
data (portrayed by the red circles in Figure 9). The mean (M) ± standard deviation (SD)
of cluster distance of the members in each cluster: (a) Cluster 1 (M = 0.95 ± SD = 0.19),
(b) Cluster 2 (M = 0.51 ± SD= 0.32), and (c) Cluster 3 (M = 0.46 ± SD = 0.32) indicated
that members of Cluster 3 were closer to the center followed by members in Cluster 2
and Cluster 1. Results from the ANOVA with the cluster mean square, error mean
square, F values, and the significant p-value (p < .05) indicated that all variables
contributed equally and significantly in forming the three clusters (Table 14). ANOVA
results from CA revealed significant differences among the three clusters for all
variables: (a) PM’s LB (F = 5.33, p = .006), (b) PM’s LMR (F = 78.60, p< .001), (c)
PM’s PP (F = 3.38, p <
.001), and (d) project completion status (F = 534.74, p < .001). Across the entire sample
of 214 cases, the total mean Silhouette score > .50 (total mean Silhouette score = .54)
indicated that the chosen number of three clusters was correct, and the grouping of data
points into clusters was meaningful and displayed acceptable quality.
In Table 15, I have displayed the final cluster centers of variables and the total
number of cases in each cluster resulting from the k-means clustering. In Table 16, I have
displayed the frequency analysis results by variable categories (0, 1) in the three clusters.
There were 35 PMs in Cluster 1, 78 in Cluster 2, and 101 in Cluster 3 (Table 15). Most of
134
the PMs in Cluster 2 and 3 displayed ROLB, with cluster centers for PM’s LB closer to 1
(Table 15). They also experienced similar and favorable situations with cluster centers
closer to 1 for all contingency predictor variables (Table 15).
Frequency analysis indicated that 89.70% of PMs in Cluster 2 and 94.10% in
Cluster 3 experienced favorable situations (belonged to Octants I, II, and III). Only
10.30% of PMs in Cluster 2 and 5.90% of PMs in Cluster 3 experienced moderately
favorable situation (belonged to Octant V). Besides, all PMs (100%) in Cluster 3 and
none of the PMs (0%) in Cluster 2 completed their DT projects successfully (Table 15
and Table 16). A significant percentage (77%) of the PMs in Cluster 1 were RO leaders,
out of which 62.90% experienced unfavorable situation (belonged to the octants VII and
VIII) and a 37.10% experienced moderately favorable situation (belonged to the octants
IV and VI). About half (54.30%) of PMs in Cluster 1 completed their DT projects
successfully and 45.71% did not (Table 15 and Table 16). Out of the PMs who did not
complete their DT projects in Cluster 1, 31.25% experienced moderately favorable
situation and 68.75% experienced unfavorable situation. Out of the PMs who completed
their DT projects in Cluster 1, 42.11% experienced moderately favorable situation and
57.89% experienced unfavorable situation. The final cluster center in Cluster 1, for PM’s
PP was = .49, PM’s TS was = .14, and PM’s LMR was = .23.
Figure 9
Dendrogram From Hierarchical Cluster Analysis
135
Table 14
Results of Dispersion Analysis of Cluster Formation
Variable Cluster
mean
square
df Error mean
square df FSignificance
Project completion status 22.01 2 .041 211 534.74 <.001
PM’s LB .42 2 .078 211 5.33 .006
136
PM’s LMR 7.08 2 .090 211 78.60 <.001
PM’s TS 5.99 2 .165 211 36.38 <.001
PM’s PP 3.32 2 .073 211 45.28 <.001
Note. PM = Project manager, LB = Leadership behaviors, LMR = Leader member relations, TS = Task
structure, PP =Position power.
Table 15
Final Cluster Centers and Number of Cases in Each Cluster
Cluster centers
Variable Cluster 1 Cluster 2 Cluster 3
Project completion status .54 .00 1.00
PM’s LB .77 .95 .93
PM’s LMR .23 .90 .94
PM’s TS .14 .77 .79
PM’s PP .49 .95 .97
Total cases 35 78 101
Note. PM = Project manager, LB = Leadership behaviors, LMR = Leader member relations, TS = Task
structure, PP =Position power.
Table 16
Frequency Analysis of Variables by Clusters
Cluster 1 Cluster 2 Cluster 3
Variable Category Frequency Valid
percent Frequency Valid
percent Frequency Valid
percent
Project
completion status
0 16 45.7 78 100.0 0 0.0
1 19 54.3 0 0.0 101 100.0
PM’s LB 0 8 22.9 4 5.1 7 6.9
1 27 77.1 74 94.9 94 93.1
PM’s LMR 0 27 77.1 8 10.3 6 5.9
1 8 22.9 70 89.7 95 94.1
PM’s TS 0 30 85.7 18 23.1 21 20.8
1 5 14.3 60 76.9 80 79.2
PM’s PP 0 18 51.4 4 5.1 3 3.0
1 17 48.6 74 94.9 98 97.0
Total - 35 100.0 78 100.0 101 100.0
Note. PM = Project manager, LB = Leadership behaviors, LMR = Leader member relations, TS = Task
structure, PP =Position power.
137
BLR Analysis Results by Cluster Membership. I performed BLR analysis of
the data by the cluster membership. In Cluster 2 and Cluster 3, the dependent variable
had only one category and a constant (Table 15); hence, a BLR analysis was impossible. I
performed BLR on the member PMs in Cluster. Some of Cluster 1 member PMs
experienced unfavorable and others moderately favorable situational control. The
nonsignificant p-value, of 1 of the Hosmer and Lemeshow test for the two situations: (a)
unfavorable (χ2 = .00, p = .1.00) and (b) moderately favorable (χ2 = .00, p = 1.00)
indicated, that the BLR model predicted the data exceptionally well in both situations.
Within Cluster 1 members, the β for the interaction effects of the independent variables
(Factor 1 * Factor 2) were not significant at the 5% level in both unfavorable (β = -77.70,
p = .055) and moderately favorable situations (β = .00, p = .999). However, the
interaction effect was almost significant (p = .055) and negative in the unfavorable
situation. The negative value of the coefficient estimate (β) with a p-value less than .05 of
Factor 1 (-152.27, p = .049) and Factor 2 (-132.23, p = .040) and the zero odds ratio for
the individual effects in the unfavorable situation (Table 17) indicated that changing
values individually of Factor 1 (PM’s LB and PM’s PP grouped) or Factor 2 (PM’s TS
and PM’s LMR grouped) from 0 to 1 would significantly decrease (by 100%) the chances
of PMs successfully completing the DT project in unfavorable situation.
In the moderately favorable situation, the positive value with a p-value less than
.05 of β (.77, p = .043) and the odds ratio larger than 1 (exp(β) = 2.16) for the individual
effect of Factor 1 indicated that changing values of Factor 1 (PM’s LB and PM’s PP
138
grouped) individually from 0 to 1 would significantly increase the chances of PMs
completing the DT project successfully. The positive value with a p-value almost close to
.05 of β (1.61, p = .053) (Table 17) and the odds ratio larger than 1 (exp(β) = 5.02) for
the individual effect of Factor 2 in the moderately favorable situation indicated that
changing values of Factor 2 (PM’s LMR and PM’s TS grouped) individually from 0 to 1
would almost significantly increase the chances of PMs successfully completing the DT
project in the moderately favorable situation. As indicated by Fiedler (1967) and Fiedler
and Chemers (1984) in the CTL model, this study’s results indicated that PMs with
ROLB have better chances of completing the DT projects successfully in moderately
favorable situations. Also indicated by the results is that in the moderately favorable
situation, improving the PM’s PP, PM’s LMR, and PM’s TS would significantly improve
the chances of the PMs completing the DT projects successfully.
Table 17
BLR Analysis Results of Cluster Members
Situational
favorability
Bootstrap
significance of β (2-
tailed)a
95% CI of βa
Predictor β Std.
Error LL UL
Factor 1 -152.27 137.14 .049b-435.00b7.40b
Unfavorabl
e
Factor 2 -132.23 125.08 .040b-379.89b25.52b
Factor 1 * Factor
2-77.70 72.52 .055b-227.67b6.55b
Constant -278.66 264.97 .041b-804.06b53.79b
Factor 1 .77 29.89 .043d-41.87c63.49c
Moderately
favorable
Factor 2 1.61 88.51 .053d-143.61c184.39c
Factor 1 * Factor
2.00 15.39 .999d-25.10c25.10c
Constant 3.27 129.74 .044d-201.08c274.59c
Note. CI = Confidence interval. LL = Lower level, UL = Upper level, β = coefficient estimate.
a Bootstrap result are based on 6000 bootstrap samples. b based on 4378 samples. c based on 4473 samples.
139
Summary of Findings
The reliability and validity analysis results confirmed that the instruments used in
this study had good psychometric properties, and the data collected based on these
instruments were of adequate quality. Given the validity and reliability of the
instruments, the inferential results based on the collected data were valid and reliable.
The nonparametric Spearman’s rho correlations as well as the categorical Phi correlations
between the DT project completion status and PM’s LB indicated that the relationship
was nonrandom as shown by Fiedler (1964, 1967) in his CTL model. The successful DT
project completion status was negative and decreased (moved from 1 towards 0) as PM’s
LB became more RO (higher PM’s LB indicate higher LPC scores) at favorable and
unfavorable situations. The successful DT project completion status was positive and
increased as PM’s LB became more RO and remained high and positive in moderately
favorable situations (Figure 2). Results agreed with the CTL theory for DT projects in
LICs in the United States.
FA with the PCA component extraction method resulted in a superior factor
model with significant factor loadings (> .40) for all the independent variables,
significant communalities (> .50) of all the variables in the factors, highly nonsignificant
Pearson correlation coefficient (.000, p = 1.00) between the two factors, anti-image
correlation coefficients ≥ .57 for all variables in the factors, adequate KMO test measure
(.59) indicating sampling adequacy, and highly significant Bartlett’s sphericity test
measure (χ2 [6, N = 214] = 39.14, p < .001) indicating adequate correlations for FA.
140
Factor 1 combined PM’s LB and PM’s PP, with PM’s LB as the major variable in the
factor with a .85 loading and explained 38.05% of the total variation in the dependent
variable. Factor 2 combined PM’s TS and PM’s LMR and explained 24.83% of the total
variation in the dependent variable. FA allowed the detection of relevant combinations of
variables without multicollinearity and the extraction of two valuable factors from the
data set.
BLR analysis of data with the extracted factors indicated that the PMs' LB and the
PMs' contingency situation (represented by PM's LMR, PM's PP, and PM's TS)
moderated each other's effect and determined together the PMs' successful completion of
the DT projects in LICs. Significant relationship existed between PM’s LB and the DT
project completion status, dependent on the contingency situation of the PM leading to
rejection of the null hypothesis for the RQ1 (H011) and acceptance of the alternate
hypothesis (H111) with 95% confidence. In the favorable and unfavorable contingency
situations, the PMs with ROLB did not complete their DT projects successfully as
indicated by the negative and significant (p < .05) regression coefficients for interaction
effects. The results indicated that these two situations depend on PMs with TOLB to
successfully complete the DT projects, supported, and confirmed the application of CTL
theory by Fiedler (1967) for DT projects in LICs in the United States.
Cluster analysis of data revealed the existence of three distinct clusters: (a) PMs
with ROLB who experienced favorable contingency situations but did not complete their
DT projects successfully (Cluster 2), (b) PMs with ROLB who experienced favorable
141
contingency situations and completed their DT projects successfully (Cluster 3), and (c)
PMs who experienced moderately favorable to unfavorable contingency situations whom
54% completed their DT projects successfully (Cluster 1). The significant p-value ( p <
.05) for all variables from the dispersion analysis indicated that all variables contributed
significantly to forming the clusters. Thus, the results led to the rejection of the null
hypothesis (H021) and the acceptance of the alternate hypothesis (H121) for RQ2 with
95% confidence. The favorable contingency situation helped some PMs with ROLB
(PMs in Cluster 2) complete their DT projects successfully but did not help some other
PMs with ROLB (PMs in Cluster 3) complete their DT projects successfully. Results
indicated the possible influence of situational variables other than the three variables
indicated in the CTL on the PMs completing the DT projects successfully.
Further analysis indicated that in Cluster 1 in the unfavorable situation, the
situational variables almost significantly moderated the PMs' LB and decreased the
chances of PMs with ROLB successfully completing their DT projects. In the moderately
favorable situation in Cluster 1, the PMs' LB and the situational control individually
contributed significantly and positively to enhancing the ability of PMs with ROLB to
successfully complete the DT projects. The situational variables and PMs' LB did not
interact significantly in the moderately favorable situation. Seventy-seven percent (77%)
of the PMs in Cluster 1 displayed ROLB. The above results indicated that CTL theory
applies to DT projects in LICs in the United States.
142
Discussion on Findings
This study provided important information regarding the relationship between
PMs' LB and the DT project completion status under the three contingency situation
(favorable, unfavorable, and moderately favorable to the PMs) described in the CTL
model by Fiedler (1967). Most PMs (91%) in this current study who responded to the
survey questions on the Fiedler's LPC self-assessment scale indicated that they were RO
leaders. According to the BLR results, ROLB (high LPC) of DT PMs significantly
decreased the successful completion of DT projects in LICs in favorable and unfavorable
situations of the leader. It is imperative from this study’s results that the TOLB (low
LPC) of DT PMs are preferred and the best LB in favorable and unfavorable situations of
the leader for the successful completion of DT projects in LICs in the United States. BLR
analysis of participants in Cluster 1 indicated that in the moderately favorable situation,
PM’s ROLB and PM’s PP significantly increased the chances of PMs successfully
completing the DT projects in LICs in the United States. This study’s results indicated
that PMs with ROLB have better chances of completing the DT projects successfully in
moderately favorable situations. Also indicated by the results of this study is that in the
favorable and moderately favorable situations, improving the PM’s PP, PM’s LMR, and
PM’s TS would significantly improve the chances of the PMs completing the DT
projects successfully. Overall, the results supported the application of the CTL theory by
Fiedler (1964, 1967) for DT projects in LICs in the United States.
143
According to the cluster analysis results of this study, a group of PMs with ROLB
did not complete their DT projects successfully in favorable situation. In contrast,
another group of PMs with ROLB completed DT projects successfully in similar
favorable situation. When managing complex projects involving DT, some team
members might expect their leaders to engage more in TOLB (i.e., clarifying purpose,
defining goals, setting direction, and training coaching to accomplish DT tasks),
especially at the beginning. When the project cycle progressed, team members might
expect leaders to express more ROLB (i.e., listening, showing interest, consideration, and
autonomy- delegation) from leaders. Therefore, PMs with a socio-independent (middle
LPC) LB might be appropriate to enforce the most effective LB at the right time to
successfully complete the evolving, task-intensive, and volatile DT projects. According
to Fiedler (1967) and Fiedler and Chemers (1984), socio-independent leaders also
performed well in favorable and moderately favorable situations. According to Henkel et
al. (2019), a situational LB approach of PMs with a distribution pattern of the TOLB and
ROLB and not an either-or LB is helpful for successful project completion. Future
research of socio- independent PM’s LB on the successful completion of DT projects
would benefit business leaders in LICs to devise strategies for the successful completion
of DT projects applicable for all situations.
Chemers et al. (1985) in their study have shown that measures of group
effectiveness such as (a) leader's job satisfaction, (b) leaders' job stress, and (c) leaders'
supervisory performance impacted the LB of the leaders, thereby, the strength of the
144
Fiedler’s CTL model. According to Chemers et al. (1985), leaders' member relationships
correlated significantly with leaders' experience of stress with their subordinates and
leaders' other perceived job stress. According to Chemers and Ayman (1985), low LPC
leaders showed a significantly stronger relationship between performance measures and
job satisfaction than high LPC leaders. Fiedler and Chemers (1984) and Fiedler and
Garcia (1987) found that the leaders’ experience and training correlated with the leaders'
LPC scores. According to Albadawi and Salha (2022), males and females differed in
their LB, implying that gender impacts the type of LB; female supervisors scored less
than males on the LPC scale, indicating TOLB, while males were motivated by ROLB.
Female supervisors prioritized getting work done and controlled and directed the
subordinates to accomplish tasks over human relationships than male counterparts;
females also used more power to protect their stand with the subordinates (Albadawi &
Salha, 2022). Scholars and IT practitioners must consider incorporating measures of
group effectiveness (leader's job satisfaction, leaders' job stress, and leaders' supervisory
performance, leader's experience), group members' job satisfaction, and gender of PMs
along with the contingency situational variables into Fiedler's CTL model in future
research of PMs’ LB on successful completion of DT projects in LICs in the United
States.
According to Fiedler (1967), high LPC leaders (RO leaders) behaved more
considerately toward group members in moderately favorable conditions where there was
a need for maintaining relationships and not in favorable or unfavorable situations than
145
low LPC leaders (TO leaders). The TO leaders behaved more considerately than RO
leaders in favorable and unfavorable situations where they felt in control and challenged.
The above results and results from this study indicate that the PM's LB reflected the PM's
values and goals (i.e., the need for task accomplishment and maintaining relationships
with people), which served as the motivational forces behind the PM's actions in a
situation. Many LICs in the United States lack PMs with effective LBs who can lead their
team to the desired level of productivity by providing supportiveness and directness
according to the given situation of subordinates and their level of motivation when
different situations need handling differently since every situation has its characteristics,
as indicated by Fiedler (1967) in CTL.
Henkel et al. (2019), using Fiedler's CTL model and the LPC scale, found that
effective PMs adapt their LB to meet the needs of the project team members and the
situational environment. Their results indicated that most leaders have primary and
secondary LB when influencing team members, and they revealed both TO leadership
and RO leadership behaviors as needed during the life cycle of a project. The goals may
differ for highly volatile and complex DT projects at different stages. When task
interdependence is high, leaders' ROLB is necessary to enhance both the team and
individual processes and outcomes; in contrast, when the task complexity is high,
leaders' TOLB is essential to strengthen both the team and individual processes and
projects (Brown et al., 2021; Warner & Wager, 2019). Future studies using Fiedler's CTL
model
146
to identify the PM's LB requirements at different stages of DT projects would benefit the
business community in LICs in the United States.
Also indicated by the results in this study, is that in the favorable and moderately
favorable situations, improving the situation through improving PM’s PP, PM’s LMR,
and PM’s TS would significantly improve the chances of the PMs completing the DT
projects successfully. Therefore, leaders should possess the skills to manage followers at
different hierarchical levels, stages of projects, and individual characteristics of followers
for DT projects' success. Therefore, always, PMs must understand the complexity of their
DT projects and their employees' skill level and design projects' task structures
effectively. They must set clear project goals, maintain good relationships with team
members, motivate, coach, and control the team to accomplish tasks and successfully
implement the DT projects (Popp & Hadwich, 2018). Past research indicated that no
single situation equally applied to DT projects in all organizations; still, DT projects'
effectiveness in organizations depends on a fit or match between the technology, culture,
people, environmental volatility, and the organizational structure's features (Makhlouf &
Allal-Chérif, 2019). Leaders must understand and handle these factors together for the
successful completion of the DT projects.
Application to Professional Practice
The results of this study are significant to the business leaders of the LICs in the
United States, as they enhance their understanding of the relationship between the LB of
the PMs managing DT projects and the successful completion of the DT projects in the
147
LICs. Approximately 76% of DT projects fail to achieve the desired results, and of the
$1.3 trillion spent on DT in 2018, $900 billion went to waste (Correani et al., 2020;
McCarthy et al., 2024; Reeves et al., 2018) primarily in LICs. LICs operate in the
industrial environment where digital technologies such as software, AI, cloud
computing, IoT, big data, and intelligent manufacturing are major driving forces for
growth, innovation, and product differentiation is a competitive necessity (Ghosh et al.,
2022; Reeves et al., 2018), and survival of LICs depends mainly on their DT capabilities.
The unsuccessful DT project implementation negatively impacts business performance,
leading to significant inefficiency and losses in financial and product quality (C.-H. Lee
et al., 2021). The successful completion of the DT projects is a big challenge as the
volatile and highly competitive business environment calls for effective and proactive
strategies for successfully implementing DT projects in LICs’ for continued viability.
Therefore, this study’s findings on leadership effectiveness for successful DT project
implementation in LICs may provide valuable insight that informs organizational strategy
and business practice.
Using this study’s results, organizations can successfully complete their DT
projects, enhancing their competitive business landscape to grow and expand their
businesses. Successful completion of DT projects can enable organizations to make the
concurrent rise of speed and complexity of processes possible and manageable, quickly
overcome the hard limits to scaling and coordinate efficiently across many organizations
and value chains, break the rigid boundaries between engineering domains and vertical
148
specializations, and leverage data analytics horizontally across siloed categories (Benbya
et al., 2020; Russ, 2021). Successful completion of DT projects also can help
organizations enhance resiliency (Nkomo & Kalisz, 2023), improve competitiveness
(Cennamo, 2021; Kraus et al., 2021a; Llopis-Albert et al., 2021), reduce costs, and
increase revenue (Bush, 2020; Llopis-Albert et al., 2021) in a challenging economic
environment. The successful DT completion and the associated structural transformation
and improved work processes in the LICs will create safer work environments for
employees, especially those with disabilities through smart and enhanced technologies
(Albukhitan, 2020; Llopis-Albert et al., 2021; Sousa & Rocha, 2019).
Results from this study indicated that the DT PMs' project situation, including DT
project tasks' structure set forth by task complexity and project goals, PMs' member
relations set forth by the organizational culture, and the team member's ability to support
their PM per members' skill level, training, and readiness, and the PMs' PP vested by the
organization significantly mediated DT PMs' LB impacting PMs completing the DT
projects successfully. One of the critical success factors for the successful completion of
DT projects is the capabilities of the acting people, especially of the managing people of
the DT projects, the first imperative of DT (Brunner et al., 2023; Dubey et al., 2020;
Klein, 2020; Müller et al., 2024; Shao, 2019). Completing an organization's DT projects
is primarily driven by the project managers' LB and the group members' skills, training,
commitment, and well-being (Leavy, 2020; Nkomo & Kalisz, 2023). Using this study's
results, business leaders in LICs can devise strategies to match the DT PMs' LB and their
149
project situation and enhance team performance and output (supported by studies,
Fernandez-Vidal et al., 2022; Fiedler, 1967; Fiedler & Chemers, 1984; Müller et al.,
2024), resulting in the successful completion of the DT projects leading to
significant growth, expansion, competence, and survival in LICs in the United
States.
This study's results further indicated that the PMs with ROLB could not complete
(indicated by the statistically significant negative coefficients) their DT projects
successfully in the favorable (strong PM's PP, good PM's LMR, and structured or
unambiguous tasks) and unfavorable (weak PM's PP, poor PM's LMR, and unstructured
or ambiguous tasks) situations. Results indicated that these situations require TOLB PMs
to complete DT projects successfully. Thus, there is a need for business leaders in
organizations to identify and match PMs' effective LB per situation for successful DT
project completion for organizational growth and survival (supported by studies [Henkel
et al., 2019; Klein, 2020; Müller et al., 2024]). The potential negative impacts and
substantial losses of unsuccessful DT project completion and the potential benefits of
successful DT project completion create a need for business leaders to have the
knowledge and capabilities to devise strategies to apply leadership match for successful
DT project completion per the leaders' PP, TS, and LMR. Because most of the LICs are
still at the beginning of their DT process, there is a lack of a shared understanding and a
standard model of effective LB of leaders of DT projects related to the leaders' project
situation (Kane et al., 2019; Klein, 2020; Philip & Gavrilova Aguilar, 2022). Business
leaders in LICs can use this study's results to determine the type of PMs' LB required
150
based on the PMs' project situation to complete the DT projects successfully for
organizational growth, improvement, and survival.
This study's results also indicated that the contingency situational variables and
PMs' LB did not interact significantly in situations moderately favorable to the PMs.
However, the PMs' LB and the situation individually contributed significantly and
positively to enhancing the ability of PMs with ROLB to successfully complete the DT
projects. According to this study's results, (a) engagement and participation of team
members with relevant skill sets and expertise are necessary for the complex and volatile
DT project's success (Gilli et al., 2024; Guinan et al., 2019; Radhakrishnan et al., 2022)
and (b) DT PMs must understand the complexity of DT projects and design their task
structures effectively, set clear project goals, maintain good relationships with team
members, and motivate, direct, coach, and control the team members to accomplish tasks
to complete the DT projects successfully (Fernandez-Vidal et al., 2022; Popp &
Hadwich, 2018). Concern over the successful completion of DT projects in LICs is
increasing as the higher levels of internationalization and the bigger dimensions of
businesses that require more advanced stages of DT demand more effective management
characteristics (McCarthy et al., 2024) in LICs. In recent years, business leaders, IT
practitioners, and researchers of DT projects' success emphasized the importance and
influence of leadership in DT projects' successful implementation, suggesting that
leadership, especially in middle project management where the action mostly is
(Nadkarni & Prügl, 2021) plays a crucial role in DT's success and without which the rest
151
of the DT efforts will be rendered meaningless (McCarthy et al., 2024; Nadkarni & Prügl,
2021; Porfírio et al., 2021). This study's results can help business leaders in LICs in the
United States devise effective strategies to improve PMs' DT project situation and
successfully complete their DT projects.
Implications for Social Change
Global technological changes evolve rapidly, transforming operations, customer
preferences, and market shares, necessitating organizations to transform digitally to stay
competitive. Subsequently, DT shapes organizations, work environments, and processes,
creating new management challenges for leaders (Brunner et al., 2023; Müller et al.,
2024). The results from this study might help LICs in the United States to complete the
DT projects on time, within an approved budget, and per quality expectations
significantly enhancing the processing power of LICs in the United States through
increased productivity and revenue (Albukhitan, 2020; Peng & Tao, 2022; Sousa &
Rocha, 2019), reduced cost, enhanced processing efficiency and faster time-to-market,
improved product and service quality, and added value through dedicated services (Sousa
& Rocha, 2019) reshaping industry competition, and allowing these companies to stay
competitive (Albukhitan, 2020; Sousa & Rocha, 2019; Zhai et al., 2022). Subsequently,
the LICs will become more digitally advanced and offer the scope to generate new
projects (Albukhitan, 2020; Sousa & Rocha, 2019), allowing them to transform their
structure of supply chains to deliver advanced higher-quality digital products and
services (Faruquee et al., 2021; Kraus et al., 2021a; Sousa & Rocha, 2019). These LICs
can attract
152
more customers by gaining their trust and increasing market share (Doukidis et al., 2020;
Nasiri et al., 2020; Verhoef et al., 2021).
The improved productivity and efficiency of business processes in these
organizations, and the subsequent higher revenue growth by increased sales (Albukhitan,
2020; Sousa & Rocha, 2019; Zhai et al., 2022), and expanded market coverage (Kraus et
al., 2021a; Zaki, 2019), can enhance their ability to provide services remotely, through
outsourcing and the deployment of virtual customer care centers (Doukidis et al., 2020;
Kraus et al., 2021a) allowing the LICs to effectively manage contingent emergencies
such as pandemics and natural disasters and prevent economic decline by leveraging
digital technologies such as electronic commerce (e-commerce) channels, remote
working, and intelligent manufacturing (Datta & Nwankpa, 2021).
The increased productivity of digital products and services can improve the living
standards of people because of the introduction of advanced services and applications
such as Internet information searches, e-commerce, distance education, digital healthcare,
digital financial and other services, more broadband services, the IoT, AI, and other
innovative products, and social networks. Subsequently, work, commerce, entertainment,
and social interactions will undergo technological transformation, shifting from physical
to virtual platforms and digital technologies. This physical-to-digital transformation will
enable the LICs to secure a better business position, increase productivity, and minimize
financial losses (Zhai et al., 2022), enhancing community living standards.
153
The increased revenue in digitally transformed companies can attract
several high-tech jobs locally and from other regions, thereby significantly
increasing employment creation in the United States. Additional demand for skilled
workers in advanced digital product and service development may result in an
increase in compensation, spending power, and increasing the economy in the
United States.
Improved employee satisfaction, performance growth, and productivity in LICs would
translate to enhanced income and socio-economic empowerment of the employees, their
families, and others in the community (Trenerry et al., 2021). Improved growth and
productivity also translate into enhanced employment effects, thus reducing
unemployment through long-term sustainable employment practices. Improved
employment conditions and well-being can boost employee morale and family
relationships, leading to healthy societies.
Technological changes in organizations impact employees at both personal and
professional levels, necessitating the strategic implementation of DT rollouts and DT
projects effectively managed to lower the impact of technostress on employees,
strengthen resilience, and improve their performance (Nkomo & Kalisz, 2023).
Subsequently, the LB of DT PMs play a critical role in strategizing tasks and influencing
the employees (Bunjak et al., 2022; N. T. Nguyen & Hooi, 2020) and encouraging,
coaching, and motivating employees during the challenge of technological change
(Wolff et al., 2019; Zulu & Khosrowshahi, 2021). Besides, according to Cortellazzo et
al. (2019), some of the most common problems generated by organizations' DT are
worker
154
alienation and weak social bonding; therefore, leaders must support and help followers
deal with the challenges of DT, such as greater autonomy and increased job demands, by
adopting LB, such as coaching, motivating, promoting employee development, and
providing resources for better task handling.
Results from this study can help PMs manage DT projects effectively per
employees' readiness level and situation by strategizing tasks and encouraging, coaching,
and motivating employees, reducing technostress and employee alienation, and
improving employees' performance (Bunjak et al., 2022), driving the continued
successful completion of DT projects. Employee financial stability caused by sustainable
employment practices, improved working and living conditions, and a more innovative
and collaborative culture caused by reduced job stress and advanced digital processing
and communication infrastructures may create highly engaged and motivated employees,
support their families, and actively build a sustainable society.
Recommendations for Action
This study’s results indicated that both the PMs’ LB and the contingency
situation of the PMs (measured as PM’s PP, PM’s LMR, and PM’s TS) significantly
moderated the effect of each other and significantly interactively impacted the successful
completion of DT projects in LICs in the United States. According to this study’s results,
the ROLB of DT PMs significantly decreased the successful completion of DT projects
in LICs in the United States in favorable and unfavorable situations. Therefore, I
recommend the following to the business leaders in LICs in the United States when
hiring, training, and
155
evaluating PMs for the DT projects and when devising strategies for continued successful
completion of the DT projects to improve their business practices.
1. When the situation is favorable to the DT leader, that is, when DT projects
are with tasks that are structured and highly interdependent, DT employees
are trained, with appropriate skill sets, and know what to do, and DT PMs are
vested with relatively strong position power and have the support of their
group members, PMs must be quick to reward and punish team members, act
authoritatively, focused on task completion, give directions and prescribe the
work to team members, and not concerned with or sensitive to the feelings of
their team members to influence and control the team members to get the DT
projects’ tasks completed (Fiedler, 1967; Fiedler & Chemers, 1984); this
situation is best for PMs with TOLB and business leaders from LICs in this
situation must consider hiring or assigning PMs with TOLB (low LPC) to
manage the DT projects.
2. When the situation is unfavorable to the DT PM, that is, when the DT
projects' task complexity is high, and tasks are unstructured, the leader's
authority or vested position power is weak, and the group members are
untrained, beginners, and unable to support their PMs, PMs must be
diplomatic, subject matter experts, able to initiate tasks, clarify purpose,
define goals, forceful and give directions to team members, and enjoy the
challenge; this situation is best, and organizations must consider hiring or
156
assigning PMs with TOLB (low LPC) who would be best at training and
coaching their team members, providing the needed directions, and
influencing the team members on task accomplishments (Fiedler, 1967;
Fiedler & Chemers, 1984; Warner & Wager, 2019).
3. When the situations have mixed problems such as: (i) where the group
members are supportive, have appropriate skills, and trained but the DT
projects’ tasks are relatively ambiguous and unstructured, and the PM’s vested
position power is weak, or (ii) where tasks are structured and precise cut and
the PMs have high position power, but the group members are unskilled,
untrained, and unsupportive; this situation is best for the relationship-
motivated (high LPC) leaders (Fiedler, 1967; Fiedler & Chemers, 1984) and
organizations in this situation must consider hiring or assigning PMs with
ROLB (high LPC) with the following traits, who are diplomatic and
concerned with the feelings and readiness level of group members to get their
corporation for task accomplishment, treat their group members like they are
significant, include team members in decision making, act like a peer with
team members, speak positively, reward team members for their excellent
work ethic but do not punish frequently, and always have an open-door
policy; these strategies can increase employee morale and corporation which
can improve DT productivity (Fernandez-Vidal et al., 2022). According to
157
Fiedler and Chemers (1984), a socio-independent (middle LPC) leader may also
perform well in this situation.
4. According to this study’s results favorable situation represented by strong PP,
good LMR, and high TS can enhance the DT projects’ successful completion.
Therefore, business leaders from the LICs in the United States must ensure the
following strategies are in place to improve the contingency situation of the
DT PMs for continued DT projects’ success.
a. Ensure that the DT PMs continuously improve their LMR through
effective employee communication, such as one-on-one meetings,
corporate meetings, and weekly, monthly, and yearly performance
evaluations. Improved LMR can enhance the PMs’ and the group
members’ satisfaction at work, foster collaboration, enhance productivity,
and lead to the successful completion of the DT projects.
b. Ensure an excellent recognition and reward program exists for the
employees' continued dedication. Reward them with intrinsic and extrinsic
gifts to motivate them to perform their best when accomplishing tasks to
complete DT projects successfully. Also, render appropriate position
powers to DT PMs to reward or punish the employees; this can keep the
PMs feeling in control of the situation.
c. Ensure a continuous performance monitoring program exists for the PMs
managing the DT projects, along with a clear articulation of goals,
158
objectives, and expected outcomes, for evaluating and monitoring their
projects’ TS and performance.
d. Ensure a coaching and mentorship program exists to develop DT
employees' DT-related skills and practices and encourage their
creativity and willingness to contribute to the organization's DT
objectives.
5. I further recommend that the business leaders of LICs in the United States
have scales or indicators to measure the LB of PMs who manage DT projects
and monitor their LB continually. I strongly suggest using Fiedler’s (1967)
LPC scale for measuring PM’s LB.
The findings of this doctoral study will be disseminated in the following ways to
enhance knowledge among the business and academic community. Upon request for
consideration and potential adoption within business practice, the participants will
receive a high-level summary of the findings. Walden University will publish the results
of this study in the University’s ScholarWorks and ProQuest databases for the benefit of
future scholars. Along with my doctoral study chair and second committee member, I
will work to publish this study’s research results in a peer-reviewed, reputable journal.
Recommendations for Further Research
While several limitations existed within this doctoral study, other researchers may
expand on the findings and contribute to the literature. I have discussed some ways in
which future researchers can expand the findings of this study to contribute to the
literature and improve the knowledge of leadership for successful DT project completion.
159
First, in this study, I investigated the effect of PM's LB, and the contingency situational
variables Fiedler (1967) introduced, namely, PM's LMR, PM's PP, and PM's TS, on
completing the DT projects successfully. Future research could broaden the scope of this
study by including additional variables such as PMs' education, experience, gender, job
stress, and job satisfaction in a quantitative study within the same CTL framework that
may explain additional variation in the dependent variable, successful completion of the
DT projects in LICs in the United States. Investigators from previous studies have shown
that measures of leaders' group effectiveness such as (a) leader's job satisfaction, (b)
leaders' job stress, and (c) leaders' supervisory performance (Chemers & Ayman, 1985;
Chemers et al., 1985; Fiedler & Chemers, 1984), leaders' gender (Albadawi & Salha,
2022), and leaders' experience and training (Fiedler & Garcia, 1987) correlated with the
leaders' LB and impacted the leaders' project management capabilities. Adding these as
additional independent variables may provide more insights and structural models about
the influence of the PM's LB on the successful completion of the DT projects in LICs in
the United States, applicable for all contingency situations of the PMs (favorable,
unfavorable, and moderately favorable) and at different stages of the DT project
(beginning, middle, end) to help build dynamic capabilities for DT.
Second, in this study, I investigated the contingency leadership of the PMs of DT
projects using the CTL by Fiedler (1967) as the framework. Future studies should
investigate the effect of other leadership styles of PMs on DT projects, such as
transformational leadership of PMs using the transformational leadership theory by Burns
160
(1978) and situational leadership of PMs using the situational leadership theory by
Hersey and Blanchard (1970, as cited in Benmira & Agboola, 2021), along with the
contingency leadership of the PMs using Fiedler's (1967) CTL in LICs in the United
States. Situational leaders are flexible and change and adapt their LB to meet the needs
of employees and display a distribution pattern of the TOLB and ROLB and not an
either-or LB to manage projects depending on the followers' maturity and readiness
levels (Henkel et al., 2019). Transformational leaders efficiently manage external crises
and events that could lead a company to transform, such as forceful DT (Philip, 2021),
and inspire followers using the strength of their vision, personality, and charismatic
behavior, motivating them to work toward completing the DT project's tasks for
successful transformations (Schiuma et al., 2022). Together or separately, these different
leadership styles may help PMs manage and complete DT projects effectively at different
stages, with team members' diverse skills, readiness levels, and varying contingency
situations.
Third, the selection of samples in this study was limited to the industrial sector
within the United States. As DT impacts various other sectors such as sales (Alavi &
Habel, 2021), pharmaceutical (Kulkov, 2021; Ma et al., 2023), food and beverage (I. Ali
& Aboelmaged, 2022), banking and financial services (Tsindeliani et al., 2022),
healthcare (Kraus et al., 2021b), and many more, investigating the influence LB of PMs
on successful completion of DT projects including these sectors may be of increasing
relevance and concern for both scholars and practitioners in the field; future research
161
should leverage the insights from this study in other sectors within the United States and
a broader global context.
Fourth, DT imposes tremendous challenges not only for individual companies
but also for national economies (Švarc et al., 2021). The global scale issues make DT
complex and difficult to comprehend (Ćukušić, 2021; Hanelt et al., 2021). At the same
time, DT also presents tremendous opportunities for growth at the global level, where
digital societies and smart cities interact to improve the lives of people of all nations and
benefit national governments and global companies (Ćukušić, 2021; Hanelt et al., 2021).
Additional research involving more extensive samples of participants from around the
globe might help enhance the national economy in the United States and help validate
this study’s findings.
Fifth, the LB of the PMs managing the DT projects was analyzed only through
the lens of the PMs who managed or were managing the DT projects during the data
collection for this study. Future research should include the perceptions of business
leaders in the supervisory positions of the DT PMs and the project team members in the
subordinate positions of the DT PMs, of the PMs' LB on the successful completion of the
DT projects. The followers' and supervisors' perceptions of the PM's LB may differ and
have different influencing effects (Philip, 2021). Montenegro et al. (2021) found that
PMs' internal and external stakeholder relationships linked to PMs completing
construction projects successfully. Future research, including the PMs' supervisors'
perceptions, PMs' perceptions, and group members' perceptions interactively on the DT
162
PMs' LB in a mixed method study approach, may provide significantly better insight into
the effective PM's LB (Müller et al., 2024) for the DT projects' successful completion and
help devise effective strategies for the DT's success at varying project cycles and
circumstances for organizational growth and survival in the LICs in the United States.
Sixth, there is little systematic insight into the application of DT in the public
sector. Very little is known about the effects of PMs’ LB on the successful
implementation of DT projects in the public sector (Demircioglu & Chowdhury, 2021),
although the effect of LB on the successful completion of DT projects in the public
sector is becoming critical. DT initiatives in the public sector can help government
agencies deliver high-value, real-time digital services and satisfy the growing people's
expectations for advanced digital services in the public sector (Mergel et al., 2019). In
response to people's changing expectations, governments increasingly invest in DT to
improve public service delivery (Mergel et al., 2019). Finally, future studies could also
encompass controlled conditions in an experimental or quasi-experimental design to gain
better insights into the causal agents of the successful completion of DT projects.
Reflections
The DBA journey significantly shaped my personal, academic, and career life and
fulfilled my dream of a doctorate. I realize that accomplishing a DBA degree is a
significant life achievement. I share my reflections as a testament to the power of
perseverance, resilience, commitment, and hard work in navigating the doctoral journey.
The DBA journey was a very fulfilling and rewarding self-discovery that widened my
163
mind, enriched my intellectual faculties, and transformed me emotionally and spiritually
to be a better and able person who could be a positive example for future scholars.
The doctoral journey has provided a wealth of knowledge and valuable and
rewarding experience, a mix of scholarly and practical, and enhanced my positive
energy, determination, and dedication toward accomplishing anything challenging but
rewarding. I further realize that pursuing a DBA degree requires an unrelenting spirit, a
healthy appetite for learning, a positive mindset, and commitment to the process. The
valuable and suitable support network full of enthusiastic and knowledgeable mentors,
loving family members and friends, and colleagues I gathered during the journey
facilitated and significantly contributed to a positive outcome. I recognize that my
mental, emotional, and physical health has been intrinsically linked and contributed to
my success.
This doctoral study gave me significant knowledge, insights, and comprehension
of the research topic. Through this study, I gained significant knowledge and insights into
leadership’s importance and value in the success of DT projects. Further, the doctoral
journey enhanced my research, learning, analysis, communication, and presentation skills
and contributed to my professional development as a scholar-practitioner. I intend to
disseminate the knowledge and skills acquired through teaching and mentoring scholar-
practitioners and informing business leaders, project managers, and other stakeholders in
the industry. Through applying quantitative research methodology, I learned to collect
large amounts of reliable data from multiple participants using existing instruments,
carefully checked their reliability and validity, and inferred the results from the data to a
164
broader audience. I learned to use survey tools for data collection for a doctoral study
through anonymous surveys, thereby minimizing researcher bias; through conducting an
anonymous survey, I ensured that my beliefs did not influence the study’s findings,
promoting objectivity in the research processes. By practicing constant reflexivity and
carefully self-examining biases and assumptions during design, analysis, and
interpretation, I further eliminated preconceived personal biases creeping into the
research.
I learned to analyze large amounts of data meticulously through advanced
statistical methods, test pre-defined hypotheses, confirm research theories, and answer
research questions, which enhanced my critical analysis, logical reasoning, and
presentation skills and provided critical guidance in producing trustworthy predictions. I
became exposed to choosing the proper methods to collect the data, employ the correct
analysis methods, perform stringent statistical analyses, and effectively present the
results. Statistics allowed me to make decisions based on data, make predictions from the
results, learn from the data reliably, use significance levels to differentiate between
reasonable and erroneous conclusions, challenge assumptions, and understand the subject
much more deeply.
Conclusion
Global technological changes evolve rapidly, transforming operations, customer
preferences, and market shares, necessitating organizations to transform digitally to
stay competitive. LICs operate in the industrial environment where globalization, rapid
165
technological advancement, and subsequent product differentiation are a competitive
necessity. According to a recent trend, over 70% of the DT initiatives failed in LICs,
negatively impacting profitability, competitive advantage, and companies' sustainability,
demonstrating that most LICs lack DT competency. As found within this research study,
successfully completing an organization's DT projects is primarily driven by a match
between the PMs' LB and the project situation, including DT project tasks' structure set
forth by task complexity and project goals, PMs' member relations set forth by the
organizational culture, and the team member's ability to support their PM per members'
skill level, training, motivation, and readiness, and the PMs' PP vested by the
organization. Using the results of this study might assist business leaders in improving
the successful completion of DT projects that can enhance economic growth and
profitability, deliver enhanced digital products and services to communities, and make
informed decisions about successful digital leadership aligned with business
improvement and continuity. Successful business leaders might return value to society
through stable and increased employment in the sector, improved financial health, and
better quality of people's lives.
