689
1 INTRODUCTION
Global shipping is operating in a rapidly changing
environment in which innovation, competitiveness,
and networking have emerged as critical determinants
of long-term sustainability. The concentration of firms
and institutional actors around geographic and
functional maritime hubs has fostered the
development of so-called “maritime clusters,” which
generate added value for participating organizations
and for the wider maritime ecosystem [1], [2].
Greece possesses the world’s largest merchant fleet
in terms of deadweight tonnage (dwt), and shipping
constitutes the country’s most strategic export-oriented
sector. Nevertheless, the concept of an organized
maritime cluster remains relatively underdeveloped
domestically, characterized by limited institutional
connectivity and the absence of integrated cluster
management. The Maritime Hellas initiative,
established by institutional stakeholders in 2016, seeks
to address this gap by promoting cooperation, outward
orientation, and a unified representation of the Greek
maritime sector.
Against this background, the present study aims to
assess the role, effectiveness, and prospects for
strengthening Maritime Hellas, drawing on the views
of industry professionals and company
representatives. Using a quantitative approach based
on descriptive statistical analysis of questionnaire data,
the study examines key trends, expectations,
Evaluating the Greek Maritime Cluster: The Maritime
Hellas Initiative and Its Implications for Cluster Policy
V. Korkidis
1
& M. Boviatsis
2
1
University of Piraeus, Piraeus, Greece
2
University of Athens, Athens, Greece
ABSTRACT: This paper evaluates the Greek maritime cluster through the case of the Maritime Hellas initiative,
examining its structure, effectiveness, and strategic role within the broader maritime ecosystem. Drawing on
cluster theory and international maritime cluster models such as Singapore, Rotterdam, and Oslo, the study
investigates how governance, institutional coordination, related industries, advanced factor conditions, and
competitive pressures influence perceived value and formal participation within the Greek maritime cluster.
Using a quantitative cross-sectional methodology based on 574 responses from maritime stakeholders, the
research applies descriptive statistics, structural modelling, and Firth penalized logistic regression to explore the
determinants of perceived support and registration within Maritime Hellas. The findings reveal that advanced
factor conditions, competitive pressure, and firm size positively affect perceived value, while institutional
maturity and related-supporting industries may reduce the perceived incremental value of intermediary
coordination mechanisms. Formal participation is primarily associated with governance quality, advanced
capabilities, and organizational capacity. The study highlights the importance of institutional trust, targeted
upgrading services, and differentiated cluster policies for strengthening the competitiveness and sustainability of
the Greek maritime ecosystem.
http://www.transnav.eu
the International Journal
on Marine Navigation
and Safety of Sea Transportation
Volume 20
Number 3
September 2026
DOI: 10.12716/1001.20.03.16
690
challenges, and proposals associated with the
initiative’s operation and, more broadly, with the
development of the Greek maritime cluster.
2 THEORETICAL BACKGROUND
The concept of collaborative formations, or “clusters”,
has been the subject of extensive academic debate over
recent decades, particularly since Michael Porter (1990,
1998) introduced the argument of spatially grounded
competitive advantage. At the core of this approach
lies the proposition that the geographic and thematic
concentration of complementary firms and institutions
creates a system that enhances efficiency, innovation,
and sectoral growth dynamics. Cluster theory has
influenced not only industrial policy but also national
and regional competitiveness strategies, emphasizing
the idea of a broader “value ecosystem” within which
firms, institutions, and enabling conditions to co-
evolve [1], [3].
2.1 Definition and Structure of Clusters
The term cluster refers to a geographically
concentrated agglomeration of firms and institutions
operating within a specific domain and developing
simultaneous relationships of cooperation and
competition. What differentiates a cluster from a mere
co-location of firms is the degree of interconnection,
knowledge exchange, and capacity for collective
action. More mature clusters typically include core
industry firms alongside specialized supporting actors,
such as legal services, financial intermediaries, and
logistics providers, as well as education and research
institutions and public or intermediary organizations
that facilitate coordination and upgrading processes
[4].
Sölvell et al. (2003) describe a cluster as a
“cooperation system” grounded in mutual trust, while
the European Commission (2016) defines clusters as
groups of interconnected firms and institutions that
collaborate and enhance competitiveness through
synergies. Across these perspectives, particular
emphasis is placed on the dynamic relationship
between firms, innovation, and knowledge—features
that distinguish clusters from traditional industrial
zones or conventional trade associations [5], [6], [7].
2.2 Advantages and Operating Dynamics of a Cluster
Clusters can generate multiple benefits at both national
and local levels. First, they create positive externalities,
including lower information costs, accelerated
diffusion of innovation, and greater availability of
specialized human capital. In addition, clusters
strengthen firms’ ability to innovate and adapt to
changing market conditions by improving access to
new technologies, universities, and research centers
[1].
Geographic proximity also supports faster and
more effective synergy formation, reducing response
times and facilitating collective representation of
sectoral interests. At the same time, clusters can
exercise greater influence on public policy, as they may
act as organized collective actors with a structured
strategic agenda [8].
In shipping, clusters can enhance national
competitiveness by providing an institutional
environment in which the supply chain, ports, financial
institutions, and marine insurance services interact
systematically. Cases such as the Rotterdam Maritime
Cluster, Maritime Singapore, and the Hamburg
Maritime Cluster illustrate how strong cluster
configurations can attract investment, stimulate
entrepreneurship, and strengthen the international
profile of a maritime economy [9], [10].
2.3 Barriers and Constraints to Cluster Development
Despite their documented benefits, clusters are not free
from challenges. A frequently cited weakness concerns
the absence of a cooperative culture. In contexts
characterized by intense competition and distrust, the
development of meaningful collaboration remains
limited. This is often observed in traditional sectors
such as shipping, where firms tend to operate
autonomously and rely heavily on individualized
bargaining and relationship-based practices [11].
A further barrier relates to bureaucracy and
information asymmetries between the private and
public sectors. Many clusters fail to function
strategically due to weak institutional representation
or limited administrative capacity. Moreover, the
absence of an integrated development strategy and the
fragmentation of initiatives can undermine long-term
planning and systematic performance evaluation [12],
[6].
Equally important is the difficulty of effectively
linking clusters with academia and research. Low
innovation absorption limited joint research projects,
and the lack of mechanisms for knowledge transfer
constitute structural deficits that restrict the long-term
competitiveness of collaborative formations,
particularly in smaller countries such as Greece, where
scale and coordination constraints can be more
pronounced.
3 MODELS AND INTERNATIONAL
APPROACHES TO MARITIME CLUSTERS
The study of successful international maritime cluster
cases offers valuable insights for policy design and for
strengthening competitiveness at the national level.
Internationally, several clusters have evolved into
global maritime hubs by integrating multiple
functions, ranging from shipowning and port services
to education, finance, innovation, and technology. The
performance of these systems is not incidental; rather,
it reflects sustained institutional planning, cross-
sectoral collaboration, and continuous investment in
research and development.
3.1 The Case of Maritime Singapore
Singapore constitutes a leading example of a successful
maritime cluster in Asia and globally. The emergence
of Maritime Singapore is widely understood as the
strategic outcome of long-term planning that began in
691
the 1970s, when the government prioritized maritime
activity as a core pillar of national development. The
cluster comprises more than 5,000 firms, spanning
shipping and logistics companies, legal and financial
services, ship inspection and certification providers,
R&D centers, and higher-education institutions [2].
The Maritime and Port Authority of Singapore
(MPA) functions not only as a regulator but also as an
active policy strategist and an accelerator of
innovation. Through initiatives such as the Maritime
Cluster Fund and the Port Innovation Ecosystem
Reimagined at NUS (PIRE@NUS), Singapore has
supported the adoption of technologies including
artificial intelligence, document digitalization, and
autonomous shipping applications [13]. At the same
time, strong emphasis is placed on human capital
development through certified professional training
programs delivered in collaboration with the National
University of Singapore.
Overall, Maritime Singapore is anchored in a model
of governed cooperation, in which public
administration, industry, and academia act in a
coordinated manner under a shared strategic umbrella.
Public intervention is primarily facilitative rather than
intrusive, aiming to ensure efficiency and long-term
orientation in cluster development. The outcome is a
highly outward-looking and innovation-driven
maritime economy that has integrated a substantial
share of global maritime trade into its service
ecosystem [14].
3.2 The Rotterdam Maritime Cluster
The Netherlands, and particularly the Port of
Rotterdam, constitutes a European benchmark for an
integrated maritime cluster. Rotterdam is the largest
port in Europe and serves as the nucleus of an
extensive cluster comprising more than 1,800 firms,
academic organizations, and innovation actors. The
cluster’s development model is frequently described in
terms of a maritime industrial complex, reflecting the
close integration of industrial production, logistics,
energy infrastructure, and maritime services [10].
The strategic agenda of the Rotterdam Maritime
Cluster is commonly structured around four core
pillars: (a) technological innovation, (b) environmental
sustainability, (c) embeddedness in local society, and
(d) cross-sectoral collaboration. Innovation ecosystems
such as RDM Rotterdam and the Maritime Innovation
Quarter host research programs in areas including
offshore energy, ship robotics, and smart port systems.
Collaboration with institutions such as Erasmus
University and Delft University of Technology further
supports technology transfer and the diffusion of
applied knowledge into industry practice [15].
Cluster governance in Rotterdam is polycentric but
effectively coordinated through the Rotterdam
Maritime Board, which includes representatives from
municipal authorities, port institutions, universities,
and the business community. A key innovation is the
application of the Quadruple Helix model, which
incorporates civil society actors and places emphasis
on participation, transparency, and the social
legitimacy of port-related policies [16]. Taken together,
Rotterdam functions as an integrated hub of sea–land
connectivity, digital transition, and green growth, with
performance underpinned by institutional stability,
specialized knowledge concentration, and adaptive
capacity in response to global shifts.
3.3 The Oslo Maritime Cluster
The Norwegian case and especially, the Oslo Maritime
Cluster, illustrates the role of advanced technology and
strong private-sector engagement in the successful
development of a cluster. Norway is recognized for
integrating maritime activity with energy and
technology sectors, creating a multi-dimensional
cluster extending from shipbuilding and marine
insurance to renewable energy and offshore
technology [17].
The cluster’s core is associated with the GCE Blue
Maritime Cluster (headquartered in Aalesund), while
its footprint extends nationally through the
participation of more than 220 firms and 12 research
centers. R&D investments exceed 6% of turnover, and
firms participate actively in the design and delivery of
education and training programs in collaboration with
the Norwegian University of Science and Technology
(NTNU) [18].
Innovation is reflected not only in technical
domains, such as hybrid propulsion, autonomous
navigation, and emissions-reduction technologies, but
also in governance arrangements, including the
establishment of thematic task forces aligned with
strategic priorities. Importantly, corporate
participation tends to be substantive rather than
symbolic, supporting the durability of collaboration
and collective upgrading. In this way, the Oslo
Maritime Cluster has contributed to positioning
Norway as a global leader in “green shipping,”
supported by responsible governance, active
participation, and technological capability [19].
3.4 Comparative Assessment
A comparison of the above clusters highlights critical
characteristics that distinguish them from less
successful initiatives. First, institutional stability and
political commitment to long-term strategic planning
emerge as a common denominator. In each case, public
administration operates less as a bureaucratic
interventionist mechanism and more as a facilitator
and accelerator of collaboration and innovation.
Second, proximity and structured interaction
between firms and academic actors are decisive.
Knowledge, education, and innovation are treated not
as external inputs but as integrated components of the
cluster’s daily functioning. Third, governance is
organized around a clear vision and a framework of
shared objectives, with active participation by firms,
local authorities, universities, and (where applicable)
civil society.
Overall, international experience suggests that
successful maritime clusters are not defined merely by
the co-location of firms; rather, they constitute full
ecosystems characterized by administrative capacity,
institutional coherence, and cross-sectoral dynamism.
These features set demanding benchmarks for the
development of the Greek maritime cluster and, in
692
particular, for Maritime Hellas, which is examined in
the following section.
4 THE MARITIME HELLAS CASE: STRUCTURE,
FUNCTION, AND STRATEGIC SIGNIFICANCE
OF THE GREEK MARITIME CLUSTER
Greece hosts one of the world’s most prominent
maritime clusters, not only due to its shipowning
capacity but also because of the concentration of a
complex, functionally differentiated, and outward-
oriented maritime ecosystem. “Maritime Hellas” is not
geographically confined to Piraeus; rather, it extends
across multiple cities and regions, encompassing hubs
of business activity, education, expertise, and
institutional support. As such, the Greek maritime
cluster constitutes a nationally scaled and strategically
significant mechanism for economic dynamism,
technological innovation, and geoeconomic influence
[20].
4.1 Structure and Geographic Scope of the Greek
Maritime Cluster
Maritime Hellas comprises a dense web of
interconnected subsystems, including shipowning
companies, maritime service providers, banking and
insurance institutions, technical and legal firms,
shipyards, port authorities, maritime education
institutions, and research centers. It should be noted
that the Greek maritime fabric encompasses more than
1,500 companies and organizations, with activity
extending to cities such as Piraeus, Athens,
Thessaloniki, Volos, Heraklion, Chania, Patras, and
Kavala, areas that concentrate specialized services and
maritime human capital [20].
The cluster’s structure is polycentric, albeit with a
strong core in the Attica region. Shipowning firms are
mainly located in Athens and Piraeus, while
shipbuilding and technical services extend to
Skaramangas, Syros, Elefsina, and shipyards in
Northern Greece. Meanwhile, education and research
centers are distributed across several regions (e.g., the
University of the Aegean, the University of Piraeus, the
Merchant Marine Academies of Macedonia and
Epirus).
Notably, the Greek cluster also exhibits a
“diasporic” element: many Greek maritime enterprises
maintain offices in major international maritime
centers (e.g., London, Dubai, Singapore), forming a
transnational network of knowledge and business
intelligence [21]. Maritime Hellas, therefore, should not
be understood as a local cluster with fixed geographic
boundaries, but rather as a networked, transnational
system headquartered in Greece and operating on a
global scale.
4.2 Strategic Importance and Economic Contribution of
the Cluster
The strategic significance of the Greek maritime cluster
arises from a set of structural and systemic parameters.
First, its economic contribution is substantial. Greek
shipping accounts for more than 7% of GDP and
represents approximately 60% of the European Union’s
total shipping capacity [22]. Shipping is also a leading
source of foreign exchange inflows, traditionally
exceeding tourism—and is directly linked to the
creation of thousands of jobs, both directly and
indirectly.
Second, accumulated know-how and managerial
expertise, built over decades of operation in
international markets, reinforce the global character of
Maritime Hellas. The Greek fleet is a leader in key
segments such as tankers and bulk carriers, while the
operational performance of Greek ship management
often compares favorably with international
benchmarking practices [23], [24].
Third, Greece’s geoeconomic location, combined
with infrastructure investments (e.g., COSCO’s
investment in Piraeus and developments linked to the
port of Alexandroupolis), positions the country as a
connective gateway between East and West in global
trade. Greece is a critical link in European transport
corridors (TEN-T), with the cluster supporting not only
maritime transport but also broader connectivity
through port, rail, and logistics infrastructure.
Although Maritime Hellas has not been formally
institutionalized as a unified cluster by the state, it
operates in practice as a functional value chain in
which specialization, spatial proximity, cultural
cohesion, and historical continuity reinforce external
economies of scale and its international reach [6], [25].
4.3 Contemporary Dynamics and Emerging Challenges
Today, the Greek maritime cluster faces a set of
complex challenges related both to internal
organization and external competition. The absence of
an institutionalized national cluster governance
framework creates coordination difficulties among
firms, institutions, and educational actors.
Collaboration among companies often remains ad hoc
or reliant on personal networks rather than being
supported by strategic guidance or formal institutional
mediation [26].
In addition, the technological and green transition
generates mounting pressures for new skills, increased
R&D investment, and the transformation of established
business models. While notable initiatives exist (e.g.,
Bluegrowth, Isalos.net, and university-based MSc
programs), university industry linkages remain
limited, as does the systematic integration of ESG
principles into business strategy [27].
At the same time, intensified international
competition from emerging maritime hubs (e.g., China
and South Korea), combined with European Union
pressures for greener and more transparent value
chains, requires the Greek cluster to re-position itself
within the global maritime landscape through deeper
collaboration, stronger innovation capacity, and
enhanced institutional maturity.
5 RESEARCH METHODOLOGY
The systematic study of a dynamic production system
such as the Greek maritime cluster (“Maritime Hellas”)
requires a research design capable of capturing
stakeholders’ perceptions, evaluations, and
693
behavioural intentions/commitments in a manner that
is both comparable and reproducible. To this end, the
present study adopts a quantitative, cross-sectional
research design based on a structured questionnaire
and Likert-type scaled variables (1 = low/disagree to 5
= high/agree). This approach enables, first, a
descriptive profiling of respondents’ assessments and,
second, an inferential examination of relationships
between theoretically grounded cluster dimensions
and key outcomes [28], [29].
5.1 Theoretical framing and conceptual mapping
(Porter/cluster theory)
The study is grounded in contemporary cluster and
competitiveness theory (Porter/diamond logic), as
further developed in the literature on regional
competitiveness, cluster structure, and mechanisms of
collective action. Recent research emphasizes that
ecosystem performance depends not merely on the
“co-location” of activities, but also on the cluster’s
composition, cross-industry linkages, and the presence
of institutions and organizational mechanisms that
facilitate coordination, knowledge diffusion, and
upgrading [30], [31].
In line with this perspective, questionnaire
variables are organized into three functional “cluster
conditions” dimensions that reflect core drivers of
competitiveness and collective effectiveness:
− F1 - Cluster Governance / Institutional Thickness:
institutional density, coordination capacity,
collective action, governance quality, and the
intermediary mechanisms (cluster intermediaries)
that organize collaborations and joint initiatives.
The literature highlights the role of cluster
initiatives and intermediary organizations as
“hubs” that reduce coordination frictions and
support upgrading processes [32].
− F2 - Related & Supporting Industries Base: the
presence and functionality of related and
supporting industries (e.g., specialized services,
suppliers, infrastructure, complementary activities)
that strengthen agglomeration economies and
cross-sector connectivity within the cluster [31].
− F3 - Advanced Factor Conditions: advanced inputs
such as knowledge, skills, innovation/technology,
and access to information/finance, which act as
drivers of upgrading and differentiation.
Contemporary cluster research links
competitiveness to such “advanced factors” beyond
basic inputs [30].
In addition, the PORTER variable (competitive
pressure) captures the perceived environment of
competitive intensity. Conceptually, competitive
pressure may either activate collective mechanisms
(e.g., collaboration for joint upgrading) or prompt
strategic distancing; hence, it constitutes a critical
exogenous force shaping stakeholders’ behaviours and
attitudes within the ecosystem [30], [31].
5.2 Research design: from descriptive profiling to
inference
The analytical design follows two stages:
1. Descriptive profiling of assessments (distributions,
central tendencies, dispersion) to establish an
overall picture of the ecosystem/cluster.
2. Explanatory (inferential) modelling, examining the
relationships between cluster conditions (F1 - F3),
competitive pressure (PORTER), and two key
outcomes:
a. perceived support/benefit from Maritime Hellas
(support_composite), and
b. registration/institutional commitment
(registered, binary outcome).
This structure is consistent with contemporary
methodological approaches to ordinal data and scaled
measurements, where descriptive analysis is treated as
a necessary precursor to modelling [28].
5.3 Purpose and quantitative research questions
The study aims to examine quantitatively how
perceptions of the operating conditions/maturity of the
Greek maritime cluster, together with competitive
pressure, relate to: (i) evaluations of the
support/benefit provided by Maritime Hellas and (ii)
the probability of registration as an indicator of formal
participation/commitment to the initiative.
The research questions are formulated in testable
quantitative terms:
− RQ1 (Support model): To what extent do ecosystem
dimensions F1 - F3 and PORTER competitive
pressure predict support_composite, controlling
firm characteristics (e.g., employees,
company_age)?
− RQ2 (Registration model): Which factors
(support_composite, PORTER, selected dimensions
F1/F3, and controls) are associated with the
likelihood of registration (registered = 1)?
− RQ3 (Mechanism/indirect logic): Do the data
indicate that cluster conditions (F1 - F3) and
competition (PORTER) translate into perceived
value (support) and, subsequently, into
institutional activation (registration)?
5.4 Population, sampling frame, and sample
The target population comprises firms and
organizations operating within the broader Greek
maritime ecosystem (shipping/ship management,
shipbuilding and repair activities, and supply-chain
and supporting service providers).
The sampling frame was constructed from the list of
firms registered with the Piraeus Chamber of
Commerce and Industry (PCCI), focusing on categories
connected to maritime and related productive
activities. The institutional role of PCCI as a chamber
and a provider of business registry and information
services in Piraeus is documented in the organization’s
official information, as well as in its sectoral and
research initiatives mapping maritime activities in the
wider Piraeus area (PCCI, n.d.; PCCI/INEMY, 2017).
The final analytical sample includes N = 574 valid
responses, supporting statistically robust descriptive
analysis and the estimation of multivariate models
with controls under standard assumptions and data-
quality tests.
694
5.5 Data collection instrument: questionnaire structure
and rationale
Data were collected using a structured 40-item
questionnaire, designed to capture, comparably and
reproducibly, stakeholders’ perceptions regarding (a)
critical “operating conditions” of the cluster and (b)
evaluation/commitment toward a coordinating
intermediary organization (Maritime Hellas).
Questionnaire development was informed by
contemporary theoretical and empirical work on
clusters and on measuring competitiveness/cluster
strength, drawing on the recent literature on cluster
composition and performance [30], [31], as well as
European initiatives for the mapping and
documentation of clusters [7].
5.5.1 Question types and measurement scales
Most items were closed-ended, primarily using a
five-point Likert scale (1 = low/disagree; 5 =
high/agree), alongside multiple-choice questions
capturing organizational characteristics. A limited
number of open-ended questions were also included to
elicit more complex views and policy suggestions;
these are used in a supplementary manner (e.g., to
support interpretation and to enrich the descriptive
framing of the quantitative findings). The use of five-
point Likert scaling is consistent with established
methodological guidance for analysing Likert-type
data in both descriptive statistics and multivariate
modelling, particularly in large samples [33], [34].
6 THEMATIC AXES AND CONCEPTUAL
COHERENCE
To ensure construct coherence and to generate data
corresponding to distinct, yet theoretically related,
dimensions of the ecosystem, the questionnaire items
were organized into five thematic axes:
1. Cluster structure and composition (relationships,
collaborations, institutions, coordination
mechanisms)
2. Innovation and technology (digital transformation,
R&D, ESG as a framework for change/upgrading)
3. Education and human capital (skills, supply/quality
of human capital)
4. Institutional and regulatory framework (regulation,
institutional support, governance)
5. Development, outward orientation, and strategic
outlook (upgrading, internationalization, collective
competitiveness)
This architecture aligns with more recent
scholarship that conceptualizes clusters as systems
whose performance depends on (i)
institutional/organizational coordination capacity, (ii)
the presence of related and supporting industries, and
(iii) advanced factor conditions [30], [31]. It is also
consistent with the European framework for cluster
documentation and “mapping” used to assess cluster
strength and inform cluster policies [7], [23].
6.1 Hypotheses (H1 - H6) Porter/cluster-theory wording
The empirical analysis tests the following hypotheses:
− H1 (Institutional Thickness & Governance →
Support/Benefit). Perceived institutional thickness
and governance quality of the Greek maritime
cluster (F1) is positively associated with perceived
support/benefit and satisfaction from Maritime
Hellas (support_composite). This expectation
follows cluster theory, where strong institutions
and coordination strengthen collective action and
upgrading mechanisms [30], [31].
− H2 (Related & Supporting Industries →
Support/Benefit). Perceived adequacy of related
and supporting industries (F2) is positively
associated with support_composite, as dense
within-cluster linkages facilitate knowledge
spillovers and collaboration and increase
“absorption” of collective services and initiatives
[35].
− H3 (Advanced Factor Conditions →
Support/Benefit). Advanced factor conditions (F3)
(e.g., human capital, specialized
knowledge/infrastructure) are positively associated
with support_composite, because they enhance the
ability to leverage cluster networks and initiatives
[36].
− H4 (Porter competitive pressure →
Support/Benefit). Competitive pressure (PORTER)
is positively associated with support_composite:
the stronger the competitive pressures, the greater
the perceived value of collective mechanisms
(cluster platform) for access, coordination, and
upgrading [30].
− H5 (Support/Benefit → Registered). Higher
perceived support/benefit and satisfaction from
Maritime Hellas (support_composite) increases the
likelihood that an organization is registered
(registered = 1), reflecting organizational
commitment and integration into the cluster
platform.
− H6 (Direct and indirect effects on Registered).
Cluster conditions (F1 - F3) and competitive
pressure (PORTER) affect registration (registered)
directly and/or indirectly via perceived benefit
(support_composite), controlling core
organizational characteristics (e.g., firm size/age).
6.2 Measures
6.2.1 Ecosystem dimensions (eco*
→
F1, F2, F3)
The eco* items capture perceived evaluations of the
Greek maritime ecosystem across multiple
components, consistent with the cluster-
competitiveness logic and the “upgrading” role of
institutions, factor conditions, and related industries
[30], [31].
− F1 - Institutional Thickness & Governance
− Items (8): eco35, eco36, eco44, eco50, eco53,
eco64, eco65, eco67
− Scale: 1 - 5 (low/disagree to high/agree)
− Aggregation (descriptives): item mean (row-
mean)
− Reliability: Cronbach’s alpha [37]
− Modelling note: standardized factor scores
(F1_z) from the three-factor solution were used
in the inferential models.
− F2 - Related & Supporting Industries
− Items (4): eco28, eco31, eco32, eco58
− Aggregation: row-mean (descriptives) and
factor score (modelling)
− Reliability: Cronbach’s alpha
695
− F3 - Advanced Factor Conditions
− Items (3): eco34, eco43, eco66
− Aggregation: row-mean (descriptives) and
factor score (modelling)
− Reliability: Cronbach’s alpha
Because the eco* items comprise 3 - 5 response
categories, modelling choices for ordinal data (e.g.,
robust continuous approaches vs. categorical
estimators such as DWLS/WLSMV) entail different
assumptions and sensitivities. In this study, the ordinal
nature was retained for CFA/SEM where feasible,
while descriptive reporting and composite-based
modelling were used where required for numerical
stability [28], [38].
6.2.2 PORTER - Competitive pressure (porter8 -
porter18)
The PORTER construct captures perceived
competitive pressure and structural conditions of
competition within the cluster. In contemporary cluster
research, such pressures are linked to incentives for
upgrading, collaboration, and collective action [30].
− Items (11): porter8 to porter18
− Aggregation: mean of the 11 items
(PORTER_mean), followed by standardization
(PORTER_z) for modelling
− Reliability: Cronbach’s alpha
6.2.3 Support/Benefit - Perceived support/benefit from
Maritime Hellas
The support_composite index reflects perceived
benefit/satisfaction with the initiative as a cluster-
platform value proposition.
− Items (2):
− support_need (“Need/usefulness of support”)
− support_benefit (“Benefit/satisfaction with
Maritime Hellas”)
− Aggregation: mean of the two items (row-mean),
standardized to support_z for modelling
− Reliability: Cronbach’s alpha for a two-item scale
6.2.4 Outcome: Registration (registered)
− registered (0/1): 1 = registered, 0 = not registered
− Modelling note: Because the registration outcome
exhibited indications of (quasi)separation under
standard logistic regression, a penalized (Firth-
type) logistic regression was used to obtain more
stable estimates [39].
7 DATA ANALYSIS AND STATISTICAL
PROCEDURES
7.1 Data preparation and missing-data tests
Data quality was assessed with respect to (a)
completeness and (b) consistency of variable types. The
binary outcome registered (0/1) contained no missing
values. In contrast, indices derived from multiple
questionnaire items, most notably support_composite,
exhibited missingness, leading to analysis-specific
effective sample sizes (see Table 1 and the notes to
Tables 2A–2B). To reduce information loss in scale
construction, composite variables were computed
using row means (i.e., the mean of available items
within the same construct). This practice is widely used
in survey research when item-level missingness is
limited and the interpretation remains aligned with the
original response scale (1 - 5).
All primary indices entering the multivariable
models were subsequently standardized to z-scores
(e.g., support_z, F1_z–F3_z, PORTER_z, emp_z,
age_z), enabling direct comparability of coefficients as
standardized effect sizes and improving numerical
stability.
7.2 Descriptive statistics and scale reliability
Prior to inferential modelling, descriptive statistics
(Mean, SD, Min, Max) were computed for the study
variables, and the distribution of registered was
documented. As shown in Table 1, 57.1% of
respondents were not registered (registered = 0) and
42.9% were registered (registered = 1). Table 1 further
reports the central tendency and dispersion of the main
indices (support_composite, PORTER_mean, F1_comp
- F3_comp, employees, company_age) to establish the
empirical profile of the sample before model
estimation.
Internal consistency of the multi-item scales (F1–F3,
PORTER) was assessed using Cronbach’s alpha (Table
1). Alpha is treated as an indicator of reliability (not
validity) and is interpreted cautiously, particularly for
short scales and/or ordinal items [37].
7.3 Index construction, ordinal Likert measurement, and
an “SEM-informed” strategy
The eco* and porter* items were measured on 5-point
Likert-type scales (1 - 5) and are therefore ordinal by
design. In an initial stage, the analysis attempted
ordinal CFA/SEM using categorical estimators (e.g.,
WLSMV/DWLS), as commonly recommended for
categorical indicators [28]. However, model testing
revealed sparse/empty response categories in some
items and (near-)perfect polychoric correlations,
conditions that can induce numerical instability, a
singular information matrix, and identification
problems in categorical SEM.
Accordingly, the study adopted a pragmatic and
publishable SEM-informed, two-stage strategy: (a)
preserving the theoretical structure by using
composites/parcels, and (b) locating the main
inferential claims at the structural (regression/path)
level. While parceling/composite approaches are
defensible when item-level latent modelling becomes
unstable, they require explicit justification because
they may smooth over heterogeneity at the indicator
level [40].
All analyses were implemented in R. The lavaan
package was used for path-type specifications and
reproducibility [38], while the registration analysis
employed Firth penalized logistic regression using the
logits package [41].
7.4 Model 1: Determinants of perceived support/benefit
(support model)
The first equation examined determinants of perceived
support/benefit from Maritime Hellas. The outcome
696
support_composite was defined as the mean of
support_need and support_benefit and subsequently
standardized (support_z). The model was estimated as
a linear structural model (path-SEM, OLS-equivalent),
with predictors F1_z–F3_z, PORTER_z, and
organizational controls (emp_z, age_z). Results are
reported in Table 2A.
Findings indicate that competitive pressure
(PORTER_z), advanced factor conditions (F3_z), and
firm size (emp_z) are positively associated with
perceived value/benefit, whereas firm age (age_z) is
strongly negatively associated (Table 2A). The negative
associations of F1_z and F2_z with support_z are
consistent with a substitution interpretation: where
institutional/governance conditions and related-
supporting industry bases are perceived as already
strong, the incremental value of an intermediary
platform may be perceived as lower—conditional on
the coding direction implying “higher = stronger
evaluation.”
This path specification is just identified (df = 0);
therefore, fit indices such as CFI/RMSEA/SRMR are not
diagnostically informative. Interpretation focuses on
effect sizes, standard errors, and explanatory power
(R²) (Table 2A).
7.5 Model 2: Probability of registration (registered) using
Firth penalized logistic regression
The second equation modelled registered = 1 as a
binary outcome. Standard logistic regression exhibited
indications of (quasi-)separation (fitted probabilities
approaching 0/1 and convergence difficulties), a
condition that may yield non-finite or unstable
maximum-likelihood estimates. In such settings, Firth
penalized likelihood is recommended as a systematic
approach for bias-reduced, finite parameter estimates
[42].
The model included support_z, PORTER_z, F1_z,
F3_z, emp_z, and age_z as predictors. Results are
reported as odds ratios (ORs) with 95% confidence
intervals and p-values in Table 2B. In summary, the
odds of registration increase with F3_z and particularly
with firm size (emp_z), while they decrease markedly
with PORTER_z; support_z is not statistically
significant (Table 2B). Given separation and penalized
estimation, some confidence intervals are exceptionally
wide and/or asymmetric; interpretation therefore
emphasizes directionality and theoretical relevance
rather than literal precision of OR magnitudes (Table
2B) [42].
7.6 Reporting approach and linkage to limitations and
robustness
Results are reported following standard journal
conventions: (a) descriptives and reliability (Table 1)
and (b) multivariable models (Tables 2A - 2B). The two-
stage, SEM-informed strategy is documented as a
necessary adaptation to the properties of the data
(ordinal indicators, sparse categories, near-collinearity)
(Rhemtulla et al., 2012; Marsh et al., 2013). Similarly,
the use of Firth logistic regression is justified as an
appropriate remedy for separation, supporting more
stable inference [43].
Table 1. Descriptive statistics and scale reliability
Note: Means/SDs are reported on the original 1 - 5 scale. Reliability
is reported as Cronbach’s alpha [37].
Panel A. Registration distribution (registered) (N = 574)
N
%
328
57.1
246
42.9
Panel B. Descriptives of key indices
N
Mean
SD
Min
Max
451
3.368
0.646
1.000
4.500
574
2.294
0.399
1.455
3.273
574
2.072
0.270
1.210
2.883
574
2.376
0.608
1.000
4.000
574
2.000
0.309
1.333
2.667
574
4.214
1.781
1.000
6.000
574
2.777
0.576
1.000
3.000
Panel C. Scale reliability (Cronbach’s α)
k
items
Cronbach’s
α
N
(complete
cases)
8
0.725
533
4
0.837
512
3
0.697
550
11
0.746
549
Table 2. Main model results
Note: All predictors are standardized (z-scores). lavaan was used
for path-type specification [38]. Firth penalized logistic regression
was estimated with logistf (CRAN).
Table 2A. Support/Benefit model (DV: support_z)
Linear structural model (path-SEM, OLS-equivalent). Just identified
(df = 0): fit indices are not diagnostically informative.
Predictor
B
SE
z
p
Std. β
(Std.all)
F1_z
-0.435
0.027
-16.047
< .001
-0.280
F2_z
-0.390
0.026
-14.829
< .001
-0.251
F3_z
0.534
0.044
12.089
< .001
0.344
PORTER_z
0.788
0.038
20.698
< .001
0.508
emp_z
0.812
0.018
44.230
< .001
0.523
age_z
-1.161
0.032
-35.819
< .001
-0.748
Model fit: R²(support_z) = 0.870.
Table 2B. Registration model (DV: registered) - Firth penalized
logistic regression
Penalized ML (Firth) due to (quasi-)separation. OR = exp(b).
Confidence intervals may be very wide/asymmetric under
separation [42].
Predictor
b (SE)
OR
95% CI OR
(Lower)
95% CI OR
(Upper)
p
support_z
0.048
(0.506)
1.050
0.779
1.808e+06
.650
PORTER_z
-5.468
(0.734)
0.0042
9.66e-04
1.31e-03
< .001
F1_z
16.838
(2.144)
2.05e+07
1.59e+03
1.18e+10
.021
F3_z
1.619
(0.792)
5.047
1.73e-07
12.866
< .001
emp_z
4.854
(0.811)
128.271
1.052
497.607
< .001
age_z
-0.765
(0.998)
0.465
0.300
0.300
< .001
Overall tests: Likelihood ratio test = 581.51 (df = 6), Wald test =
156.50 (df = 6), p < .001.
7.7 Limitations and robustness tests
7.7.1 Limitations
Overall, while data properties (ordinal scales,
sparse categories, and separation in the binary
697
outcome) necessitated specific analytical choices,
sensitivity tests indicate that the core conclusions are
stable in terms of the direction of relationships across
both sub-models (Tables 2A - 2B). Against this
backdrop, the present section summarizes key
limitations and documents the robustness tests
undertaken.
First, the data are based on self-reported
questionnaire responses (single-source, single-
respondent), which may increase the risk of common
method variance (CMV) and perceptual bias,
particularly because predictors and outcomes were
collected at the same time using the same method [43],
[44]. To mitigate this risk, procedural remedies were
implemented, including anonymity/confidentiality,
explicit clarification that there are no “right” or
“wrong” answers, and thematic blocking of items to
reduce consistency pressure in responding [45], [46].
Second, although Likert-type measures are ordinal,
sparse/empty categories and near-perfect polychoric
correlations limited the feasibility of fully latent ordinal
CFA/SEM at the item level. Such conditions are known
to generate numerical instability, singular matrices,
and identification issues in categorical SEM [28].
Consequently, the study employed an SEM-informed
two-stage strategy (composites/parcels followed by
structural modelling). This approach is defensible as a
pragmatic solution when item-level modelling is
unstable but remains a limitation relative to an “ideal”
fully latent specification because it may smooth over
indicator-level heterogeneity [29], [40].
Third, the binary outcome registered exhibited
(quasi-)separation under standard logit/probit (non-
convergence and fitted probabilities approaching 0/1),
rendering conventional ML estimates potentially non-
finite or unstable. Firth penalized logistic regression
was therefore applied, consistent with methodological
recommendations for separation and yielding bias-
reduced finite estimates [42]. Nonetheless, under
separation it is expected that some ORs and CIs will be
extremely wide or asymmetric; thus, interpretation
should prioritize direction and theoretical relevance
over literal magnitude when intervals are extreme
(Table 2B) [42].
Fourth, support_composite contained missing
values, reducing the analytic sample for the support
model to N = 451 (vs. N = 574 overall). While row-mean
scoring reduces information loss at the scale level, the
smaller N implies that support-model inferences apply
to the subset with available support data and may be
affected if missingness is not random [47].
Finally, the study is cross-sectional; therefore,
results support theory-consistent associations under a
cluster/Porter lens but do not permit causal claims or
inference about dynamics over time.
7.7.2 Robustness tests
To strengthen inferential credibility, robustness
tests were conducted via sensitivity specifications
focusing on two potential drivers of results: (a) the
strong association of firm age (age_z) and (b)
inclusion/exclusion of F2_z.
1. Support model (Table 2A). The model was re-
estimated:
− excluding age_z, to assess whether the strong
negative age association disproportionately
shaped other coefficients; and
− excluding F2_z, to test whether conclusions
depended on that dimension.
Across specifications, the direction of core
associations remained substantively stable:
PORTER_z and F3_z retained positive relationships
with perceived benefit, whereas F1_z remained
negative (and F2_z negative when included) (Table
2A). This stability suggests that the empirical
pattern does not hinge on a single specification
choice, but reflects a consistent correlation structure
within the sample.
2. Registration model (Table 2B). Given separation,
robustness focused on alternative Firth-logit
specifications:
− with/without support_z, to test whether
perceived value meaningfully altered ORs of the
key ecosystem dimensions; and
− with/without age_z, to assess whether firm age
changed the direction of core associations.
The principal pattern remained robust: F1_z and
F3_z were positively associated with registration,
PORTER_z negatively, and emp_z positively (Table
2B). Because estimation occurs under separation,
very wide intervals in some parameters are treated
as an expected technical feature rather than
grounds to reject directional inference, provided
that reporting explicitly notes the need for caution
when intervals are extreme [42].
8 DISCUSSION
This study empirically examines how perceptions of
key conditions within a maritime ecosystem/cluster,
namely governance and institutional thickness, the
base of related and supporting industries, and
advanced factor conditions, together with competitive
pressure, are associated with (a) perceived
value/benefit derived from Maritime Hellas and (b)
formal institutional commitment via registration
(registered). In the cluster-policy literature,
competitiveness and collective performance are
expected to strengthen when robust coordination and
governance structures coexist with a supportive
industrial base and upgrading capabilities (skills,
knowledge, innovation), particularly when credible
intermediaries orchestrate collective action [31], [48].
8.1 What generates perceived value/benefit from Maritime
Hellas?
The support/benefit sub-model (support_z) exhibited
exceptionally high explanatory power (R² = 0.870;
Table 2A), indicating that the selected dimensions
capture substantive mechanisms through which
respondents evaluate the cluster platform. Competitive
pressure (PORTER_z) and advanced factor conditions
(F3_z) were positively associated with perceived
benefit (Table 2A). This pattern is consistent with the
view that more demanding and competitive
environments heighten the salience of collective
services, such as access to information, knowledge,
technological upgrading, and internationalisation, that
698
function as cluster-level upgrading mechanisms [31],
[7].
Firm size (emp_z) was also strongly and positively
related to perceived benefit (Table 2A), suggesting that
larger organisations are better positioned to leverage
institutional and networked arrangements (e.g.,
participation in initiatives, access to networks, and
greater absorptive capacity). This aligns with the
broader discussion on cluster management and
intermediary effectiveness, where engagement often
presupposes organisational resources and the capacity
to participate meaningfully in collective structures [48].
By contrast, F1 (governance/institutional thickness)
and F2 (related and supporting industry base) were
negatively associated with perceived benefit (Table
2A). While counterintuitive relative to “more is better”
assumptions, this sign pattern is theoretically
interpretable via a substitution mechanism: when
actors perceive the ecosystem as already institutionally
mature or supported by effective complementary
structures, the incremental value attributed to a central
intermediary may be lower because the perceived
“coordination gap” is smaller. In such contexts, the
platform’s added value may be expected to shift
toward more specialised, differentiated services rather
than general support [7], [48].
Finally, firm age (age_z) was strongly negatively
related to perceived benefit (Table 2A), consistent with
path dependence and organisational inertia: more
established firms may rely on entrenched practices and
longstanding networks that reduce the perceived
necessity of a new institutional platform.
8.2 What drives formal participation/registration
(registered)
For the registration outcome (registered), estimation
relied on Firth penalized logistic regression due to
(quasi-)separation (Table 2B). Substantively, the
pattern indicates that registration, as a form of
institutional commitment, is driven primarily by (i)
governance/institutional thickness (F1_z), (ii)
advanced factor conditions (F3_z), and (iii)
organisational capacity/size (emp_z), while
competitive pressure (PORTER_z) is strongly negative
(Table 2B).
The contrast, PORTER positive for perceived
benefit yet negative for registration, is particularly
informative. Competitive intensity may increase the
perceived usefulness of collective services, while
simultaneously amplifying strategic distancing,
opacity, or reluctance to formalise membership if
organisations fear information leakage or erosion of
competitive advantage. Such dual dynamics are
consistent with contemporary cluster-policy
perspectives, which stress that cooperation under
competition depends critically on trust-supporting
institutions, clear participation rules, and credible
governance of collective action [48].
Moreover, the non-significance of support_z in the
registration model (Table 2B) suggests that formal
participation is not merely a function of positive
evaluation. Rather, registration appears to be shaped
by “harder” preconditions of capability and
institutional fit, governance capacity, advanced
upgrading conditions, and organisational resources,
implying that registration constitutes a higher-cost
commitment requiring readiness to engage and
confidence in a rules-based, trustworthy institutional
environment. This is broadly consistent with the
argument that clusters perform better when they
embed institutional infrastructures for collective action
and mechanisms for upgrading [31].
Finally, the presence of extremely large odds ratios
and very wide confidence intervals for some predictors
should be interpreted primarily as evidence of a strong
directional signal rather than a literal effect magnitude,
reflecting separation and the properties of penalized
estimation [42].
9 CONCLUSIONS AND IMPLICATIONS FOR
POLICY AND PRACTICE
Overall, the findings indicate that Maritime Hellas is
assessed through two complementary lenses: (a) as a
platform delivering value and services (support) and
(b) as an institution enabling collective action
(registered). Perceived value/benefit increases under
stronger competitive pressure and where upgrading
capacity is stronger (advanced factors), whereas formal
registration appears to require a combination of
institutional maturity/governance, advanced factors,
and organisational resources (Table 2A - 2B). This
distinction is central for cluster policy design:
persuading actors that a platform “creates value” is
analytically and operationally different from
motivating them to make a formal commitment
through institutional participation. Contemporary
cluster-policy agendas emphasise precisely this shift,
from broad visibility and generic coordination to
targeted upgrading services, credible governance, and
measurable added value [7], [48].
Practical and administrative recommendations for
Maritime Hellas
1. Invest in “advanced factors” as the core of the value
proposition.
The initiative should prioritise services that
enhance access to skills, technology,
data/information, innovation support, and
internationalisation, because these dimensions are
positively associated with both perceived benefit
and registration (Table 2A–2B).
2. Strengthen institutional trust and rules of
participation.
Given that higher competitive pressure reduces the
odds of registration, Maritime Hellas should
reinforce safeguards that enable cooperation
“without strategic exposure.” This implies clear
confidentiality rules, structured knowledge-sharing
protocols, and credible mechanisms for protecting
members’ sensitive information [7].
3. Adopt a differentiated membership strategy that
balances capacity and inclusion.
Because firm size is strongly related to both
perceived benefit and registration, engaging larger
firms can increase institutional weight and
credibility. At the same time, the initiative should
create explicit “on-ramps” for smaller and newer
organisations (e.g., tiered membership fees,
thematic working groups, project-based
699
participation), to avoid reinforcing representational
imbalance and to widen participation.
4. Redesign the notion of added value where
institutional and supporting conditions are already
strong.
The negative association of F1/F2 with perceived
benefit suggests that, in more mature or well-
networked environments, value must become more
specialised and differentiated (e.g., international
missions, joint training programs, innovation
procurement, and targeted upgrading
instruments), rather than relying primarily on
generic coordination under a broad umbrella [48].
Finally, this study contributes by showing that the
core “drivers” of cluster performance can influence
platform-oriented attitudes (perceived value) and
commitment-oriented behaviours (registration) in
different, and sometimes opposing, ways. This
distinction strengthens ongoing debates on how cluster
policies should be evaluated, not only in terms of
recognition and acceptance, but also in terms of their
capacity to translate favorable evaluations into
institutional participation and collective action. In this
sense, the findings reinforce the argument that
effective cluster initiatives must couple visible value
creation with governance arrangements that enable
credible commitment and sustained cooperation [7],
[31], [48].
APPENDIX A
A1. Ecosystem evaluation items (eco) & Porter pressure
items (porter) (Short label = the bracketed text from your
original codebook)
Construct
Var
Short item label
F1 – Institutional
Thickness &
Governance
eco35
Tax incentives
eco36
Tax exemptions
eco44
Cluster policy formulation
eco50
Government support for
international
exposure/presence
eco53
Collective action mechanisms
eco64
Strategic planning
eco65
Policy interventions
eco67
Coordination among
organizations
F2 – Related &
Supporting
Industries
eco28
Availability of suppliers
eco31
Cooperation & networking
among firms
eco32
Specialized services
eco58
Support from
shipyards/repair facilities
F3 – Advanced
Factor Conditions
eco34
Availability of specialized
human capital
eco43
Academic/research
infrastructure
eco66
Digital infrastructure
PORTER –
Competitive
Pressure
porter8
Competition among firms
porter9
Buyer power
porter10
Supplier power
porter11
Threat of substitutes
porter12
Threat of new entrants
porter13
Cost/price pressures
porter14
Pressure for
differentiation/innovation
porter15
Regulatory pressure
porter16
International competition
porter17
Market
instability/uncertainty
porter18
Speed/time pressure (time-to-
market)
Support –
Maritime Hellas
support/benefit
support_need
Support/Need (item)
support_benefit
Support/Benefit–Satisfaction
(item)
Outcome /
Covariates
support_composite
Mean(support_need,
support_benefit)
registered
1 = registered, 0 = not
registered
employees
Ordinal (6 levels, increasing)
company_age
Ordinal (3 levels, increasing)
respondent_role
(All values missing → not
used in the final models)
REFERENCES
[1] Porter, M. E. (1998). Clusters and the new economics of
competition. Harvard Business Review, 76(6), 77–90.
[2] Song, D.-W., Panayides, P. M., & Yeo, G. T. (2011).
Maritime logistics as a trade competitiveness enabler.
Maritime Policy & Management, 38(6), 605–632.
https://doi.org/10.1080/03088839.2011.624999.
[3] Porter, M. E. (2000). Locations, clusters, and company
strategy. The Oxford handbook of economic geography,
253, 274.
[4] Ketels, C. H. M. (2004). Cluster-based economic
development: What have we learned? UK Department of
Trade and Industry.
[5] Sölvell, Ö., Lindqvist, G., & Ketels, C. H. M. (2003a). The
cluster initiative greenbook. Ivory Tower AB.
[6] Sölvell, Ö., Lindqvist, G., & Ketels, C. H. M. (2003b).
Cluster initiative greenbook 2.0. Ivory Tower AB.
[7] European Commission. (2016). Smart guide to cluster
policy. European Commission.
[8] Panayides, P. M., & Song, D.-W. (2013). Maritime logistics
as an emerging discipline. Maritime Policy &
Management, 40(3), 295–308.
https://doi.org/10.1080/03088839.2013.782963
[9] Haezendonck, E. (2001). Essays on strategy analysis for
seaports. Garant.
[10] Verhetsel, A., Sel, S., & Vanelslander, T. (2009). Port
competitiveness and the role of hinterland connections:
The case of the Port of Antwerp. European Journal of
Transport and Infrastructure Research, 9(2), 121–142.
[11] Psarros, G., & Notteboom, T. (2022). Revisiting the
relationship between ports and regional development.
Research in Transportation Business & Management, 45,
100818. https://doi.org/10.1016/j.rtbm.2022.100818
[12] Camagni, R. (1995). The concept of innovative milieu and
its relevance for public policies in European lagging
regions. In M. Cohendet, P. Llerena, H. Stahn, & G.
Umbhauer (Eds.), Technology and the future of European
regions (pp. 121–142). Springer.
[13] Wong, P. K. (2022). The development of Singapores
innovation and entrepreneurship ecosystem. In Clusters
of Innovation in the Age of Disruption (pp. 206-244).
Edward Elgar Publishing. DOI:
https://doi.org/10.4337/9781800885165.00019.
[14] Maritime & Port Authority of Singapore. (2023). Annual
report 2022–2023: Building a future-ready maritime
Singapore. https://www.mpa.gov.sg.
[15] van Tuijl, E., & Otgaar, A. (2016). RDM campus
(Rotterdam). In Delivering Sustainable Competitiveness
(pp. 162-183). Routledge.
700
[16] Carayannis, E. G., & Campbell, D. F. J. (2009). “Mode 3”
and “Quadruple Helix”: Toward a 21st century fractal
innovation ecosystem. International Journal of
Technology Management, 46(3/4), 201–234.
https://doi.org/10.1504/IJTM.2009.023374
[17] Erikstad, S. O., & Eide, M. S. (2017). Life cycle assessment
of maritime transportation: A review. The International
Journal of Life Cycle Assessment, 22, 832–847.
https://doi.org/10.1007/s11367-016-1191-9
[18] Meyer, C., Philipp, R., & Gerlitz, L. (2020, October).
Reinforcing innovation and competitiveness of SMEs by
new maritime clustering initiatives in South Baltic Sea
Region. In International Conference on Reliability and
Statistics in Transportation and Communication (pp. 633-
648). Cham: Springer International Publishing. DOI:
https://doi.org/10.1007/978-3-030-68476-1_59.
[19] Hammervoll, T., Halse, L. L., & Engelseth, P. (2014). The
role of clusters in global maritime value networks.
International Journal of Physical Distribution & Logistics
Management, 44(1-2), 98-112. DOI:
https://doi.org/10.1108/IJPDLM-11-2012-0335.
[20] Notteboom, T., & Haralambides, H. E. (2020). Port
management and governance in a post-COVID-19 era:
Quo vadis? Maritime Economics & Logistics, 22, 329–352.
https://doi.org/10.1057/s41278-020-00162-7.
[21] Lagoudis, I. N., & Laloumis, D. (2013). Port hinterland
development: An overview of the logistics-based
approach. In T. Notteboom, P. Konings, & H.
Haralambides (Eds.), Port management (pp. 201–216).
ITMMA Press.
[22] UNCTAD. (2023). Review of maritime transport 2023.
https://unctad.org/webflyer/review-maritime-transport-
2023
[23] European Cluster Observatory. (2020). Cluster
performance and competitiveness in the EU. European
Commission.
[24] Panayides, P. M. (2023). Shipping Performance
Management. Informa Law from Routledge. DOI:
https://doi.org/10.4324/9781315717845.
[25] European Cluster Observatory. (2020). Emerging
industries: Report on the methodology for measuring
emerging industries. European Commission.
https://ec.europa.eu/growth/industry/policy/cluster/obse
rvatory_en.
[26] Anagnostou, A. (2018). Clusters and competitiveness:
The case of the Piraeus port cluster. Maritime Policy &
Management, 45(6), 819–834.
https://doi.org/10.1080/03088839.2018.1468937.
[27] Sotiriadou, M., & Karachle, P. K. (2022). Blue economy
and maritime clusters. Marine Policy, 137, 104951.
https://doi.org/10.1016/j.marpol.2021.104951.
[28] Rhemtulla, M., Brosseau-Liard, P. E., & Savalei, V. (2012).
When can categorical variables be treated as continuous?
Psychological Methods, 17(3), 354–373.
[29] Little, T. D., Rhemtulla, M., Gibson, K., & Schoemann, A.
M. (2013). Why the items versus parcels controversy
needn’t be one. Psychological Methods, 18(3), 285–300.
[30] Ketels, C. (2013). Recent research on competitiveness and
clusters: What are the implications for regional policy?
Cambridge Journal of Regions, Economy and Society,
6(2), 269–284. https://doi.org/10.1093/cjres/rst008.
[31] Delgado, M., Porter, M. E., & Stern, S. (2014). Clusters,
convergence, and economic performance. Research
Policy, 43(10), 1785–1799.
[32] Mitchell, Rebecca; Boyle, Brendan; Waterhouse, Jennifer;
McNeil, Karen; Burgess, John (2009). Institutional
thickness and inter-organisational collaboration in
clusters. Open Research Newcastle. Conference
contribution. https://hdl.handle.net/1959.13/919919.
[33] Norman, G. (2010). Likert scales, levels of measurement
and the “laws” of statistics. Advances in Health Sciences
Education, 15(5), 625–632.
[34] Boone, H. N., & Boone, D. A. (2012). Analyzing Likert
data. Journal of Extension, 50(2).
[35] Wudu, A., Singh, K., & Kassahun, S. (2024). Industry
clusters and firm performance: Evidence from the leather
product industry in Addis Ababa. Heliyon, 10(20).
[36] Li, H., de Zubielqui, G. C., & O’Connor, A. (2015).
Entrepreneurial networking capacity of cluster firms: a
social network perspective on how shared resources
enhance firm performance. Small business economics,
45(3), 523-541. DOI: https://doi.org/10.1007/s11187-015-
9659-8.
[37] Tavakol, M., & Dennick, R. (2011). Making sense of
Cronbach’s alpha. International Journal of Medical
Education, 2, 53–55.
https://doi.org/10.5116/ijme.4dfb.8dfd.
[38] Rosseel, Y. (2012). lavaan: An R package for structural
equation modeling. Journal of Statistical Software, 48(2),
1–36.
[39] Guan, Y., & Fu, G. H. (2022). A double-penalized
estimator to combat separation and multicollinearity in
logistic regression. Mathematics, 10(20), 3824. DOI:
https://doi.org/10.3390/math10203824.
[40] Marsh, H. W., Lüdtke, O., Nagengast, B., Morin, A. J., &
Von Davier, M. (2013). Why item parcels are (almost)
never appropriate. Psychological Methods, 18(3), 257–
284. https://doi.org/10.1037/a0032773.
[41] Gregson, N., Crang, M., & Antonopoulos, C. N. (2017).
Holding together logistical worlds: Friction, seams and
circulation in the emerging ‘global warehouse’.
Environment and Planning D: Society and Space, 35(3),
381-398. DOI: https://doi.org/10.1177/0263775816671721.
[42] Kosmidis, I., & Firth, D. (2021). Jeffreys-prior penalty,
separation and bias reduction in binomial-response
generalized linear models. Biometrika, 108(1), 71–82.
[43] Chang, S.-J., van Witteloostuijn, A., & Eden, L. (2010).
From the editors: Common method variance in
international business research. Journal of International
Business Studies, 41(2), 178–184.
[44] MacKenzie, S. B., & Podsakoff, P. M. (2012). Common
method bias in marketing: Causes, mechanisms, and
procedural remedies. Journal of Retailing, 88(4), 542–555.
[45] Podsakoff, P. M. (2024). Common method bias: The
remains of the day. Annual Review of Organizational
Psychology and Organizational Behavior, 11, 295–323.
[46] Podsakoff, P. M., MacKenzie, S. B., & Podsakoff, N. P.
(2012). Sources of method bias in social science research.
Annual Review of Psychology, 63, 539–569.
[47] Enders, C. K. (2010). Applied missing data analysis.
Guilford Press.
[48] Parrilli, M. D. (2024). Cluster policy: the challenging and
complex horizon in the 2020s. European Planning
Studies, 32 (9), 1868-1884.