University of Southern California Dissertations and Theses

The coevolution of multimodal, multiplex, and multilevel organizational networks in development communities

(USC Thesis Other)

The coevolution of multimodal, multiplex, and multilevel organizational networks in development communities

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content
THE COEVOLUTION OF MULTIMODAL, MULTIPLEX, AND
MULTILEVEL ORGANIZATIONAL NETWORKS IN DEVELOPMENT
COMMUNITIES

by

Seungyoon Lee

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(COMMUNICATION)

August 2008

Copyright 2008 Seungyoon Lee

ii
DEDICATION

To My Parents

iii
ACKNOWLEDGEMENTS
This dissertation is a culmination of a wonderful five-year journey with
many brilliant people at the Annenberg School. I am blessed to have had the
opportunity to work with them over the past years. First and foremost, I would like to
pay my deepest appreciation to Dr. Peter Monge, my advisor and mentor, for
stimulating me to pursue theories and ideas, teaching me the rigor of research
methods, and introducing me to the world of social networks and evolutionary
thinking. Without the pleasure of achievement and the value of self-confidence I
learned from him, this entire journey would have not been completed.
I also thank the other members of my dissertation committee: Dr. François
Bar, for helping me engage in critical thinking about the impacts of technologies, and
Dr. Thomas Valente, for providing me with training in social network studies. In
addition, I would like to thank Dr. Janet Fulk, Dr. Hernan Galperin, and Dr. Manuel
Castells for inspiring me in the fields of global organizations, development, and
network society, respectively, which have been built into this dissertation.
A special thanks goes to Namkee Park for his invaluable help, support, and
advice on every bits of my graduate student life. Thanks are also due to Seung-A Jin,
Hayeon Song, Younbo Jung, and Jae Eun Chung for making the Annenberg School
and Los Angeles a relaxing and exciting place to stay. I look forward to our
friendship and colleagueship in the many years to come. I also greatly enjoyed my
time with colleagues in the Annenberg Networks Network team who I worked,
played, and traveled with. A special acknowledgement is due to Chiyoung Seo for his

iv
wonderful work on data coding and programming, which has enabled the extensive
amount of data analyses presented in the dissertation. An acknowledgement also
goes to Garry Robins for dealing with my questions along the way of learning the
cutting edge network methodologies.
Finally, I would like to express my sincere heartfelt gratitude to my precious
family, especially my dad and mom, for their unconditional love and support. Your
emails, phone calls, visits, and our fascinating travels to many wonderful parts of the
world have sustained me throughout the days and years at the graduate school. My
thanks also goes to my brother, sister-in-law, and my little niece, who make me smile
and laugh. Thank you all for standing by me. I love you so much.

v
TABLE OF CONTENTS
Dedication ii

Acknowledgements iii

List of Tables viii

List of Figures x

List of Abbreviations xi

Abstract xiii

CHAPTER 1: INTRODUCTION 1
Purpose of the Study 1
Chapter Summaries 7

CHAPTER 2: EVOLUTIONARY PROCESSES AND
MULTIDIMENSIONAL NETWORKS 8
The Intersection of Evolutionary and Network Theories 8
Multidimensional Networks 12
Multimodal Networks 12
Multiplex Networks 15
Multilevel Networks 18
Coevolutionary Processes of Variation, Selection, and Retention 22
Resource Space and Niches in Organizational Communities 25
Spatial Locations 27
Organizational Populations 29

CHAPTER 3: NETWORK STRUCTURE, DYNAMICS, AND
EFFECTS 32
Information and Communication Technology for Development 32
The Role of ICTs in Development 35
Populations in the Community 37
The Importance of Interorganizational Networks 41
Two Major Types of Organizational Networks 43
Implementation Networks 44
Knowledge-Sharing Networks 45
Network Structure: Changes at the Global and Subgroup Levels 47
Two Structural Topologies 49
Resource Relationships at the Subgroup Level 52
Network Dynamics: Tie Formation at the Dyadic and Triadic Levels 56

vi
Tie Variation 57
Tie Selection 58
Endogenous Tie Structure 58
Exogenous Environments 62
Exogenous Nodal Attributes 66
Tie Retention 67
Network Effects: Sharing of Technologies and Applications 68

CHAPTER 4: METHOD 80
Data Collection and Coding 80
ICT for Development Projects 81
Project Attributes 84
Collaborating Organizations 85
Organizational Attributes 86
Data Transformation 87
Analysis 91
Global Network Structure (Hypothesis 1) 91
Blockmodeling Analysis (Hypotheses 2 and 3) 93
Multiplex Network Analysis (Hypotheses 4, 5a, 5b, 5c, 6a, 6b, 7,
and 8) 95
Niche Overlap and Structural Equivalence (Hypotheses 9a, 9b,
and 9c) 104
Visualization 106

CHAPTER 5: RESULTS 108
Descriptive Results 108
Network Structure Analysis 116
Hypothesis 1 116
Hypothesis 2 119
Hypothesis 3 127
Network Dynamics Analysis 133
Hypothesis 4 139
Hypotheses 5a, 5b, and 5c 139
Hypotheses 6a and 6b 140
Hypothesis 7 142
Hypothesis 8 143
Visualization 144
Network Effects Analysis 149
Hypotheses 9a, 9b and 9c 149

CHAPTER 6: CONCLUSION 158
Discussion 158
Network Structure 158

vii
Network Dynamics 162
Network Effects 167
Implications for Evolutionary Theory and Social Network
Theory 168
Implications for the ICT for Development Community 171
Limitations and Future Research 173
Limitations 173
Directions for Future Research 175
Competitive Dynamics 175
Network Dysfunctionalities and Failure 176
Network Outcome 177
External Environments 178
Broader Development Communities 179
Conclusion 179

BIBLIOGRAPHY 181

APPENDICES 201
Appendix A: Glossary 201
Appendix B: Topic Keywords for Each Sector (AiDA) 206
Appendix C: Sources of Information in AiDA 207

viii
LIST OF TABLES
Table 1: Multimodal, Multiplex, and Multilevel Approaches in
Interorganizational Networks Literature.............................................21
Table 2: How ICTs Can Impact the MDGs (ITU, 2003)..................................36
Table 3: A Summary of Hypotheses .................................................................79
Table 4: Visual Representation of XPNet Parameters (Adapted from
Wang et al., 2008) .............................................................................103
Table 5: Implementation Projects Breakdown by Attributes..........................110
Table 6: Descriptive Statistics ........................................................................111
Table 7: Organizations Breakdown by Attributes...........................................113
Table 8: Degree Centrality of Organizations in I- Network ...........................114
Table 9: Degree Centrality of Organizations in K- Network..........................116
Table 10: Centralization and Heterogeneity Measures over Time .................117
Table 11: Significance Tests of Differences in Degree Distribution over
Years..................................................................................................119
Table 12: Goodness of Fit Indices for the Models of Choice in I- Network ..121
Table 13: p* Analysis within Blocks of Geographic Regions: I- Network ....122
Table 14: p* Analysis across Blocks of Geographic Regions: I- Network ....123
Table 15: Goodness of Fit Indices for the Models of Choice in K-
Network.............................................................................................124
Table 16: p* Analysis within Blocks of Geographic Regions: K- Network...125
Table 17: p* Analysis across Blocks of Geographic Regions: K- Network...126
Table 18: Goodness of Fit Indices for the Models of Choice in I- Network ..128
Table 19: p* Analysis within Blocks of Populations: I- Network ..................129
Table 20: p* Analysis across Blocks of Populations: I- Network ..................129

ix
Table 21: Goodness of Fit Indices for the Models of Choice in K-
Network.............................................................................................130
Table 22: p* Analysis within Blocks of Populations: K- Network ................131
Table 23: p* Analysis across Blocks of Populations: K- Network.................131
Table 24: Summary of Significant Parameters in Two Time Points
Between and Across Blocks..............................................................132
Table 25: Parameter Values for Selected Models, 1997-1999........................136
Table 26: Parameter Values for Selected Models, 2000-2002........................137
Table 27: Parameter Values for Selected Models, 2003-2005........................138
Table 28: Means, Standard Deviations, and Correlations of Matrices in
MRQAP ............................................................................................149
Table 29: MRQAP Results for a Regression Model Predicting Idea
Sharing between Projects ..................................................................150
Table 30: MRQAP Results for a Regression Model Predicting Structural
Equivalence.......................................................................................151
Table 31: A Summary of Hypothesis-Testing Results....................................156

x
LIST OF FIGURES
Figure 1: Research Framework and Relevant Variables.....................................6
Figure 2: Hypothesized Relationships between Niche Overlap, Structural
Equivalence, and Sharing of Technologies and Applications
across Projects.....................................................................................78
Figure 3: Number of Projects and Organizations by Year..............................109
Figure 4: Visualization of I- Network and K- Network, 1997-1999 ..............146
Figure 5: Visualization of I- Network and K- Network, 2000-2002 ..............147
Figure 6: Visualization of I- Network and K- Network, 2003-2005 ..............148
Figure 7: MRQAP Results for the Relations between Niche Overlap,
Structural Equivalence, and Sharing of Technologies and
Applications across Projects .............................................................152
Figure 8: Visualization of 2-Mode Network of Projects and Organizations
(I- Network, All years combined) .....................................................154

xi
LIST OF ABBREVIATIONS
-2LPL -2 Log Pseudo-Likelihood
AiDA Accessible Information on Development Activities
APC Association for Progressive Communications
CIDA Canadian International Developmental Agency
CSO Civil Society Organizations
DANIDA Danish Development Assistance Agency
DBF Dedicated Biotechnology Firms
DFID Department for International Development
DOT Force Digital Opportunities Task Force
ECOSOC Economic and Social Council
ERGM Exponential Random Graph Model
GAID Global Alliance for ICT and Development
GKP Global Knowledge Partnership
GO Governmental Organizations
ICT Information and Communication Technology
IDRC International Development Research Centre
IGO Intergovernmental Organizations
IICD International Institute for Communication and Development
INGO International Nongovernmental Organizations
ITU International Telecommunications Union
MCMC Monte Carlo Markov Chain

xii
MDG Millennium Development Goals
MRQAP Multiple Regression Quadratic Assignment Procedure
MTML Multitheoretical Multilevel
NGO Nongovernmental Organizations
OECD Organisation for Economic Co-operation and Development
QAP Quadratic Assignment Procedure
SIDA Swedish International Development Cooperation Agency
TNC Transnational Corporations
UN United Nations
UNCTAD United Nations Conference on Trade and Development
UNDP United Nations Development Programme
UNECA United Nations Economic Commission for Africa
UNESCO United Nations Organization for Education, Science
and Culture
UNITeS United Nations Information Technology Service
USAID United States Agency for International Development
V , S, R Variation, Selection, Retention
WSF World Social Forum
WSIS World Summit on Information Society
WWF World Wildlife Fund

xiii
ABSTRACT
The current study is an attempt to understand organizational networks and
organizational community in the field of Information and Communication
Technology (ICT) for Development by exploring the coevolutionary processes of
multidimensional networks. The study addresses three major areas of network
research: network structure (the longitudinal evolution of network structure),
network dynamics (endogenous and exogenous mechanisms that lead to tie
formation), and network effects (the impact of interorganizational networks on
outcome). The study draws theories and hypotheses from the intersection of
evolutionary theory and social network theory. The intersection of the two fields can
be found in their common focus on resources. The current study contributes to the
integration and extension of these two distinct theoretical frameworks.
The study suggests that the coevolutionary dynamics can be more
comprehensively understood when considering the intricacies of multidimensional
networks. Three aspects of multidimensional networks are explored: multimodal,
multiplex, and multilevel networks. The findings from the study demonstrate that
multimodal (within and across organizational populations) and multiplex (two types
of networks) dynamics are significant drivers of tie formation. In addition, the study
emphasizes that the evolution of a community needs to be explored at multiple levels,
including the global and subgroup levels of network structure, dyadic and triadic
levels of tie formation, and nodal level of organizational position.

xiv
The investigation focuses on organizational populations in the ICT for
Development community, which plays a key role in the global efforts to extend the
benefits of ICTs toward enhancing development capacities in the underdeveloped
regions of the world. The study explored the evolution of network structure in the
community over time, and the network dynamics behind this structural evolution.
The results provided support for the decentralization of global network structure and
the increase of within-region and within-population ties over time. Further, the study
examined the effect of these organizational networks on a collective outcome in the
community. The outcome is reflected through the sharing of ideas, represented by
technologies and applications adopted in development projects. The results
complement the literature that examines the effect of niche overlap and structural
equivalence on knowledge sharing.

Keywords: community ecology, evolutionary theory, Information and
Communication Technology for Development, knowledge sharing, multidimensional
networks, multilevel, multimodal, multiplex, network dynamics, network effects,
network structure, niche, organizational networks, population ecology, replicability,
resource, scalability

1
CHAPTER 1: INTRODUCTION
Purpose of the Study
Various inquiries can be made from the integration of evolutionary theories
and network theories to explore how an organizational community evolves as its
interorganizational networks develop. For example, an organizational community is
formed around various organizations dedicated to enhancing the benefits of ICTs
toward development goals. How do organizations in numerous populations, such as
governmental organizations (GOs), intergovernmental organizations (IGOs),
nongovernmental organizations (NGOs), and for-profit corporations create linkages
with each other to leverage resources? How do different types of collaborative links
among organizations, such as joint project implementation and knowledge-sharing,
influence each other? How do micro level dyadic linkages shape the macro level
network properties, for instance, a centralized or a clustered structure? At the same
time, how does the macro level network structure operate as an opportunity for or a
constraint on partner selection? How do these mechanisms evolve over time?
These questions are not only interesting in terms of exploring network
dynamics, an area of inquiry raised by a number of scholars (Baum, Rowley, &
Shipilov, 2004; Powell, White, Koput, & Owen-Smith, 2005), but also have practical
implications for a community’s collective outcome. The outcomes vary considerably,
including the interorganizational transfer of ideas and knowledge (Podolny, Stuart, &
Hannan, 1996), organizational founding (Audia, Freeman, & Reynolds, 2006; Baum
& Singh, 1994c) and mortality (Baum & Oliver, 1991; Baum & Singh, 1994d),

2
community partnerships and service (Provan, Veazie, Staten, & Teufel-Shone, 2005;
Rangan, Samii, & van Wassenhove, 2006), and the inclusion and exclusion of
various organizational bodies in the network society (Castells, 2000b). In regard to
the field of ICT for Development, sharing of ideas and learning across projects are
crucial organizational outcomes, which may be facilitated or inhibited by
organizational ties. The current study attempts to answer the above questions, which
are elaborated on in the theories and hypotheses in the following chapters.
This study is motivated by three practical issues commonly addressed in the
ICT for Development community. First, the sustainability of partnerships has been a
major issue in the community (Unwin, 2005). In particular, the divide of
organizational collaboration structure with regard to world regions and
organizational types has been pointed out as a barrier for ICT for Development
initiatives (Creech & Willard, 2004; Mudhai, 2004; Wilson, 2004). In response, this
study examines changes in these patterns of divide over time. A second challenge in
the field is the discrepancy between project implementation and knowledge-sharing.
Whether organizations can transfer their learning from development projects towards
the accumulation of knowledge in the field has been a major issue (GKP, 2003). To
investigate this problem, this study addresses whether organizations collaborating
with each other for the goal of project implementation engage in collaborations for
knowledge-sharing purposes as well. In other words, the patterns in which these two
network domains are associated are examined. Third, a key metric for successful ICT
for Development projects is considered to be the ideas and lessons shared across

3
various projects, leading to replicability, scalability, and sustainability (Cernea, 1988;
Harris & Rajora, 2006; Tongia, Subrahmanian, & Arunachalam, 2005). The question
worthy of investigation is whether projects that face similar resource demands have
similar organizational partnership networks through which ideas can flow.
The study attempts to answer these questions based on the frameworks of
social networks and ecological theories. These two theoretical frameworks, when
combined, are capable of addressing the issues mentioned above by modeling the
longitudinal evolution of network relations, and the antecedents and consequences of
these relations. Hypotheses about functional relationships among multiple
populations are derived from theories of evolution and community ecology. Network
theories and methodologies are used to model and predict these relationships. While
taking the intersection of the evolutionary paradigm and the network paradigm as a
theoretical backbone, the current study proposes that community network dynamics
can be examined more thoroughly by considering the coevolutionary processes of
multidimensional networks. In addition to drawing attention to the relational features
of an organizational community, evolutionary theories provide a framework for
examining organizational networks in a richer and a more detailed way. In particular,
this study considers multidimensional networks in an organizational community
from three aspects: multimodal, multiplex, and multilevel network dynamics.
The above three issues in the ICT for Development community parallel three
aspects of interorganizational networks: network structure, dynamics, and effects.
Specifically, the study focuses on patterns of relationship structure and formation,

4
and implications of these relationships. Research questions and hypotheses follow a
multilevel framework of organizational networks that has been proposed by several
scholars (Monge & Contractor, 2003). The evolution of an organizational community
can be viewed at multiple levels, including the network position of individual
organizations, tie formation at the dyadic and triadic levels, and structural topologies
at the global level.
In each of these three aspects of network research, the current study is
theoretically distinct from previous studies in the following ways. First, the study
examines the global structure at the macro level by describing properties of the entire
network and subsets of nodes. In the process, the study considers organizational
community in the context of both geographic location and functionally interacting
populations (Aldrich & Reuf, 2006). In other words, multimodal network structure is
explored.
Second, the study explores local network processes that lead to global
outcomes, emphasized by Robins and his colleagues (e.g. Robins, Snijders, Wang,
Handcock, & Pattison, 2007; Wasserman, Robins, & Steinley, 2007). Among these
processes, this study focuses on finding whether multiplex dynamics are significance
predictors of the observed network structure, and the determinants of multiplex tie
formation. Multiplex networks deal with more than one type of network. While a
number of studies have examined the effect of resource dynamics on network
formation, a majority of them have treated all ties as equal, not giving consideration
to the distinct types of resources transferred through linkages. Several scholars in

5
recent years have expressed concern about the lack of knowledge about this issue.
For example, Provan, Fish, and Sydow (2007) raised attention to whether structures
differ across networks that have different functions. The current study is a response
to this call and emphasizes the idea that network content leads to differences in
network structures. Specifically, the study examines two network types in the ICT for
Development community: implementation and knowledge-sharing networks.
Third, the study examines the intersection between projects and
organizations, which has been an area of interest since Breiger (1974)’s earlier notion
of “the duality of persons and groups.” Multimodal networks are an extension of
one-mode network and deal with more than one set of nodes (Wasserman & Faust,
1994). Network studies can address more complex social dynamics by considering
ties among nodes with two or more types of attributes, for example, among people
and events.
With regard to methodology, the analysis is based on probabilistic network
analysis, making use of the recent developments in the statistical modeling of the
emergence and structure of social networks (Pattison & Robins, in press; Wang,
Robins, & Pattison, 2008).
The overall research framework and relevant variables of the study are
illustrated in Figure 1.

6
Figure 1: Research Framework and Relevant Variables

7
Chapter Summaries
Chapter 2 explains each of the three multidimensional network extensions to
the modeling of network structure and dynamics, and introduces how these ideas
have been infused with previous literatures in organizational networks. Further, their
theoretical association with community ecology theories is discussed. Chapter 3
discusses the ICT for Development community, followed by examinations of major
types of organizational populations and interorganizational networks in the
community. Theories and concepts related to the three areas of network research are
discussed: network structure, network dynamics, and network effects. Hypotheses in
each of these three areas follow. Chapter 4 describes data sources, collection
processes, and analyses methodologies used to empirically test the hypotheses.
Chapter 5 presents a descriptive view of the community, followed by the hypothesis-
testing results. Chapter 6 discusses the implications from the findings. Limitations of
the study and future research directions are suggested.
Important concepts and technical terms that appear in the dissertation are
listed and explained in Appendix A.

8
CHAPTER 2: EVOLUTIONARY PROCESSES AND MULTIDIMENSIONAL
NETWORKS
The Intersection of Evolutionary and Network Theories
The application of evolutionary theories to the study of organizations has
significantly contributed to understanding the longitudinal dynamics of
organizational change (Aldrich & Reuf, 2006; Baum & McKelvey, 1999; Monge,
Heiss, & Margolin, in press). Evolutionary theories, interchangeably referred to as
ecology theories, are well suited for explaining the founding, transformation, and
failure of organizations, and further, of populations and communities. The theory of
evolution has placed an emphasis on variation, selection, and retention as a central
force for organizational evolution (Campbell, 1965). Further, evolutionary theories
have explained various processes including, but not limited to, demographic
processes of size, age, and density dependence (Hannan & Freeman, 1977, 1984);
competitive processes of niche segregation and resource-partitioning (Carroll, 1985;
Carroll & Hannan, 2000); ecological processes of symbiosis and commensalism
(Aldrich & Reuf, 2006; Hawley, 1986; Hunt & Aldrich, 1998); institutional processes
of inertia, legitimacy, learning, and imitation (Hannan & Freeman, 1984); and
environmental processes of imprinting, landscapes, and complexity (Baum &
McKelvey, 1999; Kauffman, 1993). Through these processes, which are not
independent of each other, organizations and organizational populations undergo
selection and replacement in response to resource constraints and environmental
pressure.

9
Earlier evolutionary theories have focused on the population ecology of
organizations (Carroll & Hannan, 2000; DiMaggio & Powell, 1983; Hannan &
Freeman, 1977, 1984). Population ecology primarily deals with evolutionary
dynamics at the level of organizational populations, which are defined as collections
of organizations that share some common characteristics and adapt to environmental
situations (Hannan & Freeman, 1977). Further theoretical developments have been
made as the community ecology perspective has emerged, which expanded the focus
from populations to the broader context of organizational communities (Aldrich &
Reuf, 2006; Astley & Fombrun, 1987; Baum & McKelvey, 1999; Baum & Singh,
1994a; DiMaggio, 1994; Hawley, 1986; Hunt & Aldrich, 1998; Ruef, 2000). An
organizational community can be considered as a diverse set of interacting
populations that form functional relationships or interdependencies with each other
in a shared environmental space. This environmental space consists of limited
resources, and populations of organizations share as well as compete for the
resources (Aldrich & Reuf, 2006).
Coevolutionary theories claim that organizational populations are “mutually
interdependent and exert reciprocal influence on one another” (Hunt & Aldrich, 1998,
p. 267). In other words, organizational populations coexist and coevolve in a
community as they occupy environmental niches that contain resources. Moreover,
organizations themselves act as environments for other organizations in the
community. In essence, interdependence among members and interpopulation cross-
effects are fundamental for community ecology theory (DiMaggio, 1994; Hawley,

10
1986), which imply that evolutionary processes exist across populations,
communities, and environments as well as within populations. For example, linkages
provide organizations and populations with access to other populations or
environmental resources. To turn these conceptual implications into a more explicit
proposition, a community ecology perspective has a strong structural logic at its core,
proposing that the evolution of networks is fundamental to the collective survival of
the community. Particularly in the context of organizational communities,
evolutionary theories suggest the importance of interorganizational network
dynamics in understanding the evolution of a community, as exemplified by Powell
et al.’s (2005) study of the biotechnology field.
The intersection between evolutionary and network approaches can be found
in their common focus on resources. From a community ecology perspective,
resource relationships among organizations, populations, communities, and the
environment influence the evolution of ecological communities through processes of
variation, selection, and retention. From a social network approach, flows of
resources shape the micro level formation of linkages and the macro level emergence
of network structure. Despite the notable intersection between the two fields,
research adopting both perspectives has remained unexplored until recently.
Criticism of the lack of research from this perspective goes back to Aldrich (1979).
He suggested that while network concepts have been adopted to account for
organizational complexity caused by the nature of the environment as consisting of
multiorganization aggregates, studies have been limited to “analogies and allusions

11
to social networks” instead of “a full-blown network analysis” (p. 324). Suggesting
that population ecology raises the issue of network emergence and stabilization,
Aldrich discussed the way network approaches can be jointly considered with the
issues of adaptation of new organizations to environments, selection pressures,
stabilizing forces, and retention. A consideration of the evolutionary aspect of
interorganizational networks was subsequently proposed by Powell (1990), who
brought attention to network forms of organizations as distinct from market
transactions or hierarchical arrangements. Further, DiMaggio (1994) encouraged
scholars to “apply population models not to organizations, but to relations among
organizations in different populations or networks” (p. 447).
Responding to these calls, studies that empirically bridge the fields of
evolutionary theories and network theories have emerged in the past decade. For
example, scholars have studied the effect of interorganizational collaboration on
firms’ access to information, resources, and learning experience, and subsequently,
on the growth of the biotech industry (Powell, Koput, & Smith-Doerr, 1996); the
effect of geographic co-location and organizational forms on the advantageous role
networks play for innovation (Owen-Smith & Powell, 2004); the attachment logics
and evolutionary dynamics in the field of biotechnology (Powell et al., 2005); and
the contribution of interorganizational networks to information transfer and
organizational founding in the instrument manufacturing industry (Audia et al.,
2006).

12
Multidimensional Networks
Most network research has employed elementary formulations of networks,
for example, exploring advice networks among employees within a firm at the
interpersonal level. In this sense, Wasserman and Faust (1994) have pointed out that
most social network studies have focused on “one-mode networks with a single,
usually dichotomous and nondirectional relation” (p. 729). The relatively limited
scope of network studies in the past provides opportunities for further conceptual
developments. Recently, scholars have become attentive to these possible extensions.
Monge and Contractor (2003) suggested a multilevel approach to networks,
proposing that organizational network studies can be enriched by approaching
networks at the individual, dyad, triad, and global levels. Pattison and Wasserman
(1999) and Robins and Pattison (2006) suggested multiple social networks as crucial
to the understanding of organizational systems. Adopting these calls for a wider
range of network research, this study considers multidimensional networks,
specifically, multimodal and multiplex relations at multiple levels based on the
theoretical framework of community ecology. These three approaches respectively
extend the original developments of a one-mode, single type of network, and single
level approach.
Multimodal Networks
A multimodal network is a network with two or more modes. Wasserman
and Faust (1994) defined the term “mode” as “a distinct set of entities on which the
structural variables are measured” (p. 29). The simplest formulation of a multimodal

13
network is a two-mode network, which is defined as the existence of two sets of
actors of different types and a relation being directed from actors in one set to actors
in the other (Wasserman & Faust, 1994). In the majority of earlier network literature,
multimode refers to affiliation or membership networks. A two-mode network of
affiliation is created when “one set of actors is measured with respect to attendance
at or affiliation with, a set of events or activities” (Wasserman & Faust, 1994, p. 40).
For example, Anheier and Katz (2004) studied NGOs’ participation in the self-
organized events of the World Social Forum (WSF) and found that the global civil
society exhibited a core-periphery structure. Multimodal affiliation networks apply
well to the community ecology theory, since they may reflect the relationship
between populations and their external environments. While organizations secure
resources by linking with other organizations, they simultaneously access resources
by attending events or joining memberships in the community environment.
A multimodal network can be conceptualized as a network that is formed
within and across multiple sets of populations as well. An example of a two-mode
network appears in Galaskiewicz and Wasserman’s (1989) study on the
interorganizational networks of financial resource flows between two populations in
the Minneapolis/St. Paul metropolitan area, which included corporations and non-
profit organizations. Baum and Oliver’s (1991) study provides an example of a three-
mode relation, which explores the benefit of institutional linkages of child care
service organizations to government agencies and community institutions on
organizational survival. Further, Powell et al. (2005) studied linkages among

14
multiple populations in the biotechnology industry, including dedicated
biotechnology firms (DBFs) and six types of partner organizations. They
conceptualized one-mode ties as those among DBFs and two-mode ties as those
between DBFs and non-DBF organizations.
In this case, the notion of a multimodal network is inherently connected to
the theory of community ecology, which establishes the functional relationships
between multiple sets of populations as its core proposition. The extension from
population ecology to community ecology is parallel to the extension from a
unimodal network to multimodal networks. In other words, the emphasis is placed on
the idea that resources flow though networks that exist not only within a single
population, but across multiple populations (Audia et al., 2006). When viewed from
the perspective of organizational forms, these relationships within and across
populations raise the issue of symbiotic and commensalistic linkages, which are key
to community ecology theories (Aldrich & Reuf, 2006; Hunt & Aldrich, 1998).
Symbiotic linkages are by definition multimodal ties, as they connect different
populations. If populations perform different functional roles and complement each
other in terms of resources and capabilities, they are likely to benefit from the
presence of others by engaging in a symbiotic relationship. On the other hand,
commensalistic linkages are unimodal ties. They represent relations among
organizations within a population that share the same niche space and therefore,
either compete against or collaborate with each other. In this case, organizations
often perform functionally similar roles.

15
Multiplex Networks
A multiplex network is defined as a node linking to another node with more
than one relation (Wasserman & Faust, 1994). Multiplex networks among
organizations are created by applying more than one type of relation on the same set
of nodes, which involves flows of different types of resources. Tie multiplexity has
increasingly been explored as a fundamental aspect of social relations (Lewicki,
McAllister, & Bies, 1998), and particularly, organizational networks (Dacin,
Ventresca, & Beal, 1999; Robins & Pattison, 2006).
Whereas the stricter definition of multiplexity refers to the existence of more
than one relation between a dyad, some have explored the issue of multiplexity from
a broader viewpoint. Organizations are situated in organizational environments
comprised of multidimensional ties overlaid on each other (Baum & Ingram, 2002;
Gulati & Gargiulo, 1999; Robins & Pattison, 2006). In this case, multiplexity does
not presuppose all nodes to be completely overlapping, but incorporates the case
where more than one type of relations are present in a population composed of a set
of nodes. Multiplexity from this point is “the fact that organizations may be tied to
one another in many possible ways across different types of networks” (Lomi &
Pattison, 2006, p. 315). Powell et al.’s (2005) study of the biotechnology industry
can be viewed as an example of a multiplex network from this perspective, where
four types of linkages coexist and weave a community: basic research, finance,
licensing intellectual property, and sales and marketing ties. Similarly, Lomi and
Pattison provided an example of interorganizational dependencies across multiplex

16
networks including supply, technology transfer, and equity networks in the
transportation manufacture industry in Southern Italy.
A consideration of multiplexity highlights the community ecology
perspective as it emphasizes the multifaceted nature of resources and
interorganizational relationships. A central proposition about multiplexity is that
multiple types of network ties are interdependent, and ties in one network influence
the formation or dissolution of ties in other networks (Galaskiewicz & Bielefeld,
1998; Lomi, 1997; Lomi & Pattison, 2006; Lorrain & White, 1971; Robins &
Pattison, 2006). When the same sets of nodes are connected by multiple relations,
one network of interaction becomes the context for a set of other networks. An
earlier account of multiplexity of relations has been raised by Aldrich (1979), who
suggested that multiplexity of relations increases the stability of relationships.
The interdependent nature of multiple networks has been articulated in the
theory of embeddedness (Granovetter, 1985, 1994), which asserts that economic
relations are influenced by social structural and personal relations. Particularly,
structural embeddedness focuses on the way the structure of an organizational
network and an organization’s structural position within it influence economic
outcomes by influencing opportunities for new social alliances and networks (Gulati,
1999; Gulati & Gargiulo, 1999; Uzzi, 1996, 1997). Uzzi (1997) argued that through
increases in social values, attitudes, trust, and reciprocity, this process leads to
increased exposure of networks partners to a broader social and economic life,

17
leading to firms “embedding the economic exchange in a multiplex relationship
made up of economic investments, friendship, and altruistic attachments” (p. 48).
Multiplexity has various effects. Empirical evidence for the effect of
multiplex ties has been provided in the context of a diverse set of industries. For
example, Uzzi (1999) explored embeddedness in the banking industry based on two
features, the duration of a relationship and multiplexity. He found that the degree of
embeddedness of a relationship has a significant effect on reducing the cost of
financial capital. Ingram and Roberts (2000) revealed that the multiplex character of
relationships, particularly in friendship ties among managers of competing hotels
increased their performance. The positive effects included increases in the chance of
joint problem solving, mitigation of competition, and information exchange
stemming from trust, empathy, and social control. Padgett and McLean (2006)
elaborated the process in which the dynamic evolution of multiple social networks
gave birth to a new organizational invention, the partnership system in Renaissance
Florence. Three networks were central to the process: the economic network of
companies composed of partnership ties, the kinship network of marriage and
genealogy ties, and the political network of clientage ties among factions. Creating a
pattern of social embeddedness (Granovetter, 1985), these “cross-domain
connections, through people, regulate[d] the reproductive formation of constitutive
and relational ties” (Padgett & McLean, 2006, p. 1470).

18
Multilevel Networks
A multilevel approach suggests that network research can be enriched by
moving from a single level to multiple levels of analysis, and further, to effects that
occur across levels(Brass, Galaskiewicz, Greve, & Tsai, 2004; Contractor,
Wasserman, & Faust, 2006; Monge & Contractor, 2003). For example, Brass et al.
suggested that network research deals with multiple levels of analysis including the
interpersonal, interunit, and interorganizational levels, and that the realization of
networks is affected by cross-level dynamics. In their multitheoretical multilevel
(MTML) framework, Monge and Contractor suggested broadening the units of
analysis of networks to various levels, including the individual, dyad, triad, and
global levels.
There are two theoretical implications of considering multiple levels. First,
scholars have increasingly suggested a cross-level approach to networks(e.g. Monge
& Contractor, 2003). Robins, Pattison, and Woolcock (2005) have documented the
process in which local subnetwork patterns driven by individual nodes agglomerate
to create the global structure of the entire network. One level influences another level.
Second, some studies speak to the possibility of observing different outcomes by
taking networks at multiple levels into consideration. Provan and Sebastian (1998)
examined clique structure in networks among mental health agencies and their client
outcome. The authors concluded that integration among small cliques of agencies
was associated with effectiveness, while the overall level of network integration was
negatively related to effectiveness. Baldassarri and Diani (2007), following the

19
claims of Hedström (2005), suggested in the context of civic networks that both the
macro level configurations and the micro level dynamics of the network need to be
studied. In summary, these studies build up a common argument that network
structures at multiple levels, as well as the levels of network outcomes and
effectiveness, should be examined.
A similar approach to multilevelness can be found in evolutionary theories
which operate at various levels ranging along the spectrum of organizations,
populations, and communities (Baum & Singh, 1994b; McKelvey, 1994). Moreover,
it has been emphasized that multiple levels are not independent of each other. Baum
(1999) provides a summary of core propositions in multilevel coevolutionary
dynamics: Organizational evolution takes place simultaneously at multiple levels.
These levels are “nested one within the other, wholes composed of parts at lower
levels of organization, and are themselves parts of more extensive wholes” (p. 114).
Several studies on the ecology of organizational communities have taken up
this perspective. For instance, McPherson and Ranger-Moore (1991) proposed that
the selection of members at the individual level produces adaptation in communities
of organizations. Usher and Evans (1996) adopted a multilevel approach for
examining the evolutionary processes of founding, failure, and transition in a gas
station population. The authors considered coevolution at two nested levels, the unit
level of individual gas service stations and the organizational level of oil companies,
as both levels strive to survive. The study concluded that the Darwinian selection and
replacement at a lower unit level can be represented as the Lamarckian adaptation at

20
a higher level, and that they jointly drive the replication of successful forms. Peli and
Nooteboom (1999) identified a cross-influence of two levels in resource-partitioning
theory, where “organizational events appear as cumulative outcomes at the
population level” (p. 1138).
These two approaches to multilevel analysis from the fields of networks and
ecology have slightly different nuances, yet are closely related. A micro level
dependence structure is represented at the organizational dyadic level. Macro level
structural properties are observed at the population and community level. A cross-
level approach suggests that micro level dynamics affect the macro level structure,
and vice versa. Since community ecology has expanded the level of analysis beyond
populations to communities, a multilevel approach has substantial potential to be
extended to the study of organizational network evolution as well. As an example,
Shumate, Fulk, and Monge (2005) examined network evolution in the international
HIV/AIDS community at the organizational, population, and community levels. In
their study, community ecology provided the theoretical foundation for examining
evolutionary influences on network formation at multiple levels.
The multimodal, multiplex, and multilevel perspectives serve as a useful
framework for categorizing previous literature on interorganizational networks. Table
1 summarizes representative examples of studies which vary along these three
dimensions. The table provides a sense of how one or a combination of these three
multidimensional network approaches can be adopted to examine organizational
network dynamics.

21
Table 1: Multimodal, Multiplex, and Multilevel Approaches in Interorganizational
Networks Literature
Multimodal networks Multiplex networks Multilevel Networks
Field of study Networks among
multiple populations
Networks of multiple
types of relations
Analysis at multiple
network levels
Podolny,
Stuart, &
Hannan
(1996)
Worldwide
semiconductor
industry
Unimodal: 1
population
(Semiconductor
producers: merchant
producers, major
captive producers)
Uniplex: Patent
citation network
Single level:
Organizational level
Powell,
Koput,
Smith-
Doerr
(1996)
Biotechnology
industry
Multimodal: 7
populations coded/ 2
populations analyzed:
Dedicated biotech
firms (DBF), non-
DBF partners
Multiplex: 8 types
coded/ Analysis
based on 2 types
(R&D alliances and
non-R&D alliances)
Single level:
Organizational level
Powell,
White,
Koput,&
Owen-
Smith
(2005)
Biotechnology
industry
Multimodal: 7
populations coded (1
DBF, 6 non-DBF
partners)
Multiplex: 4 types
(Licensing, R&D,
finance,
commercialization)
Single level:
Organizational level
Shumate,
Fulk, &
Monge
(2005)
International
HIV/AIDS
community
Multimodal: 5
populations (4
INGOs, 1 IGO) coded
/ 2 populations
analyzed (NGO,
IGO)
Uniplex: Alliance
linkage
Multilevel:
Organizational level,
Population level,
Community level
Audia,
Freeman, &
Reynolds
(2006)
US Instrument
manufacturers
Multimodal: 2
populations
(instruments
manufacturers/
related org.s, other
Standard Industrial
Classification sectors)
Multiplex: 2 types
(Commensalistic ties,
symbiotic ties)
Single level:
Organizational level
Bryant &
Monge
(2008)
Children’s TV
community
Multimodal: 8
populations
(educational content
creators,
entertainment content
creators, content
programmers,
toy/licensed product
manufacturers,
advertisers, advocacy
groups, gov. bodies,
philanthropic org.s)
Multiplex: 2 types
(Mutual ties,
competitive ties)
Multilevel: Dyad
(Organizational)
level, Community
(Global) level

22
Coevolutionary Processes of Variation, Selection, and Retention
While studies increasingly suggest that considering the complex nature of
multidimensional networks is essential to the understanding of interorganizational
fields, these networks have mostly been taken into account separately from each
other. In other words, a large portion of studies that adopt a multimodal, multiplex,
and multilevel network approach have not examined the coevolutionary processes in
detail, with a few exceptions (e.g. Lomi & Pattison, 2006; Shumate et al., 2005).
The proposed study explores the dynamics of multidimensional networks.
Evolutionary theory provides a framework for this attempt. The approach can be
enhanced by the theoretical mechanisms of organizational ecology, particularly the
processes of variation, selection, and retention (V , S, R: Campbell, 1965). Campbell
explained that these three processes, similar to the natural selection processes in
biological evolution, lead to the “evolution in the direction of better fit to the
selective system” (p. 27). Based on this perspective, the process of network
formation can be examined through variation, selection, and retention mechanisms.
Applied to the aspect of interorganizational ties, alternatives for prospective network
ties are explored in the variation stage, particular ties are chosen in the selection
stage, and ties with higher fitness are maintained in the retention stage.
Several interorganizational network studies follow this stream of research.
At the variation stage, the environment influences the evolution of organizations and
organizational networks. While Campbell (1965) originally emphasized blind
variation, which suggests that random processes generate variations, other authors

23
have proposed constrained variation, in which environmental conditions
systematically influence variation processes (see Rao & Singh, 1999 for a review). A
key idea of this perspective is “the existence of unfilled ecological niches” (Rao &
Singh, 1999, p. 69). While endogenous effects within the community play a role, the
possibilities of tie formation simultaneously stem from environmental events which
provide resource munificence. For example, in a study of the evolution of the
children’s television community, Bryant and Monge (2008) found that major
environmental-level effects such as technological innovations, transformations of
norms and values, and new regulatory regimes led population members to initiate
new ties. Koka, Madhavan, and Prescott (2006) suggested that increases in
environmental munificence, in other words, the availability of abundant resources,
led to increases in tie formation. They proposed that the environment, such as
technological changes, culture, and regulations, affects patterns of network change in
predictable ways. In an earlier study, Madhavan, Koka, and Prescott (1998)
examined the effect of industry events and suggested that industry events can have
either structure-reinforcing or structure-loosening impacts on the formation of
interfirm relationships.
At the selection stage, certain types of variations are selected for, while
certain types are selected against. This corpus of research has investigated the
exogenous organizational attributes and endogenous network properties which both
facilitate and constrain tie formation. Studies have revealed various determinants of
interorganizational network formation, including the level of strategic

24
interdependence (Gulati, 1995; Gulati & Gargiulo, 1999); preferential attachment
and clustering (Levine & Kurzban, 2006); and the propensity of organizations to
follow the dominant trend of tie formation in the industry (Powell et al., 2005).
Retention is defined as a process where positively selected variants are
preserved over time. Structural inertia is a key concept of evolutionary theory,
particularly as a constraining force that leads to the retention of organizational forms
and their structure (Hannan & Freeman, 1984). When applied to network changes, it
can be defined as “a persistent organizational resistance to changing
interorganizational dyadic ties or difficulties that an organization faces when it
attempts to dissolve old relationships and form new network ties” (Kim, Oh, &
Swaminathan, 2006, p. 704). Many studies found that prior network ties predict
future relationships as they provide information about the needs, reliabilities, and
capabilities of potential partners, and reduce uncertainty and hazards associated with
new alliances (Ahuja, 2000b; Granovetter, 1985; Gulati, 1995).
If extended to the longitudinal process of evolution, structural embeddedness
is attributed to the effect of inertial forces in tie formation. Granovetter (1985)
suggested that the strength of relations is dependent upon satisfactory past relations.
Gulati and Gargiulo (1999) found that prior direct and indirect alliances between
organizations are likely to enhance further alliances between organizations. Monge
and Contractor (2003) suggested that the focal network at a previous point of time
acts as an exogenous force influencing tie formation. Shumate et al. (2005) examined
an evolutionary process favoring past partners, finding that international NGOs

25
(INGOs) that have formed alliances with each other in the recent past are more likely
to continue those alliances in the future than to change to new alliances.
As shown in these processes, community ecology theories help better
examine how ties evolve in association with one another in multiple populations, in
multiplex networks, and across levels. For the purpose of this examination, the
discussion of resource space in the organizational community is fundamental, which
will be presented in the next section.
Resource Space and Niches in Organizational Communities
Together with the V , S, R mechanism, consideration of the effect of resource
and niche dynamics on organizational communities provides the key to examining
the dynamics of multiple network coevolution. Evolutionary theory and
interorganizational network research share a common emphasis on resources as a
central concept that affects the relationships among populations. In evolutionary
theory, the struggle of populations over scarce resources for survival and growth
drives the evolution of the community (Aldrich, 1979; Aldrich & Reuf, 2006; Monge
et al., in press). In other words, resource dynamics among populations not only affect
the survival of one population but are also closely tied to the collective survival and
growth of the community. An important issue is whether the growth of one
population strengthens or undermines another population, and the community as a
whole. For example, by examining two historical cases of human rights NGOs, Stohl
and Stohl (2005) noted that the growth of NGOs complemented rather than competed
with nation states. NGOs, by filling structural holes in the global organizational

26
sphere, not only brokered information but also became identity managers and
support systems.
A large number of network studies demonstrated the significance of
resources (Chung, Singh, & Lee, 2000; Gulati, 1995; Gulati & Gargiulo, 1999).
Since networks with external constituents allow organizations to accrue valuable
resources that can complement their internal resources (Gnyawali & Madhavan,
2001), the effect of networks on organizational outcome has been an important topic
(Arya & Lin, 2007; Gulati, 2007). In a review of network paradigms in
organizational research, Borgatti and Foster (2003) classified network studies into
two broad categories: studies of network “topologies” and studies of network
“flows” (p. 1002). The former focuses on the structural patterns of interconnection.
The latter primarily considers network content, or the resources that flow through
networks. In the same vein, Smith-Doerr and Powell (2005) restated Emirbayer and
Goodwin’s (1994) as well as Stinchcombe’s (1990) criticism that network studies
have neglected the network content, therefore treating all ties as comparable.
Perspectives emphasizing resource relations are also elaborated in another strand of
social science theorizing: resource exchange and resource dependence theories (Blau,
1964; Homans, 1950; Pfeffer & Salancik, 1978). In interorganizational networks,
resource exchange and dependence is used as one of the dominant theoretical
frameworks for explaining the processes of network formation (Aldrich, 1976;
Pfeffer & Nowak, 1976; Salancik, 1995). Based on this theory, organizations form
ties with other organizations based on the resources both they can offer and obtain

27
from them. In this sense, Salancik (1995) summarized that networks perform a
central role in “maintaining stable collective structures by enabling coordination
among interdependent parties” (p. 346).
The dynamics of network structure and evolution are determined by patterns
of resource distribution in the larger community. In community ecology theory, a
community is regarded as a major locus of resources in an organizational field. The
concept of community located in both spatial and functional relations has not been
explored to a great extent. Aldrich and Reuf (2006) stated that the spatial aspect of
community has been emphasized by early human ecologists, yet has been neglected
due to the heavier emphasis on the relational aspects of symbiosis and
commensalism suggested by Hawley (1986). Recently, Freeman and Audia (2006)
re-explored this concept and conceptualized communities in two different yet
relevant ways: first, spatially, as “places where organizations are located in resource
space or in geography” and second, functionally, as “sets of relations between
organizational forms” (p. 145). Following these conceptualizations, Audia and
colleagues (e.g. Audia et al., 2006) paid attention to the dynamics of geography in
organizations and communities. The current study follows this stream and argues that
these two conceptualizations offer an insightful dimension in examining
organizational niches, as discussed in the next section.
Spatial Locations
In organizational ecology, geographic location in the macro level
environment plays a role as a resource space. Spatial differentiation and

28
interdependency affect the dynamics of populations and communities by providing
opportunities and constraints on the availability of scarce and valued resources
(Freeman & Audia, 2006; Laumann, Galaskiewicz, & Marsden, 1978). Based on this
proposition, the literature on spatial ecology suggests that geographical propinquity
structures are an essential element of organizational dynamics (Freeman & Audia,
2006; McPherson, Smith-Lovin, & Cook, 2001). A common geographic location
entails similarities in demographic, cultural, and policy-related aspects of the
environment. Therefore, being in the same geographic community increases the
likelihood of sharing an environmental resource space. Some studies have assessed
the effects of geographical proximity and local density on organizational founding
and failure. In this context, Carroll and Wade (1991) argued that density dependence
processes apply differently according to spatial locations. Lomi (1995) suggested
that the level of spatial concentration affected organizational founding rates. Another
example of spatial approaches to modeling the ecology of organizations is provided
in the context of automobile manufacturers (Carroll & Hannan, 2000; Hannan,
Carroll, Dundon, & Torres, 1995). These authors studied the two nested levels,
national and world regional, and found that the processes of legitimation and
competition operate differently in the two levels: whereas competition operates at the
national level, legitimation operates at a broader level across nations.
Applied to network dynamics, the effect of location on interorganizational
network formation has been examined in various studies. Since networks channel
resources, whether geographically distant organizations will engage in competitive or

29
mutualistic ties is dependent on their resource relations. Audia et al. (2006) examined
why some locations experience increased initial founding rates and concluded that
information flow through social networks is the key determinant. Similarly, research
has examined why some locations facilitate a larger number of interorganizational
networks than others. Particularly, the importance of geographical location as a
resource space has been emphasized in international organizational networks.
Scholars have examined asymmetrical organizational networks between the North
and the South, and different propensities of networking activities depending on the
geographical location of organizations (Katz & Anheier, 2005).
In summary, spatial location influences organizational dynamics and the
ways networks evolve. Resource relationships among organizations are determined
by whether organizations are located in resource-rich or resource-poor locations, and
in a relative sense, in resource-complementary or resource-competitive locations.
Subsequently, network ties will reflect these different resource patterns.
Organizational Populations
In addition to spatial influences, functional differentiation and
interdependence is central to the discussion of resource dynamics. An organizational
population is understood as a group of organizations possessing a common
organizational form, which implies that they operate in the same resource space
(Aldrich & Reuf, 2006). From this perspective, organizations in a population have
their own structural properties and network relations with organizations in other
populations of the community (Monge et al., in press). Sharing the same

30
organizational form, which is defined in several ways including similarities in
mission and core technology (Hannan & Freeman, 1977) and structural equivalence
(DiMaggio, 1986), determines the resource relationships between organizations and
their positions in the niche space.
These resource distribution patterns in the environment, including
organizations coexisting in the community, influence the formation and coevolution
of multiple ties in the community. As described in the previous section, two
representative types of ties linking populations in a community are commensalism
and symbiosis (Aldrich & Reuf, 2006; Hawley, 1986; Hunt & Aldrich, 1998).
Commensalist relations stem from the similarity between nodes. If organizational
populations overlap in their niche space in terms of their form or core technology, it
means that they are structurally equivalent actors and are likely to create ties with
each other and/or similar others (Galaskiewicz & Zaheer, 1999). On the other hand,
symbiosis assumes dissimilarity between nodes and thus, organizational populations
with “complementary differences” (Barnett, 1990, p. 40) are more likely to form
mutually interdependent ties and mobilize functionally diverse resources toward a
collective goal. A similar argument is made based on exchange theory, where
organizations enter into networks to acquire the resource they need but do not
possess (Galaskiewicz & Marsden, 1978). Consequently, patterns of network
formation are determined, in part, by organizational forms. Gulati (1995) and Gulati
and Gargiulo (1999) revealed that organizations occupying different niches are likely
to be strategically interdependent and to seek alliances with each other.

31
Another distinction between organizational forms with regard to resources is
the width of niche space an organization covers. Based on niche theory, scholars
(Aldrich, 1979; Freeman & Hannan, 1983; Hannan & Freeman, 1977) argued that
population ecology will lead to the distinction of two types of organizations, one of
which they called generalist, and the other, specialist. Generalist organizations
simultaneously rely upon a wide variety of resources in the niche space. Specialist
organizations, on the other hand, concentrate on a narrow homogeneous niche. The
dynamics between these two population types are articulated in resource partitioning
theory (Carroll, 1985), which suggests that as populations evolve, the concentration
of generalist organizations will increase. At the same time, narrow niches that are not
covered by generalist organizations will be released to specialist organizations. The
distinction between generalist and specialist organizations is often associated with
organizational size. Specifically, generalist organizations are assumed to be larger
than specialist organizations (Carroll, 1985).

32
CHAPTER 3: NETWORK STRUCTURE, DYNAMICS, AND EFFECTS
Information and Communication Technology for Development
This study explores the role of multidimensional organizational networks in
the field of ICT for Development. The potential of ICTs as a tool for empowering
less developed countries has emerged as a vital issue in the international
development community, leading to the explosion in the number of organizations
involved in relevant strategies and actions. The evolution of the community can be
summarized by two trends: the growth of development initiatives led by IGOs
accompanied by the expansion of partnership activities, and the increase of global
forums devoted to the issue, such as the World Summit on Information Society
(WSIS).
Since the mid 1990s, major IGOs have launched a series of initiatives that
aim for enhancing the benefit of ICTs for developing countries (Creech & Willard,
2004; Wilson, 2004). These initiatives include the infoDev by the World Bank, the
ICT Task Force by the United Nations (UN), the ICT for Development by the
Organisation for Economic Co-operation and Development (OECD), the G8’s
Digital Opportunities Task Force (DOT Force), and the United Nations Information
Technology Service (UNITeS). With the increase of awareness from a broader
international community, numerous development projects have begun around the
world. These issues have also drawn attention from the perspective of partnerships in
particular (UN-ICT Task Force, 2003; UNESCO, 2006). These initiatives have been
followed by another notable trend, represented by the expansion of partnerships

33
formed by the participation from multiple sectors. These include the Global Alliance
for ICT and Development (GAID), the Partnership on Information and
Communication Technologies, and the Global Knowledge Partnership (GKP), to
name a few.
Several environmental events have led to the rapid growth of the ICT for
Development community since the late 1990s (Creech & Willard, 2004).
Organizations have formed links with each other by participating in activities such as
global and regional associations, meetings, conferences, and web-based networks.
These initiatives have aimed for building knowledge bases and creating an enabling
policy environment for ICTs (GKP, 2003). Among international conferences and
gatherings on issues of ICT and development, the two phases of WSIS organized by
the UN have served as a major environmental catalyst. The first phase was held in
2003, in Geneva, Switzerland, followed by the 2005 summit in Tunis, Tunisia. The
WSIS aimed at enhancing organizations’ interest in partnerships that use ICT for
Development (UNESCO, 2006). The first major legitimacy of the forum was an
inclusive decision-making process that bridges developed and developing countries,
public sector and private sector, and the civil society (MacLean, 2003). Second, it
aimed for promoting a development-oriented Information Society.
Despite its scale and scope, the summit has led to controversies about its
achievements. Some pointed out that the actual outcome did not fulfill the ambition,
particularly when evaluated from the perspective of civil society (Klein, 2004;
Wilson & Best, 2003; Toure, 2004). Some argued that the WSIS failed to address the

34
sociopolitical context of the broader development divide that is critical for the ICT
development agenda (Hamelink, 2004). The demise of the International
Telecommunications Union (ITU) as the central entity in the telecom regime was
criticized in the sense that developing countries became deprived of their stage to
speak (MacLean, 2003). A number of events and global conferences have taken place.
For example, IGOs held a central role and organized events such as the United
Nations Information and Communication Technologies Task Force open
consultations and the United Nations Conference on Trade and Development
(UNCTAD). Conferences such as the UN Roundtable on Communication for
Sustainable Development in 2004 and the World Congress on Communication for
Development in 2006 have taken place.
Further, there has been a growing participation of organizations in a
knowledge-sharing network as well. Some scholars referred to these as a newly
emerging structure of network-like organizations (Smith, 2004), that often exist in
the form of online networks. One example is the Association for Progressive
Communications (APC), which coordinates the civil society organizations dedicated
to using ICTs for social justice and development. Such networks have primarily
linked various initiatives and research projects and assisted the establishment of a
collective information commons.
Given this series of environmental events that have driven the evolution of
the community, it is worth examining the structural changes of the community
represented by network relations among organizations. In particular, the question of

35
whether the community has achieved its goal of facilitating participatory networks of
various stakeholders remains to be answered.
The Role of ICTs in Development
Discussions about the role of ICTs in development have prospered during
the past couple of decades. A widely understood position nowadays is that ICTs
should not be seen as the ultimate development goal but as complementary with
other development needs such as health, food, and sanitation. Scholars have argued
that ICTs should be a tool for broad-based development, which can accelerate the
development efforts in other sectors. The diffusion of ICT infrastructure, once
penetrated, can be used in numerous ways. In this sense, scholars and practitioners in
the field have suggested viewing the potential impact of ICTs in terms of its
contribution to the eight Millennium Development Goals (MDGs) declared by the
UN (ITU, 2003; Tongia et al., 2005). For each goal, ICTs have the potential to be an
accelerator of development. For example, ITU (2003) suggested that each sector has
representative projects as reported in Table 2. In a similar manner, a majority of past
and ongoing ICT for Development projects can be categorized according to the
sector(s) they target.

36
Table 2: How ICTs Can Impact the MDGs (ITU, 2003)
MDG Indicator Impact
Goal 1.
Eradicate
extreme poverty
and hunger
Increase in income
from ICTs
A 1999 study of Village Pay Phone (VPP) owners
in Bangladesh found that profits from providing
phone service constitutes 24% of these households’
total income.
Goal 2. Achieve
universal
primary
education
Primary school
teachers trained by
ICT-based education
In Nepal, 4,430 people were trained as primary
school teachers using radio-based distance
education in 2001. Based on the current student-to-
teacher ratio of 40, an additional 176,616 new
primary school students could be enrolled once
these teachers complete their training. This would
raise the net primary school enrolment rate 5.7%.
Goal 3. Promote
gender equality
and empower
women
Females enrolled in
ICT-based education
as percentage of total
female tertiary
enrolment
Open Learning Australia (OLA) offers higher
education through a combination of distance and
on-line teaching. In 2002, there were 6,129
students enrolled in OLA of which 56.9% were
female. This is higher share than in overall higher
education (54.9%). As a result of OLA enrolment,
female tertiary school enrolment is 0.8% higher.
Goal 4. Reduce
child mortality
Percentage of parents
of small children
using ICT-based
health tools
Baby CareLink is a telemedicine program for
parents of infants in the United States. A 1997-99
evaluation of 56 patients found those parents who
used Baby CareLink reported a 10% higher quality
of care than those who did not use Baby CareLink.
Goal 5. Improve
maternal health
Percentage of
maternal health
workers using ICTs
A July 1999 evaluation of a maternal health project
in the Tororo district of Uganda based on radio
technology, found that maternal mortality dropped
50%.
Goal 6. Combat
HIV/AIDS,
malaria and
other diseases
Percentage of adult
population adopting
health lifestyle after
exposure to ICT-
based health
information
A September 1998 evaluation of an entertainment-
education radio soap opera on HIV prevention in
St. Lucia found that condom imports rose 143%
after the program was aired.
Goal 7. Ensure
environmental
sustainability
Teleworkers as
percentage of total in
employment
There are 38,700 teleworkers in Ireland (2.3% of
total in employment). As a result, CO
2
emissions
from car use are 2% less. If all those in Ireland who
say their job lends itself to teleworking (28% of
total employment) could telework, there would be a
30% reduction in CO
2
emissions

The first goal, eradicate extreme poverty and hunger, is related to projects
that explicitly target economic development, enhanced access to market and labor
information, and electronic commerce and transaction. The second goal, achieve
universal primary education, has been addressed by a number of tele-education

37
initiatives around the world. Third, the potential power of ICTs to promote gender
equality and empower women has been evidenced in numerous projects including the
Grameen Phone in Bangladesh. The next three goals, reduce child mortality, improve
maternal health, and combat HIV/AIDS, Malaria and other diseases, relate to e-
health initiatives which aim for better access to medical information and
infrastructure in developing areas. The seventh goal, ensure environmental
sustainability, has been assisted by projects that implement information and
technologies for resource management and eco-tourism. The last goal, develop a
global partnership for development, has been another prevalent issue in the ICT for
Development agenda. In this category, there have been efforts to link multiple
stakeholders and resources through enhanced use of communication technologies.
Populations in the Community
A diverse array of organizational populations has coexisted in the
development community during the past couple of decades. The ICT for
Development community is not exceptional. There are four organizational
populations that play a critical role in the community: GOs, IGOs, NGOs, and for-
profit corporations.
First, governments, local governments, and provincial authorities in the
public sector constitute an organizational population. Both in the developed and
developing nations, nation states have had a significant role in the
telecommunications sector, especially when dealing with large-scale infrastructure
and policy issues. In general, governments are endowed with authority, credibility,

38
and economic capability in the partnership contexts (Rangan et al., 2006). At the
local government level, partnerships with local stakeholders is considered essential
for gaining a sense of ownership through local wellbeing, culture of trust with
villagers, and knowledge about the community (Ballantyne, 2003; Cecchini & Scott,
2003; Ó Siochrú & Girard, 2005). These local beneficiaries play a critical role in
supplementing the ICT infrastructure and technologies, such as by leveraging
existing facilities and public services and by arranging ICT policies that are
favorable to development goals (UNESCO, 2006).
Second, as interactions across nations have expanded with globalization,
multilateral and bilateral agreements among governments have proliferated (Smith,
2004). While the international system was dominated by states and bilateral relations
until the First World War, the second half of the 20
th
century has seen an extensive
growth of IGOs as the prominent global actor (Madon, 2000). IGOs are defined as
“organizations that meet regularly, are formed by treaty, and have three or more
states as members” (Ingram, Robinson, & Busch, 2005, p. 825). Major players in
this population are the OECD, and various UN agencies such as the United Nations
Development Programme (UNDP), the United Nations Organization for Education,
Science and Culture (UNESCO), the World Bank, and the ITU. This wide variety of
IGOs play an important role in the community by addressing cross-national issues
that are difficult for individual national governments to manage.
Third, NGOs, often referred to as Civil Society Organizations (CSOs) in
recent days (Salamon, Sokolowski, & List, 2004), constitute an organizational

39
population. With globalization, the growth of NGOs’ power and prevalence has been
documented in numerous literatures (e.g. Scholte, 2000). CSOs, which are generally
non-profit organizations, have no participation or representation of government, and
therefore signify civic participation in social and political process (Putnam, 2000).
Representatively, these organizations often pursue humanitarian, progressive, and
watchdog activities. A societal approach to the definition of NGO emphasizes that
the primary aim is to promote common goals, by working for the promotion of
public goods (Martens, 2002). McAdam, McCarthy, and Zald (1996) suggested that
NGOs’ function encompasses both filling in gaps in services provided by the
established institutions and advocating agendas of marginalized groups. This
population includes a wide range of organizations including those operating at the
community level, national level, and international level. This category also embraces
other types of nonprofit organizations, such as foundations, charities, research
institutions, educational institutions, and labor unions. The primary aim of these
organizations is not geared toward promoting common political goals such as
influencing governmental actors or implementing policies. Research institutions in
the technology and social science field have played a critical role in creating and
building knowledge and expertise on ICT and development (Wilson & Best, 2003).
Foundations and charities have emerged as substantial donors in the development
field, as was true with GOs and IGOs in the past.
As global politics have moved beyond the boundary of nation states (Kaldor,
Anheier, & Glasius, 2004), these civil society organizations, primarily based on

40
voluntarism, have moved to the international arena. These INGOs are viewed as a
foundational social infrastructure for the global diffusion of ideas, norms, social
movements, and world politics (Kaldor, Anheier, & Glasius, 2003; Madon, 2000;
Smith, 2004). The definition of INGOs was first given in resolution 288 (X) of The
Economic and Social Council (ECOSOC) on February 27, 1950 as “any international
organisation that is not founded by an international treaty”
1
. The growth of INGOs in
international development areas has been well documented (Lewis & Madon, 2004;
Lewis & Wallace, 2000). For example, large-scale multilateral development agencies
such as the International Development Research Centre (IDRC), the International
Institute for Communication and Development (IICD), and the Development
Gateway Foundation are in the INGO population. Organizations such as the
International Committee of The Red Cross, and World Wildlife Fund (WWF) are
examples of INGOs. The vital role of NGOs has been acknowledged by their intense
arrangements for consultative relationships with the UN and other IGOs.
Finally, an acceleration of private sector participation has been noted in the
development community (Reinicke & Deng, 2000). Among private sector
organizations, there are for-profit organizations including both transnational and
national corporations. Particularly, transnational corporations (TNCs) have been
active in working together with INGOs to address global problems (Kaldor et al.,
2003). The corporate sector, particularly in technology areas, has engaged in digital
divide issues by investing ICT infrastructure technologies to development initiatives

1
http://esango.un.org/paperless/Web?page=static&content=committee

41
(Wilson, 2004). For example, Intel, Microsoft, and Cisco have dedicated their
corporate resources to assist ICT projects in developing countries.
The Importance of Interorganizational Networks
Interorganizational networks have become imperative due to the nature of
ICT for Development initiatives being built on a collection of financial,
administrative, technical, and local resources and expertise (UNESCO, 2006).
Required skills and specialties cannot be acquired from one population. Moreover,
ICT for Development strategies cannot rely solely on national or governmental
efforts, thus requiring a collaborative structure that involves cross-sectoral linkages
among the public, private, and civil society sectors (Mudhai, 2004; UNESCO, 2006;
Wilson, 2004). Organizations involved in the field are distributed over a wide range
of geographic locations as well, providing both opportunities and challenges to
forming a sustainable partnership community. For these reasons, the ICT for
Development field is a suitable context for empirically testing the hypotheses on
interorganizational networks and organizational and geographic communities located
in the resource space.
The ICT for Development field also warrants attention due to the fact that
both non-profit and for-profit institutions are simultaneously present. While the
importance of interorganizational networks has been emphasized in various
organizational populations, a majority of studies have been conducted in the context
of for-profit organizations (Arya & Lin, 2007). Most of these studies have focused on
the dynamics associated with strategic alliances among firms (e.g. Ahuja, 2000a;

42
Gulati, 1995, 1999; Oliver, 1990). Nevertheless, interorganizational networks in non-
profit sectors have started to attract attention in the past several years. There is a
growing literature on networking activities among civil society and social movement
organizations (e.g. Diani, 2003a, 2003b; Katz & Anheier, 2005), NGOs (e.g.
Shumate et al., 2005; Stohl & Stohl, 2005), and public and nonprofit community
organizations (e.g. Provan et al., 2005; Taylor & Doerfel, 2003). Some have looked
at partnerships across multiple populations, such as between private firms and public
institutions (e.g. Powell et al., 2005; Rangan et al., 2006).
The emphasis on collective patterns of survival, growth, and sustainability of
interconnected organizations rather than the success of a single unit (Osborn &
Hagedoorn, 1997; Provan et al., 2007) well applies to the context of the nonprofit
sector, including the international development community. The need for
organizations to come together to accomplish goals has grown extensively in these
complex problems, as single organizations do not have the capacity to address the
demands in the field (Lusthaus & Milton-Feasby, 2006). Provan et al. (2005)
emphasized that multi-organizational partnerships are critical in addressing
community and social problems, economic development, and health and human
services. Madon (2000) emphasized that INGOs have two goals: first, implementing
development projects and second, strengthening partnerships with national and local
development agencies in order to influence global policy (p. 4). Despite the
significance of networks for such purposes, few studies have explored this topic, and
even fewer have done so from a systematic network perspective.

43
In addition, bringing an evolutionary perspective to the field is fruitful
because it highlights the resource and network dynamics among populations as well
as between populations and the external environment. Scholars contended that
nonprofit research can benefit from taking an ecological perspective (DiMaggio,
2003; DiMaggio & Anheier, 1990). Central concepts in the ecological research
paradigm, such as environmental conditions, niche and carrying capacity, and the
mutual impact of vital rates across multiple populations provide insights for
understanding the nonprofit sector in relation to its constituencies. The current study
on the ICT for Development community aims to fill these gaps by taking both
network and ecological perspectives.
Two Major Types of Organizational Networks
Understanding networks in the ICT for Development community can be
enhanced by considering the content of partnerships that exist among organizations.
Major types of relationship building between INGOs, TNCs, and IGOs in the global
society include information exchange, project collaboration, participation in
meetings and forums, or joint membership in advocacy coalitions (Katz & Anheier,
2005). This study classifies network relations that have emerged from the core
activities of the field into two types: a network oriented to project implementation,
and a network oriented to knowledge-sharing. A description of the two networks is
provided below.
The categorization scheme follows a shared perspective that appears in
literatures of NGO, social movement, development, and network studies. Baldassarri

44
and Diani (2007) summarized that most of the network literatures in the fields of
civil society and social movement have focused on two major types of networks:
interorganizational alliances and ties created by multiple memberships. In their study,
these two networks were referred to as “transactions”, which is geared toward
exchange of resources and pursuit of collective goals, and “social bonds”, created by
shared core members or individuals’ participation in multiple activities (p. 737).
Laumann et al. (1978) have differentiated these two types as linkages being based on
“resources transfers” versus “interpenetration of organizational boundaries” (p. 463).
The latter often involves goals of solidarity maintenance and bonding. In a similar
sense, GKP (2003) suggested that two topologies of partnerships in the ICT for
Development community are implementation-oriented and design-oriented. A similar
distinction has been adopted in the context of the biotechnology industry, which is
characterized by its reliance on extensive collaborative relationships for acquiring
knowledge and resources. Powell (1998) suggested that network relationships have
been approached in two different ways: first, an exchange-oriented view which puts
transaction as the central feature, and second, collaboration that is noted by
“continuous communication and organizational learning” (p. 229).
Implementation Networks
A number of partnerships have been established among organizations for the
purpose of implementing ICT for Development projects (Unwin, 2005). These
partnerships aim at extending the benefit of various technologies and applications to
the developing world. This type of network can be considered as implementation-

45
oriented partnerships. These are activity-focused, project-based networks which rely
on partnerships to draw on expertise and resources such as funding, knowledge,
learning, and physical technology and infrastructure from various stakeholders
(Unwin, 2005). This approach applies to the ICT for Development community where
projects are implemented by numerous stakeholders composed of project donor,
implementing organization, technology provider, and regulatory bodies. To leverage
complementary strengths and core competencies from multiple populations,
organizations have been engaged in extensive collaboration networks.
Activity focus theory (Corman & Scott, 1994; Feld, 1981; McPhee &
Corman, 1995) articulates that much interaction in modern organizations is based on
joint activities and common projects. The activity-based approach to organizational
communication proposes that networks are dependent on the larger context of
focused activities among members beyond the hierarchy of organization. Therefore,
the network is a dynamic object that is continuously evolving and developing
(McPhee & Corman, 1995; Scott, 2005).
Knowledge-Sharing Networks
The community shares a goal of creating knowledge for development by
learning from ICT experiments (Wilson & Best, 2003). This type of networks can be
considered as knowledge sharing networks, which serve the purpose of creating an
enabling environment for ICTs, including policy, strategy, and regulatory regime.
International organizations and treaties have facilitated international cooperation and
networking among individuals and groups (Smith, 2004). These networks are often

46
formed through affiliation to common events, such as global and regional
committees, forums, conferences, and publication activities (Fillip, 2003; Katz &
Anheier, 2005; Madon, 2000). In addition, a frequent practice in the development
field is for IGOs to invite representatives of NGOs to their conferences, whether as
experts, full participants or observers, so as to gain substantive legitimacy (UIA
2
).
The goal of this network is research, documentation, and dissemination of
lessons and experiences drawn from development work. These networks enable
organizations to be informed of partner’s and community’s activities as a whole.
Consequently, the networks facilitate organizations to approach prospective partners
and knowledge in the field (Gulati, 1995; Powell et al., 1996). Ultimately, these
networks give organizations access to new networks (Galaskiewicz & Bielefeld,
1998).
A recent development in the field of organizational networks enlightens this
aspect: the role of networks as a conduit for knowledge access and flow (Ahuja,
2000a; Owen-Smith & Powell, 2004; Podolny et al., 1996; Powell et al., 1996).
Knowledge is regarded as a strategically important resource of firms (Grant, 1996).
Consequently, a growing emphasis has been given to the value of networks as a locus
through which knowledge held by individual parties can be synthesized. The
centrality of knowledge and information sharing in organizations has been
increasingly discussed in the context of development agencies (King & McGrath,
2003) and NGOs (Lewis & Madon, 2004; McAdam et al., 1996; Meyer, 1997). This

2
http://www.laetusinpraesens.org/docs/ngocivil.php#

47
importance applies to the ICT for Development field, where knowledge is dispersed
across various organizations and populations. Knowledge and experiences often exist
in a form of small-scale success stories, which are yet to affect development
strategies more broadly (Madon, 2000). Shared knowledge is crucial to ease the
process of locating information, knowledge, and expertise in the field. In this sense,
studies have emphasized the importance of organizational learning and knowledge
management between development agencies in different regions, most
representatively, the North and the South (e.g. Hovland, 2003).
Network Structure: Changes at the Global and Subgroup Levels
The inquiry into the structural patterns of interconnection constitutes a
central paradigm in network research (Borgatti & Foster, 2003). Previous studies
have documented the significance of global level structures in understanding
networks (Katz & Anheier, 2005; Mizruchi & Galaskiewicz, 1994) from the vantage
point that the social structure of ties in which organizations are embedded influences
subsequent actions by providing both opportunities and constraints (Granovetter,
1985). Interorganizational networks can be considered as an evolutionary outcome of
organizational action that is embedded in the formation of network structure (Baum
& Ingram, 2002). In this sense, examination of network at the global level provides
understanding about how the structural properties of a social system have evolved
and at the same time, have exerted influence on members of the system.
Despite this importance, a broad review of literature on interorganizational
network studies revealed that the whole-network level has been seldom empirically

48
researched (Provan et al., 2007). Earlier, Jones, Hesterly, and Borgatti (1997) insisted
that organizational studies need to consider not only exchange dyads but also a
network’s overall structure. Recognizing how network relationships are functioning
and evolving is crucial for community members to facilitate the coordination of
diverse resources, knowledge, and skills (Provan et al., 2005). Along similar lines,
Baldassarri and Diani (2007) also insisted that the overall properties of network
structure need to be investigated. The usefulness of examining the global level
structure is further enhanced when a longitudinal perspective is added. Examination
of the whole network level allows the understanding of how structures evolve and
how the evolution leads to a collective outcome for the community of organizations.
In the context of global interorganizational network research, the patterns of
interconnection have significant implications (Anheier & Katz, 2004). Resource
relationships, which are manifested through network structure, influence the
collective patterns of survival, growth, and sustainability of interconnected
organizations. Especially, when viewing organizations as being situated in a
geographically and functionally interacting community, the patterns of interaction
provide a valuable insight into the degree to which organizations are visible and
active in the community. Nevertheless, in contrast to the extensive references made
to the network metaphor in the discussion of global civil society, there has been scant
research available which approaches global networks in a systematic and analytical
way (Townsend, 1999; Waterman, 2000).

49
Given this background, the major goal of this section is to explore the
changing structure of global participation of organizations in the ICT for
Development community and to examine its implications. First, taking a multilevel
approach, investigating both the overall and subgroup network structure is necessary
for achieving this goal. At the global level, the study examines how the observed
network fits in to the two contrasting structural topologies of core-periphery and
clusters. At the subgroup level, hypotheses about ties among and within blocks
defined by two organizational attributes, geographic location and organizational
population, are tested. In other words, this study explores multimodal networks that
are formed across organizational populations in two types of networks. The current
research addresses the debate that frequently appears in development communities:
whether geographic divide in the partnerships among various stakeholders has
lessened, and whether the participation of civil society sectors in the global sphere
has increased over time. These network structures provide insight to the dynamics of
inclusion and exclusion of various entities of the global system.
Two Structural Topologies
Along with the proliferation of studies on globalization, scholars have
assessed the role played by organizational populations in the global civil society. In
particular, in the past few decades, various forms of transnational and
nongovernmental organizations have emerged, interacting with nation states
(Rosenau, 2003; Stohl & Stohl, 2005). Some have viewed these global networks as a

50
new form of emergent structure that contrasts the two traditional forms of market and
hierarchy (e.g. Fulk, 2001; Powell, 1990).
Facilitated by the process of globalization and the blurring of national
boundaries, the foremost structural change in global networks is heightened
connectivity (Held, McGrew, Goldblatt, & Perraton, 1999; Monge & Matei, 2004).
Held et al. argued that the global society can be described by increases in the
following three characteristics: intensity, which is represented as the overall density
of connections among various nodes; extensity, in other words, the overall spread of
the network; and velocity, measured by the frequency of connections among nodes.
Along similar lines, Castells (2000a) argued that at the core of globalization is the
network society, in which nodes and hubs in the space of flows construct the
organization of the society. Behind this transformation are technological
developments such as telecommunications and the Internet, the increase of political
mobility, and facilitation of economic interactions among nation states.
While a number of studies have discussed the overall connectivity of the
global society, it is important to examine the structural patterns of connections. A
major theme that has emerged in this context is the contrast between a centralized,
hierarchical view of networks versus pluralistic, complex structures (Barnett,
Salisbury, Kim, & Langhorne, 1999; Danowski, 2000; Galtung, 1971). These two
contrasting network structures of world systems are theoretically grounded in the
political economy literature. The traditional view of the world system emphasized
the asymmetry between information-rich and information-poor countries, which is

51
affected by the country’s geopolitical position. This view originates from the
frameworks of dependency and world systems theory (Barnett, 1999; Kick & Davis,
2001; Van Rossem, 1996; Wallerstein, 1974). Some studies of organizational
networks at the global level revealed such patterns. An example is Anheier and
Katz’s (2004) study on NGOs’ participation in the self-organized events of the World
Social Forum (WSF), which revealed that the global civil society exhibited a core-
periphery structure. From this view, organizational structure reproduces the existing
power structure of the political and economic sphere. As a representation of this
view, Smith (2002) noted that transnational organizations reproduces existing
structure by reflecting “the predominant power inequalities that exist between the
global North and the South that persist in the government and business sectors” (p.
506).
On the other hand, more recent studies have suggested a perspective counter
to the reinforcing forces of centralization and polarization posited by the world
systems theory. In this body of research, the global network structure is characterized
by mobility and flexibility. Theoretically, these trends are contrary to the nature of
network dynamics driven by preferential attachment and rich-get-richer phenomenon.
Castells (2000a) suggested the idea of “decentralized concentration”, where a
multiplicity of interconnected tasks takes place in different sites. In studies along this
line, regionalization appears as one of the major transformations. Subsystems are
formed by geographical, social, or cultural homophily and interdependence. These
changes have been found in various global networks, such as the telecommunications

52
networks (Barnett, 2001; Danowski, 2000; Lee, Monge, Bar, & Matei, 2007; Matei,
2006).
Anheier and Katz (2004) found that NGO networks show a property of small
worlds, where there is a high degree of local clustering despite the overall sparseness
observed in large networks. In their study, the small world is a structure in which
“relatively small and distant clusters are connected by relatively small numbers of
bridging actors” (p. 211). A recent work by Baldassarri and Diani (2007) added to
this stream of literature by revealing that civic networks in two British cities
represent a polycentric structure consisting of multiple, interconnected centers.
Based on these findings, the current study tests whether these observed trends of
decentralization are evidenced in the networks of organizations in the ICT for
Development community as well:
Hypothesis 1. Interorganizational networks in the ICT for Development
community will show less centralization over time.
Resource Relationships at the Subgroup Level
Following the earlier discussion on multilevel cross-effects, resource
relationships at the local level act as a driving force behind the formation of global
network structure and community evolution. From a community ecology perspective,
interorganizational networks across populations play a role as a vital channel of
resource flows (Laumann et al., 1978), as the environment is a space where resources
are located and controlled by more than one organization and population. A
proposition directly derived from this theory is that the probability of network

53
formation within and across populations increases if there are resources to obtain
from potential partners (Astley & Fombrun, 1987; Oliver, 1990).
Based on the framework of community ecology and resource space proposed
in Chapter 2, this study considers two aspects of community, which are geographic
location and organizational population. The following hypotheses examine whether
collaborative ties between organizations of different regions and populations have
changed over time. First, studies have emphasized the importance of geographical
location as a resource space. Organizations in the same geographic locations have
access to similar geography-based resources. By contrast, being located in different
geographical locations implies that organizations have different resource
environments, therefore a possibility for forming interdependent ties to pool
resources (Astley, 1985; Gulati, 1995).
In development and social movement literatures, the geographic aspect has
received a substantial amount of debate. Several scholars have discussed the uneven
distribution of the global civil society (Anheier & Katz, 2003; Kaldor et al., 2003;
Katz & Anheier, 2005). Resource dependence perspective (Pfeffer & Salancik, 1978),
which suggests that organizations establish relationships with others to obtain needed
resources, serves as a theoretical background in these studies. For example, Katz and
Anheier (2005) studied transnational networks of social movement organizations and
proposed a pattern of uneven geography, with organizational networks extending
asymmetrically across regions. They noted a divide between the North (resource-
rich) and the South (resource-poor), and suggested that Southern organizations were

54
more likely to connect to Northern organizations to enhance their access to resources.
The divide between the North and the South lies in the fact that Northern
organizations have their roots in industrialized countries, while Southern
organizations exist in aid recipient countries (Lewis, 1998). Whereas Northern
organizations have more abundance in material resources, Southern organizations
have better access to local situations and understandings.
This study examines a recent trend in the network relationships in
development communities, which is the facilitation of regional integration. The
theory of spatial proximity, or homophily with regard to location, has been evidenced
in the context of development communities. In community ecology, these ties are
represented as commensalistic ties formed within organizations of same attributes. In
an earlier study, Boli and Thomas (1999) pointed out that regional organizing has
advantages of shared language, culture, and history. In literatures on transnational
social movement, attention has been paid to the localized contexts, in which
organizations can bridge local and regional concerns with broader international
issues (Smith, 2004). Smith also offered an interesting finding that relates to the
association between spatial location and resource space. She suggested that regional
IGOs and universal IGOs occupy different network positions, with the former being
more central. This finding was accredited to the differential resource dynamics
driven by location, as regional institutions were viewed as having a more immediate
impact on local conditions than universal institutions which deal with the diverse
interests of their global membership. The practical implication of these changes is

55
the improved potential of coordination among organizations that are situated in
similar regions. Further, the growth of the Southern organizations in terms of their
presence, resources, and activities in the global community would support the trend
toward resource exchange, and thus, the facilitation of regional ties:
Hypothesis 2. Interorganizational networks in the ICT for Development
community will show larger increase of within-region ties than across-region
ties over time.
The second conceptualization of community is grounded in the relationships
between organizational populations. In community ecology theory, dynamics among
multiple populations are represented by competitive and cooperative relationships.
Different populations are embedded in heterogeneous resource environments and
subsequently, pursue linkages in different manners. Multimodal ties imply a
symbiotic resource relationship between organizations. Assuming that organizations
operate for the same goal or mission, if the distance of organizations to its potential
partner in its resource space increases, they will likely have complementary
differences. Such difference leads to an increase in the potential gain from
networking with each other, and thus, to an increased likelihood of an exchange
relationship formation. On the other hand, homophily assumes that commensalistic
ties are formed between those with similar traits such as organizational forms.
The extent to which these various sectors are involved in the global process
has important implications for the ICT for Development community. A long-standing
criticism has been that most of the development projects are conducted in a

56
hierarchical manner. Development initiatives have been provided with a vertical flow
of development aid from official donors, mostly IGOs and national aid departments
(UNESCO, 2006).
In contrast, the recent years have seen a larger presence of various
nontraditional donors such as private foundations, charities, and NGOs. Recent
studies have emphasized the emergence of new patterns of ICT-based partnership
combining private sector, government, and civil society (UNESCO, 2006). The
tremendous growth of NGOs active at both international and domestic levels has
been documented (e.g. Anheier & Katz, 2003; Williams, 2003). The growth of both
nonofficial donors and NGOs will lead to an increase of resource exchange
relationships between the two sectors. These changes in resource dynamics,
represented by the decrease of asymmetric dependence of NGOs on IGOs and GOs
will lead to structural changes at the global network level. Specifically, it is expected
that there will be relatively larger increases in active linking within nongovernmental
and nonprofit actors as opposed to linkages across NGOs and IGOs or GOs. This
argument leads to the following prediction:
Hypothesis 3. Interorganizational networks in the ICT for Development
community will show larger increase of within-population ties than across-
population ties over time.
Network Dynamics: Tie Formation at the Dyadic and Triadic Levels
The multitheoretical, multilevel (MTML) analytic framework suggested by
Monge and Contractor (2003) theorizes that both endogenous and exogenous

57
mechanisms need to be modeled in network structuring processes. Endogenous
mechanisms imply that networks are governed by their own structural logic. Local
network properties such as reciprocity, transitivity, and balance are significant
mechanisms behind the observed network structure. In contrast, exogenous
mechanisms assess the influence of external variables on the focal network, such as
other types of networks, the focal network in previous time points, and nodal
attributes.
This section focuses on multiplex mechanisms, which fall into the category
of exogenous variables by examining first, other types of networks and second, the
focal network in previous time points. The distinctive aspect of the current study is
that the approach is longitudinal, since patterns of network formation both at present
and at earlier points of time are expected to influence the evolution of
interorganizational networks. This section attempts to show that taking a multimodal
and multiplex approach can reveal insights that are not captured by examining
unimodal and uniplex ties alone. Therefore, hypotheses exploring the coevolutionary
dynamics of networks in an organizational community are presented below. They are
based on the evolutionary mechanisms of variation, selection, and retention applied
to networks (Monge et al., in press; Zajac & Olsen, 1993), and include a
consideration of resource space dynamics.
Tie Variation
Variation of ties is closely related to the variation of nodes. When new
organizations or organizational populations emerge, these new nodes create potential

58
for tie creation, which implies increases in variation (Monge et al., in press). In this
stage, organizations explore potential partners for linking. Past research (Bryant &
Monge, 2008; Koka et al., 2006; Madhavan et al., 1998) has found that an initial
creation of variation results from ecological niches. Based on these ideas, the number
of ties is likely to increase in response to environmental events that create
opportunities for the broadening of resource space. This increase in resource
munificence will motivate organizations to form ties with each other.
Tie Selection
This section examines the coevolution of multimodal and multiplex
networks at the tie selection stage. According to the evolutionary approach, selection
occurs through the judgment of alternatives that are best suited to the resource niche
(Hannan & Freeman, 1977). In the same way, the selection process of ties will be
determined by their fitness to the resource base. Also called natural selection, ties
that are not fit will be selected out and may be replaced by other ties. Linkages with
high fitness will survive as they are capable of transferring substantial resources
between nodes. On the contrary, linkages with low fitness are lower in their
capability of resource transfer and may subsequently fail. Hypotheses in this section
identify network dynamics that increase the probability of tie selection.
Endogenous Tie Structure
First, the study examines the dynamics between more than one type of tie,
which is one of the endogenous factors behind the tie formation process. It attempts
to answer whether there are systematic processes that drive the evolution of multiple

59
types of interorganizational networks. Multiplexity can take two forms: first,
multiple ties between the same dyads, and second, multiple ties coexisting in a
community between any dyads of organizations. The latter notion can be interpreted
as the diversity of ties an organization forms in the community as a whole. The study
examines the dynamics of tie formation across multiple networks when resources
flow through more than one type of network. Powell et al. (2005) have explored this
issue and demonstrated that multiconnectivity, or the tendency to connect with others
through plural pathways, was a fundamental driver of network evolution in the
biotechnology field. Lomi and Pattison (2006) examined networks among producers
of transportation in Southern Italy and found interorganizational dependencies across
three different types of networks: supply, technology transfer, and equity networks.
Both studies indicated that one reason the selection procedure could be
associated across multiplex networks was the accumulation of organizational
learning. Knowledge exchange leads to an enhanced flow of ideas and skills, and
therefore, organizations attend to partners more intensively (Powell et al., 2005). In a
sense, organizational learning is history-dependent (Levitt & March, 1988), meaning
that organizations learn from past behavior and apply their knowledge to their
linking behavior across multiple networks. Powell (1990) pointed out that useful
information is obtained from “someone you have dealt with in the past and found to
be reliable” (p. 304). Particularly, in fields where knowledge development is rapid
and sources of expertise are dispersed, there are uncertainties about the best approach
to a problem. Therefore, an organization is likely to utilize its experiences and

60
learnings from a past or current linkage with it partner to make decisions about
forming another tie. In other words, multiplex mechanisms will increase the
possibility of identifying ties with potentially higher fitness. Therefore, an
organization selected as a partner in one network relation will have a higher
possibility of also being selected for in another network relation.
In the context of the current study, multiplex linkages are formed when two
organizations collaborate in both implementation and knowledge-sharing projects. If
two organizations have ties to each other by collaborating on a knowledge-sharing
project, it will enhance the probability of those organizations being aware of each
other’s existence and capabilities as a potential partner in project implementation. It
is expected that the same dynamics will apply in the opposite direction as well, in
which co-participation in an implementation project will lead to tie formation in the
knowledge-sharing network. Therefore, the following hypothesis is suggested:
Hypothesis 4. The existence of a tie in one network type will increase the
likelihood of tie formation among the same organizations in another network
type.
The logic of local dependencies (Lomi & Pattison, 2006) suggests that the
formation of dyadic interorganizational ties is determined by the local social
neighborhood. One mechanism that falls into this logic is that of structural
equivalence, which is described by the effect of a similar relational patterns with
other alters (Burt, 1976). In network terms, this leads to the increased possibility of
tie formation between structurally equivalent organizations. Shumate et al. (2005)

61
empirically tested this proposition in the process where a common third-party shapes
the alliance partner choice in HIV/AIDS field. Specifically, INGO dyads that have
similar relations with IGOs were found to be more likely to subsequently form
alliances than INGO dyads that have dissimilar relations with IGOs.
This effect is partly, although not exclusively, explained by the role of
common ties to a third-party (Granovetter, 1973; Gulati, 1995). A common third-
party tie is defined as two organizations having a common connection to a third-
party (Gulati, 1995). Gulati (1995) showed that in the absence of prior direct ties
between two firms, the larger the number of their common third-party ties, the more
likely they are to form new alliances with each other. Studies have offered several
theoretical explanations for this mechanism. First, a common partner can be a source
of legitimacy and reliability. In addition, from the perspective of organizational
learning, participation in collaborative activities with other organizations provides
information about new opportunities (Levinthal & March, 1994; Powell et al., 1996).
An empirical study by Uzzi (1997) suggested that one of the mechanisms behind the
development of embedded ties is associated with third-party referral networks, which
assume the role of a “go-between” (p. 48). These networks transfer expectations and
opportunities of an existing network to a newly matched network through a
transitivity mechanism. In a similar sense, Burt (1992) referred to the role of such
linkages as brokerage, which bridges two nodes that are unconnected to each other.
The current study follows this line of thinking. In the context of the ICT for
Development community, this mechanism may predict the likelihood of tie formation

62
between two organizations if they have ties with a common third organization. It is
hypothesized that this mechanism will exist both within and across networks:
Hypothesis 5a. The existence of a common third-party will increase the
likelihood of tie formation in the same network.
Hypothesis 5b. The existence of a common third-party will increase the
likelihood of tie formation across different types of networks.
Further, past research demonstrated that the logic of embeddedness
(Granovetter, 1985) applies to the above dynamics of transitivity (Gulati, 1995;
Gulati & Gargiulo, 1999). In other words, when two organizations share many
organization partners in common, they have higher likelihood of forming ties with
each other. The current study predicts that this mechanism will extend to multiplex
networks. For example, when organizational dyads have a knowledge-sharing link,
they are likely to share multiple common partners in the implementation network.
The opposite direction may exist as well: having multiple common third-party ties in
the implementation network may increase the chance of tie formation in the
knowledge sharing network. This logic of embedded patterns of transitivity in
multiplex networks is summarized in the following prediction:
Hypothesis 5c. Organizational dyads are likely to share multiple common
third- parties across different types of networks.
Exogenous Environments
Whereas past research emphasized interdependency among multiplex ties, a
competing line of thought exists on this issue. Several network studies have found

63
that the structure and evolution of multiple networks reflect differential resources
flowing in each of the networks. For instance, Laumann et al. (1978) emphasized
“relation-specific structures” (p. 463), where links of one type between organizations
do not necessarily imply other types of links. More recently, Contractor et al. (2006)
stated that if resource dimensions in multiple networks differ, corresponding tie
formation mechanisms and network structures will differ. Lazega and Pattison (1999)
provided an example of these dynamics. While acknowledging the interlocking of
various types of relations, the authors suggested that transfers or exchanges of each
type of resource will be subject to different constraints, which individually determine
the precise form of these interdependencies. For example, they found extra-dyadic
interdependence between collaboration and advice-seeking networks. This
generalized exchange pattern, which is not limited to immediate dyadic reciprocity,
occurs because advice-seeking is likely to be associated with problems induced by
collaboration. On the other hand, they found that advice and friendship networks are
likely to show both dyadic and extra-dyadic interdependence because friendship ties
may be formed to mitigate the negative effects of status competition. People will
create advice-seeking ties in difficult situations. From this perspective, various types
of content flowing through the network drive differential motivation behind network
formation processes, subsequently creating distinctive structural signatures.
These competing perspectives, one emphasizing the association of multiple
networks and the other highlighting relation-specific networks, call for an
exploration of moderating conditions that determine the likelihood or intensity of

64
coevolutionary processes occurring across multiplex ties. Community ecologists
have posited that niche position affects resource relationships, therefore leading to
organizations’ dependence on each other (Astley, 1985; Baum & Singh, 1994d;
Fombrun, 1986). A proposition that applies this theory to the context of networks is
that organizational relationships in resource environments are likely to influence tie
formation. Arguments have been made that organizations in different niches possess
complementary sets of skills and resources, therefore showing greater
interdependence (Gulati, 1995). This argument relates to the symbiotic linkages
formed across different organizational populations (Aldrich & Reuf, 2006).
Nevertheless, in the context of multiplex ties, there have rarely been studies
on whether these resource dynamics will hold true. The current study brings
ecological and network theories to predict this dynamic. A well-examined
proposition is that multiplex ties are driven by embedded relationships to a greater
extent than uniplex ties are (Granovetter, 1985; Uzzi, 1996, 1997). In other words,
accumulated knowledge and trust about partners facilitates tie formation across more
than one type of network. In a distinct stream of research on organizational ecology,
scholars have emphasized that knowledge tends to be shared more easily between
organizations in similar resource spaces (Audia et al., 2006; McEvily & Zaheer,
1999). As McEvily and Zaheer (1999) summarize, geographic clusters support
“extensive interfirm networks supporting frequent and repeated knowledge sharing
and collaborative innovation” (p. 1999). A combination of these two propositions
leads to the argument that when two organizations are located in similar resource

65
spaces, they will have a higher probability of exchanging resources and subsequently,
forming ties of multiple types. In other words, homophily and spatial proximity
mechanisms are expected to be applied to resource dynamics. Therefore, the
following prediction is proposed:
Hypothesis 6a. Organizational dyads in the same resource space are more
likely to form multiplex linkages than a dyad in which the two organizations
are in different resource spaces.
The second hypothesis about external environment considers the width of
space from which organizations seek resources from. Wasserman and Faust (1994)
raised the question of which nodes are likely to be involved in many relations and
which in few, emphasizing that it is one of the substantive issues to be answered in
multiple relational models. Based on the formulation of the resource relationship as a
central driver of tie formation, it is assumed that the moderating condition lies in the
dynamics of resource relationships. First, at the variation stage, alternative
possibilities of multiplex tie formation are explored. The width of the niche space, or
the degree to which an organization concentrates on specific resource space, leads to
the distinction between a generalist organization and a specialist organization
(Aldrich, 1979). As generalist organizations seek a relatively wider range of resource
spaces than specialist organizations, it is expected that they are able to leverage their
resource base across multiple types of networks. Therefore, the following hypothesis
is proposed:

66
Hypothesis 6b. Generalist organizations are more likely to form multiplex
linkages than specialist organizations.
Exogenous Nodal Attributes
The next category of exogenous variable is nodal attributes. The following
hypothesis examines whether the structural positions of organizations in a network
influence multiplex tie formation. Various scholars have studied how structural
position affect alliance formation in a single network context (Gimeno, 1994; Gulati
& Gargiulo, 1999). Network position is associated with perceived influence and
prominence of nodes (Freeman, 1979). Prior research has found that central
organizations are typically more visible to potential partners and also have higher
learning capabilities, therefore gaining organizational prominence (Gulati, 1999;
Powell et al., 1996). This leads to the logic of preferential attachment, also called the
rich-get-richer phenomenon, which favors organizations with high centrality
(Barabasi, 2002).
Prior research has examined the association between an organization’s
structural positions in multiple types of networks. For example, Powell et al. (1996)
identified a multiplex dynamic, in which a firm’s centrality in a network increases its
number of subsequent exchange relationships by facilitating common understandings
and shared principles of cooperation (p. 122). Specifically, they found that the
greater the number of research and development alliances, the greater the number of
other types of collaborations the firm subsequently pursued and the more diverse its
future portfolio became. Involvement in one project opens up opportunities of access

67
to knowledge and resources in the field. From a similar point of view, Gardini (2004)
tested the effect of multiplexity on partner selection, suggesting that organizational
dyads with multiplex relationships partner together more frequently. Her findings
showed that the presence of a production relationship facilitated the presence of a co-
financing relationship in the Hollywood film industry. In a later study, Powell et al.
(2005) suggested that as growing organizations develop various functions, they tend
to pursue preferences for diversity and multiconnectivity, which fuels accumulative
advantage and the attainment of a more central position in the network.
In summary, based on the idea that organizations learn to take advantage of
different types of alliances, it is expected that nodal-level network positions will
influence the selection of ties across different types of networks. Thus:
Hypothesis 7. The greater the combined centrality of two organizations, the
greater the likelihood of multiplex tie formation between those organizations.
Tie Retention
Lastly, coevolutionary dynamics are considered for deriving predictions
about the retention stage. Factors that influence the retention mechanism include
trust, increased knowledge, and reduced uncertainty of potential partners (e.g. Ahuja,
2000; Kim, Oh, & Swaminathan, 2006; Gulati, 1995b). Organizational routine is a
central concept in evolutionary theory (Hannan & Freeman, 1994). As social
networks evolve and structural embeddedness increases over time, it is expected that
the strength of association in the structural similarities across multiplex networks
will increase. As the duration of relationships extends, accumulated experience and

68
solidarity with organizational partners will become greater. The evolutionary
mechanism of self-sufficiency also supports the idea (Barnett, 1990; Hawley, 1986).
As a community evolves and there is a higher depletion of environmental resources,
organizations and populations rely more on each other than on the environment to
acquire resources (Astley, 1985; Astley & Fombrun, 1987). In other words,
organizational populations will increase their engagement in collective action,
depending to an increasing degree on within-community resource exchange and
linkage formation as a field evolves (Bryant & Monge, 2008). Subsequently, there
will be an increasing amount of information exchange and interaction among
populations in multiple dimensions of networks. In addition, based on the idea that
the establishment of routines increases the efficiency of organizational behavior
(Nelson & Winter, 1982; March, Schultz, & Zhou, 2000), ties will be routinized as
networks age and there will be a stronger inertial force (Hannan & Freeman, 1984) to
retain partners across networks. From the context of organizational learning and
knowledge-sharing, establishing a long-term relationship has been identified to be
beneficial for placing the exchange of know-how within a learned and shared code
(V on Hippel, 1988). Therefore, it is hypothesized that:
Hypothesis 8. There will be an increase in the number of repeated dyad
formations for the same organizations across multiplex networks over time.
Network Effects: Sharing of Technologies and Applications
An essential feature of ICT for Development projects is that they are both
local and global in nature. Whereas developing regions are faced with common

69
difficulties and barriers, development efforts need to be tailored to local settings that
are socially and culturally sensitive (Ballantyne, 2003; Cecchini & Scott, 2003; Lee
& Chib, 2008; W2i, 2003). Resulting from this need for heterogeneous adaptation
strategies to match local conditions, ICT for Development tasks have generally been
case-by-case projects with relatively low levels of standardization. Pilot projects tend
to be small scale, small funded, and dispersed around wide locations. As a result,
projects often fail to achieve sustainability and therefore, suffer from lack of broader
programming strategy (Cernea, 1988). This point addresses one of the most pressing
tasks of the ICT for Development community: to find a way for technologies and
applications to be replicated and scaled up from what appears to be “a perpetual pilot
syndrome” (Harris & Rajora, 2006, p. 24).
In this sense, there is a clear benefit of global networking in this
organizational community. First, micro-initiatives and experiences can be replicated
and scaled up when information, knowledge, and learnings are shared. Replicability
means that projects are “reasonably recreated with new applications, in new
environments or in other countries” and scalability means that projects are “scaled up
and/or rolled out on a national level or in other countries” (UNESCO, 2006, p. 85).
In the development literature, there is a growing emphasis on project initiatives to
exchange knowledge with each other and learn from successes and failures
experienced in the region, and the broader set of developing countries. Second, ICT
infrastructure and technologies are characterized by strong scale economies and
network externalities (Best & Maclay, 2002). When new projects create synergies

70
with other existing projects, these projects take advantage of not only learnings but
also technological infrastructure, equipments, and applications. For example, if a
low-cost wireless laptop is deployed in a number of projects, network effect can be
created through bulk provision and low cost sharing of both hardware and ICT
contents (UNESCO, 2006). In this sense, identifying potential linkages among
various projects that are implemented in the field is valuable for the collective
growth of the community. The lack of literature in this area, despite its importance,
motivates examination of the following hypotheses. The current section focuses on
ICT projects as the unit of analysis. The overarching research question is to explore
the relationships between niche overlap among projects, structural equivalence of
projects in regard to their organizational affiliation networks, and the sharing of
technologies and applications across projects.
This study employs the concept of niche for modeling the dynamics of
projects. The concept of a population’s niche is fundamental in evolutionary theory
(Hannan & Freeman, 1977). The concept becomes particularly useful when modeling
the evolutionary path of organisms as they interact with resource environments.
Niche space, or the resource dimension over which organisms compete, can be
characterized in various ways across contexts. In organizational ecology, niche is
defined as “each distinct combination of resources and other constraints that suffices
to support an organizational form” (Aldrich, 1979, p. 112). For instance, niche space
can be the taste of consumers for wine producers (Swaminathan, 1995, 2001),
children of various age ranges for day care centers (Baum & Singh, 1994c, 1994d),

71
and demographic characteristics of members for voluntary organizations (McPherson
& Ranger-Moore, 1991). Environmental resources determine niche space, which
exhibits a carrying capacity in the sense that resource exhaustibility sets a certain
limit on population density (Aldrich & Reuf, 2006).
Organizations or organizational populations with similar resource demands
share the same niche, thus creating niche overlap. Niche overlap leads to a variety of
organizational consequences. When new niches open up, organizations begin
populating the niches and taking advantage of abundant resources and as a
consequence, typically form cooperative relationships (Bryant & Monge, 2008;
Monge & Poole, in press). As density in niche spaces increase and two or more
organizations occupy the same niche, they face similar resource requirements and as
a result, develop competitive relationships (Astley, 1985; Hannan & Freeman, 1977).
Organizational niches affect the dynamics of symbiosis and commensalism, therefore
influencing the rate of founding (Audia et al., 2006; Baum & Singh, 1994c) and
mortality (Baum & Singh, 1994d). Baum and Singh (1994c), based on their analysis
of the founding of day care centers, asserted that density dynamics need to be
disaggregated according to the presence of niche overlap. In other words, the study
found that overlap density negatively affected founding rate as it stimulates direct
competition for resources, while nonoverlap density positively affected founding rate
by increasing the legitimacy and social acceptance of organizational form.
Venkatraman and Lee (2004) elaborated on this idea in the context of the U.S. video
game sector, and suggested that the likelihood of game developers launching

72
products decreased with niche crowding measured by overlap density. In a similar
sense, Podolny et al. (1996) suggested that niche crowding, by strengthening the
competitive effect between organizations, impairs an organization’s life chances.
Some have paid attention to the relationship between niche overlap and alliance
formation. For example, Stuart (1998) found that semiconductor firms in crowded
niche positions are more likely to form alliances than those in sparse niche positions.
As suggested above, a majority of earlier studies examined the effect of
niche overlap in the context of competitive, for-profit industries (Gimeno, 1994;
McPherson, 1983). Such a perspective follows the concept of niche in population
biology, which focuses on competition between species, selection of particular traits,
and adjustment to the environment by species. Yet, whether the commensalistic
aspect of niche overlap results in competition depends on whether species in the
environment need the same resources that are in short supply (Langton, 1987). This
study attempts to expand niche theory by focusing distinctively on the collaborative
aspect of niche overlap. This relationship is referred to as mutualism, when “two
populations in overlapping niches benefit from the presence of the other” (Aldrich,
1999). Especially, when it comes to the context of knowledge-sharing and innovation,
niche overlap can be conceptualized as enhanced opportunities for sharing ideas.
Audia et al. (2006)’s research is based on this idea that information is a nonrival
resource, which can be utilized by multiple users without being diminished or
depleted.

73
Powell, Koput, and Smith-Doerr (1996) emphasized that learning is a social
construction process that exists in a community of organizations formed by exchange
and collaboration relationships. Interorganizational networks can serve not only for
knowledge transfer but also for innovations by which organizations synthesize and
build on other’s knowledge. Niche spaces in technological settings have been
explored by Podolny and his colleagues. Podolny and Stuart (1995) viewed
inventions in semiconductor industry as a technological network, in which nodes
represent inventions and ties represent technological commonalities. The existence of
a tie between two nodes indicates that the present invention shares common features
with the antecedent invention, therefore being able to build upon prior processes of
knowledge creation. The usefulness of adding network components to the study of
niche has been supported by others as well (e.g. Audia et al., 2006; Burt, 1992).
A similar conceptualization is applicable to the examination of networks
among ICT for Development projects. The idea of niche overlap in organizational
contexts (Baum & Singh, 1994c, 1994d) can be directly applied to the case of
projects. Within a community, here defined as communities in developing countries,
development projects occupy distinct resource spaces. When there are multiple
projects in the same niche, the niche is more densely inhabited than others. In the
same way as organizational populations emerge (Monge et al., in press), when one or
more projects enter a niche space previously uninhabited by other projects, it can be
considered that a population of projects emerged. Over time, new populations of ICT
for Development projects emerge.

74
In community ecology theory, community is considered as a
multidimensional resource space (Audia et al., 2006; Baum & Singh, 1994d). Based
on this view, niche can be identified along multiple dimensions (Galaskiewicz,
Bielefeld, & Dowell, 2006; Podolny et al., 1996). In this study, the diverse nature of
ICT for Development projects are classified according to two major dimensions,
which are geographic location and project goals. Niche overlap between two projects
is the degree to which they have commonalities in their features. When multiple
projects occupy fully or partially overlapping niches, it implies that the same
dimensions are pursued by these projects. In other words, there may be more than
one project working on e-health initiatives in Africa. While this is likely to lead to
competition in industry settings (Podolny et al., 1996), it may also facilitate effective
collaboration, based on mutualism (Stuart, 1998). Stuart grounded his argument on
the idea of absorptive capacity, by which organizations can more effectively learn
from the know-how of those who are technologically similar to themselves. If there
were no collaboration, investments would be made independently, creating redundant
efforts (Stuart, 1998). One of the most integral components of ICT for Development
projects is the technology and application pursued by the project. Consequently, it is
implied that projects in similar resource spaces have a greater potential of sharing the
same ideas about technologies and/or applications associated with the
implementation of projects. Based on these arguments, the following hypothesis is
suggested:

75
Hypothesis 9a. The greater the niche overlap between projects, the greater
the possibility that they will share technologies and applications between
them.
Projects that occupy the same niche may have mutual, commensalistic
resources they can exchange with each other. The chance of obtaining useful
information is higher when nodes tie with those that are socially and physically
proximate to them (Lazer & Andre-Clark, 2000), because they are in similar
situations and have relevant experiences. Owen-Smith and Powell (2004) found that
when ties were embedded in a dense regional web of both formal and informal
affiliations, the accessibility of information transmitted through formal linkages
became higher. As these studies imply, when projects have commonalities in one or a
combination of niche dimensions, it is expected that these projects are likely to
potentially seek similar types of organizational resources and therefore, have similar
relational patterns with collaborating organizations. These relational patterns are well
captured by the network measure of structural equivalence, which indicates having a
similar pattern of relations with other nodes in the network (Gnyawali & Madhavan,
2001; Wasserman & Faust, 1994). This reasoning leads to the following hypothesis:
Hypothesis 9b. The greater the niche overlap between projects, the greater
their structural equivalence.
The next hypothesis considers the benefits projects will get from their
organizational collaborations. The effect of network relations on the knowledge
aspect of organizations has been investigated by several scholars. Ahuja (2000a)

76
explored the impact of direct ties, indirect ties, and structural holes on innovation in
the context of interfirm collaboration. He found that both direct ties and indirect ties
have positive impact, while structural holes can have both positive and negative
impacts. At the interpersonal level, Singh (2005) investigated the effect of networks
on patterns of knowledge diffusion. In this study, the likelihood of a patent citing an
earlier patent was a function of similarity of attributes between patents including
region, technology category, and parent firm, and network variables including past
collaboration, common collaborator, and direct and indirect social links. In the realm
of knowledge management and community development, Fesenmaier and Contractor
(2001) examined knowledge networks among rural development practitioners, in
which nodes represented knowledge held by people and ties represented shared
knowledge in the system. They suggested that in a distributed knowledge network,
knowledge network support tools have a potential to facilitate locating individuals’
skills and expertise.
This study extends these literatures by addressing a feature of network
characteristics that has not been examined: the effect of structural equivalence on
sharing of ideas. Structural equivalence means having ties to similar sets of
organizations or having similar network patterns (Wasserman & Faust, 1994),
through which organizations can learn about new opportunities. Therefore, if projects
have similar relations with a set of organizations, there is a greater opportunity of
adopting the same ideas. In other words, structural equivalence in this particular
context of development projects may imply that there is a larger potential of sharing

77
technologies and applications, and thus, a greater possibility of replicability and
scalability of projects. In summary, as phrased by Audia et al. (2006), it is expected
that interorganizational networks will become a vehicle of information transfer.
Therefore, the following hypothesis is proposed:
Hypothesis 9c. The greater the structural equivalence between projects, the
greater the likelihood of technologies and applications shared between them.
Figure 2 illustrates a summary of predicted relations between matrices
examined in Hypotheses 9a, 9b, and 9c.

78
Figure 2: Hypothesized Relationships between Niche Overlap, Structural
Equivalence, and Sharing of Technologies and Applications across Projects

A complete list of hypotheses is presented in Table 3.

Structural
Equivalence
Niche overlap
(sector)
Niche overlap
(region)
Sharing of
technologies and
applications
Temporal
proximity
H9c
H9a
H9b
H9a
H9b

79
Table 3: A Summary of Hypotheses
Network Structure
Hypothesis 1. Interorganizational networks in the ICT for
Development community will show less centralization over time.
Hypothesis 2. Interorganizational networks in the ICT for
Development community will show larger increase of within-region
ties than across-region ties over time.
Hypothesis 3. Interorganizational networks in the ICT for
Development community will show larger increase of within-
population ties than across-population ties over time.
Network Dynamics
Tie Selection Hypothesis 4. The existence of a tie in one network type will increase
the likelihood of tie formation among the same organizations in
another network type.
Hypothesis 5a. The existence of a common third-party will increase the
likelihood of tie formation in the same network.
Hypothesis 5b. The existence of a common third-party will increase
the likelihood of tie formation across different types of networks.
Hypothesis 5c. Organizational dyads are likely to share multiple
common third-parties across different types of networks.
Hypothesis 6a. Organizational dyads in the same resource space are
more likely to form multiplex linkages than a dyad in which the two
organizations are in different resource spaces.
Hypothesis 6b. Generalist organizations are more likely to form
multiplex linkages than specialist organizations.
Hypothesis 7. The greater the combined centrality of two
organizations, the greater the likelihood of multiplex tie formation
between those organizations.
Tie Retention Hypothesis 8. There will be an increase in the number of repeated dyad
formations for the same organizations across multiplex networks over
time.
Network Effects
Hypothesis 9a. The greater the niche overlap between projects, the
greater the possibility that they will share technologies and
applications between them.
Hypothesis 9b. The greater the niche overlap between projects, the
greater their structural equivalence.
Hypothesis 9c. The greater the structural equivalence between projects,
the greater the likelihood of technologies and applications shared
between them.

80
CHAPTER 4: METHOD
Data Collection and Coding
Studies of organizations and organizational networks have often been based
on archival research methods, which involve the analyses of documents, texts, and
electronic databases that are produced by and about organizations (Diani, 2002;
Ventresca & Mohr, 2002). In particular, archival methods have contributed to
ecological research by enabling the collection of life histories of organizations and
organizational populations (Ventresca & Mohr, 2002). The current study is based on
the records of interorganizational collaboration activities in the field of ICT for
Development. Among databases that contain records about development projects, the
study used the Accessible Information on Development Activities (AiDA) database
as its main source. AiDA
3
is an online directory of development aid activities by
sectors, provided by the Development Gateway Foundation
4
in cooperation with the
OECD, the UNDP and the World Bank. Development Gateway Foundation is an
international independent non-profit organization working on development through
information technology. The database is an online directory of official development
aid activities, and is a major source of public information on ICT for Development
projects (Tongia et al., 2005). The database harvests information from major bilateral
donors, multilateral development banks, and UN agencies.

3
http://aida.developmentgateway.org/
4
(http://www.dgfoundation.org/).

81
Data collection was conducted in two steps. In the first step, a pre-
categorized list of ICT for Development projects provided by the AiDA was coded.
The data cover projects in the field of ICTs between 1987 and 2008. These projects
are categorized into ten different themes: ICT capacity building, ICT use in health
sector, ICT use in education sector, ICT initiatives, ICT empowerment, ICT
infrastructure, ICT and rural development, ICT policy, strategies and plans, ICT e-
readiness, and ICT e-commerce. For each of these categories, the AiDA uses unique
sets of keywords to extract relevant projects (see Appendix 1). In the second step,
data were collected from a broader category defined as the Communications sector to
capture the projects that are not pre-categorized into the above ten themes. Projects
listed in the AiDA database were screened based on whether they met the criteria of
ICT for Development projects. Details of these criteria are explained below.
ICT for Development Projects
The study selected projects that have ICT for Development components,
guided by the following principles. First, projects were judged upon the type of
technologies implemented. In previous literatures, ICTs have been described as
encompassing a variety of forms (Chib, 2007). This ranges from advanced modern
technologies, such as the Internet, mobile telephony, computer-based applications,
and satellite communication, along with relatively older technologies such as radio,
television, land-line telephones, video and audio cassettes, multimedia CD-ROM,
and print (OECD, 2001). This study chose to take the former set of advanced
technologies, based on telecommunications networks, into consideration. Therefore,

82
projects exclusively focused on traditional communication media were excluded
from the selection process. Based on this decision, the following media met the
criteria for ICTs: computer, Internet, wireless, mobile phones, satellite connectivity,
and multimedia materials such as CD. On the other hand, the following types did not
meet the criteria and were excluded from the boundary of current research: TV ,
broadcasting, radio, video, land-line telephones, and print.
Second, ICT for Development projects included in the current research were
defined as those having ICTs as the core component of the project. In other words,
this study excluded projects in which ICTs or telecommunications are only of
peripheral concern. For example, projects with an overarching goal such as national
capacity building, privatization, and utility reform were excluded from the selection
process.
The third criterion defined the geographic scope of a project. In this study,
projects targeting specific rural communities or the larger nation were both included.
For example, some projects exclusively focused on a small community area. On the
other hand, some projects had a geographically broader scope, such as improving the
performance of national telecommunications sectors and restructuring regulatory
environments.
The study took two different types of projects into consideration, as
described in Chapter 3. For implementation projects, projects on telecommunications
regulations, telecommunications policy development and implementation, and
restructuring of the developing country were included. Projects on training,

83
education, creating ICT infrastructure including websites, and creating networks
among organizations in developing countries site were included as well. Also,
projects that relate to support for research system, networks, dispatch of experts, and
human resource development using ICTs in the developing countries site were
included. In addition, although a project did not directly deal with the physical
installation of ICTs or its use, capacity building projects such as local language
computing were included.
For knowledge-sharing projects, any types of forums on ICT for
Development agenda satisfied the selection criteria. This included projects that are
related to information and knowledge dissemination, either with or without the use of
ICTs as a tool. Projects in this boundary included the followings: conference,
workshop, meeting, seminar, and joint-publication. Most of these projects had one or
more of the following goals: “dissemination of results to policymakers and
stakeholders”, “propagate the results of its research in other countries of the region”,
“creating training materials”, “information exchange”, “communicate research
findings”, and “facilitating information sharing”. In other words, the commonality
among these projects is that they aim for knowledge sharing and dissemination. To
eliminate projects that do not focus on the knowledge-sharing aspect, projects
primarily focused on program evaluation, monitoring, case study, policy making,
grant, and award were excluded from the scope of the current study.

84
Project Attributes
Once the ICT projects that met the criteria were identified, information about
the projects was coded. These data were collected for two purposes: first, to extract
organizational collaboration networks that were used for the analyses of network
structure (Hypotheses 1 to 3) and dynamics (Hypotheses 4 to 8) and second, to
analyze Hypotheses 9 proposed at the project level. Each project listing included
information on a project’s donor group, funding agency, period, and a link to the
donor webpage that contained detailed information about projects. To maintain the
comparability among numerous donor websites, the following common project
attributes were extracted: project title, year started, year ended, location, project
description, and collaborating organizations. Supplementary information was
gathered from secondary sources including project documents and relevant press
releases.
For the coding of project information, the following principles served as
guidance. First, start year and end year were coded according to the database. When
specified, end year was coded based on planned or actual completion year
(completion date can be 2008 or later). Second, ongoing projects with no indication
of completion date were coded as “ongoing”. At the stage of data transformation,
both cases were treated as having an end year of 2008, therefore producing a right-
censored dataset.

85
Collaborating Organizations
Major analyses of the current study are based on the networks of
collaborating organizations in the ICT for Development projects. Therefore,
information about collaborating organization was coded in the next stage. This
included recipient institution, implementing agency, contractor organizations, and
sponsors. Information about a more complete list of partners was supplemented by
project descriptions and public documents available through the database and project
website. The following rules served as a guideline for coding information on
collaborating organizations.
First, when the partner was an individual (e.g. consultant, researcher) with
no information about affiliated organizations, it was excluded from the category of
collaborating organizations. Second, when a project had two or more separate sectors
involved (for example, telecommunications and electricity), collaborating
organizations that were not directly related to ICT component of the project were
excluded from coding. For instance, a contractor for the electricity component of a
project, such as those working on office furniture was excluded from coding. These
cases appeared in ICT infrastructure projects funded by World Bank and
implemented by national governments. Third, for university collaborations, sub-
departments or centers of universities were coded as the university at large. For
example, Berkman Centre of Harvard Law School was coded as Harvard University.
For government, specific department or ministry was coded separately.

86
Organizational Attributes
For all organizations identified as collaborating organizations, organizational
attributes were coded. Primarily, these were coded from the project information
provided via donor websites. Information on organizational attributes that were not
available from the project database were supplemented by secondary sources,
including organization websites and other publicly available documents. The
information includes organization name, institution type, geographic location,
geographic scope, and ICT-specialized versus general. All of these variables were
used to test hypotheses regarding network structure (Hypotheses 1 to 3) and
dynamics (Hypotheses 4 to 8).
For testing Hypotheses 3, 6a, 9a, and 9b, organizations were classified into
four types, as defined in Chapter 3. These are GO, IGO, NGO, and private/for-profit
corporations.
The geographic location of organizations was based on the location of the
organizational headquarters. When a collaboration partner for a project was an
international branch of a multinational organization, the branch country was coded
instead of the headquarters location. Individual countries were then collapsed into
categories to construct the geographic region variable used for Hypotheses 2, 6a, 9a,
and 9b. World regions were categorized into the following six regions: Africa, Asia,

87
Europe, North America, Latin America and the Caribbean, and Oceania. The
categorization followed data from the UN Statistics Division
5
.
With regard to geographic scope, this study categorized organizations into
three groups: international, regional, and national organizations. The categorization
was determined by whether the organization had any international presence. In other
words, if an organization had worldwide offices and staffs, it was coded as an
international organization. If these offices and staffs were from the region, the
organization fell into the category of a regional organization.
Lastly, to test Hypothesis 6b, organizations were coded either as ICTs-
specialized or general. If an organization is exclusively focused on ICT strategies
such as information technologies and telecommunications, it was coded as ICT-
specialist. If the organization’s activities were not confined to ICT fields, the
organization was coded as belonging to the general category. For example, an
organization focused on traditional media (e.g. BBC World Trust) was coded as a
general organization.
Data Transformation
What constitutes a link is an open question in network studies. In
interorganizational network studies, the term network is used in a variety of contexts
including “partnerships, strategic alliances, interorganizational relationships,
coalitions, cooperative arrangements, or collaborative agreements” (Provan et al.,
2007, p. 480). In both types of networks considered in this study, the definition of

5
http://unstats.un.org/unsd/methods/m49/m49regin.htm

88
link was based on organizations’ affiliations with projects. Another assumption of
linkage in the study was that collaborating organizations were treated as being
networked throughout the duration of the project. In other words, all collaborating
organizations involved in a project were coded as having links with each other from
the start year to the end year. This assumption is not free of limitations, for instance,
in a case where an organization enters into collaboration at a different stage of the
project. Yet, due to the limited details provided by the data, this study adopted a less
restrictive approach to defining linkages.
Information extracted from the database was the formal networks among
organizations for the goal of implementing development projects or sharing
knowledge. The dataset created from the coding process of projects and collaboration
organizations forms an affiliation network, in other words, a two-mode data. Projects
form one set of social unit and organizations form another set of social unit, in which
relations are formed between the two units (Wasserman & Faust, 1994).
For analysis of organization to organization networks, two-mode data were
reorganized to one-mode data. In other words, two-mode networks between
organizations and implementation projects, and between organizations and
knowledge-sharing projects, were transformed to one-mode network among
organizations. This network represents ties among organizations created by jointly
working on implementation or knowledge-sharing projects. Consequently, the dataset
created was a valued symmetric network, with the value indicating the number of
times each organization jointly participated in a project with each other. For both

89
types of networks, one-mode networks of relationship among projects were created
as well. These networks among projects provide information about links between
projects which have common participating organizations. Further, the analysis can
reveal networks among isolated projects that have commonalities in terms of project
attributes, therefore, providing suggestions for transfer of lessons and experiences for
potential replicability and scalability of projects.
In terms of time span, the network data were organized in different ways to
best represent the theoretical and methodological considerations of the hypotheses in
each section. For longitudinal comparison of network structure in Hypothesis 1 and
the analyses of subgroup network structure and network dynamics in Hypotheses 2
and 3, networks were constructed into 3-year windows and data from these time
periods were compared. The strategy of taking multiple years instead of single years
has been adopted in several studies in the past (e.g. Baum et al., 2004). Using
multiple year windows captures interorganizational networks more accurately than
examining a single year because it takes the duration of ties into consideration.
Three-year moving periods were chosen based on the consideration of the average
length of a project: the average length of an implementation network was 4.7 years
and of a knowledge-sharing project was 2.8 years. The first 4 and 2 years from each
network type were excluded from analyses considering the average length of project
duration, so that the analysis periods capture only those years in which both tie
formation and dissolution can be freely made. The last three years were excluded
because the records of projects have not been fully harvested in the database yet. In

90
consequence, three-year windows between 1991 and 2005 were used for
implementation projects and between 1997 and 2005 were used for knowledge-
sharing projects. The above selection procedures generated a final dataset of total
325 projects, among which 212 were implementation projects and 113 were
knowledge-sharing projects.
For the analyses of multiplex dynamics (Hypotheses 4, 5a, 5b, 5c, 6a, 6b, 7,
and 8), the study considered three periods: 1997-1999, 2000-2002, and 2003-2005.
By using multiple year windows, the study attempted to capture the multiplex
dynamics that are present in a single year and have lagged effects for a maximum
two years.
For the analysis of network effects (Hypotheses 9a, 9b, and 9c), this study
considered all years in which the implementation projects existed, in other words, in
the interval from 1987 to 2008. The temporal dimension of the data was considered
in order to create a matrix that represents the differences in years of project initiation.
This procedure yielded a sample of 222 implementation projects and 596 unique
organizations included in the analysis.
The networks were transformed to binary data which indicates the presence
or absence of ties between nodes, without regard to the strength of ties. This
transformation was conducted in order to model and estimate the networks in
MultiNet and PNet, which run based on a dichotomized data format.

91
Analysis
Network analysis (Monge & Contractor, 2003; J. Scott, 2000; Wasserman &
Faust, 1994) provides a systematic approach to capture the complex relations among
a given set of nodes. Several network analysis software programs developed within
the field were used in this study.
Global Network Structure (Hypothesis 1)
Hypotheses 1 involved the comparison of global-level network measures over
time. Degree distribution was used for the analysis, which captures the pattern of
inequality in ties across nodes in a network. Measures of centralization and
heterogeneity are representative of this property (Wasserman & Faust, 1994). A
centralized network implies a hierarchical structure with disproportionate number of
ties for selected nodes, in which ties are integrated around one or a set of nodes. It is
characterized by a densely connected core group and sparsely connected peripheral
nodes. On the other hand, a decentralized, polycentric structure indicates that there is
comparatively higher homogeneity in the degree distribution of nodes. This structure
is represented by a small world graph with the characteristics of high clustering and
small average path length between pairs of nodes (Baldassarri & Diani, 2007; Watts,
1999).
First, the centralization measure is suitable for testing the extent to which a
network manifests a centralized structure (Wasserman & Faust, 1994). While
centralization captures the overall distribution, it is not normalized and cannot be
compared across networks over time. Therefore, a normalized heterogeneity measure

92
was used to examine changes over time between networks of different size and
density. The heterogeneity measure represents diversity, and is also known as a
concentration ratio, the Hirschman-Herfindahl index, or the Herfindahl index
(Borgatti, 2006). The normalized measure divides the Heterogeneity measure by 1-
1/n to control for the size of the network. The measure reaches its minimum value of
0 when the network consists of a single component. On the other hand, when the
network is maximally fragmented, in other words, all isolates, the measure shows its
maximum value of 1 (Borgatti, 2006).
To determine whether the focal networks differ from random networks, this
study compared centralization and heterogeneity measures of the observed network
with random networks that are generated from the same number of nodes and ties.
This method has been proved to be useful in previous literatures on structural
properties (e.g. Baldassarri & Diani, 2007; Baum et al., 2004; Bearman, Moody, &
Stovel, 2004). This method also allows the comparison of networks that are of
different size and density, including evolving networks over multiple years
(Baldassarri & Diani, 2007). In each time period, the comparison of parameters
suggested whether the observed network exhibits significantly different properties
than a network operating on random attachment logics.
The hypothesis predicted changes over time, which involves testing whether
there are statistically significant differences in the equality of tie distribution. This
test can be conducted by examining a node statistic and computing heterogeneity
statistics, such as the standard deviation and variance measures on that variable to

93
reveal the distribution of links (Borgatti, 2005, UCINET Users Group
6
). This study
computed normalized degree centrality scores, which controls for network size, and
compared the standard deviation to test whether two different networks differ in
terms of degree distribution. An F-test was conducted to compare the difference in
standard deviation.
For testing the hypothesis, UCINET6 (Borgatti, Everett, & Freeman, 2002b)
was used. UCINET6 is useful for deriving descriptive measures of network
properties and computing both nodal and global level network properties. UCINET6
was used to compute centrality and centralization. As degree centrality and
centralization are conventionally defined for non-valued data, the network was
dichotomized to a binary network so that normalized degree centrality could be
calculated (Borgatti, Everett, & Freeman, 2002a). Random networks were created
through UCINET6 routines (Random>Erdos-renyi) and variances were calculated in
SPSS based on 1,000 observations.
Blockmodeling Analysis (Hypotheses 2 and 3)
As a method of analyzing local network structure, p* analysis (Crouch,
Wasserman, & Contractor, 1998; Wasserman & Pattison, 1996), a family of
exponential random graph model (ERGM), is widely used. The p* analysis is based
on Markov models, which assume that the probability of a tie between i and j
depends on other ties that involve the same actor(s), i and j (Frank & Strauss, 1986).
In other words, p* models “model nonindependence among dyads by including

6
http://tech.groups.yahoo.com/group/ucinet/message/775

94
parameters for structural features that capture hypothesized dependencies among
ties” (Faust & Skvoretz, 2002, p. 274). In contrast, statistical models based on
traditional Bernoulli random graph distribution are based on the assumption of
dyadic independence. Therefore, this Markov assumption better represents the
complex dependence relations in social networks than the Bernoulli distribution does
(Wasserman et al., 2007). The p* analysis tests whether hypothesized structural
signatures significantly increase the likelihood of the realization of certain network
patterns. Several analysis methods have been developed based on p* analysis to
account for intricate network structures.
MultiNet for Windows 4.55 (Richards & Seary, 2006) is one of the methods
that adopt the p* analysis. The process detects a local level structure of dyads, triads,
k-stars and k-triangles surrounding the focal organization. These parameters show
whether these local processes are operative in the observed network in a significantly
different way than in a network that would have been created by random chance.
Multilevel analysis suggests looking at networks not only at the nodal or
global levels, but also at the population level. A comparable level of analysis from
the network approach is subgroup analysis. MultiNet incorporates blockmodeling
techniques (Seary & Richards, 2000), by using actor attribute information. The
technique is designed to detect the relationships among subgroups of a network
(DiMaggio, 1986; Fienberg, Meyer, & Wasserman, 1981) and estimates inter- and
intra- subgroup parameters. By specifying block structure in the model, it also allows
for testing the likelihood of ties being present between subcategories, for example,

95
within and between specific regions or populations. A significant and positive B
parameter indicates a statistical significance of a given graph configuration. In other
words, when organizations are partitioned into mutually exclusive and exhaustive
subgroups based on external attributes that designate their population membership,
both across-population and within-population networks can be estimated.
In this study, geographic region and population were used as exogenous
attributes for blockmodeling. Further, as the dataset was a symmetric network, the
Edge (Choice) parameter and the Edge within and across blocks parameter were
estimated throughout the analysis. MultiNet also gives overall estimation of model
fit by providing – 2 pseudo likelihood measures based on logistic regression
procedures. This measure allows for testing comparison between the baseline model
and more advanced model with additional parameter combinations. The significance
of a given parameter was tested by comparing whether the decrease in – 2 pseudo
likelihood measure is larger than the chi-square value required for df = 1 at p < 0.01
significance level (Lee et al., 2007; Monge & Matei, 2004).
Multiplex Network Analysis (Hypotheses 4, 5a, 5b, 5c, 6a, 6b, 7, and 8)
Multiplex networks capture the idea that network relations vary in
organizational communities. When organizations engage in more than one type of
linkage, they are treated as forming multiplex ties. Wasserman and Faust (1994)
introduced models that measure the interrelatedness of multiple relations, as used in
numerous studies by Wasserman and colleagues (e.g. Fienberg et al., 1981;
Wasserman, 1987).

96
Based on these earlier studies, p* models have been extended in a variety of
ways over the past decade (Wasserman et al., 2007). PNet (Robins et al., 2007) is a
recently developed program. PNet uses Monte Carlo Markov Chain (MCMC)
maximum likelihood estimation to obtain convergent estimates that will fit the
observed network (Robins et al., 2007). The MCMC maximum likelihood process
simulates random graphs from a starting set of parameter values and subsequently
refines the values by comparing the observed network with random network until the
parameter stabilizes. Therefore, the method gives more reliable standard errors for
the estimates than the pseudo-likelihood parameter estimates (Robins et al., 2007). In
addition, the advantage of PNet over earlier Markov random graph models is that it
includes a wider range of higher order parameters, especially those of transitivity,
which allow fitting models for many cases including complex social networks
(Robins et al., 2007). These new specifications have been found to be successful in
overcoming degeneracy (Igarashi, Robins, & Pattison, 2006).
More recently, network analysis models and programs have been developed
for the purpose of analyzing multiplex networks. In particular, the interdependency
among multiplex networks can be tested by a multivariate p* analysis as suggested
by Robins and Pattison (2006) and exemplified in Lomi and Pattison’s (2006) study
of multiplex tie interdependencies. With its multivariate extension, the p* method
allows testing of the dependence of one set of network tie variables on another set of
network tie variables at the levels of dyads and triads. The current study used XPNet
for this purpose, which is an extension of PNet that incorporates the analysis of

97
multivariate networks. Parameters were selected to test hypotheses that examine the
dynamics between multiple networks.
Hypotheses in this section were tested by examining parameter estimates,
which indicates the comparison between the structural properties of observed and
random networks. Positive and significant parameter estimates imply that the
hypothesized graph configurations are present in the observed networks with a
statistically significant difference than what would be expected by random chance.
On the other hand, significant negative parameter estimates indicate that fewer
configurations are present than what would be expected by chance (Robins et al.,
2007). PNet/XPNet allows inclusion of both structural and attribute parameters in
model estimation (Pattison & Robins, in press; Wang et al., 2008). Since network ties
in the current study were non-directional, parameters for nondirected graph were
used for the analysis of all hypotheses.
PNet/XPNet estimation aims for obtaining a parameter or a combination of
parameters that capture the observed structural characteristics. When the parameter
estimates do not stabilize, the model is considered as nonconvergent, or degenerate
(Wasserman et al., 2007). For all hypotheses, the basic model estimation started with
the inclusion of a structural parameter for tie formation, in other words, density,
which is represented by the Edge parameter. The K-star and K-triangle parameters
are suggested as graph statistics that are useful for fitting most of the standard
networks. These parameters have been found to improve the odds of model
convergence (Robins et al., 2007).

98
It is suggested that many possible parameter combinations should be
explored to find the model that best empirically represents the given dataset
(Goodreau, 2007). Parameter combinations were estimated in a stepwise function,
starting from a set of basic parameters and adding each of a set of parameters to
determine which contributes to the improvement of model fit. Parameters that
yielded significance was added to the next model run, while those that were
nonsignificant were dropped from subsequent model-building process (Goodreau,
2007). To increase the odds of model convergence, different multiplication factors up
to 1,000 were examined, which implies that the program tries a larger number of
trials for fitting parameter estimates. As suggested by Harrigan (2008), parameters
estimated from a previous model were updated in the subsequent model to improve
the model fit.
Model selection was based on two strategies. First, a model with a good fit
should show convergence of parameter values and be non-degenerate as well. For a
good converged model, it is suggested that the t-ratios for all parameter estimates be
less than or close to 0.1 in absolute value (Robins et al., 2007). It is to be noted that
this convergence t-ratio is distinct from the conventional t-statistic, which is defined
as estimate divided by standard error. Second, PNet/XPNet has a goodness of fit
function to evaluate the model fit. The goal is “to investigate how well the model
parameters succeed in replicating features of the observed graph that are not
explicitly modeled” (Robins et al., 2007, p. 206). With the parameters set to the same
values in the converged model, goodness-of-fit tests generate new networks

99
according to the probability distribution implied by the fit model (Goodreau, 2007).
It is suggested that t-statistics for every parameter be below 2 (absolute value) for a
good fit, including parameters that are not estimated in the original model (Robins,
Pattison, & Wang, 2008).
All of the following hypotheses were tested with XPNet. Hypothesis 4
examined whether a tie in one network type will lead to tie formation in another
network type. In other words, the hypothesis tests whether two organizations which
co-participate in knowledge-sharing project are likely to be collaborators in
implementation projects as well, and vice versa. The estimation was conducted with
the EdgeAB parameter. A significant and positive EdgeAB would suggest that
organizational dyads with one type of tie are likely to have another type of tie as well.
Hypothesis 5a tested whether the existence of a common third-party
increases the likelihood of tie formation between organizational dyads. In other
words, the hypothesis tests whether the structural property of triads is present in the
observed network. This hypothesis was tested in a unimodal context, considering
only one type of tie at a time. The TriangleA and TriangleB parameter was used to
estimate this effect for each network separately from each other. A significant and
positive TriangleA and TriangleB would suggest that if organization A ties with
organization C and organization B ties to organization C, it is likely that organization
A and B have a tie with each other. Hypothesis 5b extended the network dynamics of
Hypothesis 5a to the multiplex context. This estimation was conducted with the
TriangleAAB parameter. A significant and positive TriangleAAB would suggest that

100
the above mechanism is observed across two types of networks. Hypothesis 5c can
be tested with a higher order transitivity structure, captured by the K-TriangleABA
parameter. This parameter captures dense regions, in which “configurations with k
separate triangles sharing one edge” are present (Wasserman et al., 2007, p. 50).
The next set of hypotheses modeled exogenous attributes at the node level.
Hypothesis 6a examined whether dyads of organizations in the same resource space
are likely to form a larger number of multiplex ties than those dyads that exist in
different resource spaces. For attribute parameters, two categorical attributes were
used: geographic location and organizational population. Parameters for categorical
attributes were adopted for this analysis, in which the categories were geographic
locations and organizational populations. The estimation was done with the Same-
category-AB parameter. A significant and positive Same-category-AB would suggest
that a tie is more likely to exist between two organizations that are in the same
category of attributes. Hypothesis 6b examined whether generalist organizations
show higher likelihood of forming multiplex ties than specialist organizations. In
order to operationalize generalist versus specialist organizations, two organizational
attributes were adopted: first, the geographic boundary of organization’s scope, and
second, whether an organization’s activities were general or ICT-focused. Based on
this formulation, two binary nodal attributes were created: international versus non-
international (regional and national), and generalized versus specialized. In other
words, if organizations have general functions and span a geographic scope
exceeding a single nation, they were treated as generalist organizations. The RAB

101
parameter was used to test this hypothesis. A positive and significant RAB parameter
would suggest that there is a higher likelihood of tie formation for organizations that
possess the hypothesized binary attribute.
Hypothesis 7 suggested that the structural position of organizations would
influence the likelihood of forming ties with partner organizations across multiple
relations. In particular, the hypothesis examined whether the summed centrality of
organizational dyads influences the likelihood of multiplex tie formation. In order to
make comparison across different sized networks possible, normalized degree
centrality was used. Normalized degree centrality is a centrality measure divided by
the maximum possible ties for a network of that size, and was obtained from
UCINET6 for all nodes in both implementation and knowledge-sharing networks in
each time period. Centrality of nodes in a dyad was treated as continuous node
attributes. The SumAB parameter was used and a significant and positive SumAB
would suggest that a tie is more likely to exist between two organizations that have
higher centrality collectively.
Lastly, Hypothesis 8 predicted an increase in the number of repeated dyad
formations for the same organizations across multiplex networks over time. The
parameter used for the analysis was the EdgeAB parameter. While standardized
parameter estimates can directly be compared across networks (Faust & Skvoretz,
2002), parameters from XPNet are not standardized, therefore not comparable across
networks of different size as in this case. Therefore, the comparison was conducted
by examining the empirical frequency of the multiplex tie configuration in each year,

102
a method suggested by Wasserman and Faust (1994). For this purpose, the counts of
ties were obtained from the goodness of fit run in XPNet. The study compared the
percentage of the realized count of the EdgeAB graph statistics out of the total
possible EdgeAB (multiplex ties) configuration in the graph, which is n(n-1)/2 for a
nondirectional network given that n is the number of common nodes in two networks.
All parameters used in the study are summarized and illustrated in Table 4.
For each hypothesis, one or more parameters were selected from the XPNet
specifications. XPNet parameters are intended to investigate multiplex dynamics, as
represented in the co-existence of black and red edges in each structural illustration.

103
Table 4: Visual Representation of XPNet Parameters (Adapted from Wang et al.,
2008)
Legend
Network A

Network B

Nodes with attribute

Nodes without attribute

Graph Statistics
Hypothesis Parameter Illustration
Hypothesis 4 EdgeAB

Hypothesis 5a
TriangleA
TriangleB

Hypothesis 5b
TriangleAAB

TriangleABB

Hypothesis 5c K-TriangleABA

Hypothesis 6a Same-category-AB

Hypothesis 6b RAB

Hypothesis 7 SumAB

Hypothesis 8 Increase in EdgeAB

+

104
Niche Overlap and Structural Equivalence (Hypotheses 9a, 9b, and 9c)
The testing of Hypotheses 9a, 9b, and 9c was preceded by three steps in
order to measure the three major variables in the hypotheses: niche overlap,
structural equivalence, and sharing of technologies and applications. First, niche
overlap was measured by similarities in the two project features: regions and goals.
Data on project regions were obtained from the information coded on project
attributes. Regions were categorized into five areas as mentioned earlier: Africa, Asia,
Europe, Latin America and the Caribbean, and Oceania. Project goals were extracted
from content analysis of project descriptions. Based on the MDGs discussed in
Chapter 3 (Table 2), project goals were coded into the following five categories:
economic development, education, women empowerment, health, and environment.
In addition, two categories were added: governance/policy and enhancement of
infrastructure and systems. All projects were assigned to one or more categories in
these two dimensions. If a project spanned multiple categories, such as being
involved in more than one region or more than one goal, it was assigned to all of the
categories. Subsequently, these two dimensions, project regions and goals, were used
to produce matrices that represent niche overlap respectively. For all possible pairs of
projects, overlap in the categories was computed, yielding a project-to-project matrix
with niche overlap that indicates similarities in resource space.
Second, structural equivalence between pairs of projects was measured in
UCINET 6 (Network>Roles&Positions>Structural>Profile). Correlation coefficient
measures were used to capture the extent of structural equivalence of pairs of nodes

105
in a valued network. Pearson product-moment correlation coefficient ranges from -1
to +1, where +1 means that two actors have exactly the same tie to other actors
(Hanneman & Riddle, 2005; Wasserman & Faust, 1994). In a 2-mode network
between project and organizations, structural equivalence measures of projects
indicates the extent to which two projects have similar relational patterns to
organizations. The output was formatted as a project by project matrix.
Third, types of technologies and applications adopted in each project were
extracted from the content analysis of project descriptions. Once all keywords
relevant to the technologies and applications adopted in the projects were recorded, a
consolidated list was created to represent meaningful groups of distinct technologies
and applications.
7
All projects were assigned to one or multiple types, based on the
specific form of deployment that was the primary concern of each project. Lastly, a
project-to-project similarity matrix was created based on the common occurrence of
these technology and application types in all possible pairs of projects.
In addition to the three matrices suggested in the hypotheses, it was
important to control for other variables that may influence the dependent matrices.
The study considered the period effect in the course of organizational evolution,
which involves the impact of historical events on organizations (Aldrich & Reuf,

7
The content analysis yielded the following types of technologies and applications: wireless, mobile
phone, satellite, community telecentres, Distance Learning Center (DLC), web-based tools and
applications, ICT literacy/ local content, electronic library, school networking, information systems/
computer networks, distance education, Internet exchange point, last-mile, universal access/service,
cellular licenses, payphone, spectrum management, rural telecom development fund, private sector
participation/ privatization, open source, and computer recycling.

106
2006). The study assumed that this previously-known antecedent of evolutionary
effect may influence the structural equivalence and sharing of ideas between projects.
Therefore, the study included a matrix to represent the temporal proximity of
projects. The matrix was created based on the similarity of years in which each pair
of projects was initiated.
As the next step, Hypothesis 9a, 9b, and 9c tested the relationship between
these multiple matrices. A popular methodology of testing whether there is an
association between two networks on a dyadic basis is the Quadratic Assignment
Procedure (QAP). An extended model, Multiple Regression Quadratic Assignment
Procedure (MRQAP)(Krackhardt, 1987, 1988), tests if there are dependency
relations among multiple networks. This analysis is comparable to a multiple
regression in which a dependent matrix is regressed on one or more independent
matrices. The MRQAP procedure was performed in UCINET 6 (Tools>Testing
Hypotheses>Dyadic (QAP)>QAP Regression> Double Dekker Semi-Partialling
MRQAP) to calculate the association between matrices.
Visualization
For visualization of global and subgroup networks, Netdraw, a computer
routine built into UCINET6, was used. Further, visualization of two-mode data was
conducted with the tools readily available in UCINET6 (Netdraw>Open>UNINET
dataset>2 mode network). A two-mode graph shows both organizations and projects
and the connections between those two sets of nodes. Organizations that are close
together represent that they have similar profiles of projects. Projects that are close

107
together indicate that they are similar in terms of having organizations in common
(Hanneman & Riddle, 2005).

108
CHAPTER 5: RESULTS
Descriptive Results
Figure 3 shows the descriptive statistics of the dataset. The graph represents
the number of projects started as well as the number of projects present in a given
year between 1991 and 2005, which were included in the final analyses of the current
study. The line graph indicates the number of organizations that were present in each
type of project. The start year of implementation projects coded from the dataset
spanned from 1991 to 2005. The number grew considerably in the late 1990s, with
the largest number of projects (32) initiated in 2000. Year 2001 had the largest
number of total projects present (127). In terms of knowledge-sharing projects, years
of project initiation spanned from 1995 to 2005, with the largest number of projects
(23) started in 2003. In the same year, 49 projects were present in total. The total
count of projects by year showed a fairly similar trend in the two networks, with the
largest number recorded between 2001 and 2003. As was explained in the method
section, the dataset used in the study was both left- and right-censored. In addition,
the decline of the number of projects in the later years is a result of the nature of the
database, since completed projects have higher likelihood of being harvested in the
database than ongoing projects do. For the final dataset of 325 projects used in the
analyses of network structure and dynamics, the average length of projects was 4.6
years for implementation projects and 3.0 for knowledge-sharing projects.

109
Figure 3: Number of Projects and Organizations by Year
Note. I-project refers to implementation project; K-project refers to knowledge-sharing project
0
20
40
60
80
100
120
140
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
Projects
0
50
100
150
200
250
300
350
400
450
Org.s
I-project Start I-project Total
K-project Start K-project Total
I-project Orgs Total K-project Orgs Total

110
For the analyses of network effects, the study coded two major attributes of
implementation projects: geographic location and project sector. Table 5 presents
frequency counts and percentages of project regions and sectors. Over 45% of
implementation projects were involved in the African region. Over 25% of the
projects were implemented in Asia, and over 15% in Latin America and the
Caribbean. In terms of project sector, infrastructure/systems (26.4%), education
(22.6%), and economic development (21.6%) were the major goals adopted in the
implementation projects. Governance and policy (18.8%) was the next largest issue
of the implementation projects.
Table 5: Implementation Projects Breakdown by Attributes
Variables Description Categories I- Network
Total # of Unique
Projects
212
Geographic Region Asia 73 (29%)
Africa 119 (47%)
Europe 11 (4%)
Latin America & Caribbean 40 (16%)
North America 1 (0%)
Oceania 11 (4%)
Project Sector Economic Development 63 (22%)
Education 66 (23%)
Women Empowerment 6 (2%)
Health 16 (6%)
Environment 9 (3%)
Governance, Policy 55 (19%)
Infrastructure, Systems 77 (26%)
Note. The frequency counts include multiple-category codings, therefore the sum exceeds
the total number of unique projects.

Table 6 summarizes the number of organizations, linkages, and density of
the two networks by year. The density of the network is shown to decrease and

111
stabilize around years 2002, 2003, and 2004, and to increase from 2005 for both
types of networks. Two types of networks show differences in the number of
collaborating organizations. The total number of unique organizations was 578 in the
implementation network and 174 in the knowledge-sharing network. The average
number of organizations in a project was 4.63 for an implementation project and 3.45
for a knowledge-sharing project, although there was a substantial variance between
projects.
Table 6: Descriptive Statistics
Number of nodes Number of ties Density
I-
Network
K-
Network
I-
Network
K-
Network
I-
Network
K-
Network
1991 18 27 0.177
1992 32 92 0.186
1993 39 115 0.155
1994 46 136 0.131
1995 59 5 178 6 0.104 0.600
1996 80 6 278 9 0.088 0.600
1997 115 18 598 30 0.091 0.196
1998 155 33 698 66 0.058 0.125
1999 249 56 1211 138 0.039 0.089
2000 322 57 1566 133 0.030 0.083
2001 392 65 2158 142 0.028 0.068
2002 410 67 2433 156 0.029 0.070
2003 399 104 2463 377 0.031 0.069
2004 390 103 2339 385 0.031 0.071
2005 308 87 1647 325 0.035 0.085
Note. I- Network refers to implementation network; K- Network refers to knowledge-
sharing network

Presented in Table 7 is a breakdown of collaborating organizations in each
network by four types of attributes: geographic region, geographic scope of activity,
institution type, and general versus ICT-focused organizations. The composition of

112
organizations by geographic scope differed between the two networks. Specifically,
the percentage of domestic organizations was higher in the implementation networks
(64%) than in the knowledge-sharing networks (43%), while the percentage of
international organizations was higher in the knowledge-sharing networks (39%)
than in the implementation networks (30%). The study included organizations from
six continents. While the composition of organizations by geographic location was
quite similar, a couple of differences were noticeable. The knowledge-sharing
network showed a relatively higher percentage of Latin American and Caribbean
organizations (15%) than in the implementation network (10%). On the other hand,
the percentage of Asian organizations was higher in the implementation network
(24%) than in the knowledge-sharing network (16%). The knowledge-sharing
networks had a higher percentage of collaborating organizations that were IGOs
(17%) or NGOs (53%) compared to those in the implementation networks (7% and
40% respectively). The representation of for-profit sector organizations was higher in
the implementation networks (28%) than in the knowledge-sharing networks (9%).
In respect to general versus ICTs-focused organizations, the two networks showed
relatively similar percentage representation of both types.

113
Table 7: Organizations Breakdown by Attributes
Variables Description Categories I- Network K- Network
Total # of Unique
Organizations
578 174
Geographic Region Asia 138 (24%) 27 (16%)
Africa 152 (26%) 48 (28%)
Europe 101 (18%) 33 (19%)
Latin America &
Caribbean
57 (10%) 26 (15%)
North America 117 (20%) 38 (23%)
Oceania 13 (2%) 2 (1%)
Institution Type Governmental 149 (26%) 36 (21%)
Intergovernmental 38 (7%) 30 (17%)
Nongovernmental 229 (40%) 92 (53%)
Private - For-profit 162 (28%) 16 (9%)
Geographic Scope International 174 (30%) 68 (39%)
Regional 34 (6%) 31 (18%)
National 370 (64%) 75 (43%)
General vs. ICTs-
focused
General 405 (70%) 123 (71%)
ICTs-focused 173 (30%) 51 (29%)

Major players in the implementation and knowledge-sharing networks are
shown in Tables 8 and 9. Table 8 shows five organizations with the highest degrees
in the implementation network in each of the three-year time periods. Measures of
degree and normalized degree are shown as well. Across all time periods, the most
central organization was World Bank. From the mid 1990s to the mid 2000s, most of
the central organizations were governmental developmental agencies such as the
United States Agency for International Development (USAID), the Swedish
International Development Cooperation Agency (SIDA), the Canadian International
Developmental Agency (CIDA), and the Danish Development Assistance Agency
(DANIDA). Among IGOs, the UNDP was the most prominent organization.

114
Table 8: Degree Centrality of Organizations in I- Network
Year Organization Name Degree Normalized
degree
1991-1993 World Bank 38 100
Government of France 10 26.316
Government of Indonesia 10 26.316
Perumtel / Postel 10 26.316
Sumitomo Trust And Banking Co. Ltd. 10 26.316
1994-1996 World Bank 51 60.714
Swedish International Development
Cooperation Agency (SIDA)
22 26.19
Danish Development Assistance Agency
(DANIDA)
20 23.81
International Development Research Center
(IDRC)
19 22.619
Canadian International Developmental Agency
(CIDA)
14 16.667
1997-1999 World Bank 112 44.444
International Development Research Center
(IDRC)
103 40.873
United States Agency for International
Development (USAID)
55 21.825
United Nations Development Programme
(UNDP)
46 18.254
Swedish International Development
Cooperation Agency (SIDA)
28 11.111
2000-2002 World Bank 194 41.991
International Development Research Center
(IDRC)
188 40.693
United States Agency for International
Development (USAID)
78 16.883
Canadian International Developmental Agency
(CIDA)
63 13.636
United Nations Development Programme
(UNDP)
60 12.987
2003-2005 World Bank 198 43.326
International Development Research Center
(IDRC)
175 38.293
United States Agency for International
Development (USAID)
97 21.225
Canadian International Developmental Agency
(CIDA)
63 13.786
McCarthy Tétrault LLP 61 13.348

115
Table 9 shows five organizations with the highest degrees in the knowledge-
sharing network in each of the three-year time periods. In all three time periods,
IDRC was the most central organization represented in the knowledge-sharing
network. IDRC was followed by a number of IGOs including the UNESCO, the
United Nations Economic Commission for Africa (UNECA), the World Bank, and
the ITU. The Department for International Development (DFID) was ranked as the
second highest in degree centrality in the later two time periods. Overall, the
knowledge-sharing network was largely dominated by IGOs than GOs compared to
the implementation-network, reflecting that IGOs have been playing a vital role in
various conferences and meetings that have been organized in the ICT for
Development field.

116
Table 9: Degree Centrality of Organizations in K- Network
Year Organization Name Degree Normalized
degree
1997-1999 International Development Research Center
(IDRC)
44 77.193
World Bank 21 36.842
United Nations Educational, Scientific and
Cultural Organization (UNESCO)
13 22.807
Commonwealth of Learning (COL) 12 21.053
International Telecommunication Union (ITU) 10 17.544
2000-2002 International Development Research Center
(IDRC)
64 71.91
Department for International Development
(DFID)
23 25.843
United Nations Educational, Scientific and
Cultural Organization (UNESCO)
19 21.348
World Bank 12 13.483
United Nations Economic Commission for
Africa (UNECA)
10 11.236
2003-2005 International Development Research Center
(IDRC)
93 67.391
Department for International Development
(DFID)
42 30.435
United Nations Economic Commission for
Africa (UNECA)
33 23.913
United Nations Educational, Scientific and
Cultural Organization (UNESCO)
30 21.739
World Bank 29 21.014

Network Structure Analysis
Hypothesis 1
Hypothesis 1 asserted that interorganizational networks will show greater
equality over time. The hypothesis was tested by measures of network centralization,
heterogeneity, and degree distributions. Centralization shows “how variable or
heterogeneous the actor centralities are” (Wasserman & Faust, 1994, p. 176). When
the graph resembles a star graph by one node completely dominating the other actors,
the centralization measure equals 1. When all nodes in a graph have the same

117
centrality index, the network centralization measure equals 0 (Wasserman & Faust,
1994, p. 177). An initial exploration demonstrated that centralization and
heterogeneity scores are higher than random networks in all time periods, indicating
that observed networks exhibit a highly centralized structure overall (Table 10).
Comparison of normalized heterogeneity measure revealed changes over
time. In the implementation network, the measure decreased from 2.89% in 1991-
1993 to 0.37% in 2003-2005. The same pattern was found in the knowledge-sharing
network, in which the measure decreased from 2.74% in 1997-1999 to 1.26% in
2003-2005. The result implies that the equality in the ties among interorganizational
network has increased over time. In other words, the polarization between the core
and the periphery has decreased.
Table 10: Centralization and Heterogeneity Measures over Time
I- Network K- Network
Centralization Heterogeneity
b
Centralization Heterogeneity
b

Obser-
ved
Ran-
dom
a

Obser-
ved
Ran-
dom
a

Obser-
ved
Ran-
dom
a

Obser-
ved
Ran-
dom
a

1991-
1993
89.05
%
14.96
%
2.89% 0.39%
1994-
1996
54.03
%
8.44% 1.19% 0.17%
1997-
1999
40.97
%
4.07% 0.56% 0.04% 70.86
%
10.08
%
2.74% 0.32%
2000-
2002
39.73
%
2.57% 0.40% 0.02% 68.00
%
7.16% 2.52% 0.23%
2003-
2005
40.91
%
2.70% 0.37% 0.02% 62.78
%
5.95% 1.26% 0.10%
Note. a. Average of measures was obtained from 250 random networks of the same size and
density as the observed network in each time period.
Note. b. All heterogeneity measures shown in the table are normalized.

118
The statistical significance of differences in degree distribution over years
was tested by the standard deviation of the degree centrality of nodes. The
significance of the differences between the two standard deviations was tested by
computing an F test value, where the larger of the two variances is divided by the
smaller of the two variances. For example, in order to compare between 2000-2002
and 2003-2005, an F-test value was calculated (F = (20.398)
2
/(10.88)
2
= 3.5149). The
difference between the two standard deviations was statistically significant at p < .01
level, with the F value larger than the critical F value of 0.77 required with df = 252
for the numerator and df = 462 for the denominator. The results indicate that in 2000-
2002, the variance of degree was smaller than in 1997-1999. In other words, the
degree distribution became more equal over time. The same analysis was conducted
for other time periods as well. Table 11 reports that the differences between years
were significant except for the implementation network between the two time
periods of 2000-2002 and 2003-2005.
Overall, the results from both descriptive (centralization, heterogeneity) and
statistical (standard deviation of degree centrality) analyses supported Hypothesis 1,
indicating that the network at the global level became less centralized over time in
both implementation and knowledge-sharing networks.

119
Table 11: Significance Tests of Differences in Degree Distribution over Years
I- Network K- Network
Variance of
degree
centrality
Sig.
a
Variance of
degree
centrality
Sig.
a

1991-
1993
264.15
1994-
1996
63.332 Decrease (p
< .01)

1997-
1999
20.398 Decrease (p
< .01)
120.143
2000-
2002
10.88 Decrease (p
< .01)
65.805 Decrease (p
< .01)
2003-
2005
11.402 Increase (n.s.) 52.71 Decrease (p
< .01)
Note. a. Significance of difference compared to the immediately preceding time period.

Hypothesis 2
Hypothesis 2 tested ties across and within regions. In particular, the
hypothesis tested whether the structural tendencies toward choice across and within
blocks of geographic region were significant predictors of the observed network
structure.
The analysis was conducted separately on the implementation and
knowledge-sharing networks. First, the implementation network was examined. The
results of a hierarchical fitting of global parameters provided in Table 12 show that
the log likelihood measures (a measure of badness of fit) decreased significantly
when choice within blocks and across blocks parameters were added. In the time
period 1991-1993, Model 2 and Model 3 were compared against Model 1. Model 1 is
the baseline model reflecting the hypothesis that probabilities of graph realization do
not exhibit specific structural properties. The baseline badness of fit measure was

120
1279.312 and decreased as successive models produced a better fit. A statistical
assessment was performed to determine whether the best fit was simply a result of
more explanatory variables, by comparing the decrease in -2 log pseudo-likelihood (-
2LPL) value with the χ
2
associated with the change in degrees of freedom due to the
process of adding parameters. For blocks of geographic region, the decrease in -
2LPL with the addition of the within blocks parameter from Model 1 to Model 2 was
15.65. This difference was larger than 13.28 points required by the χ
2
critical value
for p < .01 significance at df = 4, showing that choice within blocks was an important
property. When the parameter for choice across geographic region was added in
Model 3, the -2LPL measure decreased by 274.21 points to 1005.10, larger than the
24.32 points for p < .01 significance at df = 7. For the time period 2003-2005, the
same analysis was conducted. The comparison between the reduction of -2LPL
against the χ
2
associated with the change in degrees of freedom revealed that both
choice between and across blocks of regions produced a better fit than the baseline
model, by reducing the -2LPL by 2039.81 and 2971.42 respectively.

121
Table 12: Goodness of Fit Indices for the Models of Choice in I- Network
Note. Each p value represents the significance of the model compared to the baseline model
represented in Model 1.

Once the significance of the model specification was tested, the individual
parameters were investigated. Due to the differences in network size, the B
parameters across time periods cannot be directly compared in terms of their effect
size. Therefore, the analysis compared whether a given parameter was a statistically
significant predictor of relational ties in each time period. In 1991-1993, the positive
and significant predictors of the observed implementation network were: ties within-
Europe (B
Europe-Europe
= 0.86, p < .01), between Asia and Europe (B
Asia-Europe
= 1.07, p
< .01), between Asia and Latin America (B
Asia-L. America
= 2.96, p < .01), between
Africa and Latin America (B
Africa-L. America
= 2.12, p < .01), between Europe and Latin
America (B
Europe-L. America
= 3.22, p < .01), and between Latin America and North
America (B
L. America-N. America
= 0.52, p < .01). Therefore, a majority of across-region
ties positively predicted the network in 1991-1993. On the other hand,
interorganizational ties in 2003-2005 revealed a large contrast. In this time period,
most of the within-region ties were significant. B

values for Asia, Europe, Latin
Model
Number of
Parameters (df)
-2LPL
(Badness of
fit)
p
Geographic region (1991-1993)
1 Choice 1 (1) 1279.31
2 Choice + Choice within blocks 2 (5) 1263.67 0.01
3 Choice + Choice across blocks 2 (8) 1005.10 0.001
Geographic region (2003-2005)
1 Choice 1 (1) 50422.89
2 Choice + Choice within blocks 2 (7) 48383.08 0.001
3 Choice + Choice across blocks 2 (16) 47451.47 0.001

122
America, North America, and Oceania were 1.46, 0.66, 1.43, 1.44, and 2.92
respectively, all significant at p < .01, except those within Africa. All of the across-
region ties were negative and significant in 2003-2005, showing that the
interorganizational networks for project implementation moved towards a stronger
within-regional pattern compared to the earlier period. The parameters and p values
are shown in Tables 13 and 14.
Table 13: p* Analysis within Blocks of Geographic Regions: I- Network
B S. Error PLWald p exp(B)
1991-1993 Global -1.76 0.08 440.55 <0.01 0.17
Asia 0.23 0.29 0.65 >0.25 1.26
Africa -0.27 0.26 1.07 >0.25 0.76
Europe 0.86 0.25 12.17 <0.01 2.37
L. America N/A
a

N. America 1.07 0.62 3.00 <0.10 2.92
2003-2005 Global -3.95 0.02 48505.15 <0.01 0.02
Asia 1.46 0.04 1282.06 <0.01 4.32
Africa -0.03 0.07 0.23 >0.25 0.97
Europe 0.66 0.07 94.24 <0.01 1.93
L. America 1.43 0.08 360.68 <0.01 4.18
N. America 1.44 0.04 1185.09 <0.01 4.23
Oceania 2.92 0.21 199.71 <0.01 18.61
Note. a. Parameters with no/rare counts and excluded from model fit.

123
Table 14: p* Analysis across Blocks of Geographic Regions: I- Network
B S. Error PLWald p exp(B)
1991-1993 Global -2.12 0.13 263.90 <0.01 0.12
Asia-Africa -1.33 0.38 12.18 <0.01 0.26
Asia-Europe 1.07 0.21 26.68 <0.01 2.92
Asia-L. America 2.96 0.37 64.58 <0.01 19.37
Asia-N. America N/A
a

Africa-Europe -0.91 0.32 7.97 <0.01 0.4
Africa-L. America 2.12 0.31 47.71 <0.01 8.3

Africa-N.
America
N/A
a

Europe-L.
America
3.22 0.39 68.78 <0.01 24.91

Europe-N.
America
N/A
a

L. America-N.
America
2.12 0.52 16.78 <0.01 8.3
2003-2005 Global -2.87 0.02 17666.06 <0.01 0.06
Asia-Africa -3.24 0.14 511.49 <0.01 0.04
Asia-Europe -1.35 0.07 389.31 <0.01 0.26
Asia-L. America -2.33 0.13 304.69 <0.01 0.1
Asia-N. America -0.78 0.05 245.97 <0.01 0.46
Asia-Oceania -0.87 0.14 37.30 <0.01 0.42
Africa-Europe -1.02 0.06 318.68 <0.01 0.36
Africa-L. America -3.71 0.25 217.77 <0.01 0.02

Africa-N.
America
-0.56 0.04 161.36 <0.01 0.57
Africa -Oceania -4.23 0.71 35.76 <0.01 0.01

Europe-L.
America
-1.39 0.10 213.89 <0.01 0.25

Europe-N.
America
-0.39 0.05 69.42 <0.01 0.68
Europe -Oceania -0.68 0.15 22.19 <0.01 0.5

L. America-N.
America
-0.62 0.06 99.20 <0.01 0.54

L. America-
Oceania
-1.26 0.24 27.97 <0.01 0.28

N. America-
Oceania
-0.33 0.11 8.54 <0.01 0.72
Note. a. Parameters with no/rare counts and excluded from model fit.

124
Second, knowledge-sharing networks were compared between the two time
periods as well. It was shown that the choice parameters between and across blocks
significantly decreased the -2LPL value (Table 15). In 1997-1999, the addition of
choice between region parameter reduced the -2LPL from 1965.28 to 1928.24. The
difference of 36.34 was larger than the critical χ
2
value of 18.47 for p < .001
significance at df = 4. Choice across region parameter was shown to increase the
model fit to a significant level (decrease of -2LPL value was 151.88, larger than
29.59 required for df = 10 at p < .001 level). For the time period 2003-2005, both
choice between and across regions parameter significantly contributed to the model
fit.
Table 15: Goodness of Fit Indices for the Models of Choice in K- Network
Note. Each p value represents the significance of the model compared to the baseline model
represented in Model 1.

Changes in the parameter values were examined. As suggested in Table 16,
in the period of 1997-1999, within-region parameters were positive and significant
for Latin America (B
L. America-L. America
= 0.70, p < .01) and North America (B
N. America-N.
Model
Number of
Parameters (df)
-2LPL
(Badness of
fit)
p
Geographic region (1997-1999)
1 Choice 1 (1) 1965.28
2 Choice + Choice within blocks 2 (5) 1928.94 0.001
3 Choice + Choice across blocks 2 (11) 1813.40 0.001
Geographic region (2003-2005)
1 Choice 1 (1) 1965.28
2 Choice + Choice within blocks 2 (4) 1944.48 0.001
3 Choice + Choice across blocks 2 (4) 1941.266 0.001

125
America
= 1.22, p < .01). Across regions, only the parameter for ties between Europe
and North America was positive and significant (B
Europe-N. America
= 0.42, p < .01). By
contrast, in 2003-2005, within-region ties were significant in four regions, Asia
(B
Asia-Asia
= 0.48, p < .01), Africa (B
Africa-Africa
= 0.32, p < .025), Europe (B
Europe-Europe
= 1.08, p < .01), and North America (B
N. America-N. America
= 1.30, p < .01). For across-
region ties, most of the parameter values were negative and significant as shown in
Table 17, indicating that across-region ties did not significantly explain the observed
knowledge-sharing network. In summary, both types of networks showed transitions
towards the increase of within-region ties over time, supporting the predictions of
Hypothesis 2.
Table 16: p* Analysis within Blocks of Geographic Regions: K- Network
B S. Error PLWald p exp(B)
1997-1999 Global -2.49 0.07 1148.31 <0.01 0.08
Asia 0.70 0.45 2.46 >0.10 2.02
Africa 0.03 0.23 0.02 >0.25 1.03
Europe N/Aa
L. America 0.70 0.22 9.78 <0.01 2.02
N. America 1.22 0.21 34.34 <0.01 3.37
2003-2005 Global -3.06 0.04 6134.80 <0.01 0.05
Asia 0.48 0.16 8.40 <0.01 1.61
Africa 0.32 0.13 5.75 <0.03 1.38
Europe 1.08 0.11 99.04 <0.01 2.93
L. America -1.58 0.71 4.93 <0.05 0.21
N. America 1.30 0.10 184.00 <0.01 3.66
Oceania N/Aa
Note. a. Parameters with no/rare counts and excluded from model fit.

126
Table 17: p* Analysis across Blocks of Geographic Regions: K- Network
B S. Error PLWald p exp(B)
1997-1999 Global -1.94 0.11 296.01 <0.01 0.14
Asia-Africa -1.77 0.43 17.16 <0.01 0.17
Asia-Europe -1.06 0.52 4.05 <0.05 0.35
Asia-L. America -1.93 0.52 13.93 <0.01 0.14
Asia-N. America -0.15 0.26 0.34 >0.25 0.86
Africa-Europe -1.09 0.34 10.03 <0.01 0.34
Africa-L. America -2.89 0.51 31.53 <0.01 0.06

Africa-N.
America
0.02 0.18 0.02 >0.25 1.02

Europe-L.
America
-0.63 0.32 3.82 <0.10 0.54

Europe-N.
America
0.42 0.24 3.13 <0.10 1.52

L. America-N.
America
0.15 0.19 0.62 >0.25 1.16
2003-2005 Global -2.23 0.05 1680.18 <0.01 0.11
Asia-Africa -2.07 0.22 87.69 <0.01 0.13
Asia-Europe -1.09 0.15 53.75 <0.01 0.34
Asia-L. America -2.07 0.32 41.22 <0.01 0.13
Asia-N. America -0.77 0.13 36.38 <0.01 0.46
Asia-Oceania -0.95 0.51 3.41 <0.10 0.39
Africa-Europe -1.05 0.13 65.20 <0.01 0.35
Africa-L. America -2.35 0.32 53.31 <0.01 0.09

Africa-N.
America
-0.54 0.11 26.19 <0.01 0.58
Africa -Oceania -0.81 0.42 3.73 <0.10 0.44

Europe-L.
America
-1.95 0.27 50.40 <0.01 0.14

Europe-N.
America
0.03 0.09 0.11 >0.25 1.03
Europe -Oceania -0.44 0.37 1.44 >0.10 0.64

L. America-N.
America
-0.51 0.14 12.58 <0.01 0.6

L. America-
Oceania
-1.14 0.72 2.48 >0.10 0.32

N. America-
Oceania
-0.27 0.33 0.66 >0.25 0.76

127
Hypothesis 3
Hypothesis 3 tested ties across and within populations. In particular, the
hypothesis tested whether the structural tendencies toward choice across and within
blocks of organizational populations were significant predictors of the observed
network structure.
The testing of Hypothesis 3 proceeded in the same manner as Hypothesis 2.
First, in the implementation network, a χ
2
test was conducted to determine whether
models with additional block parameters significantly increased the model fit
compared to the baseline model with the choice parameter. Table 18 shows the
results. In 1991-1993, the addition of choice between populations parameter
decreased the -2LPL value by 68.52, which was larger than the χ
2
value of 13.82
associated with the change in df = 2 at p < .001 due to the process of adding
parameters. Choice across populations parameter was significant as well, reducing
the -2LPL value by 158.33, which was larger than 16.26, the χ
2
value for df = 3 at p
< .001. These two parameters contributed to the model fit in the time period of 2003-
2005 as well.

128
Table 18: Goodness of Fit Indices for the Models of Choice in I- Network
Note. Each p value represents the significance of the model compared to the baseline model
represented in Model 1.

In the implementation network, a comparative analysis was conducted to
find out the parameters that were significant predictors in each period. As listed in
Table 19, the results showed that the 2003-2005 period was significantly explained
by ties within organizational populations (B
Gov.-Gov.
= 0.32, p < .01; B
IGO-IGO
= 1.32, p
< .01; B
NGO-NGO
= 0.47, p < .01; B
For-profit-For-profit
= 0.83, p < .01). This structural
pattern is in contrast to the earlier period of 1991-1993, in which no within-
population parameters were significant predictors of the network. For across-
population ties, the time period 1991-1993 was predicted by ties between GOs and
IGO (B
Gov.-IGO
= 2.53, p < .01) and between IGO and NGO (B
IGO-NGO
= 2.19, p < .01).
On the other hand, in the later time period, parameter estimation for across-
population ties were negative and significant, indicating that there was a lower
likelihood of across-population ties representing the network compared to a random
network (Table 20).
Model
Number of
Parameters (df)
-2LPL
(Badness of
fit)
p
Organizational population (1991-1993)
1 Choice 1 (1) 1279.31
2 Choice + Choice within blocks 2 (3) 1210.79 0.001
3 Choice + Choice across blocks 2 (4) 1120.98 0.001
Organizational population (2003-2005)
1 Choice 1 (1) 50422.89
2 Choice + Choice within blocks 2 (5) 49887.11 0.001
3 Choice + Choice across blocks 2 (7) 49520.59 0.001

129
Table 19: p* Analysis within Blocks of Populations: I- Network
B S. Error PLWald p exp(B)
1991-1993 Global -1.12 0.10 132.99 <0.01 0.33
Governmental -1.20 0.15 60.71 <0.01 0.3
Intergovernmental N/A
a

Nongovernmental 0.21 0.36 0.34 >0.25 1.23
2003-2005 Global -3.85 0.02 43499.96 <0.01 0.02
Governmental 0.32 0.06 29.63 <0.01 1.38
Intergovernmental 1.32 0.15 79.39 <0.01 3.74
Nongovernmental 0.47 0.04 177.04 <0.01 1.6
For-profit 0.83 0.04 467.23 <0.01 2.3
Note. a. Parameters with no/rare counts and excluded from model fit.

Table 20: p* Analysis across Blocks of Populations: I- Network

B S. Error PLWald p exp(B)
1991-1993 Global -2.19 0.11 396.81 <0.01 0.11
Gov.-IGO 2.53 0.22 137.64 <0.01 12.51
Gov.-NGO 0.28 0.18 2.29 >0.10 1.32
IGO-NGO 2.19 0.39 30.95 <0.01 8.93
2003-2005 Global -3.27 0.02 24969.99 <0.01 0.04
Gov.-IGO -0.02 0.08 0.10 >0.25 0.98
Gov.-NGO -0.43 0.04 123.06 <0.01 0.65
Gov.-For-profit -0.73 0.05 230.93 <0.01 0.48
IGO-NGO -0.25 0.06 14.90 <0.01 0.78
IGO-For-profit 0.27 0.06 21.76 <0.01 1.31
NGO-For-profit -1.05 0.04 579.94 <0.01 0.35

The analysis was conducted in the knowledge-sharing network as well. First,
the block parameters significantly increased the model fit in both time periods, as
indicated in Table 21.

130
Table 21: Goodness of Fit Indices for the Models of Choice in K- Network
Note. Each p value represents the significance of the model compared to the baseline model
represented in Model 1.

The comparison of parameters showed that in accordance with findings from
the implementation network, within-population ties were more likely to predict the
networks in 2003-2005 than in 1997-1999. While the earlier network was
represented by within-IGO ties only (B
IGO-IGO
= 1.24, p < .01), the later time period
showed that ties within GOs (B
Gov.-Gov.
= 0.69, p < .01), IGOs (B
IGO-IGO
= 2.25, p
< .01), and for-profit organizations (B
For-profit-For-profit
= 1.52, p < .01) were all positive
and significant (Table 22). The results can be summarized by the increase of the
significance of within-population ties in predicting the network structure. Changes in
across-population ties are shown in Table 22. Ties between GOs and IGOs were
significant predictors of the network structure in both time periods (B
Gov.-IGO
= 0.52, p
< .01 in 1997-1999 and B
Gov.-IGO
= 0.29, p < .01 in 2003-2005). On the other hand, all
other ties across organizational populations were negative in both time periods, as
presented in Table 23.
Model
Number of
Parameters (df)
-2LPL
(Badness of
fit)
p
Organizational population (1997-1999)
1 Choice 1 (1) 8187.85
2 Choice + Choice within blocks 2 (6) 7962.22 0.001
3 Choice + Choice across blocks 2 (16) 7770.07 0.001
Organizational population (2003-2005)
1 Choice 1 (1) 8187.85
2 Choice + Choice within blocks 2 (5) 7706.52 0.001
3 Choice + Choice across blocks 2 (7) 7987.24 0.001

131
Table 22: p* Analysis within Blocks of Populations: K- Network
B S. Error PLWald p exp(B)
1997-1999 Global -2.37 0.08 839.57 <0.01 0.09
Governmental 0.03 0.27 0.01 >0.25 1.03
Intergovernmental 1.24 0.26 22.93 <0.01 3.45
Nongovernmental -0.10 0.14 0.50 >0.25 0.91
2003-2005 Global -3.09 0.04 4915.13 <0.01 0.05
Governmental 0.69 0.16 18.70 <0.01 2
Intergovernmental 2.25 0.10 547.69 <0.01 9.49
Nongovernmental 0.03 0.08 0.15 >0.25 1.03
For-profit 1.52 0.19 64.71 <0.01 4.55

Table 23: p* Analysis across Blocks of Populations: K- Network
B S. Error PLWald p exp(B)
1997-1999 Global -2.31 0.09 611.00 <0.01 0.1
Gov.-IGO 0.52 0.19 7.06 <0.01 1.68
Gov.-NGO -0.50 0.17 8.90 <0.01 0.61
IGO-NGO 0.14 0.16 0.84 >0.25 1.15
2003-2005 Global -2.49 0.05 2986.45 <0.01 0.08
Gov.-IGO 0.29 0.11 7.58 <0.01 1.34
Gov.-NGO -0.44 0.09 24.05 <0.01 0.65
Gov.-For-profit -1.29 0.26 25.24 <0.01 0.28
IGO-NGO -0.81 0.10 68.14 <0.01 0.44
IGO-For-profit -1.37 0.26 28.59 <0.01 0.25
NGO-For-profit -0.99 0.13 55.59 <0.01 0.37

In summary, Hypothesis 3 was partially supported. In general, ties within
organizational populations rather than across populations significantly predicted the
network structure in recent periods. The over time change indicates that across-
population ties represented the network in earlier periods better than in the later
period in the implementation networks, while there were no differences in the
knowledge-sharing networks. In other words, over time, the networks demonstrated a

132
weaker structural tendency toward forming multimodal ties, which are formed
between organizations in different regions and populations. Table 24 summarizes the
above findings regarding Hypotheses 2 and 3.
Table 24: Summary of Significant Parameters in Two Time Points Between and
Across Blocks
Implementation network Knowledge-sharing network
1991-1993 1997-1999
Within-Europe
Within-N.America
Across Asia-Oceania
Across Asia-Europe
Across Africa-L.America
Across Europe-L.America
Across L.America-
N.America

Within-L.America
Within-N.America
Across Europe-N.America

2003-2005 2003-2005
Geographic region
Within-Asia
Within-Europe
Within-L.America
Within-N.America
Within-Oceania

Within-Asia
Within-Africa
Within-Europe
Within-N.America
1991-1993 1997-1999
Across Gov.-IGO
Across IGO-NGO

Within-IGO
Across Gov.-IGO
Across Gov.-NGO

2003-2005 2003-2005
Organizational
population
Within-Gov.
Within-IGO
Within-NGO
Within-For-profit
Across IGO-For-profit
Within-Gov.
Within-IGO
Within-For-profit
Across Gov.-IGO

133

Network Dynamics Analysis
In the analyses of network dynamics, the current study focused on the
propensity of multiplex tie formation and various exogenous conditions that affected
the likelihood of selection and retention of multiplex ties. The results from
hypotheses were obtained from XPNet models at three different time periods: 1997-
1999, 2000-2002, and 2003-2005. At each time period, the networks were
constructed into 292*292, 513*513, and 549*549 nondirectional organization-to-
organization matrices respectively. Implementation networks and knowledge-sharing
networks were fitted together in the XPNet estimation by matching the number of
nodes in two networks. Organization attributes were constructed for the same set of
nodes in each time period. The model building started with a simple model based on
a Bernoulli distribution that assumes independency of ties, with only the Edge
parameters to capture the density of network. The results showed that the EdgeA and
EdgeB parameters were negative and significant across all converged models (Model
1 in Tables 25, 26, and 27). As suggested by Robins et al. (2007), the large negative
Edge parameter implies that there is a low baseline propensity to form ties net of
other effects. In other words, in both implementation and knowledge-sharing
networks, the likelihood of tie formation was significantly lower than in random
networks. The result corresponds to the overall low density found in both types of
networks.

134
Subsequent models considered Markov parameters, which assume non-
independency of ties when they share a node in common. Before the hypothesized
parameters were tested, a baseline analysis was conducted to select structural
parameters that significantly represent the network and eliminate those that are not
present in the network. First, three types of low order parameters, 2-Star, 3-Star, and
Triangle, were not significant in both of the observed networks. Therefore, these
three parameters were excluded from consideration in the current study. Second,
higher order graph statistics were estimated, including the K-Star and K-Triangle
parameters. In the two types of networks respectively, these parameters were not
significant. Therefore, only the EdgeA and EdgeB parameters that were significant
were included in subsequent models that are presented in Tables 25, 26, and 27.
Parameter estimates were obtained from converged models in three time
periods respectively. Tables 25, 26, and 27 summarize the estimation of the baseline
model with structural parameters, as well as advanced models with both structural
and nodal attribute parameters in three time periods. Each table includes parameter
values for selected models with standard errors in parentheses. Parameter estimates
that are more than twice their standard error can be considered significantly different
from zero (Robins et al., 2007). This indicates that the parameter is regarded
statistically significant in terms of representing the observed network. As mentioned
in Chapter 4, the t-ratios reported in the Tables 25, 26, and 27 indicate the
convergence of models. An absolute value of less than or close to 0.1 for these t-
ratios implies a good fit (Robins et al., 2007). The estimation results from the five

135
models with unique combinations of parameters are explained and discussed in detail
in the following sections.

136
Table 25: Parameter Values for Selected Models, 1997-1999
Models
Model 1
(H4)

Model 2
(H5c)

Model 3
(H6a)

Model 4
(H6b)

Model 5
(H7)

Estimates Value t Value t Value t Value t Value t
Structural parameter
EdgeA
-3.55*
(0.03)
-0.07 -3.54*
(0.03)
-0.02 -3.54*
(0.03)
0.01 -3.55*
(0.03)
-0.04 -3.54*
(0.03)
-0.02
EdgeB
-5.82*
(0.09)
-0.08 -5.76*
(0.10)
-0.05 -5.76*
(0.08)
-0.11 -5.84*
(0.09)
0.04 -5.84*
(0.09)
-0.06
EdgeAB
1.83*
(0.24)
-0.05
K-TriangleABA 0.29 (0.18) 0.07
Categorical nodal attribute
Same-category-AB
(Region)
1.34*
(0.50)
0.08
Same-category-AB
(Org. type)
1.78*
(0.40)
0.02
Binary nodal attribute
RAB (Geographic scope-
international)
0.97*
(0.19)
0.00
RAB (Functional scope-
general)
0.50 (0.29) -0.02
Continuous nodal attribute
SumAB (Centrality in I-
Network)
0.05 (0.03) -0.04
SumAB (Centrality in K-
Network)
0.30*
(0.07)
-0.06
Note. t-ratios indicate the convergence statistics of each parameter estimates
Note. Standard errors in parentheses
Note. * Indicates significant parameter at p < .05

137
Table 26: Parameter Values for Selected Models, 2000-2002
Models
Model 1
(H4)

Model 2
(H5c)

Model 3
(H6a)

Model 4
(H6b)

Model 5
(H7)

Estimates Value t Value t Value t Value t Value t
Structural parameter
EdgeA
-3.92*
(0.02)
-0.03 -3.97*
(0.02)
-0.01 -3.91*
(0.02)
0.01 -3.93*
(0.02)
0.08 -3.91*
(0.02)
-0.01
EdgeB
-6.70*
(0.07)
-0.01 -6.96*
(0.10)
-0.01 -6.56*
(0.07)
-0.02 -6.66*
(0.08)
0.08 -6.61*
(0.07)
-0.06
EdgeAB
2.90*
(0.15)
-0.02
K-TriangleABA
1.50*
(0.13)
0.03
Categorical nodal attribute
Same-category-AB
(Region)
1.98*
(0.31)
-0.03
Same-category-AB
(Org. type)
2.34*
(0.28)
-0.00
Binary nodal attribute
RAB (Geographic scope-
international)
0.84*
(0.16)
-0.03
RAB (Functional scope-
general)
1.52*
(0.20)
-0.06
Continuous nodal attribute
SumAB (Centrality in I-
Network)
0.07*
(0.03)
0.07
SumAB (Centrality in K-
Network)
0.43*
(0.10)
0.05
Note. t-ratios indicate the convergence statistics of each parameter estimates
Note. Standard errors in parentheses
Note. * Indicates significant parameter at p < .05

138
Table 27: Parameter Values for Selected Models, 2003-2005
Models
Model 1
(H4)

Model 2
(H5c)

Model 3
(H6a)

Model 4
(H6b)

Model 5
(H7)

Estimates Value t Value t Value t Value t Value t
Structural parameter
EdgeA
-4.03*
(0.02)
0.05 -4.11*
(0.02)
0.04 -4.01*
(0.02)
0.10 -4.02*
(0.02)
0.00 -4.01*
(0.02)
0.07
EdgeB
-5.80*
(0.05)
-0.00 -6.00*
(0.06)
-0.02 -5.73*
(0.05)
0.14 -5.80*
(0.05)
0.00 -5.76*
(0.05)
-0.03
EdgeAB
2.36*
(0.12)
-0.06
K-TriangleABA
1.20*
(0.09)
-0.02
Categorical nodal attribute
Same-category-AB
(Region)
2.03*
(0.24)
0.04
Same-category-AB
(Org. type)
1.71*
(0.23)
0.03
Binary nodal attribute
RAB (Geographic scope-
international)
0.60*
(0.13)
0.00
RAB (Functional scope-
general)
1.42*
(0.15)
0.00
Continuous nodal attribute
SumAB (Centrality in I-
Network)
0.01 (0.03) 0.03
SumAB (Centrality in K-
Network)
0.42*
(0.06)
0.05
Note. t-ratios indicate the convergence statistics of each parameter estimates
Note. Standard errors in parentheses
Note. * Indicates significant parameter at p < .05

139
Hypothesis 4
Hypothesis 4 tested the likelihood of tie formation in multiplex networks,
that is, across multiple networks. As represented in Model 1, the models converged
and the EdgeAB parameter showed a significant and positive value in all three time
periods. The parameter estimates were EdgeAB
1997-1999
= 1.83, t = -0.05, EdgeAB
2000-
2002
= 2.90, t = -0.02 and EdgeAB
2003-2005
= 2.36, t = -0.06, which were statistically
significant. The result indicated statistically significant evidence that ties in the
implementation and knowledge-sharing networks are likely to co-occur. In other
words, the multiplex mechanism between the two networks suggested in Hypothesis
4 was supported.
Hypotheses 5a, 5b, and 5c
Hypotheses 5a, 5b, and 5c predicted a structural equivalence mechanism
based on a common third-party. This mechanism suggests that organizational dyads
that are structurally equivalent by having a common third-party will have a higher
likelihood of forming both uniplex and multiplex ties. To test these hypotheses, the
Triangle parameter was estimated in both of the single networks and in the multiplex
network. First, the results showed that the Triangle parameters for single networks
were not significant, indicating that Hypothesis 5a was not supported. In other words,
the structural equivalence mechanism did not hold true within each uniplex network
respectively. Hypothesis 5b modeled the same theoretical mechanism in multiplex
networks. Contrary to prediction, the TriangleAAB and TriangleABB parameters
were negative and nonsignificant. Therefore, Hypothesis 5b was not supported.

140
To test Hypothesis 5c which assumed complexity in the multiplex networks,
the higher order parameter, K-TriangleABA, was estimated. Model 2 in Tables 25, 26,
and 27 summarizes the results of this estimation. The results showed that the models
which incorporated the parameter converged, and that the K-TriangleABA parameters
were positive and significant in the later two time periods, 2000-2002 and 2003-2005
(K-TriangleABA
2000-2002
= 1.49 at t = 0.03, and K-TriangleABA
2003-2005
= 1.20 at t = -
0.02). In contrast, the 1997-1999 network yielded a converged model but the
parameter was not significant (K-TriangleABA
1997-1999
= 0.29, t = 0.09). Therefore,
Hypothesis 5c was partially supported. The results indicate that this mechanism is
observed more strongly in later time periods. The simultaneous support for
Hypothesis 5c and the rejection of Hypothesis 5b, in other words, the co-occurrence
of nonsignificant Triangle and significant K-Triangle parameters is worth paying
attention to. This finding implies that the triangle effects are not necessary to explain
the data once the higher order transitivity is accounted for (Robins et al., 2007).
Hypotheses 6a and 6b
Hypothesis 6a predicted that the likelihood of multiplex tie formation would
be higher when organizations shared the same resources. The hypothesis suggested
two organizational attributes of geographic locations and organizational populations.
As shown in Model 3 (See Tables 25, 26, and 27), the hypothesis was supported. For
geographic region, in 1997-1999, Same-categoryAB was estimated and the parameter
estimate was positive and significant (Same-categoryAB
1997-1999
= 1.34, t = 0.08). In
2000-2002, the parameter was significant as well (Same-categoryAB
2000-2002
= 1.98, t

141
= -0.03). In 2003-2005, the parameter was significant as well (Same-categoryAB
2003-
2005
= 2.03, t = 0.04). All of the three models showed either full convergence (2000-
2002) or near convergence (1997-1999 and 2003-2005). In terms of organizational
populations as the categorical attribute, the parameter was positive and significant in
all three time periods (Same-categoryAB
1997-1999
= 1.77, t = 0.02; Same-
categoryAB
2000-2002
= 2.35, t = -0.00; Same-categoryAB
2003-2005
= 1.71, t = 0.03).
These findings suggest that resource similarity based on organizational function
affects tie formation. In summary, the result provided support for Hypothesis 6a by
showing that organizational dyads in the same resource spaces, in terms of both
location and population, are likely to form multiplex ties.
Hypothesis 6b suggested that generalist organizations are more likely to
form multiplex ties than specialist organizations. Model 4 (Tables 25, 26, and 27)
shows the estimation results. For geographic scope, the results showed that
organizations with international scope were likely to form multiplex ties in all three
time periods evidenced by positive and significant parameters (RAB
1997-1999
= 0.97, t
= 0.00; RAB
2000-2002
= 0.85, t = - 0.02; RAB
2003-2005
= 0.61, t = 0.00). For functional
scope, the RAB parameter was significant in the two later time periods (RAB
2000-2002

= 1.52, t = - 0.06; RAB
2003-2005
= 1.43, t = 0.00). In other words, organizations with
functional activities not specialized in ICTs were likely to form multiplex ties. Yet,
such mechanism was not supported in 1997-1999. Therefore, the results showed a
partial support for Hypothesis 6b.

142
Hypothesis 7
Hypothesis 7 tested the relationship between structural position measured by
centrality and multiplex tie formation. Specifically, it was predicted that central
organizations will form a larger number of ties that spans across both implementation
and knowledge-sharing networks. Parameter values obtained from the estimation are
shown in Model 5 in Tables 25, 26, and 27. First, the effect of centrality in the
implementation-network was examined. The parameter estimate varied slightly
across time periods. In the 1997-1999, the parameter was positive and nearly
significant (SumAB
1997-1999
= 0.05, t = -0.04). In the 2000-2002 time period, the
parameter was positive and significant (SumAB
2000-2002
= 0.07, t = 0.07). In the 2003-
2005, the parameter was positive but not significant (SumAB
2003-2005
= 0.01, t = 0.05).
In summary, the results show a partial support for the claim that there is a higher
likelihood of multiplex tie formation for organizational dyads that are central in the
implementation network.
The same mechanism was tested for the centrality of organizations in the
knowledge-sharing network as well. The result showed stronger and coherent
support for the hypothesis. The parameter values were SumAB
1997-1999
= 0.30 at t = -
0.06, SumAB
2000-2002
= 0.43 at t = 0.05, and SumAB
2003-2005
= 0.43 at t = 0.03, all
significant and positive. In other words, the results suggested that the combined
centrality of organizational dyads in the knowledge-sharing network is a significant
predictor of multiplex tie formation. The findings generally supported Hypothesis 7.

143
Hypothesis 8
Hypothesis 8 examined longitudinal changes in the magnitude of multiplex
effects, predicting that there will be increases in multiplex tie formation over time.
To test Hypothesis 8, the EdgeAB parameter was investigated. As found in the results
of Hypothesis 4, the parameters were significant and positive across all three time
periods. To test whether the likelihood of multiplex network dynamics have
increased over time, the percentage of the EdgeAB configuration out of total possible
multiplex edges across two networks was calculated. The maximum possible number
of edges were calculated as n(n-1)/2, given n being the number of common
organizations in the implementation and knowledge-sharing networks. The count of
the parameter configuration was obtained in the goodness of fit statistics. In 1997-
1999, the count of the EdgeAB configuration was 22. As the number of common
nodes was 19, the total possible EdgeAB formation was (19*18)/2= 171.
Consequently, the percentage of the EdgeAB formation in the observed network was
(22/171)*100 = 12.87%. In 2000-2002, the observed count of the EdgeAB
configuration was 57, and the maximum number of possible EdgeAB formation was
(40*39)/2= 780, with the number of common nodes being 40. Therefore, the
percentage was (57/780)*100 = 7.31%. In 2003-2005, out of total possible number of
EdgeAB counts, (48*47)/2=1128, the observed number of the EdgeAB configuration
was 84, therefore the percentage being (84/1128)*100 = 7.45%. The comparison of
the EdgeAB ratio in the three time periods did not show increases in the probability
of multiplex tie formation over time. Therefore, Hypothesis 8 was not supported.

144
Visualization
Figures 4, 5, and 6 present the discrete-time visualizations of the
implementation and knowledge-sharing networks in each of the three time periods.
The graphs were laid out using a spring-embedding algorithm through Netdraw. The
color of the nodes reflects their organizational population, and the shape of the nodes
indicates their geographic region. Node size is scaled to reflect the network degree of
each organization.
In each time period, both implementation and knowledge-sharing networks
are simultaneously presented. First, it is noticeable that there is less centralization in
the network over time. Moreover, the graphs illustrate the trends of increasing
clustering in the network. For example, the grey nodes in diamond and down triangle
forms in the lower left corner of the graph represent clusters of for-profit
organizations in European and North American regions. The clusters in the lower
right corner composed of nodes in red square indicate NGOs in African region.
These clusters become more visible in the later time periods, which support the
finding that there are more unimodal ties formed within-regions and populations over
time.
The color of the ties represent network types: orange indicates
implementation and blue indicates knowledge-sharing ties. Multiplex ties formed
across both networks are represented by grey color. A comparison of the 2000-2002
(Figure 5) against the 1997-1999 (Figure 4) period shows that there is an increasing
visibility of multiplex tie formation over time. In addition, the comparison between

145
2000-2002 (Figure 5) and 2003-2005 (Figure 6) illustrates that knowledge-sharing
ties become more important over time. Further, several inferences can be made about
the emergence of particular network structures in Figures 5 and 6. These structures
are attributable to the multiplex dynamics observed in the above analyses. For
example, a closer examination of the triad-level structure reveals that there are a
number of embedded triangles across the two network types, which lead to dense
local clusters (Hypothesis 5c). The graphs also show that central organizations,
represented by larger node sizes, have a larger number of multiplex ties (Hypothesis
7).

146
Figure 4: Visualization of I- Network and K- Network, 1997-1999

Note. Node color represents organizational population (Blue: Governmental, Black: IGO, Red: NGO, Grey: For-profit)
Note. Node shape represents geographic location (Circle: Asia, Square: Africa, Diamond: Europe, Box: Latin America and the Caribbean,
Down triangle: Northern America, Plus: Oceania)
Note. Node size represents degree centrality
Note. Link color represents network type (Orange: Implementation, Blue: Knowledge-sharing, Grey: Both)

147
Figure 5: Visualization of I- Network and K- Network, 2000-2002

Note. Node color represents organizational population (Blue: Governmental, Black: IGO, Red: NGO, Grey: For-profit)
Note. Node shape represents geographic location (Circle: Asia, Square: Africa, Diamond: Europe, Box: Latin America and the Caribbean,
Down triangle: Northern America, Plus: Oceania)
Note. Node size represents degree centrality
Note. Link color represents network type (Orange: Implementation, Blue: Knowledge-sharing, Grey: Both)

148
Figure 6: Visualization of I- Network and K- Network, 2003-2005

Note. Node color represents organizational population (Blue: Governmental, Black: IGO, Red: NGO, Grey: For-profit)
Note. Node shape represents geographic location (Circle: Asia, Square: Africa, Diamond: Europe, Box: Latin America and the Caribbean,
Down triangle: Northern America, Plus: Oceania)
Note. Node size represents degree centrality
Note. Link color represents network type (Orange: Implementation, Blue: Knowledge-sharing, Grey: Both)

149
Network Effects Analysis
Hypotheses 9a, 9b and 9c
Hypotheses 9a, 9b, and 9c tested the effects of niche overlap and structural
equivalence between projects on the sharing of technologies and applications. Each
of these hypotheses was tested by MRQAP analyses. Table 28 reports descriptive
statistics and correlations from QAP analysis of matrices.
Table 28: Means, Standard Deviations, and Correlations of Matrices in MRQAP
Parameters Niche
overlap
(region)
Niche
overlap
(sector)
Tem-
poral
proxi-
mity
a

Struc-
tural
Equi-
valence
Common
-alities
Den-
sity
SD
Niche overlap
(region)
-- 0.430 0.516
Niche overlap
(sector)
-0.001
(0.481)
-- 0.354 0.502
Temporal
proximity
a

-0.040
(0.092)
0.023
(0.177)
-- 3.680 3.179
Structural
Equivalence
0.005
(0.400)
-0.019
(0.209)
-0.134*
(0.000)
-- 0.135 0.192
Commonalities
in technologies
and
applications
0.011
(0.260)
0.084*
(0.000)
0.036
(0.057)
0.043*
(0.034)
-- 0.278 0.467
Note. a. Temporal proximity was measured as differences in years
Note. p-values in parentheses
Note. * Significant at p < .05

First, Hypothesis 9a predicted that the greater the niche overlap between
projects, the greater the likelihood that they will share technologies and applications.
As shown in Table 29, the MRQAP procedure suggested that niche overlap of project
sector is significantly and positively related to commonalities in technologies and
applications (β = 0.078, p = 0.000). In other words, if pairs of projects had
commonalities with regard to project goals, it was likely that they adopted the same

150
types of technologies and applications. On the other hand, the results showed that
niche overlap of project region was not related to the sharing of technologies and
applications (β = 0.011, p = 0.225). Therefore, being implemented in the same region
did not influence the types of technologies and applications adopted in the project. In
summary, the results provided partial support for Hypothesis 9a.
Table 29: MRQAP Results for a Regression Model Predicting Idea Sharing between
Projects
Hypothesis Dependent
Variable
R
2
Independent
Variables
Standardized
Coefficient
(Un-
standardized)
Sig.
H9a Commonalities
in technologies
and
applications
0.011 Niche overlap
(region)
0.011
(0.012)
0.225
H9a Niche overlap
(sector)
0.078*
(0.084)
0.000
H9c Structural
Equivalence
0.120*
(0.050)
0.019
Temporal
proximity
a

0.006*
(0.041)
0.036
Note. a. Temporal proximity was measured as differences in years
Note. * Significant at p < .05

Hypothesis 9b suggested that the greater the niche overlap between projects,
the greater the structural equivalence among projects. In other words, pairs of
projects with greater niche overlap were predicted to have similar relations with
organizations. A MRQAP correlation procedure was conducted between three
matrices, among which two matrices represented niche overlap in the two
dimensions of geographic regions and project goals, and one matrix represented

151
structural equivalence among projects. Regression coefficients in Table 30 indicate
the observed value between pairs of networks. Neither niche overlap in terms region
(β = 0.000, p = 0.242) nor sector (β = -0.006, p = 0.503) among projects were
significantly related with structural equivalence. Therefore, the results did not
support Hypothesis 9b.
Table 30: MRQAP Results for a Regression Model Predicting Structural Equivalence
Hypothesis Dependent
Variable
R
2
Independent
Variables
Standardized
Coefficient
(Un-
standardized)
Sig.
H9b Structural
Equivalence
0.018 Niche overlap
(region)
0.000
(0.000)
0.242
H9b Niche overlap
(sector)
-0.006
(-0.016)
0.503
Temporal
proximity
a

-0.008*
(-0.134)
0.001
Note. a. Temporal proximity was measured as differences in years
Note. * Significant at p < .05

Hypothesis 9c predicted that structural equivalence of projects will affect
their commonalities in technologies and applications. The analysis supported the
hypothesis, as shown in the positive and significant regression coefficient (β = 0.120,
p = 0.019) in Table 29 presented above. In other words, the result suggested that
pairs of projects that had similar relations with collaborating organizations were
more likely to share commonalities in the technologies and applications.
The results also suggested a significant effect of temporal proximity on
structural equivalence (β = -0.008, p = 0.001), as presented in Table 30. As temporal
proximity was measured by the differences in years projects were initiated, the signs

152
of coefficients were interpreted in the opposite direction. In other words, a negative
and significant coefficient indicted that when two projects were implemented in a
similar time period, they had similar relationship patterns with collaborating
organizations. The results indicated that temporal proximity had a positive
coefficient on commonalities of technologies and applications (β = 0.006, p = 0.036).
Contrary to evolutionary predictions, the result suggested that pairs of projects were
more likely to adopt similar types of technologies and applications if they had a
larger time gap. The combined results of the MRQAP analyses are shown in Tables
29 and 30. Figure 7 shows a summary of relations between matrices examined in
Hypotheses 9a, 9b, and 9c.
Figure 7: MRQAP Results for the Relations between Niche Overlap, Structural
Equivalence, and Sharing of Technologies and Applications across Projects

Note. Solid lines indicate significant and supported hypotheses; Dotted lines indicate
insignificant or unsupported hypotheses
Note. Indicated coefficients are standardized
Note. * Significant at p < .05

Structural
Equivalence
Niche overlap
(sector)
Niche overlap
(region)
Sharing of
technologies and
applications
Temporal
proximity
0.120*
0.078*
-0.008*
0.011
-0.006
0.000
0.006

153
Figure 8 shows a visualization of the implementation network. The 2-mode
network view shows both projects and organizations that existed between year 1987
and 2007.

154
Figure 8: Visualization of 2-Mode Network of Projects and Organizations (I- Network, All years combined)

Note. Blue squares represent projects; Red Circles represent organizations
Note. Node size represents degree centrality

155
Table 31 provides a summary of all hypothesis-testing results in the
evolution of organizational networks over time. With regard to network structure, the
study found overall decentralization at the global level and increase of within-region
ties at the subgroup level. As for network dynamics, the study found supports for
multiplex mechanisms. Further, conditions that lead to the selection and retention of
multiplex ties were examined. The findings included support for the positive effects
of resource similarity on multiplex tie formation. In addition, the study provided
partial support for the positive effects of multiple common third-parties, generalist
orientation of organizations, and centrality of organizations on the likelihood of
forming multiplex ties. In terms of network effects, the study found a positive effect
of niche overlap on both structural equivalence and the sharing of knowledge and
applications in the ICT for Development projects.

156
Table 31: A Summary of Hypothesis-Testing Results
Hypothesis Results
Network Structure
Hypothesis 1. Interorganizational networks in the
ICT for Development community will show less
centralization over time.
Supported
Hypothesis 2. Interorganizational networks in the
ICT for Development community will show larger
increase of within-region ties than across-region
ties over time.
Supported
Hypothesis 3. Interorganizational networks in the
ICT for Development community will show larger
increase of within-population ties than across-
population ties over time.
Partially
Supported
Network Dynamics
Tie
Selection
Hypothesis 4. The existence of a tie in one
network type will increase the likelihood of tie
formation among the same organizations in
another network type.
Hypothesis 5a. The existence of a common third-
party will increase the likelihood of tie formation
in the same network.
Hypothesis 5b. The existence of a common third-
party will increase the likelihood of tie formation
across different types of networks.
Hypothesis 5c. Organizational dyads are likely to
share multiple common third-parties across
different types of networks.
Hypothesis 6a. Organizational dyads in the same
resource space are more likely to form multiplex
linkages than a dyad in which the two
organizations are in different resource spaces.
Hypothesis 6b. Generalist organizations are more
likely to form multiplex linkages than specialist
organizations.
Hypothesis 7. The greater the combined centrality
of two organizations, the greater the likelihood of
multiplex tie formation between those
organizations.
Supported

Not
Supported

Not
Supported

Partially
Supported

Supported

Partially
Supported

Partially
Supported

157
Table 31: A Summary of Hypothesis-Testing Results (Continued)
Tie
Retention
Hypothesis 8. There will be an increase in the
number of repeated dyad formations for the same
organizations across multiplex networks over
time.
Not
Supported

Network Effects
Hypothesis 9a. The greater the niche overlap
between projects, the greater the possibility that
they will share technologies and applications
between them.
Hypothesis 9b. The greater the niche overlap
between projects, the greater their structural
equivalence.
Hypothesis 9c. The greater the structural
equivalence between projects, the greater the
likelihood of technologies and applications shared
between them.
Partially
Supported

Not
Supported

Supported

158
CHAPTER 6: CONCLUSION
Discussion
This study examined a series of research questions related to the three major
aspects of interorganizational networks in the ICT for Development community.
Results of the analyses articulated above provide insights about interorganizational
network structure, dynamics, and effects in the community.
Network Structure
Network structure analysis conducted to test Hypothesis 1 found structural
changes at the global level. To summarize, the study found evidence for an overall
shift towards decentralization and heterogenization in both implementation and
knowledge-sharing networks. These changes are in accordance with the patterns of
structural shifts that have been found in various global networks in the recent decade
(Barnett, 2001; Danowski, 2000; Lee et al., 2007; Matei, 2006; Monge & Matei,
2004). The increasing trends of decentralization imply that the patterns of
organizational participation in the ICT for Development field have become more
equal and heterogeneous, with less concentration on a small number of organizations.
To examine these structural changes in further detail, in Hypotheses 2 and 3,
the study moved to the subgroup level of analysis and examined the tie structure
among the subsets of networks. Overall, attention to the subgroup level was found to
be useful for explaining the network structure, as shown in the increased model fit
with the addition of parameters both across and within blocks (see Tables 12, 15, 18,
and 21). Analyses of block parameters over time revealed the following finding:

159
while the boundary of the ICT for Development community encompasses the entire
global system, it was generally shown that there are increasingly more clusters of
organizations within the same regions and organizational populations (see Table 24
for a summary of results). In other words, the results suggested that unimodal ties
were better predictors of the network than multimodal ties over time. In terms of
evolutionary dynamics, the results imply that there are increasing tendencies for
organizations to form commensalistic relationships with other organizations in
overlapping resource spaces in the ICT for Development community. As suggested
by Aldrich and Reuf (2006), commensalistic relationships can range from full
mutualism to full competition depending on whether the existence of one population
has positive or negative effects on another population. While the current study found
increases of commensalistic relationships between organizational populations, it did
not provide predictions about which types of relationship are more dominant in the
ICT for Development community. Future research is encouraged to incorporate
measures of the growth of organizational populations so that the effects of inter-
population dynamics can be identified.
Specifically, the results showed the evolution of unimodal and multimodal
ties within and across regions. In both implementation and knowledge-sharing
networks, the study found increases in within-region ties. The increase of
commensalistic ties within organizational populations that are located in similar
geographic resource space implies the potential for self-sustaining regional networks
among organizations in developing countries. In particular, the results found

160
increases of within-region ties in less developed areas. This finding implies a
potential for resource transfer between organizations in less developed regions,
empowered by the growth of resources they possess.
Attention to the interorganizational network structure in the knowledge-
sharing network suggested several implications. The findings demonstrated that there
has been a lack of connection between Southern and Northern organizations
throughout the years observed in the study (Hypothesis 2). A long-standing criticism
has been that Southern organizations, while locally influential, have been largely
excluded from the global governance processes such as conferences, forums, and
agreements. Consistent with Mudhai’s (2004) assertion, the results from this study
imply that it is imperative for Northern and Southern organizations to become part of
the partnership practices. This would ensure that the local contexts of developing
regions are reflected in the global discussion of ICT for Development agenda.
Second, the results found changing patterns of multimodal ties within and
across organizational populations. The results showed that ties across blocks of
organizational populations were significant predictors of the networks more in earlier
than recent years. Over time, several changes have been noticed in terms of across-
population ties. In the implementation networks, increases in the ties between IGOs
and for-profit organizations were found, implying the increase of the presence of
private corporations. In the knowledge sharing networks, ties across GOs and IGOs
were found to be a significant predictor of the overall network structure throughout
the years.

161
In particular, the results showed that there is a lack of NGO presence in the
knowledge-sharing network in the recent time periods, both in symbiotic and
commensalistic forms (Hypothesis 3). This finding is in line with the current
discussions about whether the civil society sector has managed to increase its voice
in the global dialogue. As Smith (1998) noted, scholars have criticized that
“international institutions generally exclude nonstate actors from any formal role in
decision-making processes” (p. 102). Similarly, with regard to the role of INGOs in
global advocacy, Madon (2000) pointed out that most INGOs are considered as
“mere implementers of projects” as opposed to participants in the policy dialogue (p.
5). Consistent with these views, the current study showed that NGOs have been
relatively excluded from the process of knowledge-sharing in the ICT for
Development community.
The critical role of NGOs’ participation in the global public sphere has been
emphasized with the acceleration of globalization, based on the expectation that they
are capable of counterbalancing GOs and IGOs. NGOs have been increasingly
known for extensive networking with external agencies (Fulk, 2001). As the capacity
of NGOs grows, they are facing several tasks. These include: improved
responsiveness to local situations, partnership with external agencies, and impact on
policy and practice of governments and multilateral institutions (Hovland, 2003).
Results from the current study provided support for the call for a more active
participation of NGOs in the global forums of the ICT for Development community.

162
Network Dynamics
In this section, the current study examined network dynamics at the dyadic
and triadic levels of ties. The research attempted to fill in the gap in the past
literature on interorganizational tie formation, which has exclusively focused on
identifying the determinants of uniplex ties but neglected those of multiplex ties.
Overall, the results suggested that the interlocking relationship between the
implementation and knowledge-sharing networks is a significant pattern in the ICT
for Development community. The findings for Hypothesis 4 showed a strong positive
effect of the EdgeAB parameter. This finding implies that there is a significant
chance for organizations to form multiplex collaborative ties in both implementation
and knowledge-sharing networks, compared to what can be found in a random
network.
The following hypotheses examined endogenous and exogenous
mechanisms that drive multiplex tie formation. First, the study examined endogenous
mechanisms behind multiplex tie formation. The results did not support the
theoretical arguments of a common third-party tie in uniplex and multiplex settings
(Hypotheses 5a and 5b). Yet, empirical support was evidenced for more complex
patterns of embeddedness in multiplex dynamics, as indicated by the positive and
significant K-TriangleABA parameters (Hypothesis 5c). The mechanism was found in
the later two time periods of 2000-2002 and 2003-2005, implying the increase in the
magnitude of embedded structure over time. This suggests that organizations which
collaborate in project implementation for multiple times are likely to form ties in

163
knowledge-sharing projects as well. On the other hand, the results did not suggest
that multiple common third parties in the knowledge-sharing project lead to ties in
the implementation network. This contrasting finding can be explained by the
principles of the strength of weak ties suggested by Granovetter (1973). As
mentioned in Chapter 3, the implementation networks resemble transactional and
exchange-oriented types of relations, which can be seen as weak ties. On the other
hand, the knowledge-sharing networks represent social bonding and communication-
oriented types of relations, with can be considered as strong ties (Baldassarri & Diani,
2007; Powell, 1998). Consequently, the effects of common-third parties are likely to
be present more in the implementation network than in the knowledge-sharing
network, as was shown in the results.
The significance of the K-TriangleABA parameter has two implications. First,
this graph configuration implies that the network is best represented by a complex
pattern in which multiple triangles are embedded (Robins et al., 2007). In other
words, the mechanism of a common third-party tie is present in the ICT for
Development community in conjunction with the logic of embeddedness. Second,
while it is expected that there are triangles in the network, it is likely that they are not
homogeneously scattered through the network (Robins et al., 2007). Rather, the K-
TriangleABA configuration indicates that the network has dense regions built upon
cliques, in which triangles are grouped together by sharing one edge (Wasserman et
al., 2007). This finding suggests implications for multilevel dynamics, in which
dyadic and triadic tie formation leads to the emergence of global level network

164
structure. In this case, the embedded triangles were drivers of the increased
formation of dense clusters in the overall network.
The study also found that exogenous environments have significant
influences on multiplex tie formation. The results from testing Hypothesis 6a
explained the effect of resource space similarity on multiplex tie formation.
Organizations that are physically co-located with each other were more likely to
form multiplex ties, confirming the effect of spatial proximity. In addition, the study
found that organizations that share the same organizational type had higher
probability of forming multiplex ties. These findings support the emphasis on
organizational dynamics at the smaller geographic and functional level of analysis
(Aldrich & Reuf, 2006; Haveman & Romanelli, 1997). The impact of geography in
ecological processes has received increasing attention in the recent decade (e.g.
Audia et al., 2006; McEvily & Zaheer, 1999). Overall, these results supported the
theory of homophily (e.g. McPherson et al., 2001) in the context of resource
similarity. In other words, it has been emphasized that organizations engage in
ecological relationships with those that pursue the same resources and in their local
environment. The current study contributed to this theory by suggesting that the
homophily mechanism extends to the multiplex context. In contrast to the effect of
resource complementarity on uniplex tie formation (Astley, 1985; Baum & Singh,
1994d; Fombrun, 1986; Gulati, 1995), the study suggested that commensalistic
relations are stronger drivers of multiplex tie formation than symbiotic relations are.

165
In Hypothesis 6b, the study found partial support for the claim that generalist
organizations have higher likelihood of forming multiplex ties than specialist
organizations. Specifically, the study found that organizations operating at the
international scope formed a larger number of multiplex ties than those at the
regional or national scope. In terms of organizational function, the study found that
organizations specialized in the ICT field have smaller chance of forming multiplex
ties, more significantly in the later time periods. The results show that overall,
generalist organizations have a larger tendency to engage in both implementation and
knowledge-sharing networks.
The study found that network positions in a single network affect the
likelihood of tie formation. The results showed mixed support for Hypothesis 7. In
the knowledge-sharing networks, the effect of centrality on multiplex tie formation
corresponded to previous explanations about the advantages of central, prominent
positions on tie formation (Barabasi, 2002; Gulati, 1999; Powell et al., 1996). In
particular, the findings extended the network literature in the sense that these
dynamics are applicable to the multiplex context. On the other hand, this mechanism
was not fully supported in the implementation networks. These results imply that the
effect of network centrality on facilitating multiplex tie formation was more evident
in the knowledge-sharing networks, which may suggest that prominent organizations
are more visible in the knowledge-sharing projects.
The results from testing Hypothesis 8 failed to indicate that multiplexity is
observed more strongly in later time periods. In other words, the prediction that there

166
will be a mechanism of accumulated knowledge and learning about partners as the
community evolves was not empirically observed in the current study.
In summary, the analyses of network dynamics have found the following
dynamics: the propensity towards multiplex tie formation (Hypothesis 4), the effects
of common third-party ties and structural embeddedness across multiple networks in
the later two periods (Hypothesis 5c), the effects of resource similarity in terms of
geographic location and organizational type on multiplex tie formation (Hypothesis
6a), the effects of generalist orientation with regard to geographic scope on multiplex
tie formation (Hypothesis 6b), and the positive effects of centrality in the knowledge-
sharing network on multiplex tie formation (Hypothesis 7). These results have
implications for organizations’ networking strategies. Particularly, the dynamics of
multiplex networks explored in the study suggest that organizations can take
advantage of their positions in one network towards building their positions in other
networks. Second, efforts for selecting potential partners can be facilitated across
multiple networks, as the presence of ties in one network is expected to increase
potential tie formation in another network. Third, the implications are applicable to
multilevel dynamics of network structure. For the purpose of mitigating the uneven
structure of global networks, it is expected that intervention in one network will
potentially influence another network. For example, it is likely that the effort to
increase the participation of Southern organizations in the implementation networks
will enhance their structural position in the knowledge-sharing networks as well.

167
Network Effects
The last part of the study examined the intersection between projects and
organizations. The results from Hypothesis 9a suggested that projects that are in the
same resource space defined by project sectors were likely to adopt the same types of
technologies and applications. The finding highlighted a mutualistic aspect of niche
overlap, in which projects in similar resource environment have a greater likelihood
of sharing ideas. Yet, this mechanism was not supported in terms of niche overlap
defined by geographic region. In addition, the results did not support Hypothesis 9b;
it failed to show that niche overlap was a significant predictor of projects’ forming
similar relational patterns with collaborating organizations.
The study examined whether organizational networks have effects on the
types of technologies and applications adopted in the projects. The results provided
empirical support for this argument, which suggested that structural equivalence
between projects had a significant effect on commonalities in technologies and
applications adopted in projects (Hypothesis 9c). In other words, it is found that
having similar structural relations with collaborating organizations is associated with
the sharing of ideas. In summary, this study supports the proposition that
interorganizational networks play a role as conduits of knowledge and information
(Audia et al., 2006; Podolny et al., 1996) in the ICT for Development field. By
forming organizational collaborations, the projects are more likely to have higher
potential of creating synergies with existing projects, through which a successful
choice of technologies and applications can be replicated or scaled up.

168
Lastly, the study examined the period effect that is often considered as a core
determinant of evolutionary processes (Aldrich & Reuf, 2006). The study found a
positive effect of temporal proximity in terms of project initiation on the likelihood
of projects having similar relational patterns with collaborating organizations. This
finding implies that there may be an evolutionary pattern of partnership structures in
the ICT for Development community over time. For example, while certain
organizations may be favored as partners in one year, different organizations may be
favored in the next year. On the other hand, the effect of temporal proximity on idea
sharing was not supported, suggesting that the choice of technologies and
applications in projects is independent of the time periods of project implementation.
Implications for Evolutionary Theory and Social Network Theory
The study explored the dynamics of interorganizational network evolution
from a community ecology perspective. Findings from the study suggest that
evolutionary and community ecology theories provide a useful framework for
explaining network changes over time. Evolutionary theories provide a framework
for examining networks at multiple levels including the entire community,
populations, and individual organizations. Each of these levels corresponds to the
networks at the global, subgroup, and dyadic/triadic levels. In particular, the study
supported the argument that tie formation is contingent on both exogenous and
endogenous resource dynamics. The study emphasized the attention to resources that
has been proposed by both evolutionary theories (Aldrich, 1979; Aldrich & Reuf,

169
2006; Monge et al., in press) and social network theories (Borgatti & Foster, 2003;
Burt, 1992).
For explaining resource dynamics, the current study showed the advantages
of using the framework of evolutionary theories over other previously used theories,
such as resource dependence (e.g. Pfeffer & Salancik, 1978) and transaction cost (e.g.
Williamson, 1981). First, evolutionary theory expands the horizon beyond the focal
organization and immediate partners and accounts for local neighborhoods,
populations, community, and environment. In other words, evolutionary theory
emphasizes a multiple level approach. Second, evolutionary theory is better suited to
account for the fact that organizations can be tied via multiple types of relations.
Further, the study added to the current development of knowledge on
multidimensional networks, which include multiplex, multilevel, and multimodal
networks. First of all, relatively little is known about multiplex networks, their
structural differences, and dependencies across them (Pattison, 2006; Robins &
Pattison, 2006). Moreover, determinants of multiplex tie formation are a largely
unexplored area of network research. This study examined both endogenous (tie
structure) and exogenous (environment and organizational attributes) factors behind
multiplex tie formation. In particular, the study found that various determinants that
have been found to influence alliance formation, such as structural equivalence (Burt,
1976), network centrality (Gimeno, 1994), and resource width (Carroll & Hannan,
2000), were also applicable to multiplex contexts.

170
The dissertation suggested the usefulness of multilevel approach in network
studies. First, as was discussed in Hypotheses 1, 2, and 3, parameters at the subgroup
level significantly contributed to the explanation of a given network structure. These
subgroup level parameters included both within- and across-block ties, indicating
that population-level examination is useful in network studies. Second, the study
added to the current stream of research on cross-level influences (Robins et al., 2005)
by showing that local network processes lead to global network structure. In
particular, the significance of embedded transitivity mechanisms found in the later
years explained the global level patterns of decentralization and the emergence of
clusters over time.
The current study contributed to the literature on multimodal networks in
two major ways. First, the study considered a traditional approach to 2-mode
affiliation networks by investigating ties between ICT for Development projects and
collaborating organizations. Second, the study examined multimodal networks from
a novel perspective combined with ecological concepts. In particular, the study
considered two dimensions (organizational populations and geographic locations)
taken from the conceptualization of community ecology to examine the evolution of
multimodal ties.
Past empirical evidence has demonstrated the importance of niche overlap in
organizational dynamics (Baum & Singh, 1994c, 1994d; Carroll, 1985; Freeman &
Hannan, 1983). The study added to the literature on niche overlap in two ways: first,
by emphasizing the collaborative aspects of niche overlap, and second, by moving

171
the level of analysis from organizations to projects. Most of the past studies on niche
overlap emphasized competitive relations among organizations that seek the same
limited resources (Gimeno, 1994; McPherson, 1983). In contrast, the current study
suggested that the collaborative aspect of niche overlap needs to be emphasized as
well, especially when resources in the community do not diminish with the increase
of consumption. In the current study, development projects targeting the same sector,
therefore having similar resource demands, had a higher chance of sharing ideas.
The study applied a novel class of methodologies in social network analysis
to explore multidimensional networks. In particular, the study demonstrated the
usefulness of the ERGM approach (Crouch et al., 1998) for modeling multiplex
network dynamics. The current study is one of the few studies (e.g. Lazega &
Pattison, 1999; Lomi & Pattison, 2006) that have applied the ERGM to the multiplex
network environment in order to explain networks of interlocking domains. The
study supported the argument that high-order parameters can effectively explain
complex network patterns (Robins et al., 2008), as evidenced in the K-Triangle
parameters. Overall, the study contributed to the literature of empirical studies that
adopt the statistical modeling of network dynamics (e.g. Pattison & Robins, in press;
Robins et al., 2007; Wasserman et al., 2007).
Implications for the ICT for Development Community
The study, driven by three practical issues with regard to organizational
partnerships in the community (see Chapter 1), introduced a research framework for
examining these issues. The current study is the first to empirically investigate

172
interorganizational network structure, its antecedents, and its effects in the ICT for
Development community. The study has shown the usefulness of evolutionary and
network theories in understanding the changes of organizational partnership patterns
in the field. Resource dynamics are of key importance to the community because
organizations in the ICT for Development field face the needs to pool resources
together for effective collaboration (UNESCO, 2006). The study provided
understandings about the community by highlighting interorganizational network
dynamics and the collaborative processes of project implementation and knowledge
sharing. From a practical viewpoint, network dynamics observed in the study provide
implications for organizations’ networking strategies in the ICT for Development
community. Ultimately, the findings can serve as a reference point for thinking about
how to proceed in implementing partnerships (Mansell & Wehn, 1998; UNESCO,
2006).
This study provided an initial attempt to utilize data on ICT for Development
projects for understanding the evolution of organizational partnerships and the larger
community. While there are numerous case studies that examine partnerships (e.g.
Unwin, 2005), there has been a dearth of a structured approaches to a quantitative
investigation of organizational collaborations in the field. One of the practical
contributions of the current study is that it suggested a systematic framework for
documenting and analyzing organizational partnerships that have been formed in the
past. This knowledge repository can be used to assist the choice of potential

173
collaboration partners based on information such as the type of projects, geographic
location, and organizational characteristics.
Limitations and Future Research
Limitations
This study has some empirical limitations due to the nature of the data. First,
this study is not representative of all organizational populations in the ICT for
Development community. The major sources of the dataset used in the study are
IGOs and other traditional donor organizations. In consequence, projects designed
and implemented by other sectors including private corporations, smaller NGOs, and
non-traditional donor organizations such as private philanthropic foundations are
underrepresented. This fact limits the possibility of making generalizable inferences
about population dynamics in the ICT for Development community. However, the
limitation is homogenous across years and therefore, does not impede the
longitudinal investigation of shifts in organizational populations and networks. Once
a more extensive and complete dataset becomes available, the study will be able to
provide more accurate analyses about organizations and organizational populations
in the ICT for Development community.
Second, the data source lacks detailed information on collaborating
organizations. For example, in some cases, the names of private organizations are not
listed for confidentiality reasons. In addition, larger organizations tend to have a
larger presence, and the names of local and community-based NGOs are often left
out from the data. In the case of knowledge-sharing projects, records of collaborating

174
organizations are often limited to host and sponsor organizations and do not include a
more extensive list of participating organizations. These limitations are likely to have
led to imperfect representation of collaborating organizations in development
projects.
Third, the dataset did not provide details on the nature of ties, such as the
substance, direction, and duration of relations within each project. Therefore, the
current study treated ties as collaborative, symmetric, and being existent throughout
the duration of each project. Once these details of information become available, the
current study can be extended in substantial ways to account for the effects of tie
content and direction on various outcomes. These extensions may reveal useful
insights about the direction of resource flows, power relations among organizations,
collaborative versus competitive ties, and the intensity of relations.
There are some methodological limitations as well. First, the study did not
consider multiplex dynamics in a strict longitudinal sense. Analysis of multiplex tie
formation over time will allow examining meaningful distinctions about the
directions of multiplex tie formation over time. For instance, it is worthwhile to
explore whether ties in the implementation network lead to ties in the knowledge-
sharing network or the opposite direction is more prevalent. Such analyses can be
made possible by more sophisticated tools for simultaneously modeling both
multivariate and longitudinal network dynamics, such as XLPNet, which is currently
in the development phase (Wang et al., 2008). In terms of testing the network effects,
the study did not consider the temporal aspect, and, therefore, could not suggest

175
interpretations about the diffusion aspect. If a temporal component is built into the
network of learning, the results can show whether knowledge about technologies and
applications is transferred from earlier projects to those implemented later in time,
which can be regarded as the diffusion of ideas.
Second, this study used MultiNet and PNet/XPNet for the analysis of
subgroup network structure and multiplex network dynamics. Both programs have
room for further development. In MultiNet, the pseudo-likelihood measure is
considered to be misleading in some cases when estimated models are close to
degeneracy or when there is a strong dependency exhibited in the observed data
(Wasserman et al., 2007). While PNet overcomes this unreliability by fitting models
with MCMC simulation procedures, it has some drawbacks. For example, PNet does
not give a measure of overall model fit which can be compared across networks
(Harrigan, 2008; Wang et al., 2008).
Directions for Future Research
This study provides the basis for a wide range of interesting future theorizing
and research. These issues are discussed in details below.
Competitive Dynamics
This study primarily focused on collaborative aspects of organizational
networks. While there are commensalistic and symbiotic linkages among
organizational populations, competitive relations are important for understanding
population dynamics as well (Burt, 1992; Monge et al., in press). In development
communities, there is a high level of competition for scarce resources among NGOs

176
to obtain financial resources, political opportunity, and media attention (McAdam et
al., 1996; McCarthy & Zald, 1977). In the context of the ICT for Development
community, competition can exist both at the organizational and project level.
Multiple projects may compete for implementing technologies and services in a
particular region. Various organizations may compete for collaborating in a particular
project. Future research needs to examine the way in which the competitive
dynamics between organizations and organizational populations influence the
evolution of the community.
Network Dysfunctionalities and Failure
Networks can be detrimental, and this aspect has important implications in
the context of civil society and development communities in particular. For example,
there have been debates about the negative aspects of NGOs’ dependence on donor
funding due to the fact that they may be less able to pursue radical agenda (Lister,
2004). Issues about the dependence of recipient organizations on government have
also been subject to debate (Lister, 2004). In general, Smith (2004) pointed out that
“ties with external actors can also be interpreted as a weakness because they can
undermine the autonomy of an organization” (p. 272). The current did not consider
network failure. Podolny and Page (1998) stressed that there has been scant
empirical research examining the fact that network forms of organization, such as
strategic alliances, research consortia, and outsourcing agreements, can lead to
dysfunctionalities and failure. In the future, there need to be studies examining the
conditions under which networks achieve the desired goals versus those under which

177
networks fail to achieve their goals. From the perspective of evolutionary theories,
partnerships that fail or terminate before their intended end will provide important
insights about strategies for successful partnership practices.
Network Outcome
This study examined the sharing of technologies and applications as the
collective outcome of the community. In future studies, the effects of network
structure and dynamics on more diverse outcome variables can be explored. Studies
about the effect of network ties in non-commercial contexts encompass a diverse
topic area. For example, the outcome of a collaboration network for a project can be
measured by various success indicators such as media attention, funding, and
mobilization. Ingram et al. (2005) studied the effect of IGO connectedness on trade.
Authors found that trade benefits occur through the strength of IGO connections
created by joint membership between the countries, including IGOs in not only
economic but also social and cultural realms. As another example, Baldassarri and
Diani (2007) tested the contribution of civic networks to democratic outcomes.
The question of how different network structures lead to effective outcomes
has been subject to debate (Baldassarri & Diani, 2007). Further, scholars have
suggested that the effectiveness of organizational networks needs to be evaluated at
multiple levels: community, network, and organization and participant levels (Katz &
Anheier, 2005; Provan & Milward, 2001). The multilevel network approach
employed in the study can provide insights for pursuing this idea.

178
External Environments
One aspect of evolutionary theory that has not been considered in the current
study is the effect of external environmental conditions. Across the evolutionary
stages of the emergence, growth, transformation, and demise of a community,
environmental events can exert critical influences (Baum & Singh, 1994a; Carroll &
Hannan, 2000). For example, in the ICT for Development community, major
technological innovations throughout the history would have had impacts on the
implementation of projects. New technologies will lead to variations in the projects,
and these projects will be either selected for or selected out, and the more successful
ones will be retained over time. These evolutionary processes relate to the central
issue of the community, that is the replicability and scalability of projects.
Major conferences, such as the WSIS, can have influences on the dynamics
of network ties as well. Moreover, the munificence or lack of resources at the
collective level of the community is important, which can be measured by the total
amount of funding or development aid allocated to the field. At the national level, the
level of economic development, human development, and telecommunications
diffusion and adoption can be relevant measures of resource munificence. Changes in
the resource space over time will influence the evolutionary dynamics at various
levels, including organizations, projects, and networks. While these issues are
beyond the scope of this dissertation, they are topics that deserve further research.

179
Broader Development Communities
The analyses undertaken in the current study can be applied to the
examination of partnerships in the broader development practice. First of all, the
current study has examined one subset of ICT for Development community, which
has been determined by the selection criteria with regard to the types of technologies
(e.g. focus on advanced networked technologies) and projects (e.g. focus on projects
in developing country locations). These choices may have impacted the results by
influencing not only the boundary of organizational populations but also the
networking patterns among them that have been taken into account in the analyses.
Future research is encouraged to take a broader sample of ICT for Development
projects and investigate whether different selection criteria lead to differences in the
findings and implications about network structure, dynamics, and effects. In addition,
the current study can be applied to other development sectors in which organizational
partnerships practices play an important role.
Conclusion
This dissertation presents a research framework for examining
multidimensional network evolution in the ICT for Development community. The
findings on network structure and dynamics, in combination, suggest that
multimodal and multiplex dynamics are significant drivers of tie formation at the
dyadic level and the emergence of network structure at the global level. The study
emphasizes that both endogenous ties and exogenous environmental and nodal
attributes influence multiplex dynamics. Overall, the study builds on the past

180
literatures on social networks and evolutionary theories by examining the conditions
under which multiplex ties are selected for and retained. The study showed a
significant relationship between organizational networks and knowledge sharing,
supporting the idea that interorganizational networks play an important role in the
way knowledge and ideas are shared. The dissertation provides substantive
implications for understanding the current organizational network structure in the
ICT for Development community and facilitating collaborative partnership designs
for sustainable development initiatives.

181
BIBLIOGRAPHY
Ahuja, G. (2000a). Collaboration networks, structural holes, and innovation: A
longitudinal study. Administrative Science Quarterly, 45, 425-455.
Ahuja, G. (2000b). The duality of collaboration: Inducements and opportunities in
the formation of interfirm linkages. Strategic Management Journal, 21(3),
317-343.
Aldrich, H. (1976). Resource dependence and interorganizational relations: Relations
between local employment service offices and social services sector
organizations. Administration and Society, 8, 419-453.
Aldrich, H. (1979). Organizations and environments. Englewood Cliffs, NJ:
Prentice-Hall.
Aldrich, H. (1999). Organizations evolving. Thousand Oaks, CA: Sage Publications.
Aldrich, H., & Reuf, M. (2006). Organizations evolving. London: Sage Publications.
Anheier, H., & Katz, H. (2003). Mapping global civil society. In M. Kaldor, H.
Anheier & M. Glasius (Eds.), Global civil society 2003 (pp. 206-221).
London: Sage Publications.
Anheier, H., & Katz, H. (2004). Network approaches to global civil society. In H.
Anheier, M. Glasius & M. Kaldor (Eds.), Global civil society 2004/5 (pp.
206-221). London: Sage Publications.
Arya, B., & Lin, Z. (2007). Understanding collaboration outcomes from an extended
resource-based view perspective: The roles of organizational characteristics,
partner attributes, and network structures. Journal of Management, 33(5),
697-723.
Astley, W. G. (1985). The two ecologies: Population and community perspective on
organizational evolution. Administrative Science Quarterly, 30, 224-241.
Astley, W. G., & Fombrun, C. J. (1987). Organizational communities: An ecological
perspective. In Research in the sociology of organizations (V ol. 5, pp. 163-
185). Greenwich, CT: JAI Press.
Audia, P. G., Freeman, J., & Reynolds, P. D. (2006). Organizational foundings in
community context: Instruments manufacturers and their interrelationship
with other organizations. Administrative Science Quarterly, 51, 381-419.

182
Baldassarri, D., & Diani, M. (2007). The integrative power of civic networks.
American Journal of Sociology, 113(3), 735-780.
Ballantyne, P. (2003). Ownership and partnership: Keys to sustaining ICT-enabled
development activities (Research Brief No. 8). The Hague, Netherlands:
International Institute for Communication and Development (IICD).
Barabasi, A. L. (2002). Linked: The new science of networks. Cambridge, MA:
Perseus.
Barnett, G. (1999). The social structure of international telecommunications. In H.
Sawhney & G. Barnett (Eds.), Progress in communication science (V ol. 15.
Advances in telecommunications, pp. 151-186). Stanford, CT: Ablex.
Barnett, G. (2001). A longitudinal analysis of the international telecommunication
network, 1978-1996. American Behavioral Scientist, 44, 1638-1655.
Barnett, G., Salisbury, J., Kim, C., & Langhorne, A. (1999). Globalization and
international communication networks: An examination of monetary,
telecommunications, and trade networks. The Journal of International
Communication, 6(2), 7-49.
Barnett, W. P. (1990). The organizational ecology of a technological system.
Administrative Science Quarterly, 35(1), 31-60.
Baum, J. A. C. (1999). Whole-part coevolutionary competition in organizations. In J.
A. C. Baum & B. McKelvey (Eds.), Variations in organizations: In honor of
Donald T. Campbell (pp. 113-135). Thousand Oaks, CA: Sage.
Baum, J. A. C., & Ingram, P. (2002). Interorganizational learning and network
organizations: Toward a behavioral theory of the interfirm. In M. Augier & J.
G. March (Eds.), Economics of Choice, Change, and Organization. Essays in
the memory of Richard M. Cyert. Cheltenham UK: Edward Elgar.
Baum, J. A. C., & McKelvey, B. (1999). Variations in organizations: In honor of
Donald T. Campbell. Thousand Oaks, CA: Sage.
Baum, J. A. C., & Oliver, C. (1991). Institutional linkages and organizational
mortality. Administration Science Quarterly, 36(2), 187-218.
Baum, J. A. C., Rowley, T. J., & Shipilov, A. V . (2004). The small world of Canadian
capital markets: Statistical mechanics of investment bank syndicate networks,
1952-1990. Journal of Administrative Sciences, 21, 307-325.

183
Baum, J. A. C., & Singh, J. V . (1994a). Organization-environment coevolution. In J.
A. C. Baum & J. V . Singh (Eds.), Evolutionary dynamics of organizations (pp.
379-401). New York: Oxford University Press.
Baum, J. A. C., & Singh, J. V . (1994b). Organizational hierarchies and evolutionary
processes: Some reflections on a theory of organizational evolution. In J. A.
C. Baum & J. V . Singh (Eds.), Evolutionary dynamics of organizations (pp. 3-
20). New York: Oxford University Press.
Baum, J. A. C., & Singh, J. V . (1994c). Organizational niches and the dynamics of
organizational founding. Organization Science, 5(4), 483-501.
Baum, J. A. C., & Singh, J. V . (1994d). Organizational niches and the dynamics of
organizational mortality. American Journal of Sociology, 100, 346-380.
Bearman, P., Moody, J., & Stovel, K. (2004). Chains of affection: The structure of
adolescent romantic and sexual networks. American Journal of Sociology,
110, 44-99.
Best, M., & Maclay, C. (2002). Community Internet access in rural areas: Solving
the economic sustainability puzzle. In In World Economic Forum, The global
information technology report 2001-2002: Readiness for the networked world
(pp. 76-88). New York: Oxford University Press.
Blau, P. M. (1964). Exchange and power in social life. New York: Wiley.
Boli, J., & Thomas, G. (Eds.). (1999). Constructing world culture: International
nongovernmental organizations since 1875. Stanford: Stanford University
Press.
Borgatti, S. P. (2006). Identifying sets of key players in a social network.
Computational & Mathematical Organization Theory, 12(1), 21-34.
Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002a). Ucinet 6 for Windows
Reference Guide. Harvard, MA: Analytic Technologies. Available with the
Ucinet 6 Software.
Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002b). Ucinet 6 for Windows:
Software for Social Network Analysis [Computer program]. Harvard, MA:
Analytic Technologies.
Borgatti, S. P., & Foster, P. (2003). The network paradigm in organizational research:
A review and typology. Journal of Management, 29(6), 991-1013.

184
Brass, D. J., Galaskiewicz, J., Greve, H. R., & Tsai, W. P. (2004). Taking stock of
networks and organizations: A multilevel perspective. Academy of
Management Journal, 47(6), 795-817.
Breiger, R. (1974). The duality of persons and groups. Social Forces, 53(2), 181-190.
Bryant, J. A., & Monge, P. (2008). The evolution of the children's television
community, 1953-2003. International Journal of Communication, 2, 160-192.
Burt, R. (1976). Positions in networks. Social Forces, 50, 93-122.
Burt, R. (1992). Structural holes: The social structure of competition. Cambridge,
MA: Harvard University Press.
Campbell, D. T. (1965). Variation and selective retention in socio-cultural revolution.
In H. R. Barringer, G. I. Blanksten & R. W. Mack (Eds.), Social change in
developing area: A reinterpretation of evolutionary theory (pp. 19-48).
Cambridge, MA: Schenkman.
Carroll, G. R. (1985). Concentration and specialization: Dynamics of niche width in
populations of organizations. American Journal of Sociology, 90, 1262-1283.
Carroll, G. R., & Hannan, M. (2000). The demography of corporations and industries.
Princeton, NJ: Princeton University Press.
Carroll, G. R., & Wade, J. (1991). Density dependence in the organizational
evolution of the American brewing industry across different levels of analysis.
Social Science Research, 20, 271-302.
Castells, M. (2000a). The rise of the network society (2nd ed.). Oxford: Blackwell.
Castells, M. (2000b). Toward a sociology of the network society. Contemporary
Sociology, 29(5), 693-699.
Cecchini, S., & Scott, C. (2003). Can information and communications technology
applications contribute to poverty reduction? Lessons from rural India.
Information Technology for Development, 10, 73-84.
Cernea, M. (1988). Non-governmental organizations and local development (World
Bank Discussion Paper No. 40). Washington, D.C.: World Bank.
Chib, A. (2007). Network influences in health initiatives: Multimedia games for
youth in Peru. Unpublished Doctoral Dissertation, University of Southern
California, Los Angeles, CA, Los Angeles, CA.

185
Chung, S., Singh, H., & Lee, K. (2000). Complementarity, status similarity, and
social capital as drivers of alliance formation. Strategic management journal,
21(1), 1-22.
Contractor, N. S., Wasserman, S., & Faust, K. (2006). Testing multi-theoretical
multilevel hypotheses about organizational networks: An analytic framework
and empirical example. Academy of Management Review, 31(3), 681-703.
Corman, S. R., & Scott, C. R. (1994). Perceived networks, activity, foci, and
observable communication in social collectives. Communication Theory, 4,
171-190.
Creech, H., & Willard, T. (2004). Notes on trends and niches in the ICT4D/K4D
debate. Winnipeg, Canada: International Institute for Sustainable
Development.
Crouch, B., Wasserman, S., & Contractor, N. (1998). A practical guide to fitting p*
social network models via logistic regression. Connections, 21, 87-101.
Dacin, T. M., Ventresca, M. J., & Beal, B. D. (1999). The embeddedness of
organizations: Dialogue and directions. Journal of Management, 25(3), 317-
356.
Danowski, J. (2000). Arab countries' global telephone traffic networks. In H. Amin &
L. Gher (Eds.), Civic discourse & digital age communications in the Middle
East. Norwood, NJ: Ablex.
Diani, M. (2002). Network analysis. In B. Klandermans & S. Staggenborg (Eds.),
Methods of social movement research (pp. 173-200). Minneapolis, MN:
University of Minnesota Press.
Diani, M. (2003a). "Leaders" or brokers: Positions and influence in social movement
networks. In M. Diani & D. McAdam (Eds.), Social movements and
networks: Relational approaches to collective action (pp. 105-122). Oxford,
UK: Oxford University Press.
Diani, M. (2003b). Networks and social movements: A research programme. In M.
Diani & D. McAdam (Eds.), Social movements and networks: Relational
approaches to collective action (pp. 299-319). Oxford, UK: Oxford
University Press.
DiMaggio, P. J. (1986). Structural analysis of organizational fields: A blockmodel
approach. In B. M. Staw & L. L. Cummings (Eds.), Research in
organizational behavior (V ol. 8, pp. 335-370). Greenwich, CT: JAI Press.

186
DiMaggio, P. J. (1994). The challenge of community evolution. In J. A. C. Baum & J.
V . Singh (Eds.), Evolutionary dynamics of organizations (pp. 299-319). New
York: Oxford University Press.
DiMaggio, P. J. (2003). An ecological perspective on nonprofit research. In H. K.
Anheier & A. Ben-ner (Eds.), The study of the nonprofit enterprise: Theories
and approaches (pp. 311-320). New York: Kluwer Academic/Plenum
Publishers.
DiMaggio, P. J., & Anheier, H. (1990). The sociology of nonprofit organizations and
sectors. Annual Review of Sociology, 16, 137-159.
DiMaggio, P. J., & Powell, W. W. (1983). The iron cage revisited: Institutional
isomorphism and collective rationality in organizational fields. American
Sociological Review, 48, 147-160.
Emirbayer, M., & Goodwin, J. (1994). Network analysis, culture, and the problem of
agency. American Journal of Sociology, 99(6), 1411-1454.
Faust, K., & Skvoretz, J. (2002). Comparing networks across space and time, size
and species. In R. Stolzenberg (Ed.), Sociological Methodology (V ol. 32).
Cambridge, MA: Basil Blackwell.
Feld, S. (1981). The focused organization of social ties. American Journal of
Sociology, 86(5), 1015-1035.
Fesenmaier, J., & Contractor, N. (2001). Inquiring Knowledge Networks on the Web
(IKNOW): The evolution of knowledge networks for rural development. The
Journal of the Community Development Society, 32, 160-175.
Fienberg, S. E., Meyer, M., & Wasserman, S. (1981). Analyzing data from
multivariate directed graphs: An application to social networks. In V . Barnett
(Ed.), Interpreting multivariate data (pp. 289-306). London: John Wiley &
Sons.
Fillip, B. (2003). ICT4D - Information and Communication Technologies for
Development: Self-paced learning materials. Retrieved November 15, 2006,
from http://www.knowledgefordevelopment.com/ICT4D03SP/
Fombrun, C. J. (1986). Structural dynamics within and between organizations.
Administrative Science Quarterly, 31, 403-421.
Frank, O., & Strauss, D. (1986). Markov graphs. Journal of American Statistical
Association, 81, 832-842.

187
Freeman, J., & Audia, P. G. (2006). Community ecology and the sociology of
organizations. Annual Review of Sociology, 32, 145-169.
Freeman, J., & Hannan, M. (1983). Niche width and the dynamics of organizational
populations. American Journal of Sociology, 88, 1116-1145.
Freeman, L. C. (1979). Centrality in networks: 1. Conceptual clarification. Social
Networks, 1, 215-239.
Fulk, J. (2001). Global organizational networks: Emergence and future prospects.
Human Relations, 54(1), 91-99.
Galaskiewicz, J., & Bielefeld, W. (1998). Nonprofit organizations in an age of
uncertainty: A study of organizational change. New York: Aldine de Gruyter.
Galaskiewicz, J., Bielefeld, W., & Dowell, M. (2006). Networks and organizational
growth: A study of community based nonprofits. Administrative Science
Quarterly, 51, 337-380.
Galaskiewicz, J., & Marsden, P. V . (1978). Community structure as
interorganizational linkages. Annual Review of Sociology, 4, 455-484.
Galaskiewicz, J., & Wasserman, S. (1989). Mimetic and normative processes within
an interorganizational field: An empirical test. Administrative Science
Quarterly, 34, 454–480.
Galaskiewicz, J., & Zaheer, A. (1999). Networks of competitive advantage. In S. B.
Andrews & D. Knoke (Eds.), Research in the sociology of organizations (pp.
237-261). Greenwich, CT: JAI Press.
Galtung, J. (1971). A structural theory of imperialism. Journal of Peace Research, 8,
81-118.
Gimeno, J. (1994). Competition within and between networks: The contingent effect
of competitive embeddedness on alliance formation. Academy of
Management Journal, 47(6), 820-842.
GKP. (2003). Multistakeholder partnerships: Issue paper. Kuala Lumpur: Global
Knowledge Partnerships.
Gnyawali, D. R., & Madhavan, R. (2001). Cooperative networks and competitive
dynamics: A structural embeddedness perspective. Academy of Management
Review, 26(3), 431-445.

188
Goodreau, S. M. (2007). Advances in exponential random graph (p*) models applied
to a large social network. Social Networks, 29, 231-248.
Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology,
78, 1360-1380.
Granovetter, M. (1985). Economic action and social structure: the problem of
embeddedness. American Journal of Sociology, 61(3), 481-520.
Granovetter, M. (1994). Business groups. In N. J. Smelser & R. Swedberg (Eds.),
The handbook of economic sociology. Princeton, NJ: Princeton University
Press.
Gulati, R. (1995). Social structure and alliance formation patterns: A longitudinal
analysis. Administrative Science Quarterly, 40(4), 619-652.
Gulati, R. (1999). Network location and learning: The influence of network resources
and firm capabilities on alliance formation. Strategic Management Journal,
20, 397-420.
Gulati, R. (2007). Managing network resources: Alliances, affiliations, and other
relational assets. Oxford: Oxford University Press.
Gulati, R., & Gargiulo, M. (1999). Where do interorganizational networks come
from? American Journal of Sociology, 104(5), 1439-1493.
Hannan, M., Carroll, G. R., Dundon, E. A., & Torres, J. C. (1995). Organizational
evolution in a multinational context: Entries of automobile manufacturers in
Belgium, Britain, France, Germany, and Italy. American Sociological Review,
60, 509-528.
Hannan, M., & Freeman, J. (1977). The population ecology of organizations.
American Journal of Sociology, 85, 929-964.
Hannan, M., & Freeman, J. (1984). Structural inertia and organizational change.
American Sociological Review, 49, 149-164.
Hanneman, R. A., & Riddle, M. (2005). Introduction to social network methods,
from http://www.faculty.ucr.edu/~hanneman/nettext/
Harrigan, N. (2008). PNet for dummies: An introduction to estimating exponential
random graph (p*) models with PNet, from
http://www.sna.unimelb.edu.au/pnet/pnet.html

189
Harris, R., & Rajora, R. (2006). Empowering the poor: Information and
communications technology for governance and poverty reduction - A study
of rural development projects in India. Bangkok, Thailand: UNDP-APDIP
(United Nations Development Programme - Asia-Pacific Development
Information Programme).
Haveman, H., & Romanelli, E. (1997). Geography and organizational founding:
Unpublished manuscript, Johnson Graduate School of Business, Cornell
University.
Hawley, A. H. (1986). Human ecology: A theoretical essay. Chicago: University of
Chicago Press.
Hedström, P. (2005). Dissecting the social: On the principles of Analytical Sociology.
Cambridge: Cambridge University Press.
Held, D., McGrew, A., Goldblatt, D., & Perraton, J. (1999). Global transformations:
politics, economics, and culture. Stanford, CA: Stanford University Press.
Homans, G. C. (1950). The human group. New York: Harcourt Brace.
Hovland, I. (2003). Knowledge management and organisational learning: An
international development perspective. (Working Paper No. 224). London:
Overseas Development Institute (ODI).
Hunt, C. S., & Aldrich, H. (1998). The second ecology: Creation and evolution of
organizational communities. Research in Organizational Behavior, 20(267-
301).
Igarashi, T., Robins, G., & Pattison, P. (2006). Longitudinal changes in friendship
networks: An approach from exponential random graph models, NetSci2006.
Bloomington, IN.
Ingram, P., & Roberts, P. W. (2000). Friendships among competitors in the Sydney
hotel industry. American Journal of Sociology, 106(2), 387-423.
Ingram, P., Robinson, J., & Busch, M. L. (2005). The intergovernmental network of
world trade: IGO connectedness, governance and embeddedness. American
Journal of Sociology, 111(3), 824-858.
ITU. (2003). World telecommunication development report 2003. Geneva,
Switzerland: International Telecommunication Union.

190
Jones, C., Hesterly, W. S., & Borgatti, S. P. (1997). A general theory of network
governance: Exchange conditions and social mechanisms. Academy of
Management Review, 22, 911-945.
Kaldor, M., Anheier, H., & Glasius, M. (2003). Global civil society in an era of
regressive globalisation. In M. Glasius, M. Kaldor & H. Anheier (Eds.),
Global civil society 2003 (pp. 3-17). London, UK: Sage Publications.
Kaldor, M., Anheier, H., & Glasius, M. (2004). Introduction. In H. Anheier, M.
Glasius & M. Kaldor (Eds.), Global civil society 2004/5 (pp. 1-22). London,
UK: Sage Publications.
Katz, H., & Anheier, H. (2005). Global connectedness: The structure of transnational
NGO networks. In M. Glasius, M. Kaldor & H. Anheier (Eds.), Global civil
society 2005/6 (pp. 240-262). London, UK: Sage Publications.
Kauffman, S. (1993). The origins of life: Self-organization and selection in evolution.
New York: Oxford University Press.
Kick, E., & Davis, B. (2001). World-system structure and change: An analysis of
global networks and economic growth across two time periods. American
Behavioral Scientist, 44(10), 1561-1578.
Kim, T.-Y ., Oh, H., & Swaminathan, A. (2006). Framing interorganizational network
change: A network inertia perspective. Academy of Management Review,
31(3), 704-720.
King, K., & McGrath, S. (2003). Knowledge sharing in development agencies:
Lessons from four case studies (The World Bank Operations Evaluation
Department). Washington, DC: World Bank.
Koka, B. R., Madhavan, R., & Prescott, J. E. (2006). The evolution of interfirm
networks: Environmental effects on patterns of network change. Academy of
Management Review, 31(3), 721-737.
Krackhardt, D. (1987). QAP- Partialing as a test of spuriousness. Social Networks, 9,
171-186.
Krackhardt, D. (1988). Predicting with networks: Nonparamentric multiple
regression analysis of dyadic data. Social Networks, 10, 359-381.
Langton, N. (1987). Niche theory and social movements: A population ecology
approach. The Sociological Quarterly, 28(1), 51-70.

191
Laumann, E. O., Galaskiewicz, J., & Marsden, P. V . (1978). Community structure as
interorganizational linkages. Annual Review of Sociology, 4, 455-484.
Lazega, E., & Pattison, P. (1999). Social capital, multiplex generalized exchange and
cooperation in organisations: A case study. Social Networks, 21, 67-90.
Lazer, D., & Andre-Clark, A. (2000). The knowledge in the network. Cambridge,
MA: Program on Networked Governance, John F. Kennedy School of
Government, Havard University.
Lee, S., & Chib, A. (2008). Wireless for the poor: No strings Attached? A framework
for wireless initiatives connecting rural areas. In N. Carpentier & B. D. Cleen
(Eds.), Participation and media production. Critical reflections on content
creation. ICA 2007 Conference Theme Book. Newcastle, UK: Cambridge
Scholars Publishing.
Lee, S., Monge, P., Bar, F., & Matei, S. A. (2007). The emergence of clusters in the
global telecommunications network. Journal of Communication, 57(3), 415-
434.
Levine, S. S., & Kurzban, R. (2006). Explaining clustering in social networks:
Towards an evolutionary theory of cascading benefits. Managerial and
Decision Economics, 27, 173-187.
Levinthal, D. A., & March, J. G. (1994). The myopia of learning. Strategic
Management Journal, 14, 95-112.
Levitt, B., & March, J. G. (1988). Organizational learning. Annual Review of
Sociology, 14(319-340).
Lewicki, R. J., McAllister, D. J., & Bies, R. J. (1998). Trust and distrust: New
relationships and realities. Academy of Management Review, 23(3), 438-458.
Lewis, D. (1998). Development NGOs and the challenge of partnership: Changing
relations between North and South. Social Policy & Administration, 32(5),
501-512.
Lewis, D., & Madon, S. (2004). Information systems and nongovernmental
development organizations: Advocacy, organizational learning, and
accountability. The Information Society, 20(2), 117-126.
Lewis, D., & Wallace, T. (Eds.). (2000). New roles and relevance: Development
NGOs and the challenge of change. Hartford, CT: Kumarian Press.

192
Lister, S. (2004). The future of international NGOs: New challenges in a changing
world order. London, UK: British Overseas NGOs for Development (BOND).
Lomi, A. (1995). The population ecology of organizational founding: Location
dependence and unobserved heterogeneity. Administrative Science Quarterly,
40, 111-144.
Lomi, A. (1997). Markets with hierarchies and the network structure of
organizational communities. Journal of Management and Governance, 1, 49-
66.
Lomi, A., & Pattison, P. (2006). Manufacturing relations: An empirical study of the
organization of production across multiple networks. Organization Science,
17(3), 313-332.
Lorrain, F., & White, H. C. (1971). The structural equivalence of individuals in social
networks. Journal of Mathematical Sociology, 1, 49-80.
Lusthaus, C., & Milton-Feasby, C. (2006). The evaluation of inter-organizational
relationships in the not-for-profit sector: Some observations. Montreal:
Ottawa, Canada: Universalia.
MacLean, D. (2003). The quest for inclusive governance of global ICTs: Lessons
from the ITU in the limits of national sovereignty. Information Technologies
and International Development, 1(1), 1-18.
Madhavan, R., Koka, B. R., & Prescott, J. E. (1998). Networks in transition: How
industry events (re)shape interfirm relationships. Strategic Management
Journal, 19, 439-459.
Madon, S. (2000). International NGOs: Networking, information flows and learning
(The Development Informatics Working Paper No. 8). Manchester, UK:
Institute for Development Policy and Management.
Mansell, R., & Wehn, U. (1998). Knowledge societies: Information technology for
sustainable development. New York: Oxford University Press.
Martens, K. (2002). Mission impossible? Defining nongovernmental organizations.
Voluntas: International Journal of Voluntary and Nonprofit Organisations,
13(3), 271-285.
Matei, S. A. (2006). Globalization and heterogenization: Cultural and civilizational
clustering in telecommunicative space (1989-1999). Telematics and
Informatics, 23, 316-331.

193
McAdam, D., McCarthy, J. D., & Zald, M. N. (Eds.). (1996). Comparative
perspectives on social movements: Political opportunities, mobilizing
structures, and cultural framings. Cambridge, UK: Cambridge University
Press.
McCarthy, J. D., & Zald, M. N. (1977). Resource mobilization and social
movements: A partial theory. American Journal of Sociology, 82(6), 1212-
1241.
McEvily, B., & Zaheer, A. (1999). Bridging ties: A source of firm heterogeneity in
competitive capabilities. Strategic Management Journal, 20, 1133-1156.
McKelvey, B. (1994). Evolution and organizational science. In J. A. C. Baum & J. V .
Singh (Eds.), Evolutionary dynamics of organizations (pp. 314-326). New
York: Oxford University Press.
McPhee, R. D., & Corman, S. R. (1995). An activity-based theory of communication
networks in organizations, applied to the case of a church. Communication
Monographs, 62, 132-151.
McPherson, M. (1983). An ecology of affiliation. American Sociological Review,
48(4), 519-532.
McPherson, M., & Ranger-Moore, J. R. (1991). Evolution on a dancing landscape:
Organizations and networks in dynamic Blau space. Social Forces, 70(1), 19-
42.
McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather:
Homophily in social networks. Annual Review of Sociology, 27, 415-444.
Meyer, C. A. (1997). The political economy of NGOs and information sharing. World
Development, 25(7), 1127-1140.
Mizruchi, M. S., & Galaskiewicz, J. (1994). Networks of interorganizational
relations. In S. Wasserman & J. Galaskiewicz (Eds.), Advances in social
network analysis: Research in the social and behavioral sciences (pp. 230-
253). Thousand Oaks: Sage Publications.
Monge, P., & Contractor, N. (2003). Theories of communication networks. Oxford,
UK: Oxford University Press.
Monge, P., Heiss, B., & Margolin, D. (in press). Communication network evolution
in organizational communities. Communication Theory.

194
Monge, P., & Matei, S. A. (2004). The role of the global telecommunications
network in bridging economic and political divides, 1989 to 1999. Journal of
Communication, 54, 1-21.
Monge, P., & Poole, M. S. (in press). The evolution of organizational communication.
Journal of Communication.
Mudhai, O. F. (2004). Possible impacts of NGO-divide on ICT4D agenda: SSRC IT
& International Corporation's IT & Civil Society Network publications.
Ó Siochrú, S., & Girard, B. (2005). Community-based networks and innovative
technologies: New models to serve and empower the poor (UNDP Report
Series: Making ICT Work for the Poor). New York: Poverty Group, Bureau of
Development Policy, UNDP.
OECD. (2001). Digital opportunities for poverty reduction: Addressing the
international digital divide. Paris, France: Organisation for Economic
Cooperation and Development.
Oliver, C. (1990). Determinants of interorganizational relationships: Integration and
future directions. Academy of Management Review, 15(2), 241-265.
Osborn, R. N., & Hagedoorn, J. (1997). The institutionalization and evolutionary
dynamics of interorganizational alliances and networks. The Academy of
Management Journal, 40(2), 261-278.
Owen-Smith, J., & Powell, W. W. (2004). Knowledge networks as channels and
conduits: The effects of spillovers in the Boston biotechnology community.
Organization Science, 15(1), 5-21.
Padgett, J. F., & McLean, P. D. (2006). Organizational invention and elite
transformation: The birth of partnership systems in Renaissance Florence.
American Journal of Sociology, 111, 1463-1568.
Pattison, P. (2006). Multiple networks, NetSci 2006. Bloomington, Indiana.
Pattison, P., & Robins, G. (in press). Probabilistic network analysis. In T. Rudas (Ed.),
Handbook of probability theory with applications: Sage Publications.
Pattison, P., & Wasserman, S. (1999). Logit models and logistic regressions for social
networks, II. Multivariate relationships. British Journal of Mathematical and
Statistical Psychology, 52, 169-193.
Peli, G., & Nooteboom, B. (1999). Market partitioning and the geometry of the
resource space. American Journal of Sociology, 104(4), 1132-1153.

195
Pfeffer, J., & Nowak, P. (1976). Joint venture and interorganizational
interdependence. Administrative Science Quarterly, 21, 398-418.
Pfeffer, J., & Salancik, G. R. (1978). The external control of organizations: A
resource dependence perspective. New York: Harper Row.
Podolny, J. M., & Stuart, T. E. (1995). A role-based ecology of technological change.
American Journal of Sociology, 100(5), 1224-1260.
Podolny, J. M., Stuart, T. E., & Hannan, M. (1996). Networks, knowledge, and
niches: Competition in the worldwide semiconductor industry, 1984-1991.
American Journal of Sociology, 102(3), 659-689.
Powell, W. W. (1990). Neither market nor hierarchy: Network forms of
organization. In B. M. Staw & L. L. Cummings (Eds.), Research in
Organizational Behavior (V ol. 12, pp. 295-366): JAI Press.
Powell, W. W. (1998). Learning from collaboration: Knowledge and networks in the
biotechnology and pharmaceutical industries. California Management Review
(Special Issue on Knowledge and the Firm), 40(3), 228-240.
Powell, W. W., Koput, K., W., & Smith-Doerr, L. (1996). Interorganizational
collaboration and the locus of innovation: Networks of learning in
biotechnology. Administrative Science Quarterly, 41(1), 116-145.
Powell, W. W., White, D. R., Koput, K. W., & Owen-Smith, J. (2005). Network
dynamics and field evolution: The growth of interorganizational collaboration
in the life sciences. American Journal of Sociology, 110(4), 1132-1205.
Provan, K. G., Fish, A., & Sydow, J. (2007). Interorganizational networks at the
network level: A review of the empirical literature on whole networks.
Journal of Management, 33(3), 479-516.
Provan, K. G., & Milward, H. B. (2001). Do networks really work? A framework for
evaluating public-sector organizational networks. Public Administration
Review, 61(4), 414-423.
Provan, K. G., & Sebastian, J. G. (1998). Networks within networks: Service link
overlap, organizational cliques, and network effectiveness. Academy of
Management Journal, 41(4), 453-463.
Provan, K. G., Veazie, M. A., Staten, L. K., & Teufel-Shone, N. I. (2005). The use of
network analysis to strengthen community partnerships. Public
Administration Review, 65(5), 603-613.

196
Putnam, R. (2000). Bowling alone: The collapse and revival of American community.
New York: Simon & Shuster.
Rangan, S., Samii, R., & van Wassenhove, L. N. (2006). Constructive partnerships:
When alliances between private firms and public actors can enable creative
strategies. Academy of Management Review, 31(3), 738-751.
Rao, H., & Singh, J. V . (1999). Types of variation in organizational populations: The
speciation of new organizational forms. In J. A. C. Baum & B. McKelvey
(Eds.), Variations in organizations: In honor of Donald T. Campbell (pp. 63-
77). Thousand Oaks, CA: Sage.
Reinicke, W. H., & Deng, F. (2000). Critical choices: The United Nations, networks,
and the future of global governance. Ottawa, Canada: International
Development Research Centre.
Richards, W. D., & Seary, A. J. (2006). MultiNet 4.75 for Windows [Computer
Program]. Vancouver, British Columbia, Canada: Simon Fraser University.
Robins, G., & Pattison, P. (2006). Multiple networks in organisations. Retrieved
March 10, 2006, from
http://www.sna.unimelb.edu.au/publications/publications.html
Robins, G., Pattison, P., & Wang, P. (2008). Closure, connectivity and degrees: New
specifications for Exponential Random Graph (p*) models for directed social
networks, from http://www.sna.unimelb.edu.au/publications/publications.html
Robins, G., Pattison, P., & Woolcock, J. (2005). Small and other worlds: Global
network structures from local processes. American Journal of Sociology, 110,
894-936.
Robins, G., Snijders, T., Wang, P., Handcock, M., & Pattison, P. (2007). Recent
developments in exponential random graph (p*) models for social networks.
Social Networks, 29(2), 192-215.
Rosenau, J. (2003). Distant proximities: Dynamics beyond globalization. Princeton,
NJ: Princeton University Press.
Ruef, M. (2000). The emergence of organizational forms: A community ecology
approach. American Journal of Sociology, 106(3), 658-714.
Salamon, L. M., Sokolowski, W., & List, R. (Eds.). (2004). Global civil society:
Dimensions of the nonprofit sector (V ol. 2). Bloomfield, CT: Kumarian
Publishers.

197
Salancik, G. R. (1995). Wanted: A good network theory of organization.
Administrative Science Quarterly, 40(2), 345-349.
Scholte, J. A. (2000). Globalization: A critical introduction. New York: St. Martin's
Press.
Scott, C. R. (2005). Knowledge workers: Social, task and semantic network analysis.
Corporate Communications: An International Journal, 10(3), 257-277.
Scott, J. (2000). Social network analysis: A handbook. (2nd ed.). London: Sage.
Seary, A., & Richards, W. (2000). Fitting to p* models in MultiNet. Connections,
23(1), 84-101.
Shumate, M., Fulk, J., & Monge, P. (2005). Predictors of the HIV/AIDS INGO
network over time. Human Communication Research, 31, 482-510.
Singh, J. (2005). Collaborative networks as determinants of knowledge diffusion
patterns. Management Science, 51(5), 756-770.
Smith-Doerr, L., & Powell, W. W. (2005). Networks and economic life. In N. J.
Smelser & R. Swedberg (Eds.), The Handbook of Economic Sociology (pp.
379-402): Princeton University Press.
Smith, J. (1998). Global civil society? Transnational social movement organizations
and social capital. American Behavioral Scientist, 42(1), 93-107.
Smith, J. (2002). Bridging global divides? Strategic framing and solidarity in
transnational social movement organizations. International Sociology, 17(4),
505-528.
Smith, J. (2004). Exploring connections between global integration and political
mobilization. Journal of World-Systems Research, 10(1), 255-285.
Stinchcombe, A. L. (1990). Weak structural data (Review of Mizruchi and Schwartz).
Contemporary Sociology, 19, 380-382.
Stohl, C., & Stohl, M. (2005). Human rights, nation states, and NGOs: Structural
holes and the emergence of global regimes. Communication Monographs,
72(4), 442-467.
Stuart, T. E. (1998). Network positions and propensities to collaborate: An
investigation of strategic alliance formation in a high-technology industry.
Administrative Science Quarterly, 43(3), 668-698.

198
Swaminathan, A. (1995). The proliferation of specialist organizations in the
American wine industry, 1941-1990. Administrative Science Quarterly(40),
653-680.
Swaminathan, A. (2001). Resource partitioning and the evolution of specialist
organizations: The role of location and identity in the U.S. wine industry.
Academy of Management Journal, 44(6), 1169-1185.
Taylor, M., & Doerfel, M. L. (2003). Building interorganizational relationships that
build nations. Human Communication Research, 29(2), 153-181.
The Wireless Internet Institute(W2i). (2003). The wireless Internet opportunity for
developing countries. Boston: The Wireless Internet Institute, World Bank
infoDev & The United Nations ICT Task Force.
Tongia, R., Subrahmanian, E., & Arunachalam, V . S. (2005). Information and
communications technology for sustainable development: Defining a global
research agenda. Bangalore, India: Allied Publishers.
Townsend, J. G. (1999). Are non-governmental organisations working in
development a transnational community? Journal of International
Development, 11, 613-623.
UN-ICT Task Force. (2003). Challenges and partnerships: Opening up ICT to the
world. New York: Secretariat of the UN ICT Task Force.
UNESCO. (2006). Partnerships in development practice: Evidence from multi-
stakeholder ICT4D partnership practice in Africa. Paris, France: United
Nations Educational, Scientific and Cultural Organization.
Unwin, T. (2005). Partnerships in development practice: Evidence from multi-
stakeholder ICT4D partnership practice in Africa. Paris, France: UNESCO.
Usher, J. M., & Evans, M. G. (1996). Life and death along gasoline alley: Darwinian
and Lamarckian processes in a differentiating population. Academy of
Management Journal, 39(5), 1428-1466.
Uzzi, B. (1996). The sources and consequences of embeddedness for the economic
performance of organizations: The network effect. American Sociological
Review, 61, 674-698.
Uzzi, B. (1997). Social structure and competition in interfirm networks: The paradox
of embeddedness. Administrative Science Quarterly, 42, 35-67.

199
Uzzi, B. (1999). Embeddedness in the making of financial capital: How social
relations and networks benefit firms seeking financing. American
Sociological Review, 64(4), 481-505.
Van Rossem, R. (1996). The world system paradigm as general theory of
development: A cross-national test. American Sociological Review, 61, 508-
527.
Venkatraman, N., & Lee, C.-H. (2004). Preferential linkage and network evolution: A
conceptual model and empirical test in the U.S. video game sector. Academy
of Management Journal, 47(6), 876-892.
Ventresca, J. J., & Mohr, J. W. (2002). Archival research methods. In J. A. C. Baum
(Ed.), The Blackwell Companion to organizations (pp. 805-828). Malden,
MA: Blackwell.
Wallerstein, I. (1974). The modern world-system. New York: Academic Press.
Wang, P., Robins, G., & Pattison, P. (2008). PNet: Program for the Simulation and
Estimation of P* Exponential Random Graph Models. User Manual, from
http://www.sna.unimelb.edu.au/pnet/pnet.html
Wasserman, S. (1987). Conformity of two sociometric relations. Psychometrika,
52(1), 3-18.
Wasserman, S., & Faust, K. (1994). Social network analysis: Methods and
applications. New York: Cambridge University Press.
Wasserman, S., & Pattison, P. (1996). Logit models and logistic regressions for social
networks: 1: An introduction to Markov random graphs and p*.
Psychometrika, 60, 401-425.
Wasserman, S., Robins, G., & Steinley, D. (2007). Statistical models for networks: A
brief review of some recent research. In E. M. Airoldi, D. M. Blei, S. E.
Fienberg, A. Goldenberg, E. P. Xing & A. X. Zheng (Eds.), Statistical
Network Analysis: Models, Issues, and New Directions: ICML 2006
Workshop on Statistical Network Analysis, Lecture Notes in Computer
Science (V ol. 4503, pp. 45-56). Berlin / Heidelberg: Springer-Verlag.
Waterman, P. (2000). Social movements, local places and globalised places:
Implications for globalisation from below. In B. K. Gills (Ed.), Globalisation
and the politics of resistance. London: McMillan.
Watts, D. J. (1999). Small worlds: The dynamics of networks between order and
randomness. Princeton, NJ: Princeton University Press.

200
Williams, M. (2003). Global civil society: Expectations, capacities and the
accountability of international NGOs, from
http://www.21stcenturytrust.org/williams2.html
Williamson, O. (1981). The economics of organization: The transaction cost
approach. American Journal of Sociology, 87(3), 548-577.
Wilson, E. J. (2004). The information revolution and developing countries.
Cambridge, MA: MIT Press.
Wilson, E. J., & Best, M. L. (2003). What can we expect from the World Summit?
Information Technology and International Development, 1(2), 3-5.
Zajac, E. J., & Olsen, C. P. (1993). From transaction cost to transactional value
analysis: Implications for the study of interorganizational strategies. Journal
of Management Studies, 30(1), 131-145.

201
APPENDICES
APPENDIX A: GLOSSARY
Blockmodeling Blockmodeling is a technique designed to examine the
relationships among subsets of a network
partitioned based on nodal attributes.
Centrality The centrality of a node measures the extent to which a
node is central to a network.
Centralization Centralization is a group-level measure of the
variability or heterogeneity of the node centralities.
It represents the extent to which a graph has
disproportionate number of ties for selected nodes.
Commensalism Commensalism denotes the relationships among
organizational populations which overlap in their
niche space. Commensalist relations stem from the
similarity between nodes.
See: Symbiosis
Common third-party
tie
Two nodes have a common third-party tie when they
have a common connection to one or more alters.
Degree The degree of a node is its number of direct ties with
other nodes.
Degree Centrality The degree centrality of a node is its centrality
measured by its degree.
See: Degree, Centrality
Density The density of a network is the total number of ties
expressed as a proportion of the maximum possible
number of ties.
Dichotomous network A dichotomous network represents ties as either
present or absent for each pair of nodes, without
regard to the strength of ties.
See: Valued network

202
APPENDIX A: GLOSSARY (Continued)

Dyadic ties Dyadic ties are formed between a pair of nodes.
See: Triadic ties
ERGM (Exponential
Random Graph
Model)
The class of ERGMs, also known as p* models, is a
technique to examine a variety of structural
tendencies of a particular network.
See: p* models
Generalist
organizations
Generalist organizations rely upon a wide variety of
resources in the niche space.
See: Specialist organizations, Resource
partitioning
Heterogeneity Heterogeneity measure is an index of the extent to
which ties are concentrated on one node.
Markov models Markov models are based on the assumption that the
probability of a tie between i and j depends on
other ties that involve the same actor(s), i and j.
Mode Mode is a distinct set of entities on which the structural
variables are measured.
MRQAP (Multiple
Regression Quadratic
Assignment
Procedure)
MRQAP is a methodology to test the dependency
relations among multiple network matrices.
MTML
(Multitheoretical,
Multilevel)
MTML is an analytical framework to examine
networks based on diverse competing and/or
complementary theories and various levels of
analysis.
Multidimensional
networks
Multidimensional networks refer to the three
extensions (multimodal, multiplex, and multilevel)
to the basic network formulation (unimodal,
uniplex, and unilevel).

203
APPENDIX A: GLOSSARY (Continued)

Multilevel networks Multilevel networks refer to networks examined at
more than one level (e.g. dyadic, triadic, subgroup,
and global levels).
See: Multidimensional networks
Multimodal networks Multimodal networks refer to networks with more than
one set of nodes.
See: Multidimensional networks
Multiplex networks Multiplex networks refer to networks with more than
one type of relation.
See: Multidimensional networks
Niche Niche is a distinct combination of resources and other
constraints that suffices to support an
organizational form.
Niche overlap Organizations or organizational populations create
niche overlap when they have similar resource
demands and share the same niche space.
See: Niche
Normalized degree
centrality
Normalized degree centrality is a centrality measure
divided by the maximum possible ties for a
network of that size.
See: Degree centrality
Organizational
community
Organizational community is a diverse set of
interacting populations that form functional
relationships or interdependencies with each other
in a shared environmental space.
p* models p* models measure nonindependence among nodes by
including parameters for structural features that
capture hypothesized dependencies among ties.
QAP (Quadratic
Assignment
Procedure)
QAP is a methodology to test whether there is an
association between two network matrices on a
dyadic basis.

204
APPENDIX A: GLOSSARY (Continued)

Replicability Replicability in the context of ICT for Development
projects indicates that projects can be reasonably
recreated with new applications, in new
environments or in other countries.
Resource munificence Resource munificence is characterized by an
abundance of resources in the niche space.
See: Niche
Resource partitioning Resource partitioning is an ecological process in which
the concentration of generalist organizations
increases as populations evolve. At the same time,
narrow niches that are not covered by generalist
organizations are released to specialist
organizations.
See: Generalist organizations, Specialist
organizations
Retention Retention is an evolutionary process in which
positively selected variants are preserved over
time.
See: Variation, Selection
Scalability Scalability in the context of ICT for Development
projects indicates that projects are scaled up and/or
rolled out on a national level or in other countries.
Selection Selection is an evolutionary process in which certain
types of variations are selected for, while certain
types are selected against.
See: Variation, Retention
Specialist
organizations
Specialist organizations concentrate on a narrow range
of resources in the niche space.
See: Generalist organizations, Resource
partitioning

205
APPENDIX A: GLOSSARY (Continued)

Structural
embeddedness
Structural embeddedness focuses on the way the
structure of an organizational network and an
organization’s structural position influence
economic outcomes by influencing opportunities
for new social alliances and networks.
Structural equivalence Two nodes are structural equivalent if they have similar
relational patterns with other alters.
Structural inertia Structural inertia is an evolutionary concept that
explains a persistent organizational resistance to
change. In the context of interorganizational ties,
structural inertia refers to organizational resistance
to dissolve old relationships and form new network
ties.
Symbiosis Organizational populations are in a symbiotic
relationship when they have complementary
differences and form mutually interdependent ties.
See: Commensalism
Transitivity Transitivity refers to the structural patterns of triples of
nodes in a graph. A relation is transitive if
whenever i→j and j→k, then i→k.
See: Triadic ties
Triadic ties Triadic ties are formed among a triple of nodes.
See: Dyadic ties
Valued network A valued network represents ties along a range of
values that indicate the strength and/or frequency
of relations.
See: Dichotomous network
Variation Variation is an evolutionary process in which new
entrants are introduced to the pool of organisms (as
by mutation, trial, etc.).
See: Selection, Retention

206
APPENDIX B: TOPIC KEYWORDS FOR EACH SECTOR (AIDA
8
)
Sector Keywords
ICT and Rural
Development
radio and rural, broadcast and rural, satellite and rural,
ICT and rural, ICT and rural education, ICT and
agriculture, radio and agriculture, broadcasts and
agriculture, internet and agriculture, internet and rural,
telecenters, computer and rural, telecommunication and
rural.
ICT Capacity
Building
computer skill, ICT and training, computer training,
information technology and capacity building, media
training, distance learning
ICT in Health Sector ICT and health; information technology and health;
video and health; radio and health; internet and health;
communication and health; satellite and health; mass
media and health; telemedicine
ICT in Education
Sector
distance education; distance learning; school and
internet; education and information technology;
computer and learning; computer and training; computer
and teaching
ICT Initiatives PAN Asia, SDNP, Leland, Acacia, IICD, Cisco, Info21,
Infodev, Development Gateway, Bellanet, GKP, GDLN,
APC,
ICT Empowerment community access, telecenter, telecentre, public access,
universal access and ICT, community radio, kiosks,
ICT Infrastructure telecommunication, telecommunications, telecom,
internet connectivity, data network, information
infrastructure, telephone service
ICT Policy, Strategies
and Plans
internet and policy, ICT policy, communication policy,
information policy, telecommunication policy,
telecommunication and policy, internet policy, ICT
strategy,
E-Commerce e-commerce, electronic commerce, internet and
commerce
E-Readiness internet service, internet access, telephone service,
readiness and technology

8

http://aida.developmentgateway.org/TopicWindowView.do?viewtype=aida&tid=130&search_GO.x=1
1&search_GO.y=10

207
APPENDIX C: SOURCES OF INFORMATION IN AIDA
9

Primary sources of donor information:
– Inter-American Development Bank (IDB)
– IDRC Development Research Information System
– International Monetary Fund (IMF)
– John D. and Catherine T. MacArthur Foundation
– Natural Resources Information System (NARSIS)
– United Kingdom Department of Foreign International Development
(DFID)
– United Nations Capital Development Fund (UNCDF)
– United Nations Population Fund (UNFPA)
– United States Agency for International Development (USAID)
– World Bank Projects

9
http://aida.developmentgateway.org/aida/AidaSourcesDesc.do

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

The evolution of multidimensional and multilevel networks in online crowdsourcing

PDF

Ecology and network evolution in online innovation contest crowdsourcing

PDF

Communicating organizational knowledge in a sociomaterial network: the influences of communication load, legitimacy, and credibility on health care best-practice communication

PDF

The evolution of knowledge creation online: what is driving the dynamics of Wikipedia networks

PDF

The evolution of multilevel organizational networks in an online gaming community

PDF

Effects of economic incentives on creative project-based networks: communication, collaboration and change in the American film industry, 1998-2010

PDF

A multitheoretical multilevel explication of crowd-enabled organizations: exploration/exploitation, social capital, signaling, and homophily as determinants of associative mechanisms in donation-...

PDF

The evolution of scientific collaboration networks

PDF

Understanding alliance building and communication across multiple technology platforms among California's immigrant-serving NGOs

PDF

Intermedia agenda setting in an era of fragmentation: applications of network science in the study of mass communication

PDF

Crisis and stasis: understanding change in online communities

PDF

Community foundations and new governance networks: three studies exploring the role of regionally-networked philanthropic organizations in local problem-solving

PDF

Mapping out the transition toward information societies: social nature, growth, and policies

PDF

The Silicon Valley startup ecosystem in the 21st century: entrepreneurial resilience and networked innovation

PDF

Heterogeneous graphs versus multimodal content: modeling, mining, and analysis of social network data

PDF

Mobile communication and development: a study of mobile phone appropriation in Ghana

PDF

A view from below: the development and role of organizational social capital in neighborhood regeneration in Los Angeles

PDF

The formation and influence of online health social networks on social support, self-tracking behavior and weight loss outcomes

PDF

Knowledge, motivation, and organizational influences within leadership development: a study of a business unit in a prominent technology company

PDF

A tale of two principals: the complexity of fostering and achieving organizational improvement

Asset Metadata

Creator Lee, Seungyoon (author)

Core Title The coevolution of multimodal, multiplex, and multilevel organizational networks in development communities

School Annenberg School for Communication

Degree Doctor of Philosophy

Degree Program Communication

Publication Date 07/11/2008

Defense Date 06/18/2008

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag community ecology,evolutionary theory,information and communication technology for development,knowledge sharing,multidimensional networks,multilevel,multimodal,multiplex,network dynamics,network effects,network structure,niche,OAI-PMH Harvest,organizational networks,population ecology,replicability,resource,scalability

Language English

Advisor Monge, Peter R. (committee chair), Bar, Francois (committee member), Valente, Thomas W. (committee member)

Creator Email yoonlee@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1336

Unique identifier UC1302306

Identifier etd-Lee-20080711 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-87196 (legacy record id),usctheses-m1336 (legacy record id)

Legacy Identifier etd-Lee-20080711.pdf

Dmrecord 87196

Document Type Dissertation

Rights Lee, Seungyoon

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu