University of Southern California Dissertations and Theses

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

(USC Thesis Other)

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

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

A MULTITHEORETICAL MULTILEVEL EXPLICATION OF
CROWD-ENABLED ORGANIZATIONS:
EXPLORATION/EXPLOITATION, SOCIAL CAPITAL,
SIGNALING, AND HOMOPHILY AS DETERMINANTS OF
ASSOCIATIVE MECHANISMS IN DONATION-BASED CROWDFUNDING

by

Zhiming Xu

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

August 2020

Copyright 2020 Zhiming Xu

ii
Acknowledgements

I wish to express great appreciation to my dissertation committee for their time and effort
devoted to reading drafts of this project. Special thanks go to Dr. Janet Fulk for her role as
committee chair, who provided detailed, timely, and constructive feedback from the
dissertation’s early stages until its completion. Over the years, Dr. Fulk has exerted positive
influence on my career choice, pedagogy, and, inevitably, my language style. I heartily thank her
for her invaluable help.
To write a dissertation on networks, I also owe a debt of gratitude to Dr. Peter Monge, the
first person I spoke to at USC-Annenberg. I wish to thank him for his continuing encouragement
and straight-forward criticisms, both of which contributed to an improved dissertation and are
beneficial to my professional development. My gratitude also goes to Dr. Matt Sargent from
RAND Corporation, who served as an external member. His erudition in global history and
economy broadened my vision and motivated me to think outside the box.
The dissertation was completed, defended, and submitted in the first half of 2020. Many
years later, elementary schoolers in their history class are to study this period for quite some time
in order to understand what happened to their ancestors and to humanity. Hoping that there
would be explanations that make sense to them by then, I wish to thank my family, friends, and
mentors for their company in these confusing times.

iii
Table of Contents

Acknowledgements ii
List of Tables v
List of Figures vi
List of Abbreviations vii
Abstract viii
Chapter 1: Introduction 1
Chapter 2: Conceptual Framework 9
2.1. Crowd, Network, and Organization 9
2.1.1. From “Collective” to “Connective” 9
2.1.2. The Network Forms of Crowd 13
2.1.3. Theoretical Positioning of the Dissertation 15
2.2. An Empirical Example: Networks in Donation-based Crowdfunding 17
2.2.1. Introduction to Donation-based Crowdfunding 17
2.2.2. Multidimensional Networks in Donation-based Crowdfunding 19
2.2.3. A Multitheoretical Multilevel Approach of Hypothesis Development 24
Exploration/exploitation Tradeoff 26
Social Capital 30
Signaling 36
Homophily 38
Chapter 3: Method 44
3.1. Data Retrieval and Measurement 44
3.2. Network Construction 47
3.3. Analytical Procedures 49
Chapter 4: Results 52
4.1. Hypothesis Testing from H1 to H4 52
4.2. Hypothesis Testing from H5 to H8 57
4.3. Evaluation of Goodness-of-fit 61
Chapter 5: Discussion and Conclusion 68
5.1. Summary of Findings 68
5.1.1. Fundraising Exploitation 72
5.1.2. Project Signaling 74
5.1.3. Donor Homophily 76
5.1.4. Notes on Unsupported Hypotheses 78

iv
5.2. General Thoughts on Crowd Theorizing 80
5.3. Limitations 81
5.4. Future Research 84
5.5. Conclusion 87
References 88

v
List of Tables

Table 1. A MTML Explication of Typical Structural Properties 23
Table 2. Results from Multilevel ERGMs (H1 to H4) 63
Table 3. Results from Multilevel ERGMs (H5 to H8) 65
Table 4. Multilevel ERGMs Goodness-of-fit (H1 to H4) 67
Table 5. Multilevel ERGMs Goodness-of-fit (H5 to H8) 67
Table 6. Summary of Findings 70

vi
List of Figures

Figure 1. Theoretical Positioning of the Dissertation 17
Figure 2. Multidimensional and Multilevel Crowdfunding Networks 49
Figure 3. Network Configuration for H1 (TriangleXAX) 54
Figure 4. Network Configuration for H2 (X2StarADifference) 55
Figure 5. Network Configuration for H3 (XEdgeB) 56
Figure 6. Network Configuration for H4 (XEdgeB) 56
Figure 7. Network Configuration for H5 (XEdgeA) 58
Figure 8. Network Configuration for H6 (XEdgeA) 59
Figure 9. Network Configuration for H7 (TriangleXAX) 60
Figure 10. Network Configuration for H8 (X2StarADifference) 61

vii
List of Abbreviations

CDC Centers for Disease Control and Prevention
COVID-19 Coronavirus Disease 2019
DyNAM Dynamic Network Actor Models
ERGM Exponential Random Graph Models
eWOM Electronic Word-of-mouth
GOF Goodness-of-fit
FCC Federal Communications Commission
HR Human Resources
ICTs Information and Communication Technologies
MCMCMLE Markov Chain Monte Carlo Maximum Likelihood Estimation
MMORPG Massively Multiplayer Online Role-Playing Game
MTML Multitheoretical Multilevel
Multilevel ERGMs Exponential Random Graph Models for Multilevel Networks
REM Relational Event Models
SAOM Stochastic Actor-Oriented Models
TERGM Temporal Exponential Random Graph Models
WoC Wisdom of Crowds

viii
Abstract

A growing body of research has examined how emergent technological artifacts influence human
communication in crowd-enabled organizations, yet questions have also arisen regarding how
humans and technologies are conditionally connected to enable organizing, networking, and the
achievement of collective goals. To develop a comprehensive research agenda, this dissertation
examines donation-based crowdfunding as an example of crowd-enabled organizations, in which
human users and technological artifacts co-exist, co-function, and co-evolve to facilitate coherent
organizing, effective networking, and successful fundraising. The theorizing of crowds as
organization emphasizes the formation and impact of multidimensional and multilevel networks
that connect multiple types of nodes on various occasions to form diverse network structures,
which are associated with different crowdfunding processes and outcomes. In light of the large
scale and high complexity of these multidimensional crowdfunding networks, a multitheoretical
multilevel (MTML) approach to hypothesis development is adopted to provide theoretical and
analytical advantages. Theoretically, four empirically-driven perspectives
(exploration/exploitation tradeoff, social capital, signaling, and homophily) are employed to shed
light on the associative mechanisms that give rise to the typical local network configurations in
donation-based crowdfunding. Analytically, both individual and collective levels (i.e., project) of
agents are incorporated in the multilevel models to capture the global network structure of
crowdfunding. Data from a major donation-based crowdfunding platform were obtained and
analyzed to explicate the associative mechanisms leading to the formation of crowdfunding
networks, as well as the contingencies of these mechanisms. Exponential random graph models
for multilevel networks (multilevel ERGMs) were used to test eight hypotheses that predicted the

ix
emergence of specific network configurations corresponding to certain processes in fundraising
and donating behaviors. Results showed that a number of salient networking patterns, including
fundraising exploitation, project signaling, and donor homophily, were associated with the
hypothesized network configurations as estimated in multilevel ERGMs. Apart from adding
empirical evidence to crowdfunding research, the dissertation has important theoretical and
practical implications for fundraising management and online organizing in crowd-based
contexts. Overall, it suggests that in order to enable functional and self-sustaining crowd-enabled
organizations, it is critical to understand both the overarching and the microscopic coordinating
processes as represented in the global and local multidimensional network structures. Scholars
and practitioners may be further informed by the nuanced multilevel associative mechanisms in
crowdfunding networks, which form heterogeneous connections, determine the course of action,
and catalyze the emergence of the “connected crowd” and “organization in the crowd”.

Keywords: crowd, crowdfunding, donation-based crowdfunding, crowd-enabled organizations,
exploration/exploitation tradeoff, social capital, signaling, homophily, networks,
multidimensional networks, multilevel networks, MTML

1
Chapter 1: Introduction

“There has been a particular way of looking at crowds at every period of history”
(McClelland, 2010, p.1). Although the rudimentary understanding of crowd psychology and
behavior was already revealed in the ancient Greek scholarship, the development of formal
theories did not start until the late 19
th
century (McClelland, 2010). Ever since, the evolving
conception of crowd has been a distinctive component of the literature on collective behavior and
has motivated generations of scholars to craft disparate arguments about the principles of mass
social gatherings (McPhail, 1989). Among the pioneering crowd theorists, Gustave Le Bon was
known for his influential piece Psychologie des Foules (1897), in which he associated crowds
with negative traits such as impulsiveness, incapability to reason, and the absence of critical
spirit. As he put it, “[c]ivilisations as yet have only been created and directed by a small
intellectual aristocracy, never by crowds” (Le Bon, 1897, p. xviii). Following Le Bon’s tradition,
the so-called “madding crowd” hypothesis (see McPhail, 2017) postulates that crowds have the
potential to “de-individuate” people, such that conscious and innocuous individuals could be
transformed into irrational and dangerous mobsters when coalesced into crowds. At large, this
hypothesis based itself on the reasoning that crowds would enable the reduction of the members’
inner constraints, and individuals could therefore indulge in behavior from which they are
usually refrained (e.g., Festinger, Pepitone, & Newcomb, 1952). In line with this pessimistic
view of crowds, a steady stream of research in the 20
th
century (e.g., Allport, 1924; Blumer,
1969; Canetti, 1984; Horkheimer & Adorno, 1944/1972; Martin, 1920; Moscovici, 1985; Park,
1972; Tarde, 1903; Tilly, Tilly, & Tilly, 1975) examined crowds from copious yet vastly
different theoretical perspectives. This line of research provided important contributions to the

2
understanding of how identity, norms, behavioral intent, interpersonal relationships, group
dynamics, information flow and exchange, as well as power structure are affected when
individuals become a crowd. Despite the different foci in these foundational works, however, a
notable amount of classic crowd research happens to incorporate a somewhat paradoxical
question: if people are born to pursue their diverse interests, when will they begin to act in
concert in a crowd (McPhail, 2017)?
Whereas this long-existing question still holds its relevance in contemporary contexts, the
derogatory connotations associated with the term “crowd” seem to start dissolving to some
extent. Neologisms such as “crowdfunding”, “crowdsourcing”, “wisdom of crowds”
(Surowiecki, 2005), and “crowd work” (e.g., Kittur et al., 2013) have provided an alternative yet
increasingly prevalent narrative to viewing crowds as merely disorganized and disruptive mobs.
With numerous forms of modern crowds starting to “act in concert” by undertaking diverse,
meaningful, and consequential tasks, the society has witnessed the elevated potentials for
positively acting crowds to produce the common good and bring about social change (Bennett &
Segerberg, 2012; Howe, 2006; Mollick, 2014). Specifically, extensive media coverage and
scholarly attention have been laid on a variety of crowd-enabled activities, which may provide an
updated image of the crowds of our time. Examples include the pro-democracy uprisings in the
Arab Spring (Tufekci & Wilson, 2012), the grassroots fundraising efforts for the Flint water
crisis in the United States (Garcia, 2016), the online knowledge sharing and online collaboration
on question-and-answer websites (Bighash, Oh, Fulk, & Monge, 2018), the novel artistic designs
emerging from crowd competition (Brabham, 2010), and, more recently, the various global
initiatives to rely on crowds to fight the coronavirus outbreaks (Kulish, 2020; Lefkowitz, 2020).
Some scholars believe that crowds may be considered a signature characteristic of the society

3
and the changing communication landscape, as crowds of today could provide opportunities to
enable new modes of information processing and collective thinking, and to facilitate organizing
and remote coordination both online and offline (Bennett, Segerberg, & Walker, 2014; Stohl,
2014).
A number of researchers have associated the potentials and capabilities of modern crowds
with the ubiquitous networks whereby the distribution of information and connection of
individuals are made possible (Bimber, Flanagin, & Stohl, 2005; Castells, 2011; Easley &
Kleinberg, 2010). Furthermore, these networks are believed to be a crucial factor that
differentiates the atomized and disordered “LeBon-ian crowds” from the “connected crowds”
(Starbird, 2012) in which messages can be transmitted via networks instantly, and individuals,
tasks, and resources can also be efficaciously linked and organized. In the connected crowds, as
emergent technologies are employed to enhance the effectiveness of communication, the
distinctions between face-to-face and virtual communication have been blurred; consequently,
mass and remote coordination may be achieved (Stohl, 2014). In light of the coordination and
organizing that take place, the connected crowds have also been referred as “crowd-enabled
organizations” (e.g., Agarwal et al., 2014), which distinguish themselves from traditional forms
of organizations by fluid structure/membership and personalized information exchange.
Increasingly, the rise of crowd-enabled organizations is believed to have initiated a wave of
organizational evolution in a wide array of political (e.g., Bennett, Segerberg, & Walker, 2014),
business (e.g., Kittur et al., 2013), civic (e.g., Davies, 2015), and other domains.
Despite the prevalence of crowd-enabled organizations, the concept itself still requires
further theoretical elucidation and empirical investigation (Della Porta, 2014). Particularly, the
promise of applying this theoretical framework is accompanied by the challenges to the

4
theorization of crowds as organization. A major challenge in conceptualizing crowd-enabled
organizations lies in disentangling the interactions between human actors and nonhuman
resources that constitute and complicate modern crowds. This challenge is also represented in the
difficulty of determining the overarching and microscopic coordinating process through which
large-scale crowds achieve coherent organizing (Bennett, Segerberg, & Walker, 2014). As
Mueller (2010) noted, crowd-driven collectivities are usually unbounded compared to formal
organizations composed of fixed structural and relational constituents. This means that
understanding the “organization in the crowd” phenomenon (Bennett et al., 2014) requires a
unique and comprehensive analytical framework that can encompass and account for a variety of
volatile crowd-organizing mechanisms in action. Overall, these mechanisms tackle how
technological platforms facilitate the dynamic integration of the localized “dense, fined-grained
networks of individuals” (Bennett & Segerberg, 2013, p.13) to achieve massive coordination in
crowds at scale.
Another challenge has to do with the multilevel nature of crowds, in which networks at
individual, team, group, project, organizational, and societal levels play a critical role in
activating, structuring, and maintaining a successful crowd-enabled organization (Agarwal et al.,
2014). These multilevel networks are different from traditional social networks, as multiple types
of nodes, relations, and attributes are present (see Contractor, Monge, & Leonardi, 2011). Hence,
other than connecting two humans, it would also be viable to incorporate sociotechnical
connections by drawing ties between humans and artifacts. These sociotechnical ties are
important for the understanding of how technology use influences social connections, and vice
versa. Furthermore, in these networks, artifacts can also be connected to reflect the informational
and technological infrastructures through which devices and tools are linked. As the empirical

5
analyses of the dissertation show, connections of artifacts provide new theoretical possibilities to
conceptualize how human participation in crowds is managed. Overall, these multilevel networks
may provide very useful information on how crowd-enabled organizations orchestrate common
goals by allowing technology-equipped individuals to act synchronously (Bennett & Segerberg,
2013). However, to date empirical work that theorizes and analyzes these multilevel crowd
networks is still meager in the research literature.
Lastly, it appears that a more balanced and neutral view of the capabilities of crowd
organizations should be adopted, in an attempt to avoid both the extreme pessimistic and
optimistic tendencies that have emerged. Adopting a balanced view means that individual actors
in crowds should not been conceptualized as either atomized (as some pioneering crowd theorists
would claim) or connected (as some technological determinists would claim), whose connections
and actions should be instead viewed as conditional on a variety of factors. As a result, the
theoretical framework of crowd-enabled organizations may benefit from a more nuanced
depiction of the conditionalities under which organizing is facilitated or disrupted. Thus,
empirical studies that draw from real-life crowd network data would be particularly helpful for
reducing the conceptual ambiguity and improving the explanatory power of conceptualizing
crowds as organization.
Based on above, the ultimate goal of this dissertation is to bridge the gaps between the
prevalence of crowd-enabled organizations and the promising yet nascent theorizing of this
phenomenon. To identify the coordinating processes and shed light on the complex relationships
among humans and technologies as components of crowds, the dissertation develops a
corresponding theoretical framework and provides empirical evidence to explicate when and
how different associative mechanisms facilitate the connections of these crowd components. As

6
a central concept throughout the dissertation, associative mechanisms are defined as the
processes by which humans and nonhumans are conditionally connected to enable and sustain
the functioning of crowds, which grapple with “myriad inputs, demands, and events” to achieve
coherent organizing (Bennett et al., 2014, p.234). Put alternatively, associative mechanisms deal
with the structural forces that drive network formation and change in crowds, which explain the
conditions under which unconnected nodes become connected. Perhaps one of the most
important reasons why these associative mechanisms matter is that they determine how
individuals in crowds form relationships, how teams and social groups are assembled, how tasks
are assigned, and resources allocated, as well as how technologies are embedded to facilitate
connection, collaboration, and innovation. Also described as the “stitching mechanisms”
(Arvidsson et al., 2015; Bennett et al., 2014), the associative mechanisms fabricate and suture
smaller, local connections into large crowd-enabled networks (Della Porta, 2014).
Regardless of sharing similar characteristics, various forms of crowd-enabled
organizations are not a monolithic totality. As crowds are assembled for different purposes, by
different organizers, and in different contexts, inconsistencies are to be found across, for
example, crowdsourcing, crowdfunding, crowd-computing, crowd-based activism campaigns,
and so forth. To provide a generalized overview as well as more specific descriptions of crowd-
enabled organizations, the dissertation first delineates the principles of crowd organizing and
then evaluates if and how these principles apply in a particular type of crowd. The type of crowd
that this dissertation empirically investigates is donation-based crowdfunding, a prevailing
fundraising model that allows a large number of donors to provide monetary contributions
without getting any form of return (Belleflamme, Lambert, & Schwienbacher, 2014). As an
innovative fundraising strategy and a widespread social practice, donation-based crowdfunding

7
has become an alternative source of funding for a variety of personal and public purposes (Xu,
2018). Digital trace data obtained from a major donation-based crowdfunding platform were
used in the dissertation, which recorded fundraising and donation networks as well as various
attributes in these networks. The analyses employing exponential random graph models for
multilevel networks (multilevel ERGMs, see Wang, Robins, Pattison, & Lazega, 2013) were
performed to demonstrate how relationships among fundraisers, donors, and projects are formed
in donation-based crowdfunding. Results from these analyses suggested that networks may serve
as a critical organizing tool to structure the conditional connections between humans and
technological artifacts in crowds, thus providing empirical support to expounding the
theorization of crowd-enabled organizations. The findings may inform scholars and practitioners
by showing the contingencies of crowd organizing and networks, in order for them to
(re-)consider both the promise and challenges that modern crowds have created. Particularly,
these findings may be conducive to establishing a balanced and neutral view of crowds.
The rest of the dissertation is organized as follows. Chapter 2 elaborates the conceptual
framework that guides this dissertation. The first part of the chapter is dedicated to connecting
the concepts of crowd, network, and organization, the three underpinnings of this dissertation.
Special regard is paid to delineating the academic trajectory from “collective action” to
“connective action” (Bennett & Segerberg, 2012) as related to how modern crowds take the form
of networks in organizing. Chapter 2 also explicates the structures and formation of
multidimensional networks in crowds, with particular focus on the theoretical references and
empirical implications of the multimodality and multiplexity in multidimensional networks. The
last part of Chapter 2 centers on multidimensional networks in donation-based crowdfunding.
Employing a multitheoretical, multilevel (MTML) approach (Monge & Contractor, 2003) to

8
hypothesis development, it incorporates multiple empirically-driven theoretical perspectives,
including exploration/exploitation tradeoff (e.g., March, 1991), social capital (e.g., Coleman,
1988), signaling theory (e.g., Spence, 1978), and homophily (e.g., McPherson, 2001) to develop
eight hypotheses pertaining to the associative mechanisms in donation-based crowdfunding.
Chapter 3 introduces the methodology of the dissertation, including details about data retrieval,
measurement, and analytical procedures. The chapter provides details on the multidimensional
and multilevel crowdfunding networks, as well as how ERGMs for these networks were
implemented. Chapter 4 provides results from hypothesis testing and goodness-of-fit statistics of
the models. The dissertation concludes with Chapter 5, which discusses network patterns
emerging from the results, including fundraising exploitation, project signaling, and donor
homophily. Furthermore, the chapter also covers the implications for crowd organizing and
networks, theoretical and methodological limitations, topics for future research, and the general
conclusion on how this dissertation contributes to crowd research.

9

Chapter 2: Conceptual Framework

2.1. Crowd, Network, and Organization
2.1.1. From “Collective” to “Connective”
Crowds are an important component of human history and everyday life. While the term
“crowd” may be more readily tied to a varied array of mobs, riots, and carnivals, in fact it has
been used in history to describe an incredibly wide spectrum of collectivities and forms of
behavior from the unruly to the organized (Holton, 1978). When urban overthrowing forces
loomed as matter of deep concern in the wake of the French Revolution, conservative and elitist
intellectuals (e.g., Le Bon, 1897; Tarde, 1903) tended to regard all types of crowd as a hostile
symbol for the irrationality of the street mobsters. This intellectual tendency had its long-lasting
effects on histories and political thoughts especially during the world wars, and it was not until
the 1960s (e.g., Canetti, 1960) that more explicit and elaborate theoretical interest started to
emerge (McClelland, 2010). The changing conceptions of crowd mind and behavior were
introduced in psychology, sociology, political science, and other social science disciplines.
Especially in the second half of the 20
th
century, the term “crowd” was more often used in the
contexts of social protests and activist campaigns, in which crowd members were believed to
have some but limited capabilities to respond to ambiguous situations and to negotiate emergent
norms (e.g., Turner & Killian, 1957). At the time, crowds perhaps started to undertake a more
neutral connotation when scholars attempted to disassociate the word with its stereotypes. As
argued by McPhail (2017), instead of anti-social entities crowds should be understood as
temporary and mass gatherings, which provide opportunities for people to act collectively. In

10
line with this conception of crowd, more recent scholarship has viewed crowds are also a
quintessential example of collective behavior and action (Snow & Owens, 2013).
While research on collective behavior is often concerned with noninstitutionalized
gatherings and their influence (e.g., mass hysteria), collective action often takes place with a
shared interest among people (McPhail, 2017). Although the two terms are sometimes
interchangeably used, both collective behavior and collective action have been of long-standing
interest to social scientists. According to Sandler (2015), collective action “…arises when the
efforts of two or more individuals or agents (e.g., countries) are required to accomplish an
outcome” (p.196). Behind this rather broad definition is a wide array of aspects that scholars
have examined, leading to research topics such as factors guiding decision-making in political
and social movements (e.g., Melucci, 1996; Ostrom, 1998), public goods and social dilemma in
groups (Olson 1965), contagion and hypnotic influence in collective behavior (e.g., Le Bon,
1897; Zomeren, Postmes, & Spears, 2008), as well as interorganizational alliances and
information exchange (e.g., Monge et al., 1998). Some early theoretical approaches assumed that
collective action is often caused by an objective state of disadvantage (e.g., lack of material
resources), which motivates individuals to form and act in groups and organizations (Zomeren et
al., 2008). For decades, an important line of research of collective action has revolved around
how rationality affects individual decision-making and the collective outcomes (Gilbert, 2006).
In contrast with the earlier assumption that individuals sharing the common interests will
automatically act collectively, however, Olson (1965) maintained that individual rationality is
not always sufficient for collective rationality: as members may attach greater importance to self-
interests than group-interests, this tendency often ends up undermining the common entities.
Therefore, the question of cui bono (who benefits? see Blau & Scott, 1962) has been an

11
important concern for scholars to address how this benefit structure influences behavior in
organizations and the society. Following this research tradition, a prominent area of scholarship
has analyzed the formation and functions of interest groups in organizations and collective action
(Ganesh & Stohl, 2014). As Bimber and colleagues (2012) put it, this research agenda on interest
groups in collective action examines the “emerging goals and identities of organizations,
mobilizing people and resources, building alliances, and shaping ideologies and cultural frames
to support and sustain collective action” (p. 81). In contemporary forms of collective action (e.g.,
crowd-enabled social movements) where competing and ideologically opposed interest groups
are present, the research agenda on examining cui bono has become particularly relevant yet
increasingly complex.
Expedited by the advent of the world wide web, the development of information and
communication technologies (ICTs) since the late 20th century has exerted great impact on the
scale and forms of collective action. From e-movements to e-communities, the Internet has
facilitated the emergence of collective action scalable anywhere on the spectrum between
individuals and organizations (Dolata & Schrape, 2016). The connectivity enabled by advanced
technologies has enhanced people’s ability to cross communication boundaries in order to
interact with each other and engage with organizations (Bimber et al., 2012). The increased
interconnectedness and interdependencies in the society has motivated scholars to rethink the
assumption of individuals forming crowds as cognitively and socially isolated actors (e.g., Le
Bon, 1897). As Castells (2011) described, networks may have become a “key feature of social
morphology” (p. 5), which may have affected how individuals, goods, and markets function. In
collective action, remaining connected is to bear the consequences of one’s relative position

12
within the global network, and to disconnect from these flows is to bear the consequences of
isolation (Castells, 2011).
The simultaneous loosening of organizational boundaries and the thickening of
interpersonal connectedness in society have reconfigured the landscape of collective action
(Stohl, 2014). Bennett and Segerberg (2012) argue that large-scale crowd-driven actions
distinguish the familiar logic of collective action from the less familiar but the increasingly
manifest logic of “connective action”. The logic of connective action emerges when digital
media replace traditional organizational structures and mechanisms (e.g., bureaucracy, command
hierarchies) with more personalized content sharing and communicative activities that are either
self-organizing or weakly enabled by formal organizations (Bennett & Segerberg, 2012). As a
contemporary example of connective action, crowd-enabled organizations and networks are
formed via dispersed face-to-face and digitally assisted communication infrastructure. They are
often leaderless (or have distributed and/or collective leadership due to the decentralized
structures of digital media), but some crowd organizations can display coordination and
efficiency comparable to conventional organizational forms (Contractor, et al., 2012; Payne,
2012; Syrek, 2012). This structural characteristic is referred by some scholars as “organized
informality” (Berdou, 2011; Dobusch & Quack, 2011; Dolata & Schrape, 2016). The term aptly
describes how orders and rules are flexibly and dynamically established in crowds with little
implementation of the formal organizing strategies as in conventional collectivities. Hence, the
conceptual transition from collective action to connective action is pushed by the increasing
networking possibilities available to the crowds (Howe, 2006). As summarized by Bennett and
Segerberg (2012), the networks in connective action “are typically far more individualized and
technologically organized sets of processes that result in action without the requirement of

13
collective identity framing or the levels of organizational resources required to respond
effectively to opportunities” (p. 750).

2.1.2. The Network Forms of Crowd
Prior to the recent stream of research that theorizes crowds as organization (e.g., Agarwal
et al., 2014; Arvidsson et al., 2016; Bennett & Segerberg, 2012; Bennett et al., 2014, Dolata &
Schrape, 2016; Starbird, 2012), discussions were initiated on similar forms of boundaryless, non-
bureaucratic collective entities ranging from virtual organizations, in which humans and
technological artifacts are organized by electronic means (e.g., DeSanctis & Monge, 1999; Fulk
& DeSanctis, 1995) to network forms of organization (e.g., Poldony & Page, 1998; Powell,
1990). For the latter, however, the research has burgeoned as scholars are increasingly intrigued
by the plethora of emerging organizational forms since the 1980s that do not conform to the
conventional market-hierarchy continuum (e.g. Baker, Nohria, & Eccles; Granovetter, 1985;
Miles & Snow, 1986; Monge & Fulk, 1995; Poldony & Page, 1998; Powell, 1990). A network
form of organization (also known as “network organization” for short) is defined as “any
collection of actors (N ≥ 2) that pursue repeated, enduring exchange relations with one another
and, at the same time, lack a legitimate organizational authority to arbitrate and resolve disputes
that may arise during the exchange” (Podolny & Page, 1998, p. 59). According to Powell (1990),
network organizations are typified by “lateral or horizontal patterns of exchange, independent
flows of resources, and reciprocal lines of communication” (p. 296). Possessing their own
organizational structures and processes, network organizations have been shown to have distinct
advantages over pure free markets and strict hierarchies, such as effectiveness in learning and
economic efficiencies (Bradach & Ecceles, 1989; Uzzi, 1997).

14
Network organizations provide a pertinent explanatory framework to illustrate the
morphology and dynamics of crowd-enabled activities, as modern crowds may well take the
forms of networks to exist and function (Agarwal et al., 2014). To borrow the terminology from
political communication scholars, the dense networks connecting people and artifacts behave as
a “hybrid media system” (Chadwick, 2017), which plays a central role in allocating roles,
managing resources, promoting actions, and forming an organized entity (Starbird, 2012). In the
hybrid media system, individuals and technological artifacts are “… articulated by complex and
ever-evolving relationships based on adaptation and interdependence and concentrations and
diffusion of power” (Chadwick, 2017, p.xi). Therefore, the concepts of “network organization”
and “hybrid media system” converge upon the relational dynamic view of the emerging forms of
organizing. Likewise, the “organization in the crowd” phenomenon makes a quite similar
statement. According to Bennett et al. (2014), crowd organizing shares the fundamental
capacities with conventional bureaucratic organizations, including “resource mobilization,
responsiveness to external conditions, and long-time adaptation or change” (p.240). However,
because of the fuzzy boundaries, the fluidity of membership, and the decentralization and
deinstitutionalization of the structures, modern crowds may sometimes outperform conventional
organizations in promoting innovations and mass collaborations while limiting the negative
impact of dominance hierarchies and fixed practices. For instance, in Stack Overflow, a website
that features crowdsourced problem-solving related to computer programming, the median time
of getting a question answered was only 11 minutes (Mamykina et al., 2011), which can be
possibly attributed to the interest-driven practice and fluid membership that shape the crowds
into a dynamically and functionally connected network. Crowds have also been identified as
potential problem solvers for well-established organizations. In 2020, the Centers for Disease

15
Control and Prevention (CDC) in the United States has worked with the CDC Foundation to
launch an international crowdfunding campaign to combat the global COVID-19 (Coronavirus
Disease 2019) pandemic, and by the time the dissertation is completed, the campaign has
received over 49 million dollars of donations (CDC Foundation, 2020). Although these funds are
not an ultimate solution to the COVID-19 infection crisis, they may be used to expand and
accelerate public health response to the pandemic. These examples suggest the potentials for
networks forms of crowd to allow for more problem-focused organizing and the provision of
quicker responses than conventional forms organizations. In other words, networks may provide
unique organizing benefits, through which both information exchange and resource allocation
could become more efficient in modern crowds.

2.1.3. Theoretical Positioning of the Dissertation
A Venn diagram may be helpful to visualize how this dissertation is guided by the
preexisting literature. In Figure 1, three circles that represent crowd, organization, and network
divide the focal area into seven subfields. Subfields I, II, and III can be construed as traditional
and relatively independent theoretical areas in which the three subjects are studied. These three
areas might represent the primitive theorizing of the constructs, in which the theoretical
connections among crowd, organization, and network were largely neglected. For example,
despite its large volume, early research on crowd behavior tended to emphasize the structural
instabilities of crowds, with inadequate attention to the organizing and networking possibilities
of crowds; individuals forming a crowd were often naturally considered disorganized and
disruptive (Le Bon, 1897). Similarly, early organization and network studies tended to focus on
more structured collectivities, such as factories and schools, than crowds as an appropriate

16
subject matter. These theoretical tendencies were altered when scholars were motivated by
emergent organizational forms to consider the inherent relatedness of these concepts.
Intersectional concepts and theories such as “network organization” (e.g., Powell, 1990),
“connected crowd” (e.g., Starbird, 2012), and “organization in the crowd” (Bennett et al., 2014)
have been proposed in an attempt to account for the increased organizing capabilities and
networking possibilities in crowds. In the Venn diagram, they can be situated in subfields IV, V,
and VI to show the efforts in bridging the realms of crowd, organization, and network.
Ultimately, this dissertation is informed by the above perspectives and contends that crowd-
enabled organizations and networks cannot be fully understood if any of the three theoretical
underpinnings is missing. Subfield VII, which is the overlap of the three circles, suggests a
viable and comprehensive approach for contemporary research on this topic to draw on a mixture
of perspectives.

17
Figure 1. Theoretical Positioning of the Dissertation

2.2. An Empirical Example: Networks in Donation-based Crowdfunding

2.2.1. Introduction to Donation-based Crowdfunding
Roughly one out of four Americans has participated at least once in crowdfunding (Pew
Research Center, 2016). It is estimated that the exponential growth curve of this business will hit
$300 billion annual revenue by 2030 (Shepherd, 2020), surpassing venture capital and Angel
investment as one of the most prominent ways of financing individuals, organizations, and
communities (Barnett, 2016). As an alternative source of capital encouraged by the US Congress
(H.R. 4564, 2014), crowdfunding has become a rapidly growing area of entrepreneurial activity,

18
government action, philanthropy, and civic engagement (Mollick, 2014; Özdemir et al., 2015;
Stiver et al., 2015).
Regardless of sharing the same title, crowdfunding campaigns differ in their goals. A
widely used typology provided by Belleflamme and colleagues (2014) categorizes crowdfunding
projects as equity-, reward-, and donation-based. In equity- and reward-based projects,
contributors (also known as backers or funders) are compensated with financial or non-financial
benefits to repay their giving, whereas in donation-based projects the donors are detached from
any kind of benefits in return. From overcoming medical crises to rescuing homeless animals,
eliminating community problems to reallocating educational resources, donation-based
crowdfunding has covered an incredibly broad spectrum of topics. By 2017, although there were
more than 2,000 global crowdfunding platforms (Drake, 2017), platforms such as Gofundme,
Donate Kindly, Charidy, Watsi, and Crowdrise are known for their focus on donation-based
campaigns. In 2017, Facebook launched “Fundraisers”, an internal tool allowing users to raise
money for nonprofits, which may further expand the frontiers of this already pervasive social
phenomenon, given its over 2.5 billion monthly active users (Constine, 2020).
The social causes leading to the rise of donation-based crowdfunding are diverse. Davies
(2015) believed that a large number of campaigns result from the failure to fulfill public services
or the under-provision of public goods. Likewise, Wash and Solomon (2014) maintained that
many donation-based crowdfunding initiatives involve some public good component being
funded, such as education sectors (e.g., Lukk, Schneiderhan, & Soares, 2018) and journalism
(Jian & Usher, 2014). During community disruptions, such as the Flint, Michigan water crisis
started in 2014 in which effective governmental actions were delayed, crowdfunding played an
important supporting role in providing resources and services to residents (Garcia, 2016). In

19
addition to public goods, research has shown that crowdfunding has been used for private goods.
In a recent survey (Hellmann, 2020), 20 million Americans reported that they have used
crowdfunding to tackle the crippling burden of medical costs. Apart from medical care,
Kenworthy and colleagues (2020) also noted that crowdfunding has been increasingly used to
support medical research. Xu (2018) found that donation-based crowdfunding has been used to
raise money for pets and funerals, among other personal causes. With a wide array of charitable
purposes attained through crowdfunding, Özdemir et al. (2015) believe that donation-based
crowdfunding is changing the landscape of traditional philanthropy due to its participation mode
and extensive reach. Aligned with the models of individual charitable giving and social
entrepreneurship (i.e., entrepreneurship for social, cultural, and environmental missions, see
Dees, 2012), donation-based crowdfunding achieves old goals through new means (Frydrych et
al., 2014; Meer, 2014).

2.2.2. Multidimensional Networks in Donation-based Crowdfunding
A recent approach to studying the heterogeneous agents and relations in networks is to
conceptualize them as multidimensional and multilevel networks. According to Contractor,
Monge, and Leonardi (2011), multidimensional and multilevel networks are characterized by
multimodality and multiplexity. Multimodality describes the existence of multiple kinds of
objects in a network, which include not only humans but also technological artifacts;
multiplexity explains the heterogeneous relationships that the objects can have with each other.
Relationships can be formed between humans, between artifacts, and also between humans and
artifacts (Contractor, 2009; Contractor et al., 2011). For example, in a niche social media
community like Linkedin where users are gathered online for career-related purposes, nodes that

20
can exist in the multidimensional and multilevel networks include: a jobseeker, a human
resources (HR) director looking for job candidates, a job listing that includes the details for a
position, the company in which the HR director works, the university from which the job seeker
graduated, personal and organizational webpages containing information about the persons, the
companies, and the universities, as well as the numerous multifunctional buttons that the users
can click on to navigate the website, to name a few. The job seeker can follow a company’s
webpage, receive updates, and connect with the HR director who can also check the candidate’s
Linkedin webpage and send personal messages to obtain more information. In this example,
relations exist not only between the job candidate and the HR director but also between
numerous forms of technological artifacts that facilitate richer communication between
individuals. In other words, the example shows that the approach of multidimensional networks
makes technological artifacts endogenous to the networks. As noted by Contractor et al. (2011),
“…instead of asking how technologies might change social networks (or vice versa, or both), the
more appropriate question is, ‘What happens when a new technology becomes a part of a social
network?’” (p. 684). At first glance, adding multimodality and multiplexity into a network seems
to largely complicate the theoretical framework and empirical analysis, whereas in reality this
addition is necessary for restoring the network to what it is really like.
In the context of donation-based crowdfunding, a diverse spectrum of nodes is at play in
the multidimensional network, including humans that undertake different roles (fundraisers,
donors, or observers who can become fundraisers or donors) and an assemblage of nonhuman
artifacts (projects, technologies, narratives, texts, videos, pictures, hyperlinks, and money). These
nodes and the relations among the nodes form a network that may bring together and stabilize
actions that have the potential to call into being a crowd-enabled organization (Agarwal et al.,

21
2014). A person initiates a project, writes a story about an event, inserts a video and a few
pictures, shares it on social media where it is seen by the donors who are mobilized, donate
money, and share the story to their friends. All these actions can be represented in
multidimensional and multilevel crowdfunding networks, which showcase how crowd
organizations are constructed, maintained, or changed. These networks, in turn, become a pivotal
organizing tool to integrate the human and nonhuman components of crowds. Moreover, they
allow for the formation of a flexible and adaptable social structure in response to future
organizational events and individual actions in the crowds (Bennett et al., 2014).
Table 1 lists some of the typical structural properties in multidimensional networks.
These properties, which form relationships at the dyadic and triadic levels, contribute to the
formation of larger crowd-enabled networks (e.g., Bennett et al., 2014). To begin with, dyads can
be found between two humans when they develop a relationship. The relationship can be as
traditional as kinship and friendship, or it can also be some form of online interactions. Online
follower-followee relationships may be an example of the “virtual” connections that can exist
between two users. In that way, individuals do not have to be closely connected offline to have a
relationship. Their online connections can become the basis for the individuals to exchange
information and resources. A variety of theoretical perspectives may be applicable, including
homophily, contagion, and theory of social influence, which may account for the contingencies
for these interpersonal relationships to form. Multidimensional networks also allow for the
connections of technological artifacts, although existing theory does not explicitly stipulate the
conditions under which two artifacts can be connected by an edge. As an example, Contractor,
Monge, and Leonardi (2011) treated the extent to which two artifacts (software and hardware)
were compatible with each other as their basis to draw a tie between them. Although

22
technological compatibility is not always relevant across different contexts, it can be
extrapolated that two artifacts can be treated as having a relationship if they share certain
attributes. These attributes include but are not restricted to category, hashtag, or other online
taxonomy tools that seek to facilitate the classification and retrieval similar artifacts. Homophily
and institutional theory may be applicable to explain the reasons leading to artifacts’ shared
attributes. Furthermore, a dyad that connects a human and an artifact can represent an affiliative
connection (e.g., a user is a member of an online forum), or it can also represent an action
performed on an artifact by a human (e.g., a user donates money to a crowdfunding project). A
number of possible theoretical perspectives, including signaling theory (Spencer, 1973) and
homophily may be applicable.
Based on the above dyadic relationships, more complex triadic structural properties can
be introduced. Table 1 lists some of the typical and interpretable triadic network signatures.
These triads can be within-level, cross-level, or both (multilevel). The triads can also be fully
(closed) or partially connected (open). Depending on the research contexts, a variety of theories
may be applicable. These theories include but are not limited to cognitive consistency theories
(e.g., balance theory, dissonance theory), collective action theories (e.g., public good), and self-
interest theory (e.g., social capital, structural holes).

23
Table 1. A MTML Explication of Typical Structural Properties
Structural
Property
Level Agent Interpretation Theoretical
Possibility

Dyadic

Within Human Kinship; friendship;
direct communication
Homophily; contagion;
social influence

Within Artifact Shared attributes (e.g.,
category, hashtag)
Homophily; institutional
theory

Cross Human
& Artifact
Affiliation; performing
an action (e.g., donation)
Signaling; homophily
Triadic

Within Human Brokerage Balance/cognitive
dissonance; structural
holes; social capital

Within Human Closure Balance/cognitive
dissonance; structural
holes; social capital

Within Artifact Shared attributes (e.g.,
category, hashtag)
Homophily; institutional
theory

Cross Human
& Artifact
Unconnected individuals
share same affiliation or
performing same action
Collective action; public
good

Cross Human
& Artifact
Individual affiliated with
or acting on
heterophilous artifacts
Exploration/exploitation;
homophily

Multi Human
& Artifact
Individual affiliated with
or acting on
homophilous artifacts
Exploration/exploitation,
homophily

Multi Human
& Artifact
Connected individuals
share same affiliation or
performing same action
Collective action; public
good; homophily

24
2.2.3. A Multitheoretical Multilevel Approach to Hypothesis Development
Monge and Contractor (2003) made the observation regarding network literature that
most of the research only uses a single theory and often constrains the level of analysis to
individual, dyad, or the entire network level. Although more network scholars have started to
address this problem (e.g., Contractor et al., 2011), the volume of the research output has yet to
match the high speed that complex networks emerge in the society. This dissertation continues
the pursuit of a more systematic and comprehensive research framework in which the complexity
of the multidimensional networks can be adequately examined. Hence, a multitheoretical,
multilevel (MTML) approach (Monge & Contractor, 2003) is employed to bring together
relevant theoretical lenses that examine the particular aspects in which they are most applicable.
MTML is a framework that incorporates a range of local and global structural properties and
patterns, which enables comprehensive examination of the fine-tuned changes in organizations
and networks over time (Ognyanova & Monge, 2013). For example, Contractor and Monge
(2003) proposed a MTML framework to study the creation and dissolution of organizational
social networks. They described the theoretical advantages to examine both the exogeneous and
endogenous network mechanisms within the same analytical framework. Likewise, Ahmad,
Borbora, Srivastava, and Contractor (2010) used the MTML approach to derive models to make
link predictions in a Massively Multiplayer Online Role-Playing Game (MMORPG). The study
also showed the potential for the MTML approach to synthesize multiple theoretical aspects that
were otherwise scattered across different networks. Despite the different contexts in which the
framework has been applied, the MTML approach is particularly suitable to exploring the
network forms of organizing (Matei, 2006), which this dissertation explores.

25
The MTML framework employed in this dissertation uses four social theories and
perspectives, including the exploration/exploitation tradeoff (e.g., March, 1991), social capital
(e.g., Coleman, 1988), signaling theory (e.g., Spence, 1978), and homophily (e.g., McPherson,
2001). To overview, the perspective of the exploration/exploitation tradeoff is applied to
examine the adaptive processes of fundraisers finding (or failing to find) the optimal strategies
over time. Social capital theory is used to account for the impact of the network advantages from
existing and activated ties in crowdfunding processes. Signaling theory is incorporated in the
framework to explain the effects of communication modality (e.g., video, picture) cues that the
fundraisers use to signal and inform the donors that their donations are worthwhile. Homophily
theory is applied to describe the aggregative donating patterns occurring during the fundraising
campaigns. The four theories are selected because they cover the sender
(exploration/exploitation, social capital), content (signaling), and receiver (homophily) aspects of
crowdfunding communication, exploring both fundraising and donation activities. Furthermore,
the selected theories address important questions in the crowdfunding literature from a network
perspective. These questions include but are not limited to how fundraisers initiate campaigns,
what message strategies they use to maximize donations, as well as how potential donors
respond to crowdfunding messages. Hence, the combination of the four theories may provide a
well-rounded picture of the crowdfunding processes while allowing for the use of appropriate
theoretical lenses to examine the corresponding structural properties and associative
mechanisms.

26
Exploration/exploitation Tradeoff
The exploration/exploitation tradeoff describes a phenomenon that in organizational
settings members often try to maintain a balance between the certainties of sticking to existing
strategies (exploitation) and the benefits from taking alternative opportunities that are potentially
more advantageous in the long run (exploration) (Levinthal & March, 1993). “Exploration
includes things captured by terms such as search, variation, risk taking, experimentation, play,
flexibility, discovery, innovation. Exploitation includes such things as refinement, choice,
production, efficiency, selection, implementation, execution” (March, 1991, p.71). Although the
perspective has been used to analyze the strategic learning processes in a variety of
organizational contexts (see the overviews by Gupta, Smith, & Shalley, 2006, as well as Lavie &
Rosenkopf, 2006), it is particularly helpful for delineating the organizing processes in
crowdfunding where fundraisers have to make decisions in regard to campaign launching. In this
context, the exploration/exploitation tradeoff is represented in the dilemma of maximizing
returns while reducing uncertainties associated with such factors as employing new and untried
strategies, making choices that offer only the possibility of long-term returns, and relying on
unacquainted donors.
As donation-based crowdfunding projects often require prompt responses (emergency,
crisis, etc), many fundraisers may consciously or unconsciously resort to the most convenient
method that requires least effort (Zipf, 1949/2016) and refrain from experimenting with
alternative ideas that diverge from existing routines and strategies. This phenomenon echoes
what March (1991) refers as the “vulnerability of exploration”, which indicates that returns from
exploration are less certain, more delayed and distant from the locus of action compared to
returns from exploitation. Levinthal and March (1993) provided a similar explanation that the

27
fruits of successful exploration are often public goods, whereas the risks and costs of exploration
are often private goods. Thus, short-term and incremental adaptive processes may improve
exploitation more rapidly than exploration. Likewise, raising money from unspecified groups of
donors is a highly uncertain process, as the crowdfunding system resembles a constantly shifting
heterogeneous assemblage that contains significantly greater fluidity. Given the sources of
uncertainty in the volatile networks such as the ever-changing body of donors or the numerous
ways of crafting a fundraising message, fundraisers may decide to avoid adding more complexity
to the already unpredictable situations by following known strategies, especially when under the
short-term pressure of fundraising.
Existing literature on crowdfunding has started to delve into the phenomenon of the
exploration/exploitation tradeoff. For example, Stanko and Henard (2017) conceptualize
exploration in crowdfunding as the number of alternative external sources (breadth) that a
fundraiser draws upon, whereas exploitation is viewed as the extent to which a fundraiser draws
intensively and repeatedly from external sources (depth). Findings from their research suggested
that crowdfunding may be an organizational niche that fosters exploitation, such that an average
fundraiser tends to seek efficiency and value appropriation rather than to pursue remote and
unclear advantages from exploring unconventional choices. For example, if a fundraiser has a
record of launching successful campaigns for paying medical bills, it is likely for this person to
exploit the returns by launching recurring campaigns of this type (see Proelss, Schweizer, &
Zhou, 2020). In fact, this exploitive tendency may be deeply rooted in the social structure that
warrants drawing intensively and repeatedly from existing resources. Horvát, Uparna, and Uzzi
(2015) found that recurring crowdfunding transactions are more likely to take place among
friends, who often provide more and faster monetary contributions than strangers. In other

28
words, the often-unconditional support from family and friends may be an important reason for a
fundraiser to exploit a fundraising category such as medical expenses, as recurring donations are
likely to be provided by close-knits contacts.
Apart from what existing research has revealed, there may be other scenarios in which a
fundraiser decides to choose to exploit a project category instead of creating projects in new
categories. For example, a person who is passionate about animal rescue projects is likely to
repetitively launch campaigns in the same category, even though the projects themselves are
different from each other. One possible explanation is that the consistent use of a category shows
the project initiator’s commitment to the particular type of initiative. Another explanation is that
by sticking with the same category, fundraisers can obtain and secure resources and retrieve
network benefits more easily, as the category may be filled with established social connections
from previous interactions (Liu, Yang, Wang, & Hahn, 2015). Based on all above, the first
hypothesis is raised to test the effect of exploitation on project initiation:

H1: Projects initiated by the same fundraisers increasingly exploit the same
categories.

Apart from strategies of leveraging project categories, another aspect of
exploration/exploitation tradeoff may be found in crowdfunding message design strategies,
which pertains to the principle of least effort. The principle of least effort was initially developed
by Zipf (1949/2016) in the field of human ecology. The principle has been used to analyze other
similar social phenomena. An example is Oh and Monge (2013), who examined how the
principle of least effort is represented in social tagging systems, which are designed to minimize

29
both the encoding and decoding efforts for web users. Similarly, Liu and Lang (2004) found that
the Internet was more frequently used than other library resources by college students because it
involved “the least effort”. The principle of least effort is also revealed in recent research on
message design in donation-based crowdfunding, which shows that the majority of campaigns
only employ the most rudimentary text-based descriptions to deliver fundraising messages (Xu,
2018), although multimedia modalities are more helpful in attracting donations (Mollick, 2014;
Wheat et al., 2013). However, because producing simple texts requires less time and energy than
creating multimedia content, a fundraiser may continue exploiting this default, widely used, but
sub-optimal strategy instead of exploring alternative possibilities. On the other hand, if an
individual chooses to explore more advanced and sophisticated fundraising strategies, such as
producing videos to communicate the fundraising ideas, the rewards may sustain the future
exploitation of these strategies. Xu (2018) noted that donation-based crowdfunding projects are
different from entrepreneurial projects in modality use patterns, and they are different in
objectives, expectations, motivations for participation, and rules. For example, in for-profit
crowdfunding communities such as Kickstarter, fundraising has to follow the “all-or-nothing”
rule that projects failing to reach preset goals cannot get any funds. On the contrary, in donation-
based crowdfunding the rule usually does not apply, and the goals only serve as technical
references for fundraisers to inform the public how much money is needed. Even if the goals are
not attained, fundraisers can get all the donations. On some platforms (e.g., Gofundme)
fundraisers will still be able to receive donations after the goals are met (see Gofundme, 2020).
Therefore, as some return is more or less guaranteed, donation-seeking fundraisers may possibly
use less effort instead of being sufficiently motivated to try alternatives. Therefore, an average
fundraiser may choose to exploit the cumulative advantages by employing the same modality use

30
patterns. These patterns can be evaluated by the extent to which a fundraiser is consistent in the
type of modalities used in crowdfunding messages. For example, if a fundraiser only uses text in
a project, the person is less likely to use other modalities such as videos and pictures in other
projects. Similarly, if a fundraiser uses a full range of multimedia modalities, the person is likely
to maintain this pattern across different projects. Therefore, the second hypothesis is raised to
test is tendency:

H2: Projects initiated by the same fundraisers increasingly exploit similar
modality use patterns.

Social Capital
The concept of social capital and its derivative theories have been one of the most salient
research themes in social sciences (Lin, Cook, & Burt, 2001). Although the sporadic uses of the
term “social capital” can be traced back to more than a century ago (e.g., Hanifan, 1916), it did
not become widely used in books until late 1980s (Google Ngram, 2017). Among the scholars
who contributed to the popularization of social capital research are Bourdieu (1986), Coleman
(1988), and Putnam (1995), to name a few. Because of its interdisciplinary applications, the
definitions of social capital have been diverse. Whereas some of the definitions are similar, they
still express significant nuances on whether the focus is placed on the substance, the sources, or
the effects of social capital (Robison, Schmid, & Siles, 2002). Adopting a different taxonomy,
Alder and Kwon (2002) categorized the definitions as focusing on external social ties (also
known as bridging social capital) versus internal social ties (also known as bonding social
capital). An example of the former category would be Bourdieu (1986), who defined social

31
capital as “the aggregate of the actual or potential resources which are linked to possession of a
durable network of more or less institutionalized relationships of mutual acquaintance and
recognition” (p. 248). Bourdieu’s flexible definition is in contrast with Coleman’s (1990) more
restricted concept that emphasized the underlying social structures, which stated that social
capital is “a variety of entities with two elements in common: they all consist of some aspect of
social structure, and they facilitate certain actions of actors...within the structure” (p. 302). Some
definitions fall into the middle ground. For example, Nahapiet and Ghoshal (1998) provided an
integrated definition that social capital is “the sum of the actual and potential resources
embedded within, available through, derived from the network of relationships possessed by an
individual or social unit. Social capital thus comprises both the network and the assets that may
be mobilized through that network” (p. 243).
In existing research, bridging and bonding are considered two major subtypes of social
capital (Norris, 2002). Bridging social capital is often provided by loosely connected individuals,
such as friends of friends or acquaintances, who may provide useful information or novel ideas;
alternatively, bonding social capital exists among close-knit contacts such as immediate family
members and friends (Ellison et al., 2007). Based on specific research contexts, scholars have
also proposed new subtypes, such as maintained social capital (Ellison et al., 2007), linking
social capital (Aldrich, 2012), and consummatory and instrumental social capital (Portes, 1998).
However, these subtypes are not as widely used as the bridging/bonding distinction in research.
Network scholars have long been interested in the complex relationships between
different types of social capital and corresponding network attributes and structures. Attention
has been paid, for example, to how the strength of interpersonal ties influences the configuration
of social capital. Early attempts include Granovetter (1973), who, though he did not use the term

32
“social capital”, examined the utility of “weak ties”, which coincides with the idea of bridging
social capital. He argued that too much attention was laid on the “strong ties” that confines the
applicability of theory to small, well-defined groups, instead of the “weak ties” that possess
greater opportunities for mobility. Alternatively, Krackhardt (1992) argued that the strength-of-
weak-ties hypothesis has left some theorization and measurement issues, and that strong ties,
particularly the affective aspects they own, may become important when they are spread out
among the players. The strength-of-strong-ties argument also receives support from research
(e.g., Bian, 1997; Nelson, 1989), but regardless of the disagreement on whether the strong versus
weak ties matter more, tie strength is often considered an important factor that influences social
capital.
Another related stream of research on social capital has focused on the embeddedness of
actors in segments of networks. Granovetter (1985) suggested that market and hierarchical
relations are usually embedded in social relations, which, in turn, affect institutions and behavior.
Integrating the ideas of tie strength and embeddedness, Burt (1992) offered the conceptualization
of structural holes, which exist when there is a lack of connection between two agents bridged by
a broker; thus, an actor who is able to span structural holes is more likely to get non-redundant
information from different clusters. Brokerage and closure are two foundational mechanisms for
work on social capital (Burt, 2000). Networks with closure (clusters of strong ties) are argued to
be a source of social capital, because it affects access to information and facilitates sanctions that
make it less risky for the actors to trust one another (Burt, 2000; Coleman, 1990). On the other
hand, brokerage (structural holes) is also deemed a source of social capital because it separates
non-redundant sources of information and creates a competitive advantage for an actor whose
relationships span the holes.

33
With the development of the Internet, some scholars have examined the brokerage and
closure signatures in the virtual realms. For example, Shen, Monge, and Williams (2014) studied
the connections in an online game community and found whereas a player’s brokerage positively
predicted task performance, closure increased the level of trust. In a similar vein, Ganley and
Lampe (2009) demonstrated that newer users tended to connect with individuals in different
clusters (brokerage), whereas more involved individuals were found to tightly connect with one
another (closure). Overall, the assumptions of relationships between certain network
configurations and forms of social capital are complex but tightly associated. Nahapiet and
Ghoshal (1998) suggested considering the structural, relational and cognitive dimensions of
social capital as a way to better connect research of different traditions. While the structural
dimension of social capital may be a result of the properties of the social system and the network
of relations as a whole, the relational dimension focuses on the particular relationships between
people, such as friendship and personal influence (Granovetter, 1985; Nahapiet & Ghoshal,
1998; Tsai & Ghoshal, 1998). Furthermore, the cognitive dimension of social capital is embodied
in attributes, such as a shared code, that facilitate the common understanding among individuals
(Tasi & Ghoshal, 1998). This dimension echoes what Coleman (1988) described as “the public
goods aspect of social capital” (S. 116), which “…is an important resource for individuals and
may affect greatly their ability to act and their perceived quality of life” (p. S118). The taxonomy
proposed by Nahapiet and Ghoshal (1998), along with empirical findings that fall into each
dimension, illustrates that social capital is a mulitheoretical, multidimensional, and multilevel
concept. Hence, researchers must navigate different theories, dimensions and levels in order to
understand the complex dynamics of social capital.

34
Social capital is a concept that has been widely applied to explain how resources residing
in crowdfunding networks can be transformed into financial capital (Lu, 2014). According to
McKenny et al. (2017), who interviewed 54 out of 121 editorial review board members of
Entrepreneurship Theory and Practice, network theory and human/social capital theory were
identified as the most promising theories for future crowdfunding research. There is also no lack
of recent research that looks at crowdfunding dynamics with social capital theory. Some studies
show the direct effects of social capital on crowdfunding outcomes. For example, Butticè,
Colombo, and Wright (2017) found that social capital developed through incorporating digital
social links to other peer fundraisers on the crowdfunding platform makes fundraisers more
successful because the links can increase information exchange that is critical to the community.
This finding was supported by another study conducted by Skirnevskiy, Bendig, and Brettel
(2017). Other studies demonstrate the possible moderating effect of social capital. For example,
Giudici, Guerini, and Rossi-Lamastra (2017) showed that the effect of altruism on fundraising
success was moderated by the level of social capital, such that the positive effects of altruism can
be magnified by higher level of social capital. Furthermore, scholars have also differentiated
internal social capital from external social capital in crowdfunding research (Butticè et al., 2017).
External social capital is the network benefits obtained from family and friends, as well as
personal acquaintances including social media contacts (Agrawal, Catalini, & Goldfarb, 2011;
Borst et al., 2017; Mollick, 2014), whereas internal social capital is developed inside the
crowdfunding community through successful social interactions with other community members
(Butticè et al., 2017). What should be noticed is the above dichotomy uses crowdfunding
platforms as the reference point that anything outside the platforms is considered “external”,
which is different from traditional conceptions of viewing primary groups (strong ties) as

35
internal and secondary groups (weak ties) as external. According to the dichotomy,
crowdfunding platforms are not only sites where fundraisers’ social capital from off-platform
networks can be “cashed in”, they are also sites where new social relationships are formed
through increased involvement in the communities.
Existing research has already shown that both internal social capital (e.g., Butticè et al.,
2017) and external social capital (e.g., Borst et al., 2017) contribute to crowdfunding success, yet
little is known about the relationship between social capital and fundraising actions such as the
choice to initiate a campaign. In other words, although numerous studies have confirmed the
positive relationship between higher levels of social capital and crowdfunding performance, we
still need to develop an understanding of whether occupying advantageous network positions
motivate fundraisers to act purposively to leverage the resources they have. This insufficiency in
the crowdfunding literature echoes some classical social media scholars such as Coleman (1988)
and Lin (2002), who raised the important question about if and how structural properties in
networks propel individual actions. Therefore, H3 and H4 are proposed to establish a relationship
between external/internal social capital and purposive fundraising actions:

H3: Users with higher levels of external social capital are more likely to initiate a
project than users with lower external social capital.
H4: Users with higher levels of internal social capital are more likely to initiate a
project than users with lower external social capital.

36
Signaling
The formation of signaling theory in economics started in the context of labor markets,
where employers lack information about the quality of job candidates; therefore, the candidates
have to use a variety of signals to communicate their competence (Spence, 1973). One of the
most commonly used signals by job candidates is education credentials, which suggest that the
degree holders possess the ability of withstanding the rigors of higher education and are thus
more likely to succeed in their careers. However, although education credentials do not always
guarantee good performance of job candidates, they are still used by employers as a universal
standard to judge the quality of candidates due to a lack of better ways of getting that
information (Weiss, 1995). Hence, signaling theory is fundamentally concerned with reducing
information asymmetry between two parties: the message sender and the receiver (Spence,
2002). According to the theory, a signal serves as an implicit way of communicating the
information known to the sender but not to the receiver. Therefore, as noted by Connelly et al.
(2011), the influence of signaling theory “lies in ascribing costs to information acquisition
processes that resolve information asymmetries in a wide range of economic and social
phenomena.” (p. 42). Numerous studies have used signaling theory to understand different
organizational phenomena. For example, Kirmani and Rao (2001) found high-quality firms often
choose to signal their true quality whereas low-quality firms often refuse to leverage these
signals in their communication. This is explained by the perceived payoff associated with
signaling such that those who can get greater advantages of signaling are more motivated to do
so. Similarly, Lewis (2011) found that the voluntary disclosure of private information increases
the perceived quality and prices of used cars sold online, because the disclosure provided the
opportunity and the incentives to remedy information asymmetries between sellers and buyers.

37
Although most research on signaling theory includes quality as a central theoretical
component, it is defined and interpreted in many different ways (Connelly et al., 2011). For
example, quality has been used to refer to the unobserved ability of an individual that can be
signaled by education credentials (e.g., Spence, 1973) or the unobserved quality of an
organization that can be signaled by financial structures and managerial incentives (e.g., Ross,
1977). Such signals can be linked with other established indicators such as prestige and
reputation of the signaler (e.g., Certo, 2003).
In crowdfunding, there are great information asymmetries between the fundraisers and
donors. Because the campaigns are often immediately relevant to the fundraisers, they usually
have a much better understanding and clearer evaluation of the causes of the projects and the
needs for fundraising. The donors, on the other hand, often have to rely on the cues embedded in
fundraising project listings to judge if the donations are worthwhile, and if the fundraisers are
honest and ethical (Xu, 2018). Scholars have found that a variety of modality cues can signal
high quality of the campaigns and the preparedness of the fundraiser, such as the length of the
descriptions and number of updates provided by the fundraisders (e.g., Feldmann & Gimpel,
2016), the embeddedness of the visual cues (e.g., Courtney, Dutta, & Li, 2017; Mollick, 2014;
Wheat et al., 2013; Xu, 2018). Likewise, other reactive signaling cues were found to function
similarly. These reactive signals document the public reactions to the fundraising projects,
including comments (Mollick, 2014) and social media shares (Cordova, Dolci, & Gianfrate,
2015; Greenberg et al., 2013; Thies,Wessel, & Benlian, 2014), as well as endorsement cues such
as the “Likes” and “Favorites” (Courtney et al, 2017; Xu, 2018). Unlike the modality cues that
are embedded by the fundraisers themselves, reactive cues are provided by the public, which
may be equally if not more important to serve as effective signals to convince the donors. Thies

38
et al. (2014) maintained that the reactive “social buzz” that the fundraising projects generate may
be a form of electronic word-of-mouth (eWOM, see Chu & Kim, 2011). They further contended
that “…project quality from these signals leads to informational cascades, an information-based
explanation for herd behavior that occurs when individuals who face a certain decision choose to
follow the actions of others” (p. 2). Hence, to assess how modality and reactions signals affect
fundraising outcomes H5 and H6 are proposed:

H5: Projects that have more modality cues will receive more donations than projects that
have fewer modality cues.
H6: Projects that trigger more reactions will receive more donations than projects that
trigger fewer reactions.

Homophily
According to Rogers and Bhowmik (1970), homophily describes the tendency that “pairs
of individuals who interact are similar with respect with certain attributes” (p. 526). Famously
known by the expression, “birds of a feather flock together”, the homophily principle centers on
the notion that similarity breeds social connection (McPherson, Smith-Lovin, & Cook, 2001).
For decades, scholars (e.g., Aral, Muchnik, & Sundararajan, 2009; Centola, Willer, & Macy,
2005; Ibarra, 1992; Kossinets & Watts, 2009; McPherson et al., 2001) have examined a wide
range of demographic, cultural, geographic, and network factors as well as their effects on the
formation or dissolution of social ties. Early examples include Lazarsfeld and Merton (1954)
who studied the effects of homophily on friendship and distinguished two types of homophily,
status and value. Status homophily refers to the sociodemographic dimensions of personal

39
characteristics, such as race, ethnicity, sex, religion, occupation, and education. Value
homophily, on the other hand, is the collection of internal states such as attitudes, abilities,
beliefs, and aspirations, which are presumed to influence and guide behavior (McPherson et al.,
2001). In a similar vein, Kandel (1978) examined the influence of homophily on adolescent
relationship formation and dissolution as well as how continued association reinforced
homophily, showing that shared attributes and network ties may influence each other
longitudinally.
More recently, scholars have examined the effects of homophily in the context of social
networking sites. Halberstam and Knight (2016) used Twitter data to show homophily had
significant impact on information flow and production so that individuals might be
disproportionally exposed to like-minded information online. Also relying on Twitter data,
Dehghani and colleagues (2016) observed that linguistic and moral homophily were associated
with increasing likelihoods of forming online connections. Barnett and Benefield (2017)
examined homophily at the international level. They found that culturally homophilous counties
(countries that have shared border, language, civilization, etc) facilitated the formation of online
friendship.
As a concept that often co-occurs with homophily, heterophily describes the tendency
that “pairs of individuals who interact are different with regard to certain attributes” (Rogers &
Bhowmik, 1970, p. 526). While some scholars believe heterophily is the “mirror opposite” of
homophily (e.g., Lazarsfeld & Merton, 1954), others argue that these two mechanisms are
complementary (e.g., Barranco, Lozares, & Muntanyola-Saura, 2019; Lozares et al., 2014). Their
argument is largely based on the reasoning and evidence that individuals may choose to form
social relationships with those that have complementary qualities and competencies. Although

40
the principle of heterophily has received far less attention than homophily (Barranco et al.,
2019), this tendency has been shown in the literatures on business alliances (e.g., Chung et al.,
2000) and collaborative work (e.g., Uzzi, 2008), where scholars have explored the importance of
having heterophilous individuals and teams to enhance creativity and collaboration.
In the contexts of crowdfunding, scholars have started to explore if and how patterns of
homophily apply. For example, Greenberg and Mollick (2015) examined the effects of gender
match on fundraising outcome and found female backers disproportionately support women-led
projects in areas where women are historically underrepresented (e.g., technology and video
games). In a follow-up study (Greenberg & Mollick, 2017), the concept of homophily was
extended to include the shared structural barriers stemming from a common social identity.
Given that, the researchers believed that homophily provides a useful explanation gender effects
on fundraising outcomes. Similarly, Muller et al. (2014) found that a variety of “interpersonal
attributes-in-common” had a profound influence on collaborations in crowdfunding. Specifically,
they identified three homophily factors, including fundraiser and donor from the same country,
the same group, and the same company division. Lin and Viswanathan (2015) refer to this
tendency as “home bias”, which suggests that transactions are more likely to occur between
parties within the same boundaries. Apart from home bias, researchers (e.g., Zvilichovsky et al.,
2015) have also demonstrated the tendency for fundraisers to back projects that are similar to
their own projects in terms of category and size. In general, this line of research centers on how
the characteristics of donors that are shared with fundraisers influence the aggregative patterns in
donating behavior.
Alternatively, another stream of research is emerging in recent literature, which focuses
on the capacities for homophilous nonhumans to influence crowdfunding outcomes (e.g., Koch

41
& Siering, 2015; Kuppuswamy & Bayus, 2018; Xu, 2018). Instead of asking the classic question
that “Do birds of the same feather flock together?”, this line of research seems to be interested in
examining “Will houses that are constructed similarly attract the same people?”. Put
alternatively, this line of research tries to understand if technological artifacts that have the same
properties will be treated similarly. Therefore, homophily is not only a concept that deals with
the effects of human similarities and differences on relationships, it may also have a nonhuman
component that could influence organizational structures and outcomes (see Latour, 2005; Sayes,
2014). In crowdfunding, nonhuman homophily can be translated into how projects sharing
certain characteristics can trigger the same tie formation and decay mechanisms, such as
attracting the same set of donors. The effects of projects characteristics on crowdfunding
outcomes are extensively studied in literature. Notably, researchers have noticed that fundraising
outcomes are likely to vary simply by project category (e.g., Borst et al., 2018; Briggman, 2014;
Dey et al., 2017; Mollick, 2014). For example, Mollick (2014) reported that Kickstarter music
projects were almost twice more likely (59.7%) to reach fundraising goals than fashion projects
(31.0%). Furthermore, scholars have observed that category effects may interact with modality
effects such that in some categories certain modality configurations are more effective than in
others. Recent examples include Dey et al. (2017) who found in different categories of projects
that the use of videos produced uneven effects on fundraising outcomes. The differences are
attributed to the characteristics of different categories, such as popularity of the topics,
complexity of the messages, relevance to the donors, and urgency of the projects (Aprilia and
Wibowo, 2017; Berliner and Kenworthy, 2017; Mollick, 2014). Based on the literature, Xu
(2018) summarizes that there are the two major types of project characteristics: modality and

42
category. In line with Xu (2018), the following two hypotheses are proposed to test the effects of
modality and category homophily on the aggregative patterns of donors:

H7: A donor is more likely to contribute to projects that belong to the same category.
H8: A donor is more likely to contribute to projects that employ the same modality.

While H7 and H8 correspond to the theoretical perspective of homophily, there are
similarities between this set of hypotheses and exploration/exploitation hypotheses (H1 and H2).
While it is obvious that the two sets of hypotheses deal with fundraising and donation activates
respectively, it should also be noted that the theoretical mechanisms are very different despite
their surface similarity. The exploration/exploitation tradeoff was proposed to understand the
strategies that organizational members use to gain advantages (March, 1991). Like all other
strategic theories, the theory assumes that both exploration and exploitation are purposively
implemented (e.g., Stadler, Rajwani, & Karaba, 2014). On the contrary, homophily theory does
not necessarily assume intention, and, quite often, homophilous aggregations could take place as
natural tendencies instead of purposive actions (e.g., Lazarsfeld & Merton, 1956). This critical
dissimilarity between exploration/exploitation and homophily can also be found in how
fundraising and donation activities differ. For example, while a fundraiser has to stipulate the
category of a project every time s/he initiates one, a donor may contribute to two projects in the
same category without even being aware of it. Likewise, although a fundraiser needs to make
decisions in regard to what modality cues to use to craft the project descriptions, donors do not
need to use the same levels of elaboration to process these cues. Therefore, while it is more
theoretically sound to assume that fundraising is always an intended behavior, it would be less so

43
to assume that the aggregative donating patterns are always intended as well, as a variety of
reasons other than intention may explain homophily. Based on this distinction, it is important to
note that H7 and H8 tackle a quite different aspect of crowdfunding from what H1 and H2 do.

44
Chapter 3: Method

3.1. Data Retrieval and Measurement
The dissertation used data obtained from a major US-based crowdfunding platform. The
platform was launched in 2010, which allows its users to raise money for events such as
emergencies, medical crises, and other challenging circumstances. The platform also holds
community-based fundraising campaigns launched by individuals and nonprofit organizations.
During the fundraising process, the platform allows users to create webpages for their
fundraising cause. Users can also share these webpages via social media such as Facebook and
Twitter. Over the past years, the platform has adopted several important changes, including
canceling the platform fee (5%, charged before 2017), changing its project categorization system
for several times, and updating its terms of use and privacy policy (the platform did not indicate
that web crawling was forbidden in their terms of use until 2016).
The publicly available digital trace data were retrieved using a Python web crawler in
October 23, 2015. As the website was scraped at one time point, the data were cross-sectional. In
total, 10,566 crowdfunding projects were included in the sample. The numbers of donors
participating in these crowdfunding projects ranged from 0 to 7,158 (M = 59.78, Med = 20, SD =
209.60). The following measures were also obtained for each project: category, modality use,
and reaction. Besides, measures for the fundraisers’ internal and external social capital were also
retrieved. The detailed explanations for the above measures are as follows.
The projects in the sample belonged to 26 different categories, including “Accident”,
“Animal”, “Baby”, “Boston”, “Business”, “Carolina”, “Celebration”, “Community”, “Colorado”,
“Competition”, “Creative”, “Dream”, “Education”, “Funeral”, “Ferguson”, “Medical”,

45
“Mission”, “Nepal”, “Non-profit”, “Oklahoma”, “Miscellaneous”, “Philippine”, “Sandy”,
“Sports”, “Volunteer” and “Wedding”. These categories were stipulated by the platform to cover
general (e.g., “Accident”, “Animal”, “Baby”) and specific fundraising topics (e.g., the category
“Ferguson” refers to the unrest in Ferguson, Missouri after the fatal shooting of Michael Brown
by police officer Darren Wilson in August 2014; the category “Sandy” refers to crowdfunding
projects related to Hurricane Sandy in 2012). Although some the categories seem to overlap,
fundraisers could only select one category that best described their projects by the time the date
were collected.
In line with previous literature (e.g., Courtney, Dutta, & Li, 2017; Feldmann & Gimpel,
2016; Liang, Hu, Jiang, 2020; Mollick, 2014; Wheat et al., 2013; Xu, 2018), modality use was
operationalized by two sub-variables, the number of videos and the number of pictures in project
descriptions. They were entered in the models separately, as described in Chapter 4. Reaction
was operationalized by three sub-variables including the number of comments, the number of
“Favorites”, and the number of shares, which existing literature (Cordova, Dolci, & Gianfrate,
2015; Courtney et al, 2017; Greenberg et al., 2013; Xu, 2018) identified as the possible measures
of this variable. Likewise, these variables were not combined but were input in the models
separately.
To review, internal and external social capital are concepts developed by crowdfunding
researchers (e.g., Butticè et al., 2017) that seek to contextualize and adapt the traditional
taxonomies of social capital (e.g., bridging vs. bonding; brokerage vs. closure). In general,
internal social capital deals with the network advantages obtained from the crowdfunding
platform, and external social capital pertains to the network benefits rooted in family and friend.
As the concepts are relatively new, there is a lack of established measures for internal and

46
external social capital, although scholars have provided rationales for how they quantify these
two measures. For example, Butticè et al. (2017) used the number of Facebook friends at the
point of campaign launch as a proxy of external social capital. They justified the use of this
measure with three reasons: 1) its public availability, 2) the information that this measure
conveys about kinship and friendship important in the early stages of crowdfunding campaigns,
and 3) the consistency with previous research (e.g., Agrawal, Catalini, & Goldfarb, 2011;
Mollick, 2014). Similar patterns were seen with the operationalizations of internal social capital.
Colombo et al. (2015) measured internal social capital by “the number of Kickstarter projects
that a proponent had backed at the time of launching her own crowdfunding campaign” (p. 83).
Likewise, they argued that this measure, which was visible to the public, represented the degree
to which a crowdfunding user has established connections within the community. Obviously,
existing research tries to delimit the theoretical boundaries of the concepts while also attempting
to use measures that are publicly available. In line with existing research, the current dissertation
uses two measures that were publicly available on the studied platform to quantify internal and
external social capital. Specifically, a fundraiser’s external social capital in the dissertation was
measured by the amount of donations raised from social media sharing. This measure was
consistent with Butticè and colleagues’ (2017) use of social media contacts but also provided
further information about to what extent existing social ties were activated by the campaigns. In
a similar vein, the dissertation measured internal social capital by the number of projects that the
fundraisers initiated prior to the data collection. The underlying argument resembles the logic by
Colombo et al. (2015) that repeat participations may be helpful for establishing new connections
in the community.

47
3.2. Network Construction
A typical set of multilevel networks include Network A and Network B as within-level
networks and a bipartite Network X as the cross-level network (Wang et al., 2013). In the
dissertation, Network A is defined as the within-level individual network formed by the social
connections among the platform users. Unfortunately, the platform does not provide information
about how the individuals are directly connected with each other in the form of friendship or
follower-followee relationship on the platform, possibly due to privacy concerns (see Gonzales,
Kwon, Lynch, & Fritz, 2018). Network B is defined as the project-level network. An undirected
tie was drawn if the projects belonged to the same category as determined by the crowdfunding
platform. Lastly, two types of cross-level ties exist. If a human node represents a fundraiser, an
undirected tie was drawn if the person initiated a project; similarly, if a human node represents a
donor, an undirected tie was drawn if a donor contributed to a project.
Following the above logic, two multilevel networks, (A X B)1 and (A X B)2 were
constructed. Both networks, independent of each other, were sampled from the large dataset
consisting of 10,566 projects. (A X B)1, the fundraising network, consists of fundraisers and
projects. A threshold was set as 3, meaning that only fundraisers who launched three and more
projects were selected and included in network (A X B)1. Applying this threshold yielded a 59
(fundraisers) x 202 (projects) multilevel network. In the network, the number of projects a
fundraiser launched ranged from 3 to 10 (M = 3.12, Med = 3, SD = .45). Likewise, (A X B)2, the
donating network, applied the same threshold, selecting donors who contributed to at least 3
different projects, yielding an 81 (donors) x 266 (projects) multilevel network. In (A X B)2, the
number of projects a donor contributed to ranged from 3 to 8 (M = 3.04, Med = 3, SD = .21).

48
There were both theoretical and methodological reasons for applying the threshold to
select repeat fundraisers and donors for the subsequent analyses. Theoretically, three out of four
perspectives in the MTML framework directly assumed repeated actions in theorizing, including
exploration/exploitation tradeoff, social capital, and homophily. As March (1991) described in
his original theoretical piece, the exploration/exploitation tradeoff takes place as an adaptive
process, in which individuals are engaged in organizational learning through comparing the
value of repeating an action versus starting a new action. Meanwhile, social capital theorists also
argue that personal networks are built and reinforced by repeated human interactions. As
Fukuyama (2001) argued, “…if individuals interact with each other
repeatedly over time, they develop a stake in a reputation for honesty and reliability” (p. 16). In a
similar vein, homophily is fundamentally concerned with how network ties are being reinforced
(or weakened) through repeated interactions among people who are similar (or dissimilar) (see
McPherson et al., 2001). For this reason, it is important to select repeat fundraisers and donors in
order to test the hypotheses (H1, H2, H3, H4, H7, and H8) derived from the theories that assume
repeated actions.
On the other hand, setting the threshold as 3 to eliminate users that did not initiate/donate
to at least three projects was due to a methodological consideration.
This treatment has to do with the computational constraints associated with the current
algorithms and analytical tools. According to Stivala, Robins, and Lomi (2020), the largest
sample size for single-level ERGMs seen in the literature was 2,209 nodes in a study by Hunter,
Goodreau, and Handcock (2008). As multilevel models require exponentially more
computational power, all the published articles analyzing multilevel networks as of today had a

49
few hundred nodes at most. Therefore, by setting the threshold as 3, the trimmed networks (A X
B)1 and (A X B)2 could be handled by the current analytical programs.
Figure 2 summarizes the theoretical model of the two multilevel networks and what the
multiplex connections indicate. In the figure, round (red) nodes represent individuals (fundraisers
and donors); square (blue) nodes represent crowdfunding projects. Network A consists of within-
level individual connections, which are absent in the data. Network B consists of within-in level
project ties connecting those belonging to the same categories. Individuals can initiate projects,
as shown in Network (A X B)1, and they can donate to projects, as shown in Network (A X B)2.

Figure 2. Multidimensional and Multilevel Crowdfunding Networks

3.3. Analytical Procedures
As network data are interdependent among the observations, they violate the
independence assumption as stipulated in many standard statistical models. Therefore, analyzing
network data requires statistical models that take into account the relational nature of network
data. Exponential random graph models (ERGMs) are a family of statistical models for analyzing

50
relational network data, which can be used to compare the probability of generating a network tie
to what would occur based on random chance (Robins, Pattison, Kalish, & Lusher, 2007).
Besides, through simulation ERGMs allow the consideration of other alternative networks that
share similar network properties with the observed network. Currently, most of the ERGMs are
based on the algorithm of Markov chain Monte Carlo maximum likelihood estimation
(MCMCMLE) (Robins et al., 2007), although alternative algorithms have also been developed to
enhance the computational efficiency of ERGMs (e.g., Stivala et al., 2020). MCMCMLE is an
algorithm that simulates a distribution of random networks from a set of parameter values and
then redefines the estimated parameter values by comparing this distribution with the observed
network (Snijders, 2002; Zappa & Lomi, 2015). Therefore, ERGMs allow for the testing of the
proposed hypotheses that aim to examine the local network formation mechanisms.
An extension of the ERGMs was developed by Wang, Robins, and Matous (2013) for the
analysis of multilevel networks. The extension allows the modeling of three networks in a
complete set of multilevel networks. According to Wang et al. (2016), a multilevel ERGM can
be expressed as

Pr(𝐴 = 𝑎,𝑋 = 𝑥,𝐵 = 𝑏) =
1
𝑘(𝜃)
exp34
𝜃
!
𝑧
!
(𝑎)+𝜃
!
𝑧
!
(𝑥)+𝜃
!
𝑧
!
(𝑏)+
𝜃
!
𝑧
!
(𝑎,𝑥)+𝜃
!
𝑧
!
(𝑏,𝑥)+𝜃
!
𝑧
!
(𝑎,𝑥,𝑏)
8
!

In the equation, A, X, and B can be treated as network random variables at three different levels
of networks (A and B as within-level and X as cross-level), with their realization respectively
denoted by a, x, and b; Q defines a network configuration in which the tie variables are assumed
to be conditionally dependent. zQ is the count of the network configuration of Q, with θQ being

51
the estimated parameter of zQ. k(θ) is a normalizing constant defined based on the graph space of
networks of a given size and the actual model specification.
All hypotheses other than H1 and H7 include certain nodal attributes. Therefore, the
testing of these hypotheses requires the inclusion of nodal attributes as exogeneous covariates to
evaluate whether the emergence of the hypothesized network structures is affected by these
attributes. Based on the above equation, Wang et al. (2016) express a multilevel ERGM with
nodal attributes as

Pr(𝐴 = 𝑎,𝑋 = 𝑥,𝐵 = 𝑏 | 𝑌
"
= 𝑦
"
,𝑌
#
= 𝑦
$
)
=
1
𝑘(𝜃)
exp3<𝜃
!
𝑧
!
(𝑎,𝑥,𝑏)+𝜃
%
𝑧
%
(𝑎,𝑥,𝑏,𝑦
"
,𝑦
#
)=
!,%

In the equation, zQ (a, x, b) represents graph statistics involving only network tie variables as in
multilevel ERGMs. zΛ (a, x, b, y
A
, y
B
) are statistics involving interactions among tie variables
and attribute values. θQ and θΛ are the associated parameters for their corresponding network
statistics.
The analyses were performed using MPNet, a program developed for multilevel ERGMs
that is capable of modelling both one-mode and bipartite networks (Wang et al., 2013).
Specifically, MPNet was used to estimate specific ERGM parameters for the hypothesized
network configurations. Furthermore, it was used to test the goodness-of-fit by comparing the
simulated network statistics against the modeled networks. The network files were prepared
according to the requirements as detailed in the manual of MPNet as adjacency matrices.
Separate nodal attribute files were also prepared in order to enter the covariates into the models.

52
Chapter 4: Results

4.1. Hypothesis Testing from H1 to H4
H1, H2, H3, and H4 predicted the emergence of specific network mechanisms associated
with Network (A X B)1, where the cross-level ties indicate that individuals initiate fundraising
projects. To test the four hypotheses, four multilevel ERGMs were fitted (see Table 2). Across
Model 1 to Model 4, three structural tendencies were included as the baseline predictors. These
three tendencies include the within-level edge terms for cross-level Network X (XEdge) and
within-level Network B (EdgeB), as well as a cross-level configuration (XASB). The edge terms
were included in the models to function similarly as the “control variables” to represent the
simplest possible network models. As described by Butts (2019), always including an edge term
in any ERGM is an advised practice in order to be able to compare the observed network with
the general homogenous Bernoulli random graph as its reference. Specifically, XEdge can be
interpreted as the occurrence of a fundraiser initiating a project. Likewise, EdgeB can be
interpreted as the occurrence of two projects belonging to the same category. The cross-level star
configuration, XASB, can be interpreted as a fundraiser initiating multiple projects. These
structural tendencies were included in the models because they were associated with the overall
probability of the most common network configurations in the observed network. Other
unobserved network configurations were excluded from the models. For edge terms in ERGMs,
nonsignificant estimates indicate that edges are likely to form at random, whereas significant
estimates indicate otherwise. Across four models, the estimates for XEdge (Model 1 est. coeff. =
-9.21, SE =.49; Model 2 est. coeff. = -8.36, SE =.46; Model 3 est. coeff. = -8.92, SE =.37; Model
4 est. coeff. = -7.99, SE = .41) and Edge B (Model 1 est. coeff. = -3.34, SE =1.12; Model 2 est.

53
coeff. = -2.91, SE =.73; Model 3 est. coeff. = -5.45, SE =1.92; Model 4 est. coeff. = -3.05, SE =
1.00) were all statistically significant, meaning that these within-level and cross-level ties were
unlikely to form at random in both Network B and Network X.
Guided by the perspective of exploration/exploitation tradeoff, H1 stated that a fundraiser
is more likely to initiate projects that belong to the same category, which corresponds to the
network configuration (TriangleXAX) of a closed triad consisting of one fundraiser and two
projects. In this configuration, if a fundraiser initiates a project, a cross-level tie is drawn to
connect them; if the two projects belong to the same category, they are also connected by a
within-level tie. Hence, by including this network configuration in the model, the ERGM
evaluates the likelihood for this structural tendency to occur by chance in the multilevel network.
In Model 1, the TriangleXAX configuration had a parameter estimate of 1.11 (SE = .40).
According to Wang et al. (2016), multilevel ERGMs consider the absolute value of a parameter
estimate greater than twice the size of the estimated standard errors as the sign of statistical
significance. Therefore, it can be concluded that the likelihood of observing the pattern as
predicted in H1 was significantly different from the null positing random occurrence. H1 was
therefore supported, suggesting the tendency for fundraisers to limit the activities of category
spanning by launching multiple projects that belong to the same categories.

54
Figure 3. Network Configuration for H1 (TriangleXAX)

Note: This multilevel network configuration is documented in Wang et al. (2014). In the analysis
for H1, a red circle represents a fundraiser (human); a blue square represents a crowdfunding
project (artifact). A blue within-level tie that connects two crowdfunding projects indicates that
they belong to the same category. A black cross-level tie that connects a fundraiser and a
crowdfunding project indicates that the project is initiated by the fundraiser.

Likewise, H2 was built upon the exploration/exploitation tradeoff but was raised to test
the mechanism in another aspect of the fundraising message design strategy. It stated that
projects initiated by the same fundraiser would be more likely to have similar rather than
different modality use. As described in Chapter 3, modality use was operationalized by the
number of video and picture counts. As continuous nodal attributes, the differences of video and
picture counts can be calculated to evaluate if a fundraiser’s modality use patterns change across
different projects. This network configuration corresponds to an open triad (X2StarADifference)
consisting of a fundraiser and two projects. The differences of the nodal attributes (video and
picture counts) were incorporated in the network configuration as well. In Model 2, the
configuration of X2StarADifference had two parameter estimates respectively for video (est.
coeff. = .43, SE = .25) and picture use (est. coeff. = .05, SE = .14). Both parameters were not
significantly different from the null stating the patterns would occur by chance. In other words,

55
there was inadequate evidence to support that fundraisers regulate their strategies by sticking to
the same modality use patterns. Therefore, H2 was not supported.

Figure 4. Network Configuration for H2 (X2StarADifference)

Note: This cross-level network configuration is documented in Wang et al. (2014). A circle
represents a fundraiser (human); a square represents a crowdfunding project. A solid icon
represents a node whose attributes are incorporated; an outline icon represents the otherwise. In
the analysis for H2, the nodal attributes included were, respectively, video count and picture
count in the project description. A black cross-level tie that connects a fundraiser and a
crowdfunding project indicates that the project is initiated by the fundraiser. A negative sign
represents the numerical difference of a nodal attribute. A larger node represents a higher level of
a nodal attribute than a smaller node.

Both H3 and H4 were guided by social capital theory and the adaptation of the theory in
crowdfunding research (e.g., Butticè et al., 2017; Skirnevskiy, et al., 2017). Specifically, H3
predicted the positive association between external social capital and the likelihood to initiate a
crowdfunding project. The hypothesis corresponds to a network configuration (XEdge B) of a
dyadic relation between a fundraiser and a project. In this network configuration, a cross-level tie
indicates that a user initiates a fundraising project; internal and external social capital can be
treated as nodal attributes of a user. Hence, this configuration allows for the testing of H3 and H4
in ERGMs. The configuration was entered in Model 3, which produced a parameter estimate of

56
-.07 (SE = .04). The same configuration was entered in Model 4 with the attribute of external
social capital replaced by internal social capital, which had a parameter estimate of .001 (SE
= .003). Hence, both H3 and H4 were unsupported, suggesting that no statistically significant
relationship between external/internal social capital and fundraising activities was found.

Figure 5. Network Configuration for H3 (XEdgeB)

Note: This cross-level network configuration is documented in Wang et al. (2014). A circle
represents a fundraiser (human); a square represents a crowdfunding project. A solid icon
represents a node whose attributes are incorporated; an outline icon represents the otherwise. In
the analysis for H3, the nodal attribute included was external social capital of a fundraiser, as
measured by the donations that the fundraiser obtained via social media sharing. A black cross-
level tie that connects a fundraiser and a crowdfunding project indicates that the project is
initiated by the fundraiser.

Figure 6. Network Configuration for H4 (XEdgeB)

Note: This cross-level network configuration is documented in Wang et al. (2014). A circle
represents a fundraiser (human); a square represents a crowdfunding project. A solid icon
represents a node whose attributes are incorporated; an outline icon represents the otherwise. In
the analysis for H3, the nodal attribute included was internal social capital of a fundraiser, as
measured by number of projects that the fundraiser launched. A black cross-level tie that
connects a fundraiser and a crowdfunding project indicates that the project is initiated by the
fundraiser.

57

4.2. Hypothesis Testing from H5 to H8
Following the same logic, four additional models (Model 5 to Model 8) were fitted using
Network (A X B)2, the network consists of 81 donors and 266 crowdfunding projects (see Table
3). Correspondingly, these four models were used to test H5 to H8. Across the four models, the
basic structural tendencies stayed the same as for Network (A X B)2 to compare the observed
networks with the reference models. For Network (A X B)2, although EdgeB still meant that two
projects belonged to the same category, the other two structural tendencies had different
interpretations. Respectively, XEdge described the occurrence of a donor contributing to a
project, with the tie indicating the action of donation. The configuration of XASB could be
interpreted as a donor who contributed to multiple projects, with the ties indicating donations
also.
The third theory in the MTML framework detailed in Chapter 3 was signaling theory
(e.g., Spence, 1973), from which H5 and H6 were derived. H5 postulated that projects that had
more modality cues, which could serve as effective signals, would receive more donations. This
hypothesis corresponded to a dyadic cross-level relationship (XEdgeA) between a donor and a
project, with modality cues treated exogenous covariates associated with the project. Model 5,
which was similar to Model 2 in terms of its separation of video and picture counts as separate
nodal attributes, produced two parameter estimates for video (est. coeff. = .21, SE = .07) and for
picture (est. coeff. = -.001, SE = .007). While the former estimate met the statistical significance
threshold, the latter failed to do. Thus, H5 was only partially supported, suggesting that projects
that had more videos were likely to receive donations, whereas a similar pattern did not hold for
pictures in the project descriptions.

58

Figure 7. Network Configuration for H5 (XEdgeA)

Note: This cross-level network configuration is documented in Wang et al. (2014). A circle
represents a donor (human); a square represents a crowdfunding project. A solid icon represents
a node whose attributes are incorporated; an outline icon represents the otherwise. In the analysis
for H5, the nodal attributes included were, respectively, video count and picture count in the
project description. A black cross-level tie that connects a donor and a crowdfunding project
indicates that the donor gives money to the project.

As for the other hypothesis about signaling, H6 stated that more user reactions on the
crowdfunding pages could be another effective signal to facilitate the increase in donations.
While the attribute was different, the network configuration (XEdgeA) held the same as for H5.
Chapter 3 already described how reactions were measured, with comment, like, and share counts
as three sub-variables. Therefore, Model 6 included the three measures and generated parameter
estimates for each. Interestingly, although the estimates did not reach statistical significance for
comment (est. coeff. = -.009, SE = .05) and like (est. coeff. = .12, SE = .07), a significant
positive effect was observed for the number of shares (est. coeff. = .15, SE = .06). This means
that among the different types of reactions, only the number of shares on social media functioned
as an effective signal to facilitate donations. H6 was thus partially supported.

59

Figure 8. Network Configuration for H6 (XEdgeA)

Note: This cross-level network configuration is documented in Wang et al. (2014). A circle
represents a donor (human); a square represents a crowdfunding project. A solid icon represents
a node whose attributes are incorporated; an outline icon represents the otherwise. In the analysis
for H5, the nodal attributes included were, respectively, comment count, “Favorite” count, and
share count. A black cross-level tie that connects a donor and a crowdfunding project indicates
that the donor gives money to the project.

The last associative mechanism that the MTML framework proposed to investigate was
homophily, which aimed to explore whether donors exhibited the same donating behavior across
similar projects. Following the theory, H7 hypothesized that a donor was more likely to provide
monetary contributions to projects that belong to the same category. The hypothesis can be
represented by a cross-level closed triad (TriangleXAX) consisting of a donor and two projects.
In this configuration, the within-level tie should be interpreted as two projects that belong to the
same category, while the cross-level ties should be interpreted as actions of donation. Model 7
was fitted to test H7, which produced a parameter estimate of .39 (SE = .14) for the hypothesized
configuration. Therefore, it can be concluded that H7 was supported.

60

Figure 9. Network Configuration for H7 (TriangleXAX)

Note: This multilevel network configuration is documented in Wang et al. (2014). In the analysis
for H7, a red circle represents a donor (human); a blue square represents a crowdfunding project
(artifact). A blue within-level tie that connects two crowdfunding projects indicates that they
belong to the same category. A black cross-level tie that connects a donor and a crowdfunding
project indicates that the donor gives money to the project.

In line with H7, H8 furthered the investigation of the homophily phenomenon in
crowdfunding by positing that donors were attracted by the projects that shared the same
modality use patterns. Despite using a different network, this hypothesis corresponded to the
same network configuration (XStarADifference) as proposed in H2. Model 8 incorporated the
configuration and separated the estimates for the differences of videos (est. coeff. = -.02, SE
= .02) and pictures (est. coeff. = -.01, SE = .02), both of which were non-significant. To
conclude, H8 was not supported.

61
Figure 10. Network Configuration for H8 (X2StarADifference)

Note: This cross-level network configuration is documented in Wang et al. (2014). A circle
represents a donor (human); a square represents a crowdfunding project. A solid icon represents
a node whose attributes are incorporated; an outline icon represents the otherwise. In the analysis
for H2, the nodal attributes included were, respectively, video count and picture count in the
project description. A black cross-level tie that connects a donor and a crowdfunding project
indicates that the donor gives money to the project. A negative sign represents the numerical
difference of a nodal attribute. A larger node represents a higher level of a nodal attribute than a
smaller node.

4.3. Evaluation of Goodness-of-fit
Apart from the estimates generated in the multilevel ERGMs, it is also important to
evaluate the goodness-of-fit of the models. As the ERGMs used in MPNet implement the
algorithm of MCMCMLE (see Snijder, 2002 and Wang et al., 2013), the goodness-of-fit of
models can be assessed through comparing the network statistics of the estimated model against
the modelled network. For the estimated parameters, the algorithm calculates the “t-ratios” as
indicators of goodness-of-fit. For configurations included in the modelled network, t-ratios
smaller than .1 in absolute value indicate model convergence and good fit; for configurations that
are not included in the model, t-ratios smaller than 2.0 in absolute value indicate adequate fit
(Wang et al., 2016). Table 4 and Table 5 showed the statistics (means and standard deviations) of

62
the configuration counts across the simulated graphs as well as the goodness-of-fit statistics
respectively for Model 1 to 4 and Model 5 to 8. In Table 4, most of the t-ratios were smaller
than .1, with the exception of the XEdgeB configurations corresponding to H3 (t-ratio = .28) and
H4 (t-ratio = .23). Similarly, in Table 5, large t-ratios were found only for the two
X2StarADifference configurations (t-ratios = 4.23 and 2.49) hypothesized in H8. These relatively
high t-ratios suggested the potential problem by including the network configurations that
produce poor fit. Therefore, for the networks analyzed in the dissertation, better-fitting models
could be obtained by dropping network configurations such as XEdgeB and X2StarADifference
from the models.

63
Table 2. Results from Multilevel ERGMs (H1 to H4)
Model 1 Model 2 Model 3 Model 4
Configurations Effects
(Interpretations)
Estimate
(SE)
Estimate
(SE)
Estimate
(SE)
Estimate
(SE)
Structural Effects
XEdge
(A fundraiser initiates
a project. )
-9.21
*

(.49)
-8.36
*

(.46)
-8.92
*

(.37)
-7.99
*

(.41)

EdgeB
(Two projects belong
to the same category.)

-3.34
*
(1.12)

-2.91
*
(.73)

-5.45
*
(1.92)

-3.05
*
(1.00)

XASB
(A fundraiser initiates
multiple projects.)

.25
*

(.12)

.17
*

(.08)

.29

(.18)

.54
*
(.08)

Hypothesized Effects

TriangleXAX
(H1: Projects initiated
by the same
fundraisers
increasingly exploit
same categories.)

1.11
*
(.40)

X2StarADifference
(H2: Projects initiated
by the same
fundraisers
increasingly exploit
similar modality use
patterns.)

.43
(.27)
a

.05
(.14)
b
XEdgeB
(H3: Fundraisers with
higher levels of
external social capital
are more likely to
initiate a project.)

-.07
(.04)

64
XEdgeB
(H4: Fundraisers with
higher levels of
internal social capital
are more likely to
initiate a project.
.001
(.003)
Note:
*
= statistical significance at 95% confidence interval; a = video count; b = picture count

65
Table 3. Results from Multilevel ERGMs (H5 to H8)
Model 5 Model 6 Model 7 Model 8
Configurations Effects
(Interpretations)
Estimate
(SE)
Estimate
(SE)
Estimate
(SE)
Estimate
(SE)
Structural Effects
XEdge
(A donor donates to a
project. )
-3.94
*

(1.05)
-4.52
*

(.98)
-4.56
*

(1.20)
-5.22
*

(.79)

EdgeB
(Two projects belong to
the same category.)

-2.38
*
(1.12)

-4.00
(2.96)

-5.08
*
(2.21)

-2.03
*

(.98)

XASB
(A donor donates to
multiple projects.)

.56
*

(.02)

.77
*
(.34)

.49
*
(.04)

.54
*
(.22)
Hypothesized Effects
XEdgeA
(H5: Projects that have
more modality cues
will receive more
donations.

27
*

(.02)
a

-.001
(.007)
b

XEdgeA
(H6: Projects that
trigger more reactions
will receive more
donations.)

-.009
(.05)
c

.12
(.07)
d
.15
*
(.06)
e

TriangleXAX
(H7: A donor is more
likely to contribute to
projects that belong to
the same category.)

.39
*
(.14)

-.02
(.02)
a

66
X2StarADifference
(H8: A donor is more
likely to contribute to
projects that employ the
same modality use.
-.01
(.02)
b

Note:
*
= statistical significance at 95% confidence interval, which is achieved when ; a = video
count; b = picture count; c = comment count; d = “Favorite” count; e = share count

67
Table 4. Multilevel ERGMs Goodness-of-fit (H1 to H4)
Effects Interpretations Mean SD t-ratio
XEdge Edge term X 259.42 32.56 -.08
EdgeB Edge term B 12.12 1.88 .09
XASB Multiple initiations 8.86 1.42 .04
TriangleXAX H1 (Projects same category) 3.90 .98 -.06
X2StarADifference(1) H2 (Video count difference) 17.24 2.32 .07
X2StarADifference(2) H2 (Picture count difference) 16.02 1.99 .09
XEdgeB(1) H3 (External social capital) 273.41 29.03 .28
XEdgeB(2) H4 (Internal social capital) 266.57 27.38 .23

Table 5. Multilevel ERGMs Goodness-of-fit (H5 to H8)
Effects Interpretations Mean SD t-ratio
XEdge Edge term X 200.96 98.88 -.04
EdgeB Edge term B 9.87 1.03 .01
XASB Multiple donations 7.12 .89 .02
XEdge(1) H5 (Video count) 197.34 28.74 .09
XEdge(2) H5 (Picture count) 225.32 35.97 .09
XEdgeA(3) H6 (Comments) 208.58 102.42 -.07
XEdgeA(4) H6 (Favorites) 188.62 76.98 -.03
XEdgeA(5) H6 (Shares) 200.01 89.96 -.07
TriangleXAX H7 (Projects same category) 3.91 .48 .01
X2StarADifference(1) H8 (Video count difference) 14.03 6.62 4.23
X2StarADifference(2) H8 (Picture count difference) 12.93 7.53 2.49

68
Chapter 5: Discussion and Conclusion

5.1. Summary of Findings
The study of crowds represents “…both the promises and perils of our collective past and
the dreams and the dangers of global connectedness” (Stohl, 2014, p. 1). At the intersection
where the classical perspectives of collective behavior meet the emergent forms of mass
mobilization, the dissertation considers how the evolving notion of crowd brings about new
possibilities of large-scale networking and organizing. Drawing upon the literatures on collective
behavior, online community and social networks, the dissertation is guided by the theorizing of
crowds as organization (Agarwal et al., 2014; Bennett et al., 2014). The empirical inquiry of the
dissertation is enabled by a MTML framework that connects four theoretical perspectives
(exploration/exploitation tradeoff, social capital, signaling, homophily), from which eight
hypotheses were derived and tested. As an example of crowd-enabled organizations,
crowdfunding was analyzed to identify the forces that facilitate the formation of
multidimensional and multilevel networks. Digital trace data from a major donation-based
crowdfunding platform were retrieved and analyzed using exponential random graph models for
multilevel networks. In general, there were three major findings from the analyses. First,
fundraisers are likely to restrict crowdfunding campaign initiations within a limited number of
categories, as compared to exploring diverse categories of projects. Second, projects that
incorporate effective signals (i.e., more video production and wider social sharing) are more
likely to receive donations. Lastly, donors have the tendency to fund similar projects that belong
to the same categories. To facilitate the discussion of the three major findings in the next section,
they are respectively labeled as “fundraising exploitation”, “project signaling”, and “donor

69
homophily”. The three major findings correspond to the three types of agents in crowdfunding,
respectively fundraisers, projects, and donors. All results from hypothesis testing, as detailed in
Chapter 4, are summarized in Table 6 to give a detailed review of both the supported, partially
supported, and unsupported hypotheses.

70
Table 6. Summary of Findings
Hypothesis Theory Level

Configuration Dyadic
/triadic
Attributes Result
H1: Projects initiated
by the same
fundraisers
increasingly exploit
same categories.

Exploration/
exploitation
tradeoff
Multi

Triadic No Supported
H2: Projects initiated
by the same
fundraisers
increasingly exploit
similar modality use
patterns.

Exploration/
exploitation
tradeoff
Cross

Triadic Yes Unsupported
H3: Users with higher
levels of external
social capital are
more likely to initiate
a project than users
with lower external
social capital.

Social
capital
Cross

Dyadic Yes Unsupported
H4: Users with higher
levels of internal
social capital are
more likely to initiate
a project than users
with lower internal
social capital.

Social
capital
Cross

Dyadic Yes Unsupported

H5: Projects that have
more modality cues
will receive more
donations.

Signaling

Cross

Dyadic

Yes

Partially
supported
H6: Projects that
trigger more reactions
will receive more
donations.

Signaling Cross

Dyadic Yes Partially
supported

71
H7: A donor is more
likely to contribute to
projects that belong
to the same category.

Homophily Multi

Triadic No Supported
H8: A donor is more
likely to contribute to
projects that employ
the same modality
use.

Homophily Cross

Triadic Yes Unsupported

72
5.1.1. Fundraising Exploitation
An interesting phenomenon that the dissertation reveals is fundraising exploitation in
donation-based crowdfunding projects. As elaborated in Chapter 2, exploitation in crowdfunding
can be defined as the tendency for fundraisers to extensively draw on refined and efficient
fundraising routines rather than attempting to employ alternative fundraising strategies. This
phenomenon is represented by the supported H1 as shown in Chapter 4, which reveals that repeat
fundraisers tend to cluster their projects in fewer categories rather than span the projects across
multiple categories. In other words, fundraising is more likely to be exploitive than explorative
for repeat fundraisers in the sample.
A possible explanation for the pattern of fundraising exploitation has to do with the
fundraisers’ pursuit of value appropriation. As different types of crowdfunding projects (e.g.,
reward-, equity-, and donation-based) are launched for entirely different purposes, whether a
fundraiser pursues value creation versus appropriation is largely determined by the purpose of
the project. For example, Stanko and Henard (2017) found that in innovation-focused
crowdfunding projects launched on Kickstarter, fundraisers may seek to draw from a wide array
of alternative resources (exploration) instead of drawing intensively from the existing resources
(exploitation). They further contended that the exploration in innovation-focused crowdfunding
allows for “the possibility of new variations through the consideration of divergent knowledge
bases and broad-based, expansive conversations” (p. 789). On the contrary, the current
dissertation found the opposite phenomenon in donation-based crowdfunding: when fundraisers
launch projects to ask for donations the default strategy may be focused on value creation instead
of value appropriation. This inconsistency can be interpreted through a simple economic
perspective. After all, reward- and equity-based crowdfunding are essentially a form of

73
investment that seeks to maximize the mutual benefits for both fundraisers and funders (backers),
whereas the donation-based campaigns are more or less “zero-sum” and do not necessarily create
any new values during fundraising. It thus makes sense to reexamine the goal of a given
crowdfunding project in order to understand where the fundraiser’s strategy implementation falls
on the continuum of exploration and exploitation. As more studies have started to reveal the
disparate landscape of donation-based crowdfunding as compared to entrepreneurial
crowdfunding (e.g., Berlinger & Kenworthy, 2017; Kenworthy, 2019; Xu, 2018), it should be
acknowledged that the conclusions drawn from for-profit crowdfunding might not always hold
true in nonprofit projects, particularly those launched by individual fundraisers.
A second explanation for the presence of fundraising exploitation is network inertia. Kim,
Oh, and Swaminathan (2006) defined network inertia as “a persistent organizational resistance to
changing interorganizational network ties or difficulties that an organization faces when it
attempts to dissolve old relationships and form new network ties” (p. 704). They identified
several factors that would increase the likelihood of network inertia, including the lack of
structural holes (non-overlapping contacts, see Burt, 1992) in networks. In other words, the
perspective of network inertia asserts that it is more difficult for an already well-connected and
cohesive network to change. Another factor that Kim et al. (2006) identify is competition, which
may reduce network inertia and lead to voluntary network change. It has been shown that a
considerable amount of monetary contributions come from family, friends, and close-knit
contacts (Horvát et al., 2015; Snyder, Crooks, Mathers, & Chow-White, 2017). In this type of
crowdfunding networks, both structural holes and competition may be to a large extent scarcer
than in for-profit campaigns that have more outreach concerns and practices. Hence, it stands to

74
reason that the lack of structural holes and competition in donation-based campaigns may
cultivate network inertia, which further results in the increase of fundraising exploitation.
A third explanation for fundraising exploitation can be drawn from the theory of resource
mobilization (e.g., McCarthy & Zald, 1977), which analyzes how individuals and organizations
engaged in collective action optimize the strategy of obtaining needed resources in a cost-
effective manner. The theory assumes that individuals are rational but also perform constant
cost-benefit analysis. In donation-based crowdfunding, if the ultimate benefit is to maximize
donations, a major cost may come from employing sophisticated message design strategies that
are more time-consuming, less convenient, and, potentially, expensive. To avoid such costs,
simple yet less “viral” fundraising messages may be crafted and distributed only within smaller
social circles, where family and friends may feel obligated to contribute regardless of the quality
of the fundraising messages. This pattern is especially salient in medical crowdfunding, where
recurring projects about the same cause are often launched by individuals who employ
crowdfunding to pay medical bills (Proelss, Schweizer, & Zhou, 2020). Recurring projects as
such may partially explain why some fundraisers choose to replicate the selection of the same
categories in hope to continually draw on the resources residing in the existing networks. In this
way, individuals may still be able to enjoy the benefits of crowdfunding while avoiding incurring
the costs.

5.1.2. Project Signaling
Consistent with existing research (e.g., Ahlers, Cumming, Günther, & Schweizer, 2015;
Lu, 2014; Xu, 2018), the dissertation shows that project signaling plays an important role in
attracting donors in crowdfunding. Particularly, empirical analysis identified two effective

75
signals that were positively related to fundraising success, including the richness of video
production as a modality cue and the number of social shares as a cue of “social buzz” (Thies et
al., 2014, or social proof, see Cialdini, 2001). Although it is not surprising to find that these
signals contribute to fundraising success given how the patterns have been replicated in
crowdfunding research (e.g., Feldmann & Gimpel, 2016; Koch & Siering, 2015; Lu, 2014;
Mollick, 2014; Xu, 2018), the project contributes to this literature by showing what signals are
particularly appealing to repeat donors. Repeat donors differentiate themselves from one-time
donors by their high involvement and rich experience in backing projects. Consequently, we may
expect repeat donors to be more sensitive to informative signals than less involved donors. This
dissertation found that the numbers of pictures, comments, and favorites did not appeal to repeat
donors, whereas the number of videos and shares did. Here, it would be interesting to assess how
much effort is required to generate a signal and see if there is a relationship between signaling
effort and effectiveness. Obviously, videos are more difficult to produce than pictures, and social
shares may require more effort than comments and favorites as the sharers are transferring both
the crowdfunding messages and the fundraising obligations to other social media platforms.
Hence, it may be the case that not all signals are equal, and some are more informative and
“robust” in suggesting the quality of the projects than others.
The tendency of project signaling would raise another major question when considered
with fundraising exploitation: if project signaling requires more effort from fundraisers, and
fundraising exploitation requires less effort from them, why do the two mechanisms co-exist? In
fact, this paradoxical question is a crucial yet unresolved problem pertaining to a refined
understanding of crowdfunding management. In existing literature, despite the existence of
research on fundraising exploration/exploitation (e.g., Stanko & Henard, 2017) and project

76
signaling (e.g., Courtney et al., 2017; Feldmann & Gimpel, 2016) separately, the current
dissertation is the first study that discusses the symbiotic relationship between the two
mechanisms from a network perspective. The interaction of fundraising exploitation and project
signaling can be construed as a dynamic process that allows fundraisers to draw benefits from
the existing networks while requiring them to pay a certain amount of effort to embed enough
signals in their projects. These signals may help maintain the community and crowdfunding
networks from which they draw benefits. This problem resembles what Olson (1965) described
as the “free rider problem”, which states that individuals benefiting from shared resources
without paying for them can often lead to the depletion of the shared resources. Following
Olson’s perspective, if donation-based crowdfunding can be conceptualized as a source of both
public goods (platforms, communities) and private goods (donations), the existence of “free
riders” may hinder the sustaining of crowdfunding communities. The free riders in donation-
based crowdfunding can be defined as the users who only care about maximizing the donations
they receive, without reciprocating the donations, backing other projects, or contributing to
community development by obeying the tacit rules. In medical crowdfunding, the free rider
problem and its negative consequences have started to emerge. Suspicion, distrust, and even the
collapse of the networks may arise in response to a small number of free riders, who fail to
contribute to the maintenance of the crowdfunding systems by using the services honestly and
responsibly (Kenworthy, 2018).

5.1.3. Donor Homophily
In Chapter 2, the dissertation proposed a question that “Will houses that are constructed
similarly attract the same people?” Based on the finding of donor homophily, the answer seems

77
to be positive, but with a contingency as represented by the tendency for donors to fund projects
that are in the same category. While this finding itself is not particularly counter-intuitive, it has
important implications for the development of a more refined understanding of the effects of
categorization in crowdfunding. Existing research (Oh & Monge, 2013; Sargent, Clark, Monge,
& Fulk, 2018; Xu, 2019) has already documented how categorization on crowd-based platforms
contributes to the organizing of crowds and influences individual behavior. As Zuckerman
(2017) explained, users can feel strong pressure to shape themselves to fit in existing categories.
Thus, categories may serve as the cognitive foundations for web users to make sense of the
environment (Xu, 2019). Although the categories in crowdfunding are often subjectively
designated by the platforms, they serve to restructure the large crowdfunding networks into
smaller segments, each holding homophilous participants that share certain interests or
characteristics. Therefore, H7 from another angle confirmed the categorization effect, which
suggest that donors could be attracted and, from a network perspective, rather stably attached to
certain categories. Donors’ attachment to certain categories may also result from their limited
cognitive and financial capabilities that restrict them from spanning multiple categories.
While the tendency of donor homophily certainly has its benefits in crowd organizing, it
also means that both human and material resources could also be trapped into certain categories
and could hardly flow across categories (see Liu, Yang, Wang, & Hahn, 2015). The
segmentation of donors may be a signal of a well-established subcommunity, but it may also be a
sign of the difficulty in effective resource mobilization. A question that can be asked is if we
assume that dog owners will never fund cat projects, will they ever visit the webpages of any
projects in the category of “cat”?

78
If the category is instead created as “animal” that holds projects for dogs, cats, as well as
other animals, one may be more hesitant to answer the above question. In fact, the studied
crowdfunding platform has undergone several major changes in categorization in the past years,
potentially in order to establish communities through leveraging donor homophily while
minimizing the negative influence of donor segmentation. For example, in 2016 the platform
adopted a strategy by using several keywords in one category, such as “Accident-Personal-
Crisis”, and “Babies-Kids-Family” (Xu, 2018). This attempt of re-categorization might facilitate
blurring the artificial boundaries between similar projects and thus reduce the potential negative
impact of donor homophily on resource mobilization.

5.1.4. Notes on Unsupported Hypotheses
While the previous sections have discussed the findings that were fully or partially
supported, it is important to discuss the possible reasons why other hypotheses (H2, H3, H4, H8)
were not supported. To begin with, H2 was derived from the perspective of
exploration/exploitation tradeoff, which predicted that fundraisers were more likely to be
consistent in modality use across the projects they launched. The rejection of the hypothesis
suggested that there could be inconsistencies in terms of modality use even among projects by
the same fundraisers. Interestingly, this finding does not necessarily contradict the theory of
exploration/exploitation, as March (1991) argued that organizational learning is a long-term and
adaptive process. In this process, trial and error may occur, and fundraisers can use the feedback
they get from the community (e.g., donations, reactions, etc.) to help them refine the strategy.
Thus, although the supported H1 may suggest that fundraisers exploit the same categories, the
unsupported H2 could mean that fundraisers may explore the optimal modality use strategies over

79
time. This inconsistency between H1 and H2 may suggest the possibility for some fundraisers to
be engaged in both exploration and exploitation, but with regard to different aspects of the
fundraising strategies.
H3 and H4 were raised to establish relationships between external/internal social capital
and project initiation, both of which were not supported. In other words, we could not draw the
conclusion from the sample that the fundraisers who possessed higher levels of resources would
seek to translate their existing network benefits into actual financial capital. Several explanations
may be pertinent. First, the network benefits that the fundraisers had might be sufficient that they
no longer needed to launch more crowdfunding projects to retrieve more resources. Second, it is
possible that using existing measures available in the datasets as proxies was a less than ideal
solution. To review, the dissertation used the donations from social sharing as a measure for
external social capital and the number of projects the fundraiser launched as a measure for
internal social capital. Although these measures to some extent capture the network properties of
individuals, they might still be insufficient to fully represent social capital as a complex and
contested notion in social science. Part of the reason for this potential incongruity is rooted in the
lack of individual-level network data, which, if available, can provide richer information about
what individuals occupy advantageous network positions.
Lastly, H8 stated that a donor was more likely to contribute to projects that employed the
same modality use, which was not supported. To understand why the hypothesis was rejected, it
may be helpful to juxtapose H8 with H7. Unlike H8 that predicted that homophily could be found
in modality use, H7 predicted that homophily was seen in category repetition and was supported.
To compare the two hypotheses, it seems that homophilous projects could be better represented
by the ones that share the same category than those that apply similar modality use strategies. It

80
makes sense to think that two cat projects are more homophilous than two projects that both use
videos for storytelling. An alternative explanation is if donors are committed to donate to
multiple or even recurring projects, they may already be highly motivated and convinced of the
value of the projects. Therefore, modality cues may become less important.

5.2. General Thoughts on Crowd Theorizing
Apart from the findings that specifically deal with donation-based crowdfunding network
mechanisms, the dissertation also provides implications for crowd theorizing. In general, the
dissertation aims to present a theoretical proposition
that crowds, however large and complex, are comprised of basic constituent units and
structures—people who participate, artifacts that people interact with, and networks that hold
people and artifacts together. Thus, different configurations of the constituents determine the
nature, properties, and functions of the crowds. Through countless rounds of dynamic practices
over time, diverse sorts of social players stitch the multidimensional and multilevel networks that
serve as the infrastructure of crowd-enabled organizations. At the macro level, these innovative
types of organizations take the form of deinstitutionalized and decentralized network
organizations consisting of both material entities and social relations (Stohl, 2014). At the micro
level, miniature social structures such as dyads, triads, groups, teams, and projects are constantly
formed and shuffled due to different associative mechanisms. Over time, networks become both
the media and the outcomes of crowd organizing.
In the article where the expression “organization in the crowd” is proposed, Bennet et al.
(2014) maintained that “…it is important to consider differences in how participants engage
them and what this means for the development of the broader networks of networks that

81
constitute the crowd” (p.249). Specifically, the dissertation explicates various crowd associative
mechanisms by delving into eight different structural tendencies in crowdfunding processes
focusing on repeat fundraisers and donors. Empirically, it analyzes recurring fundraising and
donation patterns using theories and perspectives that shed light on the adaptive organizational
processes and cumulative network benefits from these activities. Although repeat users are not
the only players in crowds, they may be the “core contributors” (Bennett et al., 2014) that occupy
central network positions due to their high levels of participation and connectedness to other
users and projects. In comparison, occasional or one-time fundraisers and donors may be the “co-
producers” (Bennett et al., 2014) and occupy more peripheral positions in crowd-enabled
networks. While the empirical analyses in this dissertation focus on how the core contributors
play a major role in structuring the community and setting the rules, it also opens the discussion
for future research to explore how the co-producers contribute to crowd organizing.
With the theorizing and evidence provided in the dissertation, it would be interesting to
revisit the guiding question. The question states that “if people are born to pursue their own
diverse interests, when can they begin to act in concert in a crowd?” (McPhail, 2017). While
there is no easy answer to such a grand question, this dissertation shows that at least
multidimensional networks theory and analysis lends itself a viable approach to providing a
systemic answer to better explain the phenomenon of “organization in the crowd” (Bennett et al.,
2014).

5.3. Limitations
Despite the implications and contributions, the dissertation has several limitations. First,
although the dissertation treats exploration and exploitation as two competing fundraising

82
strategies (i.e., “orthogonal”, see Gupta et al., 2006), it does not consider the possibility that the
two strategies could be complementary. Although this treatment was consistent with March
(1991) who argued that exploration and exploitation were fundamentally incompatible, other
studies have provided alternative views (e.g., Gupta, et al., 2006; He & Wong, 2004; Levinthal &
March, 1993). For example, Gupta and colleagues (2006) believed that whether or not
exploration and exploitation are mutually exclusive largely depends on the availability of
resources to sustain both strategies at the same time; when organizational resources are ample, it
is possible to employ both strategies at the same time. Likewise, He and Wong (2004) used the
term “ambidexterity” to describe the possibility and benefits to embrace both explorative and
exploitive strategies. Levinthal and March (1993) stated that “[t]he basic problem confronting an
organization is to engage in sufficient exploitation to ensure its current viability and, at the same
time, to devote enough energy to exploration to ensure its future viability (p.105). All these
alternative perspectives suggest that treating exploration and exploitation as orthogonal variables
could lead to an incomplete representation of how the two strategies are actually implemented
and balanced. Hence, in the context of crowdfunding, it is possible that exploration and
exploitation strategies are more complex than what this dissertation presents, which can be seen
in other aspects than category and modality use. When taking into account other factors such as
prior success or failure in implementing either strategy, the situation can be further complicated.
For example, highly successful fundraisers may be capable of using both strategies while
maintaining the equilibrium, whereas less successful fundraisers can learn from others’
exploration to minimize the costs, which this dissertation does not take into account.
Second, according to Wang et al. (2013), a complete set of multilevel networks would
consist of Network A, Network B, and Network A X B. a complete set of multilevel networks

83
would consist of all macro (project-project), meso (project-individual), and micro (individual-
individual) levels of connections (see Wang et al., 2013). Whereas the datasets used for the
dissertation had both the macro (B) and meso (X) networks, they did not include the micro level
connections that show how individuals are connected with each other. Should the micro level
network data be available, it would allow for additional hypotheses to be tested, particularly
those related to social contagion. Although on some crowd-based platforms (e.g., Wikipedia,
Kaggle) the social connections among individuals are publicly available and easily obtainable, in
crowdfunding such information is harder to acquire. An important reason such data are often
missing is possibly due to the privacy concerns that a significant portion of crowd activities
come from participants who wish to remain anonymous. Regardless, the platforms may have
richer data than the publicly available data for internal use only. Therefore, this limitation also
suggests that it is necessary for scholars and practitioners to work out protocols that facilitate
collaboration.
The third limitation is the relatively restricted sample size in ERGMs. Although the
sample sizes in the dissertation were one of the largest in existing studies using ERGMs for
multilevel networks, it should be acknowledged that the introduction of refined algorithms (e.g.
equilibrium expectation, see Stivala, Robins, & Lomi, 2020) is increasing the efficiency of
computation, making fitting large sample ERGMs an attainable goal in the near future. A large
sample size will also allow for the examination of additional network configurations that the
observed networks did not possess, which may open new possibilities for network modeling and
crowd theorizing.
The fourth limitation is that the analyses assume all fundraisers and donors in the sample
were individuals and does not consider the possibility that sometimes organizations can launch

84
or contribute to campaigns on behalf of individuals. Unfortunately, the datasets used the in
dissertation do not contain further information about whether there was an organizational force
behind the fundraiser/donor. Should this information be available, it would have been helpful for
evaluating if fundraising and donation strategies are different between individuals and
organizations.
Lastly, the data used in the dissertation were treated as cross-sectional to facilitate
network analyses, although the actions involved in those crowdfunding campaigns (e.g.,
fundraising, donating) did not happen at the same time. It should be acknowledged that earlier
fundraising and donating behaviors may have direct impact on the later ones. Hence, a more
nuanced understanding of the crowdfunding dynamics can be achieved by examining the over-
time changes of the networks by taking into account the effects of time.

5.4. Future Research
The dissertation has informed future research in several ways. First, a variety of other
crowd-enabled organizations than donation-based crowdfunding can be studied using the MTML
framework but with different sets of theories that better suit the subjects of study. According to
Monge and Contractor (2003), as alternative social theories can make differential or even
contradictory predictions about networks, it is important to use appropriate and complementary
theories to constitute a multitheoretical framework that accounts for the corresponding properties
and mechanisms in the networks. As various crowd-enabled networks differ in structure, size,
and formation/decay mechanisms, it is important to ensure that the selected theories correspond
to the observed network patterns and available network attributes. In other words, the MTML
approach allows for both theoretical comprehensiveness and fluidity. An example that uses

85
MTML to study crowd-enabled organizations is provided by Xu et. (2018), in which an MTML
framework is used to investigate a crowdsourcing artistic design community. The study uses
public goods theory, balance theory, and evolutionary and ecological theory to examine the
effects of exogenous and endogenous variables in the multidimensional and multilevel networks
of crowds. Using ERGMs, the paper shows promises in augmenting the theoretical power of the
MTML framework through applying advanced network analytical techniques. Another piece of
research (Margolin & Markowitz, 2018), while not addressing the network effects directly,
examines the large amount of text data on a crowdsourcing review system (Yelp) using the
MTML approach. As similar studies are emerging, the MTML may emerge as a particularly
applicable framework in crowdsourcing, crowdfunding, and other types of crowd-enabled
organizations.
Second, it is also promising for future research to develop a more comprehensive yet
nuanced understanding of the levels of analysis in crowd-enabled organizations, in which a
dichotomy of “individual versus collective” might be sometimes too general to capture the
hierarchical structures in such systems. In acting crowds, individuals often are nested in local
structures and diverse types of subcommunities. At the same time, the individuals can interact or
create various forms of artifacts. The crowd-enabled organizations and platforms can also
designate categories, hashtags, and labels to organize or reclassify artifacts and individuals. This
means that the complexity of the actual multidimensional and multilevel networks can far
transcend the theoretical and methodological boundaries of crowd research as of today, spanning
a wide spectrum of levels from a small hashtag to the entire system. Although adding more
levels of analysis undoubtedly increases the analytical difficulty, the real challenge lies in
developing the theoretical justifications for the attempt. In fact, there have been many vigorous

86
debates regarding the appropriate ways to conceptualize multilevel models (Klein & Kozlowski,
2000), but most of the debates are still unfolded within the traditional trichotomy consisting of
the micro-, meso-, and micro- levels of analysis. Although this trichotomy still has its validity to
some extent, it would be timely for future research to redefine which types of agents belong to
what levels of analysis, so as to develop multilevel models that are both applicable to explicating
crowd dynamics and are up to date.
Third, future research can take full advantage of the time stamps, if available, in the
digital trace datasets pertaining to crowd and collective behavior. The presence of time variables
allows for a longitudinal examination and modeling of the formation and changes in networks.
Meanwhile, analytical techniques that take into account the inter-temporal dependence are on the
rise. Network modeling techniques such as temporal exponential random graph models
(TERGM), stochastic actor-oriented models (SAOM), relational event models (REM), and
dynamic network actor models (DyNAM) can be used to give a better account of how networks
dynamically form and change over time.
Lastly, while digital trace data provide unprecedented opportunities for researchers to
explore many aspects of online crowds, it should also be acknowledged that usually there is a
lack of cognitive and motivational information, which is critical for the understanding of crowd
behavior. To ameliorate this built-in weakness, mixed-method approaches can be used to provide
in-more depth explanations of the patterns revealed by the “big data” about crowds. Both
experimental and qualitative methods might provide additional information to explain the
motivations of crowd behavior.

87
5.5. Conclusion
In today’s society, the constantly emerging and updating information and communication
technologies (ICTs) have enabled innovative forms of mass mobilization and mediated
communication, which have posed thought-provoking questions about the century-old negative
view of crowds. The increasing prevalence of crowd-enabled organizations should be
accompanied by a deeper understanding of them. It is thus critical for researchers to disentangle
the complex interactions of crowd components and fathom how large-scale and effective
organizing is enabled and sustained through the “stitching mechanisms” in crowd-enabled
networks (Bennett et al., 2014). The current dissertation shows that a multitheorietcal multilevel
(MTML) framework may possess theoretical and analytical advantages over approaches that
restrict research to only one theory or single-level of analysis. The explanatory power of the
MTML framework is further augmented by the advances in multidimensional networks theory
and analytical tools on which the dissertation relies to draw empirical inferences from rich digital
trace data on crowd communication and behavior. Overall, the dissertation suggests that instead
of treating a crowd as a collection of individuals whose attributes and networks are neglected, ill-
defined, or unnecessarily homogenized, it is more theoretically sound and, at the same time,
empirically feasible to explicate how large numbers of heterogeneous individuals and
technologies are embedded and functioning in the social fabric of crowd-enabled organizations.
In other words, as the capacity to collect and analyze massive amounts of crowd data has
allowed researchers to reexamine many classical hypotheses, it has become more promising to
empirically investigate how the “organization in the crowd” phenomenon takes place.

88
References

Adler, P. S., & Kwon, S. W. (2002). Social capital: Prospects for a new concept. Academy of
Management Review, 27(1), 17-40.
Agarwal, S. D., Bennett, W. L., Johnson, C. N., & Walker, S. (2014). A model of crowd enabled
organization: Theory and methods for understanding the role of twitter in the occupy
protests. International Journal of Communication, 8, 27.
Ahlers, G. K., Cumming, D., Günther, C., & Schweizer, D. (2015). Signaling in equity
crowdfunding. Entrepreneurship Theory and Practice, 39(4), 955-980.
Ahmad, M. A., Borbora, Z., Srivastava, J., & Contractor, N. (2010, December). Link prediction
across multiple social networks. In 2010 IEEE International Conference on data mining
workshops (pp. 911-918). IEEE.
Aldrich, D. P. (2012). Building resilience: Social capital in post-disaster recovery. University of
Chicago Press.
Allport, F. H. (1924). The group fallacy in relation to social science. American Journal of
Sociology, 29(6), 688-706.
Aral, S., Muchnik, L., & Sundararajan, A. (2009). Distinguishing influence-based contagion
from homophily-driven diffusion in dynamic networks. Proceedings of the National
Academy of Sciences, 106(51), 21544-21549.
Arvidsson, A., Caliandro, A., Airoldi, M., & Barina, S. (2016). Crowds and value. Italian
directioners on Twitter. Information, Communication & Society, 19, 921-939.
Baker, W., Nohria, N., & Eccles, R. G. (1992). The network organization in theory and
practice. Classics of Organization Theory, 8, 401.

89
Barnes, J. A., & Harary, F. (1983). Graph theory in network analysis. Social Networks, 5(2),
235-244.
Barnett, C., (2016). Trends show crowdfunding to surpass VC after 2016. Retrieved from:
https://medium.com/startup-grind/trends-show-crowdfunding-to-surpass-vc-in-2016-
65df924d8a82.
Barnett, G. A., & Benefield, G. A. (2017). Predicting international Facebook ties through
cultural homophily and other factors. New Media & Society, 19(2), 217-239.
Barranco, O., Lozares, C., & Muntanyola-Saura, D. (2019). Heterophily in social groups
formation: a social network analysis. Quality & Quantity, 53(2), 599-619.
Belleflamme, P., Lambert, T., & Schwienbacher, A. (2014). Crowdfunding: Tapping the right
crowd. Journal of Business Venturing, 29, 585-609.
Bennett, W. L., & Segerberg, A. (2012). The logic of connective action: Digital media and the
personalization of contentious politics. Information, Communication & Society, 15, 739-
768.
Bennett, W. L., Segerberg, A., & Walker, S. (2014). Organization in the crowd: peer production
in large-scale networked protests. Information, Communication & Society, 17(2), 232-
260.
Berdou, E. (2011). Organization in open source communities. New York/London: Routledge.
Berger, P. L., & Luckmann, T. (1966). The social construction of reality: A treatise in the
sociology of knowledge (No. 10). Penguin UK.
Berliner, L. S., & Kenworthy, N. J. (2017). Producing a worthy illness: Personal crowdfunding
amidst financial crisis. Social Science & Medicine, 187, 233-242.

90
Bian, Y. (1997). Bringing strong ties back in: Indirect ties, network bridges, and job searches in
China. American Sociological Review, 366-385.
Bighash, L., Oh, P., Fulk, J., & Monge, P. (2017). The Value of Questions in Organizing:
Reconceptualizing Contributions to Online Public Information Goods. Communication
Theory, 28(1), 1-21.
Bimber, B., Flanagin, A. J., & Stohl, C. (2005). Reconceptualizing collective action in the
contemporary media environment. Communication Theory, 15, 365-388.
Bimber, B., Flanagin, A., & Stohl, C. (2012). Collective action in organizations: Interaction and
engagement in an era of technological change. Cambridge University Press.
Blau, P. M., & Scott, W. R. (1962). Formal organizations: A comparative approach. Stanford
University Press.
Blumer, H. (1969). Collective behavior. In A. M. Lee (Ed.), Principles of sociology (pp. 165–
221). New York, NY: Barnes and Noble.
Borgatti, S. P., Mehra, A., Brass, D. J., & Labianca, G. (2009). Network analysis in the social
sciences. Science, 323(5916), 892-895.
Borst, I., Moser, C., & Ferguson, J., (2018). From friendfunding to crowdfunding:
Relevance of relationships, social media, and platform activities to crowdfunding
performance. New Media & Society, 20, 1396-1414.
Bourdieu, P. (1986). The forms of capital. Cultural theory: An anthology, 1, 81-93.
Brabham, D. C. (2010). Moving the crowd at Threadless: Motivations for participation in
a crowdsourcing application. Information, Communication & Society, 13(8),
1122-1145.

91
Bradach, J. L., & Eccles, R. G. (1989). Price, authority, and trust: From ideal types to plural
forms. Annual Review of Sociology, 15(1), 97-118.
Burt, R. S. (1992). Structural holes: The social structure of competition. Harvard University
Press.
Burt, R. S. (1997). A note on social capital and network content. Social Networks, 19(4), 355-
373.
Burt, R. S. (2000). The network structure of social capital. Research in Organizational
Behavior, 22, 345-423.
Butler, B., Joyce, E., & Pike, J. (2008, April). Don't look now, but we've created a bureaucracy:
the nature and roles of policies and rules in Wikipedia. In Proceedings of the SIGCHI
conference on human factors in computing systems (pp. 1101-1110). ACM.
Butticè, V., Colombo, M. G., & Wright, M. (2017). Serial crowdfunding, social capital, and
project success. Entrepreneurship Theory and Practice, 41(2), 183-207.
Butts, C. T. (2019). A dynamic process interpretation of the sparse ERGM reference model. The
Journal of Mathematical Sociology, 43(1), 40-57.
Calic, G., & Mosakowski, E. (2016). Kicking off social entrepreneurship: How a sustainability
orientation influences crowdfunding success. Journal of Management Studies, 53(5),
738-767.
Cialdini, R. B. (2001). Influence: Science and practice (4th ed.). Boston, MA: Allyn & Bacon.
Callon, M. (1986). The sociology of an actor-network: The case of the electric vehicle.
In Mapping the dynamics of science and technology (pp. 19-34). Palgrave Macmillan,
London.

92
Callon, M. & Latour, B. (1981). Unscrewing the big leviathan: How actors macro-structure
reality and how sociologists help them to do so. In: Knorr-Cetina K and Cicourel AV
(eds) Advances in Social Theory and Methodology: Towards an Integration of Micro-and
Macro Sociologies. Boston, MA: Routledge, pp. 277–303.
Canetti, E. (1984). Crowds and power. Macmillan.
Carley, K. M., Diesner, J., Reminga, J., & Tsvetovat, M. (2007). Toward an interoperable
dynamic network analysis toolkit. Decision Support Systems, 43(4), 1324-1347.
Castells, M. (2000). Materials for an exploratory theory of the network society1. The British
Journal of Sociology, 51, 5-24.
Castells, M. (2011). The rise of the network society (Vol. 12). John Wiley & Sons.
CDC Foundation. (2020). CDC Foundation launches crowdfunding campaign to extend and
accelerate public health coronavirus response. Retrieved from:
https://www.cdcfoundation.org/pr/2020/cdc-foundation-launches-coronavirus-covid19-
crowdfunding-campaign-charidy
Certo, S. T. (2003). Influencing initial public offering investors with prestige: Signaling with
board structures. Academy of Management Review, 28(3), 432-446.
Centola, D., Willer, R., & Macy, M. (2005). The emperor’s dilemma: A computational model of
self-enforcing norms. American Journal of Sociology, 110(4), 1009-1040.
Chu, S. C., & Kim, Y. (2011). Determinants of consumer engagement in electronic word-of-
mouth (eWOM) in social networking sites. International Journal of Advertising, 30(1),
47-75.
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.

93
Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of
Sociology, 94, S95-S120.
Coleman, J. S. (1990). Foundations of social theory. Cambridge, MA: Harvard University Press.
Colgren, D., (2014). The rise of crowdfunding: Social media, big data, cloud
technologies. Strategic Finance, 95, 56-57.
Connelly, B. L., Certo, S. T., Ireland, R. D., & Reutzel, C. R. (2011). Signaling theory: A
review and assessment. Journal of Management, 37(1), 39-67.
Constine, J. (2020). Facebook hits 2.5B users in Q4 but shares sink from slow profits.
Retrieved from: https://techcrunch.com/2020/01/29/facebook-earnings-q4-2019/
Contractor, N. (2009). The emergence of multidimensional networks. Journal of
Computer-Mediated Communication, 14(3), 743-747.
Contractor, N. S., DeChurch, L. A., Carson, J., Carter, D. R., & Keegan, B. (2012). The
topology of collective leadership. The Leadership Quarterly, 23(6), 994-1011.
Contractor, N. & Monge, P. R., (2003, August). Using multi-theoretical multi-level
(MTML) models to study adversarial networks. In Dynamic Social Network
Modeling and Analysis, Proc. Dynamic Social Network Modeling and Analysis
Workshop, US National Research Council, Committee on Human Factors (pp. 325-
344).
Contractor, N., Monge, P., & Leonardi, P. M. (2011). Multidimensional networks and the
dynamics of sociomateriality: bringing technology inside the network. International
Journal of Communication, 5, 39.

94
Contractor, N. S., Wasserman, S., & Faust, K. (2006). Testing multitheoretical, multilevel
hypotheses about organizational networks: An analytic framework and empirical
example. Academy of Management Review, 31(3), 681-703.
Cordova, A., Dolci, J., & Gianfrate, G. (2015). The determinants of crowdfunding success:
evidence from technology projects. Procedia-Social and Behavioral Sciences, 181, 115-
124.
Courtney, C., Dutta, S., & Li, Y. (2017). Resolving information asymmetry: Signaling,
endorsement, and crowdfunding success. Entrepreneurship Theory and Practice, 41(2),
265-290.
Davies, R., (2015). Three provocations for civic crowdfunding. Information, Communication &
Society, 18, 342-355.
Dees, J. G. (2012). A tale of two cultures: Charity, problem solving, and the future of social
entrepreneurship. Journal of Business Ethics, 111(3), 321-334.
Dehghani, M., Johnson, K., Hoover, J., Sagi, E., Garten, J., Parmar, N. J., ... & Graham, J.
(2016). Purity homophily in social networks. Journal of Experimental Psychology:
General, 145(3), 366.
Della Porta, D. (2014). Comment on organizing in the crowd. Information, Communication &
Society, 17(2), 269-271.
Desanctis, G., & Monge, P. (1999). Introduction to the special issue: Communication processes
for virtual organizations. Organization Science, 10(6), 693-703.
Dolata, U., & Schrape, J. F. (2016). Masses, crowds, communities, movements: Collective action
in the internet age. Social Movement Studies, 15(1), 1-18.
Drake, D., (2017). 2,000 global crowdfunding platforms to choose from by 2016: Top 5 growth

95
indicators. Retrieved from: https://www.huffingtonpost.com/david-drake/2000-global-
crowdfunding-_b_8365266.html
Easley, D., & Kleinberg, J. (2010). Networks, crowds, and markets: Reasoning about a highly
connected world. Cambridge University Press.
Ellison, N. B., Steinfield, C., & Lampe, C. (2007). The benefits of Facebook “friends:” Social
capital and college students’ use of online social network sites. Journal of Computer‐
Mediated Communication, 12(4), 1143-1168.
Feldmann, N., Gimpel, H., (2016). Financing projects through Enterprise crowdfunding:
Understanding the impact of proposal characteristics on funding success. In ECIS (pp.
Research-in).
Festinger, L., Pepitone, A., & Newcomb, T. (1952). Some consequences of de-individuation in a
group. The Journal of Abnormal and Social Psychology, 47, 382.
Frydrych, D., Bock, A.J., & Kinder, T., (2014). Exploring entrepreneurial legitimacy in reward-
based crowdfunding. Venture Capital, 16, 247-269.
Fukuyama, F. (2001). Social capital, civil society and development. Third World
Quarterly, 22(1), 7-20.
Fulk, J., & DeSanctis, G. (1995). Electronic communication and changing organizational
forms. Organization Science, 6(4), 337-349.
Ganesh, S., & Stohl, C. (2014). Community organizing, social movements, and collective
action. The Sage Handbook of Organizational Discourse, 743-766.
Ganley, D., & Lampe, C. (2009). The ties that bind: Social network principles in online
communities. Decision Support Systems, 47(3), 266-274.

96
Garcia, A., 2016. 65 GoFundMe accounts raising money for Flint water crisis. Available at:
http://money.cnn.com/2016/01/23/technology/flint-water-gofundme-
campaigns/index.html
Giddens, A. (1979). Agency, structure. In Central problems in social theory (pp. 49-95).
Palgrave, London.
Gilbert, M. (2006). Rationality in collective action. Philosophy of the Social Sciences, 36(1), 3-
17.
Giudici, G., Guerini, M., & Rossi-Lamastra, C. (2017). Reward-based crowdfunding of
entrepreneurial projects: the effect of local altruism and localized social capital on
proponents’ success. Small Business Economics, 1-18.
Gofundme. (2020). Answers to common fundraising questions. Retrieved from:
https://www.gofundme.com/c/questions
Gonzales, A. L., Kwon, E. Y., Lynch, T., & Fritz, N. (2018). “Better everyone should know our
business than we lose our house”: Costs and benefits of medical crowdfunding for
support, privacy, and identity. New Media & Society, 20(2), 641-658.
Goode, E. (1992). Collective behavior. Fort Worth, TX: Harcourt Brace Jovanovich.
Google Books Ngram Viewer (2017). Retrieved from: https://books.google.com/ngrams.
Granovetter, M. S. (1973). The strength of weak ties. American Journal of Sociology, 78(6),
1360-1380.
Granovetter, M. (1985). Economic action and social structure: The problem of
embeddedness. American Journal of Sociology, 91(3), 481-510.

97
Greenberg, M. D., Pardo, B., Hariharan, K., & Gerber, E. (2013, April). Crowdfunding support
tools: predicting success & failure. In CHI'13 Extended Abstracts on Human Factors in
Computing Systems (pp. 1815-1820). ACM.
Greenberg, J., & Mollick, E. (2015). Leaning in or leaning on? Gender, homophily, and activism
in crowdfunding. In Academy of Management Proceedings.
Greenberg, J., & Mollick, E. (2017). Activist choice homophily and the crowdfunding of female
founders. Administrative Science Quarterly, 62(2), 341-374.
Gupta, A. K., Smith, K. G., & Shalley, C. E. (2006). The interplay between exploration and
exploitation. Academy of Management Journal, 49(4), 693-706.
Hanifan, L. J. (1916). The rural school community center. The Annals of the American Academy
of Political and Social Science, 67(1), 130-138.
Hardin, G. (1998). Extensions of" the tragedy of the commons". Science, 280(5364), 682-683.
He, Z. L., & Wong, P. K. (2004). Exploration vs. exploitation: An empirical test of the
ambidexterity hypothesis. Organization Science, 15(4), 481-494.
Hellmann, J. (2020). Survey: 20 million Americans have crowdfunded to help pay medical bills.
Retrieved from: https://thehill.com/policy/healthcare/483668-survey-20-million-
americans-have-crowdfunded-to-help-pay-medical-bills
Holton, R. J. (1978). The crowd in history: Some problems of theory and method. Social History,
3(2), 219-233.
Hong, Y., Hu, Y., & Burtch, G. (2015). How does social media affect contribution to public
versus private goods in crowdfunding campaigns? In 2015 International Conference on
Information Systems: Exploring the Information Frontier, ICIS 2015 Association for
Information Systems.

98
Horkheimer, M., & Adorno, T. W. (1972). Dialectic of Enlightenment. 1944. Trans. John
Cumming. New York: Herder.
Horvát, E. Á., Uparna, J., & Uzzi, B. (2015, August). Network vs market relations: The effect of
friends in crowdfunding. In Proceedings of the 2015 IEEE/ACM international conference
on advances in social networks analysis and mining 2015 (pp. 226-233).
Howe, J. (2006). The rise of crowdsourcing. Wired magazine, 14(6), 1-4.
Hunter, D. R., Goodreau, S. M., & Handcock, M. S. (2008). Goodness of fit of social network
models. Journal of the American Statistical Association, 103(481), 248-258.
H.R.4564. (2014). Equity Crowdfunding Improvement Act of 2014. Retrieved from:
https://www.govtrack.us/congress/bills/113/hr4564
Ibarra, H. (1997). Paving an alternative route: Gender differences in managerial networks. Social
Psychology Quarterly, 91-102.
Jian, L., Usher, N., (2014). Crowd-funded journalism. Journal of Computer-Mediated
Communication, 19, 155-170.
Jopson, B., Why are so many Americans crowdfunding their healthcare? Available at:
https://www.ft.com/content/b99a81be-f591-11e7-88f7-5465a6ce1a00
Kenworthy, N. (2018). Drone philanthropy: Global health crowdfunding and the anxious futures
of partnership. Medical Anthropology Theory, 5(2), 168-87.
Kenworthy, N. J. (2019). Crowdfunding and global health disparities: An exploratory conceptual
and empirical analysis. Globalization and Health, 15(1), 1-13.
Kenworthy, N., Dong, Z., Montgomery, A., Fuller, E., & Berliner, L. (2020). A cross-sectional
study of social inequities in medical crowdfunding campaigns in the United States. PloS
One, 15(3), e0229760.

99
Kim, T. Y., Oh, H., & Swaminathan, A. (2006). Framing interorganizational network change: A
network inertia perspective. Academy of Management Review, 31(3), 704-720.
Kirmani, A., & Rao, A. R. (2000). No pain, no gain: A critical review of the literature on
signaling unobservable product quality. Journal of Marketing, 64(2), 66-79.
Kittur, A., Nickerson, J. V., Bernstein, M., Gerber, E., Shaw, A., Zimmerman, J., ... & Horton, J.
(2013, February). The future of crowd work. In Proceedings of the 2013 conference on
Computer supported cooperative work (pp. 1301-1318).
Klein, K. J., & Kozlowski, S. W. (2000). Multilevel theory, research, and methods in
organizations: Foundations, extensions, and new directions. Jossey-Bass.
Kossinets, G., & Watts, D. J. (2009). Origins of homophily in an evolving social
network. American Journal of Sociology, 115(2), 405-450.
Krackhardt, D. (1992). The Strength of Strong Ties: The Importance of Philos in Organizations.
In N. Nohria & R. Eccles (eds.), Networks and Organizations: Structure, Form, and
Action. Boston: Harvard Business School Press, pp. 216–239.
Kulish, N. (2020) “People need immediate relief”, and online donors make it happen. Retrieved
from: https://www.nytimes.com/2020/03/16/business/coronavirus-bills-charity.html
Latour, B. (1987). Science in action: How to follow scientists and engineers through society.
Harvard University press.
Latour, B. (2005). Reassembling the social: An introduction to actor-network-theory. Oxford
University Press.
Lavie, D., & Rosenkopf, L. (2006). Balancing exploration and exploitation in alliance
formation. Academy of Management Journal, 49(4), 797-818.

100
Lazarsfeld, P. F., & Merton, R. K. (1954). Friendship as a social process: A substantive and
methodological analysis. Freedom and control in modern society, 18(1), 18-66.
Lazer, D., Pentland, A. S., Adamic, L., Aral, S., Barabasi, A. L., Brewer, D., ... & Jebara, T.
(2009). Life in the network: the coming age of computational social science. Science
(New York, NY), 323(5915), 721.
Le Bon, G. (1897). The Crowd: A study of the popular mind. T. Fisher Unwin.
Lefkowitz, M. (2020). Crowdsourcing app aims to fill gaps in coronavirus data. Retrieved from:
https://news.cornell.edu/stories/2020/03/crowdsourcing-app-aims-fill-gaps-coronavirus-
data
Leonardi, P. M. (2009). Crossing the implementation line: The mutual constitution of technology
and organizing across development and use activities. Communication Theory, 19(3),
278-310.
Leonardi, P. M. (2011). When flexible routines meet flexible technologies: Affordance,
constraint, and the imbrication of human and material agencies. MIS Quarterly, 35, 147-
167.
Leonardi, P. M. (2012). Materiality, sociomateriality, and socio-technical systems: What do
these terms mean? How are they different? Do we need them. Materiality and
organizing: Social interaction in a technological world, 25.
Levinthal, D. A., & March, J. G. (1993). The myopia of learning. Strategic Management
Journal, 14(S2), 95-112.
Lewis, G. (2011). Asymmetric information, adverse selection and online disclosure: The case of
eBay motors. American Economic Review, 101(4), 1535-46.

101
Liang, X., Hu, X., & Jiang, J. (2020). Research on the Effects of Information Description on
Crowdfunding Success within a Sustainable Economy—The Perspective of Information
Communication. Sustainability, 12(2), 650.
Lin, M., & Viswanathan, S. (2016). Home bias in online investments: An empirical study of an
online crowdfunding market. Management Science, 62(5), 1393-1414.
Lin, N. (2002). Social capital: A theory of social structure and action (Vol. 19). Cambridge
University Press.
Lin, N., Cook, K., & Burt, R. (2001). Social capital: Theory and research. Aldine de Gruyter.
Liu, Z., & Yang, Z. Y. L. (2004). Factors influencing distance-education graduate students' use
of information sources: A user study. The Journal of Academic Librarianship, 30(1), 24-
35.
Liu, J., Yang, L., Wang, Z., & Hahn, J. (2015). Winner takes all? The “blockbuster effect” in
crowdfunding platforms. Retrieved from:
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3007075
Lozares, C., Verd, J. M., Cruz, I., & Barranco, O. (2014). Homophily and heterophily in personal
networks. From mutual acquaintance to relationship intensity. Quality & Quantity, 48(5),
2657-2670.
Lu, L. (2014). How Social and Human Capital Create Financial Capital in Crowdfunding
Projects (Doctoral dissertation, University of Southern California).
Lukk, M., Schneiderhan, E., & Soares, J. (2018). Worthy? Crowdfunding the Canadian Health
Care and Education Sectors. Canadian Review of Sociology/Revue canadienne de
sociologie, 55(3), 404-424.

102
Lundin, R. A., & Söderholm, A. (1995). A theory of the temporary organization. Scandinavian
Journal of Management, 11(4), 437-455.
Mamykina, L., Manoim, B., Mittal, M., Hripcsak, G., & Hartmann, B. (2011, May). Design
lessons from the fastest Q&A site in the west. In Proceedings of the SIGCHI conference
on Human factors in computing systems (pp. 2857-2866). ACM.
March, J. G. (1991). Exploration and exploitation in organizational learning. Organization
Science, 2(1), 71-87.
Margolin, D., & Markowitz, D. M. (2018). A multitheoretical approach to big text data:
comparing expressive and rhetorical logics in Yelp reviews. Communication Research,
45(5), 688-718.
Martin, E. D. (1920). The behavior of crowds: A psychological study. Harper.
Matai, S. A. (2006). Network forms of organizing. Retrieved from:
http://matei.org/ithink/2006/04/05/networx-forms-of-organizing-a-mtml-perspective/
McCarthy, J. D., & Zald, M. N. (1977). Resource mobilization and social movements: A partial
theory. American Journal of Sociology, 82(6), 1212-1241.
McClelland, J. S. (2010). The Crowd and the Mob (Routledge Revivals): From Plato to Canetti.
Routledge.
McKenny, A. F., Allison, T. H., Ketchen, D. J., Short, J. C., & Ireland, R. D. (2017). How
Should Crowdfunding Research Evolve? A Survey of the Entrepreneurship Theory and
Practice Editorial Board. Entrepreneurship Theory and Practice, 41(2), 291-304.
McPhail, C. (1989). Blumer's theory of collective behavior: the development of a non-symbolic
interaction explanation. The Sociological Quarterly, 30(3), 401-423.
McPhail, C. (2017). The myth of the madding crowd. Routledge.

103
McPherson, M., Smith-Lovin, L., & Cook, J. M. (2001). Birds of a feather: Homophily in social
networks. Annual Review of Sociology, 27(1), 415-444.
Meer. J., (2014). Effects of the price of charitable giving: Evidence from an online crowdfunding
platform. Journal of Economic Behavior and Organization, 103, 113-124.
Melucci, A. (1996). Challenging codes: Collective action in the information age. Cambridge
University Press.
Miles, R. E., & Snow, C. C. (1992). Causes of failure in network organizations. California
Management Review, 34(4), 53-72.
Mollick, E. (2014). The dynamics of crowdfunding: An exploratory study. Journal of Business
Venturing, 29(1), 1-16.
Monge, P. R., & Contractor, N. S. (2003). Theories of communication networks. Oxford
University Press.
Monge, P. R., Fulk, J., Kalman, M. E., Flanagin, A. J., Parnassa, C., & Rumsey, S. (1998).
Production of collective action in alliance-based interorganizational communication and
information systems. Organization Science, 9(3), 411-433.
Moscovici, S. (1985). The age of the crowd: A historical treatise on mass psychology.
Cambridge University Press.
Mukherjee, A., Xiao, P., Wang, L., & Contractor, N. (2018). Does the opinion of the crowd
predict commercial success? Evidence from Threadless. In Academy of Management
Proceedings (Vol. 2018, No. 1, p. 12728). Briarcliff Manor, NY 10510: Academy of
Management.

104
Muller, M., Geyer, W., Soule, T., & Wafer, J. (2014, February). Geographical and organizational
distances in enterprise crowdfunding. In Proceedings of the 17th ACM conference on
Computer supported cooperative work & social computing (pp. 778-789).
Müller, M., & Schurr, C. (2016). Assemblage thinking and actor‐network theory: conjunctions,
disjunctions, cross‐fertilisations. Transactions of the Institute of British
Geographers, 41(3), 217-229.
Nahapiet, J., & Ghoshal, S. (1998). Social Capital, Intellectual Capital, and the Organizational
Advantage. The Academy of Management Review, 23(2), 242-266.
Nelson, R. E. (1989). The strength of strong ties: Social networks and intergroup conflict in
organizations. Academy of Management Journal, 32(2), 377-401.
Norris, P. (2002). The Bridging and Bonding Role of Online Communities. The Harvard
International Journal of Press/Politics, 7(3), 3–13.
Ognyanova, K., & Monge, P. (2013). A multitheoretical, multilevel, multidimensional network
model of the media system: Production, content, and audiences. Annals of the
International Communication Association, 37(1), 67-93.
Oh, P., & Monge, P. (2013). The functionality of social tagging as a communication
system. International Journal of Communication, 7, 28
Olson, M. (1965). Logic of collective action: Public goods and the theory of groups (Harvard
economic studies. v. 124). Harvard University Press.
Özdemir, V., Faris. J., & Srivastava, S. (2015). Crowdfunding 2.0: the next‐generation
philanthropy. EMBO Reports, 16(3), 267-271.
Orlikowski, W. J. (1992). The duality of technology: Rethinking the concept of technology in
organizations. Organization Science, 3(3), 398-427.

105
Orlikowski, W. J. (2007). Sociomaterial practices: Exploring technology at work. Organization
Studies, 28(9), 1435-1448.
Ostrom, E. (1998). A behavioral approach to the rational choice theory of collective action:
Presidential address, American Political Science Association, 1997. American Political
Science Review, 92(1), 1-22.
Park, R. (1972). The crowd and the public, and other essays. Chicago: University of Chicago
Press.
Payne, J. (2012). Teaching ethnography for user experience: A workshop on Occupy Wall Street.
Retrieved from http://ethnographymatters.net/2012/08/31/teachingethnography-for-user-
experience-a-workshop-on-occupy-wall-street
Pew Research Center (2016). Shared, collaborative, and on demand: The new digital economy.
Retrieved from: http://www.pewinternet.org/2016/05/19/collaborative-crowdfunding-
platforms/
Podolny, J. M., & Page, K. L. (1998). Network forms of organization. Annual Review of
Sociology, 24(1), 57-76.
Portes, A. (1998). Social capital: Its origins and applications in modern sociology. Annual
Review of Sociology, 24(1), 1-24.
Powell, W. W. (1990). Neither market nor hierarchy: Network forms of organization. Research
in Organizational Behavior, 12, 295-336.
Proelss, J., Schweizer, D., & Zhou, T. (2020). Economics of philanthropy—evidence from health
crowdfunding. Small Business Economics, Forthcoming.
Putnam, R. D. (1995). Bowling alone: America's declining social capital. Journal of
Democracy, 6(1), 65-78.

106
Reagle, J. M. (2010). Good faith collaboration: The culture of Wikipedia. MIT Press.
Ring, P. S., & Van de Ven, A. H. (1994). Developmental processes of cooperative
interorganizational relationships. Academy of Management Review, 19(1), 90-118.
Robins, G., Pattison, P., Kalish, Y., & Lusher, D. (2007). An introduction to exponential random
graph (p*) models for social networks. Social Networks, 29(2), 173-191.
Robison, L. J., Schmid, A. A., & Siles, M. E. (2002). Is social capital really capital?. Review of
Social Economy, 60(1), 1-21.
Rogers, E. M., & Bhowmik, D. K. (1970). Homophily-heterophily: Relational concepts for
communication research. Public Opinion Quarterly, 34(4), 523-538.
Ross, S. A. (1977). The determination of financial structure: The incentive signaling structure.
Bell Journal of Economics, 8, 23-40.
Sargent, M., Clark, J. A., Monge, P., & Fulk, J. (2018). The impacts of spanning implicit
categories in online markets: Mapping stylistic categories from keywords. Paper
presented at the 68th annual conference of the International Communication Association,
Prague, Czech Republic.
Sayes, E. (2014). Actor–Network Theory and methodology: Just what does it mean to say that
nonhumans have agency? Social Studies of Science, 44(1), 134-149.
Shen, C., Monge, P., & Williams, D. (2014). Virtual brokerage and closure: Network structure
and social capital in a massively multiplayer online game. Communication
Research, 41(4), 459-480.
Shepherd, M. (2020). Crowdfunding statistics (2020): Market size and growth. Retrieved from:
https://www.fundera.com/resources/crowdfunding-statistics

107
Shirky, C. (2008). Here comes everybody: The power of organizing without organizations.
Penguin.
Singleton, V., & Michael, M. (1993). Actor-networks and ambivalence: General practitioners in
the UK cervical screening programme. Social Studies of Science, 23(2), 227-264.
Skirnevskiy, V., Bendig, D., & Brettel, M. (2017). The influence of internal social capital on
serial creators’ success in crowdfunding. Entrepreneurship Theory and Practice, 41(2),
209-236.
Snijders, T. A. (2002). Markov chain Monte Carlo estimation of exponential random graph
models. Journal of Social Structure, 3(2), 1-40.
Snijders, T. A., Van de Bunt, G. G., & Steglich, C. E. (2010). Introduction to stochastic actor-
based models for network dynamics. Social Networks, 32(1), 44-60.
Snow, D. A., & Owens, P. B. (2013). Crowds (Gatherings) and Collective Behavior
(Action). The Wiley-Blackwell Encyclopedia of Social and Political Movements.
Snyder, J., Crooks, V. A., Mathers, A., & Chow-White, P. (2017). Appealing to the crowd:
ethical justifications in Canadian medical crowdfunding campaigns. Journal of Medical
Ethics, 43(6), 364-367.
Spence, M. (1978). Job market signaling. In Uncertainty in Economics (pp. 281-306).
Spence, M. (2002). Signaling in retrospect and the informational structure of markets. American
Economic Review, 92(3), 434-459.
Stanko, M. A., & Henard, D. H. (2017). Toward a better understanding of crowdfunding,
openness and the consequences for innovation. Research Policy, 46(4), 784-798.

108
Starbird, K. (2012). Crowdwork, crisis and convergence: How the connected crowd organizes
information during mass disruption events (Doctoral dissertation, University of Colorado
at Boulder).
Statista, (2017). Number of monthly active Facebook users worldwide as of 4th quarter 2017 (in
millions). Retrieved from: https://www.statista.com/statistics/264810/number-of-
monthly-active-facebook-users-worldwide/
Stivala, A., Robins, G., & Lomi, A. (2020). Exponential random graph model parameter
estimation for very large directed networks. PloS One, 15(1), e0227804.
Stiver, A., Barroca, L., & Minocha, S., et al. (2015). Civic crowdfunding research: Challenges,
opportunities, and future agenda. New Media & Society 17(2), 249-271.
Stohl, C. (2014). Crowds, clouds, and community. Journal of Communication, 64, 1-19.
Surowiecki, J. (2005). The wisdom of crowds. Anchor.
Syrek, S. (2012). Why am I talking? Reflecting on language and privilege at Occupy Wall Street.
Critical Quarterly, 54(2), 72–75.
Tarde, G. (1903). The laws of imitation. H. Holt.
Thies, F., Wessel, M., & Benlian, A. (2014). Understanding the dynamic interplay of social buzz
and contribution behavior within and between online platforms: Evidence from
crowdfunding. In The 35th International Conference on Information Systems. ICIS 2014:
Building a Better World through Information Systems. Association for Information
Systems. AIS Electronic Library (AISeL).
Tilly, C., Tilly, L., & Tilly, R. (1975) The Rebellious Century, 1830–1930. Harvard University
Presss.

109
Tsai, W., & Ghoshal, S. (1998). Social capital and value creation: The role of intrafirm
networks. Academy of Management Journal, 41(4), 464-476.
Tufekci, Z., & Wilson, C. (2012). Social media and the decision to participate in political protest:
Observations from Tahrir Square. Journal of Communication, 62(2), 363-379.
Turner, R. H., & Killian, L. M. (1957). Collective behavior (Vol. 3). Englewood Cliffs, NJ:
Prentice-Hall.
Uzzi, B. (1997). Social structure and competition in interfirm networks: The paradox of
embeddedness. Administrative Science Quarterly, 35-67.
Uzzi, B. (2008). A social network's changing statistical properties and the quality of human
innovation. Journal of Physics A: Mathematical and Theoretical, 41(22), 224023.
Váques, C. & Plourde., M. (2017). Materiality and organizing. In Scott, C. R., & Lewis, L.(Eds.),
The International Encyclopedia of Organizational Communication. John Wiley & Sons,
Inc. (pp 1-15).
Wang, P., Robins, G., Pattison, P., & Koskinen, J. (2014). Program for the simulation and
estimation of (p*) exponential random graph models for multilevel networks. Retrieved
from:
https://static1.squarespace.com/static/57a1436215d5dbbcd2031828/t/581ac475e58c6
2432be83acf/1478149242644/MPNetManual.pdf
Wang, P., Robins, G. L., Pattison, P., & Lazega, E. (2013). Exponential random graph models
for multilevel networks. Social Networks, 35(1), 96–115.
Wang, P., Robins, G., Pattison, P., & Lazega, E. (2016). Social selection models for multilevel
networks. Social Networks, 44, 346–362.

110
Wang, P., Robins, G., Matous, P. (2016). Multilevel network analysis using ERGM and its
extension. In E. Lazega & T.A.B. Snijders (Eds.), Multilevel network analysis for the
social sciences: Theory, methods and applications (pp. 125–143). Cham, Germany:
Springer.
Wash, R., Solomon, J., 2014. Coordinating donors on crowdfunding websites. Proceedings of the
17th ACM Conference on Computer Supported Cooperative Work & Social Computing,
38-48.
Weiss, A. (1995). Human capital vs. signalling explanations of wages. Journal of Economic
perspectives, 9(4), 133-154.
Wheat, R.E., Wang, Y., Byrnes, J.E., 2013. Raising money for scientific research through
crowdfunding. Trends in Ecology and Evolution, 28, 71-72.
Whittle, A., & Spicer, A. (2008). Is actor network theory critique? Organization Studies, 29(4),
611-629.
Xu, A., Yang, X., Rao, H., Fu, W. T., Huang, S.W., & Bailey, B.P., (2014). Show me the money!
An analysis of project updates during crowdfunding campaigns. In Proceedings of the
SIGCHI conference on human factors in computing systems (pp. 591-600). ACM.
Xu, L. Z. (2018). Will a digital camera cure your sick puppy? Modality and category effects in
donation-based crowdfunding. Telematics and Informatics, 35(7), 1914-1924.
Xu, Y. (2019). The evolution of multidimensional and multilevel networks in online
crowdsourcing (Doctoral dissertation, University of Southern California).
Yurieff, K., (2017). Facebook launches personal fundraising tool. Available at:
http://money.cnn.com/2017/03/30/technology/facebook-personal-fundraising/index.html

111
Zappa, P., & Lomi, A. (2015). The analysis of multilevel networks in organizations: models and
empirical tests. Organizational Research Methods, 18(3), 542-569.
Zheng, H., Li, D., Wu, J., & Xu, Y. (2014). The role of multidimensional social capital in
crowdfunding: A comparative study in China and US. Information &
Management, 51(4), 488-496.
Zipf, G. K. (1949/2016). Human behavior and the principle of least effort: An introduction to
human ecology. Ravenio Books.
Zomeren, M., Postmes, T., & Spears, R. (2008). Toward an integrative social identity model of
collective action: A quantitative research synthesis of three socio-psychological
perspectives. Psychological Bulletin, 134(4), 504.
Zuckerman, E. W. (2017). The categorical imperative revisited: Implications of categorization as
a theoretical tool. Research in the Sociology of Organizations, 51, 31–68.
Zvilichovsky, D., Inbar, Y., & Barzilay, O. (2015). Playing both sides of the market: Success and
reciprocity on crowdfunding platforms. Available at SSRN
https://ssrn.com/abstract=2304101

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

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

PDF

The evolution of scientific collaboration networks

PDF

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

PDF

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

PDF

Staying ahead of the digital tsunami: strategy, innovation and change in public media organizations

PDF

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

Asset Metadata

Creator Xu, Zhiming (author)

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

School Annenberg School for Communication

Degree Doctor of Philosophy

Degree Program Communication

Publication Date 07/14/2020

Defense Date 04/27/2020

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

Tag crowd,crowd-enabled organizations,crowdfunding,donation-based crowdfunding,exploration/exploitation tradeoff,homophily,MTML,multidimensional networks,multilevel networks,networks,OAI-PMH Harvest,signaling,social capital

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Fulk, Janet (committee chair), Monge, Peter (committee member), Sargent, Matthew (committee member)

Creator Email lawrencexu110@gmail.com,zhimingx@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c89-330923

Unique identifier UC11663268

Identifier etd-XuZhiming-8680.pdf (filename),usctheses-c89-330923 (legacy record id)

Legacy Identifier etd-XuZhiming-8680.pdf

Dmrecord 330923

Document Type Dissertation

Rights Xu, Zhiming

Type texts

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

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA