University of Southern California Dissertations and Theses

Governance networks: internal governance structure, boundary setting and effectiveness

(USC Thesis Other)

Governance networks: internal governance structure, boundary setting and effectiveness

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content GOVERNANCE NETWORKS:
INTERNAL GOVERNANCE STRUCTURE, BOUNDARY SETTING AND
EFFECTIVENESS
Copy Right 2015
By
Weijie Wang
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
(POLICY, PLANNING AND DEVELOPMENT)
August 2015
Weijie Wang
Dedication
To my parents, Ji de Wang and Chunhua Liu,
for their unconditional support and love that made me who I am today
ii
Acknowledgement
I am deeply grateful to the members of my dissertation committee: Terry Cooper, Shui Yan Tang
and Peer Fiss. When I was reading Professor Cooper's book in the library of Wuhan University
back in 2006, I never expected that, three years later, I would become a student of his. He has
been extremely supportive of all my endeavors in the past six years. He led me to the research on
homeowners' movement in China and then gave me the maximum degree of freedom to pursue
the research topics that I am interested in. Professor Tang is a role model for many Asian PhD
students at the Price School. He not only gave me invaluable comments on my dissertation but
also shared with me the experience of doing research. Professor Fiss is an extraordinary scholar
and one of the best professors that I have had at USC. He always listens to me and read my work
carefully. I can easily tell how much thought and time he has put in my papers. He also helped
me tremendously to use the fsQCA method in the dissertation.
In a dissertation that studies collaboration, I also want to thank the professors who taught me and
thus participated in the collaboration of training me to be a scholar: Elizabeth Graddy, Tridib
Banerjee, Dowell Myers, David Sloane, Paul Lichterman, John Odell, Tom Valente, Tim Biblarz,
Peter Monge, Peter Radchenko, Juliet Musso, and Kiros Berhane. Although they are not directly
involved in this dissertation, what they taught me over the past six years is indispensable for this
dissertation.
iii
My special thanks go to Hui Li and Henry Wai-Hang Yee. I am fortunate to have these wonderful
friends who selflessly read and critique my work, brainstorm and share ideas with me. I greatly
appreciate James Polk for editing the dissertation. I know how much patience it is needed to go
through the dissertation line by line and clear all grammar mistakes. I also want to thank
Professor Youhong Chen of Renmin University for helping me to arrange interviews. I deeply
appreciate those whom I interviewed in Beijing for their assistance and understanding. They
shared their stories and perspectives with me and answered my questions with great patience.
I am forever indebted to the University of Southern California and USC Price School of Public
Policy for giving me the opportunity and generous financial support to study at the one of the
finest universities in the US. I was amazed each day by the valuable academic resources and
fantastic culture on the campus. The past six years at USC not only made me a scholar but also a
die-hard USC football fan. I am extremely proud to be a graduate of this great university and a
member of the Trojan Family. I also want to acknowledge the USC Graduate School for
awarding me the Dissertation Completion Fellowship, which allows me to focus on my
dissertation writing in the past year. The USC US-China Institute offered me a fieldwork grant in
the summer of2011, which became the seed of this dissertation.
Last but not least, I would like to dedicate this dissertation to my parents, Jide Wang and
Chunhua Liu. They are my best teachers. Their unconditional love and support made me who I
am today.
iv
Table of Contents
Dedication .................................................................................................................................. ii
Acknowledgement ................................................................................................................... iii
List ofTables ............................................................................................................................ vii
List of Figures ........................................................................................................................ viii
Abstract ..................................................................................................................................... ix
Chapter 1 : Introduction ............................................................................................................. 1
Governance Networks: An Overview ..................................................................................... 3
An Overview of the Dissertation .......................................................................................... 13
Chapter 2 : Power Balance and Institutionalization ................................................................. 20
Introduction .......................................................................................................................... 20
Theoretical Framework ........................................................................................................ 22
Research Context .................................................................................................................. 29
Research Design and Data Collection .................................................................................. 34
A Typology of Governance Structure ................................................................................... 35
Discussion and Conclusion .................................................................................................. 5 2
Chapter 3: The Boundary Setting of Collaborative Governance Regimes ............................. 58
Introduction .......................................................................................................................... 58
Theoretical Framework ........................................................................................................ 60
Research Context .................................................................................................................. 68
Research Methods, Data and Operationalization of Variables ............................................. 71
Sampling ........................................................................................................................... 73
Data collection .................................................................................................................. 74
Operationalization of variables ......................................................................................... 75
Analysis and Results ............................................................................................................ 78
Discussion and Conclusion .................................................................................................. 84
Contribution and Limitations ............................................................................................... 91
Appendix One ................................................................................................................... 93
Chapter 4: Exploring the Determinants of Network Effectiveness ........................................ 97
Introduction .......................................................................................................................... 97
Theoretical Framework ........................................................................................................ 99
Research Context: Neighborhood Governance Networks .................................................. 105
Research Methods, Data and Operationalization of Variables ........................................... 109
v
Research Design .............................................................................................................. 109
Sampling and Data Collection ........................................................................................ 110
Operationalization of the outcome and causal conditions .............................................. 111
Analysis and Results .......................................................................................................... 120
Summary statistics .......................................................................................................... 120
Linear Modeling .............................................................................................................. 121
Analysis with the fsQCAmethod ....................................................................................... 123
Discussion and Conclusion ................................................................................................ 127
Appendix One ..................................................................................................................... 137
Appendix Two .................................................................................................................... 141
Appendix Three .................................................................................................................. 143
Calibration ....................................................................................................................... 143
Robust Check .................................................................................................................. 146
Chapter 5 : Conclusion ........................................................................................................... 15 2
Theoretical Contributions ................................................................................................... 152
Overall Theoretical Reflection ........................................................................................... 155
Methodological Contributions ............................................................................................ 159
Policy Design hnplications ................................................................................................ 161
Bibliography .......................................................................................................................... 166
vi
List of Tables
Table 1.1 Comparison Between Three Types OfNetworks ....................................................... 9
Table 2.1 Predicting the Governance Structures of All Governance Networks ....................... 54
Table 3.1 Descriptive Statistics ................................................................................................ 79
Table 3.2 Configurations For The Inclusion Of Civic Organizations ...................................... 82
Table 3.3 Configurations for The Exclusion OfHOAs ........................................................... 84
Table 3.4 Results of Robustness Checks .................................................................................. 96
Table 4.1 Tests of Statistical Significance with Bootstrap Standard Errors ........................... 122
Table 4.2 Configurations Leading To Network Effectiveness ............................................... 125
Table 4.3 Configurations Leading To Network Effectiveness ............................................... 126
Table 4.4 Configurations Leading To Network Ineffectiveness ............................................ 127
Table 4. 5 Multitrait - Multimethod Correlation Matrix ........................................................ 140
Table 4.6 Summary Statistics ................................................................................................. 141
Table 4.7 Network Density And Centralization ..................................................................... 141
Table 4. 8 Correlation Coefficients Between Variables .......................................................... 142
Table 4.9 Results of Robust Check ........................................................................................ 149
Table 4.10 Truth Table For The Analysis of Network Effectiveness ..................................... 150
Table 4.11. Truth Table for the Analysis of Network Effectiveness ...................................... 150
Table 4.12 Truth Table for the Analysis of Network Ineffectiveness .................................... 151
vii
List of Figures
Figure 2.1 A Typology of Internal Governance Structures ...................................................... 37
Figure 2.2 The Network of Shang-Di Neighborhood ............................................................. .41
Figure 2.3 The Network of Feng-Dan Neighborhood ............................................................ .45
Figure 2.4 The Network of Rong-Feng Neighborhood ........................................................... 49
Figure 2.5 The Network of Chao-Yang Garden ....................................................................... 52
Figure 4.1 The Provan -Milward Model of Network Effectiveness ...................................... 100
viii
Abstract
This dissertation uses a mix-methods approach, which includes regression analysis, fuzzy set
Qualitative Comparative Analysis (fsQCA), social network analysis and constant comparative
method, to explore the following research questions: what factors influence the
inclusion/exclusion of certain organizations in governance? What may explain different internal
governance structures? What factors may affect the effectiveness of governance networks? This
dissertation is based on neighborhood governance networks in Beijing, China. I collected data of
22 neighborhood networks, each of which consists of public, business and civic organizations.
This provides an ideal environment to address the above research questions.
The first chapter is an introduction chapter that reviews the concept of governance networks, and
then provides an overview of the entire dissertation. Chapter 2 uses resource dependence theory
and institutional theory to explain different internal governance structures to develop a new
framework to explain how different types of governance structures are formed. The balance of
power and degree of institutionalization are considered to be two important dimensions to
explain how organizations interact with one another. The chapter develops a typology of
governance structures. Four ideal types of governance structures were identified: shared
governance, inertial governance, insurgent coalition domination and lead organization
governance. This chapter explains the power dynamics in each governance structure and
illustrates with cases of neighborhood governance in Beijing.
ix
Chapter 3 studies the boundary settings of governance networks. The dynamic process of
inclusion and exclusion of certain organizations in governance networks is defined as boundary
setting. Drawing on the literature from general organization theory, urban regime theory,
nonprofit management and network research, this chapter attempts to transcend both the
resource-based explanation and power-based explanation. This chapter develops a
capacity-threat framework to explain the boundary setting of collaborative governance. When
assessing potential members to include, extant members will consider not only resources that
they may bring but also the threats that they may pose to their interests. Extant members
especially those that benefit from the status quo will try to block new members if they are
perceived as serious threats. However, this does not mean that these organizations will always be
excluded. To be successfully included, these potential members have to have high organizational
capacity so that they can overcome the resistance from extant members. If they lack the
organizational capacity, for example, they lack strong leadership or have a low degree of
formalization, then they will likely to be excluded. Therefore, the boundary setting of
governance networks is a political and dynamic process.
Chapter 4 examines the determinants of the effectiveness of governance networks. Evaluating
network effectiveness and studying its determinants have been an important topic in the network
research. Based on the models proposed by Provan and Milward (1995) and Provan and Kenis
(2008), this chapter employed a mixed-methods approach to study the determinants of the
effectiveness of governance networks. Linear regression was used to identify independent
x
variables that exert statistically significant influence over network effectiveness, and the fuzzy
set Qualitative Comparative Analysis was used to investigate the complex interactions between
explanatory variables. The chapter revealed the causal complexities of network effectiveness: the
analysis found different but functionally equivalent configurations of causal conditions that lead
to network effectiveness, and showed that configurations of factors leading to network
effectiveness are different from those leading to network ineffectiveness. The results also
suggested that network structural characteristics such as network centralization and density are
neither sufficient nor necessary conditions for network effectiveness. However, in contract to
Provan and Milward's (1995) findings, the results suggested that network density is more
important than network centralization in affecting effectiveness in small networks. Resource
munificence was identified as an "almost always" necessary condition for network effectiveness.
To summarize, this dissertation makes several important theoretical contributions and advances
our understating of collaborative governance. First, it proposes a new theoretical framework to
explain different types of internal governance structures of networks, which has often been
neglected in the current literature. The framework is especially suitable for explaining the
internal governance structures of serendipitous networks that we often see in the real world.
Second, it offers the first systematic study on the boundary setting of governance networks. A
capacity-threat framework was proposed to explain the dynamic process of boundary setting
based on the systematic investigation of neighborhood governance networks in Beijing. Third,
this dissertation develops a configurational theory of network effectiveness and better explores
xi
the causal complexities of network effectiveness with a mixed methods approach. It sheds new
light on the relationship between network structures and network effectiveness.
The dissertation also makes a number of methodological contributions. First, it employs an
innovative mixed-methods approach, which better utilizes the strengths of each method to
address different research questions. One method that is particularly worth mentioning is the
fsQCA method, which has not been widely used in public administration research, although a
few scholars have started to use it to develop configurational theories (Raab et al., 2013; Verweij
et al., 2013). The fsQCA method helps to find out the complex interactions between causal
conditions of network effectiveness and boundary setting. Second, this dissertation also uses an
innovative a multitrait-multimethod approach to measuring network effectiveness. The previous
research often used perceived network effectiveness as a proxy. However, subjective measures of
performance are prone to contamination and may contain sizeable random errors. Therefore, I
supplement the subjective evaluation with external evaluation based on photographic evidence.
This dissertation may be the first one to use the multitrait-multimethod approach to measuring
effectiveness. Using statistical techniques such as Pearson product-movement correlation
coefficients, Cronbach's alpha and Intraclass correlation coefficients (ICC), this approach gives
us a more comprehensive and less biased evaluation.
xii
Chapter 1 : Introduction
In the US, the collaboration between public, business and nonprofit organizations has been
widely accepted and supported as the way to solve a wide range of thorny policy problems, such
as AIDS pandemic, terrorism and gang control. Many of the policy problems are not only
technologically complex but also cut across conventional organizational and sectorial boundaries
(Koppenj an & Klijn, 2004 ). It is thus beyond the capability of one organization or sector to
effectively address these problems. On the other hand, political and economic resources are
widely dispersed in modern societies (Dahl, 1961). The mismatch between power and resources
makes it necessary for government to seek the assistances from business and civic organizations
in order to govern effectively.
The transformation from government to governance and the surge of governance networks
consisting of governments, business and civic organizations has become a notable phenomenon.
Governance networks exert important influences in the governing of megaregions (Innes, Booher,
& Di Vittorio, 2010), cities (Filip De Rynck & Voets, 2006) and neighborhoods (Chaskin & Garg,
1997). One notable example is Atlanta - though the composition of alliances went through
several rounds of changes over the period of 1946 to 1988, the city government had to cooperate
with different business and civic organizations so as to govern the city (Stone, 1989).
1
This dissertation does not aim to build a grand theory of collaborative governance or governance
networks. Instead, using a mix-methods approach that includes regression analysis, fuzzy set
Qualitative Comparative Analysis (fsQCA), social network analysis and constant comparative
method, this dissertation explores the following research questions: what factors influence the
inclusion/exclusion of certain organizations in governance? What may explain different internal
governance structures? What factors may affect the effectiveness of governance networks? This
dissertation is based on neighborhood governance networks in Beijing, China. I collected data of
22 neighborhood networks, each of which consists of public, business and civic organizations.
This provides an ideal environment to address the above research questions.
The first chapter is an introduction chapter that reviews the concept of governance networks, and
then provides an overview of the entire dissertation. Chapter 2 uses resource dependence theory
and institutional theory to explain different internal governance structures; Chapter 3 focuses on
the boundary setting of governance networks; Chapter 4 uses regression analysis and fsQCA to
study factors that influence network effectiveness. Chapter 5 highlights the theoretical and
methodological contributions of this dissertation and also provides some policy
recommendations.
2
Governance Networks: An Overview
Governance networks from the perspective of collaborative governance
There has been a wealth of literature on collaborative governance. For example, there are not
only some very good literature reviews that help us to gain a broad picture of the entire research
area (Ansell & Gash, 2007; Bryson, Crosby, & Stone, 2006; K. Emerson, Nabatchi, & Balogh,
2012; Provan, Fish, & Sydow, 2007) but also some empirical research on specific topics such as
interorganizational power (Choi & Robertson, 2013; Purdy, 2012), collaborative governance
processes (Thomson & Perry, 2006), inclusion/membership (Hendriks, 2008; Huxham & Vangen,
2000; Johnston, Hicks, Nan, & Auer, 2010), the effectiveness of collaboration (Provan & Kenis,
2008; Provan & Milward, 1995) and the application of collaborative governance in different
policy areas (Durant, Chun, Kim, & Lee, 2004; Koch Jr, 2005; Tett, Crowther, & O'Hara, 2003).
Collaborative governance was defined in various ways. K. Emerson et al. (2012, p. 2) provided a
definition that is broad enough to cover research on various aspects of collaborative governance:
"the processes and structures of public policy decision making and management that engage
people constructively across the boundaries of public agencies, levels of government, and/or the
public, private and civic spheres in order to carry out a public purpose that could not otherwise
be accomplished." One key characteristic of collaborative governance is the blurring of sector
boundaries between public, business and nonprofit organizations and cross-sector collaboration.
Collaborative governance and governance networks are two concepts that are closely connected.
3
S0fensen and Torfing (2005b) argued that governance networks were a particular form of
governance and a particular kind of network. The organizations involved in collaborative
governance and the interconnections between them form interorganizational structures, and these
structures are understood as governance networks or broadly defined "collaborative governance
regimes" (K. Emerson et al., 2012; Klijn, 2008; Klijn & Skelcher, 2007; Vangen, Hayes, &
Cornforth, 2014). In other words, collaborative governance is the process that takes place within
governance networks (Klijn, 2008). This is consistent with Giddens' (1979) structuration theory,
which suggests that the structural properties of social systems are often the outcome of
interactions between individuals and groups.
Governance networks from a network perspective
Any studies on networks have to answer the fundamental question of "what is a network."
Networks in the real world have a variety of forms, histories and characteristics, and they are
defined in various ways in the literature. Provan et al. (2007) commented that "a shared language
with definite, concrete meanings in the study of networks has not been developed." Some
literature perceives a cluster of organizations within certain policy areas or geographical areas as
"networks" with little consideration of the interactions between these organizations. "Network"
has thus become a loose and popular metaphor (Isett, Mergel, LeRoux, Mischen, & Rethemeyer,
2011; O'Toole & Meier, 2004), which often causes confusion. It is thus necessary to define
network clearly in order to further the development of network theories.
4
Some scholars defined networks narrowly by highlighting formal contracts or coordinating
mechanisms as the defining characteristic of networks. For example, Barringer and Harrison
(2000) defined networks as constellations of organizations that came together through the
establishment of social contracts. By comparison, some scholars especially social network
analysts defined networks rather broadly as "a set of nodes and the set of ties representing some
relationship, or lack of relationship, between the nodes" (Brass, Galaskiewicz, Greve, & Tsai,
2004). Though broad definitions were theoretically enlightening and made connections to social
network analysis which provided rich theories and methods, they lost their accuracy and made
little sense in the context of public policy and management.
Horizontality, a key structural feature of a network, has been well acknowledged by both broad
and narrow definitions. The middle-range definition of networks, based on the understanding of
the root of networks in the real world, recognized the complexities of forms that networks might
take and highlighted "interdependency" as the second key characteristic of networks (Agranoff
& McGuire, 2001; Kickert, Klijn, & Koppenjan, 1997; Koppenjan & Klijn, 2004; R. A. Rhodes,
1997). For example, O'Toole (1997) defined network as "structures of interdependence involving
multiple organizations or parts thereof, where one unit is not merely the formal subordinate of
the others in some larger hierarchical arrangement" (p.45). In public policy and management
areas, many complex policy problems such as poverty reduction and climate change were
beyond the capability of any one single public or private agency (Koppenjan & Klijn, 2004; R. A.
Rhodes, 1997). Inter-sectoral cooperation, which may take a wide variety of forms, was
5
necessary to address these "wicked problems." The mam reason that organizations formed
networks was resource dependency - network members may not have sufficient resources to
achieve their goals and thus they have to depend on others for critical resources. It is
interdependency that really connected organizations together and that greatly influenced the
strategies that each organization take in their interactions with other network members (Hardy &
Phillips, 1998).
The study of networks in public management and policy literature had a long history. F. S. Berry
et al. (2004) traced three traditions of research in the current network literature in public
administration: sociological tradition, political science tradition and public management
tradition.
Sociological research on networks could be traced back to the "gestalt" tradition of psychology
and was further developed by a group of anthropologists at Manchester University and some
sociologists at Harvard University in the 1960s (J. Scott, 2012). Research of this tradition
focused on network structure and how structure affected network outcomes. Sociological
network research also developed network analysis techniques which provided important research
tools for network researchers in public administration. Some studies in policy networks
employed network analysis techniques to visualize policy networks and to analyze the structural
characteristics and their influences (Raab, 2002; Rethemeyer & Hatmaker, 2007; Weible &
Sabatier, 2005).
6
The political science tradition focused on the study of policy networks which originated in the
study of policymaking. Generally it reflected the policymaking processes in western democracies
that involved a wide range of interest groups and government agencies (Klijn, 1996). Early
research could be traced back to Lowi's (1964) discussion of iron triangles in American policy
making - interest groups exerted important influences over federal government agencies and
congress committees. Similar concepts such as sub-systems (Freeman, 1965) and
sub-governments (Ripley & Franklin, 1984) were along this line. Recio (1992) criticized this
elitist description of policymaking and proposed issue networks as an alternative. Issue network
is a loose network of government agencies, legislative committees, interest groups, nonprofit
organization and media that are interested in a certain policy area. Rhodes (1986) defined a
policy network as a cluster or complex of organizations connected to one another by resource
dependencies.
The third research tradition studied collaborative networks which are mechanisms to provide
public services by government, private and nonprofit organizations. With the rise of the New
Public Management movement, service provision and service production were differentiated.
Governments contracted out many social services to private firms or nonprofit organizations in
order to enhance efficiency. Nowadays, government agencies rely heavily on business and
nonprofit organizations to produce social services, which is described as " hollow state"
(Milward & Provan, 2000). The interdependency between these public, private and nonprofit
organizations in service delivery makes them collaborative networks.
7
Similarly, Isett et al. (2011) argued that there were three main streams of literature on networks
in public administration: policy networks, collaborative networks and governance networks.
Policy networks and collaborative networks corresponded to the above discussion of political
science tradition and public management tradition, governance networks were considered as a
separate category.
Governance networks, or area-based policy networks (Filip De Rynck & Voets, 2006), are a web
of interdependent public, private and nonprofit organizations that work together to address a
wide range of policy problems within a certain geographical area (Klijn, Steijn, & Edelenbos,
2010; S0fensen & Torfing, 2005b). Partly due to the wide distribution of political and economic
resources in modern society, it is very difficult for governments or other entities to govern a city
or neighborhood effectively without cooperating with others. Governance networks are thus
prevalent and exert important influences over the governance of megaregions (Innes et al., 2010),
cities (Filip De Rynck & Voets, 2006) and neighborhoods (Chaskin & Garg, 1997).
Though governance networks share some common features of other networks, such as
interdependency, this type of networks had some unique characteristics as Table 1 illustrates.
Unlike policy network which is made up of intersectoral organizations that share interests in a
particular policy issue or collaborative networks which deliver public services and goods,
governance networks combine policy formulation and implementation together in the governing
of an area such as neighborhoods or cities (Isett et al., 2011; Klijn & Skelcher, 2007; Rethemeyer
8
& Hatmaker, 2007). The object of governance network is a geographical area; by comparison,
the object of policy networks is a policy area and that of collaborative networks is a public
service. Governance networks are usually naturally grown, and may have a quite low degree of
formalization. There are usually no agreements or contracts that bind these organizations
together - they just interact with one another on an ongoing process, just like the "serendipitous
network" described by (Kilduff & Tsai, 2003). Partly due to their informal nature, members of
governance networks may lack a clear common goal, and thus their internal conflicts may be
high because each member tried to maximize their own interests in their interactions with other
network members (S0fensen & Torfing, 2009).
Table 1.1 Comparison Between Three Types Of Networks
object function Goal Internal Degree of
consensus conflicts formalization
Policy A policy area Policy Medium/high Medium/low Low/medium
network formulation
Collaborativ A public service Policy high low high
e network implementation
Governance Geographical Policy low high low
network area formulation &
implementation
Note: table made by the author
Governance networks in practice
Public managers face many wicked social problems (Rittel & Webber, 1973), such as climate
change, poverty reduction and gang control. The wickedness of these problems lies not only in
technological complexity but also in the blurred organizational and sectorial boundaries
9
(Koppenj an & Klijn, 2004 ). It is beyond the capability of one organization or sector to
effectively address these wicked social problems.
Another feature of modern society is that political and economic resources are widely dispersed
(Dahl, 1961 ). Though the debate over who actually govern cities has been going on for decades,
there is little dispute over the privileged position of business organizations which have the
structural and instrumental (Harding, 2009; Lindblom, 1977). In addition, citizens who have
political resources - votes - can also play an important role in urban governance if they are
properly organized (Ferman, 1996; Kantor, Savitch, & Haddock, 1997). Due to the mismatch
between power and resources, governments have to seek the assistances from business and civic
organizations in order to govern effectively.
The cross-sectorial nature of wicked social problems and the wide dispersion of political and
economic resources make it necessary for governments to collaborate with business and civic
organizations in the process of governing (O'Toole, 1997; R. A. Rhodes, 1997). Governments are
no longer the sole center of decision-making. The transformation from government to
governance and the surge of governance networks consisting of governments, business and civic
organizations has become a notable phenomenon in the governing of cities and megaregions.
One notable example is Atlanta - though the composition of governing alliances went through
several rounds of changes over the period of 1946 to 1988, governments had to cooperate with
different business and civic organizations so as to govern the city (Stone, 1989). Governance
10
network also play a prominent role in governing megacities or megaregions such as Northern
California (Innes et al., 2010).
Governance networks arise not only in the US and other western countries but also in developing
countries such as China. One notable example is the rise of governance networks in urban
neighborhoods. Due to the entry of new actors, government is no longer the sole center of
governance in neighborhoods after China's housing reform. Homeowners who are concerned
about maintaining their property values have strong incentives to engage in neighborhood affairs;
property management firms are hired by homeowners or developers for housing maintenance,
sanitation, safety and other issues pertaining to property management. In some new
neighborhoods, developers are important stakeholders because they are owners of unsold
housing units. Street Offices and Residents Committees keep working as the street-level
government entities with new responsibilities m neighborhood governance. Urban
neighborhoods in China have thus become a place where plural actors govern together.
One key feature of neighborhood governance network is interdependence. No single actor has
the resources or capability to govern by itself; each of them has to depend on one another in the
process of governing. For example, government entities do not have the absolute power that they
used to have in urban neighborhoods because they are not property owners. Many government
operations have to get property owners' consent as a precondition. For instance, setting up a
billboard for the purpose of outreach or propaganda has to be agreed by homeowners. Depending
11
on their relationships, governments may even have to pay certain fees to HO As for the billboard.
On the other hand, homeowners are well aware of that local governments including Street
Offices can affect their property rights in various ways. For example, governments have
considerable power in issues such as urban planning and zoning, and thus can affect their
neighborhoods' surrounding environment and land use. Governments' programs in public safety
and social security also directly affect many homeowners' life quality. Therefore, HO As may not
really charge fees or obstruct government operations in the expectation that governments would
reciprocate their goodwill when needed. Though there are usually not any formal contracts,
complex interdependence connects these organizations. The interdependency makes them
constitute governance networks.
These networks combine policy making and implementation. They may work closely with one
another to make decisions and then implement the decisions. For example, to maintain
neighborhood safety and reduce property and violent crimes, HO As, Residents' Committees, and
management firms may need to work together to decide whether certain measures need to be
taken, such as installing access control systems at entrances or checking IDs manually. Once a
decision is made, management firms may be responsible for the implementing the decision,
while HO As may provide necessary financial resources and monitor the services. Residents'
Committees and Street Offices may be responsible for communicating and coordinating with
police agencies on neighborhood safety. The lack of efforts from any part would compromise the
neighborhood safety service.
12
Similar to other governance networks, interorganizational relationships are not necessarily
cooperative; in many cases, the relationships are dynamic, political and conflictive. These
organizations take different strategies of engagement in order to maximize their own interests
and influence. Organizations may choose not to cooperate in order to compromise others' efforts.
For example, partly due to the poor enforcement of laws and regulations related to property
rights, government entities sometimes cooperate with property management firms or developers
to infringe on homeowners' communal properties. They may gain huge revenue from using
communal properties without giving homeowners' compensation, and in many cases they thus
try to stop homeowners from establishing HOAs in order to keep their dominating status in
neighborhoods.
An Overview of the Dissertation
The current literature on collaborative governance and governance networks is enlightening in
many aspects, but it is not without its problems. The purpose of this dissertation is to further the
research on governance networks by resolving three understudied questions: boundary setting,
internal governance structures, and the determinants of effectiveness.
Chapter 2 focuses on how governance structures of collaborative networks came into being.
Organizations may come into collaborative governance with "different role positions and carry
different weights" (Agranoff & McGuire, 2001, p. 315). With different authority and resources,
13
organizations in networks may occupy different power positions. Powerful organizations may
dominate networks and advance their own agenda (Purdy, 2012). Klijn and Teisman (1997)
conceptualized the complex interactions between all involved actors as "games" in which actors
try to maximize their own interests. The complex interactions between organizations results in
diverse governance structures in these networks. Unfortunately the governance structures of
these networks did not receive much attention (Provan & Kenis, 2008). As Provan and Kenis
(2008) argued, people often ignore that networks vary considerably in terms of structural patterns
and assume that the structures of governance networks are undifferentiated. Provan and Kenis
(2008) identified three forms of governance structures, including shared network governance,
lead organization governance and network administrative organization governance. Each
governance structure has different power dynamics and may affect network effectiveness in
different ways. Limited research has been conducted to further Provan and Kenis' (2008)
typologies of forms of governance structures.
Provan and Kenis' (2008) typologies were based on "goal-directed" networks and may not work
for the large number of "serendipitous networks". In serendipitous networks, network change is
primarily driven by serendipity, and "network trajectories develop haphazardly from the
interactions of individual actors" (Kilduff & Tsai, 2003, p. 89). This chapter attempts to propose
a new framework to explain what factors affect the governance structures of serendipitous
networks. This knowledge is crucial for designing successful collaborative governance regimes
(Ansell & Gash, 2007; K. Emerson et al., 2012).
14
This chapter develops a new framework to explain how different types of governance structures
are formed. The balance of power and domain consensus are considered to be two important
dimensions to explain how organizations interact with one another. The chapter develops a
typology of governance structures. Four ideal types of governance structures were identified:
shared governance, inertial governance, insurgent coalition domination and lead organization
governance. This chapter explains the power dynamics in each governance structure and
illustrates with cases of neighborhood governance in Beijing.
Chapter 3 studies the boundary settings of governance networks. The dynamic process of
inclusion and exclusion of certain organizations in governance networks is defined as boundary
setting (van Nuenen, 2007). Getting the "right" participants in collaborative governance is
contended to be the "first condition of successful collaboration" (Chrislip & Larson, 1994; K.
Emerson et al., 2012). Despite the strong normative and instrumental foundations of inclusion,
the reality is that some organizations are often partially or wholly excluded from collaborative
governance. A number of studies well documented the exclusion of important stakeholders. For
example, Hendriks' (2008) study of Dutch energy transition networks showed that government
and industry elites dominated the reform while small businesses and societal organizations were
excluded. The study of urban governance networks Los Angeles showed that civic organizations
are "politically vulnerable" and are often circumscribed by city administration (Musso, Weare,
Bryer, & Cooper, 2011 ). In the collaborative governance literature, boundaries are often taken as
a given with little attention on how they are formed in the first place.
15
What may explain the inclusion of certain actors in collaborative governance? Conventional
explanations are often based on either or resources or power. Resource-based explanation
maintains that stakeholders will be included if they have indispensable resources for achieving
network goals (Scharpf, 1978). Boundary setting is thus considered as a rational process.
Power-based explanation highlights the political process of inclusion and argues that powerful
organizations may intentionally exclude certain stakeholders in order to maintain their
dominating status (Purdy, 2012; Tett et al., 2003 ).
Drawing on the literature from general organization theory, urban regime theory, nonprofit
management and network research, this chapter attempts to transcend both the resource-based
explanation and power-based explanation. Neither framework helps us to understand the
mechanism of boundary setting. This chapter develops a capacity-threat framework to explain
the boundary setting of collaborative governance. When assessing potential members to include,
extant members will consider not only resources that they may bring but also the threats that they
may pose to their interests. Extant members especially those that benefit from the status quo will
try to block new members if they are perceived as serious threats. However, this does not mean
that these organizations will always be excluded. To be successfully included, these potential
members have to have high organizational capacity so that they can overcome the resistance
from extant members. If they lack the organizational capacity, for example, they lack strong
leadership or have a low degree of formalization, then they will likely to be excluded. Of course,
our analysis also suggests that if these new members are not perceived as threats, then they may
16
also be included in governance. Therefore, the boundary setting of governance networks is a
political and dynamic process. The boundary is not static but may be constantly shifting
depending on the changes in relative power of extant and potential members. Inclusion in
governance networks is not permanent. Members may be deactivated if their agenda are
conflicting with the others', or if they are perceived by others as a wrong partner (Agranoff &
McGuire, 2001).
Chapter 4 examines the determinants of the effectiveness of governance networks. Evaluating
network effectiveness and studying its determinants have been an important topic in the network
research (Meier & O'Toole, 2003; O'Toole & Meier, 2004; Provan & Kenis, 2008; Provan &
Milward, 1995). Provan and Milward (1995) proposed a preliminary theory of network
effectiveness by examining the effects of environmental factors and network structural
characteristics. A series of studies followed their original research and examined the effects of
factors such as forms of network governance, network age, and internal trust (Provan & Kenis,
2008; Provan & Milward, 1995; Provan & Sebastian, 1998; Raab, Mannak, & Cambre, 2013).
Another school of scholars studied the roles of agency with a special focus on network
management (Juenke, 2005; Meier & O'Toole, 2003; O'Toole & Meier, 2004). A number of
variables such as managerial networking (Meier & O'Toole, 2001, 2003), management quality
(Meier & O'Toole, 2002), personnel stability (O'Toole & Meier, 2004) and network leadership
(McGuire & Silvia, 2009) have been examined.
17
The current literature is enlightening in many ways, but it is not without its problems. One
problem is that, it once relied heavily on single case study or comparative case study to develop
theory (Isett et al., 2011; Meier & O'Toole, 2003; Provan et al., 2007), though recently more and
more studies have used medium to large-N samples (O'Toole & Meier, 2004; Raab et al., 2013;
Verweij, Klijn, Edelenbos, & van Buuren, 2013). A related problem is that configurational
theories of network effectiveness are still in its nascent stage. Conventional regression models
are limited in testing the complex interactions between explanatory variables (Fiss, 2011 ).
However, these configurational propositions are difficult to test partly due to the lack of
appropriate statistical models. Though scholars have started to employ new methods such as
set-theoretic methods to build configurational theories (Raab et al., 2013; Verweij et al., 2013),
more research is needed in order to cover different research contexts and to test the effects of
different factors.
Based on the models proposed by Provan and Milward (1995) and Provan and Kenis (2008), this
chapter employed a mixed-methods approach to study the determinants of the effectiveness of
governance networks. The chapter is based on 22 neighborhood governance networks in Beijing,
and each network was made up of public, business and civic organizations. Linear regression
was used to identify independent variables that exert statistically significant influence over
network effectiveness, and the fuzzy set Qualitative Comparative Analysis was used to
investigate the complex interactions between explanatory variables. The chapter revealed the
causal complexities of network effectiveness: the analysis found different but functionally
18
equivalent configurations of causal conditions that lead to network effectiveness, and showed
that configurations of factors leading to network effectiveness are different from those leading to
network ineffectiveness. The results also suggested that network structural characteristics such as
network centralization and density are neither sufficient nor necessary conditions for network
effectiveness. However, in contract to Provan and Milward's (1995) findings, the results
suggested that network density is more important than network centralization in affecting
effectiveness in small networks. Resource munificence was identified as an "almost always"
necessary condition for network effectiveness.
19
Chapter 2 : Power Balance and Institutionalization
-- How Forms of Network Governance are Determined
Introduction
In the public administration literature, governance and networks are closely intertwined.
Governance refers to "governing styles in which boundaries between and within public and
private sectors have become blurred" (Stoker, 1998, p. 17). The definition implies the existence
of a web of public, business and nonprofit organizations that engage in the process of governing.
These organizations form the structural part of governance, or networks. To some degree,
governance can be understood as the process that takes place within governance networks (Klijn,
2008). This network form of governance is often contrasted with market and hierarchy as a new
mode of social coordination (Klijn & Skelcher, 2007; Lowndes & Skelcher, 1998). The
implications of network governance to democracy and society have been studied extensively
(Bogason & Musso, 2006; Fung, 2006; Klijn & Skelcher, 2007; McGuire, 2002; Provan &
Milward, 1995).
The focus of this paper is on the governance of these networks per se rather than the governance
of society through networks (Vangen et al., 2014). The structural part of governance, or networks,
involves organizations from public, business and nonprofit sectors. The governance of these
20
networks per se is to work out ways to jointly make decisions and make rules to govern
behaviors and relationships (Thomson & Perry, 2006). Although collaborative governance seems
to indicate that these organizations share goals and work together voluntarily, actually they may
come into play with "different role positions and carry different weights" (Agranoff & McGuire,
2001, p. 315). With different authorities and resources, organizations in networks may occupy
different power positions. Powerful organizations may dominate networks and advance their own
agenda (Purdy, 2012). Klijn and Teisman (1997) conceptualized the complex interactions
between all involved actors as "games" in which actors try to maximize their own interests and
influences. The complex interactions between organizations may result in diverse governance
structures in these networks.
Understanding the internal governance structures has important implications for the study of
governance networks. For example, each governance structure has different power dynamics and
may affect network effectiveness in different ways (Provan & Kenis, 2008). Unfortunately, the
governance structures of networks received relatively little attention. As Provan and Kenis (2008)
argued, people often ignore that networks vary considerably in terms of structural patterns and
assume that the structures of governance networks are undifferentiated. Provan and Kenis (2008)
identified three forms of governance structures, including shared network governance, lead
organization governance and network administrative organization governance. Limited research
has been conducted to further Provan and Kenis' (2008) typologies of forms of governance
structures.
21
Provan and Kenis' (2008) typologies were based on "goal-directed" networks and may not work
for the large number of "serendipitous networks." In serendipitous networks, network change is
primarily driven by serendipity, and "network trajectories develop haphazardly from the
interactions of individual actors" (Kilduff & Tsai, 2003, p. 89). This paper proposes a new
framework to explain how power dynamics and institutionalization affect the governance
structures of serendipitous networks. This knowledge may help to design successful
collaborative governance regimes (Ansell & Gash, 2007; K. Emerson et al., 2012).
Theoretical Framework
One fundamental assumption of this paper is that power differentials and the degree of
institutionalization in interorganizational networks affect how organizations interact with one
another and thus give rise to different network governance structures (Benson, 1975). When
people talk about collaborative governance, they often assume that organizations cooperate
voluntarily and share power. However, interorganizational relationships may also be
characterized by conflict, domination and exploitation (Agranoff & McGuire, 2001).
Organizations may come from public, business and civic sectors and thus have different
authority and resources. Organization theory has provided several powerful frameworks such as
the resource dependence theory to explain interorganizational relationships. Power has long been
recognized as an important variable to explain interorganizational interactions (Agranoff &
McGuire, 2001; Choi & Robertson, 2013; R. M. Emerson, 1962; Gray & Hay, 1986). Emerson
22
(1962) pointed out that power resides in the other's dependency: the power of A over B is equal
to the dependence of B on A. Pfeffer and Salancik (1978) further developed this exchange-based
view of power into a theory of resource dependence in order to analyze interorganizational
power and organizations' management of their environments. This theory generally argues that
organizations are interdependent because they depend on other organizations for critical
resources. Interactions between organizations must be explained ultimately at the level of
resource acquisition (Benson, 1975). Organizations which have critical resources that are not
available from anywhere else will have more power over dependent organizations, and they thus
may take advantage of their power when interacting with others. Organizations, especially those
at disadvantaged positions, adopt different tactics to manage their relationships with others and
to change power dynamics. Their strategies may include merger and acquisition. Hardy and
Phillips (1998) took a broader view of dependence and argued that organizations may depend on
others not only for resources but also for other things. They further discussed three sources of
interorganizational power - formal authority, control of critical resources and discursive
legitimacy. Formal authority refers to "the recognized, legitimate right to make a decision"
(p.219); for example, government agencies usually enjoy this form of power. The control of
critical resources becomes one source of power due to the dependence of other organizations on
these resources, which is consistent with the resource dependence theory. Resources may include
money, personnel, information or techniques that can be transformed into power and influence
when interacting with other organizations. Organizations with discursive legitimacy have power
because they are considered to have the expertise and legitimacy to speak about a specific issue.
23
For example, Greenpeace is often considered to have discursive legitimacy in environmental
issues. These different sources are not mutually exclusive, and one organization may possess
authority, resources and discursive legitimacy simultaneously. This broader view of power
provides a more fine-grained framework to analyze power relations in organizational domains.
Agranoff and McGuire (2001) argued that power concerns should be the center of any general
network management theory because it has important implications for the patterns of interactions
in networks. A balanced power distribution is deemed to be favorable for interest representation,
deliberation and achieving cooperative participation (Choi & Robertson, 2013; Hardy & Phillips,
1998). In contrast, imbalanced power distribution may lead to a number of negative outcomes.
Powerful organizations with more access to authority, resources or discursive legitimacy may use
their power to exclude certain stakeholders, silence different voices and advance their own
interests. Organizations at weaker power positions may take different measures to mitigate power
imbalances. Emerson (1962) maintained that organizations may reduce their dependence on a
certain organization by exploring alternative sources of resources or by building coalitions with
others to change the relative distribution of resources. Hardy and Phillips (1998) proposed four
strategies that organizations can use in their engagement with other organizations: collaboration,
compliance, contestation and contention. For example, in the UK refugee system, the Refugee
Legal Center became a separate organization because of government support and also received
funding from the government, so it chose to "compliance" as its strategy to interact with the
government.
24
Institutional theory provides another theoretical perspective on interorganizational relationships.
Institutional theory has different variants and thus defines institution in different ways with
different emphases (W. R. Scott, 1987). Barley and Tolbert (1997, p. 96) defined institutions as
"shared rules and typifications that identify categories of social actors and their appropriate
activities or relationships." Goodin (1998, p. 52) highlighted the construction of roles in his
definition of institution as "organized patterns of socially constructed norms and roles."
Organizations are embedded in an "institutional" context of values, norms, beliefs and
taken-for-granted assumptions that prescribe appropriate and socially legitimate ways of doing
things (Barley & Tolbert, 1997; Greenwood, Hinings, & Whetten, 2014). Institutionalism
recognizes the richness of human history, culture and norms in defining not only the means
(socially legitimate ways of doing things) but also the ends (what is desirable) of individuals and
collectives. Institutions structure our behaviors, lead us to interpret the world in certain ways,
and more importantly, shape our understanding of what is the "legitimate" ways of doing things
(Geertz, 1973; Meyer & Rowan, 1977; Phillips, Lawrence, & Hardy, 2000). The institutional
context exerts profound influence over organizational forms and structures, the diffusion of
innovation, and organizational survival (Kennedy & Fiss, 2009; Meyer & Rowan, 1977; Tolbert
& Zucker, 1983; Westphal, Gulati, & Shortell, 1997).
Institutions are created, in the contexts of previous culture and history, through repeated
interactions between actors (Giddens, 1979, 1984). Based on the work of Giddens, Barley and
Tolbert (1997) described the recursive relationship between institutions and actions: Actors
25
first encode existing institutional principles through socialization and then enact scripts which
are actors' interpretations of institutions in a particular setting. Actors' behaviors' may either
revise or replicate the scripts, leading to possible institutional change if scripts are altered. The
last moment is the externalization of patterned behaviors, which gives these patterns a normative
status. DiMaggio and Powell (1983, p. 148) provided four indictors of instutitionalization: "an
increase in the extent of interaction among organizations in the field; the emergence of sharply
defined interorganizational structures of domination and patterns of coalition; an increase in the
information load with which organizations must contend, and the development of a mutual
awareness among participants in a set of organizations that are involved in a common
enterprise.''
The institutional theory gives us a fresh perspective to understand collaborative governance as a
new institution. We have been witnessing the hollowing out of the state and the transformation
from government to governance over the past few decades in response to a wide range of societal
changes (Milward & Provan, 2000; R. A. Rhodes, 1997). The contextual change provides
opportunities and incentives for organizations to question rather than replicate scripted patterns
of behaviors (Barley & Tolbert, 1997). People may question the ineffective or inefficient
government-centered approach of governing and start to regard cross-sector collaboration as the
standard approach to tackle the so-called "wicked problems" - for example, we have reached the
point where "it is difficult to imagine successfully addressing global problems, such as the AIDS
pandemic or terrorism, and domestic concerns, such as the educational achievement gap between
26
mcome classes and races, without some sort of cross-sector understanding, agreement, and
collaboration" (Bryson et al., 2006). The institutionalization of collaborative governance is
defined as the wide acceptance of collaborations between governments, business and nonprofit
organizations as a legitimate way of governing complex modem societies.
In any specific cases of collaboration, the degree of institutionalization may vary. The reason is
that collaboration happens in a context where there is no legitimate authority and the
relationships are negotiated (Phillips et al., 2000; Thomson & Perry, 2006). Various aspects of
collaborative relationships needs to be negotiated: the definition of the problem, the appropriate
roles and scope of an organization, and the legitimate responses to the problem (Benson, 1975;
Phillips et al., 2000). The negotiations may lead to institutionalization because they touch on
fundamental and complex issues and thus may give rise to lead to new rules, norms and
understandings (Lawrence, Hardy, & Phillips, 2002). Although through repeated interactions,
organizations may reach shared understandings of these critical aspects of collaboration and
develop common rules regarding problem and role definitions, these practices and behavioral
patterns are not equally institutionalized (Barley & Tolbert, 1997). The degree of
institutionalization may depend on "how long an institution has been in place and on how widely
and deeply it is accepted by members of a collective" (Barley & Tolbert, 1997, p. 96). Lawrence
et al. (2002)argued that collaboration has institutional effects because it facilitates the
structuration process described by DiMaggio and Powell (1983): The intensity of
27
interoganizational interactions increases, coalitions can form, information can be exchanged, and
mutual awareness of involvement can develop.
Lowly institutionalized collaboration may have a high possibility of internal conflicts. The
reason is that member organizations have not yet reached a common understanding of key
aspects such as problem definitions, the appropriate roles and scope of an organization,
coordination procedures, and interorganizational relationships. The situation may be worse if
organizations come from different sectors because they may act on their own knowledge or
scripts which are deeply influenced by the institutional logics of their sectors in the process of
institutionalization (Alexander, 1998; Barley & Tolbert, 1997; Phillips et al., 2000). Without
sufficient negotiations to reconcile these different institutional logics, conflicts may be created.
The lack of wide support and legitimacy in these key aspects of collaboration may cause a
number of problems. For example, different problem definitions mean different solutions and
stakeholders. Black and Rose (2002) studied the case of mental health community and found that
an organization attempted to redefine mental health problem as a social problem rather than a
"disease" as it is conventionally defined. This new definition clearly implied a new solution and
needed to incorporate new stakeholders, greatly changing the existing distribution of resources
and power in the domain. Even if there are clear definitions of the problem and who should be
involved, consensus on the proper roles of each participant is also critical. Clear definitions of
the role and position of an organization in a domain directly affect its resources. Benson (1975)
pointed out that "authority and money flow to an agency on the basis of its sphere of activities -
28
services provided, clients served, and so forth" (p.236). Conflicts will arise if an organization's
sphere of activities is encroached upon by other organizations.
Research Context
This study focuses on neighborhood governance networks in China. Urban areas in China have
gone through significant changes because a large number of commercially-developed
neighborhoods have been established after the Housing Reform in the late 1990s. Due to the
establishment of private property rights, the previous government-dominated neighborhood
governance structure has lost its legitimacy. New types of neighborhood organizations such as
Homeowners' Associations become important stakeholders in neighborhood governance.
However, laws and covenants regulating behaviors of neighborhood organizations are either not
in place or are not well enforced. New common understandings of how to govern these
commercially-developed neighborhoods have not been fully established. Diverse neighborhood
organizations with different institutional principles, goals, interests, resources and power
cooperate with or challenge one another, trying to maximize their interests. This gives rise to
highly spontaneous and dynamic neighborhood orders and provides an ideal context to study
how different modes of neighborhood governance structures arise.
China's urban neighborhoods used to be firmly controlled by the government in the planned
economy era. Street Offices (SOs) were the lowest level of governments. Residents' Committees
29
(RCs) were the most basic unit of social organization and were controlled by SOs. Housing was
provided by state-owned enterprises or local governments. The public housing system could
hardly keep up with people's increasingly higher standards of living and thus could be sustained.
Housing Reform, an important part of China's overall economic reform, was launched in the late
1990s and a housing market was created. Commercially-developed neighborhoods have become
the main type of neighborhoods in most cities.
In these commercial-developed neighborhoods, RCs and SOs have the public authority. RCs,
which are supposed to be "self-management, self-education and self-service mass organizations",
are actually firmly controlled Street Offices. They are responsible for a variety of administrative
functions, such as family planning and neighborhood safety. Street Offices are the lowest level of
governments in urban areas. Beijing Municipal Bureau of Statistics (2013) showed that, as of
2012, the City of Beijing had 143 Street Offices overseeing 2,816 Residents' Committees.
Though SOs usually cover a larger area and may have dozens of neighborhoods, current laws
and regulations give SOs a direct role in neighborhood governance. According to the regulation
in Beijing, homeowners have to form their HOAs under the direction of SOs. They have to first
establish a preparatory group and then submit a formal letter of application to SOs. SOs then
designates the head of the preparatory group who is usually the director the RC in the
neighborhood. Homeowners also have to submit all their application materials to SOs. Therefore,
SOs and RCs have considerable influence over the process of establishing HOAs - they can
affect how fast and smooth the entire process goes. Of course, the scope of their authority goes
30
far beyond this - when homeowners have conflicts with developers or property management
firms over properties, SOs work as mediator and even arbitrator according to Chinese laws.
Therefore, SOs and RCs are very powerful players mainly because of their political authority.
Urban residents buy their own homes in the fast-developing housing market. Homeowners are
allowed to establish Homeowners' Associations (HO As) to represent their interests in managing
communal properties and protecting their property rights. The Property Right Law is often
weakly enforced and other organizations such as developers often infringe on homeowners'
communal properties to gain huge revenues. Therefore, although legally HOAs should be
important participants in all the decisions regarding communal properties, homeowners have
great difficulties in establishing HOAs in the first place. Even if HOAs are established, their
roles are often greatly weakened. The major reason is that HOAs do not have the political and
economic resources as local governments and business interests do. In an authoritarian country
like China, citizens have very limited voting power and almost no mechanisms to hold local
officials accountable. Therefore, they do not have the most important political resources as
citizens in democracies do. They may use the legal approach to protect their rights, but the legal
system is not very friendly to ordinary people and also may be manipulated as one of our cases
suggests. In most cases homeowners have limited political resources except that they have
important political connections, such as having friends working at high-level governments.
Established HOAs also have fewer economic resources compared with developers or even
property management firms. They may get some revenues from sharing property management
31
fees with property management firms or from renting out communal properties, but the total
revenues are usually much fewer compared with the economic resources of property developers
or management firms. Due to the limited opportunities that citizens have to engage in politics,
homeowners in these neighborhoods often lack democratic skills and have the very passive
"authoritarian personality" - for example, many homeowners do not dare to question the
govermnent (W. Wang, Li, & Cooper, 2015).
Business organizations also become important stakeholders in commercially-developed
neighborhoods. Developers develop properties and then sell homes to buyers. Theoretically,
developers should pull out of these neighborhoods once all units are sold. In reality, their
interests are deeply entrenched in these neighborhoods. Weakly enforced Property Rights Law
gives developers plenty of loopholes to take advantage of. In order to make more money,
developers often exaggerate the actual areas of apartments or leave many construction quality
problems. Many developers hire their son companies as property management firms and charge
high property management fees that do not match service quality. These son companies also
work closely with developers to hide all the left-over problems. Business organizations generally
have great economic resources. The booming real estate industry is highly profitable and
developers can generate a significant amount of revenues. Many developers are giants in the real
estate industry in China, and even small developers can mobilize considerable financial
resources in development processes. Their economic resources can be transformed into political
resources. As Lindblom (1977) pointed out, business interests have both structural and
32
instrumental power. Their structural power comes from their role as tax bases. Local
governments in China get huge revenues from real estate development, so they often forms
informal pro-growth coalitions with developers (Zhang, 2002). Business organizations'
instrumental power comes from their bribery and interests exchange with local government
officials.
The previous governance structure has been changed greatly with new players actively engaging
in neighborhood affairs. Governmental, business and civic organizations have diverse
institutional logics, interests, goals and sources of legitimacy in neighborhood affairs, and they
actively participate in these affairs in order to maximize their interests. Although there have been
some regulations on the legitimate roles of these players and their relations, in many cases,
common understandings of roles and interorganizational relationships have not been reached.
The actual governance structures differ significantly. For example, in some cases a shared
governance structure is formed in which all organizations participate in neighborhood
governance, while in some cases one or two organizations dominate neighborhood governance.
Neighborhood governance in China can be characterized as highly spontaneous and lowly
structured, which provides us with an ideal field to study how different governance structures are
formed.
33
Research Design and Data Collection
This paper employed comparative case method as the major research design. Since my major
objective was to study the governance structure in each neighborhood, the unit of analysis is
neighborhood. The success of comparative case method depends on obtaining cases that vary in
governance structures so that meaningful comparisons can be conducted. The relative frequency
distribution of each governance structure is not important. G King, Keohane, and Verba (1994)
argued that random sampling might worsen the problem of selection bias in small-N studies.
Therefore, this research adopted snowball sampling in order to get cases that varied in their
governance structures. Our long-term collaborator in Beijing, who is also the director of a local
nonprofit organization specializing in homeowners' rights protection, helped me to connect with
the first seven neighborhoods. I visited each one of them and then asked them to recommend
neighborhoods that had similar or different governance structures. I stopped data collection after
visiting 22 neighborhoods because I felt that I had reached the point of data saturation (Small,
2009). Additional cases provided little new information on governance structures or power
dynamics at that point. In the end, four major governance structures were identified: shared
governance, inertial governance, insurgent domination and lead organization governance.
Different types of cases allowed me to compare the interorganizational power relations and
degree of institutionalization and see how these differences give rise to different modes of
governance structures.
34
My research methods are qualitative. In order to have a comprehensive understanding of
interorganizational power relations and institutionalization, I conducted interviews with
homeowner activists, government officials from both RCs and SOs, and property management
firm executives. Each interview lasted 90-120 minutes. In order to collect network data, I asked
interviewees to nominate the organizations that they often collaborated with on neighborhood
affairs. The nomination was then verified in subsequent interviews. If there were inconsistencies,
I asked interviewees to clarify how they collaborated and decided whether it counted as a
collaborative relationship. Since collaboration is mutual in nature, in the end 22 symmetrical
collaborative networks were constructed based on the network data. Network data analysis was
conducted by using UCINET. I also used participant observation to collect data. I attended
HOAs' meetings to collect information on HOA's governance, their relationships with other
players in neighborhood governance and other issues to answer the above research questions.
A Typology of Governance Structure
Based on my fieldwork and the theoretical framework outlined above, two dimensions, balance
of power and degree of institutionalization, were identified as two critical factors that may affect
interactions between organizations and governance structures. The degree of institutionalization
describes the degree to which members of governance networks have reached common
understandings on problem definition, roles of each member and the direction(s) to go.
High-institutionalized governance and low-institutionalized governance can be differentiated
35
based on the degree of institutionalization. Based on DiMaggio and Powell (1983), three
indictors were used to evaluate the degree of institutionalization of collaborative networks. The
first indicator is role definition, which assesses whether organizations have been founded to take
their legally-defined roles in neighborhood governance. The second indicator is role acceptance,
which evaluates the degree to which organizations respect role definitions and accept other
organizations to fulfill their respective roles. The third indicator is the density of relationships,
which measures how closely organizations share information and work together. We would
expect that in high-institutionalized neighborhoods, through repeated negotiation, governance
structures and interorganizational relationships are clearly defined and widely accepted. The
governance structures thus enjoy high legitimacy, and the possibility of interorganizational
conflicts is low. The balance of power describes the degree that the distribution of power in a
network is balanced. The analysis will consider three sources of power: authority, resources and
discursive legitimacy (Hardy & Phillips, 1998) and also organizations' capabilities to use these
sources of power to their advantage. Balanced and imbalanced governance structures can be
identified along this continuum. We would generally expect that if power is balanced in a
neighborhood network, organizations are more likely to compromise with one another in order to
solve conflicts; in contrast, some organizations may dominate neighborhood governance if power
is highly unbalanced. The interactions between the two dimensions thus give rise to a 2x2
typology of governance structures, as Figure 2.1 shows.
36
High
Balance of
Power
Low
Inertial Governance Shared Governance
Insurgent Domination Lead Organization Governance
Low High
Institutionalization
Figure 2.1 A Typology of Internal Governance Structures
Shared Governance (High power balance, high degree of institutionalization)
If power is more or less balanced in an interorganizational network, meanmg that each
organization has authority, resources or discursive legitimacy to resist the demands from others,
then it is likely that each organization has the opportunity to participate in the governance of
these networks. In the meantime, if these organizations also have a high degree of consensus on
problem definition, the roles of each organization and the directions to go, the possibility of
conflicts may be low and each organization will take on what they need to do. Under these
conditions, a shared governance structure may emerge.
37
The Shang-Di neighborhood is a good case of shared governance. This neighborhood was
developed soon after China's Housing Reform and was one of the first few commercially
developed neighborhoods in Beijing. The developer was a large state-owned real estate
development firm and this project was just one of their many projects. This project was not a
major source of revenue, and the developer soon pulled out of the neighborhood after all units
were sold. Therefore, the developer had the economic resources but did not use these resources
to gain influence or power in this neighborhood. All members of the RCs were also property
owners, so they were quite supportive of establishing the HOA. SOs did not set too many
obstacles either, so the political authority did not become a major barrier. Led by some visionary
homeowners who were lawyers, scholars and high-level managers in international companies,
homeowners soon established one of the first few HO As in Beijing. They also did a number of
experiments, including establishing a monitoring committee to monitor the work of HOA, in
order to enhance the organizational capacity. The HOA firmly held the position as the legitimate
representative of homeowners in the governance network. They used a bidding process to hire a
property management firm to provide services. What we can see is that public, business and civic
organizations had relatively equal power in this network, and no organization had the power to
exclude or marginalize others.
In addition, this network was highly institutionalized. Organizations successfully negotiated the
roles and scopes of each organization, and they understood and respected other organizations'
38
turfs. For example, unlike some Residents' Committee (RC) that saw HOAs as a competitor and
a trouble maker, the director of their RC told me that:
"Some people hold a negative view of HOAs, thinking that they messed things up. Some say
they are different from us (RC), (but) I don t think so. I think they are similar to us. HOAs are
established because people need them. I think HOA members bring us some fresh air. For
example, Mr. Guo (HOA director) is a professional manager, and he shares with me some good
management experience.. We want (all organizations) to use each other's strengths and work
together to make things better. "
Similarly, Mr. Guo, the HOA director, told me that the HOA and RC had a good working
relationship although they did not agree on everything:
"The HOA and RC complement each other: sometimes the RC takes the initiative to do
something that they are not supposed to do, (in this circumstance) I usually take a step back; of
course whenever they do not do what are supposed to, I will take the lead."
The property manager also had a very clear understanding of their role:
"Legally, we are the service provider and the HOA is the service receiver. As the provider, we
listen to their needs. As the property management firm, our responsibility is to serve the
39
homeowners. Of course we serve for fees. Sometimes when some problems arise and they do not
cost too much, I usually ask my boss: shall we use our money and labor to solve the problems?
My main objective is to do our job well."
As DiMaggio and Powell (1983) pointed out, another indicator of high institutionalization is the
increase in the extent of interaction among organizations in the field. A network analysis was
conducted and the density score of the network was 1, suggesting that every organization was
cooperating with everyone else.
As the above discussion indicates, the relative balance of power in this network made it possible
for all members to participate. A high degree institutionalization led all organizations to focus on
what they were supposed to do. As a result, a shared governance structure was formed, as Figure
2.2 indicates. The network had a high density score of 1 and a low centralization score of 0,
suggesting it was a completely decentralized network. In this neighborhood, organizations
respected each other and collaborated quite well. For example, the HOA always invited leaders
from the RC and the property management firm to attend their regular meetings in order to
enhance mutual understanding.
40
Figure 2.2 The Network of Shang-Di Neighborhood
Inertial governance (High power balance, low degree of institutionalization)
A second governance structure is inertial governance which arises when power is balanced but
the degree of institutionalization is low. Under this circumstance, each organization has its power
in the network, though the power may come from different sources. All organizations have the
opportunities to participate. However, problem definitions, the roles and scopes of each
organization are not well negotiated or accepted. The missing of common understanding means
that the legitimacy of the governance structures is relatively low. Organizations may thus have
conflicts with one another. Since no organization has the ability to impose their preferences to
other organizations, they cannot do what they believe is right or necessary. They may soon lose
interests in participation or may take the minimum level of effort to work with others.
41
Feng-Dan neighborhood was a case in point. It was a high-end neighborhood with both
apartment buildings and sing-home houses. The power distribution in the neighborhood was
relatively balanced, and no organization had the capability to exclude others. Some homeowners
who owned sing-home houses were high-level government officials and had powerful political
connections. Therefore, homeowners successfully established their HOAs in 2005 with little
obstacle from the RC or SO. The HOA became the legal representative of homeowners and
successfully hired property management firms through bidding. However, as time went by many
HOA members lost their commitments to their HO As and the HOA became weaker and weaker
in terms of organizational capacity. The RC and SO, though did not set barriers to homeowners
when they established their HOA, started to fill the vacuum left by a weaker HOA and played an
increasingly important roles in the neighborhood. The property management firm was hired by
the HOA, but they had the expertise and economic resources to manage properties and even to
evade HO A's monitoring. The property management firm even often challenged the weaker HOA
after being hired.
The degree of institutionalization was low in this network especially regarding the roles of each
organization. First, homeowners themselves were not clear about the role of the HOA and what
should they do, which to some degree weakened the organizational capacity of the HOA. The
HOA failed to serve as the legal representative of homeowners. For example, as one HOA
member said: "Our HOA could not work because people have very different opinions. We could
42
not hold any meetings because people keep quarreling (in these meetings)." When asked why the
property management firm dared to challenge the HOA given that it was hired by the HOA, the
interviewee explained that "This has nothing to do with the fact that they were hired by us. Our
HOA is too weak. We ourselves do not have a consensus; internally homeowners have too many
different opinions. " Second, even if the property management firm and the local government
understood their roles, in many cases they did not do what they were supposed to do and
sometimes they intentionally engaged in extra-role behaviors. For example, decision-making
regarding property management and hiring property management firms should be the HOA's
responsibility, and the RC and SO only had a vague legal role to facilitate this process. The SO
nevertheless managed to manipulate the bidding process and hired a property management firm
that was ranked third or fourth by homeowners. The property management firm, unlike the one in
the Shang-Di neighborhood, did not consider itself just as a service provider and should listen to
the needs of homeowners. One HOA members complained that the management firm charged
high management fees but failed to fulfil their responsibilities:
"We told them that homeowners were not satisfied with their work, but they simply did not listen
to us. In their eyes, we (HOA) are just nothing .. ... we are really pathetic. It is like we should
kneel ourselves to beg them to work: please do your job, and you can charge as much money
as you want. "
43
Another indicator of the low degree of institutionalization was the low density score of 0.583,
which showed that only 58.3% of all possible collaborative relationships existed in this network.
The organizations were not collaborating closely with each other, resulting in the inertial
governance structure. Figure 2.3 shows the network structure.
In this neighborhood, although conflicts between organizations may arise due to the low degree
of institutionalization, the relatively balanced power structure reduced the possibility or severity
of conflicts. As much as some organizations wanted to impose their problem or role definitions
to others, they did not have the power to do so. For example, even though the SO once
manipulated the bidding process, it could not completely marginalize the HOA on neighborhood
affairs. The relatively balanced power structure actually created a deadlock in this network.
Organizations gradually lost interests in taking the trouble to collaborate with one another. They
just took the minimum amount of effort to fulfill their responsibilities or cooperate with each
other, resulting in the so-called collaboration inertia (Huxham & Vangen, 2000). Benson (1975, p.
235) argued that this type of non-cooperative network is frequently encountered because "none
can muster power sufficient to dictate terms to the others."
44
so
PMF _ ___ _____ _ __ HOA
RC
Figure 2.3 The Network Of Feng-Dan Neighborhood
Insurgent Domination (low balance of power, low degree of institutionalization)
Imbalanced power and low degree of institutionalization may give rise to a third governance
structure: insurgent domination. Previous research has suggested that asymmetrical power
relations in a network may cause a number of problems, such as the exclusion of certain
members and the marginalization of different voices (Agranoff & McGuire, 2001). Powerful
organizations have a higher possibility of dominating governance in these networks, and
sometimes they may even form coalitions in order to gain absolute power advantage in a network
(Cook, 1977). Low degree of institutionalization may worsen power struggles in these
imbalanced networks because organizations lack common understanding of what the real
problem is and what their appropriate roles are. Of course, low degree of institutionalization is
not bad to everyone; some organizations may benefit from vague problem and role definitions
because they can profit from engaging in extra-role behaviors. One strategy for these insurgent
45
organizations to gain absolute power advantage and maintain the status quo is to form coalitions,
resulting in an insurgent domination structure that is not easy to break down. Powerful
organizations may force weak organizations to accept their terms, but the major problem with
this governance structure is the lack of legitimacy.
Rong-Feng is a good case of insurgent dominance. The developer was a small-to-medium firm
that seldom succeeded in its other projects. As a result, the developer relied heavily on the
revenues generated from this neighborhood. To maximize their revenue, the developer
established a son company to take charge of property management in Rong-Feng. They charged
a high fee while providing very poor services. I visited the neighborhood and found that the fire
facilities, lawns, gardens and elevators were extremely poorly maintained. The developer also
sold homeowners' communal properties for tens of millions without giving homeowners any
compensation. To maintain the revenues, the developer used its economic resources and political
connections to influence local governments' decision-making. The RC and SO had the political
authority to at least help homeowners to protect their communal properties, yet they chose to
work closely with the developer partly because the RC also occupied 300 square meter's office
space which was homeowners' communal property and partly because of the developer's
political connections. Though homeowners were led by some experienced and persistent
homeowner activists, they were much weaker with almost no economic resources and little
political resources compared with the business and government organizations.
46
To make things worse, the degree of institutionalization was very low in this neighborhood. The
imbalanced power dynamics undermined any basis for negotiating roles and interorganizational
relationships. The developer and property management firm apparently did not see themselves
just as service providers because they seriously violated homeowners' property rights. They were
not willing give up their interests, so their strategy was to dominate the network by forming a
coalition with the SO and RC. Homeowners wanted to establish their HOA in order to better
represent their interests and protect their property rights. However, the RC and SO were not
cooperative partly because of the coalition with business interests and partly because of their
fundamentally different problem definitions than homeowners'. They saw homeowners'
organizing as a threat to social stability. For example, one official at the SO told me:
"The HOA is a good thing, but it is not very compatible with the contemporary Chinese political
system. People in high-end neighborhoods are better because they can accept this. In some other
neighborhoods, it's OK if you don't tell these homeowners. Once you tell them that these
(communal) properties are theirs, then they want these properties, but they did not know how to
orderly claim and use these properties .. Social stability is gone."
This government official clearly saw stability as a priority and worried that homeowners'
organizing may cause instability. One interviewee at the Rong-Feng neighborhood told me that:
47
"The government always held us up whenever we took one step forward. Once I went to the
government, the deputy director of the SO said, 'This homeowner from Rong-Feng came to
create troubles again. 'They could not care less about our interests."
The fundamental difference in problem definition became another source of conflicts between
homeowners and local governments, making it even harder for homeowners to gam
governments' support. The business interests and local governments both benefited from
violating homeowners' property rights. Therefore, neither of them had the motivation to clarify
roles of each organization or seek an agreement on problem and role definitions that all parties
can accept. Instead, they formed a coalition in order to gain absolute power over homeowners.
Homeowners, who were the legitimate property owners, were too weak to effectively challenge
the coalition. They were thus excluded from the governance of their neighborhood. "Insurgent"
organizations with little legitimacy became the dominant parties in property management,
resulting in this insurgent domination governance structure. As Figure 2.4 shows, there were
clearly two cliques of organizations in the network. One was made up of the business interest and
local governments, including the developer, property management firm, the RC and the SO. The
other clique comprised of the HOA and its consultant firm. The network density score was only
0.47 which was much lower than the density scores of the previous two networks.
48
so
~
Consult:ant
Developer
PMF
Figure 2.4 The Network of Rong-Feng Neighborhood
Lead organization governance (low balance of power, high degree of institution)
The last governance structure is lead organization governance, which is formed under the
conditions of low balance of power and high degree of institutionalization. As the above
discussion suggests, asymmetrical power relationship is likely to cause powerful organizations to
dominate network governance. However, high degree of institutionalization and shared
understanding of role and problem definitions may check power and keep powerful organizations
from infringing upon others' sphere of activities. Therefore, powerful organizations may have
49
significant influence in a network, but they usually do not impose their problem and role
definitions to weaker organizations. The governance structure has a high degree of legitimacy.
Chao-Yang Garden had this lead organization governance structure. Chao-Yang Garden was
developed by a Hong Kong developer soon after China's housing reform in 1998. It was a
high-end neighborhood which was originally developed for foreigners in Beijing. The developer
pulled out of the neighborhood after all units were sold and had little influence in the
neighborhood. The RC and SO had the political authority to monitor the HOA, yet they could not
benefit much from engaging in neighborhood affairs. They thus chose to be neutral. The HOA
was established in 2001 and was one of the first few HO As in Beijing. As the representative of
homeowners, the HOA was led by some experienced activists and had been working well since it
was founded. The HOA successfully hired property management firms on its terms through
bidding processes, and designed a good mechanism to share revenue with property management
firms. The HOA had the upper hand in its relationship with the property management firm. With
plenty of economic resources, the HOA hired a full-time secretary to take care of daily affairs. In
this neighborhood, the HOA was the most powerful organization with both economic resources
and discursive legitimacy. The RC even requested the HOA for some funding to support cultural
activities. The HOA refused the request citing that this was not the HOA's responsibility.
This network was highly institutionalized. Appropriate role and problem definitions and
interorganizational relationships had been worked out. The RC and SO understood that HOA was
so
responsible for all decisions regarding properties, so they chose not to intervene. They just
focused on their own responsibilities such as delivering social services. The property
management firm hired by the power HOA understood its role very well too. The firm
considered itself as a service provider and the HOA as the employer, so they shared all their
financial records with the HOA and worked closely with homeowners. As the representative of
homeowners, the HOA had regular meetings that were very well organized to make decisions
regarding property management. Figure 2.5 shows this network had a fairly high density score of
0.73, showing that organizations were collaborating with one another quite closely.
Although the figure cannot show the power relations between organizations, my field work
suggested the HOA played a central role. It controlled not only discursive legitimacy but also
considerable economic resources, which made it the most powerful player in the network.
However, it did not infringe upon others' sphere of activities or impose its problem and role
definitions to others. It generally respected other organizations but firmly defended its legal roles.
Compared with the abovementioned "insurgent domination" governance structure, this
governance structure also has an imbalanced power structure; however, one key difference is that
this type governance is accepted by organizations and has legitimacy. In these networks,
organizations with the most political authority, economic resources or discursive legitimacy may
become the most powerful one. The high degree of institutionalization becomes an institutional
force that helps to regulate how organizations interact with each other and stop the most
powerful one from infringing upon others' sphere.
51
PMF
HOA
/
so
RC
Figure 2.5 The Network Of Chao-Yang Garden
Discussion and Conclusion
Drawing on the resource dependence theory and the institutional theory, this paper develops a
new framework to explain how different types of governance structures are formed. The balance
of power and degree of institutionalization are two important variables to affect how
organizations interact with one another, and their interactions give rise to different governance
structures. Based on neighborhood governance networks in Beijing, this paper explains how the
interactions between the balance of power and degree of institutionalization give rise to four
different governance structures: shared governance, inertial governance, insurgent coalition
52
domination and lead organization governance. The governance structure of collaborative
networks is a surprisingly understudied area (Provan & Kenis, 2008). The theoretical framework
developed in this paper may provide building blocks for further theoretical development.
To further test the validity of this theoretical framework, I applied it to predict the governance
structures of all the other 18 neighborhood networks. Results are shown in table 1. The
framework made correct prediction in 15 governance networks, achieving a success rate of
83.3%. Three governance networks had internal governance structures that were different from
theoretical predictions, which offers opportunities to further examine the framework. Dong-Mao
had a low degree of institutionalization and a low degree of power balance, which should lead to
an insurgent domination structure. However, some homeowner activists organized several
demonstrations against the developer, which forced the developer to give in. Homeowners also
tried to work with the Street Office and achieved a limited degree of success. Evidence did not
suggest a coalition between the developer and local governments. The neighborhood had a
relatively short history, so a clear governance structure had yet to be established. It is possible
that the developer may seek to establish a coalition with the SO and RC if homeowners became a
serious threat to their interests. Similarly, the Atlantic neighborhood was also undergoing a
transition. They managed to achieve a relatively high level of institutionalization and power
balance, yet the governance structure was not shared governance but transitional because the
conflicts between the property management firm and the HOA became big barriers for
collaboration. Another interesting case was Yue-Yuan. The high degree of institutionalization and
53
power balance should lead to a shared governance structure. However, it was a very small
neighborhood with only two apartment towers - it was so small that no property management
firms wanted to sign a formal contract with them due to limited potential for profits. SOs and
RCs lacked interests in the neighborhood either. Therefore, the HOA had to work hard to govern
the neighborhood even though it clearly was not the most powerful organization.
Table 2.1 Predicting the governance structures of all governance networks
name i nsti tu ti onalizati on power balance Governance Structure
Peng_ Lai low low insurgent domination
Wang_ Fu low high inertial governance
Dong_ Mao low low transitional structure*
Hong-Yuan low low insurgent domination
LI Du low low insurgent domination
YI Mei high high shared governance
Yue Yuan high high lead organization*
Rui Du low low insurgent domination
Guan Zhu low high inertial governance
Fei Cui low low insurgent domination
Tian Tian high high shared governance
Yi Shui low low insurgent domination
Bai Zi low low insurgent domination
Atlantic high high Transitional structure*
Jian_Xiang low low Insurgent domination
Guan Hu Inter high low lead organization
Mei Ii Yuan high low lead organization
Wan_Quan low low insurgent domination
Note: * representing predictions that are not consistent with the reality.
One key contribution of this paper is that it deepens our understanding of the governance
structures of serendipitous networks. Serendipitous networks are more common in the real world
compared with goal-directed networks, yet they are often left unstudied in the public
54
administration literature. Provan and Kenis (2008) studied the governance structure of
goal-directed networks, but their typology may not work for serendipitous networks in which the
network change is primarily driven by serendipity (Kilduff & Tsai, 2003). In addition,
serendipitous networks usually do not have externally administrative organizations to govern
these networks or network-level goals to coordinate organizations. Organizations usually interact
with others on the basis of their interests and take a "tit-for-tat" strategy. The behaviors of
organizations are more natural and may cover most of the cases in the real world. Therefore, the
framework developed in this paper may deepen our understanding of how organizations interact
with one another in natural settings and fill a critical gap in the current literature.
Another contribution of this paper is that it furthers the research on the relationship between
power and the institutionalization process. Previous research has studied the relationship between
power and institutionalization (Phillips et al., 2000). The basic argument is that the rules,
interpretations and problem definitions of powerful members are more likely to be
institutionalized (Phillips et al., 2000). Some scholar went so far as to argue that
institionalization is a political process that reflects the interests of powerful members (Maguire,
Hardy, & Lawrence, 2004; Seo & Creed, 2002). Despite its theoretical insight, this argument
ignores the possible effect of institutionalization on power dynamics. This research studied
different modes that power may interact with the institutionalization process and highlighted a
reciprocal relationship: a high degree of institutionalization may also serve as a check to power.
The shared normative beliefs about the roles and scopes of each organization as well as
55
taken-for-granted assumptions may force organizations to conform to these normative beliefs and
constrain or even stop powerful organizations from infringing upon the sphere of activities of
other organizations.
Drawing on the institutional theory and resource dependence theory, this research also offers rich
practical implications. Based on the institutional theory, Alexander (1998, p. 349) argued that
"actors' knowledge of their social context is the basis for the dual interaction between social
structures and action." Actors enact the social knowledge or scripts from their previous
institutional logics, and their actions deeply affect how they engage with other organizations.
Therefore, their knowledge or scripts plays a key mediating role. In order to facilitate the
institutionalization of collaboration as a new way of governing, it is important to raise the
awareness of all involved actors. Alexander (1998, p. 349) recommended "enlightening the
potential participants in an interorganizational system with an awareness of their interdependence,
and revealing to them their potential mutual objectives and common goals." In the case of
neighborhood governance in China, RCs and SOs are usually strongly influenced by the
institutional logics of governments and they may have a hard time getting used to the new
institutional logic of collaboration. It is important to raise their awareness about how property
rights set boundaries to their power and to lead them to take HOAs as partners rather than
challengers. Of course, education or enlightenment is no easy task and it does not always work.
Benson (1975) pointed out that agreements on matters of collaboration such as problem and role
definitions can be reached on the condition that an organization's interests are not threatened.
56
This paper also has some limitations. One is that like all typologies, the four types of governance
structures are just some ideal types, which may not be able to capture all the nuances in reality.
In the real world, many neighborhoods may not be categorized in each type neatly, and their
boundaries are more blurred. The typologies developed in this paper are by no means an
exhaustive or perfect reflection of all governance structures in reality. The goal is to deepen our
understanding of how governance structures are formed from the perspective of power balance
and institutionalization. The messy real-world cases were simplified in constructing these ideal
types. I would argue that this framework is useful in explaining existing governance structures
and predicting what structures may arise, but unfortunately there will be a margin of error and in
some cases the margin may be significant. Another limitation is that the networks under study
were quite small - each network had about 4 to 7 organizations. The relationships between
organizations may increase exponentially as more organizations are involved in networks. Large
network may have different patterns of interactions between organizations. Therefore, caution is
needed when generalizing the conclusions to larger networks. Third, as a qualitative research, it
is hard for this paper to provide "airtight proof for a causal inference" (Odell, 2001, p. 176) due
to the fact that this method cannot control all possible causal factors. The strength of this paper is
not to establish causal relationship in any sense but to explore the "mechanisms" through which
governance structures arise.
57
Chapter 3 : The Boundary Setting of Collaborative Governance Regimes
-- Why Are Civic Organizations Often Excluded?
Introduction
Collaborative governance has drawn wide attention in both theory and practice. It has been
defined in different ways, but one key feature is that it brings together organizations from public,
business and civic sectors in policy making and implementation (Ansell & Gash, 2007; Bryson et
al., 2006; K. Emerson et al., 2012). Almost all current theoretical models of collaborative
governance regard inclusiveness as an important element. For example, Ansell and Gash (2007)
consider inclusiveness as part of institutional design in their model; K. Emerson et al. (2012)
argue that the inclusion of members is a precondition of principled engagement in collaborative
processes. Broad participation from non-state parties is regarded as both a defining feature and
an advantage of collaborative governance (Ansell & Gash, 2007; Klijn & Skelcher, 2007).
Getting the "right" participants in collaborative governance is the "first condition of successful
collaboration" (Chrislip & Larson, 1994; K. Emerson et al., 2012). Including all relevant parties
may increase the "output legitimacy" by making policy making and implementation more
effective (Gunton, 2003; Hendriks, 2008). In addition, broad inclusion also enhances "input
legitimacy" by increasing the degree of descriptive representativeness (Hendriks, 2008).
Collaborative governance involves the exercise of power, and thus from a procedural justice
perspective all parties whose interests may be affected should be able to participate (Hicks, 2002).
58
Collaborative governance may help strengthen democracy by incorporating stakeholders who are
not part of the traditional iron triangle (Bogason & Musso, 2006; Klijn & Skelcher, 2007). For
example, citizens or civic organizations may be better connected to policy-making processes.
Despite the strong normative and instrumental foundations of inclusion, the reality is that some
organizations, especially civic organizations, are often partially or wholly excluded. A number of
studies have documented the exclusion of important stakeholders in collaborative governance.
For example, Hendriks' (2008) study of Dutch energy transition networks showed that
government and industry elites dominated the reform while small businesses and societal actors
were excluded. The study of urban governance networks in Los Angeles showed that civic
organizations were "politically vulnerable" and were often bypassed by city administration
(Musso et al., 2011). A study of the National Coal Policy Project also showed that important
stakeholders such as United Mine Workers were excluded from policy discussions (Gray & Hay,
1986). Similarly, Lowndes and Skelcher (1998) found that voluntary and community
organizations were often excluded in urban regeneration partnerships in the UK.
What may explain the inclusion/exclusion of certain actors m collaborative governance?
Conventional explanations are often based on either resources or power. Resource-based
explanations maintain that stakeholders will be included if they have indispensable resources for
achieving network goals (Scharpf, 1978). Boundary setting is thus considered as a rational
process. Power-based explanations highlight the political process of inclusion and argue that
59
powerful organizations, m order to maintain their own dominating status, may intentionally
exclude certain stakeholders (Purdy, 2012; Tett et al., 2003). Scholars have also pointed out other
possible reasons: in some cases, organizations are excluded on the basis of reducing group size
and maintaining a balance between efficiency and inclusiveness (Huxham & Vangen, 2000;
Vangen et al., 2014). Organizations may also self-exclude due to the lack of resources or
expertise (Hendriks, 2008); some organizations may use alternative venues instead of engaging
in collaborative processes in order to achieve their goals.
The dynamic process of inclusion and exclusion of certain organizations in governance is defined
in this paper as boundary setting (van Nuenen, 2007). In the collaborative governance literature,
boundaries are often taken as a given without much attention to how they are set in the first place.
This important question requires systematic investigation: Why are some organizations included
in governance networks while others are excluded? Drawing on the literature from general
organization theory, urban regime theory, nonprofit management and network research, this
paper attempts to build a new theoretical framework to explain the boundary setting of
collaborative governance and test the framework with data collected from neighborhood
governance networks in Beijing.
Theoretical Framework
Although some scholars argue that governance and networks are very closely related because
governance implies the existence of networks of actors (Kickert et al., 1997; R. A. Rhodes, 1997;
60
Stoker, 1998), a distinction between collaborative governance and governance networks should
be made in order to clarify the confusion surrounding the two concepts. Contrary to the concept
of government which emphasizes formal political institutions, collaborative governance, as a
species of democratic governance, is a process of governing in which organizations interact with
one another in policy formulation and implementation (Ansell & Gash, 2007; Klijn, 2008; Stoker,
1998). The organizations involved in collaborative governance and the interconnections between
them form interorganizational structures, and these structures are understood as governance
networks or broadly-defined "collaborative governance regimes" (K. Emerson et al., 2012; Klijn,
2008; Klijn & Skelcher, 2007; Vangen et al., 2014). In other words, collaborative governance is
the process that takes place within governance networks (Klijn, 2008). This is consistent with
Giddens' (1979) structuration theory, which suggests that the structural properties of social
systems are often the outcome of interactions between individuals and groups. On the other hand,
the structural properties greatly affect the patterns of interactions and how the processes work
(Vangen et al., 2014). For example, if some stakeholders whose interests are affected are
excluded from governance, then their interests and voices may not be taken into account in
decision-making (Ansell & Gash, 2007; Chrislip & Larson, 1994). Setting boundaries of
collaborative governance is to decide which organization should be included in the process of
governmg.
The inclusiveness of governance networks has important implications for democratic governance.
From the perspective of procedural justice, those who are affected by a policy should have the
61
right and opportunity to participate in its formulation and implementation; it is hard for those
who are excluded to have their voices heard and concerns taken into account (Bogason & Musso,
2006; Hendriks, 2008; S0fensen & Torfing, 2005a). Hendriks (2008) argued that inclusion has
implications for both "input legitimacy" and "output legitimacy": it helps make better policies by
considering a broader knowledge base and helps reflect the will of all constituents. Inclusion has
at least two levels: symbolic inclusion and substantive inclusion (Guo & Musso, 2007).
Symbolic inclusion means that "network participants should reflect or mirror affected
populations" (Hendriks, 2008, p. 1012). Affected populations should at least have some
organizations to represent them. Substantive inclusion goes one step further and suggests that
participants can affect policy making in a meaningful way.
Scholars have studied the inclusion of collaborative governance from two major perspectives.
For some scholars, inclusion is a rational process largely driven by the needs for resources.
Scharpf (1978), for instance, argued in a pioneer study that organizations will be included in
policy-making if they have indispensable resources for achieving network goals. In his view, it is
important to correctly identify "necessary participants" in policy-congruent networks so that
policy formulation and implementation can be successful. The activation of certain actors in
governance networks will help to incorporate necessary resources such as money, information
and expertise. Selective activation can be used as an important network management technique
in order to change network dynamics, shift the influences of existing actors and facilitate fluid
leadership roles (Agranoff & McGuire, 2001; Klijn & Teisman, 1997). Johnston et al. (2010)
62
used multiagent modeling to study how to best manage the inclusion processes in order not to
destabilize collaborative networks. They tested two strategies of managing the inclusion of new
members in collaborative networks: deliberative planning and thoughtful inclusion. They found
that networks can maintain effectiveness if new members have sufficient time to observe the
collaboration before participating and if new members can build rapport with a subset of network
members.
In contrast, inclusion has often been considered as a political process which is fraught with
conflicts. Collaborative governance seems to indicate an idealized model of consensus-based
decision making and implementation, yet in reality coordinating organizations with different
resources and interests is conflictual in nature (Tett et al., 2003). Klijn and Teisman (1997)
conceptualized the complex interactions between all involved actors as "games" in which actors
often have different or even conflicting goals and interests. Actors tend to behave strategically in
order to protect and maximize their interests. Their strategies may involve excluding certain
actors from governance primarily because of the imbalances of power and resources among
network members (Ansell & Gash, 2007; Bradford, 1998). Organizations come into play with
different resources and thus occupy different power positions. In order to maximize their
interests and influences, powerful organizations may use their power to restructure networks by
adding new members or blocking certain members (Ansell & Gash, 2007; Klijn & Teisman,
1997). Hardy and Phillips (1998) argued that the identification of participants is not an objective
process but determined by whether dominant members allow less powerful members to
63
participate. Therefore, inclusion is largely determined by power dynamics within an
interorganizational domain (Gray & Hay, 1986). Gray and Hay's (1986) study of the National
Coal Policy Project showed that the decision on whom to include was mainly based on the
consideration of how new members may change the power balances of the interorganizational
collaboration. Network research also suggests that "closedness" is a salient feature of governance
networks - network members intentionally or unintentionally exclude other members from
network processes (Koppenjan & Klijn, 2004; Schaap & van Twist, 1997).
Among all the public, business and c1v1c stakeholders in collaborative governance, c1v1c
organizations are most likely to be excluded. Their constituency is diffuse and has diverse
interests, making it difficult for them to speak with one voice (Echeverria, 2000). Civic
organizations are thus in a weak position compared with more organized interests such as local
governments and business organizations. In addition, civic organizations do not have the political
authority or the economic resources that government and business organizations have (C. S. King,
Feltey, & Susel, 1998). As already noted, the exclusion of civic organizations in governance is
well documented in the literature. This paper studies the boundary setting of governance
networks by specifically focusing on why civic organizations are often excluded.
What factors may explain the inclusion/exclusion of civic organizations? One factor is external
pressure. Organizations are open systems that are subject to environmental influence (R. W. Scott,
2003). Public organizations may be more susceptible to external influence because they have to
64
deal with multiple stakeholders and conflicting demands (Bryer, 2007). In the case of
bureaucracy-citizenship relationships, the bureaucratic responsiveness literature has long argued
that stakeholder pressure, which may come from elected officials, media, nonprofits, business,
and citizens, forces bureaucrats to respond to citizens' needs (Yang & Callahan, 2007). Yang and
Pandey's (2007) empirical test confirmed that pressure from elected officials and the
public/media are positively related to bureaucratic responsiveness. Although collaboration is one
step beyond responsiveness and inclusion is arguably one step beyond collaboration (Quick &
Feldman, 2011; Vigoda, 2002), it is reasonable to expect that civic organizations may be included
if external pressure is strong enough to overcome barriers.
The network management literature has proposed that resources and their opposite - threats -
may also explain the boundary setting of networks. Resource dependence theory maintains that
each organization may have information, capital or knowledge that other organizations need for
survival or goal achievement (Pfeffer & Salancik, 1978). Organizations depend on others for
critical resources, and those in possession of important resources have more power over others
(Benson, 1975; R. M. Emerson, 1962). Therefore, certain actors can be included if they have
indispensable resources for achieving network goals. On the other hand, existing network
members may be selectively deactivated if they are perceived as threats (Scharpf, 1978). The
inclusion of new members with new resources may change the current distribution of power, and
as a result, some members are likely to lose their privileged positions. In addition, new
participants may bring their own agenda and lead networks to stray from their original directions
65
(Huxham & Vangen, 2000). Therefore, current members in privileged positions may try to block
the inclusion of new members if these new members are perceived as threats. In the case of civic
organizations, their political and economic resources are often negligible compared with business
organizations and the government, but their threat can be substantial. They may have diverse
demands and greatly increase the costs of negotiation and coordination. Therefore, civic
organizations with fewer economic or political resources and diffuse interests are often excluded
in policy processes (Hendriks, 2008).
Another factor that may influence the inclusion of civic organizations 1s the existence of
government-business coalitions. Lindblom (1977) argued that business interests enjoyed a
privileged position in urban governance because of their structural and instrumental power.
Structural power originates from business interests' status as tax bases and job creators;
instrumental power comes from their support to politicians' political campaigns. In contrast,
civic organizations usually have far fewer resources and do not have such structural or
instrumental power. The difficulty to organize collective actions may worsen the problem.
Government and business interests often establish informal coalitions, or urban regimes, to
govern cities. A number of studies have documented the exclusion of civic or community
organizations from urban regimes (Haughton & While, 1999; Sites, 1997; Stone, 1989). Stone's
(1989) study of the governance of Atlanta suggested that civic organizations of African
Americans exerted little influence over urban governance until they were given the rights to vote.
66
The fourth factor is the formalization of civic organizations. Since inclusion is largely affected
by power dynamics, organizations often need to fight hard in order to be included in
collaborative governance (Hardy & Phillips, 1998). Highly formalized civic organizations are
more likely to have the capacity to win the fights. Civic organizations are often informal: they
are often volunteer-based and have consensus-oriented decision-making processes that do not
emphasize accountability or efficiency (Kreutzer & Jager, 2011 ). The informal nature of civic
organizations may be favorable for maximum volunteer participation (Hall, 2006), but it comes
with certain costs: organizational decisions are unstable, and they often lack the capacity to
represent citizens and the expertise to engage in technical activities. The formalization of civic
organizations through professionalizing staff and adopting rational management practices have
been found to increase organizational capacity in terms of attracting government funding (Suarez,
2011), mission fulfillment (Eisinger, 2002), sustaining the organization themselves, maintaining
social movements and building coalitions with other organizations (Staggenborg, 1988).
Similarly, more formalized civic organizations are more likely to be perceived as having
legitimacy and expertise and are thus included in governance networks. Therefore, high
organizational capacity may help civic organizations to fight for membership in governance
networks.
The fifth factor is civic organizations' nonjinancial resources, especially the social and human
capital of members. If members have abundant human and social capital, they may be more
skilled in mobilizing their constituents to put pressure on extant network members and they may
67
have political connections to make their voices heard. Connections to other political or civic
organizations may serve as sources of external pressure or as bridges to increase mutual
understanding between potential and extant network members. The social and human capital may
thus help civic organizations to be included in governance. The role of nonfinancial resources in
facilitating collaboration has attracted scholars' attention. For example, Weiss, Anderson, and
Lasker (2002) studies how nonfinancial resources, measured by items such as skills and expertise,
and connections to political decision makers, government agencies, or other organizations or
groups, may contribute to partnership synergy when organizations are collaborating with each
other.
Research Context
A systematic investigation of factors influencing the boundary setting of governance networks is
yet to be conducted. This is an important research topic not only in democratic societies for the
purpose of strengthening democracy, but also in authoritarian states like China. The focus of this
paper is on the inclusion of Homeowners' Associations (HO As) in neighborhood governance
subsequent to the sweeping housing reform in China.
China's neighborhoods used to be firmly controlled by the government. In the planned economy
era, danwei (work unit) had the responsibility of providing housing for employees. The housing
bureaus attached to danwei or cities were responsible for maintaining these public housing
complexes. Street Offices (SOs) were the lowest level of governments and usually were in
68
charge of several Residents' Committees. Residents' Committees (RCs ), which were supposed to
be "self-management, self-education and self-service mass organizations," were actually strictly
controlled by SOs. RCs were responsible for a variety of administrative functions delegated from
SOs, such as family planning, neighborhood safety and organizing local elections.
Housing reform, an important part of China's overall economic reform, greatly changed the
previous governance structure. The reform was launched in the late 1990s and a housing market
was created. Commercially developed neighborhoods have since become the dominant form of
neighborhoods. New players emerge and actively engage in neighborhood affairs. Homeowners
can establish Homeowners' Associations (HO As) to address the problem of managing communal
properties and protecting property rights. Business interests including developers and property
management firms (PMFs) become important stakeholders in neighborhood governance.
Interorganizational relationships are highly dynamic. The Property Rights Law and relevant
regulations are not very well enforced in China. Developers and property management firms
often gain enormous revenues from infringing upon homeowners' communal properties, for
example, by selling or renting out communal properties without giving homeowners any
compensation. To claim or protect their communal properties, homeowners have to establish
HOAs as their legal representatives. This is challenging for several reasons. First, homeowners
have to overcome collective action problems. In new urban neighborhoods which are "societies
of strangers" (Forrest & Yip, 2007), it is very difficult to get homeowners organized. Second,
69
many SOs are not cooperative and become a major obstacle for homeowners. They either deem
homeowners' organizing efforts as threats to social stability or form informal coalitions with
developers in order to benefit from infringing on homeowners' communal properties. The current
regulations and laws give street-level goverrunents too much influence in establishing HOAs.
According to relevant regulations in Beijing, the founding of HOAs has to be done under the
direction and supervision of SOs. SOs are responsible for naming the head of preparatory groups
and for ensuring that all application materials are correct. SOs thus have the authority to stop
homeowners from organizing, or at least to restrain their efforts considerably.
Neighborhood organizations with different goals, interests, and resources use different strategies
to engage with one another in order to maximize their interests. The latest numbers show that
about 75% of the commercially-developed neighborhoods in Beijing do not have HOAs. Without
any organizations to represent their interests, homeowners have a hard time getting their voices
heard. Only 25% of all neighborhoods have HO As, and some of them work effectively and play
important roles in neighborhood governance. This provides an ideal field to study what caused
the difference in the inclusion of HO As in neighborhood governance networks. The findings may
shed new light on how boundaries of governance networks are set.
70
Research Methods, Data and Operationalization of Variables
In order to study the determinants of the inclusion of civic organizations m neighborhood
governance, I employed fuzzy set Qualitative Comparative Analysis (fsQCA) to study 22
neighborhood networks in Beijing. The fsQCA method is based on Boolean algebra and can
systematically compare cases in order to find out configurations of factors that lead to certain
outcomes.
Compared with the conventional regression analysis or case study method, the fsQCA method
has several unique strengths. First, it is very suitable for exploring causal complexity. The
mainstream regression analysis aims to isolate the net effect of each independent variable on the
dependent variable, holding all other independent variables constant (Fiss, 2007; Ragin, 2008).
The assumptions are that, first, these independent variables compete with one another to explain
variances in the dependent variable (Fiss, 2007), and second, one variable alone is sufficient to
produce influence over the dependent variable. However, this is hardly the case in the real world.
It is usually the combination of factors, rather than one single factor, that produces influence over
a certain dependent variable (Ragin, 1989). It is also likely that when combined together, factors
may produce synergistic effects over dependent variables. The multivariate regression method
tries to capture synergistic effects by using two-way or three-way interaction terms. However,
this approach is not without its problems. The first problem is that it can work only on the
condition that all interaction effects are known in advance so that interaction terms can be
71
constructed (Ragin, 1989). The second problem is that terms that go beyond two-way interaction
become extremely difficult to interpret, which makes three-way interaction terms become the
actual boundary of regression analysis (Fiss, 2007). However, as Fiss (2007) has pointed out,
there are no good reasons to exclude the possibility of more than three causal factors working
together to influence outcomes. Therefore, regression analysis may be suitable to examine the net
effect of economic development on democracy while holding other factors such as culture
constant, but it may not be able to investigate how the combination of economic development,
culture, and geographical politics affects democratization (Ragin, 1989).
The fsQCA method is especially well suited for dealing with causal complexity. It treats cases as
configurations of characteristics and takes a holistic view toward cases (Fiss, 2007; Ragin, 1989).
Unlike multivariate statistical analysis, "the relations between parts of the whole cases are
understood with the context of the whole" (Ragin, 1989, p. x). The combination of characteristics
is not understood as a collection of different values of independent variables but as a unique
situation that affects the outcome (Ragin, 1989). The fsQCA method can be used to discover
configurations of characteristics that produce a certain outcome and thus better capture the
synergistic effects. In addition, it acknowledges equifinality which means that there may be more
than one path leading to a certain outcome (Fiss, 2007, 2011; George & Bennett, 2005).
Another major strength of the fsQCA method is that it can deal with medium numbers of cases
that is difficult for traditional qualitative case method and large-N quantitative studies to analyze.
72
For example, the fsQCA method can process 15 to 40 cases, which are too many for the
comparative case method to manage and too few for regression analyses. This feature is very
useful for network research because it is time- and resource- consuming to collect large-N data of
networks which usually consist of a large number of organizations. Conventional case studies
offers insightful conclusions, but its generalizability is limited due to the small number of cases
studied.
Sampling
The fsQCA method is based on the logic of combinatorial causation and has different
requirements for case selection. As Ragin (1989) argued, "when causal arguments are
combinatorial, it is not the number of cases but their limited variety that imposes constraints on
rigor" (p.13). In other words, the fsQCA method does not require a large number of cases as the
mainstream statistical analysis does; Instead, cases should exhibit as many logically possible
combinations of characteristics as possible. The relative frequency distribution of different types
of cases is not important. For example, the fsQCA method will consider the two combinations of
conditions that produce a certain outcome as valid accounts even if one of them has a very high
frequency but the other does not (George & Bennett, 2005; Ragin, 1989). The configuration of
conditions that has a low frequency of occurrence still represents a possible causal path that
carries implications in the real world, but the mechanism can easily be ignored by multivariate
analysis. Again, as Ragin (1989) argued, "more important than relative frequency is the variety
of meaningful patterns of causes and effects that exist" (p. 52).
73
Based on the above understanding of the fsQCA method, I did not use the random sampling
approach that can cause selection bias in small-N studies (G King et al., 1994). In order to deal
with the limited diversity problem that negatively affects the application of the fsQCA method, I
used purposeful sampling to select 22 cases that exhibited different combinations of the outcome
variable and causal conditions. Homeowners were included in neighborhood governance at
different degrees in these neighborhoods. This sampling method helped to capture the wide range
of variation in the outcome and causal conditions.
Data collection
Semi-structured interviews were used in data collection. These interviews were based on an
interview guideline with standardized questions on the outcome and causal conditions. The use
of this structured component allowed for "the comparability of data across individuals, times,
settings and researchers, and are thus particularly useful in answer variance questions" (Maxwell,
2005, p. 80). However, data collection did not stop when interviewees assigned values to all the
questions. A deep understanding of each case and the context was necessary for the successful
application of the fsQCA method. In order to gain a deeper understanding of the cases, I asked
interviewees for their reasons behind their evaluations, which helped to stimulate in-depth
dialogues about neighborhood governance. Open-ended questions were raised in order to gain
information beyond the variables. In addition, I also visited each of the 22 neighborhoods in
order to get first-hand experience and to double check the information gained from interviews.
74
Operationalization of variables
Inclusion The outcome variable is "inclusion," which is used to measure both symbolic
inclusion and substantive inclusion. In this research context, homeowners play dramatically
different roles in governing their neighborhoods. In some cases, homeowners could not establish
Homeowners' Associations (HOAs) to represent their interests due to barriers set by street-level
governments or property management firms, and thus could not achieve even symbolic inclusion,
while in some cases organized homeowners could exert substantial influence in decision makings
regarding neighborhood affairs. Many neighborhoods fall somewhere in between these two
extremes. "Inclusion" is thus a multi-dimensional variable that cannot be measured by a single
unambiguous external indicator. It was thus measured by an index that tries to capture both levels
of inclusion: first, symbolic inclusion was measured by whether HOAs are established;
establishing HOA is a precondition for homeowners to have a presence in governance; if HO As
are established, homeowners are in a stronger bargaining position and are thus more likely to
play a meaningful role. Substantive inclusion was measured by three items: the first is whether
HOAs have ever succeeded in claiming any communal properties that were previously usurped
by other parties in their neighborhoods; the second is whether HOAs hire property management
firms, which is one of the legal authorities ofHOAs; the last item is whether other organizations
actively consult homeowner leaders on neighborhood affairs. A positive answer to the questions
was assigned a value of "l", and a negative answer was assigned a value of "O". Therefore, the
index score ranged from 0 to 4.
75
The first causal condition is threat to other parties. In this project, HOAs' threats to current
members were measured by the value of properties that were in dispute between homeowners
and other parties. If the value of properties in dispute was large, organized homeowners would
claim these properties and thus greatly harm other parties' interests. As a result, homeowners are
very likely to be perceived as threats by other organizations. The parties benefiting from the
status quo may try their best to block or at least resist the participation of HO As in neighborhood
decision-making processes. Another factor worth considering is resources according to the
resource-based framework. However, in reality, homeowners do not have voting power as
citizens m democracies do; they also have far fewer economic resources compared with
developers and property management firms. Their participation could not contribute many
resources to neighborhood governance, so this factor was not included in analysis.
The second causal condition is government-business coalition. This was treated as a dummy
variable with 1 representing the existence of a government-business coalition and 0 otherwise.
This information was obtained from interviews with homeowners and verified in my fieldwork.
It was almost impossible to get a definitive answer from either developers or SOs about whether
they had an informal coalition. However, homeowners had pretty good ideas about the
relationships between governments and businesses based on how the two reacted to
homeowners' organizing. My fieldwork in each neighborhood also provided valuable
information on how government interacted with business.
76
Another relevant factor is pressure. External pressure on SOs and business interests may come
from courts and the media. Pressure was thus measured by counting the number of times that
homeowners petition higher-level governments (district- and municipality-level governments)
and the number of times that homeowners sue local governments and business organizations.
Generally, the sum of the two numbers is positively related to the external pressure that is put on
current network members.
The fourth factor is the formalization of HO As, which was measured by the following indicators:
whether there are established organizational rules, whether HOAs have regular meetings,
whether their organizations have stayed intact for a certain amount of time, and whether HO As
have paid professional workers. Paid professional workers in voluntary organizations such as
HOAs can help organizations to work effectively and efficiently. Positive answers to the above
questions received a point of 1 and negative answers received a point of 0. Therefore, the value
of this causal condition ranged from 0 to 4.
The fifth causal factor is the neighborhood socio-economic status (SES). This variable serves as
a proxy of nonfinancial resources, or the availability of human and social capital, which is hard
to be measured directly. Neighborhood SES may be particularly relevant in this research context.
Homeowners living in high-end neighborhoods may be lawyers, journalists and professors.
These people are better educated and have wider social networks. They are usually more aware
of rights and thus are more likely to participate in neighborhood affairs. They bring their skills
77
and expertise m orgamzmg homeowners and participating m neighborhood affairs.
Neighborhood socioeconomic status was measured by average price per square meter (APPSM)
in recent home transactions. Average price per square meter (APPSM) was used in other studies
as a predictor of neighborhood SES (F. Wang, Yin, & Zhou, 2012).
Analysis and Results
Table 3.1 presents the descriptive statistics and the bivariate correlation between variables. As we
expected, threat and government-business coalitions were negatively related to inclusion. lf
HO As pose a serious threat to other network members, then they are less likely to be included in
governance networks. lf there is a government-business coalition, then HO As are also less likely
to be included. Formalization and neighborhood SES were positively related to inclusion,
suggesting that highly formalized HOAs and HOAs in high-SES neighborhoods are more likely
to be included. One puzzling result was that external pressure is negatively related to inclusion,
suggesting that HO As are less likely to be included if they put more pressure on other network
members. This was contrary to our expectation, but of course this was a bivariate correlation that
did not control for other possible confounding variables. The highest correlation coefficient
between causal conditions was -0.64, suggesting that multicolinearity was not a big threat.
78
Table 3.1 Descriptive Statistics
Variable Mean s.d. 1 2 3 4 5 6
I.Threat
8.2 12.1 1
(in millions yuan)
2.Government-business
coalitions
0.45 0.51 0.22 1
3. Pressure 0.82 1.14 0.01 0.40 1
4. Formalization 2.18 1.5 -0.32 -0.64 -0.04 1
5.Neighborhood SES
42.5 16.7 -0.40 -0.06 0.44 0.46 1
(in thousand yuan)
6.Inclusion 1.91 1.72 -0.53 -0.71 -0.18 0.77 0.28 1
To better capture the complex interactions between variables, the fsQCA analysis was conducted
with the software fsQCA 2. 5. As a first step of analysis, raw data needed to be calibrated based
on social knowledge, social science theories and substantive knowledge of cases (Ragin, 2008).
Raw data were converted to values ranging from 0 (full nonmembership of a set) to 1 (full
membership of a set) after calibration.
According to Ragin (2008), the key tool to systematically compare cases is the truth table. The
truth table lists all logically possible combinations of causal conditions (the total number of
combinations ~ 2\ where k is the number of conditions). In fuzzy set analysis, cases may have
different membership scores (ranging from 0 to 1 after calibration) in each set (causal condition).
We need to examine whether the degree of membership of a combination of conditions is a
consistent subset of the degree of membership in the outcome. The consistency cut-off is usually
defined by researchers, and Ragin (2008) recommended that the cut-off value should not be
79
smaller than 0. 75, meaning that at least 75% of the cases with a certain combination of causal
conditions should exhibit the outcome.
The fsQCA analysis can result in three different types of solutions: complex, intermediate and
parsimonious, depending on different approaches in counterfactual analysis. One problem that
fsQCA faces is limited diversity because naturally occurring social phenomena usually do not
exhibit all logically possible combinations of causal conditions (Ragin, 2008). Some
combinations of factors lack empirical cases. There are three approaches to deal with the
problem. The conservative strategy treats the combinations without empirical cases as false and
then excludes them from analysis, resulting in so-called complex solutions. The complex
solutions are often needlessly complicated. One way to simplify the complex solutions is to
incorporate some "easy counterfactuals" which "refer to situations in which a redundant causal
condition is added to a set of causal conditions that by themselves already lead to the outcome in
question" (Fiss, 2011, p. 403). Incorporating easy counterfactuals results in intermediate
solutions that maintain a balance between barring counterfactuals completely and incorporating
difficult counterfactuals. Intermediate solutions can be further simplified by incorporating
difficult counterfactuals that refer to "situations in which a condition is removed from a set of
causal conditions leading to an outcome on the assumption that this condition is redundant "
(Fiss, 2011, p. 403). Using difficult counterfactuals would result in the parsimonious solutions. I
have presented the intermediate solutions and parsimonious solutions in this paper (For a
detailed explanation of counterfactual analysis, see Ragin (2008)).
80
A detailed explanation of the truth table approach to analyzing fuzzy sets can be found in
Ragin (2008). In this research, I set the threshold of frequency at 1 due to the small sample size.
The consistency threshold was set at 0. 80, which is greater than the minimum requirement of
0.75. I used the notation system introduced by Ragin and Fiss (2008) to present the results. Black
circles represent high membership of a causal condition, and circles with "x" in them indicate
low membership of a condition. Large circles indicate core conditions, which are those that
appear in both parsimonious and intermediate solutions; small ones represent peripheral
conditions, which are those that appear only in the intermediate solutions (Fiss, 2011 ). Blank
spaces indicate that the conditions are irrelevant. According to Fiss (2011 ), core conditions play a
stronger role in affecting outcomes, and peripheral conditions play a weaker role.
Table 3.2 shows two configurations of factors that lead to the inclusion of HOAs in governance.
One was high neighborhood SES combined with high formalization and no government-business
coalitions. High formalization was the core causal condition in the path. Threat to other parties
and external pressure were not relevant in this situation. This combination indicated that in high
SES neighborhoods where HOAs had high formalization and where no business-government
coalitions existed, HO As were very likely to be included in governance. HO As were included in
98% of the cases with this configuration, and this configuration alone covered 48% of all the
cases in which homeowners were included. The second configuration was the combination of
low neighborhood SES, no government-business coalitions, low external pressure on
81
governments and low threat. This meant that in low-SES neighborhoods where there was no
government-business coalition and low external pressure, the HOA still could be included if they
were not perceived as a serious threat by other organizations. The core condition was low threat
to the status quo. This causal path did not happen very often in the real world because the unique
coverage of this causal configuration was only 0.07.
Table 3.2 Configurations For The Inclusion Of Civic Organizations
configuration solutions
Threat
Government-business
®
coalitions
Pressure
Formalization
e
Neighborhood SES
•
Consistency 0.98 1.00
Raw coverage 0.66 0.25
Unique coverage 0.48 0.07
Overall Solution
consistency
0.98
Overall solution
0.73
coverage
Note: Black circles represent high membership of a causal condition, and circles with " x" in it indicate
low membership of a condition. Large circles indicate core conditions, and small ones represent
peripheral conditions.
The fsQCA method is built on causal asymmetry, which means that factors leading to the
presence of an outcome may be different from factors that cause the absence of an outcome (Fiss,
82
2011; Ragin, 2008). It is thus necessary to analyze the negation of the outcome - exclusion in the
governance. Table 3.3 presents two configurations that lead to the exclusion of HOAs in
governance. Two core causal conditions were present in each causal path: high threat and low
formalization. Two peripheral causal conditions, the presence of government-private coalition
and low external pressure, were interchangeable in causing the exclusion of HOAs. The first
configuration suggested that HOAs would be excluded from governance if their participation
posed a serious threat to other parties' interests, the HOA had low formalization and they failed
to put any pressure on other parties. Neighborhood SES and government-business coalition were
irrelevant in the above situation. The configuration had a consistency score of 1 and a unique
coverage of 0.17. The second configuration indicated that homeowners would be excluded from
governance if their participation posed serious threat to other parties' interests, they had low
formalization, and there were government-business coalitions. The second solution had a unique
coverage score of 0.37, meaning that this solution alone could explain 37% of all cases in which
HOAs were excluded from governance. The two solutions accounted for 84% of all cases m
which citizens were excluded from governance, and the consistency score was as high as 0.96.
83
Table 3.3 Configurations for The Exclusion OfHOAs
configuration solutions
Threat
e e
Govenunent-business
•
coalitions
Pressure
0
Formalization
®
Neighborhood SES
Consistency 1.00 0.95
Raw coverage 0.46 0.66
Unique coverage 0.17 0.37
Overall Solution
0.96
consistency
Overall solution
0.84
coverage
Note: Black circles represent high membership of a causal condition, and circles with " x" in it indicate
low membership of a condition. Large circles indicate core conditions, and small ones represent
peripheral conditions.
Discussion and Conclusion
The above analysis revealed nvo configurations of factors leading to the inclusion of HOAs in
governance and nvo different causal recipes leading to the exclusion of HOAs. Both inclusion
and exclusion are highly contingent on various environmental and organizational characteristics.
The configurations of factors revealed nonadditive and nonlinear relationships benveen causal
conditions and the outcome variable.
84
High degree of formalization was identified as a core causal condition in the analysis of
inclusion. The Beautiful Garden neighborhood was such a case. It is a high-end neighborhood
located in the southwest part of Beijing. The developer and its property management firm
seriously violated homeowners' property rights by intentionally overstating the actual areas of
apartments and by charging extremely high management fees. Led by some homeowners who
were professors, lawyers and retired government officials, homeowners pioneered establishing
Homeowners' Association in Beijing. The HOA had strong leadership and exceptional degree of
formalization: they had regular meetings to discuss important neighborhood affairs, and also
hired a paid secretary to deal with daily affairs. Therefore, the HOA has high organizational
capacity. Although the SO and RC were not very supportive of the HOA, they did not form a
coalition with the developer or property management firm. The HOA first sued the developer and
won the lawsuit. The developer lost the case but did not want to give in easily. In order to force
the HOA to compromise, the developer and its property management firm left the neighborhood
abruptly without giving any notice. As a result, the HOA was under great pressure to provide
security, sanitation and facility maintenance service, but it managed to find another property
management firm to provide these services before officially firing the developer-owned property
management firm. The HOA then hired a new one based on the HO A's terms through a bidding
process. Since that time, the HOA has played a central role in neighborhood affairs. Although the
HOA posed a serious threat to the developer's interests, the high organizational capacity helped
homeowners to overcome the resistance. The core of the second causal path to inclusion is low
85
degree of threat. The Tian-Tian neighborhood was such a case. The developer pulled out of the
neighborhood long ago and was no longer involved in neighborhood affairs. The SO did not have
direct interests in the neighborhood either. Therefore, there was no government-business
coalition since neither party could benefit from engaging in neighborhood affairs. This created
plenty of room for the HOA to manage neighborhood affairs even if they did not put external
pressure on governments.
In contrast, low formalization and high threat were identified as core causal conditions in leading
to exclusion. Low external pressure and the existence of government-business coalitions were
peripheral conditions given the two core conditions. The Wan-Quan and Bai-Zi neighborhoods
can help to illustrate the causal paths. HOAs in both neighborhoods posed serious threats to
developers: In Wan-Quan, the developer overstated the actual areas of the apartments and
deceived homeowners into overpaying. The properties in dispute were about 20 million yuan; in
Bai-Zi, the developer and its property management firm received annual revenue of about 10
million yuan by selling and renting out homeowner's communal properties without giving
homeowners any compensation. Organized homeowners may challenge business interests and
cut off the latter's huge revenues derived from violations of homeowners' properties. The
business interests were strongly motivated to exclude homeowners. HOAs were established after
overcommg numerous difficulties. However, they were gradually paralyzed in both
neighborhoods for various reasons, including developers' threats, co-optation and members'
losing interests and commitments m runnmg their HOAs. There were differences too:
86
Homeowners in the Wan-Quan neighborhood sued the developer for deceiving homeowners into
paying for overstated areas. The fact was clear and homeowners had a good chance of winning
the case. However, the government-business coalition was so strong that the court inexplicably
refused to hear the case. This situation had been in deadlock for years at the time of this study. In
Bai-Zi, the problem was that the paralyzed HOA could not put any external pressure on local
governments by organizing meaningful protests or demonstrations. As a result, HOAs were
excluded from governance in both cases.
In the current literature, the boundary setting of governance networks is often explained by a
resource-based framework or a power-based framework. Resource-based explanations maintain
that organizations will be included if they have indispensable resources for achieving network
goals (Scharpf, 1978). This is a useful explanation based on the resource dependence theory and
recognizes the interdependence of organizations in addressing complicated problems. However,
this explanation does not give enough attention to how the inclusion of new resources may
change the power dynamics in networks. The resource dependence theory also points out that
power resides in the other's dependency (R. M. Emerson, 1962); therefore, if a network of
organizations depends on a potential member for critical resources, then this new member will
have immense power over extant members. Including this new member will probably change the
extant distribution of power and influence in the governance network. It is hard to imagine that
privileged organizations will be willing to give up or share their power especially if they have
their own agendas to advance. Therefore, the power-based explanation maintains that the
87
inclusion of new members in collaborative governance is hardly a rational process based purely
on rational calculation but rather a political process based on interest calculation (Agranoff &
McGuire, 2001; Purdy, 2012). The findings in this paper partly support the argument that
powerful organizations may exclude organizations if they are perceived as threats.
This paper attempts to transcend both the resource-based explanation and the power-based
explanation. Neither framework gives us a comprehensive explanation of boundary setting.
Based on findings in the paper, a capacity-threat framework to explain the boundary setting of
collaborative governance is proposed. When assessing potential members to include, extant
members will consider not only the resources that they may bring but also the threats that they
may pose. Extant members, especially those that benefit from the status quo, will try to block
new members if these are perceived as serious threats. However, this does not mean that these
organizations will always to be excluded. To be successfully included, these potential members
have to have high degree of formalization, which can increase organizational capacity, so that
they can overcome the resistance from extant members. lf they lack the organizational capacity,
then they will not be able to overcome the resistance and will be excluded. Of course, our
analysis also suggests that if these new members are not perceived as threats, then they may also
be included in governance. Therefore, the boundary setting of governance networks is a political
and dynamic process. The boundary is not static but may be constantly shifting depending on the
changes in relative power of extant and potential members. Inclusion in governance networks is
88
not permanent. Members can be deactivated if their agendas conflict with those of others, or if
they are perceived by others as a wrong partner (Agranoff & McGuire, 2001).
This political and dynamic process of inclusion has important implications for c1v1c
organizations and democratic governance. The exclusion of c1v1c organizations m both
democratic and authoritarian countries may not be coincidental but structural. With agendas
focusing on social justice, the environment and democracy, c1v1c organizations are often
perceived as threats by pro-growth coalitions formed by local governments and business interests.
These pro-growth coalitions may thus have strong motivations to exclude or marginalize civic
organizations in governance networks. A number of documented cases support the above claim,
including issues related to the governance of Ghent, Belgium (F. De Rynck, 2006), Los Angeles,
California (Musso et al., 2011) and to the Netherlands's energy reform networks (Hendriks,
2008). It is difficult for civic organizations to be included in governance networks unless they
have high capacity to overcome the resistance. However, civic organizations have several hurdles
to overcome in building their capacities: citizens need to overcome the collective action problem
and get themselves organized in the first place; second, citizens are so diverse and their interests
are so diffused that it is hard for civic organizations to speak with one strong voice (Echeverria,
2000); third, civic organizations often lack sufficient resources to enhance their organizational
capacity. In addition, even if some civic organizations are included, they may not be
representative of the full spectrum of citizens' interests in governance processes. Therefore,
contrary to the argument that collaborative governance may strengthen democracy (Bogason &
89
Musso, 2006; Klijn & Skelcher, 2007), we tend to agree that collaborative governance does not
fare well against the ideal of democracy (Dryzek, 2007; Hendriks, 2008).
This may be a more serious problem in China because of both the authoritarian rule and the
needs for economic growth. At the city level, research has suggested that the exclusion of
citizens in urban regimes is quite normal in China, and that the citizens' voices, concerns and
interests are often ignored or even sacrificed (Zhang, 2002). Things are no better at the street
level. In an authoritarian state like China, citizens have very limited opportunities to vote, so they
have even fewer political resources than do people in democracies. It is very difficult for them to
hold government officials accountable. Street-level governments are often not neutral in Chinese
neighborhoods. They establish informal coalitions with business interests on the basis of interest
exchange or the need to promote economic growth. These coalitions can be very stable and
persistent. We can call these coalitions neighborhood regimes. These regimes dominate
decisions-making regarding the distribution of interests in these neighborhoods, and decisions
are often made at the cost of homeowners.
Therefore, designers of collaborative governance initiatives should pay special attention to the
role of civic organizations in terms of both symbolic representation and also substantial
representation (Guo & Musso, 2007). In the phase of institutional design, it is important to
actively seek participation from civic organizations or citizen groups whose interests may be at
stake. Designers should bear in mind that these citizen groups are less organized and thus they
90
are at a high risk of being excluded. Some civic organizations may even self-exclude due to the
lack of expertise in the discussion of technical details (Fung, 2005). Studies have suggested that
the participation of civic organizations is more likely to happen if some assistance of resources is
offered (S0fensen, 2006). In addition, having civic organizations sitting on the table does not
mean that they have a meaningful role to play - internal power dynamics can marginalize civic
organizations, especially when their interests are not compatible with the interests of government
and business. Designers should be aware of the hidden power dynamics in policymaking or
deliberation processes and actively solicit voices and opinions from civic organizations. Scholars
have proposed several strategies to strengthen civic organizations' role in collaborative
governance, such as integrating mandatory deliberative forums in decision-making processes and
reframing policy discourse in a way that is more conducive to public participation and more
closely tied to the impacts on communities (Hendriks, 2008).
Contribution and Limitations
This paper makes several contributions to the current literature. It is one of the first attempts to
systematically investigate the boundary setting of governance networks. The paper transcends
the resource-based and power-based explanations and establishes a capacity-threat framework to
explain the dynamic process of boundary setting. The new framework helps us to better
understand mechanisms of inclusion/exclusion. In addition, the capacity-threat framework sheds
new light on why civic organizations are often excluded from governance. Citizens' interests are
often conflicting with the interests of pro-growth coalitions, and the organizational capacity of
91
civic organizations is often much lower compared with the capacity of local governments and
business organizations. The paper thus provides new evidence and theoretical explanations for
why collaborative governance may not be favorable for democratic governance.
The paper also has some limitations. One is that the small sample size compromises the
generalizability of conclusions. We studied only 22 neighborhood networks, so caution is needed
when generalizing the conclusions to other cultural and institutional contexts. More research is
needed to cover different contexts. The second limitation is that, like conventional regression
analysis, no causal mechanism can be established by using the fsQCA method. The causal
language used in this paper is the terminology of the fsQCA method (Greckhamer, Misangyi,
Elms, & Lacey, 2008).
92
Appendix One
Calibration
Ragin (2008) argued that calibration should be based on theories and substantive knowledge, and
substantive knowledge should take precedence if theoretical knowledge are not available. Direct
method of calibration was employed in this research.
Inclusion ranged from 0 to 4. The full membership of inclusion was set at 4, showing that the
HOA played a central role in their neighborhoods. The full nonmemership was set at 1, which
was the 36 percentile. A membership score of 1 suggested that the HOA played a marginalized
role in neighborhood governance. The crossover point was set at 1. 5, which was the 50 percentile,
showing a maximum degree of ambiguity with regard to whether the HOA was included or not.
Neighborhood SES was measured by housing pnce - the pnce per square meter m recent
apartment sales. The average housing price in Beijing at the time of my fieldwork at the end of
2012 was 28,000 RMB (approximately $4,500) per square meter. The full membership threshold
of high SES neighborhood was set at 60,000 RMB (86 percentile), which was more than two
times of the average price in Beijing. Neighborhoods with housing price above this level
included high-end apartment neighborhoods and townhouse neighborhoods. The full
nonmembership threshold was set at 20,000 RMB (13 percentile), which was lower than the
average price in Beijing. Neighborhoods with housing price lower than this level were usually at
93
the peripheral part of the city. People living in these neighborhoods usually came from low-SES
background: they were farmers who lost their land in commercial development processes, retired
workers or people migrated from other places to work in Beijing. The crossover point was set at
29,000 ( 45 percentile), which was slightly higher than the city average price.
The formalization of HO As had values ranging from 0 to 4. The full membership threshold was
set at 2. 5, which is the 70 percentile. The full nonmembership threshold was set at 0, which
suggested that these networks did not have an established HOA at all. Three networks had this
score. The crossover point was set at 1.5, which was the 45 percentile. Organizations at this point
had some degree of formalization, but they were not highly formalized yet.
Pressure was thus measured by counting the number of times that homeowners petitioned
higher-level governments (district- and municipality-level governments) and the number of times
that homeowners sued local governments or business organizations. The full membership
threshold was set at 2, which was the 72 percentile. In these neighborhoods, homeowners either
sued extant network members or petitioned higher-level governments multiple times; they thus
could exert significant pressure on extant network members. The full nonmembership threshold
was set at 0, suggesting that these homeowners put almost no pressure on extant network
members. The crossover point was set at 1, which suggested that homeowners tried to exert some
pressure.
94
HO As' threats to extant members were measured by the value of properties that were in dispute
between homeowners and other parties. The values ranged from 0 to as much as 40 million yuan.
The large values in dispute were due to high housing prices in some neighborhoods that were in
very good locations or due to the large areas that were in dispute. The full membership was set at
1.8 million. Although this value is much smaller than 40 million, it is not a small number to
many small-to-medium developers or property management firms. Homeowners would be
perceived as serious threat to extant network members at or beyond this point. The full
nonmembership threshold was set at 0.2 million, which was a relatively small amount of money
in the eyes of most organizations and thus would not cause serious conflicts. The crossover point
was set at 0.5 million, which was hard to tell whether it was big enough to cause extant members
to perceive HOAs as threats.
Governance-business coalition was a dichotomous variable, so it did not need to be calibrated.
Robust Check
Robustness checks were conducted in order to see how the results would change against different
specifications of calibration thresholds. I selected the thresholds that made the best sense to me
besides the one used in the original data analysis. I used the recalibrated values of one variable
and held the other variables unchanged in each robustness check. Results were shown in Table
3.4. The results were generally robust to different specifications of the crossover points. The
biggest change in consistency or coverage was not more than 5%, and no new causal paths
emerged in robustness checks. I also did a new analysis with all newly calibrated variables and,
95
again, no new causal paths emerged. However, the consistency score of the first causal path, the
combination of high SES, high formalization and no government-business coalition, was reduced
from 0.98 to 0.86.
Table 3.4 Results of Robustness Checks
Causal Alternative Alternative full Alternative Changes Biggest
condition Full nonmembership Crossover point in causal change in
membership paths Consistency
or coverage
Inclusion 4 0 2.5 No 5%
Pressure 4 0 0.8 No 5%
Formalization 4 0 l.5 No 3%
Neighborhood 55,543 38,000 18,272 No 2%
SES
Threat 3 0.5 0 No 1%
96
Chapter 4: Exploring the Determinants of Network Effectiveness
Introduction
Evaluating network effectiveness and studying its determinants have been an important topic in
network research (Meier & O'Toole, 2003; Provan & Kenis, 2008; Provan & Milward, 1995).
Provan and Milward (1995) proposed a preliminary theory of network effectiveness by
examining the effects of environmental factors and network structural characteristics. A series of
studies followed their original research and examined the effects of factors such as forms of
network governance, network age, and internal trust (Provan & Kenis, 2008; Provan & Milward,
1995; Provan & Sebastian, 1998; Raab et al., 2013). Another school of scholars studied the roles
of agency with a special focus on network management (Juenke, 2005; Meier & O'Toole, 2003;
O'Toole & Meier, 2004). A number of variables such as managerial networking (Meier &
O'Toole, 2001, 2003), management quality (Meier & O'Toole, 2002), personnel stability (O'Toole
& Meier, 2004) and network leadership (McGuire & Silvia, 2009) have been examined.
The current literature is enlightening in many ways, but it is not without its problems. Though
recently an increasing number of studies have used medium to large-N samples (O'Toole &
Meier, 2004; Raab et al., 2013; Verweij et al., 2013), single case studies or comparative case
studies were once the dominant approach to develop theory (Isett et al., 2011; Meier & O'Toole,
97
2003; Provan et al., 2007) simply because it is time- and resource-consuming to collect data on
networks, each of which involves a large number of organizations. Conclusions drawn from case
studies are inspiring, but their generalizability is compromised. A related problem is that
configurational theories of network effectiveness are still in the nascent stage. Conventional
regression models are limited in testing the complex interactions between explanatory variables
(Fiss, 2011). For example, with current statistical techniques, it is difficult to test the complex
contingency propositions that Provan and Kenis (2008) developed. Though scholars have started
to employ new methods such as set-theoretic methods to build configurational theories (Raab et
al., 2013; Verweij et al., 2013), more research is needed in order to cover different research
contexts and to test the effects of different factors. Thirdly, most theoretical development in the
field has been based on the US or European contexts. We know little about whether these models
can hold in broader contexts.
This paper attempts to further the research on network effectiveness by employing a
mixed-methods approach to better address causal complexities. The fuzzy sets Qualitative
Comparative Analysis (fsQCA) enables us to go beyond simple linear understandings and
explore equifinality, causal asymmetry and complex interactions among explanatory variables.
The aim is to further the research on configurational theories of network effectiveness (Provan &
Milward, 1995; Raab et al., 2013; Verweij et al., 2013). In addition, the paper focuses on
neighborhood governance networks in Beijing, which provides an opportunity to test the external
validity of models developed in American or European contexts.
98
Theoretical Framework
Although considerable progress has been made in the current literature on network effectiveness,
there are hardly any well-established and generalizable theories. However, we do have some
influential models of network effectiveness to draw on. Provan and Milward (1995) did a pioneer
study on four service-delivery networks and developed some configurational propositions
involving structural and environmental factors (See Figure 4.1 for the full model). The model
was further developed by Provan and his colleagues (Provan & Kenis, 2008; Provan & Sebastian,
1998). For example, Provan and Kenis (2008) furthered the research by proposing a contingency
theory of network effectiveness. They explored the best fits between forms of network
governance and other important network characteristics such as trust, number of participants,
goal consensus and the need for network-level competencies.
Another school of scholars paid more attention to network management (Agranoff & McGuire,
2001; Rethemeyer & Hatmaker, 2007). For example, O'Toole and Meier (1999) tested the effects
of a number of network management variables, including managerial networking (Meier &
O'Toole, 2001, 2003), management quality (Meier & O'Toole, 2002), personnel stability and
management stability (O'Toole & Meier, 2004) , the time that managers spend in networks and
management tenure (Juenke, 2005). Scholars have also investigated network management
variables such as network leadership (McGuire & Silvia, 2009) and the identification of and
connection to crucial actors (Klijn et al., 2010).
99
The Provan-Milward model was employed as the major theoretical framework of this paper for
several reasons: First, the model focuses on the interaction between structural characteristics and
environmental factors, which is also the major focus of this paper, and it is probably the best
known model in the network effectiveness literature, providing a solid theoretical foundation for
this research and offering opportunities to further development. Secondly, unlike some studies
that focus on dyads or substructures (Chen & Graddy, 2010; Meier & O'Toole, 2003), this paper
uses the network as the unit of analysis, as does the Provan-Milward model. One objective of this
paper is to further the research on networks at the network level and to respond to the lack of
systematic studies on networks as a whole (Isett et al., 2011; Provan et al., 2007).
NETWORKSTRUCTUAL NETWORK
CHARACTERISTICS ... EFFECTIVENSS
'\
-
-Centralized integration
-Direct, nonfragmented
external r.nntrn 1
NETWORK
CONTEXT
-System stability
-Hi!Ih resource
Figure 4.1 The Provan -Milward Model Of Network Effectiveness
Source: Provan and Milward (1995)
Network integration
"Structure matters" has long been a central tenet of network research. Network integration is the
most commonly studied structural variable in the network effectiveness literature. Provan and
100
Sebastian (1998) argued that integration must occur if networks are to perform well, although
scholars have not reached a consensus on the relative importance of different forms and levels of
integration. Integration contributes to effectiveness because it can reduce fragmentation
(Jennings Jr & Ewalt, 1998; Provan & Milward, 1995), create common norms, enhance
communication and contain opportunistic behaviors. Two types or logics of integration have
often been studied: density-based integration and centralized integration. Density-based
integration is measured by a network density score and identifies how cohesive a network is. In a
dense network, member organizations are closely connected to each other and are thus more
likely to collaborate. Centralized integration measures the degree to which a network's ties are
focused on one organization or one person (J. Scott, 2012). Organizations that occupy central
positions can better coordinate all member organizations to achieve network goals (Provan &
Milward, 1995; Raab et al., 2013). Density-based integration and centralized integration are two
different or even opposite logics - the former is a mechanism of decentralized collaboration,
while the latter is a mechanism of centralized coordination. It may be difficult to coordinate
densely-connected organizations integrated in a decentralized way (Morrissey et al., 1994;
Provan & Milward, 1995). Dual simultaneous integration through the two mechanisms may
make the system unnecessarily complex and cause it to be less effective (Provan & Milward,
1995). Since networks may be more effective if they are integrated through one mechanism, an
interesting question arises: which mechanism is more effective? Provan and Milward's (1995)
original model and some other studies showed that centralized networks are more effective than
networks that are integrated through density (Provan & Milward, 1995; Raab et al., 2013).
101
Stability
The resource dependence theory argues that organizations have to manage their environments in
order to reduce uncertainties and maintain a dependable supply of critical resources (Pfeffer &
Salancik, 1978). A turbulent environment may hurt organizations' efforts to maintain predictable
supplies of resources and greatly increase the difficulties of management. The environment
produces similar impacts on networks. Unstable environments first threaten the flow of resources
to all network members, compromising their abilities to fulfill respective responsibilities in
service delivery or governance. Second, unstable environments may impede the collaboration
between network members. For example, attempts to make systemwide changes in the
distribution of resources may create uncertainties and disrupt existing communication channels
or collaborative relationships, leading to ineffectiveness (Provan & Milward, 1995). Third,
interorganizational trust may be quite low in uncertain environments (Hicklin, 2004; Lambright,
Mischen, & Laramee, 2010), which may hinder collaboration. Networks in unstable
environments may have a high turnover rate of member organizations. Organizations may be
uncertain about whom to collaborate with on a long-term basis and thus lack the incentive to
invest in relationship building.
Resource munificence
Resource munificence 1s of obvious importance to organizations and interorganizational
networks. Resources in the form of funding and expertise enable network members to fulfill their
responsibilities, coordinate action and collaborate with each other in service provision. We would
102
generally expect resource munificence to have a positive influence on network effectiveness,
which is supported by a number of studies (Bazzoli et al., 2003; Conrad et al., 2003). Provan and
Milward (1995) argued that this relationship may be nonlinear because structural characteristics
and environmental factors may moderate the relationship. Raab, Mannak and Cambre's (2013)
study partly supported resource munificence as an INUS condition (Insufficient but Necessary
part of a condition which is itself Unnecessary but Sufficient) and found that the configuration of
resource munificence, centralized integration, network stability and at least 3 years of network
history would lead to effectiveness.
Neighborhood SES
The Provan-Milward model was modified to fit this research context. Neighborhood
socioeconomic status (SES) was added to the model because it may be particularly important in
influencing the effectiveness of neighborhood governance networks. Homeowners and HOA
board members in high-SES neighborhoods are usually lawyers, journalists, professors or other
professionals with high levels of human capital. High-SES neighborhoods may be able to hire
high-end property management firms. Therefore, agents in these neighborhood networks tend to
have higher levels of skills, and expertise in resolving conflicts and making collaboration work.
In addition, these agents often have the social capital that may contribute to network
effectiveness in various ways. For example, when neighborhood organizations run into thorny
problems, some agents may come to the rescue by bringing in new resources, knowledge or
actors. We heard more than once in our fieldwork that some homeowners in high-SES
neighborhoods sought help from powerful friends in high levels of government in order to put
103
pressure on their uncooperative Street Offices. Some scholars conceptualize the social and
human capital as nonfinancial resources that may affect the effectiveness of collaboration (Weiss
et al., 2002).
The original Provan-Milward model included the existence of external control as a key
explanatory variable; however, in this research context, all networks were initiated by network
members who were relatively independent of each other; no external organizations controlled
these networks. This variable was thus dropped. Network governance was considered to affect
network effectiveness (Provan & Kenis, 2008), but it was also dropped because all these
neighborhood governance networks had almost the same internal governance structure - shared
governance. Some other factors, such as formalization, rules, and network size, have been found
to affect network effectiveness (Turrini, Cristofoli, Frosini, & Nasi, 2009; Zakocs & Edwards,
2006). These variables were naturally controlled in this research context. For example, the
influence of institutions was controlled because all cases under study were in Beijing and were
thus subject to the same national and municipal laws and regulations. Network sizes were usually
quite small, ranging from four to seven members. Based on the examination of the
Provan-Milward model and other potential variables, the above four variables, network
integration, stability, resource munificence and neighborhood SES, were used as key explanatory
variables in this research.
104
Research Context: Neighborhood Governance Networks
Governance networks, or area-based policy networks (Filip De Rynck & Voets, 2006), are webs
of interdependent public, business and nonprofit organizations that work together to address a
wide range of policy problems within certain geographical areas (Isett et al., 2011; Klijn et al.,
2010; S0fensen & Torfing, 2005b). Governance networks combine policy formulation and
implementation together in the governing of an area such as neighborhoods or cities (Isett et al.,
2011; Klijn & Skelcher, 2007; Rethemeyer & Hatmaker, 2007). This type of network is usually
naturally grown, and may have a quite low degree of formalization. There are usually no
agreements or contracts that bind these organizations together - they just interact with one
another on an ongoing process, just like the "serendipitous network" described by (Kilduff &
Tsai, 2003). Partly due to the fact the governance involves the distribution of resources, the
internal conflicts of these networks may be high. Each member tried to maximize their own
interests in their interactions with other network members.
This study focuses on neighborhood governance networks in Beijing, formed after the Housing
Reform. In the planned economy era, housing used to be provided through employment by the
government or state-owned enterprises. Street Offices (SOs) and Residents' Committees (RCs ),
which are government and quasi-government entities, dominated neighborhood governance.
Street Offices are the lowest level of government, and are responsible for the administration of
small parts of urban areas. Residents' Committees are theoretically self-governing civic
105
organizations, but they are usually controlled by Street Offices and work as the most basic unit of
social management.
Housing reform, which was launched in the late 1990s and created a real estate market, gave rise
to new organizations in urban neighborhoods. Commercially developed condominium
neighborhoods have become the dominant form of neighborhood in contemporary Beijing.
People living in condominium neighborhoods are allowed to establish Homeowners'
Associations to represent them in the management of communal properties such as lawns, shared
elevators and stairways. Property management firms are hired to provide management services.
Together with developers, these private organizations, which are based on property rights and
market transactions, have become important stakeholders in urban neighborhoods. Government
entities have to work with these new players and operate in a network environment.
These neighborhood governance networks have several defining features. The first and most
important feature is interdependence. Interdependence arises partly from the complex
interactions between political power and property rights in neighborhoods. Theoretically, private
property rights set boundaries to political power, but political power still exerts significant
influence over property rights because property right laws are not well enforced. In these
neighborhood networks, no organization has the resources or power to govern on its own;
organizations have to cooperate with one another in decision making and service delivery. For
example, government entities do not have the absolute power that they used to have in the
106
planned economy era because now they are not property owners. Many government operations
have to get the consent of property owners as a precondition. For instance, setting up a billboard
for outreach or propaganda purposes requires the consent of homeowners. In some cases,
governments even have to pay certain fees to HOAs for the billboard. On the other hand,
homeowners are well aware that local governments including Street Offices can effectively
influence their property rights in various ways. For example, governments have considerable
power in urban planning and zoning, and can thus affect a neighborhood's surrounding
environment and land use. Government programs in public safety and social security also
directly affect the homeowners' quality of life. Therefore, HO As may not really charge fees or
obstruct government operations in the expectation that governments would reciprocate when
needed. Though formal contracts may not exist, complex interdependence connects these
organizations.
The second feature relates to the manner in which governance networks combine policy making
and implementation (Isett et al., 2011; S0fensen & Torfing, 2005a) to produce services, such as
sanitation, but also to decide whether and how to provide these services in the first place. For
example, to maintain neighborhood safety and reduce property and violent crimes, HOAs,
Residents' Committees, and management firms may need to decide whether certain measures
need to be taken, such as installing access control systems at entrances or checking IDs manually.
Once a decision is made, management firms may be responsible for implementing the decision,
while HO As may provide necessary financial resources and monitor the services. Residents'
107
Committees and Street Offices may be responsible for communicating and coordinating with
police agencies on neighborhood safety. The lack of effort from any part would compromise
neighborhood safety service.
The third feature of these networks is the fact that they are place-based, or more specifically,
neighborhood-based. Core organizations operate within neighborhood boundaries clearly defined
in the commercial development processes. One neighborhood has only one HOA, and the HOA
has no property rights beyond neighborhood boundaries. Within neighborhoods, these
organizations will always exist unless there are dramatic institutional changes. Therefore, the
interactions between these organizations resemble "repeated games."
Like other governance networks, interorganizational relationships are not necessarily cooperative;
in many cases, the relationships are dynamic, political and conflictive. These organizations take
different strategies of engagement in order to maximize their own interests and influence.
Organizations may choose not to cooperate in order to compromise others' efforts. For example,
partly due to the poor enforcement of laws and regulations related to property rights, government
entities sometimes form alliances with property management firms or developers to infringe on
homeowners' communal properties. These alliances may gain enormous revenue from selling or
renting out communal properties without giving homeowners compensation. They may try to
stop homeowners from establishing HOAs in order to keep their dominant status in
neighborhoods.
108
These networks produce quite different impacts on their neighborhoods. Some networks work
well and their neighborhoods are clean, safe and well maintained, while other neighborhoods are
governed badly and their property values have gone down - the average price per square meter
can be lower than that of similar neighborhoods by as much as 2,000 RMB (about $ 330). We are
thus given an ideal context to study what cause the differences in the effectiveness of
neighborhood governance networks.
Research Methods, Data and Operationalization of Variables
Research Design
This paper employed a mixed-methods approach. Linear regression was first used to identify
factors that exert statistically significant influence over network effectiveness. The fuzzy set
Qualitative Comparative Analysis (fsQCA) was then used to capture synergistic effects between
factors. This mixed-methods approach allows us to utilize the strengths of each method and to
better explore causal complexities. Findings from the two methods can be cross-validated and
provide different perspectives on the determinants of network effectiveness.
Regression analysis is limited in capturing interaction effects among variables. Complicated
interaction effects are often not known to researchers (Ragin, 1989) and thus cannot be
constructed or tested. Moreover, three-way interaction terms or higher order terms are extremely
difficult to interpret, making three-way interaction terms the actual boundary of regression
analysis (Fiss, 2007). However, as Fiss (2007) pointed out, there are no good reasons to exclude
109
the possibility of three factors combining together to affect outcomes. The fsQCA method was
employed to overcome these problems. Based on Boolean algebra, the fsQCA method is
inherently qualitative but enables structured and focused comparisons of a large number of cases
(Ragin, 2008). It treats cases as configurations of characteristics, and does not try to isolate the
effects of individual variables (Fiss, 2007; Ragin, 1989). In addition, it enables the analysis of
equifinality, which means that there may be multiple configurations of factors that lead to certain
outcomes (Fiss, 2007, 2011 ).
Sampling and Data Collection
The fsQCA method is based on the logic of combinatorial causation and has different
requirements for case selection. As Ragin (1989) argued, "when causal arguments are
combinatorial, it is not the number of cases but their limited variety that imposes constraints on
rigor" (p.13). In other words, the fsQCA method requires that cases should exhibit as many
logically possible combinations of factors as possible. Following this requirement, a purposeful
sampling was employed to ensure that cases with different combinations of network
effectiveness and four causal conditions were selected. Our collaborator in Beijing, a nonprofit
organization specializing in homeowners' advocacy, helped to contact the first six
neighborhoods. Starting from this initial sample, we asked our interviewees to introduce us to
new neighborhoods that exhibited combinations of network effectiveness and causal conditions
that we had not covered. We stopped at 22 cases when we felt that we repeatedly came across
similar cases and seldom new patterns could be identified. This moderate number of cases
110
allowed us to manage in-depth knowledge of each case and variations across cases, which is
another key requirement for the successful application of the fsQCA method (Ragin & Fiss,
2008).
Semi-structured interviews were used as the major approach to collecting data. The structured
part had standardized questions on the outcome and causal conditions. Detailed measurement of
the outcome and causal conditions were discussed in the following section. Data collection did
not stop when interviewees assigned values to all the questions. A deep understanding of each
case was necessary for the successful application of the fsQCA method. Therefore, I asked
interviewees for their reasons behind their choices, which helped to stimulate in-depth dialogues
about neighborhood governance. Open-ended questions were also raised during the interviews.
Each interview took about 2-3 hours. In addition, I visited each of the 22 neighborhoods in order
to get first-hand experience and to double-check the information gained from interviews.
Operationalization of the outcome and causal conditions
Measuring Effectiveness. Measuring effectiveness is the first major task of this research.
Network effectiveness was defined as "the attainment of positive network-level outcomes that
could not normally be achieved by individual organizational participants acting independently"
(Provan & Kenis, 2008, p. 4). Measuring effectiveness is a normative task (Kenis & Provan,
2009). The reasons are that, first, networks have multiple constituents and these constituents
have different beliefs about the criteria of effectiveness (Herranz, 2010; Provan & Milward,
111
2001 ); selecting the preferences of one group of constituents over those of another group, or
assigning weights to the preferences of different groups, is a normative decision; second, the
criteria for measuring effectiveness are normative (Kenis & Provan, 2009; Raab et al., 2013).
Simon (1976) argued that any assessment criteria are elements of value rather than elements of
facts. These elements of value cannot be derived from facts or proven empirically. Therefore,
there are hardly any scientific ways to determine if one criterion is superior to another (Kenis &
Provan, 2009). Kenis and Provan (2009, p. 444) argued that network performance is "a function
of external criteria used to assess the network." Since there are no completely scientific or
objective criteria to measure effectiveness, researchers should be conscious of the issue and be
explicit about the criteria that they select.
Provan and Milward (2001) made a major contribution to the network effectiveness literature by
specifying three levels of analysis: community, network and organization/participant. However,
no consensus has been reached on the criteria or methods to measure effectiveness. Measuring
effectiveness at three levels would be extremely burdensome, so researchers usually have to
select one level (Raab et al., 2013). The major goal of establishing public networks is to solve
some social problems or serve a certain segment of the population. As with many other public
programs, the social impacts that these networks produce should be important indicators of their
effectiveness. Provan and Milward (2001) argued that the effectiveness of these networks should
be "judged by the contribution they make to the communities they are trying to serve" (Provan &
Milward, 2001, p. 416). Raab et al. (2013) also preferred the community-level criteria because
112
"effectiveness at community level is the cumulative outcome of processes and results on the
organizational and network levels" (p. 7). Based on the above understanding, this paper measured
the effectiveness of neighborhood governance networks at the community level, focusing on the
impacts of these networks on residents' quality of life. Of course, as noted above, the decision
was normative rather than completely objective.
The next step was to find criteria that could measure community-level effectiveness.
Homeowners' priorities were selected over those of other constituency groups. Different groups
of constituents may have diverse views and priorities that are hard to reconcile. In this context,
street-level governments may prioritize neighborhood stability; residents may value things like
safety and cleanliness; developers and property management firms may care most about profits.
A choice had to be made to prioritize one group of constituents. In the study of health service
networks, Provan and Milward (1995) selected client wellbeing as the primary measure of
effectiveness because it was "most certainly a top priority" (p.8) of all constituency groups such
as clients, families, funding agencies, and policy makers. In this research, homeowners' priorities
were selected for two reasons: first, homeowners legally own these neighborhoods as the owners
of private and communal properties. They live in these neighborhoods, and both their quality of
life and property values are deeply affected by the governance networks. Compared with other
constituents, they are affected the most by these networks and they care the most about the
effectiveness of governance. In addition, homeowners have first-hand and day-to-day experience
with how neighborhood governance networks work. Second, serving these homeowners is the
113
common goal of all neighborhood organizations, though it may be symbolic for some. For
street-level govermnents, "serving the people" is the official ideology; for HOAs, servmg
homeowners is why they are founded in the first place; for developers and property management
firms, serving homeowners is how they get their revenue.
A composite measure consisting of three items: neighborhood safety, cleanliness and the
maintenance of important facilities, was used as the final measure of effectiveness. This
composite measure was used by both external and subjective evaluations of each network, and
then scores of the two evaluations were added up. The final score of network effectiveness can
range from 0 to 56. Therefore, this is a multitrait-multimethod approach (D. T. Campbell & Fiske,
1959). The three items of the composite measure are homeowners' central concerns and are
closely tied to the quality of life and property values. The services of sanitation, safety and
facility maintenance are usually jointly provided by neighborhood organizations, and thus can
reflect the effectiveness of neighborhood governance networks. For example, property
management firms may be hired to directly produce services, but HOAs provide financial
resources and monitor service qualities. Street-level governments and Residents' Committees
have programs on safety and sanitation, so they are also important participants. In addition, as
the representation of public power at the neighborhood level, these government entities play very
important moderating roles in neighborhood governance.
114
Subjective and external evaluations were integrated in order to fully capture different dimensions
of network effectiveness and to overcome the limitations of each method (Bommer, Johnson,
Rich, Podsakoff, & MacKenzie, 1995; Wall et al., 2004). For example, due to interviewees'
personal bias and bounded cognitive abilities, subjective measures of performance are prone to
contamination and may contain sizeable random errors (J. P. Campbell, 1991; S0fensen &
Torfing, 2009). External evaluations are less affected by personal bias, but evaluators do not have
the deep knowledge of these governance networks and thus may capture only the surface.
Therefore, combining the two methods may help to overcome the limitations.
Network effectiveness was first measured by homeowner leaders' subjective evaluations of
network effectiveness. Using perceived outcomes as a measure of effectiveness has been
commonly used in the current literature (Klijn et al., 2010; Provan & Milward, 1995; Zakocs &
Edwards, 2006). Elected homeowner leaders were selected as interviewees because they were
more familiar with their neighborhoods than were ordinary homeowners. In neighborhoods
where no homeowner leaders had been elected yet, leading homeowner activists were
interviewed. They were asked to rate three statements about safety, sanitation, and housing
maintenance such as "my neighborhood is safe." Responses were constructed using a 7-point
Likert scale with 7 representing strongly agree and 1 representing strongly disagree.
External evaluations were based on photographic evidence. Ideally, external evaluators should
visit each neighborhood and do the evaluations on the scene. However, limited resources made
115
this option impossible. Since conditions of sanitation and facility maintenance were visible, the
photographic method can serve as an alternative way of evaluation. Photos of specific targets
such as garbage cans, neighborhood roads, lawns, fire facilities, recreational facilities and
elevators were taken. Evaluation standards were written and made available to two independent
coders. Coders conducted evaluations by comparing the photos of these specific targets. Safety
should also be externally evaluated. The reported number of crimes in each neighborhood was an
ideal measure of safety, but after consulting with local Bureaus of Public Safety, we learned that
the statistics were not open to the public. As a remedy, I evaluated neighborhood safety on a
7-point scale based on my personal observations and the experience of entering each
neighborhood. Two key criteria that I used were whether special passes or registration were
required at the entrances and whether there were security personnel patrolling the
neighborhoods.
Different evaluations were added up to construct a composite measure, which raised questions of
measurement reliability and construct validity. Appendix One provides the full technical details
of a series of statistical tests including bivariate correlation coefficients, Cronbach's alpha and
intraclass correlation coefficients (ICC). Table 4. 5 in Appendix One shows good internal
consistency reliability with Cronbach's alphas ranging from 0. 79 to 0.93 and good inter-coder
reliability with ICC ranging from 0.61 to 0.87.
116
To establish convergent validity, the correlation coefficients of the same trait measured by
subjective and external evaluations should be statistically significant and sufficiently large (D. T.
Campbell & Fiske, 1959). The tests showed that the composite measure had a good convergent
validity. The correlation coefficients ranged from 0.433 to 0.996 and all of them were statistically
significant at the 5% level. However, we should note that the correlation coefficients between
subjective and objective evaluations ranged from 0.433 to 0.612, which were only moderately
high. The correlation coefficients found in this research were comparable to those found in the
management literature (Bommer et al., 1995; Heneman, 1986).The meta-analysis of Bommer et
al. (1995) showed that the corrected mean correlation between subjective and objective measures
in his sample of 50 published papers was only 0.389, and similarly Heneman (1986) reported a
corrected mean correlation of 0.27 in his meta-analysis. A number of other studies found similar
ranges from 0.4 to 0. 7 (Dawes, 1999; Wall et al., 2004). Since the ranges of coefficients are
common in the literature and they are considered valid (Wall et al., 2004), the moderately high
coefficients were not a serious threat to convergent validity of the integrated measure in this
research.
Causal Conditions/Independent Variables. Causal conditions in this research were based on
the above theoretical framework. One structural characteristic, network integration, and three
environmental conditions, including resource munificence, stability, and neighborhood SES,
were used as independent variables/causal conditions in the regression and fsQCA analyses.
117
The first causal condition was network integration. The effects of two forms of integration,
centralized and density-based integration, were examined separately. The network generator
question was "Can you list the neighborhood organizations that you are collaborating with or
collaborated with recently on neighborhood affairs?" Interviewees listed the organizations that
they collaborated with, which were verified in subsequent interviews with organizations that
were mentioned. Since collaboration is symmetrical in nature, it does not make much sense to
have asymmetrical collaborative relationships. If there were inconsistencies or asymmetrical
relationships, I asked interviewees to clarify how they collaborated and decided whether it
counted as a collaborative relationship. In the end, symmetrical collaborative networks were
constructed based on these interviews. Centralization scores and density scores were calculated
with UCINET. Centralization scores were based on Freeman's degree centrality which is a
measure of local centrality. Therefore, these centralization scores measure the degree to which
one or a few organizations have many links in networks.
Resource munificence measured the availability of financial resources in each network to provide
sanitation, safety and maintenance services. Resource munificence was measured by the
percentage that financial resources could cover the cost of service provision. This information
was mainly collected from the financial records that each property management publicizes and I
also double checked with homeowner leaders in each neighborhood to make sure that the
numbers were consistent with their evaluations. Resources in these governance networks come
mainly from property management fees that homeowners pay and revenues generated from
118
communal properties. In most neighborhoods, property management fee is the major source of
income. The fee rate is set in contracts that homeowners sign collectively with their property
management firms. The approval of a certain percentage of homeowners (usually 2/3) is needed
in order to change the rate. Pre-determined rates often cannot not keep up with the changes in
economic conditions (e.g. inflation), but the procedural requirement makes it difficult to adjust
the rates in a timely manner. As a result, many neighborhoods suffer from inadequate resources.
In some neighborhoods, communal properties can generate a considerable amount of revenue.
For example, some community space can be rented out to small businesses to generate rents.
Obviously this part of revenues is mainly determined by market conditions such as locations and
business opportunities. On the other hand, costs related to personnel, utilities, and facility
maintenance are incurred in the process of providing services. In our study, both revenues and
costs are exogenous to network effectiveness because they are largely determined by market
conditions or contracts.
The third causal condition was network stability. It measured the degree of stability of the
environments in which networks were embedded. The key criterion was whether network
members were proposing or making any changes to the distribution of interests and resources.
The information was gained from interviews. It was a binary variable with 1 representing the fact
that some members were proposing or facilitating such change. Apparently such changes are
very likely to threaten the interests of some network members, reduce mutual trust and create
conflicts. A typical example was the turbulence associated with HOAs' firing of property
119
management firms. In some cases, property management firms refused to cede control, requiring
that some disputes with HOAs be resolved. What typically happened was that the quality of
service plummeted; prolonged disputes greatly harmed the effectiveness of governance.
The last causal condition was neighborhood socioeconomic status (SES), which was measured
by the average price per square meter in recent apartment sales. This indicator has been used in
other studies as a predictor of neighborhood SES (F. Wang et al., 2012). We expect that people
who can afford high housing prices belong to high SES groups.
Analysis and Results
Summary statistics
Network analysis was conducted with UCINET. Table 4.6, Table 4.7 and Table 4.8 in Appendix
Two present summary statistics, density and centralization scores, and correlation coefficients
between variables. Network density ranged from 0.4 to 1 with a mean of0.65. A density score of
1 suggested that every possible link between organizations existed, indicating that organizations
were closely working together. In contrast, 0.4 meant that only 40% of the possible links existed,
indicating that organizations did not cooperate with one another very well. The network
centralization score ranged from 0 to 0.67. The mean was 0.27 and the median was 0.29. Unlike
some service-delivery networks, these governance networks usually did not have a centralized
organization that controlled key resources. Organizations were rather independent. A
120
centralization score of 0 indicated that the network was completely decentralized, and 0.67
suggested a moderately high score of centralization. Network effectiveness ranged from 10 to
51.5 with a mean of 33.3. The absolute values of bivariate correlation coefficients ranged from
0.11 to 0.63, suggesting that variables were not highly correlated. Therefore, multicollinearity
may be not a serious threat.
Linear Modeling
OLS regression was employed to tease out the true relationships between variables. The small
sample size posed a number of problems for regression analysis. For example, the assumption of
normality was not likely to hold, causing invalid hypothesis tests of coefficients. The bootstrap
approach was employed to mitigate this problem. Bootstrap is a nonparametric approach to make
statistical inferences, which does not require distributional assumptions. The basic idea is to
resample the original data set with replacement for a very large number of times, and then to use
the empirical sampling distribution to make inferences. In this research, I resampled 2,000 times
with replacement from the original sample, which, according to Efron and Tibshirani (1993), can
produce a "very safe" estimation of the true confidence intervals. Table 4.1 displays the test of
statistical significance based on bootstrap standard errors. Model 1 tried to replicate the
Provan-Milward model by using centralized integration as the structural variable. Network
centralization was statistically significant at the 5% level; however, the sign was negative,
suggesting that centralized networks were less effective, adjusting for the effects of other
variables. This result contradicted the findings of other studies (Jennings Jr & Ewalt, 1998;
121
Provan & Milward, 1995; Raab et al., 2013). Network stability was still significant and had a
positive effect on network effectiveness. As the above discussion suggests, density-based
integration is a different or even opposite logic compared with centralized integration. The
effects of centralized integration and density-based integration may contradict each other (Provan
& Milward, 1995), so it would be interesting to see which mechanism is more favorable for
network effectiveness. Model 2 replaced centralized integration with density-based integration.
Network density turned out to be statistically significant at the 5% level. The positive sign
suggested that networks with dense cooperative relationships tended to be more effective,
controlling for other independent variables.
Table 4.1 Tests of Statistical Significance with Bootstrap Standard Errors
Independent
Variables
Neighborhood
SES
Network stability
Resource
munificence
Network density
Network
centralization
Constant
Number of
observations
Replications
Adjusted-R
squared
Model 1 Model 2
Coefficients
0.19
10.15**
0.29
-18.46**
24.61
22
2,000
0.29
Bootstrap Bootstrap
SE 95% CI
0.14
4.30
12.43
9.21
[-0.09,0.45]
[2.35,19.18]
[-31.30, 19. 57]
[-36.68,-0.23]
Coefficients
0.18
3.18
3.10
22.00**
6.93
22
2,000
0.31
Note: *** p<0.01 **p<0.05 * p<0.1
Bootstrap
SE
0.15
4.48
11.94
9.74
Bootstrap
95%CI
[-0.12,0.47]
[-6.68, 11.92]
[-26.00,22.79]
[3.68,42.1 OJ
122
Analysis with the fsQCA method
The fsQCA analysis was employed for further analysis in order to better explore causal
asymmetry, equifinality and complex interactions between independent variables. Raw data have
to be calibrated in order to be analyzed by the fsQCA software. Calibrated measures not only
order cases relative to each other but also peg measures to external criteria (Ragin, 2008).
Calibrated measures are thus directly interpretable, showing qualitative degrees in the underlying
construct. The raw data were calibrated into set membership scores ranging from 0 to 1 with 0
representing full nonmembership and 1 representing full membership. Further analysis was based
k vector space comers where k is the number of causal conditions. Each corner corresponds to a
specific combination of causal conditions. The analytical software, fsQCA 2.5, was used to do
the analysis. Results are presented with the notation system developed by Ragin and Fiss (2008).
A black circle (.) suggests a high membership score in a condition, and a circle with a cross-out
C®) indicates a low membership score in a condition. A blank space suggested that this condition
was irrelevant - the presence or absence of it did not make a difference in the outcome. Tables
4.10, 4.11 and 4.12 in Appendix Three show all the truth tables that were analyzed.
I first tried to reproduce the results of Provan and Milward's (1995) study with network
centralization as the structural factor and three environmental factors: resource munificence,
stability and neighborhood SES. Table 4.2 shows the two causal paths that were obtained. Please
note that only complex solutions are reported here. This is the most conservative approach to
123
dealing with logical remainders or theoretical configurations that do not have empirical cases.
The logical remainders that were considered as "false" were not used to simplify configurations
(Ragin, 2008). One causal path was the combination of high membership scores in the sets of
resource munificence and network stability, suggesting that stable networks with sufficient
resources are very likely to be effective. Neighborhood SES and network density were irrelevant.
The consistency score of this path was 0.80, suggesting that 80% of the cases with this
configuration were effective. Its raw coverage was 0. 57, meaning that this causal path can
explain 57% of all the effective networks. Its unique coverage was 0.22, suggesting that 22% of
the effective networks were uniquely explained by this causal path. The other causal path was the
combination of high resource munificence, high neighborhood SES and low network
centralization. 91 % of the networks with this configuration were effective, and this configuration
explained 62% of all effective networks with a unique coverage of 17%. Interestingly, the
evidence suggests that network centralization is unfavorable to network effectiveness, which,
again, contradicts findings from other studies (Provan & Milward, 1995; Raab et al., 2013). The
overall solution coverage was 0.74, showing that these two paths could explain 74% of all the
effective networks, and the solution consistency was 0. 80, indicating that 80% of the networks
with the two configurations were effective.
124
Table 4.2 Configurations Leading To Network Effectiveness
configuration solutions
Resource munificence
e
Neighborhood SES
e
Network centralization
®
Network stability
e
Consistency 0.91 0.80
Raw coverage 0.62 0.57
Unique coverage 0.1 7 0.22
Overall Solution
0.80
consistency
Overall solution
0.74
coverage
Note: frequency cutoff=l, consistency cutoff=0.80
To test the impacts of network density on network effectiveness, I replaced network
centralization with density and did the analysis again. Two causal recipes leading to network
effectiveness were returned, as Table 4.3 shows. The first causal path was the combination of
resource munificence and network stability, which was also obtained in the above analysis. Its
consistency and raw coverage were also the same as those of the previous analysis. The second
path was the combination of resource munificence, network density and neighborhood SES.
Network stability was iITelevant in this causal recipe. The consistency score of this path was 0.92,
suggesting that 92% of the cases with this combination were effective. This path could explain
4 7% of effective networks with a unique coverage of 9%. A comparison of the two paths showed
125
that the first path was less consistent than the second (80% vs. 92%), but the first path had a
higher unique coverage.
Table 4.3 Configurations Leading To Netwo1·k Effectiveness
configuration solutions
Resource munificence
e e
Neighborhood SES
e
Network density
e
Network stability
e
Consistency 0.80 0.92
Raw coverage 0.57 0.47
Unique coverage 0.20 0.09
Overall Solution
0.81
consistency
Overall solution
0.67
coverage
Note: :frequency cutoff= 1, consistency cutoff=O .80
Unlike conventional regression analysis that assumes causal symmetry, the fsQCA method is
built on causal asymmetry, which assumes that factors leading to the presence of an outcome
may be different from factors leading to its absence (Fiss, 2011; Ragin, 2008). An analysis of
configurations of factors leading to network ineffectiveness was thus conducted. As Table 4.4
shows, one configuration was obtained, which was the combination oflow membership scores in
three sets: network density, network stability and neighborhood SES. If a network is unstable,
126
lack cooperation between organizations and have a low SES, then it is very likely to be
ineffective. 82% of the networks with this configuration were ineffective, and this configuration
alone could explain 50% of all ineffective networks.
Table 4.4 Configurations Leading To Network Ineffectiveness
configuration solutions
Resource munificence
Neighborhood SES
®
Network density
®
Network stability
®
Consistency 0.82
Raw coverage 0.50
Unique coverage 0.50
Overall Solution consistency 0.82
Overall solution coverage 0.50
Note: Note: frequency cutoff=!, consistency cutoff=0.80
Discussion and Conclusion
The regression analyses and fsQCA analyses produced some interesting findings. Adjusting for
the influence of other independent variables, the first regression analysis showed that network
stability is positively related to effectiveness and network centralization is negatively related to
effectiveness; the second regression analysis showed that network density is positively related to
network effectiveness. The fsQCA analyses revealed that two equifinal paths that lead to network
127
effectiveness and one causal path that leads to network ineffectiveness. Each causal path is a
configuration of factors that is difficult to capture with conventional linear models.
The first fsQCA analysis of network effectiveness showed an interesting finding of the impact of
network centralization on effectiveness. Low membership in the set of centralization combined
with high membership in the sets of resource munificence and neighborhood SES leads to
network effectiveness. In other words, a network is very likely to be effective if it has sufficient
resources, high-SES and low centralization. This is consistent with the above regression analysis
in that a lower level of network centralization is favorable for effectiveness, although the finding
here is configurational in nature. Linear regression and the fsQCA analysis revealed a similar
effect of centralization on network effectiveness; combining the two perspectives thus
complicates our understanding of the relationship. This finding, however, conflicts with other
studies that suggested a positive relationship between centralization and effectiveness (Jennings
Jr & Ewalt, 1998; Provan & Milward, 1995; Raab et al., 2013). According to these studies,
central organizations can better coordinate other organizations to overcome fragmentation and
increase efficiency.
128
The negative relationship between centralization and effectiveness found in this paper does not
necessarily invalidate previous findings but calls for a reexamination of the relationship. One
caveat is that the centralization scores were based on degree centrality, which measured the
degree to which one or a few organizations have a lot of links with neighbors. This is different
from centralization based on global measures such as betweenness centrality, which measures the
degree that one or a few organizations in the network are on the shortest pathway between other
pairs of actors. In small networks such as those under study, centralization based on degree
centrality may well capture the global network characteristic because central organizations may
be connected to all others. However, this may not be true in large networks. Whether different
types of centralization scores may have different effects on network effectiveness needs to be
further examined. The size of networks may be an important factor to moderate the relationship,
as Provan and his colleagues suggested (Provan & Kenis, 2008; Provan & Milward, 1995;
Provan & Sebastian, 1998). An organization can handle only a limited number of links. In
networks with a large number of organizations, the quantity of relationships that each
organization needs to manage may go beyond their limits, leading to chaos and ineffectiveness
(Provan & Milward, 1995). Coordinated action becomes very difficult to achieve (Provan &
Kenis, 2008). In these networks, centralized coordination may help to solve the problem because
the central organization can coordinate other organizations and free them from managing
relationships. In contrast, the governance networks under study are small, and handling all
possible links is hardly a problem for organizations. Direct coordination with other organizations
may have the edge over centralized coordination. The reason is that, through intensive
129
interactions, organizations become familiar with one another, develop trust, mm1m1ze
opportunistic behaviors and reduce transaction costs (Provan & Sebastian, 1998; Raab et al.,
2013). This causal mechanism is consistent with, though not the completely the same as, one of
the contingencies that Provan and Kenis (2008) proposed. They maintained that small networks
can be effective if they also combine pervasive trust and a decentralized or shared governance
structure in which all organizations participate in production processes on an equal basis.
A common causal path identified by the two fsQCA analyses was the combination of network
stability and resource munificence. With this configuration, neither density-based integration nor
centralized integration was relevant. The result suggested that the combination of resource and
stability can offset some negative effects of the lack of dense collaborative relationships.
Wan-Quan was such a network with abundant financial resources and stable governance
structure. Though homeowners once fought hard to establish an HOA, they were completely
defeated. The developer and property management firm formed an alliance with the Street Office
and controlled the neighborhood completely. Stability was achieved through domination in this
case, and as a result, interorganizational trust was relatively low. However, the alliance pumped
sufficient resources into the neighborhood. Though the network was at the low end of network
density, it was still governed very well with a high membership score in effectiveness (0.95).
Provan and Kenis (2008) argued that symmetric power relations are important for small networks
to achieve effectiveness because they are the basis of participant shared-governance and dense
trust. This research suggests that small networks can still be effective even if power relations
130
among organizations are asymmetric and little trust exists, as long as powerful organizations can
provide necessary resources and completely dominate networks in order to maintain stability.
Stability does not indicate consensus - it can be achieved by powerful organizations suppressing
other organizations.
The second fsQCA analysis indicated that the combination of network density, resource
munificence and neighborhood SES leads to effectiveness. Stability is not relevant if the other
three conditions are present. The Atlantic neighborhood was such an unstable network in which
homeowners were seeking to play a bigger role in governance. They were proposing to change to
the existent governance structure at the time by firing the property management firm (PMF) and
hiring a new one. The PMF refused to hand over their duties. The neighborhood was rich in
financial resources (0.95), had a high SES (0.85), and it was considerably high in network
density because some members such as the HOA and government entities were still collaborating.
Despite the uncertainties about the future of its governance structure, it was still quite high in
effectiveness (0.84). With this configuration, networks can tolerate certain degrees of uncertainty
and instability. It is thus a favorable condition for initiating changes or reforms in networks.
However, this does not mean that tolerance is unlimited. It is very likely that, beyond a certain
point, instability may reduce collaboration and decrease network density, leading networks away
from effectiveness.
131
The path leading to network ineffectiveness was not a simple negation of any of the paths
leading to network effectiveness. The combination of low stability, low network density and low
neighborhood SES will lead a neighborhood governance network to be ineffective. Resource
munificence is irrelevant in this configuration, which suggests that, even if there are sufficient
resources, the lack of collaboration and stability may prevent the network from making the best
of the resources. The Hong-Yuan neighborhood was a case of ineffective network. Homeowners
were very unsatisfied with the poor property management service, and were trying to organize
themselves to fire the management firm. However, in a low SES neighborhood, they lacked the
human and social capital to make their organizing efforts more effective. Activists gained little
support from other homeowners and they could not put much pressure on the developer/property
management firm. Little interorganizational collaboration existed in service provision. The
neighborhood was governed rather terribly with a membership score of 0.98 in the set of poorly
governed neighborhoods.
Structural characteristics and their impacts have long been the center of network research.
Salancik (1995) argued that a good network theory should "propose how structures of
interactions enable coordinated interaction to achieve collective and individual interests" (p.348).
However, our fsQCA analyses suggest that network structures may not be as important as
scholars have thought in influencing network effectiveness. Network structure, or specifically,
network integration, is neither a necessary nor a sufficient condition for network effectiveness.
Effectiveness can be achieved regardless of the condition of integration, provided that networks
132
are stable and have sufficient resources. Even when integrated structure such as high density is
present, it has to be combined with resource munificence and high neighborhood SES in order to
achieve effectiveness. The analysis of configurations leading to ineffectiveness suggested that the
lack of network integration alone is not sufficient for ineffectiveness; it has to be combined with
low stability and low SES in order to cause ineffectiveness.
The fsQCA analysis showed that resource munificence appeared in all four paths leading to
effectiveness. A test of necessary condition was conducted using the fsQCA software. A
necessary consistency score of 0.80 was obtained, which, according to Ragin (2000), suggested
that this condition is "almost always necessary." This is consistent with previous findings
(Conrad et al., 2003; Provan & Milward, 1995). Provan and Milward (1995) maintained that
network effectiveness ranges from low to high in a resource-rich environment and ranges from
low to moderate in a resource-scarce environment, depending on other network characteristics.
Thanks to the fsQCA method, we were able to identify the factors such as system stability that
combine with resource munificence to affect network effectiveness. The interview record
suggested that resource munificence was critical to keep networks working; it became a key
constraint in two cases: Yi-Mei and Shang-Di. Both networks had impressive effectiveness
scores (0.89 and 0.91), but interviewees suggested that the lack of sufficient resources was the
major obstacle for them to achieve a higher level of effectiveness. For example, some facilities
were not maintained as well as residents expected due to resource constraints.
133
This paper contributes to the current literature in several ways. One major contribution is that,
following the work of Raab et al. (2013) and Verweij et al. (2013), this study continued to use
the QCA method to develop configurational theories of network effectiveness. The fsQCA
analyses greatly deepen our understanding of the causal complexities of network effectiveness.
First, the paper revealed the complex interactions between four causal conditions that are
difficult for conventional regression analysis to capture. Second, this research found equifinal
causal paths to network effectiveness. Unlike conventional regression analysis that aims to find a
unifinal solution, the fsQCA method reveals functionally equivalent causal paths to network
effectiveness. Third, the analysis also showed that causality was asymmetrical: factors leading to
network effectiveness were different from factors leading to ineffectiveness, which gave us a
better idea of what may cause ineffectiveness.
This paper also complicates our understanding of the effects of structural characteristics on
network effectiveness. The analyses demonstrated that centralization may be unfavorable for
effectiveness, and density-based integration is positively related to network effectiveness.
Network size may have a role to play here, and this conclusion may hold in small networks.
Network structure is neither a sufficient nor a necessary condition for network effectiveness.
Last but not least, the findings of this paper also carry practical implications. The identified
causal recipes that lead to network effectiveness and ineffectiveness can be used to make more
discriminating diagnoses of networks. The findings may help managers and policy makers to
134
gain a more fine-grained understanding of how factors interact with each other, and thus design
networks in an effective way or change current networks in order to achieve effectiveness. The
equifinal causal paths to network effectiveness are especially helpful. Since there are several
functionally equivalent paths to network effectiveness, managers can select the one that incurs
the least cost or the one that is more compatible with their environments.
When it comes to implications to neighborhood governance in urban China, special attention
needs to be placed on network density and network stability. Compared with neighborhood SES
and resource munificence which are generally fixed in the short run, network density and
stability can be improved within a relatively short period, though not without difficulties. In
many neighborhoods that are poorly governed, conflicts and fights between organizations cause
instability and impede collaboration. Neighborhood organizations need to work out a
compromise regarding respective roles and responsibilities, which may serve as a basis for trust
building and collaboration. Of course, in many neighborhoods, conflicts are so severe that it is
almost impossible for these organizations to reach any compromises. Under this circumstance, it
is imperative for higher-level governments or the designers of the Housing Reform to take
measures such as strictly enforcing laws or mediating/arbitrating conflicts on a case-by-case
basis in order to solve the barriers to collaboration and to achieve stability.
This paper also has several limitations. One limitation is the approach to operationalizing and
measuring network effectiveness. This paper selected the priorities of one specific constituency
135
group, homeowners, as the criteria for measuring effectiveness. The reason was that homeowners
were both property owners and the most important consumers. As the above discussion indicated,
this was a normative decision. The priorities of other constituency groups such as street-level
governments may also need to be considered in some form. Multiple approaches and criteria can
be used to measure different dimensions of effectiveness and each approach has its strengths and
limitations (Bommer et al., 1995). The approach adopted by this paper is just one of the many
possible approaches and it is certainly not perfect. Another limitation is that the paper does not
formally test the effects of network size on network effectiveness, although the theoretical role of
network size is highlighted. The reason is that the sizes of networks under study were small and
varied too little for the fsQCA analysis. Future research may test the moderating effect of
network size with networks that have considerable variation in sizes. A last limitation is that this
paper investigates a particular type of network - governance networks - that combine decision
making and service delivery. Governance networks have their own unique internal dynamics and
thus are not completely the same as service-delivery networks. Caution is needed when
generalizing the conclusions to other types of networks.
136
Appendix One
In order to overcome the limitations of subjective and objective measures, an integrated measure
of effectiveness was constructed. HOA leaders subjectively evaluated the conditions of sanitation,
facility maintenance and safety; two independent evaluators evaluated the conditions of
sanitation and facility maintenance based on photographic evidence. Since safety can hardly be
captured by photos, the external evaluation of safety came from my personal observation and the
experience of entering each neighborhood in my fieldwork. This was a multitrait-multimethod
approach to measure effectiveness and it raised questions of construct validity and reliability.
Campbell and Fiske's (1959) correlational approach was employed to check the validity of the
integrated measure. To establish convergent reliability, the correlation coefficients of the same
trait measured by different methods should be statistically significant and sufficiently large (D. T.
Campbell & Fiske, 1959). As Table 4.5 showed, the correlation coefficients between three
evaluations of sanitation were 0.896, 0.612 and 0.580, and all of them were significant at the 5%
level. The coefficients involved subjective ratings (0.612 and 0.580) were smaller than the one
between two independent coders. The correlation coefficients between evaluations of facility
maintenance were 0.966, 0.445 and 0.476, and all of them were significant at the 5% level.
Similarly, the two correlation coefficients between subjective and external evaluations were
lower. The correlation coefficient between external and subjective evaluations of safety was
0.433 and it was also significant at the 5% level. These coefficients were comparable to, if not
higher than, the correlation coefficients found in the current literature. The meta-analysis of
137
Bommer et al. (1995) showed that the corrected mean correlation between subjective and
objective measures in his sample of 50 published papers was only 0.389, and similarly Heneman
(1986) reported a corrected mean correlation of 0.27 in his meta-analysis. A number of other
studies found similar ranges from 0.4 to 0.7 (Dawes, 1999; Wall et al., 2004). This range of
correlation coefficients was considered as having convergent validity in the literature (Wall et al.,
2004). Therefore, the moderately high correlations were not a serious threat to the convergent
validity of the integrated measure.
The reliability of measures also needs to be considered. Since this was a multitrait-multimethod
approach to measure effectiveness, internal consistency reliability of different traits measured
with the same method and inter-rater reliability (inter-method reliability in this research) should
be checked. "Effectiveness" was operationalized with three dimensions, so Cronbach's alphas
were calculated to check whether the three items were consistently measuring the same
underlying construct. The results were shown in the Table 4.5 in Appendix One. The three
Cronbach's alphas ranged from 0.79 to 0.93, indicating good reliability. Intraclass correlation
coefficients (ICC) were calculated to test whether external and subjective evaluations converged
with one another. The ICC of three measures of sanitation was 0.87, and the ICC of three
measures of facility maintenance was 0.83. The above two indicated very good inter-method
reliability. The ICC of subjective and external evaluations of safety was 0.61, which was lower
than a satisfactory one.
138
The relatively low consistency between subjective and external evaluations of safety merited
special attention. The discrepancy might come from two sources: subjective bias or gloss
external evaluation that did not capture the actual situation. To test the quality of external
evaluation of safety, I checked the Cronbach's alpha between the external evaluation of safety
and the two external evaluations of sanitation and facility maintenance to see how consistent they
were. The results were 0.884 and 0.859, suggesting a very good internal consistency. The result
showed that they were measuring the different dimensions of "effectiveness". I also checked the
ICC between external evaluation of safety and external evaluations of sanitation and facility
maintenance to check the inter-coder reliability. The results were 0.880 and 0.853, which again,
showed good reliability. We can conclude from the evidence that the external evaluation of safety
was consistent with external evaluations of sanitation and facility maintenance, though they were
conducted by different evaluators. The quality of this evaluation was as good as the other
external evaluations. Therefore, I would argue that the discrepancy in the evaluation of safety
was more likely to be caused by reporters' subjective bias. Safety is so important that any report
of crime may cause great psychological impacts on people. As the criminal justice literature
suggests, subjective measures of crime are prone to a number of bias, such as recency bias,
intensity bias, income bias, and people tend to overestimate the number and intensity of crimes
(Jahedi & Mendez, 2014; Lewis & Maxfield, 1980). Given the limitations of subjective
evaluation of safety, I would argue that this discrepancy is not a serious threat to the reliability of
the measures.
139
Table 4.5 Multitrait - Multimethod Correlation Matrix
External 1 External2 Subjective
XI X2 XI X2 X3 XI X2 X3
External 1
Sanitation(XI) 1
Maintenance (X2) 0.836
External2
Sanitation(XI) (0.896) 0.849 1
Maintenance (X2) 0.847 (0.966) 0.867 1
Safety (X3) 0.558 0.617 0.677 0.606 1
Subjective
Sanitation(XI) (0.612) 0.458 (0.580) 0.520 0.477 1
Maintenance (X2) 0.446 (0.445) 0.523 (0.476)
0.359ns
0.684 1
Safety (X3) 0.532
0.289ns
0.555
0.374ns
(0.433) 0.601 0.430 1
Cronbach's a 0.91 0.93 0.79
ICC All XI: 0.87 All X2:0.84 X3: 0.61
Note: "ns" denotes the coefficient is NOT statistically significant at the 5% level.
ICC=lntraclass correlation coefficient
Correlation coefficients of interests are in parentheses.
140
Appendix Two
Table 4.6 Summary Statistics
Mean
Effectiveness 33.3
Neighborhood SES 44.3
(in thousands
RMB)
Network Stability 0.5
Resource 0.80
Munificence
Density 0.65
Centralization 0.27
Median
35.5
46.5
0.5
0.89
0.59
0.29
Min
IO
IO
0
0.3
0.4
0
Table 4. 7 Network Density And Centralization
name network network
density centralization
Wan_Quan 0.50 0.33
Fei Cui 0.40 0.17
Feng_Dan_Li_She 0.58 0.67
Mei Ii Yuan l.00 0.00
Guan Hu Inter l.00 0.00
Guan Zhu 0.67 0.67
LI Du 0.47 0.20
Jiang_ Xiang 0.60 0.25
Atlantic 0.83 0.33
Yue Yuan 0.83 0.33
Wang_ Fu 0.50 0.33
Shang_Di l.00 0.00
YI Mei 0.70 0.50
Tian Tian l.00 0.00
Rong_ Feng 0.47 0.20
Chao_Yang 0.75 0.33
Hong-Yuan 0.50 0.33
Bai Zi 0.50 0.33
Yi Shui 0.40 0.17
Rui Du 0.40 0.17
Peng_ Lai 0.50 0.33
Dong_ Mao 0.60 0.25
Max
51.5
79.8
I
I
I
0.67
141
Table 4.8 Correlation Coefficients Between Variables
Effectiveness Neighborhood Network Resource Density Centralizatio
SES Stability Munificence n
Effectiveness 1
Neighborhood SES 0.41 1
(in thousands RMB)
Network Stability 0.52 0.32 1
Resource 0.33 0.35 0.42 1
Munificence
Density 0.55 0.13 0.63 0.27 1
Centralization -0.22 0.11 0.11 -0.13 -0.36 1
142
Appendix Three
Calibration
Raw data have to be calibrated in order to be analyzed by the fsQCA software. Calibrated
measures not only order cases relative to each other but also peg measures to external criteria.
Calibrated measures are thus directly interpretable, showing qualitative degrees in the underlying
construct. The raw data were calibrated into set membership scores ranging from 0 to 1 with 0
representing full nonmembership and 1 representing full membership.
Calibration should be conducted on the basis of the researcher's substantive and theoretical
knowledge (Ragin 2008). Ragin (2008) also argued that calibration should be based mainly on
substantive knowledge when theoretical knowledge are not available. Direct method of
calibration was employed in this research. Three important anchors structure the calibration: the
threshold for full membership in a set, the threshold for full nonmembership, and the crossover
point. The crossover point is the point where there is maximum ambiguity as to whether a case is
more in or more out of the set (Ragin, 2008).
Network effectiveness The scores of network effectiveness ranged from 0 to 56. There are hardly
any theories that set the criteria of network effectiveness, so the calibration was mainly based on
substantive knowledge of each neighborhood governance network. The threshold of full
143
membership of effective network was set at 44. Only three neighborhoods have effectiveness
scores higher than 44, which were governed clearly better than the rest. The crossover point was
set at 33 ( 45 percentile). The neighborhood with this score was the one that was the most
difficult to judge whether it was effectiveness or not. The full nonmembership was set at 25
(about 27 percentile). This point made a difference in network effectiveness - neighborhoods
with scores less than this point was clearly badly governed.
Resource munificence Resource munificence was measured by the percentage that financial
resources could cover the cost of service provision. The threshold of full membership in resource
munificence was set at 100%, and actually 10 neighborhoods had this value. In these
neighborhoods, there were sufficient resources for member organizations to provide services and
engage in related activities. The crossover point was set at 65%, which is the point of maximum
ambiguity. For example, Yi-Mei neighborhood had this level of resource munificence. The
resources enabled organizations to provide services at a very basic level and were not enough to
meet homeowners' expectations. The full nonmembership point was set at 40% (about 14
percentile). Two neighborhoods had this value and it was clear to me during my fieldwork that
these two neighborhoods did not have the basic level of resources to provide services. The
neighborhoods were very poorly maintained.
Network density Density is the number of links in a network reported as a fraction of the total
links possible. In this paper, the full membership threshold for dense network was set at 1, which
144
means that all possible links existed in these networks. This score shows very high density and it
is rare in reality (Scott 2012). Four neighborhood networks had this density score, which was
mainly due to the small sizes of these networks. The crossover point was set at 0.65 (about 60
percentile). About two thirds of all possible links existed, and it was really hard to tell whether
they are densely connected or not compared with other networks. The full nonmembership point
was set at 0.4 (about 14 percentile). Three networks had density scores smaller than this one, and
my field observation was that organizations in these networks were quite independent of one
another. No meaningful collaborations existed.
Network centralization Centralization measures the degree to which a network's ties are focused
on one organization. When calculating centralization scores, we need to look at the differences
between the centrality score of the most central point and those of other points (J. Scott, 2012).
Centralization score was calculated by dividing the actual sum of differences by the maximum
possible sums of differences. The most centralized network, which has the centralization score of
1, is a star-shaped network with all peripheral nodes linking only to the central node (J. Scott,
2012) . These star-shaped networks are actually rare in reality. In this research, the full
membership threshold of centralized network was set at 0.6, which is a fairly high centralization
score. The crossover point was set at 0.32 (50 percentile), which is the point of maximum
ambiguity with regard to centralization. The full nonmembership point was set at 0.1 (22
percentile). Four neighborhoods were completely decentralized with a centralization score of 0.
145
Neighborhood SES. Neighborhood SES was measure by housing price - the price per square
meter in recent apartment sales. The average housing price at the time of my fieldwork in the end
of 2012 was 28,000 RMB (approximately $4,500) per square meter. The full membership
threshold of high SES neighborhood was set at 60,000 RMB (86 percentile), which is more than
two times of the average price in Beijing. Neighborhoods with this level of housing price or
higher include high-end apartment neighborhoods and townhouse neighborhoods. The full
nonmembership threshold was set at 20,000 RMB (13 percentile), which is lower than the
average price in Beijing. Neighborhoods with this level of price and lower were usually at the
peripheral part of the city where housing price is lower due to location. The crossover point was
set at 29,000 ( 45 percentile), which was slightly higher than the city average price.
Network stability was a dichotomous variable, so it did not need to be calibrated.
Robust Check
To check the robustness of the key findings, I conducted several robustness checks. First, I
checked whether the findings are robust to alternative specifications of the thresholds of
calibration. Following the practice of Fiss (2011), I specified two new crossover points for each
of the causal conditions except network stability which is a dichotomous condition. The reason
to change crossover point is that it is the point of maximum ambiguity with regard to whether it
is in or out of the set. In contrast, the thresholds for full membership and full nonmembership are
146
two extreme points that have relatively little ambiguity. Then, I compared the new intermediate
solutions with the original one. The results are reported in Table 4.9. The results suggest that
generally the results were robust to different specifications of the crossover points. The only
exception is that a new causal path emerged when the crossover point of resource munificence
was changed from 0.65 to 0.75. The new path was the combination of high density and high
stability. However, when I looked the cases again, 0.65 was the crossover point that made more
sense to me. It was the point where there was a qualitative difference between neighborhood.
Neighborhoods in which the revenues could cover 75% of the costs were in a much better
condition of resource munificence. So I would stick to the original crossover point of 65% where
neighborhood organizations really had to make tough decisions regarding how to balance their
different priorities.
I then changed the consistency cutoff from 0.80 to 0.85 in all analyses, and the resulted causal
paths were the same as the reported ones. However, different patterns emerged when the
consistency cutoff was set at 0.90. For example, in the analysis of network effectiveness with
centralization as the structural characteristic, the causal path of low centralization, high SES and
high resource munificence stayed the same, the other causal path changed to high resource, high
stability and low SES. When the consistency cutoff was set at 0.95, completely different causal
paths emerged in the analyses. However, 0.9 is a high requirement for consistency and may
reduce the number of meaningful cases in analyses in a small-N study like this. A consistency
147
level of0.80 or 0.85 is above the minimum recommended level of0.75, and has also been widely
used in published papers (Fiss, 2011; Stockemer, 2013; Verweij et al., 2013). Therefore, I would
accept the conclusion as robust.
148
Table 4.9 Results of Robust Check
Round I Round 2
Original
Biggest
Causal condition crossover
Changes Biggest changes
point
Crossover
in causal
changes in Crossover Changes in
in consistency or
point
paths
consistency point causal paths
coverage
or coverage
Network
0.32 0.36 No 2% 0.42 No 10%
centralization
Network
Density
0.65 0.60 No 1% 0.55 No 2%
Resource
Munificence
65% 60% No 3% 0.75% Yes 8%
Neighborhood
29,000 36,000 No 1% 39,000 No 1%
SES
149
Table 4.10 Truth Table For The Analysis Of Network Effectiveness
(with centralization as the structural condition)
Conditions Outcome Number consistency
(effectiveness) of cases
stability centralization Resource Neighborhood
munificence SES
1 0 1 1 1 3 1.00
1 0 1 0 1 1 0.94
1 1 1 0 1 3 0.90
1 1 1 1 1 4 0.84
0 0 1 1 1 2 0.81
0 1 1 1 0 1 0.73
0 0 0 1 0 1 0.60
0 0 0 0 0 1 0.49
0 1 0 1 0 1 0.35
0 1 0 0 0 2 0.32
0 0 1 0 0 3 0.32
Table 4.11. Truth Table for the Analysis of Network Effectiveness
(with density as the structural condition)
Conditions Outcome Number consistency
(effectiveness) of cases
stability density Resource Neighborhood
munificence SES
1 1 1 0 1 3 1.00
1 1 1 1 1 5 0.93
0 1 1 1 1 1 0.90
1 0 1 1 1 2 0.90
1 0 1 0 1 1 0.87
0 0 1 1 0 2 0.73
0 0 0 1 0 2 0.53
0 0 0 0 0 3 0.37
0 0 1 0 0 3 0.31
150
Table 4.12 Truth Table for the Analysis of Network Ineffectiveness
(with density as the structural condition)
Conditions Outcome Number consistency
( Ineffectiveness) of cases
stability density Resource Neighborhood
munificence SES
0 1 0 0 1 3 0.91
0 0 0 0 1 3 0.87
0 0 0 1 0 2 0.64
0 0 1 1 0 2 0.58
1 0 1 0 0 1 0.58
0 1 1 1 0 1 0.51
1 1 1 0 0 3 0.40
1 0 1 1 0 2 0.40
1 1 1 1 0 5 0.33
151
Chapter 5 : Conclusion
Based on 22 neighborhood governance networks in Beijing, this dissertation focuses on three
important but understudied areas in the research on governance networks/collaborative
governance: internal governance structures, boundary setting and effectiveness.
Theoretical Contributions
Although this dissertation does not aim to build a grand theory of collaborative governanace, it
makes several important theoretical contributions and advances our understating of collaborative
governance.
First, the dissertation proposes a new theoretical framework to explain different types of internal
governance structures of networks. Drawing on resource dependence theory and institutional
theory, the dissertation argues that power differentials and the degree of institutionalization affect
how organizations interact with each other and thus give rise to four different modes of internal
governance: shared governance, insurgent coalition domination, inertial governance and lead
organization governance. The theoretical framework developed in this paper may provide
building blocks for further theoretical development. It is particularly useful to explain the
governance structures of serendipitous networks, which are more common in the real world but
152
are less studied compared with goal-directed networks (Isett et al., 2011 ). The dissertation also
contributes to the research on the relationship between power and institutionalization. Previous
research suggests that the interpretations and problem definitions of powerful members are more
likely to be institutionalized (Phillips et al., 2000); institutionalization is a political process that
reflects the interests of powerful members (Maguire et al., 2004; Seo & Creed, 2002). This
research studies different modes that power may interact with the institutionalization process and
highlights a reciprocal relationship: a high degree of institutionalization may also serve as a
check to power. The shared normative beliefs about the roles and spheres of activities of each
organization as well as taken-for-granted assumptions may constrain or even stop powerful
organizations from infringing upon the spheres of activities of other organizations.
Second, it offers the first systematic study on the boundary setting of governance networks.
Scholars often take networks as a given without questioning how these networks come into being
in the first place; on the other hand, although some scholars proposed the resource-based and
power-based explanations of boundary setting (Agranoff & McGuire, 2001; Scharpf, 1978), they
ignore important variables and have not captured the complexities in the decisions to include or
exclude certain organizations. Why civic organizations are often excluded from governance is
also a question that has not been well studied. In this dissertation, a capacity-threat framework
was proposed to explain the dynamic process of boundary setting based on the systematic
investigation of neighborhood governance networks in Beijing. When assessing whether to
include potential members, extant members will consider not only the resources that potential
153
members may bring but also the threats that potential members may pose. Extant members,
especially those that benefit from the status quo, will try to block new members if these members
are perceived as serious threats. However, this does not mean that these potential members will
always be excluded. To be successfully included, these organizations have to have high
organizational capacity so that they can overcome the resistance from extant members. If
potential members lack the organizational capacity, then they will not be able to overcome the
resistance and will be excluded. This framework also helps to explain why civic organizations
are often excluded. Citizens' interests are often conflicting with the interests of pro-growth
coalitions, and the organizational capacity of civic organizations is often much lower compared
with the capacity of local governments and business organizations. Therefore, under some
circumstances, collaborative governance may not be favorable for democratic governance.
Third, based on the theoretical model proposed by Provan and Milward (1995), this dissertation
employs a mixed methods approach to studying the determinants of the effectiveness of
neighborhood governance networks. "Structure matters" has long been a central tenet of network
research. The dissertation greatly complicates our understanding of the impact of network
structure on network effectiveness. First, our fsQCA analyses suggest that network structures
may not be as important as scholars have thought in influencing network effectiveness. Network
structure, or specifically, network integration, is neither a necessary nor a sufficient condition for
network effectiveness. Effectiveness can be achieved regardless of the condition of integration,
provided that networks are stable and have sufficient resources. Even when an integrated
154
structure such as high density is present, it has to be combined with resource munificence and
high neighborhood SES in order to achieve effectiveness. Second, contradicting studies that
suggest a positive relationship between centralization and effectiveness (Jennings Jr & Ewalt,
1998; Provan & Milward, 1995; Raab et al., 2013), the regression analyses and fsQCA analyses
converge in finding a negative relationship between network centralization and effectiveness.
This negative relationship does not necessarily invalidate previous findings but calls for a
reexamination. The small network size may be one explanation, but it needs further examination.
In addition, following the work of Raab et al. (2013) and Verweij et al. (2013), this dissertation
continues to use the QCA method to develop configurational theories of network effectiveness.
The fsQCA analyses greatly deepen our understanding of network effectiveness by revealing the
complex interactions between four causal conditions, which are difficult for conventional
regression analysis to capture.
Overall Theoretical Reflection
This dissertation is built on the assumption that collaborative governance is a political process,
and it emphasizes power as an important variable to affect interorganizational relationships,
which has often been downplayed in collaborative governance research. Although more and
more studies have recognized that organizations come into play with different resources and
authorities, some people still have the romantic idea that collaborative governance indicates a
model of consensus-based decision making and implementation (Tett et al., 2003). Of course,
155
power has gained increasing attention in the collaborative governance literature. Theories of
interoganizational power are nothing new (Benson, 1975; R. M. Emerson, 1962; Hardy &
Phillips, 1998; Pfeffer & Salancik, 1978), and these theories have often been used to investigate
collaborative governance (Choi & Robertson, 2013; Purdy, 2012; Rodriguez, Langley, Beland, &
Denis, 2007; Thomson & Perry, 2006). A number of studies have associated power with
representation, voice, inclusion/exclusion of certain members in governance processes, and the
internal governance structures (Agranoff & McGuire, 2001; Ansell & Gash, 2007). For example,
powerful organizations may exclude less powerful ones in order to advance their own agenda
(Tett et al., 2003). However, these studies are often conceptual in nature (Agranoff & McGuire,
2001; Ansell & Gash, 2007; Bryson et al., 2006; K. Emerson et al., 2012). With a few exceptions
of case studies (Gray & Hay, 1986; Purdy, 2012; Rodriguez et al., 2007), power has not been
treated as an explanatory variable to examine network-level outcomes such as boundary setting.
Few efforts have been made to measure or quantify power and then systematically study how it
affects outcomes. This dissertation adds to the literature that uses power as an explanatory
variable. Specifically, the dissertation focuses on the role of power in shaping internal
governance structures: power disparities may give rise to balanced or imbalanced governance
structures, and powerful organizations may shape internal governance of the networks to their
advantage. However, the degree of institutionalization - the social acceptance of roles and
responsibilities of each organization - may serve as a check to power. In addition, this
dissertation, which is built on the assumption that the boundary setting is a political process,
systematically investigated how threats and capacities of organizations may affect this political
156
process. One final theoretical implication is that, collaborative governance may represent a
desirable way of governing modern societies given the complexities of many policy problems,
but its internal power dynamic and how power may affect representation and voice should be
paid more attention to.
Another theoretical contribution is that this dissertation uses institutional theory to explain how
organizations interact with one another and offers some interesting theoretical insights. One core
theme that goes through the entire dissertation is that collaborative governance is a new
institution. I also used the degree of institutionalization as a key variable to explain the rise of
different internal governance structures. The boundary setting, internal governance structures and
even the effectiveness of governance networks are all related to the institutionalization of this
new way of governing. This dissertation understands institution in the sense of sociological
institutionalism rather than rational choice institutionalism; the former takes a broader view of
institutions and emphasizes the social aspects of institutions. Barley and Tolbert (1997, p. 96)
defined institutions as "shared rules and typifications that identify categories of social actors and
their appropriate activities or relationships." Collaborative governance is not only a new way of
governing that involves organizations from different sectors but also a new social construction.
The formal and informal rules on the new roles of organizations and interorganizational
relationships are just one aspect of this new institution; another, and arguably more important
part is the social acceptance of collaborative governance as the right way of governing in
contemporary society. It is the social acceptance that gives collaborative governance social
157
legitimacy. Unfortunately scholars of collaborative governance have not widely employed
institutional theory and thus miss its rich theoretical and practical implications. For example,
institutional theory may shed light on why conflicts arise when organizations from different
sectors interact with one another. First, different sectors have their unique institutional principles,
which are not always compatible with one another. For example, government stresses legitimacy
and procedures, while business organizations value efficiency. Organizations enact the
institutional principles of their sectors (Alexander, 1998; Barley & Tolbert, 1997; Phillips et al.,
2000) and thus may conflict with each other. Second, existing institutional arrangements such as
bureaucracy may repel new institutional logics such as collaborative governance (Edelenbos,
2005). Organizations are used to the traditional way that business is conducted, and thus may not
be self-motivated or committed to new institutions. Therefore, in some cases or in early stages,
similar to the decoupling observed by Meyer and Rowan (1977), structures of collaborative
governance may be decoupled from the real political processes - collaborative governance only
plays a symbolic or ceremonial role. For example, the study of neighborhood council reform in
Los Angeles (Musso et al., 2011) and interactive governance in Dutch local governments
(Edelenbos, 2005) suggest that politicians are often not committed to these initiatives that aim to
include citizens in decision-making processes and may try to bypass them.
158
Methodological Contributions
This dissertation employs an innovative mixed-methods approach by combining the fuzzy set
Qualitative Comparative Analysis (fsQCA), regression analysis, social network analysis, and
qualitative methods such as constant comparative method. This mixed-methods approach better
utilizes the strengths of each method to address different research questions. One method that is
particularly worth mentioning is the fsQCA method, which has not been widely used in public
administration research, although a few scholars have started to use it to develop configurational
theories (Raab et al., 2013; Verweij et al., 2013). The fsQCA method helps to find out the
complex interactions between causal conditions of network effectiveness and boundary setting.
The configurations may involve three or more causal conditions, which is difficult for
conventional regression analysis to capture. Second, this research finds equifinal causal paths to
network effectiveness. Unlike conventional regression analysis that aims to find a unifinal
solution, the fsQCA method reveals multiple functionally equivalent causal paths to network
effectiveness. Third, the analysis also shows that causality is asymmetrical: factors leading to
outcomes such as network effectiveness and inclusion are different from factors leading to their
absence, which gives us a better idea of causal relationships. In addition, the convergence of
findings by two different methods provides stronger support for conclusions. For example, both
the fsQCA analyses and regression analyses find that network centralization may impede
network effectiveness. Although one method is linear in nature and the other is configurational,
159
the consistency in findings is makes the conclusions more persuasive than those based on a
single method.
This dissertation also uses an innovative a multitrait-multimethod approach to measunng
network effectiveness. This approach combines both subjective and external evaluations, and
thus may give us a more comprehensive and less biased evaluation. Network effectiveness is a
multilevel and multidimensional concept, which is very difficult to measure (Herranz, 2010;
Provan & Milward, 2001 ). The previous research often used perceived network effectiveness as
a proxy (Klijn et al., 2010; Provan & Milward, 1995; Zakocs & Edwards, 2006; Zheng, Yang, &
McLean, 2010). However, due to respondents' personal bias and bounded cognitive abilities,
subjective measures of performance are prone to contamination and may contain sizeable
random errors (J. P. Campbell, 1991; Feldman, 1981; S0fensen & Torfing, 2009). Therefore, I
supplement the subjective evaluation with external evaluation based on photographic evidence.
This dissertation may be the first one to use the multitrait-multimethod approach to measuring
effectiveness. Using statistical techniques such as Pearson product-movement correlation
coefficients, Cronbach's alpha and Intraclass correlation coefficients (ICC), the method also help
to check the validity and reliability of measures.
160
Policy Design Implications
This dissertation is based on neighborhood governance in urban China; its findings and
theoretical discussions have important implications for policy design.
As noted above, the transformation from government to governance is a process of
institutionalization, which needs not only new laws and rules that give it legal legitimacy but also
acceptance and consensus that give it social legitimacy. The social acceptance of collaboration
between public, business and civic organizations as the way of governing makes collaborative
governance legitimate. This transformation is relatively easy in the US and other western
countries, although decoupling is still a problem in many cases. The transformation from
government to governance has been going on in the US and other Western countries for a
relatively long period of time. The rise of governance as a new paradigm and conceptualization
of governing may be associated with the influence of the New Public Management (NPM)
movement (Stoker, 1998). The NPM movement emphasizes the involvement of business and
civic organizations in the coproduction of public services, and a number of well-known terms,
such as "reinventing government" and "governing the hollow state", have become influential in
shaping how people think about the governing of modern societies (Milward & Provan, 2000;
Osborne & Gaebler, 1992). Of course, over the years, governance has acquired much broader
meaning than simply contracting out services. In addition, in democracies, citizens and civic
organizations have the right and opportunities to participate in governance, which can be seen
161
from an increasing number of experiments of civic engagement in urban governance (J. M. Berry,
Portney, & Thomson, 1993; Cooper & Musso, 1999). Citizen participation also has a long
tradition in the US (Tocqueville, 1840). Thomson and Perry (2006) argued that collaboration
between public and private sectors is rooted in two competing political traditions in the US:
classic liberalism and civic republicanism. This tradition can help collaborative governance to be
accepted by societal actors. It has been argued that intersectoral collaboration has been
institutionalized in the US (Cheng, 2006), or at least it has reached the point where "it is difficult
to imagine successfully addressing global problems, such as the AIDS pandemic or terrorism,
and domestic concerns, such as the educational achievement gap between income classes and
races, without some sort of cross-sector understanding, agreement, and collaboration" (Bryson et
al., 2006).
162
The situation in China is drastically different. Collaborative governance as a new institution at
the neighborhood level faces low degrees of both legal and social legitimacy. Although the
Property Right Law and relevant local regulations allow homeowners to establish Homeowners'
Associations, and also provides legal foundations for developers and property management firms
to engage in neighborhood affairs, these laws and regulations have been weakly enforced. The
conflicts between different institutional logics also come to play in neighborhood governance:
The RCs and SOs are so used to the absolute power they enjoyed in the planned economy era
that they don't want to share their power with other parties in neighborhoods. My fieldwork
suggests that some officials at SOs see homeowners as trouble makers rather than possible
collaborators. Therefore, they often set barriers to homeowner 's organizing. Only about 25% of
neighborhoods in Beijing had established their HO As as of December, 2012. Although laws give
HOAs the legal legitimacy to represent homeowners, these laws are not well enforced or
respected in reality. Another, and arguably bigger problem, is the lack of social legitimacy: in
many cases collaborative governance hasn't gained wide social support and acceptance.
Neighborhoods are one of the few areas in contemporary China that have witnessed a
transformation in which social actors have the legal foundation to participate in governance.
However, many Chinese people still have the so-called "authoritarian personality" and they even
have doubts whether government makes mistakes (Adorno, Frenkel-Brunswik, Levinson, &
Sanford, 1950; W. Wang et al., 2015). Many of them still consider governments as the sole or
most important decision maker in their neighborhoods. Therefore, HOAs and even business
163
organizations lack the wide support or legitimacy to work as important players in neighborhood
governance.
The low degree of institutionalization of collaborative governance in many neighborhoods is
causing problems: homeowners, as the legitimate owners of private and communal properties,
are often excluded from neighborhood governance, and many neighborhoods are very poorly
governed. The conflicts between homeowners and the coalitions of government and business in
urban neighborhoods seriously threaten social stability. In some extreme cases, the conflicts
turned violent: some homeowner activists were assaulted; and many homeowners held
demonstrations and blocked roads in order to make their voices heard.
I would argue that local governments and the designers of the Housing Reform have a lot to do
with the problems in neighborhood governance. Policymakers provided legal foundations for
HOAs, but the laws were weakly enforced. Government officials, at least those at local levels,
are deeply influenced by traditional institutional logics and are not used to share their power with
HOAs in neighborhoods. A new model of collaborative governance has not been established,
although it seems to be the blueprint designed by policy makers. Therefore, the
institutionalization of collaborative governance should be a major task.
Since the participation of property owners in neighborhood governance is both necessary and
inevitable, it is important to strengthen HO As' legal legitimacy as the representative of
164
homeowners. Streamlining the process of founding HO As and reducing unnecessary government
involvement may help homeowners to better organize themselves.
Actors enact the social knowledge or scripts from their previous institutional logics, and their
actions deeply affect how they engage with other organizations. Therefore, their knowledge or
scripts plays a key mediating role. In order to facilitate the institutionalization of collaboration as
a new way of governing, it is important to raise the awareness of all involved actors. Alexander
(1998, p. 349) recommended "enlightening the potential participants in an interorganizational
system with an awareness of their interdependence, and revealing to them their potential mutual
objectives and common goals." In the case of neighborhood governance in China, RCs and SOs
are usually strongly influenced by the institutional logics of governments and they may have a
hard time getting used to the new institutional logic of collaboraion. It is important to raise their
awareness about how property rights set boundaries to their power and to lead them to take
HO As as partners rather than challengers. Therefore, some types of education or training is need
to raise awareness, which may help to establish a new institutional logic and help all
organizations to accept collaboration as the way of government. This will help increase the social
legitimacy of collaborative governance and contribute to its institutionalization. Eventually this
will help to solve problems with inclusion and effective neighborhood governance.
165
Bibliography
Adorno, Theodor W, Frenkel-Brunswik, Else, Levinson, Daniel J, & Sanford, R Nevitt. (1950).
The Authoritarian Personality. Oxford, England: Harpers.
Agranoff, Robert, & McGuire, Michael. (2001 ). Big questions in public network management
research. Journal of public administration research and theory, 11(3), 295-326.
Alexander, E.R. (1998). A structuration theory of interorganizational coordination: Cases in
environmental management. The International Journal of Organizational Analysis, 6( 4),
334-354.
Ansell, C., & Gash, A. (2007). Collaborative Governance in Theory and Practice. Journal of
Public Administration Research and Theory, 18( 4), 543-571.
Barley, Stephen R, & Tolbert, Pamela S. (1997). Institutionalization and structuration: Studying
the links between action and institution. Organization Studies, 18(1 ), 93-117.
Barringer, Bruce R, & Harrison, Jeffrey S. (2000). Walking a tightrope: Creating value through
interorganizational relationships. Journal of Management, 26(3), 367-403.
Bazzoli, Gloria J, Casey, Elizabeth, Alexander, Jeffrey A, Conrad, Douglas A, Shortell, Stephen
M, Sofaer, Shoshanna, ... Zukoski, Ann P. (2003). Collaborative initiatives: Where the
rubber meets the road in community partnerships. Medical Care Research and Review,
60( 4 suppl), 63S-94S.
Benson, J Kenneth. (1975). The interorganizational network as a political economy.
Administrative science quarterly, 229-249.
Berry, Frances S, Brower, Ralph S, Choi, Sang Ok, Goa, Wendy Xinfang, Jang, HeeSoun, Kwon,
Myungjung, & Word, Jessica. (2004). Three traditions of network research: What the
public management research agenda can learn from other research communities. Public
Administration Review, 64(5), 539-552.
Berry, Jeffrey M, Portney, Kent E, & Thomson, Ken. (1993). The rebirth of urban democracy.
Washington, D.C.: Brookings Institution Press.
Black, Bruce L, & Rose, Stephen M. (2002). Advocacy and empowerment: Mental health care in
the community. Boston, MA: Routledge.
166
Bogason, P., & Musso, J. A. (2006). The Democratic Prospects of Network Governance. The
American Review of Public Administration, 36(1), 3-18. doi: 10.1177/0275074005282581
Bommer, William H, Johnson, Jonathan L, Rich, Gregory A, Podsakoff, Philip M, & MacKenzie,
Scott B. (1995). On the interchangeability of objective and subjective measures of
employee performance: Ameta - analysis. Personnel Psychology, 48(3), 587-605.
Bradford, Neil. (1998). Prospects for Associative Governance: Lessons from Ontario, Canada.
Politics & Society, 26( 4), 539-573.
Brass, Daniel J, Galaskiewicz, Joseph, Greve, Henrich R, & Tsai, Wenpin. (2004). Taking stock
of networks and organizations: A multilevel perspective. Academy of Management
Journal, 47(6), 795-817.
Bryer, Thomas A. (2007). Toward a Relevant Agenda for a Responsive Public Administration.
Journal of Public Administration Research and Theory, 17(3), 479-500.
Bryson, John M, Crosby, Barbara C, & Stone, Melissa Middleton. (2006). The Design and
Implementation of Cross - Sector Collaborations: Propositions from the Literature.
Public Administration Review, 66(sl), 44-55.
Campbell, Donald T, & Fiske, Donald W. (1959). Convergent and discriminant validation by the
multitrait-multimethod matrix. Psychological bulletin, 56(2), 81.
Campbell, J.P. (1991). Modeling the performance prediction problem m industrial and
organizational psychology. In M. D. Dunnette & L. M. Hough (Eds.), Handbook of
industrial and organizational psychology, Vol. 2 (2 ed., Vol. 1, pp. 687-732). Palo Alto,
CA: Consulting Psychologists Press.
Chaskin, R. J., & Garg, S. (1997). The Issue of Governance in Neighborhood-Based Initiatives.
Urban Affairs Review, 32(5), 631-661. doi: 10.1177/107808749703200502
Chen, Bin, & Graddy, Elizabeth A. (2010). The effectiveness of nonprofit lead - organization
networks for social service delivery. Nonprofit Management and Leadership, 20( 4),
405-422.
Cheng, Antony S. (2006). Build It and They Will Come-Mandating Collaboration in Public
Lands Planning and Management. Natural Resources Journal, 46( 4 ), 841-85 8.
167
Choi, Taehyon, & Robertson, Peter J. (2013). Deliberation and Decision in Collaborative
Governance: A Simulation of Approaches to Mitigate Power hnbalance. Journal of
Public Administration Research and Theory, 24(2), 495-518.
Chrislip, David D, & Larson, Carl E. (1994). Collaborative Leadership: How Citizens and Civic
Leaders Can Make a Difference. San Francisco, CA: Jossey-Bass
Conrad, D. A., Cave, S. H., Lucas, M., Harville, J., Shortell, S. M., Bazzoli, G J., ... Margolin, F.
(2003 ). Community care networks: linking vision to outcomes for community health
improvement. Med Care Res Rev, 60(4 Suppl), 95S-129S. doi:
10.1177/1077558703259096
Cook, Karen S. (1977). Exchange and power in networks of interorganizational relations*. The
Sociological Quarterly, 18(1 ), 62-82.
Cooper, Terry L, & Musso, Juliet A. (1999). The potential for neighborhood council involvement
in American metropolitan governance. International Journal of Organization Theory and
Behavior, 2, 199-232.
Dahl, Robert A. (1961). Who governs?: Democracy and power in an American city. New Haven,
CT: Yale University Press.
Dawes, John. (1999). The relationship between subjective and objective company performance
measures in market orientation research: further empirical evidence. Marketing
bulletin-Department of Marketing Massey University, 10, 65-75.
De Rynck, F. (2006). Democracy in Area-Based Policy Networks:
American Review of Public Administration,
10.1177 /0275074005282585
The Case of Ghent. The
36(1), 58-78. doi:
De Rynck, Filip, & Voets, Joris. (2006). Democracy in Area-Based Policy Networks The Case of
Ghent. The American Review of Public Administration, 36(1), 58-78.
DiMaggio, Paul, & Powell, Walter. (1983). The iron cage revisited: Institutional isomorphism
and collective rationality in organizational fields. American Sociological Review, 48(2),
147-160.
Dryzek, John S. (2007). Networks and democratic ideals: Equality, freedom, and communication.
In E. S0fensen & J. Torfing (Eds.), Theories of Democratic Network Governance (pp.
262-273). London: Palgrave.
168
Durant, Robert F, Chun, Young-Pyoung, Kim, Byungseob, & Lee, Seongjong. (2004). Toward a
new governance paradigm for environmental and natural resources management in the
21st Century? Administration & Society, 35(6), 643-682.
Echeverria, John D. (2000). No success like failure: The Platte River collaborative watershed
planning process. Wm. & Mary Envtl. L. & Pol'y Rev., 25, 559.
Edelenbos, Jurian. (2005). Institutional hnplications of Interactive Governance: Insights from
Dutch Practice. Governance, 18(1), 111-134.
Efron, Bradley, & Tibshirani, Robert J. (1993). An introduction to the bootstrap (Vol. ). New
York, NY: Chapman & Hall.
Eisinger, Peter. (2002). Organizational Capacity and Organizational Effectiveness among
Street-level Food Assistance Programs. Nonprofit and Voluntary Sector Quarterly, 31(1),
115-130.
Emerson, Kirk, Nabatchi, Tina, & Balogh, Stephen. (2012). An integrative framework for
collaborative governance. Journal of Public Administration Research and Theory, 22(1),
1-29.
Emerson, Richard M. (1962). Power-dependence relations. American Sociological Review, 27(1),
31-41.
Feldman, Jack M. (1981). Beyond attribution theory: Cognitive processes m performance
appraisal. Journal of Applied psychology, 66(2), 127.
Ferman, Barbara. (1996). Challenging the growth machine: Neighborhood politics in Chicago
and Pittsburgh. Lawrence, KS: University Press of Kansas.
Fiss, Peer. (2007). A set-theoretic approach to organizational configuration. Academy of
Management Review, 32( 4), 190-208.
Fiss, Peer. (2011). Building better causal theories: A fuzzy set approach to typologies m
organization research. Academy of Management Journal, 54(2), 393-420.
Forrest, Ray, & Yip, Ngai-Ming. (2007). Neighbourhood and neighbouring in contemporary
Guangzhou. Journal of Contemporary China, 16(50), 47-64.
Freeman, JL. (1965). The Policy Process (Vol. 11). New York, NY: Doubleday Publication.
Fung, Archon. (2005). Deliberation before the Revolution: Toward an Ethics of Deliberative
Democracy in an Unjust World. Political Theory, 33(3), 397-419.
169
Fung, Archon. (2006). Varieties of participation in complex governance. Public Administration
Review, 66(sl), 66-75.
Geertz, Clifford. (1973). The Interpretation of Cultures: Selected Essays. New York, NY: Basic
books.
George, A.L., & Bennett, A. (2005). Case studies and theory development in the social sciences.
Cambridge, MA: MIT Press.
Giddens, Anthony. (1979). Central Problems in Social Theory: Action, Structure, and
Contradiction in Social Analysis (Vol.). Berkeley, CA: University of California Press.
Giddens, Anthony. (1984). The Constitution of Society: Outline of the Theory of Structuration.
Berkeley, CA: University of California Press.
Goodin, Robert E. (1998). The Theory of Institutional Design. New York, NY: Cambridge
University Press.
Gray, Barbara, & Hay, Tina M. (1986). Political limits to interorganizational consensus and
change. The Journal of Applied Behavioral Science, 22(2), 95-112.
Greckhamer, T., Misangyi, V. F., Elms, H., & Lacey, R. (2008). Using Qualitative Comparative
Analysis in Strategic Management Research: An Examination of Combinations of
Industry, Corporate, and Business-Unit Effects. Organizational Research Methods, 11( 4),
695-726. doi: 10.1177 /1094428107302907
Greenwood, Royston, Hinings, CR, & Whetten, Dave. (2014). Rethinking institutions and
organizations. Journal of Management Studies, 51(7), 1206-1220.
Gunton, Thomas I. (2003). The Theory and Practice of Collaborative Planning in Resource and
Environmental Management. Environments: A Journal of Interdisciplinary Studies, 31(2).
Guo, Chao, & Musso, Juliet A. (2007). Representation in Nonprofit and Voluntary Organizations:
A Conceptual Framework. Nonprofit and Voluntary Sector Quarterly, 36(2), 308-326.
Hall, Peter Dobkin. (2006). A Historical Overview of Philanthropy, Voluntary Associations, and
Nonprofit Organizations in the United States, 1600-2000. In W. Powell & R. Steinberg
(Eds.), The Nonprofit Sector: A Research Handbook (Vol. 2, pp. 32-65). New Haven, CT:
Yale University Press.
170
Harding, Alan. (2009). The history of community power Theories of Urban Politics (2nd Edition)
(pp. 27-39). Thousand Oaks, CA: Sage.
Hardy, Cynthia, & Phillips, Nelson. (1998). Strategies of engagement: Lessons from the critical
examination of collaboration and conflict in an interorganizational domain. Organization
Science, 9(2), 217-230.
Haughton, G, & While, A. (1999). From Corporate City to Citizens City?: Urban Leadership
After Local Entrepreneurialism in the United Kingdom. Urban Affairs Review, 35(1),
3-23. doi: 10.1177/10780879922184266
Recio, Hugh. (1992). Issue networks and the executive establishment. In R. J. Stillman (Ed.),
Public administration: Concepts and cases (5th ed., Vol. ). Boston,MA : : Houghton
Mifflin.
Hendriks, Carolyn M. (2008). On Inclusion and Network Governance: The Democratic
Disconnect of Dutch Energy Transitions. Public Administration, 86(4), 1009-1031. doi:
10. lll l/j.1467-9299.2008.00738.x
Heneman, Robert L. (1986). The relationship between supervisory ratings and results - oriented
measures of performance: Ameta - analysis. Personnel Psychology, 39( 4), 811-826.
Herranz, Joaquin. (2010). Multilevel performance indicators for multisectoral networks and
management. The American Review of Public Administration, 40( 4), 445-460.
Hicklin, Alisa. (2004). Network Stability: Opportunity or Obstacles? Public Organization Review,
4(2), 121-133.
Hicks, Darrin. (2002). The promise (s) of deliberative democracy. Rhetoric & Public Affairs, 5(2),
223-260.
Huxham, Chris, & Vangen, Siv. (2000). Ambiguity, complexity and dynamics in the membership
of collaboration. Human Relations, 53(6), 771-806.
Innes, Judith E., Booher, David E., & Di Vittorio, Sarah. (2010). Strategies for Megaregion
Governance. Journal of the American Planning Association, 77(1), 55-67. doi:
10.1080/01944363.2011.533640
Isett, K. R., Mergel, I. A., LeRoux, K., Mischen, P.A., & Rethemeyer, R. K. (2011). Networks in
Public Administration Scholarship: Understanding Where We Are and Where We Need to
Go. Journal of Public Administration Research and Theory, 2l(Supplement 1), il57-il 73.
doi: 10 .1093/jopart/muq06 l
171
Jahedi, Salar, & Mendez, Fabio. (2014). On the Advantages and Disadvantages of Subjective
Measures. Journal of Economic Behavior & Organization.
Jennings Jr, Edward T, & Ewalt, Jo Ann G (1998). Interorganizational coordination,
administrative consolidation, and policy performance. Public Administration Review,
417-428.
Johnston, Erik W, Hicks, Darrin, Nan, Ning, & Auer, Jennifer C. (2010). Managing the Inclusion
Process in Collaborative Governance. Journal of Public Administration Research and
Theory, 21 ( 4 ), 699-721.
Juenke, E. G (2005). Management Tenure and Network Time: How Experience Affects
Bureaucratic Dynamics. Journal of Public Administration Research and Theory, 15(1),
113-131. doi: 10.1093/jopart/mui006
Kantor, Paul, Savitch, Hank V, & Haddock, Serena Vicari. (1997). The Political Economy of
Urban Regimes A Comparative Perspective. Urban Affairs Review, 32(3), 348-377.
Kenis, Patrick, & Provan, Keith G (2009). Towards an exogenous theory of public network
performance. Public Administration, 87(3), 440-456.
Kennedy, Mark Thomas, & Fiss, Peer Christian. (2009). Institutionalization, framing, and
diffusion: The logic of TQM adoption and implementation decisions among US hospitals.
Academy of Management Journal, 52(5), 897-918.
Kickert, Walter JM, Klijn, Erik-Hans, & Koppenjan, Johannes Franciscus Maria. (1997).
Managing Complex Networks: Strategies for the Public Sector. Thousand Oaks, CA:
Sage.
Kilduff, Martin, & Tsai, Wenpin. (2003). Social networks and organizations. Thousand Oaks, CA:
Sage.
King, Cheryl Simrell, Feltey, Kathryn M, & Susel, Bridget O'Neill. (1998). The Question of
Participation: Toward Authentic Public Participation in Public Administration. Public
Administration Review, 59( 4), 317-326.
King, Gary, Keohane, Robert 0, & Verba, Sidney. (1994). Designing social inquiry: Scientific
inference in qualitative research. Princeton, NJ: Princeton University Press.
172
Klijn, Erik-Hans. (1996). Analyzing and managing policy processes in complex networks a
theoretical examination of the concept policy network and its problems. Administration &
Society, 28(1), 90-119.
Klijn, Erik-Hans. (2008). Governance and governance networks in Europe: An assessment often
years of research on the theme. Public Management Review, 10( 4), 505-525.
Klijn, Erik-Hans, & Skelcher, Chris. (2007). Democracy and Governance Networks: Compatible
or Not? Public Administration, 85(3), 587-608.
Klijn, Erik-Hans, Steijn, Bram, & Edelenbos, Jurian. (2010). The hnpact of Network
Management on Outcomes in Governance Networks. Public Administration, 88(4),
1063-1082. doi: 10.llll/j.1467-9299.2010.01826.x
Klijn, Erik-Hans, & Teisman, Geert R. (1997). Strategies and games in networks. In W. J. M.
Kickert, E.-H. Klijn & J. F. M. Koppenjan (Eds.), Managing complex networks.
Strategies for the public sector (pp. 98-118). Thousand Oaks, CA: Sage.
Koch Jr, Charles H. (2005). Collaborative Governance in the Restructured Electricity Industry.
Wake Forest L. Rev., 40, 589.
Koppenjan, Joop, & Klijn, Erik-Hans. (2004). Managing uncertainties in networks: a network
approach to problem solving and decision making New York, NY: Routledge.
Kreutzer, Karin, & Jager, Urs. (2011 ). Volunteering versus Managerialism: Conflict over
Organizational Identity in Voluntary Associations. Nonprofit and Voluntary Sector
Quarterly, 40(4), 634-661.
Lambright, Kristina T, Mischen, Pamela A, & Laramee, Craig B. (2010). Building Trust in Public
and Nonprofit Networks Personal, Dyadic, and Third-Party Influences. The American
Review of Public Administration, 40(1), 64-82.
Lawrence, Thomas B, Hardy, Cynthia, & Phillips, Nelson. (2002). Institutional effects of
interorganizational collaboration: The
Management Journal, 45(1), 281-290.
emergence of proto-institutions. Academy of
Lewis, Dan A, & Maxfield, Michael G (1980). Fear in the neighborhoods: An investigation of
the impact of crime. Journal of Research in Crime and Delinquency, 17(2), 160-189.
Lindblom, Charles Edward. (1977). Politics and markets: the world's political economic systems.
New York, NY: Basic Books
173
Lowi, Theodore J. (1964). American business, public policy, case-studies, and political theory.
World politics, 16(04), 677-715.
Lowndes, Vivien, & Skelcher, Chris. (1998). The dynamics of multi - organizational
partnerships: an analysis of changing modes of governance. Public Administration, 76(2),
313-333.
Maguire, Steve, Hardy, Cynthia, & Lawrence, Thomas B. (2004). Institutional entrepreneurship
in emerging fields: HIV/AIDS treatment advocacy in Canada. Academy of Management
Journal, 47(5), 657-679.
Maxwell, Joseph A. (2005). Qualitative research design: An interactive approach (Vol. 41 ).
Thousand Oaks, CA: Sage.
McGuire, Michael. (2002). Managing networks: Propositions on what managers do and why they
do it. Public Administration Review, 62(5), 599-609.
McGuire, Michael, & Silvia, Chris. (2009). Does leadership m networks matter? Public
Performance & Management Review, 33(1), 34-62.
Meier, Kenneth J, & O'Toole, Laurence J. (2001). Managerial strategies and behavior in
networks: A model with evidence from US public education. Journal of Public
Administration Research and Theory, 11(3), 271-294.
Meier, Kenneth J, & O'Toole, Laurence J. (2002). Public management and organizational
performance: The effect of managerial quality. Journal of Policy Analysis and
Management, 21(4), 629-643.
Meier, Kenneth J, & O'Toole, Laurence J. (2003). Public management and educational
performance: The impact of managerial networking. Public Administration Review, 63(6),
689-699.
Meyer, John W, & Rowan, Brian. (1977). Institutionalized organizations: Formal structure as
myth and ceremony. American Journal of Sociology, 83(2), 340-363.
Milward, H Brinton, & Provan, Keith G (2000). Governing the hollow state. Journal of Public
Administration Research and Theory, 10(2), 359-380.
Morrissey, Joseph P, Calloway, Michael, Bartko, W Todd, Ridgely, M Susan, Goldman, Howard
H, & Paulson, Robert I. (1994). Local mental health authorities and service system
174
change: Evidence from the Robert Wood Johnson Foundation Program on Chronic
Mental Illness. The Milbank Quarterly, 72(1), 49-80.
Musso, Juliet, Weare, Christopher, Bryer, Thomas, & Cooper, Terry L. (2011 ). Toward "Strong
Democracy" in Global Cities? Social Capital Building, Theory - Driven Reform, and the
Los Angeles Neighborhood Council Experience. Public Administration Review, 71(1),
102-111.
O'Toole, Laurence J. (1997). Treating networks seriously: Practical and research-based agendas
in public administration. Public Administration Review, 57(1), 45-52.
O'Toole, Laurence J, & Meier, Kenneth J. (1999). Modeling the impact of public management:
Implications of structural context. Journal of Public Administration Research and Theory,
9( 4), 505-526.
O'Toole, Laurence J, & Meier, Kenneth J. (2004). Public Management in Intergoverrnnental
Networks: Matching Structural Networks and Managerial Networking. Journal of Public
Administration Research and Theory, 14( 4), 469-494. doi: I 0.1093/jopart/muh032
Odell, John S. (2001). Case study methods in international political economy. International
Studies Perspectives, 2(2), 161-176.
Osborne, David, & Gaebler, Ted. (1992). Reinventing government: How the entrepreneurial
spirit is transforming government. Reading, MA: Adison - Wesley.
Pfeffer, J. , & Salancik, GR. (1978). The external control of organizations: A resource
dependence perspective. New York, NY: Harper & Row, Publishers.
Phillips, Nelson, Lawrence, Thomas B, & Hardy, Cynthia. (2000). Inter - organizational
collaboration and the dynamics of institutional fields. Journal of Management Studies,
3 7(1 ), 23-43.
Provan, Keith G, Fish, A., & Sydow, J. (2007). Interorganizational Networks at the Network
Level: A Review of the Empirical Literature on Whole Networks. Journal of
Management, 33(3), 479-516. doi: 10.1177/0149206307302554
Provan, Keith G, & Kenis, P. (2008). Modes of Network Governance: Structure, Management,
and Effectiveness. Journal of Public Administration Research and Theory, 18(2), 229-252.
doi: 10.1093/jopart/mum015
175
Provan, Keith G, & Milward, H Brinton. (1995). A preliminary theory of interorganizational
network effectiveness: A comparative study of four community mental health systems.
Administrative Science Quarterly, 1-33.
Provan, Keith G, & Milward, H Brinton. (2001). Do networks really work? A framework for
evaluating public - sector organizational networks. Public administration review, 61(4),
414-423.
Provan, Keith G, & Sebastian, Juliann G (1998). Networks within networks: Service link overlap,
organizational cliques, and network effectiveness. Academy of Management Journal,
41(4), 453-463.
Purdy, Jill M. (2012). A Framework for Assessing Power in Collaborative Governance Processes.
Public Administration Review, 72(3), 409-417.
Quick, Kathryn S, & Feldman, Martha S. (2011). Distinguishing Participation and Inclusion.
Journal of Planning Education and Research, 31(3), 272-290.
Raab, Jiirg. (2002). Where do policy networks come from? Journal of Public Administration
Research and Theory, 12( 4), 581-622.
Raab, Jiirg, Mannak, Remco S, & Cambre, Bart. (2013). Combining Structure, Governance, and
Context: A Configurational Approach to Network Effectiveness. Journal of Public
Administration Research and Theory, mut039. doi: 10.1093/jopart/mut039
Ragin, Charles C. (1989). The comparative method: Moving beyond qualitative and quantitative
strategies. Berkeley, CA: University of California Press.
Ragin, Charles C. (2000). Fuzzy-set social science. Chicago, IL: University of Chicago Press.
Ragin, Charles C. (2008). Redesigning social inquiry: Fuzzy sets and beyond. Chicago, IL:
University of Chicago Press
Ragin, Charles C, & Fiss, P. (2008). Net effects analysis versus configurational analysis: An
empirical demonstration. In C. C. Ragin (Ed.), Redesining social inquiry: Fuzzy sets and
beyond (pp. 190-212). Chicago, IL: University of Chicago Press.
Rethemeyer, R. K., & Hatmaker, D. M. (2007). Network Management Reconsidered: An Inquiry
into Management of Network Structures in Public Sector Service Provision. Journal of
Public Administration Research and Theory, 18( 4), 617-646. doi:
10.1093/jopart/mum027
176
Rhodes, Rod AW. (1997). Understanding governance: Policy networks, governance, reflexivity
and accountability (Vol. Philadelphia, PA): Open University Press.
Rhodes, Roderick Arthur William. (1986). The national world of local government. London,
UK: : Allen & Unwin.
Ripley, Randall B, & Franklin, Grace A. (1984). Congress, the bureaucracy, and public policy.
Homewood, IL: Dorsey Press.
Rittel, Horst WJ, & Webber, Melvin M. (1973). Dilemmas in a general theory of planning. Policy
sciences, 4(2), 155-169.
Rodriguez, C., Langley, A., Beland, F., & Denis, J. L. (2007). Governance, Power, and Mandated
Collaboration in an Interorganizational Network. Administration & Society, 39(2),
150-193. doi: 10.1177/0095399706297212
Salancik, Gerald R. (1995). Wanted: A good network theory of organization. Administrative
Science Quarterly, 40(2), 345-349.
Schaap, Linze, & van Twist, Mark JW. (1997). The dynamics of closedness in networks. In W.
Kickert, J.M., E.-H. Klijn & J. Koppenjan (Eds.), Managing complex networks:
Strategies for the public sector (pp. 62-78). Thousand Oaks, CA: Sage.
Scharpf, F.W. (1978). Interorganizational policy studies: issues, concepts and perspectives. In K.
Hanf & F. W. Scharf (Eds.), Interorganizational policymaking: Limits to coordination
and central control. Thousand Oaks, CA: Sage.
Scott, John. (2012). Social Network Analysis Thousand Oaks, CA: Sage.
Scott, Richard W. (2003). Organizations: Rational, Natural, and Open systems. Upper Saddle
River: N.J.: Prentice Hall.
Scott, W Richard. (1987). The adolescence of institutional theory. Administrative Science
Quarterly, 32(4), 493-511.
Seo, Myeong-Gu, & Creed, WE Douglas. (2002). Institutional contradictions, praxis, and
institutional change: A dialectical perspective. Academy of Management Review, 27(2),
222-247.
Simon, Herbert Alexander. (1976). Administration behavior: a study of decision-making
processes in administrative organization. New York, NY: Free Press.
177
Sites, W. (1997). The Limits of Urban Regime Theory: New York City under Koch, Dinkins, and
Giuliani. Urban Affairs Review, 32( 4), 536-557. doi: 10.1177/107808749703200405
Small, Mario Luis. (2009). How many cases do I need? On science and the logic of case
selection in field-based research. Ethnography, 10(1), 5-38.
Sorensen, Eva. (2006). Metagovernance: The Changing Role of Politicians m Processes of
Democratic Governance. The American Review of Public Administration, 36(1), 98-114.
Sorensen, Eva, & Torfing, Jacob. (2005a). The Democratic Anchorage of Governance Networks.
Scandinavian Political Studies, 28(3), 195-218.
Sorensen, Eva, & Torfing, Jacob. (2005b). Network governance and post-liberal democracy.
Administrative Theory & Praxis, 2 7(2), 197-237.
Sorensen, Eva, & Torfing, Jacob. (2009). Making Governance Networks Effective and
Democratic through Metagovernance. Public Administration, 8 7(2), 234-258. doi:
10.1111/j.1467-9299.2009.01753.x
Staggenborg, Suzanne. (1988). The Consequences of Professionalization and Formalization in
the Pro-choice Movement. American Sociological Review, 53( 4), 585-605.
Statistics, Beijing Municipal Bureau of (2013). Beijing Statistics Yearbook 2012 Retrieved
from http://www.bjstats.gov.cn/nj/main/20 13-t jnj/index.htm
Stockemer, Daniel. (2013). Fuzzy Set or Fuzzy Logic? Comparing the Value of Qualitative
Comparative Analysis (fsQCA) Versus Regression Analysis for the Study of Women's
Legislative Representation. European Political Science, 12(1), 86-101.
Stoker, Gerry. (1998). Governance as theory: Five propositions. International Social Science
Journal, 50(155), 17-28.
Stone, Clarence N. (1989). Regime politics: governing Atlanta, 1946-1988. Lawrence, KS:
University Press of Kansas
Suarez, David F. (2011 ). Collaboration and Professionalization: The Contours of Public Sector
Funding for Nonprofit Organizations. Journal of Public Administration Research and
Theory, 21(2), 307-326.
Tett, Lyn, Crowther, Jim, & O'Hara, Paul. (2003). Collaborative partnerships m community
education. Journal of Education Policy, 18(1), 37-51.
178
Thomson, Ann Marie, & Perry, James L. (2006). Collaboration processes: Inside the black box.
Public Administration Review, 66(sl), 20-32.
Tocqueville, Alexis De. (1840). Democracy in america (H. C. Mansfield & D. Winthrop, Trans.).
Chicago, IL: University of Chicago Press.
Tolbert, Pamela S, & Zucker, Lynne G (1983). Institutional sources of change in the formal
structure of organizations: The diffusion of civil service reform, 1880-1935.
Administrative Science Quarterly, 28(1 ), 22-39.
Turrini, Alex, Cristofoli, Daniela, Frosini, Francesca, & Nasi, Greta. (2009). Networking
Literature About Determinants of Network Effectiveness. Public Administration, 88(2),
528-550. doi: 10. llll/j.1467-9299.2009.01791.x
van Nuenen, Martine. (2007). Beyond Boundaries: A Cognitive Perspective of Boundary Setting.
In T. Gossling, L. A. G Oerlemans & R. Jansen (Eds.), Inside Networks: A Process View
on Multi-Organisational Partnerships, Alliances and Networks. Northampton, MA:
Edward Elgar Publishing.
Vangen, Siv, Hayes, John Paul, & Cornforth, Chris. (2014). Governing Cross-Sector,
Inter-Organizational Collaborations. Public Management Review( ahead-of-print), 1-24.
Verweij, Stefan, Klijn, Erik - Hans, Edelenbos, Jurian, & van Buuren, Arwin. (2013). What
Makes Governance Networks work? A Fuzzy Set Qualitative Comparative Analysis of 14
Dutch Spatial Planning Projects. Public Administration. doi: 10.1111/padm.12007
Vigoda, Eran. (2002). From Responsiveness to Collaboration: Governance, Citizens, and the
Next Generation of Public Administration. Public Administration Review, 62(5), 527-540.
Wall, Toby D, Michie, Jonathan, Patterson, Malcolm, Wood, Stephen J, Sheehan, Maura, Clegg,
Chris W, & West, Michael. (2004). On the validity of subjective measures of company
performance. Personnel psychology, 57(1), 95-118.
Wang, Feng, Yin, Haitao, & Zhou, Zhiren. (2012). The Adoption of Bottom - up Governance in
China's Homeowner Associations. Management and Organization Review, 8(3), 559-583.
Wang, Weijie, Li, Hui, & Cooper, Terry. (2015). Civic Engagement and Citizenship Development:
The Case of Homeowners's Participation in Neighborhood Affairs in Beijing.
Administration & Society. doi: 10.1177/0095399715581041
179
Weible, Christopher M, & Sabatier, Paul A. (2005). Comparing policy networks: Marine
protected areas in California. Policy Studies Journal, 33(2), 181-201.
Weiss, E. S., Anderson, R. M., & Lasker, R. D. (2002). Making the Most of Collaboration:
Exploring the Relationship Between Partnership Synergy and Partnership Functioning.
Health Education & Behavior, 29(6), 683-698. doi: 10.1177/109019802237938
Westphal, James D, Gulati, Ranjay, & Shortell, Stephen M. (1997). Customization or conformity?
An institutional and network perspective on the content and consequences of TQM
adoption. Administrative Science Quarterly, 42(2), 366-394.
Yang, Kaifeng, & Callahan, Kathe. (2007). Citizen Involvement Efforts and Bureaucratic
Responsiveness: Participatory Values, Stakeholder Pressures, and Administrative
Practicality. Public Administration Review, 67(2), 249-264.
Yang, Kaifeng, & Pandey, Sanjay K. (2007). Public Responsiveness of Government
Organizations: Testing a Preliminary Model. Public Performance & Management Review,
31(2), 215-240.
Zakocs, R. C., & Edwards, E. M. (2006). What explains community coalition effectiveness?: a
review of the literature. Am J Prev Med, 30(4), 351-361. doi:
10.1016/j.amepre.2005.12.004
Zhang, T. (2002). Urban Development and a Socialist Pro-Growth Coalition in Shanghai. Urban
Affairs Review, 37(4), 475-499. doi: 10.1177/10780870222185432
Zheng, Wei, Yang, Baiyin, & McLean, Gary N. (2010). Linking organizational culture, structure,
strategy, and organizational effectiveness: Mediating role of knowledge management.
Journal of Business Research, 63(7), 763-771.
180

Abstract (if available)

Abstract This dissertation uses a mix-methods approach, which includes regression analysis, fuzzy set Qualitative Comparative Analysis (fsQCA), social network analysis and constant comparative method, to explore the following research questions: what factors influence the inclusion/exclusion of certain organizations in governance? What may explain different internal governance structures? What factors may affect the effectiveness of governance networks? This dissertation is based on neighborhood governance networks in Beijing, China. I collected data of 22 neighborhood networks, each of which consists of public, business and civic organizations. This provides an ideal environment to address the above research questions. ❧ The first chapter is an introduction chapter that reviews the concept of governance networks, and then provides an overview of the entire dissertation. Chapter 2 uses resource dependence theory and institutional theory to explain different internal governance structures to develop a new framework to explain how different types of governance structures are formed. The balance of power and degree of institutionalization are considered to be two important dimensions to explain how organizations interact with one another. The chapter develops a typology of governance structures. Four ideal types of governance structures were identified: shared governance, inertial governance, insurgent coalition domination and lead organization governance. This chapter explains the power dynamics in each governance structure and illustrates with cases of neighborhood governance in Beijing. ❧ Chapter 3 studies the boundary settings of governance networks. The dynamic process of inclusion and exclusion of certain organizations in governance networks is defined as boundary setting. Drawing on the literature from general organization theory, urban regime theory, nonprofit management and network research, this chapter attempts to transcend both the resource-based explanation and power-based explanation. This chapter develops a capacity-threat framework to explain the boundary setting of collaborative governance. When assessing potential members to include, extant members will consider not only resources that they may bring but also the threats that they may pose to their interests. Extant members especially those that benefit from the status quo will try to block new members if they are perceived as serious threats. However, this does not mean that these organizations will always be excluded. To be successfully included, these potential members have to have high organizational capacity so that they can overcome the resistance from extant members. If they lack the organizational capacity, for example, they lack strong leadership or have a low degree of formalization, then they will likely to be excluded. Therefore, the boundary setting of governance networks is a political and dynamic process. ❧ Chapter 4 examines the determinants of the effectiveness of governance networks. Evaluating network effectiveness and studying its determinants have been an important topic in the network research. Based on the models proposed by Provan and Milward (1995) and Provan and Kenis (2008), this chapter employed a mixed-methods approach to study the determinants of the effectiveness of governance networks. Linear regression was used to identify independent variables that exert statistically significant influence over network effectiveness, and the fuzzy set Qualitative Comparative Analysis was used to investigate the complex interactions between explanatory variables. The chapter revealed the causal complexities of network effectiveness: the analysis found different but functionally equivalent configurations of causal conditions that lead to network effectiveness, and showed that configurations of factors leading to network effectiveness are different from those leading to network ineffectiveness. The results also suggested that network structural characteristics such as network centralization and density are neither sufficient nor necessary conditions for network effectiveness. However, in contract to Provan and Milward’s (1995) findings, the results suggested that network density is more important than network centralization in affecting effectiveness in small networks. Resource munificence was identified as an “almost always” necessary condition for network effectiveness. ❧ To summarize, this dissertation makes several important theoretical contributions and advances our understating of collaborative governance. First, it proposes a new theoretical framework to explain different types of internal governance structures of networks, which has often been neglected in the current literature. The framework is especially suitable for explaining the internal governance structures of serendipitous networks that we often see in the real world. Second, it offers the first systematic study on the boundary setting of governance networks. A capacity-threat framework was proposed to explain the dynamic process of boundary setting based on the systematic investigation of neighborhood governance networks in Beijing. Third, this dissertation develops a configurational theory of network effectiveness and better explores the causal complexities of network effectiveness with a mixed methods approach. It sheds new light on the relationship between network structures and network effectiveness. ❧ The dissertation also makes a number of methodological contributions. First, it employs an innovative mixed-methods approach, which better utilizes the strengths of each method to address different research questions. One method that is particularly worth mentioning is the fsQCA method, which has not been widely used in public administration research, although a few scholars have started to use it to develop configurational theories (Raab et al., 2013

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

A study of Chinese environmental NGOs: policy advocacy, managerial networking, and leadership succession

PDF

Governing regional collaboratives: institutional design, management and leadership

PDF

Civic associations, local governance and conflict prevention in Indonesia

PDF

The social functions of private neighborhood associations: the case of homeowner associations in urban China

PDF

The Wenchuan earthquake recovery: civil society, institutions, and planning

PDF

Quango reforms and challenges in South Korea: social relations, informal networks, and hidden actions

PDF

Essays on fiscal outcomes of cities in California

PDF

Measuring the drivers of economic, energy, and environmental changes: an index decomposition analysis

PDF

Institutional contestation, network legitimacy and organizational heterogeneity: interactions between government and environmental nonprofits in South Korea

PDF

Does collaborative R&D policy work? The effect of collaborative R&D policy on innovative activities of firms in Korea

PDF

Governance and urban design: Omaha by design

PDF

China's environmental reform: ecological modernization, regulatory compliance, and institutional change

PDF

Three essays on the causes and consequences of China’s governance reforms

PDF

Why go green? Cities' adoption of local renewable energy policies and urban sustainability certifications

PDF

How does collaborative governance work? The experience of collaborative community-building practices in Korea

PDF

Collaborating to provide microfinance to caregivers of orphans and vulnerable children in Ethiopia

PDF

Essays on government intervention in financial crises

PDF

The effects of interlocal collaboration on local economic performance: investigation of Korean cases

PDF

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

PDF

Foreign-related activities of the Chinese local governments and agents of globalization: a case study of 31 provinces in mainland China

Asset Metadata

Creator Wang, Weijie (author)

Core Title Governance networks: internal governance structure, boundary setting and effectiveness

School School of Policy, Planning and Development

Degree Doctor of Philosophy

Degree Program Policy, Planning, and Development

Publication Date 07/10/2017

Defense Date 05/04/2015

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

Tag boundary setting,collaborative governance,governance networks,internal governance structure,interorganizational collaboration,network effectiveness,OAI-PMH Harvest

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Cooper, Terry (committee chair), Fiss, Peer C. (committee member), Tang, Shui Yan (committee member)

Creator Email weijiewa@usc.edu,weijiewang427@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-592721

Unique identifier UC11300535

Identifier etd-WangWeijie-3591.pdf (filename),usctheses-c3-592721 (legacy record id)

Legacy Identifier etd-WangWeijie-3591.pdf

Dmrecord 592721

Document Type Dissertation

Format application/pdf (imt)

Rights Wang, Weijie

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