University of Nevada Reno System of Care for Children's Behavioral Health: Implementation from a Networked Perspective A

University of Nevada Reno

System of Care for Children’s Behavioral Health: Implementation from a Networked Perspective

A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Political Science

Jill M. Manit

Dr. William L. Eubank, Dissertation Advisor

December, 2018

THE GRADUATE SCHOOL

We recommend that the dissertation prepared under our supervision by

JILL M. MANIT

Entitled

System Of Care For Children's Behavioral Health: Implementation From A Networked Perspective

be accepted in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

William Eubank, Ph.D, Advisor

Aleksey Kolpakov, Ph.D., Co-advisor

Eric Herzik, Ph.D., Committee Member

Thomas Harris, Ph.D., Committee Member

Denise Montcalm, Ph.D., Graduate School Representative

David W. Zeh, Ph. D., Dean, Graduate School December, 2018

Abstract

Governance models influence the approach public service organizations take when implementing programs, policies and practices. The networked model of governance supports the involvement of multiple actors who span organizational boundaries and roles to implement solutions to address complex social problems. This study examines the implementation of a statewide System of Care for children’s mental health from a networked governance perspective. System of Care is an evidence-based framework that aims to coordinate services in a culturally and linguistically appropriate way and places youth and families at the center of decision making on their care as well as in the design of the system. This descriptive study examines the extent to which the effort to implement a statewide System of Care resembles a networked governance perspective and describes the patterns of relationships that comprise the structure of the network.

The study employs a mixed-method approach to identify actors in the network and to describe their patterns of interactions across four network relations. Archival data are collected to first identify the actors in the network (N=107). An “affiliation network” and a “name roster” are constructed for the second phase of the research, a survey of network members. The survey yielded a 53% response rate.

Multi-level network analysis techniques are employed to identify and describe the patterns of relationships in the network. Four network relations are the focus of the study: operations (i.e. working together on the implementation), sharing information,

ii sharing resources and trust. The data are analyzed at the individual actor level to identify those actors who are most central in the network, at the dyadic level to describe any connections between two actors, at the triadic level to identify the types of relations between three actors and at the whole network level. The data are analyzed using UCINET, Pajek and Excel software programs to run the appropriate analyses at multiple levels. Additionally, UCINET’s NetDraw program is utilized to generate visual depictions of the network including a distinction of the central actors.

The findings reveal the network is operating under to two network model types; shared-governance and lead organization. For the shared-governance model, the network is moderately dense for the network relations of operations and sharing information. It becomes sparser and more centralized under the lead organization model for the network relations of sharing resources and trust. The network also has a

“team” of central actors consisting of a parent, caregiver, professional staff and administrator. This team maintained high centrality scores across all four of the measured network relations.

This study supports the utility of the networked governance perspective in public policy implementation for describing how a network structure can shape opportunities and constraints associated with public policy implementation. However, further analysis is needed. The paper concludes with recommendations for further study of the multiple sub-structures that can be found in a network and the impact the structural configurations can have on achieving public policy and administration goals.

iii

Acknowledgments

I am proof that it takes a team to complete a Ph.D. program. While I knew this going in to the program, I still thought I could get it done on my own. This is probably why it took me so long to get it done! It wasn’t until I embraced a support system and surrounded myself with committed professionals that I was finally able to kick in to gear and complete my dissertation. As such, I am indebted to my central “team” for the completion of this project. Without you, I would not have been able to finish this journey. I thank you for all of your support, encouragement and enthusiasm.

Dr. Eubank: I took my first Political Science course from you as a graduate special student a long time ago. I took the course wondering if this was the right field for me. It was your course that confirmed my decision to enter the program and provided the foundation for where I went after that. I thank you for stepping in as my

Chair and keeping the fire lit beneath me. You pushed me when I needed and pulled me when I needed that too. You helped me set my priorities with the goal of completion in mind. I appreciate your skill in forcing me to rethink and refine my research … often with a single thought provoking question. I am absolutely certain that I would not be here without you. Thank you for your belief in me.

Dr. Kolpakov: I took on a Network Analysis study because I have been arguing its benefits for the study of public policy implementation for years. I didn’t know anything about the approach when I embarked on this research project, but I knew I wanted it to become the basis for my research agenda in the future. I was so pleased to have met you and then I was intimidated when I learned that you had been trained by the best in

iv the field of network analysis. You helped my move this project through the final phase and I am so appreciative of the time you took to teach me about the study of networks and how they exist everywhere!

Dr. Ostergard: I want to thank you for lighting the fire under me and giving me a deadline to finish my program. It is what I needed, and it is what has lead me to this point today. I appreciate your support during the program and as I neared completion.

Your calm reassurance helped me to get through some of those challenging final moments.

Dr. Montcalm: Thank you for serving as a committee member and for creating the “space” that I needed to complete my dissertation. I felt like we were in this together and I appreciate your support in making this happen. I appreciate the gentle reminders to keep things moving along and for your patience as I shifted my priorities.

Dr. Herzik and Dr. Harris: Thank you for your time and commitment in serving on my committee. I appreciate your support and guidance on this project.

The Nevada System of Care: To the youth, family members, professionals and administrators involved in developing and implementing a System of Care to improve access and quality of behavioral health services for children and their families … you have taught me so much about being person-centered in the design of public services; you have taught me what it means to advocate for quality and efficient services and you have shown me the tenacity it takes to embark on a long journey to align public service systems. I admire your commitment to the youth and families of this state and I ask that you keep forging the way. Thank you for allowing me to be a participant with your

v group and thank you for trusting me with your information. Thank you for listening to my research ideas and for testing out my survey. I look forward to our continued work together on implementing a true System of Care in Nevada.

To my husband: Thank you, thank you, thank you. From proof reading my essays for admission to the final completion of my dissertation, you have been there for me. Thank you so much for taking care of the kids while I spent hours at my computer, for listening to me when I was excited about my surveys coming in and for your unwavering belief that I can get this done. I needed to lean on you and I am so glad you were a pillar for me. I really feel like this became a family affair and don’t think I could have completed it without everyone embracing it. You helped to make it happen and I thank you for that.

To my sweet kids: You weren’t even born yet when I started this journey so, in a sense, you were born to do this! You both were so amazing during the last push of my dissertation. You did such a great job taking care of yourself, doing your homework and leaving me alone so I could keep working on it. I found your song “We Don’t Need No

Dissertation” especially funny and motivating. Thank you for asking me good questions about my research and for learning about Network Analysis along side of me. I am so proud of you both.

To both sets of my parents: I don’t even know where to begin with my gratitude for what you have done for me. You stepped in to help with my family and parent my children while I was mostly absent. I am sorry that I missed so many gatherings and I

vi thank you for your understanding. I would not have been able to get this done without your support and encouragement.

To my brother and sister: You two have always been and always will be a motivation for me to push hard, be smart and do good things. My whole life I have looked up to you and that has never stopped. Thank you for continuously being role models for me. The good that you do in this world motivates me to do the same.

Social Work Colleagues: Mary, Gloria and Candace … As you know, this has been a long time coming for me. I can’t thank you enough for your ongoing encouragement and support. Thank you for being excited for me, for listening to me when I was ready to give up and for supporting me when I had to “disappear” to get my dissertation done.

Whether it was talking about my research design and data analysis to just trying to figure out how to balance a Ph.D. program with work, it was such a relief to be able to talk through it with smart and hard-working women. Thank you for being amazing colleagues and my friends.

vii

Table of Contents

Chapter 1: Introduction 1 Chapter 2: Literature Review 8 § Governance 10 § Implementation 18 § Networked Implementation 24 § Networks 26 § Types of Networks 29 § Network Power and Effectiveness 31 § Basic Network Components 34 § Theoretical Explanations of Network Behavior 36 § Network Analysis 43 § Sources of Data and Data Collection in Network Analysis 46 § Common Measures in Network Analysis 47 Chapter 3: Study Context 57 § System of Care 59 § Nevada System of Care 60 § Nevada System of Care as a Network 63 Chapter 4: Methodology 67 § Research Design 67 § Phase One, Purpose One: Affiliation Network Matrix 68 § Phase One, Purpose Two: Name Generation for Survey 71 § Phase Two: Survey 71 § Data Analysis Plan 81 Chapter 5: Results 90 § Research Question 1 90 § Research Question 2 101 § Research Question 3 153 Chapter 6: Discussion, Limitations and Conclusion 167 References 179 Appendices 192

viii

List of Tables

Table 2.1 Common Terms to Describe Multiple Actors in Policy Implementation

Table 4.1 Threats to Validity Considered in Survey Design

Table 5.1 Survey Respondent Attributes – Geographic Region by Respondent Role

Table 5.2 Survey Respondent Attributes –Length of Time Involved by Respondent Role

Table 5.3 Survey Respondent Attributes – Agency Type by Respondent Role

Table 5.4 Overall Network Density by Relationship Variable

Table 5.5 Overall Network Degree Centrality by Relationship Variable

Table 5.6 Top Five Actors for Degree Centrality and Betweenness Centrality Scores, Operations Relationship

Table 5.7 Top Five Actors for Degree Centrality and Betweenness Centrality Scores, Information Relationship

Table 5.8 Top Five Actors for Degree Centrality and Betweenness Centrality Scores, Resources Relationship

Table 5.9 Top Five Actors for Degree Centrality and Betweenness Centrality Scores, Trust Relationship

Table 5.10 Network Reciprocity by Relationship Variables

Table 5.11 Network Transitivity by Relationship Variables

Table 5.12 Triadic Census of the Nevada SOC Operations Network

Table 5.13 Triadic Census of the Nevada SOC Information Network

Table 5.14 Triadic Census of the Nevada SOC Resource Network

Table 5.15 Triadic Census of the Nevada SOC Trust Network

Table 5.16 External-Internal Index of Ties Between Network Attributes and Network Relationships

Table 5.17 External-Internal Index of Ties Between Network Attribute Groups and Network Relationships

Table 5.18 Summary of Correlations Between Relational Variables

Table 5.19 Summary of Correlations Between Actor Attributes and Relational Variables

Table 5.20 Network Relationship Regressions

Table 5.21 Summary of Nevada SOC Whole Network Level Findings

Table 5.22 Survey Respondent Attributes – Perception of Effectiveness by Respondent Role

Table 5.23 Survey Respondent Attributes – Perception of Effectiveness by Length of Time

List of Figures

Figure 1.1 Study Framework

Figure 1.2 Structure of Literature Review

Figure 2.1 Matland’s (1995) Ambiguity-Conflict Matrix of Implementation

Figure 2.2 Sixteen Types of Triads

Figure 3.1 Nevada SOC as a Network

Figure 3.2 Nevada SOC within a Larger Network

Figure 5.1 Role Representation in Nevada SOC

Figure 5.2 Geographic Region of Representation

Figure 5.3 Length of Time Involved in Nevada SOC

Figure 5.4 Organizational Type

Figure 5.5 Nevada SOC Network Emergent to Strengthened Relationship Tie

Figure 5.6 Nevada SOC Network Operations Relationship

Figure 5.7 Nevada SOC Network Operations Relationship, Gower View

Figure 5.8 Nevada SOC Network Operations Relationship, Gower View, Fully Connected Star Depicted Figure 5.9 Respondent Matrix Operations Relationship, Network Members Grouped by Whom they Represent

Figure 5.10 Respondent Matrix Operations Relationship, Network Members Grouped by Geographic Location of Representation

Figure 5.11 Nevada SOC Network Information Sharing Relationship

Figure 5.12 Nevada SOC Network Information Sharing Relationship, Gower View

Figure 5.13 Nevada SOC Network Resource Sharing Relationship

Figure 5.14 Nevada SOC Network Resource Sharing Relationship, Gower View

Figure 5.15 Nevada SOC Network Trust Relationship

Figure 5.16 Nevada SOC Network Trust Relationship, Gower View

Figure 5.17 Nevada SOC Network Summary of Emergent to Strengthened Relationship Tie Findings

Figure 5.18 Network Actors’ Perception of Effectiveness

Figure 5.19 Operations Relationship with Effectiveness Attribute

Figure 5.20 Operations Relationship with Effectiveness Attribute, Gower View

Figure 5.21 Information Relationship with Effectiveness Attribute

Figure 5.22 Information Relationship with Effectiveness Attribute, Gower View

Figure 5.23 Resources Relationship with Effectiveness Attribute

Figure 5.24 Resources Relationship with Effectiveness Attribute, Gower View

Figure 5.25 Trust Relationship with Effectiveness Attribute

Figure 5.26 Trust Relationship with Effectiveness Attribute (Respondent Matrix), Gower View

xii

List of Appendices

Appendix A Multilevel, Multitheoretical Framework to Test Hypotheses About Organizational Networks.

Appendix B System of Care Values and Principles

Appendix C Nevada System of Care Network Formation, Abbreviated Timeline

Appendix D Affiliation Network Development, 2017 List of Public Meetings (Events)

Appendix E Modified Copy of Nevada SOC Survey (with roster names removed)

Appendix F Network Actor Degree and Betweenness Centrality Scores

Appendix G Questions for Communities Based on Network Analysis, Table 2 of Provan, Veasie, Staten & Teufel-Shone (2005, p. 606).

Chapter One: Introduction

Implementing efficient and effective public services is at the core of the public administrator’s duties. In responding to complex social problems, administrators are frequently called upon to implement policy solutions that are vague and span multiple public and private entities. In doing so, the type of governance model utilized to guide implementation influences incentives, opportunities and constraints which become the context of the policy implementation. This then influences the extent to which policy or program outcomes are achieved. Attention to the implementation process is important for providing a contextual understanding of the outcomes.

The networked governance model has been gaining attention in public administration as it attends to both the complexity and the boundary spanning issues of policy implementation. Measuring the impact of this type of governance structure is as complex as the implementation itself. However, the rise of analytical tools and prolific publication of network analysis research has resulted in increased attention to analyzing policy implementation from a networked governance perspective. Monge and

Contractor (2003) state “complex systems analysis explores the behavior of a network … the network is important because it provides the context or the environment for the individual agents” (p. 86).

The purpose of this study is to describe a statewide “System of Care” (SOC) for publicly funded children’s behavioral health services from a networked governance perspective. It aims to describe the network actor attributes and the patterns of

2 relationships that comprise the overall network structure which influences implementation of the “Nevada System of Care” (hereinafter “Nevada SOC”). Figure

1.1 below presents a framework for the basis of this study.

Figure 1.1. Study Framework

Policy Tool = System of Care Approach Governance Implementation Performance Approach

• Networked • Incentives • Process Governance • Opportunities Outcomes • Constraints • Program/policy Outcomes

Who (Actors) & Impacts goal Influences how network structure attainment

The study is based on the premise that the governance approach utilized during the implementation of a public policy or program influences the outcomes of the policy or program. More specifically, the patterns of interactions within a networked governance approach shapes the incentives, opportunities and constraints of implementation which, in turn, impacts the outcomes. This study employs multi-level analyses to describe the structure of the Nevada SOC network, utilizes network actor attributes to explain aspects of the network structure and analyzes the actors’ perceived outcomes of the network.

A mixed-method approach is employed to describe the network structure and to attribute the structure to network effectiveness as perceived by the respondents. The study focuses on three research questions:

1. Who are the network actors in the implementation of the Nevada System of

Care?

2. What are the patterns of interactions within and amongst these actors?

3. To what extent do these patterns of interaction contribute to the perceived

outcomes of the policy?

According to the Substance Abuse and Mental Health Services Administration

(SAMHSA), “by 2020, behavioral health disorders will surpass all physical diseases” and

“13% to 20% of children in the United States (up to 1 out of 5 children) experience a mental disorder each year” (SAMHSA, no date). Children and their families experiencing such needs can enter any given system, such as school or child welfare systems, and then typically crossover multiple systems as they access services. Consider the following case as an example:

Jenni is a 12-year-old 7th grade student who lives in a low socioeconomic status

neighborhood. Her parents work multiple shift jobs and the family relies on

public transportation. Jenni was recently arrested for truancy. Her local juvenile

probation department identified a possible learning disorder that leaves Jenni

feeling unengaged with school, causing her to become truant. The juvenile

probation departments refers Jenni back to her school for an assessment of the

disability. The school confirms a learning disorder and also identifies a severe

emotional disturbance, so the school refers Jenni to the children’s behavioral

health system.

Families such as Jenni’s navigating these systems will typically have multiple caseworkers and multiple case plans specific to each system them encounter. It can become difficult for the child and family to follow through with the requests of all the case workers, manage transportation to and from multiple appointments and manage the multiple bills for any services rendered. When faced with multiple demands, families will often give up hope and will even consider giving up custody of their child feeling the child welfare system may be able to handle the situation better.

In recognition of this inefficiency and risk of harm to families, SAMHSA began to fund and promote the SOC approach. The approach mandates partnerships with youth and families and also mandates coordination across systems to serve the child and family holistically. SOC is an evidence-based framework utilized by states to coordinate and align public and private entities for the purpose of improving access to behavioral health treatment for children with severe emotional disturbances and to prevent children from having to go outside of the state to obtain treatment. Within a SOC, services are highly coordinated so families are not faced with multiple (and sometimes competing) caseworkers and case plans. SAMHSA (2013) states their grantees receive funding “to establish a comprehensive spectrum of mental health and other necessary services and supports organized into a coordinated network...” (p. 1). To develop and implement this coordinated network, states are expected to formalize cross agency partnerships to align funding and interventions as well as intentional partnerships with

5 youth and families to ensure their voice is incorporated in to the planning process.

Within a SOC, collaboration amongst a multitude of public and private partners across systems interacting with children and families is necessary for the success of the project.

SOC is not a specific treatment service; it is a framework for implementing a comprehensive system of children’s behavioral health services.

Given the emphasis SOC places on the coordination and collaboration of a diverse set of stakeholders, it is important to observe and map the pattern of interactions in order to gain an understanding of the extent to which this coordination is established in the structure of implementation. Increased understanding of the network structure aids the implementing agency and the administrative manager in the overall development and governance of publicly-funded children’s behavioral health services. This paper reports the findings of the Nevada SOC network structure as it implements a SOC.

Chapter two is a summary of the literature. The literature is organized by first describing the broad perspective of governance and policy implementation. Then, it narrows to describe types of governance models such as centralized governance and networked governance. Networked governance is applied to policy implementation with a discussion of the current state of the literature on networked approaches to solving social problems. The literature review concludes with a more detailed description of networks, types of networks, basic components of networks and supporting theories.

Chapter three provides a description of the study context. First, a background of

“System of Care” is provided as the evidence-based framework the state is currently implementing. A description of the incremental nature of the state’s efforts in implementing this framework is provided. The chapter concludes with a description of how the effort is defined as a “network.”

Chapter four describes the methodology employed in the study. It first describes

“phase one” of the study, which is to develop an affiliation matrix of the network based on individual attendance at SOC meetings. It then describes how archival data is utilized to generate a name roster for the survey portion of the study. The chapter describes

“phase two” of the study, which is the development and implementation of a survey of

Nevada SOC network participants. The chapter concludes with a description of how the data from both phases of the methodology is analyzed.

Chapter five presents the results of the study organized by the research questions described in chapter four. First, the results describe the Nevada SOC network actors according to specific actor attributes such as the type of organization the actor represents and their role within the SOC (i.e. parent, professional staff, etc.). To describe the patterns of interactions that comprise the network structure, results from multi-level analyses are presented including visual depictions of the network across four relationship variables (operations, sharing information, sharing resources and trust).

The chapter concludes with a description of the network actors’ perceptions of effectiveness of the Nevada SOC.

Chapter six presents a discussion of the findings, limitations of the study and a conclusion. Examples of how the results can be utilized by the SOC manager and implementing agency to change or strengthen the network structure are presented.

The chapter and the paper conclude with a discussion on the utility of utilizing the networked governance model to describe and analyze implementation of policies to address complex social problems.

Chapter Two: Literature Review

Resolving complex social problems is a primary goal of publicly funded services and the task placed before public administrators. The formula at the center of this task is typically: resources, plus an evidence-based practice equals efficiency and effectiveness while attaining outcomes intended to reduce or eliminate the targeted social problem. However, such a formula is rarely so simple. The reality is most social problems are so “wicked,” it is not possible to break them down into discrete parts with straightforward solutions.

There are three dimensions of wicked problems (Weber and Khademian, 2008).

First, wicked problems are unstructured. It is difficult to define exact causes and exact solutions. The problems evolve over time so defining the actual target is difficult.

Second, wicked problems are cross-cutting; spanning stakeholders and perspectives.

Relentless is the third dimension of a wicked problem, meaning the problem is rarely solved or ended. In response to wicked social problems, there is typically the formation of special task forces, committees, work groups, public-private partnerships, etc. who organize to coordinate a response (O’Toole, 1997). There is “emerging recognition” of collaborative arrangements enhancing effectiveness “in ways that would not be possible through the traditional governance mechanisms of market or hierarchy” (Milward &

Provan, 1998, p. 388-389).

As depicted in Figure 1.2 below, the literature review first examines the broad perspective of governance and policy implementation. Then, it narrows to types of governance models such as centralized governance and networked governance. It

9 discusses how certain governance models, such as the networked model, are responsive to the complexity of the “wicked problems.” Networked governance is then applied to policy implementation with a discussion of the current state of the literature on networked approaches to solving social problems. The literature review continues with a more detailed description of networks. This includes a discussion on types of networks, basic components of networks and theories commonly referenced in network approaches. The literature review concludes with a discussion of network analysis as a methodology commonly employed in the study of networked governance.

Figure 1.2: Structure of Literature Review

•State-Centric •Liberal-Democratic Governance Models •Network

•Top down/bottom-up approaches Policy •Ambiguity-conflict model implementation •Networked implementation

•Network types •Network models Networks •Network components •Network analysis

Governance and Policy Implementation

Governance

Governance is not synonymous with government; governance is an overarching structure and meaning of government as opposed to the functions of government itself

(Rhodes, 1996). Governance is more about the “process” of governing as opposed to the institutions themselves (Bevir, M., 2012). In their summary of the literature on governance, Or and Aranda-Jan (2017) describe governance as encompassing a movement to pull back from state-based perspectives while adding increased involvement of citizens. While there are “imprecise” definitions of the term

“governance” (Rhodes, 1996, p. 652), the International Encyclopedia of Political Science defines it as “the pursuit of collective interests, with the state as a coordinating and enabling actor” and “at its most basic level, governance means the capacity to steer the economy and society toward collective goals” (Pierre, J., 2011, p. 984-985; Peters, G., p.

996).

Governance is a process that is both internal to the state as well as external as it interacts with “other actors in society” (Pierre,2011). The increasing complexity of social problems requires the capacity of multiple governments and partners to solve problems through formal and informal network partnerships. There are three models of governance: state-centric, liberal-democratic, and network-based (Pierre, 2011). The models describe the set of institutions and the arrangements and relationships that impact how policies are developed and implemented. The following describes the

11 models of governance and concludes with a description of the networked model of governance, which serves as a basis of this study.

State-Centric Governance Model. This governance model primarily describes heavy influence from the state (i.e. the central or federal government) over its jurisdiction, including interjecting itself into markets if there is justification the intervention will benefit the state (Pierre, J., 2011). In this hierarchical model of governance, policy implementation rests with the state and includes sufficient resources and capacity necessary for such implementation. From this perspective, the role of the state is pivotal for achieving good governance (Or & Aranda-Jan, 2017). Centralized governance is a form of control centralized with a single authority as opposed to a more dispersed style of governance (Hoogh & Marks, 2003). In a review of the meanings of governance, Rhodes (1996) states “central government is no longer supreme.” Rather, the role of government is to “encourage many and varied arrangements” (p. 657). In the “network era” of public administration, Agranoff (2011) describes the role of public agencies as retaining their centric tasks, but in a “conductive” manner. Essentially, the governance model becomes one of linking and partnering in various forms with non- governmental organizations (Agranoff, 2011). From this perspective, implementation can still be viewed as state-centric as it is coordinated or directed by the state while direct services are provided by collaborative partners. From the network administrative perspective, the state is not replaced by external partners; rather, its role has shifted

(Agranoff 2011).

Liberal-Democratic Governance Model. The liberal-democratic governance model primarily rests on democratic theory with distinction between the state and society (Pierre, 2011). In this model, the role of the state is to serve in a partnership with society, but still retain decision making authority (Pierre, 2011). In other words, the state can still impose power as a central authority, but rarely has to do so. Rather, government involvement is balanced with societal freedoms.

Network Governance Model. Older generations of bureaucracy aimed to solve problems that were more specific and discrete (Agranoff & McGuire, 2001). In doing so, there was greater ease in developing and implementing policy instruments focused on insulated goals and objectives and implemented with hierarchical control. However, with increased attention to more complex problems where solutions are either unknown or not feasible, there has been a shift in governance to a networked approach.

This approach becomes one of “governance with government” (Weber & Khademian,

2008, p. 341). Whereby government retains an obligation to the public good while acknowledging government, alone, cannot solve problems. Rather, it can serve as a catalyst and facilitator of the networked model.

The network model of governance is responsive to complex social problems, or

“wicked problems,” such as children’s access to mental health services. The response to such problems typically involves the implementation of complex policy solutions in partnership with multiple individuals and organizations across public and private sectors

(Agranoff, 2017). “It is at the local level where policies and programs are implemented, where the routines build policy, and where the enduring challenges of promoting

13 economies, eliminating poverty, integrating immigrants, and building democracy take place” (Agranoff, 2014, p. 57S). As such, the networked approach provides a structural mechanism for shared information, resources and expertise across policy partners

(Weber & Khademian, 2008). There has been a rise in the networked governance approach since the global financial crisis of 2008 when resources for public services were constrained (Or & Oranda-Jan, 2017). In response, flexible and collaborative service arrangements across public and private sectors became a strategy for maximizing limited resources.

A basic premise of the networked governance model is the state does not carry the sole burden of solving the problem and it involves actors who are close to the problem in the solution. This model includes a shift from the state being the sole provider of services to an increasing trend where the central government is providing funds to “regional actors” to implement policy solutions (Klaster, Wilderom & Muntslag,

2017, p. 676). The model includes a certain level of agreement amongst actors for the policy solutions; making implementation swifter with regular engagement between the institutions and partners (Berg-Schlosser & Badie, 2011). The nature and structure of this coordination is referred to as a “network.” Whereas, the characteristics of the network, such as the patterns of relationships amongst the actors, can impact the outcomes of the policy on the social problem.

In their review of networked governance research, Klign and Koppenjan (2012) identify three types of research: policy networks, service delivery and implementation, and managing networks. From the network perspective, the authors state “governance

14 refers to the horizontal interactions by which various public and private actors at various levels of government coordinate their interdependencies in order to realise public policies and deliver public services” (p. 594). The “horizontal interactions” are contrasted with a “vertical accountability structure” found in traditional governance forms (p. 596). In the horizontal model, the accountability for policy outcomes transpires across the network members and not just up and down the vertical accountability structure. Hall and O’Toole (2000) describe the phenomenon of multiple organizations working together on all or parts of the program as “networked arrays” of implementation (p. 673). Networked implementation assumes “outcomes and performance result from interactions between a variety of actors rather than from the actions and policy of one actor alone” (Klign & Koppenjan, 2012, p. 589).

Not all networks are self-organizing and, at times, governments become a driving force in either mandating or facilitating networks (O’Toole, 2014). When a lead organization managing the network is a public-sector entity, its members and activities can create a “shadow of hierarchy” (Klaster, Wilderom & Muntslag, 2017, p. 678). This shadow is sometimes necessary in big, complex networks as it is difficult for such networks to govern themselves. However, network management by a public entity

“looms continuously” over the network and bounds the network to certain parameters defined by the organizing entity (Klaster, Wilderom & Muntslag, 2017, p. 678).

Collaborative or networked governance shifts policy implementation from the sole public entity responsible for the policy to a collaborative network consisting of actors from public and private entities who have a stake in the outcome of the policy.

However, the network perspective doesn’t necessarily replace the bureaucratic perspective (O’Toole, 2014). Instead, it adds layers of “structural complexity” (O’Toole,

2014, p. 361). The rise of networks in policy implementation is a result of trends in dispersing resources across public and private entities and shifts from hierarchical to a more collaborative decision-making processes amongst actors involved in the development and implementation of policy solutions (Knoke, 2011).

Network governance and network management. For the purpose of this study, network governance is conceived as a type of governance structure a public entity adopts when embarking on the implementation of a policy, practice or program. This position steers the direction of implementation. Essentially the central public entity responsible for the implementation, commits to a collaborative process across individuals and organizations who have a stake in the policy (Kapucu, Hu & Khosa,

2017). From this perspective, the actions of the public entity can support or withdraw from the collaborative process.

It is important to distinguish network governance from network management.

Network management would be the day to day implementation of the policy whereby management is not hierarchical; rather, it consists of managing mutuality between network members’ expertise, knowledge and resources (Agranoff, 2007). In describing intergovernmental management (the administrative process employed in a networked governance perspective), Agranoff (2017) specifies the need to focus on managing daily transactions and interdependencies of multiple entities and the manager works within a process guiding actors in taking actions to achieve the goals (p. 117). While network

16 management factors can contribute to overall effectiveness, the study of a network structure does not necessarily include the study of network management (O’Toole,

2014).

Network management. An understanding of the structural properties of a network can “enhance a community’s capacity to combine diverse knowledge and skills to come up with effective solutions to complex problems…” (Provan, Veazie, Staten &

Teufel-Shone, 2005, p. 604). Organization managers can use this understanding to make decisions about where their organization lies or ought to lie within the context of the other actors in the network (Provan, Veazie, Staten & Teufel-Shone, 2005). For example, trust is a critical component of network effectiveness (described further below). Thus, the network manager should understand the state of trust amongst members within the network and then take appropriate steps to build trust (O’Toole,

1995 as cited in O’Toole, 1997). Agranoff (2011) states “being able to create and leverage participation in network-designed and -delivered solutions is becoming a critical organizational and leadership capability” and “trust fosters this commitment and cements the network partnership” (p. 273).

In their early work, Agranoff and McGuire (2001) describe a set of “network management behaviors” as key elements utilized by network managers (pgs. 13-16).

The behaviors can manifest simultaneously, each requiring attention and adjustment through the lifecycle of the network. Those behaviors include: activating the network by identifying stakeholders and bringing them together, framing the context of the

17 network, mobilizing the actors toward the goals and synthesizing the network by maneuvering conflict and priorities.

Agranoff (2011; 2012) describes the shifting role of public managers to one of being a conductor who coordinates, links and collaborates across organizational boundaries. Much like the role of a conductor in a symphony, the public manager of the network unites the members of the network and sets the pace (i.e. beat) and tone of the network’s actions. In framing the actions and structures of collaborative management, Agranoff (2012) states “network activity is thus a variant of collaboration”

(p. 15). He builds upon the above behaviors by identifying eleven “emergent collaborative management functions” (Agranoff, 2011). The functions include providing leadership on internal and external communications, continuous promotion of the network to leverage and control the flow of information across partners, broker collaborative activities, promote mutual and continuous learning amongst network partners, organizing internal and external partners, coordinate information that helps the network in decision making processes (i.e. compiling existing needs assessments), promotion of staff who are skilled in the behaviors required of conductive network management, development of partners who represent diverse interests while sharing the same commitment to the problem to be solved and building communities of knowledge from client-based perspectives (p. 286-288).

Given the distinction described above, this study aims to describe the implementation of a policy under a networked governance structure. While the findings can inform the management of the network, this is not the primary aim of the study.

The findings of this study are useful to the network manager by gaining a baseline of the structure of the network which will allow for the network manager to take steps to strengthen the network function and structure. As this study is interested in exploring policy implementation from a networked governance perspective, the following describes literature in policy implementation.

Implementation

When viewed as a process, stakeholders or actors engage in public policy across phases ranging from problem identification and agenda setting to implementation and evaluation (Birkland, 2001; Weible & Sabatier, 2018). This section focuses on policy implementation as a critical component of the process. I argue governance structures such as those described above impact the capacity, resources and approach to policy implementation. In turn, this influence on implementation becomes a factor in the success or failure of a public policy. The following describes various perspectives and approaches to the study of policy implementation. These perspectives inform the standpoint(s) from which research questions are proposed and influenced the development of the research questions in this study.

Implementation is a “problem-solving activity that involves behaviors that have both administrative and political content” (Goggin, 1986, p. 330). As opposed to a single event, Goggin clarifies policy implementation as a process occurring over time. When coupled with the decisions and behaviors of the multiple actors involved with the implementation and the policy context, the result is numerous variables for a single

19 policy under study. He argues implementation behaviors and their associated outcomes should not be confused with actual programmatic outcomes, as they typically are. He clarifies successful implementation doesn’t necessarily mean the policy or program is successful. In his description of the different types of implementation, he describes

“coordinated implementation” as one of the more successful types of implementation

(p. 331).

The early stages of implementation research primarily utilized a case study approach (Goggin, 1986; Sabatier, 1986). Implementation research started with single case designs, which typically resulted in “pessimistic conclusions about the ability of governments to effectively implement their programs” (Sabatier, 1986, p. 21). This era of research was characterized by an increased “interdependence of several levels of government,” which resulted in more complex implementation arrangements (Chubb,

1984, p. 994). Thus, it was difficult for researchers to link the implementation structures with the outcomes.

A second generation of implementation research is characterized as a “top- down” perspective where the analysis starts with the policy decision then examines the extent to which mandates in the policy are achieved (Sabatier, 1986). Within this top- down approach, policy designers are the central actors (Matland, 1995).

Implementation research from this perspective focuses on the degree to which the policy goals are achieved and typically calls for the development of clear goals at the outset. Thus, a challenge in conducting implementation research from this perspective is the originating policy frequently does not contain clear and consistent program goals

(Sabatier, 1986). Additionally, this perspective does not fit implementation scenarios that are missing a single policy or originating entity. Rather, the implementation may include multiple government directives and actors.

The next phase of implementation research unfolded as the “bottom-up” perspective where analysis begins at the frontline where the policy is operationalized

(Matland, 1995). DeLeon & DeLeon (2002) “argue that policy implementation has too often been practiced as a top-down or governing-elite phenomenon and that its study and practice would be much better served were its practitioners to adopt a more participatory, more directly democratic orientation” (p. 468). In this era of implementation research, Sabatier (1986) describes bottom-up approaches as first identifying the “network of actors involved in service delivery” and then asking them about their involvement in the planning and execution of the policy as well as their connection to other actors involved in the implementation (p. 32). From this perspective, it is those who are delivering the services who ultimately share the outcomes of the policy (Matland, 1995). Implementation from this perspective typically recommends flexible strategies allowing those on the front lines of implementation to adapt their approaches and services according to their direct service experiences

(Matland, 1995). From the bottom up perspective, implementation of a federal policy can vary from state to state and is more susceptible to contextual influences. This means outcomes for the same policy can vary widely under differing implementation structures.

This “top-down” vs. “bottom-up” perspective has been framed as a dichotomy in past approaches to the study of policy implementation (Hall & O’Toole, 2000). While a public entity may be responsible for the implementation of a policy, research on public administration remains distinct from research on policy implementation (Hall & O’Toole,

2000). From the public administration perspective, the study of implementation places the public agency at the center of the analysis (i.e. top-down). Whereas, the policy implementation research perspective places the public program itself at the center of the analysis (i.e. bottom-up). This placement of the unit of analysis either with the organization or the program carries implications for how the implementing structures are studied and how outcomes are defined. For example, outcomes from a top-down perspective would be defined as the extent to which the original policy goals are achieved; whereas outcomes from a bottom up perspective may center more on the general effects of the program (Matland, 1995).

Understanding the critique of implementation literature from this policy- administration dichotomy perspective recognizes the difficulty in separating politics from administration and vice versa. Implementation research focused only on the public administrative side risks missing a frontline perspective of implementation and an understanding of the extent to which a social problem is actually impacted. Likewise, implementation research at the programmatic level risks missing an understanding of the structural mechanisms that supported attainment of programmatic goals.

Following this era of implementation research centered on an originating point, there were efforts to combine the top-down and bottom-up perspectives into single

22 approaches. The problem continued to be the production of lengthy lists of variables that impact implementation (Matland, 1995). In a seminal piece on organizing the implementation literature into a structure of variables, Matland (1995) presents the

“ambiguity-conflict model” of implementation (p. 155-160). In this model, policy ambiguity occurs when there are unclear goals of the policy and/or there are unclear means or resources associated with those goals. Whereas, policy conflict occurs when there are differences in perspective of the policy goals and/or the activities of the policy.

The ambiguity typically results in or is caused by misunderstandings and uncertainty.

Matland (1995) describes four implementation types in a matrix of policy ambiguity and policy conflict. The four types are: administrative implementation, political implementation, experimental implementation and symbolic implementation.

Within each implementation type, Matland identifies the central principle having the greatest impact on outcomes (p. 160). In “Administrative Implementation, the outcomes would be determined by resources that are appropriated. Power is the central principle that determines outcomes in “Political Implementation.” In the

“Experimental Implementation” type, contextual conditions are the primary determinant of outcomes and in the “Symbolic Implementation” type the outcomes are determined by coalition strength. Figure 2.1 below presents Matland’s four implementation types in the ambiguity-conflict matrix with the central principle noted in parentheses and a short list of characteristics associated with that implementation type and central principle.

Figure 2.1. Matland’s (1995) Ambiguity-Conflict Matrix of Implementation

Conflict

Low High

Low Administrative Implementation Political Implementation (Resources) (Power) ● Resource availability determines ● Actors’ goals are clear, but outcomes. conflict with one another. ● Authority lies at the “top” and ● Power impacts information flows down. implementation (i.e. which ● Closed system. actor has the power to have ● Implementation runs like a their goals prioritized?). machine where tools are put ● Open system. into action. Failure occurs when ● Implementation activities the machine “sputters.” include negotiating and bargaining.

High Experimental Implementation Symbolic Implementation (Contextual Conditions) (Coalition Strength)

● Which actors are involved, and ● Local level coalitions impact the intensity of their involvement outcomes through influences the outcomes. determination of the course Ambiguity ● Outcomes are difficult to predict of the policy and control of as they are dependent on the the resources. varied actors, resources and ● Conflict arises due to implementation sites. conflicting determinations of ● Varied implementation may make policy goals and competition it difficult to assess extent to for resources. which outcomes are achieved. ● Coalition strength and Formative evaluation is used to opposition coalitions impact describe the process. outcomes. ● Open system is influenced by the ● Implementation includes environment. ● With specific program outcomes intense involvement of being difficult to define, the actors with their influence process or learning outcomes tied to the strength of the become the essence of coalition. implementation. Source: Matland, R. (1995). Synthesizing implementation literature: The ambiguity-conflict model of policy implementation. Journal of Public Administration Research and Theory, 5(2), 145-174.

Matland’s ambiguity-conflict matrix frames a theoretical structure of implementation literature. It organizes factors and patterns of influence that impact policy outcomes through varied implementation approaches. Matland concludes this model “provides a map [for the researcher] to those elements expected to most greatly influence policy outcomes” (p. 171). This theoretical structure is re-visited in Chapter

Six (Discussion, Limitations and Conclusion) below.

There are other influential factors in the study of policy implementation.

Mischen and Jackson (2008) describe “historicity” as a factor of policy implementation.

Considering implementation as a process occurring over time, time generates history with unique organizational and implementation sites. Historical factors set the context for implementation. “Implementation researchers often feel stymied by the wide variation in policy outcomes in different locations” (Mischen & Jackson, 2008, p. 322).

Monge and Contractor (2003) explain the “magnitude of the relationship” between the actors in a network is impacted by the history of the system and where the status or structure of the network is at in the beginning of its formation or at the end (p. 84).

Networked Implementation

Implementation of public programs from the networked perspective essentially means “funding and implementation will not be concentrated in a single government entity” (Weber & Khademian, 2008, p. 341). Within networked implementation

“interorganizational linkages” become a central component of implementation whereby

“governance is about managing networks” (Rhodes, 1996, p. 658). Under this model, a

25 government entity may take on the role of creating and managing networks for problem solving and service delivery as opposed to directly providing services. By nature of the complexity of wicked problems, it is not always feasible to achieve policy or program outcomes in the exact manner initially prescribed. Thus, the benefit of networked implementation can be its responsiveness to programmatic needs and resources. The federal government has options for implementing in such ways, such as demonstration and waiver programs tying innovative service provision to research for demonstration of both process and program outcomes (Agranoff, 2012; 2017). However, networked implementation presents unique challenges for conducting research on the effectiveness of the network activities.

Dispersed service delivery under a networked implementation model presents a challenge as there is reduced control over the specific program activities. Private providers may be less inclined to abide by strict implementation rules; thus, making it difficult to achieve accountability and performance outcomes (Agranoff, 2017; Rhodes,

1996). Describing or predicting how a network’s structure impacts outcomes becomes as complex as the implementation itself. Networked implementation is neither top- down nor bottom-up. It is multi-level and involves multiple actors across those levels in a flexible and adaptable manner. It is not a stagnant nor prescriptive approach to implementation. Analyzing policy implementation from a networked perspective acknowledges the inherent complexity and impact of context. At the same time, it also acknowledges the difficulty in measuring implementation from this perspective (Ward,

Stovel & Sacks, 2011). The following will provide more detail on networks and network

26 analysis. This study embraces the networked approach to governance and implementation while accessing recent advancements in network analysis techniques to describe the network structure and its possible impact on outcomes.

Networks

A network is comprised of actors (individuals, groups, organizations) who use

“flexible, dynamic communication linkages” to span boundaries and work collaboratively on complex issues (Contractor, Wasserman & Faust, 2006, p. 681). From those efforts, patterns of relationships amongst the actors emerge (Kapucu, Hu &

Khosa, 2017). The patterns of relationships occur at different levels: the individual actor, between two actors (dyads) and relationships between small groups (triads)

(Yang, Keller & Zheng, 2017). The patterns of relationships across the different levels comprise the overall structure of the network (O’Toole, 1997; Ward, Stovel & Sacks,

2011; Wasserman & Faust, 1994). Essentially, the network structure is the collection or pattern of smaller structures within the network. The network structure represents an interdependence amongst the actors where hierarchy is not simply the result of a formal position or bureaucratic structure (O’Toole, 1997).

One of the earliest known studies of a social network was by psychiatrist Jacob

Moreno in 1933 (Ward, Stovel & Sacks, 2011; Yang, Keller & Zheng, 2017). The New

York Times (1933) reported on his study of relationships amongst girls within a school.

In his results he was able to identify girls who are isolated, cliques, as well as popular and unpopular girls. The article reported that Dr. Moreno predicted two of the girls

27 would run away from the school and he ended up being correct in his prediction. In this article he was quoted as stating “such an invisible structure underlies society and has its influence in determining the conduct of society as a whole” (The New York Times, 1933).

In applying Moreno’s statement to the study of policy implementation, the “invisible structure” would underlie the context of implementation thereby influencing the outcome of the policy.

Policy networks consist of “a set of public and private corporate actors linked by communication ties for exchanging information, expertise, trust and other political resources” (Kenis & Schneider, 1991, as cited in Knoke, 2011, pg. 211). The network itself is a form of social capital whereby it becomes the “conduit” for the flow of information and resources and is impacted by the size and range of the network (Burt,

1992, p. 12). Given a network is comprised of individuals and organizations connected through various types of relationships, network analysis is the study of those patterns of relationships that comprise the structure of the whole network. While this notion of collaborative policy implementation is not new, the use of network analysis to empirically describe and predict network outcomes is on the rise. Table 2.1 below defines some common terminology used to describe the involvement of multiple actors in policy implementation efforts.

Table 2.1

Common Terms to Describe Multiple Actors in Policy Implementation

Term Description(s) Source(s)

Collaborative Deliberative management and Agranoff management coordination of multiple sectors across (2012). varied levels of government and private sectors to solve complex problems.

Conductive The role of public agencies in a network Agranoff (2011; Organization era where the agency facilitates, 2012). coordinates and links network partners to achieve and remain accountable for objectives.

Federalism A set of governments nested within Hooghe, L. & each other with allocated and shared Marks, G. authority and resources extended to (2003). the local level.

Intergovernmental Partnerships amongst governments, Agranoff (2017, management non-government organizations, and p. 3, 9) other groups to transform policies in to public programs. Government is not replaced but increases the number of external actors involved in implementing programs.

Intraorganizational Coordination occurs but is managed Agranoff, R. & through hierarchies and strict lines of McGuire, M. communication. (2001).

Multi-level System of nested governments Hooghe, L. & Governance engaging in continuous negotiation at Marks, G. multiple levels. (2003, p. 234).

Network The structure is dispersed, not Agranoff, R. & centralized or hierarchical. McGuire, M. (2001).

Types of Networks

As networks are responsive to contexts, it is difficult to define a single type of network. The structures evolve over time and are influenced by the network’s goals and resources. Provan and Kenis (2007) propose the network type or structure can predict network-level outcomes; thus, suggesting the need to measure and describe the network type.

In a classic study of public organization networks, Agranoff (2007) presents four types of networks: informational, developmental, outreach and action networks. The informational network primarily focuses on the exchange of information where partners come together to exchange technology, policies and solutions. This type of network is also identified by Lecy, Mergel and Schmitz (2014) as a cooperative network. Agranoff’s

(2007) developmental network includes the exchange of technologies, but with the aim of increasing the network members’ capacity to implement solutions. The outreach network includes the exchange of information and technologies, but also coordinates an array of solutions to increase access to the opportunities. Finally, the action network consists of formalizing collaborations amongst network members who adjust within their own agencies to promote program delivery, exchange of information and technology. Lecy, Mergel and Schmitz (2014) refer to this type of network as a coordination network where network organizations take steps to align policies and maximize resources. The following describes additional models of network types.

Shared governance or collaborative models. The shared governance or collaborative model of networks describes the working relationship amongst the

30 members of the network. In this model, the network members make all decisions collectively without coordination from a governing entity (Cristofoli, Markovic &

Meneguzzo, 2014). Through their collaboration, network organizations exhibit and rely on their interdependence (Lecy, Mergel & Schmitz, 2014). This type of network is also known as a “participant-governed” network model that is decentralized with formal and informal processes involving most or all of the network members (Provan & Kenis, 2007, p. 234). There is some debate on the extent to which dispersed network relations contribute to network effectiveness (Klaster, Wilderom & Muntslag, 2017).

Lead organization models. The lead organization network model designates an organization to serve as the lead in network activities. This network model contrasts with the shared governance model in that a coordinating agency serves as a highly centralized entity and actively coordinates the brokering and communication roles of the network (Provan & Kenis, 2007). The lead organization provides administrative support for the network such as facilitation and background work (Cristofoli, Markovic &

Meneguzzo, 2014). While the organization serves as a lead, they also operate as a member of the network. This network model is considered to be more efficient than other models as the involvement of a lead organization lends legitimacy to the network

(Klaster, Wilderom & Muntslag, 2017). However, the involvement of a lead organization can create tension amongst network members if the lead organization is perceived to have an “agenda” they are imposing on network members.

Network Administrative Organization. A slightly different version of the lead organization model is the Network Administrative Organization (NAO) model (Cristofoli,

Markovic & Meneguzzo, 2014; Provan & Kenis, 2007). This network model also has a lead organization providing administrative and facilitation support, but the distinguishing difference is the organization does not simultaneously serve as a member of the network too. This model is appropriate for large and highly complex networks.

Out of this model, smaller boards tend to emerge as leadership groups who set the direction for the network and facilitate decision making necessary in order to maintain the momentum of the network (Provan & Kenis, 2007).

Affiliation networks. An affiliation network is identified by the actors’ linkage to other actors through events, such as participation in meetings. Analysis of affiliation networks consists of examining the existence of “co-relationships.” This co-relationship creates opportunity for communication and collaboration through a social tie at or with the event (Borgatti &Halgin, 2011; Mulvaney, Lee, Hook & Prokopy, 2015). An affiliation network presumes attendance at the same event represents a tie as it is an opportunity for the actors to share information and develop a relationship (Borgatti, Everett &

Johnson, 2013).

Network Power and Effectiveness

Network power. Power within a network arises from the actor’s unique position in the network as well as the number and type of connections the actor has (Slaughter,

2009 as cited in Agranoff, 2017). “Different actors occupy different role positions and carry different weights within networks” (Agranoff & McGuire, 2001, p. 19). Unequal weighting can contribute to power differentials amongst the members of the network.

Social capital from a network perspective is “access to people with specific resources, which creates a correlation between theirs and yours” (Burt, 1992, p. 11). This social capital can equate to power within a network (i.e. the more you are able to connect with others who have resources, the more power you will have in the network).

Measuring power within a network will be discussed further in the Network Analysis section and Common Measures in Network Analysis.

Network Effectiveness. Network effectiveness is broadly defined as the achievement of network goals (Klaster, Wilderom and Muntslag, 2017). However, defining “effectiveness” can be difficult as it is dependent on the context and the perspective of the actor as well as their experience within the network (Klaster,

Wilderom and Muntslag, 2017). While program effectiveness at the individual provider level is necessary to measure, it is just as important to collect outcomes at the network level (Provan & Milward, 1995). However, network analysts are still defining mechanisms for defining a strong versus weak network structure and attributing network structure to network effectiveness and goal attainment.

Examining “network structure may capture important contours of opportunity and constraints that shape social, political or economic behavior” (Ward, Stovel and

Sacks, 2011, p. 246). A weak network structure is sparse with few connections between the members of the network; whereas, a strong network is dense and highly connected

(Ward, Stovel and Sacks, 2011). Whereas, Provan and Kenis (2007) propose that as a network grows in size and as trust becomes less dispersed, the network structure

33 becomes a lead or NAO model and is likely to be more effective than less centralized network types such as the shared-governance model (p. 237).

In their study of mental health service provider networks, Provan and Milward

(1995) examine the connection between the structure of the networks and network effectiveness. The findings of the study confirmed the challenges in identifying strategies for defining network outcomes and attributing those outcomes to the network’s structure. For example, the authors employed strategies to gather multiple views of client outcomes and still did not have a single measure of client outcomes emerge. It was simply multiple views bringing multiple perspectives. To measure network structure, the authors examined the cohesiveness of the network (i.e. network density) and the network power and control (i.e. centralization). Provan and Milward’s

(1995) study of four large mental health networks concluded with the authors proposing

“networks will be effective under structural conditions of centralized integration and direct, non-fragmented external control” (p. 23).

Thus far, I have broadly described models of governance, networked implementation and types of networks involved in implementation. I have also discussed considerations of power and effectiveness within networks. The following literature will focus more on the basic components of a network, theoretical explanations of network behavior and network analysis as a methodology. The remaining literature review also includes a description of common sources of data and analysis techniques at different levels of measurement.

Basic Network Components

Given the array of network types, the following describes basic components of a network commonly referenced in network analysis.

Nodes. The members of the network, who become the focus of the analysis are often referred to as the “actors” (Wasserman & Faust, 1994). In the public policy field, they may also be considered key stakeholders. As stakeholders, network members can include funders, clients, public employees and members of the general public (Klaster,

Wilderom & Muntslag, 2017, p. 677). In network analysis, the actors are described as

“nodes” (Lecy, Mergel & Schmitz, 2014; Mischen, 2008). In visual depictions of networks, nodes are typically represented by a circle or dot. In those visual depictions, colors, shapes and size can be assigned to the nodes to illustrate certain characteristics of those nodes.

Ties. Ties are the types of relations that connect nodes throughout the network

(Yang, Keller & Zheng, 2017). How a tie is defined is often dependent on the nature of the study. Wasserman and Faust (1994) describe “network researchers often define actor set boundaries based on the relative frequency of interaction, or intensity of ties among members as contrasted with non-members” (p. 31). Common approaches in defining ties can include physical presence together such as working together in an organization, inclusion or proximity to geographic area, interactions such as attending events at the same time (Lecy, Mergel & Schmitz, 2014) and communication (Mischen,

2008). Ties can represent specific dimensions of interorganizational relations, which includes trust, support and the exchange of resources (Knoke, 2011).

From a theoretical perspective, Granovetter (1973) describes positive and symmetric ties as either being “strong, weak, or absent” (p. 1361). The more time two or more actors spend together, the more likely they are to have strong ties

(Granovetter, 1973). By contrast, weak ties have less frequent connections or contact

(Granovetter, 1973). From a diffusion of innovation perspective, Granovetter argues weak ties are actually more likely to diffuse information as there are more paths (i.e. connections) for information to travel. Whereas, strong ties tend to form cliques where the information flow stays within the clique as opposed to spreading through multiple weak ties. As such, individuals with multiple weak ties serve as bridges (see next section) and often play a more central role in the diffusion process.

Burt (1992) introduces the role of “holes” in networks. He describes cohesion in networks as the direct connections or ties between network actors. Whereas, structural equivalence is when an actor is tied indirectly to another actor (i.e. information flows from one through another). The disconnects between actors are “holes” in the network. For example, there may be clusters of actors tied together in the network, but then there is a space between those clusters. According to Burt, it is Granovetter’s

“weak tie” that spans those spaces or holes (p. 27). He adds there is strength or power in the actor who bridges the structural hole. The structural hole position generates information and control benefits in the network (Burt, 1992).

Bridges. When a network is visually depicted, it is possible to observe clusters of network members tied as groups or cliques. Given the clusters, there may be some network members connected to more than one group; they span groups. These

36 network members are referred to as the “bridge.” Granovetter (1973) describes the important role of bridge actors as they can serve as the only information route between other actors who would otherwise not be connected together.

Theoretical Explanations of Network Behavior

As described above, the overall network structure (whole network) is comprised of smaller structures within the network such as relationships between two (dyadic) or three (triadic) actors. Thus, network analysis can measure patterns of relationships at the actor, dyadic, triadic or whole network levels (Borgatti, Everett & Johnson, 2013;

Yang, Keller & Zheng, 2017). While this study is primarily concerned with describing the structure of the Nevada SOC network at the whole network level, it also includes analyses at the other levels. The following describes theories of network behavior at the different levels of analyses.

Contractor, Wasserman and Faust (2006) broke new ground in the study of networks by linking theoretical constructs with specific network hypotheses at the varying levels of measurement. The authors propose ten hypotheses for testing at the actor, dyadic, triadic and the whole network levels (see Appendix A). As such, the authors contend network analysis provides an opportunity to not only describe the structural characteristics of the network, but to also explain them by testing the probability of ties as present or absent within a network. The hypotheses are connected to “families” of theories such as collective action, social exchange theories, cognitive consistency theories and theories of self-interest (Contractor, Wasserman & Faust,

2006). While this study does not aim to test hypotheses, the theoretical constructs are useful in determining appropriate measures for describing multi-level network behavior.

Collective Action Theory – Whole network/global level. Collective action theory describes the behaviors of actors within groups and organizations as they participate in acting for the common good. As a theoretical construct, it is applied to networks at the global level through the property of network centralization which is discussed further below (Contractor, Wasserman & Faust, 2006). In describing groups and organizations from the collective action perspective, organizations are “expected to further the interests of their members” (Olson, 1965, p. 6). This includes an expectation of the state to “further the common interests of its citizens” (Olson, 1965, p. 7). Applied to networks, the “collective outcomes we wish to explain are negotiated agreements and their implementation” (Sabatier, Leach, Lubell & Pelkey, 2005 p. 186). Thus, network analysis identifies the structure through which agreements can be negotiated while also impacting the flow of information or public goods to other members of the network.

As rational actors, members of groups make decisions about their participation according to their analysis of the cost of participation versus the perceived benefit of participation; their participation is dependent on their relational experiences in the network (DeMarrais & Earle, 2017; Olson, 1965). This can result in the “free-rider” problem where individuals may not participate because they perceive the problem already has or can resolved by the other members of the group or that the actor doesn’t have to do anything about the problem. In networks, the free-rider actors may be less central to the network. However, there is a social incentive in the actor’s decision-

38 making process of whether or not to participate in the group (Olson, 1965). This “social pressure” may induce actors to participate toward the group’s goals (Olson, 1965, p.

60). Essentially, the actor’s relative position in the network becomes the frame of reference or context for their decision to participate in the network. For example, if the actor’s primary ties are to individuals who do not perceive the functions of the network to be valuable, then the actor themselves may not perceive value either.

Collective action theory explains the probability of ties occurring or not occurring based on properties of centralization (Contractor, Wasserman and Faust, 2006).

Centrality is discussed further below under network measures, but it is the notion that a network has a set of actors who are more central within the network (as measured by the number of ties they have with other members of the network). From a collective action perspective, the network is more likely to obtain a “collective good if the network is centralized” (Contractor, Wasserman & Faust, 2006, p. 689; Marwell, Oliver & Prahl,

1988).

Cognitive consistency theories – Triadic level. Cognitive consistency theories such as Balance Theory explain network behavior at the triadic level of measurement

(ties between groups of three actors) and is applied to networks through the property of network transitivity which is discussed further below (Contractor, Wasserman & Faust,

2006). Describing the patterns of relationships between groups of three actors reveals network sub-structures such as “a friend of my friend is also my friend,” but when applied to formal organizations it is also an indicator of rank or hierarchy in the network as “the boss of my friend is also my boss” (Contractor, Wasserman & Faust, 2006; Nooy,

Mrvar & Batagelj, 2011). Essentially, cognitive consistency theories capture the actors’ tendency to seek balance in their network relations and form clusters in doing so (Nooy,

Mrvar & Batagelj, 2011). For example, an incomplete or unbalanced triad can occur in a social group where two actors share the same belief but the third actor does not

(Contractor, Wasserman & Faust, 2006). Balance theory suggests that the third actor, feeling the pressure of an imbalance, would be driven to complete the triad by adjusting their belief in order to be in alignment with the other two actors in the triad (Nooy,

Mrvar & Batagelj, 2011). Thus, cognitive consistency theories explain the transitive behaviors of actors in a network. Balanced clusters in a network is an indication of cohesive subgroups (Nooy, Mrvar & Batagelj, 2011).

The structural implications of Balance Theory in networks originates with Simmel

(1922) who initially raised the question of the impact of triads within a network and was further developed by Cartwright and Harary (1956) who identified the following set of four assumptions that must be present in a balanced network (as cited in Rawlings,

2017, p. 512):

1. A friend of a friend is a friend,

2. A friend of an enemy is an enemy,

3. An enemy of a friend is an enemy, and

4. An enemy of an enemy is a friend.

In network analysis, Balance Theory identifies different models of network structures where certain types of triads are permitted or forbidden according to the four assumptions listed above (Rawlings, 2017). The models describe the overall

40 structure of the network such as the extent to which the network is clustered or hierarchical (Nooy, Mrvar & Batagelj, 2011). Triad types are discussed further below in under common network analysis techniques.

Balanced clusters take time to form (i.e. it takes time for the phenomena of recognizing imbalance and then adjusting to seek balance). Thus, testing for models of balance is best for longitudinal network studies (Nooy, Mrvar & Batagelj, 2011;

Rawlings, 2017). The frequency of triadic types found in the Nevada SOC network is reported in this study as the initial point in time (not over time) and can be compared with data collected in future studies.

Social Exchange Theories – Dyadic level. Collaboration is an essential function in networks (Agranoff & McGuire, 2001; Cigler, 2001) and “social network analysis is based on an assumption of the importance of relationships among interacting units”

(Wasserman & Faust,1994, p. 4). Social exchange theories explain network behavior at the dyadic level of measurement (ties between two actors) and is applied to networks through the property of network reciprocity which is discussed further below

(Contractor, Wasserman & Faust, 2006). This theoretical perspective explains the tendency of ties occurring mutually between two actors within the network. For example, this study investigates the extent to which actors in the Nevada SOC network share information and resources. Reciprocity measures the number of ties that occur when such relationships are reciprocated amongst the network actors (Borgatti, Everett

& Johnson, 2013). A network with a large number of reciprocated ties has implications for the overall network functioning (Contractor, Wasserman & Faust, 2006).

Information sharing and trust are common factors impacting exchange within network behavior (Agranoff, 2017). The following elaborates on information sharing and trust from a network perspective.

Information sharing. As noted above, sharing information is a structural feature constituting a “tie” with other members within the network (Mischen & Jackson, 2008;

Klaster, Wilderom & Muntslag, 2017.) In their study of the role of trust in economic transactions, Dyer and Chu (2003) identify information sharing as a “value-creating behavior” ultimately resulting in reduced transaction costs (p. 61). The authors further studied the directional impact of trust and information sharing. They find mutual causality with information sharing creating trust and trust resulting from information sharing (p. 66). Information sharing differs across network members depending on the extent to which they share the same vision and values. By the very nature of wicked problems, network members are diverse in the experiences, backgrounds and values.

This “variance in value” difference is assigned to the information shared by the individual network members (Weber & Khademian, 2008, p. 338).

Trust. Trust is an important feature within networks as it is necessary for network members to trust the information, resources and ideas they share with one another to solve complex problems. “If partners report good relationships with a particular agency to which they are actually linked (i.e. a task-specific tie), then that agency is likely to be trusted” (Provan, Huang & Milward, 2009, p. 882). In one of the first cross-national studies of trust in supplier-buyer relations, Dyer and Chu (2003) tested the role of trust in reducing transaction costs in economic exchanges. The

42 authors define trust as “one party’s confidence that the other party in the exchange relationship will not exploit its vulnerabilities” (p. 58). The authors continue to describe the connection between trust and information sharing whereby trust reduces the need to formalize relationships, thus reducing the cost of such transactions and increasing the efficiency. For example, one actor’s trust of another actor allows them to freely exchange information and produce effort for one another without the need for formal agreements, negotiations and contracts. Trust reduces “strategic uncertainty” and

“facilitates investments” across actors with different policy frames (Klijn and Koppenjan,

2012, p. 594).

As a characteristic of network activity, “game-like interactions, rooted in trust and regulated by rules of the game negotiated and agreed by network participants”

(Rhodes, 1996, p. 66). In the reverse, a lack of trust can result in actors suppressing or not sharing information. Trust is considered an indicator of network effectiveness but is noted to be a “soft” indicator as trust alone does not necessarily conclude effectiveness of the network as a whole structure (Klaster, Wilderom & Muntslag, 2017, p. 677).

It is important to note that certain types of relationships may be more evident in the early formation of a network versus a network that is long standing. Essentially, some types of relationships such as trust take more time for members to develop than other types of relationships such as sharing information. Provan and Lemaire (2012) describe those initial relationships as “emergent” where network members are initially forming relationships to a more “strengthened” relationship of trust (p. 641-642). This

43 concept of emergent to strengthened relationships is re-visited in Chapter Five where it serves as a framework for describing the results of this study.

Theories of homophily – Actor level. At the actor level of analysis, the specific actor attributes such as age, gender organizational role and length of time in the network can explain the propensity of ties to occur or not occur (Contractor,

Wasserman & Faust, 2006). As such, theories of homophily explain network behavior by describing the extent to which actors tend to form ties with other actors who share the same attributes (Contractor, Wasserman & Faust, 2006). This is discussed further below under measures of homophily.

The theories described above form the basis of explaining network behavior from a multi-level perspective. Network analysis as a methodology is discussed further below. The description of common measures and data analysis techniques expands on the multi-level theories further.

Network Analysis

Network analysis is the study of the patterns of interactions among actors within a policy or administrative arena (Provan, Veazie, Staten & Teufel-Shone, 2005).

Network analysis “provides a language and methodology to examine relationships in order to facilitate the achievement of goals, such as implementing policy, or to identify roadblocks to successful implementation” (Mischen & Jackson, 2008, p. 324). An understanding of the network structure can provide insights into the potential successes, barriers and changes necessary for a network to function effectively.

In Lecy, Mergel and Schmitz’ review of network research literature in public administration (2014), the authors defined three primary areas of network research.

The first is a focus on policy formation, which is primarily concerned with inputs and activities aimed at influencing policy makers from agenda setting to policy development.

The second area of network research identified is policy governance. This area of research focuses primarily on the structure of governance as a network as opposed to more traditional bureaucratic formations and the impact this structure has on outcomes. Research questions within this area of inquiry tend to focus on representation within the network, accountability and equity. The third research cluster identified by Lecy, Mergel and Schmitz is policy implementation networks, which is the focus of this study. This area of focus is primarily concerned with the delivery of services. Efficiency, effectiveness and management strategies are the focus of research questions within this area of inquiry. The authors note some overlap between the governance and implementation network research, which was explained as a logical process as implementation through networks becomes possible from a networked governance perspective.

In any network analysis, it is important to define the boundaries of the network to be studied. However, it can be difficult to draw an absolute boundary as the network actors are “relatively bounded” (Wasserman & Faust, 1994, p. 31). There are two kinds of network research designs: whole network and individual or ego network designs

(Borgatti, Everett, & Johnson, 2013; Marsden, 2011; Yang, Keller & Zheng, 2017). The whole network research design studies the ties amongst all nodes within the network

45 and is concerned with how those ties comprise an overall network structure (Marsden,

2011; Yang, Keller & Zheng, 2017). Whereas, the individual or ego network design places the individual or node in the center of the analysis and is concerned with the ties connected to a particular node (Marsden, 2011; Yang, Keller & Zheng, 2017). This study is based on the whole network research design.

There are two primary types of variables included in network analysis: structure and composition (Wasserman and Faust, 1994). Structural variables are the distinct transactions or ties between pairs of actors within the network; whereas, composition variables measure certain attributes of the actors such as basic demographics and geographic locations (Wasserman & Faust, 1994). Thus, conducting a network analysis requires a determination of whether to only measure the structure or to also include an analysis of the composition of the structure. My study contains both structural variables and composition variables. This is described further in the methodology chapter

(Chapter Four) below.

There are two primary “modes” within network analysis: one-mode and two- mode (Wasserman & Faust, 1994). A one-mode network consists of one set of actors and the study of relationships amongst those actors (Wasserman & Faust, 1994). For example, studying the relationships amongst one set of children within a classroom setting would be a one-mode network study. A two-mode network study consists of analyzing the relationships between two groups of actors and how those relationships span across the two groups (Wasserman & Faust, 1994). My study includes both a one- mode and two-mode network analysis. The two-mode network is called an “affiliation

46 network” (described above in “types of networks”). In this type of network, one mode is the set of actors who are affiliated with each other through the second mode which is the “event” the actors belong to (Wasserman & Faust, 1994). An example of a two- mode or affiliation network would be a group of providers who are tied together through their mutual participation in an event such as a training.

Sources of Data in Network Analysis

In their review of network analysis studies within public administration research,

Kapucu, Hu and Khosa (2017) identified the use of surveys as the most common approach to collecting network data. These primary sources of data ask respondents directly via interview or survey about their ties with others in the network. In noting limitations relating to respondent recall bias and access to all members of a network, the authors recommend a mixed-method approach combining data collected from primary and secondary sources. “Because social network data are relational, the survey instruments used to collect the data are formatted quite differently from traditional surveys” (Mischen & Jackson, 2008, p. 324). The surveys typically include a list or roster of names of actors known to be in the network and then asks respondents specific questions about their relational ties to each of those actors. In an open ended or free recall format, respondents may be prompted to generate their own list of actors and then respond to questions about their relational ties to the actors they named.

Common threats to the validity in the design and collection of primary network data include: omission errors, commission errors, edge/node attribution errors, data

47 collection and retrospective errors (Borgatti, Everett, and Johnson, 2013). These threats were taken in to consideration during the design of the survey employed for this study and are discussed further in the methodology section.

Secondary sources such as government records, newspaper articles and online resources are common sources of data when identifying affiliations such as attendance at events (Borgatti & Halgin, 2011). Other secondary sources included in large network studies include patterns of interaction in social media platforms and online behaviors such as commenting or liking certain web pages. As discussed in Chapter Four below, this study utilized public meeting minutes as a secondary source of data for identifying

Nevada SOC network actors.

Common Measures in Network Analysis

The following describes common measures of analyzing the network structure at the different levels of analyses. Some of these measures were briefly mentioned in the theoretical explanations above. As network analysis is the study of relationships among specified actors, those relationships are not independent from one another nor is the sample representative of the general population. Therefore, common statistical methods such as analysis of variance are not typically employed in network research

(Contractor, Wasserman & Faust, 2006; Borgatti, Everett & Johnson, 2013; Monge &

Contractor, 2003). In measuring the patterns of relationships at multiple levels there is a reliance on “graph realizations” where the number of actors in a network are graphed

48 with the number of possible ties with other members of the network (Contractor,

Wasserman & Faust, 2006, p. 685).

While Chapter Five presents the results of my study, it is worthwhile to highlight that the Nevada SOC network analysis initially produced 107 identified actors in the network. Thus, the number of graph realizations, or the possible number of ties all 107 actors could have with the other actors in the network, would be 11,342 (107x106) ties.

Adding to the complexity of this type of analysis, this study also measured ties through the use of dichotomous variables (the tie was present or not). This means the 11,342 possible number of ties in this study would actually exist in “one of two states” or 211,342

(Contractor, Wasserman & Faust, 2006, p. 685). Essentially, this is an infinite number of total possible ties.

Given the mathematical complexity of an infinite number of possible ties, network analysis software programs are expanding the reach of empirically supported network measures. The level of sophistication can range from descriptive analysis to more complex predictions of network behavior. As described above, measurement of the structural properties of networks can be multi-level: focusing on the actor level, relations between two actors (dyadic), small groups (triadic) and at the whole network level (Contractor, Wasserman & Faust, 2006).

Density. The density of a network is a whole network level measure of the total observed ties in relation to the total number of possible ties (Yang, Keller & Zheng,

2017). It describes the overall connectedness of the network members (Milward,

Provan, Fish, Isett & Huang, 2010; Provan, Veazie, Staten & Teufel-Shone, 2005; Yang,

Keller & Zheng, 2017). Density is the sum of all the network ties divided by the maximum possible number of direct ties (Klaster, Wilderom, & Muntslag, 2017, Yang,

Keller & Zheng, 2017). This produces an analysis of the network structure ranging from

“sparse,” where there are not many ties between nodes, to “dense” indicating a large number of ties (Contractor, Wasserman & Faust, 2006; Monge & Contractor, 2003).

High levels of density are not always indicators of network effectiveness as the increase in number of ties general requires increased costs in managing those ties (Milward et al., 2010). It may also influence an actor’s decision to participate in the network. For example, if it takes too many resources (i.e. time) for the actor to gain access to information, then this sets the context for their participation behavior.

Centrality. Centrality is an actor level and a whole network level measure (Yang,

Keller & Zheng, 2017). Within a network, central nodes are more influential in the flow of information, opportunities, constraints and resources important to policy implementation. Crucial to this centrality are characteristics of trust, reputation and influence (Provan, Huang & Milward, 2009). At the whole network level of analysis, overall centralization represents the distribution of ties amongst the actors in the network (Yang, Keller & Zheng, 2017). In other words, a highly centralized network would be represented by a small number of people with the greatest number of ties.

Network structures with high degree of centralization are more likely to achieve their collective goals (Contractor, Wasserman & Faust, 2006).

At the actor level of measurement, centrality is the actor’s unique position within the network and is an indication of power (Borgatti, Everett & Johnson, 2013). It

50 is a measure of connectivity among the actors within the network (Borgatti & Halgin,

2011; Contractor, Wasserman & Faust, 2006). Identification of actors who are more

“central” in the network can indicate who has increased power through their access to others. Additionally, central actors can inhibit or promote the spread of knowledge and innovation (Mulvaney, Lee, Hook & Prokopy, 2015). From a collective action theory perspective, a centralized network means there would be one or a small group of actors responsible for leading the collective action. The further an actor is from those central actors or the fewer ties an actor has to centralized actors (i.e. the outlier), the less likely they are to see value in their network participation or network outcomes. Borgatti

(2018a) describes four common actor level measures of centrality: degree, betweenness, closeness and eigenvector. While this study analyzed degree and betweenness centrality, the following briefly describes all four measures of centrality.

The “degree measure of centrality” is the number of direct ties an actor has with other actors in the network and is an indication of power and prestige within the network (Borgatti & Halgin, 2011; Contractor, Wasserman & Faust, 2006; Yank, Keller &

Zheng, 2017). This includes the number of ties the actor sends out to others as well as the number of ties coming in to the actor from others. This pattern of ties reveals how central the actor is within the network. In network communications, a large number of ties can increase communication and mutual trust (Mulvaney, Lee, Hook & Prokopy,

2015). When utilizing survey data, centrality is based on the respondents’ perception of how central a node is within the network and is measured by the number of direct or

51 indirect ties a network member has (Klaster, Wilderom, & Muntslag, 2017; Provan,

Veazie, Staten & Teufel-Shone, 2005).

The “betweenness centrality” measure captures the position of an actor as a bridge between others in the network; where they serve to connect network members who otherwise would not have been connected (Monge & Contractor, 2003; Mulvaney,

Lee, Hook & Prokopy, 2015; Kolpakov, Agranoff & McGuire, 2016). It is a measure of the frequency a given actor falls along the shortest path between other actors (Borgatti,

Everett & Johnson, 2013, p. 174). The “path” is the ties between the network actors.

For example, if Actor A needs to get information to Actor Z, the “path” is the shortest route the information must travel to reach Actor Z. Each actor along the route plays a central role in the travel of the information, which impacts their centrality to the network as a whole. An outlier or “isolate” would have a betweenness score of zero because they are not along the shortest path between the other network actors

(Borgatti, Everett & Johnson, 2013, p. 174). A higher “betweenness” score would indicate an actor is along the shortest path between the other nodes. For example, a person standing in the middle of the crowd would have easier access to the others in the crowd versus someone standing at the edge of the crowd or outside of the crowd.

Thus, the person in the center of the crowd would have more power because others would have to go through that person in order to get to the others because they are

“between” the actors.

The “closeness” measure of centrality as the extent to which a node is directly or indirectly tied to all the other nodes in the network (Monge & Contractor, 2003).

Understanding this score for an actor within a network can inform how efficiently an actor has access to the network such as access to information and innovation (Monge and Contractor, 2003; Yang, Keller & Zheng, 2017). A high closeness centrality score would indicate that an actor is has direct or “close” ties to all other members in the network (i.e. their path to all other network members is short). This measure is sensitive to network size as if it is a small highly connected network, then we would see high closeness scores (Yang, Keller & Zheng, 2017).

Eigenvector is the fourth construct of centrality and is essentially a measure of power by not just looking at the number of ties an actor has, but also who are they tied to (Borgatti, Everett & Johnson, 2013). It is a weighted centrality score of an actor according to the centrality of the others they are tied to. In other words, “a node is only as central as its network” (Borgatti, Everett & Johnson, 2013, p. 168). For example, an actor may have many ties, but if they are tied to other actors who don’t have many ties themselves, then the power can only travel so far. However, an actor who has few ties, but they are tied to actors who have many ties still impacts their centrality within the network by way of their connection to those who are more connected. Borgatti (2018a) describes the eigenvector in the sense of a popularity measure where there are those who are “in” and those who are out. Borgatti states, as an example, a kid may not have many friends, but if their one friend is popular, then they are popular.

Transitivity. At the triadic level of measurement, analyzing the structure of relationships amongst groups of three actors and the frequency those structures are present throughout the network reveals the network’s sub-structures and is an

53 indication of the whole network structure (Nooy, Mrvar & Batagelj, 2011). When a network has a high level of transitivity, “they tend to have a clumpy structure” containing “knots of nodes that are all interrelated” (Borgatti, Everett & Johnson, 2013, p. 155).

The frequency of certain types of triadic structures, or a triadic census, is the measure of transitivity in a network (Borgatti, Everett & Johnson, 2013; Nooy, Mrvar &

Batagelj, 2011; Wasserman & Faust, 1994). There are sixteen structural types of a triad which are depicted in Figure 2.2 on page 55 below (Borgatti, Everett & Johnson, 2013;

Nooy, Mrvar & Batagelj, 2011; Wasserman & Faust, 1994). The labeling scheme of the triad types, evident in Figure 2.2, is originally from Holland and Leinhardt (1970) and

Davis and Leinhardt (1972) who derived the labels according to the presence of “dyadic states” contained within each triad (as cited in Wasserman & Faust, 1994). Each dyadic state consists of the number of mutual relationships between each of the three actors, the number of asymmetric relationships and/or the number of null or lack of a tie between the actors. The dyadic states are also evident in Figure 2.2 below according to the arrangement and direction of the arrows connecting each node (i.e. mutual, asymmetric or null).

As the dyadic states can be mutual, asymmetric or null, analyzing the patterns of small sub-structures can reveal structural components of hierarchy within the network

(Contractor, Wasserman & Faust, 2006; Nooy, Mrvar and Batagelj, 2011; Kolpakov,

2012; Wasserman & Faust, 1994). For example, asymmetric ties between two actors

(represented by one arrow between two nodes in Figure 2.2 below) can be considered

“friends;” whereas, mutual dyads are can be considered “close friends” and is represented by two arrows going each direction between the two nodes (Holland &

Leinhardt, 1975 as cited in Wasserman & Faust, 1994). For example, in Figure 2.2 below, item 7-111D below has 1 mutual relationship between two nodes, 1 asymmetric relationship and no relationship (null) between two nodes. So, within the triad there would be two actors who are “close friends,” two actors who are “friends” and two actors who are not friends at all.

Counting the number and types of triads in a network allows researchers “to test the propensity of public management networks to exhibit signs of hierarchy, resource exchange, and stability” (Kolpakov, 2012, p. 178). The triadic census produces a report of the type and prevalence of triadic states depicted in Figure 2.2 below. Identifying the types of triads most prevalent in the network is a measure of the extent to which the network is balanced, hierarchical, transitive or clustered (Nooy, Mrvar, Batagelj, 2011).

As described above in theoretical explanations of network behavior, measuring transitivity is best conducted over time as a network evolves. Thus, the triadic census conducted in this study provides an inventory of the frequency of triadic types in the

Nevada SOC network for a single point in time.

Triadic Census All possible triads (Wasserman & Faust) Sequence numbers as used in Pajek 55

Figure 2.2: Sixteen Types of Triads

1 - 003 2 - 012 3 - 102 4 - 021D

5 - 021U 6 - 021C 7 - 111D 8 - 111U

9 - 030T 10 - 030C 11 - 201 12 - 120D

13 - 120U 14 - 120C 15 - 210 16 - 300

Figure Source: Institute for Mathematics, Physics and Mechanics. Retrieved from http://www.educa.fmf.uni-lj.si/datana/pub/networks/pajek/triade.pdf

Mutuality and reciprocation. At the dyadic level of measurement, mutuality and reciprocity measure the extent to which actors exchange information and resources and the extent to which the exchange is reciprocated (Contractor, Wasserman & Faust,

2006). While a dyadic level of analysis, this measure can also reveal elements of the global structure of the network as it can indicate levels of cohesion amongst the actors

56 in the network (Contractor, Wasserman & Faust, 2006; Hanneman & Riddle, 2011b).

The absence and presence of mutual and reciprocated ties can reveal clusters of relationships and expose patterns of the control or flow of information, opportunities, constraints and resources.

As discussed in this chapter, the governance approach utilized during the implementation of a public policy or program shapes incentives, opportunities and constraints thereby impacting the outcomes of the policy or program. The networked governance approach is a structure of governance that shifts the burden of implementation from a single state entity to a collaborative process with public and private entities. Measuring the extent to which a network is present in implementation and whether the structure of the network impacts outcomes is difficult. Network analysis is a methodology used to describe network structure at multiple levels of analysis. The following reports a study conducted to describe the implementation of the Nevada SOC from a networked governance perspective and to report the patterns of interactions that comprise the structure of the network. Chapter Three below describes the context of the study.

Chapter Three: Study Context

As cities, counties and states continue address complex social problems by engaging in interdependent partnerships spanning public and private sectors, it is apparent the structure of such partnerships would differ across geographic regions and policy arenas. As noted in the literature, network analysis and interpretation of findings is very context dependent (Goggin, 1986; Klaster, Wilderom & Muntslag, 2017). Thus, this chapter is focused on describing the context of the study. First, an overview of the evidence-based framework that is the focus of implementation is provided. Then, a description of the incremental process of developing a network involved with implementation in Nevada is presented.

The setting of this study is the state of Nevada divided into three regions common in public service provision for the state. The regions include the two most populated and urban counties in the state, Washoe and Clark, and the third region constitutes the remainder of the counties in the state. This third region is frequently referenced more generally as the “rural region.” While there is no designated children’s mental health authority, the Division of Child and Family Services (DCFS) within the

Nevada Department of Health and Human Services is one of public entities primarily responsible for children's mental health services.

As illustrated in the case of “Jenni” on page 3 of this report, children’s mental health is a “wicked problem” social issue in that each case manifests differently, requires varying resources and the problem can change over time. Children and their families need quick access to high quality assessment services and then an array of

58 evidence-based interventions tailored to the child’s unique needs. Families need to be actively included in the determination of services and there is no one formula for all children and families. Additionally, children and their families who have mental health needs typically interact with multiple systems such as child welfare, school systems and juvenile justice.

The three primary regions of the state (Washoe County, Clark County and rural region) influence the organizational structure of public services for children and families in the state. For example, the DCFS must work closely with child welfare agencies operated at the county level for Washoe and Clark but is responsible for direct child welfare services in the rural region. For juvenile justice, the DCFS must partner with county juvenile probation departments, but is responsible for state facilities for youth detention and youth parole. The DCFS is also a provider of children’s mental health services in the two urban counties of the state and coordinates with separate state entity for the provision of children’s mental health services in the rural regions of the state (Nevada Department of Health and Human Services, Division of Child and Family

Services, 2018). This setting presents a unique opportunity to study the network of individuals and organizations concerned with restructuring the systems associated with the provision of publicly funded behavioral health services for children, youth and their families or caregivers.

Stimulated with funding from the federal government’s Substance Abuse and

Mental Health Services Administration (SAMHSA), the state has been working toward the development and implementation of a statewide “System of Care” (SOC). The

59 following describes the SOC as an evidence-based framework for addressing the behavioral health needs of children and families. Then a summary of efforts, to date, in the creation and implementation of this framework throughout the state is provided.

System of Care

SOC is a guiding framework for communities and states in organizing behavioral health services for children and their families (Pires, 2002). A SOC is “a comprehensive spectrum of mental health and other necessary services which are organized into a coordinated network to meet the multiple and changing needs of severely emotionally disturbed children and adolescents and their families” (Stroul & Friedman, 1988b, p. 15).

The framework mandates the coordination and cooperation of providers, public entities, families and youth for the planning and implementation of services according to a defined set of SOC principles (see Appendix B for a list of the SOC values and principles).

Spearheaded by funding from SAMHSA, the framework has been implemented to some extent in “nearly all communities across the nation” (Stroul, Blau, & Friedman, 2010 as cited in Stroul & Friedman, 2011, p. 1).

Implementation of a statewide SOC aims to coordinate services in a culturally and linguistically appropriate placing youth and families at the center of decision making on their care as well as in the design of the system. A major premise of the framework is the child and family have access to a wide array of services in their home community or as close to home as possible (Stroul & Friedman, 1988a). The SOC framework mandates extensive coordination between public agencies and community-based private providers for both the planning and execution of services. In the case of “Jenni,”

60 she and her family would have one case plan and one case manager or service coordinator supporting them in navigating the other systems and accessing services as opposed to the multiple plans and case managers they would have if they were on their own. In turn, this extensive coordination results in reduced service fragmentation, increased efficiency in use of public funds, increased fidelity to evidence-based services and improved outcomes for children and families (Pires, 2002). Pires explains the SOC framework starts with the population of focus and then works with partners involved with the population. This standpoint upholds the value of youth and family centered approaches. It also localizes the efforts geographically. The result is a series of subsystems involved in the SOC network ranging from local providers, school districts and public entities.

Federal programs in the United States are increasingly mandating cross collaboration amongst public entities (O’Toole, 2014). As a federal agency, the

Substance Abuse and Mental Health Services Administration (SAMHSA) spearheaded the SOC framework through their Children’s Mental Health Initiative. Under this SOC framework, SAMHSA is funding states to establish, update or align their systems in order to support implementation of collaborative services. They have sponsored SOC planning, implementation and expansion grants since 1993 (SAMHSA, 2015).

Nevada System of Care

Nevada has been working to develop a SOC since 1998 when the state received a SAMHSA grant to establish some of the core supports necessary for a SOC such as the

“wraparound” case management approach and family peer support services (State of

Nevada, 2016). In 2016, the state received a grant from the SAMHSA to implement the

SOC in the state. In doing so, the state tapped in to existing public commissions and sub-groups to serve as the “network” for implementation. The state DCFS is the recipient of the grant and serves as the lead organization in the network.

The Nevada SOC network features a combination of public and private organizations who are concerned with improving access to and outcomes of publicly funded children’s behavioral health services within the state. The network also includes parent and family stakeholders who represent the impacted children and families. The formation of the network is the result of an incremental process stimulated by needs, laws, special commissions and funding awards. The following provides a summary of those events.

In 1985, the Nevada state legislature created the “Commission on Mental Health and Mental Retardation” to operate within what is currently known as the state’s

Department of Health and Human Services, Division of Public and Behavioral Health

(NRS, Chapter 672, 1985). In 2009, the legislature added a special subcommittee on children’s mental health to the commission (NRS, Chapter 181) and in 2013, the legislature changed the name to the “Commission on Behavioral Health and

Developmental Services.” The Commission is charged with providing “policy guidance and oversight of Nevada’s public system of integrated care and treatment of adults and children with mental health, substance abuse and developmental disabilities/related conditions” (State of Nevada, 2013, p. 1).

In 2013, Nevada Governor Brian Sandoval established, by Executive Order, the

“Governor’s Council on Behavioral Health and Wellness.” He charged the Council with creating a plan and making recommendations on improving the delivery of behavioral health services in the state. In his Executive Order, the Governor establishes the need to examine the relationships amongst the providers and important stakeholders such as law enforcement and to “facilitate cooperation” amongst the providers to improve the provision of behavioral health services in the state (Nevada Executive Order No. 2013-

26, 2013). According to posting of public meeting agendas and minutes, the Council met during the calendar year of 2014 and does not appear to have gathered since that time (http://dpbh.nv.gov/Boards/BHWC/Meetings/Meetings/).

During the 2017 Nevada state legislative session, the state passed Assembly Bill

366 establishing “behavioral health policy boards” in four specific regions of the state

(State of Nevada, DPBH, 2018). The boards are to advise the state Department of Public and Behavioral Health on the behavioral health needs of the region and to propose plans to improve services. As of this writing, the specific connection from the boards to the Nevada SOC is unknown as the boards are relatively new. However, there have been initial communications to initiate a collaboration between the boards and the

Nevada SOC. Appendix C provides an abbreviated timeline of the events described above.

At the time of this study, the Nevada SOC is funded by a SOC implementation grant from SAMHSA. Under the grant, the DCFS partnered with multiple public and private sector service providers, youth and families as key stakeholders. Utilizing

63 existing collaborative structures (see next section), the DCFS developed a Strategic Plan and a Communication Plan to guide implementation efforts (State of Nevada, DCFS,

2018). The goals of the grant target the two most populated counties in the state with some services extending to the remaining rural counties of the state.

The Strategic Plan includes plans to redesign the structure of publicly funded children’s mental health services in the state. An element of the proposed redesign includes a shift in the provision of services from the state to external non-governmental organizational partners (State of Nevada, DCFS, 2016). The ongoing partnership and execution of the Strategic Plan and Communication Plan is carried out through a series of public collaborative groups (see next section). Thus, the “event” within this study is the set of workgroups and committees that focus on the implementation of a System of

Care within Nevada.

Nevada System of Care as a Network

Agranoff (2012) describes sixteen criteria for identifying when it is the right time to form a network. The Nevada SOC meets all of those criteria. The following lists a sampling of those criteria (p. 134):

• Problems are complex, and potential solutions lie with many public agencies,

organizations, programs, and services. No single agency can approach the

problem;

• Top administrators and decision makers recognize the complexity of the

problem and are willing to lend their resources;

• Potential partners see the opportunity to learn and adapt, and to develop new

competencies and new approaches to problems;

• Potential ability to manage uncertainty, solve complex problems;

• Those who control/participate are willing to engage in constructive

deliberative dialogue/engagement leading to the development of new

possibilities; and

• An initial willingness to align services and programs impacting others.

Thus, the timing is right to describe the Nevada SOC efforts as a networked approached to implementing the SOC framework in the state.

Figure 3.1 below describes some of the basic components of the Nevada SOC as a network. Utilizing Agranoff’s (2007) typology and description of public management, the Nevada SOC is an “action network.” The network follows the “lead organization model” of networks and carries out most of the network business during public meetings in accordance with state open meeting laws (Provan & Kenis, 2008). The lead coordinating entity for the Nevada SOC is the state’s Division of Child and Family

Services (DCFS), which is one of six divisions housed under the Nevada Department of

Health and Human Services (DHHS). At the time of this writing, the DCFS has seven programs in operation (DCFS, 2018). The implementation and management of the programs are managed by seven Deputy Administrators who all report to a single

Division Administrator who reports to the Director of the DHHS. One of those programs is children’s mental health. Under this program, the state provides direct services to

65 children and families in both inpatient and outpatient modalities. The state also coordinates community-based contracted services for children’s mental health (DCFS,

2018).

Figure 3.1 Nevada SOC as a Network

Network Type Purpose Enabling Authority Primary Agencies (Agranoff, 2007)

•Planning and •Action Network •State government •Nevada Division implementation •Reviews, •Nevada System of of Child and of system develops and Care Children's Family Services alignment recommends Behavioral Health •Nevada PEP (as strategies to policies that Sub-Committee formal grant increase access impact service partner for and quality of delivery and youthe and family services for monitoring. voice) children's mental •Nevada Division and behavioral of Health Care health. Finance and Policy (Medicaid) •County-based child serving agencies •Non- governmental child serving agencies

The Nevada SOC is embedded within a larger system of other networks such as the regional consortia and the behavioral health policy boards (as discussed above and presented in Appendix C, Abbreviated Timeline). Figure 3.2 below depicts the other networks. While they are not distinguished as separate entities for this study, it is important to note the layered effect of the subnetworks on the System of Care network.

There is a need for future studies to delve deeper in to the, regional consortia, for example as distinct structural components of the larger System of Care network.

Figure 3.2 Nevada SOC within a Larger Network

SOC Special Populations Workgroup

Commission on Behavioral SOC Health, Children's System of SOC Workforce Communications Care (SOC) Behavioral Development Workgroup Health Subcommittee Workgroup (NRS 232.482)

State of Nevada Rural Children's DPBH Commission on Mental Washoe Behavioral Health and Health RBHPB Developmental Services Consortium (NRS 433.047)

Regional Statewide Behavioral Rural Consortium Health Policy RBHPB on Children's Boards (RBHPB) Mental Health AB366 Clark Children's Mental Health Northern Washoe Southern Consortium RBHPB Children's RBHPB (Frontier) Mental Health Consortium

Chapter Four: Methodology

This descriptive study is an examination of the Nevada SOC network. It aims to describe the structure of the network and to attribute the structure to network effectiveness as perceived by the respondents. The guiding research questions are:

1. Who are the network actors in the implementation of the Nevada System of Care?

2. What are the patterns of interactions within and amongst these actors?

3. To what extent do these patterns of interaction contribute to the perceived

outcomes of the policy?

Essentially, this is a study of “relational properties” of a statewide network as it designs and implements a SOC for the state (Contractor, Wasserman & Faust, 2006, p.

685). Support for the study was obtained by the agency administrator, deputy administrator and the network manager through informal oral presentations of the concept and purpose.

Research Design

To answer the research questions, this study employs a mixed method approach.

Phase one of the study is a secondary analysis of public meeting minutes for two purposes. First, the analysis is used to develop and an affiliation network matrix.

Second, the affiliation analysis generated a network member name roster for use in the second phase of the study. Phase two of the study is the collection of primary data through the development and distribution of an electronic survey completed by network members. The phases are discussed further below.

This study is a whole-network study, which is the study of the patterns of relationships amongst all the members of the network. A whole-network level of analysis is common in public administration studies for describing power or position of network members as well as describing structural characteristics of the network

(Kapucu, Hu & Khosa, 2017). In this study, the boundary of the network is also a policy domain, children’s mental health in the state of Nevada. Knoke (2011) states “the twin concepts of policy network and policy domain can be reconciled by recognizing that a policy domain delineates a bounded system whose members are interconnected by multiple policy networks” (pg. 211). The defined boundaries of the network include the state of Nevada with a focus on including individuals and organizations who are concerned with publicly funded children’s mental health services and have attended public meetings related to the implementation of the SOC in Nevada (see Chapter

Three, “Study Context”). As a network study of a single policy context, this study is descriptive and exploratory in nature.

Phase One, Purpose One: Affiliation Network Matrix

An affiliation network description is used to answer research question one,

“Who are the network actors in the implementation of the Nevada System of Care?” To describe and represent an affiliation network, Wassermann and Faust (1994) recommend first developing an “affiliation network matrix.” This matrix maps ties between network members through their mutual attendance at selected events. To develop the affiliation matrix, actors and events are coded in a “g x h” matrix where the

69 actor (g) is the row and the events(h) are the columns (Wassermann & Faust, 1994).

This matrix is also described as a sociomatrix and records each actor as they are affiliated with specific events (Ward, Stovel & Sacks, 2011).

To collect this data, a list of events associated with the Nevada SOC is obtained from a publicly available list of meetings from the State of Nevada, Department of

Health and Human Services, Division of Child and Family Services website for the 2017 calendar year. This list is comprised of publicly identified committees, workgroups and consortia associated with children’s mental and/or behavioral health as a topic of focus for the network. The list was narrowed to public meetings specifically associated with the Nevada System of Care. Meeting Minutes were obtained for each of the identified events from publicly available sources. This includes hard copies of minutes obtained from my attendance at meetings and attachments included in electronic distributions of meeting minutes. The minutes were organized by the month and year of the event.

Meeting minutes that did not contain the names of members present or absent are not included in the analysis. See Appendix D (Affiliation Network Development) for a list and brief description of the events included in the development of the affiliation network matrix.

During calendar year 2017, there were 38 identified public meetings related to the Nevada System of Care. Those meetings consisted of the following groups and sub- groups:

1. Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee

2. Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee - Communications Workgroup

3. Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee - Workforce Development Workgroup

4. Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee - Special Populations Workgroup

5. Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee - Provider Standards and Evidence-Based Practices Workgroup

6. Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee - Governance Workgroup

As is common in the networked governance approach, the network evolves over time.

Therefore, it is noted that the “Provider Standards and Evidence-Based Practices” workgroup and the “Governance” workgroups (items 5 and 6 above) were modified and/or subsumed under other workgroups. Therefore, they discontinued meeting as distinct groups on or around February 2017.

The analysis of this data is considered a “two-mode” network analysis as the data can only infer a tie by mutual attendance at a meeting rather than asking the network members directly about their ties (Borgatti, Everett & Johnson, 2013). In this study, the actors are the first mode and the events they attend are the second mode.

To collect this data and develop the affiliation matrix, each set of meeting minutes is reviewed for the inclusion of the name of an actor. In the “g x h” matrix described above, the actor is coded with the number one if they were affiliated with the event. In

71 this study, affiliation is defined as being named in the meeting minutes as having attended the meeting. An actor who is not affiliated with a specific event is coded as a zero for the event.

While data analysis for this study focused on findings from the survey (discussed further below), future research could further analyze the affiliation data. In doing so, the next step would be to utilize the UCINET software program to translate the two- mode matrix into a new one-mode matrix. This matrix contains all the events each actor co-attended with another actor. From there, it is possible to employ network analysis techniques to this data set of co-attendance at meetings. As discussed next, the affiliation data were primarily used to identify actors in the Nevada SOC network.

Phase One, Purpose Two: Name Generation for Survey

The development of the affiliation matrix produces a database of the events, names of actors involved with those events and the number of times each actor is affiliated with the events. From this database, a roster of network members is generated to identify the sample for the survey. For this study, 107 actors were identified in the affiliation matrix and all were initially included in Phase Two of the study, the survey.

Phase Two: Survey

The Nevada SOC Network Analysis survey is used to answer research question two, “What are the patterns of interactions within and amongst these actors?” and research question three “To what extent do these patterns of interaction contribute to

72 the perceived outcomes of the policy?” Since this phase of the study involves human subjects, an application for exempt status was submitted to the University of Nevada

Reno, Research Integrity office for review according to human subject research ethical guidelines. The determination from the Research Integrity Office is that the study is exempt from human subjects review.

Survey Design

Sample. The survey sample is a census sample which includes adults, age 18 or older who were listed in public meeting minutes during the calendar year 2017 as having attended at least one SOC meeting (as described in phase one of the study above). Of the 107 individuals identified in phase one of the study, the survey was distributed to 96 individuals. Incorrect or unavailable electronic mail addresses account for the difference between the original defined network and the distributed surveys. This sample includes parents of children with behavioral health needs, behavioral health providers, youth representatives (who are 18 or older), state employees who are employed by divisions associated with the Nevada SOC (i.e. the

Nevada Department of Health and Human Services, Division of Health Care Finance and

Policy, Division of Child and Family Services) and state employees who are staff and administrators associated with implementation of the Nevada SOC.

All survey participants respond in their individual capacity as opposed to on behalf of their organization. Thus, the unit of observation is the actor themselves. This means the data gathered in phase two of the study comes directly from the actor while

73 the unit of analysis remains the whole network (Wasserman & Faust, 1994). The study did not employ a sampling technique beyond the census as the focus was to obtain measurement of actors contained within the network.

Survey content. The survey consists of a consent page and four measures: basic characteristics, structural variables (closed ended), structural variables (open ended) and perceived effectiveness. SurveyMonkey is the electronic survey platform utilized to develop, test, distribute, collect and perform some basic analysis of the survey data.

The SurveyMonkey platform offers certain design features to increase likelihood of respondent completion. As a service, SurveyMonkey offers an automated “genius” to score the survey design, estimate completion rates (not response rates) and estimate time for completion. The “genius” results for this survey are: “perfect” for overall survey design, 71% estimated completion rate and estimated 10-minute response time to complete the survey. The specific survey elements are discussed further below. See

Appendix E for a modified copy of the survey (the roster of names was removed from the survey copy).

Table 4.1 below describes the threats to validity taken in to consideration in the design of the survey.

Table 4.1

Threats to Validity Considered in Survey Design

Threat to Validity Design Recommendation Sample Survey Question

Data collection and Do not ask about a specific Who are the people you retrospective errors: point in time. Rather, ask a have interacted with the Respondents “are more question that prompts the most about the accurate in reporting long- respondent to think about a implementation of the term patterns of behaviors pattern over time. Nevada System of Care? than behaviors at some point in time” (pg. 38)

Omission errors: forgetting Reduce cognitive demand The following list of to include some members by including a close-ended individuals was developed of the network. question that contains a from public records of roster of individuals already meetings and other Nevada known to be within the System of Care documents. network. From the list below, please select the people with whom you are directly or indirectly involved as part of working on the implementation of the Nevada System of Care.

Non-response bias: Ethnographic background The survey was tested with sensitivity of relational research to determine types two respondents who questions of questions that might be provided feedback on the appropriate. Pre-test the length of time it took to questions with potential take the survey, specific respondents. questions or response options they didn’t understand and their reaction to sensitive questions such as the trust question. *developed according to recommendations from Borgatti, Everett, and Johnson (2013)

Consent. Participants enter the survey via a link included in the electronic mail invitation. Once the participant clicks on the link, they are taken to the first page of the survey, which is a consent page. The consent page is an abbreviated version of the electronic mail message that contained the full consent information and general survey instructions. Since the data generated in phase one of the study (the affiliation network matrix) is also used in the final description of the network, participants are informed in the information sheets that non-participation in the survey does not necessarily mean they excluded from the study.

After reviewing the consent information and survey instructions, participants are prompted to “agree” or “disagree” to participate in the study. If the participant selects both “agree” and the “next” option, they are taken to the survey. If the participant selects both the “disagree” and “next” option, they are taken to the end of the survey with no opportunity to complete the survey.

Confidentiality. The nature of network analysis is to collect data from specific, named members of the network and to map the pattern of relationships amongst those actors. The dataset contains personally identifiable information from participants by requesting their name on the survey. Thus, respondent names are included in the data collection and, therefore, are not anonymous. However, the following precautions are implemented to protect the confidentiality of the respondents.

Participant names remain within the original survey results and the network analysis data set. However, data entered into the network analysis software program are assigned a respondent code. Only the codes are reported in the final analysis.

Additionally, data are reported in aggregate form so individual responses to survey questions remain confidential. Through the information sheet, participants are notified the findings will be reported as a network diagram depicting who is connected to whom.

Although specific names are not included in the diagram, participants are informed it may be possible for a participant to identify or suppose themselves or others in the diagram. In this case, respondents are informed they may be able to identify themselves in the final results and it may be uncomfortable for the respondent to view their position within the network.

Survey Testing. The survey was tested with two prospective respondents. They initially described the trust question within a matrix format as too sensitive and it created a feeling of mistrust and fear with the research results. The change in format to the “advanced branching logic” reportedly alleviated the fear. However, the survey testers reported some concern or discomfort with being asked to provide their names within the survey. The testers also identified items they felt did not have a response option applicable to them, items that were confusing to them and general errors such as typos. Additionally, the survey content was reviewed with the agency Administrator to solidify support for the study.

Survey Distribution and Response Collection. A risk in any research design that relies on surveys responses is the actual response rate. In traditional studies, there is a dependence on high response rates to produce the statistical power needed for accurate results. While true to a certain extent for network analysis (particularly in instances when the researcher wants to report the predictive value of the network

77 structure), it is most important to seek responses from as many network members defined in the “whole network” as possible. This is a slight departure from a mere focus on the number of respondents.

A strategy to reduce non-response is to develop rapport with the potential responders prior to administering the survey (Johnson, 1990 as cited in Borgatti,

Everett, and Johnson, 2013). Borgatti, Everett and Johnson (2013) emphasize rapport building by stating “we advocate making as much contact with potential respondents as possible” (pg. 55). For this study, my rapport with the network members was established through my prior relationship with the SOC. I had attended meetings and provided facilitation services during SOC meetings. Thus, potential respondents are already familiar with me.

The survey was initially distributed via a direct electronic mail function available within the SurveyMonkey platform. This function requires direct entry of electronic mail addresses which were gathered from publicly available sources such as electronic mail distributions of meeting announcements and general Internet searches for individuals.

Once the electronic mail addresses are entered, SurveyMonkey generates the electronic mail message to each participant. The messages are sent to the participants individually and it is not possible for the participants to see who received the survey announcement.

This feature provides a tracking mechanism for monitoring responses. This method initially produced a 28% response rate. From there, I sent individual electronic mail messages directly to non-respondents. The message contained a reminder to complete the survey, a link to the survey and an invitation to ask questions about the survey if

78 necessary. This direct link to the survey via personal invitation method produced an additional 35% response rate. Further discussion of the survey response rate is presented in Chapter Five (Results).

Survey Items

Basic Characteristics (Attributes). Certain variables such as demographics, roles, positions in organizations or geographic location are basic characteristics (Wasserman &

Faust, 1994) or attributes (Mischen & Jackson, 2008) that can help explain why a tie may or may not exist. For this study, the following attribute variables are included in this measure:

1. First and last name

2. Role/perspective of involvement in the Nevada System of Care (i.e. parent,

professional, etc.)

3. Type of organization the individual represents

4. Geographic region the individual represents

5. Length of time involved

6. Affiliation with other workgroups and committees connected to the Nevada SOC

The SurveyMonkey platform allows for certain questions to be designated as requiring an answer. Thus, the first and last name questions were designated as such.

Participants who attempted to skip this question received the following prompt “In order to accurately map the "network" for the Nevada System of Care, your name is needed. Your responses will be treated confidentially and your name will not be

79 included in any final reports.” This allowed for more accurate tracking of network members’ completion of the survey.

Structural Variables (closed ended). Structural variables (Wasserman & Faust,

1994) describe the structure of the network, revealing the patterns of relationships amongst the actors. The survey design in this study collected data on the structural variables utilizing two strategies: roster format and free recall (Borgatti, 2008). The roster format is a close-ended list of names of known network actors generated from public records. This roster is developed by obtaining names of individuals from public ally available meeting agendas and minutes (see “Phase One” above). Within this roster format, respondents are asked a set of questions that require them to review the roster of names and respond to the structural variable questions for each name on the list. The survey design collects binary data on the following relationship variables:

1. Direct or indirect involvement with the identified network member

2. Sharing information with the network member

3. Sharing resources with the network member

4. Presence of a trusting relationship with the network member

Respondents report the presence of a tie with each name on the roster by selecting “all that apply” within the context of the survey item. To decrease respondent fatigue, the

“advanced branching logic” feature available within the SurveyMonkey platform is utilized. Under this survey logic, participants are provided with a full list of the 107 names developed during phase one of the study. If a list of names is long, it can be a barrier for participation as survey respondents have to review the entire list for each

80 relationship question. Thus, the “advanced branching logic” allows for participants to initially select the names of individuals they work closely with. From there, the survey platform automatically narrows the response options on the remaining questions to only those selected individuals as opposed to the whole list.

Structural Variables (open ended). The free recall format is employed to allow respondents to recall whomever they choose in response to prompts. The questions included in this format aim to measure who respondents have frequent contact with, who they view as influential (Mischen & Jackson, 2008) and who they go to for information. Within this strategy, the respondent is asked to identify who they go to if they: want to get something done or improved, get a true reading of what is happening with the Nevada SOC and who they have interacted with the most. Responses are then entered in to the matrix by creating a column for each of the items (i.e. frequent contact, influential, etc.). The rows in the matrix consist of the respondent. A tie is indicated by presence (one) or absence (zero) of the relationship according to their free recall responses.

Perceived Effectiveness. Provan and Milward (1995) describe difficulty in assessing network effectiveness. They highlight the work of Zammuto (1984) in recommending consideration of the stakeholder’s own views of network effectiveness

(as cited in Provan & Milward, 1995). In later work, Provan, Veazie, Staten, & Teufel-

Schone (2005) recommend including items on a survey that assess the respondents’ expectations of the network. Network expectation was collected from one item in the

81 survey asking respondents to rate how effective or ineffective they feel the implementation of the Nevada SOC has been.

Data Analysis Plan

Research Question One

Two sources of data are utilized to answer research question one “Who are the network actors in the implementation of the Nevada System of Care?” First, public meeting minutes were reviewed to collect who attended each meeting and to develop an affiliation network matrix. Second, frequencies and crosstabulation reports of the survey data are utilized to describe the attributes of the actors in the network.

To create the affiliation matrix, data are entered in an Excel spreadsheet consisting of the actors (rows) and events (columns). The “who” in this case is the actor and the “meeting” is the event. For each actor the number one is entered for each event they participate in and a zero is entered for events they did not participate in. Actor names, frequencies and averages are calculated from this dataset. Matrices are developed from the data set and imported in to the UCINET software program for analysis. First, a two-mode analysis of actors-by-events is run (see results for depictions of the network). This produces a two-mode incidence matrix (Borgatti & Halgin, 2011).

To analyze the network structure properties of the affiliation matrix, the rate of participation is calculated within Excel for the number of events each actor is affiliated.

The results are calculated by the row totals for each actor in the affiliation network, indicating the number of times the actor attended events throughout the time period.

The size of each event is calculated by the column totals for each event in the affiliation network, indicating the number of participants at each event. A mean of the events is calculated by taking the event totals for each column and dividing them by the total number of events. The calculations are derived from Wasserman and Faust (1994).

Additionally, the average number of meetings attended by survey respondents vs. non- respondents is calculated from this affiliation network data.

Research Question Two

To answer the question “What are the patterns of interactions within and amongst these actors?” survey data is analyzed at multiple levels and across multiple network relation types. In order to prepare the data for analysis, it must first be converted to matrices. Hanneman and Riddle (2011) explain “representing data as matrices is the basis of manipulating data and calculating the measures” (p. 332). The matrix is necessary for describing the relationships between actors. The following describes the steps taken to download the survey data from SurveyMonkey, clean the data and prepare the data for analysis by creating matrices. This chapter concludes with a basic discussion of the data analysis strategies employed to answer the research questions, while Chapter Five will discuss the findings of the analysis.

Data Download and Preparation. The following details the steps employed to extract the survey response data from SurveyMonkey, utilize Excel to clean and transform the data and then import the data into UCINET for further analysis.

1. Data download: All survey data is downloaded directly from SurveyMonkey to

Excel. The data were downloaded with the actual answer text within the Excel

sheet cells. All data was included in one Excel sheet.

2. Data cleaning: Surveys where the respondent marked as “disagree” for their

consent to participate were removed from the data set. For some surveys, the

respondents marked “agree,” but did not respond to any items in the survey.

Those responses were removed as well. Finally, one respondent had two survey

entries, so one of the entries was removed as well. At this point in the study, it

was discovered that there were some actors included in the survey roster items

that were duplicates. For example, an individual’s formal name was used as one

selection and the same individual was included as another response selection

using a shortened name. Thus, the duplicate selection was removed as it had

not been selected by any respondents.

3. A code list is used to de-identify the respondent data prior to analyzing the data

within the Network Analysis software program. Each actor is assigned a unique

code, which is used for the final analysis and reports.

4. Data transformation (Borgatti, 2016): Since the data download included the

actual answer text within the cells, the responses had to be transformed in to

binary data. Utilizing the “find and replace” function in Excel, respondent names

were transformed to the number one. Empty cells, where respondents did not

select the named actor, were filled with a zero. For the open-ended questions

where respondents can fill in an actor’s name, a code was assigned to the actor.

If it was an actor already included in the data set, then the same code was

entered. All other categorical response data were transformed to numbered

responses utilizing a code book.

5. “Complete Response Matrix”: A square matrix is developed according to survey

respondents in both the rows and columns. This matrix was created by

removing non-respondent actors from the columns so that it matched the

corresponding rows of the matrix. The data are transposed so the rows

(respondents) match the columns (the other actors to which the respondent is

tied to). The complete square matrix for the respondent only data is 51 x 51.

The matrix is asymmetrical as the ties reported in the matrix are not necessarily

reciprocated although they can be. For example, one respondent may report the

presence of a trusting relationship with another network member while that

network member may not reciprocate the presence of the trusting relationship.

In a symmetrical matrix, the relationships would be reciprocated. Four sub-

matrices are extrapolated from the respondent matrix based on the relationship

variables: involvement with the named actor, information sharing, resource

sharing and trust.

6. UCINET Import: The matrices described above were individually imported into

UCINET (2002) by copying the data set directly from Excel and pasting it to the

“Matrix Editor” in the UCINET system. Each new file was saved individually.

The matrices developed from item five above produced two-dimensional matrices. In order to conduct multi-level analyses, the layers or “slices” of data need to

85 be extrapolated (Borgatti, Everett & Johnson, 2013, p. 73). As this study collected data on four relationship items (involvement with the actor during implementation, sharing information, sharing resources and trust), a matrix is developed for each type of tie.

Four separate relationship matrices were extrapolated from the dataset (item 5 above).

Visualizations. Visual depictions of networks are useful in telling the story of the data. To visually depict the patterns of interactions within the Nevada SOC, UCINET’s

NetDraw program was utilized (Borgatti, 2002). Visual depictions of the network according to the relationship variables of operations, sharing information, sharing resources and trust are included in this analysis. The following steps are employed to develop the visual depictions:

1. Within NetDraw, the relationship variable is selected, and the program develops the

initial depiction according to the ties indicated in the matrix. This depiction is useful

for spotting initial trends, but the visual depiction does not provide any indication of

the actor’s location in the network.

2. In order to gain a deeper understanding of the visualization, attributes can be

embedded in to the data. In a sense, it is like “layering” the attributes on top of the

tie data. Built-in functions such as color coding and adding shapes allows for

customization of the depiction and tailoring of the representation of the data.

3. The analysis function within NetDraw is then employed to run a degree centrality

analysis, which identifies the nodes most central within the network. Further

customization of the depiction allows for setting the visual representation of the

node size to correspond with the centrality measures.

4. Further manipulations are employed to gain a better understanding of the data. For

example, attribute-based scatter plots are useful to visualize the relationships of the

actors according to certain attributes such as whom they represent when they

participate in the Nevada SOC or their geographic location (Borgatti, Everett &

Johnson, 2013). NetDraw is then used to produce visual depictions that represent

multi-dimensional scaling of the data utilized the Gower procedure. Chapter Five

(Results) presents visual depictions of the network from various options

perspectives.

Statistical tests. Multi-level analyses such as those described in Chapter Two above were employed to answer the research questions. Excel, UCINET (with NetDraw) and Pajek are the primary data analysis tools utilized to analyze the network data for phase one and two of the study. A combination of whole network, triadic, dyadic and actor level analyses are employed to describe the patterns of interactions amongst network actors (Borgatti, 2018a,b; Kolpakov, 2012).

UCINET is utilized to conduct the actor level and whole network analysis including density, degree centrality, betweenness centrality and reciprocity. The analyses describe structural characteristics such as cohesion or connectedness. Actor level analysis of degree centrality measures the position or power of the individual actors within the network, while betweenness centrality measures how a node influences the flow of information (Kapucu, Hu and Khosa, 2017; Kolpakov, Agranoff &

McGuire, 2016). A program within UCINET, NetDraw, conducts the same degree centrality analysis but produces the network outcomes as a visual representation of

87 nodes (actors) and their corresponding ties. In such depictions, the size of the node represents the centrality of the actor.

Another network analysis software program, Pajek, is utilized at the triadic level to analyze the patterns of ties amongst groups of three nodes. As discussed in Chapter

Two above, analyzing a network’s sub-structures in by groups of three is a strategy to understand the overall structure of the network (Nooy, Mrvar & Batagelj, 2011). Pajek conducts a triadic census, which counts the presence of each of the sixteen types of triads presented in Figure 2.2 above (Nooy, Mrvar & Batagelj, 2011; Mrvar & Batagelj,

2016). The triadic census results in a report of the actual frequency of each type of triad, the frequency that the triad would be expected by chance and the relative difference between the actual versus chance frequencies (Nooy, Mrvar & Batagelj,

2011). From this report, I am able to determine which types of triads are most prevalent in the Nevada SOC network which then allows me to determine the characterization of the overall network structure (i.e. hierarchical, clustered, transitive, etc.).

At the dyadic level of analysis, UCINET is utilized to conduct the Quadradic

Assignment Procedure (QAP) Correlation and Regression (MRQAP) procedures describe and predict the occurrence of ties between two nodes (Dekker, Krackhardt & Snijders,

2007). By the very nature of a network, network data is not random or independent. In fact, “each node’s action not only affects its relations with other nodes in the network, but also exerts ripple effects to other nodes in the same network” (Yang, Keller & Zheng,

2017, p. 88). Thus, in order to simulate randomization, the QAP and MRQAP procedures

88 conducted in UCINET utilize a permutation approach to compare the observed network with randomly computed graphs to test for significance (Borgatti, Everett & Johnson,

2013; Yang, Keller & Zheng, 2017).

Data Preparation

In order to further prepare the data for the above analyses, additional attribute matrices are extrapolated from the data set to use for the correlation and regression procedures. The attribute matrices indicate ties between nodes that share the same designated attribute. Ten separate matrices were initially extrapolated for each of the following attributes:

a. Same effectiveness: Respondents selected the same choice on the effectiveness

item.

b. Same whom: Respondents selected the same role in identifying whom they

represent when they participate in the Nevada SOC such as a parent, clinician or

professional staff.

c. Same organization: Respondents selected the same organizational type such is

public or private, non-profit.

d. Same time: Participants selected the same option indicating the length of time

they have been involved in the Nevada SOC.

e. Same geographic region: Respondents selected the same option of either Clark

or Washoe counties, rural Nevada or a statewide option.

Research Question Three

To answer Question Three, “To what extent do these patterns of interaction contribute to the perceived outcomes of the policy?” the SurveyMonkey results are analyzed for basic response frequencies. A crosstabulation report run directly from the

SurveyMonkey platform is used to describe the respondent’s perception of effectiveness by their role in the network (i.e. parent, caregiver, staff, etc.) and also by the length of time they have been involved in the SOC.

For the network analysis of the perception of effectiveness, an attribute matrix is created for the “perceived effectiveness” item. To create the attribute matrix, the survey response data is downloaded from the software program is converted from the actual answer text to the response codes recorded in the code book. The data set is then imported to UCINET and converted to an attribute matrix. Utilizing UCINET’s

NetDraw program the attribute matrix is embedded in to a visualization of the network to examine the patterns of perceived effectiveness within the complete response network structure.

Chapter Five: Results

The results of this study provide a point-in-time description of the overall structure of the Nevada SOC network. The following results are reported according to the Phase One and Phase Two research design described above and the following guiding research questions:

1. Who are the network actors in the implementation of the Nevada System of

Care?

2. What are the patterns of interactions within and amongst these actors?

3. To what extent do these patterns of interaction contribute to the perceived

outcomes of the policy?

Utilizing the research questions as a guide, the following results report who is in the

Nevada SOC network with a description of some of their basic attributes (question one).

The patterns of interaction are reported as the “structure” of the network (question two). The structural results are presented according to the different levels of analysis: whole network, triadic level, dyadic level and actor level. Visual depictions of the network are presented in support of the whole network level results and results of the actors’ perception of effectiveness (question three).

Research Question One

To describe the actors involved with the implementation of the Nevada SOC, the following reports the findings of the initial affiliation network development (phase one of the research design). Additionally, characteristics or “attributes” of the network

91 actors are reported (phase two of the research design). It is important to note the findings only represent one calendar year for the affiliation network and a point in time for the survey respondents. Given the fluid nature of network structures, a fluctuation in the numbers would be expected at any other point in time.

Network size. The size of the network as defined by the number of actors who attended at least one Nevada SOC meeting during the calendar year of 2017. There were 35 meetings included in this analysis with 107 unique attendees. The actors

(attendees) and the meetings they attended are recorded as the affiliation network. Of this initial group, the average number of times individuals attended a meeting is seven.

The range for meeting attendance was 1-33 with the mode being one meeting and the median is four meetings. The average total attendance at the meetings according to the minutes reviewed was 21 individuals. Of the 107 network actors identified: 31% attended one meeting, 44% attended 2-10 meetings, 16% attended 10-20 meetings and

9% attended more than 20 meetings. The average number of meetings attended by survey respondents was ten and the average number of meetings attended by survey non-respondents was five. This is an indication that the survey respondents may be more familiar with the Nevada SOC than the survey non-respondents given their higher average attendance at the meetings.

Network member attributes. Invitations to complete the survey were sent to 96 of the 107 network members. The difference of 11 invitations is due to “bounced back” invitations from invalid electronic mail addresses or inability to locate an electronic mail address for the network actor. Given the sample size of 96 survey invitations, 61

92 surveys were returned (64% response rate). Three respondents did not agree to participate in the study and six respondents did not respond to the questions after they entered the survey. One survey was completed two times by the participant. Thus, of the 61 opened surveys, 51 were usable for the study (53% response rate).

The survey respondents were primarily professional staff who consider themselves as representing the whole state when they participate in the Nevada SOC.

The following figures present the frequencies and crosstabulations of attribute data describing the survey respondents. A comparison of the attribute data with survey non- respondents is also included in some of the figures.

Figure 5.1 Role Representation in Nevada SOC

30 28 27

15 11 10 9 7 7 5 4 4 5 3 1 1 0 0 0 0 0

Youth Parent Missing Caregiver Clinician Administrator Family Member Professional Staff Non Respondent (N=56) Respondent (N=51)

As indicated in Figure 5.1 above, the majority of survey respondents stated they consider themselves “professional staff” involved in the development of the Nevada

SOC (55%). This is followed by individuals who describe themselves as an

“Administrator” or someone who is in a decision-making capacity related to the SOC

(22%) and a “Clinician” who is a practitioner licensed to diagnose and treat children

(14%). The remaining respondents identified as parents (6%), caregivers (2%) and youth

(2%). As visually represented in Figure 5.1 above, the trends of role representation are similar to those of the survey non-respondents, except there was an underrepresentation of youth in the survey respondents. Thus, youth results are subject to question.

Professional staff constituting the majority of respondents is important for understanding the results of this study. As described in the types of networks above, the “lead organization model” of networked implementation identifies an organization to serve as a “lead” in the network activities. Thus, the Division of Child and Family

Services (DCFS) as a lead organization plays a central role in the network activities and management. The majority of respondents identifying as professional staff is evidence of the lead organization model as the type of network for the Nevada SOC. This is discussed further in Chapter Six below. It is important to note that respondents who identify as professional staff may be from other public organizations represented in the

Nevada SOC network, not just the DCFS. Specific organization names were not collected in the survey.

Figure 5.2 Geographic Region of Representation

25 23 21 20 20 19

10 9 7 5 5 4

0 0 0 Clark County Washoe County Rural Counties Statewide Missing Non Respondent (N=56) Respondent (N=51)

Figure 5.2 above depicts the geographic region of the respondents. The number of respondents representing a “statewide” (38%) or Clark County (37%) geographic areas constitute the majority of respondents (75%). This is followed by Washoe County

(17%) and the rural region (8%). As visually represented in Figure 5.2 above, the trends of geographic region representation are similar to those of the survey non-respondents.

Table 5.1 below presents the crosstabulation of respondents’ geographic representation by their identified role within the Nevada System of Care. Clinicians are more represented in Clark County than Washoe County whereas professional staff are more represented in Washoe County than Clark County. The highest number of

95 professional staff identify as having a “statewide” representation. While the rural region does not have a high number of participants involved in the Nevada SOC, there is an even representation of roles for those who are involved (although not all roles are represented).

Table 5.1

Survey Respondent Attributes – Geographic Region by Respondent Role

Parent Caregiver Youth, Clinician Professional Administrator (N=3) (N=1) 18+ (N=7) Staff (N=11) (N=1) (N=28) Geographic 1 ------5 3 Region- 11% 56% 33% Washoe (N=9) Geographic ------6 8 5 Region-Clark 32% 42% 26% (N=19) Geographic 1 -- 1 1 1 -- Region-Rural 25% 25% 25% 25% (N=4) Geographic 1 1 -- -- 14 3 Region- 5% 5% 74% 16% Statewide (N=19) Zero respondents identified as representing the “for-profit” sector. Item response rate, N=51 Percentages are by row.

Figure 5.3 below represents the length of time respondents have been involved with the Nevada SOC. Most of the respondents (29%) report being involved in the

Nevada SOC for 10 years or more. The next largest group of respondents have been involved for one year or less (24%). So, over half of the survey respondents have either been involved for a lengthy period of time (historicity) or they are brand new to the

96 project. This is followed by those who have been involved for at least two years which is the time the state received its most recent System of Care grant (20%) and those who have been involved for longer than 2 years (18%). Some respondents reported they were involved at some point in time but are no longer involved (2%) or they don’t consider themselves as being involved with the Nevada SOC (8%).

Figure 5.3 Length of Time Involved in Nevada SOC

16 15

12 11 10 10 9 8

6 4 4

2 1 1

0 1 year or At least 2 Longer 10+ years No longer Not Missing less years than 2 involved involved years Respondent (N=51)

As depicted in Figure 5.4 below, the majority of the respondents reported they represent a public organization such as city, county or state government (77%). The next largest group represents non-profit organization (13%) followed by those not representing an organization at all when they participate in the Nevada SOC (10%). As visually represented in Figure 5.4 below, the trends for the type of organization the

97 network member represents are similar to those of the survey non-respondents, except there is a discrepancy between the number of survey respondents and non-respondents who represent the non-profit sector. There were more non-respondents from the non- profit sector. Thus, results related to the non-profit sector should be treated with caution.

Figure 5.4 Organizational Type

45 40 40

35 30 30

20 18

10 7 6 5 5 1 0 0 0 0 Public Non-Profit For-Profit No Org. Missing Non Respondent (N=56) Respondent (N=51)

Table 5.2 below presents a cross tabulation of the respondents’ role with their length of time involved in the Nevada SOC. The majority of respondents who report being involved in the Nevada SOC for one year or less are professional staff (73%). Most of the respondents who have been involved for 10 years or longer are also professional staff (44%) followed by clinicians and administrators (27% each). This means about half

98 of the professional staff have some history of the Nevada SOC due to their length of time involved in its activities while the other half recently joined the Nevada SOC network.

Table 5.2

Survey Respondent Attributes –Length of Time Involved by Respondent Role

Parent Caregiver Youth, Clinician Professional Administrator (N=3) (N=1) 18+ (N=7) Staff (N=11) (N=1) (N=28) 1 Year or Less 2 -- 1 -- 8 -- (N=11) 18% 9% 73% 2 Years ------7 3 (N=10) 70% 30% Longer than 2 1 -- -- 2 4 2 years 11% 22% 44% 22% (N=9) 10 Years or -- 1 -- 4 6 4 Longer 7% 27% 40% 27% (N=15) No longer ------1 -- involved 100% (N=1) Don’t consider ------1 2 1 myself as 25% 50% 25% being involved (N=4) Item response rate, N=50 Percentages are by row.

Table 5.3 below presents a cross tabulation of the respondents’ role in the

Nevada SOC and the type of agency they represent in the SOC. Professional staff comprise the majority of network actors who represent the public sector (63%) while parent representatives primarily represent the non-profit sectors (33%). Administrators

99 have been involved for longer periods of time while parents and a youth have more recent involvement. The parent and youth involvement could be attributed to a tendency for turnover in voluntary community participation in collective action groups.

Table 5.3

Survey Respondent Attributes – Agency Type by Respondent Role

Parent Caregiver Youth, Clinician Professional Administrator (N=3) (N=1) 18+ (N=7) Staff (N=11) (N=1) (N=28) Public -- 1 -- 5 25 9 Organization 3% 13% 63% 23% (N=40) Non-Profit 3 ------2 1 (N=6) 50% 33% 17% No agency -- -- 1 2 1 1 representation 20% 40% 20% 20% (N=5) Zero respondents identified as a “family member.” Item response rate, N=51 Percentages are by row.

In summary, the Nevada SOC network is primarily comprised of professional staff and of actors who represent the whole state or Clark County when they participate in the network. Additionally, there is one group of actors that have been involved in

System of Care activities in the state for ten years or more while another group of actors has been involved for one year or less. Most of the Nevada SOC network actors indicate their organization type as public with less representation from the non-profit sector and no representation from the private sector. Youth and parents do not appear to have

100 presence in the network based on the number of actors. However, the parent involvement will be discussed further below in the centrality findings.

The four categories of attribute variables reported above (role, geographic region, length of time in Nevada SOC and organization type) are important for describing the patterns of interactions amongst the network actors. Thus, the findings presented below refer to these categories as “attributes.” At times, attributes are reported as actors who share the same attribute (i.e. two actors are tied together by virtue of sharing the same geographic representation) or they are reported as small group categories. The results below represent the attributes with shortened descriptions as follows:

• “Whom” = Identification of the role the respondent indicates they represent

when they participate in the Nevada SOC (i.e. parent, professional staff,

clinician),

• “Geo” = Identification of the geographic region the respondent indicates they

represent when they participate (i.e. Clark County, Washoe County, rural Nevada

or statewide),

• “Time” = Identification of a category that describes the length of time the actor

has been involved in the Nevada SOC, and

• “Org” = Identification of the organization type the actor belongs to.

101

Research Question Two

The following reports the of patterns of interaction amongst actors in the

Nevada SOC network, which comprise the overall network structure. As discussed in

Chapter Two and the data analysis plan in Chapter Four, the results are multi-level and include visual representations of the network. The following framework guides the presentation of findings and support the overall description of the Nevada SOC network.

As discussed in the theoretical explanations of network behavior in Chapter Two above, a network at the global level is comprised of “many overlapping sets of dyadic relationships that collectively make up the full network” and these multi-layered relationships must be attended to by the network administrators (Provan & Lemaire,

2012, p. 641). Network relations can develop over time from an “emergent” nature where network members are initially forming relationships (or ties) to a more

“strengthened” relationship of trust (Provan & Lemaire, 2012, p. 641-642). Figure 5.5 below depicts the relationship types measured in this study on a continuum from emergent to strengthened relationships. This will serve as a framework for describing the results of this study.

Figure 5.5 Nevada SOC Network Emergent to Strengthened Relationship Ties

Operations Sharing Sharing (working Trust Information Resources together)

Emergent Relationships Strengthened Relationships

102

Thus, the following results present findings amongst the patterns of four relationship variables: operations (working together), sharing information, sharing resources and trust. The operations and sharing information relationships are considered emergent, while the sharing resources and trust relationships are considered strengthened relationships. The results are also presented according to specific attributes of the network actors as discussed above.

The patterns of interaction that comprise the network structure are reported below according to the findings of the multi-level data analysis techniques described in

Chapter Four above. The findings are presented first at the whole network or global level. At this level of analysis, whole network patterns of density, centrality and cohesiveness are reported. Centrality is reported as a whole network level measure as well as an individual level measure, which is discussed further below. Visualizations of the network at this whole level are also included as an illustration of the findings at this whole network level of analysis. The data are then presented at the triadic level of measurement with a report of the frequency of triadic types evident in the network (as discussed in Chapter Two and depicted in Figure 2.2 above). Additionally, patterns by which network members tend to form ties with others who share the same attributes

(homophily) is included in the individual level of measurement.

Global or Whole Network Level of Measurement

From a collective action perspective, analyzing patterns of relationships of the network as a whole can indicate how the network is operating, as a whole, to reach a common goal (Provan & Lemaire, 2012). To describe the patterns of interaction at the

103 whole network level of analysis, survey data were used to measure density, centrality

(degree and betweenness) and cohesion (reciprocity and transitivity). The following reports those findings.

Density. The network density measure allows us to analyze the global structure of the network (Hanneman & Riddle, 2005). Density is the measure of the number of actual ties within the network divided by the number of possible ties (Yang, Keller &

Zheng, 2017). UCINET calculates the density for binary matrices by dividing the total number of ties observed by the total number of possible ties (Borgatti, Everett &

Freeman, 2002). The closer a density score is to one, the higher the density is of the network (indicating a high number of ties amongst the network actors). Table 5.4 below presents the overall density scores for the Nevada SOC broken down by the four relationship variables.

As presented in Table 5.4 below, the Nevada SOC is a moderately dense network. The network is most dense for the operations relationship (.407) and decreases by nearly half as we move from the emergent to strengthened relationships.

Density is lowest for the trust relationship (.201). From the multi-layered patterns of relationships perspective described above, the ties amongst the Nevada SOC network actors (i.e. density) decreases as we move from emergent to strengthened relationship types.

104

Table 5.4

Overall Network Density by Relationship Variable

Operations Information Resources Trust

Density Density .407 .262 .229 .201 # of Ties 1039 668 583 512

Centrality. As discussed in Chapter Two above, degree centrality at the actor level describes how central an actor is within the network by calculating the number of ties the actor has (Yang, Keller & Zheng, 2017). Analyzed from the whole network level, centralization is the distribution of ties across the network whereby a higher overall centralization score indicates a greater difference between centrality scores amongst the individual actors (Yang, Keller & Zheng, 2017). Essentially, a centralization score closer to one indicates a highly centralized network meaning there is a smaller group of central actors within the context of the whole network.

The degree measure of centrality includes an “OutDegree” number which is the number of ties the actor sends out to others and the “InDegree” number which is the number of ties coming in to the actor from others in the network (Borgatti, Everett &

Johnson, 2013; Kolpakov, 2012; Wasserman & Faust, 1994). In this study, the

OutDegree is measured by Respondent X reporting their ties with other actors in the network. For example, when Respondent X reports they share information with other actors in the network, the resulting number of ties would be the OutDegree score for

Respondent X. The InDegree is measured by the other actors (also respondents) who

105 report their ties with Respondent X. InDegree provides a measure of how central an actor is from the perspective of the other actors in the network (Klaster, Wilderom &

Muntslag, 2017). For example, the number of actors who report sharing information with Respondent X would be the InDegree score for Respondent X. The OutDegree ties typically represent influence while the InDegree ties can indicate popularity, prestige or power.

Table 5.5 below reports the degree centralization scores for the Nevada SOC network at the whole network level. The closer a centralization score is to one, the higher the centrality the network (indicating the number of network ties is concentrated to a smaller number of actors). The scores indicate, for actors reporting their ties going

“out” to others in the network (OutDegree), the network is more centralized at the strengthened relationship of trust and least centralized at the emergent relationship of operations. This is an indication that as the relationship tie strengthens, centrality is less dispersed and becomes more focused on a core group of actors. This pattern was not the same for the number of ties coming in to an actor (InDegree) indicating a greater distribution of ties going to actors across the network as opposed to going out from the actors.

106

Table 5.5

Overall Network Degree Centrality by Relationship Variable

Operations Information Resources Trust

Centralization .502 .630 .664 .693 (outdegree) Centralization .461 .488 .317 .326 (indegree)

From an implementation standpoint, the central actor(s) have the “power” to control opportunities and constraints that impact implementation (Borgatti, Everett &

Johnson, 2013) and to understand how equal that distribution of power may be across network actors (Yang, Keller & Zheng, 2017). In the network visualizations below, the centrality of the actors is depicted by the size of the node. Tables 5.6-5.9 below report the top five actors with the highest degree centrality scores for each relationship.

Scores for betweenness centrality are also included in tables 5.6-5.9 below.

Betweenness centrality scores are another measure of power within a network. This score reports the extent to which an actor falls on the path between pairs of other actors (Borgatti, Everett & Johnson, 2013). For example, it is an indicator of power within a network as the actor is in a position to benefit from the flow of information and resources. As discussed in Chapter Two above, network actors can access social capital and serve as brokers in the network by filling “structural holes” (Burt, 1992). The betweenness centrality score captures the extent to which an actor is in a position of power given their location between other central actors (Yang, Keller & Zheng, 2017).

107

Again, actors with this type of power have the potential to influence the opportunities and constraints involved with implementation.

While tables 5.6-5.9 below report the top five central actors according to degree and betweenness centrality, Appendix F includes the centrality scores for each of the identified network actors. In Appendix F, survey non-respondents will not have an

OutDegree number in the Appendix, but they may have an InDegree number if other actors indicate a tie to them. Actors with high degree and betweenness centrality scores are shaded in green in the table located within the Appendix F.

The degree and betweenness centrality scores for the top five actors within each relationship variable reported below reveal a core “team” of actors central to the

Nevada SOC network across each of the relationship variables (actor numbers 57, 55, 29 and 107). This is discussed further in the descriptions for each table presented below.

This core team is also visually evident in the network figures below.

108

Table 5.6

Top Five Actors for Degree Centrality and Betweenness Centrality Scores, Operations Relationship

Operations Operations Operations Out Degree Centrality In Degree Centrality Betweenness Centrality Rank Actor # of Ties Rank Actor # of Ties Rank Actor # of Ties

1 57 79 1 57 43 1 57 565

2 55 70 2 55 40 2 55 268

3 29 63 3 29 37 3 29 259

4 107 63 4 107 36 4 31 228

5 75 59 5 75 36 5 107 213

Table 5.6 above reports the top five actors with the highest centrality scores for the operations network relationship. It reveals a core “team” for the Nevada SOC network also visible in the network depictions below. This team includes a caregiver, a professional staff, a parent and an administrator (actors 57, 55, 29 and 107).

109

Table 5.7

Top Five Actors for Degree Centrality and Betweenness Centrality Scores, Information Relationship

Information Information Information Out Degree Centrality In Degree Centrality Betweenness Centrality

Rank Actor # of Ties Rank Actor # of Ties Rank Actor # of Ties

1 57 79 1 57 37 1 57 1052

2 55 66 2 55 34 2 55 487

3 31 46 3 29 27 3 29 259

4 73 45 4 73 26 4 79 254

5 81 44 5 10 25 5 31 234

Table 5.7 above reveals three actors of the core team retain high degree centrality and betweenness centrality scores (actors 57, 55 and 29). However, there are other actors who emerge in the top five for this relationship variable (actors 31, 73, 81,

10 and 79). Interestingly, actor 107 ranks in the top five central actors for all relationship variables except this information sharing relationship. It is noted that this actor had a low attendance rate at the meetings that were analyzed for the affiliation network described above, which may have impacted their information sharing centrality

(i.e. if they are not present at the meetings, how can they share information?).

110

Table 5.8

Top Five Actors for Degree Centrality and Betweenness Centrality Scores, Resources Relationship

Resources Resources Resources Out Degree Centrality In Degree Centrality Betweenness Centrality Rank Actor # of Ties Rank Actor # of Ties Rank Actor # of Ties

1 57 79 1 57 27 1 57 711

2 55 67 2 55 25 2 29 447

3 29 63 3 29 25 3 55 427

4 14 52 4 10 24 4 84 241

5 107 46 5 73 22 5 10 207

Interestingly, Actor 29, who identifies as a parent in the Nevada SOC network, typically falls at rank three in the degree centrality and betweenness centrality scores.

However, as presented in Table 5.8 above, the actor rises to rank two in the betweenness centrality score for resources. This is an indicator that this actor serves as a bridge or fills those structural holes between others who share resources. This could translate to Actor 29 having social capital. Actor 29 also identified as a “parent” in their role in the Nevada SOC network. As reported earlier, while there are not a high numbers of parents participating in the network, there is a parent that plays a central role in the network.

111

Table 5.9

Top Five Actors for Degree Centrality and Betweenness Centrality Scores, Trust Relationship

Trust Trust Trust Out Degree Centrality In Degree Centrality Betweenness Centrality Rank Actor # of Ties Rank Actor # of Ties Rank Actor # of Ties

1 57 79 1 55 26 1 57 1081

2 55 54 2 57 24 2 55 490

3 36 36 3 29 23 3 84 311

4 29 35 4 28 21 4 107 242

5 10 35 5 10 20 5 29 220

Table 5.9 above reveals actors 55 and 57 retaining their central positions for the trust relationship. However, new actors continue to emerge (actors 36, 10, 28 and 84).

As was evident in the network depictions, the core “team” of a parent, caregiver, professional staff and an administrator retained rankings in the top five for degree centrality and betweenness centrality scores. Kolpakov (2012) states a network is considered centralized when there are one or more actors with high centrality scores.

Therefore, the tables above indicate the Nevada SOC is a centralized network with a core group of actors who retain high degree centrality and betweenness centrality scores. The caregiver and professional staff (actors 55 and 57) remained in the top two ranks across all four network relationships except the information relationship (see discussion of Table 5.5 above). The parent (actor 29) remained in ranks three, four or

112 five for most of the relationship variables (see discussion of Table 5.6 above). An administrator (actor 107) was ranked in the top five in all network relations except information sharing.

As discussed earlier, Provan and Lemaire (2012) describe the importance of attending to multi-layered relationships within network structures. In the Nevada SOC network, the network is denser at the emergent relationship types of operations and information indicating a more connected network. The density decreased at the strengthened relationship types of sharing resources and trust indicating a network structure still developing. This is evident in the degree centrality and betweenness centrality scores reported above as the core team is present in the operations and information relationships, but other actors emerge in the top five for the sharing resources and trust relationships. The degree centrality and betweenness scores also reveal “rising stars” or those actors emerging as also being central to the Nevada SOC network. For example, Actor 10 is present in the top five across three of the four relationships, particularly the resources and trust variables, which are considered to be in the earlier stages of network formation. Kolpakov (2012) posits the addition of new actors and the increased centrality of the core team is also an indication of a network at an implementation stage. A network manager can use this data by positioning Actor 10

(a professional staff) more prominently in network activities to foster further development of the network.

113

Cohesion

As described in Chapter Two above, measures of cohesion include reciprocity and transitivity. The following reports the results of measures at the whole network level.

Reciprocity. The reciprocity values reported in Table 5.10 below indicate the propensity of dyads within the network that are reciprocal. In general, reciprocal relationships are an indicator of trust within a network (Kolpakov, 2012). Thus, Table

5.10 below reports the network has a higher percentage of reciprocal relationships at the emergent relationship types of operations and information as opposed to the strengthened relationships of resources and trust.

Table 5.10

Network Reciprocity by Relationship Variables

Operations Information Resources Trust Dyad-based .586 .529 .450 .418 Reciprocity

Transitivity. Network structure can also be described by measuring the patterns of relationships amongst groups of three actors within the network that occur as a transitive triad (Kolpakov, 2012; Wasserman & Faust, 1994). As discussed in Chapter

Two above, a transitive triad within organizations is “the boss of my friend is also my boss” and it is an indicator of hierarchy in a network. Transitivity at the whole network level is the percentage of transitive triads that comprise the whole network structure and is an indication of how hierarchical a network is (Borgatti, Everett and Johnson,

114

2013). A higher score indicates the presence of more transitive triads than what would be expected given the density of the network. The higher the percentage, the more hierarchical the network. Table 5.11 below presents the transitivity percentage for the whole network within each of the four relationship variables. The results indicate a moderate presence of transitive triads occurring in the Nevada SOC network with the highest percentage of triads within the operations relationship (66%) and the lowest percentage within the trust relationship (51%).

Table 5.11

Network Transitivity by Relationship Variables

Operations Information Resources Trust Triplet .659 .549 .545 .514 Transitivity

While still moderate, the results in Table 5.9 above indicate the network structure exhibits moderate hierarchy at the emergent relationships of operations and information and becomes less hierarchical at the strengthened relationship levels of resources and trust. Transitivity is reported above as a measure of the network at the global level. Additional detail on the types of transitive triads observed in the Nevada

SOC network is reported below in the triadic level of measurement.

The following presents visual depictions of the global or whole network structure. Presenting the data in this way provides an opportunity to visualize the network in light of the results reported above. The "Operations Relationship" depicted in Figure 5.5 below represents the survey respondents’ ties to one another based on

115 who they indicated they were directly or indirectly involved with when they work on the implementation of the Nevada SOC. Each colored shape represents a node or actor in the network. The node’s color represents the “whom” attribute and the shape represents the “geo” attribute. The size of the node represents the degree centrality of the actor. The color, shape and size representations continue in the following depictions of the Nevada SOC network according to each of the network relations.

116

Figure 5.6 Nevada SOC Network Operations Relationship

Pink = Parent Circle = Clark County Light blue = Caregiver Square = Washoe County Purple = Youth Triangle = Rural Yellow = Clinician Quadrant Box = Statewide Green = Professional staff Red = Administrator *Size of node indicates number of connections.

117

In Figure 5.6 above, the density is observed by the number of black lines (ties) amongst the nodes within the network. Additionally, the degree centrality measure in the UCINET NetDraw program measures how central an actor is within the network. In the figure above, the actors’ degree centrality is represented by the size of the shape representing the node. Thus, the pattern of interaction for the operations relationship reveals professional staff, some administrators, a caregiver and a parent are the most central actors in this network.

In determining the overall network structure, it is important to visually examine the structure from multiple perspectives. UCINET’s NetDraw program offers a “Gower” function that presents the nodes as data points represented by their dimensional space of dissimilarity. This function allows the researcher to rotate the data in order to and view the structure from different angles. The figure below offers the Gower view of the same operations relationship network depicted in Figure 5.6 above.

Figures 5.6 (above) and 5.7 (below) also illustrate the “clinician” groups as an outlier group (depicted by the yellow dots) to the network with professional staff

(green) and administrators (red) holding more central roles within the network.

Network members from the rural areas (triangles) are also located as an outlier in the network.

118

Figure 5.7 Nevada SOC Network Operations Relationship, Gower View

119

While the density of the ties amongst the nodes makes it difficult to see an overall structure of the above network depiction (Figure 5.7), it is possible to see structural patterns in the network. Figure 5.8 below is a repeated image of Figure 5.7, but is enhanced with a visual aide to see a fully connected star structure. This indicates a structure of subgroups connected by a small group of central actors who serve as a

“bridge” between the subgroups.

Figure 5.8 Nevada SOC Network Operations Relationship, Gower View, Fully Connected Star Depicted

120

In Figure 5.8 above, bright blue circles and lines have been added to enhance the visibility of the fully connected star structure of the network. This is an ideal structure for information sharing as information can travel through the central node or around the network as well.

In the depictions above, the central nodes are evident by the size of the shape associated with them. This is also supported in the individual level results reported below, which reports the most central actors across the four relationship variables. In their structural role, the central node(s) can control the flow of information (i.e. messaging), impede the travel of information (i.e. “bottlenecking information), or promote the expedited travel of information.

Given the structure is also circular in nature, information doesn’t have to flow through the central node as it can travel around the network circle. While this is beneficial for access to information, it may travel slower (as it goes around the perimeter) and there is risk of the information getting distorted as it is not traveling through the control mechanism.

121

Figure 5.9 Nevada SOC Network Operations Relationship, Network Members Grouped by Whom they Represent

In continuing an examination of the operations relationship, the network visualization in Figure 5.9 above clustered the actors by the “whom” attribute. We know the largest group of respondents in the survey consisted of professional staff and employees, followed by clinicians and youth. This is evident in the depiction above. The

122 benefit of such depiction is that one can see the patterns of interactions amongst those actors. For example, network administrators (red) are mostly tied to the professional staff (green) and have fewer ties to the parent (pink) or clinician groups (yellow). There is also a parent representative highly tied to the professional staff, clinicians and administrator groups. By “pulling” the network depiction apart as in Figure 5.9, one can gain a better understanding of how connected the nodes are. For example, the SOC framework emphasizes parent “voice” in the network activities. While there are few parent representatives (pink) in the network, they do have multiple ties to other network members. One parent representative in particular (labeled as node number

29) has numerous ties to other network members. This means Node 29 carries a large responsibility or important role in maintaining those connections. If Node 29 were to leave the network, then the network’s patterns of interaction with parent representatives would be impacted.

In another view of the same operations variable, the network members are groups by the geographic location they represent when they participate in the Nevada

SOC. In Figure 5.10 below, we can see the majority of the actors in the network view themselves as representing the whole state followed by Clark county and then Washoe county. It is not surprising to find a small number of rural representatives in the network as the current SOC grant has an explicit focus on Clark and Washoe Counties with an intention to expand some services to the rural areas as an implicit benefit.

Within the sub-groups it is evident there is one parent representative within each cluster, except the Clark county cluster. This doesn’t necessarily mean there is no

123 parent representation for Clark county, it may be a parent representative respondent selected “statewide” as their response to the survey item when they really span the state and the Clark county target area.

Figure 5.10 Nevada SOC Network – Operations Relationship, Network Members Grouped by Geographic Location of Representation

124

Figure 5.11 below represents the network according to the “sharing information” relationship variable. The slight reduction in density of ties (as reported above) is evident in the representation with patterns of degree centrality resting with professional staff, administrators and one parent remaining. In this depiction, one can ascertain the professional staff as the “hub” of the information flow. From an implementation perspective, the actors in their hub positions have the power to influence the opportunities and constraints of information flow. There are some administrators located within the hub of the information flow while others (presumably from systems outside of the primary network) possess fewer information ties.

125

Figure 5.11: Nevada SOC Network Information Sharing Relationship

Figure 5.12 below represents the same information relationship network from the Gower view. This depiction of the network again highlights the clinicians (yellow) as spatially located on the periphery of the network and the youth representative is far

126 removed from the network. This information is applicable for the network manager in deciding what steps they can take to strengthen the connections of those outlier groups within the network.

Figure 5.12 Nevada SOC Network Information Sharing Relationship, Gower View

127

Figure 5.13 below represents the network according to the “sharing resources” network relation. Actors who have higher degree centrality scores (larger size nodes) also indicate a statewide representation (quadrant box shape). This pattern supports the one of the aims of the SOC implementation which is to aligning resources across state systems (i.e. central actors have a statewide focus).

Figure 5.13 Nevada SOC Network Resource Sharing Relationship

Presence of transitive Pink = Parent Circle = Clark County triads. Light blue = Caregiver Square = Washoe County Purple = Youth Triangle = Rural Yellow = Clinician Quadrant Box = Statewide Green = Professional staff Red = Administrator *Size of node indicates number of connections.

128

Transitivity is a measure of transitive triads in a network structure and it indicates the extent to which a network structure is hierarchical (Hanneman & Riddle,

2011b; Kolpakov, 2012). While a complete triadic census will be discussed further below, Figure 5.13 above provides a good visual example of the presence of transitive triads for the “resource sharing” relationship variable. For example, in the highlighted box in Figure 5.13, actor numbers 92, 103, and 48 are an example of a transitive triad where actor 92 has an asymmetric relationship with actor 103 and 48 (depicted by a one-way arrow) while 103 and 48 have a mutual relationship (depicted by a two-way arrow).

Figure 5.14 below depicts the same resource sharing network from the Gower view. As discussed earlier, a network with high levels of transitivity, “tend to have a clumpy structure” containing “knots of nodes that are all interrelated” (Borgatti, Everett

& Johnson, 2013, p. 155). The depiction below has been enhanced with bright blue circles to highlight the clusters or “clumpy structure” evident in the network.

129

Figure 5.14 Nevada SOC Network – Resource Sharing Relationship, Gower View

130

Figure 5.15 below represents the network according to the “trust” relationship variable. The reduction in density of ties continues to be evident as the network relation moves from emergent to strengthened. The patterns of degree centrality continue with the same professional staff, administrators and one parent remaining central to the network. However, the overall density pattern is sparser. When trust is less dispersed, there is evidence of a lead organization model network (Provan & Kenis,

2007).

131

Figure 5.15 Nevada SOC Network Trust Relationship

Pink = Parent Circle = Clark County Light blue = Caregiver Square = Washoe County Purple = Youth Triangle = Rural Yellow = Clinician Box = Statewide Green = Professional staff Red = Administrator

132

As discussed above, the Gower view network images represent the actors’ spatial location within the network while also depicting their degree centrality to the network. Figure 5.16 below reveals clusters of actors for the trust relationship. For example, actors 55,57 and 107 exhibit high degree centrality scores noted by the larger node size and it is visually evident the other actors cluster spatially near them. The

Gower depiction also presents more integration of the Clinician sub-group than what was evident in the figures above, although they still remain spatially “outside” the network.

133

Figure 5.16 Nevada SOC Network Trust Relationship, Gower View

Again, the overall density of the network has decreased at this strengthened relationship type. As discussed earlier, the network density (meaning the numbers of ties among the members of the network) decreases as the relationship type moves from

134 emergent to strengthened. The Gower depiction above reveals tightly clustered patterns of interaction with a few central actors depicted as being “outside” the clusters indicating their role as boundary spanners as well.

In summary, the depictions presented above reveal a network density structure that decreases as the multiple layers of relations change from working together

(emergent relations) to sharing information and resources to trust (strengthened relations). The network structures also reveal a “team” of a central set of actors consisting of a parent, care giver, professional staff and administrator. There is a repeating pattern of some actors who are structurally centralized within the network giving them the opportunity to control the flow of information and resources. A closer examination of this pattern is discussed below in the measure of degree centrality for network participants. The patterns of relationships also reveal a centralized network structure with fully connected star patterns and clusters of “hubs” for the flow of information and resources. The clinicians do not present as being highly central to the network. Overall, the professional staff (particularly those with the statewide emphasis) appear to serve as the “glue” for the network as they are either highly centralized or they serve as bridges or boundary spanners.

From a collective action theory perspective, the organizer may ask what the basis for an actor’s decision is to participate or not participate. For example, the clinician group appears to be on the outskirts of the network. What information do they rely on to determine the costs and benefits of their participation in the Nevada SOC network? Clinicians frequently express the difficulty in participating in public action

135 meetings because their time away from their clinical practice is money lost because they are not billing for services. From a networked governance perspective, the public entity would revisit the goals of the network and if clinician involvement was necessary, then they would act to increase their degree centrality within the network.

Triadic Level of Measurement

As described above, transitivity in a network is the patterns of relationships amongst three actors and is an indication of hierarchy within the network. As reported at the global or whole network level of measurement, the Nevada SOC network is moderately transitive with triads comprising 51% (trust relationship) to 66% (operations relationship) of the network ties. As discussed in Chapter Two above, there are varying types of triad relationships. For example, one pattern amongst three actors could be two actors having a mutual relationship with each other while each has an asymmetric relationship with the third actor in the triad. A triadic census analysis increases our understanding of the frequency of triadic types and further explains the overall structure of the network. For example, in Figure 2.2 on page 55, triad type 16-300 is fully reciprocal (each actor has a mutual relationship with the other two in the triad), a high percentage of this type of triad an indication of a sustainable network (Kolpokov,

2012).

In order to conduct a triadic census, a different network analysis program is carried out using the Pajek software program. The sixteen types of triads included in

Figure 2.2 on page 55 are included in Pajek and the program will test for the presence of

136 the triadic types (Nooy, Mrvar, Batagelj, 2011; Mrvar & Batagelj, 2016). Tables 5.12-

5.15 below report the type of triad (as depicted in Figure 2.2), the frequency the triad occurs in the network (column labeled “Number of Triads (ni)”), the frequencies as they would be expected by chance (column label “Expected (ei)”), and the relative difference between the actual versus expected presence of the triad which is calculated by taking the difference between the actual and expected and then dividing it by the expected

(column label (ni-ei)/ei).

As discussed in Chapter Two above, Balance Theory identifies different models of network structures where certain types of transitive triads are permitted or forbidden according to the four assumptions listed in Chapter Two (Rawlings, 2017). The models describe the overall structure of the network such as the extent to which the network is clustered or hierarchical (Nooy, Mrvar & Batagelj, 2011). As presented in Tables 5.12-

5.15 below, the findings of the triadic census across all four network relations reveal a high frequency of triadic type 201 (it is an “open” triad where Actor A has a mutual relationship with Actor B and Actor C, but no relationship exists between Actor C and

Actor B). This particular type of triad is considered “forbidden,” meaning it violates the rules of balance theory and therefore the findings of the triadic census do not fit any balance theoretical models (Nooy, Andrej & Mrvar, 2011). However, it was also noted in

Chapter Two above that a triadic census is best for longitudinal studies as network relations change over time. Thus, Tables 5.12-5.15 present the frequencies of triad types across all four relations as a point-in-time description of the presence of triads in

137 the Nevada SOC network. The triad types with the highest frequencies are noted in bold font in the following tables.

Table 5.12

Triadic Census of the Nevada SOC Operations Network

Triad Type* Number of Triads Expected Number ni-ei/ei Observed (ni) of Triads (ei) 3-102 3438 1279 1.69 16-300 1607 95 15.86 1-003 3716 901 3.12 4-021D 358 1279 -.72 5-021U 430 1279 -.66 9-030T 223 1758 -.87 12-120D 356 605 -.41 13-120U 526 605 -.13 2-012 3327 3719 -.11 14-120C 247 1209 -.80 15-210 1674 831 1.01 6-021C 475 2557 -.81 7-111D 1203 1758 -.32 8-111U 1595 1758 -.09 10-030C 9 586 -.98 11-201 1641 605 1.71 Chi-square: 44988.8157***

As reported in Table 5.12 above, the Operations relation network primarily consists of triad types 3-102, 16-300, 1-003, 15-210 and the forbidden type 11-201. The most prevalent triad type is 16-300 which is a fully mutual triad and is observed 15.86 times more than what would be expected by chance. A large number of fully mutual triads in a network is a sign that the network is “less vulnerable to exogenous shocks”

(Kolpakov, 2012, p. 181). The Chi-square is significant at the .001 level indicating a significant difference in the frequencies of the observed triads and the expected triads.

138

Table 5.13

Triadic Census of the Nevada SOC Information Network

Triad Type* Number of Triads Expected Number ni-ei/ei Observed of Triads (ni) (ei) 3-102 3414 1272 1.68 16-300 589 7 86.52 1-003 7355 3366 1.19 4-021D 372 1272 -.71 5-021U 225 1272 -.82 9-030T 120 903 -.87 12-120D 138 160 -.14 13-120U 243 160 .52 2-012 4064 7168 -.43 14-120C 125 321 -.61 15-210 728 114 5.40 6-021C 345 2544 -.86 7-111D 694 903 -.23 8-111U 1328 903 .47 10-030C 8 301 -.97 11-201 1077 160 5.72 Chi-square: 3396.0663***

Table 5.13 above reports a similar pattern of triad types for the Information

Sharing relation network, which primarily consists of triad types 3-102, 16-300, 1-003,

15-210 and the forbidden type 11-201. The most prevalent triad type is 16-300 which is observed at 86.52 times more than what would be expected by chance. This continues the pattern of observing a stable network structure based on a high number of fully mutual triads. The Chi-square is significant at the .001 level indicating a significant difference in the frequencies of the observed triads and the expected triads.

139

Table 5.14

Triadic Census of the Nevada SOC Resource Network

Triad Type* Number of Triads Expected Number ni-ei/ei Observed of Triads (ni) (ei) 3-102 2775 1156 1.4 16-300 407 3 136 1-003 8389 4387 .91 4-021D 860 1156 -.26 5-021U 154 1156 -.87 9-030T 145 685 -.79 12-120D 70 102 -.31 13-120U 406 102 3 2-012 4034 7802 -.48 14-120C 92 203 -.55 15-210 512 60 7.5 6-021C 333 2312 -.86 7-111D 405 685 -.41 8-111U 1604 685 1.34 10-030C 3 228 -.99 11-201 636 102 5.26 Chi-square: 74442.0434***

As reported in Table 5.14 above, the Resource Sharing relation network primarily consists of triad types 3-102, 16-300, 13-120U, 15-210 and forbidden types 8-111U and

11-201. The most prevalent triad type is 16-300 which is observed at 136 times more than what would be expected by chance. This continues the pattern of observing a stable network structure based on a high number of fully mutual triads and it is observed at notably higher levels than the previous two relations. There is also a relatively high number of type 15-210 which is a hierarchical triad type with Actor A having a mutual relationship with Actor B and Actor C with an asymmetric relationship

140 between Actor B and C. This triad type is observed at 7.5 times than would be expected by chance. The Chi-square is significant at the .001 level indicating a significant difference in the frequencies of the observed triads and the expected triads.

Table 5.15

Triadic Census of the Nevada SOC Trust Network

Triad Type* Number of Triads Expected Number ni-ei/ei Observed of Triads (ni) (ei) 3-102 2732 1028 1.66 16-300 264 1 192.48 1-003 8917 5427 .64 4-021D 554 1028 -.46 5-021U 240 1028 -.77 9-030T 165 516 -.68 12-120D 92 65 .42 13-120U 284 65 3.38 2-012 4667 8181 -.43 14-120C 80 130 -.38 15-210 385 33 10.81 6-021C 339 2055 -.84 7-111D 471 516 -.09 8-111U 1088 516 1.11 10-030C 2 172 -.99 11-201 545 65 7.40 Chi-square: 68568.5845***

As reported in Table 5.15 above, the Trust relation network follows the same pattern as the Resource Sharing relation reported above and primarily consists of triad types 3-102, 16-300, 13-120U, 15-210 and forbidden types 8-111U and 11-201. The most prevalent triad type is 16-300 which is observed at 192 times more than what would be expected by chance. This continues the pattern of observing a stable network

141 structure based on a high number of fully mutual triads and it is observed at notably higher levels than the previous relations. There is also a high number of type 15-210 observed at 11 times than would be expected by chance. The Chi-square is significant at the .001 level indicating a significant difference in the frequencies of the observed triads and the expected triads.

While the triadic census did not result in an application to balance theoretic models due to the presence of forbidden triad types, it is notable that the most frequently observed triad type is 16-300, the fully mutual triad. Again, this is an indication of a sustainable network (Kolpakov, 2012). Also, the frequency of this triad type was higher at the strengthened relationship types of resource sharing and trust than it was at the emergent relations of operations and sharing information.

Individual Level Network Measures

Homophily. At the individual network level of analysis, it is important to examine the patterns of homophily (i.e. tendency to “stick together”) amongst the sub- groups embedded in the network (Hanneman & Riddle, 2005). The “External-Internal

Index” (E-I Index) within UCINET measures the number of ties among the actors as embedded in their own sub-groups or groups external to their own (homophily). The E-I

Index score is calculated by subtracting the observed external ties from the observed internal ties and then divide that sum by the total possible ties (Hanneman & Riddle,

2005). A score closer to a negative one indicates groups are mostly tied to members of their same group; whereas, a score closer to a positive one indicates the actors are tied to others outside of their groups (Hanneman & Riddle, 2005). The permutation test

142 described above for the Quadratic Assignment Procedure is also used in the E-I Index tests. The following reports the results of the E-I Index.

Table 5.16

External-Internal Index of Ties Between Network Attributes and Network Relationships

Operations Information Resources Trust (1,310 Ties) (874 Ties) (804 Ties) (722 Ties) Attribute Whom Internal 1052 (80%) Internal 740 (85%) Internal 686 (85%) Internal 594 (82%) External 258 (20%) External 134 (15%) External 118 (15%) External 128 (17%) E-I Index -.162+ E-I Index -.565+* E-I Index -.617+* E-I Index -.584+ Geo Internal 778 (59%) Internal 574 (66%) Internal 522 (65%) Internal 482 (67%) External 532 (41%) External 300 (34%) External 282 (35%) External 240 (33%) E-I Index -.104+* E-I Index -.314+* E-I Index -.299+* E-I Index -.335+*

Time Internal 692 (53%) Internal 460 (53%) Internal 440 (55%) Internal 384 (53%) External 618 (47%) External 414 (47%) External 364 (45%) External 338 (47%) E-I Index .104 E-I Index -.053+ E-I Index -.095+ E-I Index -.064+ Org Internal 992 (76%) Internal 700 (80%) Internal 626 (78%) Internal 580 (80%) External 318 (24%) External 174 (20%) External 178 (22%) External 142 (20%) E-I Index -.321+* E-I Index -.602+* E-I Index -.557+* E-I Index -.607+* E-I Index reported is re-scaled index to account for demographic constraints of size and density. + Preponderance of internal over external ties (i.e. tendency toward group closure). *p<.05, significant difference between expected and observed. Permuted 10,000 times. Percent is by individual cell.

Table 5.16 above describes tendency for network actors to form ties with actors who share the same sub-group attributes. The findings above indicate the Nevada SOC network actors tend to form ties with members of their own groups, particularly those

143 who share the same role and same organization type. Each cell in Table 5.16 above reports the actual and percent of observed internal ties as well as the actual and percent of observed external ties for each network relation and according to network actor attribute sub-groups. For example, for the information sharing relation (third column from left in Table 5.16 above), we can see that 85% of the network actors’ ties are with others who share the same role (“whom”) and 15% of their ties are with actors who share the same organization type. In Table 5.16 above, in the same cell for information sharing by network actor role, the E-I Index is -.565 indicating a preference to form information sharing ties with other actors who share the same role. Overall,

Table 5.16 above reports a tendency for the NevadaSOC network actors to form ties with other actors who share the same attributes, particularly those who share the same role, geographic location and organization type. To examine this further, the attributes are broken down further and examined for external versus internal ties. For example, the sub-group of “Role” is broken down to the specific roles to see if there is a tendency to form ties even more specifically to sub-groups.

Table 5.17 below reports those sub-groups and their tendency toward external versus internal ties. The results offer more clarification of the patterns of interaction. It is important to note the data reported in Table 5.17 below are not statistically significant. However, there are some interesting trends of external ties and internal

While Table 5.16 above indicates overall preponderance of internal ties over external, in

Table 5.17 below we can see this number is primarily due to the professional staff and public organization sub-groups’ tendency to form internal ties. Since professional staff

144 are also the greatest number of actors in the network (as reported in Research Question

One above), their results “tip the scale” for the overall results reported in 5.16 above.

Other than professional staff and public organization sub-groups, the other subgroups included in the Table 5.17 below trend toward a preponderance of external ties indicating a more open group.

Table 5.17

External-Internal Index of Ties Between Network Attribute Groups and Network Relationships

Operations Information Resources Trust (1,310 Ties) (874 Ties) (804 Ties) (722 Ties) Attribute Sub-Group E-I Index, Raw Whom Parent .840 .918 .926 .915 Caregiver 1.000 1.000 1.000 1.000 Youth 1.000 1.000 1.000 1.000 Clinician .686 .614 .662 .600 Professional Staff -.178+ -.211+ -.180+ -.150+ Administrator .557 .568 .610 .570

Geo Clark County .226 .172+ .242 .154 Washoe County .531 .439 .350 .389 Rural .750 .647 .657 .636 Statewide .209 .086 .068 .110

Time 1 year or less .643 .615 .664 .603 At least 2 years .686 .692 .678 .765 More than 2 years .659 .673 .658 .733 No longer involved 1.000 1.000 1.000 1.000 Longer than 10 years .234 .243 .205 .154 Not involved 1.000 1.000 1.000 1.000

Org Public -.716+ -.775+ -.745+ -.776+ Non-profit .821 .836 .846 .806 No organization .844 .846 .862 .840

+ Preponderance of internal over external ties (i.e. tendency toward group closure). Permuted 10,000 times.

145

Table 5.17 above reports all sub-groups within the Role attribute tend to form ties with others outside of their group, except for Professional Staff who tend to form internal ties. A score closest to one in table 5.17 above indicates a preponderance of ties with actors who are external to their subgroup. This result is likely influenced by the higher number of professional staff in the network as opposed to the other roles

(i.e. the other roles have to form ties with others outside of their group by default as there are fewer actors who share their same attribute). While further exploration into homophily is necessary such a pattern could mean that professional staff as “front line” workers do not perceive their position to include one of building network ties (i.e. just focusing on getting the “work” done), whereas Administrators, for example, have to be the “face” of the organization in building those collaborative relationships.

This information is important from a collective action perspective as an actors’ decision to participate in the network may be dependent on the sub-groups’ behavior the actor is primarily tied to. Take, for example, the clinician sub-group. In the visual depictions of the network, the clinicians tended to be more spatially located outside or on the edges of the network. If their ties to other clinicians influence a choice to participate, then one can estimate the other clinicians may not perceive a benefit to their participation as the experience of their “friends” in the network is less central to the network anyway. If a sub-group is less likely to be open to other sub-groups, this can result in the “why bother?” or free-rider mentality in collective action groups.

146

Dyadic Level of Measurement

At the dyadic level of measurement, the following reports patterns of relationships by exploring whether the patterns of ties in one matrix is similar to the pattern of ties in another matrix for the respondent matrix data set. In other words, are those who share resources related to those who state they trust each other? In network analysis, the Quadratic Assignment Procedure (QAP) is used to calculate correlations and regressions. According to Borgatti, Everett and Johnson (2013), this procedure can essentially simulate randomization to correlate two matrices by taking one of the matrices and randomly rearranging the rows and matching columns. The

UCINET software can run correlations and regressions using this QAP procedure and it will also run it thousands of times to “stabilize” the p-value (p. 129). For the correlations reported below, the UCINET program permuted the data 10,000 times for each procedure.

Table 5.18 reports the QAP Correlation for the four relationship variables included in the study: operations (working together), sharing information, sharing resources and trust.

147

Table 5.18

Summary of Correlations Between Relational Variables

Operations Information Resources Trust Operations 1.000 .718*** .657*** .604*** Information 1.000 .763*** .703*** Resources 1.000 .657*** Trust 1.000 *** p<.001, **p<.01, *p<.05 Permuted 10,000 times.

Table 5.18 above reports the Pearson correlations among the relational variables of operations, information, resources and trust. The results reveal statistically significant correlations between the kinds of ties present in each of the relational variables. The results support when one of the relationship variables is present, it is likely that the others are also present. In providing a guide for the interpretation of correlation coefficients, Mukaka (2012) offers a “rule of thumb” for interpreting correlation results (p. 71). In this guide, the author describes a magnitude of .30-.50 is considered a low correlation, .50-.70 is a moderate, .70-.90 is a high correlation and above .90 is a very high correlation.

Given this, the magnitude of the correlations in Table 5.18 above are considered to be a moderate to high magnitude. The correlation between the information relationship with the other relationships carries the greatest magnitude at above .70.

This means the network actors’ access to information sharing with other actors is essential to developing the other relationships of working together (operations), sharing

148 resources and trusting one another. The findings are important to the network administrator as an individual who shares one of the network relations with others, is also likely to share the network relations. For example, if you increase interactions between actors (i.e. operations) then you can influence sharing information, resources and trust.

Table 5.19 below reports the Pearson correlations among the relational variables and attribute variables of same geographic location, same organization type, same length of time involved in the Nevada SOC and the same role of who they represent when they participate (i.e. parent, professional, etc.).

149

Table 5.19

Summary of Correlations Between Actor Attributes and Relational Variables

Operations Information Resources Trust Whom .132* .102* .072 .054 Geo .120*** .137*** .130*** .101** Time .091** .068* .059* .062* Org .253*** .247*** .187** .200** *** p<.001, **p<.01, *p<.05 Permuted 10,000 times.

The results in Table 5.18 above indicate a significant relationship among the network actors’ attributes and their network relationship ties. However, given

Mukaka’s (2012) guidelines for interpreting correlations, the magnitude of the correlations reported above is low. With regard to correlations, the relationship types remain more closely correlated to one another as opposed to a correlation between the attributes of the actor to the relationship types.

Predicting Relationships

To further understand the patterns of interactions, each relationship variable in the respondent data set is treated as a dependent variable to regress on the other variable matrices as independent. Table 5.20 below reports the findings of the regressions for each relationship variable. The regressions examine the extent to which having the same attribute variables and ties in the other relationship variables predict a certain relationship behavior. The same permutation procedures described for the QAP

150 analysis above is utilized in calculating the regressions. For the regressions reported below, the UCINET program permuted the data 10,000 times for each procedure.

Table 5.20

Network Relationship Regressions

Dependent: Operations Information Resources Trust

Model Fit (.563) (.704) (.627) (.541) (R-Squared) Unstandardized Unstandardized Unstandardized Unstandardized Coefficient Coefficient Coefficient Coefficient Same Whom .049* .009 -.009 -.025 Same Geo .011 .019 .020* -.004 Same Time .049* .003 -.004 .006 Same Org .066* .040* -.018 .010 Operations ___ .259*** .159*** .121*** Information .477*** ___ .468*** .367*** Resources .255*** .408*** ___ .238*** Trust .173*** .285*** .213*** ___ *** p<.001, *p<.05 Number of observations = 2,550 Permuted 10,000 times.

Overall, the models above indicate having a relationship tie in the other variables significantly predicts having the targeted tie for the dependent variable. We see a weak but statistically significant effect. The R-squared values indicate the models presented above predict a high amount of variance. Additionally, actors from the same organization, same amount of time and the same role (i.e. professional staff or parent)

151 also suggest weak but statistically significant connection to the Operations network relation as a dependent variable.

Summary of Results for Question Two

The findings presented above reports of the patterns of relationships that comprise the network structure at the whole network or global level, at the triadic level, dyadic and at the individual levels (research question two). The results of the visual network depictions support the data analyses as well. Table 5.21 below summarizes the whole network level results reported above. The findings are then interpreted in Figure

5.17 below according to the emergent and strengthened relationship framework discussed above.

Table 5.21

Summary of the Nevada SOC Whole Network Level Findings

Operations Information Resources Trust

Centralization .502 .630 .664 .692 (outdegree) Centralization .462 .488 .318 .325 (indegree) Density .407 .262 .229 .201 Reciprocity .586 .529 .450 .418 Transitivity .659 .549 .545 .514

As discussed earlier, Provan and Lemaire (2012) describe the multi-layer types of relationships occurring within a network noting that emergent types of relationships are those initially formed within a network such as working together and sharing information. The authors then describe certain network relationships such as trust

152 becoming strengthened and then sustained. This perspective is consistent with the findings of this study in that the operations and information relationships in the Nevada

SOC are emergent types and indicate a network structure that is moderately dense with decreased centralization, consistent with a network structure in a planning stage or still developing (Kolpakov, 2012). Whereas, the resource sharing and trust relationships indicate a network structure at an implementation stage of a policy (Kolpakov, 2012;

Kolpakov, Agranoff & McGuire, 2016). Figure 5.17 summarizes the findings according to the emergent and strengthened relationship framework.

Figure 5.17 Nevada SOC Network Summary of Emergent to Strengthened Relationship Tie Findings

Sharing Information Operations (influences Sharing (working Trust presence of Resources together) other network relations)

Evidence of a network structure Evidence of a developed that is still developing under a network structure under a lead

shared governance model. organization model. • Increased density • Decreased density

• Decreased centralization • Increased centralization • High presence of mutual • Higher presence of mutual

triads (sustainability) triads (sustainability)

Emergent Relationships Strengthened Relationships

153

As depicted in Figure 5.17 above, the findings suggest that there is not one structure for the Nevada SOC network. Rather, there are two types of network governance models and sub-structures that are dependent on the type of network relation assessed. The higher density and lower centralization for the emergent network relations is indicative of a shared-governance model; whereas, the network structure found for the strengthened relations exhibits higher density and lower centralization is indicative of a lead organization model (Provan & Kenis, 2007). I argue this is because, the Nevada SOC network has been operating formally and informally for a long period of time. Additionally, the mandate of the current funding source of the

Nevada SOC is to fund implementation. So, in a sense, we see network members coming together to make that happen even as they are also still building their network.

As indicated by the respondents’ length of time they are involved in the Nevada SOC, we had one group involved for more than 10 years and then another group involved for one year or less. This could explain the presence of a developing network for some relationships and a more centralized and less dense network for others.

Research Question Three

The following reports the patterns of relationships amongst the network actors according to their perception of effectiveness. Survey respondents were asked to rate their perception of how effective or ineffective the Nevada SOC has been. Figure 5.18 below presents the survey findings of this item.

154

Figure 5.18: Network Actors’ Perception of Effectiveness (N=46)

20 19 18 16

14 13 12

10 8 8

6 4 4 2 2

0 Very Effective Somewhat Neutral Somewhat Very Ineffective Effective Ineffective

Half of the survey respondents report they feel the Nevada SOC is effective with

41% reporting it is “somewhat effective” and 9% reporting “very effective.” This is followed by those who remained “neutral” (28%) or found it to be “somewhat ineffective” (17%) or “very ineffective” (4%). It is important to note the number of respondents who skipped this question is higher than the other questions (15 respondents skipped). This could either be due to the sensitive nature of the question and/or the question was the last item on the survey and the respondents had experienced fatigue by the time they reached the question. The number of respondents selecting “neutral” is notable as well and could also be attributed to the sensitive nature of the question. Table 5.22 below presents the crosstabulation of respondents’

155 perception of network effectiveness by their identified role within the Nevada System of

Care.

Table 5.22

Survey Respondent Attributes – Perception of Effectiveness by Respondent Role

Parent Caregiver Youth, Clinician Professional Administrator 18+ Staff (N=3) (N=1) (N=1) (N=5) (N=26) (N=10) Very Effective 1 ------1 2 (N=4) 25% 25% 50% Somewhat 2 1 -- 1 8 7 Effective 11% 5% 5% 42% 37% (N=19) Neutral -- -- 1 3 8 1 (N=13) 8% 23% 62% 8% Somewhat ------1 7 -- Ineffective 13% 87% (N=8) Very ------2 -- Ineffective 100% (N=2) Item response rate, N=46 Percent totals by row.

As presented in Table 5.22 above, the majority of respondents who perceive the

Nevada SOC network as very or somewhat effective identify as professional staff and administrators. Given the centrality of actor types within the network, it appears the actor’s position in the network provides them with greater access to information and resources, which may influence their perception of effectiveness. Of the respondents who stated the Nevada SOC is “somewhat ineffective,” 87% are professional staff. So, the access to information and resources can also negatively influence perception of effectiveness. The majority of clinicians remained “neutral” on their perception of

156

effectiveness. Notably, this is a sub-group that was somewhat disconnected from the

overall network.

Table 5.23 below presents the cross tabulation of respondents’ perception of

network effectiveness by their identified role within the Nevada System of Care.

Table 5.23

Survey Respondent Attributes – Perception of Effectiveness by Length of Time

10 years 2 years 2 years 1 Year No Not or longer or longer (N=10) or Less longer involved (N=13) (N=8) (N=11) involved (N=2) (N=1) Very Effective 2 2 ------(N=4) 50% 50% Somewhat Effective 5 -- 7 6 -- -- (N=19) 27% 37% 32% Neutral 6 2 1 2 1 1 (N=13) 46% 15% 8% 16% 8% 8% Somewhat -- 2 2 3 -- 1 Ineffective 25% 25% 38% 13% (N=8) Very Ineffective -- 2 ------(N=2) 100% Item response rate, N=45 Percent totals by row.

According to Table 5.23 above, respondents who report their perception of

effectiveness as “very effective” have also been affiliated with the Nevada SOC for a

longer length of time which would afford them some history of its efforts and progress

toward achieving goals. Respondents who rate the network as “somewhat effective” or

somewhat ineffective” have been involved in the efforts for a shorter length of time.

157

To examine the actors’ perception of effectiveness from the context of their patterns of relationships, the network relationships are diagramed with UCINET’s

NetDraw program (see Figures 5.19-5.26 below). The basic network view and Gower views (for spatial representation) are presented according to each of the four relationship variables (operations, sharing information, sharing resources, trust). The actor’s degree centrality is included in the depictions and is evident by the size of the node. The actors’ perception of effectiveness as an attribute that is included in the diagram and denoted by colors (blue is “very effective,” light green is “somewhat effective,” yellow is “neutral,” orange is “somewhat ineffective” and red is “very ineffective). The following presents varied views of the four network relations with the inclusion of the actors’ perception of effectiveness. The visual depictions below reveal a tendency for actors who are more central within the network (depicted as a larger node size) to perceive the Nevada SOC as effective, while actors who are less central perceive the network as less or not effective.

158

Figure 5.19 Operations Relationship with Effectiveness Attribute

Blue = Very Effective Green= Somewhat Effective Yellow = Neutral Orange = Somewhat Ineffective Red = Very Ineffective Black = No response

*Size of node indicates number of connections.

159

Figure 5.20 Operations Relationship with Effectiveness Attribute, Gower View

Blue = Very Effective Green= Somewhat Effective Yellow = Neutral Orange = Somewhat Ineffective Red = Very Ineffective Black = No response

*Size of node indicates number of connections.

160

Figure 5.21 Information Relationship with Effectiveness Attribute

Blue = Very Effective Green= Somewhat Effective Yellow = Neutral Orange = Somewhat Ineffective Red = Very Ineffective Black = No response

*Size of node indicates number of connections.

As evident in Figure 5.21 above and Figures 5.23-5.26 below, the actors with higher levels of degree centrality in the network generally report a higher perception of effectiveness for the Nevada SOC (displayed by the blue and green colors). In the above depiction, an actor who is central in an information network generally has more access

161 to the flow of information than actors who are less central. Thus, it is logical that they would perceive higher levels of effectiveness based on this access to information.

Figure 5.22 Information Relationship with Effectiveness Attribute, Gower View

Blue = Very Effective Green= Somewhat Effective Yellow = Neutral Orange = Somewhat Ineffective Red = Very Ineffective Black = No response

*Size of node indicates number of connections.

162

Figure 5.23 Resources Relationship with Effectiveness Attribute

Blue = Very Effective Green= Somewhat Effective Yellow = Neutral Orange = Somewhat Ineffective Red = Very Ineffective Black = No response

*Size of node indicates number of connections.

163

Figure 5.24 Resources Relationship with Effectiveness Attribute, Gower View

Blue = Very Effective Green= Somewhat Effective Yellow = Neutral Orange = Somewhat Ineffective Red = Very Ineffective Black = No response

*Size of node indicates number of connections.

164

Figure 5.25 Trust Relationship with Effectiveness Attribute

Blue = Very Effective Green= Somewhat Effective Yellow = Neutral Orange = Somewhat Ineffective Red = Very Ineffective Black = No response

*Size of node indicates number of connections.

165

Figure 5.26 Trust Relationship with Effectiveness Attribute, Gower View

Blue = Very Effective Green= Somewhat Effective Yellow = Neutral Orange = Somewhat Ineffective Red = Very Ineffective Black = No response

*Size of node indicates number of connections.

As depicted in the images above, the more central actors perceive higher levels of effectiveness than those who are not as central. Some clustering can be seen of actors that share similar perceptions of effectiveness. In general, it appears that access

166 to other actors, information and resources may impact one’s perception of effectiveness. Additionally, note that one of the more central network actors perceives the Nevada SOC to be very effective. This would make sense as they likely have the benefit of full knowledge of the network and its resources.

167

Chapter Six: Discussion, Limitations and Conclusion

This study described the network structure present in the implementation of the

Nevada SOC. Embracing a networked governance approach to policy implementation carries potential for the network administrator to draw upon the rich literature in network management in order to strengthen the overall functioning of the network and achieve its goals. By doing so, the implementing organization acknowledges the power and influence the network structure, as well as the individual actors within it, has to shape the opportunities and constraints of implementation. In turn, this increases our understanding of the process of implementation and to evaluate how the process impacts the targeted outcomes. As presented in the literature review, some core concepts and assumptions of governance networks are: actors, interdependency and frames; interactions and complexity; institutional features and network management

(Klign and Koppenjan, 2012, p. 591). The following discusses the results of this study according to those core concepts.

Actors, interdependency and frames. This core component is addressed by the first question in this study: Who are the network actors in the implementation of the

Nevada System of Care? As reported above, the majority of network members are professional staff, represent public organizations, and identify as representing Clark

County or a Statewide perspective when they participate in the Nevada SOC network.

The study revealed a centralized “team” of actors evident in all four network relations.

The members of the team also had high “betweenness centrality” scores, which means they also play a role in the network as bridges between other sets of actors. The team

168 represents the non-profit and public organization sectors as well as serving in the roles of parent, caregiver, professional staff and administrator. Thus, the team is poised to serve collaboratively as the “network manager” for the Nevada SOC. This is consistent with the SOC as an evidence-based framework which sets forth inclusion of youth and families in decisions about the system changes as a critical principle of the approach.

Interactions and complexity. The core component of interactions and complexity is addressed by the second research question in this study: What are the patterns of interactions within and amongst these actors? As described in Chapter Five above, the Nevada SOC network is not just one structure. Rather, it is multiple structures operating across various levels of analysis and across network relations. As such, the higher density and decreased centrality observed in the emergent network relations of operations and sharing information indicate a network structure similar to a planning stage in network formation (Kolpakov, 2012). The decreased density and increased centrality observed in the strengthened network relations of sharing resources and trust indicates a network structure similar to an implementation stage and is considered more likely to be effective (Contracter, Wasserman & Faust, 2006;

Kolpakov, 2012). The study also revealed a tendency toward group closure specifically by professional staff and those who represent the public sector. This is an indication that the network actors tend to form ties with others who share similar attributes.

Institutional features and network management. This core feature of networks is addressed by research question two (named above) and research question three: To what extent do these patterns of interaction contribute to the perceived outcomes of

169 the policy? In looking at the institutional features, the strengthened network relation of trust is less dispersed, which is evidence of a lead organization model and has a higher chance of realizing network outcomes (Provan & Kenis, 2007; Klaster, Wilderom &

Muntslag, 2017). Additionally, the results of the triadic census found a high number of fully mutual triadic types across all four network relations indicating a network structure that is resistant to contextual factors (Kolpakov, 2012). The high number of fully mutual triads is indicative of a network structure that is “stable at their core” (Provan &

Lemaire, 2012).

While network leadership can arise from the core team and from the network’s lead organization, caution should be taken as there is a risk of the “shadow of hierarchy” stifling the horizontal function and benefits of the network (Klaster, Wilderom &

Muntslag, 2017, p. 678). In short, the risk is that there is an appearance of network governance when the reality is traditional hierarchical function remains. For example, if network actors are heavily engaged in the design of a program and then suddenly the lead organization returns to the network with a different program design and a decision that has already been made; network members will begin to lose trust, which will impact information and resource sharing. From a collective action theory perspective, the risks will influence the context by which members make their decisions on whether or not to participate and their perception of the effectiveness of the network. The findings also indicate that actors who are more central in the network or clustered with central actors report a higher perception of network effectiveness. Thus, the public entity serving as the lead organization can enact network management approaches that

170 balance retaining their centric tasks, but implements them in a “conductive” manner

(Agranoff, 2011).

Network analysis findings are particularly useful to the network manager as the coordinating entity. Given the SOC framework demands resource alignment across systems, efforts can be taken to increase the ties amongst actors in the resource sharing and trust relationship variables. The correlation results indicate when one of the network relation ties is present, it is likely that the others are present as well.

Specifically, as reported in Table 5.18 above and illustrated in Figure 5.17 above, the information sharing network relation is highly correlated with the presence of the other network relations. Additionally, the regression results indicate the presence of the relationship ties is related to a presence in the targeted relationship tie. This means that if the “team” and lead organization implement strategies to increase the information and resource ties, then they will also increase trust. Conversely, if they increase trust ties, then they will likely increase resource sharing.

Limitations

As a field study design, “there is less control over various threats to the studies’ reliability and validity” (Borgatti, Everett & Johnson, 2013, p. 25). Therefore, the following highlights some of the limitations of the study.

Missing data. The distribution of the survey relied on sending electronic mail to recipients; therefore, missing or inaccurate electronic mail addresses accounted for some network respondents not receiving the survey link. Additionally, the initial survey

171 was distributed via SurveyMonkey. There is a chance this distribution strategy caused some of the messages to be determined as “junk mail” by the recipients’ electronic mail systems. Missing data also occurs when respondents refuse to participate in the survey or did not answer all questions in the survey. Missing data can impact the accuracy of reporting ties between actors in the network.

Respondent accuracy. The design of this survey asks respondents to recall information about their relationships with other actors in the network. Relying on respondent recall can present limitations as the respondent may not accurately recall the information (Wasserman & Faust, 1994). Additionally, some respondents may not have felt safe in truthfully responding to the perception of effectiveness survey item. As many of the respondents were public employees with knowledge that their supervisors were also included in the survey; they may have not truthfully responded to the question. In the future, the findings may be more robust if the participants would have been asked a question of expectancy as opposed to direct effectiveness (i.e. “how hopeful are you that the System of Care will achieve its goals”).

Point in time. The data in this study were collected at a point in time. Thus, it only represents the actors and ties at the point of time this study is conducted. As noted above, the Nevada SOC has evolved over time and will continue to do so.

Repeating this study over time and collecting retrospective data will depict a more accurate network structure as it changes over time. Importantly, at the time of this study the Nevada SOC network is grant-funded. Once the funding ends, the sustainability of the network and its actors will be tested and there will likely be a shift

172 in overall network structure as the actors maneuver their position to adjust to the new context; this would be an opportune time to conduct a follow-up study of the Nevada

SOC network.

Descriptive, sociometric approach. Most network analysis studies in political science utilize a descriptive sociometric approach (Ward, Stovel & Sacks, 2011). This study did the same. Given the majority of network members are from public organizations, Clark County and self-identified as professional staff involved in the implementation of the Nevada SOC, there was a limitation in analyzing group cohesion as measured by the extent to which actors were tied to other actors outside of their sub-groups. The attribute variables included in the survey did not produce enough descriptors to fully assess for homophily within the network. There simply was not enough subgroup variation to get a strong measure of this. Future research could aim to define more “hidden” subgroups to run such tests, which would allow for further predictive tests of network relations. Further analysis of the networked approach in policy implementation should aim to continue conducting multi-level network analysis in order to get a fuller picture of the network structure (Contractor, Wasserman &

Faust, 2006).

Single item measures. Most of the measures included just one or two survey items. This presents some concerns with the validity and reliability of the results.

Further, this study primarily described “soft indicators” of network effectiveness such as trust and frequency of contact. Further analyses and data collection are necessary to

173 analyze the “hard indicators” such as goal attainment, innovation and efficiency (Klaster,

Wilderom & Muntslag, 2017, p. 677).

Conclusion

This study began on the premise that the type of governance structure a public entity employs influences how the policy or program is implemented, which then influences outcomes achieved. The concept of networked governance provides a foundation from which to explain policy implementation from a networked perspective.

This study asserts the patterns of relationships among actors within an implementation network comprise the overall structure of the network. The resulting network structure(s) shapes the incentives, opportunities and constraints that become the context of the policy or program implementation. In order to fully understand the extent to which a network structure contributes to the attainment of goals, it is necessary to employ network analysis techniques to describe the patterns of relationships within the network.

This study described the Nevada SOC network at a point in time the network is implementing a SOC in the state. Given the SOC approach is a “framework” consisting of principles, values and a recommended array of services, the implementation can vary from state to state and community to community and is responsive to the implementation context. The framework is useful as it is adaptable given the nature of children’s mental health as a “wicked problem.” However, empirically analyzing the efforts of the Nevada SOC within a changing environment is difficult. While it is

174

Nevada’s goal to involve stakeholders in the SOC implementation, the outcome is influenced by the intensity of involvement and relationship patterns within the network.

This study identified the actors involved in the network, described the patterns of interaction among those actors and described the actors’ perception of the overall effectiveness of the Nevada SOC. The Nevada SOC network is operating under two network model types; shared-governance and lead organization. For the shared- governance model, the network is moderately dense at the emergent network relations of operations and sharing information. It becomes sparser and more centralized under the lead organization model at the strengthened network relations of sharing resources and trust. Overall the Nevada SOC network has a “team” of central actors consisting of a parent, care giver, professional staff and administrator. This team maintained high centrality scores across all four of the measured network relations.

The Nevada SOC network is an “action network” type (Agranoff, 2007) and its implementation is most closely aligned with Matland’s (1995) experimental implementation type in that implementation is highly dependent on the context of the implementation environment and actors involved. The Nevada SOC’s central team of actors, by way of their positions in the network, can greatly influence the contextual environment. Under this implementation type, the policy ambiguity is high while policy conflict is low. From the perspective of collective action theory, these contextual conditions become the environment in which the individual actors conduct their cost- benefit assessment in making decisions on the extent to which and how they participate in the network. Additionally, the perceptions of those they are tied to within the

175 network also contribute to the contextual conditions of their decisions to participate or not.

While this study contributes to the literature by employing a mixed-method and multi-level network analysis, further research can examine the network structures according to the framework of emergent and strengthened relationship types. Data collected at additional points in time would confirm the network structures over time or describe how they have evolved. Additionally, the balance theoretical models did not apply to the findings of the triadic census due to the presence of a “forbidden” triad type (type 11-201). As this triad consists of two mutual relationships, zero asymmetric relationships and one null or non-existing relationship, the triadic structure is considered open. Burt’s (1992) “structural holes” theory may offer additional explanation of the presence of open triads in a network. The structural holes are important in a network as an opportunity for boundary spanners to broker new ideas between network sub-groups (Provan & Lemaire, 2012). Thus, further analysis of the structural holes evident in the Nevada SOC network may reveal additional explanations of the network actors’ ties.

Goggin (1986) describes a distinction between policy implementation processes and policy outcomes. As such, further study is needed to examine the extent to which network structure predicts network outcomes. Is a dense or centralized network more likely to achieve outcomes than a sparse and de-centralized network? For example, the structure of the network itself (i.e. a circle, chain or star pattern) influences the speed at which information and resources can flow. Additionally, central actors have the power

176 to influence incentives, create opportunities or enforce constraints on the others in the network. Thus, to what extent is this effective?

The patterns of interaction analyzed in this study indicate the source of power and influence over opportunities and constraints of implementation is centralized to a few actors within the public system. From a collective action perspective, those few actors become the catalyst through which to engage the other actors in collectively addressing societal issues. For the Nevada SOC, this role of centralized actors and network managers is a different skill set than what may be typical for the lead organization. If the DCFS were to fully embrace this networked governance perspective, training the “team” and the other professional staff in network management skills carries the potential to increase the overall effectiveness of the network.

Applying techniques of network analysis to policy implementation research draws upon the theoretical developments from the 1980s (i.e. top down vs. bottom up approaches) and utilizes technological advancements to understand how a network structure can shape the incentives, opportunities and constraints within the implementation of a public policy. By employing network analysis techniques to the study of policy implementation, we can pull the complexity of the actors’ interactions into a separate study that can, through multi-level network analysis, be used to explain and predict network behaviors.

While network analysis is beneficial in describing the patterns of relationships amongst those who are involved with the implementation of policies, policy implementation is still impacted by other contextual factors such as training, formal

177 rules and evaluation (Mischen & Jackson, 2008). With the differing types of implementation models, network types, and actor relationships, network analysis still faces the problem of not just having too many variables, but a need to develop more standardized measures of network relations in policy implementation research. For example, this study analyzed four network relations (operations, sharing information, sharing resources and trust). There are other variables that can also play a predictive role in the structure of the network such as history of working together. Further research is necessary to identify specific network relations that should be included in network analysis studies as well as the development of standardized measures for those network relations.

The results of this study benefit the organizing agency in that, as recommended by O’Toole (2014), public administrators “would be well advised to conduct regular scanning of the networks in which (or with which) they work to inventory their principal contingencies and alliances, including indirect ties” (p. 367). To further analyze the network, Provan, Veazie, Staten and Teufel-Shone (2005), present a set of eight

“questions for communities based on network analysis” (See Appendix G). The questions prompt a more in-depth examination of the network relationships. This would be beneficial in future research as more detailed inclusion of actor attributes will offer opportunities to measure predictive relationship patterns.

This study demonstrates that even when implementation of a public policy or program happens through a “networked array” of actors, there is still a need to have a centralized network that is managed for efficiency and flow of relationships. While the

178 public sector may be getting out of the business of directly implementing the specific activities of a program, I do not believe this will limit government. Rather, there will be a shifting role of government toward becoming lead organizations in networked implementation. From a networked governance perspective, the public entity becomes the lead organization that serves as the conductor who articulates the common interests of the citizens, unites the network actors and sets the standards for the collective action of both public and private sectors who organize to address wicked social problems. In this model, public employees assume roles of centralized actors and networked managers. Thus, it is the collective work of the network actors in achieving the outcomes and is not simply dictated by or solely implemented by one public entity.

So, if networked governance becomes the standard perspective by which policies are implemented, then this first step in empirically describing the network structure of policy implementation becomes the foundation for the next research question, which is

“under what conditions does each governing structure work effectively” and efficiently

(Rhodes, 1996).

179

References

Agranoff, R. (2007). Managing within networks: Adding value to public organizations.

Washington, D.C.: Georgetown University Press.

Agranoff, R. (2011). Collaborative public agencies in the network era. In Menzel, D.,

White H. (Eds.). The state of public administration: Issues, Challenges and

opportunities (pp. 272-294). New York, NY: Taylor & Francis.

Agranoff, R. (2012). Collaborating to manage: A primer for the public sector.

Washington, D.C.: Georgetown University Press.

Agranoff, R. (2014). Local governments in multilevel systems” Emergent public

administration challenges. American Review of Public Administration, 44(4S),

48S-62S.

Agranoff, R. (2017). Crossing boundaries for intergovernmental management.

Washington, D.C.: Georgetown University Press.

Agranoff, R. McGuire, M. (2001). After the network is formed: Process, power, and

performance. In Mandell, M. (Ed.). Getting results through collaboration:

Networks and network structures for public policy and management (pp. 11-29).

Westport, CT: Quorum Books.

Bevir, M. (2012). Governance: A very short introduction. Oxford, UK: Oxford University

Press.

Birkland, T. (2001). An Introduction to the Policy Process: Theories, Concepts, and

Models of Public Policy Making. Armonk, NY: M.E. Sharpe, Inc.

180

Borgatti, S.P, Everett, M.G., Freeman, L. C. (2002). UCINET 6 for Windows: Software for

social network analysis. Harvard, MA: Analytic Technologies.

Borgatti, S.P, Everett, M.G., Johnson, J. (2013). Analyzing social networks. Los Angeles,

CA: SAGE Publications.

Borgatti, S.P. (2002). NetDraw network visualization. Harvard, MA: Analytic

Technologies.

Borgatti, S.P. (2008). Network Data Collection: MGT 780, Spring 2010. Power Point

Presentation retrieved from

http://www.analytictech.com/mgt780/slides/survey.pdf.

Borgatti, S.P. (2016, July 6). Import survey data from Excel [video file]. Retrieved from

https://www.youtube.com/watch?v=iwXo6EXYrQs.

Borgatti, S.P. (2018a, February 24). mgt 780 centrality 1 [video file]. Retrieved from

https://www.youtube.com/watch?v=FKM0f9QzlhU.

Borgatti, S.P. (2018b, March 7). Class feb 23 2016 hyp testing I [video file]. Retrieved

from https://www.youtube.com/watch?v=Zt6BopIDbqk&t=1668s.

Borgatti, S.P., Halgin, D. (2011). Analyzing affiliation networks. In Scott, J., & Carrington,

P. (Eds.), The SAGE handbook of social network analysis (417-433). Los Angeles,

CA: SAGE Publications.

Burt, R. (1992). Structural holes: The social structure of competition. Cambridge, MA:

Harvard University Press.

181

Chubb, J. (1985). The political economy of federalism. The American Political Science

Review 79(4), 994-1015.

Cigler, B. (2001). Multiorganizational, multisector, and multicommunity organizations.

In Mandell, M. (Ed.). Getting results through collaboration: Networks and

network structures for public policy and management (pp. 11-29). Westport, CT:

Quorum Books.

Contractor, N., Wasserman, S., Faust, K. (2006). Testing multitheoretical, multilevel

hypotheses about organizational networks: An analytical framework and

empirical example. The Academy of Management Review, 31(3), 681-701.

Cristofoli, D., Markovic, J., Meneguzzo, M. (2014). Governance, management and

performance in public networks: How to be successful in shared-governance

networks. Journal of Management and Governance, 18(1), 77-93.

Dekker, D., Krackhardt, D., Snijders, TA (2007). Sensitivity of MRQAP tests to collinearity

and autocorrelation conditions. Psychometrika, 72(4), 563-581.

DeLeon, P., DeLeon, L. (2002). What ever happened to policy implementation? An

alternative approach. Journal of Public Administration Research and Theory,

12(4), 467-492.

DeMarrais, E., Earle, T. (2017). Collectie action theory and the dynamics of complex

societies. Annual Review of Anthropology, 46. 183-201.

182

Dyer, J., Chu, W. (2003). The role of trustworthiness in reducing transaction costs and

improving performance: Empirical evidence from the United States, Japan, and

Korea. Organization Studies, 14(1), 57-68.

Goggin, M. (1986). The “too few cases/too many variables” problem in implementation

research. The Western Political Quarterly, 39(2), 328.347.

Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology,

78(6), 1360-1380.

Gross Stein, J. (2009). The politics and power of networks: The accountability of

humanitarian organizations. In Kahler, M. (Ed.), Networked politics: Agency,

power and governance. Ithica, NY: Cornell University Press.

Hall, T., O’Toole, L. (2000). Structures for policy implementation: An analysis of national

legislation, 1965-1966 and 1993-1994. Administration & Society, 31(6), 667-686.

Hanneman, R.A., Riddle, M. (2011a). A brief introduction to analyzing social network

data. In Scott, J., & Carrington, P. (Eds.), The SAGE handbook of social network

analysis (331-339). Los Angeles, CA: SAGE Publications.

Hanneman, R.A., Riddle, M. (2011b). Concepts and measures for basic network

analysis. In Scott, J., & Carrington, P. (Eds.), The SAGE handbook of social

network analysis (332-369). Los Angeles, CA: SAGE Publications.

Hanneman, R.A., Riddle, M. (2005). Introduction to social network methods. Riverside,

CA: University of California, Riverside (published in digital form at

http://faculty.ucr.edu/~hanneman/ )

183

Hemphill, J. (2003). Interpreting the magnitudes of correlation coefficients. American

Psychologist, 58(1), 78-80.

Institute for Mathmatics, Physics and Mechanics, Ljubljana, Slovenia. Department of

Theoretical Computer Science (date unknown). Triadic census: All possible triads

(Wasserman & Faust), sequence numbers as used in Pajek. Retrieved from

http://www.educa.fmf.uni-lj.si/datana/pub/networks/pajek/triade.pdf

Hooghe, L., Marks, G. (2003). Unraveling the central state, but how? Types of multi-

level governance. American Political Science Review, 97(2), 233-243.

Kapucu, N., Hu, Q., Khosa, S. (2017). The state of network research in public

administration. Administration & Society, 49(8), 1087-1120.

Klaster, E., Wilderom, C., Muntslag, D. (2017). Balancing relations and results in regional

networks of public-policy implementation. Journal of Public Administration

Research and Theory: J-PART, 27(4), p. 676-691.

Klign, E., Koppenjan, J. (2012). Governance network theory: past, present and future.

Policy & Politics, 40(4), 587-606.

Knoke, D. (2011). Policy networks. In Scott, J., & Carrington, P. (Eds.), The SAGE

handbook of social network analysis (pp. 340-369). Los Angeles, CA: SAGE

Publications.

Kolpakov, A. (2012). Structural development of public management networks over time:

Where process meets structure (Doctoral dissertation). Ann Arbor, MI: ProQuest

LLC.

184

Kolpakov, A., Agranoff, R., McGuire, M. (2016). Understanding interoperability in

collaborative network management: The case of Metro High School. Journal of

Health Science, 4, 318-332.

Lecy, J., Mergel, I., Schmitz, H. (2014). Networks in public administration. Public

Management Review, 16(5), 643-665.

Mann, C., Hyde, P. (2013). Joint CMCS and SAMHSA informational bulletin: Coverage of

behavioral health services for children, youth, and young adults with significant

mental health conditions. Substances Abuse and Mental Health Services

Administration and Center for Medicaid and CHIPS Services. Retrieved February

7, 2018 from: https://www.medicaid.gov/federal-policy-

guidance/downloads/cib-05-07-2013.pdf.

Marsden, P. (2011). Survey methods for network data. In Scott, J., & Carrington, P.

(Eds.), The SAGE handbook of social network analysis (340-369). Los Angeles, CA:

SAGE Publications.

Marwell, G., Oliver, P., Prahl, R. (1988). Social networks and collective action: A theory

of the critical mass. American Journal of Sociology, 32, 503-534.

Matland, R. (1995). Synthesizing implementation literature: The ambiguity-conflict

model of policy implementation. Journal of Public Administration Research and

Theory, 5(2), 145-174.

Milward, H., Provan, K. (1998). Measuring network structure. Public Administration, 76,

387-407.

185

Milward, H., Provan, K., Fish, A., Isett, K., Huang, K. (2010). Governance and

collaboration: An evolutionary study of two mental health networks. Journal of

Public Administration Research and Theory: J-PART, 20(supp. 1), i125-i141.

Mischen, P., Jackson, S. (2008). Connecting the dots: Applying complexity theory,

knowledge management and social network analysis to policy implementation.

Public Administration Quarterly, 32(3), 314-338.

Monge, P.R., Contractor, N.S. (2003). Theories of communication networks. New York,

NY: Oxford University Press.

Mrvar, A., Batagelj, V. (2016). Analysis and visualization of large networks with program

package Pajek. Complex Adaptive Systems Modeling, 4(6), 1-8.

Mulvaney, K.K., Lee, S., Hook, T.O., Prokopy, L.S. (2015). Casting a net to better

understand fisheries management: An affiliation network analysis of the Great

Lakes Fishery Commission. Marine Policy, 57, 120-131.

Mukaka, M.M. (2012). Statistics corner: A guide to appropriate use of Correlation

coefficient in medical research. Malawi Medical Journal, 24(3), 69-71.

Nevada Executive Order No. 2013-26 (2013). Retrieved February 7, 2018 from

http://gov.nv.gov/News-and-Media/Executive-Orders/2013/EO_-2013-26-

Establishing-the-Governor_s-Behavioral-Health-and-Wellness-Council/

Nevada Revised Statutes § 433.316 (2013).

186

Nooy, W., Mrvar, A., Batagelj, V. (2011). Exploratory social network analysis with Pajek.

New York, NY: Cambridge University Press.

Olson. M. (1965). The logic of collective action: Public goods and the theory of groups.

Cambridge, MA: Harvard University Press.

O’Toole, L. (1997). Treating networks seriously: Practical and research-based agendas

in public administration. Public Administration Review, 57(1), 45-52.

O’Toole, L. (2014). Networks and networking: The public administrative agendas.

Public Administration Review, 75(3), 361-371.

Or, N., Aranda-Jan, A. (2017). The dynamic role of state and nonstate actors:

Governance after global financial crisis. Policy Studies Journal, 45(S1), S67-S81.

Peters, G. (2011). Governance. In Berg-Schlosser, D., Badie, B. (Eds). International

Encyclopedia of Political Science, Volume 1-8, Sage. EBook, EBSCO Publishing.

Retrieved via University of Nevada Reno.

Pierre, J. (2011). Governance. In Berg-Schlosser, D., Badie, B. (Eds). International

Encyclopedia of Political Science, Volume 1-8, Sage. EBook, EBSCO Publishing.

Retrieved via University of Nevada Reno.

Pires, S. (2002). Building systems of care: A primer. National Technical Assistance

Center for Children’s Mental Health, Georgetown University Child Development

Center. Retrieved February 7, 2018 from:

https://gucchd.georgetown.edu/products/PRIMER_CompleteBook.pdf

187

Provan, K. & Kenis, P. (2007). Modes of network governance: Structure, management

and effectiveness. Journal of Public Administration Research and Theory: J-

PART, 18(2), p. 229-252.

Provan, K. & Lemaire, R. (2012). Core concepts and key ideas for understanding public

sector organizational networks: Using research to inform scholarship and

practice. Public Administration Review. 72. 638-648.

Provan, K. & Milward, H. (1995). A preliminary theory of interorganizational network

effectiveness: A comparative study of four community mental health systems.

Administrative Science Quarterly, 40, p. 1-33.

Provan, K. & Milward, H. (2001). Do networks really work? A framework for evaluating

public-sector organizational networks. Public Administration Review, 61(4), 414-

423.

Provan, K., Huang, K., Milward, B. (2009). The evolution of structural embeddedness

and organizational social outcomes in a centrally governed health and human

services network. Journal of Public Administration Research and Theory: J-PART,

19(4), 873-893.

Provan, K., Veazie, M., Staten, L. & Teufel-Shone, N. (2005). The use of network analysis

to strengthen community partnerships. Public Administration Review, 65(5),

603-613.

Rawlings, C., Friedkin, N. (2017). The structural balance theory of sentiment networks:

Elaboration and test. American Journal of Sociology, 123(2), 510-548.

188

Rhodes, R.A.W. (1996). The new governance: Governing without government. Policy

Studies XLIV, 652-667.

Sabatier, P. (1986). Top-down and bottom-up approaches to implementation research:

a critical analysis and suggested synthesis. Journal of Public Policy, 6(1), p. 21-48.

Sabatier, P., Leach, W., Lubell, M., Pelkey (2005). Theoretical frameworks explaining

partnership success. In Sabatier, P., Focht, W., Lubell, M., Trachtenberg, Z.,

Vedlitz, A. & Matlock, M. (Eds.), Swimming upstream: Collaborative approaches

to watershed management (pp. 173-199). Cambridge, MA: The MIT Press.

Scott, J., & Carrington, P. (Eds.). (2011). The SAGE handbook of social network analysis.

Los Angeles, CA: SAGE Publications.

State of Nevada, Department of Health and Human Services, Division of Public and

Behavioral Health (2013). Nevada Commission on Behavioral Health and

Developmental Services Bylaws. Retrieved 2/3/18 from:

http://dpbh.nv.gov/uploadedFiles/dpbh.nv.gov/content/Boards/xxx/Bylaws-9-

13.pdf

State of Nevada, Commission on Behavioral Health, Children’s System of Care Behavioral

Health Subcommittee (2016). Nevada System of Care, Implementation Grant

Strategic Plan.

189

State of Nevada, Department of Health and Human Services, Division of Child and Family

Services (DCFS) (2017). 2017 Public Meetings and Notices. Retrieved February,

11, 2018 from: http://dcfs.nv.gov/Meetings/2017/2017_Meetings/

State of Nevada, Department of Health and Human Services, Division of Child and Family

Services (DCFS) (2018). Retrieved February 19, 2018 from: http://dcfs.nv.gov.

State of Nevada, Department of Health and Human Services, Division of Public and

Behavioral Health (DPBH) (2018). Nevada regional behavioral health policy

boards: Fact sheet. Retrieved April 30, 2018 from:

http://dpbh.nv.gov/uploadedFiles/dpbhnvgov/content/Boards/RBHPB/Meetings

/2018/Behavioral%20Health%20Board%20Fact%20Sheet.pdf.

Stroul, B., Friedman, R. (1988a). Principles for a system of care. Children Today, 17(4),

11-15.

Stroul, B., Friedman, R. (1988b). Putting principles into practice. Children Today, 17(4),

15-17.

Stroul, B., Friedman, R. (2011). A report on the study of strategies for expanding the

System of Care: National evaluation of the comprehensive community mental

health services for children and their families program. Atlanta, GA: ICF Macro.

Retrieved June 30, 2018 from:

https://gucchdtacenter.georgetown.edu/publications/SOC%20Expansion%20Stu

dy%20Report%20Final.pdf

190

Substance Abuse and Mental Health Services Administration (SAMHSA), (no date). State

and local government partnerships. Retrieved February 19, 2018 from:

https://www.samhsa.gov/state-local-government-partnerships

Substance Abuse and Mental Health Services Administration (SAMHSA), (2013). The

comprehensive community mental health services for child with severe emotional

disturbances, Evaluation findings: report to Congress. Retrieved February 22,

2018 from: https://store.samhsa.gov/shin/content//PEP15-CMHI2012-

2013/PEP15-CMHI2012-2013.pdf

Substance Abuse and Mental Health Services Administration (SAMHSA), (2015). The

comprehensive community mental health services for child with severe emotional

disturbances program, Report to Congress 2015. Retrieved April 30, 2018 from:

https://www.samhsa.gov/sites/default/files/programs_campaigns/nitt-ta/2015-

report-to-congress.pdf1

The New York Times (1933, April 3). Emotions mapped by new geography. Charts Seek

to Portray the Psychological Currents of Human Relationships. First Studies

Exhibited. Colored Lines Show Likes and Dislikes of Individuals and of Groups.

Many Misfits Revealed. Dr. J.L. Moreno Calculates There Are 10 to 15 Million

Isolated Individuals In Nation. Retrieved January 2, 2018 from:

http://query.nytimes.com/mem/archive-

free/pdf?res=9903E0D7173BEF3ABC4B53DFB2668388629EDE

191

Ward, M., Stovel, K., Sacks, A. (2011). Network analysis and political science. Annual

Review of Political Science, 14, 245-264.

Wasserman, S., Faust, K. (1994). Social network analysis: Methods and applications.

Cambridge, NY: Cambridge University Press.

Weber, E., Khademian, A. (2008). Wicked problems, knowledge challenges, and

collaborative capacity builders in network settings. Public Administration

Review, 68(2), 334-349.

Weible, C., Sabatier, P., McQueen, K. (2009). Themes and variations: Taking stock of the

advocacy coalition framework. The Policy Studies Journal, 37(1), 121-140.

Weible, C., Sabatier, P. (Eds.) (2018). Theories of the Policy Process (4th ed.). New York,

NY : Routledge.

Yang, S., Keller, F., Zheng, L. (2017). Social network analysis : Methods and examples.

Thousand Oaks, CA : SAGE Publications, Inc.

Ysa, T., Sierra, V., Esteve, M. (2014). Determinants of network outcomes: The impact of

management strategies. Public Administration, 92(3), 636-655.

192

Appendices

Appendix A: Multilevel, Multitheoretical Framework to Test Hypotheses About Organizational Networks.

Contractor, N., Wasserman, S., Faust, K. (2006). Testing multitheoretical, multilevel hypotheses about organizational networks: An analytical framework and empirical example. The Academy of Management Review, 31(3), p.687.

193

Appendix B: System of Care Values and Principles

System of Care Deﬁnition and Philosophy Deﬁnition A system of care is: A spectrum of effective, community-based services and supports for children and youth with or at risk for mental health or other challenges and their families, that is organized into a coordinated network, builds meaningful partnerships with families and youth, and addresses their cultural and linguistic needs, in order to help them to function better at home, in school, in the community, and throughout life.

Core Values Systems of care are: 1. Family driven and youth guided, with the strengths and needs of the child and family determining the types and mix of services and supports provided 2. Community based, with the locus of services, as well as system management, resting within a supportive, adaptive infrastructure of structures, processes, and relationships at the community level 3. Culturally and linguistically competent, with agencies, programs, and services that reﬂect the cultural, racial, ethnic, and linguistic differences of the populations they serve to facilitate access to and utilization of appropriate services and supports

Guiding Principles Systems of care are designed to: 1. Ensure availability of and access to a broad, flexible array of effective, evidence-informed, community-based services and supports for children and their families that addresses their physical, emotional, social, and educational needs, including traditional and nontraditional services as well as informal and natural supports 2. Provide individualized services in accordance with the unique potential and needs of each child and family, guided by a strengths-based, wraparound service planning process and an individualized service plan developed in true partnership with the child and family 3. Deliver services and supports within the least restrictive, most normative environments that are clinically appropriate 4. Ensure that families, other caregivers, and youth are full partners in all aspects of the planning and delivery of their own services and in the policies and procedures that govern care for all children and youth in their communities, states, territories, tribes, and nation 5. Ensure cross-system collaboration, with linkages between child-serving agencies and programs across administrative and funding boundaries and mechanisms for system-level management, coordination, and integrated care management 6. Provide care management or similar mechanisms to ensure that multiple services are delivered in a coordinated and therapeutic manner, and that children and their families can move through the system of services in accordance with their changing needs 7. Provide developmentally appropriate mental health services and supports that promote optimal social and emotional outcomes for young children and their families in their homes and community settings 8. Provide developmentally appropriate services and supports to facilitate the transition of youth to adulthood and to the adult-service system as needed 9. Incorporate or link with mental health promotion, prevention, and early identification and intervention to improve long-term outcomes, including mechanisms to identify problems at an earlier stage and mental health promotion and prevention activities directed at all children and adolescents 10. Incorporate continuous accountability mechanisms to track, monitor, and manage the achievement of system of care goals; fidelity to the system of care philosophy; and quality, effectiveness, and outcomes at the system level, practice level, and child and family level 11. Protect the rights of children, youth, and families and promote effective advocacy efforts 12. Provide services and supports without regard to race, religion, national origin, gender, gender expression, sexual orientation, physical disability, socioeconomic status, geography, language, immigration status, or other characteristics; services should be sensitive and responsive to these differences

Stroul, B., Blau, G., & Friedman, R. (2010). Updating the system of care concept and philosophy. Washington, DC: Georgetown University Center for Child and Human Development, National Technical Assistance Center for Children’s Mental Health.

Retrieved from: https://gucchd.georgetown.edu/products/Toolkit_SOC_Resource1.pdf

194

Appendix C: Nevada System of Care Network Formation, Abbreviated Timeline

Date or Event Source Approximate Time Period

May 27, 1975 Nevada Revised Statutes, Chapter 745 https://www.leg.state.nv.us created the “Mental Hygiene and Mental /Statutes/58th/Stats197507. Retardation Advisory Board.” html#Stats197507page1593

June 14, 1985 Nevada Revised Statutes, Chapter 672 https://www.leg.state.nv.us created the “Commission on Mental Health /Statutes/63rd/Stats198510. and Mental Retardation” html#Stats198510page2262

June 11, 2013 Nevada Revised Statutes, Chapter 489 https://www.leg.state.nv.us changed the name of the Commission to /Statutes/77th2013/Stats20 “Commission on Behavioral Health” 1318.html#Stats201318page 3003

December 13, Nevada Governor Brian Sandoval http://dpbh.nv.gov/uploade 2013 establishes, by Executive Order, the dFiles/01%20EO_2013- “Governor’s Council on Behavioral Health 26.pdf and Wellness.” He charges the Council with creating a plan and making recommendations on improving the delivery of behavioral health services in the state.

May 22, 2009 Nevada Revised Statutes, Chapter 181 (NRS https://www.leg.state.nv.us 232.482) established a subcommittee of /Statutes/75th2009/Stats20 the Commission on the mental health of 0907.html#Stats200907page children. Set forth the subcommittee shall 662 review the plans of the regional consortia and create a statewide plan.

September 30, State of Nevada, Department of health and 2015 Human Services, Division of Child and Family Services awarded a statewide Systems of Care implementation grant from the Substance Abuse and Mental Health Services Administration

2017 Assembly Bill 366 establishes four regional http://dpbh.nv.gov/uploade behavioral health boards in the state. dFiles/dpbhnvgov/content/B oards/RBHPB/Meetings/201 8/Behavioral%20Health%20 Board%20Fact%20Sheet.pdf

195

Appendix D: Affiliation Network Development, 2017 List of Public Meetings (Events)

Code Name of Event Date

1EBP17 Commission on Behavioral Health, Children’s System of Care 1/3/17 Behavioral Health Subcommittee - Provider Standards and Evidence Based Practices Workgroup

2GOV17 Commission on Behavioral Health, Children’s System of Care 1/4/17 Behavioral Health Subcommittee - Governance Workgroup

3SOC17 Commission on Behavioral Health, Children’s System of Care 1/5/17 Behavioral Health Subcommittee

4COM17 Commission on Behavioral Health, Children’s System of Care 1/6/17 Behavioral Health Subcommittee - Communications Workgroup

5SP17 Commission on Behavioral Health, Children’s System of Care 1/18/17 missing Behavioral Health Subcommittee - Special Populations Workgroup

6EBP17 Commission on Behavioral Health, Children’s System of Care 1/31/17 Behavioral Health Subcommittee - Provider Standards and Evidence-Based Practice Workgroup

7GOV17 Commission on Behavioral Health, Children’s System of Care 2/1/17 Behavioral Health Subcommittee - Governance Workgroup

8COM17 Commission on Behavioral Health, Children’s System of Care 2/3/17 Behavioral Health Subcommittee - Communications Workgroup

9SP17 Commission on Behavioral Health, Children’s System of Care 2/15/17 Behavioral Health Subcommittee - Special Populations Workgroup

10EBP17 Commission on Behavioral Health, Children’s System of Care 2/28/17 Behavioral Health Subcommittee - Provider Standards and Evidence-Based Practice Workgroup

11COM17 Commission on Behavioral Health, Children’s System of Care 3/10/17 Behavioral Health Subcommittee - Communications Workgroup

196

Code Name of Event Date

12SP17 Commission on Behavioral Health, Children’s System of Care 3/21/17 missing Behavioral Health Subcommittee - Special Populations Workgroup

13GOV17 Commission on Behavioral Health, Children’s System of Care 3/22/17 missing Behavioral Health Subcommittee - Governance Workgroup

14EBP17 Commission on Behavioral Health, Children’s System of Care 3/30/17 Behavioral Health Subcommittee - Provider Standards and Evidence-Based Practices Workgroup

15SOC17 Commission on Behavioral Health, Children’s System of Care 4/6/17 Behavioral Health Subcommittee

16COM17 Commission on Behavioral Health, Children’s System of Care 4/21/17 Behavioral Health Subcommittee - Communications Workgroup

17WD17 Commission on Behavioral Health, Children’s System of Care 4/26/17 Behavioral Health Subcommittee - Workforce Development Workgroup

18SP17 Commission on Behavioral Health, Children’s System of Care 4/27/17 Behavioral Health Subcommittee - Special Populations Workgroup

19SP17 Commission on Behavioral Health, Children’s System of Care 5/22/17 Behavioral Health Subcommittee - Special Populations Workgroup

20WD17 Commission on Behavioral Health, Children’s System of Care 5/31/17 Behavioral Health Subcommittee - Workforce Development Workgroup

21SOC17 Commission on Behavioral Health, Children’s System of Care 6/1/17 Behavioral Health Subcommittee

22COM17 Commission on Behavioral Health, Children’s System of Care 6/16/17 Behavioral Health Subcommittee - Communications Workgroup

23WD17 Commission on Behavioral Health, Children’s System of Care 6/28/17 Behavioral Health Subcommittee - Workforce Development Workgroup

197

Code Name of Event Date

24SP17 Commission on Behavioral Health, Children’s System of Care 7/17/17 Behavioral Health Subcommittee - Special Populations Workgroup

25WD17 Commission on Behavioral Health, Children’s System of Care 7/26/17 Behavioral Health Subcommittee - Workforce Development Workgroup

26SOC17 Commission on Behavioral Health, Children’s System of Care 7/3/17 Behavioral Health Subcommittee

27COM17 Commission on Behavioral Health, Children’s System of Care 8/18/17 Behavioral Health Subcommittee - Communications Workgroup

28SP17 Commission on Behavioral Health, Children’s System of Care 8/21/17 Behavioral Health Subcommittee - Special Populations Workgroup

29WD17 Commission on Behavioral Health, Children’s System of Care 8/23/17 Behavioral Health Subcommittee - Workforce Development Workgroup

30SP17 Commission on Behavioral Health, Children’s System of Care 9/18/17 Behavioral Health Subcommittee - Special Populations Workgroup

31WD17 Commission on Behavioral Health, Children’s System of Care 9/27/18 Behavioral Health Subcommittee - Workforce Development Workgroup

32SOC17 Commission on Behavioral Health, Children’s System of Care 10/5/17 Behavioral Health Subcommittee

33SP17 Commission on Behavioral Health, Children’s System of Care 10/16/17 Behavioral Health Subcommittee - Special Populations Workgroup

34COM17 Commission on Behavioral Health, Children’s System of Care 10/20/17 Behavioral Health Subcommittee - Communications Workgroup

198

Code Name of Event Date

35WD17 Commission on Behavioral Health, Children’s System of Care 11/28/17 Behavioral Health Subcommittee - Workforce Development Workgroup

36SOC17 Commission on Behavioral Health, Children’s System of Care 12/7/17 Behavioral Health Subcommittee

37COM17 Commission on Behavioral Health, Children’s System of Care 12/15/17 Behavioral Health Subcommittee - Communications Workgroup

38SP17 Commission on Behavioral Health, Children’s System of Care 12/18/17 Behavioral Health Subcommittee - Special Populations Workgroup

Coding Pattern: Event number + Meeting Name + 2-digit year

Meeting Name Codes Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee = SOC

Commission on Behavioral Health, Children’s System of Care Behavioral Health

Subcommittee - Communications Workgroup = COM

Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee - Workforce Development Workgroup = WD

Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee - Special Populations Workgroup = SP

Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee - Provider Standards and Evidence-Based Practices Workgroup = EBP

Commission on Behavioral Health, Children’s System of Care Behavioral Health Subcommittee - Governance Workgroup = GOV

199

Appendix E: Modified Copy of Nevada SOC Survey (with roster names removed)

Welcome

Thank you for your interest in this study to analyze the network structure of individuals and organizations involved in the implementation of the “Nevada System of Care.” Please refer to the the e-mail for additional information about this study.

Your participation in this study is completely voluntary and confidential. You may stop at any time. Declining to participate or stopping your participation will not have any negative effects on your involvement in the System of Care. Your responses are saved and submitted each time you click the “Next” or “Done” button on each page of the survey. Responses don’t automatically save as each question is answered—they are saved and submitted page by page as you progress through the survey.

In order for this survey to be effective, we need participation from as many people as possible by April 13th.

You may ask questions of the researcher at any time by calling the student research, Jill Manit, at 775-846-5468 or e-mailing at [email protected]. You may contact the Principal Investigator, Dr. William Eubank by sending an email to [email protected].

The researchers and the University of Nevada, Reno will treat your identity and the information collected about you with professional standards of confidentiality and protect it to the extent allowed by law. You will not be personally identified in any reports or publications that may result from this study. The US Department of Health and Human Services, the University of Nevada, Reno Research Integrity Office, and the Institutional Review may look at your study records.

* 1. Clicking on the "agree" button below indicates that: • you have read the above information and the information distributed at a meeting and/or via e-mail • you voluntarily agree to participate • you are at least 18 years of age

If you do not wish to participate in the research study, please decline participation by clicking on the "disagree" button.

Agree

Disagree

200

Characteristics

* 2. What is your first name?

* 3. What is your last name?

4. When participating in Nevada System of Care meetings, which of the following best reflects whom you represent?

Parent

Caregiver

Family Member

Youth (18 or over)

Clinician (a practitioner with a license to diagnose and treat children)

Professional staff (a professional who is involved in the development and/or implementation of the System of Care)

Administrator (a professional who is in a decision-making capacity related to the System of Care)

5. What type of organization do you represent when you are involved in the Nevada System of Care?

Public (City, County or State government)

Non-Profit

For-Profit

I do not represent an organization when I participate in the System of Care

6. When you participate in the Nevada System of Care, what region of the state best describes the geographic region that you represent?

Clark County

Washoe County

Rural counties (any other county in Nevada besides Clark and Washoe)

Statewide

201

7. Which of the following best describes the length of time you have been involved in the Nevada System of Care?

I have been involved for 1 year or less

I have been involved for at least 2 years (since 2015, when Nevada was awarded its most recent System of Care Implementation grant)

I have been involved longer than 2 years (prior to 2015, when the state was continuing to build upon previous System of Care efforts)

I was involved at some point during the past 2 years, but am no longer involved

I have been involved for 10 or more years (when the state had its first System of Care grant)

I don't consider myself as being involved with the System of Care

8. Which of the following group(s) are you most involved with? (you may select all that apply)

Nevada Commission on Behavioral Health

Nevada Commission on Behavioral Health, Children's System of Care Behavioral Health Subcommittee

Nevada Children's Behavioral Health Consortium

Clark County Children's Mental Health Consortium

Washoe County Children's Mental Health Consortium

Nevada Rural Children's Mental Health Consortium

I am not involved with any of these groups

202

Connection to Others

The following list of individuals was developed from public records of meetings and other Nevada System of Care documents.

9. From the list below, please select the people with whom you are directly or indirectly involved as part of working on the implementation of the Nevada System of Care. Laura Adler Kathy Mayhew Sarah Adler Christy McGill

Krisann Alvarez Kevin McGrath

Babak Amirikhorhe Ken McKay

InformaRtoisosn A rSmhstarorning g Mary Meeker

Lorea Arostegui Michelle Metheny

The folLloauwryin gAv qeruyestion gathers information about nature oSuf syieo Muirll errelationship with other individuals involved in the implementation of the Nevada System of Care. Jennifer Bevacqua Kristen Melton 10. Select the names of individuals that you share information with. Jay Bonomo Dinisha Mingo TLaraucreay A Bdloewr les Heather Mivshek CSaraohle A Bdrloeer rsma Dan Musgrove

Karistiea nBnr uAblvaakreerz Kellie Nesto

JBoahbna kB Aurmeikrikhorhe Aric Neumann

ResourBRceoevses Sr Alyrh mBausrtrritonngg Laura O'Neil

KLoarrean ACrhoasntedgleuri Zarek O'Neil

Lauryn Avery The folElolizwaibnegth qCuhreisstitainosnen gathers information about nature oAuf sytion uOrls roenlationship with other individuals involveKJder insinteif net hrC Belee mivmaecnpqtslue-aNmolelentation of the Nevada System of CTaiffraen.y Ontiveros

Jay Bonomo 11. SCehleerci Dt athye names of individuals that you share resourceNsa wthiathn .Orme KLTaraurclraae yDA eBdllogewar dleos Cara Paoli GSCaarbaorhlie A l Bdrilo eCerhrsiamra Yvonne Penkalski

LKirasistaiae Dn Bunrr uAebltvtaeakreerz Joe Pritchard

RBJoaohbnae krBt AuDrmueirkreikttheorhe Amber Reid

CRBehovrsies rAElyrm mBpusetrrtyong Kim Riggs

RelatioSLKnohasryrehloani ApECrnhodasrntiesdgleuri Paige Ritzman

JLCeahsuiasryricana A Evrenrsyter Kristen Rivas

The folJEloelinzwanibifnegrth B qCeuvhareicssqtituaianosnen gathers information about nature of your relationship with other individuals involved in the implementation of the Nevada System of Care. JKariys tBeonn Colmemo ents-Nolle 12. Select the names of individuals that you feel you have a trusting relationship with. TCrhaecrei yD Bayowles Laura Adler CKarloal eD Berlgoaedrsoma

Sarah Adler KLiastaie D Burruebttaeker

Krisann Alvarez JGoahbnr ieBlu drei Ck hiara

Babak Amirikhorhe BReovberrlty D Buurretottne

Ross Armstrong KCahreisn E Cmhpaenydler

Lorea Arostegui CShyialora Endris

Lauryn Avery EJeliszsaibceat hE rCnhsrteisrtiansen Jennifer Bevacqua Kristen Clements-Nolle

Jay Bonomo Cheri Day

Tracey Bowles Karla Delgado

Carole Broersma Lisa Durette

Katie Brubaker Gabriel di Chiara

John Burek Robert Durette

Beverly Burton Chris Empey

Karen Chandler Shylo Endris

Chiara Jessica Ernster

Elizabeth Christiansen

Kristen Clements-Nolle

Cheri Day

Karla Delgado

Lisa Durette

Gabriel di Chiara

Robert Durette

Chris Empey

Shylo Endris

Jessica Ernster 203

Additional Relationships

13. If you wanted to get something improved or done related to the Nevada System of Care, who would you contact?

First and last name of the first person you would go to:

First and last name of the next person you would go to:

14. If you wanted to get a true reading on where the Nevada System of Care was headed, who would you talk to?

First and last name of the first person you would go to:

First and last name of the next person you would go to:

15. Who are the people you have interacted with the most about the implementation of the Nevada System of Care?

First and Last Name:

16. Overall, how effective or ineffective do you feel the implementation of the Nevada System of Care has been?

Very effective

Somewhat effective

Neutral

Somewhat ineffective

Very ineffective

Thank you very much for your participation in the survey. Please click on "done" below to finish the survey.

204

Appendix F: Network Actor Degree and Betweenness Centrality Scores

Operations Information Resources Trust

OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw

Actor

1 0 32 0 0 23 0 0 17 0 0 17 0

2 0 6 0 0 5 0 0 6 0 0 5 0

3 22 16 5 22 11 16 11 12 7 20 10 37

4 0 41 0 0 32 0 0 19 0 0 25 0

5 0 9 0 0 5 0 0 5 0 0 4 0

6 0 1 0 0 0 0 0 0 0 0 1 0

7 28 5 3 15 2 3 15 3 6 13 1 1

8 0 14 0 0 10 0 0 8 0 0 5 0

9 0 0 0 0 0 0 0 0 0 0 0 0

10 48 28 85 36 25 152 23 24 207 35 20 189

11 25 5 3 7 2 1 7 3 5 6 1 0

12 51 26 69 31 19 58 22 18 71 10 15 37

13 0 1 0 0 1 0 0 1 0 0 0 0

14 54 29 92 27 18 28 52 19 185 21 13 34

15 31 12 8 14 10 6 14 10 6 12 7 0

16 0 17 0 0 13 0 0 9 0 0 8 0

17 40 20 69 40 15 91 40 11 198 17 8 29

18 0 24 0 0 15 0 0 16 0 0 9 0

19 3 10 0 2 10 0 1 6 0 0 6 0

20 0 0 0 0 12 0 0 10 0 0 4 0

205

Operations Information Resources Trust

OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw

Actor

21 52 31 182 12 0 0 8 0 0 5 0 0

22 0 10 0 0 5 0 0 3 0 0 2 0

23 0 27 0 0 15 0 0 17 0 0 8 0

24 35 25 36 13 20 10 4 16 11 23 17 143

25 8 14 2 3 8 1 5 11 57 3 3 0

26 0 12 0 0 8 0 0 10 0 0 7 0

27 0 8 0 0 4 0 0 4 0 0 3 0

28 47 32 67 19 24 52 33 19 77 17 21 61

29 63 37 259 43 27 259 63 25 447 35 23 220

30 0 8 0 0 7 0 0 7 0 0 6 0

31 57 29 228 46 23 234 26 20 153 19 19 87

32 0 1 0 0 0 0 0 0 0 0 0 0

33 13 24 7 11 12 4 0 11 0 9 10 2

34 37 29 39 7 18 5 7 15 16 14 15 31

35 35 32 69 22 15 80 21 12 54 35 11 196

36 42 26 47 36 19 64 36 14 53 36 14 75

37 31 22 68 4 10 1 5 9 1 6 7 3

38 0 12 0 0 6 0 0 8 0 0 6 0

39 0 0 0 0 0 0 0 0 0 0 0 0

40 24 11 5 24 4 3 1 6 1 1 2 0

41 0 0 0 0 0 0 0 0 0 0 0 0

206

Operations Information Resources Trust

OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw

Actor

42 0 1 0 0 1 0 0 1 0 0 1 0

43 0 26 0 0 16 0 0 12 0 0 9 0

44 0 11 0 0 5 0 0 4 0 0 4 0

45 0 0 0 0 0 0 0 0 0 0 0 0

46 36 16 66 16 8 59 11 9 54 21 5 50

47 22 18 12 8 6 3 0 7 0 0 6 11

48 25 12 13 13 9 3 13 9 6 17 5 3

49 0 0 0 0 0 0 0 0 0 0 0 0

50 23 9 4 15 7 10 10 6 1 7 7 7

51 0 15 0 0 9 0 0 12 0 0 9 0

52 45 18 33 24 13 27 19 8 11 6 9 2

53 0 16 0 0 9 0 0 8 0 0 8 0

54 0 6 0 0 2 0 0 2 0 0 2 0

55 70 40 268 66 34 487 67 25 427 54 26 490

56 4 12 4 4 8 4 4 5 2 4 6 7

57 79 43 565 79 37 1052 79 27 711 79 24 1081

58 0 2 0 0 1 0 0 1 0 0 1 0

59 0 0 0 0 0 0 0 0 0 0 0 0

60 0 2 0 0 1 0 0 2 0 0 2 0

61 0 32 0 0 26 0 0 24 0 0 21 0

62 0 2 0 0 1 0 0 1 0 0 1 0

207

Operations Information Resources Trust

OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw

Actor

63 0 1 0 0 0 0 0 0 0 0 0 0

64 0 14 0 0 11 0 0 11 0 0 8 0

65 0 20 0 0 9 0 0 6 0 0 8 0

66 0 20 0 0 10 0 0 9 0 0 12 0

67 0 1 0 0 1 0 0 1 0 0 1 0

68 22 8 18 8 5 98 6 5 98 8 4 95

69 7 3 1 2 2 0 2 2 0 3 2 1

70 0 20 0 0 15 0 0 14 0 0 14 0

71 37 30 37 13 18 6 8 14 1 3 17 5

72 4 5 0 4 2 0 0 4 0 0 1 0

73 47 35 106 45 26 142 46 22 196 34 18 91

74 0 2 0 0 2 0 0 2 0 0 1 0

75 59 36 204 39 22 82 33 19 122 26 16 59

76 20 32 45 11 17 46 4 12 25 3 9 4

77 0 23 0 0 14 0 0 14 0 0 9 0

78 0 0 0 0 0 0 0 0 0 0 0 0

79 36 35 85 32 20 254 9 14 95 0 16 0

80 0 8 0 0 6 0 0 5 0 0 1 0

81 44 27 51 44 21 94 33 17 70 31 17 124

82 16 8 1 6 4 0 5 3 0 1 2 0

83 0 14 0 0 6 0 0 7 0 0 6 0

208

Operations Information Resources Trust

OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw

Actor

84 34 25 99 24 19 134 27 19 241 29 15 311

85 0 1 0 0 0 0 0 0 0 0 0 0

86 23 16 16 16 11 10 6 14 32 19 5 12

87 0 4 0 0 3 0 0 3 0 0 2 0

88 33 9 66 19 6 12 16 7 26 17 6 58

89 0 1 0 0 0 0 0 0 0 0 0 0

90 0 0 0 0 0 0 0 0 0 0 0 0

91 0 9 0 0 5 0 0 7 0 0 4 0

92 18 15 4 12 8 2 14 9 6 15 9 40

93 0 1 0 0 1 0 0 0 0 0 0 0

94 0 2 0 0 2 0 0 2 0 0 2 0

95 0 32 0 0 23 0 0 19 0 0 13 0

96 20 11 3 8 8 2 11 9 7 7 6 3

97 0 20 0 0 10 0 0 11 0 0 10 0

98 38 26 45 38 16 90 38 11 46 10 11 38

99 0 0 0 0 0 0 0 0 0 0 0 0

100 7 8 1 7 4 1 7 5 1 7 3 3

101 0 4 0 0 1 0 0 2 0 0 2 0

102 0 6 0 0 2 0 0 3 0 0 1 0

103 27 17 10 16 10 11 11 9 3 0 10 0

104 0 13 0 0 7 0 0 6 0 0 6 0

209

Operations Information Resources Trust

OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw OutD InD Cbtw

Actor

105 0 18 0 0 10 0 0 9 0 0 9 0

106 0 22 0 0 12 0 0 9 0 0 11 0

107 63 36 213 35 23 188 46 17 183 35 20 242

OutD = Raw Out Degree Centrality Score InD = Raw In Degree Centrality Score Cbtw = Betweenness Centrality

210

Appendix G: Questions for Communities Based on Network Analysis, Table 2 of Provan, Veasie, Staten & Teufel-Shone (2005, p. 606).

1. Which community agencies are most central in the network, and are these agencies essential for addressing community needs?

2. Which core network members have links to important resources through their involvement with organizations outside the network?

3. Are critical network ties based solely on personal relationships or have they become formalized so that they are sustainable over time?

4. Are some network relationships strong while others are weak? Should those relationships that are weak be maintained as is, or should they be strengthened?

5. Which subgroups of network organizations have strong working relationships? How can these groups be mobilized to meet the broader objectives of the network?

6. Based on comparative network data over time, has reasonable progress been made in building community capacity through developing stronger network ties?

7. What is the level of trust among agencies working together, and has it increased or decreased over time? If it has declined, how can it be strengthened?

8. What have been the benefits and drawbacks of collaboration, have these changed over time, and how can benefits be enhanced and drawbacks be minimized?