THE STRUCTURE AND FUNCTION OF COORDINATION NETWORKS IN COLLABORATIVE PARTNERSIHPS by JOHN C. CALANNI B.A., University of Texas at Austin, 1993 M.S., University of Colorado Denver, 1997 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Doctor of Philosophy Public Affairs 2014
ii This thesis for the Doctor of Philosophy degree by John C. Calanni has been approved for the School of Public Affairs b y Christopher Weible, Dissertation Chair Tanya Heikkila, Chair of Examination Committee Paul Teske William Leach January 21 2014
iii Calanni, John C. (PhD, Public Affairs) The Structure and Function of Coordination Networks in Collaborative Partnerships Thesis directed by Professor Chris Weible ABSTRACT Recently, regulatory bodies responsible for managing coastal natural resource areas have turned to more inclusive forms of governanc e. This has given rise to c ollaborative partnerships which tend to rely on consensus based decision making to addr ess issues of mutual interest to the participants As a part of this process multiple stakeholders with divergent beliefs and interests are brought together to address scientific and/ or policy problems with which they are confronted Within such partner ships, stakeholders selectively coordinate with one another to varying degrees to achieve both individual and shared objectives. The ensuing networks of coordination may be associated with various attributes of a partnership, such as perceptions of trust and the facilitation of scientific and policy learning. These attributes, in turn, can influence the ability of the partnership to achieve the objectives that it was designed to accomplish Using interview and questionnaire data from ten m arine aquaculture partnerships, three propositions regarding how individuals within such partnerships decide with wh om to coordinate were tested These included belief homophily (individuals will coordinate with those who hold similar beliefs) trust (individua ls will coordinate with those whom they trust) and resources individuals will coordinate with those who hold needed resources) Of equal interest are the actual coordination networks that are formed within such partnerships and the association of the ens uing network structure with important partnership characteristics. Therefore, t his study also explored the existence and degree of association between network structure ( specifically partnership network density and individual centrality) and variables sho wn to be critical to success in collaborative natural resource management approaches: learning and interpersonal trust Results of this study showed that individuals tend to rely on aspects of trust and needed resources when decid ing with
iv whom to coordina te rather than shared beliefs Structural network variables of density and out degree centrality were positively associated with learning, while in degree centrality and eigenvector centrality were negatively associated with learning and trust. Keywords: collaboration, coordination networks, ideology, beliefs, homophily, trust, resources, Advocacy Coalition Framework, Resource Dependence Theory, Social Capital, marine aquaculture, eigenvector centrality, in degree centrality, out degree centralit y, density learning, partnerships The format and content of this abstract are approved. I recommend its publication. Approved: Chris Weible
v DEDICATION I dedicate this work to my daughters Amber and Sasha. I hope I have shown them that though the pursuit of knowledge and understanding is an enduring battle, it is one th at is valiant and wort hwhile. May they continue the fight with the force of new perspective when the scholars of today abate. Of course they are only 6 and 10, so they thing. is work to you since you were just as much of a part of it as I was. We did this together and it is ours like it or not
vi ACKNOWLEDGMENTS good reason that anyone in their right mind would choose to pursue a PhD in Public Affairs with two young children at home while at the same time trying to carve out a career in the environmental sciences. Such an endeavor would not only be ill advised, it would simply be indicative of a larger psychological issue at hand or a desperate attempt to soothe some buried sense of athletic inadequacy. Either way, I did it. I had no idea what I was getting into. Of course ignorance is bliss and were it not for the support of a wonderful academic cohort, it would all have been for naught. Through the continual banter and cutting exchanges of ideas, fears, failures and successes with my scholarly brethren, we found the determination to carry on. Thou gh our paths have diverged, I want to thank them each, for without them I would not have endured: Scott Mendlesburg, Saba Siddiki, An dy Pattison, Paula Robinson, Joa nn e Shupe, Robin Phelps and Laurie Mandrino, as well as our adopted members, Sarah Miller and Robyn Mobbs I certainly want to thank the members of my dissertation committee, Chris Weible, William Leach, Tanya Heikkila and Paul Teske. I am honored that you chose to stay the course with me through this process. Each of you has played a criti cal and continuing role in my education. Chris Weible showed me that there is a place for the empirical scientist in the study of learning and decision making. He has guided me in the development of critical thinking, helped me to see the questions in th e ether just waiting to be asked, and has always challenged me to rise above that which I believed I was capable Without his friendship, support, and constant nagging I would not have gotten to this point. Will iam Leach showed me that there is indeed po etry to be found in the world of public affairs research. He has influenced my writing, enlightened my understanding of th e persuasive argument and most importantly has sho wn me that the researcher never has to put aside his passion for the pursuit of public ation. They are not
vii exclusive. R ather they are syner gistic. Not only did Tanya Heikkila provide countless hours of dissertation counseling, academic advice, as well as classroom instruction, she served as a relentless archer, land ing arrow after arrow into ideas I was sure were impenetrable. I cannot thank her enough for her contributions to this dissertation, as well as my approach to research. Paul Teske is why I entered the program in the first place. His advice and counselin g early on in my tenure with the School of Public Affairs gave me the courage to push onward, even in the face of several other alluring paths. Most importantly, each member of this committee, though giant in their academic accomplishments, always found t he time to support my academic endeavors. For that I am grateful. I would like to thank the members of the Aquaculture Partnerships Project Team. If there ever was a prime example of the collaborative research process, this was it. I could not have asked for a better team with which to carve out my dissertation. Specifically, I would like to thank Saba Siddiki. Her positive outlook, unwavering willingness to help and fierce attention to detail were instrumental to the success of this team, as well as my morale. I would like to thank the team of scholars inside and outside of the University of Colorado who were instrumental (either directly or indirectly) in delivering to me the skills that would take me to this point: Linda DeLeon, Christine Mart ell, Nancy Leech, Lynne Swackhammer, Lloyd Burton, Jody Fitzpatrick, Xavier Basurto, Adam Henry, Mark Lubell, Paul Sabatier, and many others. I would specifically like to thank Peter DeLeon for his advice, support and most of all his unwavering humor. Having come to the completion of this stage of my academic career, I am honored and humbled that each of you were a part of it. Thank you for everything.
viii TABLE OF CONTENTS CHAPTER I. INTRODUCTION AND SURVEY OF THE TO PIC ................................ ................................ 1 II. STUDY CONTEXT ................................ ................................ ................................ ................ 15 III. STAT EMENT OF THEORIES AN D HYPOTHESES TO BE T ESTED ............................. 20 The Advocacy Coalition Framework ................................ ................................ ........................ 20 Social Capital Theory ................................ ................................ ................................ ................ 22 Resource Dependence Theory ................................ ................................ ................................ ... 24 Ne twork Structure, Learning and Trust ................................ ................................ .................... 25 IV. METHODOLOGY ................................ ................................ ................................ ................ 38 Design ................................ ................................ ................................ ................................ ....... 38 Sample Population Aq uaculture Partnerships ................................ ................................ ........ 39 Pacific Aquaculture Caucus ................................ ................................ ................................ .. 39 California Aquaculture Development Committee ................................ ................................ 40 Shellfish Aquaculture Regulatory Committee ................................ ................................ ....... 41 Florida Marine Net Pen Working Group ................................ ................................ ............... 41 Florida Aquaculture Review Council ................................ ................................ .................... 42 Maine Aquaculture Advisory Council ................................ ................................ ................... 42 Maine Fish Health Technical Committee ................................ ................................ .............. 43 Maryland Aquaculture Coordinating Council ................................ ................................ ....... 43 New Jersey Aquaculture Advisory Council ................................ ................................ .......... 44 The Rhode Island Coastal Resources Management Council ................................ ................. 44
ix Operationalization Measuring Coordination Networks in Aquaculture Partnerships ............ 45 Operationalization Measuring the Importance of Coordination Factors ................................ 47 Operationalization Measuring the Policy Core Beliefs ................................ .......................... 49 Operationalization Measuring Learning Within Partnerships ................................ ................ 51 Operationalization Measuring Trust ................................ ................................ ....................... 52 Statistical Approach and Level of Analysis ................................ ................................ .............. 52 Interview Data ................................ ................................ ................................ ........................... 57 V. RESULTS DEMOGRAPHICS ................................ ................................ ............................ 58 Survey Response Rates ................................ ................................ ................................ ............. 58 Respondent Characteristics ................................ ................................ ................................ ....... 59 Ag e ................................ ................................ ................................ ................................ ......... 59 Education ................................ ................................ ................................ ............................... 59 Political Leanings ................................ ................................ ................................ .................. 60 Environmental Perceptions ................................ ................................ ................................ .... 60 Affiliation ................................ ................................ ................................ .............................. 61 VI. RESULTS COORDINATION NETWORK S ................................ ................................ ..... 64 VII. RESULTS LEARNING WITHIN PART NERSHIPS ................................ ...................... 79 Modeling Results Changes in Opinion on Scientific or Technical I ssues .......................... 84 Modeling Results Changes in Opinion on Policy I ssues ................................ ..................... 85 VIII. RESULTS TRUST WITHIN PARTNER SHIPS ................................ ............................. 88 IX. RESULTS NETWORK CENTRALITY A ND DENSITY ................................ ................ 92 Network Structure Centrality ................................ ................................ .............................. 92
x Network Structure Density ................................ ................................ ............................... 102 X. RESULTS NETWORK STRUCUTRE AN D PARTNERSHIP VARIAB LES ................ 104 Network Density and Learning ................................ ................................ ............................ 104 Network Density and Trust ................................ ................................ ................................ .. 112 Ne twork Centrality and Learning ................................ ................................ ........................ 115 Network Centrality and Trust ................................ ................................ .............................. 122 XI. DISCUSION OF RE SULTS ................................ ................................ ............................... 127 Factors Influencing the Formation of Coordination Networks ................................ ............... 128 Learning and Network Structure ................................ ................................ ............................. 135 Trust and Network Structure ................................ ................................ ................................ ... 140 XII. STUDY LIMITATI ONS ................................ ................................ ................................ .... 143 XIII. CONCLUSIONS A ND FUTURE RESEARCH ................................ .............................. 150 BIBLIOGRAPHY ................................ ................................ ................................ ....................... 155
xi LIST OF TABLES TABLE I.1 Coordination Hypotheses and Levels of Analysis. ................................ ................................ ... 8 I.2 Hypotheses and Levels of Analysis. ................................ ................................ ....................... 14 V.1 Response Rates by Partnership. ................................ ................................ ............................ 59 V.2 Environmental Perceptions. ................................ ................................ ................................ .. 61 V.3 Overall Respondent Affiliations. ................................ ................................ .......................... 62 VI.1 Coordination Citations by Organization. ................................ ................................ ............. 66 VI.2 Mean S cores for Importance of Coordination Factors. ................................ ....................... 67 VI.3 Coordination Factors by Organizational Affiliation. ................................ ........................... 70 VI.4 Marine Aquaculture Policy Beliefs by Organizational Affiliation. ................................ ..... 74 VI.5 Summary of Results Hypotheses 1 3. ................................ ................................ ................. 77 VII.1 Reported Learning by Partnership. ................................ ................................ ..................... 80 VII.1 Reported Learning by Organizational Affiliation. ................................ ............................. 82 VII.2 Changed Opinion by Organizational Affiliation. ................................ ............................... 83 VII.3 Logistic Regression Predicting Opinion Change on Scientific/Technical Issues. ............. 85 VII.4 Logistic Regression Predicting Opinion Change on Policy Issues. ................................ ... 86 VIII.1 Reported Interpersonal Trust by Partnership. ................................ ................................ ... 89 IX.1 Comparison of Centrality Measures by Affiliation. ................................ .......................... 101 X.1 Simple Linear R egression Analysis Summary for Network Density Predicting Learning (Scale Variable). ................................ ................................ ................................ ......................... 110 X.2 Weighted Least Squares Regression Summary for Network Density Predicting Learning (Scale Variable). ................................ ................................ ................................ ......................... 110 X.3 Simple Linear Regression Analysis Summary for Network Density Predicting Learning (understanding stakeholder perspectives). ................................ ................................ .................. 111
xii X.4 Simple Linear Regression Analysis Summary for Network Density Predicting Learning (understanding of aquaculture science). ................................ ................................ ..................... 112 X.5 Logistic Regression Predicting Opinion Change on Significant Policy Issues. ................. 117 X.6 Simple Linear Regression Analysis Summary for Individual In Degree Centrality Predicting Learning (Scale). ................................ ................................ ................................ ......................... 117 X.7 Simple Linear Regression Analysis Summary for Individual In Degree Centrality Predicting Learning (understanding of stakeholder perspectives). ................................ .............................. 119 X.8 Simple Linear Regression Analysis Summary for Individual In Degree Centrality Predicting Learning (understanding of aquaculture science). ................................ ................................ ...... 119 X.9 Logistic Regression Predicting Opinion Change on Scientific/Technical Issues. .............. 120 X.10 Logistic Regression Predicting Opinion Change on Scientific/Technical Issues. ............ 122 X.11 Simple Linear Regression Analysis Summary for In Degree Centrality Predicting Trust (Scale). ................................ ................................ ................................ ................................ ........ 123 X.12 Simple Linear Regression Analysis Summary for In Degree Centrality Predicting Trust (willingness to listen and understand). ................................ ................................ ........................ 124 X.13 Simple Linear Regression Analysis Summary for In Degree Centrality Predicting Trust (trustworthy). ................................ ................................ ................................ .............................. 125 XI.1 Summary of Results and Qualitative Support for Hypotheses. ................................ ......... 127
xiii LIST OF FIGURES FIGURE VI.1 Coordination Citations Percentage of all citations received. ................................ ............ 65 VII.1 Reported Learning by Organizational Affiliation. ................................ ............................. 82 VIII.1 Overall Trust by Organizational Affiliation. ................................ ................................ ..... 91 IX.1 Distribution of Respondent Out Degre e Centrality. ................................ ............................ 93 IX.2 Mean Out Degree Centrality by Organizational Affiliation. ................................ ............... 94 IX.3 Distribution of Respondent In Degree Centrality. ................................ ............................... 96 IX.4 Mean In Degree Centrality by Organizational Affiliation. ................................ ................. 97 IX.5 Distribution of Respondent Eigen vector Centrality. ................................ ........................... 99 IX.6 Mean Eigenvector Centrality by Organizational Affiliation. ................................ ............ 100 IX.7 Network Density by Partnership. ................................ ................................ ....................... 102 X.2 Scatter Plot of Network Density and Learning. ................................ ................................ .. 106 X.2 Network Density and Learning. ................................ ................................ .......................... 107 X.3 Network Density and Learning for Six Partnerships ................................ ......................... 108 X.4 Network Density and Learning for Four Partnerships ................................ ....................... 108 X.5 Scatter Plot of Network Density and Trust. ................................ ................................ ........ 113
xiv LIST OF ABBREVIATIONS ACF Advocacy Coalition Framework CADC California Aquaculture Development Committee CRMC Coastal Resources Management Council EPA Environmental Protection Agency FARC Florida Aquaculture Review Council FMNPWG Florida Marine Net Pen Working Group MAAC Maine Aquaculture Advisory Council M ACC Maryland Aq uaculture Coordinating Council MFHTC Maine Fish Health Technical Committee NGO Non governmental Organization NJAAC New Jersey Aquaculture Advisory Council NOAA National Oceanic and Atmospheric Administration PAC Pacific Aquaculture Caucus RIAWG Rhode Island Aquaculture Working Group SARC Shellfish Aquaculture Regulatory Committee
1 CHAPTER I INTRODUCTION AND SUR VEY OF THE TOPIC To cope with the intractability of modern social problems, public administration has been trending towards governan ce strategies designed to be inclusive of a wide variety of interests and that employ flexible and adaptive processes (Alter and Hage 1993; 2006). Collaboration is an example of such a governance strategy. With its roots in civic republicanism, collaboration is intended to be a deliberative and integrative process that incorporates multiple individuals with the intent of achieving a common goal (Thompson and Perry 2006). Conceptually, collaboration relies on understanding, a collective will, trust and sympathy and the implementation of shared Scholars and prac titioners frequently prescribe incorporating collaboration into governance approaches for issues that exhibit technical, legal, and political complexity and that transect jurisdictional boundaries of multiple government agencies and multiple levels of gove rnment (Innes and Booher 2010; O'Leary et al. 2009; Steelman 2010). Environmental and natural resource conflicts often exhibit high levels of social and scientific complexity and typically transect governmental jurisdictions, making them prime targets for collaborative governance efforts Collaborative governance typically entails bringing together representatives from industry, government agencies, non governmental organizations, academia, and the interested general public to address complex problems (Ansell and Gash 2008). The overarching goal is to generat e policy solutions that are durable, represent the interests of the governed, and are based on deliberation and consensus building as opposed to a purely top down approach (Hall and
2 2000; Agranoff and McGuire 2001; Leach et al. 2002 ; Schneider 2003; Hall and ). These efforts often attain a degree of formality, with elements such as a charter or bylaws, regular meetings, written agendas, a discernible organizational structure, a public website, and a name that reifies the group as an entity unto itself (Leach et al. 2002). Manifestations of collaboration have emerged across several policy arenas with the specific intent of fostering coordination amon g people and organizations from a variety of sectors to engender trust among stakeholders, provide a platform for information exchange and facilitate learning, with the larger goal of improv ing policy outcomes (Himmelman 1992) They have emerged as predom inant governance approaches in areas including public health (Roussos 2000) sustainable tourism (Inskeep 1991; Wight 1998; Hall 2013), community education (Tet 2003), and climate policy (Ingold 2011). In environmental and natural resource policy, manifes tations of collaboration have been coined as watershed partnerships (Sabatier et al. 2005 ), c ollaborative environmental management (Koontz 2004; Koontz and Thomas 2006), c ollaborative institutions (Lubell 2003), integrated environmental management (Marger um and Born 1995), c ollaborative based management (Layzer 2008), and c ollaborative resource management institutions (Heikkila and Gerlak 2005). Very recently, stakeholders have us ed similar st rategies to address natural resource dilemmas associated with development of the marine aquaculture industry in the Uni ted States. Briefly, marine aquaculture refers to the production of marine and/or anadromous species such as oysters, clams, mussels, shrimp and salmon in selected aquatic environments (coastal or offshore) or land based facilities. Efforts to address th e associated natural resource dilemmas have r esulted in the formation of aqua culture partnerships within coastal states These
3 partnerships, which provide the policy context for the empirical focus of this study might vary in composition and purpose. Ho wever, they share an inclusive structural attribute that results in multiple stakeholders from a variety of backgrounds and/or organizational affiliations being brought together to solve issues associated with development of the aquaculture industry in a c ollaborative manner. Aquaculture partnerships might include representatives from the aquaculture industry (ranging from fish and shellfish production to feed and supply services), agencies associated with natural resource governance (on the federal, state and local levels), non governmental organizations (such as environmental groups), educational institutions (university scientists and researchers), shoreline developers as well as interested citizens Even within partnerships, actors tend to engage othe rs in a selective manner, such as coordinating with some and not with others, or coordinating extensively with some and in a limited fashion with others (Wood 1991). Scholars have referred to the resulting structures of interaction as coordination network Sabatier 2005). Prior to moving forward, we need to clarify the idea of coordination and provide some context for the ensuing networks that are referenced above For example, the idea of coordination is widely cited in public management and administration literature but is often used interchangeably with collaboration and cooperation (Isett 2011). Some studies have explicitly defined coordination, with the intent of providing a basis for common under standing within the research community. For example, public management scholars view coordination as a base form of joint activity to harmonize actions (Giesecke 2012) which may include exchanging ideas, providing access to needed resources, or alternatin g activities (Montiel Overall 2005) For the purposes of this study, I base my notion of coordination on that put fort h by Zafonte and Sabatier
4 alters its own political strategies to accommodate the activity of others in pursuit of similar and aligning political behavior, as well as developing, communic ating, and implementing a common plan of action. For clarity it should be understood that c oordination networks can emerge within and outside of collaborative partnerships. The context of this study is specific to coordination networks as they exist wit hin collaborative partnerships. With that understanding, this study will take a two stage approach to analyze networks within collaborative partnerships. In the first stage I will explore the factors that drive the formation of coordination networks usi ng self reported survey responses As will be discussed below, those factors include shared beliefs, interpersonal trust and resource deficits and are specific to individual actors as heuristics for making coordination decisions The following flow diag ram is intended to illustrate the first stage of this study Stage 1 Flow Diagram : Coordination Network Resource Deficits Belief Homophily Interpersonal Trust
5 In the second stage of this study I will explore the impact that coordination network structure ( individual centrality and network density by partnership ) has on outcome variables that are considered important for the successful function of collaborative partnerships (i.e., outcomes that have been found to be important for partnerships to accomplish that which they were intended). These outcome variables include l earning and trust among actors within the collaborative partnership Each of t hese items will be discussed in greater depth below The following flow diagram is intended to illustrate the second stage of this study: Stage 2 Flow Diagram : T rust is cons idered in both stages of this study; in Stage 1 as a self reported individual level factor for coordination (i.e., as a basis for forming network ties) and in Stage 2 as a measure of partnership function (i.e., perceptions of trust within the partnership reported by survey respondents aggregated to the partnership level) Chapter IV details specifically how trust was measured and/or represented in each instance. When dealing with attributes of network structure ( i.e., centrality and density), the level of analysis becomes an important consideration as well. The reader will see that the level of analysis changes between the two. This is by necessity. Centrality is a characteristic of an individual in a network (i.e. o ne individual may occupy a more central position within the network than another individual). Reported Learning Reported Trust Individual Centrality Network Density
6 Conversely, density (the total number of network ties as a proportion of the total possible number of ties among members of a partnership) is a characteristic of a partnership (i.e., the density of a network may be less or more dense than that of another partnership). Given this, when considering centrality, the level of analysis is at the individual. When considering density, the level of analysis is at the level of the partnership. Efforts have been made to clearly show which level of analysis is applicable throughout the study. Coordination As indicated above, we see that coordination networks emerge within partnerships, and that even within the partnership structure actors are selective in their coordination behavior. This leads us to the question individuals when they are deciding with whom to coordinate ? Social science literature has indicated that variables expla ining relationships and behavior within partnerships are not necessarily all that well understood (Agranoff and McGuire 2001; Imperial 2005; Jones 1997; Kramer 1999). This is not to say that there is a complete absence of such research. For example, poli cy network literature has identified multiple reasons that actors within collaborative environments form network connections. These may include having existing relationships, mutual interests, or reputation (Powell, 1990; Knoke, 1998), the need for inform ation and advice (Knoke, 1996; Schneider, 2003), resource exchange (Provan & Milward, 1995; Knoke, 1996), the need to gain influence and/or access to influential officials (Knoke, 1998), and to increase the potential for success in advancing policy objecti ves (Zafonte & Sabatier, 1998). Scholars have also suggested in some cases that preferences and belief systems tend to drive the formation of and serve as the glue to hold coalitions together (Weible & Sabatier 2005; Ingold 2011).
7 Given the need to further explore variables explaining relationships and behavior within partnerships, the presence of a relevant framework or existing set of theoretical explanations in this area is necessary to guide this study. In fact, policy scholars have repeatedly s tressed the need for theory driven inquiry into the policy process, including explanations of network formation as well as implications of network structure (Sabatier 1991, Ibarra 1993). Given this, it is clear that more research is needed that attempts t o empirically evaluate the rationale behind individual coordination network formation and does so using a rigorous multiple theory testing approach This study seeks to answer those calls by elicit ing the factors that individuals rely upon when they decid e with whom to coordinate and doing so by testing three theories that rely on complementary explanations for individual and organizational behavior : belief homophily (the idea that individuals tend to have ties with others who are similar to themselves i n socially significant ways; in this case beliefs being the underlying basis for similarity) as advanced by the Advocacy Coalition Framework coalition formation (Sabatier and Jenkins Smith 1993), interpersonal trust as suggested in Social Capital Theory (Coleman 1990; Putnam 1993), and resource deficits as suggested in Resource Dependence Theory (Pfeffer and Salancik 1978). Table I.1 provides the hypotheses related to coordination (H1 H3) that will be evaluated in the first stage of this study, the level of analysis and the approximate sample number.
8 Table I.1 Coordination Hypotheses and Levels of A nalysis. Hypothesis Level of Analysis Sample Number H1: In collaborative partnerships, agreement on major policy issues between members will be an important factor in the formation of individual coordination networks. Individual n=110 H2: In collaborative partnerships, interpersonal trust will be an important factor in the formation of individual coordination networks Individual n=110 H3: In collaborative partnerships, perceived resource attributes between members will be an important factor in the formation of individual coordination networ ks. Individual n=110 Returning again to the flow diagram for Stage 1 of this study, we can incorporate the individual hypotheses for additional clarity. Stage 1 Flow Diagram: On its own, i nquiry into factors influencing individual coordination network formation in partnerships should naturally lead to the follow on question so what? T hough coordination factors are of interest (i.e., helping to answer the question why ), of equal interest is the association of the resulting coordination network structure with variables linked to the ability of Coordination Network H3: Resource Deficits H1: Belief Homophily H2: Interpersonal Trust
9 a partnership to accomplish its goals which, in turn, can impact the effectiveness of inclusive natural resource governance strategies overall This brings us to Stag e 2 of the study. I f we measure the relationship between the network structure and partnership function (i.e., aspects of the partnership that contribute to its success) we can shed light on the so what? lend insight into the association of network structure on attributes associated with successful natural resource governance. Given this, after addressing how stakeholders form coordination networks within partnerships, this study investigates a second question: is the structure of t hese coordination networks associated with T he concept of social networks has gained much attention among scholars of natural resource management processes (Holling 1978; Schneider et al. 2003; Anderies et al. 2004; Olsson 2004; Ostrom 2005), and has widely been considered a n observable phenomenon that can be analyzed using quantitative methods (Marsden 1990; Scott 2000 ; Freeman 2004 ). Studies have shown that the existence of social networks between stakeholders can have substantial impacts on the environment within which they are constructed, such as building community resilience and increasing natural resource adaptive capacity (Thompkins et al. 2004). Other research has identified specific features of a network structure (on both the individual and partnership levels) that can impact outcomes of natural resource management efforts. For example, having high number of social ties for an individual may enhance development of knowledg e and understanding through access to an expanded body of information This can lead to greater opportunities for individual learning and understanding which in turn may increase the potential for joint action to address natural resource dilemmas (Isaac e t al. 2007; Conley and Udry 2001). Research has also shown that learning can be influenced by partnership level network attributes as well. For example, in their study of partnership processes, Newig, et al.
10 (2010) indicated that learning within collabora tive groups should be enhanced as the number and density of connections between group members increases. Additionally, natural resource governance literature has shown that network density is positively correlated with the potential for collective action due to increased communication, reciprocity and mutual trust which may reduce the potential for stalemates regarding resource conflicts (Pretty and Ward 2001; Janssen and Ostrom 2006) Conversely, there is evidence that suggests a limit to the benefits of network density and density of interactions For example, it has been suggested that within very dense networks (where interactions are frequent) knowledg e homogenization can occur where members of a group can adopt very similar perceptions of a probl em or issue at hand. This can greatly inhibit the transfer of new ideas or information (Bodin and Norberg 2005; Leavitt 1951; Shaw 1981) and inadvertently inhibit the potential for innovative solutions (Crona and Bodin 2006) Management theory has also i dentified the importance of social networks, primarily focusing on structural properties and how those structures may explain organizational outcomes For example, studies indicate that individual positions within networks may confer certain advantages, s uch as increasing power (Brass 1984), influence (Friedin 1993) and promotion potential (Burt 1992) ; or influence the specific role an individual may have within an organization (Ibarra 1993) What is evident from this body of literature is that network st ructure matters, and the attributes of structure are worthy of inquiry. Though past research has shown that network structure appears to make a difference in natural resource governance, scholars indicate the need to evaluate network characteristics that can benefit inclusive governance approaches including facilitation of learning and engendering
11 trust among stakeholders (Bodin et al 2009). These two characteristics (i.e., learning and trust ) are seen as critical to successful natural resource governance strategies for fostering environments that allow for information exchange and provide opportunities for learning among a group of participants (Innes an d Booher 2010; Weible and Sabatier 2009). This sort of group learning within partnerships has been coined by some scholars as collaborative learning. Heikkila and Gerlak (2013) put forth a framework of collective learning that defines not only process, b ut also products. For instance, the collective process may include knowledge development (or acquisition), interpretation and evaluation, followed by dissemination across members of that collective (2013). Collective products are generated as a product o In 1999, Pierre Dillenbourg assembled what was then a comprehensive review of what learning in a collaborative environment might entail. His review of the literature indicated a general agreement that collaborative learning includes notions such as studying and evaluating information to address unknowns, joint problem solving (though learning might be a byproduct of joint problem s olving), or acquisition of knowledge from various individuals within a professional community as a part of collaborative work (Dillenbourg 1999). Muro and Jeffrey (2008) performed a similar critical review of social learning research specific to collabor ative natural resource management processes. In this review they discovered a common thread throughout the literature regarding learning in social environments. Specifically, they indicated that existing models of social learning have at their core a pro cess of
12 the generation of new knowledge, acquisition of new skills (both social and technical), and the generation of social capital in collaborative environments may lead to a clearer, shared definition of the problem or issue being addressed and enhance the potential for understanding and subsequent collective action (W ebler 1995; Maarleveld 1999; Woodhill 2004; and Keen 2005). Learning can also broaden the possibilities of conflict remedies, thus increasing the ability of a group to address natural resource dilemmas more efficiently and effectively (Isaac et al. 2007; Conley and Udry 2001). What of trust? Historically scholars have espoused the benefits of trust among individuals and wi thin an organization. High levels of trust can facilitate the emergence of cooperative behavior in addressing collective action dilem mas (Coleman 1990; Lubell 2007; Ostrom 1990); improve the flow and interpretation of information (Gnyawali and Madhavan 2001); facilitate the negotiation process (Berardo 2009); reduce political transaction costs (Scholz and Lubell 1998); and reduce the em ergence of opportunistic behavior, which can have a positive feedback effect by leading to an increase in levels of trust (Putnam 1995). Likewise, understa nd views of other members, make compromises, and to fully participate in the process of solving natural resource dilemmas (Pretty and Ward 2001; Schneider et al. 2003; Thompkins et al. 2004; Janssen and Ostrom 2006). In addition to the network variables discussed above, scholars have identified others that influence learning and trust within partnerships. Interestingly, the literature is somewhat limited when it comes to this topic as well However, the studies that have been performed have helped to place bounds on the problem and develop a research agenda. Through their extensive review
13 Muro and Jeffrey (2008) and Gerlak and Heikkila (2011) identifi ed structural partnership traits that are important to facilitate collaborative learning and enhance trust For instance, diverse participation can facilitate learning within partnerships (Brummel 2010) allowing for the incorporation of information from diverse sources that might be associated with unique perspectives (Muro and Jeffrey 2008). Likewise procedures and processes within a partnership that are viewed as fair and legitimate can provide for greater learning opportunities (Muro and Jeffrey 2008; Susskind et al. 1999) and may also lead to greater levels of interpersonal trust and serve to gene rate social capital among stakeholders (Leach and Sabatier 2005). Just as perceived fairness can lead to greater interpersonal trust, Walker and Senecah (2011) found that greater levels of interpersonal trust can likewise facilitate learning, allowing for participants to feel that others are communicating openly, honestly, and negotiating in good faith. Leach et al. (2013) put forth a cursory model of learning within partnerships using data generated as a part of the Aquaculture Partnerships Project. They found that new knowledge is highly correlated with partnership level traits (such as procedural fairness, trustworthiness of other participants, level of scie ntific certainty, and diverse participation) as well as individual traits (such as norms of consensus and scientific or technical competence). Thus we see that learning and trust are important for successful natural resource governance approaches. The l iterature has also shown that there are several individual and partnership level variables that have the potential to impact learning and trust. In order to continue this line of inquiry and to provide additional external validity to the concepts of netwo rk structure and collaborative process outputs, this study will explore the existence and degree of association between network density and individual centrality on learning and
14 perceptions of trust within collaborative partnerships. Note that this approac h is intended to build on the model put forth by Leach (2013), which identified partnership and individual traits of associated with knowledge acquisition within aquaculture partnerships based on the same APA data used in this study. Table I. 2 provides the hypotheses to be evaluated (H4 H7) related to learning and trust the level of analysis and the approximate sample number. Table I. 2 Hypotheses and Levels of A nalysis. Level of Analysis Sample Number H4: The higher the network density fo r a collaborative partnership, t he higher the level of reported learning on average across members of the partnership. Partnership N=10 H5: The higher the network density for a collaborative partnership, the higher the level of reported trust on average across members of the partnership. Partnership N=10 H6: In collaborative partnerships, the higher the network centrality of an individual, the higher the reported learning by those same individuals. Individual n =110 H7: In collaborative partnerships, the higher the network centrality of an individual, the higher the reported level of trust reported by those same individuals. Individual n = 110 Returning again to the flow diagram for Stage 2 of this study, we can incorporate the individual hypotheses for additional clarity. Stage 2 Flow Diagram: Reported Learning Reported Trust H6/H7: Individual Centrality H4/H5: Network Density
15 CHAPTER II STUDY CONTEXT For the past decade, policy process scholars have studied watershed partnerships (Lubell et al. 2009) and similar ecosystem scale processes (Heikkila and Gerlak 2005). This study investigates several partnerships engaged in a relatively new debate involving the development of marine aquaculture. Aquaculture is defined by the National Oceanic and Atmospheric Administration (NOAA), the proposed lead federal agency on marine aquaculture regulatory and the propagation and rearing of aquatic organisms in selected aquatic environments (coastal and offshore) or land based facilities for any commercial, recreational, or public purpose. 2011 a 1 ). The term marine aquaculture refers to the production of marine and/or anadromous species such as oysters, clams, mussels, shrimp, and salmon. In 2011, the De partment of Commerce and NOAA concurren tly released draft aquaculture policies which are currently under public comment (public comment period ends April 11, 2011). Growing U.S. and worldwide dema nd for seafood is likely to conti nue as a result of increases in projected to meet increased demand even with rebuilding efforts, future increases in supply are lik ely to come either from foreign aquaculture or increased domestic aquaculture production, or some combination of both a 4) re is also enable sustainable aquaculture that provides domestic jobs, products, and services and that is in harmony with healthy, productive,
16 (NOAA 2011 a 1) In 2009 the combination of marine and freshwater aquaculture accounted for half of all seafood consumed worldwide (NOAA 2011 b 3). T he disparity between domestic and foreign aquaculture produced seafood is striking. In 2011 the national Marine Fisheri es Service Office of Science and Technology estimated total United States aquaculture production at about $ 1.2 billion annually, compared to worldwide production of $70 billion (NOAA 2 0 11 b ). This has translated into a situation where the United States imp orts roughly 84% of its seafood. Half of that is produced via aquaculture, only 5% of which is actually produced in the United States (NOAA 2011a). Given the broad based goals and objectives that NOAA has proposed in its draft aquaculture policy (i.e., covering topics that might be expected to afford NOAA with broad support from various interests including food supply, ocean stewardship, dwindling wild stocks, and economic development ), coupled with the substantial seafood trade deficit mentioned above, one might ask why a comprehensive aquaculture development policy has been so slow on the uptake in the United States. One reason is that the overall regulatory landscape for marine aquaculture is quite complex and somewhat piecemeal, including 18 applica ble federal regulations that are housed in a variety of federal agencies, as well as multiple state and local regulations that may overlap or conflict with federal statutes (Firestone 2004; Wirth 1999). The piecemeal regulations often discourage new busin esses from entering the industry, or existing operations from expanding. An intimidating regulatory environment coupled with stiff opposition to marine aquaculture from many environmental groups (concerned about natural resource degradation, impacts to na tive fish species, and visual pollution), food safety advocates (concerned with mercury and other environmental contaminants), fishermen (concerned about seafood market competition), homeowners (concerned about visual pollution), and coastal land
17 developer s (concerned about coastal property values), has left the United States in the position of laggard when it comes to marine aquaculture development (Firestone 2003; Mazur and Curtis 2006; McDaniels 2006; Wirth 1999). Given this backdrop, it is easy to see the challenge confronting NOAA when the agency was charged in 1998 aquaculture policy and strategy to provide a context for agency activities for the next ten to 2011b, 1 ). Thirteen years later, in conjunction with the release of he stage is set for a c ollaborative governance process that brings together a wide variety of individuals, groups, and interests, providing a prime opportunity to study individual behavior in a socially relevant policy context. The global debate over development of the aquaculture industry is complex, covering issues rooted in economics, ecology, food safety, government jurisdiction, and regulatory structure (Fi restone 2004). Advocates for rapid industry development, as well as those demanding restraint to insure sustainable industry development, see a role for aquaculture in helping to satisfy the global demand for seafood (which is currently outpacing supply) and also providing critical economic opportunities for remote and rural communities worldwide (Katranidis 2003). Proponents for industry development also take an ecological perspective, viewing aquaculture as a potential remedy for dwindling populations o f wild fish due to overfishing, pollution, habitat loss, and an entrenched fisheries management process that cannot accommodate the rising global demand for seafood (Firestone 2004; Katranidis 2003). Specifically, marine aquaculture is seen as a mechanism to help preserve threatened aquatic biodiversity in two ways: by rebuilding endangered fish stocks and relieving the pressure placed on wild fish stocks for food consumption (Francik 2003).
18 Opponents have indicated that, from an ecological perspective, marine aquaculture is an unsustainable endeavor, at least in its current form. Specifically they cite the use of farmed and wild fish stocks (such as herring) for feeding larger farm raised species, often at a protein consumption rate that is higher than production (i.e. more fish protein is consumed in the form of feed than is produced in the form of marketed fish). Other arguments against industry development include the propagation of diseases (such as sea lice in salmon) from farmed fish to wild fish stocks; the biological and environmental accumulation of chemical and pharmaceutical products used for disease control at aquaculture facilities; the potential for farmed fish to escape and breed with wild fish stocks (referred to as genetic pollution of wild fish stocks); competition between farmed fish escapees and wild stocks for habitat, food, and mating opportunities; and degradation of natural resources in the area of farming activities, including sedimentation, concentrated biological wastes emanati ng from farming operations, as well as visual pollution (Black 2001; Francik 2003; Treece 2002; Naylor 2000; Mazur 2006). Conflicts regarding aquaculture development have been documented on a global scale, ranging from degradation of mangroves and sensit ive wetlands due to shrimp farming in Asia (Kaiser and Stead 2002), to the relative cessation of salmon farming development in the Pacific Northwest of the US due to environmental and aesthetic opposition (Mazur 2006). Social science research has shown th at much of the conflict is based on risk perception disparity between experts and lay persons in populations affected by aquaculture development (Mazur 2006) as well as public uncertainty regarding the use of scientific information and the ability of techn ology to solve industry challenges (Petts and Leach 2000; Kaiser and Stead 2002). To advance the discussion on marine aquaculture, there has been a call globally to improve the image of the industry by taking inclusive approaches to natural resource plann ing in local
19 communities (i.e., including a variety of stakeholders in planning) as well as changing the way information is communicated to the public (Mazur 2006; Burbridge 2001). Studies of public perception of aquaculture worldwide support this call, indicating that public risk perceptions and divergent expectations regarding management and allocation of resources to maximize public benefit are substantial contributors to community resistance to aquaculture development (Mazur 2004). European and Ameri can scholars have suggested that there is a need to communicate accurate cost and benefit information regarding sustainable aquaculture development to the public, natural resource managers, and policy makers (Burbridge 2001); establish pro active approache s to coastal resource management that is inclusive of a wide variety of stakeholders and interests (Mazur 2006; Burbridge 2001; Firestone 2004); and management of the marine environment in a transparent and equitable fashion (Firestone 2004, In the United States, several partnerships have emerged to grapple with these issues. Similar to partnerships in watersheds and other complex socio ecological syst ems, marine aquaculture partnerships feature government and non government entities engaging in consensus based deliberations with an emphasis on finding win win solutions. Like other c ollaborative processes, aquaculture partnerships are intended to serve as a crucible for mitigating political disagreement and ensuring some coordination among stakeholders. This study will help to determine whether or not this is actually occurring.
20 CHAPTER I I I STATEMENT OF THEORIE S AND HYPOTHESES TO BE TESTED To furthe r explore the formation of coordination networks within aquaculture partnerships hypotheses generate d from three theoretical lenses (each identifying a specific factor to explain how actors might form coordination networks ) were tested The three factors a nd their affiliated theories are belief homophily in the Advocacy Coalition Framework theory of coalition formation ; interpersonal trust in Social Capital Theory; and resource deficit in Resource Dependence Theory. These lenses were selected since they provide theory based and testable rationale for individual decision making. The questions that naturally surface include the following: A re participants in aquaculture partnerships relying on shared beliefs interper sonal trust or attempting to satisfy a resource deficit (such as information, influence or funds) when deciding with whom to coordinate? I s there an interaction between some or all of these? Even though these coordination factors have a theory based explanation for individual decision making, there is the potential that external mandates could be influencing coordination. Therefore, I also incorporated the concept of a legal requirement into this study, which is further discussed below. The Advocacy Coalition Framework Developed by Sabatier and Jenkins Smith in the 1980s, the ACF employs a model of the ind ividuals as fundamentally limited in the information that they can encounter, consider, and subsequently use in decision making processes (Simon 1955). Given these limitations,
21 individuals must use short cuts, or heuristics to simplify and process the inf ormation they encounter. According to behavioral theory, they do this by filtering incoming information based on how it lines up with their own beliefs and precognitions so as to reduce the mental discomfort (also referred to as cognitive dissonance) new information creates (Lord et al. 1979; Festinger existing beliefs (causing unco mfortable psychological dissonance). Using this model of the individual, the ACF supports a theory that deals with formation, structure, and stability of advocacy coalitions as well as the behaviors of individuals within coalitions Within this theory, beliefs are a causal driver for behavior and the primary heuristic on which individuals rely for political decision making. The most important type of beliefs in shaping political behavior in the ACF is policy core beliefs. Policy core beliefs are subsystem wide in scope and are the foundation for forming coalitions, establishing alliances, and coordinating activities among subsystem members. Policy core beliefs represent the major issue wide attributes of a policy area, such as seriousness of a p The ACF would thus hold that when policy core beliefs are in dispu te, coordination networks within partnerships will be based on shared policy core beliefs. In fact, several ACF studies have shown that there is an association between belief congruence or shared ideology and network ties (McPherson et al. 2001; Weible 20 05; Weible and Sabatier 2005; Henry et al. 2010). In a study evaluating social network structures and power relations within global climate
22 conflict and coope enemy networks [investigated through cooperation and conflict relations] thus seem to be a close However, in a study juxta posing belief systems and social capital as drivers of policy network structure, Henry (2010) found that though policy networks may be structured based on policy beliefs, bonding and social capital also plays an important formative role as well. In fact, interact because of an aversion to a common opponent, although bonding forms of social capital are critical in order for these like Other studies howev er, tend to support the importance of beliefs as a network driver. I n a n ACF study of the Swedish Carnivore Management System, perceived belief correspondence was found to be the driving mechanism behind political coordination when compared to perceived i nfluence (Matti and Sandstrom 2011) Likewise, in a study of conflict and cooperation networks associated with Swiss climate policy, it was found that the structure of those networks was a significant and close predictor of coalition belief systems (Ingol d 2011). Given the body of evidence suggesting the importance of beliefs, t his leads us to the first testable hypothesis, which emphasizes shared beliefs as a factor for coordination: Hypothesis 1 (H1) Belief Homophily : In c ollaborative partnerships agreement on major policy issues between members will be an important factor in the formation of individual coordination networks Social Capital Theory Social capital social re and has been identified by organizational scholars as a topic of key concern. It has been identified as a driver of individual career success ( Gabbay and Zuckerman 1998), or ganizational effectiveness ( Tsai and Ghoshal 1998 ), resource exchange and problem solving (Gabbay and Zuckerman
23 1998), and the creation of intellectual capital (Hargadon and Sutton 1999). Social capital has been defined in a variety of ways, including the facilitation of cooperation due to trust, social norms, and expectations of reciprocity (Coleman 1990; Putnam 1993) In this study, we focus on the role of trust, since it serves as the foundation of social capital theory (Putnam 1995) and its methods f or its conceptualization and measurement are more highly developed in the existing literature (Lubell 2007). A n argument can be made that trust and belief homophily interact in such a way that make it very difficult to elicit a difference between the two. For example, individuals may tend to trust others that share their beliefs. Or, conversely, individuals may tend to take on the beliefs of those they trust. In order to untangle these concepts, I have employ ed a methodological approach (through survey questions) that explicitly de couples the two and employ ed a statistical approach that can independently test f or a difference between the two. This will be discussed further in Chapter IV Levi and Stoker (2000) view trust as relational (between multiple individuals or groups), conditional (context specific), based on a transaction cost model (i.e., a reduction i n transaction costs associated with monitoring and enforcement activities), and reliant on beliefs about individuals and/or groups. In their definition of trust, Levi and Stoker (2000) indicate that trust and trustworthiness have two dimensions: a commit ment to act in the best interest of the truster on the basis of moral values (values that align incentives, a shared definition and value of promise keeping, and a general concern for the welfare of the truster); and competence (trustworthy entities have t he aptitude to act in a trustworthy fashion). Alternatively, trust has also been defined more in line with rational choice, where individuals act in a benefit maximizing and cost minimizing fashion. For example, Hardin (1993) indicates that knowledge
24 and incentives are central to trust. Therefore, if a group or individual has the knowledge of both in such a way that living up to the expectations of the othe r group/individual is in the best interest of both groups (self maximizing behavior). In both cases presented above, it is evident that interests play an essential role in the generation of trust. A salient aspect of trust for this study is thus how trust translates to coordination between individuals and groups. Trust has been shown to be an important aspect of coordination, cooperation and group participation in policy endeavors which incorporate multiple competing interests (Cook 2005). If we take the transaction cost minimizing definition of trust, as well as the rational choice explanation of trust, it would follow that in c ollaborative policy environments, the minimization of transaction costs or self maximizing behavior would influence group coordi nation Given this idea of trust, the following hypothesis was tested : Hypothesis 2 (H2) Trust : In c ollaborative partnerships trust between members will be an important factor in the formation of individual coordination networks Resource Dependence Theory Advanced by Jeffrey Pfeffer and Gerald Salancik (1978), Resource Dependence Theory has its roots in sociology and is based on a power maximization model of organizations originated by Max Weber in 1947. Resource Dependence Theory has been applied t o both coalitions within organizations and coalitions among organizations (Ulrich 1984). These coalitions are viewed as continually changing their form and relationships primarily to decrease ce other organizations have on them. Additionally, and critical to this theory, resources are seen as not only essential to organizational power and survival, but also scarce and difficult to acquire (Ulrich 1984). Therefore, organizations will coordinat e with other organizations to reduce the uncertainty of
25 obtaining critical resources. Organizations and coalitions within organizations are expected to strive to maximize their power by controlling resources and limiting their reliance on other organizati ons for critical resources. Just as Resource Dependence is scalable from the organizational level to the inter and intra organizational level, we have extended the scale to the individual level for the purposes of this study. Conceptually this should not pose a substantial problem since, in Resource Dependence Theory; the organization is modeled after certain attributes of rational choice actors (i.e., self maximizing entities acting in a strategic manner) and, conversely, individuals included in this study represent organizations The extension to the individual level should therefore not be problematic. The model of motivation outlined in Resource Dependence Theory differs from those presented in our two previous theories, where trust and beliefs ar e the central behavioral impetus. Though trust and behavior might be a component of Resource Dependence Theory (via the ease with which relationships are forged), controlling resource availability will dictate with whom organizations or those speaking on their behalf, collaborate. In the organizational literature this translates to firms forming partnerships and joint ventures, forging personal ties and/or contacts with critical supply firms (Provan 1980), and non profit organizations forming joint ventu res or other collaborations (Alter and Hage 1993). With this power based model of organization and coordination behavior, a Resource Dependence hypothesis was tested along with the ACF and Social Capital hypotheses. Hypothesis 3 (H3) Resource Deficit : In c ollaborative partnerships perceived resource attributes between members will be an important factor in the formation of individual coordination networks Network Structure Learning and Trust Having tested three complementary explanations for the formati on of coordination networks within partnerships this study next explore d the association between network structure
26 and important partnership variables (as discussed in the introduction) S ocial network analysis prov ides a set of tools that can help us accomplish that task The specific aspect s of coordination ne tworks that were evaluated include network density (measured at the partnership level) and centrality ( measured at the individual level). Network density (at the partnership level) is defined as the total number of network ties as a proportion of the total possible number of ties among members of a partnership (Scott 2000) and has been identified as an important structural variable in natural resources management literature (see Bodin et al. 2006 for a review). The maximum density (where every node is connected to every other node) would be given a value of 1. The minimu m density would be 0. According to Ingold ( 2011 ), density values above 0.5 are considered high, 0.25 0.5 medium, and below 0.25 low. Conceptually, the presence of a network tie is an indicat or of interaction. The greater the number of network ties, the greater the number of interactions, or potential interactions is the structure of these coordination networks associated with For this study, the partner ship function variables of interest are learning and trust As discussed earlier, n atural resource management l iterature is mixed when it comes to the impact of network density on learning and trust For example, Newig, et al. (2010) found that learning within collaborative groups should be enhanced as the density of connections between group m embers increases, citing the opportunity for information exchange as the reason. The idea of opportunity for information exchange is persistent in n etwork literature, and relates not only learning but also to trust. For example, scholars in this tradition view increasing do 2009; 179) At the same time dense networks can
27 enhance the trustworthiness of that information since any instances of misinformation can be communicated throughout the network quickly resulting in reputation impacts, group sanctions, or other penalti es (Burt 2001) Borgatti and Cross (2003) established some constraints on learning and information seeking and sharing in social networks. They introduced a model that views information exchange as a function of 1) knowing what a person knows, 2) valuin g that information, and 3) being able to access that information. Access is an important aspect of this model, and as network density increases, so do the opportunities for access (Borgatti 2003). Thus, as density increases, so do the opportunities for o rganizational learning (Borgatti 2003) and the spread of information through increased accessibility (Bodin 2006) Conversely, there is evidence that suggests a limit to the benefits of network density. For example, it has been suggested that very high l evels of network density can result in knowledge homogenization, which can inhibit the transfer of new ideas or information (Bodin and Norberg 2005; Leavitt 1951; Shaw 1981). a density of inter action among individuals that leads to a situation in which all individuals tend to adopt similar perceptions of issues at hand may allow for access to a vast body of information that access may not nec essarily lead to learning. For example, when faced with large quantities of information, human cognitive understanding (Festinger 1957), or may simply overload new information and force them to rely on their preexisting notions of a problem (Bodin 2006).
28 Research has shown that network density can have an impact on individual behavior in a general sense with a theoretic al basis in Social Capital For example, Coleman suggests that dense social networks are critical for the establishment, existence and monitoring of social norms (1988) This is partly because dense networks enable reputation effects (i.e., that unsavory behavior will be publicized across a network) and the possibility of sanctions by members of a network for unethical or dishonest behavior (1988). Studies performed on the New York diamond exchange tend to support this, indicating t hat in situations of dense, close knit networks (such as the network of diamond exchange merchants) individual behavior could easily be monitored and any instances of unethical actions would quickly be transmitted through the community, resulting in sever reputation (Burt and Knez 1995). The combination of close surveillance, quick network broadcasting and a reliance on professional reputation resulted in a positive association of network density with perceptions of trust within th e community (Burt and Knez 1995). Given this, dense network s ma y increase the likelihood of consensus on network related issues, while at the same time nudging behavior of network actors (such as judgment, perception, intent and subsequent actions) toward s what might be considered moral intensity (Jones 1991). The general consensus of the literature indicates that there should be a significant and positive relationship associated with network density and trust. However, as indicated above, there do appea r to be limitations. In order to further investigate the effect of network density on learning and trust the following hypotheses were developed : Hypothesis 4 (H4) Network Density and Learning (at the partnership level) : The higher the network density for a collaborative partnership the higher the level of reported learning on average across members of the partnership
29 Hypothesis 5 (H5) Network Density and Trust (at the partnership level) : The higher the network den sity for a collaborative partnership the higher the level of reported trust on average across members of the partnership As discussed earlier, centrality is another important variable to consider when looking at network structure. Centrality has historically been represented in terms of power in network analytics. However, recent advances in social network methods have sug gested that the use of power is somewhat of an oversimplification, since power is more accurately described as a by product of network position that can be influenced by other variables (Hanneman 200 5 ) Instead centrality, when viewed as the meas ure of i ndividual position with respect to the center of activity within a network, is considered a more appropriate approach when evaluating network structure Centrality t ends to be described in multiple ways, each using a different aspect of network position as the focus. One common measure of centrality is termed degree centrality Degree centrality is a measure of the number of ties an indi vidual has to others within a network. Individuals with many ties may find themselves in an advantageous position for multiple reasons: they are less dependent on others to meet their needs ; they may have broader access to greater network resources ; and since they have multiple ties, they are likely to serve as int ermediaries in network transactions, from which there is the potential for benefit (Borgatti 1992). In directed networks, degree centrality is typically represented in terms of in degrees (i.e., ties directed towards an individual) and out degrees (ties emanating from an individual to others ). A visual representation of in degree and out degree centrality might look something like this:
30 In this case we have four network nodes: A through D. Node A has three incoming connections (in our case these are citations for coordination), yielding an in degree centrality value of 3. Similarly, n ode B has 3 incoming network connections (though from different nodes than A) yielding an in degree centrality value of 3 Node C has two incoming connections, and D has only one incoming network connection, yielding in degree centrality values of 2 and 1 respectively. Conversely, node A has 2 outgoing network connections (in our case node A has cited C and D for coordination ), yielding an out degree centrality value of 2. Node B has two outgoing citations, resulting in an out degree centrality of 2 as well Nodes C and D have cited four others for coordination, resulting in out degree centrality values of 4 for each. In tabular form centrality would look like this:
31 Node In degree Centrality Out degree Centrality A 3 2 B 3 2 C 2 4 D 1 4 In this example, the most central network member when considering in degree centrality would be nodes A and B (with 3 each) and the least central node D (only 1 incoming citation for coordination ) When considering out degree centrality the most central network member would be nodes C and D (4 each), and the least central nodes A and B (2 each) An individual with a high level of in degree centrality is said to be prominent, or h old a certain level of prestige. It is an individual with whom other actors are trying to make ties and individual which is often the object of communication rather than the source (Knoke 1983) A n individual with a high level of out degree centrality has the opportunity for multiple exchanges with other individuals within the network, providing the chance to make other network members aware of their views (Borgatti 1992). Individuals that hold a high level of out degree centrality can be thought of as influential (i.e., able to influence a large number of others) Though some what straightforward and simple, in degree and out degree centrality are considered robust indicator s of relative importance in a network. Where degree centrality focuses on the immediate ties an individual has, it does not consider the indirect ties to others within a network. In other words, an individual may have a high degree centrality, but may be relatively isolated within the network as a whole. Eigenvector centrality addresses this problem by taking into considerat ion the centrality of the adjacent nodes in a network Eigenvector centrality is used to elicit the most central individuals within a network in terms of the overall network structure, at the expense of smaller, local networks (Hanneman 2005). For instanc e, two individuals may have similar levels of closeness, even though one may
32 be very close to a small, isolated group within a network, and somewhat far from the rest of the network; while the other may be at a relatively moderate distance from all individ uals within the network. In this case though the level of centrality might be similar between the two, the latter is more than likely a more central figure in the network since he/she could more easily access the entire network (Hanneman 2005). The follo wing undirected network graphic illustrates the concept of eigenvector centrality (Bonacich 1987) : In this case we have four network nodes: A through D. Looking at node B, we see that there are four connections to other network nodes (either incoming or outgoing), yielding a degree centrality value of 4. Node A only has three connections to other nodes, yielding a degree centrality of 3. However, notice that A is also connected to other sub networks through nodes C and D, each of which have a degree ce ntrality value of 4. Based on degree centrality, node A appears to be the least central actor in this network. However, when taking into consideration the network as a whole (i.e,. the eigenvector centrality of the node), node A is
33 actually the most cent ral (with a value of 0.182), with the remaining nodes less so (each with an eigenvector value of 0.091). Employing a factor analysis approach, the eigenvector centrality of an individual node is essentially calculated as the sum of the centrality values of the nodes that are connected to it. Thus the centrality of each individual node reflects the centrality of all adjacent nodes, identifying dimensions associated with the distances among actors throughout the entire network. For each dimension of distance, an individual eigenvalue is calculated. The collection of the many i ndividual eigenvalues makes up the eigenvector, which can be assigned to each network node (Hanneman 2005) Given these multiple approaches to network centrality (in degree, out degree and eigenvector) there is a sound reason to retain all three for inclusion in the study Degree centrality is a non weighted measured of centrality. In other words, every network contact is weighted equally. The benefit to using degree centrality is that it is a direct measure of incoming and outgoing contacts for an individual is robust a gainst missing network data and is one of the standard measures of centrality in studies employing social network analytics (Borgatti 1992). The drawback of using only degree centrality is that it only considers immediate incoming and outgoing connections In other words (as discussed above), it does not take into account the position of the individual in the network as a whole only immediate ties (Hanneman 2005). T o address this, we can also retain eigenvector centrality which relies on a factor anal ysis approach to identify and evaluate the dimensions of distances [i.e., eigenvalues] between actors across the entire network This allows us to generate a measure of the most central actors across the network as a whole ( Hanneman 2005). Given this, I have retained degree (out degree and in degree) centrality and eigenvector centrality as the network measures of interest.
34 Having identified the network variable of interest (centrality), we can turn to the potential association with learning within a partnership. As discussed earlier, research has shown that higher network density may lead to generation of trust within groups, and also facilitate learning or information exchange (though not in all cases). In a study of learning and dispersion of information within social networks, Jabbabaie, et al (2013) found that the rate of learning and subsequent informational exchange within a social network is positively associated with the centrality of individuals within that network. Likew ise, in a large scale study of students in a team based MBA program, Baldwin, et al (1997) found a positive and significant correlation between student network centrality, cognitive performance (grades), and overall student perceptions of learning and prog enabled students to avail themselves of resources and support to a greater degree than their less learning, Russo, et al (2007) found that network prestige (in degree centrality) and the level of external outreach (out degree centrality) were each positively correlated with cognitive learning. Further, this study showed that the lionshare of cognitive learning was associated with out degree centrality, and much less so in degree centrality. Studies of networks within collaborative partnerships have also indicated that in the case of individuals, a high level of centrality can enhance learning for that individual within his or her network (Hanneman 2005 ; Berdardo 2009 ). Similar to our discussion of density, this is related to access. For example, one might consider that an individual with a high level of centrality, hin a partnership, and would have the opportunity to interact with many individuals when compared to a less central actor. The actor with a high level of centrality would benefit from that position in terms of access to information, understanding of
35 opini ons within the network; each of which would increase the potential for learning (Berdardo 2009) In other words, a highly central individual might have the benefit of learning from multiple actors within a network based on an increased potential for inter action. Increased potential for learning due to centrality may have several implications. It could influence how knowledge and understanding flows within a partnership which can impact the relative power and/or influence an individual wields For inst ance, an actor located in a central central (Berdardo 2009, 180 181; H anneman 2005). Given this, it was selected as a variable of interest for this study leading to the following hypothesis: Hypothesis 6 (H 6 ) Centrality and Learning (at the individual level) : The higher centrality of an individual among members of c ollaborative partnerships, the higher the level of reported learning among those same members of the partnership This leads us to the final hypothetical that was tested as a part of this study. As mentioned before, individuals with a high level of centrality within a network may benefit by having a high level of power and/or influence ( Berardo 2009; Hanneman 2005) increased cognitive learning (Russo 2007) and a potential for greater organization al satisfaction (Baldwin 1997) However, less is known at the individual level whether centrality explicitly translates to increased levels of trust In terms of partnerships, does centrality impact individual perceptions of trust for members of the part nership? Research indicate s th at it may. Brass and Burkhardt have suggest ed that the informational advantage created by the centrality of an individual within a network should result in enhanced opportunities for informational exchange which may in turn be the foundation of
36 building trust with other network actors (1993). Following this line of inquiry, Berdardo employed structural equation modeling on an organizational level to test if centrality of respondents to a National Estuaries Program survey (S chneider et al. 2003) was positively related to the level of trust towards others (2009). Similar to Brass and Burkhardt, h e found that higher centrality yielded higher level s of trust towards others and this trust was based on interactions associated wi th information exchange (as suggested by Brass and Burkhardt) The following hypothesis was tested to advance the discussion on the impact of individual centrality on perceptions of trust within partnerships : Hypothesis 7 (H 7 ) Centrality and Trust (at the individual level): The higher the centrality of an i ndividual among member s of c ollaborative partnerships, the higher the level of reported trust among those same members of the partnership As can be seen in the flow diagrams presented in Chapter I, there is the potential for the arrows of influence to be reversed. In other words, perhaps trust and learning are influencing network centrality and density. For this study, this is not expected to be the case. First, the existence of the partnership itself can be thought of as the experimental treatment. Prior to the partnership, partnership networks would not be present. It makes logical sense then for the networks to form, and from those ne twork ties, trust and learning to follow. This assertion is generally supported in the literature. For instance, as Newig found, the existence of dense networks provides for greater opportunity for information exchange and subsequent learning (2010). Li kewise, as indicated earlier, a high level of individual centrality can enhance learning for that individual within her network, due to enhanced opportunity to interact with other network nodes (Hanneman 2005; Berardo 2009). Similarly, Berardo (2009) asse rted that actors with a high level of centrality would benefit from that position in terms of access to information, understanding of opinions within the network; each of which would increase the potential for
37 enhancing understanding and learning. Even wi th support from the literature, this is not to say that it would be impossible for the causal arrows to be reversed. Again, there is a distinct possibility that learning and trust can influence the partnership network density and individual centrality. G iven this, this results presented later in this study should be interpreted with care.
38 CHAPTER I V METHODOLOGY Design The research design was a cross sectional on line survey of individual members of 10 marine aquaculture partnerships across the United States coupled with preliminary and formal interviews For the purposes of this study, marine aquaculture partnerships were defined as organizations that include governmental and non governmental groups that collaborate on policy or research (or both) to support the development or regulation of the marine aquaculture industry. To be considered for inclusion in the study, the partnership must be c ollaborative in the sense that it must have some combination of the following: consensus based rules, a focus on prob lem definition, deliberative processes open to participation, and employ fair rules of negotiation. A cursory list of partnerships was identified by conducting an iterative search on t he and membership lists. P reliminary interviews were conducted with individuals knowledgeable about marine aquaculture policy, including representatives from industry, academia, government, nongovernmental organizations and non affiliated citizens. Preliminary interviews were used to provide a reality check regarding the partnership list (i.e., those that are intended for inclusion in the study ) to develop a deeper understanding of the marine aquaculture policy landscape and to assist in generation of an advisory committee. The advisory committee consisted of individuals identified via review of relevant policy literature (inclusive of researchers, policy ad v ocates, and government officials), information obtained from internet searches, and information obtained from specific government and non government web sites. Members of the advisory committee
39 were interviewed to determine relevant policy issues appropriate ness of operationalized measures and also to provide an evaluation of the survey prior to distribution to the sample population. Modifications to the survey were made based on input from the advisory committee as well as results of preliminary interviews Formal interviews were conducted with approximately 2 5 members of each partnership. Once the partnership list was finalized, each partnership coordinator was contacted to explain the purpose of the study, verify that the organization ha d be en accurately identified as a partnership, gain acceptance for inclusion in the study, obtain names of members for survey distribution, and to provide additional partnership names (using a snowball sampling approach) if appropriate. Sample Population Aquaculture Partnerships A brief discussion of each of the 10 partnerships selected for this study is included below. Note that the descriptions were taken from the Aquaculture Partnerships Project Summary Report which was developed in December 2011. Pac ific Aquaculture Caucus The Pacific Aquaculture Caucus (PAC) is an aquaculture collaborative in Washington State that was formed circa 1998 with th e assistance and funding of NOAA The PAC was one of several caucuses that NOAA funded with the general goal of assisting the development of aquaculture at the regional (i.e., multistate) level throughout the United States. Originally intended to encompass the entire pacific region (Washington State, California, Hawaii, Oregon, and Alaska), the PAC has been pre dominantly active in the state of Washington. The mission and freshwater aquaculture for the Pacific region through sound public policy and best available science.
40 strategies: (1) assist local governments with aquaculture regulations, (2) support production systems that address economic and environmental systems, (3) encourage bes t management practices, (4) encourage collaboration, (5) promote and assist with scientific research, (6) utilize collective expertise, (7) take on specific tasks to resolve issues around aquaculture, and (8) provide a central point for information dissemi nation. California Aquaculture Development Committee The California legislature created the Aquaculture Development Committee through the California Aquaculture Promotion Act of 1995 (Assembly Bill 1636). The Committee consists of 12 members representing various state agencies, the University of California, and all sectors of the freshwater and saltwater aquaculture industry. The committee is advisory to the director of the Department of Fish and Game, and is chaired by the Department's Aquaculture Coordi nator. By law, the Committee is charged with identifying opportunities for "industrial development" and "regulatory relief." The membership was informally expanded in 2008 to include two organizations that are critical of aquaculture industry development The Ocean Conservancy and the Monterey Bay Aquarium. In recent years, the Committee has focused on implementation of the Sustainable Oceans Act of 1996 (SB 201), which prohibits marine finfish aquaculture in state waters without a lease from the Fish and Game Commission. This legislation made California the first state in the nation to enact stringent environmental standards for marine finfish aquaculture. The act requires the Department to consult with the Aquaculture Development Committee to prepare a Programmatic Environmental Impact Report for existing and potential commercial aquaculture operations. The report expected to be finalized in late 2011.
41 Shellfish Aquacul ture Regulatory Committee The Shellfish Aquaculture Regulatory Committee (SARC) w as formed by the Washington State legislature through House Bill 2220 (Chapter 216, Laws of 2007). The permit process for all current and new shellfish aquaculture projec ts and activities that integrates all applicable existing local, state, and federal regulations and is efficient both for the regulators nascent industry of geoduck cl am farming in Washington State. The Committee issued its final report in 2009 and held its last meeting in June 2010. Florida Marine Net Pen Working Group The Florida Marine Net Pen Working Group (FMNPWG) was formed by the Florida Division of Aquaculture (seated within the Florida Department of Agriculture) to develop best management practices associated with the installation and operation of marine net pens within Florida state waters, and to address any operational and management issues that could impact marine resources, site selection, feed management, nutrients, escape, solid waste, and general facility management. Additionally, the Division of Aquacult ure saw a general need to develop guidance for fish farming that incorporated both economic and environmental considerations. The Division of Aquaculture created the FMNPWG partly as a proactive measure in anticipation of potential federal regulation of n et pens (commensurate with the 2005 release of the National Offshore Aquaculture Act). The FMNPWG consists of approximately 15 20 members.
42 Florida Aquaculture Review Council The Florida Aquaculture Review Council (FARC) was created within the Florida De partment of Agriculture and Consumer Services (FDACS) under Florida Statute 597.005 in the late 1980s. According to the FDACS, the FARC was created to provide a forum for communication between the aquaculture industry and the Department. The FARC consists of nine members, including the aquaculture representative on the State Agricultural Advisory Council, the chair of the Aquaculture Interagency Coordinating Council, an alligator farmer, a food fish farmer, a shellfish farmer, a tropical fish farmer, an aqu atic plant farmer, a representative of the commercial fishing industry, and a representative of the aquaculture industry at large. The FARC meets a minimum of four times per year. According to the FDACS, the responsibilities of the FARC are to "recommend r ules and policies governing the business of aquaculture to the Commissioner of Agriculture and to annually submit a list of recommendations for short term research projects designed to solve problems designated in the state aquaculture plan. The FARC revie ws and discusses problems that serve as barriers to the growth and development of aquaculture and has been a key in the continued growth of the aquaculture industry in Florida" ( www.floridaaquaculture.com ). Maine Aqu aculture Advisory Council T he Maine Aquaculture Advisory Council (MAAC) was created in 1995 under the Chapter 24 Regulations of the Maine Department of Marine Resources (DMR). The formal creation of the MAAC was preceded by an ad hoc industry bas ed advisory committee. The purpose of the ad hoc advisory committee was to convene a group of individuals from the aquaculture community who could provide insight on various aquaculture related issues associated with the increasingly prominent aquaculture industry. The MAAC consists of four to
43 five members, including aquaculture producers and government representatives. Its primary purpose is to provide recommendations to the Commissioner of Marine Resources on issues identified by the DMR, members of the p ublic, and/or other members of the aquaculture industry. Maine Fish Health Technical Committee The Maine Fish Health Technical Committee (MFHTC) was created in 1994 in response to the growing aquaculture industry in Maine and related concerns regarding fis h health and public safety. The MFHTC serves both the Maine Department of Inland Fisheries and Wildlife (IFW) and the Maine Department of Marine Resources (DMR) and deals specifically with providing recommendations for managing disease outbreaks and issues of fish health in the State of Maine and in the region more broadly. For example, the MFHTC makes recommendations on establishing testing requirements, site selection requirements, and aquaculture facility management techniques to prevent the spread of ha rmful fish health diseases. The MFHTC was originally formed as an ad hoc advisory committee to advise public officials on issues relating to fish health and was later formally mandated under statute in the Chapter 24 Regulations through the Maine Departme nt of Resources. The MFHTC is required to have representatives from private industry, academia, and state and federal government. Maryland Aquaculture Coordinating Council Maryland State legislation enacted in 2005 created the Maryland Aquaculture Coordin ating Council (MACC), comprised of 17 designated members from industry, academia, regulatory, and political categories. The MACC guides the responsible development of the aquaculture industry within the state The duties of the MACC include making annual proposals to the Governor and General Assembly for advancing the industry, conducting studies of projects
44 and products that will lead to expanding the industry, developing best management practices (BMPs), and providing for the establishment of Aquaculture Enterprise Zones. The MACC also periodically reviews state regulations impacting aquaculture and makes recommendations on any necessary or advisable regulatory changes. New Jersey Aqua culture Advisory Council The New Jersey Aquaculture Advisory Council (NJAAC) was created under the New Jersey Aquaculture Development Act in 1998. Members of the aquaculture industry seeking more representation in the regulatory decision making processes impacting their practices supported the creation of the NJAAC. The t wo state agencies involved with managing the Council are the New Jersey Department of Agriculture and the New Jersey Departme nt of Environmental Protection The primary purpose of the NJAAC is to provide a forum in which stakeholders from multiple sectors can share information and make recommendations regarding aquaculture issues and policies. The NJAAC is r egarded as an advisory entity and is comprised of 20 40 individuals representing state governmental officials, university researchers, and aquaculture producers The Rhode Island Coastal Resources Management Council approach to aquaculture has been driven by the dual goals of protecting the public trust while at the same time encou raging a sustainable aquaculture industry that respects the traditions of the state. In the past decade the Rhode Island aquaculture industry has been growing at double digit s and the waters above them. The CRMC, believing that including as many interests as practical in the regulatory process is essential to achieving a durable consensus on contentious issues, formed the Rhode
45 Island Aquaculture Working Group (RIAWG) in 2000. The RIAWG routinely met until 2001, when the participants decided that the critical issues driving its formation had been explored to Narragansett Bay and resulted in changes to CRMC regulations. The initial meetings also set a precedent for increased communication between regulators and the aquaculture industry which aquaculture l membership has ranged from 10 25 members. O perationalization Measuring Coordination Networks in Aquaculture Partnerships The survey measured indicate the groups that you regularly coordinate with on aquaculture policy issues related to the [partnership name] roster containing the following groups known to be involved with aq uaculture policy : Native American Tribe, US Department of Agriculture, National Oceanic and Atmospheric Administration, Army Corps of Engineers, US Fish and Wildlife Service, US Environmental Protection Agency, US Food and Drug Administration, State Enviro nmental Protection Agency, State Department of Agriculture, State Department of Natural Resources/Division of Wildlife, Local Governments (city/county/districts), Finfish Aquaculture, Shellfish Aquaculture, Other Aquaculture (e.g., aquatic plants), Commerc ial Fisheries, Recreational Fishing, Shoreline Developers, Environmental Groups, Consultants, University Cooperative Extension, University Researchers and News Media. Respondents were also provided an opportunity to manually enter group names if needed. The roster approach, coupled with an opportunity for manual entry for organizations inadvertently left off of the roster, is an effective way to alleviate problems associated with recall
46 bias that typically haunt network data collection efforts (Henry 2 012). However, use of the roster does run the risk of omitting important and strong linkages within the network, depending on how comprehensive the roster is. During development of the survey associated with this study, in addition to extensive aquacultu re policy research, we incorporated informal interviews as well as an advisory committee to develop the roster. This approach coupled with a write in option, ensured that network omissions would be minimized. Use of the roster approach has been suggested in circumstances where the networks are likely to have well defined boundaries with the potential for many ties (Henry 2012). The completed rosters were then used to generate a coordination network for each respondent and for each partnership Coordination network data was used to directly evaluate hypotheses H4 H7 In order to do this, a network density variable was generated at the partnership level (n=10) for use in evaluating hypotheses H4 and H5 Likewise, an eigenvector centrality variable was generated at the individual level (n=110) for use in evaluating hypotheses H6 and H7 Both the network density variable (at the partnership level) and the eigenvector centrality variable (at the individual level) were generated using UCInet a social network analytical software package The following schematic illustrates how the network data were used to evaluate H4 H7 : Survey: Roster Results Post Processing: UCINet to generate Coordination Networks Independent Variables : H4/H5: Partnership Density H6/H7: Individual Centrality Dependent Variables: Reported Learning Perceptions of Trust Statistical evalaution of hypotheses
47 Coordination network data were also used to provide additional evaluation of H1 (the belief homophily hypothesis) The importance of coordination factors are measured as described below (Operationalization Measuring the Importance of Coordination Factors). However, in addition to this, the network data were also used to evaluate whether or no t individual coordination networks were associated with shared policy core beliefs (in an effort to further evaluate the belief homophily hypothesis, as discussed in Operationalization Measuring the Policy Core Beliefs). The following schematic shows ho w the network data were used to further evaluate H1 : Operationalization Measuring the Importance of Coordination Factors H ypotheses H1, H2 and H3 (i.e., the relative importance of multiple factors influencing coordination network formation ) were add ressed by asking (based on a five point scale with 1 being not important at all, and 5 being very important). Respondents were presented with statements to represent factors associated with belief homophily, trust, and resources. Survey Results: Coordination Roster Policy Core Beliefs Post Processing: UCINet to generate Coordination Networks and Beliefs Matrix Variables : Policy Core Beliefs Coordination Networks Statistical evalaution to further test H1
48 Belief Homophily Belief homophily was operationalized using the following statement: They share my beliefs on major aquaculture policy issues. The belief homophily statement was developed based on ACF surveys historically used in watershed and marine policy subsystem research where belief networks are created by asking respondents to indicate which groups they tend to agree with on specific polic y issues (Weible 2005; Zafonte and Sabatier 1998). Trust Trust was captured through a battery consisting of the following three statements: I trust them to keep their promises. They are professionally competent. I have worked with them in the past. The attributes of trust selected for inclusion in the survey were informed by past studies of trust in collective action settings (Lubell 2007) and evaluations of social and institutional trust (Lubell 2004). Note that perceptions of trust within the part nership which were measured to address hypotheses H5 and H7 (see below) are distinctly different that the measure of trust used to determine factors driving coordination ( H1 H3 ) Resource Deficits Resources were operationalized using a battery consistin g of the following four statements: They have influence outside the partnership They have influence inside the partnership They have access to financial resources
49 They have access to expertise on major aquaculture policy issues. Resource attributes wer e developed based on a review of historical Resource Dependence literature (Pfeffer and Salancik 1978) as well as more recent operationalizations of various resource attributes, including influence, access to funding, as well as policy expertise (Casciaro and Piskorski 2005; Weible 2005). Legal Mandate Finally, respondents were asked to rate the following statement with regard to its importance in coordination decision making: There is a legal requirement. This was done to determine if there exists an overarching legal mandate for coordination due to operational rules, or legal prescription that would dominate coordination network formation over and above the behavioral attributes listed above. For the trust and resources hypotheses, we conceptualize the multiple survey questions as discrete reasons for coordination rather than complementary facets of a single underlying concept (which might be captured in terms of internal reliability, such as factor analysis) In other w d reason for coordination This approach is a very direct way (i.e., explicit question) of measuring factors underlying network or identi fy particular network partners. Operationalization Measuring the Policy Core Beliefs As indicated earlier, H ypothesis H1 was further evaluated by measuring respondent beliefs on major aquaculture policy issues Drawing from the ACF literature as wel l as prior
50 surveys historically used in watershed and marine policy subsystem research (Weible 2005; Zafonte and Sabatier 1998), policy core beliefs (i.e., beliefs on aquaculture policy issues) are the ive belief systems (the others being deep core and secondary). Policy core beliefs geographic breadth of a policy subsystem; their specificity makes them ideal for forming coalitions and coordinating acti vities among policy susbsystem members. These policy core Smith 1999, pp. 133) and are the primary mechanism binding advocacy coalitions together (Zafonte and Sabatier 1998, Sabatier and Jenkins Smith 1999). Given the support in the literature regarding the role of secondary beliefs in coalition formation and coordination behavior, they are valuable measures with which to test coordination behavior. B eliefs on aquaculture policy iss ues (policy core beliefs) were measured by providing a battery of the 12 following statements regarding marine aquaculture policy in the United States, and asking each respondent to indicate their level of Marine shellfish aquaculture must be expanded in U.S. waters Marine finfish aquaculture must be expanded in U.S. waters Existing marine shellfish aquaculture facilities in the United States are ecologically sustainable Existing marin e finfish aquaculture facilities in the United States are ecologically sustainable The best strategy for managing marine aquaculture involves sustained dialogue among all stakeholders The U.S. marine aquaculture industry is already too heavily regulated Adverse risks to the natural environment outweigh the benefits of marine aquaculture External verification and certification programs provide the necessary incentives to develop a sustainable marine aquaculture industry
51 The expansion of U.S. marine aqua culture will provide a significant supply of sustainable and healthy seafood offsetting the trade deficit Marine aquaculture will diversify coastal economies Marine aquaculture threatens the livelihood of commercial fishers Marine aquaculture allows for the continuation of maritime heritage As indicated earlier, survey responses to the policy core beliefs battery were entered into UCInet to generate a beliefs matrix for respondents. The beliefs matrix was used in conjunction with the coordination netw orks (discussed above) to determine the association between beliefs and network structure. Specifics are discussed in the Statistical Approach and Level of Analysis section. Operationalization Measuring Learning Within Partnership s H ypotheses H4 and H 6 ( i.e., the impact of network density and individual centrality on learning) were addressed by asking individuals to indicate their level of agreement (ranging from 1 = Strongly Disagree to 5 = Strongly Agree) with the following statements: Participation i n the partnership has given me a better understanding of aquaculture science. Participation in the partnership has given me a better understanding of aquaculture policy, law, or regulations. Participation in the partnership has given me a better understanding of aquaculture economics or business. Participation in the partnership has given me a better understanding of other stakeholder perspectives T o further explore the nature and extent of learning taking place, individuals were also prompted to respond professional opinion on any significant scientific or technical issues related to marine t least partially through your participation in the partnership, have you changed your professional opinion on any significant policy
52 Operationalization Measuring Trust T o address hypothes e s H5 and H7 (i.e., the i mpact of network density and individual centrality on trust) p erceptions of trust within the partnership were measured by asking respondents to indicate whether the following statements apply to none, few, half, most, or all partnership participants: Are h onest, forthright, and true to their word Are willing to listen and sincerely try to understand other points of view Reciprocate acts of good will or generosity Are trus t worthy These metrics of trust, associated with perceptions of partnership members, are based on the concepts put forth in social capital and collaborative natural resource governance literature, and are based on the concepts of social trust advanced by Putnam (19 93), Stoker (2000) and Lubell (2004). S tatistical Approach and Level of Analysis Survey responses were evaluated using SPSS and STATA statistical software package s For clarity, I have re stated each hypothesis below and included a statement regarding the stat istical methodology that was employed to test each hypothesis, as well as the level of analysis : --------------------------------------------------------------------------------------------------------------------(H1) In collaborative partnerships agreement on major policy issues between members will be an important factor in the formation of individual coordination networks (H2) In c ollaborative partnerships trust between members will be an important factor in the formation of individual coordination networks (H3) In c ollaborative partnerships perceived resource attributes between members will be an important factor in the formation of individual coordination networks
53 Statistical approach: The first three hypotheses were tested using one way analysis of variance (ANOVA) for mean scores of reported level of importance for each variable. A social network analysis tool called the Q uadratic A ssignment P rocedure (QAP) a technique for calculating the association between two m atrices (Dekker et al. 2008) was used to determine the existence and degree of correlation between individual responses to the policy core belief battery of questions and coordination network matrices. The QAP is a tool (associated with UCINet) common ly used in social network analysis and is particularly useful since it does not rely on the assumption of independence of observations. For example, with social network data (such as the coordination data associated with this study), one would reasonably assume that if individual x reported some level of coordination with y, then individual y would probably report some level of coordination with x. Thus, the observations cannot be assumed to be independent (this comes up again the following sections). I n the context of this study, the QAP will be used to determine if policy core beliefs held by a particular participant align with those with whom s/he is coordinating. So how does the QAP approach this problem? Fundamentally the QAP is comparing two mat rices The goal is to determine the significance and size of the correlation between two matrices when compared to that which would result when randomly permuted. Using this approach, the QAP runs thousands of iterations (i.e., permuting one of the matri ces [same permutation for rows and columns] and holding the other constant) to determine if there is a significant difference between the distribution of random permutations versus the original. The QAP returns a Pearson R coefficient (effect size) as wel l as a p value to indicate the significant of the effect size. Similar to Ordinary Least Squares output, an insignificant result indicates that
54 no significant relationship exists between the two matrices (i.e., the null hypothesis is supported). The QAP is discussed further in Chapter VI. Level of analysis: Individual responses to the questions regarding the importance of coordination factors were combined and evaluated (via means testing) for all respondents Responses were also evaluated at the partnership level and by respondent affiliation For purposes of the QAP, individual responses to the policy core beliefs battery and individual responses to the coordi nation network question were used to generate a single correlation coefficient for all respondents combined (N=110) as well as for multiple organizational breakdowns to test for data sensitivity --------------------------------------------------------------------------------------------------------------------(H4) The higher the network density for a collaborative partnership, the higher the level of reported learning on average across members of the partnership (H5) The higher the network density for a c ollaborative partnership the higher the level of reported trust on averag e across members of the partnership Statistical Approach: Network density values were calculated for each partnership using UCInet Individual responses for each question regarding learning and reported trust within the partnership were combined to generate a mean score for those variable s for each partnership. Linear regression modeling was used to determine if network density was significantly related to reported learning and/or trust at the partnership level Regression models were evaluated for each individual learning and trust question as well as for scale variables created for each trust and learning battery of questions (i.e., one mean score per partnership for all trust questions and one mean score per partnership for all learning quest ions ). Due to the small sample number associated with this portion of the study (i.e., 10 partnerships) in addition to performing regression analysis, a descriptive evaluation of the data will be performed as well to complement results of regression mod eling. However, that being
55 said, in spite of the small sample number in this portion of the study (10 partnerships) I remain confident in the validity of regression modeling results. Following the approach taken by Maggioni et al (2012), there is a theor etical basis for this assertion : the aquaculture partnerships included in this study represent almost the entire population of aquaculture partnerships in the United States. Therefore, issues associated with obtaining convergence between both sample and population parameters should be minimized since we are dealing with what is close to the entire sample population. Second we are employing simple linear regression modeling, rather than multiple regression approaches. Therefore, the regression model wi ll only rely on one predictor variable (density) rather than multiple This effectively addresses the large predictor/small sample size issue typically cited for small n studies relying on multiple regression techniques Third, since we do have a small sample size to work with for this portion of the study, violations of hetero and homoskedasticity (typical for small n situations) can impact ordinary least squares regression results. To address this, I employed a weighte d least squares regression technique that is designed to downweight residuals associated with observations that have large variances. In this case, the weighted least squares estimators were weighted on the network density variable. This variable showed the greatest amount of variation (compared to the learning and trust variables) and should therefore provide conservative model estimates. Fourth and outside of the robustness argument, I have incorporated the use of some standard descriptive analytical techniques to support the regression modeling approach With this combination, the results observed in this study should stand. Level of Analysis: These hypothes e s were tested at the partnership level (n=10) -------------------------------------------------------------------------------------------------------------------(H6) The higher the centrality of an individual among members of c ollaborative partnerships, the higher the level of reported learning among those same members of the partnership
56 (H7) The higher the centrality of an i ndividual among members of collaborative partnerships, the higher the level of reported trust among those same members of the partnership Statistical Approach: Individual centrality values (in degree, out degree and eigenvector) were generated using a combination of approaches, depending on the centrality measure used. Out degree centrality was based on the number of citations emanating from an individual (i.e., the number of times a respondent cited an organization ). In degree centrality was based on the number of times an individual received a citation from another individual. Since the survey was anonymous, individuals were asked to cite an organization with which they coordinated on major aquaculture policy iss ues. This resulted in an in degree centrality value for each organization or affiliation. The in degree centrality value for an organization/affiliation was then assigned to each respondent reporting that affiliation on the survey. This approach, though necessary given the survey design, requires an innate assumption that when an affiliation is cited (such as aquaculture industry), all respondents reporting that affiliation were being cited equally. Eigenvector centrality was calculated by generating a citation matrix (similar to that for in degree centrality on an affiliation by affiliation basis), entering that matrix into UCInet and selecting the centrality measures tool. Individual responses to the learning and trust batteries were then used for s imple linear regression analysis (employing a cluster approach to control for observation independence) to determine if centrality i s a significant predictor of reported learning and/or trust. S tatistical analysis was performed on each individu al learning and trust question. Additionally, a scale variable was constructed to indicate a general level of trust (i.e., mean score for all trust questions) and learning (i.e., a mean score for all learning questions) a nd subsequently used in linear reg ression analysis
57 Logistic regression models (also employing a cluster approach to control for observation independence) were used to determine the explanatory power of individual centrality on learning hen asked if they had changed their opinion on any technical or policy issue associated with marine aquaculture see Operationalization Measuring Learning ). Level of Analysis: These hypothes e s were tested at the individual level Sample number (N) wa s approximately 110. Interview Data As discussed above, interviews were conducted with approximately 2 5 members of each partnership ( for a total of 47 ). The purpose of the interviews was multifold: provide study context, provide feedback on survey development, increase overall study response rate, and provide qualitative data to enhance assessment of quantitative data. Note that the questions used for this study are a part of a larger in terview protocol. Specifically, the interview was used to provide insights regarding tendencies of coordination for individuals (see H1, H2 and H3 ) by posing the following questions and probes: Which groups in this partnership do you tend to coordinate wi th most frequently? Probe: Why do you tend to coordinate with those groups and not with others? Probe: Can you tell me the types of coordination activities that you engage in? Interview responses were used on a qualitative basis, looking for general respon se patterns across the interviews to support, clarify, or refute finding from the survey data. I nterview data were not evaluated on a quantitative basis
58 CHAPTER V RESULTS DEMOGRAPHICS This section includes an overview of the survey response rates (o verall and for each individual partnership), respondent characteristics (sex, age, level of education, reported competencies, environmental perception and political leanings ) and respondent affiliation. Survey Response Rates Of the 198 individuals to whom the survey was sent, 123 individuals responded, yielding a 62% response rate overall. Response rates for each partnership ranged from 52% to 86%. Survey response rates as well as individual question response rates are presen ted in order to determine if there is the potential for non response bias that might impact study validity. In a study of journals across multiple disciplines, survey response rates of published research efforts typically varied between 24 and 82%, with a n average of 42% (Carley Baxter 2009). Response rates observed for the survey as well as for individual survey questions are well within the range presented by Carley Baxter. However, in a portion of this study data were aggregated to the partnership lev el. Therefore, suboptimal response rates may impact the ability to validly measure network densities (on the partnership level), even though network measures tend to be robust against low response rates (Borgatti 2002). Even so, results should be interpr eted with that understanding. Table V.1 displays the percentage of responses per partnership.
59 Table V .1 Response Rates by Partnership Partnership Responses Total Response Rate Maryland Aquaculture Coordinating Council 10 17 59% Pacific Aquaculture Caucus 25 43 58% California Aquaculture Development Committee 18 28 64% New Jersey Aquaculture Advisory Council 12 23 52% Rhode Island Aquaculture Working Group 16 25 64% Shellfish Aquaculture Regulatory Committee 13 22 59% Florida Net Pen Working Group 10 14 71% Florida Aquaculture Review Council 5 9 56% Maine Aquaculture Advisory Council 6 7 86% Maine Fish Health Technical Committee 8 10 80% Total 123 198 62% Respondent Characteristics Of the 123 individuals who responded to the survey, 21 (17.1%) of these were female and 89 (72.4%) were male. Thirteen respondents (10.6%) did not answer this question. Non responses are included for declarative purposes and to indicate the general potential for non response bias issues. As stated earlier, the response rates observed in this study are well within the range observed for published survey research (Carley Baxter 2009). A ge Of the 123 individuals who responded to the survey, one was between 20 29 years of age (0.8%), 13 were 30 39 (10 .6%), 14 were 40 49 (11.4%), 57 were 50 59 (46.3%), 20 were 60 69 (16.3%), and 6 were over the age of 70 (4.9%). Twelve respondents (9.8%) did not answer this question. E ducation Respondents were asked to indicate their highest level of education. Among t he 123 individuals who responded, one (0.8%) was not a high school graduate, four (3.3%) graduated
60 6 (29.3%) earned a PhD, MD, or JD. Eleven respondents (8.9%) did not answer this question. Respondents were also asked to indicate the disciplinary fields in which they are most competent on a scale from 1 5, with 1 having no competence and 5 being very competent. The field options included: Oceanography, Ecology/Biology, Engineering, Business or Economics, Policy/Law/Planning, and Fish/Shellfish Culture. The highest reported competencies included ecology/biology (mean reported competency of 3.7), as wel l as fish/shellfish culture (3.6), followed by policy, law, or planning (3.0). Oceanography and business/economics followed with mean reported competencies of 2.6 and 2.4 respectively. The lowest reported competency was in engineering (1.9). Political Leanings Respondents were asked to indicate their political leanings, ranging from Very Liberal to Conservative. Overall, survey r espondents are primarily moderate to liberal, with 52 (42.3%) reporting moderate political beliefs, 38 (30.9%) reporting eith er very liberal (4 ) or liberal (34) leanings. Seventeen respondents (13.8%) report ed conservative political beliefs and none indic ate very conservative. Sixteen respondents (13.0%) did not answer this question Environmental Perceptions To determine the general perception s that partnership participants have regarding the environment, respondents were asked to rate their level of agreement with a series of statements about the environment using a scale from 2 to 2, with 2=Strongly Disagree to 2=Stron gly
61 Agree. Table V 2 shows the mean levels of agreement with each statement among all respondents. Table V.2 Environmental Perceptions Environmental Statement Overall Mean (N=112 114) Plants and animals exist primarily for use by people. 0.7 The balance of nature is very delicate and easily upset by human activities. 0.5 The so called "ecological crisis" facing humankind has been greatly exaggerated. 0.3 We are approaching the limit of the number of people the Earth can support. 0.6 Scale: 2=Strongly Disagree; 2=Strongly Agree As Table V. 2 shows, partnership participants tend to hold what might be considered pro environmental views, disagreeing with the ideas that plants and animals exist primarily for use by people ( ( 0.3); and agreeing with the ideas that the balance of nature is delicate and easily upset by human activities (0.5), and that we are approaching the limit of the number of people the Earth ca n support (0.6). Affiliation Respondents were asked to indicate their respective organizational affiliation as far as their participation in each partnership was concerned. As is displayed in Tab le V .3 the majority of respondents are representatives f rom State Government (30.6%), Industry (28.1%), and Research/Extension (20.7%). Affiliations that are less well represented include Environmental Groups, Fisheries, and Consultants (9.9% combined), Federal Government (8.3%), and Local Government (2.5%).
62 Table V.3 Ove rall Respondent Affiliations Group Affiliation Number of Respondents Percent of Total Federal NOAA NMFS 7 5.8% US Department of Agriculture 1 0.8% US Fish and Wildlife Service 1 0.8% Native American Tribe 1 0.8% Subtotal 10 8.3% State State Department of Agriculture 9 7.4% State Department of Natural Resources/ Division of Wildlife 17 14.0% State Environmental Protection Agency 3 2.5% State Health Department 5 4.1% State Legislature 3 2.5% Subtotal 37 30.6% Local Local Governments (city/county/districts) 3 2.5% Research/ Extension University Researchers 14 11.6% Marine Research Not for Profit 2 1.7% Cooperative Extension 9 7.4% Subtotal 25 20.7% Industry Shellfish Aquaculture Producer 19 15.7% Finfish Aquaculture Producer 13 10.7% Aquaculture Feed 2 1.7% Subtotal 34 28.1% Other Environmental Groups 5 4.1% Commercial Fisheries 2 1.7% Consultants 5 4.1% Subt otal 12 9.9% Note: N=121 Compared to sample populations reported in earlier watershed studies (Leach 2006), the participants of this study appears relatively homogenous in several areas. For instance, respondents were primarily ma le and over the age of 50. Participants reported organizational affiliation s primarily with industry and government, with limited representation from environmental groups or from organizations that might stand in opposition to development of the marine aquaculture industry (such as homeowners, shoreline developers or marine fishermen). Similarly, the level of education of participants is generally high, with the majority holding an
63 advanced degree of some kind. Political leanings of respondents are also generally at or to the left of center, with very f ew reporting either very liberal or very conservative viewpoints.
64 CHAPTER VI RESULTS COORDINATION NETWORK S This section presents an overview of coordination with the partnerships as well as the results for Stage 1 of the study (hypotheses 1 3 associa ted with factors influencing the formation of coordination networks). The following flow diagram presents the overall scheme for this stage: Prior to addressing hypotheses 1 3, we will look at where coordination ties are being made on a broad level so t he factors that might be influencing the form ation of coordination networks within partnerships can be better understood. As indicated earlier, coordination networks were developed by allowing respondents to indicate (using a list prompt) which groups they tend to coordinate with on aquaculture issues. Figure V I .1 shows the percent of all citations that specific organizatio ns received by respondents for all partnerships combined. As can be seen from the figure, the most frequently cited organizations include University Researchers (8.5% of all citations), the Shellfish Aquaculture Industry (7.9%), the State Department of Na tural Resources (7.1%), and the University Cooperative Extension (6.6%). The least cited organizations (as a percentage of all citations) include the Army Corps of Engineers (0.4%), Coordination Network H3: Resource Deficits H1: Belief Homophily H2: Interpersonal Trust
65 Shoreline Developers (1.1%), the US Environmental Protection Agency (2.4%) and Native American Tribes (2.5%). Figure V I .1 Coordination Citations Percentage of all citations received. Taking this one step further, we will next evaluate who is coordinating with whom (rather than simply looking at which organization is receiving coordination citations). In Table V I 1 the affiliations on the top row are the citing affiliations and the affiliations on the left are the cited affiliations. All numbers are percentages of all citations for each citing affiliation to each cited affiliation. For example, looking down the first column in Table VI. 1, the 13 finfish industry representative s had a total of 68 citations for coordination, of which 19% went to finfish industry, 9% to shellfish industry, 3% to other aquaculture industry, etc As can be seen in Table V I 1 similar t o the results shown in Figure V I .1, most of the coordination citations went to three main groups: industry (pri marily finfish and shellfish with an average of 12 and 16% of coordination citations respectively), government (primarily state government with an average of 29% of citations) and science/research entities including researchers and university extension (wi th an average of 21% of citations). 2.5% 4.8% 6.0% 0.4% 4.3% 2.4% 2.7% 5.7% 5.9% 7.1% 4.1% 6.0% 7.9% 2.8% 4.7% 3.2% 1.1% 5.7% 3.9% 6.6% 8.5% 3.5% Native American Tribe US Department of Agriculture National Oceanic and Atmospheric Administration Army Corps of Engineers US Fish and Wildlife Service USEnvironmental Protection Agency US Food and Drug Administration StateEnvironmental Protection Agency State Department of Agriculture Local Governments (city/county/districts) Finfish Aquaculture Shellfish Aquaculture Other Aquaculture (e.g., aquatic plants) Commercial Fisheries Recreational Fishing Shoreline Developers Environmental Groups Consultants University Cooperative Extension University Researchers News Media
66 Table V I 1 Coordination Citations by Organization Finfish Shellfish Other Aqua Federal Gov State Gov Local Gov Com/Rec Fishing Science/ Research Consultants Env. Groups Other Average # of Respondents 13 19 2 10 37 3 2 25 5 5 2 NA # of Cites 68 90 32 263 213 47 89 172 44 65 53 NA Finfish 19% 8% 16% 13% 12% 13% 10% 9% 11% 9% 9% 12% Shellfish 9% 20% 9% 16% 18% 19% 18% 18% 11% 17% 17% 16% Other Aqua 3% 1% 0% 1% 0% 0% 0% 0% 2% 0% 0% 1% Federal Gov 12% 6% 9% 13% 4% 2% 7% 8% 7% 5% 8% 7% State Gov 26% 33% 34% 25% 36% 26% 33% 31% 25% 29% 21% 29% Local Gov 0% 3% 0% 0% 2% 6% 0% 1% 5% 5% 4% 2% Com/Rec Fishing 1% 1% 0% 1% 1% 2% 3% 1% 2% 2% 4% 2% Science/ Research 21% 20% 22% 23% 18% 19% 22% 24% 20% 20% 26% 21% Consultants 4% 2% 3% 3% 4% 4% 1% 3% 9% 5% 4% 4% Env Groups 3% 3% 3% 2% 2% 6% 2% 3% 5% 8% 4% 4% Other 1% 2% 3% 2% 2% 2% 3% 1% 2% 2% 4% 2% Total 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% 100% NA Based on this overview of coordination citations within these partnerships, the question that next surfaces is what factors are important when individuals are deciding with whom they will coordinate? This question (a s discussed in Chapter IV Methodology ) was addressed by asking coordinate with on aquaculture policy issues For clarity, the specific hypotheses associated with this question include : (H1) In collaborative partnerships agreement on major policy issues between members will be an important factor in the formation of individual coordination networks (H2) In collaborative partnerships trust between members will be an important factor in the for mation of individual coordination networks (H3) In collaborative partnerships perceived resource attributes between members will be an important factor in the formation of individual coordination networks
67 Table V I 2 includes each coordination factor the mean score of importance (ranging from 1 Not Important at All to 5 Very Important) the hypothesis associated with each coordination factor, as well as the groupings based on significant difference between means (discussed below) Table V I 2 Mean Scores for Importance of Coordination Factors Coordination Factor Associated Hypothesis Mean Scale of Importance (N=109 114) Significance Group They are professionally competent H2 4.2 A They have access to expertise on major aquaculture issues H3 3.8 AB I trust them to keep their promises H2 3.2 BC They have influence outside the Partnership H3 3.1 BC There is a legal requirement. N/A 2.8 CD I have worked with them in the past H2 2.4 DE They have influence in the Partnership H3 2.3 DEF They have access to financial resources H3 1.9 EF They share my beliefs about major aquaculture issues H1 1.8 F Scale: 1=Not Important at All; 5=Very Important As Table V I 2 s h ows, the most important factors for respondents when deciding with whom to coordinate are associated with trust ( professional competence mean score of 4.2; and trust to keep promises mean score of 3.2) as well as resources ( access to expertise on major aquaculture iss ues mean score of 3.8; and influence outside the partnership mean score of 3.1 ). The presence of a legal requirement (included specifically to determine if there is an institutional requirement that could be driving coordination) had a mean slightly bel ow the scale midpoint (2.8), followed by having worked with them in the past (2.4), influence in the partnership (2.3), and access to financial resources (1.9). The least important factor for respondents when deciding with whom to coordinate is belief homophily (i.e., shared beliefs about major aquaculture issues ; mean score of 1.8)
68 A one way ANOVA with post hoc testing was performed to determine if the differe nces in mean scores for coordination factors are statistically significant from one another The results of a one way ANOVA are considered reliable as long as the following assumptions are met: indepe ndence of observations, normality (or approximate norma lity) and eq uality of variances. Each of these assumptions was evaluated with the following violations noted: T he assumption of normality was not met for four of the factors shared beliefs (normality factor of 5.1) professional competence (normalit y factor of 5.7) access to expertise (normality factor of 3.8) and access to financial resources (normality factor of 4.0 ) ; and T he assumption of equality of variances was not met. Normality was evaluated using the SPSS software package by generati ng a normality factor The normality factor was calculated by dividing the skewness value by the standard error for each variable. When a normality factor is above |2.5|, the distribution of a variable is considered significantly different from normal (p <0.01). However, ANOVA is especially robust agai nst the assumption of normality (Leech, 2010) Therefore, the observed violation of normality is not expected to impact the reliability of the results. T he violation of the assumption of equality of varianc es can be remedied by selection of the appropriate post hoc test (in this case using Games see below ). Results of the A NOVA showed a statistically significant difference among the coordination factors F ( 8 990) = 51.1 p < 0.001 (assuming an alpha of 0. 01 for significance) Post Hoc Games Howell tests (used since equal variances cannot be assumed) were performed to determine significant differences between mean scores for coordination factors. Table V I 2 includes a row tit intended to show the statistically significant differences between mean scores based on the
69 Games Howell tests. Mean scores included in the same grouping group ) are not significantly differ ently from one another. They are, however, different from the means included in other groupings and so on ) As can be seen from the table professional competence w as significantly more important than all other factors (p<0.01), with the exception of access to expertise on major aquaculture issues Likewise, influence within the partnership access to financial resources and shared beliefs about major aquaculture issues were significantly less important than all other factors (p<0.01). The coordination factors with the highest mean scores in the grouped data were somewhat consistent when broken down for individual partnerships as well. For example, the coordination factors with the highest mean scores for all groups combined ( professional competence and access to expertise on major aquaculture issues ) had the highest mean score for every individual partnership with the excepti on of two ( NJAAC and SARC ). In those two partnerships, professional competence still had the highest mean score, followed by trust to keep promises (NJAAC) and a legal requirement (SARC) Similar to the data combined for all groups shared beliefs and ac cess to financial resources had the lowest mean scores for importance in six of the ten partnerships, with shared beliefs scoring lowest in eight of the 10 partnerships. T o determine the impact of organizational affiliation on the results observed in the grouped data, respondents were grouped by reported affiliation (based on f ive umbrella groups). Table V I 3 lists the general organizational affiliations across the top row, and coordination factors along the far left row. The five affiliations include t he aquaculture industry (finfish, shellfish, and feed/other aquaculture), government (federal, state and local officials as well as
70 elected officials), scientists and consultants (university and non university researchers, university extension representati ves and consultants), environmental groups, and other (commercial/recreational fishermen and Native American Tribes) All reported numbers are means, with coordination factors listed in order of decreasing importance when considering the mean of all respon dents (Grouped Data ; included for reference and comparison ). Means are cast in bold typeface to indicate the four most important coordination factors for all data combined (grouped data column). The means for those same coordination factors are also bold ed for the organizational affiliation columns as well for reference. Table V I 3 Coordination Factors by Organizational Affiliation Coordination Factors Aquaculture Industry Government Scientists/ Consultants Environmental Groups Other Grouped Data They are professionally competent 3.9 4.1 4.5 3.8 4.5 4.2 They have access to expertise on major aquaculture issues 3.7 3.7 3.8 3.6 4.5 3.8 I trust them to keep their promises 3.5 2.8 3.9 3.0 4.0 3.2 They have influence outside the Partnership 3.7 2.5 3.5 3.4 2.3 3.1 There is a legal requirement. 2.5 3.3 2.2 2.3 2.3 2.8 I have worked with them in the past 3.2 2.2 2.4 1.3 1.8 2.4 They have influence in the Partnership 2.3 2.0 2.7 2.0 3.0 2.3 They have access to financial resources 2.0 1.7 2.1 1.0 2.0 1.9 They share my beliefs about major aquaculture issues 2.4 1.5 1.7 1.0 2.0 1.8 Scale: 1=Not Important at All; 5=Very Important As indicated in Table V I 3 the coordination factors that had the highest mean scores in the grouped data were somewhat consistent for organizational affiliations. For example, the factors that had the highest four mean scores for all groups combined also had the highest four mean scores for three out of five organizational affiliations (aquaculture industry,
71 scientists/co nsultants and environmental groups). Respondents who reported a government affiliation retained professional competence and access to expertise as the factors with the highest mean scores (similar to the grouped data), but, not surprisingly, indicated the presence of a legal requirement was the next most important factor. Similar to the grouped data, those professional competence access to expertise and trust to keep promises as the most important coordination factors, with influence in the partnership following. Equally as interesting is the fact that shared beliefs was the least important coordination factor for three of the five organizational affiliations (government, scientists/consultants and environmental groups), an d was below the scale midpoint for all groups. Overall these results provide little support for the belief homophily hypothesis (i.e., that agreement on aquaculture policy issues will drive the formation of coordination networks in aquaculture partnersh ips In fact, shared beliefs appear to be one of the least important factors when individuals decide with whom to coordinate. Alternatively, attributes of trust and perceived resource s appear to play a more important role in forming coordination networks in aquaculture partnerships This is captured in the scoring of professional competence an attribute of trust (Lubell 2004), as one of, if not the most important coordination factor for respondents, followed by the resource attribute of access to expert ise the trust attribute of trust to keep promises and finally the resource attribute of influence outside the partnership Though slightly lower than the scale midpoint, havin g worked with someone in the past (a trust attribute) was significantly more important than access to financial resources and shared beliefs Thus, there appears to be some support for the resources and trust hypotheses, and given which specific attributes appear to be most important (a nd since the mean of the scores are not significantly different from one another), there may be an interaction (or relationship) between
72 the two. For example, respondents reported significantly higher importance for professional competence and access to e xpertise Though they are intended to operationalize specific concepts (i.e., trust and resources respectively), they share the idea of the expert; one relating to access to expertise and the other relating to the competence of that expertise. Finally, the presence of a legal requirement was significantly lower than professional competence and access to expertise This was especially important since the presence of a mandate for coordination could have had a substantial confounding effect on the result s of this study. Even for individuals reporting an affiliation with government, the presence of a legal requirement though highest for this group of respondents, was still significantly less important than professional competence and access to expertise Further Testing of Belief Homophily Even though respondents indicated that shared beliefs was a relatively unimportant factor when it came to deciding with whom to coordinate, there is the possibility that shared beliefs in fact, influence coordination choices unbeknownst to the respondent. In other words, perhaps beli e fs are subconsciously driving coordination behavior, influencing which actors are perceived as professionally competent, or influencing which actors are perceived as expert resources. This is a distinct possibility, since results for several previous ACF studies have indeed showed support for the belief homophily hypothesis. In an ACF study of the Swedish Carnivore Management System, perceived belief cor respondence was found to be the driving mechanism behind political coordination when compared to perceived influence (Matti and Sandstrom 2011). Likewise, in a study of conflict and cooperation networks associated with Swiss climate policy, it was found t hat the structure of those networks were a significant and close predictor of coalition belief systems (Ingold 2011). Given this, further investigation into the belief
73 homophily hypothesis is warranted. As indicated earlier, to address this possibility a social network analytical tool called the QAP was employed a technique used to determine the association between two matrices (Dekker et al. 2008). It is particularly useful for the analysis of social networks since it does not rely on the assumption of independence of observations. For example, with the coordination data generated in this study, one would reasonably assume that if an individual x reported some level of coordination with individual y then individual y might report some coordination wit h individual x Thus, the observations cannot be assumed to be independent. This is critical since other parametric statistical tests do require such an assumption. Since the QAP is a nonparametric technique that does not rely on this assumption, it is a useful tool for the analysis conducted in this section. Prior to running the QAP, it is necessary to first determine if there are significant differences in beliefs among the affiliations regarding marine aquaculture policy issues. In other words, if there are no significant difference in beliefs, then application of the QAP would be irrelevant. Thus, there would be no significant variability in the data to measure. A one way ANOVA was conducted with post hoc Games Howell testing to address this iss ue Table V I 4 lists organizational affiliations across the top row and marine aquaculture policy beliefs in the far left column. All reported numbers are means. The results indicate that overall there are significant differences between affiliations re garding the level of agreement with 10 of the 12 policy statements F ( 4 112) = 5.7 16.7 p < best strategy for manag For example respondents from aquaculture industry and scientists/consultants indicated significantly higher levels of agreement with regard to expanding
74 finfish and shellfish aquaculture in U.S. waters when compared to environmental groups and government representatives. Similarly, scientists/consultant s indicated a significantly higher level of agreement with the potential for marine aquaculture to diversify coastal economies when compared to government and environmental groups. Government and environmental groups also indicated a significantly lower l evel of agreement than industry with the statement that the aquaculture industry i s already too heavily regulated. Similar results (though opposite in direction) were observed with regard to adverse risks to the natural environment posed by aquaculture fa cilities (i.e., government and environmental groups showed a significantly higher level of agreement with the statement that adverse risks to the environment outweigh the benefits of aquaculture). Table V I 4 Marine Aquaculture Policy Beliefs by Organizati onal Affiliation Policy Statements Aquaculture Industry Government Science/ Consultants Environmental Groups Other Total Marine shellfish aquaculture must be expanded in U.S. waters 4.7 3.8 4.7 2.4 4.7 4.2 Marine finfish aquaculture must be expanded in U.S. waters 4.5 3.6 4.5 2.0 4.3 4.0 Existing marine shellfish aquaculture facilities in the United States are ecologically sustainable 4.7 4.0 4.4 2.8 4.3 4.2 Existing marine finfish aquaculture facilities in the United States are ecologically sustainable 3.9 3.5 3.8 1.8 3.7 3.6 The best strategy for managing marine aquaculture involves sustained dialogue among all stakeholders 4.2 4.1 4.2 4.4 4.0 4.2 The U.S. marine aquaculture industry is already too heavily regulated 4.0 3.0 3.6 1.6 3.0 3.4 Adverse risks to the natural environment outweigh the benefits of marine aquaculture 1.5 2.1 1.8 3.0 1.3 1.9 External verification and certification programs provide the necessary incentives to develop a sustainable marine aquaculture industry 2.8 2.8 2.8 3.6 4.0 2.9
75 Table V I 4 (Cont.) Marine Aquaculture Policy Beliefs by Organizati onal Affiliation The expansion of U.S. marine aquaculture will provide a significant supply of sustainable and healthy seafood offsetting the trade deficit 4.6 3.6 4.4 2.4 4.7 4.0 Marine aquaculture will diversify coastal economies 4.5 4.0 4.6 2.2 4.7 4.2 Marine aquaculture threatens the livelihood of commercial fishers 1.3 2.1 1.5 3.2 1.7 1.8 Marine aquaculture allows for the continuation of maritime heritage 4.4 4.0 4.4 3.0 4.7 4.2 Suffice it to say that since their appears to be variability in policy beliefs by affiliation, using the QAP will allow us to determine if there is a significant correlation between actors with similar beliefs, and those with similar coordination ties. In this case I used the QAP to determine if individuals that share common beliefs tend to share similar coordination citation tendencies. The beliefs matrix was developed based on responses to the 12 statements regarding marine aquacult ure policy presented in Table V I 4 and represent what can be considered policy core beliefs according to ACF conventions. The belief homophily hypothesis would predict that there should be a strong, positive and significant correlation across belief and coordination network s. T o conduct the QAP, two matrices were created with 11 rows (one for each of the affiliations listed in Table V I 1 and either 11 (citation matrix) or 12 (beliefs matrix) columns. Data in the citation matrix were normalized by dividing by the total num ber of respondents in a coefficients between each row to create two 11 by 11 similarity matrices (i.e., generating symmetrical matrices from non symmetrical metrical m atrices based on row similarities [using The QAP allows us to establish the existence, degree and significance of a correlation between the belief and coordination networks m entioned above by generating thousands of
76 random permutations of the independent matrix links (in this case beliefs). Subsequently, the procedure computes the proportion of coefficients generated from the random permutations that appear as extreme as the coefficient between the matrices. For the purposes of this study, significance is defined as the condition when a random permutation of links reveals a stronger correlation than that observed between the two networks less than 5% of the time. In other wo rds, 95% of the time the randomly permuted correlation is weaker than the actual observed. For purposes of the QAP, w e have defined the beliefs matrix as independent, leaving the coordination matrix as dependent. Thus, the QAP will fix the structure of the beliefs network and randomly permute the coordination network. This will allow us to test whether beliefs explain the respondent coordination network. This approach could have been reversed just as easily (i.e., fixing the coordination network and permuting the beliefs matrix) and would yield the same results. Results of the QAP showed no significant correlation bet ween the beliefs and coordination networks (Pearson R=0.02 p =0.32 ), supporting the results from the means testing performed earlier suggesting there is not a significant relationship between beliefs and coordination networks among respondents in this stud y. In addition to the five organizational affiliation grouping presented above, the QAP was also performed on the data when grouped into 5 organizational affiliations (as presented in Table V I.4 ) and on an individual basis (for 121 respondents). This wa s done to test the robustness of the results across different data groupings (this is equivalent to an evaluation of the sensitivity of the QAP). Similar to the results for the 11 organizational affiliation grouping, the QAP showed no significant correlat ion between the beliefs and coordination networks when considering the 5 organizational affiliation grouping (Pearson R=0.0 3 p =0. 3 5 ) or when considering individual
77 responses (R=0.0 3 p =0. 37 ), supporting the results observed for the 11 organizational affil iation grouping as well as the means testing results observed earlier. Results of Stage 1 of this study generally showed support for hypotheses 2 and 3, which rely on trust and perceived resource attributes as the impetus for coordination network formation Hypothesis 1, belief homophily, was not supported by the data collected in this study. These results will be discussed in more detail in Chapter XI. Table VI.5 recaps each hypothesis as well as the level of support observed. Table VI.5 Summary of Results Hypotheses 1 3 Hypothesis Supported Some Support Not Supported H1: In collaborative partnerships, agreement on major policy issues between members will be positively associated with the formation of individual coordination networks. H2: In collaborative partnerships, trust between members will be positively associated with the formation of individual coordination networks H3: In collaborative partnerships, perceived resource attributes between members will be positively associated with the formation of individual coordination networks. Having analyzed hypotheses 1 3 in Stage 1 of this study, we now move to Stage 2. Stage 2 focuses on hypotheses 4 7, which addresses how network position (individual centrality and network density) shapes learning and perceptions of trust within partnerships. The following flow diagram illustrates Stage 2 of the study:
78 We begin by providing a description of learning and trust observed within the partnerships in Chapters VII and VIII, respectively, followed by a description of individual centrality and network density in Chapter IX. Having established an understanding of learning, trust, individual centrality and network density within these partnerships, Chapter X focuses on an evaluation of hypotheses 4 7, which are aimed at evaluation how network position shapes learning and trust. Reported Learning Reported Trust H6/H7: Individual Centrality H4/H5: Partnership Density
79 CHAPTER VII RESULTS LEARNING WI THIN PARTNERSHIPS Having tested th ree complementary explanations for the formation of coordination networks (Stage 1) we now turn to learning within partner s hips (a component of Stage 2) As can be seen in the flow diagram below, we need to understand le arning within partnerships in order to address hypotheses 4 and 6 (i.e., how network position shapes learning). As discussed in Chapter IV Methodology learning in the partnership was captured by asking respondents to indicate their level of agreement (ranging from 2 = Strongly Disagree to 2 = Strongly Agree) with the following statements: Participation in the partnership has given me a better understanding of aquaculture science. Participation in the partnership has given me a better underst anding of aquaculture policy, law, or regulations. Participation in the partnership has given me a better understanding of aquaculture economics or business. Reported Learning H6: Individual Centrality H4: Partnership Density
80 Participation in the partnership has given me a better understanding of other stakeholder perspect ives to assess whether the four learning items could be combined to form a reliable overall learning variable This approach is commonly used to determine the appropriateness of combining multiple Likert type s cale variables into one (i.e., genera ting a single learning variable using responses from each of the individual learning questions). The alpha for the four items was 0.87 (with inter item correlations ranging from 0.59 to 0.68), indicating that the lear ning questions can be combined to make a scale variable with reasonable internal consistency reliability. Therefore, responses to the learning questions were averaged to form a single variable (referred to as overall learning). Table V II 1 shows the mean levels of agreement with the individual learning question as well as the overall learning variable for each of the 10 partnerships included in this study. Table V II 1 Reported Learning by Partnership Partnership Better understanding of other perspectives Better understanding of aquaculture science Better understanding of aquaculture policy, law, or regulations Better understanding of aquaculture economics or business Overall Learning CADC 1.17 .78 1.06 .89 .97 FARC 1.80 1.60 1.40 1.20 1.50 FMNPWG .90 .80 .60 .70 .75 MAAC 1.20 .80 .80 .60 .85 MFHTC 1.29 .71 1.14 .86 1.00 MACC 1.44 .89 1.33 1.00 1.17 NJAC 1.30 .30 1.10 .20 .73 PAC .68 .73 .68 .55 .66 RIAWG 1.40 .80 1.20 .60 1.00 SARC .38 .46 .54 .23 .40 Total 1.06 .74 .94 .64 .84 Scale: 2 = Strongly Disagree; 2 = Strongly Agree
81 As the mean scores from Table V II 1 indicate, there is general agreement that participation in these partnerships has provided opportunities for learning, with the most agreement in the areas of stakeholder perspectives (1.06) ; aquaculture policy, law or regulations (0.94); and aquaculture science (0.74). The least amount of learning (though still positive) was associated with aquaculture economics or business (0.64). Mean scores for individual partnerships were positive for each learning question, ranging from 0.23 to 1.80 Partnerships differed in mean score for overall learning, with the least amount observed in the SARC (0.40) and the highest in the FARC (1 .5). Figure V II 1 includes mean scores for the overall learning variable by organizational affiliation (separated into five distinct groups). As shown in the figure, some level of learning (i.e., positive mean scores for the overall learning variable ) was observed for eac h organization al affiliation with the highest level for respondents from environmental groups (1.10), followed by government officials (federal, state and local; 0.97), scientists /consultants (0.74) and aquaculture industry representatives (0.73). The gr (0.25) This group was comprised of three individuals: one representative from a Native American Tribe ( 1.25 mean for the overall learning variable) and two commercial fishermen (each with a mea n of 1.00 for the overall learning variable).
82 Figure VII.1 Reported Learning by Organizational Affiliation. Survey results indicate that learning is occurring within the se aqu aculture partnerships, specifically in the form of better understanding (i. e., better understanding of stakeholder perspectives, aquaculture science, aquaculture policy/law/regulations, and aquaculture business/economics). Though better understanding is certainly an indicator of learning, whether or not this leads to actual belief change is an equally important indicator. T o further explore this aspect of learning individuals were st partially through your participation in the partnership, have you changed your professional opinion on any significant scientific or technical issues related to marine you changed your Of the participants who responded to these questions, 41 percent indicated they had changed their opinion on a significant scientific or technical issue 46 percent indicated they had changed 1.10 0.97 0.74 0.73 0.25 0.84 0.00 0.20 0.40 0.60 0.80 1.00 1.20
83 their opinion on a significant policy issue and 56 percent indicated they had changed their opinion on either a scientific/technical or policy issue. Additionally, as Table V II 2 shows, respondents indicating organi zational affiliations with environmental groups had the highest level of opinion change (80%), followed by other (67%, driven by commercial fishermen). 58% and 57% of Scientists/consultants and government respondents respectively indicated they had change d their opinion on either science or policy issues. Respondents from the aquaculture industry reported the lowest level of changed opinion (50%) when compared to other groups. This is a clear indication of learning among partnership participants. Table V II 2 Changed Opinion by Organizational Affiliation Percentage of respondents reporting they have changed their professional opinion on the following issues related to marine aquaculture : Organization al Affiliation Scientific/ Technical Policy Issues Either Environmental Groups 60% 80% 80% Other 67% 67% 67% Scientists/Consultan ts 27% 50% 58% Government 50% 46% 57% Aquaculture Industry 34% 35% 50% Total 41% 46% 56% Even though we have indications of learning within partnerships, for the next portion of this report we need to understand if there is a relationship between the learning metrics better un derstanding and changed opinion In other words, does an increase in understanding actually lead to changed professional opinion? T his is important to ascertain since I am using the concept of changed professional opinion as a metric of learning in addition to better understanding In the event that an increase in understanding did not lead to a changed professional opinion, it would indicate that an important aspect of learning may have been inadvertently omitted from the survey which could confound results For example, the survey did not capture the idea of
84 reinforcement of existing professional opinion, only change. Since learni ng can also be represented in terms of reinforcement of professional opinion, its omission could potential ly confound the results observed in this study. Note that this assessment was conducted initially in Leach, et al, 2013, as a part of the development of a model of knowledge acquisition within collaborative partnerships. To further demonstrate the validity of the learning measures employed in this study, b inary logistic regression modeling was conducted to assess whether responses to the learning que stions (independent variables, also referred to as predictor variables) significantly predicted whether or not an individual reported a change in professional opinion (dependent variables) For each model, variables were retained that might be expected to influence opinion change. For changes in opinion on scientific or technical issues, the predictor variables included better understanding of stakeholder perspectives and better understanding of aquaculture science For changes in opinion on policy issue s, the predictor variables included better understanding of stakeholder perspectives ; better understanding of aquaculture policy, law or regulations ; better understanding of aquaculture economics and business and better understanding of aquaculture science Results for each model are included below. Modeling Results Changes in Opinion on Scientific or Technical I ssues When both predictor variables ( better understanding of stakeholder perspectives and better un derstanding of aquaculture science ) are considered together, they significantly predict whether or not a respondent changed their opinion on a scientific or technical issue, 2 =29.23, df =2, N =114, p <0.001. Table V II 3 presents the beta values, standard er rors, odds ratios, and significance values, which suggest that the odds of a significant change in opinion on scientific
85 or technical issues are increasingly greater as understanding of stakeholder perspectives and aquaculture science increases. Table V II 3 Logistic Regression Predicting Opinion Change on Scientific/Technical Issues Variable B SE Odds Ratios p Better understanding of: Stakeholder perspectives 0.97 0.44 2.64 0.027 Aquaculture science 0.76 0.29 2.14 0.009 Constant 2.1 0.54 0.12 0.000 These results are not necessarily surprising, and indicate that an increase in understanding other stakeholder perspectives as well as aquaculture science may lead to changes in professional opinion on scientific or technical issues. In other words, as an individual learns more regarding stakeholder perspectives and/or specific scientific aspects of aquaculture science, the odds of them changing their opinion on a scientific or technical issue increases. Modeling Results Changes in Opinion on Policy I ssues When all four predictor variables ( better understanding of stakeholder perspectives better understanding of aquaculture policy, law or regulations ; better understanding of aquaculture economics and business and better understanding of aquaculture science ) are considered together, they significantly predict whether or not a respondent changed their opinion on a n aquaculture policy issue, 2 = 29.35 df =4 N =114, p <0.001. Table V II 4 presents the beta values, standard errors, odds ratios, and significance values, which suggest that the odds of a significant change in opinion on policy issues are increasingly greater as understanding of stakeholder perspectives and aquaculture science increases.
86 Table V II 4 Logistic Regression Predicting Opinion Change on Policy Issues Variable B SE Odds Ratios p Better understanding of: Stakeholder perspectives .985 .47 7 2.67 .0 39 Aquaculture policy 055 .3 52 1. 06 876 Aquaculture economics and business .107 .3 30 .899 747 Aquaculture science .795 .356 2.21 .026 Constant 2.1 5 .5 47 .1 16 .000 Results of binary logistic regression modeling for opinion change on policy issues provide some additional insight into the learning measure T hough the model as a whole is significant, the significant predictor variable s in the model are better understanding of stakeholder perspectives and aquaculture science One might expect that an increase in understanding of aquaculture policy and economi cs/business would lead to change in policy opinion. Rather, opinion change appears to be more directly related to an increased understanding of stakeholder perspectives and aquaculture science. Though this result seems somewhat confounding, the aspects u nderlying aquaculture policy tend to be centered on issues of science. For instance, a s indicated in Chapter II Study Context areas of policy debate are related to aspects such as genetic impairment of wild stocks, water pollution sustainability of feeds, etc., each of which are integral to the science of aquaculture. Therefore, understanding of aquaculture scien ce should be expected to play an important role in opinion change for both the policy and science/technical arenas. Th ese results indicate that l earning is happening within these partnerships and among the various organizations taki ng part in these partnerships ; which, in some cases, is leading to professional opinion change on scientific/technical and policy issues assoc iated with aquaculture. Given this, the use of a change in professional opinion as a metric of learning appears
87 appropriate. These results will feed directly into Chapter X, where we will determine how learning within the partnership is shaped by network position.
88 CHAPTER VIII RESULTS TRUST WITHIN PARTNER SHIPS Having looked at the levels of learning reported among aquaculture partnerships, we now turn to trust As can be seen in the flow diagram below, we need to understand trust within partnerships (a component of Stage 2) in order to address hypotheses 5 and 7 (i.e., how network position shapes perceptions of trust within partnerships). As discussed in Chapter IV Methodology trust within partnerships was captured by asking respondents to i ndicate whether the following statements apply to none, few, half, most or all of the participants in the partnership: Are honest, forthright, and true to their word Have the same values and priorities that you do Have reasonable motives and concerns Are w illing to listen, and sincerely try to understand other points of view Reciprocate acts of good will or generosity Are trustworthy Reported Trust H7: Individual Centrality H5: Partnership Density
89 Participant responses were re coded using a five point scale (1 = none, 2 = few, 3 = half, 4 = most, and 5 = all ) Similar to the battery of learning questions was generated to assess whether the six learning items could be combined to form a reliable overall trust variable The alpha for the six items was 0.80 (with inter it em correlations rang ing from 0.47 to 0.74 ), indicating that the trust questions can be combined to make a scale variable with reasonable internal consistency reliability. Therefore, responses to the trust questions were averaged to form a single variable (referred to as over all trust) for use in regression modeling Table V III .1 shows the mean responses to individual trust question as well as the overall scale trust variable for each of the 10 partnerships included in this study. Table V III .1 Reported Interpersonal Trust by Partnership Partnership Are honest, forthright and true to their word Have same values and priorities as you Have reasonable motives and concerns Are willing to listen and try to understand other points of view Reciprocate acts of goodwill or generosity Are trust worthy Overall trust CADC 4.17 3.39 4.00 3.94 3.76 3.94 3.87 FARC 4.80 4.00 4.40 4.60 4.60 4.60 4.50 FMNPWG 4.33 3.33 4.40 4.20 4.00 4.33 4.10 MAAC 4.20 3.80 4.40 4.20 4.20 4.20 4.1 7 MFHTC 4.29 3.86 4.29 4.29 4.14 4.29 4.19 MACC 4.22 3.44 4.22 4.11 3.89 4.13 4.00 NJAC 4.00 3.30 3.80 3.70 3.60 4.00 3.73 PAC 4.45 4.09 4.32 4.32 4.41 4.41 4.33 RIAWG 4.00 3.43 4.14 4.07 3.79 3.93 3.89 SARC 3.77 3.23 3.54 3.77 3.46 3.69 3.5 8 Total 4.20 3.58 4.12 4.09 3.95 4.12 4.0 1 Scale: 1 = None; 5 = All As indicated in Table V III .1 overall perceptions of trust among partnership participants are generally high, with most or all participants (average of 4.0 or above) viewed as: h onest, forthright and true to their word (4.2 0 ) having reasonable motives and concerns (4.12)
90 willing to listen and try to understand other points of view (4.09) trustworthy (4.12) Between half and most participants (average between 3.0 and 4.0) were viewed as: having the same values and priorities (3.58) reciprocating acts of good will and generosity (3.95) Mean scores for individual partnerships were above the scale midpoint (3, half of participants) for each trust question, ranging from 3.30 to 4.80. Overall, reported levels of trust were generally high among the partnerships, ra nging from 3.73 for the NJAAC to 4.50 for the FARC. Figure V III 1 includes mean scores for the overall trust variable by organizational affiliation (separated into five distinct groups). As shown in the figure, generally high levels of trust (means above the scale midpoint) were observed for each organizational affiliation, with the highest level for respondents from government (4.05 ), followed by scientists/consultants 3.99 ), ( 3.83; this group i s comprised of three individuals: one representative from a Native American and two commercial fishermen ). The organizational affiliation with the lowest level of reported trust (though still above the scale midpoint) was environmental groups (3.79). N ote that the sample size is different between organizational affiliations in this figure. This influences how much of an effect the level of trust reported for an organization has on the average for all respondents, making the results presented in the grap h appear odd. For example, the average level of reported trust for all groups combined is 4.01. However, the levels of trust observed for are somewhat lower than the other groups. One would think these values would bri ng down the total for all respondents more than appears
91 in the figure However, the number of responses from those two groups (7 total) is much less than the others (103 responses) Therefore, the lower trust scores for those two groups do not impact the overall trust score all that much Figure VIII 1 Overall Trust by Organizational Affiliation. These results indicate that there is a generally high level of trust (i.e., mean scores for each individual question and for the scale trus t variable are above the scale midpoint) within these partnerships and among the various organizations taking part in these partnerships. However, though trust scores are generally high, there is variability among the partnerships and organizations. These results will feed directly into Chapter X, where we will test hypotheses 5 and 7 to determine how perceptions of trust within the partnership are shaped by network position. 4.05 3.99 3.98 3.83 3.79 4.01 3.65 3.70 3.75 3.80 3.85 3.90 3.95 4.00 4.05 4.10
92 CHAPTER IX RESULTS NETWORK CENTRALITY AND DENSITY Having investigated learning and trust within partnerships, we now turn to network position (individual centrality and network density by partnership ), which a component of Stage 2. As can be seen in the flow diagram below, we need to understand network position in order to address hypotheses 4 through 7 (i.e., how network position shapes learning and trust bolded in the diagram below ). This section will evaluate individual centrality and network density at the partnership level Network Structure Centrality As outlined in Chapter IV three measures of centrality were calculated for survey respondents: out degree centrality, in degree centrality and eigenvector centrality. The c entrality measures were generated by first allowing respondents to indicate (using a list prompt) which groups they tend to coordinate with on aquaculture issues (see F igure V I .1 and Table V I 1 ) and entering the ensuing coordination network matrices into UCInet . Once in UCInet, the centrality measures options can be selected either in the main menu, or once network schema have been generated. Specifics regarding each centrality measure are given below. Reported Learning Reported Trust H6/H7: Individual Centrality H4/H5: Partnership Density
93 Out degree centrality was calculated for each respondent (using UCInet) as the number of citations emanating from an ind iv idual direc ted towards an organization By doing this, each respondent is assigned their o wn out degree centrality value. Figure IX 1 shows the distribution of out degree centrality values for respondents with the count of each centrality value presented on the x axis, and the frequency on the y axis Figure IX 1 Distribution of Respondent Out Degree Centrality. As indicated in the figure, out degree centrality values ranged from 1 to 21, with a mean of 9.7 and a median of 9.0. This means that respondents had an average of close to 10 organizations they reach ed out to for coordination purposes The next logical question that arises is which organizations have the highest level of out degree centrality. In other words, which
94 organizations are the most central when considering how many times they sought others for coordination purposes Figure IX 2 shows the mean out degree centrality value for each of 11 general organizational affiliations. Figure IX 2 Mean Out Degree Centrality by Organizational Affiliation. As indicated in the figure, the organizations reaching out the most to others for coordination purposes include consultants, federal government, shellfish industry and members of the scientific/research community. Conversely, the organization s reaching out the least include other aquaculture, environmental groups, and representatives from local government. In degree centrality was calculated as the number of citations received by an organization in a partnership by surve y respondents. For e xample, in the SARC the shellfish aquaculture industry received 12 coordination citations from partnership participants. Therefore each respondent that indicated a shellfish aquaculture industry affiliation in the SARC was assigned an in degree centrality value of 12. Since there were far more organizational options provided in the list prompt than were represented within each partnership, respondent affiliations were condensed into 11 general organizations as indicted in Table V I 1 By doing this, each 3.5 6.8 7.0 8.0 8.0 9.3 9.8 10.0 10.4 10.6 10.6 0.0 2.0 4.0 6.0 8.0 10.0 12.0
95 individual in a partnership could receive one of 11 in degree centrality values, based on the number of citations received by that individual s general organization al affiliation Obviously this is not ideal, since this approach assumes that when an individual cites an organization, he is citing all members of that organization equally. Ideally, one would prefer to have each survey respondent explicitly state with whom they coordin ate, thereby generating specific individual in degree centrality values. D ue to the constraints of survey administration, such as anonymity, this approach was not possible. However, by generating in degree centrality values on a partnership by partnershi p basis using 11 organizational affiliations (i.e., using the largest number of organizational affiliations as possible given survey responses) we can ameliorate the situation and establish centrality values that are adequate for analysis. Figure IX 3 sh ows the distribution of in degree centrality values assigned for respondent s with the count of each centrality value presented on the x axis, and the frequency on the y axis
96 Figure IX 3 Distribution of Respondent In Degree Centrality. As indicated in the figure in degree centrality values ranged from 2 to 16, with a mean of 9.8 and median of 10.0 This means that respondents were recipients of coordination efforts (i.e., cited by) from an average of about 10 individuals. Similar to what was done for out degree centrality; we next look at which organizations have the highest level of in degree centrality. In other words, which organizations are the most central when considering the number of times individuals sought them out for coordination purpo ses Figure IX 4 shows the mean in degree centrality value for each of 11 general organizational affiliations
97 Figure IX 4 Mean In Degree Centrality by Organizational Affiliation. As indicated in the figure, the most central individuals (based on in degree centrality) are members of organizations such as local and federal government, industry as well as representatives from the science/research community. Conversely, the organizations that appear least central include commercial/recreational fish ing and other aquaculture. The final centrality measure considered in this study was eigenvector centrality. E igenvector centralities were generated for each respondent by developing a symmetric citation matrix for each partnership that included only the participants in that partnership. For example, in the SARC, there were 13 respondents overall representing 7 organizations. Therefore, following the approach used by Faust (1997) only the organizations cited that were also represented within the partner ship were included in the symmetrical citation matrix Similar to in degree centrality, this approach has some limitations. For example, it assumes that when an individual cites an organization as a coordination tie, he is citing a member of that organi zation that is actively participating within that partnership and responding to the survey Additionally, if an organization was cited by a respondent, but that organization did not respond to the survey 4.5 5.0 8.0 9.0 9.8 10.0 10.3 10.6 11.0 11.2 12.3 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0
98 (either because they did not choose to, or they wer e simply not a part of the partnership), then that tie was not considered in the coordination matrix. In other words, if a member of the news media was cited as a coordination network tie, but there were no news media representatives in the partnership, then that coordination tie would be omitted from consideration when developing the symmetrical network matrices. This approach (i.e., using symmetrical network matrices) can greatly limit the size of networks since the maximum size o f the matrix axes are equal to the number of respondents within a partnership. Secondly, it ignores the contribution of exogenous coordination ties (i.e., ties with individuals not within the partnership) in preference of endogenous ties. Ideally, one wo uld prefer to have each survey respondent explicitly state with whom they coordinate within the partnership, thereby generating specific individual coordination networks from which very accurate centrality values could be calculated. However, due to the c onstraints of survey administration, such as anonymity, this approach was not possible. Even so, by using three measures of centrality (i.e., evaluating in degree, out degree and eigenvector centrality versus relying on just one measure), we can lessen th e impact that the limitations of one centrality measure might have on the study as a whole. Once developed, the symmetrical coordination matrices (specific to each partnership) were entered into UCInet (using the spreadsheet entry tool) Eigenvector cen trality values were then generated by selecting t menu, Results were captured by UCInet in a text file containing the eigenvector centrality values for each organ ization Eigenvector centrality values were then entered into SPSS for subsequent analyses.
99 Figure IX 5 shows the distribution of eigenvector centrality values assigned to respondents, with the frequency of each centrality value presented on the x axis, and the eigenvector centrality values on the y axis Figure IX 5 Distribution of Respondent Eigenvector Centrality. As indicated in the figure, eigenvector centrality values ranged between 12 and 107. The figure also indicates the potential of a bimodal distribution for the variable (or at least a departure from normality). Though problematic, this aspect of the distribution will be addressed later in the study (as a part of regression mod eling). Figure IX 6 shows the mean eigenvector centrality values for each of 11 organizational affiliation s
100 Figure IX 6 Mean Eigenvector Centrality by Organizational Affiliation. As shown in the figure th e organizations with the highest mean eigenvector centralities are those associated with industry (finfish and shellfish), government (federal and state), and the science/research community. Conversely, organizations with the lowest eigenvector centrality include other aquaculture, commercia l fishing, and consultants. Next we can look at the three measures of individual centrality (by organizational affiliations) to see differences among the groups based on the measure. Table IX.1 ranks the relative centrality for each affiliation, with 1 be ing the most central, and 11 being the least for each measure (organizations are listed based on out degree centrality). 21.6 28.3 31.0 44.1 46.2 46.2 58.7 62.9 67.3 71.8 72.4 0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0
101 Table IX .1 Comparison of Centrality Measures by Affiliation Organizational Affiliation Out Degree In Degree Eigenvector Science/Research 1 3 5 Shellfish 2 2 3 Federal Government 3 1 1 Consultants 4 7 9 Finfish 5 5 2 State Government 6 9 4 Other 7 8 8 Commercial/Recreational Fishing 8 11 10 Local Government 9 6 6 Environmental Groups 10 4 7 Other Aquaculture 11 10 11 As Table IX.1 shows, if we compare the level of centrality (as an average for organizational affiliations), Federal Government appears to be the most central organization overall, ranking in the top for in degree and eigenvector and third for out degree. Likewise, Science/Research appears quite central, ranking first in out degree, third for in degree and fifth for eigenvector. Conversely, commercial/recreational fishing and other aquaculture appear to be the least central organizations somewhat consistently for all three measures. However, the measures of centrality provide a bit more insight regarding how organizations operat e within their networks. For example, as will be discussed later in this study, out degree centrality might be thought of in terms of information seeking or attempts to influence others within a network. Given this, information seeking or influencing or g anizations include Science/Rese arch, Shellfish and Federal Government. Conversely, in degree centrality might be an indicator of popularity or power within a network. In other words, those organizations are connection targets. Similar to out degree, the most central in degree organizations include Federal Government, Shellfish, and Science/Research. Thus we see these three organizations as hubs of centrality both inwardly and outwardly. Finally, if we look at
102 eigenvector centrality as an overarching ce ntrality measure (i.e., centrality of a node in the network as a whole), then the sto ry is slightly different. The Federal Government appears to be the most central organization, followed by aquaculture industry (finfish and shellfish). Again, this is ba sed on average centrality across organizational affiliations. Network Structure Density N etwork densities were generated by developing a 2 mode coordination matrix for each partnership which was subsequently entered into UCINet, using the spreadsheet d ata entry This process was repeated for each partnership, resulting in a total of 10 den sity values. Figure IX.7 shows density values for each partnership. As discussed earlier, UCINet calculates network density by dividing the total number of reported ties by the total number of possible ties, which results in a unitless value between 0 and 1. Figure IX 7 Network Density by Partnership. 0.36 0.37 0.37 0.39 0.39 0.43 0.44 0.45 0.58 0.61 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 MFHTC SARC FMNPWG PAC NJAAC MACC CADC RIAWG MAAC FARC
103 As shown in the figure, network density values ranged from 0.36 to 0.61, with the highest level of network densi ty observed in the FARC (0.61) and MAAC (0.58 ) The next tier of partnership densities included the RIAWG (0.45), CADC (0.44) MACC (0.43) NJAAC (0.39) and PAC (0.39) The lowest network densities were observed in the FMNPWG (0.37), SARC (0.37) and MFHT C (0.36). In this chapter, we have presented a descriptive overview of individual centrality and network density by partnership each of which are integral to Stage 2 of this study. In the next chapter, we will bring together the variables presented in Chapters VII through IX to address hypotheses 4 through 7 (i.e., how network position shapes learning and trust).
104 CHAPTER X RESULTS NETWORK STRUCUTRE AN D PARTNERSHIP VARIAB LES This section includes the results of testing for hypotheses 4 th r ough 7 (restated below), which seek to elicit how network density and individual centrality shape learning and trust within partnerships. The following flow diagram illustrates how each of these variables (shown in bold) has been brought together for Stage 2 of this study Network Density and Learning Hypothesis 4 (H4) Network Density and Learning (at the partnership level) : The high er the network density for a collaborative partnership, the higher the level of reported learning on average across members of the partnership. As indicated earlier, the evaluation of how network density shapes learning was conducted at the partnership le vel, using network density and reported learning (average for each partnership) as variables. This was performed for each learning question as well as the learning scale variable. The use of partnership level data is somewhat challenging due to the relati vely small sample size (in our case n=10). This presents some complexities for regression modeling since as discussed in Chapter IV, standard simple regression techniques may be susceptible to fluctuations in error variances (Maggioni 2012). However, to evaluate the hypotheses associated Reported Learning Reported Trust H6/H7: Individual Centrality H4/H5: Partnership Density
105 with partnership level data, a combination of statistical approaches was used. As discussed in Chapter IV, descriptive statistics, simple linear regression and weighted least squares regression were used congruently to build converging lines of evidence Descriptive statistics can be helpful to give a broad picture of the relationship between the variables, and can also lend support to results seen using other statistical tools such as simple linear regression and w eighted least squares regression. The weighted least squares regression technique is designed to downweight residuals associated with observations that have large variances (typical for data associated with small sample sizes). In this case, the weighte d least squares estimators were weighted on the network density variable. This variable showed the greatest amount of variation (compared to the learning and trust variables) and should therefore provide conservative model estimates. As a result, the use of weighted least squares regression should provide conservative model estimates while addressing the small sample size problem. In addition to the theoretical basis provided for using regression techniques provided in Chapter IV, the use of multiple sta tistical tools as well as descriptive statistics should ensure the robustness of the results presented in this section Figure X. 1 shows a scatter plot of net work density versus learning, with a best fit line for reference. The figure shows the potential for a positive correlation between density and learning. In other words, as density increases, reported learning appears to as well. Obviously there is some variability in the data, with several points falling quite far from the best fit line.
106 Figure X 2 Scatter Plot of Network Density and Learning. Another way to look at the data is though bar charts, with the variables coupled Figure X.2 illustrates density and learning values for each partnership sorted from the highest density to the lowest (learning values were normalized to bring the scale closer to that of density) This approach allows us to elicit some trends that might be present in the data (such as that observed in the scatter plot) For instance, as you move from left to right in the graph (from the FARC to the MAAC) density values fall from 0.609 to 0.437, as learning values (normalized) drop from 0. 500 to 0. 283 Learning rises as we mo ve to the RIAWG, and once again falls as we get to the CADC. Likewise, learning jumps once again with the MACC, but drops as density drops
107 through the PAC. Learning rises again with the FMNPWG and drops as we get to the SARC. Finally, we get to the MFTH C, which has the lowest density, but the third highest learning value. Figure X 2 Network Density and Learning. There appears to be some evidence of a correlation between density and learning in Figure X.2, however the variability in the data does mak e that relationship more difficult to see than that presented in the scatter plot. However, if we modify what is included in the chart, the relationship observed in the scatter plot becomes more apparent. For instance, if we remove the data associated wi th four of the partnerships (MAAC, MACC, FMNPWG, and MFHTC), we can see that learning values are clearly falling as density decreases. It is likely that these six partnerships are driving the relationship observed in the scatter plot Figure X.3 shows th e modified network density and learning values for six of the ten partnerships. 0.609 0.576 0.452 0.437 0.427 0.390 0.389 0.368 0.367 0.364 0.500 0.283 0.333 0.323 0.390 0.243 0.220 0.250 0.133 0.333 0.000 0.100 0.200 0.300 0.400 0.500 0.600 0.700 Density Learning
108 Figure X 3 Network Density and Learning for Six Partnerships When we look only at the partnerships removed from Figure X.3, we see that among those, as density drops, learning does as well. Figure X.4 shows density and learning for those four partnerships. Figure X 4 Network Density and Learning for Four Partnerships Given the suggestion of a relationship between density and learning observed in the descriptive sta tistics, n ext we employ some more complex statistical modeling tools to determine 0.609 0.452 0.437 0.390 0.389 0.367 0.500 0.333 0.323 0.243 0.220 0.133 0.000 0.100 0.200 0.300 0.400 0.500 0.600 0.700 FARC RIAWG CADC NJAAC PAC SARC Density Learning 0.576 0.390 0.389 0.367 0.283 0.243 0.220 0.133 0.000 0.100 0.200 0.300 0.400 0.500 0.600 0.700 MAAC NJAAC PAC SARC Density Learning
109 if that relationship is significant. S imple linear regression was conducted to determine if network density was a significant predictor of reported learning among partnerships. T his was conducted at the partnership level, using network density and reported learning (average for each partnersh ip) as variables. This was performed for each learning question as well as the learning scale variable. The assumptions of linearity and normality were checked, and each was met, indicating that the use of simple regression techniques is appropriate for t his dataset. However, the assumption of independent observations was obviously not met, since respondents were often participants from the same partnerships. This is a common limitation in social research and was addressed by employing cluster variables in STATA to serve as a control. Network density ( M ean = 0.438 S tandard D eviation = 0.087 ) significantly predicted reported learning ( M = 0.903 SD = 0.302 ), F (1,9) = 5.64, p = 0.04 adjusted R 2 = 0.34. For clarity, the reported p value is the probability of erroneously rejecting the null hypothesis (in this case the null hypothesis is that there is no significant relationship between density and reported learning). In most cases, when p is less than 0.05, then the probability of erroneo usly rejecting the null hypothesis is sufficiently low, which leads to interpretation of a significant result. The associated R 2 value is an indicator of effect size. In this case, the R 2 value of 0.34 indicates that 34% of the variability observed in re ported learning is explained by network density. According to Cohen (1988), this is a medium to large effect size. The beta weights, presented in Table X .1 indicate that when network density increases by one unit, learning (scale) increases by 2.2 units.
110 Table X .1 Simple Linear Regression Analysis Summary for Network Density Predicting Learning (Scale Variable) Variable B SEB ** *** Network Density 2.20 0.935 0.64* Constant 0.069 0.417 Note: R 2 = 0.34, F (1,9) = 5.64, p <0.05 p <0.05 **SEB Standard Error for the regression coefficient *** slope defining the linear relationship between the independent and dependent variables Weighted least squares regression was performed to partially validate the simple linear regression analysis and address issues associated with a small sample size. Weighted least squares regression results were quite similar to simple linear regression re sults, with network density ( M ean = 0.438 S tandard D eviation = 0.087 ) significantly predicting reported learning ( M = 0.903, SD = 0.302), F (1,9) = 5.00, p = 0.046, adjusted R 2 = 0.31. The beta weights, presented in Table X.2, indicate that when network density increases by one unit, learning (scale) increases by 2.3 units. Table X.2 Weighted Least Squares Regression Summary for Network Density Predicting Learning (Scale Variable) Variable B SEB Network Density 2.2 8 0. 439 0.62 Constant 0.0 94 1.02 Note: R 2 = 0.3 1 F (1,9) = 5. 00 p <0.05 p <0.05 Th e weight of evidence suggests that there is a significant relationship between learning and network density (via regression modeling as well as an assessment of descriptive statistics) Next the type of learning that is related to density . S imple linear regression was
111 performed using each individual learning statement to determine this For clarity, those statements include: Participation in the partnership has given me a better understanding of aquaculture science. Participation in the partnership has given me a better understanding of aquaculture policy, law, or regulations. Participation in the partnership has given me a better understanding of aquaculture economics or business. Participation in the partnership has given me a better understanding of other stakeholder perspectives Network density significantly predicted reported learning for the following learning statements : better understanding of stakeholder perspectives ( M = 1.16, SD = 0.407), F (1,9) = 4.80, p = 0.05, adjusted R 2 = 0.3 0 and better understanding of aquaculture science ( M = 0.787 SD = 0. 338 ), F (1,9) = 9.18 p = 0.02 adjusted R 2 = 0. 53. The beta weights, presented in Table X. 3 and X. 4 indicate that when network density increases by one unit, learni ng increases by 2. 9 units (understanding of stakeholder perspectives) and 2.8 units (understanding of aquaculture science) Table X 3 Simple Linear Regression Analysis Summary for Network Density Predicting Learning (understanding stakeholder perspectives) Variable B SEB Network Density 2. 85 0. 580 0.61 Constant 0.0 93 1.301 Note: R 2 = 0.30 F (1,9) = 4.82 p = 0.05 p < /= 0.05
112 Table X 4 Simple Linear Regression Analysis Summary for Network Density Predicting Learning (understanding of aquaculture science) Variable B SEB Network Density 2.82 0.932 0.73* Constant 0.449 0.415 Note: R 2 = 0. 53 F (1,9) = 9.18 p <0.05 p <0.05 These regression results indicate that the type of learning that is related to network density is specific to an increased understanding of aquaculture science (the most significant variable and highest associated R 2 value) as well as an understanding of s takeholder perspectives. Network density was not a significant predictor of learning associated with a better understanding of aquaculture policy, law or regulations; or a better understanding of aquaculture economics or business. Understanding that even with the approach taken above, evaluation of small n data can be problematic. Therefore, these results should be interpreted carefully. Network Density and Trust Hypothesis 5 (H5) Network Density and Trust (at the partnership level): The higher the network density for a collaborative partnership, the higher the level of reported trust on average across members of the partnership. As indicated earlier, the evaluation of how network density shapes trust was conducted at the partnership level, using network density and perceptions of trust within the partnership (average for each partnership) as variables. This was performed for each trust question as well as the trust scale variable. For the reasons outlined preceding section to evaluate the hypotheses associated with partnership level data (small sample size) a combination of statistical approaches was used. As discussed in Chapte r IV, descriptive statistics, simple linear regression and weighted least squares regression were used congruently to build converging lines of evidence. Descriptive statistics can be helpful to give a broad picture of the relationship between the
113 variabl es, and can also lend support to results seen using other statistical tools such as simple linear regression and weighted least squares regression. In addition to the theoretical basis provided for using regression techniques provided in Chapter IV, the u se of multiple statistical tools as well as descriptive statistics should ensure the robustness of the results presented in this section. data. Figure X.4 shows a scatter plot of network density versus learning, with a best fit line for reference. The figure shows the potential for a positive correlation between density and learning. In other words, as density increases, trust appears to as well. However, compared to the learning variable, there is a substantial amount of variability in the trust data, with many points falling quite far from the best fit line. Figure X 5 Scatter Plot of Network Density and Trust.
114 Given the suggestion of a relationship between density and learning observed in the descriptive statistics, next we employ some more complex statistical modeling tools to determine if that relationship is significant. Simple linear regression was conducte d to determine if network density was a significant predictor of reported trust among partnerships. As indicated earlier, this was conducted at the partnership level, using network density and reported trust (average for each partnership) as variables. Th is was performed for each trust question as well as the trust scale variable. The assumptions of linearity and normality were checked, and each was met, indicating that the use of simple regression techniques is appropriate for this dataset. However, as indicated earlier, the assumption of independent observations was obviously not met, since respondents were often participants from the same partnerships. This was addressed by employing cluster variables in STATA to serv e as a control. Network density ( M = 0.438 SD = 0.087 ) was not a significant predictor of reported trust ( M = 4.03 SD = 0.288 ), F (1,9) = 2.89 p = 0.128 ) Though network density did not significantly predict trust (as a scale variable), each trust question was evaluated to determine if network density was a significant predictor of an individual trust metric. S imple linear regression was performed using eac h individual trust statement to determine this As indicated above perceptions of trust were measured by asking respondents to indicate whether the following statements apply to none, few, half, most, or all partnership participants: Are honest, forthrig ht, and true to their word Are willing to listen and sincerely try to understand other points of view Reciprocate acts of good will or generosity Are trustworthy
115 Network density was not a significant predictor (i.e., p <0.05) of any individual metric of tru st (F values ranging from 1.5 to 3.5 and p values ranging from 0.097 to 0.25). Weighted least squares regression was performed to verify results from simple linear regression analysis and address issues associated with a small sample size. Though this e xercise is not anticipated to yield significant results since it is more conservative than simple linear regression. Weighted least squares regression results indicate that density was not a significant predictor of trust ( F (1,9) = 2.92, p = 0.126 ) Netw ork Centrality and Learning Hypothesis 6 (H6) Centrality and Learning (at the individual level) : The higher centrality of an individual among members of collaborative partnerships, the higher the level of reported learning among those same members of the partnership. S imple linear regression was conducted to determine if network centrality was a significant predictor of reported learning among respondents. As indicated earlier, this was conducted at the individual level, using network centrality ( out degree, in degree and eigenvector) and reported learning as variables. This was performed for each learning question as well as the learning scale variable. The a ssumptions of linearity and normality were checked, and each was met, indicating that the use of simple regression techniques is appropriate for this dataset. However, as indicated earlier, the assumption of independent observations was obviously not met, since respondents were often participants from the same partnerships. This serve as a control. For simplicity, results for each centrality variable are presented ind ividually. O UT DEGREE C ENTRALITY AND L EARNING Individual out degree centrality ( M = 9. 66 SD = 4.33 ) was not a significant predictor of reported learning ( M = 0.844, SD = 0.770), F (1, 113 ) = 0.045 p = 0. 832 Though out degree
116 centrality was not a significant predictor of learning (as a scale variable), each learning question was evaluated to determine if out degree centrality was a significant predictor of an individual learning metric. S imple linear regression was performed u sing responses to each learning statement to determine this For clarity, those statements include: Participation in the partnership has given me a better understanding of aquaculture science. Participation in the partnership has given me a better underst anding of aquaculture policy, law, or regulations. Participation in the partnership has given me a better understanding of aquaculture economics or business. Participation in the partnership has given me a better understanding of other stakeholder perspect ives Out degree centrality was not a significant predictor (i.e., p <0.05) of any individual metric of learning (F values ranging from 0.048 to 0.863 and p values ranging from 0. 358 to 0. 832 ). Though out degree centrality did not significantly predict learning when measured in changed opinion (both scientific/ technical and policy opinion) For example, respondents were asked the followi ng questions: At least partially through your participation in the partnership, have you changed your professional opinion on any significant scientific or technical issues related to marine aquaculture? At least partially through your participation in the partnership, have you changed your professional opinion on any significant policy issues related to marine aquaculture?
117 B inary logistic regression was performed to determine if out degree centrality significantly predicted learning in terms of changed pro fessional opinion. Out degree centrality did not signif icantly predict whether a respondent changed their professional opinion on a technical or scientific issue 2 = 0.102, p = 0.75). However, out degree centrality did significantly predict whether or not a respondent indicated they had changed their professional opinion on a policy issue related to marine aquaculture ( 2 =6.55, df =2, N =114, p <0.05). Table X.5 presents the odds ratios, which suggest that the odds of a change in professional opinion on a significant policy issue are increasingly greater as out degree centrality increases (since the odds ratio is greater than 1). Table X. 5 Logistic Regression Predicting Opinion Change on Significant Policy Issues Variable B SE Odds Ratios p Out degree Centrality 0.12 0.05 1.12 0.014 Constant 1.2 0.49 0.29 0.012 I N DEGREE C ENTRALITY AND L EARNING Individual in degree centrality ( M = 9.83 SD = 3.78 ) significantly predicted reported learning ( M = 0. 844 SD = 0. 770 ), F (1, 111 ) = 6.19 p = 0.0 14 adjusted R 2 = 0. 05 According to Cohen (1988), this is a relatively small effect size. T he beta weights, presented in Table X.6 indicate that when in degree centrality increases by one unit, learning (scale) decreases by 1.14 units. This result is in clear opposition to the expected relationship between centrality and learning. Table X. 6 Simple Linear Regression Analysis Summary for Individual In Degree Centrality Predicting Learning (Scale) Variable B SEB Individual In degree Centrality 1.14 0.459 0.23* Constant 10.7 0.520 Note: R 2 = 0. 05 F (1, 111 ) = 6.19 p <0.05 p <0.05
118 Though it appears that there is a significant (negative) relationship between learning and individual in degree centrality, the type of learning that is related to centrality is also of interest. S imple linear regression was performed using each individual learning statement to determine this For clarity, those statements include: Participation in the partnership has given me a bett er understanding of aquaculture science. Participation in the partnership has given me a better understanding of aquaculture policy, law, or regulations. Participation in the partnership has given me a better understanding of aquaculture economics or busin ess. Participation in the partnership has given me a better understanding of other stakeholder perspectives Individual in degree centrality significantly predicted reported learning for the following learning statements: better understanding of stakehol der perspectives ( M = 1. 06 SD = 0.779 ), F (1, 111 ) = 6.91 p = 0.0 1 adjusted R 2 = 0.0 51 and better understanding of aquaculture science ( M = 0.7 40 SD = 1.01 ), F (1,111 ) = 4.91 p = 0.03 adjusted R 2 = 0. 034 The beta weights, presented in Table X. 7 and X. 8 indicate that when individual in degree centrality increases by one unit, learni ng actually decreases by 1.19 units (understanding of stakeholder perspectives) and 0.775 units (understanding of aquaculture science)
119 Table X. 7 Simple Linear Regressio n Analysis Summary for Individual In Degree Centrality Predicting Learning (understanding of stakeholder perspectives ) Variable B SEB In degree Centrality 1.194 0.454 0.24 Constant 11.06 0.597 Note: R 2 = 0.051 F (1,111 ) = 6.91 p =0.01 p <0.05 Table X. 8 Simple Linear Regression Analysis Summary for Individual In Degree Centrality Predicting Learning (understanding of aquaculture science) Variable B SEB In degree Centrality 0.775 0.350 0. 21 Constant 10.34 0.434 Note: R 2 = 0. 034 F (1, 111 ) = 4.91 p = 0.0 3 p <0.05 Similar to the results for the scale learning variable, these regression results indicate that the type of learning that is negatively associated with in degree centrality is specific to an understanding of aquaculture science as well as an understanding of stakeholder perspectives. In degree centrality was not a significant predictor of learning associate d with a better understanding of aquaculture policy, law or regulations; or a better understanding of aquaculture economics or business. In degree centrality was shown to be a significant predictor of learning when measured in degree centrality, learning was also measured in terms of changed opinion (both technical and policy At least partially through your participation in the partnership, have you changed your professional opinion on any significant scientific or technical issues related to marine aquaculture?
120 At least partially through your participation in the partnership, have you changed your professional opinion on any significant policy issues related to marine aquaculture? B inary logistic regression was performed to determine if in degree centrality significantly predicted learning in terms of changed professional opinion (tech nical /scientific and policy opinion) In 2 = 3.52 p = 0. 06 ) assuming a level of significant less than or equal to 0.05. However, with a p value of 0.06, the results. For instance, if we accept a p value of 0.06 as a significant result, in degree centrality significantly predicted whether o r not a respondent indicated if they had changed their professional opinion on a significant technical or scientific issue related to marine 2 = 3.52, df = 2 N = 121, p = 0.06) Table X. 9 presents the odds ratios, which suggest that the odd s of a significant change in professional opinion on a significant technical or scientific issue decrease as in degree centrality increases (i.e., the odds ratio for in degree centrality is less than 1) Table X. 9 Logistic Regression Predicting Opinion Change on Scientific/Technical Issues Variable B SE Odds Ratios p In degree Centrality 0.92 0.05 0.912 0.06 Constant 0.72 0.52 2.05 0.17 This result in degree centrality was found to be negatively associated with lear n ing. Finally, i n degree centrality was not a signif icant predictor of whether or not a respondent indicate d a change in professional opinion on a significant policy issue 2 = 0. 63 p = 0. 43 ). E IGENVECTOR C ENTRALITY AND L EARNING Individual eigenvector centrality ( M = 60.6 SD = 19.7 ) was not a significant predictor of reported learning ( M = 0.844, SD = 0.770), F (1, 111 ) = 0. 294 p = 0. 589 Though eigenvector
121 centrality was not a significant predictor of learning (as a scale variable), each learning question was evaluated to determine if eigenvector centrality was a significant predictor of an individua l learning metric. S imple linear regression was performed using responses to each learning statement. For clarity, those statements include: Participation in the partnership has given me a better understanding of aquaculture science. Participation in the partnership has given me a better understanding of aquaculture policy, law, or regulations. Participation in the partnership has given me a better understanding of aquaculture economics or business. Participation in the partnership has given me a better u nderstanding of other stakeholder perspectives Eigenvector centrality was not a significant predictor (i.e., p <0.05) of any individual metric of learning (F values ranging from 0. 292 to 0. 903 and p values ranging from 0. 3 44 to 0. 590 ). Though eigenvector centrality did not significantly predict learning when measured in changed opinion (both scientific/ technical and policy opinion). For example, respondents were At least partially through your participation in the partnership, have you changed your professional opinion on any significant scientific or technical issues related to marine aquaculture? At least partially through your participation in the partnership, have you changed your professional opinion on any significant policy issues related to marine aquaculture?
122 B inary logistic regression was performed to determine if eigenvector centrality significantl y predicted learning in terms of changed professional opinion (technical and policy opinion). Eigenvector centrality significantly predicted whether or not a respondent indicated they had changed their professional opinion on a significant scientific or t echnical issue related to marine aquaculture ( 2 = 5.86 df =2, N = 123 p <0.05). Table X. 10 presents the odds ratios, which suggest that the odds of a significant change in professional opinion on a scientific or technical issue decrease as eigenvector centra lity increases (i.e., the odds ratio is less than 1). Table X. 10 Logistic Regression Predicting Opinion Change on Scientific/Technical Issues Variable B SE Odds Ratios p Eigenvector Centrality 0. 02 0.0 1 0.97 0.01 9 Constant 1.2 2 0. 62 3.39 0.0 50 N etwork Centrality and Trust Hypothesis 7(H7) Centrality and Trust (at the individual level): The higher the centrality of an i ndividual among members of collaborative partnerships, the higher the level of reported trust among those same members of the partnership. S imple linear regression was conducted to determine if network centrality was a significant predictor of perceptions trust among respondents. As indicated earlier, this was conducted at the individual level (N=110) using network centrality (out degree, in degree and eigenvector) and reported trust as variables. This was performed for each trust question as well as the trust scale variable. The assumptions of linearity and normality were checked, and each was met, indicating that the use of simple regression techniques is appropriate for this dataset. However, as indicated earlier, the assumption of independent observ ations was obviously not met, since respondents were often participants from the same partnerships. This was addressed For simplicity, results for each centrali ty variable are presented individually.
123 O UT DEGREE C ENTRALITY AND T RUST Individual out degree centrality ( M = 9.66 SD = 4.33 ) was not a significant predictor of reported trust ( M = 4.00, SD = 0.492), F (1,112) = 0.413, p = 0.522. Though out degree centrality was not a significant predictor of trust (as a scale variable), each trust question was evaluated to determine if out degree centrality was a significant predictor of an individual trust metric. As indicated above, perceptions of trust were mea sured by asking respondents to indicate whether the following statements apply to none, few, half, most, or all partnership participants: Are honest, forthright, and true to their word Are willing to listen and sincerely try to understand other points of v iew Reciprocate acts of good will or generosity Are trustworthy Out degree centrality was not a significant predictor (i.e., p <0.05) of any individual metric of trust (F values ranging from 0.017 to 1.16 and p values ranging from 0. 284 to 0. 897 ). I N DEGR EE C ENTRALITY AND T RUST Individual in degree centrality ( M = 9.83 SD = 3.78 ) significantly predicted trust ( M = 4.00, SD = 0.492), F (1,110) = 4.66, p = 0.03, adjusted R 2 = 0.03. According to Cohen (1988), this is a small effect size. T he beta weights, presented in Table X.1 1 indicate that when density increases by one unit, trust (scale) decreases by 0.026 units. This result is in clear opposition to the expected relationship between centrality and trust Table X.1 1 Simple Linear Regression Analysis Summary for In Degree Centrality Predicting Trust (Scale) Variable B SEB In degree Centrality 0.026 0.012 0.202* Constant 4.25 0.126 Note: R 2 = 0.03, F (1,110) = 4.66, p <0.05 p =0.03
124 Though it appears that there is a significant (negative) relationship between trust and individual in degree centrality, each trust question was evaluated to determine if in degree centrality was a significant predictor of an individual trust metric. As i ndicated above, perceptions of trust were measured by asking respondents to indicate whether the following statements apply to none, few, half, most, or all partnership participants: Are honest, forthright, and true to their word Are willing to listen and sincerely try to understand other points of view Reciprocate acts of good will or generosity Are trustworthy Individual in degree centrality significantly predicted trust for the following statements: Are willing to listen and sincerely try to understand other points of view ( M = 4.09 SD = 0. 544 ), F (1, 110 ) = 6.25 p = 0.0 14 adjusted R 2 = 0.046 and Are trustworthy ( M = 4.12 SD = 0.599 ), F (1, 108 ) = 3.97 p = 0.0 49 adjusted R 2 = 0.027 The beta weights, presented in Table X.1 2 and X.1 3 indicate that when individual in degree centrality increases by one unit, the level of trust actually decreases by 0.033 units ( willingness to listen and understand ) and 0. 030 units ( trustworthy ). Table X.1 2 Simple Linear Regression Analysis Summary for In Degree Centrality Predicting Trust (willingness to listen and understand) Variable B SEB In degree Centrality 0.033 0.013 0.233* Constant 4.405 0.139 Note: R 2 = 0.0 46 F (1,110 ) = 6. 25 p =0.01 4 p <0.05
125 Table X.1 3 Simple Linear Regression Analysis Summary for In Degree Centrality Predicting Trust (trustworthy) Variable B SEB In degree Centrality 0.030 0. 015 0. 189 Constant 4.391 0. 156 Note: R 2 = 0.0 27 F (1, 108 ) = 3.97 p =0.0 49 p <0.05 Similar to the results for the scale trust variable, these regression results indicate that the type of trust that is negatively associated with in degree centrality is specific to a respondent s perception of how many partnership participants are willing to listen and understand other points of view as well as how many partnership participants are trustworthy. In degree centrality was not a significant predictor of trust as measured by respondent perception of how many partnership participants are honest, forthright and true to their word or reciprocate acts of good will or generosity E IGENVECTOR C ENTRALITY AND T RUST Individual eigenvector centrality ( M = 60.6 SD = 19.7 ) was not a significant predictor of reported trust ( M = 4.00, SD = 0.492), F (1,11 0 ) = 1.437 p = 0. 233 Though eigenvector centrality was not a significant predictor of trust (as a scale variable), each trust question was evaluated to determine if eigenvector centrality was a significant predictor of an individual trust metric. As indicated above, perceptions of trust were measured by asking respondents to indicate whether the following statements apply to none, few, half, most, or all partnership participants: Are honest, forthright, and true to their word Are willing to listen and sincerely try to understand other points of view Reciprocate acts of good will or generosity Are trustworthy
126 Eigenvector centrality was not a significant predictor (i.e., p <0.05) of any individual metric of trust (F values ranging from 0.0 05 to 2.54 and p values ranging from 0. 114 to 0. 945 ).
127 CHAPTER XI DISCUSION OF RESULTS This study has provided empirical and qualitative results with evidence supporting several of the hypotheses put forth, while also putting into question many others. Each hypo thesis (or group of hypotheses) will be discussed in detail in this chapter. T able XI .1 presents a summary of each hypothesis tested as w ell as a qualitative indicator regarding whether or not the results presented in Chapter V provide support for each hy pothetical contention. Table XI .1 Summary of Results and Qualitative Support for Hypotheses Hypothesis Supported Some Support Not Supported H1: In collaborative partnerships, agreement on major policy issues between members will be an important factor in the formation of individual coordination networks. H2: In collaborative partnerships, trust between members will be an important factor in the formation of individual coordination networks. H3: In collaborative partnerships, perceived resource attributes between members will be an important factor in the formation of individual coordination networks. H4: The higher the network density for a collaborative partnership, the higher the level of reported learning on average across members of the partnership. H5: The higher the network density for a collaborative partnership, the higher the level of reported trust on average across members of the partnership. H6: The higher the centrality of an individual among members of collaborative partnerships, the higher the level of reported learning among those same members of the partnership. O* I* ** E* ** H7: The higher the centrality of an individual among members of collaborative partnerships, the higher the level of reported trust among those same members of the partnership. O ** I ** E ** *O = Out degree centrality; I = In degree centrality; E = Eigenvector centrality *Not only was the hypothesis not supported, but an inverse relationship was observed.
128 Each hypothesis or group of related hypotheses will be discussed in the following section. They have been restated to streamline the discussion. Factors Influencing the Formation of Coordination Networks (H1) In collaborative partnerships agreement on major policy issues between members will be an important factor in the formation of individual coordination networks (H2) In collaborative partnerships trust between members will be an important factor in the formation of individual coordination networks (H3) In collaborative partnerships perceived resource attributes between members will be an important factor in the formation of coord ination networks Results presented in Chapter V showed that the most important factors for respondents when deciding with whom to coordinate are associated with ideas of trust ( professional competence and trust to keep promises ) as well as resources ( ac cess to expertise on major aquaculture issues and influence outside the partnership ) In fact, professional competence and access to expertise were significantly more important to respondents than all other factors considered. These results provide little support for the belief homophily hypothesis, which predicts that agreement on aquaculture policy issues will drive the formation of coordination networks in aquaculture partnerships In fact, shared belief s was reported as the least important facto r among respondents. Rather, trust and resource attributes dominated coordination decision making for respondents. For example, professional competence an attribute of trust (Lubell 2004), was identified as the most important coordination factor, follow ed by the resource attribute of access to expertise (not significantly different from professional competence), and the trust attribute of trust to keep promises and the resource attribute of influence outside the partnership So, rather than a structure of shared beliefs guiding coordination activity within partnerships, it appears that trust and access to resources are the dominant forces.
1 29 Given this, there appears to be some support for the trust and resources hypotheses. However, given the technica l nature of the partnerships included in this study, there may be an interaction or at least a blending of the ideas presented in each. For example, as indicated in Chapter V, professional competence and access to expertise each share the ideas of knowled ge acquisition as well as the expert; one relating to access to expertise and the other to the reliability (or competence) of that expertise. Access is important, but it is important that it is reliable. Though the intention was to operationalize two dis tinct concepts, they obviously share some important attributes. This finding is not necessarily surprising. For example, organizational management literature has pointed to the mediating role of expertise, competence based trust and benevolence based tr ust in organizational knowledge transfer (Leven and Cross 2004). Given this, the findings in this study seem to indicate that the interaction of competence and expertise may also extend from knowledge transfer to individual co ordination decision making, a n important external (contextual) validation of the work performed by Leven and Cross. These results were somewhat consistent across partnerships as well as organizational affiliations, rily driving results from the grouped data. Equally as important, results of the QAP showed no significant correlation between individual beliefs and coordination networks, further supporting the trust and perceived resource attributes hypotheses. What is even more striking is that these trust and resource attributes trumped an institutional mandate for coordination (i.e., the presence of a legal requirement ). One would reasonably expect that the presence of a legal requirement might dominate coordinati on activities for government representatives (i.e., the mandate that created the partnership might require an
130 inclusive, non selective process of coordination for government representatives). However, even among government respondents professional compete nce and access to expertise were still the two most important coordination factors. This may have been due to differences in the organizational mandates of each partnership (i.e., there might not have been a requirement for inclusivity for government part icipants). Even so, the importance of a legal requirement for government representatives was substantially higher than for other groups (fifth most important factor for the combined data and third most important for government representatives). Interview Responses Interviews were conducted to provide some insight into survey responses and to see if there were obvious discrepancies between and survey and interview responses regarding coordination factors I nterviewees were asked the following questions re garding coordination : Wh ich groups in this partnership do you tend to coordinate with most frequently? Probe: Why do you tend to coordinate with those groups and not with others?" Probe: Can you tell me the types of coordination activities that you engage in? Interview responses generally supported survey results which highlighted the importance of access to expertise as well as professional competence However, this tended to be linked to the need to accomplish a specific task associated with partnership activities, such as to conduct a research project, develop a plan, or evaluate a technology. Interview responses are discussed below. Though responses to this series of interview questions were quite limited (primarily due to use of multiple interview protocols during execution of the project several of which had not
131 been revised to include the coordination series of questions at the time of use), of the 18 individuals who answered the coordination interview questions, 14 ( 78%) cited an organization /gr oup or an individual that they relied upon for professional purposes, and indicated access to expertise or need for information gathering as a reason Of those 14, three also included the idea of trust (having worked with them in the past and professional competence), and one included the idea of self interest. An additional four respondents (22%) indicated coordination was driven by contractual or organizational affiliation reasons. For example, one participant indicated that Our biologists would coordinate with scientists in the group and the biology sub group in order to develop the biology plan for the group. I, however, was interested in regulation, so I coordinated with the CRMC staff. We all tended to work with each other depending on the issue Another respondent took an equally utilitarian approach stating information about siting for on The idea of expertise directed at accomplishing a par tnership objective or to fill knowledge gap was echoe d in several other responses as well including the following: "We [partnership members] all worked with each other fairly often since you needed We tended to work with th e group as a whole mainly, unless we needed additional information or to get technical understanding about something."
132 table. I worked with those folks on the biology there, to provide some expertise in the area and help get around some of the biology t depended on what you had to get to get done, and what your role was in the group. We depended on others to brin g in their perspective and expertise to the table so we s a scientist and has access to an actual facility in Puerto Rico. That way we could see how things were actually working, how they ran, and what problems they were having and how to avoid them. inly, unless we needed additional information or to get technical understanding about something. So mostly do a lot of research well In addition to access to expertise to fill knowledge gaps, three respondent s added the idea of the importance of trust ( i.e., existing working relationship and professional competence )
133 [name speak directly with the Vet for [aquaculture company name] who is our main fish closest working relationship. Our specializations are very comple mentary to each other. person. All of them on the council are really good at what they do and very competent." Interestingly, one interviewee indicated a utilitarian view of coordination (similar to thos e presented above) coupled with that of self interest, on. The biologists tend to cooperate together to answer a question, since the group needed them to. The farmers would work together since they were going to be impacted by the recommendations of this group. They were self Four interview responses fell outside of the expertise/knowledge gap reasoning for coordination, citing more formalized industry or organization based reasons,:
134 existence that I dealt with them at all. I have very little reason to do s FHTC and the State Vet. I was also joint Similar to the survey results, int erview responses indicate that indeed, access to expertise, specifically to fill a knowledge gap, was important for individuals when making decisions regarding coordination, with 78% of responses indicating so. Interestingly, trust seems to play a part as well for some respondents, in the form of existing working relationship as well as professional competence. The four responses that fell outside of this general trend are interesting in their own right. Two responses (11%) indicate that there is a struc tural attribute to coordination that is an important coordination factor for some, based on either contractual obligations or their official position within the partnership. The remaining two (11%) indicated that they tend to stay within their respective industry caucus when making coordination decisions. Interviewees indicated that the type of coordination they engaged in with other members of the partnership ranged from formal to informal. Formal activities engaged in by partnership participants was so mewhat limited, and incl ud ed the following: engaging in projects on a contract basis, developing policy recommendations or guidance per organizational mandate, conducting formal literature reviews, developing proposals or grant requests, and working on spe cific deliverables mandated by the partnership (such as guidelines or plans)
135 Informal coordination was widespread and manifested itself a variety of ways. These included casual sharing of information in person, via email or via telephone about the follo wing: upcoming projects or meetings, clarification of technical issues, environmental impact discussions, discussions of economic aspects of aquaculture (such as startup costs, markets and earning potential), interpreting research, getting policy informati on (such as lease size constraints), exchanging ideas, providing technical expertise, listening to concerns, answering technical and non technical questions, and discussions regarding specific issues being dealt with by the partnership. Interestingly, the importance of informal coordination in the partnership process was articulated by one interviewee, who said information, things we discussed, ideas and going back to my group to get their input or buy in, and take that bac k to the working group. The idea was to get consensus, and that was a back Learning and Network Structure Hypothesis 4 (H4) Network Density and Learning (at the partnership level) : The higher the network density for a collaborative partnership, the higher the level of reported learning on average across members of the partnership. Hypothesis 6 (H6) Centrality and Learning (at the individual level) : The higher centrality of an individual among members of collaborative partnershi ps, the higher the level of reported learning among those same members of the partnership. Results from this study clearly indicate that the partnerships included in th e APP are providing opportunities for learning Mean scores for l earning (operational were positive for each learning question (i.e., agreement with the learning statements provided) when grouped for all respondents, as well as when considered for each individual partnership and organizational affiliation. Learning within these partnerships appears to have translated into
136 opinion change on scientific/technical issues related to marine aquaculture (41% indicating a change in opinion) as well as policy issues (46% indicating a change in opinion). Given this, it appears that the collaborative stakeholder partnership approach, a form which has been highly favo red as a policymaking platform specifically due to its ability to stimulate le arning and discourse (Weible and Sabatier 2009), is indeed serving as somewhat of a crucible for learning. Importantly, scholars interested in collaborative learning and practitioners who rely on collaborative structures have a need to understand the ro le that structure plays in knowledge acquisition and learning. As shown in Chapter X network density (which showed an almost two fold difference among partnerships with the least dense to the densest coordination networks) appears to play a significant and substantial role in learning within these partnerships. Regression results not only showed a significant positive relationship, but the adjusted R 2 value indicates that almost 34% of the variability identified in learning was explained by network dens ity. These results show strong support for the network density and learning hypothesis, and also validate social network research efforts that have identified density as a network variable of interest. The benefits of dense network structures have been shown to be wide ranging. Having man y links to other network nodes provides opportunities for informational exchange, can help to preserve institutional knowledge (by making knowledge within networks redundant, hence reducing the loss of a network node l ess impacting; Borgatti 2003) and can also foster feelings of group identity and enhance understanding (Coleman 1990). In our case, we see that this may translate into opportunities for learning. Further, network density is an important metric since it d discussed next). Rather, it is one measure of the diversity of the network structure since it
137 considers the number of the actual network connections within the c ontext of all possible connections. This is an important aspect of network structure. R ecent r esearch conducted by Leach et al (2013) on fostering learning through collaboration (a study performed as a part of the Aquaculture Partnerships Project) showed that knowledge is directly correlated with diverse partnership participation. In an environment with diverse participation and dense network structure, the opportunities for learning from a variety of sources are increased. Along the same line of thou ght as network density, it was my contention that highly central individuals would experience a high level of learning (hypothesis H6). This was not wholly supported by the results presented in Chapter X and in two instances it was outright rejected. Th e centrality measures considered in this study (out degree, in degree and eigenvector) returned mixed results. Generally speaking, individual centrality tended to show an inhibitory effect (if any) with respect to learning at least for scientific and lea rning For example, in degree centrality was actually negatively correlated with the learning scale variable, as well as the individual learning statements regarding understanding of stakeholder perspectives and aquaculture science. A similar result was obser ved for changes in professional opinion, where results showed that the odds of a change in professional opinion on a scientific or technical issue decrease as in degree centrality increases Eigenvector centrality was not significantly related to the lear ning scale variable or any of the individual learning statements, but showed a negative relationship in terms of opinion change. Similar to in degree centrality, results showed that the odds of a change in professional opinion on a scientific or technical issue decrease as eigenvector centrality increases. These results indicate that high levels of individual centrality might actually dampen learning, specifically in technical areas. These results should be interpreted carefully however,
138 for several rea sons. Results of regression modeling showed very small effect sizes ( R 2 values ranging between 0.03 and 0.05; suggesting that only 3 to 5% variability in learning can be explained by centrality measures). Even so, these results have some support in the n etwork literature. For example, historical work performed by Leavitt (1951) and Shaw (1981) indicated that a high degree of individual centrality might give rise to what is termed c entralized earning) leading to less experiential and adaptive learning. Similarly Frank and Yasumoto (1998) found that too many links emanating from one individual may tend to anchor an actor into a political position (due to social friendships, peer pressure, etc. ), which may inhibit his ability or willingness to incorporate new information, innovate or act. The nature of the network is important to consider as well. If an actor is coordinating with many individuals, however the individuals are all like minded (p erhaps in similar types of organizations, etc.) or relatively homogenous in background, then this may lead to an inhibition of learning since the exposure to or chance for incorporation of new information may be limited Additionally, as discussed earlier highly central individuals may hold relatively powerful positions within a network and may be somewhat entrenched in their policy positions and less likely to report a change in knowledge or understanding. Likewise, highly central individuals may have s ubstantial expertise in a specific area, and may thus be less likely to exhibit changes in their understanding of policy issues or change their policy position, where individuals with less experience may report more learning simply because they have more t o learn. Out degree centrality was not a significant predictor of learning when considering the scale learning variable, individual learning statements, or a change in professional opinion on a technical or scientific issue. However, it did significantly predict whether or not a respondent
139 changed their opinion on a policy issue; showing that increased out degree centrality increased the odds of a changed opinion on a policy issue. This provided some support for the centrality and learning hypothesis, howe ver it is specific to policy opinion. Where the other measures of centrality appear to be negatively associated with learning on a scientific/technical level, out degree centrality appears to promote learning (in terms of opinion change) regarding aquacul ture policy. Again, the effect size is relatively small, so the amount of learning explained by out degree centrality is therefore small as well, however it is notable. One explanation of this finding is the nature of the centrality measure. Out degree is a measure of external node contact (i.e., reaching out to others), which could be an indicator of information seeking, attempts to fill needed resources, attempts to relay information or attempts to convince others. Through this lens, out degree centra lity might be seen as an indicator of active information seeking. In our case, the type of information seeking appears to be related to aquaculture policy. Out degree centrality is also a direct centrality measure for all survey respondents. For example, each respondent has an out responses to the coordination matrix. The other two centrality measur es could be interpreted as having a component resembling more of a proxy measure, since all members of a specific organization receive the same centrality value (as discussed earlier). Collaborative learning literature seems to support what we are seeing in this study. Small effect sizes and slight indications of inhibition of learning may be by products of influences of a larger model that must consider other critical learning variables. For instance, as a part of the Aquaculture Partnerships Project, Leach, et. al (2013) developed a model of knowledge acquisition leading to belief change, which included partnership traits such as 1) diverse participation, 2) perceptions of procedural fairness, 3) lack of adequate science for decision
140 making, and 4) tru stworthiness of other participants. These traits were shown to lead to an increase in knowledge acquisition and subsequent opinion change and support findings in previous work regarding factors for successful collaborative designs (Lubell, Leach, and Sab atier 2009) Given this, as well as the role that network density appears to play (as mentioned in this study), the aspects of network centrality and the influence on knowledge acquisition might end up somewhat muted in comparison. Trust and Network Struc ture Hypothesis 5 (H5) Network Density and Trust (at the partnership level): The higher the network density for a collaborative partnership, the higher the level of reported trust on average across members of the partnership. Hypothesis 7(H7) Centra lity and Trust (at the individual level): The higher the centrality of an i ndividual among members of collaborative partnerships, the higher the level of reported trust among those same members of the partnership. Results of this study indicate generally high levels of trust among partnership respondents, with most or all participants being perceived as honest, forthright and true to their word; having reasonable motives and concerns ; willing to listen and try to understand other points of view; and trustworthy. Additionally, between half and most participants were perceived as having the same values and priorities and willing to reciprocate acts of good will and generosity. Mean values for responses to each trust question were well above the sc ale midpoint (3.00), ranging from 3.58 to 4.20. Similar results were found for mean trust scores for each individual partnership (ranging from 3.3 to 4.8) as well as organizational affiliation (3.79 to 4.05). Since one of the goals of inclusive governanc e approaches is to engender trust among stakeholders (Bodin et al 2009), it appears that the partnerships included in this study have made great strides in achieving that
141 As mentioned earlier in this study, trust is one of the critical characteristics of successful natural resource governance strategies (Bodin et al 2009). Therefore eliciting the role that network structure has in engendering trust is an important task. Results from this study did not support the trust hypotheses (H5 and H7; the conte ntions that network density and centrality would be associated with higher levels of trust). Contrary to expectations, network density, out degree centrality and eigenvector centrality were not significantly associated with levels of trust reported among and within partnerships. Similar to what was observed for learning, the trust scale variable was actually inversely related to in degree centrality. In other words, the higher the level of in degree centrality, the lower the reported level of trust. Whe n evaluated for each individual trust question, in degree centrality was found to have the same inverse relationship This result may have a similar explanation as the results for learning, and may be rooted in the argument presented by Frank and Yasumoto (1998). Though their findings were related to learning and innovation (finding that too many links associated with one individual may tend to anchor an a ctor into a political position [ due to social f riendships, peer pressure, etc.] which may inhibit his ability or willingness to incorporate new information, innovate or act), the concept is certainly applicable to tr ust. In degree centrality can be thought of as a measure of input, connections incoming from outside nodes. A high level of in degree centrality (indicating power and popularity in some circumstances) might have a similar effect as that noted for learnin g, where an individual may end up rooted in a political position. Additionally, if an individual is the target of many network connections, then he may be exposed to a larger amount of
142 information, organizational or individual agendas, personalities, and attempts by others to convince, etc. These results should be interpreted carefully however. Results of regression modeling showed very small effect sizes ( R 2 values ranging between 0.03 and 0.05 for the scale variable and individual trust questions; suggesting that only 3 to 5% variability in trust can be explained by in degree centrality). Even so, these results have some support in the network literature. A similar result was found in an analysis of network centrality and high organizational performance (Eckenhofer 2010). In this study of three networks of 170 university students, it was found that individual in degree centrality (measured as in closeness) was inversely related to metrics of going T he Eckenhofer study looked only at associati on (through correlations) rather than causation. However, the results do indicate that in degree centrality measures have instances of negative association with trust.
143 CHAPTER XII STUDY LIMITATIONS The research design guiding implementation of this study was developed with the intent to address and mitigate factors that could bring into question study results. However, limitations and challenges still remained in spite of these efforts. This section will address issues associated with study validit y and reliability, as well as problems of endogeneity and network measure robustness. Internal validity, or the strength of the notion that the independent variable is responsible for the observed change in the dependent variable, can be impacted by seve ral factors. These include history, maturation, testing effects, instrumentation, statistical regression, selection bias, and attrition (Singleton, 2004). Briefly, history refers to an event or events that occur during implementation of the study that ca n impact the study outcome. For example, as mentioned earlier in this paper the debate over the development of the aquaculture industry encompasses a wide variety of social issues, including economic development of small coastal communities. Therefore, i t is quite possible that media coverage of a severe economic recession or large environmental disaster might impact how individuals respond to questions regarding their views on potentially costly environmental regulations or the role of government and mar kets in society. History is considered a threat to the study at hand. In order to address this threat, I incorporated an approach that used informal and formal interviews as well as a questionnaire. Use of interviews allowed the research team to pose questions to participants to discover variables outside of the study that could be influencing the observed outcomes. Additionally, to reduce the potential of a confounding event taking place during the study, partnership interviews and surveys were condu cted within a one year period. However, even with this approach, the
144 sample of partnerships was indeed impacted by a large scale environmental disaster, the Deep Water Horizon crude oil release that occurred in the Gulf of Mexico in 2010 occurred at the s ame time that an aquaculture partnership in Louisiana was going to be interviewed and surveyed as a part of this study. Therefore, in consideration of the proximity of the oil release to the partnership, an alternate group located in Florida was substitut ed. Even after taking these measures, the occurrence of the crude release may have impacted the data collected from the final group in this study. Maturation refers to the internal participant changes that can take place during a study. Maturation effects can range from permanent changes in test subjects (such as an individual maturing in different ways [physically or intellectually] if involved with longer te rm studies), to temporary ones, such as fatigue. Each of these conditions may confound the study results. Since this the design of this study includes one time surveys and interviews, neither temporary nor permanent changes in participants are considered a threat. Testing effects refer to the tendency of participants to respond in ways that might be deemed more socially responsible (or at least less socially reprehensible) than what is actually true, or answer in ways that they anticipate the researche r would like. For example, in this study participants were asked to respond to questions that relate to beliefs about the environment. Testing effects might be encountered if participants artificially inflate their responses to garner a higher estimate o f overall environmental beliefs (i.e., pro environmental) believing that it is more socially acceptable to have those beliefs. This is especially problematic when individuals are given questionnaires multiple times (i.e., a pre test/post test design). Si nce the study at hand is a post test only design, the issue of testing effects due to multiple applications of the research tool is not relevant. However, to address problems associated with the social acceptability of
145 responses, the participants were giv en information regarding the purpose of the study, the need for honest responses, and finally an assurance of anonymity. Secondly, as discussed earlier in this paper, I employed the use of the QAP as a cross checking tool to determine if individual belief s are correlated with individual coordination networks. These results were compared with the means testing results for individual coordination factors. In other words, in the event that individuals cited attributes of trust as a primary driver of coordin ation network formation, I was able to cross check those results to verify that, in fact, it was or was not beliefs that were actually influencing network formation. Instrumentation threat is associated with changes in the delivery of the research tools such as interviews or other observations. For example, as an individual performs multiple interviews the way those interviews are performed might change over time. This could be due to the comfort the researcher has with the interview script, or the in terviewer may become more observant over time. Additionally, if multiple interviewers are being used, study results could be confounded if the way the interview is performed differs from person to person. In the current study multiple interviewers were u sed. Therefore, prior to conducting interviews an interview script was developed by the research team that included specific questions to be asked. Interviewers reviewed the script to identify any areas that are not clear (regarding question intent). Th e interviewers routinely reported interview results to the larger research team with the goal of identifying any previously unforeseen problems and/or issues that could impact the data collected. An important part of the process was routinely meeting as a group to make sure the appropriate questions were being asked in a way that was consistent across the research team. Also, mock interviews were performed prior to conducting formal interviews to ensure that there was a consistent approach taken by the in terviewers. Even with this approach in place, early
146 interviews that were conducted with the PAC did not include the questions relevant to the coordination portion of this study. Therefore, interview responses were somewhat limited. Additionally, it was observed that follow on probes were not necessarily pursued by each interviewer. This also resulted in a limited amount of information that could be gleaned from interview data. Selection bias was probably the most severe threat to the validity of the study at hand. This threat occurs when groups selected for study are not equivalent, or if their equivalence is unknown prior to study implementation. This is especially true when the research d esign includes control groups. Since the current study does not include control group comparisons, this aspect of selection bias should not be a significant concern. However, comparability of the partnerships selected for study could be problematic. To address this, earlier in this paper I proposed a clear and specific definition of partnerships that assisted in identifying groups appropriate for study. In addition, review of documentation associated with the partnerships (such as by laws, charters, rul es of order, membership, etc.) allowed for identification of potential confounding inequalities among groups. Finally, the use of informal interviews as well as an advisory board enabled pilot testing not only of the survey instrument, but also (and more pertinently here) the groups to be included in the study. The advisory board and preliminary Since the current design employs a post test only approach, reg ression towards the mean (i.e., extreme scores becoming less so on consecutive tests) and attrition (loss of participants during the study) should not pose substantial threats to internal validity. External validity, or the generalizability of the study a cross multiple contexts, was addressed by incorporating 10 partnerships in the study that vary in geographical representation. In addition, there is the
147 potential to compare results from this study to those of previous studies performed on watershed partn erships. However, this does not obviate the potential problems associated with external validity. The groups selected for this study tend to be dominated by government representatives as well as industry representatives, and the partnerships were designe d to guide industry development, not to inhibit it. This could have a substantial impact on how individuals within these groups interact. For example, in groups with relatively equal representation of pro and anti minded individuals, factors influencing network formation might be substantially different from groups where there is a fairly uneven representation. The same might be said for groups that have varying levels of conflict. This was a limitation of the study at hand but also provided ample oppor tunity for additional study. Several steps were taken to bolster the reliability of the data generated as a part of this study. First, preliminary interviews were used to help generate questions for inclusion on the survey tool and on the formal intervie w protocol. Once the survey tool was developed, it was pilot tested on the advisory board as discussed earlier in this paper. This provided a level of verification as to whether or not the appropriate questions were being asked and, if so, were asked in an appropriate manner. In addition, attempts were made to maintain internal measure consistency by developing multi item scales to measure the variables of interest. The reliability of those multi item scale data was evaluated by generating inter item re liability indicators (i.e., item basis to elicit nuances among sample responses. One area that did impact the reliability of the study results is associated with learning. For example, I attempt to capture learning through a battery of questions that relates to the participation in the partnership leading to a better understanding of aquaculture science, policy, law, regulations, economics, business, and other stakeholder
148 perspectives. In addition, I also capture learning in terms of changing professional opinion. However, I do not capture the idea of reinforcement of understanding or beliefs, which is indeed an important aspect of learning. Again, this is a limitation that I was not abl e to address, however it is a gift to scholars looking for follow on study. Endogeneity presented a challenge to the evaluation of network data using ordinary least squares techniques put forth in this study. In terms of the study at hand, individual a nd partnership network characteristics (centrality and density) are assumed to be independent variables, with dependent variables being learning and trust. In other words, the underlying assumption is that network characteristics influence learning and tr ust. Given the available literature on the subject of social networks presented above, an argument can be made that, indeed, network characteristics (or position) tend to drive instrumental outcomes such as power, influence, learning and trust. Hence, th ere is support in the literature for network impacts on the gr chose to use regression techniques which rely on the definition of dependent and independent variables. Finally, I was faced with incomplete or erroneous netw ork data, bringing into question the robustness of the network measures used in this study. Specifically, the question should arise as to whether or not errors associated with network measures (related to response rates) will have such a substantial impac t on the resulting coordination networks that reliable conclusions cannot be reached. In other words, is it reasonable to compute network indices (such as centrality and density) when the network data is known to contain errors and/or omissions? Borgatti Carley
149 and Krackhardt (2006) explored the question of network measure robustness by conducting an evaluation of the impact of four types of errors (node addition, node removal, edge addition and edge removal) at varying rates (0%, 1%, 5%, 10%, 25%, and 5 0%) on known network data. Results of this study indicate that, under relatively small amounts of error (approximately 10%), network measures are actually quite robust. For example, their data showed that if the data collection method missed 5% of the ne twork ties, then the correlation between true and observed centrality (measured as degree, closeness, betweenness, and eigenvector centralities) would still th is, it seems that the network centrality m easures proposed in this study will be robust given small to modest rates of error. Thus, I should have been able to effectively discern the existence of an association between network characteristics and attribut es of trust and learning, even with suboptimal response rates. As was indicated earlier in this paper, the way coordination networks were generated was an additional challenge. The only direct network measure unique to each individual was out degree cen trality. The others required an assumption that all members of an organizational affiliation were equally cited (i.e., all members of state government received the same value for in degree and eigenvector centrality within a partnership). I attempted to m itigate this problem by generating values on a partnership by partnership basis, rather than by grouping all of that data together. In other words, state government representatives had different centrality values across each partnership. If I had grouped the data together, all state government representatives would have the same centrality values.
150 CHAPTER XIII CONCLUSIONS AND FUTU RE RESEARCH This study has continued efforts to conduct empirical, theory driven research to evaluate the factors that inf luence how individuals form coordination networks in aquaculture partnerships, how those networks are structured, and what the implications are for critical aspects of collaborative natural resource governance strategies, such as learning and trust. Each is discussed in turn. Recent work and conventional understanding of human interactions have suggested that beliefs are the primary heuristic guiding political coordination (Henry 2011). This contention is also a mainstay of foundational policy process th eory (Sabatier 1993). This study attempts to explore this assumption by evaluating it in juxtaposition to two alternative explanations for the formation of coordination networks in collaborative policy. Specifically I have evaluated the importance of bel ief homophily, trust and resources when individuals decide with whom they will coordinate. Data from this study did not support the belief homophily hypothesis (i.e., shared beliefs driving coordination behavior). Rather, individuals seem to place more importance on attributes associated with trust (professional competence and trust to keep promises) and resources (access to expertise and external influence). These results are not isolated in context. Recent studies of policy network cohesion have ind icated that although shared ideology is a strong polarizing force in collaborative environments, network cohesion is often better explained by power seeking relationships (as explored in Resource Dependence Theory) that allow actors within a network to inc rease their ability to influence policy outcomes (Henry et al. 2010). These results indicate that though long standing assumptions about individual behavior (political and otherwise) have applicability and provide value to understanding policy processes, research should be aimed at
151 testing the limits of those assumptions in order to advance the development of policy process theory, especially in the collaborative governance arena. Having said this, the role of ideology or shared beliefs should be neithe r discounted nor understated. The role of beliefs in the policy process has been documented in prior research. For example, Weible (2005) and Weible and Sabatier (2005) have shown that in situations where policy core beliefs are contested, ally networks tend to correlate with beliefs. Similarly, Henry et al. (2010) found that belief systems matter in forming collaboration networks, but more so to avoid actors with dissimilar beliefs than to link actors with similar beliefs. Put alongside the studies whi ch have highlighted the importance of belief homophily, this study provides a potential contextual boundary for the role of beliefs. An idea advanced in this study is that belief homophily may be a less important driver of coordination networks in enviro nments where policy core beliefs are not highly contested. This is not to say that conflicting beliefs were absent in the partnerships included in this study. For instance, oystermen in Rhode Island tended to view aquaculture development as a direct thre at to their livelihood, and demanded that regulatory officials chart a path forward to manage development of the industry, which brought about the RIAWG. However, the aquaculture partnerships included in this study had an underlying goal of advancing the industry, and most participants levied some agreement with the benefits of aquaculture, from economics to food production. Additionally there was overall agreement that the industry should be developed at some level. Thus, we see the genesis of some cont extual bounds based on the level of conflict: environments of collaborative partnership versus conflict over policy core beliefs.
152 Research has recognized that policymaking occurs within a variety of venues within any given policy subsystem (Baumgartner and Jones, 1993; Lubell et al 2010). These venues can be diverse with respect to composition, size, resources, structure, process and in many other ways. This aspect also provides fertile ground for additional inquiry. In order to advance policy process theory, future research could focus on comparisons between multiple policy venues with differing levels of belief conflicts as added variables of interest. This opens up the possibility for larger scale meta studies that could look across a variety of ven ues, in substance and in conflict level in order to evaluate the contention made in this portion of the study, that beliefs may be a weaker driver of coordination networks in environments where policy core beliefs are not highly contested. Results of this study have shown that, in the absence of major disagreement on policy core beliefs, individuals will coordinate with fellow participants who possess needed resources and/or those that are trustworthy to further their goals. This leads us to the question of coordination network structure and the variables of trust and learning. The importance of trust and learning for successful collaborative natural resource governance strategies has been well documented (Dervitsiotis 2006, Becerra 2002, Isaac et al 2007 Conley and Udry 2001) and should be treated as a research priority in this area of policy process inquiry. Results from this study provide some important insights and additions to the concept of learning in collaborative governance environments. Recent work performed by Leach et al. (2013) showed that knowledge and belief change are driven by partnership and individual traits. Partnership traits include diverse participation, conceptions of procedural fairness, lack of adequate science for decision mak ing as well as perceptions of trust among participants (2013). Individual traits include duration of participation in a process, level of scientific competence, norms of consensus and other demographics. This study has provided some support for
153 additiona l variables to be added to this model (on the partnership and individual levels), providing evidence for the idea that network structure is a critical variable to consider, specifically the density of a coordination network and individual centrality (out d egree). Each of these network variables appear to contribute to learning, and/or opinion change. Conversely, this study also identified the potential for certain attributes of network structure (namely in degree and eigenvector centrality) to inhibit learning and/or opinion change. Interestingly, in degree centrality was also found to be negatively associated with perceptions of trust. Though results of this study did not support the contention of a positive association of network variables and trust, the fact that in degree centrality appeared to be negatively correlated with trust and learning is intriguing, and brings about additional questions. Though the finding is supported in network literature, what is obvious is that not all centrality is equ al. Network literature has shown this time and again, but this study has shown the importance of differing centrality measures as distinct network variables. Obviously these results indicate the need for future research in a variety of areas. Trust has been shown to be a critical part of natural resource collaborative governance, and the lack of association between the network variables included in this study and perceptions of trust indicate that efforts should be made to explore how trust is engendere d within partnerships. An approach similar to that taken by Leach et al (i.e., identifying partnership level and individual level traits associated with trust) would be an important start. Additionally, critical analysis of how networks are elicited in p olicy process research while still adhering to guarantees of respondent anonymity should be conducted. As performed in this study, out degree centrality is the only centrality measure where unique and specific values could be assigned to a respondent. Th e
154 other measures of centrality rely on a proxy approach, which can end up reducing the amount of variability captured in survey results, and may obscure important relationships. This study contributes to the larger body of work related to the policy proc ess and was intended to answer the multiple calls for empirically based, theory driven research to help build and refine the theories that have been advanced in the field. What I have shown in this study is that alternative theory testing provides valuabl e insights into individual behavior in collaborative policy environments, adds to existing models of knowledge and trust and can help to establish and challenge the limits of theoretical research in the behavioral sciences. And while this study is specifi c to policy networks, the findings can certainly be translated to applications of public management theories with a focus on coordination patterns as well as variables critical to collaborative natural resource management success.
155 BIBLIOGRAPHY Agranoff, R and M. McGuire (2001). Big Questions in Public Network Management R esearch. Journal of Public Administration Research and Theory 11 (3), 295 326. Adler, Paul S. and S. Kwon (2002). Social Capital: Prospects for a New Concept. The Academy of Management Review, 27(1): 17 40. Alter, C., and J. Hage (1993). Organizations Working Together. Newbury Park, CA: Sage. Anderies, J.M., M.A. Janssen and E. Ostroom (2004). A framework to analyze the robustness of social ecological systems from an insti tutional perspective. Ecology and Society 9(1): 18. Baldwin, T. T., Bedell, M., & Johnson, J. J. (1997). Social networks in a team based MBA program: Effects on student satisfaction and performance. Academy of Management Journal 40, 1369 1397. Barda ch, E. (2001). Developmental dynamics: Collaboration as an emergent phenomenon. Journal of Public Administration Research and Theory 11 (2), 149 164. Becerra, M & Huemer, L. (2002). Moral character and relationship effectiveness: an empirical investigation of trust within organizations. In Moral Leadership in Action: Building and Sustaining Moral Competence in European Organizations Glos, UK: Edward Elfar Publis hing. Berdardo, R. (2009). Generalized Trust in Multi Organizational Policy Arenas. Political Research Quarterly, 61(1) 178 189. Black, K. (2001). Sustainabililty of Aquaculture. In Environmental Impacts of Aquaculture ed. K. Black, 199 212. Sheff ield, UK: Sheffield Academic Press. Bodin, O., B. Crona and H. Ernstson (2006). Social Networks in Natural Resource Management: What is There to Learn from a Structural Perspective? Ecology and Society, 11 (2). On line access at http://www.ecologyandsociety.org/vol11/iss2/resp2/ Bodin, O. and B.I. Crona (2009). The role of social networks in natural resource governance: what relational patterns make a difference? Global Environmental Change. 19 366 374. Bonacich, Phillip (1987) Power and Centrality: A Family of Measures. American Journal of Sociology, 92, 1170 1182. Borgatti SP (2006). On the robustness of centrality measures und er conditions of imperfect data Social networks (0378 8733) 28 (2) p. 124 Borgatti, S., M. Everett, and L. Freeman. (1992) UCINET IV Version 1.0 User's Guide Columbia, SC: Analytic Technologies.
156 Brass, D. and M. Burkhardt (1992). Centrality and power in organizations. In Network and Organizations: Structure, Form, and Action ed. Nitin Nohria and Robert Eccles, 191 215. Boston: Harvard Business School Press. Bressers, H., L.J. and J. Richardson (1995). Networks as Models of A nalysis: Water Policy in Comparative P Toole, Jr., and J. Richardson (Eds.), Networks for Water Policy: A Comparative P erspective (pp. 1 23). London: Frank Cass. Burbridge, P., V. Hendrick, E. Roth, and H. Rosenthal (2001). Social and Economic Policy Issues Relevant to Marine Aquaculture. Journal of Applied Ichthyology, 17 194 206. Burt, R. and M. Knez (1995). Kinds of Third Party Effects on Trust. Rationality and Society, 7 255 292. Burt, R. S. (2001). Bandwidth and echo: Trust, information, and gossip in social networks. In Networks and markets: Contributions from economics and sociology ed. Alessandra Casella and James Rauch, 30 74. New York: Russell Sage. Carley Baxter, Lisa R., Craig Hill, David Roe, Susan Twiddy, Rodney Baxter and Jill Ruppenkamp (2009). Survey Practice 2(7) on line journal access. Coleman, J.S. (1988 ). Social Capital in the Creation of Human Capital. American Journal of Sociology Supplement 94 S95 S120. Cook, K., R. Hardin, and M. Levi (2005). Cooperation Without Trust? New York: Russell Sage. Crona, B. and O Bodin (2006). What You Know is Who You Know? Communication Patterns among Resource Users as a Prerequisite For Co Management. Ecology and Society, 11 (2): 7. Dekker, D., Krackhardt, D. and T. Snijders. 2008. Sensitivity of MRQAP tests to collinearity and autocorrelation conditions. Psychometrika, 72 : 563 581. Dervitsiotis, K. (2006). Building trust for excellence in performance and adaption to change. Total Quality Management, 17(7), 795 810. Downs, A. ( 1957 ) An Economic Theory of Democracy New York: Harper and Row. Eckenhofer, E. M. (2010). Is centrality the key to high performance? Journal of Competitiveness 2, 58 73 Farrell, H. and J. Knight (2003). Trust, Institutions, and Institutional Change: Industrial Districts and the Social Capital H ypothesis. Politics and Society, 31(4), 537 566.
157 Faust, K. (1997). Centrality in Affiliation Networks. Social Networks 19, 157 191. Fernandez Giminez, Maria E, H. Ballard and V. Sturtevant (2008). Adaptive management and social learning in collaborative and community based monitoring: A study of five community based forestry organizations in the western USA. Ecology and Society 13 1 22. Festinger, L. (1957). A Theory of Cognitive Dissonance Stanford, CA: Stanford University Press. Firestone, J. (2003). Fish as Pollutants: Limitations of and Crosscurrents in Law, Science, Management, and Policy. Washington Law Review, 78, 693 722 Firestone, J., W. Kempton, A. Krueger, and C. Loper (2004). Regulating Offshore Wind Power and Aquaculture: Messages from Land and Sea. Cornell Journal of Law and Public Policy, 14 71 111. Francik, A. and C. Hershner (2003). Sustainable Aquaculture: Developing the Promise of Aquaculture. Aquaculture International 11 517 530. Frederickson, H. G. (1996). The spirit of public administration San Francisco: Joss ey Bass. Fukuyama, J. (1999). Essay for the International Monetary Fund Conference on Second Generation Reforms. Presented November 1999). Gabbay, S.M. and E. Zuckerman (1998). Social Capital and Opportunity in Corporate R&D: The Contingent Effect of Contact Density on Mobility Expectations. Social Science Research, 27 189 217. Gieseck e, Joan (2012). The Value of Partnerships: Building New Partnerships for Success. Faculty Publications, U niversity of N ebraska, Lincoln L ibraries. Paper 294. http://digital commons.unl.edu/libraryscience/294. Gnyawali, Devi and R. Madhavan (2001). Netwo rk Structure and Competitive Dynamics: A Structural Embededdness Perspective. Academy of Management Review, 26 431 435. Gray, B. (1989). Collaborating: Finding Common Ground for Multiparty Problems. San Francisco: Jossey Bass. Gray, B. and D.J. Woo d (1991). C ollaborative Alliances: Moving From Practice to Theory. Journal of Applied Behavioral Science, 27(1) 3 21. Hall, T. E and L.J. (2000). Structures for Policy Implementation: An A nalysis of National L egislation, 1965 1966 and 1993 1994. Administration and Society 31 667 686.
158 ( 2004 ) Shaping Formal Networks Through the Regulatory Process. Administration and Society 36(2), 186 207. Hanneman, Robert A. and M Riddle ( 2005 ). Introduction to social network methods Riverside, CA: University of California, Riverside (published in digital form at http://faculty.ucr.edu/~hanneman/ ) Hargaadon, A. and R. Sutton (1997). Technolo gy Brokering and Innovation in a Product Development Firm. Administrative Science Quarterly, 42 716 749, Hardin, R. (1993). The Street Level Epistemology of T rust. Politics and Society, 21 505 529. Henry, Adam, Mark Lubell and Michael McCoy (2012). Survey Based Measurement of Public Management and Policy Networks Journal of Policy Analysis and Management, 31(2), 432 452. Henry, Adam, M ark Lubell and M ichael McCoy (2010). Belief Systems and Social Capital as Drivers of Policy Network Structure: The Case of California Regional Planning. Journal of Public Administration Research Theory 21: 419 444. Himmelman, A. (1992). Communities Working Collaboratively for a Change Humphrey Institute of Public Affairs, University of Minnesota, Minneapolis. 74pp. Holling, C. S. (1978). Adapti ve E nvironmental A ssessment and M anagement New York: John Wiley. Ingold, Karen (2011). Network Structures Within Policy Proc esses: Coalitions, Power, and Brokerage in Swiss Climate Policy. Policy Studies Journal, 39(3), 435 459. Ibarra, H. (1993). Network Centrality, Power, and Innovation Involvement: Determinants of Technical and Administrative Roles. The Academy of Manageme nt Journal, 36 (3), 471 501 Imperial, M.T. (2005). Using Collaboration as a G overnance S trategy: Lessons From Six W atershed Management P rograms. Administration and Society, 37(3) 281 320 Innes, J. E. and David E. Booher (2010). Planning with Complexity: An Introduction to Collaborative Rationality for Public Policy New York: Routledge. Isett, K., Ines Mergel, Kelly LeRoux, Pamela Mischen and Karl Rethemeyer (2011). Networks in Public Administration Scholarship: Understanding Where We Are an d Where We Need to Go. Journal of Public Administration Research and Theory, 21, 157 173. Jones, C., Hesterly,W. S. and S.P. Borgatti (1997). A General Theory of Network G overnance: Exchange Conditions and Social M echanisms. Academy of Management Review 22 911 995.
159 Kaiser, M. and S. Stead (2002). Uncertainties and Values in European Aquaculture: Aquaculture International, 10(6), 469 490. Katranidis, S., E. Nitsi, a nd A. Vakrou (2003). Social Acceptability of Aquaculture Development in Coastal Areas: The Case of Two Greek Islands. Coastal Management, 31 (1):37 53. Keen, M, V.A. Brown and R. Dyball (2005). Social Learning in Environmental Management: Towards a Sustainable Future. London: Earthscan. Kramer, R. (1999). Trust and D istrust i n Organizations: Emerging Perspectives, Enduring Q uestions. Annual Review of Psychology, 50 569 598. Knoke, D. and R. Burt (1983). Prominence. In R. S. Burt & M. Miner (Eds.), Applied network analysis: A methodological introduction (pp. 195_ / 222). Beverly Hills, CA: Sage. Knoke, D., F. Pappi, J. Broadbent, and Y. Tsujinaka (1996). Comparing Policy Networks: Labor Politics in the US, Germany, and Japan. New York: Cambridge University Press. excerpt taken from on line source. Knoke, D. (1998). Who Steals My Purse Steals Trash: The Structure of Organizational Influence Reputation. Journal of Theoretical Politic s, 10(4) 507 530 Leach,W. D., Pelkey, N.W., and Sabatier, P. A. (2002). Stakeholder partnerships as c ollaborative policymaking: Evaluation criteria applied to watershed management in California and W ashington. Journal of Public Policy Analysis and Management 21 645 670. Leach, W.D. and P.A. Sabatier (2005). To Trust an Adversary: Integrating Rational and Psychological Models of C ollaborative Policymaking. American Political Science Review, 99(4), 491 503. Leach, W.D., C. Weible, S. Vince, S. Siddiki and J. Calanni (2013). Fostering Learning Through Collaboration: Knowledge Acquisition and Belief Change in Marine Aquaculture Partnerships. Journal of Public Administration Research and Theory, first published online March 18, 2013 doi:10.1093 /jo part/mut011 Levi, M. and L. Stoker (2000). Political Trust and T rustworthiness. Annual Review of Political Science, 3, 475 507. Levin, D. and R. Cross (2004). The Strength of Weak Ties You Can Trust: The Mediating Role of Trust in Effective Knowledge Transfer. Management Science, 50 (3), 1477 1490. Lord, C., L. Ross, and M. Lepper (1979). Biased Assimilation and Attitude Polarization: The Effects of Prior Theories on Subsequently Considered Evidence. Journal of Personality and Social Psychology 37 2098 2109.
160 Lubell, M. (2007). Familiarity Breeds Trust: Collective Action in a Policy Domain. Journal of Politics, 69, 237 250. Lubell, M. (2004). Social and Institutional Trust in a Policy Context. Submitted to Journal of Politics January 2004. Lynn, L ., C. Heinrich, and C.J Hill (2000). Studying governance and public management: Challenges and prospects. Journal of Public Administration Research and Theory 10 233 261. Maarleveld, M. and C. Dangebegnon (1999). Managing Natural Resources: A So cial Learning Perspective. Agriculture and Human Values, 16 267 280. Maggioni, Elena, Hal Nelson, Daniel A. Mazmanian (2012). Industry Influence in Stakeholder Driven State Climate Planning Efforts. Policy Studies Journal, 40(2), 234 255. Marsden, P.V (1990). Network data and measurement. Annual Review of Sociology 16: 435 463. Matti, Simon and Annica Sandstrom (2011). The Rationale Determining Advocacy Coalitions: Examining Coordination Networks and Corresponding Beliefs. Policy Studies Journal, 39(3 ), 385 410 Mazur, N. and A. Curtis (2006). Risk Perceptions, Aquaculture, and Issues of Trust: Lessons From Australia. Society and Natural Resources, 19, 791 808. McDaniels, T., H. Longstaff, and H. Dowlatabadi (2006). A Value Based Framework for R isk Management Decisions Involving Multiple Scales: A Salmon Aquaculture Example. Environmental Science and Policy, 9 423 438. McPherson, M., L. Smith Lovin, and J.M. Cook (2001). Birds of a Feather: Homophily in Social Networks. Annual of Sociology, 27, 415 444. Monteil Overall, P. (2005). Toward a Theory of Collaboration for Teachers and Librarians. Online access: http://www.ala.org/ala/mgrps/divs/aasl/aaslpubsandjournals/slmrb/slmrcontents/ volume82005/theory.cfm. Murro, Melanie and Paul Jeffrey (2008). A Critical Review of the Theory and Application of Social Learnin g in Participatory Natural Resource Management Processes. Journal of Environmental Planning and Management 51 325 344. National Oceanic and Atmospheric Administration (200 11 a ) http://aquaculture.noaa.gov A ccessed March 9, 2011
161 National Oceanic and Atmospheric Administration (2011 b ). National Oceanic and Atmospheric Administration Draft Aquaculture Policy. http://www.nmfs.noaa.gov/aquaculture/docs/noaadraftaqpolicy.pdf Accessed March 9, 2011. Naylor, R., R. Goldberg, J. Primavera, N. Kautsky, M. Beveridge, J. Clay, C. Folks, J. Lubchenco, H. Mooney, and M. Troell (2000). Effect of Aquaculture on World Fish Supplies. Nature 405 (6790): 1017 1024. Olsson, P., C. Folke, and T. Hahn (2004). Social ecological transformation for ecosystem management: the development of adaptive co management of a wetland landscape in southern Sweeden. Ecology and Society 9 (4): 2. Ostrom, Elinor. (2005). Understanding Institutional Diversity Princeton, NJ: Princeton University Press. O'Toole, Laurence J. ( 1997 ). Treating Networks Seriously: Practical and Research Based Agendas in Public Administration. Public Administr ation Review 57(1), 45 52. Phillips, N., T. Lawrence and C. Hardy (2000). Interorganizational collaboration and the dynamic of institutional fields. Journal of Management Studies 37 (1), 23 43. Pfeffer, J. and G.R. Salancik (1978). The External Control of Organizations: A Resource Dependence Perspective. Stanford, CA: Stanford University Press. Powel, W. (1990). Neither market nor hierarchy: Network forms of organization. Research in Organizational Behavior 12 295 336. Price, D.E. (2004). The Congressional Experience. Third Edition. Boulder, CO: Westview Press. Excerpts taken from internet. Provan, K. G., J.B. Beyer, and C. Kruytbosch (1980). Environmental Linkages and Power in Resource Dependence Relations Between O rganizations. Administrative Science Quarterly, 25 200 225. Putnam, R., R. Leonardi, and R. Nanaetti (1993). Making Democracy Work: Civic Traditions in Modern Italy. Princeton: Princeton University Press. Ragin, C. (1987). The Comparative Method: Moving Beyond Qua litative and Quantitative Strategies. Berkeley: University of California. Renn, O. (2006). Participatory Processes for Designing Environmental Policy. Land Use Policy, 23 34 43. Russo, T. and J. Koesten (2007). Prestige, Centrality and Learning: A Social Network Analysis of an Online Class. Communication Education, 54 :3, 254 261
162 Sabatier, P.A. and H.C. Jenkins Smith (1988). An Advocacy Coalition Model of Policy Change and the Role of Policy Oriented Learning Therein. Policy Sciences, 21 : 129 168. Sabatier, P.A. and H.C. Jenkins Smith (1993). Policy Change and Learning: An Advocacy Coalition Approach. Boulder, CO: Westview Press. Sabatier, P.A. and H.C. Jenkins Smith (199 7 The A dvocacy Coalition framework: An Assessment In Theories of the Policy Process, ed. P.A. Sabatier, Boulder, CO: Westview Press, 117 166. Schneider, M., L. Scholz, M. Lubell, D. Mindruta, and M. Edwardsen (2003). Building American Journal of Political Science, 47(1) : 143 158. Scott, J. (2000). Social network analysis: A handbook. Second ed. Newberry Park: Sage Publications. Scholz, J. T. and M Lubell (1998). Trust and Taxpaying: Testing the Heuristic Approach to Collective Action. American Journal of Political Science, 42 398 417, Simon, H.A. (1955). A Behavioral Model of Rational Choice. The Quarterly Journal of Economics 69(1); 99 118 Singleton, R. and B. Straits (2004). Approaches to Social Research Fourth Ed. Oxford: Oxford University Press. Thompkins, E.L. and W.N. Adger ( 2004). Does adaptive management of natural resources enhance resilience to climatic change? Ecology and Society 9(2): 10. Thompson, Ann Marie and J.L. Perry (2006). Collaboration Processes: Inside the Black Box. Public Administration Review (Special Edition December) 20 32 Responsible Marine Aquaculture CAB International, Stickney, R.R. and J.P. McVey (eds). Wallingford: UK. Tsai, W. and S. Ghoshal (1998). Social Capital and Value Creation: The Role of Intrafirm Networks. Academy of Management Journal, 41 464 478. Ulrich,D; Barney,J.B. (1984). Resource Depende nce, Efficiency, and Population. The Academy of Management Revie w, 9(3), 471 481. Webler, T., H. Kastenholz, and O. Renn (1995). Public Participation in Impact Assessment: A Social Learning Perspective. Environmental Impact Assessment Review, 1 (5), 443 463. Weber, M. (1947). Theory of social and economic organizations. New York: Free Press.
163 Weible, C. and P. Sabatier (2005). Comparing Policy Networks: Marine Protected Areas in California. Policy Studies Journal, 33 (2), 181 201. Wirth, F. and J. E. Luzar (1999). Environmental Management of the U.S. Aquaculture Industry: Insights from a N ational Survey. Society and Natural Resources, 12 659 672. Wood, D. J., and Gray, B. (1991). Toward a comprehensive theory of collaboration. Journal of Applied Behavioral Science 27 (2), 139 162. Woodhill, A.J. (2004). Dialogue and transboundary water resources management: towards a framework for facilitating social learning. In S. Langaas and J.G. Timmerman, eds. The Role and Use of Information in European Transboundary River Basin Management Lo ndon: IWA Publishing.