THE CONTINUED EVOLUTION OF URBAN REGIME THEORY: THE STRUCTURAL COMPOSITION OF DECISION-MAKING NETWORKS AS A DETERMINANT OF POLICY USE By Twyla Takeya Blackmond A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of Political Science  Doctor of Philosophy 2014 ABSTRACT THE CONTINUED EVOLUTION OF URBAN REGIME THEORY: THE STRUCTURAL COMPOSITION OF DECISION-MAKING NETWORKS AS A DETERMINANT OF POLICY USE By Twyla Takeya Blackmond Urban regime theory has been the predominant approach used to describe how the membership structure of collaborative decision-making networks directly impact how the network behaves, particularly in selecting policies. However, it provides a limited explanation for how the membership structure of the collaborative matters for policy. I argue that there are three network-level properties (total number of central actors, actor homogeneity, and network size) that shape the context under which individual actors are capable of exerting influence and directing development policy toward their own goals. The total number of actors measure is applied to examine the distribution of power and influence. The level of actor homogeneity and network size indicate the range of organizations and policy interests that are present within the network and must be overcome by central actors. The purpose of the dissertation is to move beyond descriptions of “who governs” by establishing a conceptual framework that offers a more complete explanation of the relationship between network structure and behavior. By developing more detailed descriptions of the governance structures, scholars can formulate more solid and testable predictions about the policies used by structurally distinct decision-making networks. This framework was tested using a sample of small to large American cities (n = 523) that responded to the 2004 International City/County Management Association’s Economic Development Survey. A three-part quantitative analysis was conducted. Hierarchical cluster analysis was employed to empirically identify structurally distinct local economic development decision-making networks. Descriptive statistical tests were used to construct structural profiles for each decision-making network. Based on the composition of the networks, hypotheses regarding policy use were developed. Several binomial logistic regression models were estimated to test the effect of network structure on the likelihood of cities using several different development policies. This statistical method was also utilized to explore the effect of several social, economic, political, and geographic variables on the type of decision-making network that exists in a city. Three structurally distinct local economic development decision-making networks were identified: municipal, local government-chambers of commerce, and broad public-private partnerships. Decision-making network type was a significant predictor of how likely cities were to use several different policies: community development corporations, microenterprise programs, nongovernmental organization partnerships, other local government partnerships, tax abatements, and one-stop permit issuance. To: Mom, Gregory and Zaria iv ACKNOWLEDGEMENTS I have been blessed with many people who have supported me in this endeavor. I will forever cherish all of the encouragement, assistance, and advocacy that I have received from my family, friends, advisors, and peers. No matter what, I can always find solace in the fact that there are many people who believe that I am more than capable of handling all that comes with this profession. I look forward to taking advantage of every opportunity that allows me to show that I was worth it. v TABLE OF CONTENTS LIST OF TABLES vii LIST OF FIGURES ix CHAPTER 1: INTRODUCTION AND LITERATURE REVIEW 1.1 Governance Theory: Public-Private Collaboration in Decision-Making 1.2 Urban Regime Theory: Gaining Access to the Decision-Making Process 1.3 A Policy Network Approach to Understanding Decision-Making in Governance Structures 1.4 Is a Regime a Network? Is a Network a Regime? 1.5 Conclusion 1 3 6 11 16 18 CHAPTER 2: CONCEPTUAL FRAMEWORK 2.1 Central Actors 2.2 Measuring the Distribution of Power: Total Number of Central Actors 2.3 Establishing the Diversity of Policy Goals: Homogeneity of Actors and Size 2.4 Conclusion 20 22 23 24 27 CHAPTER 3: DATA AND METHODOLOGY 3.1 The Complexity of Local Economic Development Decision-Making Networks 3.2 Data 3.3 Description of the Dataset 3.4 Methodology 3.5 Conclusion 37 38 41 43 51 83 CHAPTER 4: A TYPOOGY OF LOCAL ECONOMIC DEVELOPMENT DECISIONMAKING NETWORKS 84 4.1 Findings 85 4.2 Local Government-Chamber of Commerce Decision-Making Networks 86 4.3 Municipal Decision-Making Networks 91 4.4 Broad Public/Private Partnership Decision-Making Networks 95 4.5 Predicted Policy Profiles of Local Economic Decision-Making Networks 100 4.6 Conclusion 115 CHAPTER 5: THE MEMBERSHIP STRUCTURE OF DECISION-MAKING NETWORKS AS A DETERMINANT OF POLICY USE 5.1 Identifying the Dependent Variable 5.2 Decision-making Networks and Community Development Policies 5.3 Decision-making Networks and Small Business Development Policies 5.4 Decision-making Networks and Business Retention Policies 5.5 Decision-making Networks and Business Incentives Policies 5.6 Policy Profiles of Local Economic Development Decision-Making Network vi 117 119 121 128 133 140 149 5.7 Conclusion 158 CHAPTER 6: AN EXPLORATORY ANALYSIS OF FACTORS AFFECTING DECISION-MAKING NETWORK TYPE 6.1 Municipal Decision-Making Networks 6.2 Broad Public-Private Partnership Decision-Making Networks 6.3 Local Government-Chambers of Commerce Decision-Making Networks 6.4 Analysis 6.5 Conclusion 164 166 169 171 174 177 CHAPTER 7: CONCLUSION 7.1 The Underperformance of the Decision-Making Network Variable 7.2 Future Research 7.3 Conclusion 180 187 189 190 APPENDIX 191 BIBLIOGRAPHY 207 vii LIST OF TABLES Table 3.1 Independent T-Test Results: Descriptive Statistics 49 Table 3.2 Independent T-Test Results: Inferential Statistics 50 Table 3.3 Local Economic Development Actors 53 Table 3.4 Types of Local Economic Development Actors 65 Table 3.5 Local Economic Development Policies 69 Table 3.6 Description of Independent and Control Variables 75 Table 3.7 Descriptive Statistics of Variables Included in the Logistic Regression 77 Table 3.8 Description of Dependent and Independent Variables 81 Table 3.9 Descriptive Statistics of Variables Included in the Logistic Regression 82 Table 4.1 Local Government-Chamber of Commerce Networks: Central Actors 87 Table 4.2 Local Government-Chamber of Commerce Networks: Network Size 90 Table 4.3 Municipal Networks: Central Actors 92 Table 4.4 Municipal Networks: Network Size 94 Table 4.5 Broad Public-Private Partnership Networks: Central Actors 96 Table 4.6 Broad Public-Private Partnership Networks: Network Size 99 Table 4.7 Summary of Decision-Making Networks Membership Structure 103 Table 5.1 Summary of Estimated Community Development Corporations Models 122 Table 5.2 Estimated Logistic Regression Model of Community Development Corporations 123 Table 5.3 Summary of Estimated Microenterprise Programs Models 128 Table 5.4 Estimated Logistic Regression Model of Microenterprise Programs 129 Table 5.5 Model Summary: Partnering with Nongovernmental Organizations Models 134 viii Table 5.6 Estimated Logistic Regression Model of Nongovernmental Organizations 135 Table 5.7 Model Summary: Partnering with Local Governments Models 137 Table 5.8 Estimated Logistic Regression Model of Partnering with Local Governments 138 Table 5.9 Model Summary: Tax Abatements 141 Table 5.10 Estimated Logistic Regression Model of Tax Abatements 142 Table 5.11 Model Summary: One-Stop Permit Issuance 145 Table 5.12 Estimated Logistic Regression Model of One-Stop Permit Issuance 146 Table 5.13 Decision-Making Network Areas Using Local Economic Development Policies 149 Table 6.1 Description of Dependent and Independent Variables 165 Table 6.2 Model Summary: Municipal Networks 167 Table 6.3 Estimated Logistic Regression Model of Municipal Networks 168 Table 6.4 Model Summary: Broad Public Private Networks Network 169 Table 6.5 Estimated Binomial Logistic Regression Model of Broad Public Private Networks 170 Table 6.6 Model Summary: Local Government-Chambers of Commerce Networks 172 Table 6.7 Estimated Binomial Logistic Regression Model of Local Government-Chambers of Commerce Networks 173 Table 7.1 Summary of Decision-Making Network Policy Use and Presence in a Community 183 Table A3.1 List of Cities Included in the Dataset 191 Table A3.2 Participation Rates of Development Actors 197 Table A3.3 General Trends in the Use of Development Policies 198 Table A4.1: Partial Agglomeration Schedule from Hierarchical Cluster Analysis 200 ix LIST OF FIGURES Figure 2.1 Conceptual Framework: Policy Network Approach to Urban Regime Theory 21 Figure 3.1 Population Sizes of Areas Included in Sample 44 Figure 3.2 Geographic Areas Represented in Sample 45 Figure 3.3 Regional Areas Represented in Sample 46 Figure 3.4 General Trends in the Participation Rates of Specific Development Actors 54 Figure 3.5 Total Number of Actors in Development Decision-Making Networks 66 Figure 7.1 Conceptual Framework: Policy Network Approach to Urban Regime Theory 181 Figure A3.1 ICMA 2004 Economic Development Survey 201 x CHAPTER 1: INTRODUCTION AND LITERATURE REVIEW Over the years, several influential scholars have developed compelling theories and extensive analyses that examine the importance of power within policy-making. Regime theorists have demonstrated that there are many different “governing coalitions” composed of public and private actors in possession of various resources and differing in their policy interests. When present, these regimes direct local economic development decision-making in American cities. This theoretical approach emphasizes factors that impact development actors’ ability to gain access to the governing coalition and, potentially, influence policy choices. More specifically, regime theorists focus on the impact of the most resourceful, and thus powerful, actors, referred to as central actors, and formulate different typologies of potential and existing urban regimes. Regime theorists make the claim that the membership of the governing coalition has a substantial impact on the outcomes of the decision-making process. Critics argue that urban regime theory does not provide a full explanation of how the membership structure of the regime influences regime behavior, specifically policy decision-making. This limitation diminishes the capacity of regime theory to develop solid predictions about policy choices across structurally distinct decision-making regimes and networks. These critics have labeled the regime approach as an “analytical toolbox” rather than a theory. Thus, the question remains: In what way, exactly, do governing coalitions matter for policy? I argue that a complete discussion of local governance must extend beyond regime theory’s concentration on “who governs” (Dahl 1961) and acknowledge that the organizational structure of a network’s membership also offers valuable information for understanding the behavior of the regime. The central research question of this study is: How do the structural features of a decision-making network’s membership shape local governments’ policy use? 1 Although the central actors and their policy preferences are important for decision-making, network-level attributes are important determinants of network outcomes. The structure of the membership directly impacts the policy choices made by the network. I draw upon policy network concepts to expand upon regime theory and formulate a conceptual framework that strengthens scholars’ capacity to develop solid hypotheses about governing coalitions’ policy decisions. I contend that there are three structural characteristics (total number of central actors, actor homogeneity, and network size) that mediate the ability of central actors to exploit their resources. The complexity of the decision-making environment may either strengthen or weaken development actors’ ability to influence the decision-making process and sway local governments to promote specific policies. This conceptual framework extends the analysis to provide more in-depth descriptions of these governance structures, which strengthens predictions regarding how governance is related to public policy outcomes. The structuralist perspective of the network theory offers a remedy to the limitation present in urban regime theory. By expanding urban regime theory, this framework permits the simultaneous consideration of the role of individual actors, particularly their policy preferences, and network-level characteristics, which improves predictions regarding the types of policies that will emerge from structurally distinct decision-making networks. This improves regime theory’s capacity to account for how the membership of governing coalitions brings about certain policies and to develop testable as well as generalizable hypotheses about network activity. The purpose of this dissertation is to determine whether structurally distinct decisionmaking networks govern society differently. More specifically, does the membership structure of the decision-making network influence the type of policies that cities adopt? The research 2 presented in this dissertation includes three quantitative analyses focused on identifying the independent variable (decision-making network type), testing the relationship between decisionmaking network type and policy use, and exploring the factors which influence the membership structure of decision-making networks. Several steps are taken in this chapter to defend arguments that urban regime theory does not adequately explain how the membership of governing coalitions influences policy choice, and that urban regime theory would be strengthened by examining organizational structure. First, I review the foundational text of governance theory and discuss the basic characteristics of governance systems. I then explain how urban regime theory expands upon governance theory. I argue that urban regime theory provides an account for regime formation. As such, its power to explain the process of decision-making is diminished. Network theory, on the other hand, provides a governance-based explanation for how network members interact to shape policy. Urban regime theory and network theory are conceptually similar but examine different stages of the policy process. This supports the merging of urban regime theory with network-based ideas. In the last section of the chapter, I introduce a conceptual framework that considers individual actors involved as well as the structural organization of the decision-making network when predicting cities’ policy decisions. 1.1 Governance Theory: Public-Private Collaboration in Decision-Making. At various levels, the governing process has expanded beyond the institution of local government. It has been transformed into a broader horizontal system of collaboration (Peters and Pierre 1998; Rhodes 1996; Stoker 1998). This contemporary governing system, governance, involves public and private actors as influential participants in the policy-making process (Peters 3 and Pierre 1998; Rhodes 1996; Stoker 1998). Rather than focusing on the functioning of public officials and departments, the governance approach emphasizes the operations of a network of actors participating in the process of governing (Peters and Pierre 1998; Rhodes 1996, 2007; 1 Salamon 2002; Stoker 1998). This expansion of the governing sphere represented a shift from competition between public and private actors to collaboration (Salamon 2002). Development of these partnerships obscured the responsibilities and limitations of governmental and private actors within decision-making, as well as a blending of their resources (Peters and Pierre 1998; Salamon 2002; Stoker 1998). As a result, non-state actors have increasingly assumed roles traditionally performed by governmental departments. Theorists make it clear, however, that governance extends upon local government as a system of governing but does not replace it. The emergence of nongovernmental actors as influential decision-makers did not dismantle the institution of government. Kettl (2000) states that governance represents governing systems that are “both horizontal—in search of service coordination and integration with nongovernmental partners in service provision and vertical— through both traditional hierarchical bureaucracies and multilayered federalism. It is not so much that the horizontal relationships have supplanted the vertical ones; rather, the horizontal links have been layered on top of the vertical ones" (494). Therefore, state agencies and departments at various levels continued to supervise and execute formal governmental and public administration duties while creating and maintaining interorganizational linkages with private actors for collective action purposes. The expansion of the governing sphere is the direct result of various actors’ increased need to cooperate in order to achieve shared objectives (Rhodes 1996; Salamon 2002). 1 Stoker (1998) also refers to these governance systems as regimes (see page 23). 4 Interdependence between organizations stimulated the development of governance networks by motivating public and private actors to engage in collaborative governing. Governmental institutions and private actors increase their capacity to address a policy issue by combining their resources and skills (Rhodes 1996; Salamon 2002; Stoker 1998; Peters and Pierre 1998). In sum, governance networks are, essentially, collections of public and private actors connected by joint involvement in the act of governing as a result of resource interdependencies. As a result of the mutual dependence, all development actors involved in the governing 2 process lose some degree of autonomy (Saidel 1991). As government grows increasingly dependent on private actors, they relinquish power that was traditionally possessed by the state, and the private entities gain influence within the policymaking process. Stemming from the relationship between dependency and power, governance networks are self-organizing and the members are autonomous (Rhodes 1996, 2007; Stoker 1998). Actors choose to participate in these networks in an effort to gain access to critical resources, while the interdependence of actors disperses some power to all network members. Despite having access to a variety of necessary resources via collaboration, there is no one actor fully responsible for ensuring a governing system that is accountable and productive. The government acquires critical resources at the cost of its direct and complete authority over 2 In a groundbreaking article, Richard Emerson (1962) describes an inverse relationship between power and dependence. He argues that dependence on another party gives it power over the other actor. Emerson’s (1962) theory of power dependence states that “the power of A over B is equal to, and based upon, the dependence of B upon A” (33). 5 3 governmental programs (Rhodes 1996, 2007; Salamon 2002) . Thus, government focuses primarily on influencing, rather than controlling, the governing process. Government functions to guide the networks’ activities (Rhodes 1997; Stoker 1998) as well as to identify policy goals (Peters and Pierre 1998). In sum, governance refers to the process of governing that involves networks (Rhodes 2007). Public and private actors join these governing institutions because of their need for the other actors’ resources in order to successfully address a shared objective (Rhodes 1996, 2007; Salamon 2002; Stoker 1998). The governance approach views these networks, not the formal institution of government, as controlling the governing process and policy (Peters and Pierre 1998; Rhodes 1996, 2007; Salamon 2002; Stoker 1998). Although it is an altered governing system, governance networks and participants are interested in improving the local community in the same manner as government. Urban regime theorists offer substantial insight into the complexity of governance networks by further explaining the process by which decision-making networks are shaped. 1.2 Urban Regime Theory: Gaining Access to the Decision-Making Process As a governance-based theoretical approach, urban regime theory abandons the notion that formal government is sufficient for formulating and implementing policy as well as the idea 3 This makes it more difficult to sustain democratic accountability (Kettl 2000; Rhodes 1996, 2007; Salamon 2002). The non-hierarchical structure of governance networks complicates the act of governing as well as increases the length of the process and the likelihood of conflict (Kettl 2000; Rhodes 1996, 2007). Both, ultimately, reduce the effectiveness and efficiency of the governing process. 6 that economic forces unilaterally shape policy (Mossberger and Stoker 2001; Stone 1989). Regime theory is founded on two central principles. First, nongovernmental entities are necessary for policy-making. Second, politics are also important for policy decisions. This is because the membership structure of the governing coalition strongly influences the policies that cities employ (Stone 1989). With regard to the first principle, regime theorists contend that the division of labor between the state and market found in democratic capitalist societies determines the context in which local economic development policy-making transpires (Elkin 1987; Stone 1989). Local governments have a desperate need to stimulate development in order to increase revenues, yet they lack an abundance of market-related resources which diminishes the ability of local officials to effectively govern (Elkin 1987; Orr and Stoker 1994; Stone 1989, 1993). Both local government and private actors control resources necessary for successful governing (Agranoff and McGuire 1998; Clarke and Gaile 1992; Davies 2002; Eisinger 1988; Elkin 1987; Fainstein and Fainstein 1989; McGuire 2000; Stoker 1995; Stone 1989). Local officials possess the formal political power essential for policymaking, but private development actors manage an extensive stock of political resources (specifically campaign resources), as well as the market resources needed to pursue economic growth (Elkin 1987; Stone 1989). This drives local political officials to combine resources and cooperate with private actors in possession of relevant economic resources (Elkin 1987; Stone 1989; Stoker 1995). Public officials and private actors are linked by their mutual dependence for resources and policy, which creates a fragmentation of power. As a result, nongovernmental agents achieve access to the local economic development decisionmaking process. 7 These informal public-private partnerships are referred to as governing coalitions or urban regimes and work to formulate and execute policy decisions (Keating 1991; Stone 1989). According to regime theory, access to the decision-making process is contingent upon private actors’ display of two forms of power: systemic and preemptive. The former comes from nongovernmental actors’ control of a large cache of market resources. Stone states that this element of power is derived from “durable features of the socioeconomic system [and] confers advantages and disadvantages on groups in ways [that] predispos[e] public officials to favor some interests at the expense of others” (Stone 1980, 978). Governmental officials are more inclined to promote the policy ideas of private actors that exhibit higher levels of systemic power because they have increased access to critical resources. It is also very important that these private actors are capable of actually exercising the power bestowed upon them by their control of valuable resources (Elkin 1987; Stone 1989, 1993). Preemptive power is a more “intentional and active” (Stoker 1995, 64) form of power that stems from systemic power (Stone 1989). Members of the governing coalition must have the ability to exploit these resources in order to further improve the regime’s capacity to govern (Stone 1989). The central claim of urban regime theory is that the composition of local economic development governing coalitions is crucial to the policy process and has significant implications for which policies cities adopt. Regimes pursue a distinctive policy agenda that is largely determined by the membership (DiGaetano and Klemanski 1993a; Mossberger and Stoker 2001; Stone and Sanders 1987; Stoker 1995; Stone 1993, 2004, 2007). Given that power is distributed within the context of limited resources, the most resourceful actors also tend to be the most influential. As stated by Stone and Sanders (1987), “it matters greatly who is represented in 8 decision-making,” because the organizational structure of a regime’s membership directly impacts policy decisions (167). The urban regime approach, however, does not fully describe how the politics within the governing coalitions influence the policies that cities employ. Davies (2002) points out that regime theory was not intended to explain the policy-making process. Instead, this approach analyzes the formation, maintenance, disintegration, and characteristics of urban regimes (Davies 2002; Lauria 1999). The purpose of the urban regime approach is to analyze the regime, which explains regime theorists’ preoccupation with the identification of various actual and hypothetical regime types. This theoretical approach centers largely around the outputs of the structuring process (formation and evolution of regimes) rather than the outputs of the structure (policy decisions). Regime theorists strongly contend that a significant relationship exists between regime type and policy use, but it does not clearly explain why structurally distinct governing coalitions are expected to differ in their behavior, specifically policy use (Sites 1997). For this reason, urban regime theory is an incomplete theory of policymaking. Urban regime theory needs to be studied as the independent variable instead of as the dependent variable. To be a more effective theory of local policy-making, regime theorists must improve their ability to account for and predict systematic variation in economic development policy across different regime types. This approach centers on individual agency within the regime, but does not fully consider the regime as a structure (Davies 2002; Ward 1997). Regime theorists must also acknowledge that the behavior of individual actors is “structurally inscribed” (Ward 1997, 181). This approach needs to be expanded to offer an improved description of the context in which decision-making occurs and the structural limitations that powerful development agents encounter when attempting to 9 influence policy. By doing so, regime theorists would strengthen their predictions about policy use in cities governed by different regime types. Methodologically, the urban regime approach has been tested primarily through the use of case study analysis in which governance coalitions have been examined in a wide variety of contexts. The literature benefits from detailed examinations of the existence, structure, and particularly, processes of urban regimes (Dowding 2001; Ward 1996). By identifying strong similarities in the formation and processes of urban regimes in different cities, scholars are able to formulate generalizations about policy use. Any variations can be explained by the social, economic, and political conditions of the specific case. On the other hand, the sole dependence on this particular research method constrains regime theory by making it more difficult to draw comparisons and combining the findings of studies examined in various cities (Davies 2002; Mossberger and Stoker 2001;). This diminishes the reliability and validity of urban regime findings. Davies (2002) cites this as the primary 4 reason why the explanatory power of urban regime theory is challenged. The qualitative research provides valuable information regarding the functioning of specific regimes. However, regime theory needs to be extended to provide stronger generalizations about the policy outcomes of urban regimes differing in membership structure. A quantitative methodological approach would permit scholars to test hypotheses regarding policy use across a large set of cities that differ socially, economically, and politically. As such, the findings would be applicable to a much broader set of cities and lead to the development of more reliable predictions about the types of policies that structurally distinct regimes would practice. 4 Critics further argue that the localist approach utilized by regime theory severely weakens its applicability to cross-national studies (Dowding 2001; Stoker 1995; Ward 1996). 10 The purpose of this dissertation is to improve urban regime theory by providing a more complete theoretical explanation for how the membership of the regime matters for policy, and by utilizing large-N statistical analyses to develop solid generalizations for policy variations across regime types. Addressing these two significant limitations of urban regime theory would further contribute to unraveling the increased complexities of local economic development governance, by resolving the uncertainties around the functioning of these political institutions and the subsequent policy implications. I argue that network theory provides considerable clarity in the areas on which urban regime theory becomes vague. By assessing the structure of the membership, rather than the membership itself, urban regime theory may be strengthened. 1.3 A Policy Network Approach to Understanding Decision-Making in Governance Structures The body of literature examining networks is expansive and extends across different areas of study, including organization theory, interorganizational management, sociology, political science, public administration, and public policy. For this dissertation, I focus specifically on the discussion of policy networks within political science, public policy, and 5 public administration, as well as the interdisciplinary use of social network analysis. Generally, 5 Social network analysis (SNA) is a formal and quantitative approach to examining the structural relationships and processes within the network (Boragtti, Mehra, and Labiance 2009; Wasserman and Faust 1994) and employs matrix algebra as well as graph theory to depict the relationships amongst network members (Boragtti, Mehra, and Labiance 2009; Mardsen 1990). In sum, mathematical algorithms establish whether and to what extent actors are related in a variety of social contexts (Wasserman and Faust 1994). Classic studies have applied SNA to community elite decision-making (Laumann, Mardsen and Galaskiewicz 1977; Laumann and 11 there are two approaches fto examining policy networks: the network analytical approach (“interest intermediation”) and network as a form of governance (“governance”). In the network analytical approach, the term policy “network” is all-encompassing term applied to describe all instances in which public and private entities partner to develop and conduct policy activities (Adam and Kriesi 2007; Borzel 1997; Carlsson 2000; Dowding 1995; Pappi and Henning 1998; Provan and Kenis 2007; Sandstrom and Carlsson 2008). Studies that apply this perspective focus primarily on describing and comparing the structural characteristics (Provan and Kenis 2007; Wasserman and Faust 1994) of potential and existing policy networks. This led to the development of various typological classification schemes for policy networks based on different structural attributes within the relational arrangements. I, however, focus on the governance school approach, which centers on understanding the impact of network structure on network outcomes, specifically policy choice and change. Within this school of thought, the unit of analysis is the network and scholars concentrate on explaining how the network is organized and the patterns of joint-policy making (Borzel 1997; Provan and Kenis 2007). As a governance-based theoretical approach, policy network theory centers around two main contentions. First, individual actors do not determine policy. Rather, the selection of policy is the result of a process that involves multiple agents (Bressers and O’Toole 1998). Second, the membership structure of the policy network is indicative of how power and resources are dispersed throughout the decision-making environment (Borzel 1997). This influences the decision-making process by shaping the policy decisions that are considered and materialize (Borzel 1997; Carlsson 2000; Sandstrom and Carlsson 2008). In sum, network Pappi 1973), social influence (Marsden and Friedkin 1994), and power (Brass & Burkhardt 1993). 12 structure directly impacts network behavior (Knoke 1990; Marsh and Smith 2000), particularly the policy choices made. Policy networks are political institutions that symbolize persistent relations via collaborative efforts amongst individuals and organizations (Ansell 2008). Networks are characterized by two fundamental components: actors (nodes) and linkages (ties) (Kenis and Schneider 1991; Knoke 1990; Laumann and Knoke 1987; Provan, Fish and Sydow 2007; Wasserman and Faust 1994). The linkages refer to structural ties between actors that represent the presence of some formal or informal connection (Brass et al. 2004; Knoke 1990; Wasserman and Faust 1994). Analysis of networks involves identifying a set of involved actors and their attributes, including policy preferences, as well as collecting data on the relationships that exist between actors (Ansell 2008; Wasserman and Faust 1994). 6 The network approach is primarily concerned with how the pattern of interactions, associations, and affiliations amongst actors influences the networks’ behavior. The structure represents different advantages and disadvantages facing individual actors in their pursuit of impacting policy (Wasserman and Faust 1994). Outcomes are interpreted as an artifact of the network’s membership and of the structural limitations imposed on individual actors. This ensures that the individual action is understood contextually and within the constraints of the network structure (Ansell 2008). For policy network analysis, the primary concerns are the 6 In an effort to govern via collective action, actors develop connections to one another that center around communication and transmitting policy resources, including finances, information, expertise, services, trust, and social support (Kenis and Schneider 1991; Galaskiewicz and Wasserman 1994; Wasserman and Faust 1994). The linkages between actors differ in frequency and intensity (Borzel 1997). 13 position and role that various actors occupy within the network, and not the specific agent or organization involved (Galaskiewicz and Wasserman 1994; Laumann and Knoke 1987; Provan, Fish and Sydow 2007; Wasserman and Faust 1994). While considerable access to strategic resources is critical for access to the decision-making process, the structure of the network may either restrict or enhance an actor’s ability to influence the network’s behavior and shape policy decisions in its favor. Social network analysis studies have supplied the quantitative tools needed for classifying, measuring, and comparing sets of relational linkages and their structural properties (Kenis and Schneider 1991). Network theorists focus largely on characterizing the decision-making structure and the relationships between participating actors. Network theory, however, does not provide an adequate description of how individual network members will act within the decision-making environment (Ansell 2008; Borzel 1997; Peters and Pierre 1998). This theoretical approach identifies where each actor is positioned within the network and the structural obstacles it may face, but it cannot predict what individual actors will do and the policy choices they will favor. Without this information, it is difficult to develop solid predictions about which policies will be adopted by the network. For this reason, network theory has been consistently criticized by scholars in various research areas and designated as an “analytical toolbox” rather than as a complete theory (Adam and Kriesi 2007; Borzel 1997; Carlsson 2000; Dowding 1995; Kenis and Schneider 1991; Thatcher 1998; Wasserman and Faust 1999). Network theory lacks a framework that would extend the explanatory power beyond merely describing the context in which collective activities occur to actually providing testable theories that account for how network characteristics influence policy outcomes (Borzel 1997; Carlsson 2000). Critics warn that network theorists could potentially be developing the understanding of structural properties 14 that may not be strongly or systematically related to behavior and outcomes (Dowding 1995; Thatcher 1998; Sandstrom and Carlsson 2008). To expand on the metaphorical usage of the concept and advance toward a coherent theory that explains network outcomes, scholars often introduce other theories to provide the analytical support that network theory requires (Adam and Kriesi 2007; Carlsson 2000; Thatcher 1998). The inability of network theorists to account for variations in individual actors behavior, specifically policy goals, diminishes their capacity to develop solid expectations about behavioral differences across structurally distinct policy networks and explain variations in policy outcomes (Carlsson 2000; Sandstrom and Carlsson 2008). If the structural conditions of the network impact which members are more successful than others in promoting their interests, then there are considerable implications for variations in policy formation and implementation across network types. Hypotheses regarding the network’s policy choices cannot be made without understanding the policy goals of the most influential actors. For this reason, the network approach needs to consider the characteristics of involved actors, including organizational structure, goals, and resources. In sum, network theory is a useful descriptive tool and offers substantial insight into the functioning of the interorganizational collaboratives that govern the policy-making process. However, it becomes even more valuable when it is employed as an explanatory device to "better connect particular configurations of policy networks to policy dynamics" (Adam and Kriesi 2007, 237). Network theorists must demonstrate that policy networks are not only functioning features of our society, but important determinants of the policy choices that networks adopt (Borzel 1997). If partnered with an appropriate framework, the network approach could yield 15 extensive advances in the development of hypotheses regarding the policy variations across policy networks with distinct membership structures. While network theory describes the context in which policy-making occurs, urban regime theory describes the organizational resources and policy goals of network members. Individually, they suffer from theoretical limitations that prevent them from fully explaining governance systems and their effects on policy choices. Each theoretical approach offers a remedy to the weakness present in the other. I argue that urban regime theory would be greatly strengthened as a theory of policy-making if it also considered the structure of the governance network. It is clear that the different theoretical approaches do not challenge the arguments presented by the other (Adam and Kriesi 2007) and are suitable for comparison and incorporation. 1.4 Is a Regime a Network? Is a Network a Regime? Many frameworks for understanding interorganizational political and policy decisionmaking have evolved from governance theory. The precursors to urban regime and network theory include sub-governments, issue networks, policy communities, iron triangles, and 7 community power studies. The central argument of these theoretical frameworks is that resource interdependence compels various actors to collaborate in various policy areas at different levels of government (Ansell 2008; Melbeck 1998). Of particular interest to these 7 Regime theory, specifically, is strongly rooted in the community power approach and emphasizes joint policy-making amongst public and private, particularly economic, actors (Melbeck 1998; Stone 1989). 16 frameworks is the influence of the involved actors and various policy interests on the policy decision-making process. Urban regime theory and network theory have strong foundations in the same framework and present similar core arguments. Mossberger and Stoker (2001) are even more explicit in arguing that there are strong ties between these approaches. They state that "…if regimes are simply coalitions that bring together actors in a complex policy environment where the division between market and state is not a factor, then how do urban regimes differ from networks" (Mossberger and Stoker 2001, 825)? The implication is that urban regime theory and network 8 theory are, essentially, the same concept with different labels (Melbeck 1998). Theorists have argued that a policy network is a broad concept with many subcategories (Carlsson 2000; Mossberger and Stoker 2001; Rhodes and Marsh 1992; Sandstrom and Carlsson 2008). All instances of joint policy-making whether it involves public-private partnerships or not, are classified as some subcategory of policy networks (Mossberger and Stoker 2001; Rhodes and Marsh 1992). This includes urban regimes (Agranoff and McGuire 1998; Mossberger and Stoker 2001). Melbeck (1998) lists the four fundamental properties that both approaches share. First, they gauge power in decision-making according to actors’ access to productive resources in a system of mutual resource dependency. Second, the level of analysis includes all actors in the 8 Melbeck states that “…the structural characteristics leading to the examination of policy networks at the national level have been present at the local level much longer in the USA. Seen in this way, empirical policy-network studies have been carried out in the USA for a long time, though usually under a different name, for example “power elites” or “community-power” studies ” (532). 17 entire policy area. More specifically, the setting of decision-making is broader than those alliances organized by strong and natural shared interests and goals. Research involves every actor that has the capacity to act and shares a general interest in governance. Third, both urban regime theory and network theory focus on how resources are exchanged within the system of actors. Identifying those actors contributing to the pooling of resources and decision-making is important for understanding the distribution of power within the system. Lastly, the central argument of both frameworks is that the composition of the network has substantial implications for policy outcomes. The difference between urban regime theory and network theory is not a matter of fundamental conceptual differences. Rather, these concepts focus on different aspects of the decision-making network and have been employed to discuss various phases of the governing process. Regime theory focuses on unveiling the manner by which individuals and organizations exploit their resources to gain power and influence in the decision-making process; network theory focuses on the resulting set of actors and studies how they make and change policy. 1.5 Conclusion Although the urban regime approach follows governance theory and claims that the structure of the decision-making network influences how it chooses to address policy issues, urban regime theory does not fully explain how the governing coalition exactly matters for policy. If the membership structure of the decision-making network is a critical predictor of policy decisions, then a complete theory of decision-making must also account for how the structural attributes of these political institutions work to impact the policy decisions of the network. Individual actors and their policy goals as well as the network’s membership structure 18 create the institutional context that predisposes some communities to select some policies and neglect others. The inability of a theory to consider both structure and agency naturally results in severe theoretical limitations (Marsh and Smith 2000). Given urban regime theory’s overemphasis of agency and policy network theory’s structuralist perspective, neither approach is capable of single-handedly explaining how governing structures are directly linked to policy outcomes. Although urban regime theory is quite informative, it does not fully consider how the structural characteristics of the network and relationships amongst actors either constrain or promote individual actors’ policy objectives. There is considerable potential to further develop urban regime theory’s conceptualization of how governance structures influence policy choices. Analysis of urban regime theory and policy network theory’s theoretical foundations, concepts, and arguments indicate that these approaches share strong similarities but concentrate on different phases of the decision-making process. The adoption of policy network theorists’ structuralist perspective would bolster regime theory’s capacity to predict the policy outcomes of decision-making networks. The conceptual framework presented in this dissertation considers both the membership as well as the structure of the membership when predicting the policy choices of decision-making networks. 19 CHAPTER 2: CONCEPTUAL FRAMEWORK The purpose of this dissertation is to further clarify the process of governance and uncover how governing coalitions function to select policy. The central research question examined in this study is: How does the structural composition of decision-making networks influence the policies that communities select? By creating a balance between the actor-centered framework (urban regime theory) and the structure-centered framework (network theory), I construct a framework that: provides a more complex depiction of the internal dynamics within decision-making networks, and supports the development of testable hypotheses regarding the policy choices. In sum, public and private actors exploit strategic resources to gain access to the decision-making process, but the ability of each actor to influence decision-making is either enhanced or diminished by the organizational structure of the governance network. To strengthen urban regime theory as a theory of policy-making, I integrated some structural constructs into the framework. These constructs provide more detail about the structural configurations of networks, which improves the ability of scholars to distinguish amongst structurally distinct networks as well as better explain how networks function to create outcomes. In this dissertation, I argue that the ability of the most resourceful, thus powerful, individuals and organizations (central actors) to successfully promote their policy goals and influence the network’s behavior is mediated by:  number of central actors: the total number of central actors within the network  homogeneity: the extent to which network members share similar organizational goals and resources  size: the total number of organizations represented in the decision-making network. 20 These network-level structural properties essentially shape the context under which central actors are capable of exerting influence and directing development policy toward their own preferences. As displayed in Figure 1.1, this conceptual framework analyzes the effect of a single individuallevel construct and three network-level structural constructs to describe the functioning of the decision-making process and to develop hypotheses about the network’s policy decisions. The structural attributes included in this framework are based off similar attributes examined by social network analysts. The “central actors” construct identifies the development actors who are most influential in the decision-making process, while the “total number of central actors” construct quantifies these participants. The diversity of organizational goals within the network is evaluated by analyzing “homogeneity” (construct) of actors involved. The Figure 2.1 Conceptual Framework: Policy Network Approach to Urban Regime Theory “size” construct measures the number of organizations involved in the decision-making process, including the central actors. The latter two structural constructs evaluate the range of policy interests that are being considered. Heterogeneity increases the number of participants making the network more complex, which decreases the capacity of central actors to exploit their 21 resources, affect the network’s behavior, and control policy. In sum, the network-level structural constructs operate to either diminish or advance individual central actors in their pursuit to have the decision-making process reflect their policy preferences. Overall, the simultaneous analysis of the central actors, total number of central actors, heterogeneity, and size leads to a more robust characterization of decision-making networks and the environment in which development actors attempt to influence policy. 2.1 Central Actors Network theorists argue that the position of individual actors within the network represents the actors’ status in relation to other members (Ansell 2008; Brass et al 2004; Brass and Burkhardt 1993; Fowler 2006; Knoke and Yang 2008; Provan, Fish, and Sydow 2007; Rowley 1997; Sandstrom and Carlsson 2008; Wasserman and Faust 1994). According to urban regime theory, the most prominent actors have increased access to strategic policy resources, thereby wielding more influence in the decision-making environment (Stone 1989). Given their position within the decision-making network, central actors have a critical and discernible effect 9 on the overall policy behavior of the network. The activities of the network depend heavily on the involvement of these actors and their resources. Thus, central actors have more potential to influence the policy choices made by the network and shape policy in a manner that reflects their 9 Network analysts examine one individual-level structural construct, actor centrality. Social network analysts have developed several measures to assesses actor centrality, or the network position of an individual actor in relation to the other members (see Ansell 2008; Brass et al 2004; Brass and Burkhardt 1993; Fowler 2006; Provan, Fish and Sydow 2007; Rowley 1997; Wasserman and Faust 1994). 22 organizational goals. As such, the set of potential policy choices to be considered by the network is heavily influenced by the interests of actors most central to the functioning of the decisionmaking network. It is critical to examine the attributes of individual actors, especially central actors. Knowledge of the policy interests of the participants makes it possible to predict how important members will behave and in which way they will guide policy. 2.2 Measuring the Distribution of Power: Total Number of Central Actors This structural construct gauges the number of policy actors that are in positions of influence within the decision-making network and more capable of withstanding opposition. 10 The summation of central actors denotes the extent to which various actors are in a position of influence within the network. A smaller number of central actors is indicative of a governance system in which fewer actors occupy a position of status within the network. Power is concentrated within a single or small set of actors, which reduces the potential for conflict. This system resembles the policy network described by DeLeon and Varda (2009) as involving a single central actor shaping policy and coordinating network activity. As the number of central actors increases, influence is dispersed amongst a larger set of network members. A larger number of central actors is indicative of an arrangement in which power is dispersed amongst 10 A similar construct, network centralization, is utilized by social network analysts to measure the degree to which various actors are centrally located (see DeLeon and Varda 2009; Knoke and Yang 2008; Provan and Sydow 2007; Rowley 1997; Sandstrom and Carlsson 2008; Wasserman and Faust 1994). More specifically, network centralization calculates how equally or unequally network members are connected to one another (Knoke and Yang 2008; Rowley 1997; Sandstrom and Carlsson 2008; Wasserman and Faust 1994). 23 many actors and members are more equally positioned to shape the network’s behavior (Provan and Sydow 2007). The decentralization of power makes the decision-making network more complex. There are several organizations that are highly engaged in the decision-making process and simultaneously attempting to guide the network’s policy choices. The structure of the network becomes further complicated as the range of policy preferences increase across network members. 2.3 Establishing the Diversity of Policy Goals: Homogeneity of Actors and Size The last two network-level constructs, homogeneity and size, account for the different types of policy actors present in the decision-making network. The characteristics measure variation in the organizational policy goals that circulate throughout the network by identifying . the number (Adam and Kriesi 2007) and range of policy solutions (Marsh and Smith 2000) that are being considered by network members. Network homogeneity is established by calculating the number of different organizational entites within the arrangement (Adam and Kriesi 2007; Carlsson 2008; Sandstrom and Carlsson 2008; Sandstrom and Rova 2009). This structural attribute denotes the degree to which joint decision-making occurs amongst actors with varying institutional policy objectives (Sandstrom and Rova 2009). Homogeneity is important for functioning of the network (Adam and Kriesi 2007; Carlsson 2008; Sandstrom and Carlsson 2008) and has a positive impact on the efficiency of the decision-making process (Balkundi and Kilduff 2006; Borzel 1997; Sandstrom and Carlsson 2008; Sandstrom and Rova 2009). Shared similar policy preferences within the network promote higher levels of goal consensus (Provan and Kenis 2007). Those involved in decision-making will have less difficulty communicating and 24 experience less obstacles in developing trust and reciprocity in the exchange of resources (Brass et al 2004). Heterogeneity amongst actors, however, has the potential to negatively impact network efficiency by inserting competing ideas, interests, and goals into the decision-making process, which makes collective action more difficult (Sandstrom and Carlsson 2008). Solidarity around the network’s general purpose and objectives, which is more likely within a homogenous set of actors, creates an atmosphere in which development actors are more likely to be active participants and willing to collaborate (Provan and Kenis 2007). This is further supported by findings that homogenous policy networks were likely to display increased cohesiveness and strong ties between network members (Balkundi and Kilduff 2006). According to Sandstrom and Carlsson (2008), the lack of competing interests within the network leads to a decisionmaking process that requires less time. 11 The level of goal consensus within the network reflects attitudes toward central actors’ proposed policy solutions. In sum, actor homogeneity suggests that central actors should face fewer difficulties when convincing other network members, who share similar policy objectives, to support their policy ideas. Although decreased levels of variation within network membership may strengthen collaborative efforts and heighten productivity, these networks are generally closed to new ideas (Sandstrom and Rova 2009). Network ‘size’ is analyzed to determine whether decision-making occurs within the context of many or few actors (Adam and Kriesi 2007). It is simply measured as the number of actors (nodes) within the network (Adam and Kriesi 2007). As discussed at length earlier in the 11 Compared to heterogeneous governance structures, decision-making networks with low levels of diversity are better capable of setting, changing, and enforcing rules (Sandstrom and Rova 2009). 25 literature review, actors and organizaions’ involvement in interorganizational arrangements revolves around mutual resource dependency. Organizations increase their dealings with others in an effort to obtain access to more strategic resources and improve the likelihood of successfully achieving their shared goals (Berardo 2009). Increased availability of resources makes open networks more preferable than closed networks (Berardo 2009). The benefit of engaging in a more inclusive network grows immensely when the engaged actors are contributing a diverse set of resources (Berardo 2009). 12 13 However, higher levels of inclusiveness may have adverse effects on the productivity of the network (Berardo 2009; Provan and Kenis 2007). The decision-making process can become inundated by the large number of actors and the magnitude of resources as well as by policy goals flowing through a multitude of linkages within the network (Provan and Kenis 2007). Openness substantially increases the complexity of policy networks, especially when the resources are nonredundant. In these instances, decision-making would require an extended amount of time and effort when organizing the activities (Berardo 2009; Provan and Kenis 2007). Increases in collaboration enhance network performance only as long as actors’ resources and interests overlap. Larger network sizes may contribute to extreme complications in the processing of resources (Berardo 2009) and policy interests, as well as in the selection of 12 The presence of nonredundant resources fosters collaboration by filling “structural holes” within the network and linking actors who, theoretically, are in possession of varying sets of resources (Berardo 2009). 13 See also Burt (1992) and Burt (2005) for a more extensive discussion. 26 policies. The implication is that central actors will endure increased difficulties when attempting to control the flow of resources and influence policy in more inclusive networks. The disadvantages of larger networks are exacerbated amongst a heterogenous set of central actors. The relationship between homogeneity and size (as demonstrated in Table 1.2) suggests that effect of openness on network behavior is largely contingent upon the type of policy actors that are influential in the decision-making network. According to Adam and Kriesi (2007), “an increase in the number of actors and in the diversity of their composition not only increases the complexity of policy networks, but also the number of options for an adequate solution to the problem” (133). This could, potentially, decrease central actors’ capacity to induce enough policy support from network members to successfully create policy in their favor (Berardo 2009). The level of inclusiveness in a network has a smaller impact on the efficiency of decision-making when there is a homogenous group of actors and comparable policy resources are being shared. The process of obtaining goal consensus, thus decision-making, in small homogenous networks is less complicated and more manageable. In sum, network size magnifies the effects of homogeneity or heterogeneity amongst actors. 2.4 Conclusion Political science and public policy scholars have long been concerned with who participates in decision-making at various levels of government. More importantly, scholars have been especially interested in understanding who has power, how that power was obtained, and how actors use power to influence output. The discourse revolving around the seminal question “Who governs?” (Dahl 1961) contributed immensely to unraveling the complexities of 27 power distribution within decision-making. However, the relationship between who governs and which policies are selected remains unclear. Based on fundamental arguments made by urban regime theorists and policy network theorists, I construct a conceptual framework that provides a stronger explanation for the relationship between governance structures and policy. Although both of these approaches are strongly rooted in the governance framework and make similar core arguments, they focus on different elements of the decision-making process. Urban regime theory emphasizes the actions of individual policy actors, while network theory focuses on structural arrangements of the collaborative as the critical determinant of resulting policy choices. The use of similar concepts and arguments make it suitable to adopt concepts from each approach and create a more complete theory of decision-making. Bringing these theoretical approaches together creates a balance between the actor-centered and structure-centered frameworks. The main argument presented in this dissertation is that the combined structural effects (total number of central actors, homogeneity, and size) mediate the extent to which individual actors are capable of exerting influence throughout the network and affect the performance of policy networks. In their pursuit to influence policy decisions, actors must overcome networklevel structural constraints that determine the dispersal of power and range of interests present in the decision-making environment. These structural characteristics represent the potential for consensus and conflict over policy goals as well as the likelihood of central actors’ ability to affect policy. Overall, structurally distinct decision-making networks present different advantages and disadvantages for individual network members. By emphasizing the motivations of central policy actors as well as the structural configuration of the decision-making network, descriptions of governance structures can better account for the context in which joint policy- 28 making occurs. More complete profiles of the decision-making network’s structure support the formulation of stronger hypotheses regarding the policies that are generally utilized by decisionmaking networks with similar membership. The conceptual framework put forth in this dissertation has an increased ability to describe variations in policy decisions across a wide range of governance networks in different policy areas. The central research question for this study focuses on examining whether the membership structure of decision-making networks impact the policies that communities adopt. More specifically, do structurally distinct decision-making networks vary systematically in the policies they utilize to address local economic development concerns? In the process of examining this area of research, three other strongly related questions are posed. Are there structurally distinct decision-making networks that differ in governing American cities? How does the membership structure of decision-making networks differ across communities? What factors promote and hinder the existence of the various network types in a community? In other words, why do cities vary in the types of decision-making networks that exist in their communities? In Chapter 2, I outline the four-part quantitative analysis that was conducted to investigate these important questions regarding local governance and policy decisions, specifically in the area of local economic development policy. I examine cross-sectional data provided by the 2004 Economic Development Survey, which was conducted by the International City/County Management Association (ICMA). 14 First, the independent variable, decision- making network type, is examined. Original structures of development decision-making 14 The survey was sent to local administrators of municipalities in the U.S. with populations greater than 10,000 and counties with populations of 50,000 or more. 29 networks are empirically derived, classified, and analyzed. Two-step cluster analysis was executed to group cities according to similarities in the development actors that were identified by representatives of city government as involved in the decision-making process. The main assertion presented in this dissertation is based on the assumption that communities across the nation are governed by a variety of local economic development decision-making networks with distinct organizational arrangements within the membership. The results of the cluster analysis supported this claim and indicated the existence of three distinct cluster groups, or decisionmaking network types, present in the data. Cities assigned to the same network type provided very similar responses to the survey questions that were utilized a) to determine which local economic development actors were involved in decision-making, and b) the organization with primary responsibility for local economic development in the communities. A variable was created to represent each decision-making network type. The second part of the analysis involves determining how the membership structure of these decision-making networks varies. Descriptive statistical tests were used to identify the central actors, and to examine structural patterns in the network membership of the several cities within each decision-making network type, and to construct a structural profile. Whereas urban regime theory and network theory generally conduct qualitative analyses of individual governing coalitions, this dissertation involves a large-N dataset. Instead of examining a single network within each category, I have aggregated numerous decision-making networks with comparable organizational structures. The descriptions of the structural attributes (central actors, total number of actors, density, homogeneity, and size) are, essentially, summations of these characteristics across decision-making networks of the same type. Given the nature of network theory, it is not possible to hypothesize about the policy choices of local economic development 30 decision-making networks until network members and their policy preferences have been identified. After gaining a better understanding of the political context in which decision-making occurs in Chapter 3, hypotheses regarding the policy preferences of each network type are developed. Critics of both the urban regime and network theoretical approaches maintain that they need to provide more than descriptions of potential governance structures. The measures included in an analytical framework must be suitable for descriptive and empirical analysis (DeLeon and Varda 2009). It is not enough to simply state that the structural configuration and relationships amongst actors within the network have substantial implications for the behavior of the network. It is important to also test the effect of network structure and demonstrate that structurally distinct decision-making networks make different policy decisions. In the third stage of this analysis, I conducted binomial logistic regression to measure the effect of decisionmaking network type on the likelihood of a city using various local economic development policies. The purpose of estimating these models was threefold: determining whether decisionmaking networks do differ in their policy adoption at a level of statistical significance, establishing which decision-making network types were most and least likely to make use of the different policies, and identifying the other variables that influence local governments’ policy choices. Lastly, an exploratory analysis was conducted to identify the factors that influence the emergence of decision-making networks. Urban regime theory has been criticized for its diminished ability to predict the type of regime, governing coalition, or governance network that will exist in a community (see Davies 2002; DiGaetano 1997; Dowding 1999; Lauria 1997; Mossenberger and Stoker 2001; Orr and Stoker 1994;). Scholars have largely ignored the factors 31 that affect the distribution of resources and therefore lead to the materialization of a particular decision-making network type. Three binomial logistic regression models were estimated to investigate the effects of social, economic, geographic, and political variables on the formation of different types of decision-making networks in cities. The first part of the empirical analysis is presented in Chapter 3. Hierarchical cluster analysis was conducted to develop an original typology of structurally distinct local economic development decision-making networks. Three network types were identified, including municipal, local government-chambers of commerce, and broad public/private partnership development decision-making networks. Based on the structural characteristics of the decisionmaking networks, predictions about their policy use were made. In sum, municipal networks are predicted to have the most narrow local economic development policy agenda while broad public-private partnership networks are more likely to utilize an extensive set of development policies. It is hypothesized that local government-chambers of commerce networks will display policy use that includes a slightly smaller and less diverse set of policies than the broad publicprivate partnership networks. The municipal decision-making networks are characterized by a single central actor, city governments, coordinating development activities amongst a very small network of actors. As a result of decreased access to private resources, these networks are expected to make use of a smaller set of policies. This network type is also more likely to rely on development policies that city governments can maintain with minimal collaborative efforts, particularly public ownership of profit-making facilities and business incentives, and low upfront financial obligations like business subsidies (i.e. tax abatements, tax increment financing, enterprise zones, etc.). 32 The local government-chambers of commerce network and broad public-private partnership network are expected to pursue corporate-centered strategies given the strong presence of business-related organizations in very influential structural positions. The local government-chamber of commerce network is medium-sized and predicted to have adopted many local economic development policies. These networks will promote some communityoriented development policies to accommodate the few citizen-based organizations that are in moderately central positions. These policies, however, will only be utilized if they offer considerable economic advantages to the business leaders who are more prominent development actors. Broad public-private partnership networks are predicted to be stronger supporters of community-based policies given that there are many development actors in moderately central positions with interests in promoting policies that focus on revitalizing social capital and offering economic opportunities in distressed communities. Also, nonprofit development corporations are primarily responsible for managing local economic development in these cities. Given the large and diverse set of development actors involved in decision-making, the broad publicprivate partnership networks are expected to engage in a broad set of development policies. The central argument presented in this dissertation is tested in Chapter 4. Two different statistical analyses are executed to assess whether the decision-making network type is a significant determinant of a city’s policy use. First, crosstabulation is conducted to identify the dependent variables, or local economic development policies, that share a strong bivariate relationship with the independent variable, decision-making network type. Of the forty-four local economic development policies included in the survey, decision-making network was only related to sixteen policies at some level of statistical significance with p < 0.05. 33 A binomial logistic regression model was estimated for each of the sixteen development policies to determine if cities’ use of the policy is influenced by decision-making network type. The independent variable only predicted six policies (community development corporations, microenterprise programs, partnering with nongovernmental organizations, partnering with local governments, tax abatements and one-stop permit issuance) at a level of statistical significance. The relationship between decision-making network type and policy use is described in terms of odds ratios and probabilities. The findings from the full models estimating the likelihood of these six policies being used by a city support the hypothesized policy profiles of the local economic development decision-making networks that were developed in Chapter 3. There were strong similarities between policy use of the local government-chambers of commerce network and broad the public-private partnership network. Governmental organizations, primarily local government, and business-related organizations (chambers of commerce, private business/industry, and economic development corporations) are central to the decision-making processes in communities led by decision-making networks of both types. These networks are heavily engaged in business-sensitive development policies. There is a slight, yet critical, difference between these two network types. Broad public-private partnerships are larger and include a more diverse network of actors. As such, the broad publicprivate partnerships make use of a larger number of policies. Also, this network type has several moderately central actors with citizen-based and community-oriented development goals (citizen advisory boards or commissions, colleges and universities, ad hoc citizen groups, as well as private and community economic development foundations). They are slightly more engaged in community development policies, particularly community development corporation policy, and less involved in microenterprise programs which support business interests. With regard to the 34 municipal networks, they practice a smaller number of policies and display the lowest rates of use for most local economic development policies, including business incentives and other policies that generally do not demand substantial private investments. Despite only requiring the expansion of governmental services or forfeiture of potential tax revenues, municipal networks did not have a positive impact on the use of policy subsidies. Overall, decision-making network type was not a strong predictor for policy use with the exception of six development policies. However, these findings generally provide support to the predictions presented in Chapter 3 regarding the various decision-making network types’ policy preferences. For this reason, it is important to understand the factors that lead cities to differ in the type of decision-making network that develops in their communities and influences policy choices. In Chapter 5, I investigate whether cities vary systematically in the membership of the networks governing their communities. Three binomial logistic regression models were estimated to determine how different factors influenced the likelihood of each decision-making network existing in a city. There were several independent variables included in the model, but metropolitan status and locality type were the only variables that consistently influenced the likelihood of each decision-making network being present in a community. Independent municipalities were most likely to have broad public-private partnership networks managing development decision-making, which gives access to a more diverse set of development actors. Municipal and local-government chambers of commerce networks had an increased likelihood of existing in suburban communities. Central cities varied in the types of decision-making networks. These findings suggest that membership structure of the decision-making networks is most affected by the community’s place in the regional hierarchy. This is a valuable empirical 35 contribution. A theory must be developed to describe how metropolitan status and locality type are significant predictors of governance structures. The results of this exploratory analysis aid in filling a significant gap in the literature. The underperformances of the independent variables indicate that more research is required. Theoretically relevant variables need to be identified and tested. Investigating the factors that give rise to particular types of governing structures will be useful for further understanding the importance of local governance. In the Conclusion chapter, I provide a summary of the conceptual framework, methodology, findings, and discussion of the methodological limitations of this dissertation study. I also highlight the different theoretical, substantive, and methodological contributions that this study offers. In this dissertation, the conceptual framework facilitates the development of more detailed membership profiles of governance structures and the effects of different network types on policy use measured using a large-N dataset. The underperformance of the statistical models suggests that the overall impact of decision-making network type on policy use is low. I argue that structurally distinct decision-making networks may not differ significantly in whether they use the policies, but vary in the degree to which they use them. I present a plan of model re-specification to retest the relationship between decision-making network type and policy use. Future research will examine the effects of these decision-making network types on the level at which cities make use of different policy types. 36 CHAPTER 3: DATA AND METHODOLOGY The purpose of this dissertation is to uncover the significance and direction of the relationship between decision-making networks and cities’ selection of local economic development policies. This study centers on determining whether the membership structure of the decision-making network influences the type of policies that cities adopt. Are the structural properties of decision-making networks predictive of the network’s policy decisions? If the conceptual framework for understanding the relationship between governance and policy is compelling and the interpretation of previous research regarding the policy preferences of various development actors (as well as the results of the cluster analysis) are legitimate, then cities led by different decision-making networks should vary systematically in their local economic development policies. A complete understanding of the relationship between governing structures and policy choices requires an examination of the independent variable, local economic development decision-making networks. As such, a study of decision-making networks must first determine the composition of the decision-making networks that guide development. Are there distinct decision-making networks presiding over local economic development policy in cities? More importantly, what are the structural differences in the membership of the various development decision-making networks? How do these interorganizational arrangements vary in the number and types of central actors involved, density, level of inclusion, and homogeneity of actors? To further understand decision-making networks, this study also poses the following question: Do cities vary systematically in the types of decision-making networks managing development policy in their communities? What is the role of social, economic, and political factors in predicting the emergence of different networks in cities? The conceptual framework presented 37 in the previous chapter provides a description of governance that is applicable to numerous areas of policymaking. In this dissertation, the relationship between decision-making networks and policy choices is examined within the policy area of local economic development. 3.1 The Complexity of Local Economic Development Decision-Making Networks Economic development is an integral component of a community’s social and economic well-being. The general consensus across foundational urban politics theories is that local governments enact development policy primarily to: generate and retain jobs as well as attract residents, and improve per capita income and land values within their jurisdictions, which leads to increases in tax revenues, higher quality services, and municipal status (Logan and Molotch 1987; Molotch 1976; Peterson 1981; Stone 1989; Swanstrom 1985; Tiebout 1956). 15 All of these classic perspectives on urban governance and politics generally argue that local governmental officials are driven to engage in development activities to support the social and economic well-being of businesses and residents to ultimately improve their city’s status. City governments do vary considerably in the degree to which they are involved in economic development endeavors and in which policies they practice. Scholars have consistently identified five factors that influence local economic development activity, including: citizen need (Rubin and Rubin 1987; Clingermayer and Feiock 1990), fiscal stress (Fleischmann and Green 1991; Rubin and Rubin 1987; Schneider 1989; Sharp 1990), population growth 15 Some of the most seminal theories include: Tiebout’s public choice theory (Tiebout 1956), Logan and Molotch’s urban growth machine (Logan and Molotch 1987; Molotch 1976, 1988), Peterson’s city limits theory (Peterson 1981), and Elkin and Stone’s urban regime theory (Elkin 1987; Stone 1989, 1993). 38 (Rubin 1986; Rubin and Rubin 1987) and size (Green and Fleischmann 1991), regional competition (Bowman 1988; Schneider 1989), and governmental structure (Clingermayer and Feiock 1990; Fleischmann and Green 1991; Fleischmann et al. 1992; Feiock and Clingermayer 1986; Reese 1997; Rubin 1986). Those studies focusing on governmental structure as an explanatory factor of local development activities focus primarily on the formal institutions of local government, including different types of local government systems as well as the role of the various agencies, locally elected officials, and appointed officials (Stoker 1998). In this study, I analyze the informal collaborative aspects of decision-making. As stated by Reese and Rosenfeld (2002), there is a “critical distinction between the structure of government and the act of governing” (645). The governance of local economic development is not necessarily limited to local government. Scholars of political science, public policy, and public administration and management have declared that the act of governing local economic development has undergone a fundamental change as private entities have become more influential in policy process. In the late 1980s, federal devolution, globalization, and the downturn of the American economy placed a considerable amount of pressure on city budgets (Eisinger 1998; Fainstein and Fainstein 1989; Gibbs and Jonas 2000; Weir et al. 2005). While facing economic decline, cities were experiencing decreased federal assistance as well as additional policy and fiscal responsibilities (Agranoff and McGuire 1998; Kettl 2000). In response to this “government rescaling” (Gibbs and Jonas 2000), local elected officials were desperate to replace lost revenues and improve local governments’ ability to finance the policies and programs needed to enhance local economy. Increased independence compelled city governments to depend heavily upon local resources to support development ventures (Eisinger 1998). These resources, however, are 39 in short supply. Although government is the only actor with legitimate authority to employ a particular set of policy instruments, a variety of private agents have gained access to valuable financial, information, and technical resources, which increased their status in local governance via the economic development decision-making process (Agranoff and McGuire 1998; Kettl 2000; McGuire 2000). To effectively govern and address development under new circumstances, local government officials increased access to strategic resources by mobilizing and coordinating various private stakeholders (Agranoff and McGuire 1998; Clarke and Gaile 1992; Fainstein and Fainstein 1989). Local governments’ increased dependence on private resources also transferred power to private development actors. Consequently, local private actors, organizations, and institutions have become more influential in the governing of local policy. Local economic development decision-making occurs within a network of actors (Agranoff and McGuire 1998). The number of organizations with strategic resources and interest in improving the local economy has increased, thereby increasing the number of organizations that must be mobilized to effectively govern and promote local economic development (Agranoff and McGuire 1998). To further complicate the governance of local economic development, each actor has an interest in promoting development according to the organization’s economic goals (Agranoff and McGuire 1998a, 1998b; Eisinger 1998; Fainstein and Fainstein 1989; McGuire 2000). Although all stakeholders are generally concerned with improving the health of the economy, they vary in their preferred policies and strategies. For this reason, the effectiveness of local development policymaking is contingent on the lead actor’s capacity to overcome the power struggle and organize various actors and resources in 40 collective development action (Agranoff and McGuire 1998; Clarke and Gaile 1992; Eisinger 1998; Fainstein and Fainstein 1989; McGuire 2000). The types of actors engaged in development decision-making networks vary across cities. As such, there is an assortment of very complex public-private partnerships governing local economic development across the nation (Agranoff and McGuire 1998; Fainstein and Fainstein 1989; McGuire 2000; Olberding 2002). Analysis of the different local economic development decision-making networks with distinct structural arrangements facilitates an improved understanding of the effects of these governance networks on the act of governing, particularly cities’ policy choices regarding which instruments and strategies they employ to improve the local economy. 3.2 Data In this dissertation, data provided by the International City/County Management Association’s (ICMA) 2004 Economic Development Surveys were analyzed. 16 17 This organization has operated as a resource for local governments since its establishment in 1914. According to their website (http://www.icma.org), the International City/County Managers Association focuses on equipping localities with a variety of resources, including technical 16 For more information, please contact: International City/County Management Association at 777 North Capitol Street NE, Suite 500, Washington, DC 20002-4201. They may also be contacted at 202.289.ICMA (phone) or 202.962.3500 (fax). 17 This dataset was made available by Dr. Laura Reese and the Global Urban Studies Program at Michigan State University. 41 assistance, training, publications, and data. Since 1984, ICMA has conducted a version of this survey every five years. The 2004 questionnaire used in this study was sent to local administrators of municipalities in the U.S. with populations greater than 10,000 and counties with populations of 50,000 or more. 18 ICMA reported a response rate of 19.6% with N = 726. 19 Over the years, scholars have used the results of these instruments to examine how local economic development policy is influenced by form of government and administrative organization (Feiock and Kim 2001; Sharp 1991; Sharp and Elkins 1991), and by collaborative policy networks (Agranoff and McGuire 1998; Fleischmann, Green, and Kwon 1992; Kwon, Berry, and Feiock 2009; McGuire 2000). Wolman and Spitzley (1996) utilized the ICMA Economic Development survey to develop a typology of local economic development activities. The dataset was reduced twice. This study focuses on local economic development decision-making networks in cities. The survey, however, is administered to both city and county governmental officials. To limit the sample to city governments only, all communities coded as ‘2’ on the UTYPE variable were removed from the dataset. This variable was created and labeled by ICMA to represent “all U.S. counties defined by Census Bureau plus city-county consolidations that function as county governments” (ICMA 2004b, 1). A total of 89 cases were removed from the dataset. 18 Please review Appendix Figure 2.1 for a copy of the ICMA 2004 Economic Development Survey 19 This response rate was similar to those of other years in which the survey was administered: 1999 (31.5%) and 2009 (22.2%) 42 The sample was further reduced to accommodate for the marked sensitivity to missing data of the clustering method used to group cases. This mathematical tool is employed in Chapter 3 (An Application of Cluster Analysis: Empirically-Derived Typology of Local Economic Development Decision-Making Networks) to identify the structurally distinct decision-making network types that shape local economic development in sample cities. A large number of cities did not respond to the portion of the survey that focused on the cluster analysis variables. This was problematic for the software being utilized to conduct the empirical analysis and impacted the quality of the resulting typology. In an effort to avoid significant difficulties with the methodological tool employed in this study, cities that did not respond to the two questions utilized to create variables representing local economic development actors and primary responsibility for development were removed from the dataset. The omission of cases with missing data on all of the variables reduced the dataset by 114 cases. Of the 726 cities that responded to the survey, 523 cases remained in the final dataset. A list of cities included in the dataset can be found in Appendix Table 2.2. 3.3 Description of the Dataset There were five different local governments included in the sample. Most of the responding governments (79%, n = 411) were cities and 21.4% of the sample (n = 112) were described as some other type of sub-state governmental jurisdiction, including towns (9%), villages (7%), townships (4%), and boroughs (1%). With regard to the metropolitan status of the communities, 62% of the communities (n = 323) were classified as suburbs. Central cities and independent cities were evenly represented with the latter constituting 18% of the sample and 43 central cities composing 20% of the sample. 20 As shown in Figure 2.1, the sample was largely composed of smaller cities. Half of the sample reported population sizes between 10,000 and 24,999, or 49%. Twenty-five percent of sample cities had 25,000-49,999 residents, and 16% of the responding communities were home to between 50,000 and 99,999 residents. A small proportion (8%) of the cities had populations between 100,000-249,999, and less than 2% of Figure 3.1 Population Sizes of Areas Included in Sample 20 The independent municipality cities are not located within a Metropolitan Statistical Area (MSA). 44 sample cities were larger than 250,000. The findings presented in Figure 2.2 show that the dataset is representative of cities located in different geographic regions. Cities in the North Central region were most represented and accounted for 34% of the sample. Southern cities were present at a similar rate of 31%. Western cities constituted 22% of the sample, and 13% of the sample was located in the northeastern area of the nation. Further analysis shows that smaller divisions of regional areas are included at fairly similar rates. Cities located in the East North-Central (Illinois, Indiana, Michigan, Ohio, and Wisconsin) region were the most prevalent and made up 25% of the sample. The Pacific Coast (Alaska, California, Hawaii, Oregon, Washington), South Atlantic (Delaware, Florida, Georgia, Figure 3.2 Geographic Areas Represented in Sample Maryland, North Carolina, South Carolina, Virginia, West Virginia, District of Columbia) and West South-Central (Arkansas, Louisiana, Oklahoma, Texas) regions were represented at rates of 45 Figure 3.3 Regional Areas Represented in Sample 21 21 Each geographic area is composed of several states: New England (Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont), Mid-Atlantic (New Jersey, New York, Pennsylvania), East North-Central (Illinois, Indiana, Michigan, Ohio, Wisconsin), West NorthCentral (Iowa, Kansas, Minnesota, Missouri, Nebraska, North Dakota, South Dakota), South Atlantic (Delaware, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, West Virginia, District of Columbia), East South-Central (Alabama, Kentucky, Mississippi, Tennessee), West South-Central (Arkansas, Louisiana, Oklahoma, Texas), Mountain (Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming), and Pacific Coast (Alaska, California, Hawaii, Oregon, Washington). 46 12%-14% in the sample. Three areas, Mid-Atlantic, West North-Central, and Mountain regions, each constituted 8% of the cities included in the dataset. Cities located in the New England (6%) and East South-Central (5%) regions were the least represented. Several different types of local governmental systems are represented in the dataset. Cities in which a council-manager system was functioning comprised 77% of the sample. Mayor-council systems were the form of government utilized by 21% of the sample cities. Other governmental systems were represented in the data but at very low levels. Town meetings governed six (1.1%) sample cities while commission and representative town meeting systems were present in only two cities (0.4%). At the time of the survey, few of the sample cities reported experiencing economic decline within the last five years. The economic conditions of sample cities, generally, were stable or involved some degree of growth (including rapid expansion, moderate growth, and slow growth). Approximately 60% of the cities had experienced either moderate (31%) or slow (31%) economic growth over the previous five years. Seventy-six (15%) of the responding communities gave an account of rapid expansion (more than 25% growth) in the local economy. The same proportion of cities stated that the condition of their local economy had remained stable (no real growth or decline). Only 8% of the sample had experienced some economic decline, with 5% reporting slow decline. A very small percentage of cities (1%) had endured moderate or rapid decline. The cities in the sample varied in their primary economic base. Many local economies largely revolved around retail/service (28%), manufacturing (21%), residential (20%), or institutional/office industries (13%). Few sample cities listed their primary economic base as tourism/hospitality (4%), agriculture (4%), warehousing/distribution (2.8%), other (2.8%), or technology/telecommunications (2.4%). 47 Overall, the description of the dataset indicates that a broad set of cities is included in the sample. However, respondents mostly represent city governments, specifically suburban communities and those areas with at least 10,000 residents but no more than 250,000 residents. 22 The sample cities are located within various regions across the nation, although many (25%) are found in the states of Illinois, Indiana, Michigan, Ohio, and Wisconsin. With regard to governmental structure, the sample is overwhelmingly representative of cities with reformed council-manager systems. Mayor-council systems were the only other form of government to be adequately represented in the sample. At the time of the survey, sample cities varied in their primary economic base, but almost all of the sample cities reported experiencing some degree of economic growth or stability. 23 This study involves a degree of self-selection of cases, which presents some limitations by potentially causing sampling bias. Independent sample t-tests were conducted to measure the mean differences between the study sample (N =523) and the excluded cities (N=114) across several demographic variables within the ICMA data, including: locality type, metropolitan status, form of government, population, and geographic area. The findings were utilized to determine whether the set of cities excluded from the study varied significantly from the sample analyzed. This statistical test provides two sets of results: descriptive statistics (Table 3.1) and the inferential statistics (Table 3.2). The descriptive statistics indicate that groups shared the 22 This is in comparison to other sub-state governmental jurisdictions not defined as county governments, including towns, villages, townships, and boroughs. 23 For this reason, all inferences drawn in this study may not be generalizable to cities having experienced slow, moderate or rapid decline over the previous five years. 48 similar means for each variable, which is indicative of two very comparable sets of cities. The results of the inferential statistical tests were also positive. For each variable examined the Levene’s Test for Equality of Variances test statistic was greater than the .05 signifying that the Table 3.1 Independent T-Test Results: Descriptive Statistics variability of each group was equal. Assuming equal variances, the results of the t-test failed to display a statistically significant difference between the groups across all variables except locality type (t = -10.190, p = 0.000). In other words, the group of excluded communities included more towns, villages, townships and boroughs. The results of the t-test demonstrate that the final sample did include more cities than other locality types, but was quite similar to other local governments who responded to the ICMA Economic Development survey but were excluded from the analysis. 49 Table 3.2 Independent T-Test Results: Inferential Statistics 50 3.4 Methodology This dissertation research involves a three-part analysis. The central research question focuses on identifying how structural differences in the composition of decision-making networks influence variations in cities’ use of local economic development policies. There are, however, two other closely related research questions that also offer critical insight into the relationship between decision-making network type and policy choices. Prior to examining the effect of these governance structures on policy, it is essential to inspect the membership structure of local economic development decision-making networks. If decision-making network type is a significant predictor of municipalities’ policy use, then it is important to explore the various factors that influence the emergence of some network types rather than others. Many regime theorists as well as critics and other scholars have addressed this research question and developed various typologies to classify alternative “regimes” or “governing coalitions” and “policy” networks. 24 These findings and categorization schemes are insightful, however they are also problematic. These typologies were largely theoretical or had been developed based on findings from a small-N case study analysis. The inconsistencies of coalition and network type found across classification schemes make it difficult to develop general inferences about the presence and structure of decision-making networks in different communities. The first step in this study is to develop a rigorous methodology that permits the identification of local economic development decision-making networks within a large sample of cities. Original types of local economic development decision-making networks are empirically 24 A discussion of the various systems developed for classifying urban regimes and local economic development policy networks can be found in the Literature Review (Chapter 1). 51 derived, analyzed, and classified according to distinct patterns in their membership structure regarding centrality, network density, inclusion, and homogeneity of actors. Each category included in the development decision-making network typology represents a group of cities that share similarly structured decision-making networks. The use of a large-N dataset produces a typology of decision-making networks that is not theoretical in nature and is applicable to a large set of cities. This typology advances efforts to identify patterns in policy use within and across development decision-making network types. In accordance with Bruce and Witt’s (1971) suggestion that “it may be more reasonable to compare groups of cities that are similar in structure rather than individual cities” (239), the large-N dataset strengthens the inferences drawn in this study. This analysis involves two sets of variables: development actors’ participation in the decision-making network (“Who participates?”) and the organization primarily responsible for economic development (“Who leads?”). To identify those actors engaged in the decisionmaking process, this study uses the reputational method of data collection. The first group of variables was used to operationalize an individual organization’s involvement in the local economic development decision-making process. The ICMA economic development survey questioned city leaders about the perceived involvement of sixteen specific economic development actors in the local economic development decision-making process (refer to Table 25 3.3). More specifically, the survey asked respondents: “Which of the following participate in developing your local government’s economic development strategies? Check all applicable”. A 25 The 2004 ICMA Economic Development Survey included “other” development actors twice. None of the respondents indicated that a second unlisted development actor was involved in development strategies. As such, this variable was not included in the analysis. 52 dichotomous variable (0=no; 1=yes) was used to indicate whether each organization is involved in the decision-making process. Thus, the membership of the empirically-derived decisionmaking networks are characterized based on the respondents’ perceptions of individual actors’ involvement in the local economic development decision-making process. While these variables do not effectively measure the extent to which each actor participates in the decision-making process, particularly with regard to the amount of resources dedicated to policy efforts (i.e. time, money, and expertise) or their specific role, they do provide enough information to assess variations in the inclusion of each actor in local economic development decision-making networks. These variables were used to establish the horizontal structure (“Who participates?”) of the development decision-making networks. The percentage of communities that selected an organization as a contributor to their decision-making process represents the development actor’s rate of participation within the network. Comparison of these rates demonstrates the degree to which organizations are most and least involved in the various local economic development decision-making networks. Table 3.3 Local Economic Development Actors 53 Individual development actors differed greatly in the extent to which they were engaged in decision-making. Their various participation rates, which equal the proportion of cities that selected the organization as a participant in their local economic development decision-making process, are displayed in Figure 3.4. Comparison of these rates demonstrates the degree to which organizations are most and least involved in local economic development decision-making Figure 3.4 General Trends in the Participation Rates of Specific Development Actors across the nation. The findings clearly show that most of the listed development actors are at least moderately involved in development decision-making. Ten of the sixteen actors included in the survey were involved in development decision-making in at least 30% of the cities. Four 54 organizations, private/community economic development foundations, federal government, ad hoc citizen groups, and other actors, were engaged in the shaping of local economic development policies in less than 10% of the cities sampled. The second variable was concerned with leadership within the decision-making process and was examined using a categorical variable that indicated whether a local government, nonprofit development corporation, or some other organization possessed primary responsibility for economic development within local government. The hierarchical cluster analysis technique used in this study is appropriate only for count, interval or binary data. For this reason, the categorical variable was recoded and a dichotomous variable was created for each category listed. In total, there were three variables utilized to understand the type of organization (local government, nonprofit organization, or other) that supervises economic development within the municipalities and ascertain the hierarchical structure of the decision-making networks. Most often, local governments (70% of the sample) presided over the development activities in their cities. A nonprofit development corporation managed local economic development in 21% of the cities while some other organization (chambers of commerce, other, or a mixture of organizations) was primarily responsible for development in 9% of the cities. 26 26 For this question, other organizations, chambers of commerce, and mixtures were also listed. The frequency of these options was low: other (0.8%), chambers of commerce (1.3%) and mixture (6%). These categories were collapsed and recoded to be included in the “other” category. There were only three categories for the primary responsibility variable. 55 These variables are used to construct the independent variable of primary interest to this dissertation, development decision-making networks. Cluster analysis, a mathematical method used to identify groups of similar cases within a dataset, is employed to examine the existence of structurally distinct development decision-making networks within a sample of American cities. Cluster analysis involves several different mathematical methods whose purpose is to identify groups or clusters of comparable cases (Bruce and Witt 1971; Hill et al 1998; Everitt et al 27 28 2011). This study employs hierarchical clustering primarily because this method does not require the number of final cluster groups to be predetermined like the other clustering procedures. All clustering procedures involve the specification of variables, or attributes, by which similarity between cases will be evaluated and groups of homogenous cases will be identified (Bruce and Witt 1971; Hill et al 1998; Everitt et al 2011). It is important that the clustering variables, or indicators, are considered carefully. Hill, Brennan and Wolman (1998) state that “variable selection is critical…because the cluster analysis minimizes the within-group variance based on all of the variables included in the analysis and cannot distinguish between variables statistically. Therefore, variables that do not have theoretical reasons for inclusion will distort 27 For a complete discussion of the various clustering methods see: Everitt, B., Sabine Landau, Morven Leese, and Daniel Stahl (2011). Cluster Analysis. 5 ed. London: Wiley. 28 SPSS/PASW Statistics 18.0, the quantitative data analysis computer package used in this dissertation, provides three different techniques for use in clustering data: hierarchical, k-means, and two step clustering. Each method uses different criterion for measuring similarities amongst cases and merging clusters; thus, the identified cluster groups vary across clustering procedures. 56 the results of the clustering process” (1939). In the instance that some of the indicators are not important for discerning similarities and differences between cases, the resulting clusters may not be useful and fail to identify distinct groups. Given that only local government representatives responded to the survey, the results of the cluster analysis represent the membership of decision-making network types as perceived by the representative of local government that responded to the survey. In this study, cities were clustered according to similarities and dissimilarities in the horizontal (“who participates”) and vertical (“who leads”) structure of the membership within the decision-making networks that direct local economic development policy. The indicators included sixteen dichotomous variables that represent the involvement of various organizations in the development of local economic development strategies (see Table 2.4) as well as three dichotomous variables denoting whether local government, nonprofit development corporations, or some other organization possessed primary responsibility for economic development within local government. In total, there were nineteen asymmetrical binary variables included in the cluster analyses. If the variables are not related, then unstable and uninterpretable clusters will be found (Fleishman 1986). Large amounts of missing data present a considerable challenge to cluster analysis. When observations are omitted for clustering variables, there are difficulties in calculating the measure of similarity between cases (Fleishman 1986; Everitt et al 2011). In the instance of missing data, cluster analysis cannot determine whether cities are governed by decision-making networks that are comparable in structure. When the involvement of a development actor is omitted from the analysis, a city could potentially be erroneously assigned to a cluster group. With complete information for each variable included in the cluster analysis, 57 the city may be included in a different cluster group. Extensive levels of omitted data increase the likelihood of uninterpretable clustering groups (Fleishman 1986; Hill et al 1998; Everitt et al 2011). For this reason, cases with missing observations for any of the clustering variables were removed from the study. 29 Hierarchical cluster analysis entails three steps. First, the researcher has to define how the similarity between cases is measured. This mathematical procedure is based on the idea that similarity between cases is quantifiable. There are numerous measures of similarity that are used in cluster analysis to examine the proximity of one case to another. The similarity, or distance, measure is selected according to whether interval, count or binary data are being examined. In this study, the Jaccard coefficient was used. This similarity measure is commonly employed with binary data (Finch 2005; Garson 2010). The Jaccard coefficient is, essentially, a ratio of the 29 The ICMA dataset codes each community’s selection of an organization as “participat[ing] in developing your local government’s economic development strategies” as a “1”. The dataset, however, does not provide an observation for cities that did not select the actor as engaged in development decision-making. More specifically, the non-involvement of an organization is not coded and left blank. Cities with missing data for the entire set of development actor variables were considered to have skipped the question and were removed from the analysis. If the cities did select one or more of the organizations as contributing to economic development decisionmaking, they were considered to have answered the question and, thus, are included in the study. For these cities, the development actor variables with missing values were re-coded as “0” to indicate that these particular organizations were not engaged in the decision-making process. Essentially, only cities who selected at least one organization as a participant in the development decision-making network were included in the analysis. 58 number of variables for which both cases share a “1” to the frequency at which either case was coded “1” for a clustering variable. This measure, inherently, “excludes cases from the denominator where neither subject has the trait of interest” (Finch 2005, 87). A proximity matrix displaying the distance between a single case and the other cases in the dataset was created. “The basic data for cluster analysis are indices of the dissimilarity between pairs of cases” (Fleischman 1986, 374). The proximity matrix provides the data needed in the second step of hierarchical clustering, the formation of cluster groups. A method must be selected to determine how cases are coupled and merged into homogenous clusters. There are several “linkage methods” used to group cases according to their similarity. 30 The average linkage within groups, which is also referred to as within-group linkage, was employed in this study. This particular grouping algorithm combines clusters according to the average distance, or similarity, between all possible pairs within the newly joined cluster. With agglomerative hierarchical clustering, each case begins as an individual cluster. Initially, the two most similar cases as measured by their Jaccard coefficient will be merged to establish new clusters. For each subsequent stage of clustering, the grouping algorithm determines which clusters are paired according to this criterion: “The average distance between all [pairs] in the resulting cluster is made to be as small as possible” (Garson 2010, 7). During each stage of clustering, the grouping algorithm examines the fusion of every pair of existing clusters. A case will only be included in 30 SPSS/PASW Statistics 18.0, the quantitative data analysis package used in this dissertation, offers several different methods for clustering cases, including: nearest neighbor or single linkage, furthest neighbor or complete linkage, between-groups linkage or average linkage, within-groups linkage, centroid clustering, median clustering, and Ward’s method. Each linkage method uses different criterion for forming clusters and produces different cluster patterns. 59 a cluster if its average distance to the other cases in the cluster is minimal. In other words, a case will be considered for inclusion in a cluster when it is the most similar when measured against the average similarity of cases within the cluster. After the formation of new clusters, the average similarity within the cluster is computed. This process continues until one cluster remains. Given the mean distance between all cases within the cluster is minimized, “this method is therefore appropriate when the research purpose is homogeneity within clusters” (Garson 2010, 7). During the final phase of cluster analysis, the number of distinct cluster groups present in the data is determined. As mentioned previously, the process of hierarchical grouping continues until only one cluster is present. The researcher, therefore, must establish the stage at which clustering should be concluded. Examination of the agglomeration schedule has been cited as a commonly used method for identifying the appropriate number of clusters (Bruce and Witt 1971; Everitt 1993; Fleishman 1986; Garson 2010; Hill, Brennan and Wolman 1998). The agglomeration schedule displays how clusters were joined at each stage in the process. More importantly, it provides the agglomeration coefficient. This is an indicator of the distance, or dissimilarity, between the clusters being joined (Fleishman 1986; Garson 2010). 31 Changes in this coefficient over the various clustering stages measure the changes in the similarity within the cluster after the inclusion of a new cluster. Small increases in this value signify that the paired clusters are close in distance and similar. A large increase in the agglomeration coefficient, however, denotes the combination of two clusters that are dissimilar (Bruce and Witt 1971; Everitt 1993; Fleishman 1986; Garson 2010; Hill, Brennan and Wolman 1998) resulting in a 31 Hill, Brennan and Wolman (1988) define the agglomeration coefficient as “the sum of the within-group variance of the two clusters combined at each successive stage” (1942). 60 greater increase in the total variance (Hill, Brennan and Wolman 1998). At this point, none of the remaining clusters are similar enough to be merged. The number of clusters remaining unjoined equals the number of distinct clusters within the dataset. Hierarchical clustering is an agglomerative process. As such, it “begins with the same number of clusters as there are observations and proceeds to group similar observations together in a systematic fashion, until the final cluster contains all the observations…the number of stages in the process is one less than the number of observations” (Hill, Brennan and Wolman 1998, 1937). In this process, clusters are paired and combined until only one cluster remains. Once the cluster groups have been formed, however, they cannot be divided into smaller clusters. Cases are permanently assigned to the initial cluster (and effectively joined with other clusters according to the original cluster membership) and cannot be transferred to a more appropriate cluster at later stages in the process (Fleishman 1986). So, the erroneous designation of cases to clusters is not corrected throughout the process. Also, hierarchical clustering is sensitive to case order given the agglomerative nature of the process. To overcome both of these limitations, hierarchical cluster analysis was conducted two times with the cases sorted differently within the dataset to observe the number of dissimilar clusters identified by cluster analysis. 32 Consistency in these findings provided further confirmation for the proper number of cluster solutions. In sum, cities were sorted into homogenous cluster groups based on their inclusion of various organizations in the development decision-making process and the agency possessing 32 ICMA assigns a numerical identifier to each respondent city and includes it as a variable in the dataset, iMISD. The cluster analysis was conducted with the cases sorted by this variable in ascending order as well as descending order. 61 primary responsibility for the management of economic development. These cluster groups represent original, empirically-derived, structurally distinct local economic development decision-making networks. Cluster analysis effectively identified decision-making networks in which within-group variation in membership structure was minimized and across-group variation was maximized. The cluster analysis created a variable, cluster group membership that denotes the type of decision-making network leading local economic development policy in a city. 33 Each community is assigned to the cluster group for which the distance to the cluster mean is smallest. Once the cluster groups have been identified and variables are assigned to the cities to denote which group best fits their data, it is necessary to determine the nature of the cluster groups. Given the nature of the dataset utilized in this study, the resulting decision-making network types are essentially an aggregation of decision-making networks very similar in membership structure. It is important to acknowledge that the resulting cluster groups represent a type of local economic development decision-making network that includes a group of cities with governance networks whose membership structure shares a strong resemblance. Several similarly constructed decision-making networks were stacked to examine patterns in the composition of networks within each type. Each of the structural attributes analyzed in this study are based on a summation of data for each variable across cases assigned to the same cluster group. Descriptive statistical tests were conducted to examine the various structural attributes within the identified network types. 33 SPSS/PASW creates the cluster group membership variable for a range of cluster solutions. Once the number of dissimilar clusters was identified, the appropriate membership variable was used. 62 The results of these tests provide a frequency for each structural characteristic across networks within the same typological category. Central actors are those participants most involved and who displayed the highest rates of participation in the network. Total number of actors is a count of the central actors. Homogeneity measures the number of development actors with different organizational goals and development policy preferences. Network size is the sum of actors selected by the respondents as engaged in local economic development decision-making. To identify the “central actors”, or the prominence of individual actors, crosstabulation between the cluster group membership variable and the development actor variable was 34 implemented. The results were analyzed to determine the rate at which each organization was involved in the different types of decision-making networks. For each cluster group, this test provided the percentage of network areas in which the individual organization participated in the local economic development decision-making process. Variation in the actors’ rates indicated the degree to which organizations differed in their involvement within and across decisionmaking networks. Crosstabulation between cluster group membership and the primary responsibility variable was also conducted to examine leadership within the local economic development decision-making process. Results from this test indicate which organizations are most often charged with managing economic development activity within the different classes of decision-making networks. Higher rates of participation signify that the network member plays a stronger role in the decision-making processes of these particular communities and is more 34 Social network analysts use several different measures to examine the importance of individual network members. In sum, central actors are the most prominent participants because they have the largest number of ties to other network members (Knoke and Yang 2008; Wasserman and Faust 1994; Sandstrom and Carlsson 2008). 63 likely to affect the policy choices made by the network. “Total number of actors,” on the other hand, is a network-level structural construct that is employed to assess whether the governance process is centralized within a single actor or decentralized amongst a system of actors. 35 In effect, this construct is used to describe the dispersion of influence throughout the network type. The total number of central actors is the sum of actors with high rates of participation within each network type. A small number of central actors suggest that there are very few actors extensively involved in the decision-making network type. As such, influence is fairly centralized amongst a small set of development actors. A higher number of central actors represents a pluralistic governance structure in which there are several actors consistently engaged in the decision-making network type. As such, there are a larger number of development agents working to influence policy. Homogeneity within the network is derived from examining the extent to which network actors differ in their institutional organization and economic development goals. In other words, it is a measure of diversity amongst network actors and determines how many different types of development actors are participating in decision-making (see Table 2.4). The participation levels of each type of development actor will be operationalized using a continuous variable derived from the sum of actors within each category involved in development decision-making. For example, the variable “nonprofit organizations” ranges from 0 to 2 depending on whether 35 Social network analysts utilize the measure of “network centralization” to determine the extent to which the set of actors are centralized (Knoke and Yang 2008; Wasserman and Faust 1994). Network centralization calculates the overall level of prominence within the network by computing variance in individual network members’ centrality scores (Knoke and Yang 2008; Wasserman and Faust 1994). 64 private/community economic development foundations, colleges and universities, neither, or both participated in the process of forming policy. The size of development decision-making networks is measured using the average number of development actors selected as participants in the decision-making process. Each of the sixteen variables representing the involvement of various development actors was coded “1” if the specific organization was active in decision-making. The sum of these observations Table 3.4 Types of Local Economic Development Actors equaled the number of development actors involved in the development decision-making process. Crosstabulation between cluster group membership and a variable representing the total number of organizations involved in creating development policy was used to provide further 65 insight into how cities led by the same network type vary in the number of actors they included in the process of developing policy. Within the overall sample, communities incorporated an average of approximately six organizations in the process of developing policy. Findings also show that cities in the sample most often selected five development actors as involved in development decision-making. As shown in Figure 3.5, there was considerable variation in the number of development actors that sample cities included in the decision-making process. However, sample cities generally did not Figure 3.5 Total Number of Actors in Development Decision-Making Networks involve an especially large number of development actors in the decision-making process. Most development decision-making networks (81%) included between two and eight organizations. Fifty-four percent of the networks included two to five development actors. 66 After constructing membership profiles of the identified local economic development decision-making networks and analyzing their structural composition (centrality, density, homogeneity and size), I develop predictions regarding the complexity of the decision-making environment and each network’s policy use according to the conceptual framework presented in Chapter 2. The next step in the study is to examine the relationship between decision-making network type and policy use. In this study, I examine the use of development policies according to the categorization scheme that was used by ICMA on its questionnaires. The labeling of economic development policy areas and categorization of various strategies were consistent across surveys. For each type of policy area, multiple strategies were listed. When selecting economic development policies as functioning in their cities, respondents were most likely considering how the category had been labeled. The survey questioned local officials about their city’s use of over fifty development policies to improve the local economy and grouped them according to four areas of local economic development that the policy is intended to address: community development, small business development, business retention, and business incentives. The individual development policies included in each category are displayed in Table 3.5. The dependent variable in this study is the use of an economic development policy (community development, small business development, business retention, and business incentives). For each policy area, survey respondents were asked either “Which of the following activities does your local government conduct?” or “Which of the following [economic development strategies] does your local government offer?” For each question, there were 67 several specific strategies listed and respondents were instructed to “check all applicable.” 36 While these variables do not effectively measure the extent to which cities use these individual development strategies, they do indicate whether there is a policy that authorizes the city to implement the policy in an effort to address local economic development. First, crosstabulation was employed to examine the bivariate relationship between the decision-making network type and cities’ use of forty-four different economic development policies included on the ICMA survey instrument. The results of this test provided the Pearson chi-square test statistic. The chi-square was analyzed to determine whether the relationships between decision-making network type and policy use were statistically significant. Although there were four categories analyzed, this sample included a total of forty-four local economic development policies. A dichotomous variable (0=no; 1=yes) was created for each between 36 The ICMA dataset codes the selection of an economic development policy as a “1.” The dataset, however, does not provide an observation for cities that did not select the development policy. If the policy was not active in the community, it was not coded and left blank. Cities with missing data for all policies within a specific type were considered to have skipped the question or to be non-active users of policies within the category and were removed from the analysis. If the cities did select one or more of the policies, they were considered to have answered the question and, thus, included in the study. For these cities, the development policy variables with missing values were re-coded as “0” to indicate a response was provided and the policy was not operating in the community. As a result, the sample is limited to those participating cities who selected at least one active policy within each category. 68 decision-making network type. The purpose of this test was to establish which local economic development policies were used at considerably different rates by the various local economic Table 3.5 Local Economic Development Policies development decision-making networks. Those policies that were related to decision-making network type at a level of 0.05 or higher were subjected to further tests to ascertain how network type influenced cities’ development practices. 69 Table 3.5 (cont’d) Binomial logistic regression was used to uncover the strength and direction of the relationship between the membership structure of decision-making networks and policy outcomes. Those policies that identified as being significantly related to decision-making network type in the bivariate analysis were examined. While the Pearson chi-square test statistic provided by the crosstabulation indicates that decision-making network type and the use of various policies are not independent, the strength of these relationships remains unclear. To determine how influential decision-making network type and other theoretically relevant variables influence whether an economic development policy is likely to be operative within a city, logistic regression analysis was employed. In this study, the dependent variables, economic development policy use, are categorical. For this reason, ordinary linear regression is not appropriate for modeling the relationship of 70 interest, primarily because it cannot effectively predict the likelihood of an event either occurring or not occurring. More specifically, linear regression could potentially predict that the probability of the event occurring is greater than 1, which is impossible given that the probabilities are measured using only values between 0 and 1 (King 2007). The nature of these data also violates other fundamental assumptions of linear regression. Unlike linear regression, logistic regression assumes that the dependent variable is dichotomous, while not assuming that the independent variables are normally distributed and there is a linear relationship between the dependent and independent variable (King 2007). For this reason, logistic regression can be employed to predict the membership of a case in one of the two dependent variable categories according to the estimated impact of several independent variables. Logistic regression, essentially, estimates the odds of an event occurring (success) compared to the odds that the event will not occur (failure). In other words, this method is used to predict the probability that the dependent variable is a “1” rather a “0” (King 2007). This statistical method employs maximum-likelihood methods to calculate the best-fitting model. This model is considered to be the best fit because the probability of successfully sorting the observed data into the correct grouping (event occurring vs. event non occurring) is maximized given the independent variables included in the model (Myers et al. 2012). The outcome of the logistic regression, p, is not the value of the dependent variable but the probability that that the dependent variable is either 0 or 1. As such, the distribution may be skewed and requires a logistic transformation of p in order to normalize the distribution (King 2007). The logistic transformation of p, or logit (p), represents the log of the odds ratio that the event occurred (or that the dependent variable equaled 1) (Myers et al. 2012). This is important for measuring the impact of the independent variables, holding all other variables constant. 71 Logistic regression, like linear regression, provides a ‘B’ coefficient that represents the influence of the independent variable on variations within the dependent variable with all other things being equal. However, the ‘B’ coefficient in logistic regression calculates the changes in the log odds of the dependent variable occurring given the independent variable. Interpreting the log odds of an event occurring is quite complex. For this reason, the log odds are converted into an odds ratio, which is the estimated change in the likelihood of a case belonging to the dependent variable category coded as 1 for every one unit increase in the independent variable (King 2007; Myers et al. 2012). This ln (odds ratio) is represented as Exp (B) in the logistic regression output. Mason and Thomas (2010) provide a clear discussion on interpreting the Exp (B). If the odds ratio is larger than 1, then a change in the independent variable by one unit increases the likelihood (or odds) of policy use by 1 minus the odds ratio (Mason and Thomas 2010). An odds ratio that is less than 1 denotes a decrease in the likelihood of policy use by 1 minus the odds ratio for every one-unit change in the independent variable. A negative log odds (b coefficient) represents a decrease in the dependent variable by the odds ratio for a one-unit increase in the independent variable. Two logistic regression models are used when estimating the influence of decisionmaking network type and other variables on the likelihood of a policy being used by a local government. 37 37 Model 1 includes the various control variables. Model 2 represents the full This included two community development (community development corporations and job training programs), three small business development (small business development centers, management training, and microenterprise programs), four business retention (revolving loan fund programs, ombudsman programs, partnering with other nongovernmental organizations, 72 model that includes the independent variable, decision-making network type, and the control variables. Although Model 2 will be given full consideration, the estimations of both logistic regression models, specifically the likelihood ratio (or model chi-square), were examined to determine the strength of the model and estimations provided. If the test statistic is significant at a level of p < 0.05 or less, the null hypothesis is rejected, signifying that the model with the predictors varies significantly from the model with only the constant included. In other words, the estimated logit model fits the data better than the model in which all the predictor variables’ coefficients equal zero (King 2007). When independent variables are entered stepwise into the model, differences in the model chi-square are analyzed to determine improvements to the model. The likelihood ratio test, or difference between the model chi square (-2log-likelihood), is examined to determine the degree to which the added variable contributes to the strength of the model. Changes in likelihood ratio across the base and full models indicate whether the inclusion of the independent variable, decision-making network type, helps to better predict the dependent variable. This test will further confirm the importance of decision-making network type on policy use. Other methods utilized to measure the strength of the model in predicting the dependent variable were the Hosmer-Lemshow Test Statistic and Nagelkerke R2. To identify the factors that influence whether a municipality is likely to employ a local economic development policy, the logistic regression model is estimated as follows: and partnering with other local governments), and six business incentives (tax abatements, subsidized buildings, infrastructure improvements, locally designated enterprise zones, federal/state designated enterprise zones, and one-stop permit issuance). 73 where pi is the probability that the development policy is functioning in a community (or odds of the event occurring) and pi/(1- pi) is the odds ratio of a development policy being employed by local governments as opposed to the policy not being employed. N indicates the independent variable of primary interest to this study, the type of network leading local economic development decision-making. Cluster analysis was utilized to identify distinct groups of similar cases, or development decision-making networks, according to the individual public and private organizations participating in the development of economic development strategies (horizontal structure) and the organization primarily responsible for local economic development (hierarchical structure). During the cluster analysis process, a variable was created to assign cities to the decision-making network that best depicted the group of actors working in their communities to lead economic development, or the cluster group to which the city was closest. Scholars of local economic development have identified a variety of factors that contribute to local governments’ development policy decisions. Several control variables have been included in this study to account for their effects. Table 3.6 provides descriptions for the variables used in estimating the logit models, and their descriptive statistics are displayed in Table 3.8. In the logistic regression estimation formula, C represents a set of variables indicating citizen need, including unemployment and per capita personal income. In terms of citizen need, localities experiencing higher levels of poverty and unemployment tend to engage in increased levels of development and actively promote business growth (Rubin and Rubin 1987). Per capita income (Clingermayer and Feiock 1990) has also been shown to be related to development activity. Rubin and Rubin (1987) examine localities’ use of economic development incentives in Illinois and find that the poorest cities “pay more” as a result of their tendency to engage in expensive and potentially ineffective development incentives as a response to citizen need, a declining 74 Table 3.6 Description of Independent and Control Variables economy, or regional competition. This supports Peterson’s (1981) hypothesis that cities with a significant number of impoverished residents will tend to more actively promote economic development in an effort to improve the local economy and provide redistributive programs. F is a set a variables that measures fiscal stress. These include median housing value, real property tax rate, and economic condition. Several scholars argue that municipalities’ fiscal 75 stress is also an important determinant of development policy adoption (Fleischmann and Green 1991; Rubin and Rubin 1987; Schneider 1989; Swanstrom 1985). Low assessed property valuation per capita and high property tax rates provoke increased economic development policy activity (Fleischmann and Green 1991; Rubin and Rubin 1987) since both are politically unpopular. In an effort to satisfy residents, improve fiscal health, and expand the tax base, local governments will adopt more economic development policies (Fleischmann and Green 1991; Rubin and Rubin 1987; Schneider 1989; Sharp 1990). Declines in federal funds intended for local economic development in the 1980s forced cities with a heavy reliance on these funds to increase development activity (Clarke and Gaile 1992; Eisinger 1988; Rubin and Rubin 1987; Sharp 1990). Many local economies have failed to successfully adapt to the changing regional, national, and global economy. The jobs and investments lost as a result of the decline in manufacturing further spurs economic development policy implementation. S depicts the inclusion of two variables representing the size of a locality: population size and metropolitan status. A locality’s population size and metropolitan status influence its need to promote economic growth (Green and Fleischmann 1991). These two variables are indicators of a city’s place in the urban hierarchy (Fleischmann et al. 1992). Decreased population growth places pressure on local officials to promote development, attract residents and businesses, and expand the tax base (Rubin 1986; Rubin and Rubin 1987). This pressure, however, is decreased in localities experiencing rapid growth (Logan and Zhou 1990). Many theories of local economic development argue that the organization of local government, specifically political leadership, the role of local government, and bureaucratic Green 1991; Rubin 1989) and have devoted more bureaucratic staff to development efforts (Fleischmann et al. 1992; Reese 1997) tend to implement more development policies. Cities with 76 Table 3.7 Descriptive Statistics of Variables Included in the Logistic Regression 77 stronger administrative capacity tend to implement more development policies (Rubin and Rubin 1987). For this reason, P, the presence of an economic development plan, was included in the model. Another variable, T, which represents the city type, was also included to account for variations in the organization of local government. A variable, R, was also included to represent the geographic region in which the city was located. There is a set of variables denoted as L in the estimation formula that measures competition at various levels. In support of the arms race theory, scholars argue that regional competition shapes development policy. The interaction of federalism and capitalism makes the pursuit of economic development a competitive process (Bowman 1988). Localities tend to expand development efforts in response to increased development activity of neighboring areas (Bowman 1988; Schneider 1989). Regional competition, however, tends to have a more significant impact on central cities (Green and Fleischmann 1991). In this study, there are six variables measuring competition at various levels: nearby local governments, other local governments in the state, local governments in nearby states, other states, foreign countries, and other. For each form of competition, there is a dummy variable to indicate whether the city identified themselves as experiencing it or not. As mentioned earlier, crosstabulation was utilized to examine the bivariate relationship between decision-network type and active local economic development policies in cities. In addition to the Pearson chi-square, which displayed the strength of the bivariate relationship, the results also provided the proportion of cities assigned to each network type that selected the economic development policy as present in their communities. The crosstabulation provided the frequency (number and proportion) of respondents (cities) that possess the particular characteristic (policy use). These frequencies were utilized to rank the polices across network 78 types. The results were analyzed to develop a policy profile of the different network types based on the policies that were most and least actively used. In the final empirical chapter, I focus on uncovering the variables that affect the structure of the decision-making network. Scholars have yet to develop a theoretical explanation for why some environments foster the emergence of certain network types and not others. Because of the paucity of research in this area, I am not testing any particular theory of the potential predictors of decision-making networks. Instead, I conduct an exploratory analysis of variables that may increase (or decrease) the likelihood of each network type being present in a city. I investigate the effects of city demographics, local economy, land use, and political factors on network type. For each decision-making network type identified in Chapter 3, a binomial logistic regression model will be estimated as follows. 38 In this formula, pi represents the odds of the decision-making network type occurring in a city and pi/(1- pi) is the odds ratio of the network type guiding policy in a city as opposed to another network type being present. Table 3.8 shows a description and coding scheme for each independent variable. The descriptive statistics for each independent variable are displayed in Table 3.9. In the logistic regression formula, E denotes the inclusion of three variables that operationalize different economic factors. The first is economic condition, which was also included in the previous analysis and is a categorical variable used to measure the health of the 38 A dichotomous variable was created to represent whether the specific network type was present in that city (0 = No; 1 = Yes). 79 economy over the past five years. Cities had the option of selecting rapid expansion or decline, moderate growth or decline, slow growth or decline, or stable. Another variable, economic base, was included in the model to identify the primary industry operating in the communities. Respondents were also asked to identify which industry local economic development efforts were most focused on. Significant findings regarding the effects of these variables on the emergence of decision-making networks would imply that certain networks emerge to address specific economic conditions and industries. A set of demographic variables is represented by D in the logistic regression formula. This includes population size, region, and metropolitan status. P signifies three different variables intended to measure political characteristics: city type, mayor-council government, and council-manager government. The last group of variables, L, is all categorical and indicates the percentage of land devoted to a specific use. Respondents from sample cities identified the proportion of land used for non-residential purposes (including commercial, industrial and manufacturing use), residential use and vacant land. Although, each composed an individual variable in the model, the summation of the values should have equaled 100%. By further examining why network types emerge, this section of the study contributes to the development of theoretical explanations and testable hypotheses that facilitate understanding of why structurally distinct decision-making networks select different policies. If scholars can better understand what conditions lead to the presence of particular networks in cities, they would be more equipped to test if the membership structure of the network impacts policy. More specifically, researchers would be capable of determining if the structure leads the network to select policies that address issues that gave rise to its emergence. 80 Table 3.8 Description of Dependent and Independent Variables 3.5 Conclusion The purpose of this research is to test the effect of decision-making networks on local governments’ policy choices. Methodologically, the three-part quantitative analysis conducted 81 Table 3.9 Descriptive Statistics of Variables Included in the Logistic Regression 82 in this dissertation differs considerably from urban regime and policy network studies. The typology of local economic development decision-making networks is based on the statistical analysis of a large-N sample of cities. Each category within the typology can be utilized to describe the decision-making network within several cities. The increased generalizability of the typology facilitates the examination of patterns in policy use across groups of cities that are being guided by different types of decision-making networks. At this level of analysis, it is possible to quantify the effects of network type on the likelihood of a city’s policy use and strengthen the inferences drawn in this study. The exploratory analysis presented in Chapter 5 attempts to address an area of research that has been largely ignored by scholars. Urban regime theorists and policy network theorists have only offered some variables that could potentially influence the structure of the governance network in a community. The literature would be vastly improved by theoretical development regarding the social, economic and political factors that contribute to the formation of decisionmaking networks. The analysis in this study tests the effects of theoretically-relevant variables on the type of network in a city. These findings provide valuable insight into an important element of the governance process. The development of an original typology of structurally distinct decision-making networks supported the causal analyses presented in this dissertation. 83 CHAPTER 4: A TYPOLOGY OF LOCAL ECONOMIC DEVELOPMENT DECISION-MAKING NETWORKS A complete understanding of the relationship between governing structures and policy output requires an examination of the independent variable, local economic development decision-making networks. The analysis conducted in this chapter focuses on examining the membership structure of the decision-making networks that guide local economic development in a large sample of American cities. Three research questions are posed. First, are there distinct decision-making networks presiding over development in cities? Second, what are the structural differences in the membership of the various development decision-making networks? Lastly, who are the key development actors, organizations, and institutions participating in the decisionmaking process? The purpose of the analysis in this chapter is to construct and analyze the independent variable of primary interest to this dissertation, development decision-making networks. Hierarchical cluster analysis was conducted to identify groups of cities with similarly structured decision-making networks. The analysis included two sets of variables. The first set of variables was used to operationalize an individual organization’s involvement in the local economic development decision-making process. There were sixteen dichotomous variables used to represent the inclusion of city, county, state, and federal governments, as well as: private businesses and industries, chambers of commerce, economic development corporations, regional organizations, planning consortia, public/private partnerships, colleges or universities, utilities, private or community economic development foundations, citizen advisory boards or commissions, ad hoc citizen groups, or some other development actor. The second set of variables measured leadership within the decision-making network and was examined using 84 three dichotomous variables that indicated whether a local government, nonprofit development corporation, or some other organization possessed primary responsibility for economic development within local government. Thus, cities were clustered according to patterns in the horizontal (“who participates”) structure and the vertical (“who leads”) structure of the decisionmaking network. By accounting for varying policy resources and goals present in the network, one is capable of determining the potential for collaboration and conflict in decision-making. Descriptive statistical analyses were conducted to analyze each network and classify them according to distinct patterns in their membership structure. Central actors were identified to determine who was most involved in the decision-making process and their policy goals. The total number of central actors, actor homogeneity, and size were measured to assess the extent to which central actors would be capable of shaping the network’s policy decisions. Policy profiles for each local economic development decision-making network will be developed according to the structural composition of the network. 4.1 Findings The criterion used to determine the number of cluster groups was a strikingly evident increase in the distance between the clusters, which was measured using the agglomeration coefficient as the clusters were joined during each stage. A substantial change in the coefficient indicates an increase in the dissimilarity between clusters. The differences in the agglomeration coefficient 39 for the last nine stages were 0.007, 0.011, 0.006, 0.021, 0.015, 0.024, 0.038. and 0.042. The agglomerative hierarchal clustering method involved 522 stages of clustering and the 39 Refer to Appendix 4.1: Partial Agglomeration Schedule from Hierarchical Cluster Analysis 85 dissimilarity between clusters increased substantially between clustering stages 520 and 521. The results indicate a three-cluster solution. Analysis of the networks’ membership structure led to the establishment of a typology that included three structurally distinct local economic development decision-making networks: local government-chamber of commerce, municipal, and broad public/private partnerships. The local government-chamber of commerce network type was the most common and present in 54% of the sample cities. The other two network types were prevalent at similar levels. Municipal networks guided the development decision-making in 24% of cities while broad public/private partnerships governed development policy in 23% of cities. 4.2 Local Government-Chamber of Commerce Decision-Making Networks Several development actors were frequent members of this network type with eleven development actors participating in at least 40% of the local government-chamber of commerce networks (see Table 3.1). However, two development actors (city governments and chambers of commerce) displayed especially high rates of involvement. City governments led in participation with a rate of 0.971 indicating that they were contributing to almost all of the decision-making processes of communities in which these networks were present. Chambers of commerce were also quite active with a participation rate of 0.929. There were no other development actors that engaged at this level. The difference between the participation rates of the chambers of commerce and the next most involved actor was 0.40. Most development actors displayed moderate participation rates between 0.55 and 0.30. County governments were the third most active organization in this network with a participation rate of 0.532. Private business and industry (0.475), regional organizations (0.450), and public/private partnerships 86 (0.446) contributed to these particular decision-making networks at comparable rates. State governments and citizen advisory boards or commissions were equally engaged with a rate of 0.421. Other contributing network members include economic development corporations Table 4.1 Local Government-Chamber of Commerce Networks: Central Actors 87 (0.382), colleges and universities (0.346), and utilities (0.304). There were four organizations that displayed low involvement rates between 0.150 and 0.120, including the federal government (0.150), planning consortia (0.139), ad hoc citizen groups (0.129), and private or community economic development foundations (0.121). “Other” development actors (0.075) not listed on the survey were the only development actors to be involved in this class of decision-making networks at a rate below 0.10. Primary responsibility for economic development was a bit more dispersed across various organizations. City governments directed development efforts in 88% of cities led by local government-chamber of commerce decision-making networks. Other organizations not listed on the survey were fairly active in this leadership role and supervised economic development activities in 11% of these areas. Only one city (0.8%) in which local government-chambers of commerce networks were present granted management of economic development to a nonprofit development corporation. Based on participation rates, the central actors within this network type are local governments and chambers of commerce. Both actors displayed the highest rates of involvement within local government-chambers of commerce decision-making networks. These two organizations are consistently transmitting policy information throughout various local government-chambers of commerce networks. This gives them a considerable amount of influence. Nine other development actors were influential but at considerably lesser levels. Each of these organizations were present in less than half of the identified local governmentchamber of commerce networks. As a result of their diminished levels of engagement with decision-making networks of this type, these less central actors have fewer opportunities to shape local economic development policy choices and challenge the efforts of the central actors. 88 There is a small number of highly central network members exerting influence, yet there are several participating development actors stationed in moderately central positions. The relatively extensive range of actors involved in local government-chamber of commerce decision-making networks significantly increases the complexity of decision-making. There are several development actors with substantial influence and varying policy goals. The group of highly and moderately central actors is considerably diverse in their organizational goals. The two most central actors illustrate a strong governmental-business collaborative. The set of actors in moderately influential structural positions further supports this. There are several organizations representing sub-national governmental departments (county government and state government) and business-related entities (private business and industries and economic development corporations), but there are also citizen groups (citizen advisory boards and commissions) as well as public-private partnerships (public-private partnerships and regional organizations). There is substantial potential for these different types of development actors to possess different objectives for economic development policy. The constraints presented by such a diverse set of development actors are exacerbated by the moderate size of local government-chamber of commerce decision-making networks. The sizes of these networks are displayed in Table 4.2. Local government-chamber of commerce networks include two to fourteen organizations. None of the cities assigned to this network type identified a single actor as the sole participant in development decision-making and few (4.6%) chose two organizations. A substantial proportion of the networks were small to medium in size and involved three to seven individual actors. These particular sized networks comprised 66% of local government-chamber of commerce networks. Many of these particular networks (20%) had between eight and ten organization types. Several (17%) of the local 89 government-chambers of commerce networks were composed of nine or more organizations. The mean size of the set of decision-making networks was six development actor types. Only a small percentage of these networks were especially small or large. The findings indicate that local government-chambers of commerce decision-making networks were moderate in size. Table 4.2 Local Government-Chamber of Commerce Networks: Network Size 90 Overall, local government-chamber of commerce networks involve many central and moderately central actors. Local governments and chambers of commerce were the most influential, yet there were several other actors with different development policy preferences that were in lesser positions of power. The high levels of actor heterogeneity and larger network size may impede goal consensus and make it difficult for the most central actors to overcome contrasting policy goals. The variety of interests competing within the network should lead to an especially broad set of local economic development policies. 4.3 Municipal Decision-Making Networks The most interesting attribute of this class of networks is the extent to which most development actors failed to display prominence within the network. As displayed in Table 4.3, city governments were the most prevalent organization type and engaged in 91% of the municipal decision-making networks. It was the only organization to be engaged at levels above 0.500. Citizen advisory boards and commissions involved at a rate of 0.390 exhibited the next largest participation rate. All of the other development actors displayed participation rates below 0.25, including regional organizations, economic development corporations, organizations representing private business or industry, public/private partnerships, county governments, state governments, and ad hoc citizen groups. Half of the development actors, organizations, and institutions listed on the survey were included in less than 10% of the municipal networks. In conjunction with being the only development actor consistently active in municipal networks, city governments were also principally responsible for directing economic development activities in 96% of cities empirically assigned to this category of networks. Nonprofit development 91 corporations (1%) and other organizations not listed on the survey (3%) managed development policy in a very small set of these particular communities. Table 4.3 Municipal Networks: Central Actors Within municipal decision-making networks, there were no development actors that matched city governments in their rates of participation (how often they were involved) or in primary responsibility for development activities within the city. With this network type, there is 92 a single central actor managing local economic development policy. Due to their strong presence in the municipal network type, local governments’ policy goals would be the most dominant. For the most part, city officials unilaterally configure policy with little input from other development actors. Decision-making power is largely centralized within the only consistently influential development actor, local government. Citizen advisory boards and commissions are the only other actors to present medium levels of participation, which diminishes their capacity to affect the policy decisions of municipal networks. Influence over policy decisions is not equally distributed throughout the network type; decision-making power is concentrated in a single actor. In general, municipal networks are characterized by local governments dominating the decision-making process. The range of actors, homogeneity and inclusion, involved in the typical municipal networks is small to medium. Given that there is a single central actor, homogeneity is high amongst the most influential development actors. However, the set of actors participating in less than 25% of various municipal networks is quite diverse and includes citizen advisory boards and commissions, regional organizations, and economic development corporations. According to the findings displayed in Table 4.4, municipal decision-making networks are not significantly inclusive. The sizes of these specific networks varied between one and seven actors. On average, networks of this particular type involved approximately three types of participants. 40 An overwhelmingly large proportion of these particular networks (76%) included one to three organization types in the process of creating development policy. Twenty- 40 The exact mean number of development actors included in municipal development decision- making networks was 2.64. 93 five percent of these communities selected a single development actor as contributing to decision- making. Most of the cities assigned to this network type (34%) selected two organizations as involved in the process. Only 2% of these cities were led by networks that included more than six development actors. Table 4.4 Municipal Networks: Network Size 94 In sum, municipal decision-making networks are characterized by a strong central actor (local government) managing local economic development policy. With the exception of citizen advisory boards and councils, other development actors were not significantly present in municipal networks and displayed participation rates ranging between 16% and 24%. The generally small size of the municipal networks imply that a small number of organizations regularly engage in each municipal decision-making network. Small network size would be conducive to bargaining and overcoming conflict amongst different development actors. The low participation rates displayed by other actors suggest that development actors outside of city government are thinly dispersed across the various municipal networks. Without any strong competing interests influencing network members and policy decisions, local government officials and departments should have strong control over local economic development decisionmaking. 4.4 Broad Public/Private Partnership Decision-Making Networks The distinctive structural trait of these decision-making networks is the high level of participation and centrality demonstrated by a large number of development actors. The findings of the descriptive statistical tests are presented in Table 4.5. Within this category of decisionmaking networks, four organizations presented participation rates above 0.700. City governments were the most engaged in this network type, as their participation rate of 0.983 indicates that municipalities were active in the development decision-making process of almost all of the communities in which broad public-private partnership networks were identified. The next most involved organization, economic development corporations, displayed a participation rate of 0.867. Chambers of commerce joined in the process of selecting economic development 95 policy in broad public-private partnerships at a rate of 0.825. County governments were another highly involved actor and contributed to development decision-making at a rate of 0.700. Table 4.5 Broad Public Private Partnerships: Central Actors There were six organizations that functioned at moderate rates within broad publicprivate partnerships, including: private businesses and industries, colleges and universities, utilities, public/private partnerships, state government, and regional organizations. These 96 development actors were involved at rates between 0.600 and 0.300. Individuals and organizations representing private businesses and industries were the most engaged development actors within this subset of network members with a participation rate of 0.567. They were followed by colleges and universities (0.483) and utilities (0.475). Public and private partnerships and state governments shared similar rates of involvement at 0.383 and 0.367. Regional organizations joined in broad public-private partnership decision-making networks at a rate of 0.358. There was a large disparity in the participation rates of this group of moderately involved development actors and the six remaining organizations. Citizen advisory boards and commissions were next and displayed a participation rate of 0.250. Five development actors (private or community economic development foundations, planning consortia, federal government, ad hoc citizen groups, and other actors) participated in this type of decision-making network at low rates of 0.150 or below. Despite the predominance of city governments within the horizontal structure of these networks, they did not possess primary responsibility for economic development within any of the communities in which broad public-private partnerships were managing development policy. Instead, nonprofit development corporations directed development in 91% of these cities while some other organization not listed on the survey managed local economic development in the remaining areas. This type of organization, however, did not possess a strong central position within the horizontal network structure. This implies that the role of nonprofit development corporations maybe limited to managing local economic development activity within these cities. Despite the diminished levels of participation within the decision-making process, nonprofit development corporations do enjoy some degree of influence by supervising and 97 regulating local economic development projects. This is the only network type in which the development actor primarily responsible for overseeing the policy area was not centrally located within the horizontal structure of the network. Analysis of the total number of central actors reflected a decentralized network. Within the broad public-private partnership network type, there are numerous development actors in structurally influential network positions. Power in broad public-private partnership networks is more equally dispersed and several actors are potentially circulating policy ideas throughout the network. The policy interests of these development actors vary considerably. The group of the most central actors (city governments, economic development corporations, chambers of commerce, and county governments) represents a strong collaborative effort amongst regional governmental entities and market actors. The slightly less influential actors are also diverse including private businesses and industries, colleges and universities, and utilities. The diverse set of development actors may complicate the decision-making process and make it difficult to generate a consensus on policy choices. The high levels of participation displayed by development actors coincides with the larger size of the broad public-private partnership networks. The number of participants ranged from two to thirteen organizations. A substantial proportion of these communities (60%) included five to eight organizations in their development decision-making process. Twenty percent of the broad public-private partnership networks were smaller than this range and 20% of the networks were larger. None of the broad public-private partnerships were characterized by an independent organization having unilateral control over development decision-making. On 98 average, about seven development actors participated in developing policy to address economic 41 issues . Table 4.6 Various Sizes of Broad Public-Private Partnership Decision-Making Networks 41 The average size of broad public-private partnership decision-making networks is 6.66 development organizations. 99 Broad public-private partnership networks are described as having many actors with varying economic development policy priorities. The large number of central actors decreases 42 the centralization of power and decreases the potential for negotiation and collaboration . Heterogeneity amongst network members further complicates the decision-making process. Several central actors with different policy preferences may transmit conflicting goals throughout the network. If several network members do not share similar policy positions, it may be difficult to overcome conflict and disagreement on the policy goals of the network. Goal consensus is likely to be greatly diminished and gridlock could occur. The efforts of any central actor attempting to appeal to other network members, gain policy support, and sway the network to select their favored policy will be hindered by a large number of other influential actors with different policy interests. 4.5 Predicted Policy Profiles of Local Economic Decision-Making Networks The results of the hierarchical cluster analysis indicate that the three empirically-derived local economic development decision-making networks are structurally distinct with regard to the organization of their membership. Significant variations in the structural composition of the 42 Provan and Kenis state that: “Under such conditions, governance becomes extremely complex. Shared self-governance…is best suited for small networks of organizations. When problems arise in such networks, full and active face-to-face participation by partners is possible. As the number of organizations in the network gets larger, however, shared governance becomes highly inefficient, with participants either ignoring critical network issues or spending large amounts of time trying to coordinate across 10, 20, or more organizations.” (2007, 238) 100 networks, including central actors involved, homogeneity, and size, indicate potential for these decision-making networks to promote distinct policy agendas with different strategies for addressing development in their communities. Summaries of the structural characteristics of each network are represented and each typological classification is shown in Table 3.7. The structure of municipal decision-making networks should simplify the decision-making process given the high closure and low range of actors. Higher actor heterogeneity, on the other hand, complicates the transmission of policy ideas, which limits central actors’ ability to effectively promote development policies favoring their interests. As mentioned in Chapter 1, predictions about the policy agendas of each decision-making network can be made only when the network structure is analyzed in conjunction with various policy resources and goals. In this section, I develop policy profiles for each decision-making network type based on central actors’ preferred policy strategies as well as the structural constraints they face when attempting to direct the decision-making process. In the local government-chambers of commerce networks, there are several structural constraints the most prominent network members face; city government and chamber of commerce have several other participants likely to promote conflicting policy interests. Three different levels of government (city, county, and state) and three organizations representing business interests (chambers or commerce, private businesses and industries, and economic development corporations) constitute a large proportion of the total centrality in the network. These strong government-business interactions resemble the corporate (Stone 1989), progrowth (Sites 1997), and entrepreneurial (Elkin 1987), which promote largely business-favored policies. There are, however, a few key differences. 101 The first important divergence from the traditional corporate regime typology is the moderately strong presence of other levels of government, including county and state governmental organizations. Both levels of government have been generally ignored in the urban regime literature with regard to their participation in local economic development decision- making and the policies they tend to support. Scholars have mentioned, although not discussed extensively, that intergovernmental collaboration provides cities with more management resources as well as monetary assistance (Agranoff and McGuire 1998; Burns 2002). According to Agranoff and McGuire (1998), “counties often offer special tax revenue for economic development as well as other resources such as planning and zoning assistance, administration of small state programs, and general policy-making information” (154-155). State governments, on the other hand, can contribute financial resources, and by creating new revenue streams can allocate funds to support municipal bonds (Burns 2002). The intergovernmental financial aid decreases local government’s reliance on private entities for market resources. Support in managing the resources and development projects further advances city government’s pursuit to support the interests of their residents, the electorate. Similar to the corporate regime type, the local government-chambers of commerce network category includes several business-related organizations that exert high and medium levels of influence throughout the network. However, it was the local chambers of commerce that led in promoting the interests of the business community within the decision-makingprocess. As an association of local business leaders, the chambers of commerce generally do not provide financial resources, but assist in the formation of business-friendly policies (Agranoff and McGuire 1998). Their central position in the network implies that there are many policy ideas being transmitted throughout the network that concentrate on progrowth development 102 Table 4.7 Summary of Decision-Making Networks Membership Structure strategies, including business attraction and retention. Business leaders support policies focusing on improving the market by city government providing subsidized loans, tax exemptions, incentives and direct payments that benefit business interests (often downtown) (Logan and Molotch 1987; Sites 1997; Stone 1989; Stone 1992;). Imbroscio (1998) states that regimes in which the petty bourgeois were instrumental would possess conservative development policy preferences. These business interests will only support community development strategies (i.e. job-training and workforce development) as long as they directly relate to their needs (Orr and Stoker 1994, 59). Leaders of private businesses and industries, as well as economic development corporations, were also in fairly influential positions, which further strengthened the power of business-related interests within this network type. The latter development actor, which generally involves key local business leaders in leadership positions, assists in managing the 103 flow of resources and providing information (Agranoff and McGuire 1998). Thus, businessrelated organizations were strongly represented in these local government-chambers of commerce networks. Their impact on the development of potential policy options implies that these networks have an increased likelihood of pursuing policy that decrease businesses’ operating costs. The local government-chamber of commerce decision-making network does resemble a strong collaborative between city governments and business-related entities. Stone (1989) made it clear with his analysis of local decision-making in Atlanta that local government will promote business interests in an effort to secure market resources necessary to promote the local economy. On the other hand, there are also several other private, yet non-business related, organizations that were exerting moderate levels of influence in local government-chamber of commerce decision-making networks. These organizations, however, are not likely to be inclined to support corporate interests when selecting local economic development policy. Instead, citizen advisory boards and commissions as well as colleges and universities have interests in promoting community-based economic development strategies. Both development actors have strong social, economic, and political ties to the neighborhoods in which they operate. The first development actor is composed of residents and promotes the interests of citizens rather than business leaders. Higher education institutions often contribute a significant amount of resources that “facilitate community and civic revitalization, stimulate the physical revitalization of distressed areas, and reduce environmental stress, while simultaneously improving the economic and social well-being of the community” (LaMore, Blackmond, and Link 2006, 430). Both of these development actors benefit directly from improvements to the 104 social and built environment of neighborhoods within their vicinity. As such, they often take a strong progressive stance on local economic development and support policies that promote the development of social and human capital. The other moderately central actors included public private-partnerships (public-private partnerships and regional organizations), utilities, and private or community economic development foundations. These organizations generally play a more project-based role in local economic development decision-making and do not display an explicit policy position. Utilities often provide financial and informational resources to help with business start-up costs as well as technical assistance (Agranoff and McGuire 1998). Private economic development foundations most often contribute resources toward administrative and management efforts related to coordinating the exchange of interorganizational resources (Agranoff and McGuire 1998). These organizations, however, do sometimes offer fiscal resources to those projects involving physical development of community centers and other community-oriented programs (Agranoff and McGuire 1998). Although there is a strong government-business dynamic operating within local government-chambers of commerce decision-making networks, there are also several other actors with compelling interests to promote community-based development initiatives. With other levels of government, citizen advisory boards or commissions, and higher education institutions potentially supplying resources necessary for the governance of local economic development, local governmental leaders should be less reliant on local business, thereby reducing the influence of business interests. Based on the structural composition of local government-chamber of commerce networks, it is expected that this network may experience medium to high levels of conflict when 105 making decisions about local economic development policy use. According to the hypotheses developed in Chapter 1, the decentralization of power and high traffic of policy ideas should make it difficult to manage the decision-making process. The network’s ability to govern is further complicated by the large range of actors involved. More specifically, there are several actors with conflicting organizational goals and policy preferences struggling to shape policy. Given the number of central actors and diversity of interests, the local government-chamber of commerce decision-making network may have difficulty developing a consensus over the network’s policy objectives. This would increase the number of policies considered and utilized to address economic issues. Local government-chambers of commerce networks are expected to rely on a fairly large set of local economic development policies. More specifically, central actors, particularly city governments and chambers of commerce, may experience more adversity when attempting to direct the activities of the network and control the governance process. As such, those cities led by decision-making networks of this type are expected to primarily promote business-friendly strategies and strongly support policies that focus on market-based development policies. To satisfy the interests of moderately influential actors, the set of local economic development policies utilized should include some policies that support community-based initiatives. This set of policies, however, would be limited to development strategies that directly benefit businesses (i.e. job training and other employment programs) while simultaneously improving the social and economic conditions of distressed communities. Given the extreme complexity of the membership structure within the broad publicprivate partnership network type, it is possible that a wide variety of development policies are considered and selected. However, there is a very strong local government-business relationship 106 present in these networks. The five most central network members include city governments, economic development corporations, chambers of commerce, county governments, and private businesses and industries. Like the local government-chamber of commerce network, the interactions between the local and county government and various business-friendly organizations strongly resemble a corporate (Stone 1989), progrowth (Sites 1997), and entrepreneurial (Elkin 1987) development regime. In fact, this relationship is stronger because all of these network members present very high levels of centrality. Although city and county governments are capable of pursuing development policies that promote public ownership of various for-profit facilities, there are several very influential development actors promoting conservative pro-business development strategies and providing a substantial amount of development resources toward these initiatives. These actors would favor economic development policies that focus on attracting and retaining businesses by decreasing operating costs, specifically through the provision of incentives (Stone 1989,1992; Logan and Molotch 1987; Sites 1997; Olberding 2002). Given the business community’s presence within the decision-making network, local- and county-level governmental officials would consider their interests. This results in a business-friendly local economic development policy agenda. There are, however, two non-profit organizations that also possess strong structural positions within the network. Nonprofit development corporations and colleges/universities were the sixth most central network member with a participation rate of 50% and constituted 8% of the centrality within the network. Higher education institutions generally have a sociallyresponsible and progressive view of local economic development and supply communities with a considerable amount of resources to address revitalization of the built environment and social capital while improving residents’ quality of life (LaMore, Blackmond, and Link 2006). 107 Nonprofit development corporations, however, possess the strongest structural position within the network. These organizations were primarily responsible for local economic development activity within 91% of the cities led by broad public-private partnership networks. Most often, nonprofit development corporations are community-based and focus on neighborhood improvement and development (Johnson 2004; Robinson 1996), particularly in low-income communities. As agents of community revitalization, the mission of nonprofit development corporations tends to concentrate on providing quality and affordable housing, increasing homeownership, job-training, as well as socially, economically, and politically empowering community residents. Leadership of these organizations often includes “a mix of neighborhood residents, the business clients and tenants who rent from the CDC in question, grass-roots activists, clergy, social workers and local businesspersons” (Robinson 1996, 1652). Hula, Jackson and Orr (1997) describe governing nonprofits as organizations that take on traditional governmental roles and responsibilities in an effort to restructure the urban political process. This is done by coordinating broad collective interests, including underrepresented groups, into the policymaking process. Hula, Jackson, and Orr (1997) state that “governing coalitions are not extensively focused on the production of a specific product or service as are traditional nonprofits, nor do they establish a narrow strategy…Rather their goal and missions are framed on very broad social and political issues” (460). The structural position of nonprofit development corporations within the network suggests that the organization with primary responsibility for development activity promotes an open and flexible policy agenda with a particular interest in addressing societal issues, especially those affecting economically vulnerable communities. 108 The hierarchical position of nonprofit development corporations may insert support for progressive development policies that focus on community empowerment within the network. Their policy positions would likely receive overwhelming support from the many private, yet non-business, organizations moderately central to the decision-making process. Their influential position, however, does not necessarily supersede the influence of various business-related actors within the horizontal network structure of broad public-private partnerships. Business leaders do have an interest in the development of a quality labor force within the community. Rather than promoting community-based development, however, they prefer market-based development strategies. The latter set of local economic development policies provides substantial and direct benefits for business communities. However, they may “be willing to support only those [human] [capital] schemes that develop skills in potential employees that are directly related to their needs" (Orr and Stoker 1994, 59). Business leaders tend to support those policies that socially and economically empower residents but also benefit the businesses’ financial interests. The role of nonprofit development corporations as coordinators of development activity (and other progressive actors) suggests that broad public-private partnership decision-making networks involve an overall community-based approach to development. However, scholars of urban politics, including regime theorists, have criticized nonprofit development corporations, specifically community development corporations, for lacking the organizational capacity to administer community-based economic development projects (Fredericksen and London 2002; Stone 1989; 1991). This suggests that nonprofit development corporations may not have the resources necessary to single-handedly and effectively overcome businesses’ resources and interests while attempting to manage development activity and substantively represent community-based interests. Thus, these networks are not expected to act in the same manner as 109 Sites’ (1997) progressive regimes, which seek to limit business interests, particularly the expansion of downtown. The strong presence of business interests within decision-making may diminish constraints presented by the nonprofit development corporations’ leadership position and the influence of other progressive development actors. For this reason, broad public-private decision-making networks are expected to favor progressive and community-based development policies that are sensitive to business economic concerns. The structural composition of broad public-private partnership networks greatly increases the variety of policies being considered by the network. This network type is predicted to promote economic development through the use of many development policies. The municipal decision-making networks are considerably different. Local governments are responsible for a very large task: providing a viable local economy and ensuring a high quality of life for residents (Tiebout 1956, Peterson 1987). It is in the interests of city governments to increase the tax base by increasing investment within their geographic jurisdiction (Peterson 1987; Stone1989, 1991; Dowding 2001). Theories of urban governance make it clear that cities partner with private organizations to gain access to a cache of resources they would not otherwise be capable of securing but are required for effective governance of the community. As a result, private entities gain access to the decision-making process and potentially the ability to influence policy choices. However, city governments are often the only actor or one of very few actors contributing to the process of selecting development policy in cities led by municipal decisionmaking networks. The low participation rates of other actors indicate that beyond municipalities, other network members’ participation varied considerably in each city. This suggests that local governments’ partners change according to either the characteristics of the city, availability of 110 development actors, or development actors’ support for the local government’s policy agenda. The government’s diminished dependency on private resources and its role as the organization primarily responsible for development activity further solidifies city governments’ power to select policies. The municipal decision-making network strongly resembles Imbroscio's hypothetical local-statist regime. In these regimes local governmental officials have a diminished demand for a collaborative including non-state actors, because they are capable of effectively governing development without an overwhelming need to acquire private resources (Imbroscio 1998). Despite the limited collaboration with private actors, municipalities still have a strong desire to improve the local economy. There are several avenues municipalities can use to pursue economic development ventures that do not require private resources. Municipalities may create opportunities for local economic growth by concentrating on public property ownership and public profit-making, including development strategies that focus on the procurement and expansion of “airports, sports stadiums, public utilities, convention centers, and mass transit systems” (Imbroscio 1998, 238). The second most central actor within the network, citizen advisory boards and commissions, possessed a fairly weak structural position. Although they had a diminished capacity to constrain the influence of city governments, their involvement presents an interesting decision-making environment. These development actors generally share similar development goals as city government does. Both are broadly focused on supporting the economic empowerment of residents. Although the municipal decision-making network type strongly resembles Imbroscio’s local-statist regime type, there are also remnants of Imbroscio’s (1998) theoretical community-based regime present. This regime type includes neighborhood-based 111 organizations that support progressive local economic development policies which divert attention from downtown expansion and focus on policies supporting community-oriented development (Imbroscio 1998). The presence of these organizations would increase the decisionmaking networks’ likelihood of promoting the use of community development foundations and loan funds, workforce development, and affordable housing policies. This regime type suggests two important points. First, there are governance structures with potentially progressive development policy ideas. A strong criticism of Stone’s original depiction of regime theory was that it was preoccupied with corporate regimes and ignored the normative implications of alternative regimes types and development actors (Imbroscio 1998, 2003; Davies 2002). The existence of the municipal decision-making network type gives some empirical support to Imbroscio’s theoretical argument for progressive regimes that act as a caretaker of citizen-interests and focus on service delivery and decreased taxes (Mossberger and Stoker 2001). Second, there are instances of an absence of viable business leadership in local economic development activity. A strong government-business collaborative is avoidable in development policy-making (Orr and Stoker 1994). Originally, regime theorists claimed that business interests are favored in development policy-making. The distribution of market resources naturally provides a tremendous bias toward business elites who have the capacity to leverage market resources (Elkin 1987; Stone 1980, 1989; 1993; Dowding 2001). For this reason, local business elites are, naturally, the most appealing development policy partner in capitalist societies. As a result, development policy is predisposed to corporate-centered and market-based development strategies favoring the interests of the business elite, diminishing the likelihood of viable alternative regime types emerging (Elkin 1987; Stone 1989, 1993). The existence of the 112 municipal decision-making network type, however, supports critics’ claims that the unwavering superiority of business in local economic development policymaking was unrealistic (Imbroscio 1998a, 1998b, 2003, 2004; Davies 2002; Sites 1997). These findings also support the arguments presented by Hanson et al. (2010) that there have been shifts in corporate civic leadership within local governance. In other words, remnants of the corporate regime that was originally emphasized by Stone are not present in all decision-making environments. The strong centrality of city governments as well as the inclusion of citizen advisory boards and commissions in the locale creates an interesting policy situation. These networks are expected to have an increased likelihood of employing economic development strategies that concentrate on public ownership of for-profit institutions, particularly those that directly benefit residents and economically empower communities. Municipal networks may also support the provision of business incentives and subsidies that do not require substantial private resources. These development policies tend to compel local governments to sacrifice tax revenues or expand services. If the municipal network is considerably influenced by the citizen-based organization in a moderately central position, it should display the policy preferences of a caretaker and select development policies that promote improved service delivery and decreased taxes. More generally, the municipal decision-making networks are expected to use a smaller set of local economic development policies given their diminished access to private resources. 4.6 Conclusion In this chapter, I conducted hierarchical cluster analysis to identify different groups of cities that share similar sets of actors and organizations involved in the local economic development decision-making process. The resulting typology included three structurally distinct decision-making networks: municipal, local government-chamber of commerce, and 113 broad public-private partnership. The local government-chamber of commerce network type was the most prevalent and guided decision-making in approximately half of the sample cities. Given the strong government-business collaborative, these networks are expected to pursue development strategies that directly benefit the business community. However, there are several other actors in moderately central positions that strongly support initiatives intended to empower communities and residents. To accommodate these interests, local government-chambers of commerce network types will focus primarily on business-friendly policies and pursue only those community-based development strategies that offer some economic advantage to the business community. Similarly, broad public-private partnerships also center on government-business collaboration. However, county governments are also in strong central positions and nonprofit development corporations are primarily responsible for managing local economic development in these cities. Also, there are many development actors in moderately central positions with interests in promoting policies that focus on revitalizing social capital and offering economic opportunities in distressed communities. These decision-making networks are expected to draw on a broad set of development policies but concentrate on those community-based policies that include public ownership and business-friendly strategies. On the other hand, municipal decision-making networks are characterized by a single central actor, city governments, which coordinate development activities amongst a small network of actors. With decreased reliance on private resources, these networks are more likely to promote development policies that city governments can formulate and implement without any collaborative efforts. This includes public ownership of profit-making facilities and business subsidies. Decision-making networks of this type may also present caretaker-like tendencies in 114 their policy choices and concentrate their efforts on the provision of governmental services and maintaining low tax levels. The typology presented in this chapter includes decision-making networks with varying and extremely complex membership structures. Although each network type slightly resembled some of the different hypothetical urban regime and local economic development policy network types that have been previously developed by scholars, the typology presented in this dissertation offers new insights into the various governance structures in American cities. By analyzing both the individual- and network-level structural composition of development decision-making networks, I provide a more extensive account for differences across these local governance structures. Analysis of the identified decision-network types indicates that the structural composition of local economic development governance structures range considerably across network type. The findings in this chapter indicate that typologies of local economic development governing “coalitions,” “regimes,” or “networks” need to be expanded, and the evaluations of local economic development activities need to be more sensitive to the complex structural variations in decision-making networks’ membership. The typology developed in this chapter fills a gap in the literature and enhances the understanding of local governance by providing a more thorough analysis of the public and private actors’ structural positions in the decision-making network. This typology offers an opportunity to further examine governance within different decision-making environments. This classification scheme is conducive for gaining more knowledge regarding the involvement and effects of development actors on the policies cities select to improve local economy, specifically actors previously disregarded by scholars, including private, yet non-business, organizations, as well as other governmental entities, particularly county and state government. 115 This typology of local economic development decision-making networks differs from previous urban regime and policy network studies given that it is derived from a large-N quantitative analysis. These decision-making network types represent strong patterns within the governance of many American localities. Identifying this consistency in network structures across cities facilitates statistical analysis of the relationship between governing structures and policy output within a large set of cities while controlling for other variables. While previous studies mostly relied on single-case study analysis, the findings in this dissertation are generalizable to a larger set of cities. It is, however, important to confirm whether “the clusters are ‘real’ or merely artifacts of the [clustering] algorithms” (Everitt and Hothorn 2006, 254) and if the decision-making networks utilized in this study are valuable for understanding governance and policy decisions. In the next chapter, I test the effect of network type on cities’ use of various local economic development policies using logistic regression. Statistically significant relationships between the independent and dependent variable would strengthen arguments that these decision-making networks types are structurally different and that the membership configuration of networks is meaningful for policy decisions. 116 CHAPTER 5: THE MEMBERSHIP STRUCTURE OF DECISION-MAKING NETWORKS AS A DETERMINANT OF POLICY USE The central purpose of this dissertation is to uncover whether the membership structure of a decision-making network influences cities’ propensity to use various local economic development policies. As previously discussed in Chapter 1, I contend that there are individualand network-level factors that help to shape decisions made by the network regarding local economic development policy. The ability of central actors to exploit their resources and influence the decision-making process is either limited or enhanced by the structural composition of the network, specifically by number of central actors, diversity of participants (homogeneity), and the number of organizations involved (size). If this conceptual framework for understanding the relationship between governance structure and policy is compelling, then cities led by structurally distinct decision-making networks should vary systematically in the policies they rely on to improve local economy. Given the membership structure of local economic development decision-making networks identified in Chapter 4 (local government-chambers of commerce, municipal, and broad public/private partnerships) and the conceptual framework outlined in Chapter 1, policy use across network type is expected to differ. The local government-chamber of commerce decision-making networks will most likely pursue a policy agenda that focuses on utilizing market-based approaches favoring the business community. These networks are also thought to direct some resources toward community-based development policies, but only if these development strategies also directly benefit business leaders. Similarly, the broad public-private partnerships are characterized by a strong local government-business collaborative. However, there are many community- and citizen-based organizations in fairly influential structural 117 positions. Also, nonprofit development corporations manage local economic development activities in these cities. For this reason, broad public-private partnership networks are expected to promote a progressive local economic development policy agenda that concentrates on supporting community development efforts sensitive to business interests. The diversity of policy interests as well as the number of actors suggests that local government-chambers of commerce networks and broad public-private partnership networks will have much broader local economic development policy agendas and use more policies. On the other hand, power within municipal decision-making networks is concentrated in local government. The diminished involvement by other development actors compels city governments to rely on their own resources to support the governance of local economic development. Thus, municipal decisionmaking networks are presumed to partake in local economic development policies less (as a result of diminished resources) and focus their development efforts on public ownership of profit-making institutions. During the cluster analysis process, a variable was created to assign cities to the decisionmaking network that best depicted the group of actors working to lead economic development. The local government-chamber of commerce network broad public-private partnership networks were coded as ‘1’, municipal networks were coded as ‘2’, and broad public-private partnership network local government-chamber of commerce networks as ‘3’. Several binary logistic regression models were estimated to test the existence of a relationship between network type and local economic development policy use. The dependent variable, policy use, was measured using a dichotomous variable (0 = no; 1 = yes). 118 5.1 Identifying the Dependent Variable Survey respondents were questioned about their cities’ use of forty-four different policies that address various areas of local economic development, including community development, small business development, business retention, and business incentives. In an effort to decrease the number of dependent variables, crosstabulation was employed to identify local economic development policies that shared a statistically significant bivariate relationship with the decision-making network type variable. The purpose of this test was to establish which development policies were used at considerably different rates across decision-making network types. The Pearson chi-square test statistic was analyzed to explore the strength of bivariate relationships between decision-making network type and cities’ use of various development policies. These results are displayed in Table 5.8 along with the percentage of cities within each network type that selected the policy as utilized by their community. There were sixteen local economic development policies significantly related to decision-making networks. The statistical significance signifies that cities led by different network types vary greatly in the extent to which they practice these policies. Four of the development policies were significant to decision-making networks at a level of p ≤0.001 and represented two areas of development policies: community development (community development corporation) and business retention (revolving loan fund, partnering with nongovernmental organizations, and partnering with other local government). There were seven development policies that displayed statistically significant relationships with decision-making network type at a level of p ≤ 0.01. This included three small business development policies (business incubators, microenterprise programs, and management training) and four business incentive policies (tax abatements, locally designated enterprise 119 zones, federal or state designated enterprise zones, and one-stop permit issuance). Five local economic development policies were related to decision-making network at a lesser significance level of p ≤ 0.05, including job training (community development), some other small business development policy not listed on the survey (small business development), ombudsman programs (business retention), subsidized buildings, and infrastructure improvements (business incentives). The results of this descriptive statistical test highlight several important findings. First, cities managed by various network types are engaging in many of the same policies at similar rates. Of the forty-four policies examined, decision-making network type failed to be significantly related to twenty-eight of the local economic development policies. When the bivariate relationship between decision-making network type and policy use was statistically significant, it was most often at lesser degrees of significance. For the most part, decisionmaking network type is not a strong predictor of cities’ collection of local economic development policies. There are many policies that cities either favor or do not, regardless of who is participating in the governance process. However, cities managed by different decision-making networks are most distinctive in their use of policies that direct efforts toward maintaining businesses presently located in the community. Three of the four policies with which network type shared a statistically significant relationship at a level of p ≤ 0.001 focused on business retention. There were, however, several small business development policies and business incentive policies that were also significantly related to decision-making network type at a lesser level of significance of p ≤ 0.01. This suggests that decision-making network types may also be distinguished by their use of small business development policies and business incentives. 120 In this section, I examined if decision-making network type may be related to the use of development policy. In the next section, I focus on identifying how decision-making network types influence cities use of development policies and the strength of these relationships within a 43 full model . 5.2 Decision-making Networks and Community Development Policies Results of the logistic regression models estimating the effect of decision-making network type on cities’ use of community development corporation policies are shown in Table 5.1 and 5.2. The difference in the likelihood ratio of the independent variable only model (167.984) and full model (162.921) is small. At a value of 5.063 and statistically significant at a level of p < 0.10, inclusion of the decision-making network variable did slightly improve the predictions made by the model. 43 44 In fact, Model 2 correctly predicted 73% of the cases Only those estimated logistic regression models in which network type was a significant predictor of policy use in the full model are reported. These economic development policies included: one community development policy (community development corporations), one small business development policy (microenterprise programs), two business retention policies (partnering with other nongovernmental organizations and partnering with other local governments), and two business incentive policies (tax abatements and one-stop permit issuance). 44 p = 0.080 121 compared to the constant-only model, which predicted only 56% of the cases correctly. 45 The strength of the full model is further supported by a statistically insignificant Hosmer-Lemeshow goodness-of-fit test statistic of p = 0.886. 46 The results of this test lead to failure of rejecting the Table 5.1 Summary of Estimated Community Development Corporations Models null hypothesis that there is no statistically significant difference between values observed in the data and predicted by the model. Thus, the model’s prediction regarding cities’ use of community development corporations fits the actual observations quite well. Thus, the full model fit the data better than the constant-only and control-only models. 45 This is the overall percentage of correctly predicted occurrence and nonoccurrence of the event. The model’s sensitivity, or correctly predicted occurrence, equaled 81.9% while the specificity, or correctly predicted nonoccurrence, equaled 61.5%. Overall, the full model better predicted the city’s actual use of a community development corporation policy rather than nonuse. 46 Degrees of freedom = 8 122 Results of the full model indicate that the broad public-private partnership network type increases the likelihood of cities using community development corporations by a factor of 1.736 at a significance level of p = 0.054 when controlling for the effect of other variables. Cities Table 5.2 Estimated Logistic Regression Model of Community Development Corporations 123 managed by a broad public-private decision-making network are approximately two times more likely than communities in the reference group (cities led by a local government-chamber of commerce decision-making network) to make use of policies that enable governmental support of community development corporations. The 95% confidence interval ranges from 0.98 to 7.63 indicating that broad public-private partnership networks could be as much as 6.6 times more likely than local government-chambers of commerce networks to draw on community development corporations to address local economic development. However, the confidence interval does include values slightly below and greater than the value of 1, which diminishes the strength of the conclusion that a positive and statistically significant relationship exists between broad public-private partnership network types and the use of community development corporations. 47 Although municipal decision-making networks do not predict the use of CDCs at a level of statistical significance, the odds that these cities will draw upon this particular policy is diminished by a factor of 0.23. Overall, broad public-private partnership development decision-making networks have the highest propensity to use policies that support community development corporations. Municipal decision-making network types are least likely to employ support to these development institutions. This is further supported by results of the descriptive statistical tests, which show that 70% of broad public-private partnership networks relied on this development strategy, compared to 57% of local government-chambers of commerce networks and 41% of municipal network types. 47 48 48 The 95% confidence interval spanned from 0.981 to 7.631. Please refer to Table 5.8. 124 These findings strongly support the hypotheses presented in Chapter 4 regarding the preferred policies of each network type given their membership structure. Of the three decisionmaking networks, broad public-private partnerships were hypothesized to strongly pursue a local economic development policy agenda that focused heavily on progressive and community-based development policies, especially if those strategies also present some advantages to the business community. The operations of community development corporations largely revolve around citizen participation and grassroots activism (Robinson 1995; Silverman 2005). The purpose of these organizations is to connect community members with various human resources (i.e. job training, employment opportunities, quality and affordable housing, etc.) in an effort to empower citizens and promote the economic and physical revitalization of the neighborhood (Robinson 1996; Walker 2002). Although this development strategy focuses primarily on addressing the needs of residents, business leaders reap benefits, including improved quality of the workforce and the development of quality affordable housing. Many CDCs also offer various resources for commercial and business development in their targeted areas, including technical assistance, financing, building construction, and renovation (Walker 2002; Steinbach 1997). Local government-chambers of commerce network types were also expected to utilize community development policies albeit to a lesser extent. The strong central position and power of various business-related organizations within the network are slightly mediated by the vertical structural position of nonprofit development corporations and moderately influential position of citizen-based institutions within the hierarchical structure. The presence of community-based organizations within the governance process increases the likelihood that local governmentchambers of commerce networks would make use of community development policies, but only if these policies also benefit the business community. Although CDCs often provide assistance 125 and support to businesses in the targeted community, these organizations have also been cited as pursuing a progressive development agenda involving obstruction to the “growth machine forces of gentrification and downtown redevelopment” (Robinson 1995; 1648). As key members of the growth machine, business leaders may be less inclined to advocate for community development corporations as a local economic development strategy. 49 Given that CDCs do not provide the larger business community with an extensive set of developmental advantages, and have the potential to support anti-growth and anti-business development policies, local governmentchamber of commerce decision-making networks are not overwhelmingly supportive of this policy. Based on the hypothesized policy profiles of municipal networks, it is no surprise that this network type is least likely to utilize policies that promote CDCs. These organizations are generally not public entities and city governments do not accrue profits from their activities. Also, these institutions require substantial access to financial and human capital (Fainstein and Fainstein 1986). In fact, financial capital is paramount to a CDC’s impact on the community (Gittell and Wilder 1999). 50 Given that this network type depends heavily on investment resources and does not contribute to municipal revenues, cities led by a municipal decisionmaking network would have a more difficult time financing these programs. Therefore, 49 50 See Logan and Molotch 1987; Molotch 1976, 1988; also see Stone 1989, 1993. Gittel and Wilder (1999) also cite a clear mission, professional and sophisticated staff, and political influence as the most important factors influencing the effect of community development corporations on distressed communities. 126 municipal networks are less likely to make use of policies that support community development corporations. There were two control variables that also predicted the use of CDC policies at a level of statistical significance. City type was related to community development corporation policies at a statistically significant level of p < 0.01. At a significance level of p = 0.01, communities identified as “cities” rather than villages, townships, or some other type of substate jurisdiction are 4.68 times more likely to employ this economic development policy. Median home cost also predicted community development corporation policy use at a lesser level of p = 0.036. However, the odds ratio estimate equals 1.000, indicating that there is no effect. 5.3 Decision-making Networks and Small Business Development Policies The estimated logit models for cities’ use of policies that enable local governments to engage in microenterprise programs are presented in Table 5.3 and 5.4. Model 2 is the full model that includes several control variables as well as the independent variable. The likelihood ratio for Model 1 (75.278) and Model 2 (65.188) was compared to examine how inclusion of the independent variable changes the strength of the estimation. The difference in likelihood ratios equaled 10.09 and was statistically significant at a level of p < 0.01. 51 52 Both the magnitude and significance of the difference indicate that the addition of the decision-making network variable did significantly improve the strength and predictions of the model. 51 52 p = 0.006 df = 2 127 Approximately 86% of cases were correctly predicted by the logit estimation. 53 A considerably large and statistically insignificant Hosmer-Lemeshow goodness-of-fit test statistic of 0.937 further demonstrates the strength of the full logistic regression model. 54 These results signify that the predictions of the estimated logit model closely matched the actual observations. The full logistic regression model fits the data well and the predictions made by the model for cities’ use of microenterprise programs are significantly different than the values predicted by the constant only model. Table 5.3 Summary of Estimated Microenterprise Models Examination of the full model demonstrates that the use of policies creating microenterprise programs varies significantly with the independent variable, decision-making network type. The test for the effects of the overall decision-making network variable, which is 53 This is the overall percentage of correctly predicted occurrence and nonoccurrence of the event. The model correctly predicted occurrences in 52% of the cases and correctly predicted nonoccurrences in 93% of the cases. Overall, the full model was a stronger predictor of cases in which cities did not use microenterprise policies. 54 The chi-square equaled 2.959 with a degrees of freedom equal to 8. 128 denoted as “Decision-making” in Table 5.4, is statistically significant at a level of p = 0.033. Both decision-making network types included in the model predicted the governmental use of Table 5.4 Estimated Logistic Regression Model of Microenterprise Programs 129 microenterprise programs at some level of statistical significance. Broad public-private partnership decision-making networks were related to policy use at a statistically significant level of p < 0.05. 55 The existence of this network type in a city decreases the odds of microenterprise policies by 91% when compared to local government-chambers of commerce networks. Municipal decision-making networks forecast the use of these policies at a lesser level of statistical significance at p = 0.083. These network types were 95% less likely than local government-chamber of commerce networks to make use of these particular policies. Thus, microenterprise programming for economic development is most likely to be found in cities being led by a local government-chamber of commerce network. The results of descriptive statistical tests show that 23% of these cities selected microenterprise policies as functioning in their communities to promote small business development. Only 9% of municipal network communities and 10% of broad public-private partnership network communities utilized this particular development strategy. Microenterprise programs are an interesting local economic development strategy. The International City/County Management Association (ICMA) categorized this particular policy as a strategy focused on the development of small businesses. Scholars argue, however, that this particular development tool integrates aspects of economic development and economic empowerment (or community development) agendas (Servon 1993; Schreiner 1999; Cooney and Shanks 2010). Microenterprise programs have been labeled as a “market-based strategy for poverty alleviation” (Cooney and Shanks 2010, 29). These organizations support the creation and growth of small and local businesses by providing loans and training (i.e. business literacy, technical knowledge, specific job skills, etc.) (Servon 1997; Schreiner 1999; Cooney and Shanks 55 p = 0.022 130 2010). Microenterprise programs generally focus on self-sufficiency and target economically vulnerable individuals and communities (Servon 1993; Cooney and Shanks 2010). These organizations, essentially, are conduits of self-employment and job growth, and aid in the alleviation of poverty by introducing individuals into the workforce, reducing unemployment, and generating income in communities (Servon 1993; Cooney and Shanks 2010). As a social welfare policy, however, microenterprise programs suffer from two major limitations. First, these policies are market-based rather than people-based and only indirectly promote individual economic empowerment and human development (Servon 1993). Second, the opportunities made available by microenterprise programs mostly attract a “niche” population, specifically those individuals with more assets, education, skills, experience and strong support networks (Schreiner 1999), as well as prospective community leaders (Servon 1993). As Schreiner (1999) points out, “few poor people can use self-employment to escape from poverty. Microenterprise increases income little if at all” (597). Thus, these organizations’ capacity to eradicate poverty is greatly diminished (Servon 1993). In sum, microenterprise programs are primarily a market-based and business-friendly development strategy that indirectly offers citizen- and community-oriented benefits, but to a much lesser degree. Results of the logistic regression model estimating the use of microenterprise programs provide further support for the hypothesized policy profiles of local economic development decision-making networks. Given the strong presence of business-related organizations in highly central structural positions and nonprofit development organizations managing local economic development activity, local government-chambers of commerce decision-making networks were hypothesized to have a higher propensity for utilizing local economic development strategies that benefit the business community. Community-based policies may be used by this network type, 131 but only if they directly provide economic developmental advantages to the local business community. On the other hand, broad public-private partnerships were predicted to support a strong progressive citizen-based development policy agenda that concentrated efforts on improving residents’ access to economic opportunities and revitalizing distressed communities. These networks are likely to consider the use of corporate-centered and market-based policies only if they address community development goals. Although microenterprise programs do inject economic resources into distressed communities, the central purpose of this development strategy is to aid in the formation and expansion of small businesses. As such, local government-chambers of commerce decisionmaking networks are more likely to promote the use of microenterprise programs. Broad publicprivate partnership networks also use microenterprise programs, but to a lesser extent given that community-oriented development is not the primary goal of this policy. Also, since microenterprise programs are not a strong source of revenue for local governments, municipal networks were less likely than local government-chambers of commerce networks to use microenterprise supports and utilized this development policy the least. Two of the control variables predicted cities’ involvement with microenterpise programs at lesser levels of significance with p < 0.10. Perceptions of local governments in surrounding states as competition in attracting investments decreased the odds of using microenterprise programs by 84%. 56 The 95% confidence interval, however, ranged from 0.024 to 1.045. The inclusion of both positive and negative values within the confidence interval diminishes the strength of these findings. Also, cities with an economic development plan were more likely to engage in microenterprise programs and increased the odds by a factor of 4.6. The 95% 56 p = 0.056 132 confidence interval for both control variables, however, includes the value of 1.000, making it difficult to discern the direction of the relationship. The confidence interval ranges from 0.886 to 34.746, suggesting that conclusions regarding the direction of the relationship are inconclusive. Yet, the effect of economic development plans on the use of these policies may potentially be positive and quite large. 5.4 Decision-making Networks and Business Retention Policies Results of the logistic regression models estimating the likelihood of policies of partnering with nongovernmental organizations are presented in Table 5.5 and Table 5.6. The log-likelihood ratio for Model 1 (125.821) and Model 2 (121.204) differed by 4.617. This model chi-square was statistically significant at a level of p < 0.10. 57 These results indicate that the addition of the decision-making network type variable did moderately improve the strength of the model. Model 2 correctly estimated 85% of the cases, signifying that the model fits very well when comparing the predictions to actual observations. 58 Also, the Hosmer-Lemeshow Test statistic was insignificant at a level of p < 0.254 and signified that the observed and 57 58 p = 0.099 This is the overall percentage of correctly predicted occurrence and nonoccurrence of the event. The model correctly predicted occurrences in 97.3% of the cases while correctly predicting nonoccurrences in 21% of the cases. The estimated logit models were an especially strong predictor of instances in which cities did partner with nongovernmental organizations in an effort to promote economic development. 133 estimated values are not different at a level of statistical significance. 59 Results of these various tests provide strong support for the strength of the model and indicate that the full estimated model fits the data quite well. Table 5.5 Model Summary: Partnering with Nongovernmental Organizations Models There was only one category within the decision-making network variable that significantly influenced the odds of local governments partnering with nongovernmental organizations as a local economic development policy. The presence of a municipal decisionmaking network predicted the use of these policies at a statistically significant level of p < 60 0.05. Municipal decision-making networks decrease the odds of cities engaging in these partnerships by a factor of 0.275 when compared to local government-chamber of commerce networks. Moreover, government-chamber of commerce networks were fairly similar in their use of this development strategy. The 95% confidence interval indicates that municipal decisionmaking networks, when compared to the reference group, may decrease the chances of local government partnering with nongovernmental organizations by 10% to 92%. The results of descriptive statistical tests demonstrate that 66% of municipal decision-making networks made 59 60 The chi-square equaled 10.157 and the degrees of freedom equaled 8. p = 0.033 134 Table 5.6 Estimated Logistic Regression Model of Nongovernmental Organizations Partners 135 use of this policy. On the other hand, 82% of broad public-private partnership networks and 86% of local government-chambers of commerce networks partnered with nongovernmental organizations. These findings strengthen the conclusions presented in Chapter 4. On average, municipal networks involved the least number of private participants when compared with other local economic development decision-making network types. Given the composition of the various network types, municipal decision-making networks are much less likely to include private actors in the local economic development decision-making process. Nongovernmental organizations were considerably more engaged in the activities of broad public-private partnerships and local government-chambers of commerce network types. Although the relationship was insignificant, the broad public-private partnership network type had an increased likelihood to partner with nongovernmental organizations. The perception of nearby local governments as competition in attracting investments was the only control variable to display some statistical significance. At a lesser level of p = 0.078, this control variable decreased the likelihood of a city partnering with nongovernmental organizations by 79%. For this variable, however, the confidence interval included the value of 1, which lessened the strength of these findings. Results of the logistic regression models estimating the likelihood that cities partner with local governments are presented in Table 5.7 and Table 5.8. The inclusion of the decisionmaking network variable added to the strength of the model as indicated by a difference of 7.337 in the log-likelihood ratios of Model 1 (209.505) and Model 2 (202.168). 61 61 This model df = 2, or the difference in the number of independent variable categories included in the models. 136 Table 5.7 Model Summary: Partnering with Local Governments Models chi-square was statistically significant at a level of p < 0.05 suggesting that the independent variable may not have a substantial effect, but it is a significant predictor of active policies that empower city governments to partner with other local governments.62 When identifying cities that did and did not employ this particular policy, the model successfully predicted 71% of the cases.63 The insignificant Hosmer-Lemeshow test statistic64 (p = 0.505) provides additional evidence of a strong model. The various tests signify that the full model fits relatively well when comparing predicted values to the observed values of cities’ policy use. Results of the estimated logistic regression model reveal that decision-making network type predicts cities’ partnering with other local governments at a level of statistical significance. Cities in which a municipal decision-making network type was present were 70% less likely than local government-chambers of commerce networks to employ this policy. The Exp(B) equaled 0.306 and was significant at a level of p = 0.016. At a much lesser level of statistical 62 63 p = 0.026 This is the overall percentage of correctly predicted occurrence and nonoccurrence of the event. The model correctly predicted occurrence in 72% of the cases, while correctly predicting nonoccurrence is 71%. 64 The chi-square equaled 7.292 with a degrees of freedom equal to 8. 137 significance (p < 0.10), the existence of a broad public-private partnership decision-making network also decreased the likelihood of cities partnering with other local governments. In Table 5.8 Estimated Logistic Regression Model of Partnering with Local Governments 138 comparison to local government-chambers of commerce network types, broad public-private partnership networks lessened the likelihood of this policy being used by 56%. This policy was most frequently utilized by cities managed by local government-chambers of commerce networks (56%) and least popular with municipal networks (25%). Although broad publicprivate partnerships were less likely than local government-chambers of commerce networks to use this policy, there was not a large difference in the rate of policy use. Exactly one half of cities led by broad public-private partnership networks identified partnering with local governments as an economic development policy. Similar to the findings of the logit model estimating cities’ likelihood to partner with nongovernmental organizations, these findings further strengthen claims made about the structural distinctiveness of the identified development decision-making networks presented in the previous chapter. Municipal networks are characterized by low levels of participation for all development actors except local governments. State governments, as well as the federal government, displayed the lowest actor centrality scores within the municipal network type. Also, this was the only network type in which county governments did not possess a strong or moderate central position. As such, this network type has the lowest propensity to partner with other local governments for economic development efforts. There were three dichotomous control variables representing competition in attracting investments that predicted the likelihood of cities partnering with other local governments. At a statistically significant level p < 0.05, perceptions that local governments in other states (Exp(B) = 2.470 ) and other states ( Exp(B) = 2.336 ) were competition in attracting investments predicted a considerably large increased likelihood in cities’ use of this policy. The first variable increased policy use by 147% while the second variable increased the odds of policy use by 139 134%. Also, if city leaders regard nearby local governments as economic competition the odds of a city partnering with other local governments is decreased by 64% at a lesser level of significance of p = 0.059. This effect, however, is statistically insignificant given that the 95% confidence interval includes both positive and negative values. Two categories of geographic region were also statistically significant predictors of policy use. Cities in the South were more likely than cities located in the Northeast to practice this policy by a factor of 7.639. This variable was statistically significant at a level of p = 0.003. Cities in the Western area of the United States were also more likely to partner with local governments by a factor of 4 at a level of p = 0.034. Both cities in the Southern and Western parts of the nation are considerably more likely than cities in the Northeast to partner with other local governments. 5.5 Decision-making Networks and Business Incentives Policies Results of the logistic regression model estimating the likelihood of cities using policies that grant tax abatements are shown in Table 5.9 and Table 5.10. The model chi-square, which equaled 4.771, was significant at a level of p < 0.10. 65 This value represents the difference between the log-likelihood ratio of Model 1 (135.102) and Model 2 (130.133). Although it is diminished, statistical significance of the model chi-square denotes a slight improvement in the strength of the model with the inclusion of the independent variable, decision-making network type. The full model successfully predicted 76% of the cases. Strength of the model was further 65 p = 0.092 140 supported by a statistically insignificant Hosmer-Lemeshow test statistic of (p = 0.251). 66 Differences between the control variable-only model and the full model signify that the inclusion of the decision-making network variable moderately improved the strength of the logit estimations. Table 5.9 Model Summary: Tax Abatements Results of the logistic regression presented in Table 5.6 show that only one type of decision-making network was a significant predictor of a community having policies that enabled governmental provision of tax abatements. Cities led by broad public-private partnership decision-making networks were much more likely than local government-chambers of commerce decision-making networks to have a policy that supported the distribution of tax abatements. Broad public-private partnerships increased the odds by a factor of 3 at a statistically significant level of p = 0.037. Almost 75% of cities managed by this network type relied upon tax abatements as an economic development strategy, whereas only 55% of local government-chamber of commerce network cities did. Municipal decision-making networks did not differ significantly from the reference group in their use of tax abatements. The relationship, however, was positive, indicating that communities managed by a municipal network are more likely than a local government-chamber of commerce network to provide these incentives. 66 The chi-square equaled 10.203 with the degrees of freedom equal to 8. 141 Similar to the reference group, 50% of cities in which municipal decision-making networks were present selected tax abatements as a development policy used in their community. Table 5.10 Estimated Logistic Regression Model of Tax Abatements 142 As a business incentive, tax abatements are essentially governmental subsidies that municipalities distribute to businesses with the intent of attracting and maintaining their investments in the community. Eisinger (1988) stated that local governments’ provision of business incentives are intended to incentivize private investments by indirectly lessening the effects of state and local taxes on businesses. By subsidizing investment, residents benefit greatly from economic growth resulting from job growth and government experiences increases in public revenues, which improves their capacity to provide municipal services. A considerable amount of research has been conducted to test Elkin’s claims that tax abatements and other incentives are an effective development tool, and whether the effect of abatements on economic growth is substantial. Findings indicate that subsidizing businesses’ costs has had positive, negative, and mixed effects on local economies. 67 However, many other scholars have focused their attention on uncovering specific patterns in the types of cities that rely on tax abatements as an economic development tool. Studies have shown that a city’s likelihood of distributing tax abatements increases with centralization of decision-making powers in the mayor’s office (Feiock and Clingermayer 1986; Sharp 1991) as well as with fiscal stress (Rubin and Rubin 1987; Sharp 1991; Feiock 1992; Fleischmann et al. 1992; Clarke and Gaile 1998; Byrnes et al. 1999; Sands and Reese 2008), including low tax revenues. In other words, the use of incentives increases for development decision-making networks in which local government is the most central development actor and market resources are not readily available for development efforts. 67 For a brief summary, see: Reese, L., T. Blackmond Larnell, and G. Sands. (2010). “Patterns in Tax Abatement Policy: Lessons from Outliers?”. American Review of Public Administration 40(3): 261-283. 143 For this reason, municipal decision-making networks were expected to have a strongly positive effect on cities use of this business incentive. Tax abatements do not entail public ownership of profit-earning institutions or a substantial increase in governmental revenues. Local government-chambers of commerce decision-making networks were also predicted to strongly promote municipal distribution of tax abatements given that they directly diminish the costs of business operations. Instead, the findings from this study indicate that broad publicprivate partnership networks, which involve several citizen- and community-based development actors in moderately central structural positions, have the highest propensity to make use of these business incentives. According to the literature, communities with higher levels of citizen involvement in the development decision-making process are less likely to support tax abatements and other incentives (Clingermayer and Feiock 1990; Sharp and Elkins 1991). Sharp and Elkins (1991) maintain that community members tend to oppose these governmental subsidies to businesses because they contribute to higher taxes. The findings in this study indicate that the inclusion of citizen-based organizations in the decision-making process increases a city’s odds of utilizing tax abatements, which conflicts with previous research. Geographic region was the only control variable to be a significant predictor of tax abatement policy usage. Cities in the West had decreased odds of using tax abatements by a factor of 0.59 at a significance level of p = 0.003. These cities are 60% less likely than cities in the reference group, Northeastern cities, to make use of abatements. The estimated logit model also reveals that unemployment predicts the likelihood of a city using tax abatements as a business incentive at p = 0.063. A 1% increase in the unemployment rate increases the odds of a city providing businesses with some form of tax abatements by 23%. Thus, cities with a higher level of unemployment are considerably more active in engaging in the use of tax abatements. 144 One-stop permit issuance was another business incentive shown to be significantly influenced by network type. Results of the binomial logistic regression model are displayed in Table 5.7. Addition of the decision-making network variable to the logistic regression model significantly improved the fit of the model to the observed data. The difference between the likelihood ratio of Model 1 (156.991) and Model 2 (145.276) equaled 11.715 and was statistically significant at a level of p<0.01. 68 Inclusion of the decision-making network type variable significantly improved the fit of the model. The strength of the model is further supported by the model’s high rate of success (73%) in predicting the sample cities that did and Table 5.11 Model Summary: One-Stop Permit Issuance did not employ this particular policy. 69 A Hosmer-Lemeshow test statistic that was statistically insignificant with a p-value equal to 0.407 also confirms that the difference between the observed values and estimated values is not significantly significant. Thus, the predictions made by the full logistic regression model correspond strongly to the actual data. 68 69 p = 0.003 This is the overall percentage of correctly predicted occurrence and nonoccurrence of the event. The model’s sensitivity, or correctly predicted occurrence, equaled 59% while the specificity, or correctly predicted nonoccurrence, equaled 84%. 145 Table 5.12 Estimated Logistic Regression Model of One-Stop Permit Issuance The independent variable, decision-making network type, exhibited a strong influence on city use of one-stop permit issuance in the full logistic regression model. Municipal decisionmaking networks decreased cities use of policies supporting one-stop permit issuance by 89% 146 when compared to local government-chamber of commerce networks. This variable was statistically significant at a level of p = 0.002. The 95% confidence interval indicates that the municipal decision-making network may diminish the likelihood of a city using one-stop permit issuance by 55% to 92%. Municipal decision-making networks were the least likely to make use of this particular development strategy. This policy is utilized in 45% of cities led by both broad public-private partnership networks and local government-chamber of commerce networks, while only 24% of municipal decision-making network communities provide one-stop permit issuance to support development efforts. By streamlining various permit processes that businesses are required to complete, local governments reduce the complexities of business operations. The benefits of one-stop permit issuance appeal directly to the interests of business-related development actors that are centrally positioned within local government-chambers of commerce and broad public-private partnership networks. Despite variations in membership structure, these two network types utilized this policy at similar rates. Both network types have a strong presence of corporate-oriented actors, except broad public-private networks also have many community-based organizations participating in development decision-making. Given that this policy involves a consolidation of services and decreases bureaucratic activity, it decreases governmental costs and does not contribute significantly to tax stress. Therefore, citizen-based organizations may not contest business leaders and oppose one-stop permit issuance. Although this policy does offer developmental advantages to the business community, one-stop permit issuance does not directly conflict with the progressive policy agenda of citizen and community-based organizations. This local economic development policy requires governmental resources and action, but does not create a stream for municipal revenues that 147 would not exist otherwise. Although this strategy involves the expansion of governmental services and is not a costly policy, the findings show that municipal decision-making networks are less likely to make use of policies supporting one-stop permit issuance. Two of the control variables included in the full model were also significant predictors (p < 0.05) of cities using the policy of one-stop permit issuance. Cities located in the Western area of the country were more likely than the reference group (Northeastern region) to employ this policy. Cities in this region had increased odds of using one-stop permit issuance by a factor of 8.494. In other words, cities located in this part of the nation were almost 900% more likely to make use of one-stop permit issuance. Also, cities in which leaders perceived local governments in other states as competition in attracting investment were 76% less likely to make use of this policy. The continuous variable representing the unemployment rate was also statistically significant at a level of p < 0.10. Every 1% increase in a city’s unemployment rate increased the odds of a city engaging in this policy by 25%. The 95% confidence interval, however, ranged from 0.988 to 1.57, making it difficult to draw conclusions regarding the direction of this relationship. 5.6 Policy Profiles of Local Economic Development Decision-Making Networks Each of the decision-making network types, including the reference group, predicted the likelihood of cities practicing policies at a level of statistical significance in three of the six estimated logistic regression models. Communities assigned to the broad public-private partnership network type were more likely to engage in community development and tax abatement policies, but were less likely to support microenterprise programs. The presence of this network type diminishes the odds that municipal governments will practice the policies of 148 partnering with nongovernmental organizations, partnering with local governments, and providing one-stop permit issuance. The reference group, local government-chambers of commerce networks, displayed statistical significance in models that estimated cities’ use of the following policies: microenterprise programs, partnering with local governments, and one-stop permit issuance. Broad public-private partnerships and municipal networks decreased the likelihood of cities using these policies. The reference group had the most positive effect on cities’ propensity to utilize these policies. In this section, these findings are analyzed in conjunction with descriptive statistical test results to formulate policy profiles for each network. Table 5.13 Decision-Making Network Areas Using Local Economic Development Policies 149 Table 5.13 (cont’d) In comparison to other network types, broad public-private partnerships utilized many different policies at high rates. Broad public-private partnership networks were the most frequent user of twenty-one of the policies examined, including four community development 150 policies, four small business development policies, two business retention policies, and eleven business incentives. 70 Broad public-private partnership networks displayed the highest rate of policy adoption for most of the community development policies. A large proportion of these networks (39%) identified more than three out of five community development policies as a strategy utilized by their city to improve the local economy. Results of the logistic regression indicate that the existence of broad public-private partnership networks in a community increased the chances of cities employing policies authorizing community development corporations in comparison to local government-chamber of commerce networks. Almost seventy percent of these networks had adopted policies supporting CDCs, whereas only 57% of local government-chambers of commerce networks and 41% of municipal networks did the same. Although decision-making network type was not a significant predictor of community development loan funds (CDLFs) in a full model, it did approach statistical significance in the crosstabulation test. More than half of broad public-private partnership networks (53%) were using CDLFs compared to 44% of the local government-chamber of commerce networks and 35% of municipal networks. Each network type utilized many business incentive policies at similar rates, with broad public-private partnership networks displaying higher rates of use for most of the policies. Forty-three percent of broad public-private partnership communities selected at least seven business incentives as a strategy utilized by their city to improve the local economy. Tax abatements in particular were statistically significantly more likely to be employed by broad public-private partnership networks compared to the reference group in the estimated logit model. A substantial proportion of these communities (73%) distributed tax abatements to 70 There was a total of forty-five development policies examined. 151 businesses as a strategy to promote economic development. Only 55% of local governmentchambers of commerce networks and 50% of municipal networks identified abatements as utilized by their cities. Several other business incentives were used at considerably higher rates by broad public-private partnership networks but failed to display significance in the full logistic regression model (subsidized buildings, infrastructure improvements, and locally designated enterprise zone). This network type, however, decreased the probability of a community offering microenterprise programs in the full logistic regression model. Compared to local governmentchambers of commerce networks, broad public-private partnerships were considerably less likely to make use of this small business development policy. Only 10% of cities in the latter group engaged in microenterprise programs, whereas 23% of local government-chambers of commerce networks used this policy. Lack of statistical significance in the bivariate relationships of network type and the various small business development policies indicate that all of the decision-making networks had adopted these policies at similar rates. Cities managed by this network type were also very supportive of community development policies. These market-based polices direct resources to community-oriented development ventures in an effort to economically empower citizens in distressed communities, including community development corporations, community development loan funds, and job training programs. In comparison to the other types of decision-making networks, broad publicprivate partnership networks have the tendency to surpass the other types in the extent to which they practice community development policies. Broad public-private partnerships use policies that promote business interests over citizen-based development goals. More specifically, these networks engage in the distribution of 152 various business incentives, specifically tax abatements, which directly increase citizens’ tax stress yet contribute to private growth initiatives. These networks use different business development, business retention, and business incentive policies at fairly high rates. Despite the presence of a strong corporate-centered policy agenda, broad public private partnerships networks are also highly accommodating to community-oriented development strategies. In sum, this network type seems to pursue a more balanced economic development approach. As the largest and most diverse decision-making network type, broad public-private partnerships address local economic development with an extensive set of policies. Although the local government-chambers of commerce network was the reference group, it displayed statistical significance in the three logit models that examined policies promoting partnerships with local governments, microenterprise programming, and one-stop permit issuance. For each of these policies, broad public-private partnerships and municipal networks decreased the likelihood of cities utilizing these policies in comparison to the local governmentchamber of commerce type. In other words, the reference group has a more positive effect on cities’ propensity to utilize these policies in comparison to other network types. Although local governments were the most central actors for each of the decision-making network types, their participation rates were lowest for local government-chambers of commerce networks. Cities managed by this particular network type, however, are more likely to join with other municipal governments in an effort to grow their economies. This suggests that these communities have an increased likelihood to pursue regional economic development endeavors by collaborating with other local governments who may also benefit. The local government-chambers of commerce network type also increased the likelihood of communities making use of policies that promote microenterprise programs and one-stop 153 permit issuance. Both policies directly benefit the business community. Microenterprise programs direct financial and informational resources toward small businesses for development and expansion, while one-stop permit issuance lessens the difficulties that business owners experience when applying for the permits and licenses necessary to operate legally within the city. In effect, this network type is sensitive to business interests, particularly small business interests. However, both broad public-private networks and municipal networks were more likely than local government-chambers of commerce networks to use tax abatements. These policies directly decrease business costs of operations, yet only 55% of local government-chambers of commerce networks dispensed abatements to businesses. These networks did make use of many business incentives at moderate to high levels (including tax increment financing, infrastructure improvements, and zoning/permit assistance), but they were not related at a level of statistical significance. This network type did not predict cities’ use of community development corporations at a level of statistical significance, but the findings are quite interesting. Broad public-private partnership networks increased the likelihood of cities using CDCs while municipal networks decreased the odds. Although the direction of the relationship between this network type and CDC policy use is unclear, these findings suggest that local government-chambers of commerce networks are not as likely to promote these policies as the broad public-private partnerships, yet these cities are more likely than the municipal networks to make use of CDCs. Analysis of the proportion of cities within each network type that engages in the five community development policies (community development corporations, community development loan fund, job training, child care, and other) indicates that the local government- 154 chamber of commerce network type was much more similar to broad public-private partnership networks than municipal networks. Although broad public-private partnership networks practiced the first two policies (community development corporations and community development loan funds) most frequently, the difference between the proportion of broad publicprivate partnership cities and local government-chambers of commerce cities that utilized these policies was 12% and 9%. The same percentage of communities within both network types indicated that their governments were involved in job training, childcare, and other community development policies. Although the broad public-private partnership network is the most supportive of community development policies, local government-chambers of commerce decision-making networks do practice these policies at a moderate level. Community development corporations and job training policies are particularly favored by local governmentchambers of commerce decision-making networks. Overall, this decision-making network type is strongly supportive of local economic development policies that promote local business interests. They are, however, not as willing to subsidize business owners for operating within the city limits. As predicted in Chapter 3, local government-chambers of commerce communities practice community development policies to a lesser degree than broad public-private partnership networks. The latter network type has a stronger presence of citizen- and community-based organizations mediating the development goals of influential business-related organizations. Representatives of business interests do not face these limitations in local government-chambers of commerce networks, which diminishes the extent to which this network type is likely to engage in community-oriented development. Local government-chambers of commerce networks are most involved in those community development policies that directly improve the productivity of the business community by 155 offering financial and information resources through community development corporations and improvements in the quality of labor as a result of job training policies. Overall, the set of local economic development policies used by this network type is extensive but concentrates efforts on supporting the interests of the business community. Overall, cities managed by the municipal network type did not use the various policies at high levels. Municipal decision-making networks displayed the lowest rates of policy use for thirty-one of the policies. 71 For each policy area, the majority of municipal networks relied on a lesser number of development policies. There were seven development policies employed most frequently by municipal networks. 72 These policies, however, were employed at similar rates by all three decision-making networks and failed to be related to decision-making network type at a level of statistical significance. For every estimated binomial logistic regression model in which municipal decision-making network was a statistically significant predictor of policy use, this network type had a negative effect. With regard to the business retention policies (partnerships with nongovernmental organizations and partnering with local governments), broad public-private partnerships and local government-chambers of commerce networks utilize these policies at similar rates while a 71 72 A total of forty-five policies were examined. This includes one community development policy (“other”), two small business development policies (marketing assistance and other), two business retention policies (local business publicity program and other), and two business incentives (tax increment financing and training support). 156 considerably smaller proportion of municipal networks have adopted these policies. This supports the findings presented in Chapter 3. The membership structure of municipal decisionmaking networks generally does not involve several participants. For this reason, the odds of municipal networks collaborating with other governments and private organizations are diminished when compared to the other network types. City governments are the only development actor to be centrally positioned within the network and possessed primary responsibility for development activity. Therefore, the decision-making process revolves largely around the interests of city governments. For this reason, municipal networks were predicted to be more likely to support policies that do not require substantial amounts of market resources as well as development strategies supporting public ownership of profit-making institutions. The findings regarding one-stop permit issuance and tax abatement policies, however, conflict with expectations. The presence of municipal networks decreases the likelihood of cities providing one-stop permit issuance. This involves the consolidation of various permit and licensing processes. One-stop permit issuance would, essentially, contribute to a more efficient bureaucratic process. Municipal networks also offered tax abatements at the lowest level. This particular development strategy does not require private resources or municipal governments to expend funds; rather, cities incur a cost in the form of foregone future tax payments. Instead of promoting business interests using policies with low upfront cost, municipal networks reap the financial benefits of businesses contributing to municipal revenues via different permit and licensing processes as well as tax costs. Given that none of the policies examined entail governmental ownership of facilities that yield returns for the municipality, it was not possible to test the extent to which municipal 157 decision-making networks influence cities to make use of these types of policies. The findings, however, indicate that these decision-making networks use local economic development policies at lesser levels than the other network types. They only displayed the highest rate of utilizing policies that support marketing assistance, ombudsman programs, tax increment financing, and training support. This suggests that these cities lack access to the extensive set of private resources necessary to employ various policies and they are not as supportive of the many business-sensitive policies included on the survey. More generally, the lack of private actors involved in decision-making reduces local economic development efforts. 5.7 Conclusion In this chapter, several statistical methods were utilized to examine patterns in the use of local economic development policies by structurally distinct decision-making networks. The results of crosstabulation tests were analyzed to identify local economic development policies that shared a statistically significant bivariate relationship with the independent variable, decision-making network type. Binomial logistic regression was conducted to test hypotheses about the effect of structurally distinct development decision-making networks on the economic development policies that cities choose to practice. There are a several findings that have important implications for understanding the governance of local economic development policy. First, the different network types utilize a large majority of development policies at comparable rates. For the most part, a similar proportion of cities assigned to each network type identified the same policy as a part of their local economic development activities. Decisionmaking network type failed to display a statistically significant relationship with many development policies in the crosstabulation tests. These findings, as well as the 158 underperformance of the independent variable in several of the full logistic regression models, suggests that decision-making network type is not a strong predictor of cities’ decisions to use specific local economic development policies. The importance of decision-making network type became apparent with the analysis of six local economic development policy outliers (community development corporations, microenterprise programs, partnering with nongovernmental organizations, partnering with other local governments, tax abatements, and one-stop permit issuance) that were influenced by network type. Examining the relationship between governance structure and the use of these specific development policies has provided substantial insight into how policy decisions are influenced by the membership structure of the decision-making network. Second, structurally distinct decision-making network types do engage in local economic development policy at significantly different levels. Municipal development decision-making networks were the smallest and least diverse. In comparison to other network types, they utilized most of the local economic development policies the least. Results of the full logistic regression models indicate that the municipal network type decreased the odds of cities practicing most of the policies that were examined. The small size of municipal networks diminishes the availability of resources and leads to the use of fewer local economic development policies. In contrast, broad public-private partnership networks were the largest and most diverse decisionmaking network type, which supports a more extensive set of local economic development policy. This network type had the most positive effect on cities’ likelihood to utilize the policies examined (using logistic regression) and displayed the highest rates of using a large majority of local economic development policies included in the study. The extent of their policy-making efforts reflects considerable access to a broad set of development resources. Local government- 159 chambers of commerce decision-making networks were smaller and less diverse. These networks consistently used local economic development policies at rates lower than broad public-private partnership networks, but displayed rates higher than municipal networks. In general, the structural composition of decision-making networks leads to differences in the number and diversity of policies that cities use. A more comprehensive local economic development policy agenda is found in cities with larger decision-making networks that include many organizations representing different policy interests. Communities in which the decisionmaking network is less inclusive rely on a smaller set of policies to address local economic development. Third, broad public-private partnership networks and local government-chambers of commerce networks were fairly comparable in the extent to which they engage in the same policies. For many of the development policies included on the survey, the difference in the proportion of cities that practiced the policies was moderate to small. This diminished the effect of the independent variable on cities’ use of several local economic development policies, and the statistical significance of the relationship between network type and policy choice. A key difference between these two network types, however, is in their use of policies that support community development corporations and other community development programs. Business-related organizations are amongst the most central development actors within both types of these decision-making networks. They were expected to strongly support a local economic development policy agenda that was sensitive to the goals of business communities. However, citizen-based and community-oriented organizations possess influential positions within broad public-private partnership networks. In comparison to local government-chamber of commerce networks, the broad public-private partnerships were predicted to be more 160 supportive of policies that promote community development initiatives. The findings presented in this chapter indicate that structural differences in the membership of these networks increase the likelihood of broad public-private partnership networks practicing policies related to community development corporations as compared with local government-chambers of commerce networks. The latter network type does utilize community development policies but to a lesser extent, given that the business community does not directly benefit from economic advantages. Both decision-making network types strongly favor corporate-centered local economic development policies. They vary in their use of policies with community-oriented goals. Without the involvement of citizen-based and community-oriented organizations, development decision-making networks are less likely to use policies that concentrate efforts on empowering residents and neighborhoods. Fourth, findings regarding cities’ use of tax abatements directly contradict other scholars’ previous findings and the hypotheses developed in Chapter 3. Local government-chambers of commerce decision-making networks were predicted to strongly promote tax abatements because they directly reduces businesses’ cost of operation. Instead, this network type was the least likely to distribute tax abatements. Rather than sacrificing tax revenues, local governmentchamber of commerce networks are more supportive of policies that concentrate resources on improving the local economy and the business climate. It would also be in the interest of the municipal decision-making networks to distribute tax abatements. Given their decreased access to private resources and diminished capacity to finance major development projects, these cities may have a harder time attracting and maintaining local businesses. It is possible that municipal networks do perform more caretaker duties, and therefore would not be actively recruiting 161 businesses with tax abatements. When compared to local government-chambers of commerce and municipal networks, the broad public-private partnership networks significantly increase the odds of a city dispersing abatements. This network type was hypothesized to be less engaged in offering business incentives, given that they directly increase tax stress and would conflict with the interests of the citizen and community-based organizations involved in decision-making. However, the large and diverse set of actors may be more capable of garnering enough resources to provide tax abatements without considerable costs to the residents. These findings highlight a need to revisit cities use of tax abatements in an effort to support economic development efforts. In the previous chapter, the findings demonstrated that the existence of three types of structurally distinct decision-making networks guided local economic development policy in American cities. In this chapter, results of the descriptive statistical tests and logistic regression models establish that the variations in membership structure of these networks do influence cities’ adoption of some local economic development policies. The main claims of this dissertation were substantiated and it is clear that understanding the governance structure is meaningful for explaining policy choices. Given that a community’s decision-making network type has significant implications for policy, it is important to determine why cities differ in governance structures. In the next chapter, an exploratory analysis is conducted to determine which social, political, economic, and geographic factors influence the type of decision-making network that emerges in a community. 162 CHAPTER 6: AN EXPLORATORY ANALYSIS OF FACTORS AFFECTING DECISION-MAKING NETWORK TYPE Two major conclusions can be drawn from the findings and analyses presented in the previous chapters. First, there are three types of networks, varying in membership structure, that govern local economic development in American cities. Second, the type of development decision-making network present in a community has implications for cities’ decisions regarding some of the local economic development policies used. For this reason, these governance structures require further consideration. More specifically, what factors promote the presence of one decision-making network type in a community rather than another? Understanding why decision-making networks vary in membership across cities is imperative to further uncovering the importance of local governance for policy. There is considerable potential for many different factors to influence the membership structure of the decision-making network. These relationships have gone untested. Both the direction and strength of the effect remains unclear. In this chapter, I examine the characteristics of sample cities in an effort to identify the factors (see Table 6.1) that significantly increase or decrease the likelihood of decision-making networks existing in a community and managing development policy choices. The underlying assumption of this research is that some environments are more accommodating to the involvement of specific actors in decision-making and of the existence of particular governance structures. As an exploratory analysis, the goal of this study is to determine why decision-making networks have a higher (or lower) propensity to materialize in different types of cities. In this chapter, three logistic models were fitted to the data to test whether several independent variables strongly predicted the presence of each network type in a city. The 163 Table 6.1 Description of Dependent and Independent Variables analysis in this chapter involves three dependent variables: municipal, local governmentchamber of commerce, and broad public-private partnership. The cases were coded as “0” if 164 they had not been assigned to the decision-making network type and “1” if the city was categorized as being led by the decision-making network type. A binomial logistic regression model was estimated to examine the relationships between each dependent variable and city demographics (population size, metropolitan status, and geographic region), local economy (economic condition, economic base and economic focus), land use (residential, non-residential, and open), and government structure (city type, mayor-council, and council-manager). 6.1 Municipal Decision-Making Networks Results of the logistic regression model that estimated the likelihood of a municipal decision-making network existing in a community are presented in Table 6.2 and Table 6.3. The omnibus tests of model coefficients indicate that the full logit model is an improvement on the intercept-only model. With a model chi-squared equal to 66.801 and statistically significant at a level of p < 0.01, the estimated logit model fits the data better than the model in which the coefficients of all the independent variables is equal to zero. 73 The strength of the full model is further supported by a statistically insignificant Hosmer-Lemeshow goodness-of-fit test statistic of p = 0.210. 74 There is no statistically significant difference between the values predicted by the model and those observed in the actual data. The results displayed in Table 5.2 indicate that the model correctly predicted 83% of cases. 75 The model’s sensitivity, or correctly predicted occurrence, equaled 37% while the specificity, or correctly predicted nonoccurrence, equaled 73 74 75 df = 43 df = 8 This includes occurrence and nonoccurence of municipal decision-making networks. 165 95%. The full model better predicted the instances in which municipal networks were not managing local economic development in a city. Overall, the full model fit the data better than the constant-only model, but it was not a strong predictor of when cities were actually being guided by a municipal network. Table 6.2 Model Summary: Municipal Decision-Making Network There were two independent variables, metropolitan status and locality type, that predicted the presence of municipal decision-making networks at a level of statistical significance. According to the results, communities labeled as “suburbs” were significantly more likely than the reference group, independent municipalities, to have municipal networks leading local economic development decision-making in their communities. 76 77 Being categorized as a suburb increases the likelihood that a municipal network is present by a factor of 4.363 at a statistically significant level of p = .024 when controlling for the effect of other variables. Central cities were also positively related to the occurrence of municipal network type, but not at a level of statistical significance. 76 The ICMA Economic Development Survey characterizes these municipalities as cities located in the Metropolitan Statistical Area (MSA), but not the core city of the MSA. 77 The ICMA Economic Development Survey characterizes these municipalities as cities not located in the MSA. 166 Table 6.3 Estimated Logistic Regression Model of Municipal Networks 167 Locality type predicted the existence of municipal networks at a statistically significant level of p = 0.010. Communities classified as cities were more likely than other governmental jurisdictions (towns, villages, townships, boroughs, and districts) to be assigned to the municipal network category. The likelihood of municipal decision-making networks existing in a community increases by a factor of 2.597 when the locality is a city. 6.2 Broad Public-Private Partnership Decision-Making Networks The model correctly predicted 84% of all cases, including occurrences and nonoccurences. The model successfully estimated 56% of occurrences and 93% of nonoccurences. Although the full model estimated the presence of broad public-private partnership networks more effectively than the constant-only model, it was more successful at forecasting instances in which this network type was not present in a community. Several inferential test statistics were analyzed to evaluate the capacity of the overall model to estimate the existence of broad public-private partnership networks in sample cities. These results are displayed in Tables 6.4 and 6.5. The full model chi-square equaled 111.057 and was statistically significant at a level of p < 0.000, which signifies that the inclusion of the Table 6.4 Model Summary: Broad Public Private Networks Decision-Making Network 168 Table 6.5 Estimated Binomial Logistic Regression Model of Broad Public Private Networks 169 independent variables did improve power of the logit model and strengthened the estimation. 78 The statistical insignificance (p = 0.310) of the Hosmer and Lesmeshow goodness-of-fit test indicates that the observed values did not differ at a level of statistical significance. The metropolitan status variable and locality type variable influence the likelihood of cities having broad public-private partnership networks at a level of statistical significance. Metropolitan status has three categories, including central cities, suburbs, and independent municipalities. The latter is the reference group. In comparison to independent municipalities, the odds of this network type being present in suburban communities are diminished. For a community to be labeled as a suburb decreases the chances of a broad public-private partnership network existing by a factor of 0.914 at a statistically significant level of p = 0.000. Central cities also had a negative effect on cities’ use of broad public-private partnership decisionmaking networks, but at a lesser level of statistical significance of (p = 0.083). When controlling for all other variables, locality type predicted the likelihood of communities relying on this network type at a statistically significant level of p = 0.010. The odds of a broad public-private partnership network type managing local economic development is increased by a factor of 4.5 when the governmental jurisdiction is classified as a city rather than a town, village, township, borough or district. The results of the 95% confidence interval indicate that cities may be 1.5 to 20 times more likely than all other governmental jurisdictions to be governed by a broad public-private partnership network. 78 df = 43 170 6.3 Local Government-Chambers of Commerce Decision-Making Networks The findings from the logit model that estimated the odds of a city being guided by a local government-chamber of commerce network are presented in Table 5.4. In comparison to the constant-only model, the full model is a stronger predictor of cities use of this network type. Results of the omnibus test of coefficients show that the chi-square (value =69.949) was statistically significant at a level of p = 0.006. 79 The difference between the observed and predicted values was not statistically significant as demonstrated by the Hosmer and Lemeshow goodness-of-fit test statistic (p = 0.410). The overall percentage of cases, including occurrences and nonoccurences, that were correctly predicted by the model equaled 72%. More specifically, the full model accurately estimated 79% of occurrences and 64% of nonoccurences. Table 6.6 Model Summary: Local Government-Chambers of Commerce Networks Two independent variables, metropolitan status and locality type, influenced the odds of a local government-chamber of commerce network existing in a community at a level of statistical significance. Of the three categories included in the metropolitan status variable, independent municipalities (reference group) as well as suburban communities were significant determinants of whether a local government-chambers of commerce network was present. At a statistically significant level of p < 0.01, communities classified as suburbs are twice as likely as independent 79 df = 43 171 Table 6.7 Estimated Binomial Logistic Regression Model of Local Government-Chambers of Commerce Networks 172 municipalities to have a governance structure resembling this network type. Suburban communities increase the likelihood of local government-chamber of commerce network types existing by a factor of 1.3 to 7.5. The locality type variable affected the chances of this network type managing development decision-making at a high level of statistical significant with p equal to 0.000. Cities were .85 times more likely than all other types of governmental jurisdictions (towns, villages, townships, boroughs, and districts) to have a local government-chamber of commerce decision-making network. The 95% confidence interval indicates that cities increase the presence of this network type by a factor of 0.624 to 0.937. 6.4 Analysis The purpose of this study was to identify the demographic, geographic, economic, land use, and political factors the influence the type of decision-making network determining local economic development policy. As an exploratory analysis, however, the variables included in this analysis have not been tested in a full model. Several of the variables have been identified in small-N case study analyses as potentially relevant for describing variations in the structure of governance networks across cities. The other variables, theoretically, are also promising for explaining differences in decision-making network type. The findings presented in this chapter suggest that most of the independent variables are not significant determinants of local economic development decision-making network types. All three of the estimated logit models fit the data better than constant-only models, but the full models do not fit actual observations considerably well. For each estimated model, the Hosmer and Lemeshow goodness-of-fit test statistic indicated that the predicted and observed values did not differ at a level of statistical significance. These p values were small and range from 0.210 to 173 0.410. A larger p value, or higher level of statistical insignificance, would denote less difference between the values observed within the data and those estimated by the models. Weakness of the models is further supported by the moderate levels at which the estimated logit models successfully predicted the presence of decision-making network types in communities. More specifically, the models were much better predictors of nonoccurences than instances in which the network type was leading local economic development decision-making. The diminished strength of the overall models is the result of several statistically insignificant independent variables in the model. Scholars have argued that the structural composition of governance networks varies across “strategic purposes” (Agranoff and McGuire 1998, 79) and “functional specificities” (Adam and Kriesi 2007, 137). In other words, development actors are goal oriented and this motivates their participation in the governance process (see Agranoff and McGuire 1998; Stone 1989; Melbeck 1998; Borzel 1997). These factors shape the context in which local economic development decision-making occurs and strategic behavior by development actors unfolds. The involvement of specific individuals and organizations in governance, as well as network types, should vary depending on the gains they may potentially reap by guiding policy choices. Development actors, essentially, would be more involved in the decision-making process of communities in which their interests are strongly tied. For instance, cases with higher percentages of land zoned for commercial use would seemingly increase the number of business leaders committed to influencing development policy decisions. This, however, is not the case. The economic (economic condition, industrial base, and industrial focus) and land use (land zoned for commercial activities, land zoned for residential areas, and land zoned for open space) variables were not meaningful for explaining why cities differ in governance structure. 174 The metropolitan status and locality type variables did consistently display statistical significance across all of the estimated binomial logistic regression models. With regard to metropolitan status, suburban communities predicted the existence of each decision-making network type at a level of statistical significance. In comparison to independent municipalities, suburbs increase the odds of a city having municipal networks (odds ratio = 4.4) and local government-chambers of commerce networks (odds ration = 2), yet decrease the chances of broad public-private partnerships managing local economic development. Although the effects were statistically insignificant, central cities also had a positive effect on the existence of municipal networks and local government-chambers of commerce networks, but negatively influenced the likelihood of a city being led by a broad public-private partnership. Cities, in comparison to other types of localities, positively influenced the presence of each network type. These findings support Agranoff and McGuire’s (1998) contention that metropolitan status affects network structure, specifically network size. They argue that smaller and more rural localities rely on a larger set of network actors to collaborate, secure, and pool resources. On the other hand, suburban governments partner with fewer actors. The results presented in this chapter show that independent municipalities, which are outside the metropolitan area, are significantly more likely to have broad public-private partnership networks managing development decision-making. In comparison to other network types, broad public-private partnership networks involve the most diverse set and largest number of actors. Suburban governments were significantly more likely to depend on municipal networks, which include very few actors, for local economic development decision-making. Another unexpected finding was the relationship between suburban governments and local government-chambers of commerce decision-making networks. Stone (1989) claimed that 175 corporate-centered urban regimes have the strongest influence in central cities. The presence of many small businesses and large corporations, as well as high land rents, motivate business leaders to be heavily engaged in development decision-making with the intention of promoting growth. Although the findings suggest that the relationship between central cities and the most business-sensitive decision-making network type is positive, this variable does not predict the existence of local government-chamber of commerce networks at a level of statistical significance. This network type, however, is significantly more likely to guide local economic development decision-making in suburban communities as compared to independent municipalities. 6.5 Conclusion The goal of this study is to fill a gap in the literature by identifying factors that explain variations in the membership structure of decision-making network types across communities. Metropolitan status and locality type were the strongest predictors of which type of decisionmaking network would likely manage local economic development decision-making in the cases. These findings suggest that a community’s position within the regional hierarchy has an important effect on the structure of the governance network. Central cities shared a positive relationship with the independent variable in the models. This implies that larger communities range in the type of networks guiding local economic development decision-making. Communities outside of the metropolitan statistical area have a significantly higher likelihood of relying on a more inclusive decision-making network. Local economic development policy decisions in independent municipalities involve several actors with varying policy interests. This diverse network of actors leads a city to support the use of a 176 broad set of policies to address the economic issues. The inclusion of community-based organizations increases the chances that independent municipalities will have the highest propensity to utilize community development policies. Suburbs had a positive effect on the existence of municipal and local governmentchamber of commerce networks. Suburban governments are most likely to have a smaller and less diverse decision-making network. Local economic development policy in suburban governments is largely decided upon by a municipal network or local government-chamber of commerce network. Suburban cities are drawing on fewer development actors when making choices regarding local economic development. Local government and chambers of commerce are the only central actors in these network types. With regard to policy use, findings suggest that suburbs would be more likely to either strongly support business-friendly policies or minimal government presence in local economic development. Suburbs would also have a decreased likelihood to use community development policies. Despite the statistical weakness of the estimated models, these findings provide a significant empirical contribution to the literature. The results of this exploratory analysis indicate that the structure of the development decision-making network is related to a community’s location within the region. Further consideration is needed to understand why some decision-making networks have a higher (or lower) propensity to exist in certain communities. A theory must be developed to explain the direction of the relationship between decision-making network type and metropolitan status as well as locality type. Several questions emerge. Why do central cities vary considerably in the membership structures of development decision-making networks? Why do independent municipalities have a tendency to rely on broad public-private partnership networks for decision-making? Is their propensity to use large 177 and diverse networks related to their need for increased access to development resources? In a similar vein, do suburbs require fewer resources for development and therefore lend themselves better managed by smaller decision-making networks? Future research must center on theory development and model development. This line of research requires the formulation of theoretical explanations for relationships between independent variables and network type and also suitable models for testing these theories. There are significant normative implications to consider regarding the structural composition of decision-making networks and their effects on cities’ policy choices. Evaluating the factors that influence development decision-making network type would aid in understanding the policy choices that local government make. 178 CHAPTER 7: CONCLUSION The research presented in this dissertation focuses on describing the membership structure of local economic development decision-making networks as a strong predictor of cities’ policy use. Urban regime theorists have provided a compelling framework for understanding how actors exploit their resources to gain access to the decision-making process, thereby shaping the structure of these governing institutions. The central tenet of urban regime theory is that the structures of public-private collaboratives that govern local economic development directly affects which policies are practiced by communities. As a theory of structuring urban regimes (Davies 2002), urban regime theory’s explanation of the relationship between the structure of governance networks and policy choices is limited. In this dissertation, I present a conceptual framework that strengthens scholars’ capacity to construct more complete profiles of decision-making networks’ membership and develop stronger predictions regarding variations in policy use across structurally distinct decision-making networks. Urban regime theory concentrates on identifying which actors have power and how they capitalize on their resources to impact decision-making. To improve urban regime theory, three network-level structural characteristics were incorporated into the framework. These structural attributes extend our analysis of governance to include the structural advantages and constraints facing individual actors in their pursuit of impacting policy. I argue that there are three structural characteristics (total number of central actors, actor homogeneity, and network size) that affect the capacity of individual central actors to shape the network’s behavior and policy decisions. The “central actors” construct identifies development actors most engaged in the network type and, therefore, more likely to influence policy decisions. The “total number of central actors” construct is utilized to evaluate the level at which power is dispersed within the network type. 179 The diversity of interests functioning within the network is measured by analyzing the “homogeneity” of actors involved in decision-making. The “size” construct calculates the average number of organizations contributing to decision-making within each network type. These network-level structural characteristics operate to either diminish or advance individual central actors in their pursuit to have influence over the decision-making process. The conceptual framework is summarized is Figure 7.1. Figure 7.1 Conceptual Framework: Policy Network Approach to Urban Regime Theory By emphasizing the central actors involved (urban regime theory) as the critical determinant of policy decisions and analyzing the structural arrangements of the collaborative (network theory), descriptions of governance structures can better account for the context in which joint policy-making occurs. The analysis of these five network attributes (individual-level and network-level) leads to a more robust characterization of decision-making networks and the environment in which development actors attempt to influence policy. A three-part quantitative analysis was conducted to examine three critical stages of local governance: the formation, composition, and policy choices of decision-making networks. In conjunction with examining whether structurally distinct decision-making networks 180 systematically vary in the use of local economic development policy, three other important research questions are also posed in this study. Are there distinct decision-making networks presiding over local economic development policy in cities? What are the structural differences in membership of the various development decision-making networks? What is the role of social, economic, geographic, and political factors in determining the emergence of different network types in cities? There are four central findings.  Structurally distinct decision-making network types do exist.  Private, yet non-business, development actors are moderately central actors in decision-making.  Overall, the membership structure of decision-making networks is not a strong determinant of cities’ policy choices. There are, however, some instances in which structurally distinct decision-making networks vary considerably in their chances of using development policies (community development corporations, microenterprise, partnering with nongovernmental organizations, partnering with other local governments, tax abatements, and one-stop permit issuance).  A city’s location in the region is a significant predictor of which decision-making network type is likely to exist in a community. For the first part of the study, hierarchical cluster analysis was conducted to develop an original typology of local economic development decision-making networks. The results indicate that there were three significantly different categories of development decision-making networks. The resulting typology included municipal, local government-chambers of commerce, and broad public/private partnership networks. The policy preferences and factors predicting the existence of summarized in Table 7.1. 181 Table 7.1 Summary of Decision-Making Network Policy Use and Presence in A Community 182 A single central actor, city governments, characterizes the municipal decision-making networks. This is the primary development actor contributing resources and coordinating development activities. The small size of municipal networks diminishes the availability of resources needed to pursue a more comprehensive local economic development policy agenda. Although city governments possess a considerable amount of power over decision-making, the decreased access to private resources compels municipal networks to make less use of various development policies. This includes development policies that generally do not demand a substantial amount of monetary resources, but require government to expand their services, potentially forfeiting potential tax revenues. Local government-chambers of commerce networks and broad public-private partnership networks were characterized by a strong presence of business-related organizations in very influential structural positions. As such, these decision-making networks have a positive effect on the likelihood that cities utilize business-friendly policies. As the largest and most diverse decision-making network, broad public-private partnership networks are the most engaged in local economic development policy. The extent of their policy-making efforts reflects considerable access to a broad set of development resources. With the involvement of many community-oriented organizations and nonprofit development corporations supervising development activity, this network type was the most likely to promote policies that focus on partnering with commissions, colleges and universities, ad hoc citizen groups, as well as private and community economic development foundations). The broad public-private partnerships networks were slightly more engaged in community development policies, particularly community development corporation policy, and less involved in microenterprise programs. 183 This is the key difference between local government-chambers of commerce network and broad public private partnership networks. Local government-chamber of commerce networks will promote the use of community-oriented development policies if they offer considerable economic advantages to business leaders. This is to accommodate the few citizen-based organizations that are in moderately central positions. This is supported by their negative effect on cities likelihood to use community development corporations, but positive effect on cities’ use of microenterprise programs. The policy differences between broad public-private partnership and local governmentchamber of commerce highlight the influence of private, yet non-business, development actors. Stone (1989) originally claimed that these actors would be less present in decision-making given their decreased access to market-based resources. The results of this study suggest this claim needs to be reconsidered. First, business leaders are not considerably more engaged in development decision-making in comparison to other types of development actors. The presence of municipal networks in several case cities further implies that business leaders are not always extensively involved in decision-making. This supports the findings of Hanson et al. (2010) that show a decrease in corporate civic leadership. Corporate elites have become increasingly disconnected from urban governance, which has important implications for policy. Second, citizen- and community-oriented organizations were moderately central actors, particularly in broad public-private partnerships and municipal networks. In other words, there are some private, yet non-business, development actors that have gained access to productive resources and are involved in the decision-making process. Progressive decision-making networks were considered to be hypothetical. However, findings demonstrate that many cities are being 184 managed by decision-making networks that extend their efforts beyond business-sensitive policies. Scholars have argued that the structural composition of governance networks varies across “strategic purposes” (Agranoff and McGuire 1998, 79) and “functional specificities” (Adam and Kriesi 2007, 137). In other words, development actors are goal oriented and this motivates their participation in the governance process (see Agranoff and McGuire 1998; Stone 1989; Melbeck 1998; Borzel 1997). Development actors would essentially be more involved in the decision-making process of communities in which their interests are strongly tied. The results of the exploratory analysis, however, emphasize the importance of a city’s position in the regional hierarchy as a critical factor in determining the type of decision-making network that exists in a community. These findings lend support to Agranoff and McGuire’s (1998) contention that metropolitan status affects network structure, specifically network size. Smaller communities have to include more actors in their governance process to gain access to a larger set of resources, whereas larger jurisdictions involve less development agents in the decisionmaking process. Independent municipalities, which tend to be located in rural communities, are significantly more likely to depend on broad public-private partnership networks to govern local economic development policy. These large networks provide local government with access to an extensive and diverse set of development resources. Suburban communities, on the other hand, are less likely to make use of these large and diverse network types. Instead, they have an increased propensity to be managed by smaller networks, specifically municipal networks and local government-chambers of commerce networks. Converse to Stone’s (1989) findings that corporate-centered urban regimes have the strongest influence in central cities, central cities 185 were not significantly related to the existence of any decision-making network type, including local government-chambers of commerce networks. Metropolitan centers vary considerably in the types of decision-making networks that govern their communities. As an exploratory analysis, a theory must be developed to explain why metropolitan status and city type are strong determinants of the structural composition of decision-making networks in communities. 7.1 The Underperformance of the Decision-Making Network Type Variable The empirical findings presented in this dissertation provide a logical explanation for variations in decision-making network types across cities, as well as systematic differences in their policy use. However, local economic development decision-making network type was not a significant predictor of most development policies examined in this study. The membership structure of decision-making networks is an important predictor of cities’ use of some local economic development policies. Politics do matter, but only sometimes. The findings from the full models estimating the odds of these policies being used by a city do lend support to the hypothesized policy profiles of local economic development decision-making networks. These results raise an interesting question: Why does governance structure affect the “outlier” policies but not other policies? There are some potential methodological reasons for the diminished statistical power of the independent variable, decision-making network type. Of the forty-four local economic development policies included on the survey, only sixteen of the policies shared a statistically significant bivariate relationship with the independent variable. However, there were only six development policies predicted by local economic development decision-making network variable at a level of statistical significance in the full model. Many scholars would consider the overall underperformance of the decision-making 186 network variable and estimated models as indicative of a weak framework with limited explanatory power. In other words, the merging of urban regime theory and policy network theory may not provide a stronger explanation for variations in cities’ policy use. The integration of these two dominant and governance-based approaches, however, is strongly supported by the literature. The conceptual framework presented in this dissertation provides a logical and theory-based explanation for systematic variation in the policy decisions of different governance structures. I argue that the underperformance of the statistical models results, largely, from the measurement of the independent and dependent variables. The independent variable, policy use, was measured using a dichotomous variable. The estimated binomial logistic regression models simply measure how likely a city would be to use a policy given the decision-making network type present. It is quite likely that cities do not vary considerably in whether their local governments practice a policy, but differ in the extent to which they engage in the policy. This would explain the underperformance of the independent variables in the estimated logit models. This does not invalidate the conceptual framework presented in Chapter 2. To improve performance of the independent variables and strengthen the estimated models, a continuous independent variable that operationalizes the intensity of policy use is needed. There are several potential continuous measures of policy use. First, it is possible to determine the how involved cities are in a particular area (community development, small business development, business attraction, and incentives) of local economic development policy. A variable can be constructed to quantify the number of individual policies within a specific policy area to identify which areas of development are most important for decisionmaking network types. Another option requires calculating the amount of resources the city 187 devotes to each development policy or policy area. The availability of data, however, may present a significant complication. 7.2 Future Research A central purpose of this study was to improve the generalizability of a local economic development decision-making network typology. Original types of local economic development decision-making networks are empirically derived, analyzed, and classified according to distinct patterns in their membership structure regarding centrality, actor homogeneity and network size. Although the framework provides a more detailed description of these governing institutions, a more complete illustration of these networks can be accomplished with further data collection. Representatives of local government provided the data utilized to construct the development of decision-making networks. This reputational method of data collection limits the analysis because the identified networks depict the governance structure as perceived by city government. Other development actors were not surveyed. As such, it is possible that other actors have different perceptions about the membership of the decision-making network. This data would strengthen claims regarding the centrality of various actors in local economic development decision-making networks. Analysis of other development actors’ understanding of the governance structure would permit the development of a more thorough account of network structure. Also, each network type represents a set of decision-making networks with comparable organizational structures. The measurements of structural attributes (central actors, total number of actors, homogeneity, and size) are, essentially, summations of these characteristics across decision-making networks of the same type. The aggregation of networks provided important 188 information that led to substantive findings. As mentioned in Chapter 2, social network analysis included several mathematical formulas that provide measures for similar network-level structural constructs included in the framework presented in this dissertation. These tools, however, are conducive to the study of single networks. The ability to measure each individual and network-level construct within a particular network would offer further insight into how each construct operates to impact the network’s behavior and policy decisions. 7.3 Conclusion Governing structures are quite complex. The findings presented in this dissertation make it clear that the structural arrangements of decision-making networks do vary across communities and have important policy implications. The predicted policy profiles of decision-making networks, which were developed based on the conceptual framework, were supported by the results displayed in the empirical chapters. Although this research offers significant contributions to the literature, the findings prompt new questions. First, does the membership structure of decision-making networks depend on the area of policy being considered? In other words, are multiple decision-making networks governing different areas of policy within the same communities? If so, how does this impact the typology of decision-making networks? Secondly, are these governance structures stable? Does policy use change only with modifications to the structure of the decision-making network? Do changes in the network come before changes in policy use? Given the paucity of research in uncovering why structurally distinct governance structures exist in different communities, it is difficult to evaluate whether the structure of the decision-making network is “man-made” or emerges to address particular social, economic, and political issues. If so, are these decision-making networks hoarding the 189 benefits of growth for the most influential, or effectively improving the local economies of their communities? There are significant normative implications to consider regarding the policy effects of structurally distinct decision-making network type. Further analysis is required to understand why some decision-making networks behave in a certain manner and have a higher (or lower) propensity to exist in different communities. Dahl’s seminal question of “who governs” is important, because the membership of the governance structure directly affects which actors’ policy preferences are materialized and which actors’ development interests are sacrificed. This line of research is critical to understanding the governance process. 190 APPENDIX 191 Table A3.1 List of Cities Included in the Dataset ALABAMA DECATUR DOTHAN MOUNTAIN BROOK TALLADEGA ARKANSAS CABOT HOT SPRINGS SPRINGDALE ARIZONA APACHE JUNCTION BULLHEAD CITY FOUNTAIN HILLS GILBERT MESA ORO VALLEY PAYSON PEORIA SCOTTSDALE YUMA CALIFORNIA ARCADIA BAKERSFIELD BANNING BARSTOW BEAUMONT BREA BURBANK BURLINGAME CARLSBAD CARPINTERIA CHICO CLOVIS COMPTON DALY CITY DAVIS DUBLIN FOLSOM GRASS VALLEY GROVER BEACH LA HABRA LA PALMA LAGUNA HILLS LEMON GROVE LINCOLN LIVERMORE MANTECA MENLO PARK MERCED MILLBRAE MORAGA OCEANSIDE PALM SPRINGS PICO RIVERA PLEASANT HILL RANCHO MIRAGE REDDING REDWOOD CITY SAN CLEMENTE SAN JOSE SAN LEANDRO SAN PABLO SANGER SANTA CLARA SANTA CLARITA SANTA MARIA SELMA SIMI VALLEY SUNNYVALE TEMECULA TULARE VISTA WEST SACRAMENTO WHITTIER COLORADO ARVADA AURORA BRIGHTON COLORADO SPRINGS FOUNTAIN GOLDEN GRAND JUNCTION GREENWOOD VILLAGE LOVELAND PARKER STERLING CONNECTICUT 192 AVON COVENTRY EAST HAMPTON HARTFORD PLAINVILLE SOUTH WINDSOR STRATFORD FLORIDA BARTOW BOCA RATON CASSELBERRY COCOA COCOA BEACH DELRAY BEACH EUSTIS GAINESVILLE GULFPORT LAUDERDALE LAKES LEESBURG MARCO ISLAND MIAMI BEACH MIRAMAR NORTH PORT OAKLAND PARK ORLANDO PALM BAY PALM COAST PARKLAND PINELLAS PARK PUNTA GORDA SANFORD SEMINOLE TAMARAC TARPON SPRINGS WINTER SPRINGS ZEPHYRHILLS GEORGIA ACWORTH CARROLLTON DECATUR GAINESVILLE LILBURN MOULTRIE NEWNAN Table A3.1 (cont’d) ROSWELL SNELLVILLE SUGAR HILL UNION CITY IOWA ANKENY CARROLL FORT DODGE IOWA CITY MARSHALLTOWN MUSCATINE OSKALOOSA IDAHO LEWISTON POST FALLS REXBURG ILLINOIS ADDISON ARLINGTON HEIGHTS BUFFALO GROVE CAROL STREAM CHAMPAIGN DECATUR EAST MOLINE ELK GROVE VILLAGE ELMHURST ELMWOOD PARK FRANKLIN PARK HANOVER PARK HAZEL CREST HIGHLAND PARK HOFFMAN ESTATES HOMER GLEN HOMEWOOD JOLIET JUSTICE KEWANEE LA GRANGE PARK LAKE FOREST LAKE ZURICH LEMONT LINCOLNWOOD LOMBARD MATTOON MOUNT VERNON MUNDELEIN NAPERVILLE NORTH AURORA NORTHBROOK O'FALLON PLAINFIELD RANTOUL RIVERDALE ROCK ISLAND ROLLING MEADOWS SCHAUMBURG SCHILLER PARK SKOKIE SOUTH ELGIN SPRINGFIELD SYCAMORE WEST CHICAGO WESTMONT WHEATON WHEELING WILMETTE WOOD DALE WOODRIDGE WOODSTOCK INDIANA COLUMBUS FISHERS HIGHLAND INDIANAPOLISMARION COUNTY, CI LA PORTE LEBANON MARION KANSAS DERBY DODGE CITY GARDEN CITY LAWRENCE LEAVENWORTH MANHATTAN MERRIAM TOPEKA KENTUCKY 193 BOWLING GREEN HOPKINSVILLE LEXINGTON-FAYETTE MADISONVILLE LOUISIANA BATON ROUGE-EAST BATON ROUGE MASSACHUSSETS ANDOVER BEDFORD EVERETT NEEDHAM NEWTON PEMBROKE READING SUDBURY WAREHAM WESTON MARYLAND ABERDEEN ANNAPOLIS BEL AIR BOWIE COLLEGE PARK CUMBERLAND HYATTSVILLE LAUREL MAINE BANGOR BIDDEFORD GORHAM SCARBOROUGH MICHIGAN ANN ARBOR BIG RAPIDS BIRMINGHAM COLDWATER DELTA CHARTER EAST LANSING FARMINGTON HILLS GRAND HAVEN GRAND HAVEN GRAND RAPIDS Table A3.1 (cont’d) GROSSE POINTE WOODS HARPER WOODS HOLLAND KENTWOOD LANSING MADISON HEIGHTS MERIDIAN MIDLAND NORTON SHORES NOVI OAK PARK OWOSSO ROYAL OAK SOUTHFIELD STERLING HEIGHTS STURGIS MINNESOTA ANOKA COTTAGE GROVE DULUTH EAGAN ELK RIVER FAIRMONT MAPLE GROVE MOORHEAD NEW HOPE NEW ULM PLYMOUTH ROCHESTER WHITE BEAR LAKE WOODBURY MISSOURI BLUE SPRINGS CAPE GIRARDEAU CARTHAGE FULTON GLADSTONE MARYLAND HEIGHTS RAYMORE ROLLA MISSISSIPPI CLEVELAND GREENWOOD MERIDIAN MONTANA GREAT FALLS NORTH CAROLINA BURLINGTON CARRBORO CHARLOTTE CONCORD CORNELIUS GREENSBORO HAVELOCK HICKORY MONROE MOORESVILLE ROCKY MOUNT WILMINGTON WINSTON-SALEM MINOT NEBRASKA SOUTH SIOUX CITY NEW HAMPSHIRE GOFFSTOWN LONDONDERRY ROCHESTER SOMERSWORTH NEW JERSEY ATLANTIC CITY FRANKLIN LITTLE EGG HARBOR MEDFORD MILLBURN MONTGOMERY MONTVILLE OCEAN CITY ROSELLE TENAFLY NEW MEXICO ALAMOGORDO NEVADA HENDERSON LAS VEGAS NORTH LAS VEGAS RENO SPARKS 194 NEW YORK CORNING GENEVA HAMBURG MAMARONECK NEW ROCHELLE OGDENSBURG OSSINING OSWEGO POUGHKEEPSIE RYE UNION WATERTOWN OHIO ASHLAND AVON BEAVERCREEK BEDFORD HEIGHTS BLUE ASH BROOKLYN BRUNSWICK CHILLICOTHE CIRCLEVILLE CLAYTON COLERAIN TOWNSHIP CONNEAUT DELAWARE DUBLIN ELYRIA FAIRBORN FAIRFIELD FOREST PARK FRANKLIN GALION HAMILTON HUDSON KETTERING LEBANON LOVELAND NEWARK NORTH ROYALTON SPRINGBORO STEUBENVILLE TROY Table A3.1 (cont’d) UPPER ARLINGTON WEST CARROLLTON WESTERVILLE WHITEHALL OKLAHOMA BARTLESVILLE BIXBY BROKEN ARROW CHICKASHA DURANT EDMOND MOORE OKLAHOMA CITY PONCA CITY STILLWATER WOODWARD OREGON BEND CENTRAL POINT CORVALLIS GRANTS PASS HERMISTON HILLSBORO KLAMATH FALLS LA GRANDE MEDFORD SHERWOOD SPRINGFIELD ST. HELENS THE DALLES TROUTDALE PENNSLYVANIA ALLENTOWN ELIZABETHTOWN EXETER HAZLETON LOWER MERION MANCHESTER NEW BRITAIN OIL CITY PATTON PLUMSTEAD PLYMOUTH POTTSTOWN SHALER UPPER DARBY UPPER MERION WARMINSTER WEST CHESTER RHODE ISLAND COVENTRY MIDDLETOWN PORTSMOUTH SMITHFIELD WEST WARWICK SOUTH CAROLINA COLUMBIA GREENVILLE SUMMERVILLE SOUTH DAKOTA BROOKINGS YANKTON TENNESSEE ATHENS BRENTWOOD BRISTOL CLARKSVILLE COLLIERVILLE COLUMBIA FARRAGUT GOODLETTSVILLE LA VERGNE MARYVILLE MILLINGTON MURFREESBORO SPRINGFIELD TEXAS ABILENE ALVIN AMARILLO AUSTIN BEDFORD BRYAN BURLESON CEDAR HILL CEDAR PARK COLLEGE STATION COLLEYVILLE 195 DENTON DICKINSON DUMAS FLOWER MOUND FRIENDSWOOD GRAPEVINE GREENVILLE HURST IRVING JACKSONVILLE KELLER KERRVILLE KILLEEN LA MARQUE LANCASTER LEAGUE CITY LEVELLAND MESQUITE MOUNT PLEASANT NACOGDOCHES NEDERLAND PLANO RIO GRANDE CITY ROWLETT SAN MARCOS SEAGOVILLE SNYDER SUGAR LAND TERRELL THE COLONY TYLER VERNON WACO WATAUGA WAXAHACHIE UTAH CLEARFIELD MIDVALE OGDEN CITY ROY SOUTH JORDAN SPRINGVILLE TAYLORSVILLE WEST VALLEY CITY Table A3.1 (cont’d) VIRGINA BLACKSBURG CHESAPEAKE FAIRFAX HOPEWELL LYNCHBURG MARTINSVILLE POQUOSON RADFORD ROANOKE STAUNTON VIRGINIA BEACH WAYNESBORO WINCHESTER VERMONT ESSEX WASHINGTON ANACORTES COVINGTON KIRKLAND LYNNWOOD RENTON RICHLAND SUNNYSIDE UNIVERSITY PLACE WALLA WALLA YAKIMA WISCONSIN BARABOO BELOIT CALEDONIA GRAND CHUTE HOWARD JANESVILLE KAUKAUNA MARSHFIELD MENASHA MENOMONEE FALLS MOUNT PLEASANT MUSKEGO OSHKOSH PLOVER WEST ALLIS WEST VIRGINA PARKERSBURG 196 Table A3.2 Participation Rates of Development Actors 197 Table A3.3 General Trends in the Use of Development Policies 198 Table A3.3 (cont’d) 199 Table A4.1: Partial Agglomeration Schedule from Hierarchical Cluster Analysis 200 Figure A3.1 ICMA 2004 Economic Development Survey 201 Figure A3.1 (cont’d) 202 Figure A3.1 (cont’d) 203 Figure A3.1 (cont’d) 204 Figure A3.1 (cont’d) 205 Figure A3.1 (cont’d) 206 BIBLIOGRAPHY 207 BIBLIOGRAPHY Adam, S. and H. Kriesi. 2007. “The Network Approach”. In Theories of Public Policy, ed. P. Sabatier. Bolder: Westview Press, 129-149. Agranoff, Robert, and Michael McGuire. 1998. "Multietwork Management: Collaboration and the Hollow State in Local Economic Policy." Journal of Public Administration Research and Theory 8 (1): 67-91. Ansell, C. 2008. “Network Institutionalism”. In The Oxford Handbook of Political Institutions, eds. B. A. Rockman, S. A. Binder and R. A. W. Rhodes. New York: Oxford University Press, 76-89. Balkundi, P. and M. Kilduff. 2006. "The Ties that Lead: A Social Network Approach to Leadership." The Leadership Quarterly 17(2006): 419-439. Berardo, R. 2009. "Processing Complexity in Networks: A Study of Informal Collaboration and its Effect on Organizational Success." The Policy Studies Journal 37(3): 521-539. Borgatti, S. P., A. Mehra, D. J. Brass and G. Labiance (2009). "Network Analysis in the Social Sciences." Science 323: 5916, 892-895. Borzel, T. A. 1997. "What's So Special About Policy Networks? An Exploration of the Concept and Its Usefulness in Studying European Governance." European Integration Online Papers 1(16): 1-28. Borzel, T. A. 1998. "Organizing Babylon-On the Different Conceptions of Policy Networks" Public Administration 76(Summer): 253-273. Bowman, A. O. M. 1988. "Competition for Economic Development Among Southeastern Cities." Urban Affairs Review 23(June): 511-527. Brass, D. J. and M. E. Burkhardt. 1993. "Potential Power and Power Use: An Investigation of Structure and Behavior." Academy of Management Journal 36(3): 441-470. Brass, D. J., J. Galaskiewicz, H. Greve and W. Tsai (2004). "Taking Stock of Networks and Organizations: A Multi-Level Perspective." Academy of Management Journal 47(6): 795-817. Bressers, H. and L. J. O'Toole 1998. "The Selection of Policy Instruments: A Network-Based Perspective." Journal of Public Policy 18(3): 213. 208 Bruce, G. D. and R. E. Witt. 1971. "Developing Empirically Derived City Typologies: An Application of Cluster Analysis." The Sociological Quarterly 12(2): 238-246. Burns, P. 2002. "The Intergovernmental Regime and Public Policy in Hartford, Connecticut." Journal of Urban Affairs 24(1): 55-73. Burt, R. S. 1992. Structural Holes: The Social Structure of Competition. Cambridge: Harvard University Press. Burt, R. S. 2005. Brokerage and Closure: An Introduction to Social Capital. New York: Oxford University Press. Carlsson, L. 2000. "Policy Networks as Collective Action." Policy Studies Journal 28(3): 502520. Clarke, S. E. and G. L. Gaile. 1992. "The Next Wave: Postfederal Local Economic Development Strategies." Economic Development Quarterly 6: 187-198. Clingermayer, J. C. and R. C. Feiock. 1990. "Constituencies, Campaign Support, and Council Member Intervention in City Development Policy." American Politics Quarterly 16: 5567. Dahl, R. A. (1961). Who Governs? Democracy and Power in an American City. New Haven: Yale University Press. Davies, Jonathan S. 2002. "Urban Regime Theory: A Normative-Empirical Critique." Journal of Urban Affairs 24(1): 1-17. deLeon, P. and D. M. Varda. 2009. "Toward a Theory of Collaborative Policy Networks: Identifying Structural Tendencies." The Policy Studies Journal 37(1): 59-74. DiGaetano, A. 1997. "Urban Governing Alignments and Realignment in Comparative Perspective: Developmental Politics in Boston, Massachusetts, and Bristol, England, 1980-1996." Urban Affairs Review 32: 844-870. Dowding, K. 1995. "Model or Metaphor? A Critical Review of the Policy Network Approach." Political Studies 43(1): 136-158. Dowding, Keith. 2001. "Explaining Urban Regimes." International Journal of Urban and Regional Research 25 (1): 7-19. Elkin, S. L. 1987. City and Regime in the American Republic. Chicago: University of Chicago Press. Eisinger, P. K. 1988. The Rise of the Entrepreneurial State: State and Local Economic Development Policy in the United States. Madison: University of Wisconsin Press. 209 Emerson, Richard M. 1962. "Power-Dependence Relations." American Sociological Review 27(1): 31-41. Everitt, B. (1993). Cluster Analysis. London: Edward Arnold. Everitt, B. and T. Hothorn. 2006. “Cluster Analysis: Classifying the Exoplanets”. In A Handbook of Statistical Analysis Using R. 2nd edition. Boca Raton: CRC Press. 243-257. Everitt, B., S. Landau, M. Leese and D. Stahl. 2011. Cluster Analysis. Chichester: John Wiley & Sons. Fainstein, N. I. and S. S. Fainstein. 1986. Restructuring the City: The Political Economy of Urban Redevelopment. United Kingdom: Longman Group. Feiock, R. C. and J. C. Clingermayer. 1986. "Municipal Representation, Executive Power, and Economic Development Policy Administration." Policy Studies Journal 15: 211-229. Feiock, R. C. and J.-H. Kim. 2001. "Form of Government, Administrative Organization, and Local Economic Development Policy." Journal of Public Administration Research and Theory 11(1): 29-49. Finch, H. 2005. "Comparison of Distance Measures in Cluster Analysis with Dichotomous Data." Journal of Data Science 3: 85-100. Fleischmann, A. and G. P. Green. 1991. "Organizing Local Agencies to Promote Economic Development." Annual Review of Public Administration 21(March): 1-15. Fleischmann, A., G. P. Green and T. M. Kwong. 1992. "What's a City To Do: Explaining Differences in Local Economic Development Policies." Political Research Quarterly 45: 677-699. Fleishman, J. A. 1986. "Types of Political Attitude Structure: Results of a Cluster Analysis." Public Opinion Quarterly 50(3): 371-386. Fowler, J. 2006. "Connecting the Congress: A Study of Cosponsorship Networks." Political Analysis 14: 456-487. Fredericksen, P. and R. London. 2000. "Disconnect in the Hollow State: The Pivotal Role of Organizational Capacity in Community-Based Development Organizations." Public Administration Review 60(3): 230-239. Galaskiewicz, J. and S. Wasserman. 1994. “Introduction: Advances in the Social and Behavioral Sciences from Social Network Analysis”. In Advances in Social Network Analysis: Research in the Social and Behavioral Sciences. S. Wasserman and J. Galaskiewicz, eds. Thousand Oaks: Sage: xi-1. 210 Garson, D. 2010. "Cluster Analysis." Retrieved December 2, 2010, from http://faculty.chass.ncsu.edu/garson/PA765/cluster.htm. Gibbs, D. and A. E. G. Jonas. 2000. "Governance and Regulation in Local Environmental Policy: The Utility of a Regime Approach." Geoforum 31(3): 299-313. Green, G. P. and A. Fleischmann. 1991. "Organizing Local Agencies to Promote Economic Development." American Review of Public Administration 21: 1-15. Hanson, R., H. Wolman, D. Connolly, K. Pearson and R. McManmon (2010). "Corporate Citizenship and Urban Problem Solving: The Changing Civic Role of Business Leaders in American Cities." Journal of Urban Affairs 32(1): 1-23. Hill, E. W., J. F. Brennan and H. L. Wolman. 1998. "What is a Central City in the United States? Applying a Statistical Technique for Developing Taxonomies." Urban Studies 35(11): 1935-1969. Hula, R. C., C. Y. Jackson and M. E. Orr. 1997. "Urban Politics, Governing Nonprofits, and Community Revitalization." Urban Affairs Review 32(4): 459-489. Imbroscio, D. L. 1998a. "Reformulating Urban Regime Theory: The Division of Labor Between State and Market Reconsidered." Journal of Urban Affairs 20(3): 233-248. Imbroscio, D. L. 1998b. "The Necessity of Urban Regime Change: A Reply to Clarence N. Stone." Journal of Urban Affairs 20(3): 261-268. Imbroscio, D. L. 2003. "Overcoming the Neglect of Economics in Urban Regime Theory." Journal of Urban Affairs 25: 271-284. Imbroscio, D. L. 2004. "The Imperative of Economics in Urban Political Analysis: A Reply to Clarence N. Stone." Journal of Urban Affairs 26(1): 21-26. Johnson, K. 2004. "Community Development Corporations, Participation, and Accountability: The Harlem Urban Development Corporation and the Bedford-Stuyvesant Restoration Corporation." Annals of the American Academy of Political and Social Science 594(July 2004): 109-124. Keating, M. 1991. Comparative Urban Politics: Power and the City in the United Staes, Canada, Britain, and France. Aldershot: Edward Elgar. Kenis, Patrick, and Volker Schneider. 1991. "Policy Networks and Policy Analysis: Scrutinizing a New Analytical Toolbox." In Policy Networks. Empirical Evidence and Theoretical Considerations, eds Bernd Marin and Renate Mayntz. Boulder: Campus Verlag/Westview Press, 25-62 211 Kettl, Donald F. 2000. "The Transformation of Governance: Globalization, Devolution, and the Role of Government." Public Administration Review 60 (6):488-497. King, J. E. 2007. Binary Logistic Regression. Best Practices in Quanitative Methods, ed. J. W. Osborne. Thousand Oaks: Sage Publications, Chapter 24 Knoke, D. 1990. "Networks of Political Actions: Towards Theory Construction." Social Forces 68, 1041-63. Knoke, D. and S. Yang. 2008. Social Network Analysis. 2nd ed. Thousand Oaks: Sage. Kwon, T. M., F. Berry and R. C. Feiock. 2009. "Understanding the Adoption and Timing of Economic Development Strategies in U.S. Cities Using Innovation and Institutional Analysis." Journal of Public Administration Research and Theory 19: 967-988. LaMore, R. L., T. Link and T. Blackmond. 2006. "Renewing People and Places: Institutional Investment Policies that Enhance Social Capital and Improve the Built Environment of Distressed Communities." Journal of Urban Affairs 28(5): 429-442. Laumann, E. O., P. V. Mardsen and Galaskiewicz. 1977. "Community-Elite Influence Structures: Extension of a Network Approach." American Journal of Sociology 83(3): 594. Laumann, E. O. and D. Knoke. 1987. The Organizational State. Madison: University of Wisconsin Press. Laumann, E. O. and F. U. Pappi. 1973. "New Directions in the Study of Community Elites." American Sociological Review 38(2): 212-230. Lauria, M. 1999. “Reconstructing Urban Regime Theory: Regulation Theory and Institutional Arrangements”. In The Urban Growth Machine: Critical Perspectives, eds. Jonas, A.E.G. and Vilson, D. Albany: SUNY Press. Logan, J. R. and H. L. Molotch. 1987. Urban Fortunes: The Political Economy of Place. Berkely: University of California Press. Mardsen, P. V. and N. E. Friedkin. 1994. “Network Studies of Social Influence”. In Advances in Social Network Analysis, eds. S. Wasserman and J. Galaskiewicz. Thousand Oaks: Sage. Mardsen, P. V. 1990. "Network Data and Measurement." Annual Review of Sociology 16: 435463. Marsh, D. and M. Smith. 2000. "Understanding Policy Networks: towards a Dialetctical Approach." Political Studies 48: 4-21. Mason, S. and K. P. Thomas. 2010. "Tax Increment Financing in Missouri: An Analysis of Determinants, Competitive Dynamics, Equity, and Path Dependency." Economic 212 Development Quarterly 24(2): 169-179. McGuire, M. 2000. "Collaborative Policy Making and Administration: The Operational Demands of Local Economic Development." Economic Development Quarterly 14(3): 276-291. Melbeck, Christian. 1998. “Comparing Local Policy Networks.” Journal of Theoretical Politics 10(4): 531-552. Molotch, H. L. 1976. "The City as a Growth Machine." American Journal of Sociology 82(2): 309-331. Mossberger, Karen , and Gerry Stoker. 2001. "The Evolution of Urban Regime Theory: The Challenge of Conceptualization." Urban Affairs Review 36(6):810-835. Myers, R. H., D. C. Montgomery, V. G. Geoffrey and T. J. Robinson. 2012. Generalized Linear Models: with Applications in Engineering and the Sciences. Hoboken: John Wiley & Sons. Olberding, J. C. 2002. "Diving into the "Third Waves" of Regional Governance and Economic Development Strategies: A Study of Regional Partnerships for Economic Development in U.S. Metropolitan Areas." Economic Development Quarterly 16(3): 173-186. Orr, M. E. and G. Stoker. 1994. "Urban Regimes and Leadership in Detroit." Urban Affairs Review 30(1): 48-73. Pappi, F. U. and C. H. C. A. Henning. 1998. "Policy Networks: More than a Metaphor." Journal of Theoretical Politics 10(4): 553-575. Peters, B. Guy, and John Pierre. 1998. "Governance without Government? Rethinking Public Administration." Journal of Public Administration Research and Theory 8(2):223-243. Peterson, P. E. 1991. City Limits. Chicago: University of Chicago Press. Provan, K. G., A. Fish and J. Sydow. 2007. "Interorganizational Networks at the Network Level: A Review of the Empirical Literature on Whole Networks." Journal of Management 33(3): 479-519. Provan, K. G. and P. Kenis. 2007. "Modes of Network Governance: Structure, Management and Effectiveness." Journal of Public Administration Research and Theory 18: 229-252. Reese, L. 1997. Local Economic Development Policy: The U.S. and Canada. New York, Garland Publishing. Reese, L. A. and R. A. Rosenfeld. 2002. "Reconsidering Private Sector Power: Business Input and Local Development Policy." Urban Affairs Review 37(5): 642-674. 213 Rhodes, R.A.W. 1996. "The New Governance: Governing Within Government." Political Studies 44:652-667. ———. 1997. Understanding Governance: Policy Networks, Governance, Reflexivity, and Accountability. Buckingham: Open University Press. ———. 2007. "Understanding Governance: Ten Years On." Organization Studies 28 (8):12431264. Rhodes, R. A. W. and D. Marsh. 1992. "New Directions in the Study of Policy Networks." European Journal of Political Research 21(1-2): 181-205. Robinson, T. 1996. "Inner-city Innovator: The Non-profit Community Development Corporations." Urban Studies 33(9): 1647-1670. Rowley, T. J. 1997. "Moving Beyond Dyadic Ties: A Network Theory of Stakeholder Influences." Academy of Management Review 22(4): 887-910. Rubin, H. J. 1986. "Local Economic Development Organizations and the Activities of Small Cities in Encouraging Economic Growth." Policy Studies Journal 14: 363-388. Rubin, I. S. and H. J. Rubin. 1987. "Economic Development Incentives: The Poor (Cities) Pay More." Urban Affairs Review 23: 37-62. Saidel, J. R. 1991. "Resource Interdependence: The Relationship between State Agencies and Nonprofit Organizations." Public Administration Review 51(6): 543-553. Salamon, Lester M. 2002. The Tools of Government: A Guide to the New Governance. New York: Oxford University Press. Sandstrom, A. and L. Carlsson. 2008. "The Performance of Policy Networks: The Relation between Network Structure and Network Performance." The Policy Studies Journal 36(4): 497-524. Sandstrom, A. and C. Rova. 2010. "The Network Structure of Adaptive Governance: A Single Case Study of a Fish Management Area." International Journal of the Commons 4(1): 528-551. Schneider, M. 1989. The Competitive City. Pittsburgh: University of Pittsburgh Press. Sharp, E. B. 1990. Urban Politics and Administration. New York: Longman. Sharp, E. B. 1991. "Institutional Manifestations of Accessibility and Urban Economic Development Policy." Western Political Quarterly 44(1): 129-147. 214 Sharp, E. B. and D. Elkins. 1991. "The Politics of Economic Development Policy." Economic Development Quarterly 5(2): 126-139. Sites, W. 1997. "The Limits of Urban Regime Theory: New York City under Koch, Dinkins, and Giuliani." Urban Affairs Review 32(4): 536. Stoker, Gerry. 1995. “Regime theory”. In Theories of Urban Politics, eds. D. Judge, Stoker, G. and Wolman, H. London: Sage. Stoker, Gerry. 1998. "Governance as Theory: Five Propositions." International Social Science Journal 50(155): 17-28. Stone, Clarence N. 1980. “Systemic Power in Community Decision-Making: A Restatement of of Stratification Theory”. American Political Science Review 74(4): 978-990. Stone, C. N. 1989. Regime Politics: Governing Atlanta, 1946-1988. Lawrence: University of Kansas. Stone, C. N. 1993. "Urban Regimes and The Capacity to Govern: A Political Econony Approach." Journal of Urban Affairs 15(1): 1-28. Stone, Clarence N. and Sanders, H.T. 1987. The Politics of Urban Development. Lawrence: University Press of Kansas. Swanstrom, T. 1985. "Semisovereign Cities: The Politics of Urban Development." Polity 21(1): 83. Thatcher, M. 1998. "The Development of Policy Network Analyses: From Modest Origins to Overarching Frameworks." Journal of Theoretical Politics 10(4): 389. Tiebout, C. M. (1956). "A Pure Theory of Local Expenditures." Journal of Political Economy 64(5): 416-424. Ward, Kevin. 1996. "Rereading Urban Regime Theory: a Sympathetic Critique." Geoforum 27(4): 427-438. Wasserman, S. and K. Faust. 1994. Social Network Analysis: Methods and Applications. New York: Cambridge University Press. Weir, M., H. Wolman and T. Swanstrom. 2005. "The Calculus of Coalitions: Cities, Suburbs and the Metropolitan Agenda." Urban Affairs Review 40(6): 730-760. Wolman, H. and D. Spitzey. 1996. "The Politics of Local Economic Development." Economic Development Quarterly 10(2): 115-150. 215