Urban Affairs Review

Urban Affairs Review http://uar.sagepub.com/

Collaboration Networks Among Local Elected Officials: Information, Commitment, and Risk Aversion Richard C. Feiock, In Won Lee, Hyung Jun Park and Keon-Hyung Lee Urban Affairs Review published online 3 May 2010 DOI: 10.1177/1078087409360509 The online version of this article can be found at: http://uar.sagepub.com/content/early/2010/08/18/1078087409360509

Published by: http://www.sagepublications.com

On behalf of:

The Urban Politics Section, American Political Science Association

Additional services and information for Urban Affairs Review can be found at: Email Alerts: http://uar.sagepub.com/cgi/alerts Subscriptions: http://uar.sagepub.com/subscriptions Reprints: http://www.sagepub.com/journalsReprints.nav Permissions: http://www.sagepub.com/journalsPermissions.nav

Downloaded from uar.sagepub.com at UNIV OF COLORADO LIBRARY on November 10, 2010

Urban Affairs Review OnlineFirst, published on August 20, 2010 as doi:10.1177/1078087409360509

Collaboration Networks Among Local Elected Officials: Information, Commitment, and Risk Aversion

Urban Affairs Review XX(X) 1–22 © The Author(s) 2010 Reprints and permission: http://www. sagepub.com/journalsPermissions.nav DOI: 10.1177/1078087409360509 http://uar.sagepub.com

Richard C. Feiock,1 In Won Lee,2 Hyung Jun Park,3 and Keon-Hyung Lee1

Abstract Institutional collective action (ICA) problems arise from fragmentation of service responsibilities among a multitude of governments and authorities in metropolitan areas. Many previous empirical evidences suggest that fragmented governments are able to collaborate to share information even in the highly competitive economic development arena. This situation makes local officials in metro areas confront inevitable trade-offs between the potential benefits of interdependent policy action and uncertainties around selective incentives of participating governments. Especially, elected officials’ aversion to risk and short time horizons make them sensitive to possibilities of being misled by wrong or irrelevant information.The authors contend that elected officials seek partners embedded in a structure of relationships that can reduce the political risks of collaboration as well as provide information, commitment that is necessary to overcome the barriers to collective action. The authors apply an exponential random graph (p*) model to test if certain hypothesized network structures are more likely to emerge in self-organizing economic development networks among local elected officials. 1

Florida State University, Tallahassee, FL Dankook University,Yongin, Korea 3 Sungkyunkwan University, Seoul, Korea 2

Corresponding Author: In Won Lee, Department of Public Administration, School of Social Science Dankook University,Yongin, Korea Email: [email protected] Downloaded from uar.sagepub.com at UNIV OF COLORADO LIBRARY on November 10, 2010

2

Urban Affairs Review XX(X)

The results generally confirm that when there are greater commitment risks to collaboration, the actors are more likely to create mechanisms that can enhance trustworthiness to resolve cooperative problems. Keywords network embeddedness, Institutional Collective Action (ICA), political risks, Exponential Random Graph Model (ERGM), coordination and cooperation The organization of governance and public authority in metropolitan regions and its consequences have been a keen interest of urban scholars for over 40 years. Recent interest in governance and intergovernmental collaboration has introduced a new generation of scholars to this debate. Fragmentation of policy making among multiple governmental units diminishes problems of concentrated powers and can also promote competition and innovation, but it also imposes inefficiencies as decisions by one governmental unit impose positive and negative externalities on others. Urban service delivery decisions must confront scale and externality problems and other transaction risks that are exacerbated by fragmented patterns of local governments that provide customized mixes of public goods to suit the taste of the local constituency. Metropolitan regions are recognized as perhaps the most salient economic and social units in a global economy (Feiock, Moon, and Park 2008), but costs and conflicts from fragmentation are particularly acute in metropolitan areas because authority is often fragmented horizontally among competing local governments as well as vertically among overlapping specialized national, provincial, and local agencies. Thus local government officials in metro areas confront inevitable trade-offs if they seek to obtain the potential benefits of policy coordination in the face of uncertainty and commitment problems.

The Institutional Collective Action Perspective Institutional collective action (ICA) problems arise from fragmentation of service responsibilities among a multitude of governments and authorities in metropolitan areas. Local to local collaboration in response to the ICA problems can produce joint gains, but these gains may not be sufficient to motivate collective action. The ICA framework extends theories of contracting and individual-level collective action to institutional actors such as cities, counties, and government agencies (Feiock 2009; Feiock and Scholz 2009). The ICA approach builds on agency and social network theories to identify benefits and transaction costs of collaboration (Lubell et al. 2002; Schneider


Feiock et al.

3

et al. 2003; Feiock and Scholz 2009). If we assume that institutional actors act in a self-interested manner by selecting the available strategy that most enhances their (generally short-term) interests, then absent regional institutions, the collective action problem dictates that the outcomes of individual decisions will lead to collectively inefficient decisions. After reviewing the mechanisms for integrating policy decisions among fragmented actors, we focus on informal policy networks as a potential solution to ICA dilemmas. We then examine the motivations, incentives, and risks for local elected government officials when they share information and resources with officials in other governments. We argue that elected officials seek partners embedded in a structure of relationships that can reduce the political risks of collaboration as well as provide information, commitment that is necessary to overcome the barriers to collective action. An exponential random graph analysis based on an original network survey of self-organizing economic development policy networks in the Orlando metropolitan area is then presented. The analysis investigates how the career incentives and political risk aversion of elected officials influence their choices of collaboration partners in economic development. We hypothesize that elected officials minimize political risk by forming ego-centered network structures. The analysis also examines how information bridging and tightly clustered networks serve as alternative mechanisms to overcome ICA problems. In conclusion, the implications of the analysis for regional governance are discussed and an agenda for future research is developed.

Regional Integration Mechanisms There are three general mechanisms available to integrate decision making across local government units in metropolitan area: centralized authority, mutually binding contracts or agreements, and network embeddedness (Feiock 2004, 2009). Centralized authority means a third party or higher level government intervenes to consolidate authority and internalize ICAs, contracting means fragmented governments legally bind themselves to mutual action, and embeddedness means agreements among local units are coordinated and enforced through a network of social, economic, and political relationships rather than formal authority (Feiock 2009). These are not mutually exclusive and, at least in some instances, can complement each other (Feiock and Scholz 2009).

Centralized Authority When confronted with collective problems because of fragmentation, higher level institutions have the authority to resolve the problems by changing the


4


geographic or functional jurisdictions to “internalize” the externalities caused by fragmentation. For example, the amalgamation of consolidation of local government units internalizes unconsidered impacts within broad metropolitan boundaries by uniting multiple local governments into a consolidated metropolitan general purpose government. Special districts and designed networks also rely on centralized authority to internalize unconsidered impacts over a broad geographic area for specific functions. For example, states often authorize special districts governments to consolidate the delivery of services such as schools, water, and fire protection. With mandated or “designed” networks, a higher level government provides funding and mandates the formation of collaborative relations among local governmental actors or designates a lead agency or organization to manage and coordinate intergovernmental service provision (Provan and Kenis 2008). The political and administrative costs of relying on centralized authority as a regional integration mechanism often limit the scope of its use. Existing government units generally resist their loss of authority, and even if production efficiencies are possible they are achieved at the cost of reducing the ability of local units to vary the provision of services to reflect different local preferences.

Mutually Binding Agreements Contracts and regional organizations require the consent of those involved, so this institutional system preserves the autonomy of local actors while providing a more formalized mechanism for resolving externality issues of concern to all parties. To the extent that enabling legislation minimizes the transaction costs involved in developing, negotiating, and enforcing agreements, the Coase (1960, 1988) theorem suggests that contracting can resolve many of the diverse externality problems. The ability to make binding agreements can be very flexible, involving bilateral agreements between two governments or in other cases setting up a voluntary organization that binds the local governments to some degree but relies on mutual consent. They are formally constituted but voluntary in the sense that members participate at will and must approve activities. These institutions generally have limited authority to force members to do what they do not want to do, and the forces of both cooperation and competition remain within the arrangements (Gerber and Gibson 2006).

Network Embeddedness Embeddedness relies on social, economic, and political relationships rather than formal authority. Self-organizing policy networks that rely on this


Feiock et al.

5

mechanism offer several potential advantages over more formal solutions. By preserving the autonomy of the actors in a given dilemma, self-governing institutions avoid the inevitable political conflicts involved in revoking existing authority from local governments or specialized agencies. By generally requiring consent of all members, self-governing institutions enhance the search for a mutually advantageous resolution of ICA and reduce the conflicts involved when majorities can impose solutions on unwilling minorities. By ensuring sufficient flexibility for rules, procedures, and exchanges to be decided locally, self-governing institutions can customize these rules to best fit the local conditions and specific ICA situation. To the extent that self-governing institutions contribute to local social capital, the resolution of one ICA can provide the basis for resolving unrelated ICAs in the same region as well. Informal policy network structures emerge unplanned from interactions among institutional actors. Informal networks coordinate complex decisions within the formal structure. They preserve full local autonomy and require no formal authority, although federal and state programs can influence their development (Schneider et al. 2003). Policy networks can also complement central authority or contracts. The formal authority structures in political systems rely on informal, selforganized relationships among authorities for performance and stability to buffer the system from changing demands. Network interactions that tie responses to actions in an area to those in other policy areas or across time help members identify partners that are less likely to defect and build an enforcement structure to reduce the transaction costs.

Network Structure and Incentives for Collaboration While there are well-developed literatures on regional authorities and consolidations (Carr and Feiock 2004) and on intergovernmental contracting (Carr, LeRoux, and Shrestha 2008; Andrew 2009), we know very little about the role of intergovernmental policy networks in coordinating action in metropolitan areas. Moreover, the existing literature has focused almost exclusively on administrators, neglecting networks among elected officials.

Elected Officials and Political Risk The potential benefits of interdependent actions motivate elected officials to engage in ICA. However, uncertainties around selective incentives of participating governments make collaboration politically risky for these officials. In addition, the attitudes toward uncertainty and ways of addressing risks differ


6


among policy actors. Steinacker (2002) argues that elected officials are more risk averse and have shorter time horizons than do appointed officials. Politicians’ exposure to different risks depends on their distinct incentives. Local elected officials with ambition for higher office may utilize intergovernmental relations to promote themselves to a larger constituency. Progressively ambitious politicians engage in regional and intergovernmental collaboration to claim credit for services provided to individuals beyond their current constituencies (Clingermayer and Feiock 1997; Bickers, Stein, and Post 2009; McCabe et al. 2008). The consent of elected officials is typically needed for even informal collaboration and information sharing, making elected officials veto players (Tsebelis 2002). Gerber and Gibson (2006) argue that the underlying political dilemma associated with regional governance is that local officials need to give up some authority or take political risks by sharing information or coordinating action with other governments. Even if there are joint benefits of seeking information or assistance from other jurisdictions, elected officials may be unwilling to subject themselves to political risks or the potential to be misled. The fragility of electoral support in the community makes them sensitive to the political costs of their actions. Therefore, their information-seeking activities are expected to be ego centered in that they do not rely on information from only a single external source. Rather, they invest the time and resources necessary to build relationships with a diverse network of actors to be able to verify the information they receive. Creating additional relations takes a substantial amount of time, and it offers mostly redundant information, but maintaining multiple information sources is critical because the political consequences of being misled by wrong or irrelevant information are substantial, and typically elected officials do not regularly interact with each other through professional associations or affiliations. Thus, in building an ego-centered network, elected officials pursue political efficiency. Steinacker (2002) also argues that because elected officials are politically risk averse, their networking activities are constrained by their jurisdictional boundary. Access to other network actors may be limited to only their geographically neighboring governments. Therefore, elected officials’ networks are maintained in an ego-centered way when looking for links to other actors. Elected officials’ aversion to risk and short time horizons make them sensitive to the possibilities of being misled by wrong or irrelevant information. This risk can be controlled by maintaining multiple information sources. Even if creating the additional links involves additional time and resources, it provides information verification and self-constraint mechanisms. Network actors tend to look for reliable and valuable information. Sometimes this goal


Feiock et al.

7

is easily achieved when simply linking to coordinators or brokerage actors. This may be the most efficient way of collecting information when less risk is involved in collective action. However, in the absence of such a coordinating mechanism, network actors are eager to closely monitor the quality and magnitude of information. To achieve this goal, they prefer to maintain more than one information source. In other words, they try to verify the quality of information that they receive from one actor by reaching out to other actors. On the other hand, although having multiple relations is advantageous, it always entails some costs, and this is particularly burdensome when a new relationship brings redundant information. Therefore, cost considerations of network actors act as a self-constraint for network expansion as well. Thus we hypothesize, Hypothesis 1: Since elected officials are risk averse and have short time horizons, they are likely to attempt to process and screen the information directly and independently, which leads them to create ego-centered network structures. This is captured by “network activity (expansiveness)” and measured by “out-stars.”

Information and Commitment We also need to account for the need to acquire information and to ensure the credibility of commitments in investigating the intergovernmental networks of elected officials. The challenge for self-organized solutions is to overcome the transaction cost barriers to collective action. Efficient transmission of information and the building of reputations and trust provide alternative mechanisms to achieve ICA depending on whether the problem is one of coordination or cooperation (Berardo and Scholz 2007; Feiock and Lee 2008; Feiock and Scholz 2009). Even when interests are shared, search costs for information are a barrier to cooperation. The network literature indicates that certain types of structures, such as a central actor or “weak tie” relationships, facilitate the efficient transfer of information to solve coordination problems. Local officials may seek to contact someone who already possesses reliable and sufficient information by virtue of being connected to others. This leads to the emergence of actors playing an information-coordinating role. From individual actors’ viewpoint, more information about the choices of others increases the likelihood that more actors will benefit by adopting the most common approach to common problems (Berardo and Scholz 2007). The existence of popular actors is especially important when there is no


8


B + A Figure 1. Bridging network structure: 2-path

central authority but there is still a need for coordination on information flows. From a networkwide perspective, a “star” with all actors connected to only one central actor provides the most efficient information transmission mechanism. With just a single link per actor, everyone could potentially coordinate with all others by following the advice of the central coordinator; with just a single link per actor, everyone could potentially benefit from the same information and advice. Hypothesis 2a: When officials seek efficient information transmission, they are likely to count on “popular” actors who serve as central coordinators. This is measured by “in-stars.” While a “star” with all actors connected to only one popular actor provides the most efficient information transmission structure, actors who bridge structural holes also enable others to gain knowledge and the advantages of innovation that are not available within a highly clustered network. In Figure 1, actor B maintains relations with many other actors, and this may be common knowledge to every actor in the network. Then, actor A is more likely to exploit the focal actor’s (actor B) position to exchange the information rather than investing in creating many new relations on his or her own. In this sense, a bridging structure also provides a mechanism to transmit information about what others do and know. Therefore, Hypothesis 2b: When information transmission through brokerage is a critical economic development issue, the actors are more likely to count on bridging actors. This is measured by “2-paths.”


Feiock et al.

9

For policies where commitment issues are barriers to cooperation, local governments may seek more closely connected relationships that are better suited to solve problems of cooperation (Coleman 1988; Putnam 2000). Trust, mutuality, and reputations that are manifestations of networked relationships facilitate collective action (Granovetter 1985; Lubell et al. 2002). Reciprocity is the most fundamental norm in most collective action situations (Ostrom 1998; Thurmaier and Wood 2002). Putnam (1995) emphasized the importance of reciprocal ties of mutual cooperation in the development of social capital. Mutual exchanges create the shadow of the future in which defection by one actor can be punished by future defection (Axelrod 1984), thus providing a mutual deterrence on which credible commitments can develop. Confronted with uncertainty of cooperation with potential competitors, local actors attempt to overcome the credible commitment problem by exercising “the norm of reciprocity.” Even informal relationships for discussion and advice are governed by reciprocity to deter opportunism and the free-rider problem. Therefore, Hypothesis 3a: When trustworthiness and credible commitments matter in cooperative activities, the actors are more likely to forge reciprocal relations. This is measured by “reciprocal ties.” Overcoming the credible commitment problem extends the argument beyond the dyadic level to a social group dimension. For instance, from a social capital perspective, authors such as Putnam (1995) and Coleman (1988) argue that denser, “closed” networks reduce the cost by making control activities effective through overlapping information about each other’s behavior that circulates in the network. This threat of detection and punishment of noncooperative activity increases the chances of cooperation among actors. In this sense, regardless of the actors’ incentives, we expect that network actors will forge a tightly clustered network structure to enhance credible commitment. When shirking is a potential problem in the delivery of collective outcomes, the possibility of shirking imposes costs on those who have already invested their time and resources. A tightly clustered network structure is advantageous in reducing the transaction costs of enforcing and monitoring the relational obligations since any actions taken or not taken by a locality are made public (Feiock and Scholz 2009). A group with links connecting all members to each other can prevent defection more effectively because of the rich information about each other’s behavior and the possibilities for combined punishment. The enhanced detection and punishment of noncooperative


10


activity should increase credibility of commitments and cooperation among actors in such closed groups. In particular, the transitive triad serves as the most basic network structure that represents the advantages of a tightly clustered network since transitivity in a relationship may indicate social trust being built in a social context. Hypothesis 3b: When trustworthiness and credible commitments matter in cooperative activities, the actors are more likely to make groups more tightly clustered. This is measured by “transitivity.” The following sections empirically test these hypotheses based on data on informal policy networks in the Orlando, Florida, metropolitan area. Exponential random graph analysis is applied to test whether self-organizing economic development networks among local elected officials reflect the egocentered structure that minimizes political risk and whether elected officials create tightly clustered network structures to overcome ICA problems rather than information-bridging mechanisms.

Analysis The Policy Arena: Economic Development When elected officials see opportunities for gains from coordination, or they need help or assistance in their development efforts, to whom do they turn? Economic development is an ideal policy arena for examining how the incentives of elected officials shape overall configuration of collaborative relationships because the development policy arena is characterized by competition. Competition for business and a tax base, in combination with intergovernmental fragmentation, makes even the most beneficial cooperation risky for elected officials. The dilemma for local actors working on economic development is that cooperation with other jurisdictions provides tremendous potential benefit if it results in new development, but it also poses greater risks than cooperation in most other areas. This arena allows us to extend the literature that focuses on efficient transmission of information and trustworthiness as alternative mechanisms to achieve ICA. From this perspective, network structures that channel the efficient transmission of information are associated with mechanisms to solve coordination problems, while network structures that ensure trustworthiness are associated with mechanisms to resolve cooperative problems. Coordination and cooperation impose different incentives for elected officials seeking partners (Scholz, Berardo, and Kile 2008; Feiock 2009). Links


Feiock et al.

11

with popular coordinators may increase the ability of the actor to share the gains of coordinated activities but also expose the actor to risks of defection. This creates a desire to rely on members of densely connected networks capable of maintaining credible commitments for sensitive or risky information. The analysis focuses on motivations and incentives for elected local government leaders and how their propensities to be risk averse shape their willingness to engage in economic development exchanges and the structure of these networks from which they draw their partners. We apply the p* model, a powerful analytic technique of exponential random graph analysis, to test our hypotheses.

The Actor-Oriented Model Snijders (2001, 2002) advances an “actor-oriented” network model that allows us to make predictions about the structure of the network based on the incentives and the motivations of the actors. His approach provides a theoretical structure for developing and testing hypotheses linking networks and the individual incentives involved in collective action situations. The overall structure of the network emerges from the utility-maximizing choices of all actors in the policy arena, constrained by the initial, exogenously determined network structure and distribution of attributes. An exponential random graph model (aka the p* model) can then be employed to test hypotheses about the likelihood that network structures predicted by the actors’ objective function are present in the cross-section data for political and administrative development interactions. The p* model estimates the probability that the several different network structures included in the model appear at a greater frequency than would be explained by a random graph with the same number of local government actors (nodes) and development interactions (links) (Snijders et al. 2005; Snijders et al. 2006). The model is conditioned on the observed networks and takes into account the potential random relations, the relative dependence of all relationships, and the effect of actor attributes for the robust test of the structural features of the observed network of economic development interactions. The distribution of random graphs is simulated from a starting set of parameter values, and these estimated parameter values are refined by comparing the distribution of graphs to the observed graph (Wasserman and Robins 2005). Thus the estimated parameters basically indicate the likelihood of particular structural effects being observed in the data (Robins et al. 2007). By estimating the parameters that best represent an observed network, we can test the relative importance of each structural characteristic and attribute implicit in the


12


network. To test the hypotheses, we use the p* exponential random graph model to estimate several network structure statistics including reciprocity, transitivity, in- or out-stars, and so on, capturing the motivations of actors in observed policy networks in the Orlando metropolitan area.

Data and Sample A total of 40 city and country governments in the four counties (Lake, Orange, Osceola, and Seminole) within the Orlando, Florida, metropolitan area were surveyed in 2006–2007. Of these 40 governments, 31 responded (78% response rate). These city and county governments demonstrate substantial variance in their size and other socioeconomic attributes. The survey was sent to the top economic development official in each city. The survey asked economic development administrators in each of these governments to report which governments and organizations the elected officials in their governments interacted with (including discussion, advice, information sharing) on economic development issues in the past year and how frequently. These knowledgeable third-party reports, as a way to identify the network activities of elected officials, are intended to provide accurate information in an objective manner. The specific questions were originally designed to measure the intensity of network links among actors (elected officials) ranging from 0 to 5 (0 = no interaction, 1 = annual-basis interaction, and 5 = daily-basis interaction). Although network activities can occur on several different dimensions including discussion, advice, information sharing, and so on, we asked respondents to report a more general aspect of network activities. We were thus able to collect information on the level of network activities in a single dimension (informal and communication network). The information was then collapsed to a dichotomous variable coded 0 if there was no interaction and 1 if an annual or more frequent interaction occurred. In other words, only 0 (no interaction) was coded as 0, and all other responses (1–5 = from annualbasis interaction to daily-basis interaction) were coded as 1. Conceptually, this cut point captures the greatest interval. In practical terms, this coding is desirable because a large portion of our dyadic data set includes 0s, representing the absence of links based on this criterion (total dyad count = 465; mutual = 17; asymmetric = 88; null = 360); any other cut point would significantly increase the number of null dyads, leading to difficulties in obtaining good convergence diagnoses and drawing substantive meaning from the analysis. This information was translated into a 31 × 31 adjacency matrix of links for each of the participating actors, where 1s represent existing interactions


Feiock et al.

13

and 0s represent the absence of such interactions. In these adjacency matrices, when government A has identified government B as an interacting partner, row A of column B has a value of 1. In this sense, the values in the matrix represent directed relationships between two governments and the adjacency matrix is asymmetric in its structure. We can then estimate which of the hypothesized network structures are most desired by the surveyed local governments on economic development issues and investigate elected officials’ network structures. This model exclusively focuses on network structure parameters based on a lower order (Markov) model that investigates what types of network structures are more likely to occur. We fit several different models either to explore different parameter combinations if good convergence was difficult to achieve or to find the model that best represents the data (Robins et al. 2007). The convergence t-ratios reported indicate how well the network estimates converged, with a small value close to zero indicating good convergence. In fitting the models, we begin with a basic structure including degrees, reciprocity, transitivity, and in- or out-star parameters. If these models did not converge, we tried either simpler or more complex structures. Once we finalized the best fit model, we estimated the network structure parameters and tested if these values were statistically significant.

Exponential Random Graph Model The p* model in the SIENA (Simulation Investigation for Empirical Network Analysis) program is utilized to test the likelihood of the presence of the hypothesized network structures for political relationships.1 The p* model (as described in Snijders et al. 2006) uses the Metropolis–Hastings algorithm for generating random draws from an exponential random graph distribution and employs the stochastic approximation algorithm to estimate the pattern of relationships.2 It estimates the probability that the network structures included in the model appear at a greater frequency than would be explained by a random graph with the same number of local government actors (nodes) and relationships (links). Thus, in this model, local configurations predict the global structure (Steglich et al. 2010). The model must not only be conditioned on the observed networks but also take into account the potential random relations, the relative dependence of all relationships, and the effect of actor attributes for the robust test of the structural features of the observed network of economic development


14


Table 1. Hypotheses for Network Structures Hypothesis

Measure

Elected officials: Ego-centered networking Activity

Formal definition

Out-2-stars

∑ ∑ Xih Xik

Out-3-stars

∑ ∑ ∑ Xih Xik Xil

Cooperation: Tightly clustered Reciprocity

Reciprocity

∑j Xij Xji

Bonding

Transitivity triplets

Coordination: Information bridging Popularity

In-2-stars

∑ ∑ Xhi Xki

In-3-stars

∑ ∑ ∑ Xhi Xki Xli

2-paths

∑ ∑ Xhi Xki

Bridging

i h