Abstract
Networks have become a widely utilized method for dealing with emergencies and disasters. Researching this aspect of the emergency management process can effectively contribute to the outcome of disaster response and recovery by providing insight into the operations and processes of disaster management as well as ways to better utilize network resources and relationships. This article compares vertical and horizontal networks using the example of two counties in Florida. Understanding these network approaches contributes to efficient emergency management networks, thus strengthening outcomes in disaster response and recovery. The findings offer practical implications for local-level emergency management agencies.
Keywords
The catastrophic disasters of recent years have proved to show deficiencies in current emergency management systems across the world. Systems characterized by rigid boundaries, a top-down approach, and command-and-control mechanisms have proved to be less effective (Bier, 2006; Weick, 1993). As a result, the latest research has focused more on developing emergency management mechanisms that would fit the multi-faceted and complex context of disasters, as well as highlight the need for collaborative and network approaches (Comfort, 2007; Hackman, 2011; Kamensky, Burlin, & Abramson, 2004; Waugh & Streib, 2006), which is reflected in practice in the form of networked governance (Rhodes, 1996). The concept of networked governance is based on the idea that organizations are inter-dependent, and, thus, should work together to achieve a commonly desirable goal (Kapucu, 2006a; Klijn & Koppenjan, 2000; Provan, Fish, & Sydow, 2007; Thompson, 1967). The main mode of interaction and collaboration among these organizations is networks, characterized by comparatively less rigidity, more flexibility, and non-hierarchical structures targeted at loose relationships for more effective results (Bryson, Crosby, & Stone, 2006). Also, the opportunity and benefit of shared resources, personnel, information, and expertise make networks a viable and preferred approach to tackle complex problems (Lazer & Friedman, 2007; McDonald, 2008).
Collaborative networks in emergency management offer non-traditional tools to overcome the structural and relational handicaps of less flexible, command-and-control-oriented preparedness, response, and recovery systems (Kapucu, 2008; Ward & Wamsley, 2007). The emergency management field, in this regard, presents a unique case for analysis, in which the collaborative approach takes several forms, especially due to the variety of specificities at the local level across the country. The first mechanism is based on Emergency Support Functions (ESF)—a national system replicated by state and local governments. Being an inherently flexible structure for the coordination of emergency management operations, the ESF-based system envisions classification of resources and capabilities into various ESFs that are activated based on need during disasters. The second mechanism, on the other hand, is based on the Incident Command System (ICS)—a relatively more hierarchical command-and-control system with functional arrangements around key resources and capabilities similar to those in the ESF-based system.
Examining two different metropolitan counties in Florida, which use the two approaches, this article examines the micro (organizational level) and macro (network level) differences between emergency response mechanisms, and argues that the two produce different results in terms of expected structural and positional performance. The main research questions addressed in the article are as follows: What are the structural and relational differences between horizontal and hierarchical network arrangements? And, what consequences do the structural differences in the network bear in terms of overall expected performance? These questions were examined in the context of emergency management networks. Regression and network analyses were utilized to answer the research questions and derive points for methodological and theoretical discussion. The comprehensive emergency management plans (CEMPs) of the counties and surveys completed by the emergency management agencies identified in each metropolitan county were analyzed. This article seeks to provide further theoretical insight into the discussion about collaborative networks and intends to contribute to the literature as well as the practice of emergency management networks. It also seeks to add empirical rigor in conceptualizing networks and differences in network governance systems in emergency management.
Literature Review and Background
This section presents an account of previous research on networks, their application in the field of emergency management and relevant emergency management network types, as well as how performance is affected by network characteristics. The section ends with hypotheses to be tested in this study.
Collaborative Networks
Network governance entails a decision-making process by multiple stakeholders, governmental, non-governmental, and community-based actors (Brunner et al., 2005; Freeman & Peck, 2007; Hill & Lynn, 2009; Kapucu, 2012; Kapucu & Ozerdem, 2013). It is a different way of governing (Rhodes, 1996), in which public agencies directly engage non-state actors in a deliberative and consensus-oriented decision-making process (Ansell & Gash, 2008) producing a collective action (Milward & Provan, 2000). This form of governance also reflects unbinding inter-sector and inter-governmental relationships in a decentralized structure (Halachmi, 2005).
Networks can be understood as an aggregation of structural arrangements due to the belief that structure—whether horizontal or vertical—affects outcome (Hammond, 1986). They have become a preferred tool for delivering services by all sectors and levels of government today, and are seen as high-performing arrangements (Walker, Andrews, Boyne, Meier, & O’Toole, 2010). Networks, based on relational (strong dyadic relationships and reciprocated positive ties) and structural embeddedness (sharing partners, collaboration, cooperation, positive tie triangulation), serve as a means for social coordination and management of inter-organizational interactions and shared resources (Berardo & Scholz, 2010; Herranz, 2010; C. Jones, Hesterly, & Borgatti, 1997; Rhodes, 1996).
Similar views about networks are voiced by Kapucu (2009) and Provan et al. (2007) who consider networks as composed units of human and non-human entities that are tied to each other by a commonly desired and agreed-upon goal. They are systems of temporarily intertwined relationships around certain projects with informal links of communication (Mintzberg, 1996). While reasons behind the creation of networks abound, the main explanation for these types of inter-organizational arrangements, that are an alternative to, rather than a hybrid of, markets and hierarchies (Rhodes, 1996), is the fact that organizations are inter-dependent actors that extensively rely on the performance and outcomes of other actors within the same range and environment of service delivery (Bingham, O’Leary, & Carlson, 2008; Kapucu, 2006b; Klijn & Koppenjan, 2000; I. W. Lee, Feiock, & Lee, 2012).
Regardless of the type, networks are shaped by organizational, environmental, and contextual factors. It is important to note, however, that networks are dynamic structures that often change and evolve over time (Abbasi, & Kapucu, 2012; Gulati & Gargiulo, 1999). They are also inherently flexible and adjustable based on the gap between governance structure and contingencies (Emerson, Nabatchi, & Balogh, 2012; Provan & Kenis, 2008). In addition, networks might be created on the voluntary basis of actors coming together (informal and emergent) or being mandated (prescribed and formal; Bryson et al., 2006; Ibarra, 1992; Robins, Bates, & Pattison, 2011). The way networks are created and led, in turn, is a matter of network leadership, which is successful to the extent resource management, trust-building, stakeholder-oriented approaches, shared understanding, and the like are successful (Silvia, 2011; Silvia & McGuire, 2010).
Networks in Emergency Management
Networks are believed to be more effective than a market or hierarchy when the problem at hand requires an adaptive and flexible approach due to inconsistent information or uncertain conditions existing, or when the knowledge and resources needed to address the issue spans boundaries (Provan & Lemaire, 2012). Networks have become one of the most utilized tools in response to the complex events of disasters requiring the engagement of multiple stakeholders for effective results (Kapucu, 2006a; Kapucu, Garayev, & Wang, 2013; Moynihan, 2008; Waugh & Streib, 2006). Traditional methods of public management characterized by hierarchical structures, rigid boundaries, and red-tape have proven less effective when dealing with these extreme events (Bier, 2006; Kapucu, Arslan, & Demiroz, 2010; Kettl, 1997).
There is an historical explanation, however, behind the latest trends in the field. Those developments were mainly shaped by what Birkland (1997) calls “focusing events”—events of unusual size, visibility, time, and impact. Rubin (2007) claims that the trend toward collaborative approaches and practices was the result of federal-level changes in handling disasters throughout the past decades, abandoning the non-involvement policies and embracing coordinative practices. While the federal government created standards for future man-made and natural disasters, after decades-long reforms, not all practices at the local level of government were in line with the established national framework. Although regional and local diversity inevitably resulted in modified versions of the envisioned framework, two different approaches for tackling disasters have emerged at the local level: the ESF-based horizontal approach and the ICS-based vertical approach.
Horizontal Networks in Emergency Management
This structure is based on the relational and functional arrangement of available resources from all agencies responsible for managing a specific type of operation (Federal Emergency Management Agency [FEMA], 1997). Accordingly, all agencies responsible for transportation, law enforcement, or mass care, for example, are grouped under a specific type of ESF. Such an approach aimed to achieve more streamlined, efficient, and effective emergency operations characterized by an increased emphasis on relevant and connected operations between respective agencies as opposed to a mix of less relevant inter-organizational interactions.
The ESF-based framework was first introduced in the Federal Response Plan (FRP) in early 1990s when the federal government made several changes to create an improved coordinating mechanism for emergency management operations at the national level. The plan consisted of a basic plan outlining concepts, planning, response and recovery actions, and roles and responsibilities along with a list of ESFs and corresponding agencies (FEMA, 1997). With 12 ESFs, the FRP introduced a mechanism of coordination for 26 federal agencies and the American Red Cross.
Due to the FRP’s overemphasis on federal-level disaster response and less attention on state and local-level disaster management (Maniscalco & Christen, 2011), as well as a weak coordination and communication response to the September 11, 2001, terrorist attacks, the National Response Plan (NRP) was created. Accordingly, the National Response Framework (NRF) was about better coordination, communication, and collaboration during extreme events (Department of Homeland Security [DHS], 2004).
Different from the FRP, the NRP consisted of 15 ESFs and the most important aspect of change envisioned was the establishment of the National Incident Management System (NIMS) template which enabled all actors “to work together effectively and efficiently to prevent, prepare for, respond to, and recover from domestic incidents regardless of cause, size, or complexity” (DHS, 2004, p. 1).
Despite incorporating lessons learned from the failures following its creation, including Hurricane Katrina in 2005, the NRP was still considered ineffective to adequately address state and local issues (Maniscalco & Christen, 2011), which led to creation of the National Response Framework (NRF) in 2008. The framework is “built upon scalable, flexible, and adaptable coordinating structures to align key roles and responsibilities across the Nation, linking all levels of government, non-governmental organizations, and the private sector” (DHS, 2008b). Like the NRP, the NRF also consists of 15 ESFs and is a key component of the homeland security strategy, which envisions timely and effective response and recovery from incidents affecting the entire nation (Maniscalco & Christen, 2011).
Even with the national government’s attempts to create an ultimate framework for dealing with domestic incidents, some local emergency management agencies have chosen to develop their NIMS compliant coordination structure that relies on ESF-based inter-agency collaboration. The national approach has been replicated and adjusted to state and local contingencies, in which emergency response and recovery is coordinated by state or local emergency operation centers (EOCs).
Figure 1 shows an example of an ESF-based emergency management system with n ESFs. In practice, the number of ESFs in a state or local government emergency management system would vary based on local contingencies and needs. Most local CEMPs replicate the federal government’s approach and have a number of ESFs that vary between 15 and 20.
ESF-based emergency management system with n ESFs.
Vertical Networks in Emergency Management
While certain emergency management agencies have chosen to stick with the ESF-based approach, others have chosen to adapt to the NIMS requirements by focusing on ICS. While the former envisions creation of rules, norms, and principles that can be implemented across different incidents, the latter specifies a need for a standard operational structure for managing incidents (Lester & Krejci, 2007). Besides these key elements, NIMS delineates five major components it relies on, namely, preparedness (building operational capacity), communications and information management (inter-operable technologies and effective communication standards and mechanisms), resource management (standardized, flexible, and scalable resource allocation; deployment; and management), command and management (standardization of incident response structure and relationships), and ongoing management and maintenance (oversight of compliance with NIMS standards, procedures, and technical capacity; DHS, 2008a). When inter-jurisdictional and cross-sector collaboration is considered, NIMS offers a widely applicable ICS.
ICS is structured around five main functional areas, command (decision making and top management of the incident), operations (responsible for immediate reduction of threat, and response and recovery), planning (analysis, evaluation, and report of the incident situation to command), logistics (responsible for provision of necessary resources to incident scene personnel), and finance/administration (analysis of costs related to disasters, for example, recording personnel time and administering reimbursement and claims). When an incident occurs, a single incident commander (IC) is designated for the overall management and decision making of the incident. An alternative to having a single IC as incident manager is unified command (UC) which is more suitable for incidents requiring multi-jurisdictional and multi-agency response. Such an approach is characterized by joint decision making and implementation by representatives of several agencies considered as stakeholders. It is the size, scope, and severity of disasters as well as jurisdictions’ own discretion that would ultimately determine whether the management structure will be IC or UC (DHS, 2008a). Figure 2 shows the operational structure of ICS as specified in the NIMS document.
Incident command system.
It is important to note that while the main functional areas of the ICS structure are titled as sections, all other subordinate structural units are mainly determined by the scope, size and severity of disasters, the agencies involved, and the overall objectives of the incident efforts—all of which drive the resource management process. The main determinant of the sub-levels, thus, is span of control that, by standards, should be manageable in terms of resources. ICS, in this regard, offers flexibility in structural adjustments based on geographical and jurisdictional boundaries (DHS, 2008a).
Network Structure and Performance
Formal (Milward & Provan, 2006) as well as informal (Bardach, 2001; J. Lee & Kim, 2011) networks are goal-oriented inter-actor agreements structured by the individual organizations’ level of commitment with respective levels of formality of relationships, power distribution, inter-dependency, responsibility, and accountability (Kamensky et al., 2004; Kilduff & Tsai, 2003; Lazer, Mergel, Ziniel, Esterling, & Neblo, 2011). The different types of relationships, actor attributes, and positions within networks result in different types of structural arrangements as well as different levels of importance and centrality of actors (Ibarra, 1993). Those arrangements and structures are also affected by the institutional rules and environments (Meyer & Rowan, 1977), the level of coordinative understanding by the actors (Nowell, 2009), and the relationship between accountability and expectations (Romzek & Ingraham, 2000).
Structure is treated both as an outcome of functioning within a network and as a facilitator or hindrance of adequate network functioning (Ibarra, 1992; Robins et al., 2011). It is one of the key factors affecting network performance (C. Jones et al., 1997; McGuire & Silvia, 2010). However, measuring the effectiveness of networks remains a challenging task (Robins et al., 2011). Robins et al. (2011) suggest that structural properties such as “the presence of reciprocation in network exchanges, indicating relational embeddedness; and, the presence of triangulated exchanges, indicating structural embeddedness” (p. 1297) are important preconditions for effectively performing networks. Another related factor is the creation of a macro culture within the network, which helps reduce coordination costs for complex exchanges by developing a broad understanding of mutual and similar network goals and enhances the emergence of a network form of governance (C. Jones et al., 1997).
Moreover, both formal and informal aspects of networks are important to understand the true workings and functioning of networks (Robins et al., 2011). Ibarra (1992) studied network structure by examining the relationship between prescribed and emergent relationships amid network players and their patterns of interaction. Prescribed networks are formally set up relationships and partnerships that can take the form of task forces and teams while an emergent network reflects informal, voluntary interactions and relationships. In the context of emergency management, preparedness and response networks are formally structured around the ESF-based or ICS-based system. However, informal, friendship networks exist simultaneously within these formal response networks.
Ibarra (1992) discusses two archetypal forms of network governance which are hierarchical and integrative systems. The author argues that these two forms are important in understanding the different patterns of structural properties in both formal and informal or prescribed and emergent networks. In the context of emergency management, the ICS-based system is in line with the hierarchical model, while the ESF-system is in line with the integrative model of Ibarra’s classification.
The specifics of the ESF-based versus ICS-based emergency management network structures, in turn, are expected to impact the performance of the network. Therefore, in light of the ESF-based network and the ICS-based system, the main overarching assumption of this article is that centralization, whether at the network level or at the level of individual network centrality measures of the actors, would be lower in an ESF-based system when compared with that of the ICS-based system. This assumption is expected to be true across all types of networks in emergency management, whether friendship (who knows whom), advise (who collaborates with whom during preparedness), or response (who collaborates with whom during response), as well as across specific centrality measures.
Network governance research has utilized a number of methodological approaches varying from factor analysis and structural equation modeling to regression models and network analysis (Lynn, Heinrich, & Hill, 2000). The emergency management mechanisms of the two counties in this article present examples of network governance. The article utilizes the network approach for analysis and evaluation purposes. Network analysis has been suggested as a useful tool to examine the structural and relational aspects of networks (Kapucu, 2006a; J. Lee & Kim, 2011; Milward & Provan, 1998). This method is also useful in terms of increasing the visibility of hidden or informal inter-actor relationships in the network (Cross, Borgatti, & Parker, 2003). Moreover, when studying network governance and addressing collective problems, it is important to study both bilateral dyadic relationships and multi-lateral ties that reflect whole networks (Provan & Lemaire, 2012). Bilateral or micro-level analysis such as centrality measures identify egocentric qualities in a network and reflect how social interactions and relationships exist and flow in the whole network (Furst, Schuber, Rudoph, & Spieckermann, 2001; Provan et al., 2007). Structural properties of networks such as high density and centrality imply higher inter-dependency between network entities and their relationships. Other structural aspects such as hierarchy and centralization may reflect the disparity and stratification that exists within the network. These aspects can be applied to both prescribed and emergent networks (Ibarra, 1992).
Networks involved in managing emergencies are prescribed, formal and emergent, informal. Comparing the two systems, and their relative expected effectiveness, we can propose that centrality measures in the friendship networks for the ESF-based systems are higher when compared with the ICS-based system. This is assumed as the ESF-based system is more integrative and over time has allowed different functional agencies to develop informal, friendship ties with other agencies. Higher degree centrality values will depict more opportunities for interaction and more power within the network for players. Thus, the first hypothesis is as follows:
Along with degree centrality, closeness centrality is also believed to be higher in an ESF-based system which is a more horizontal and integrative system rather than a vertical or top-down approach. Closeness centrality reflects geodesic distances and paths between members of a network. Thus, higher closeness (or lower farness) centrality shows that actors within the network can reach other members at shorter paths and in a shorter distance (Hanneman & Riddle, 2005). Understanding the ESF-based system as a horizontal and team-based system, it encourages direct links, shorter paths of communication, and resource exchange between entities, and easy development of informal relationships (Kapucu & Garayev, 2012). Thus, the second hypothesis is as follows:
In line with the same argument, friendship networks for the ESF-based system will also encourage more brokerage opportunities and help to develop the capacity among players to connect members to each other and broker contacts effectively. Thus, the third hypothesis is as follows:
Finally, not only is it important to be close to all other nodes in the network, but also to be close to those with higher importance. Eigenvector centrality measures exactly this phenomenon of proximity to important and powerful nodes in the network. The assumption for friendship networks, in this regard, is that ESF-based networks allow for higher scores of eigenvector centrality. The related fourth hypothesis, thus, is as follows:
Disaster preparedness networks, on the other hand, are working networks formed when agencies are preparing to manage emergencies and disasters. They may also be described as advice networks. These networks are less prescribed and less formalized when compared with response networks; this is because ESF and ICS-based systems are essentially response systems. These preparedness networks, and the relationships formed during network existence, influence response networks directly, as agencies working with each other during the preparedness phase will develop close, embedded ties that will be utilized effectively in the response stage of managing a disaster. Similar to the previous assumptions, centrality measures in preparedness networks following the ESF-based approach will be higher when compared with the ICS-based system. The following hypotheses are related to the preparedness networks.
Last, response networks depict how the prescribed response framework and systems are implemented and come into play. These systems (both ESF- and ICS-based) act as guiding systems for appropriate response actions during disasters. It is interesting to note that the actual response networks are an amalgamation of both prescribed and emergent networks as complex situations lead to new constraints and demands that warrant flexibility and adaptability. However, routine tasks during response operations work best under the prescribed network, while tasks that require a novel, new approach work best under a flexible system. Thus, it is argued that the ESF-system, although prescribed, offers more integration and flexibility while the ICS-system offers less. Thus, the centrality measures are higher for ESF-based systems during the response stage when compared with the ICS-based system. The hypotheses explaining the response stage are as follows:
All in all, the ESF-based emergency management system measures are expected to provide a more favorable structure and actor positions in terms of power relations, information management, and resource allocation when the nature of the emergency management field and its requirements are considered. Access to information and resources or division of roles and responsibilities, in this regard, are some of the factors to be considered when the structural consequences of both systems are analyzed.
Context of the Study
The focus of this study is county-level local government, and two metropolitan counties in the State of Florida were chosen for analyzing the two approaches of managing disasters. Orange County is one of seven counties comprising Central Florida and serves a population of around 1 million. The mayor, under Section 252.38 of the Florida Statutes, directs county governments to establish an emergency management agency and has delegated his authority to manage emergencies and disasters to the Director of Emergency Management, who administers the Orange County Office of Emergency Management (OCOEM) and operates the EOC in times of disasters (OCOEM, 2011).When a local emergency occurs, the EOC is activated, depending on the level of threat, following the guidelines specified in the CEMP of the county, which is also mandated by state statutes.
Despite Orange County’s compliance with the NIMS/ICS structure, the main system of coordination, as described in the county CEMP, is based on horizontally arranged ESFs. Accordingly, it is primarily the ESFs that constitute the main terminology, jargon, and approach when dealing with emergency-related operations. The relationships and ties between agencies represented at the EOC in times of emergencies are structured around 20 ESFs and their respective primary and support agencies. The NIMS-based ICS structure, however, remains the main scheme for classification of the ESFs, as well as other parts of the Emergency Response Team (ERT), for guidance and efficiency purposes. The county mostly faces meteorological disasters like storms, tornadoes, and hurricanes, which are comparatively slowly developing in nature when compared with sudden natural disasters like earthquakes.
Duval County, on the other hand, is located at the eastern side of Florida, and serves a population of around 850,000. Emergency and disaster management in Duval County is the responsibility of the Emergency Preparedness Division (EPD) of the Fire and Rescue Department of the Consolidated City of Jacksonville/Duval County (CCJDC), which is administered by the Emergency Preparedness Division Chief. When a local emergency occurs, the EOC/Area Command is activated through the EPD to respond to the threat, pursuant to Chapter 674.203 of the CCJDC Ordinance. EOC operations would activate the Emergency Preparedness Organization (EPO) which is structured in line with NIMS (CCJDC, 2010).
As a widely practiced standard, the head of the EPO is the city mayor, who is assisted by the Jacksonville Security Coordinator, an Executive Group, and an Operations Group, also known as the EOC Management Team. The Executive Group comprises the EPD Chief and department/agency directors for advisory purposes, while the EOC Management Team consists of sections, branches, groups, and units structured in line with ICS principles and standards. Like Orange County, Duval county also mostly faces meteorological disasters, which makes the two counties comparable in terms of analyses and conclusions to be derived.
Method
This article examines the CEMPs of Orange and Duval counties in the State of Florida for organizational-level (network) analysis. CEMPs are documents created by respective county emergency management agencies to provide clear and elaborate identification of roles and responsibilities of the organizations responsible for emergency preparedness, response, and recovery in specific jurisdictions. They are created and maintained with the purpose of a unified and streamlined set of actions and operations in times of disasters. For the purposes of this study, the CEMPs of the counties were analyzed to identify emergency management actors for survey research. A roster of organizations and individual representatives was provided to the survey respondents.
Actors specified in the CEMPs were distributed a survey asking several network questions. More specifically, the actors were asked to identify among actors those (1) whom they know (friendship network), (2) with whom they collaborate during emergency preparedness (advice network), and (3) with whom they collaborate during disasters (response network). Based on the CEMP prescription, a total of 93 actors were identified in Orange County’s emergency management network, out of which 81 were accessible, resulting in a total of 56 responses, with 40 of them being valid for analysis. In the case of Duval County, a total of 66 agencies were identified, out of which 60 were accessible, resulting in a total of 30 responses, with 22 of them being complete for analysis.
This study compares two different emergency management networks with non-paired actors. Previous research has focused on several analytical techniques to compare non-equivalent networks. While some utilized the triad census method (Faust, 2006; Holland & Leinhardt, 1979), others have focused on less algebraic methods by analyzing centrality measures (Costenbader & Valente, 2003; E. C. Jones, 2003; Valente, Coronges, Lakon, & Costenbader, 2008; Watts & Strogatz, 1998). This study follows a modest approach and focuses on centrality measures developed by UCINET social network analysis software. UCINET provides statistics at three levels of analysis—micro, meso, and macro (Hanneman & Riddle, 2005). The micro level is related to specific actor positions and relationships with other network members, whereas meso-level statistics provide insight about the average and aggregate values of distribution of centrality scores (mean, variance, standard deviation, etc.). Macro-level analysis, on the other hand, focuses on the overall structural characteristics of whole network relationships (density, centralization, etc.). The statistical outputs analyzed in this study are of the micro and macro nature as centrality and centralization measures were utilized and analyzed. While centrality measures focus on micro-level analysis, centralization measures focus on the macro-level analysis (Hanneman & Riddle, 2005; Prell, 2012).
The measures chosen for micro-level analysis were degree, closeness, betweenness, and eigenvector. In the case of degree and closeness, only indegree and incloseness values were considered so that non-respondents could be included in the overall analysis. The four centrality measures were initially analyzed via a regression test combining the values of both networks in one data set and coding network type (ESF-based vs. ICS-based) as a dummy variable, ESF-based coded as 1, and ICS-based coded as 0. Regression analysis was run to see whether being in an ESF-based system versus ICS-based system impacts actor centrality measures in positive or negative way. Based on general assumptions in the literature, the 12 hypotheses identified earlier were tested through regression analysis in line with three types of networks (friendship, advice, and response) and four centrality measures, namely degree centrality (number of direct ties an actor has), closeness centrality (the extent to which an actor is close to every other actor in the network), betweenness centrality (the extent to which an actor has brokerage role between others in the network), and eigenvector centrality (the extent to which an actor has ties with most important actors of the network). Therefore, if statistically significant models had a positive sign in front of the dummy variable that would mean that being ESF-based impacts centrality measures in a positive way, while a negative sign would mean that it is ICS-based structure that affects centrality measures in a positive way. The main overall assumption is that structure matters (McGuire & Silvia, 2010).
In addition to testing hypotheses through the regression analysis, a macro-level analysis in terms of centralization measures was conducted. The measures chosen for this analysis were degree centralization (the extent to which actors’ degree centralities are distributed equally across a network), betweenness centralization (the extent to which reliance on other actors is distributed equally across a network), eigenvector centralization (the extent to which actors’ relative importance is distributed equally across a network), density (the extent to which related actors end up in dense connectivity groups across a network), and cliques (the extent to which network actors form groups for specific purposes). The values were analyzed across the three network types mentioned above, in Orange and Duval counties. The differences and similarities, as well as the related consequences are discussed in the following section (Carpenter, Li, & Jiang, 2012).
Results and Analyses
Table 1 provides a summary of the regression analyses results as well as information about the micro-level indegree, incloseness, betweenness, and eigenvector values (treated as the dependent variable) that were tested with network type (ESF-based vs. ICS-based) as an independent dummy variable. In addition, eight control variables were added into the model, all of which are related to the respondent emergency manager or to the organization he or she represents: years responder has worked in organization, years responder has worked in current jurisdiction, years responder has worked in current position, number of full-time employees in the organization, yearly budget of the organization, gender of the respondent, age of the respondent, and the highest education degree attained by the responder.
Regression Model Summaries for Three Networks and Four Network Centrality Measures.
Note. ESF = emergency support functions; ICS = incident command system.
Statistically significant at .05 level. **Statistically significant at .01 level .
As four centrality measures (degree, closeness, betweenness, and eigenvector) were tested for three different network types (friendship, advice, and response), there were a total of 12 regression models to be tested. Based on the findings, out of 12 models, 7 models had statistically significant F values at .01 level: degree (R2 = .681), closeness (R2 = .993), and betweenness (R2 = .229) measures for the friendship network; degree (R2 = .264) and closeness (R2 = .997) measures for the preparedness network; and degree (R2 = .305) and closeness (R2 = .865) measures for the response network. Therefore, only these 7 regression models were valid for interpretation.
In light of the interpretation of the regression model coefficients in the above-mentioned seven models, the type of the organization, namely whether organization is an ESF-based or ICS-based, is a statistically significant contributor in five models, namely degree (at .01 level) and closeness (at .01 level) models for the friendship network, degree (at .01 level) and closeness (at .01 level) models for the preparedness network, and only degree (at .05 level) model for the response network. Overall, in statistically significant models, having the ESF-based system leads to higher indegree values (positive relationship) but lower incloseness values (negative relationship) when compared with the ICS-based system.
In terms of control variables in the seven significant models, on the other hand, age is statistically significant (at .05 level) for the incloseness model (negative relationship), though the impact is comparatively low. In essence, it means that older people have shorter paths to other agencies when required. In addition, gender plays a role (at .01 level) in the betweenness model (negative relationship), with males contributing less to the betweenness centrality of the actors. This means that females have a higher brokerage ability at least in friendship networks. Last, the number of full-time employees in agencies responsible for emergency management in the counties positively contributes to degree models both in preparedness (at .05 level) and response (at .05 level) networks. This means that agencies with more full-time employees would have more direct ties with other agencies in the context of preparedness and response networks, which itself may be a contributor to overall performance.
Based on the regression analyses, it is clearly seen that in all three network types, the degree and closeness centralities of the network members are affected by the way the network is structured (only the closeness model in the response network is an exception). In other words, whether the network is structured more horizontally under the ESF-based system or more vertically under the ICS-based system impacts individual organizations’ network positions and the number of ties they have with others (degree centrality) as well as how close they are to other members (closeness centrality). It also appears that the R2 value for closeness centrality is substantially higher than the R2 value for degree centrality across all three networks. As both of these centralities represent a degree of reachability and access, we can say that the network structure matters in this regard. However, one should note that the higher R2 values for closeness centrality indicates that the structure and the chosen control variables account for greater part in closeness models when compared with degree models.
When regression results are analyzed in terms of the direction of the impact, it appears that the degree centrality measures of individual organizations are higher under the ESF-based system rather than under the ICS-based system (reflected through the positive coefficient in the regression model in Table 1), thus, supporting the first, fifth, and ninth hypotheses. This means that the ICS-based system is structured to limit the scope of relationships an organization might have, which corresponds to the ICS-based systems’ idea of a narrower span of control, less horizontal and more vertical control. On the contrary, the ESF-based system presents more opportunity for inter-organizational relationships as reflected by higher values of incoming ties to network actors.
In addition, the results show that the closeness centrality measures for organizations under the ESF-based system are lower when compared with those under the ICS-based system (reflected through the negative coefficient in the regression model in Table 1), thus, rejecting the second and sixth hypotheses (network structure in the 10th model is not statistically significant). The negative relationship shows that organizations in ESF-based networks are positioned further from each other when compared with the ICS-based system.
On the other hand, the betweenness and eigenvector centrality measures for individual organizations seem not to be affected by the way the system is structured. In other words, individual organizations’ brokerage power is distributed similarly in both the ESF-based and ICS-based networks. Likewise, the eigenvector centrality regression results show that no matter what the network is, an organization’s relationship with the most important and influential actors in the network is not affected. As both of these centralities represent power, we can argue that the distribution of power across network members is simply not dependent on the network structure. Accordingly, the 3rd, 4th, 7th, 8th, 11th, and 12th hypotheses are not supported.
The overall conclusion that can be drawn from the regression analyses, thus, is that network structure matters in terms of reachability or access to others, while more power and prominence-oriented centrality measures like betweenness and eigenvector are not affected by overall structure.
In addition to regression analysis, network analyses were also performed to see the overall structural characteristics of the networks. Outputs provided by UCINET software present an opportunity to analyze the emergency management network structures of the two counties, focusing on macro-level relationships. As stated earlier, the main assumption of this article is that network structure matters. It affects, shapes, and/or determines the types, nature, intensity, strength, and direction of the relationships among network actors. In terms of centralization measures, though, the main overall assumption is that the ESF-based system would produce lower centralization values when compared with that of the ICS-based system. It is important to note, however, that emergency management networks are mostly prescribed rather than naturally emergent as other networks usually observed in social contexts. CEMPs are the guiding documents for how network relationships are structured, regardless of previous or future formal/informal relationships among actors.
Degree Centralization
Based on the UCINET outputs in Table 2, it appears that overall degree centralization across the three networks is fairly high in both counties. Duval County’s centralization measure is higher in the friendship network, but lower in preparedness and response networks. This means that theoretically (based on the friendship network) actors within Duval County’s network are more centralized, whereas in practice (based on preparedness and response networks) they are relatively less centralized when compared with Orange County’s values. Thus, the practical side is essentially contradictory to the broad expectations regarding the ICS-based system being a comparatively more hierarchical and non-flexible model. The differences though, are not substantial, one should note, when reaching general conclusions.
Analytical Measures of Comparison for Orange and Duval Counties.
Nevertheless, despite small centralization value differences between the counties, it appears that, overall, both counties are characterized by a high level of relationships with only a few exceptionally important actors. Those actors are comparatively much more important in the preparedness and response operations of Orange County’s network. Such a situation might be specifically problematic because of the nature of emergency management, which favors less rigid structures and relationships, and more flexible operations on scene. The possible explanation for low levels of degree centralization in Duval County’s preparedness and response networks is the system’s reliance on the management and leadership of subordinate levels, despite the fact that differences are comparatively small, thus, being driven by internal hierarchical expectations focusing around efficiency (Romzek & Ingraham, 2000). Therefore, while Orange County’s overall degree centralization values are a surprise and not a desired situation, Duval County’s values are within the range of expectations. The essential question, then, is whether an ESF-based emergency management network is really a decentralized model as it has been argued and advocated to be?
Betweenness Centralization
Betweenness centralization, in turn, appears to have the lowest measures when compared with degree and eigenvector centralization measures. Regardless of the emergency management structure, both systems appear to result in little inequality in terms of a certain actors’ brokerage role. This means that there are many alternatives for actors in each emergency management system to reach others, without being dependent on single actors to do so. In addition, the lowest values were found in friendship networks, followed by preparedness and response networks. Concluding that, in practice, dependence on brokerage is higher when compared with a more theoretical model of the friendship network.
The difference between the two systems, though, is in favor of Duval County, which has lower values for friendship and preparedness, and a higher value for the response network. Thus, brokerage plays a more important role for Duval County when the response network is considered. The difference in values, in terms of collaboration during response networks, is almost twofold, which means that ICS-based networks have fewer boundary spanners to connect certain cliques of operations. Therefore, Duval County emergency management network actors need to invest in cross-organizational or cross-functional collaboration to reach a lower level of dependence on single specific actors in such a time-sensitive field as emergency management.
Eigenvector Centralization
On the other hand, eigenvector centralization is higher in Duval County across the three networks. Despite relatively lower measures for each county, and a small difference between the two counties, Duval County’s ICS-based structure seems to produce inequitable power relationships, giving certain actors more important roles and positions. In other words, Orange County’s network has relatively more actors that have important or central “neighbors” when compared with Duval County, thus, producing more opportunities for collaboration.
In addition, the eigenvector centralization measure is essentially about reachability to other actors (as in closeness centrality), in terms of which Orange County offers a more advantageous network model. In such a volatile and time-sensitive field as emergency management, greater reliance on others might hinder decision-making capability and timeliness of operations. Therefore, the ESF-based model with lower eigenvector centralization offers more egalitarian relationships and a more equal power distribution framework. One should note that a low eigenvector centralization value does not mean there is a lack of important and central actors within node proximity; it only means a relatively more equal distribution among actors, hence, less reliance on other groups or cliques.
Density
Despite the fact that differences appear to be small, density appears to be in favor of Orange County’s system in all three networks. Accordingly, the friendship, preparedness and response networks of Orange County produce denser relationships, and, thus, a more connected network. This situation is especially preferable when multiple sources are important for effective and quick results. In times of emergencies when source and service duplication/redundancy is extremely important as an alternative to possible disruptions, a network of denser relationships with multiple support sources becomes a panacea.
Therefore, Orange County’s ESF-based system stands as a stronger system because of the multiplexity of relationships. Multiplexity of relationships guarantees more sustained relationships, more alternative sources of resources and assets, more egalitarian relationships, less hierarchical ties, and more flexibility in terms of possible relationships. Multiplexity is essentially the strength of the relationships, and relationships are strong if there are multiple ways in which actors are connected to each other (Scott, 2013). Therefore, the ESF-based system is a stronger system with alternative ways for collaboration in times of disasters.
Cliques
Last, the cliques statistics presents an opportunity to evaluate respective networks’ role distribution. Specifically, actors that come together in the form of cliques for specific functional purposes create focus groups that concentrate on the related operation. In all three networks, Orange County’s measures are substantially higher than those of Duval County. While network size does play a crucial role in determining the ultimate number of cliques, it is also the way roles and responsibilities are distributed or shared across specific networks that impact the number of cliques. Accordingly, Orange County is characterized by, and relies on, multiple relationships shaped by multiple goals, whereas Duval County’s actors appear to have fewer and more concentrated goals. This might be an advantage in terms of clarity of roles and responsibilities, while a disadvantage in terms of diversity of services and sources when an emergency strikes. Inherently, thus, the ESF-based system allows for and fosters cross-functional relationships, whereas the ICS-based system advocates for intra-functional relationships, hence offering a focused approach. It needs to be mentioned for methodological concerns that the differences between networks may be a function of incomplete network responses (57% of the actors in Orange County network and 66% of the actors in Duval County network did not respond to the survey). Missing data in the county networks create challenges in interpreting socio-centric differences between the two networks (Costenbader & Valente, 2003). However, it should also be noted that the respondents in both networks comprise a group of primary agencies (as opposed to support agencies as classified in county CEMPs). As primary agencies are those responsible for most of the job during emergencies, their responses entails highly significant amount of information about overall network relationships. What is more, such centrality measures as indegree centrality, for example, are considered to be robust even under conditions of imperfect data (Borgatti, Carley, & Krackhardt, 2006; Costenbader & Valente, 2003).
Discussion
Without doubt, the statistics presented above might be interpreted and utilized in multiple ways. However, when structural network relationships are considered, it is mainly positions, not the actors’ attributes, that matter. When the two systems are compared, at least in light of the statistics, the ICS-based system’s reliance on multiple relationships characterized by higher density, higher number of sources, and a more egalitarian approach, does not produce a shock effect. Despite exceptions in the literature advocating the ICS-based system as non-hierarchical (Moynihan, 2009), several scholars consider it merely a command-and-control system characterized by non-flexible vertical relationships leading to dysfunctionality (Comfort, 2007; Drabek & McEntire, 2003; Kapucu, 2009, 2012). The problem is especially evident in a disaster situation when decision making at higher levels is dependent on the lower levels. The flow of information and decision making, thus, might be negatively impacted by systems favoring strict command-and-control rules, which offer little room for improvisation and alternative sources of support.
Accordingly, the ESF-based system characterized by more alternatives and relatively more flexible relationships might be a preferable option especially because of the nature of emergency management. This suggestion in no way undermines the importance of ICS-based systems, which are known for streamlined and clear roles and responsibilities as well as case-based emergency management adjusting for contingencies. However, since emergency management is more about timely decision making rather than implementation of those decisions, the ESF-based approach stands as a more viable system in the face of ever-changing and worsening disasters requiring fragility, volatility, and flexibility of densely tied network actors.
In addition, as more variation in terms of network relationships means more centralization (de Nooy, Mrvar, & Batagelj, 2011), it is essential to have a more egalitarian and dense network that would minimize the power of specific actors. Emergency management is essentially not a field to exert power, but to distribute it across all possible actors within a network. Such an approach would enhance collaboration and empower even peripheral actors whose contribution is consequential in critical times. Therefore, the ESF-based system, characterized by more desired numbers especially in the response network, seems to fit better into the emergency management setting and for the ultimate goal of the field.
The friendship network is always relatively higher in desired values when compared with the practical stages of preparedness and response, which is quite natural and within the range of expectations. The preparedness and response networks might be affected by real-world hindrances ranging from lack of resources to communication problems. Organizational politics and will are also among the factors that may negatively impact network relationships, thus, resulting in lower numbers across preparedness and response networks as opposed to the theoretical friendship network.
One should note that in all three networks, regardless of the counties’ system, degree centralization is very high, which is not desirable in the context of emergency management. Disproportionate degree centralities of actors in an emergency management network may be a hindrance to improvisation and frugal decision making, especially in systems requiring continuous approval of higher levels of decision-making authorities. In addition, centralization in this sense also means reliance on relatively fewer actors in terms of resources. Resource dependency is a threat in time-sensitive emergency situations and actors that have few or no relationships are doomed to be dysfunctional. What is more, the actor that is highly central may also become dysfunctional in times of an overwhelming situation and limited capacity. The discussion is essentially similar to Provan and Kenis’ (2008) discussion on network structures, in which the authors talk about three types of network governance varying in their degree of centralization. Accordingly, the authors state that participant-governed networks are more decentralized and egalitarian, whereas lead-organization and network administration organization models are relatively centralized and hierarchical.
This discussion, therefore, leads us to focus on two specific terms—coordination versus collaboration. It is ultimately the mode of partnership that will shape the emergency management network at the local level. More specifically, counties that focus more on coordination should and would prefer the ICS-based system, whereas those favoring collaboration would practice the ESF-based system. While this argument and evaluation might be overarching, it would be helpful for those trying to choose between the two systems. Without doubt and inevitably, this decision would be also a function of the network size. Thus, small and mid-size counties might be willing to prefer the collaborative approach, hence the ESF-based system, whereas big counties might opt-out for the coordination-oriented ICS-based system.
Conclusion
This study focused on the comparison of two widely utilized emergency management systems in the United States. The ESF-based system today is considered an alternative to command-and-control mechanisms following the NIMS/ICS structure. Proponents of the former suggest the ESF-based system’s superiority in terms of flexibility and lessened hierarchy, whereas proponents of the ICS-based system emphasize standardized operational procedures and clarity of roles and responsibilities. This study compared both systems as exemplified by two counties in Florida. The network analysis presents the ESF-based emergency management system as relatively more robust, especially in terms of power relationships and actor positioning. More specifically, in terms of practice-oriented preparedness and response stages, the ESF-based system appears to have less centralization and denser relationships, which is essential in the time-sensitive and resource-dependent field of emergency management. Overall, thus, counties that are relatively small and vulnerable in terms of resources might be in a better position to choose the ESF-based system for collaboration purposes, whereas counties requiring coordination would benefit from the ICS-based system.
Even though the study uses two counties in Florida, the two systems are widely used at local levels and can benefit other local jurisdictions in the United States and abroad. The main limitation of this study is the comparison of two non-paired networks. In addition, the response rate of network actors is also a factor that might have affected the ultimate centrality and centralization numbers presented in this study. It is our strong belief that higher response rates would present a more complete picture and analysis. Moreover, evaluation of the network’s structural attributes beyond those presented within the context of this article would be likewise helpful for additional insight at the macro level. Meso-level analyses might also shed light in terms of network centralization and actor centralities. Finally, this study should be replicated to include other, especially smaller size, counties that are following ESF- and ICS-based models to make sure the conclusions are generalizable.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research is funded by National Science Foundation (Award No.: 0943208) Title: VOSS: Creating Functionally Collaborative Infrastructure in Virtual Organizations. PI Dr. Naim Kapucu.