Mechanisms of Control in Emergent Interorganizational Networks
Marcum, Christopher Steven, Bevc, Christine A., Butts, Carter T., Policy Studies Journal
A central phenomenon of human organization is the emergence of control relations, that is, relationships in which one actor transfers his or her decision-making capacity to another (Coleman, 1990; Weber, 1958). Indeed, it has been argued that the existence of such relations is one of the defining characteristics of organizations per se (see, e.g., Perrow, 1970; Porter, Lawler, & Hackman, 1975). If "organization" is (in the words of Galbraith, 1977) "that 'something' which distinguishes any collection of 50 individuals in Kennedy International Airport from the 50 individuals comprising a football team in the National Football League," one of its core elements is clearly the willingness of organizational members to systematically cede control to other members of the group. Coincident with control relations between individual actors, control of one organization over another involves the controller organization to, in Kirsch's (1996) words, "regulate or adjust behavior" of the controllee organization.
While the emergence of institutionalized control relations (or authority relations) has been a target of intensive study at least since Spencer (1874), much remains to be learned regarding the emergence of ad hoc control over short timescales. This is particularly true at the interorganizational level, where formal authority relationships frequently overwhelm or obscure other factors (Agranoff & McGuire, 2001; Weber, 1958). Observations of interorganizational control relations during periods of relative stability are thus effective at revealing the structure and long-term evolution of formal authority but are of limited use in illuminating how control relations arise where the influence of such authority is attenuated (or altogether absent). Since the latter condition must precede the former, understanding the structural conditions in which ad hoc control relations arise can potentially shed light on the circumstances in which formal authority structures initially develop. More pragmatically, ad hoc control also acts as a "fallback" mechanism for interorganizational collaboration within settings wherein formal authority relations are ineffective or disrupted (Roethlisberger & Dixon, 1939). Moreover, ad hoc control facilitates task performance and role fulfillment when formal authority is ineffective. Thus, following Agranoff and McGuire's (2001) question of the role power plays in interorganizational networks, understanding the emergence of ad hoc control relations is of both practical and theoretical importance for the sociology of organizations.
Given the prevalence of formal authority structures in routine settings, the study of ad hoc control is more easily conducted in nonroutine settings, such as disasters. Sociologists have long observed that disasters disrupt existing social structures and practices, thereby providing researchers with an opportunity to examine mechanisms that are difficult to observe in everyday settings (Drabek, 1986; Quarantelli, 1987). As events disrupt daily routines, they also trigger the emergence of social structures than deviate from routine and/or planned patterns of interaction (Drabek & McEntire, 2002; Dynes, 1970; Quarantelli, 1996). Examination of the organizational response to disasters thus offers a glimpse into the processes by which routine social structure is reorganized and/or restored. In particular, a common effect of disasters is to render temporarily ineffective the conventional lines of authority among organizations; this is particularly true in the immediate postimpact period, during which organizations and their members must rapidly mobilize and respond within an uncertain (and often infrastructure-degraded) environment. During this period, individual and organizational task performance requires the flexible combination of planning and improvization, including the creation of novel mechanisms of coordination and control (Comfort, 2007; Dynes, 1994; Neal & Phillips, 1995, Waugh, 1993). By studying ad hoc control relations during the immediate postimpact period, we thus gain the opportunity to test competing theories regarding the processes by which such structures develop and operate.
Employing the emergent multiorganizational network (EMON) as a research site has a rich history in organizational sociology (Drabek & McEntire, 2002; Majchrzak, Jarvenpaa, & Hollingshead, 2007; Stallings & Quarantelli, 1985; Trainor, 2004, Waugh, 1993, Weick & Roberts, 1996). EMON has long been characterized as a response to the breakdown in authority and communication lines entailed in disasters (Drabek & McEntire, 2002; Tierney, Lindell, & Perry, 2001; Trainor, 2004). However, little is known about the structural mechanisms that facilitate command relations in EMONs. Indeed, the ways in which organizations delegate decisionmaking capacity and issue commands in the context of an emergent network represent a remarkable occlusion in the literature, given the emphasis on communication and coordination. We shed light on that question in this article.
This article follows the strategy outlined earlier, examining emergent control in disaster response networks to shed light on the development of ad hoc control relations. Our data are composed of a collection of networks from seven remote-area search and rescue operations from late twentieth century disasters in the United States, collected by Drabek, Tamminga, Kilijanek, and Adams (1981). To borrow a term from Provan, Fish, and Sydow (2007), we take a "whole network" approach by utilizing the Drabek studies at the network level of analysis. By doing so, we emphasize the contributions of each individual organization only as they relate to the overall process by which control mechanisms arise (Provan et al., 2007; Provan & Milward, 1991). Employing recently developed methods for analysis of error-prone network data within a Bayesian framework, we evaluate the relationship between the structure of interorganizational communication and organizations' prominence in the control structure (as evaluated by organizational informants). This relationship, in turn, is used to assess several theories regarding the process by which ad hoc control operates. Thus, we also make a methodological contribution by taking an analytic approach that bridges what Zaheer and Soda (2009) call "the structure of outcomes" with the "outcomes of structure." We used a novel model to infer the structure of communication in EMONs. We then used those networks to model how authority relations arise within the structure. This approach is general enough for many applications where ties are drawn based on informant accounts and there is uncertainty about the underlying network. Our analyses suggest that control in these networks involved both direct interaction and use of indirect contacts. As we argue later, this finding is of both theoretical and practical import.
Command, Control, and Authority in Disaster Response
Organizational practices employed in response to disasters and other extreme events fall broadly under the rubric of emergency management (Auf der Heide, 1989). The normative structure of emergency management in the United States (and the developed nations more generally) is based on a bureaucratic model, whereby response activities among a relatively decentralized collection of agents are coordinated through a centralized decision-making apparatus (Schneider, 1992; Takeda & Helms, 2006). The exercise of authority occurs "downwards" through the structure, with each agent answerable in principle to a single superordinate agent (a principle known as unity of command). As with most other modern organizations, authority relations within response organizations allocate power on a positional rather than a personal basis (Uhr & Fredholm, 2006; Weber, 1958). The design of these relations--and their effective use to coordinate activities during the response process (Comfort & Kapucu, 2006)--constitutes a family of tasks known collectively as command and control problems. Given that disasters typically involve rapid deployment of multiple units from disparate locations in a turbulent environment, it is perhaps unsurprising that effective command and control has been argued to be a core challenge for emergency management policy (Comfort, 2007; Iannella & Henricksen, 2007).
In principle, command and control problems may be solved by a combination of a priori policies and standard operating procedures. Unfortunately, however, the heterogeneous and often unpredictable environments encountered during disasters can dramatically limit the effectiveness of such systems, as Weick (1993) observed during the Mann Gulch fire. Where conventional structures and routines no longer suffice, novel (or, as we have used the term earlier, ad hoc) arrangements must be developed in response to changing circumstances (Mendonca, Beroggi, & Wallace, 2001; Webb, 2004). Failure to establish effective ad hoc control structures under such conditions can lead to conflict between organizations (e.g., due to task interference), failure to complete critical objectives (e.g., due to vital tasks being overlooked or unassigned), inefficiency (e.g., due to repeated performance of the same tasks by multiple actors), or other problems (e.g., underutilization of available personnel). Unfortunately, this process is made difficult by the limitations of interorganizational communication during disasters, differences in mission and procedures across organizations, and the short timescales within which solutions must be implemented. The development of ad hoc control structures is thus a form of "collective improvization" in which simultaneous adjustments by multiple actors within a heterogeneous environment converge to a solution that is beyond the capacity of any one actor to completely anticipate or determine.
The presence of explicitly articulated command and control concepts in disaster response within the United States may be tied back to the origins of emergency management as a distinct field of organizational practice. Prior to the creation of the Federal Emergency Management Agency in 1979 under Executive Order 12127, U.S. civil disaster response was situated largely within the Department of Defense, with additional support distributed across the Department of Commerce and the Department of Housing and Urban Development. Much of the emergency preparedness capability in the United States prior to this event stemmed from post-World War II civil defense programs. Following several devastating disasters in the mid-twentieth century (Hurricanes Carla, Betsy, and Camille; the Alaska earthquake in 1964; and the San Fernando earthquake in 1971), the U.S. government identified the need for a new independent civil agency to prepare and respond to natural hazards (Nicholson, 2003).
In contemporary emergency management, the influence of civil defense and military rigor are still apparent as former military personnel find civilian positions in paramilitary organizations, such as fire and police, and emergency management (Brooks, 2005; Nicholson, 2003). As neo-institutionalist theories would predict (DiMaggio & Powell, 1983), this exchange has promoted a certain degree of isomorphism between organizations in the military and emergency management fields. Among the practices to have diffused from the former to the latter field is a tendency for emergency response plans to be founded on a core model of hierarchical communication and coordination structures, regulated by a central authority. This hierarchical model is intended to direct actions to lower level actors based on decisions made by high-level actors, with the latter basing their decisions in part on information passed "up the chain" from units in the field. Such systems can have numerous benefits in terms of efficiency (Bavelas, 1950) and robustness to loss of noncentral units (Carley, 1992), but these benefits come at the cost of increased dependence on core units, and a rise in delays and errors stemming from the need to transmit critical information through multiple intervening parties in order to arrive at a decision (Drabek, 1985). The limitations of centralized solutions to command and control problems during disaster response have led some researchers (e.g., Dynes, 2003; Neal & Phillips, 1995; Wenger, Quarantelli, & Dynes, 1990) to call for increased use of decentralized decision making structures; the tension between the flexibility and low latency of local autonomy versus the efficiency and global robustness of centralized control is a core issue in the debate over how response operations should be structured.
Communication Network Structure and Control
Explicitly structural concepts such as kinship ties, resource flows (Phillips, Garza, & Neal, 1994), roles (Kreps, 1987, 1989; Kreps & Bosworth, 1993, 1994, 1997), and emergent behavior have been utilized by disaster researchers for several decades (Quarantelli, 1984, 1996; Stallings & Quarantelli, 1985); however, relatively few quantitative studies of communication or control structures appear in the early literature of the field. Following the 1976 Big Thompson River flood, Thomas Drabek began to study the relationships of interagency coordination and relationships (Drabek, 1985, 2002). In the disasters that followed (the Wichita Falls tornado in 1979, Hurricane Frederic in 1979, and Mount St. Helens in 1980), Drabek (2002) asserted that the measures then used to assess complex social phenomena arising in the aftermath of these events were woefully inadequate. Subsequently, Drabek and his colleagues conducted several studies that aimed to map the social relations of the multiorganizational networks that emerged among disaster response organizations (Drabek, 1985, 2002; Drabek et al., 1981; Gillespie & Colignon, 1993; Gillespie, Sherraden, Streeter, & Zakour, 1986). Following this line of research, Drabek (1987) argued that disaster response activities at the interorganizational level can be characterized via EMONs--the structured patterns of relationships among organizations collaborat-ing during the response process.
In 1978, Drabek began collecting data on EMONs resulting from remote-area search and rescue (SAR) operations mounted in response to seven different natural and anthropogenic disasters. The intent was "to prepare a set of case studies in which the multiorganizational responses could be documented" (Drabek, 1983, p. 279). Several studies focused on specific events and provide preliminary analyses (Adams, Drabek, Kilijanek, & Tamminga, 1980; Kilijanek, Drabek, Tamminga, & Adams, 1979; Tamminga, Drabek, Kilijanek, & Adams, 1979). These provide additional insights into local events, such as the July 1979 tornado that struck Cheyenne, WY (Drabek, Tamminga, Kilijanek, & Adams, 1982). The accumulated results of these studies were published by Drabek et al. (1981) and constitute the largest single collection of response operations with systematically collected and directly comparable network data to date.
The diversity of the events covered by Drabek et al. (1981)--flash floods in Texas; tornadoes in Lake Pomona (KS), Wichita Falls (TX), and Cheyenne (WY); Hurricane Frederic in Jackson County (MS); the eruption of Mount St. Helens (WA); and the report of a lost hiker on Mount Si (WA)--enable the comparison of EMON structures across context. The researchers collected data on each event in a systematic and consistent manner. While the dyadic relational information reported by in this work is limited to interorganizational communication, the authors also provide information regarding the extent to which each organization was judged by other network members to play a central role in decision-making and control activities during the initial response (Drabek et al., 1981). This information potentially makes it possible to test a number of competing hypotheses regarding the mechanisms of control employed during the early phases of each event by relating each organization's observed role in the control structure to the communication, which would be needed to obtain that role given the assumed control mechanism. We follow that strategy here.
We draw from past research and social theory to provide insight on how the realized network structure of interorganizational communication might generate control relations. Control relations are related to positions of centrality in networks, which have a long history of study in organizational science and in sociology, more generally (Bonacich, 1987; Freeman, 1979; Ibarra, 1993). Prima facie, we would expect that organizations that communicate with a lot of other organizations to play central roles in delegating authority and exerting control over the whole network. Tsai (2002) found that organizations with more ties tended to share knowledge more frequently in his study of 24 business units in a large petrochemical firm; however, the more direct control exerted by the headquarters on its subunits, the less frequently the subunits shared knowledge. This suggests that direct control, insofar that it is related to the number of ties an organization has, may hinder the communication flow in the whole network, dampening the efficacy of delegating authority.