Centralization and decentralization for resilient infrastructure and complexity

The environments in which infrastructure need to function, adapt, and thrive appear to be faced with increasing complexity as a result of accelerating volatility, uncertainty, and system dynamics (Chester and Allenby 2018, Desha et al 2009, Sharma 2019, Steffen et al 2015). Cybertechnologies are being integrated into legacy infrastructure at rapid rates, creating remarkable new opportunities but also vulnerabilities (Chester and Allenby 2020, Rinaldi et al 2001). Climate change is creating deep uncertainty for weather extremes and is threatening to exceed design envelopes of critical systems (Ayyub 2018, Bondank et al 2018, Burillo et al 2017, Chester et al 2020b, Helmrich and Chester 2020, Nasr et al 2019, Underwood et al 2017). In the United States, ageing infrastructure are increasingly confronted with these new challenges and must adapt with scant resources to meet demand and deliver services. Infrastructure are caught between goals, administrative structures, and technologies rooted in the past, and a future defined by complexity, uncertainty, and accelerating change.

The domains of infrastructure—physical networks and governing institutions—are of major importance when examining the capacities of our critical systems to adapt and transform. Concepts of centralized, decentralized, and distributed systems are pervasive across academic literature and discourse. Centralization is often associated with networks distinguished by a small number of producers serving a large number of consumers (figure 1(A)) and a top-down governance model (Alanne and Saari 2006, Albalate et al 2012, Bardhan 2002, Chandler 1977, Chester et al 2020a; Derrible 2017, Makropoulos and Butler 2010, Pagani and Aiello 2011, Quezada et al 2016, Rodrigue 2020, Tomlinson et al 2015, Wilder and Romero Lankao 2006). Generally, centrality measures the importance of a single node in a larger network, where increased emphasis on a node produces vulnerability (by, e.g. consumers relying on the operation of a single producer for a service). A decentralized system (figure 1(B)), where the ratio of producers to consumers increases, places less reliance on a few nodes, thereby, decreasing vulnerability (Baran 1964, Freeman 1978). There does not appear to be a clear definition of distributed in the infrastructure literature, but at times the concept is used synonymously with decentralization (Ackermann et al 2001, Alanne and Saari 2006, Makropoulos and Butler 2010). Following framings in computer science, a distributed system is one where nodes in a network work towards a common goal (Ge et al 2017, Srinivasa and Muppalla 2015). A distributed network, considered here to be a subset of decentralization, is where consumers are provided with a larger variety of service options (producers) and are connected with other consumers to achieve an objective (figure 1(C)). For example, in the power sector, installed residential solar panels may produce excess energy, which can supply the grid or be stored in electric vehicles (or other batteries) and used or released back to the grid a later time (especially during emergencies). Under this configuration, the consumer becomes a coordinating node to help manage supply and demand across the system (Calma 2021, Ghasempour 2019).

Zoom In Zoom Out Reset image size

In governance literature, broadly, centralization (figure 2(A)) is isolation of power to select individuals while decentralization (figure 2(B)) is the dispersion of power to many individuals of an organization (Faguet 2014, Mintzberg 1979). Decentralization can be further classified as either vertical or horizontal. Vertical decentralization is the movement of power down a chain of command, commonly found in bureaucracies, whereas horizontal decentralization is the reallocation of power across divisions (Dubois and Fattore 2009, Mintzberg 1979). Degrees of vertical and horizontal decentralization can be found in most institutions. To achieve a truly decentralized system, power would be distributed vertically and horizontally so each individual holds equivalent power. The configuration of infrastructure networks and governance has taken on a new importance under the emerging field of resilience, a field which studies a system's ability to persist, adapt, and transform to disturbances within and beyond the design envelope. Yet, there does not appear to be a concerted effort to align how centralization, decentralization, and distributed framings are used, and what different configurations mean, when adapting and transforming systems to be able to respond to increasing complexity.

Zoom In Zoom Out Reset image size

Given a growing interest in configuring infrastructure to better respond to increasingly complex environments (Baldwin and Clark 2006, Chester and Allenby 2018, Gilrein et al 2019, Rinaldi et al 2001, Tomlinson et al 2015), a coherent framing of centralization, decentralization, and distributed (hereby referred to as de/centralization) is needed. A polysemic framing (i.e. carrying multiple meanings) is particularly problematic given the importance of de/centralization within network and governance domains, how systems are physically configured and how authority is structured towards adaptive capacities (Hines et al 2015, Mintzberg 1979). In this work, infrastructure is defined as human design spaces that produce built environment systems, including their technologies and institutions of governance. Power, water, and transportation systems are the primary infrastructure sectors of focus. The upcoming section starts by reviewing key de/centralization framings across infrastructure domains. Following, de/centralization is contextualized within emerging resilient infrastructure theory. Last, a multi-dimensional framing of de/centralization through a coupled network-governance perspective with information as a mediator is proposed, emphasizing the capabilities that infrastructure will need in the Anthropocene.

2.1. Physical infrastructure as networks

2.2. Governance as networks of power

The governance of infrastructure and its associated organizational and bureaucratic structures has significant impact on system ability to manage changing conditions (Chester et al 2021, Chester et al 2020a, Mintzberg 1979, Uhl-Bien and Arena, 2018). Decentralization of governance can be classified as either vertical through the dispersion of power through a formal chain of command (i.e. strategic leadership to middle managers to front line workers), or horizontal where decision-making power is dispersed across many employees, including those outside of the chain of command, such as analysts and operators (Mintzberg 1979). Centralization exists when the decision-making power rests with a select few, for example, strategic leadership (Mintzberg 1979). Governance—a process involving collective action for resource allocation and use across multiple civic and private actors (Kooiman 1993)—is the balance between the authority, responsibility, and power of management and individuals within an organization through rules, values, and norms (Chester et al 2020a, Dubois and Fattore 2009, Faguet 2014, Mintzberg 1979). While this definition emphasizes decentralization as a process, the term is also used to reference organizational structure (Abimbola et al 2019, Dubois and Fattore 2009, Siggelkow and Levinthal 2003). Infrastructure governance is critical for resilience because human and organizational power structures, if structured appropriately, create knowledge, allocate resources, and form bureaucracies capable of handling both stability and instability.

In the power, water, and transportation sectors, the characterization of governance de/centralization depends on the jurisdiction and scale of the organizations which manage services between the producer and consumer. Power networks (e.g. the grid), water (e.g. distribution pipelines, wastewater treatment), and transportation (e.g. roadways and railways) system governance, are often characterized as centralized governance models (Albalate et al 2012, Chandler 1977, Quezada et al 2016, Rodrigue 2020) where divisional bureaucracies emphasize a concentration of power at strategic leadership (Chester et al 2020a, Mintzberg 1980). In these centralized structures, the leadership team can assess where to allocate and reallocate resources across their jurisdiction to meet demand through authoritative power. Conversely, water privatization, power generation, and transportation services are characterized by numerous, independent decision makers, and therefore decentralized (i.e. multiple actors, for example airlines, set their own policy) (Bardhan 2002, Larsen et al 2015, Quezada et al 2016, Rodrigue 2020, Wilder and Romero Lankao 2006). In decentralized structures, the decision-makers will be more directly connected with their producer and consumer needs; however, their access to resources may vary, which could elevate inequality of supply across marginalized jurisdictions with less power. Each governance structure allows infrastructure managers to address supply and demand within their jurisdiction.

De/centralization of governance within infrastructure sectors is characterized by the number of actors who hold decision-making power. Centralization is often framed as dominant in infrastructure systems, a remnant of system design emphasizing efficiency for times of stability, but infrastructure systems may take on various network-governance configurations—a system may tend towards centralized–centralized network and governance, decentralized–decentralized, centralized–decentralized, or decentralized–centralized. Each configuration can be found across infrastructure sectors, and select examples are presented in figure 3. There does not appear to be clear boundaries between centralized and decentralized configurations, but instead that a gradient is applicable where the degree of de/centralization is inherently relative to the geographic and operational scale (or level of complexity, (Snowden and Boone 2007)) of the infrastructure systems. Reframing de/centralization across networks and governance to support resilience is a necessary and timely endeavour.

Zoom In Zoom Out Reset image size

A rigorous investigation into the conditions by which de/centralization lead to improved capacities to respond to disturbances remains underexplored and is critical for understanding how to adapt and transform infrastructure for future complexity. To respond to increasing complexity, infrastructure systems will need to be resilient while maintaining services; infrastructure resilience is defined broadly as the ability of infrastructure to respond and change to known and unknown disturbances (Park et al 2013, Woods 2015). Infrastructure managers can prepare for complexity by 1) recognizing the unpredictability of a future that cannot be based on past events (i.e. non-stationarity) across social, ecological, and technological contexts and 2) understanding that traditional infrastructure design approaches, and thereby much of centralized infrastructure, relied upon assumptions of stationarity (Chester and Allenby 2018, Helmrich and Chester 2020; Klein Tank et al 2009, Milly et al 2008). By relying on stationarity, it is more likely an infrastructure design envelope will be exceeded, leading to either temporary or catastrophic failure. Given the long design lives of built infrastructure, it is important to highlight governance, which may be more responsive, in periods of instability. Within infrastructure resilience literature, decentralization has been promoted as a tool to increase system resilience (Cascio 2009, Kwasinski et al 2019, Meerow and Stults 2016, Tomlinson et al 2015). This section reviews uses of de/centralization in resilience literature and examines how de/centralization can be used to promote resilience principles.

Large-scale, centralized infrastructure systems across the United States were created to provide basic services to citizens while taking advantage of economies of scale to reduce costs, contrary to decentralized systems which leverage economies of scope to enable multifunctionality and flexibility (Faguet 2014, Gilrein et al 2019, Goldthau 2014, O'Flaherty 2005). While centralized governance can coordinate disturbance responses through top-down authority, it can be restricted to perspectives of leadership, who may be removed from local contexts; therefore, centralized governance is best equipped to manage known, rather than unknown, disturbances (Mintzberg 1979, Rinaldi et al 2001, Uhl-Bien et al 2007). Centralized networks, which are designed for a low probability of failure, may have increasingly severe failure consequences when they do fail (Gilrein et al 2019). Furthermore, centralized networks and governance are less adaptable, limited by their scale or jurisdiction (e.g. the water sector is divided into wastewater, stormwater, drinking water, etc.) (Chester and Allenby 2018, Leigh and Lee 2019, Markolf et al 2018, Mintzberg 1979). Centralized systems have been deemed more vulnerable to failure than their decentralized counterparts, due to their fail-safe approach and lack of redundancy and/or modularity (Ahern 2011, Baran 1964, Gilrein et al 2019, Leigh and Lee 2019). Decentralized infrastructure are frequently depicted as systems that service local consumers through redundancy within the network (e.g. multiple producers or pathways), and the network redundancy is seen as a principle of resilience (Ahern 2011, Zodrow et al 2017). Decentralization also provides an opportunity for modularity, or the capability to readily adapt or scale a system by reorganizing individual technical or institutional components without significantly disrupting the overall system (e.g. smart grids increase information flows to identify and respond to network component failures (Li et al 2010)). Decentralized infrastructure systems may limit cascading failures amongst infrastructure sectors by quickly recognizing and isolating the failure (Gleick 2003, Goldthau 2014, Zodrow et al 2017). However, across these studies of de/centralization and resilient infrastructure, there is recognition that, while decentralized networks are frequently promoted for increased resilience, there are situations in which centralization can also be beneficial (e.g. balancing budget, reliability, and network size) (Gilrein et al 2019, Hines et al 2015, Leigh and Lee 2019, Zodrow et al 2017). There is less exploration of de/centralization in governance; however, complexity leadership and organizational science literature assert that the ability to transition between de/centralization, vertically and horizontally, is crucial for long-term viability because this flexibility allows organizations to traverse between modes of exploitation, i.e. business as usual, and exploration, i.e. innovation (Mintzberg 1979, Siggelkow and Levinthal 2003, Uhl-Bien and Arena 2018).

3.1. Infrastructure resilience as a de/centralization spectrum

Leveraging de/centralization as a spectrum across domains increases pathways for infrastructure managers to address growing complexity and develop resilient infrastructure. Biggs et al (2012) describe seven key principles of resilience: maintain diversity and redundancy, manage connectivity, manage slow variables and feedbacks, foster complex adaptive systems thinking, encourage learning, broaden participation, and promote polycentric governance. Using these principles, and the specific characteristics attributable to each, conceptual linkages between the de/centralization spectrum and resilience are proposed (figure 4). These conceptual linkages highlight the importance of recognizing both the network and governance domains of infrastructure systems, as well as the need to consider these domains in the de/centralization spectrum. As illustrated in figure 4, both centralization and decentralization can contribute to resilience efforts. Although decentralization appears to be more closely aligned with the resilience principles, there are instances where increased centralization may be warranted. Particularly, increased centralization appears beneficial when circumstance require greater levels of coordination and/or an understanding of the entire system (rather than sub-systems or individual components). Implementing and achieving all seven of the resilience principles, or even simply a subset, requires focus on both the network and governance domains of infrastructure systems. Specifically, it is important to recognize that there may be instances where resilience efforts necessitate more centralized governance in conjunction with a more decentralized network infrastructure. This dichotomy is likely to have limits, i.e. at some point, a network can become decentralized to the point of being unmanageable by a more centralized governance structure. Therefore, infrastructure managers must reimagine how de/centralization is used and promoted in network and governance domains.

Zoom In Zoom Out Reset image size

Towards supporting infrastructure resilience, we propose a dynamic, multi-dimensional framing of de/centralization through a coupled network-governance perspective with information as a mediator (figure 5). We hope that the perspective gained from this framing motivates managers to identify and consider trade-offs of different configurations along both the network and governance spectrums in their design process. There are two important considerations when applying the framework. First, the classification of de/centralized networks and governance relies heavily on context, specifically scale, as explored in figure 2. Second, context is further complicated by information flow—coupling legacy infrastructure with emerging information communication technologies (e.g. Internet of things, artificial intelligence) accelerates the capacity for distributed networks by increasing node connections and providing near real-time information and producing adaptive capacity that was not previously achievable. For instance, traffic maps have evolved to integrate wireless and GPS data to warn users in real-time of traffic jams and provide alternative routes (Zantalis et al 2019). Similarly, smart grids are growing capacity to monitor and manage electricity consumption through smart charging, taking advantage of non-peak hours and renewable energy sources, to improve service (Calma 2021 and Ghasempour 2019). Information flows must be recognized as increasingly powerful forces that are able to forge new connections between producers and consumers (as well as other emerging stakeholders, such as activists or lobbyists), while mediating existing relationships between governing bodies, infrastructure networks, and consumers. By empowering more producers and/or consumers with information (and, therefore, sense-making), as well as decision-making power to act, a system can more readily respond to instances of instability or failure.

Zoom In Zoom Out Reset image size

By managing the contributions of networks, governance, and information towards de/centralization, infrastructure managers increase their opportunities to address instability. While de/centralization configurations have often been applied to infrastructure networks, exploration of de/centralization within the governance domain of infrastructure has largely been ignored. Decentralized governance has generally been shown to increase organizational performance in periods of instability since it brings the decision-making power closer to those experiencing instability or those with expertise on the topic of conflict, while centralized governance typically performs well in stability (Andersen 2004, Mintzberg 1979, Uhl-Bien and Arena 2018). As identified in complexity leadership theory (Uhl-Bien and Arena 2018), instability creates a need for organizations to: (a) transition between operational (centralized) and entrepreneurial (decentralized) leadership, a skill labelled as enabling leadership and (b) recognize a regime shift to initiate this leadership shift. First, enabling leadership must take advantage of vertical and horizontal de/centralization. While vertical decentralization brings the decision-making power closer to the situation, horizontal decentralization diversifies who has the ability to make decisions. Divisional bureaucracies seen in infrastructure systems offer limited horizontal and vertical decentralization by concentrating decision-making and sense-making towards the top of divisional hierarchies (Chester et al 2020a, Mintzberg 1979). Second, leadership must identify and have the capacity to respond—individually and as an institution—to shifts in the environment. If an organization continues to operate during instability with governance processes designed for efficiency under stability, they will experience failure (Snowden and Boone 2007). By recognizing the multidimensional network-governance spectrum of de/centralization, infrastructure managers are provided a more nuanced view of de/centralization that creates more solution pathways to respond to instability.

Increased flexibility of network and governance domains would increase resilience, compared to the systems that are currently rigid and locked into place—whether centralized, decentralized, or distributed. Centralized systems are dominant in infrastructure networks and governances as this configuration has served well, particularly in times of stability. Acknowledging the de/centralization spectrum and the contribution of all three configurations towards resilience, there is an opportunity to integrate more attributes of decentralized and distributed systems. Infrastructure managers must design, create, and maintain infrastructure systems that augment adaptation and respond to disturbances outside their design envelope to continue operating effectively in a world characterized by instability. In terms of infrastructure networks, an immediate action would be to consider 'maintaining diversity and redundancy.' For instance, during COVID-19, passenger airlines began transporting medical supplies as well as people, showing response diversity. This change of function, additionally, required swift institutional action, allowing passenger airlines—not only to transport cargo—but to operate with larger quantities of dry ice than previously allowed in order to maintain extremely low temperatures for vaccine shipments (Kulisch 2020). This example additionally highlights how the network and governance domains do not operate in isolation—despite being analyzed separately in much literature on de/centralization.

A spectrum of de/centralization across network and governance domains, and with consideration of information flows, provides an abundance of pathways for infrastructure managers to address complexity. Scenarios describing how change might occur can illustrate these pathways. First, there is a scenario of slow, chronic change—much like that seen in climatic conditions across the world. Here, infrastructure systems need to 'manage slow variables and feedbacks' while 'encouraging learning'. This likely requires a decentralized network to allow for experimentation but, at least some degree of, centralized governance to collect and communicate information, such as successful network adaptations or emerging climate science. However, in a second scenario, where there is a sudden, destabilizing event such as an extreme weather event or a cyberattack, the physical network cannot likely be changed in a timely manner. In this situation, decision-making power should have 'broadened participation' to allow local infrastructure managers to respond immediately while 'managing connectivity' to maximize support and provisions to the impacted region. This type of response is likely best achieved through a horizontally decentralized governance structure—a feature found in 'polycentric governance.'

Furthermore, looking across infrastructure sectors, there are likely to be different priorities and definitions of operations and failures. In the power and water sectors, there is a focus on the physical network systems and their ability to maintain constant and consistent services, including within times of failure; while in the transportation sector service is maintained to meet demand and, therefore, may change over time. (Relatedly, information flow also depends on demand—e.g. demand for information increases during times of instability when decision-makers need to respond quicky.) Viewing de/centralization as a spectrum allows infrastructure managers to consider the strengths and weaknesses of each for their particular sector and create a symbiotic relationship between configurations to respond to instability rather than to place one configuration in opposition to the other, encouraging infrastructure managers to 'foster complex adaptive systems thinking.'

As infrastructure managers pursue adaptation strategies to respond to increasing complexity in infrastructure systems and environments, they will need to confront existing configurations of networks and governance and critically examine the benefits and trade-offs of de/centralization. While existing framings of de/centralization may be polysemic, reconciling these framings expands infrastructure managers' design space to include both network and governance domains, increasing the opportunities for adaptation and transformation for infrastructure systems. It is critical that governance is observed as a domain of de/centralization. Emerging technologies that allow for rapid information sharing provide an opportunity to empower producers, as well as consumers, with decision-making capabilities to respond with speed and flexibility in instances of failure. This integration of information changes infrastructure systems from centralized or decentralized, to distributed. This increases the number of decision-makers in a system, which adds complexity along with flexibility, and confronts fundamental assumptions of who holds power. Centralization is dominant in today's infrastructure systems, but all three configurations—centralized, decentralized, and distributed—are at times appropriate for supporting resilience capacities because (1) infrastructure sectors have different priorities and, therefore, different solutions will be needed to meet each context, and (2) each configuration has varying abilities to respond to stability and instability. This reframing of de/centralization challenges infrastructure managers to develop new mental models—an evolving understanding of a system that is operational but not necessarily technically correct as it is influenced, oftentimes unconsciously, by an individual's beliefs—and, likewise, institutions to create new collective cognition (Jones et al 2011, Stevens and Gentner 1983). Reimagining de/centralization as a spectrum may help infrastructure managers navigate infrastructure systems through periods of instability by assessing, and reassessing, de/centralization configurations to increase actionable pathways forward for infrastructure resilience.

This work was in part supported by several Grants including from the United States National Science Foundation (SRN-1444755, GCR-1934933, DEB-1832016, CRISP-1832678, and CSSI-1931324).

No new data were created or analyzed in this study.