The prevalence of a bonding strategy in COVID-19 response networks: explaining intermunicipal partnering in the Santiago metropolitan region, Chile

ABSTRACT COVID-19, with its spatial particularities, has challenged the governance of institutionally fragmented regions, obliging local governments to respond collaboratively. While intermunicipal collaboration is expected, little is known about partnering strategies. Based on the institutional collective action framework, this article explains which partnering strategies (bonding or bridging) have prevailed in the city of Santiago. Using exponential random graph models, it reveals a tendency to transitive clustering, socio-territorial homophily and the importance of planned emergency intermunicipal networks. The findings demonstrate the prevalence of bonding and an absence of bridging, measured by both endogenous and exogenous factors, with implications for effective regional response.


INTRODUCTION
In July 2020, Chile's Santiago Metropolitan Region had one of the world's largest urban COVID-19 outbreaks (Canales, 2021;Gozzi et al., 2021).Two months after the first case was reported, this metropolitan city experienced a phase of uncontrolled expansion of the virus, reaching 35 cases per 100,000 inhabitants daily.It rapidly became the epicentre of the national epidemic, accounting for some 70% of total cases and more than 60% of reported deaths (MINSAL, 2021), all of which strained regional governance capacity to respond to the crisis.Paradoxically, Chile has also been recognized as achieving a successful COVID-19 response, mostly because of the rapid implementation of its vaccination campaign (Toro-Ascuy et al., 2022).In this response, the role of local governments and their collaborative capacity have been crucial (Castillo et al., 2021).
Collaborative networks have been studied as essential tools for emergency management because their more flexible, non-hierarchical structures are better suited to the multifaceted, uncertain context of disasters (Du et al., 2020;McGuire & Silvia, 2010).Indeed, collaboration has been key in COVID-19 response around the world since, by incorporating a wide range of actors in the management of the public health emergency, it expanded public capacities (Boin et al., 2021;Van Overbeke & Stadig, 2020).In contrast to floods, hurricanes or wildfires, COVID-19 is spatially invisible and cuts across jurisdictional boundaries (Boin et al., 2021;Dzigbede et al., 2020;Keough & Kaplan, 2021) and, therefore, obliges actors from different sectors and authorities at all levels of government to respond collectively.These spatial characteristics make the governance of coronavirus more complex in the context of an institutionally fragmented urban region where multiple municipal units needed to create collaborative networks in order to achieve a coordinated response.
In the Santiago Metropolitan Region, a polycentric urban system fragmented into 52 local units, the COVID-19 response was implemented partly through a collaborative intermunicipal network.While this network has been documented (Arias-Yurisch et al., 2020), the intermunicipal partnering strategies underlying the creation of the COVID-19 response network have remained unexplored.Since intermunicipal networks are formed through horizontal and voluntary collaboration (Arias-Yurisch et al., 2019;Teles, 2016), local actors have autonomy in selecting the partners with whom they collaborate.Based on the institutional collective action (ICA) framework (Feiock, 2013;Kim et al., 2022), this study argues that the partnering strategy adopted by local governments when forming the COVID-19 response networks will reflect their efforts to reduce the costs and risks of collaboration.
Local actors can reduce these risks and costs by generating different strategies based on two types of social capital: bonding and bridging (Jung et al., 2019;Shrestha, 2022).Network research has linked this distinction with the formation of specific network structures.Bonding capital is associated with the formation of connections with a few close and trusted partners to enhance credible commitment, and bridging capital with the formation of connections with a broader set of actors to ensure access to different resources (Andrew et al., 2016;Berardo, 2014;Nohrstedt, 2018;Tao & Zhang, 2020;Yeo, 2018).Previous research has defined bonding and bridging exclusively as endogenous patterns that are already embedded in a network structure.This endogenous dimension tells part of the story; actors' attributes and contextual factors, the exogenous dimension, tell the other part.Expanding previous work, this study advances an innovative proposal by embracing a comprehensive approach that conceptualizes bonding and bridging using both their endogenous and exogenous dimensions.
Given the particularities of the COVID-19 pandemic, local actors can be expected to collaborate in responding to the emergency.However, we do not know which type of strategy they may activate to select partners.In this context, the research question guiding this article is: What predominates in local actors' strategies for selecting partners for a COVID-19 response network: bonding or bridging?Using exponential random graph models (ERGMs), this article explains which intermunicipal partnering strategies (bonding or bridging) have prevailed in the formation of the COVID-19 response network created by municipal governments in Chile's Santiago Metropolitan Region.
The study of intermunicipal partnering strategies is critical in the coronavirus context since local governments, as the first point of contact with citizens, have considerable responsibility in the pandemic response but, on their own, frequently lack the resources and specialization for effective disaster management, making collaboration key in the response (Dzigbede et al., 2020;Lee, 2019).In this scenario, empirical insights into the partnering strategies of the intermunicipal network that emerged in Santiago in the context of the COVID-19 response provide lessons for future collaborative management of emergencies.
The remainder of the paper is structured as follows.The next section introduces the ICA framework and the study's main hypotheses regarding the bonding and bridging partnering strategies.The article then describes the case, methods and results before closing with a discussion of the study's implications and limitations.

INSTITUTIONAL COLLECTIVE ACTION IN EMERGENCY CONTEXTS
Research on collaboration between policy actors can be divided into two major streams: one line understands it as a formal organizational arrangement and focuses on collaborative processes and management (Ansell & Gash, 2007;Emerson & Nabatchi, 2015), while the other conceptualizes the phenomenon as a self-organized aggregation of individual decisions and studies collaborative structures (Feiock, 2013;Kim et al., 2022).Under the latter, the ICA framework focuses explicitly on the analysis of alternative mechanisms and partnering strategies that can reduce transaction costs when forming collaborative networks in fragmented systems (Feiock, 2013;Kim et al., 2022).
In particular, ICA research has been prolific in the study of fragmented metropolitan governance structures (Andrew & Carr, 2013;Andrew & Hawkins, 2013) and has been widely applied in emergency management studies to explain the formation of self-organized networks (Andrew & Carr, 2013;Feiock, 2013;Jung et al., 2019;Jung & Song, 2015;Kim et al., 2022).By considering how actors select strategies in contexts characterized by interdependence, ICA offers theoretical propositions to explain the intermunicipal partnering strategies that have prevailed in the COVID-19 response network in the fragmented context of the Santiago Metropolitan Region.
The ICA framework integrates multiple research traditions and theoretical approaches such as institutional analysis theories, social capital theory, the network approach and transaction cost theory to better understand collective action dilemmas (Andrew et al., 2016;Feiock, 2013).It assumes that actors are at least boundedly rational, situated within socially constructed environments, and base their strategies for participation in collective action on the maximization of the benefits and the reduction of risks and transaction costs (Kim et al., 2022).
From an ICA perspective, disaster situations generate collaborative risks and costscoordination, division and defectionthat can shape partnering choices (Andrew et al., 2016;Andrew & Carr, 2013;Jung et al., 2019).Under the pressure for an opportune reaction to a crisis, multiple logistical demands can overwhelm the capacity of local governments and increase miscommunication among actors, intensifying the risk of an uncoordinated response (Andrew et al., 2016).Furthermore, the indivisible nature of emergency functions makes it difficult for participants to agree on how to share the costs (Andrew & Carr, 2013).Finally, in an emergency context, there can be strong incentives for organizations to free ride on the work of others in the region (Andrew & Carr, 2013).
To explain how actors can mitigate these risks and costs, the ICA framework has incorporated the notions of bonding and bridging from social capital theory and has linked these notions with specific types of partnering strategies that are reflected in different network structures (Andrew et al., 2016;Berardo, 2014;Nohrstedt, 2018;Park & Feiock, 2006;Yeo, 2018).These strategies have been tested taking into account endogenous network configurations.This article expands this work by including exogenous aspects associated with homophily/heterophily and partnership history, which have previously been conceptualized as unrelated strategies.The remainder of this section discusses bonding and bridging strategies, incorporating the structural, actors' attributes and contextual dimensions.

Bonding strategy
A bonding strategy implies establishing strong connections with a few close and trusted partners as a means of enhancing credible commitment (Andrew & Carr, 2013;Berardo, 2014;Nohrstedt, 2018).This strategy can be studied as a complex phenomenon in which three separate social processes, based on structural, actors' attributes and contextual dimensions, converge.
The structural dimension captures transitive clustering or, in other words, the tendency of actors sharing ties with the same partners to form ties with each other (Tao & Zhang, 2020).In this sense, bonding evokes a closed, dense network structure, similar to a clique or triangle in which all three members are linked to each other (Berardo, 2014).Overlapping relationships create multiple avenues for communication of the same information and opportunities to build reputation, which can be valuable resources during an emergency (Andrew & Carr, 2013).The urgency and uncertainty of the COVID-19 crisis reduced the time actors had to assess the reliability of information and the trustworthiness of potential collaborators (Nohrstedt, 2018).In such a scenario, local governments may seek to activate bonding relationships that provide the redundant information needed to increase the credibility of communication (Andrew et al., 2016;Andrew & Carr, 2013) and the actor's reputation required to decrease enforcement and monitoring transaction costs (Abbasi, 2014;Andrew & Carr, 2013;Berardo, 2014;Jung & Song, 2015;Yeo, 2018).
A bonding structure's advantages are consistent with a homophilous relational strategy (Li & Mostafavi, 2021;McPherson et al., 2001).Putnam and Goss (2002) equated bonding with homophily, describing the former as connecting 'people who are like one another in important respects ' (p. 11).The theoretical argument of homophily is mainly instrumental (Berardo & Scholz, 2010;McPherson et al., 2001;Yeo, 2018).In a disaster context, ties with another actor with similar attributes provide a shared operational cognition for defining problems and approaches to them.This common ground may help to moderate conflicts when negotiating issues (Tao & Zhang, 2020) and increase trust-building and reciprocity (Chen et al., 2019).Emergency management studies suggest that actors may seek homophilous relations based on political, institutional or socio-territorial similarities (Li & Mostafavi, 2021;Song et al., 2018) that imply shared expectations of behaviour in the face of the disaster.
Beyond the structural and actors' attributes dimensions, the bonding argument is consistent with the contextual effects of a shared history of relationships.Under ICA, the accumulation of social relationships can mitigate collaboration risks since repeated interactions over time favour trust-building (Jung et al., 2019;Kapucu & Hu, 2016;Kim et al., 2022).During a disaster, local actors may seek to activate previous emergency agreements, collaborating with known partners with whom they have a history of intermunicipal relationships.Longstanding ties can produce a tightly clustered network, creating a strong sense of togetherness based on mutual social obligations and a shared theory of action, that facilitates the credible commitment essential in crisis management (Andrew & Carr, 2013;Tao & Zhang, 2020).
Given the above, the COVID-19 response network in the Santiago Metropolitan Region can be expected to reflect bonding strategies based on transitive clustering, homophily and a history of relationships.Therefore, the bonding hypotheses are: Hypothesis 1a.Within the COVID-19 response network, local officials are more likely to build ties with a closed triangle configuration.
Hypothesis 1b.Within the COVID-19 response network, local officials are more likely to build relationships based on political, institutional and socio-territorial similarities.Hypothesis 1c.Within the COVID-19 response network, local officials are more likely to form ties based on previous intermunicipal relationships on emergency matters.

Bridging strategy
While bonding strategies facilitate cooperation by providing credible information and commitment, they may also introduce overwhelming resource demands and decrease the capacity for innovation enabled by multiple non-overlapping relationships (Berardo, 2014;Nohrstedt, 2018).An alternative strategy is to create bridge ties that extend beyond the actors' close set of acquaintances and connect them to other diverse groups (Andrew & Carr, 2013;Berardo, 2014;Nohrstedt, 2018).This study understands bridging as a phenomenon that integrates two separate social processes based on structural dimensions and actors' attributes.
The structural dimension captures the tendency of actors to seek partners who have been disproportionately selected by many others in the domain (Tao & Zhang, 2020).From a structural perspective, bridging is similar to a star-like configuration where actors create paths between those who are not directly linked to each other, ensuring access to different resources, facilitating coordination and promoting innovation (Berardo, 2014).These advantages can be key in COVID-19 management since the unique nature of the crisis calls for a coordinated response in order to obtain a comprehensive understanding of the situation and accelerate resource transmission.
The prevalence of a bonding strategy in COVID-19 response networks in the Santiago Metropolitan Region, Chile The associational benefits of bridging structures are consistent with a heterophily strategy.Indeed, scholars have equated bridging with heterophily by describing it as connecting dissimilar people (Putnam & Goss, 2002).In the current pandemic, actors may seek ties with dissimilar others with heterogeneous backgrounds, capabilities and expertise (Jung et al., 2019) to gain access to more novel and, perhaps, valuable resources (Nohrstedt & Bodin, 2020;Siciliano et al., 2021).Resource dependence theory (Pfeffer & Salancik, 2003;Salancik & Pfeffer, 1978) has provided a theoretical argument to explain a heterophily strategy: organizations, affected by their external environments, need to exchange resources.Organizations strategically form ties to widen their stock of skills and resources, which they would not achieve through interaction with similar partners (Nohrstedt & Bodin, 2020;Siciliano et al., 2021).
In line with this, the COVID-19 response network in the Santiago Metropolitan Region can be expected to reflect bridging strategies based on the presence of starlike configurations and resource heterophily.Therefore, the bridging hypotheses are: The next section presents the context of Santiago Metropolitan Region, the case where bonding and bridging strategies hypotheses are tested.

CONTEXT OF THE STUDY
The Santiago Metropolitan Region is an urban agglomeration with an area of 15,403.20 km 2 and is home to more than 8 million people, accounting for around 40% of the country's population (CENSO, 2017).Most of the regional population is concentrated in urban zones while the outer rural area, albeit representing 64% of the region's surface area, is home to less than 14% of the population.In the last two decades, the region has been evolving from a monocentric to a more polycentric urban structure (Fuentes & Sierralta, 2004;Link, 2008;Truffello & Hidalgo, 2015).This process has brought with it urban sprawl characterized by high levels of segregation (Agostini et al., 2016;Silva & Vergara-Perucich, 2021).
Institutionally, the Santiago Metropolitan Region is organized into three tiers: regional, provincial and local.At the regional level, governmental authority lies with both the governor, who is elected directly and chairs the regional council, and the central government's regional delegate, appointed by the president of the republic.In the Santiago Metropolitan Region, there are five provincial authorities.Appointed by the president of the republic, they correspond to an intermediary authority favouring coordination between the regional and local levels.At the local level, authority lies with municipal governments.These decentralized units are responsible for the development of their territories and play a key role in the provision of local public services.Municipal governments comprise a mayor and a local council, both elected directly.
The region is fragmented into 52 local units, which vary in important respects (see Appendix A in the supplemental data online).Their mayors are from different political parties, which are organized in left-and right-wing coalitions that are almost equally represented in the region, accounting for 54% and 46% of mayors, respectively (SERVEL, 2016).In terms of socio-territorial characteristics, there are 11 rural-based local units, located in the outer area and representing 21% of the region's municipal districts (SINIM, 2016).From a fiscal standpoint, there are important disparities among the units, with a 25-fold difference between those with the lowest and the highest municipal revenues per inhabitant, indicating the extent to which some are dependent on national transfers.In the context of the current pandemic, recent studies have found a strong correlation between the city's inequalities and the uneven distribution of COVID-19 impacts (Canales, 2021;Gozzi et al., 2021;Mena et al., 2021;Villalobos Dintrans et al., 2021).COVID-19 incidence in the Santiago Metropolitan Region is shown in Figure 1.
In the face of the pandemic, Chilean municipal governments were called on to act since they are mandated to provide primary healthcare services and have co-responsibility for emergency management (Law N°18.695).In the Santiago Region, they have different institutional capacities for addressing emergencies.A total of 35 of its local units have an emergency office declared in their organizational chart, and 37 had previously established intermunicipal agreements on emergency matters (Arias-Yurisch et al., 2020).In other words, a significant number of the local units had a specialized office and planned emergency intermunicipal networks as managerial tools for handling the crisis.
In Chile, municipal governments have played a significant role in response to the pandemic (Castillo et al., 2021).They have provided a wide range of services from the basic distribution of masks to more complex actions such as the implementation of new health units and vaccination programmes (Belmar & Jofré, 2020;Castillo et al., 2021;Montecinos, 2020;Ramírez de la Cruz et al., 2020).Part of this local response was collaborative in nature taking the form of an intermunicipal network.While this network has been documented (Arias-Yurisch et al., 2020), the strategies through which the actors selected their partners remain unknown.This study explains which intermunicipal partnering strategies prevailed in the formation of the COVID-19 response network created by municipal governments in Chile's Santiago Metropolitan Region.

METHODS
To capture the COVID-19 intermunicipal network of the Santiago Metropolitan Region, an online survey was applied (approved by the Ethics Committee of the University of Santiago of Chile under Letter N°416/2020).The invitation to participate was emailed to municipal officials over the course of one month (June-July 2020) and included a link to the survey and an informed consent form, obtaining responses from 96% of the region's 52 municipal governments.Respondents could report all their collaborative partners out of a list of these 52 municipal governments.With this information, we built a matrix of intermunicipal cooperation in the context of the coronavirus pandemic.This was then symmetrized to obtain the undirected network corresponding to the dependent variable.In this network data, nodes are municipal governments and ties are the collaborative relationships among these units.
To test bonding and bridging hypotheses, we estimate a series of ERGMs using the open-source software package MPNet (Peng et al., 2009).ERGMs are tie-based models for understanding network formation, which conceive observed network structures as possible outcomes of stochastic network processes (Lubell et al., 2012;Lusher et al., 2012).They are conceptually similar to logistic regression analysis where a link between two nodes is predicted by configurations in the network that can be viewed as independent variables in the model (Nohrstedt, 2018).ERGMs attach parameters to local network configurations that represent their likelihood, given all the other configurations included in the model.Configurations that are likely outcomes of a social process within the network occur at a higher frequency than by chance when controlling for other relevant processes (Nohrstedt, 2018).
ERGMs are appropriate for understanding regional dynamics since they are applied when the object of study is all the ties among a set of actors in a territory (i.e., the whole network) rather than the ties that one actor has in a region (egocentric networks).Despite the theoretical and empirical importance of studying networks at a regional level, work of this type has been conspicuously absent in the literature, a lacuna that this research attempts to fill.
With ERGMs, endogenous and exogenous processes can be modelled to explain the formation or dissolution of ties.On the one hand, endogenous factors are considered self-organizing network structures because the likelihood of a tie between two actors is dependent on the presence or absence of other ties in the same network (Siciliano et al., 2021).On the other hand, exogenous factors are particular attributes of the node and other contextual factors that may explain the existence of a tie (Lusher et al., 2013).This property of ERGMs allows us to operationalize bonding and bridging strategies as comprising endogenous and exogenous components, in line with the conceptual proposal of this article.
The endogenous component of bonding is measured using the 'alternating triangles' parameter, a higher order statistic of the denser region of multiple triangulation, which corresponds to Hypothesis 1a.The exogenous components of bonding are based on homophily and previous relationships.The Homophily Hypothesis (1b) is measured by political, institutional and socio-territorial variables collected from the official records of different Chilean governmental agencies.Political homophily is codified by the match based on the mayor's political coalition and institutional homophily by the match based on municipal organizational structure considering the existence of an emergency office.Socio-territorial homophily is composed of both urban-rural similarity and geographical contiguity (neighbourhood).Hypothesis 1c is measured by previous intermunicipal relationships captured by local agreements on emergency matters (i.e., planned emergency intermunicipal networks).'Interaction' and 'network activity' parameters are incorporated into the models to gauge the effects of exogenous variables.The operationalization of the bonding strategy is shown in Table 1.
In the case of the bridging hypotheses, the endogenous component is measured using the 'alternating star' parameter, a higher order statistic of star-like configuration (Hypothesis 2a).Moreover, the exogenous component is based on resource heterophily, which corresponds to Hypothesis 2b.Resource heterophily is codified by the difference between municipal dependence on national transfers (%) obtained from the official records of the National System of Municipal Information.It is measured by an 'interaction' parameter in the models.Table 2 presents the operationalization of these variables.
Additionally, as control variables, we included a density parameter and basic activity configurations presented in Table 3.The density parameter is used to account for the formation of ties independently of other configurations described above.The activity configurations help to capture the likelihood that actors' attributes associated with political affiliation, institutional capacity, socio-territorial characteristics and resources will be active.
In this study, we constructed two models to test our hypotheses.Model I examines exclusively the endogenous effects of partnering strategies: the bonding structure (Hypothesis 1a) and the bridging structure (Hypothesis   2a).Model II captures the comprehensive approach of bonding and bridging strategies, examining the impact of different types of homophily: political, institutional and socio-territorial homophily (Hypothesis 1b), planned emergency intermunicipal networks (Hypothesis 1c) and resource heterophily (Hypothesis 2b).In addition, this model incorporates the effect of control variables.The following section sets out the results of the analysis.

FINDINGS
The COVID-19 response network (Figure 2) comprises 182 links and has a density of 0.069, indicating that, out of all the possible ties between local actors, only 7% were observed in the case.The network is broken into five components, with a fragmentation of 0.149, indicating that 15% of pairs of nodes remain unconnected.The main component accounts for the interactions of 92% of the region's municipal governments; only four do not participate in any of these network interactions.The degree of centralization of the COVID-19 network is 0.245, a value that reflects the presence of eight municipal governments that concentrated interactions.Finally, the transitivity measure is 0.168, indicating that, out of all relationships, 17% were tightly clustered triads.This simple description provides some hints by illustrating the presence of star-like configurations and closed triangles in the COVID-19 network structure.However, it does not offer a definitive answer to the question of which configurations predominate.For this, we turn to statistical cross-sectional analysis of the network using ERGMs.
Table 4 summarizes the ERGM results.The density parameter is negative and significant in both models (coefficients of -3.95 and -3.34, respectively), indicating the presence of fewer links in these networks than expected in random networks when taking into account the other structural configurations included in the models.Density metrics, which are extremely sensitive to the total number of nodes, describe the connectedness of the networks.Lower levels of density can be viewed as reduced potential for collective action.
The results provide evidence to support the bonding structure hypothesis (1a), indicating that there are more of these configurations than would be expected by chance.This suggests that Santiago's local governments are more likely to establish collaborative ties with those closely knitted together in the context of the COVID-19 response network.The bonding structure endogenous effect prevails even when incorporating exogenous covariates.In the simple and extended models, coefficients are positive at 0.51 and 0.32, respectively, with a significance level of 0.05.Simply put, there is a clear tendency for the COVID-19 network to exhibit transitive clustering.This is consistent with previous research showing that actors can forge tightly directed and closed relationships to gain access to redundant information and build trust, ensuring maintenance of the actors' commitment to collective solutions and reducing the risk of defection (Andrew & Carr, 2013;Tao & Zhang, 2020) This finding becomes more consistent when considering the parameters of the municipal governments' attributes and contextual factors associated with the bonding strategy.On the homophily hypothesis (1b), the analysis provides evidence of socio-territorial homophily.Rural/ urban local governments tend to collaborate with each other (coefficient of 3.13, significance of 0.05), even though rural ones are less active in the network than urban ones.Similarly, neighbouring units tend to collaborate more than those that are further apart (coefficient of 1.05, significance of 0.05).Interestingly, the coefficients The prevalence of a bonding strategy in COVID-19 response networks in the Santiago Metropolitan Region, Chile 387 REGIONAL STUDIES of both types of socio-territorial homophily are high, implying that local governments facing similar environmental conditions are much more likely to collaborate in addressing the negative consequences of the coronavirus pandemic than local governments with dissimilar characteristics and at a greater geographical distance.This result is consistent with the empirical literature in the field highlighting the important effect of socio-territorial homophily (Siciliano et al., 2021).The similarity of the problems faced by local governments with homogenous territorial characteristics favours the development of a common operational cognition to define an emergency response approach.The common ground facilitates communication and negotiation processes and increases the likelihood that collaboration will be maintained over time (Aung & Lim, 2021;Tao & Zhang, 2020).In previous research on emergency management, similarities based on political affiliation and institutional capacity were identified as important drivers of collaboration (Sapat et al., 2019;Song et al., 2018), but there is no evidence that these types of homophily played an important role in the COVID-19 response network in Chile.It should be noted that the absence of political homophily in this case coincides with other Latin American studies of intermunicipal collaboration (Mazzalay, 2011), hinting at the political particularities of the phenomenon in the region.
On the effects of previous relationships (Hypothesis 1c), the analysis supports the idea that planned emergency intermunicipal networks will influence tie formation in the actual emergency response.The network activity parameter is positive with a coefficient of 1.06 and a significance of 0.05.This positive effect was expected since repeated interactions increase actors' commitment, promote familiarity and reduce the perceived risk of collaboration.Prior relationships among local governments on emergency matters serve as a valuable resource in terms of collaborative preparedness, helping the actors to anticipate how they should behave, support each other and work together to provide a successful response (Andrew & Carr, 2013;Arias-Yurisch et al., 2020;Kapucu & Demiroz, 2011).The planned intermunicipal network structure can influence patterns of collaboration during emergency response since actors are expected to activate connections according to these interlocal plans (Abbasi, 2014;Kapucu & Hu, 2016).
In this case, bonding operates as a mechanism for predicting tie formation while the bridging strategy hypotheses are not supported by the analysis.The coefficients of bridging structure (Hypothesis 2a) and resource heterophily (Hypothesis 2b) are not significant at conventional statistical thresholds, indicating an absence of evidence of brokerage or resource dependency in the COVID-19 response.In contrast to current research (Andrew et al., 2016;Jung & Song, 2015;Nohrstedt, 2018), there is not a tendency to establish connections with popular actors or local governments that are dissimilar in terms of fiscal and administrative capacity.However, despite the absence of resource heterophily, the results show that local governments with less fiscal capacity are more active in the network than those with more resources.This suggests that more vulnerable units are more motivated to seek resources from others in the environment, whether similar or dissimilar, as a means of ensuring the local provision of services for the community during the emergency.Source: Produced using UCINET (Borgatti et al., 2002).
The overall findings help flesh out a coherent story about the actors' strategies of response to the coronavirus pandemic.The presence of a bonding strategy, measured by the endogenous effect of alternating triangles, is reinforced by the evident influence of exogenous covariates related to actors' attributes and contextual factors.This makes sense since, in the critical context of COVID-19, actors will strategically look to trustworthy partners, forming ties that are transitive, homophilous and based on previous relationships.A clustered community of high-trust cooperators can reduce the costs and risks of collaboration, particularly those related to miscommunication and defection (Nohrstedt & Bodin, 2020;Tao & Zhang, 2020;Yeo, 2018).Consistently, there is no evidence to back the bridging argument in either the endogenous or exogenous covariates.This may imply that, in an emergency, local actors may emphasize redundant, rather than diverse and newer, resources and local governments may find connecting with unfamiliar actors too risky.Taking all of this into account, the findings underscore the role of bonding relationships as a critical form of social infrastructure that facilitates the emergence of regional response to the COVID-19 crisis.

CONCLUSIONS
The COVID-19 pandemic has challenged institutionally fragmented urban regions because its spatial particularities have forced the development of collaborative responses by actors from different sectors and authorities from all levels of government (Boin et al., 2021;Dzigbede et al., 2020).Collaborative networks have been recognized as a predominant disaster response arrangement.However, we still know little about the partnering strategies behind network response, particularly at the regional level where intermunicipal cooperation can serve as a solution to institutional fragmentation.Based on the ICA framework, this study examined intermunicipal network configurations reflecting bonding and bridging strategies to explain which prevailed in the formation of the COVID-19 response network in the Santiago Metropolitan Region.We used ERGMs to determine which of these configurations predominates in the network.This study's main contribution is to provide a comprehensive perspective that permits consistent interpretation of these findings.By exploring endogenous configurations and exogenous factors, we show that the bonding effect prevailed over the bridging effect.The results reveal the structural prevalence of transitive clustering and, at the same time, the predominant tie-formation effects of socio-territorial homophily and previous intermunicipal relationships on emergency matters.These three drivers, which have been conceptually equated, share the underlying ideas of trust-building, congruent expectations and credible commitment.Together, they tell a coherent story, illustrating how local actors behave strategically under the current pandemic where uncertainty and urgency generate collaborative risks.
From a scholarly standpoint, a comprehensive approach to intermunicipal partnering strategies may be helpful in Despite the prevalence of bonding in this case, the absence of bridging effects may have implications for potential network effectiveness that should be addressed in future studies.Whereas bonding has facilitated cooperation among local actors in Santiago, it may also increase fragmentation in the regional response since the existence of closed municipal groups can hinder regional capacity to provide comprehensive and coherent policy (Norbutas & Corten, 2018).In addition, previous studies have shown that bonding may favour segregation under which local actors with certain characteristics can be excluded while others amplify their local voice, thereby reproducing regional inequalities (Gerber et al., 2013;Melamed et al., 2020).Consequently, Santiago's preexisting inequalities may be aggravated in the current crisis because of municipal bonding in the COVID-19 response.
To mitigate these potential risks of bonding, the formation of bridge ties and collaboration with dissimilar participants can be promoted.Bridging could strengthen regional capacity, favouring coordination and innovation.By facilitating the flow of resources, bridging structures can diminish the risk of a regional failure to meet community needs opportunely during a disaster (Andrew et al., 2016;Jung & Song, 2015;Nohrstedt, 2018).In addition, collaborating with different groups often brings new ideas, thereby increasing overall capacity to generate innovative solutions to the complex problems that arise during a disaster (Berardo, 2014;Burt, 2004Burt, , 2005)).
An emerging research line has emphasized that the optimal network structure combines bonding and bridging configurations (for reviews, see Siciliano et al., 2020) and, thus, they are not mutually exclusive and can appear simultaneously in any given network (Berardo & Lubell, 2016;Nohrstedt, 2018).In this regard, future collaborative management studies should explore the coexistence of bonding and bridging ties as a means of taking advantage of the benefits of both strategies in disaster contexts.Emergency plans or regional emergency planning committees can clarify coordinator roles and define the different partners' responsibilities while, at the same time, enabling the actors to get to know each other, thereby increasing their commitment and promoting trust-building.Such an integral approach would capture the complexity of effective regional disaster management in the face of a highly disruptive emergency such as the coronavirus pandemic.
Certainly, this research is not without shortcomings.First, by using ERGMs, it assumes that local actors' tiebased decisions are driven by social forces in underlying processes not directly measured through the observed network.Future research should explore these hidden forces in-depth, filling an ongoing gap in current network studies (Berardo, 2014;Siciliano et al., 2021).Second, networks are not static and may change over time.Consequently, it is important to learn more about the underlying mechanisms that make ties more or less likely to exist and persist.This could be achieved by conducting longitudinal analyses of link formation and network evolution using, for example, stochastic actor-oriented models.Finally, the findings presented in this article are highly dependent on the research context.They may be merely a snapshot of disaster management networks in Chile.To improve generalizability, the study could be replicated in other Latin American contexts as well as other continents to test the predominance of bonding relationships and their critical role as a form of social infrastructure that facilitates successful regional disaster management.
Hypothesis 2a.Within the COVID-19 response network, local officials are more likely to build ties with a star-like configuration.Hypothesis 2b.Within the COVID-19 response network, local officials are more likely to build relationships based on resource dissimilarities.

Figure 1 .
Figure 1.Distribution of COVID-19 incidence in the Santiago Metropolitan Region, July 2020.Note: Compiled based on COVID-19 data in July 2020 from the Chilean Health Ministry.Colour shades represent the level of incidence of COVID-19, measured as the number of cases divided by the municipal population; the darker the shade, the higher the incidence.The thick black line separates the 'urban zone' from the outer rural area.

Table 1 .
Bonding hypotheses: summary of operationalization of variables, configurations and sources.

Table 3 .
Control variables: summary of operationalization of variables, configurations and sources.
Source: Compiled by the authors.

Table 4 .
Simple and extended exponential random graph models (ERGMs).The prevalence of a bonding strategy in COVID-19 response networks in the Santiago Metropolitan Region, Chile 389 REGIONAL STUDIES improving the internal validity of studies of this type.This conceptual proposal, which simultaneously explores the strategies' endogenous and exogenous dimensions, seeks to deepen understanding of the collaborative logic of the actors' behaviour in a context characterized by uncertainty and interdependence.Future research should explore this multidimensional view of networks' features in more detail.Empirically, the study provides evidence about how local actors have collaborated to respond to the pandemic in a Latin American metropolitan city, which was heavily affected by COVID-19.The results of this case indicate that the importance of bonding relationships should not be underestimated in future policy responses at a level of government that is the frontline of crisis management.
Note: Baseline categories: a Left-wing coalition; b Emergency department in the institutional chart; and c Territory's rural characteristics.*Significant parameter estimates; rejects the null hypothesis of the parameter ¼ 0, p < 0.05.