Environmental Science and Policy 84 (2018) 60–68
Contents lists available at ScienceDirect
Environmental Science and Policy journal homepage: www.elsevier.com/locate/envsci
City membership in climate change adaptation networks
Sierra C. Woodruﬀ Department of Landscape Architecture and Urban Planning, Texas A&M University, 3137 TAMU, College Station, TX 77843, United States
A R T I C LE I N FO
A B S T R A C T
Keywords: Climate change adaptation Networks Collaborative Homophily Social vulnerability Capacity
By sharing best practices and lessons learned among member cities, professional and learning networks have become prominent actors in supporting and shaping local climate change adaptation. I analyze the membership of 18 highly visible adaptation learning networks to determine what cities participate and if networks attract similar cities. I ﬁnd that the formation of adaptation networks is driven by large, high-capacity cities. Adaptation networks include members of diverse sizes and planning capacity, however, cities with similar levels of social vulnerability and concern with climate change tend to participate in the same networks. Global and regional networks have diﬀerent patterns of membership. These patterns of membership have important implications for diﬀusing climate change adaptation between cities.
1. Introduction The impacts of climate change – streets ﬂooding, asthma attacks, damage from wildﬁres, or power outages – are determined not only by changing climate conditions but also the local built environment. In the U.S., many decisions about land use, infrastructure, hazard mitigation, and water resources that dictate how a city will respond to climate change are made at the local level (Nordgren et al., 2016). Local governments, consequently, are uniquely positioned to address the impacts of climate change and adaptation has largely been framed as a local issue (Shi et al., 2016; Nalau et al., 2015; Nordgren et al., 2016). In the U.S., cities have emerged as leaders of adaptation innovation and implementation (Graham and Mitchell, 2016; Shi et al., 2015). More than 40 cities and counties have created stand-alone adaptation plans (Woodruﬀ and Stults, 2016) and many others have integrated or mainstreamed adaptation into existing planning processes such as hazard mitigation, sustainability, and comprehensive planning (Rauken et al., 2014; Lyles et al., 2017). Framing climate change adaptation as a local issue, however, ignores the complex governance institutions and networks which motivate, enable, and shape local adaptation (Bulkeley and Betsill, 2013; Nalau et al., 2015). In the absence of strong federal leadership on climate change in the U.S., novel governance systems have emerged to support local adaptation (Lubell and Robbins, 2017). Learning and professional networks, that provide an opportunity for cities to exchange information about adaptation and learn from each other, have proliferated across the country. Diﬀusion of information and ideas through these types of institutions helped shape climate mitigation initiatives (Pitt, 2010; Anguelovski and Carmin, 2011; Ryan, 2015) and
E-mail address: swoodruﬀ@tamu.edu. https://doi.org/10.1016/j.envsci.2018.03.002 Received 1 September 2017; Received in revised form 2 March 2018; Accepted 3 March 2018 1462-9011/ © 2018 Elsevier Ltd. All rights reserved.
may similarly inﬂuence adaptation eﬀorts (Castán Broto and Bulkeley, 2013; Shi et al., 2016). The ﬁrst local adaptation plan in the U.S., was a collaboration between the City of Keene, NH and ICLEI-Local Governments for sustainability a network that provides member cities guides for adaptation and opportunities for shared learning. Today, high proﬁle networks like 100 Resilient Cities, the Southeast Florida Climate Change Compact, Urban Sustainability Directors, and the San Diego Regional Climate Collaborative are pushing forward the development and implementation of local adaptation. Cities voluntarily join these networks, to access information and resources (Bauer and Steurer, 2014; Busch, 2015; Vella et al., 2016; Westerhoﬀ et al., 2011). Since networks are primarily composed of municipalities, they represent a form of horizontal or polycentric governance outside formal, hierarchical structures (Fidelman et al., 2013; Kern and Bulkeley, 2009). While there is considerable variation in the scope and services of these networks, they all aim to support local climate change adaptation. Emerging adaptation networks can help local governments develop adaptation plans, policies, and programs through multiple avenues. Most importantly networks provide a forum for shared learning. Local practitioners rank learning from peers as one of the most important sources of climate adaptation information (Nordgren et al., 2016). Exchange of information among peers can proliferate information about climate vulnerabilities and potential adaptation strategies (Vella et al., 2016; Fidelman et al., 2013; Bauer and Steurer, 2014). Learning from other cities illuminates not only what is possible but also how it can be achieved (Busch, 2015). Networks also provide tools and guides to support local action. At times, networks also serve as a means for local governments to coordinate on shared vulnerabilities. As such,
Environmental Science and Policy 84 (2018) 60–68
Here, I only consider networks that (a) cities voluntarily join, (b) are horizontal or polycentric and thus constitute a form of self-governance, and (c) seek the implementation of measures through members rather than focus on lobbying or mobilization (Kern and Bulkeley, 2009). In addition, I only consider networks with formalized membership, where cities must become members to access materials or participate in network meetings. Loose cooperation or conferences are not considered (Busch, 2015). Many of the networks analyzed in this paper were initially founded to advance sustainability and climate mitigation; overtime they have taken up adaptation and become important in advancing climate preparedness in the U.S. This shift towards adaptation, mirrors the historic debate on climate change policy and the relatively recent acceptance of adaptation as necessary regardless of the success of mitigation (Busch, 2015).
adaptation networks enable member cities to consider a broader range of adaptation options and overcome barriers to adaptation (Moser and Ekstrom, 2010; Carmin et al., 2012; Nordgren et al., 2016). Despite having no direct decision-making power, adaptation networks play an important role in enabling and shaping local adaptation (Westerhoﬀ et al., 2011). As federal support for local adaptation diminishes in the U.S., networks are likely to become even more prominent in the ﬁeld. Consequently, it is critical to ask: what cities voluntarily join adaptation networks? Research on regional planning, formation of new governance institutions, and social networks suggest that a few high resource and innovative cities will drive the formation of adaptation networks (Berke et al., 2013; Lubell and Robbins, 2017). Many small cities may not have the capacity – ﬁnancial resources, personnel, and time - to participate. Moreover, there may be a tendency for cities with similar characteristics to participate in the same networks; a phenomenon known as homophily (Gerber et al., 2013). Sorting of cities into like groups may limit the ability of adaptation networks to share novel approaches with those that would beneﬁt from them most and to engender new adaptation eﬀorts. The scope of adaptation networks and whether they engage cities in a geographic region or across the globe may also inﬂuence what cities participate. Analyzing membership of adaptation networks will provide valuable insights into what cities have access to resources to help initiate and implement adaptation measures. In this paper I address three questions: (1) What communities are most active in adaptation networks? (2) Are similar communities more likely to participate in the same networks? And, (3) are the patterns of membership diﬀerent between networks that engage cities within a geographic region and those that have global membership? To address these questions, I analyze the membership of 18 highly-visible adaptation networks. I consider both “global” networks, such as 100 Resilient Cities, that include cities from across the country and “regional” networks, such as the Southeast Florida Regional Climate Change Compact, that connect cities in a deﬁned geographical area. I focus on member city’s social vulnerability and capacity, key determinants of climate impacts (Adger, 2003). Social vulnerability generally refers to the susceptibility of social groups to the impacts of hazards (Cutter, 2006). Chronic stresses like poverty and unemployment compound the risk of extreme events, leaving communities that already experience inequity more susceptible to climate change impacts (Shi et al., 2016). Capacity is deﬁned as a community’s ability to implement change that allows it to cope with climate change (Smit and Wandel, 2006). Key characteristics of capacity include economic resources, staﬃng, technical resources, communication and information sharing, and institutions (Araya-Muñoz et al. 2016; Brody et al., 2010). In the following section, I expand on how adaptation networks are deﬁned and the beneﬁts they provide to members. In addition, I explore the diﬀerence between global and regional networks. Drawing on the literature on regional planning, formation of new governance institutions, and social networks, I present a hypothesis for each research question. I then present the methods used to construct the dataset and analyze network membership. After which, I present and discuss the results. I conclude with the implications of these ﬁndings on the dissemination of adaptation and future research directions.
1.2. Beneﬁts of networks The formation of these novel governance structures can be understood as contracting process where local actors weigh the potential beneﬁts and transaction costs of diﬀerent types of interaction (Feiock et al., 2012; Lubell and Robbins, 2017). Through this lens, cities will participate in adaptation networks where the beneﬁts outweigh the transaction costs. Existing adaptation networks provide members multiple beneﬁts, but participation also comes with costs. At a minimum, participation requires staﬀ time, but in some cases cities must make binding commitments (such as creating a plan) or pay membership fees. Networks provide a forum for shared learning and can help local government overcome common barriers to adaptation including lack the technical expertise, staﬀ time, and funding (Moser and Ekstrom, 2010; Carmin et al., 2012; Nordgren et al., 2016). Global adaptation networks are largely designed as forums for shared learning. They bring together cities from across the country and world to exchange information and best practices. By collaborating communities can experiment with diﬀerent strategies, share innovation, and learn from each other’s experiences. Compared to networks that advance climate mitigation that usually promote the same mitigation actions across all member cities (e.g. fostering energy-eﬃciency), adaptation measures must be tailored to community context (Busch, 2015). The speciﬁcity of adaptation makes the transfer of knowledge more challenging and may result in networks tailored to a speciﬁc context. For example, the Mediterranean City Climate Change Consortium network connects cities in Mediterranean-climate regions since they will face many of the same climate impacts. The formation and proliferation of global networks suggest that “global” cities may share more in common with one another than with neighboring jurisdictions. New York City, for example, may beneﬁt more from collaborating with London than neighboring Newark, New Jersey. Global adaptation networks may also provide a platform for participating cities to attract investment and lobby for policy change at higher levels of government. It is widely recognized that climate mitigation – the reduction of greenhouse gas emissions - is a collective action problem (Ostrom, 2010). Cities bare the cost of reducing greenhouse gas emissions but the beneﬁts are diﬀuse, providing little incentive for action. By creating a sense of solidarity and providing assurance that others are taking action, networks can change city incentives for climate mitigation. In contrast, adaptation is generally described as a private good (Tompkins and Eakin, 2012). Only the residents of a city will beneﬁt from adaptation eﬀorts. Adaptation, however, is more complicated (Bisaro and Hinkel, 2016). Actions taken by one city often have spillover eﬀects on neighboring jurisdictions, requiring cooperation. Regional networks are grounded in managing climate impacts that span jurisdictional boundaries. Compared to global networks, regional networks focus more on shared vulnerabilities among member cities and coordination of adaptation actions. Climate change will aﬀect watersheds, transportation networks, and electrical distribution systems that span multiple jurisdictions. This scale mismatch tends to be
1.1. Adaptation networks While the adaptation literature has predominately focused on adaptation at the local level, local action is enabled and shaped by larger governance structures. Informal and voluntary networks have become increasingly important in supporting local action. Networks have been deﬁned as forums where stakeholders come together and partake in political processes outside the restraining procedures of representative democracy (Busch, 2015). In the context of this paper, networks are institutionalized spaces where local governments cooperate on and engage in climate adaptation. 61
Environmental Science and Policy 84 (2018) 60–68
networks may have lower transaction costs because it is easier to ﬁnd mutually beneﬁcial policy decisions when cities have similar policy preferences and attitudes (Gerber et al., 2013). Consequently, I hypothesize that adaptation networks will be composed of members with similar levels of social vulnerability and capacity.
the rule rather than the exception for most adaptation issues (Cosens et al., 2014). Regional networks that bring together multiple municipalities are one promising approach to achieve better scale-matching (Adger et al., 2005). Regional networks can help participating municipalities develop common goals and coordinate actions. In addition, regional networks can create economies of scales. By pooling resources to conduct vulnerability assessments and identify adaptation strategies, regional collaborations may also be able to reduce the cost of adaptation for participating cities. Local practitioners recognize that many climate impacts are best dealt with on a regional scale and call for more and better resourced regional governance structures (Nordgren et al., 2016). Regional networks also are an opportunity for cities to learn from peer cities. Local governments within the same region will tend to face the same climate impacts and the same adaptation opportunities and constraints. Consequently, local governments may learn more from other cities in their region. Regional networks can increase awareness and facilitate information ﬂow about best-practices (Vella et al., 2016). In light of these beneﬁts, multiple countries have fostered regional adaptation networks. In 2009, Natural Resources Canada established six Regional Adaptation Collaboratives across Canada to catalyze and coordinate adaptation planning, decision-making, and action. In the early 2000s nine regional adaptation partnerships emerged in the England following the United Kingdom Climate Change Impact Programme’s eﬀort to conduct regional impact studies. In the U.S. regional collaborations have also emerged such as the Southeast Florida Regional Climate Change Compact. The Compact, which is a forum for four counties and 26 municipalities to share information and coordinate actions, has been highlighted by President Obama as a model for the country (Vella et al., 2016). Greater support of regional adaptation eﬀorts was recommended by the State, Local, and Tribal Task Force on Climate Preparedness and Resilience that was convened by the White House in 2013 to help guide federal eﬀorts in support of local adaptation (Task Force, 2014).
H2. Adaptation networks include cities with similar levels of social vulnerability and capacity Regional and global networks have diﬀerent costs and beneﬁts, which may give rise to diﬀerent patterns of membership. Cities can choose between networks, joining the one that best addresses their particular needs (Busch, 2015). By addressing shared vulnerabilities and creating economies of scales, regional networks provide cities different beneﬁts than global networks. The suite of services oﬀered by regional networks may entice participation of more vulnerable and lowcapacity communities. The engagement of diverse communities within a region may lead to less homogeneous groups. Therefore, I hypothesize that regional networks will be less heterogeneous and engage lowercapacity communities. H3. Regional networks are less homogeneous and engage more vulnerable and lower-capacity communities than global networks 2. Methods 2.1. Adaptation networks and member cities Eight highly-visible global networks were included in this analysis: C40, STAR Communities, 100 Resilient Cities, ICLEI-Local Governments for Sustainability, Urban Sustainability Directors (USDN), Global Covenant of Mayors for Climate and Energy, and World Mayor’s Council on Climate Change. These networks are widely recognized as leaders in supporting local climate change policy and listed for resources for cities by the U.S. Environmental Protection Agency (US EPA, 2015) and the World Bank (World Bank, 2018). It is important to note, however, that multiple other networks exist. Consequently, this sample reﬂects a convenience sample of the most prominent networks in the U.S. As discussed previously, only networks that (a) cities voluntarily join, (b) are horizontal or polycentric and thus constitute a form of self-governance, and (c) seek the implementation of measures through members were considered for this analysis (Kern and Bulkeley, 2009). In addition, for all networks included in this analysis cities must become members to access materials or participate in network meetings (Busch, 2015). Most of the networks included in this study were created to advance sustainability or climate mitigation, but overtime have become important drivers of climate adaptation. ICLEI, for example, was founded in 1990 with the goal of advancing sustainable development goals but in the last decade has become a leader in local adaptation. In 2007, ICLEI collaborated with the City of Keene, NH to produce the ﬁrst local climate adaptation plan in the U.S. To support other local governments, ICLEI partnered with King County, WA to develop adaptation planning guidance. Since it’s release in 2007, the guidance has shaped many local adaptation plans across the country. While there is wide variation in the mission of these global networks, all provide adaptation resources such as guidance, metrics, and tools as well as opportunities for shared learning among members (for a more detailed description of each global network see Appendix A in Supplementary material). Regional networks were selected based on their engagement with the Institute of Sustainable Communities, an advocate and funder of regional adaptation eﬀorts. While this is a relatively restrictive requirement, collaboration with the Institute of Sustainable Communities indicates that these networks are well established and have a mission focused on climate adaptation. Ten regional networks are included: The Southeast Florida Regional Climate Change Compact, Public-Private Regional Resiliency Committee (P2R2), the National Capital Region,
1.3. Hypotheses on City membership in adaptation networks Participation in adaptation networks is voluntary. Consequently, there may be a self-selection bias towards actors that would likely take action regardless (Berke et al., 2013; Kern and Bulkeley, 2009). Based on an analysis of local government participation in and implementation of regional watershed planning, Berke et al. (2013) suggest that participants tend to have internal drivers that make them amenable to take action on their own. For example, communities that already had strong environmental and stormwater regulations were more likely to participate in watershed planning eﬀorts. Likewise, Kern and Bulkeley (2009) conclude that members in climate mitigation networks are pioneers looking to engage with other innovative cities. A similar pattern may occur in adaptation; cities that are already likely to prepare for the impacts of climate change are more likely to participate. Small, low-capacity cities may lack the ﬁscal and human resources to eﬀectively participate (Lubell and Robbins, 2017). I hypothesize that the most active participants in adaptation networks will be large, wealthy cities that are concerned with climate change. H1. Cities that are most active in adaptation networks will have high adaptive capacity and high concern with climate change impacts Self-selection into adaptation networks may also result in more homogeneous groups. A large body of theoretical and empirical work on social networks, suggest that actors with similar characteristics are more likely to form ties than actors with diﬀerent characteristics (Gerber et al., 2013). This phenomenon is known as homophily. More homogenous adaptation networks may have lower transaction costs and higher beneﬁts. Beneﬁts of participating in adaptation networks may be greater if members have shared issues, concerns, and goals. When coordinating actions to address shared vulnerabilities, more homogenous 62
Environmental Science and Policy 84 (2018) 60–68
Fig. 1. Map of communities that are members of adaptation networks that are included in the analysis.
selected in the analysis represent socio-economic status, household composition, employment, age, percent immigrants, and gender. To calculate an overall social vulnerability score, I rescaled and added the factors (for additional details see Appendix B in Supplementary material). Like social vulnerability, capacity is diﬃcult to measure and past studies have used a number of diﬀerent approaches and indicators (Araya-Muñoz et al. 2016; Brody et al., 2010). Here, I use median household income and city population to capture capacity. In addition, I calculate proportion population change from 2000 since shrinking cities are more ﬁscally constrained than those that have a growing tax base. In addition to social vulnerability and capacity, there are numerous factors that may inﬂuence city membership including political makeup, climate change perceptions, population density, government scale, and geography. To capture the inﬂuence of political characteristics on membership in adaptation networks, I include percent democrat deﬁned as the percent of voters in each county that voted for Barack Obama in the 2012 presidential election. I also include climate change perceptions or the percentage of residents in each county that are worried about climate change available from the Yale Program on Climate Change Communication (Howe et al., 2015). Since voting behavior and climate change perceptions were only available at the county scale, city characteristics are assumed to be the same as the county. Preliminary analysis revealed that percent democrat and climate change perceptions are highly correlated; so percent democrat was dropped from future models. Government scale – city or county – is also included as a dummy variable. The last variable included in the dataset is geographic distance, which I calculated based on euclidian distance between the centroids of each member in the network (for a full list of variables included in dataset see Appendix C in Supplementary material)
Metro Mayors Coalition, San Diego Regional Climate Collaborative, Lost Angeles Regional Collaborative for Climate Action and Sustainability, Sierra Climate Adaptation & Mitigation Partnership, Capital Region Climate Readiness Collaborative, Puget Sound Regional Council, Intergovernmental Planning Pilot Project, and Western Adaptation Alliance. As with the global networks, there is considerable variation among the regional networks (for details see Appendix A in Supplementary material), but all aim to support adaptation among member cities. Membership list for each network was collected from the networks website. Only U.S. municipalities and counties were included as members. Several networks include metropolitan planning organizations (MPOs), but MPOs were included in the membership list since they themselves lack decision-making power and act as coordinating bodies. In total, there are 418 cities and counties included in the analysis (Fig. 1). Of the 418 cities in the network, 112 (26.8%) participate in regional networks. Fifty-seven cities participate in both regional and global networks. ICLEI is largest network with 228 members. Global networks tend to be larger, on average including 81 members. The largest regional network, the National Capital Region, includes 20 members; the smallest P2R2 includes only 3. On average, regional networks include 11 members (Fig. 2).
2.2. Construction of dataset To analyze the inﬂuence of social vulnerability and adaptive capacity on city membership in adaptation networks, I created a dataset from multiple secondary sources. To measure social vulnerability, I calculated the Social Vulnerability Index (SOVI) for each municipality or county in the sample. There are a multitude of approaches to measure social vulnerability (Tate, 2012). SOVI is one of the most well-known and widely used measurements of social vulnerability. Since it was ﬁrst developed by Cutter et al. (2003) it has been applied extensively to diﬀerent geographic settings, spatial scales, and time periods. To calculate SOVI, I ﬁrst collected 26 socioeconomic and demographic variables from the American Community Survey 2010–2015 that are consistently used in social vulnerability measurements including race and age (for a full list of variables see Appendix B in Supplementary material). After standardizing and normalizing the variables, I used principal component analysis to identify independent factors that account for the majority of overall variance in the original data. The factors
2.3. Statistical analysis 2.3.1. What communities are most active in adaptation networks? To determine the most active members in adaptation networks, I calculated the number of networks each city participates in or degree. The number of networks that each city participates in was used as the dependent variable in a poisson regression with city characteristics such as social vulnerability, population, and concern with climate change as the independent variables.
Environmental Science and Policy 84 (2018) 60–68
Fig. 2. Illustration of network of adaptation networks. Each blue circle represents an adaptation network and each pink circle represents a community. The size of the pink circles represents the number of networks each community participates in, the most active community participates in 8 networks.
residents are members of adaptation networks. In fact, 60% of all cities over 250,000 residents are members of adaptation networks. These cities tend to have higher median household incomes, be more liberal, and more concerned with climate change than the U.S. average (See Fig. 3). That being said, there is considerable variation among member cities. Communities that participate in both regional and global networks have signiﬁcantly larger populations than communities that participate in only regional (diﬀerence = 315,368; p = 0.04) or global networks (diﬀerence = 344,583; p = 0.01). Communities that participate in both also have higher concern about climate change compared to communities that only participate in just regional (diﬀerence = 2.79; p = 0.03) or global networks (diﬀerence = 2.95; p = 0.002). Fig. 3 plots city characteristics as well as depicting the type of networks they participate in. The most active actors participate in 8 of 18 adaptation networks included in this analysis – this indicates membership in all but one global networks and one regional network. No city participates in more than one regional network due to their dispersion across the country. The poisson regression with the number of networks that a community participates in as the dependent variable (degree) reveals that communities with larger populations (b = 0.0000002; SE = 0.00000003; p < 0.001) and greater climate concern (b = 0.028; SE = 0.007; p < 0.001) tend to participate in more networks. Social vulnerability (b = -0.026; SE = 0.011; p = 0.01) and median household income (b = -0.000054; SE = 0.0000015; p < 0.001) have a negative relationship with degree.
2.3.2. Are similar communities more likely to participate in the same network? From the membership list, I created co-membership matrix. Essentially this converts the data from cities to dyads (or pairs of cities). For each dyad, I determine if the two communities participate in the same network. For example, if Seattle, WA and Boulder, CO are members of the same network then there would be a score of 1, if they are not members of the same network the pair would have a score of 0. Since my hypothesis focuses on similarity of cities, I calculate the absolute diﬀerence between cities on each of the attributes of interest. For example, the population for the Seattle-Boulder pair would be calculated by subtracting the population of Boulder (103,919) from that of Seattle (653,017). I used logistic regression to determine if similar cities are more likely to participate in the same network. In this model, the unit of analysis is the dyad between two communities (n = 87,153; since cities cannot have collaborations with itself, the total number of combinations is 418 × 417, because Seattle-Boulder is equivalent to BoulderSeattle you then divide by 2). The dependent variable is whether the two communities participate in the same network, if so coded 1 (40,504), if not 0 (46,649). The independent variables are absolute diﬀerence between community characteristics. If the hypothesis is supported and similar cities are more likely to collaborate, the probability of network formation will decrease as the absolute diﬀerence of the independent variables increase. 2.3.3. Are these trends diﬀerent between networks that engage cities within a geographic region and those that have global membership? To test if regional networks are more diverse and less homogeneous, I repeated the logistic regression for regional (n = 6216) and global (n = 65,703) networks separately. In addition, I ran an ANOVA to test if cities that participate only in regional networks are diﬀerent than those that only participate in global networks.
3.2. Are similar communities more likely to be members of the same networks? Based on the logistic regression of all city pairs, communities with similar levels of social vulnerability are more likely to participate in the same networks (see Table 1). If two cities are the same on all other characteristics, but are one standard deviation apart in social vulnerability the probability of them being in the same network drops from 54.0% to 50.7%. Capacity indicators have a mixed relationship with co-membership. The positive coeﬃcient (See Table 1) suggests that adaptation networks include communities with varying levels of income. Similarly,
3. Results 3.1. Network description Cities that are members of climate change adaptation networks tend to be large and liberal. All the cities in the U.S. with more than 1 million 64
Environmental Science and Policy 84 (2018) 60–68
Fig. 3. Member cities plotted on A) social vulnerability (SOVI) and population, B) SOVI and household income, and C) percent of the population worried about climate change and percent of voters that voted for Obama in 2012. Each dot represents a city, dot size represents the number of networks the city participates in, and color represents time of network the city participates in – regional, global, or both. Dotted lines indicate the national average. There is no signiﬁcant diﬀerence between cities that participate only in global and regional networks. Cities that participate in both have signiﬁcantly larger populations and are more worried about climate change. The percent of the population worried about climate change and population size signiﬁcantly increase the number of networks a city participates in, SOVI and income reduce the number of networks a city participates in.
adaptation networks appear to have members with a diversity of population sizes and population density. Communities with similar levels of concern over climate change and with similar levels of population change since 2000 are more likely to be members of the same network. There also appears to be sorting between government levels, counties are more likely to be members of the same network as other counties. Finally, communities that are geographic close tend to be members of the same adaptation networks.
Table 1 Logistic regression results for all networks, global networks, and regional networks. Coeﬃcients reported as well a standard errors in parentheses, asterisks indicate signiﬁcance at 0.05 level or greater. Variable
0.1605* (0.0019) −0.0315* (0.0020) 0.0,000,008* (0.000,002) 0.0000002* (0.000000008) −0.1543* (0.0157) 0.000,005* (0.000,002) −0.0091* (0.0014) −0.5166* (0.017) −0.0022* (0.0004) 87,153 118,945
0.5916* (0.0225) −0.0352* (0.0023) 0.0,000,003 (0.000003) 0.0000002* (0.00000001) −0.1298* (0.0161) 0.00004* (0.000003) −0.0171* (0.0016) −0.5383* (0.0204) 0.0022* (0.0005) 65,703 86,804.46
−0.468* (0.095) −0.016 (0.011) −0.000009* (0.000,002) −0.0000003* (0.00000005) 0.026 (0.143) −0.00003* (0.00,001) −0.179* (0.012) 0.321* (0.096)
SOVI Household Income Total Population Population Change Population Density Climate Change Concern City-County Distance N BIC
3.3. Are trends diﬀerent between global and regional networks? There is no signiﬁcant diﬀerence between communities that participate only in global or only regional networks on any of the characteristics measured. Repeating the logistic regression for regional and global networks separately, however, reveals remarkably diﬀerent patterns of sorting. Among global networks, communities with similar levels of social vulnerability are more likely to participate in the same networks. If two cities are one standard deviation apart in social vulnerability and all other variables are equal the probability of them being in the same network drops from 64% to 61%. Similarly, as the diﬀerence in population change between cities increase they are less likely to be members of the same global networks. There is also sorting in global networks based on concern for climate change - as the difference in climate change concern between cities increases the
Environmental Science and Policy 84 (2018) 60–68
2016). Networks may enable cities to take action that otherwise do not have resources to prepare (Westerhoﬀ et al., 2011). Sorting based on social vulnerability, however, suggests that historically disadvantaged communities that may bare the greatest cost of climate change are not members of the same networks as less disadvantaged communities. Suggesting that adaptation networks may be limited in their ability to distribute knowledge and resources to cities that need the greatest support. Yet, networks do include cities of diverse sizes and household incomes indicating that adaptation networks include members with diﬀerent levels of capacity. Adaptation networks may provide an opportunity for small, lower-income communities to learn from the experience of larger, wealthier cities that are at the forefront of adaptation planning and implementation. Further research is needed to examine if adaptation networks break down disparities between cities or reinforce them. I hypothesized that regional and global networks would have different patterns of membership, with regional networks engaging more vulnerable and low-capacity communities. There was no signiﬁcant diﬀerence between the communities that participate only in global and regional networks. Contrary to my hypothesis, regional networks appear to be more homogeneous. Communities with similar levels of household income, population size, concern for climate change, and population density are more likely to participate in the same regional networks. Global networks, on the other hand, appear to have less sorting along these characteristics. These results may reﬂect spatial auto-correlation of these variables rather than the inclusiveness of regional networks. Due to the restrictive selection criteria for regional networks, this research may not fully capture the importance and role of emerging regional networks. Many existing regional networks such as MPOs have integrated adaptation into their work and provide local governments in their jurisdictions with information and resources to begin preparing for climate change (Beiler et al., 2016). Another major limitation is that the data is based on membership alone, which may be a poor proxy for participation. In the case of the global networks, which have a diverse portfolio of programs, member cities may not be engaged in adaptation at all and joined the network for other services oﬀered such as climate mitigation. Future research should develop network data based on surveys that ask city oﬃcials which networks they participate in and how. Such an approach would extend this work to consider more holistically the complex governance systems cities are embedded in. In addition to participating in adaptation networks, cities may collaborate with local universities, attend trainings hosted by boundary networks, and apply for funding from state and federal agencies. Fully mapping this system would allow for deeper analysis into what enables and shapes local adaptation. With more complete data, network analysis could be used to address additional question about the complex government system that support adaptation. Although adaptation networks and multi-level governance are receiving increased attention, this topic remains under-studied (Busch, 2015). Multiple questions need further research, for example, do networks actually increase local adaptation eﬀorts and improve outcomes? While adaptation networks are intended to help local governments develop adaptation plans, policies, and programs there have been few empirical studies documenting the role of these networks. In particular, the strength of adaptation planning may be inﬂuenced by a cities position within governance networks by determining the type of information and resources they can access. Future research should extend this work to examine if cities participating in these networks are more likely to have plans and if they produce stronger plans than cities outside the networks. This type of work could begin to explore questions about diﬀerent structures of adaptation networks (Oberlack, 2016) such as does having more homogenous members result in more action? Finally, future research should incorporate qualitative approaches. Qualitative approaches provide more in depth understanding, which
likelihood of them participating in the same network decreases. If the percent of the population worried about climate change is one standard deviation diﬀerent in cities that are the same on all other attributes, the probability that they are in the same global network drops by 2.3%. Membership in global networks is also sorted based on government levels, with some networks appealing solely to municipalities. There is no sorting based on household income, total population, population density, or geographic distance. In contrast, membership in regional networks is sorted based on household income, total population, population density, and climate change concern. If cities diﬀer on these variable they are less likely to participate in the same network. Climate change concern has a particularly strong eﬀect: if the percent of the population worried about climate change is one standard deviation diﬀerent in cities that are the same on all other attributes, the probability that they are in the same regional network drops 18.2%. There is no sorting based on social vulnerability, population change, or government scale. The models for global and regional networks show opposite trends for population size, population density, and government level. In the regional model as the diﬀerence between cities on these characteristics increase, the likelihood of them being in the same network decreases. In global networks, on the other hand, a larger diﬀerence makes cities more likely to be members of the same network. Overall, regional networks appear to have more sorting with like cities participating in the same network. This may in part be explained by spatial autocorrelation. Since regional networks are composed of geographically clustered cities and networks are dispersed across the country, the variation within networks is likely less than between networks. For example, household income may be higher in Florida than Virginia. Consequently, the sorting in income may be a product of geography not a representation of the beneﬁts and costs for cities. In the other models, the distance variable helps partition this variation but distance was dropped from the regional model because it explains the presence of ties perfectly. 4. Discussion This research advances eﬀorts to understand the governance system that enables and shapes local action. While there have been several studies that attempt to explain the role and inﬂuence of individual adaptation networks (Burch, 2010; Fidelman et al., 2013; Vella et al., 2016), here I seek to explore patterns of membership across networks. Large, high-capacity cities are clearly the drivers of adaptation networks. Cities that participate in adaptation networks tend to be larger and wealthier than the average U.S. city. Among cities that do participate in adaptation networks, large cities tend to be the most active providing strong support for my ﬁrst hypothesis. Public support and policy preferences, however, also clearly play an important role. As the percent of residents worried about climate change increases, communities participate in more networks. Communities with similar levels of concern about climate change also tend to participate in the same adaptation networks. There is also sorting by social vulnerability and population change. These three characteristics – concern for climate change, social vulnerability, and population change – may be good indicators of city adaptation preferences. The adaptation options available to communities with high concern for climate change will diﬀer from those from communities where climate change is a more contentious issue (Adger et al., 2009). Similarly, adaptation will look diﬀerent in a growing city than a shrinking city. While the adaptation literature has not fully explored how adaptation will diﬀer across these contexts, this research suggests that diﬀerences may inﬂuence the networks cities join and the adaptation strategies they pursue. Placing the responsibility of adaptation on local governments, may exacerbate existing development gaps between large, wealthy cities and those that do not have the resource or capacity to prepare (Shi et al., 66
Environmental Science and Policy 84 (2018) 60–68
Cosens, Barbara, Gunderson, Lance, Allen, Craig, Benson, Melinda, 2014. “Identifying legal, ecological and governance obstacles, and opportunities for adapting to climate change. Sustainability 6 (4), 2338–2356. http://dx.doi.org/10.3390/su6042338. Cutter, S.L., 2006. Moral hazard, social catastrophe: the changing face of vulnerability along the hurricane coasts. Ann. Am. Acad. Polit. Soc. Sci. 604 (1), 102–112. http:// dx.doi.org/10.1177/0002716205285515. Cutter, Susan L., Boruﬀ, Bryan J., Shirley, Lynn W., 2003. Social vulnerability to environmental hazards. Soc. Sci. Q. 84 (2), 242–261. Feiock, Richard C., Lee, In Won, Park, Hyung Jun, 2012. Administrators’ and elected oﬃcials’ collaboration networks: selecting partners to reduce risk in economic development. Public Adm. Rev. 72 (s1), S58–68. http://dx.doi.org/10.1111/j.15406210.2012.02659.x. Fidelman, Pedro I.J., Leitch, Anne M., Nelson, Donald R., 2013. Unpacking multilevel adaptation to climate change in the Great Barrier Reef, Australia. Glob. Environ. Change 23 (4), 800–812. http://dx.doi.org/10.1016/j.gloenvcha.2013.02.016. Gerber, Elisabeth R., Henry, Adam Douglas, Lubell, Mark, 2013. Political homophily and collaboration in regional planning networks: POLITICAL Homophily. Am. J. Polit. Sci. 57 (3), 598–610. http://dx.doi.org/10.1111/ajps.12011. Graham, Alexandra, Mitchell, Carrie L., 2016. The role of boundary organizations in climate change adaptation from the perspective of municipal practitioners. Clim. Change(September). http://dx.doi.org/10.1007/s10584-016-1799-6. Howe, Peter D., Mildenberger, Matto, Marlon, Jennifer R., Leiserowitz, Anthony, 2015. Geographic variation in opinions on climate change at state and local scales in the USA. Nat. Clim. Change 5 (6), 596–603. http://dx.doi.org/10.1038/nclimate2583. Kern, Kristine, Bulkeley, Harriet, 2009. Cities, Europeanization and multi-level governance: governing climate change through transnational municipal networks. JCMS: J. Common Mark. Stud. 47 (2), 309–332. http://dx.doi.org/10.1111/j.1468-5965. 2009.00806.x. Lubell, Mark, Robbins, Matthew, 2017. Climate Adaptation, Sea-Level Rise and the Evolution of Polycentric Governance in the San Francisco Bay Area. Chicago, IL. . Lyles, Ward, Berke, Philip, Overstreet, Kelly Heiman, 2017. Where to begin municipal climate adaptation planning? Evaluating two local choices. J. Environ. Plan. Manage. (October), 1–21. http://dx.doi.org/10.1080/09640568.2017.1379958. Moser, S.C., Ekstrom, J.A., 2010. A framework to diagnose barriers to climate change adaptation. Proc. Natl. Acad. Sci. U. S. A. 107 (51), 22026–22031. http://dx.doi.org/ 10.1073/pnas.1007887107. Nalau, Johanna, Preston, Benjamin L., Maloney, Megan C., 2015. Is adaptation a local responsibility? Environ. Sci. Policy 48 (April), 89–98. http://dx.doi.org/10.1016/j. envsci.2014.12.011. Nordgren, John, Stults, Missy, Meerow, Sara, 2016. Supporting local climate change adaptation: where we are and where we need to go. Environ. Sci. Policy(May). http://dx.doi.org/10.1016/j.envsci.2016.05.006. Oberlack, Christoph., 2016. Diagnosing institutional barriers and opportunities for adaptation to climate change. Mitig. Adapt. Strat. Global Change(February). http:// dx.doi.org/10.1007/s11027-015-9699-z. Ostrom, Elinor, 2010. Polycentric systems for coping with collective action and global environmental change. Glob. Environ. Change 20 (4), 550–557. http://dx.doi.org/10. 1016/j.gloenvcha.2010.07.004. Beiler, Oswald, Michelle, LeylinMarroquin, McNeil, Sue, 2016. State-of-the-practice assessment of climate change adaptation practices across metropolitan planning organizations pre- and post-hurricane Sandy. Trans. Res. Part A: Policy Pract. 88 (June), 163–174. http://dx.doi.org/10.1016/j.tra.2016.04.003. Pitt, Damian Rogero, 2010. Harnessing community energy: the keys to climate mitigation policy adoption in US municipalities. Local Environ. 15 (8), 717–729. http://dx.doi. org/10.1080/13549839.2010.509388. Rauken, Trude, Mydske, PerKristen, Winsvold, Marte, 2014. Mainstreaming climate change adaptation at the local level. Local Environ. (January), 1–16. http://dx.doi. org/10.1080/13549839.2014.880412. Ryan, Daniel., 2015. From commitment to action: a literature review on climate policy implementation at city level. Clim. Change(April). http://dx.doi.org/10.1007/ s10584-015-1402-6. Shi, Linda, Chu, Eric, Anguelovski, Isabelle, Aylett, Alexander, Debats, Jessica, Goh, Kian, Schenk, Todd, et al., 2016. Roadmap towards justice in urban climate adaptation research. Nat. Clim. Change 6 (2), 131–137. http://dx.doi.org/10.1038/ nclimate2841. Shi, Linda, Chu, Eric, Carmin, Jo Ann, 2015. Global patterns of adaptation planning: results from a global survey. In: Seto, Karen C., Solecki, William D., Griﬃth, Corrie A. (Eds.), The Routledge Handbook of Urbanization and Global Environmental Change. Routledge, pp. 336–349. Smit, B., Wandel, J., 2006. Adaptation, adaptive capacity and vulnerability. Environ. Change 16 (3), 282–292. Task Force, 2014. State, Local, and Tribal Leaders Task Force On Climate Preparedness and Resilience Recommendations to the President 2014. https://obamawhitehouse. archives.gov/sites/default/ﬁles/docs/task_force_report_0.pdf. Tate, Eric., 2012. Social vulnerability indices: a comparative assessment using uncertainty and sensitivity analysis. Nat. Hazards 63 (2), 325–347. http://dx.doi.org/10.1007/ s11069-012-0152-2. Tompkins, Emma L., Eakin, Hallie, 2012. Managing private and public adaptation to climate change. Glob. Environ. Change 22 (1), 3–11. http://dx.doi.org/10.1016/j. gloenvcha.2011.09.010. US EPA, 2015. Resources for Resilience and Adaptation in New England (RAINE). US EPA Collections and Lists May (18), 2015. https://www.epa.gov/raine/resources-resilience-and-adaptation-new-england-raine. Vella, K., Butler, W.H., Sipe, N., Chapin, T., Murley, J., 2016. Voluntary collaboration for adaptive governance: the Southeast Florida regional climate change compact. J. Plan. Educ. Res. 36 (3), 363–376. http://dx.doi.org/10.1177/0739456X16659700.
can help address questions related to why patterns emerge. Interviews with participants in these networks could help illuminate why they joined speciﬁc networks, the costs of participation, and the beneﬁts. 5. Conclusion The adaptation literature has largely framed adaptation as a local concern, but this ignores the complex governance systems that enable and shape local adaptation. Moreover, emphasizing local action may exacerbate existing disparities between cities. In this paper, I examined 18 adaptation networks that support local adaptation. Speciﬁcally, I analyzed what cities participate in these networks and whether similar cities are more likely to participate in the same networks. The formation of these networks has been driven by large cities that would likely take action to prepare for climate change independently. The tendency for cities with similar levels of social vulnerability to participate in the same networks further raises concerns about whether existing institutions are redistributing knowledge and adaptation resources to cities that otherwise may not take action. To ensure that climate adaptation does not further increase disparities between small, vulnerable communities that cannot prepare for the impacts of climate change, it is important to further explore emerging governance institutions. Who participates in these institutions? And who beneﬁts from them? Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.envsci.2018.03.002. References Adger, Neil., 2003. Social capital, collective action, and adaptation to climate change. Econ. Geogr. 79 (4), 387–404. Adger, Neil, Arnell, Nigel W., Tompkins, Emma L., 2005. Successful adaptation to climate change across scales. Glob. Environ. Change 15 (2), 77–86. http://dx.doi.org/10. 1016/j.gloenvcha.2004.12.005. Adger, W., Neil, Suraje, Dessai, Marisa, Goulden, Mike, Hulme, Irene, Lorenzoni, Donald R., Nelson, Lars Otto, Naess, Johanna Wolf, Wreford, Anita, 2009. Are there social limits to adaptation to climate change? Clim. Change 93 (3–4), 335–354. http://dx. doi.org/10.1007/s10584-008-9520-z. Anguelovski, Isabelle, Carmin, Jo Ann, 2011. Something borrowed, everything new: innovation and institutionalization in urban climate governance. Curr. Opin. Environ. Sustain. 3 (3), 169–175. http://dx.doi.org/10.1016/j.cosust.2010.12.017. Araya-Muñoz, Dahyann, Metzger, Marc J., Stuart, Neil, Wilson, A. Meriwether, Alvarez, Luis, 2016. Assessing urban adaptive capacity to climate change. J. Environ. Manage. 183 (December), 314–324. http://dx.doi.org/10.1016/j.jenvman.2016.08.060. Bauer, Anja, Steurer, Reinhard, 2014. Multi-level governance of climate change adaptation through regional partnerships in Canada and England. Geoforum 51 (January), 121–129. http://dx.doi.org/10.1016/j.geoforum.2013.10.006. Berke, Philip, Spurlock, Danielle, Hess, George, Band, Larry, 2013. Local comprehensive plan quality and regional ecosystem protection: the case of The Jordan Lake Watershed, North Carolina, U.S.A. Land Use Policy 31, 450–459. Bisaro, Alexander, Hinkel, Jochen, 2016. Governance of social dilemmas in climate change adaptation. Nat. Clim. Change 6 (4), 354–359. http://dx.doi.org/10.1038/ nclimate2936. Brody, Samuel D., Kang, Jung Eun, Bernhardt, Sarah, 2010. Identifying factors inﬂuencing ﬂood mitigation at the local level in Texas and Florida: the role of organizational capacity. Nat. Hazards 52 (1), 167–184. http://dx.doi.org/10.1007/s11069-0099364-5. Bulkeley, Harriet, Betsill, Michele M., 2013. Revisiting the urban politics of climate change. Environ. Polit. 22 (1), 136–154. http://dx.doi.org/10.1080/09644016.2013. 755797. Burch, Sarah., 2010. Transforming barriers into enablers of action on climate change: insights from three municipal case studies in British Columbia, Canada. Glob. Environ. Change 20 (2), 287–297. http://dx.doi.org/10.1016/j.gloenvcha.2009.11. 009. Busch, Henner., 2015. Linked for action? An analysis of transnational municipal climate networks in Germany. Int. J. Urban Sustain. Dev. 7 (2), 213–231. http://dx.doi.org/ 10.1080/19463138.2015.1057144. Carmin, Jo Ann, Nadkarni, Nikhil, Rhie, Christopher, 2012. Progress and Challenges in Urban Climate Adaptation Planning: Results of a Global Survey. MIT, Cambridge, MA. http://web.mit.edu/jcarmin/www/carmin/Urban%20Adaptation%20Report %20FINAL.pdf. Castán Broto, Vanesa, Bulkeley, Harriet, 2013. A survey of urban climate change experiments in 100 cities. Glob. Environ. Change 23 (1), 92–102. http://dx.doi.org/10. 1016/j.gloenvcha.2012.07.005.
Environmental Science and Policy 84 (2018) 60–68
10.1038/nclimate3012. World Bank, 2018. “Actors: International Networks.” Guide to Climate Change Adaptation in Cities 2018. http://www-esd.worldbank.org/citiesccadaptation/ international.html.
Westerhoﬀ, Lisa, H. Keskitalo, E.Carina, Juhola, Sirkku, 2011. Capacities across scales: local to national adaptation policy in four European countries. Clim. Policy 11 (4), 1071–1085. http://dx.doi.org/10.1080/14693062.2011.579258. Woodruﬀ, Sierra C., Stults, Missy, 2016. Numerous strategies but limited implementation guidance in US local adaptation plans. Nat. Clim. Change(May). http://dx.doi.org/