Abstract
Sea level rise (SLR) is anticipated to be one of the most disruptive impacts of climate change for coastal communities if adaptive measures are not taken. Conceptualizing adaptation to SLR as a staged process whereby governments first choose whether to address SLR and then determine how much to address it, we use a hurdle model to examine what drives local governments to plan for SLR. Results from eighty-five coastal counties in the southeastern United States indicate that financial resources, population size, and future risk exposure increase the likelihood of addressing SLR. However, among counties addressing SLR, those with larger shares of left-leaning votes engage in more comprehensive SLR planning. Our findings align with previous research on barriers to addressing climate change and add to the growing knowledge on SLR planning.
Introduction
Climate change presents numerous challenges to society, both as a collective action issue and tall order for public administrators. As Moser and Ekstrom (2010) explain, the development of climate research has exposed adaptation deficits globally, contributing to the growth of studies on barriers to adaptation. An adaptation deficit refers to a “gap between what might be considered a well-adapted society to the existing climate and the actual and inadequate adaptation achievements of that society” (Ekstrom and Moser 2014, 55). Beliefs about climate change from the general public, support from elected officials, institutional fragmentation, financial resources, and administrative capacity have been consistently cited as general barriers (Moser and Ekstrom 2010; Krause 2011; Carmin, Nadkarni and Rhie 2012; Bierbaum et al. 2013; Yeganeh, McCoy and Schenk 2020; Hawkins, Krause and Deslatte 2023). However, because adaptation to climate change is characterized by multiple sustainability objectives with unique causal pathways to policy adoption (Hughes, Miller Runfola and Cormier 2018; Swann and Deslatte 2019), that are often led locally (Bierbaum et al. 2013; Hughes 2017), examining context-specific issues facing local governments remains critical for improving our understanding of adaptation to climate change.
Of the climate challenges facing society, SLR could be one of the most disruptive, given its long-term potential to reshape coastal communities with loss of habitable land, enhanced coastal flooding, and saltwater intrusion that could lead to migration away from low-lying coastal areas, reductions of tax revenue, demand for new infrastructure, and costly damages from extreme weather events, with the impacts being most pronounced if effective adaptive measures are not taken (Hauer 2017; Intergovernmental Panel on Climate Change 2019; Lincke and Hinkel 2021; Shi et al. 2024). Planning for these risks is challenging due to the uncertainty of long-term SLR projections and the range of service areas and infrastructure assets impacted (Haasnoot et al. 2019; Sweet et al. 2022; Grandage, Hines and Willoughby 2024). Still, coastal communities already experience visible and disruptive impacts, such as recurrent high-tide flooding that delays commuter times (Hauer, Mueller and Sheriff 2023), underscoring the importance of short-term adaptation decision making too.
Research on planning for SLR has illustrated numerous barriers to adaptation, including general beliefs about climate change, insufficient financial resources, unsupportive political environments, and lack of internal capacity (Ekstrom and Moser 2014; Hamin, Gurran and Emlinger 2014; Yusuf et al. 2016; Fu et al. 2019; John III and Yusuf 2019; Fu 2020; Siders and Keenan 2020; Milordis, Butler and Holmes 2023). This research has provided rich context on the barriers to SLR adaptation, with many studies providing in-depth case analysis of metropolitan areas. However, relatively few studies have covered wider geographic regions and used measures of SLR planning activity as dependent variables in quantifying the relative impact of these barriers (Fu et al. 2019; Fu 2020; Siders and Keenan 2020). Therefore, we provide original research on planning for sea level rise (SLR) in the southeastern United States, a region which has significant variation in SLR planning activity (Grandage, Hines and Willoughby 2024). Specifically, we measure the progress made by the eighty-five coastal counties stretching from North Carolina to Louisiana in planning for SLR and examine the impact that risk exposure, political environment, financial resources, and population size have on SLR planning practices.
Because we conceptualize adaptation as first choosing whether to address SLR and then determining how much to address it, we use a hurdle model (Mullahy 1986; Long and Freese 2014). This modeling approach suits the data well because some county governments have yet to formally incorporate SLR into their planning practices; most are in the early stages, and a few are relatively far along. We find that counties with greater per capita spending, population size, and risk exposure are significantly more likely to have nominally addressed SLR. However, the extent to which counties engage in more SLR planning is influenced by political environment, with counties having greater shares of democratic votes demonstrating broader efforts.
We begin by providing a background on climate change and planning for SLR before presenting the framework used for our analysis. Next, we describe our research methodology and data. The Results section summarizes the quantitative impacts of the variables, and the Discussion section that follows places the findings within the literature on climate change and SLR planning. We conclude the paper by explaining the key contributions of our study and outline directions for future research.
Climate Change and Planning for SLR
In its most recent summary for policymakers, the Intergovernmental Panel on Climate Change (IPCC) noted that almost half of the world's population live in areas that are highly vulnerable to climate change and highlighted a vast array of adverse impacts to water availability, food production, biodiversity and ecosystems, and damages from extreme weather events (Intergovernmental Panel on Climate Change 2023). Societies can take actions to deal with the causes of climate change via mitigation, specifically reducing greenhouse gas emissions. In addition, societies can take actions to reduce the consequences of climate change via adaptation, such as elevating homes and roads that are becoming more regularly flooded in the case of SLR.
The current and projected impacts of climate change, coupled with observations of adaptation deficits, have increased the focus on barriers to addressing climate change and the readiness of societies to mitigate and adapt. This research has provided key insights into the importance of political leadership, public support, stakeholder engagement, usable climate science, financial resources, and administrative capacity for overcoming barriers and building the capacity to manage climate risks (Moser and Ekstrom 2010; Krause 2011; Carmin, Nadkarni and Rhie 2012; Bierbaum et al. 2013; Carmin and Dodman 2013; Patt 2013; Ekstrom and Moser 2014; Ford and King 2015; Swann and Deslatte 2019; Yeganeh, McCoy and Schenk 2020; Hawkins, Krause and Deslatte 2023). To focus the scope of our literature review, we briefly discuss adaptation to climate change among local governments before analyzing SLR planning in particular.
Research on adaptation has focused strongly on the actions of local governments, given the inconsistent support from many national governments in proactively addressing climate-related issues and the eclectic nature of the impacts a locality may face (Carmin, Nadkarni and Rhie 2012; Bierbaum et al. 2013; Hughes 2017). Studies that examine the determinants of adaptation applying inferential techniques have used a variety of dependent variables, ranging from nominal measures that indicate the existence of a general climate change plan or commitment to developing one (Krause 2011; Reckien et al. 2015; Shi, Chu and Debats 2015) to count measures that capture the intensity of adaptation across different service areas (Krause 2012; Swann and Deslatte 2019). Studies tend to find that communities with greater financial resources, larger populations, and more liberal residents engage in more planning for climate change than their less resourced, smaller, and more conservative counterparts (Krause 2011, 2012; Carmin, Nadkarni and Rhie 2012; Reckien et al. 2015; Shi, Chu and Debats 2015; Yeganeh, McCoy and Schenk 2020). Other studies have found that collaboration is positively associated with adaptation and examined the factors that contribute to building the internal capacity necessary to address climate risks (Hawkins et al. 2016; Hawkins, Krause and Deslatte 2023).
Although the literature has points of convergence, a meta-analysis conducted by Yeganeh, McCoy, and Schenk (2020) highlights the need for additional research to clarify barriers to adaptation. The authors examined average elasticities for thirty-two studies and found that public support and population size are strong predictors of adaptation, whereas fiscal capacity and democratic leaning each had average effects that were positive but smaller in magnitude and had greater variation. Furthermore, because adaptation to climate change involves numerous sustainability objectives whose adoption is influenced by different factors (Hughes, Miller Runfola and Cormier 2018; Swann and Deslatte 2019), there is a need to examine context-specific issues.
With a significant portion of the world's population concentrated along coastal areas, the ability of societies to adapt to SLR has been a relatively active area of climate research. Without adaptive measures taken, SLR could lead to migration away from low-lying coastal areas and reshape coastal communities this century (Hauer 2017; Hinkel et al. 2018; Lincke and Hinkel 2021). Research on planning for SLR has illustrated the vast range of adaptive strategies available, including community outreach, accommodation of the built environment, protection with hard and soft engineering, and managed retreat (Hino, Field and Mach 2017; Siders 2019; Yusuf and John III 2021; Intergovernmental Panel on Climate Change 2022; Grandage, Hines and Willoughby 2024).
In the United States, uneven support for timely climate measures at the federal and state levels has led to an eclectic mix of coastal municipalities taking the lead on SLR (Moser 2013; Vella et al. 2016; Yusuf et al. 2016; Grandage, Hines and Willoughby 2024). Several studies have been conducted on barriers to adaptation within metropolitan areas via rich casework (Ekstrom and Moser 2014; Vella et al. 2016; John III and Yusuf 2019; Lubell and Robbins 2022; Yusuf et al. 2022; Gmoser-Daskalakis et al. 2023). In their study of the San Franciso Bay Area, Ekstrom and Moser (2014) conducted forty-three interviews and found the most frequently cited barriers to be institutional and attitudinal followed by lack of resources and supportive politics. John III and Yusuf's (2019) analysis of the Hampton Roads area of Southeastern Virginia surveyed sixty-three well-informed SLR stakeholders in the region and came to similar results, adding that barriers were present throughout the policy process, especially the development and implementation of options. Yusuf et al. (2022) also explained the importance of collaboration across organizational lines to foster whole-of-community approaches for effectively addressing SLR within this same region. Milordis, Butler, and Holmes (2023) provide an example of a study examining barriers extending beyond a metropolitan area. Their survey of ninety-six coastal municipalities in Florida found financial resources to be the most significant barrier, with roughly two-thirds reporting that inadequate funds limited their progress in addressing SLR.
However, fewer studies have examined SLR planning practices across wider geographic regions featuring numerous governments (Butler, Deyle and Mutnansky 2016; Butler, Holmes and Lange 2021; Hines, Grandage and Willoughby 2022), especially ones that use measures of SLR planning activity as dependent variables and quantify the relative impact of barriers discussed throughout this section on SLR planning activity (Fu et al. 2019; Fu 2020; Siders and Keenan 2020). Fu (2020) surveyed eighty-six coastal U.S. municipalities to create an additive index covering protection, accommodation, and managed retreat. Results indicate that greater adaptive capacity, including adequate financial resources and supportive political environments, along with population size, act as drivers for SLR planning. Siders and Keenan (2020) find that adaptation decisions are influenced by risk extent and type. Specially, they use public records to analyze coastline armoring and property acquisitions. They find that risk exposure, as measured by the proportion of residents in high risk flood areas, serves as a driver for adaptation whereas lack of financial resources was a key barrier. Finally, Fu et al. (2019) evaluated the quality of SLR vulnerability plans and found that greater resources led to better plans.
SLR Adaptation Framework
Figure 1 outlines the conceptual framework used to guide our empirical analysis. Like Patt (2013), we conceptualize adaptation as a sequential process involving two key components—first choosing whether to address climate risks and then determining how much to address them. As Figure 1 shows, we focus on the impact of risk exposure, politics, resources, and population size on SLR planning. Citations listed within the boxes are selected SLR publications that motivate our theoretical propositions and justify their inclusion in our model. Lines are drawn from these outer boxes toward the middle of the graphic to illustrate progression through stage 1 to stage 2. Although the general climate and SLR literature share key points in common together and SLR findings show some convergence on key barriers, we argue additional research is needed to quantify the relative impact of some variables and substantiate others. Below, we discuss these factors and specify our research questions for each. We begin by describing risk exposure and political environment and argue that their relationship to SLR planning is more ambiguous when compared to resources and population size.
SLR adaptation framework.
Siders and Keenan (2020) find that adaptation decisions are influenced by risk extent and type. In particular, their analysis indicates that the proportion of residents living within designated floodplains is the most important predictor for the occurrence of property buyouts and the second most influential factor for shoreline armoring. However, Fu's (2020) analysis of protection, accommodation, and retreat did not find an association between tactics implemented and the value of buildings at risk. In an experimental setting, Hines (2023) found that public works directors exhibit risk averse behavior when asked to prioritize projects protecting assets threatened by rising seas. Finally, Shi et al. (2024) found that fiscal exposure did not contribute to the prioritization of adaptive measures. In sum, the relationship between SLR adaptation and risk is not immediately clear and could depend on the measure of risk used. Therefore, we utilize different measures of risk and answer the following research questions: Does risk significantly increase the likelihood of addressing SLR and/or engaging in more comprehensive SLR planning?
Ekstrom and Moser (2014) found politics to be the fourth most cited barrier to SLR adaptation in their survey of the San Francisco Bay Area, pointing to a lack of political will among certain leaders and challenges in messaging the issue in a politically salient way. Milordis, Butler, and Holmes’ (2023) study of municipal governments in Florida found that approximately 30% of respondents indicated support from elected officials limited their progress in planning for SLR. Fu (2020) surveyed eighty-six local governments on how supportive their political environment was and found that higher levels of support contributed to more intensive SLR planning. However, it is possible that coastal residents provide greater bipartisan support for addressing local, visible climate issues such as SLR (Lee and Stecula 2021; Bromley-Trujillo et al. 2024), and that other factors such as resources and population size could play a more significant role. Thus, we argue that the relationship between political environment and SLR adaptation does not clearly emerge from the literature, specifically when put in partisan terms, and state our research questions as follows: Are political environments that are more left leaning significantly more likely to address SLR and/or engage in more comprehensive SLR planning?
In contrast to risk and political environment which were argued to have somewhat ambiguous impacts, the literature strongly converges on the positive association between resources and SLR planning activity. Survey and interview research has consistently reported financial and human resources to be one of the strongest reported barriers (Ekstrom and Moser 2014; John III and Yusuf 2019; Milordis, Butler and Holmes 2023). Furthermore, Fu (2020) found that greater resources contribute to more intensive SLR planning activity when controlling for population size, and Fu et al. (2019) found that more resources led to higher quality plans. Therefore, we state our research question as: To what extent are governments with greater resources more likely to address SLR and/or engage in more comprehensive SLR planning?
Hinkel et al. (2018) argue that in low-lying coastal areas with higher populations, it is generally highly beneficial to protect against even the more high-end SLR scenarios, whereas the protection of more rural areas provides less favorable benefit–cost ratios. In a study of smaller coastal municipalities, Hamin, Gurran, and Emlinger (2014) found that they often lacked the capacity to develop their own forecasts and focus on the ability to manage current hazards instead. Finally, Fu (2020) found that population size significantly increased the intensity of SLR planning, controlling for resources and political environment. Therefore, we state our research question as: To what extent does population size increase the likelihood of addressing SLR and/or engaging in more comprehensive SLR planning?
Data and Methods
To gather the primary data for this study on SLR planning practices, we employed a multi-method design drawing from governmental documents, specifically comprehensive plans, county commission meeting minutes, financial reports, and land-use plans. After collecting these documents from the eighty-five coastal counties stretching from North Carolina to Louisiana, we extracted text from them using keywords to identify SLR planning practices. 4 Overall, our data gives a snapshot of each county's progress in planning for SLR as of 2018. Specifically, we analyzed the most recent comprehensive and land-use plans as of 2018, county meeting minutes correspond to Fiscal Year (FY) 2017–2018, and the financial reports cover FY 2015 to 2017.
In total, we extracted 512 references to SLR and classified each reference to describe its focus area, risk type, and risk reduction tactic using an existing framework (Grandage, Hines and Willoughby 2024). This framework utilized the constant comparative method of grounded theory (Glaser and Strauss 1967), which is an inductive approach for categorizing data as it is collected, to generate the categories of focus areas, risk types, and risk reduction tactics. This allows for an iterative approach in defining categories as more is learned about the actual SLR planning practices. Importantly, a point of theoretical saturation was reached after analyzing roughly half the counties, meaning that further analysis of the data did not yield new categories beyond that point.
After collecting the documents from each county and generating the categories of focus areas, risk types, and risk reduction tactics, we coded the 512 extracted SLR references using a consensus-based process. Specifically, three authors participated in group coding meetings to reach agreement on the proper categorization for each reference. Table 1 illustrates how policy instances were coded using examples from Monroe County, Florida (the Florida Keys). Focus areas simply indicate what the government is concentrating their attention on, such as government-wide risks or specific service areas. Next, risk type describes the nature of the risk itself, and tactic identifies what is being done to address it, for example, regularly flooded roads may require elevation, which is a major focus of attention in the Florida Keys (Monroe County 2016, 2017). We created and referenced a database summarizing media articles and documentaries on SLR planning activity for each county where possible. Most often, especially for larger counties with local news outlets, this helped provide confidence in the operational validity of our data because adopted policies were being discussed in the press. As Hines, Grandage, and Willoughby (2022) explain, local news outlets in Florida cover SLR in their areas, and occasionally, discuss their progress in relation to peers. (Table 1).
Coding Examples.
The first two policies are from Monroe County's 2017 ACFR (page A6), and the second two are from their comprehensive plan (pages 6 and 11, respectively).
Figure 2 shows all of the focus areas, risk types, and risk reduction tactics, along with the percentage of instances assigned to each. Focus areas emphasize general items, such as policy development, planning and zoning, and government-wide risk. Evaluations of government-wide risk tend to highlight risks across multiple service areas but especially water and transportation. Most often, counties identify general risks to developing their communities with some going further by identifying specific threats to their portfolio of infrastructure assets and service obligations and broader implications for the natural environment and local economy. Finally, tactics for managing SLR typically involve the issuance of policy statements that provide guidance for long-term planning. However, some counties evidence more concrete actions, including intensive information gathering, regulations on land use and building codes, and investments in adaptive projects. Because of our focus on adaptation, we excluded any local policies aimed at emission reductions.
SLR policy categorization.
Figure 3 shows the number of tactics utilized by each county. As it indicates, there is significant variation in SLR planning practices, with some doing nothing, others doing a little, and some doing a lot. Therefore, we use a hurdle model to examine the factors that impact the likelihood of addressing SLR and then pursuing more comprehensive SLR planning. A logistic regression is used to predict if governments are formally addressing SLR. Among governments that choose to address SLR, a truncated Poisson is used to evaluate the breadth of planning (Mullahy 1986; Long and Freese 2014). This approach works well for studying adaptation because it can be interpreted as a sequential process whereby governments first choose whether to address SLR and then determine how much they address it. Indeed, as Patt (2013) explains, climate risk management involves two key initial hurdles—merging risk assessments with existing management plans for different service areas and then determining how to use this information.
Count of tactics used by county.
Specifically, our dependent variables are the number of focus areas, risk types, and risk reduction tactics, with the logistic regression used to predict if governments are doing anything and the truncated Poisson for how much, specifically by counting the number of areas, risks, and tactics addressed for those that have jumped the initial hurdle and decided to develop SLR policies. The hurdle model also improves statistical fit. Due to the fact that approximately 40% of counties do not have a SLR policy the variance for the count of focus areas, risk types, and tactics is larger than the mean. A standard Poisson model would likely suffer from overdispersion due to the large number of zeros. The hurdle model resolves this problem by modeling the zeros as a separate process (Mullahy 1986; Long and Freese 2014). This approach exposes more intensive forms of planning if they exist. For example, some county governments may simply identify general risks via focus areas but stop short of developing tactics to deal with them. Thus, while focus areas are key to our analysis, the tactics themselves arguably provide the most concrete form of action.
As Table 2 shows, our independent variables focus on risk, political environment, resources, population size, baseline resiliency, and social vulnerability. We include two measures of risk, one that is retrospective and another that is prospective. First, experiences with disaster may prompt organizations to adapt as they better understand their risks (Zhang, Welch, and Miao 2018). In the case of SLR, flood damages may serve as an impetus for adaptation. Therefore, we include data on flood damages from 2009 to 2018 to account for previous and recent shocks (Federal Emergency Management Agency 2023). 1 Second, as Siders and Keenan (2020) found, adaptation decisions can be driven by risk extent and type, specifically when considering the proportion of residents living in communities with high flood risk areas. Therefore, we include the percentage of the population anticipated to be impacted by annual flooding by 2050 (Hauer et al. 2021). Although left-leaning political environments are thought to be associated with greater willingness to address climate change than conservative ones (Unsworth and Fielding 2014; Yeganeh, McCoy and Schenk 2020), the relationship can be nuanced at the local level (Lee and Stecula 2021; Bromley-Trujillo et al. 2024), especially when controlling for other factors that drive adaptation (Stevens 2023). Because our primary data corresponds to 2018, we use the percent of county votes in the 2016 presidential election for Hillary Clinton (MIT Election Data and Science Lab 2018) as a proxy measure for political environment.
Descriptive Statistics for Independent Variables.
Given that SLR planning can be resource intensive, we consider financial resources from two perspectives—spending per capita and the relative size of the county's budget reserve, to account for resource utilization and fiscal slack, respectively. We calculate spending per capita using primary government expenses from the FY 2017 Annual Comprehensive Financial Reports (ACFRs) to capture the most recent spending in each county because spending across years was stable. As for the fiscal slack, we average the unassigned general fund balance as a percentage of total governmental funds’ revenue to account for normal fluctuations in reserves between FY 2015 and 2017 from the respective ACFRs (Afonso 2021; Finkler, Calabrese and Smith 2022). As the literature indicates, more populated areas are more likely to pursue adaptive measures. Therefore, we include population size for each county using the 2016–2020 American Community Survey data (Manson et al. 2023). We log transform population, spending per capita, risk exposure, and flood damages because of their skewed distribution. Finally, we control for baseline resiliency and social vulnerability. Regarding the former, we use the Baseline Resiliency Index for Communities (BRIC) which evaluates environmental hazard resilience at the county level considering social, economic, community, institutional, and environmental factors (Cutter, Ash and Emrich 2014; Hazards Vulnerability & Resilience Institute 2015). As for the latter, we use the Social Vulnerability Index (SVI) which considers socioeconomic factors and other stressors that make areas more vulnerable to environmental hazards (Cutter, Boruff and Shirley 2003; Hazards Vulnerability & Resilience Institute 2019). Both of the indices use census data collected prior to 2020 to provide a profile of each county. They are included to help answer the primary research questions while serving as control variables to ensure, for example, that financial resources are driving adaptation rather than general baseline resiliency or regional social vulnerability.
Having explained the collection of our primary data, variables used for our analysis, and the modeling strategy, we present results in the following section. 2 The Results section presents findings for each stage of the hurdle model, and the Discussion section that follows it places the findings within the climate change and SLR literature.
Results
Table 3 provides the logistic regression results for the first part of the hurdle model, which represents the most basic form of adaptation—simply addressing SLR. Consistent with the literature review, counties with greater financial resources and population sizes are significantly more likely to have addressed SLR. In particular, we find that spending per capita is a significant predictor, whereas budget reserve is not. Regarding our measures of risk, we find that prospective risk significantly increases the likelihood of addressing SLR, but that retrospective does not. Table 3 shows that political environment does not significantly impact the likelihood of addressing SLR. Finally, results indicate that neither baseline resiliency nor social vulnerability acts as significant drivers or barriers. In sum, the first stage indicates that spending per capita, population size, and prospective risk each play a key role whereas the same cannot be said for political environment, budget reserve, retrospective risk, baseline resiliency, and social vulnerability.
Logistic Regression Results.
Robust standard errors in parentheses. Standard errors are calculated in Stata using suest (Long and Freese 2014). + p < .10, * p < .05, ** p < .01.
Figures 4 to 6 provide a closer look at the average probability of addressing SLR as spending per capita, population size, and risk exposure change. Figure 4 demonstrates that increases in spending per capita significantly increase the chances of at least addressing SLR. In particular, as spending increases above the mean of $1,300 per resident, the lower bound of the interval stays above the proportion of counties that have nominally addressed SLR (62%). Next, Figure 5 illustrates that counties with larger populations are more likely to plan for SLR, with this probability increasing dramatically and then leveling off. Finally, Figure 6 shows that as the percentage of the county population projected to be exposed to annual flooding by 2050 increases, so does the likelihood that they have addressed SLR. Specifically, the probability of addressing SLR increases sharply as the percentage of the county population exposed rises from its minimum value to the median of 1.9%.
Impact of per capita spending on likelihood of addressing SLR.
Impact of population on likelihood of addressing SLR.
Impact of future risk exposure on likelihood of addressing SLR.
To help further understand the relative magnitude of resources, population, and risk exposure in driving SLR adaptation, we calculated how changes in each variable impacted the probability of addressing SLR. Specifically, we found that, on average, a 1% increase in spending per capita, population size, and risk exposure each leads to a 0.3, 0.2, and 0.1 increase in the percent chance that a county addressed SLR, respectively. Therefore, financial resources appear to have the largest impact followed by population size and risk exposure.
Each of the variables included for political environment, fiscal slack, retrospective risk, baseline resiliency, and social vulnerability were insignificant in our models. As for fiscal slack, this result must be interpreted jointly with the observation that spending per capita was found to be significant. Findings may simply reflect that counties with larger expenditures are more likely to allocate some of this spending toward resilience efforts and that fiscal slack is maintained for general economic risks. Similarly, null findings for our measure of retrospective risk, flood damages, should be considered in tandem with prospective risk. Although past flood damages may drive other adaptive actions, our forward-looking measure of risk appears to better predict whether counties are nominally planning for SLR, which is reasonable given that most of the drastic impacts from SLR will occur in the future. Although the political environment does not predict the likelihood of nominally addressing SLR, as our analysis of the second stage will show, it does influence the depth of SLR planning. To reiterate, the inclusion of BRIC and SVI as control variables allows us to infer that spending per capita, population size, and prospective risk are the factors driving efforts to address SLR rather than these composite measures of resiliency and vulnerability at the county level.
Table 4 provides results for the second stage of the hurdle model. At the summary level, the findings indicate that political environment does matter for SLR adaptation. To reiterate, three truncated Poisson models are used in the second stage to predict the count of focus areas, risk types, and risk reduction tactics for counties addressing SLR. Therefore, our results must be interpreted as indicating the impact of the independent variables on planning activity for those that have crossed the hurdle of addressing SLR. Thus, among counties that have addressed SLR, political environment is significantly associated with the number of focus areas, risk types, and risk reduction tactics. As Table 4 shows, the level of significance becomes stronger moving from general focus areas to specific risk types and then to the tactics for addressing them. On average, a 1% increase in the votes for Hillary Clinton in 2016 was associated with a 0.9% increase in focus areas, a 1.5% increase in risk types considered, and a 2% increase in the number of tactics used for counties which are addressing SLR. 3
Poisson Regression Results for Focus Areas, Risk Types, and Risk Reduction Tactics.
Robust standard errors in parentheses. Standard errors are calculated in Stata using suest (Long and Freese 2014). + p < .10, * p < .05, ** p < .01.
Within the context of our model, it can be argued that the most robust measures of adaptation are risk types and tactics because they result from information gathering and evaluation of alternatives, respectively. A simple form of adaptation could be identifying general risks via focus areas but not identifying specific risk types and stopping short of developing tactics for dealing with them. As the above elasticities indicate, political environment has roughly twice the impact on a county's use of diverse SLR tactics as compared to its focus areas addressed. However, it must be stressed that results simply indicate that counties with higher shares of democratic votes tend to engage in more SLR planning activity beyond the initial hurdle—not that strong liberal majorities are required.
When interpreting results for the second stage of the model, it is helpful to reinforce that the counties that comprise the second stage are distinct from the first because they tend to be more populated, exposed, and/or spend more per resident. Therefore, results from the second stage imply that among such counties doing something about SLR, their unique political environments are critical for understanding how much is done. For example, political environment may influence the actual resources allocated for SLR adaptation to protect a given population size and risk profile. Figure 7 provides further illustration of how political orientation impacted the number of tactics. Clearly, as the percentage of democratic votes in the 2016 presidential election increases, so does the number of adaptive strategies that governments pursue.
Impact of vote share on number of risk reduction tactics.
In sum, our results show that specific factors drive SLR adaptation but in distinct ways. Specifically, counties with greater resources, population sizes, and future risk exposure are more likely to address SLR—at least nominally. However, among counties addressing SLR, it is their distinct political environments that drive the extent of their SLR planning activity. In the following Discussion section, we first place our findings within the general climate literature and then SLR planning in particular.
Discussion
At the summary level, our results align with general research on adaptation to climate change. In their analysis of 156 U.S. municipal governments, Shi, Chu and Debats (2015) reported that approximately 60% were planning for climate change, and we found a similar proportion (62%) of southeastern coastal counties engaged in SLR planning. They also found that spending per capita and population were both significant predictors of nominally addressing climate change, consistent with the first stage of our hurdle model. Furthermore, other studies utilizing nominal measures of adaptation have also concluded that population size and financial resources drive adaptation (Krause 2011; Reckien et al. 2015). Our results differ from these studies to the extent that we do not find previous experiences with climate impacts (Shi, Chu and Debats 2015) or political environment (Krause 2011) to be significant predictors of nominally addressing SLR. Instead, we found that prospective measures of risk are better predictors than prior flood damages and that political orientation impacts how much SLR planning is done but not necessarily whether it is done.
Because we use a two-part model, it is important to distinguish between studies such as the above focusing on whether something is done to address climate versus how much is done to address it. For example, our findings from the second stage indicate that among counties addressing SLR, political environment influences the depth of planning. These results are consistent with Krause's (2012) analysis of general climate actions and Swann and Deslatte's (2019) review of different sustainability areas. When interpreting results for the second stage of the model, it is helpful to reinforce two points. First, roughly 60% of counties are planning for SLR. Second, spending per capita, population size, and risk exposure each plays a significant role in determining whether a county does such planning. Therefore, our null results from the second stage do not imply that resources, population size, and risk exposure are unimportant for understanding how much SLR planning activity a county engages in, given that they help distinguish the 60% doing something versus the 40% doing nothing. Instead, results indicate that, among counties planning for SLR, these variables do not systematically explain how much planning is actually done. This is a subtle point but a very important one for us to make because our results are ultimately consistent with others that find positive relationships when examining the impact of resources and population size on the extent of adaptation (Krause 2012; Hawkins et al. 2016; Swann and Deslatte 2019; Yeganeh, McCoy and Schenk 2020). Having placed our findings within the general climate research, we now situate our results within the SLR literature in particular.
Our findings complement other studies well. First, that local governments with greater per capita spending were found to engage in more SLR planning fits easily within studies using regression analysis (Fu et al. 2019; Fu 2020; Siders and Keenan 2020), casework of metropolitan areas (Ekstrom and Moser 2014; John III and Yusuf 2019), and a survey of numerous Floridian governments which ranked resources as the most prominent barrier (Milordis, Butler and Holmes 2023). Second, our findings that population size influences SLR planning activity aligns with Fu's (2020) study of eighty-six U.S. local governments, and other research that has shed light on the difficulties that smaller governments face in securing the resources and expertise needed for SLR planning (Hamin, Gurran and Emlinger 2014; Fu et al. 2019).
When stating our research questions, we purposefully used phrasing to emphasize that the impact of resources and population on SLR adaptation emerged relatively clear from the literature, at least in comparison to risk and political environment. Therefore, our research objectives focused on quantifying the impact of resources and population on SLR planning. In contrast, we stated our research questions on risk and politics differently as we argued these relationships to be relatively ambiguous. As detailed in the Results section, we found that spending per capita had the largest impact on the likelihood of addressing SLR, followed by population size and risk exposure but that political environment was not a significant predictor for addressing SLR.
Our findings that future risk exposure, as measured by the proportion of the population anticipated to experience annual flooding by 2050, is consistent with Siders and Keenan (2020) who found that more residents exposed to flooding risks was positively associated with coastal resiliency actions. However, Fu (2020) did not find a significant relationship between the value of buildings at risk and SLR planning activity, nor did Shi et al. (2024) when analyzing fiscal exposure and adaptive actions. Therefore, the relationship between risk exposure and SLR planning may easily depend on the measure and timeframe used for the analysis. We highlight the timeframe because our measures focus on 2050, whereas Fu (2020) and Shi et al. (2024) use SLR scenarios which correspond to later periods. However, future research is needed to better understand these relationships. Ideally, such research would consist of not just quantitative studies as described above, but also those that draw from the lived experiences of those managing these risks (Hines 2023).
Finally, our findings that political environment impacts SLR adaptation squares with Fu (2020), but a couple of points of clarification are needed. First, Fu (2020) surveyed local planners to gather data on their SLR planning activity and used self-reported measures of how supportive these local planners felt their political environments were. Second, we used governmental planning documents for SLR policies and the percentage of democratic votes. However, we must reiterate that our results do not find political leanings to be a barrier for simply addressing SLR. Instead, they indicate that among counties doing something about SLR, counties with more democratic votes tend to have planned more comprehensively for SLR. Within our study region, Milordis, Butler, and Holmes’ (2023) survey of municipalities in Florida found that approximately 30% felt the lack of prioritization from elected officials’ limited progress in addressing SLR. Therefore, our results taken in tandem with Fu (2020) provide general evidence on how political environments can impact SLR planning, and findings from the survey of Florida municipalities corroborate that political environments can impact how much progress is made within our study region.
To this point, we have structured the discussion around variables included in our model, and emphasized how findings relate to quantitative research. In the concluding section, we summarize our key findings and explain how future research can build on some of the limitations of this study.
Conclusion
Building on the work of Moser and Ekstrom (2010) on barriers to climate adaptation, Yusuf and St. John (2017) explain how SLR adaptation can become “stuck on options” where solutions are considered and potentially prioritized but not actually implemented. Consistent with this characterization, most southeastern coastal counties have addressed SLR in their plans but relatively few evidence comprehensive analysis of risk, let alone the implementation of adaptive projects (Grandage, Hines and Willoughby 2024). This study contributes to our understanding of planning for climate change by quantifying drivers to SLR adaptation.
Consistent with the general research on adaptation to climate change, we found that resources, population size, risk, and political environment each played a significant role in SLR planning (Krause 2011, 2012; Carmin, Nadkarni and Rhie 2012; Reckien et al. 2015; Shi, Chu and Debats 2015; Hawkins et al. 2016; Swann and Deslatte 2019; Yeganeh, McCoy and Schenk 2020). However, because we conceptualize adaptation as first choosing whether to address SLR and then determining how much to address it, we used a hurdle model. This modeling approach suits the data because some county governments have yet to formally incorporate SLR into their planning practices; most are in the early stages and a few a relatively far along. Furthermore, because numerous studies have highlighted that inadequate resources act as a barrier to SLR adaptation and that smaller governments often lack the capacity to address complex climate risks (Hamin, Gurran and Emlinger 2014; Yusuf and St. John 2017; Fu et al. 2019; Milordis, Butler and Holmes 2023), we focused on quantifying these relationships.
In contrast, our analysis of risk exposure and political environment focused on testing whether significant relationships could be uncovered, specifically whether counties with greater risk exposure and more liberal political environments were more likely to address SLR and engage in more comprehensive SLR planning. We found that our prospective measure of risk, the share of the population anticipated to be exposed to annual flooding by 2050, significantly increased the likelihood of addressing SLR, but that the retrospective measure of flood damages did not. Given that the most pronounced impacts of SLR will occur in the future, we consider these results sensible and supportive of the notion that adaptation decisions are influenced by risk extent and type (Siders and Keenan 2020). Finally, results indicate that political environment does not serve as a significant barrier to nominally addressing SLR but that it can be a supportive factor for engaging in more thorough planning.
Future research can build on some of the limitations of this study. First, although the eighty-five southeastern coastal counties provide great variety, the number of governments could still be expanded, especially by analyzing adaptation across different geopolitical regions. Second, our focus is limited to county governments which are just one layer of the complex intergovernmental system. Third, other explanatory variables could be included, such as intensity of community engagement, collaboration, interest group activity, and strength of advocacy coalitions, along with measures of capacity which focus on actual spending toward SLR planning and dedicated human resources. Next, our classifications measure what has been formally documented, so it is possible that some governments performed vulnerability assessments but did not incorporate such information into their documents or that new policies could have been formulated within the last few years since our data was collected. Our study is also cross sectional and future work would benefit from studying adaptation among numerous governments over time. Finally, like many studies, we focus heavily on planning and less on implementation.
Ultimately, for strategic planning to achieve its intended outcomes, attention must be focused on the challenges of implementation too, especially considering some of the major projects that are necessary to transform urban areas. To effectively manage such transformations, governments must build internal capacity in the areas of contracting and project management (Brown, Potoski and Van Slyke 2018; Grandage 2022; Greiman 2023). Otherwise, efforts to address climate change may be unnecessarily slow, expensive, and fall short of intended benefits (Flyvbjerg and Gardner 2023). We encourage future research to explore how the discipline of project management can contribute to successfully adapting to climate change and reducing implementation deficits.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
